From bd245f40eba42400283aad611e03559906411d7d Mon Sep 17 00:00:00 2001
From: csslc <22042099r@connect.polyu.hk>
Date: Thu, 4 Jan 2024 21:45:50 +0800
Subject: [PATCH] release_code

---
 .idea/vcs.xml                                 |    1 +
 .idea/workspace.xml                           |   11 +-
 README.md                                     |   68 +-
 cal_iqa/cal_iqa.py                            |  236 ++
 cal_iqa/iqa_G-STD.py                          |  111 +
 cal_iqa/iqa_L-STD.py                          |  190 ++
 ...general_deg_stablesr_realesrgan_train.yaml |   63 +
 .../general_deg_stablesr_realesrgan_val.yaml  |   64 +
 configs/model/ccsr_stage1.yaml                |  108 +
 configs/model/ccsr_stage2.yaml                |  118 +
 configs/train_ccsr_stage1.yaml                |   47 +
 configs/train_ccsr_stage2.yaml                |   49 +
 .../batch_transform.cpython-310.pyc           |  Bin 0 -> 8226 bytes
 .../batch_transform.cpython-37.pyc            |  Bin 0 -> 12024 bytes
 .../bicubic_torchvision.cpython-310.pyc       |  Bin 0 -> 2275 bytes
 .../__pycache__/data_module.cpython-310.pyc   |  Bin 0 -> 2727 bytes
 .../__pycache__/data_module.cpython-37.pyc    |  Bin 0 -> 2661 bytes
 .../__pycache__/projectdata.cpython-310.pyc   |  Bin 0 -> 3722 bytes
 .../__pycache__/projectdata.cpython-37.pyc    |  Bin 0 -> 3669 bytes
 .../__pycache__/realesrgan.cpython-310.pyc    |  Bin 0 -> 4046 bytes
 dataset/__pycache__/realesrgan.cpython-37.pyc |  Bin 0 -> 3978 bytes
 dataset/batch_transform.py                    |  374 +++
 dataset/bicubic_torchvision.py                |  107 +
 dataset/codeformer.py                         |  109 +
 dataset/data_module.py                        |   68 +
 dataset/projectdata.py                        |  121 +
 dataset/realesrgan.py                         |  183 ++
 figs/logo.png                                 |  Bin 18213 -> 30467 bytes
 inference_ccsr.py                             |  239 ++
 ldm/__pycache__/util.cpython-310.pyc          |  Bin 0 -> 6165 bytes
 ldm/__pycache__/util.cpython-37.pyc           |  Bin 0 -> 6140 bytes
 .../xformers_state.cpython-310.pyc            |  Bin 0 -> 982 bytes
 ldm/__pycache__/xformers_state.cpython-37.pyc |  Bin 0 -> 996 bytes
 ldm/data/__init__.py                          |    0
 ldm/data/util.py                              |   24 +
 .../__pycache__/autoencoder.cpython-310.pyc   |  Bin 0 -> 7866 bytes
 .../__pycache__/autoencoder.cpython-37.pyc    |  Bin 0 -> 7766 bytes
 ldm/models/autoencoder.py                     |  224 ++
 ldm/models/diffusion/__init__.py              |    0
 .../__pycache__/__init__.cpython-310.pyc      |  Bin 0 -> 173 bytes
 .../__pycache__/__init__.cpython-37.pyc       |  Bin 0 -> 170 bytes
 .../__pycache__/ddim.cpython-310.pyc          |  Bin 0 -> 9347 bytes
 .../diffusion/__pycache__/ddim.cpython-37.pyc |  Bin 0 -> 9245 bytes
 .../__pycache__/ddpm.cpython-310.pyc          |  Bin 0 -> 53110 bytes
 .../diffusion/__pycache__/ddpm.cpython-37.pyc |  Bin 0 -> 54432 bytes
 .../ddpm_ccsr_stage1.cpython-310.pyc          |  Bin 0 -> 54491 bytes
 .../ddpm_ccsr_stage2.cpython-310.pyc          |  Bin 0 -> 58092 bytes
 .../__pycache__/ddpm_psample2.cpython-310.pyc |  Bin 0 -> 58248 bytes
 .../__pycache__/ddpm_psample3.cpython-310.pyc |  Bin 0 -> 58248 bytes
 .../__pycache__/ddpm_qsample.cpython-310.pyc  |  Bin 0 -> 54428 bytes
 .../__pycache__/ddpm_qsample.cpython-37.pyc   |  Bin 0 -> 55828 bytes
 .../__pycache__/ddpm_qsample2.cpython-310.pyc |  Bin 0 -> 54422 bytes
 .../__pycache__/ddpm_qsample2.cpython-37.pyc  |  Bin 0 -> 55829 bytes
 .../ddpm_qsample_0_8.cpython-310.pyc          |  Bin 0 -> 54425 bytes
 .../ddpm_qsample_0_8.cpython-37.pyc           |  Bin 0 -> 55832 bytes
 ..._qsample_finetunedecoder24.cpython-310.pyc |  Bin 0 -> 57902 bytes
 ..._qsample_finetunedecoder30.cpython-310.pyc |  Bin 0 -> 58014 bytes
 ..._qsample_finetunedecoder45.cpython-310.pyc |  Bin 0 -> 58014 bytes
 ...m_qsample_finetunedecoder45.cpython-37.pyc |  Bin 0 -> 59692 bytes
 ...le_finetunedecoder45_22_15.cpython-310.pyc |  Bin 0 -> 58025 bytes
 ...ple_finetunedecoder45_22_15.cpython-37.pyc |  Bin 0 -> 59703 bytes
 ...le_finetunedecoder45_30_22.cpython-310.pyc |  Bin 0 -> 58025 bytes
 ...ple_finetunedecoder45_30_22.cpython-37.pyc |  Bin 0 -> 59703 bytes
 ...ple_finetunedecoder45_30_8.cpython-310.pyc |  Bin 0 -> 58037 bytes
 ...mple_finetunedecoder45_30_8.cpython-37.pyc |  Bin 0 -> 59715 bytes
 ...le_finetunedecoder45_36_15.cpython-310.pyc |  Bin 0 -> 58025 bytes
 ...ple_finetunedecoder45_36_15.cpython-37.pyc |  Bin 0 -> 59703 bytes
 ...ample_finetunedecoder45_lr.cpython-310.pyc |  Bin 0 -> 58088 bytes
 .../ddpm_qsample_lr.cpython-310.pyc           |  Bin 0 -> 54488 bytes
 ...fusion_finetune_x0_qsample.cpython-310.pyc |  Bin 0 -> 88062 bytes
 ldm/models/diffusion/ddim.py                  |  336 +++
 ldm/models/diffusion/ddpm.py                  | 1811 +++++++++++++++
 ldm/models/diffusion/ddpm_ccsr_stage1.py      | 1858 +++++++++++++++
 ldm/models/diffusion/ddpm_ccsr_stage2.py      | 2058 +++++++++++++++++
 ldm/models/diffusion/dpm_solver/__init__.py   |    1 +
 ldm/models/diffusion/dpm_solver/dpm_solver.py | 1154 +++++++++
 ldm/models/diffusion/dpm_solver/sampler.py    |   87 +
 ldm/models/diffusion/plms.py                  |  244 ++
 ldm/models/diffusion/sampling_util.py         |   22 +
 .../__pycache__/attention.cpython-310.pyc     |  Bin 0 -> 10381 bytes
 .../__pycache__/attention.cpython-37.pyc      |  Bin 0 -> 10534 bytes
 ldm/modules/__pycache__/ema.cpython-310.pyc   |  Bin 0 -> 3212 bytes
 ldm/modules/__pycache__/ema.cpython-37.pyc    |  Bin 0 -> 3186 bytes
 ldm/modules/attention.py                      |  343 +++
 ldm/modules/diffusionmodules/__init__.py      |    0
 .../__pycache__/__init__.cpython-310.pyc      |  Bin 0 -> 181 bytes
 .../__pycache__/__init__.cpython-37.pyc       |  Bin 0 -> 178 bytes
 .../__pycache__/model.cpython-310.pyc         |  Bin 0 -> 20783 bytes
 .../__pycache__/model.cpython-37.pyc          |  Bin 0 -> 21736 bytes
 .../__pycache__/openaimodel.cpython-310.pyc   |  Bin 0 -> 20651 bytes
 .../__pycache__/openaimodel.cpython-37.pyc    |  Bin 0 -> 20785 bytes
 .../__pycache__/util.cpython-310.pyc          |  Bin 0 -> 9710 bytes
 .../__pycache__/util.cpython-37.pyc           |  Bin 0 -> 9626 bytes
 ldm/modules/diffusionmodules/model.py         |  855 +++++++
 ldm/modules/diffusionmodules/openaimodel.py   |  786 +++++++
 ldm/modules/diffusionmodules/upscaling.py     |   81 +
 ldm/modules/diffusionmodules/util.py          |  270 +++
 ldm/modules/distributions/__init__.py         |    0
 .../__pycache__/__init__.cpython-310.pyc      |  Bin 0 -> 178 bytes
 .../__pycache__/__init__.cpython-37.pyc       |  Bin 0 -> 175 bytes
 .../__pycache__/distributions.cpython-310.pyc |  Bin 0 -> 3776 bytes
 .../__pycache__/distributions.cpython-37.pyc  |  Bin 0 -> 3732 bytes
 ldm/modules/distributions/distributions.py    |   92 +
 ldm/modules/ema.py                            |   80 +
 ldm/modules/encoders/__init__.py              |    0
 .../__pycache__/__init__.cpython-310.pyc      |  Bin 0 -> 173 bytes
 .../__pycache__/__init__.cpython-37.pyc       |  Bin 0 -> 170 bytes
 .../__pycache__/modules.cpython-310.pyc       |  Bin 0 -> 8440 bytes
 .../__pycache__/modules.cpython-37.pyc        |  Bin 0 -> 8516 bytes
 ldm/modules/encoders/modules.py               |  215 ++
 ldm/modules/losses/__init__.py                |    1 +
 .../__pycache__/__init__.cpython-310.pyc      |  Bin 0 -> 251 bytes
 .../__pycache__/__init__.cpython-37.pyc       |  Bin 0 -> 248 bytes
 .../__pycache__/__init__.cpython-38.pyc       |  Bin 0 -> 258 bytes
 .../contperceptual.cpython-310.pyc            |  Bin 0 -> 4174 bytes
 .../__pycache__/contperceptual.cpython-37.pyc |  Bin 0 -> 4166 bytes
 .../__pycache__/contperceptual.cpython-38.pyc |  Bin 0 -> 4235 bytes
 ldm/modules/losses/contperceptual.py          |  152 ++
 ldm/modules/losses/vqperceptual.py            |  180 ++
 ldm/modules/midas/__init__.py                 |    0
 ldm/modules/midas/api.py                      |  170 ++
 ldm/modules/midas/midas/__init__.py           |    0
 ldm/modules/midas/midas/base_model.py         |   16 +
 ldm/modules/midas/midas/blocks.py             |  342 +++
 ldm/modules/midas/midas/dpt_depth.py          |  109 +
 ldm/modules/midas/midas/midas_net.py          |   76 +
 ldm/modules/midas/midas/midas_net_custom.py   |  128 +
 ldm/modules/midas/midas/transforms.py         |  234 ++
 ldm/modules/midas/midas/vit.py                |  491 ++++
 ldm/modules/midas/utils.py                    |  189 ++
 ldm/util.py                                   |  198 ++
 ldm/xformers_state.py                         |   30 +
 model/__pycache__/callbacks.cpython-310.pyc   |  Bin 0 -> 2674 bytes
 model/__pycache__/callbacks.cpython-37.pyc    |  Bin 0 -> 2593 bytes
 model/__pycache__/ccsr_stage1.cpython-310.pyc |  Bin 0 -> 10337 bytes
 model/__pycache__/ccsr_stage2.cpython-310.pyc |  Bin 0 -> 10484 bytes
 model/__pycache__/cldm.cpython-310.pyc        |  Bin 0 -> 10352 bytes
 model/__pycache__/cldm.cpython-37.pyc         |  Bin 0 -> 10149 bytes
 .../__pycache__/cldm_decoder.cpython-310.pyc  |  Bin 0 -> 10502 bytes
 model/__pycache__/cldm_decoder.cpython-37.pyc |  Bin 0 -> 10306 bytes
 .../cldm_decoder_24.cpython-310.pyc           |  Bin 0 -> 10505 bytes
 .../cldm_decoder_30.cpython-310.pyc           |  Bin 0 -> 10505 bytes
 .../cldm_decoder_45_22_15.cpython-310.pyc     |  Bin 0 -> 10517 bytes
 .../cldm_decoder_45_22_15.cpython-37.pyc      |  Bin 0 -> 10321 bytes
 .../cldm_decoder_45_30_22.cpython-310.pyc     |  Bin 0 -> 10517 bytes
 .../cldm_decoder_45_30_22.cpython-37.pyc      |  Bin 0 -> 10321 bytes
 .../cldm_decoder_45_30_8.cpython-310.pyc      |  Bin 0 -> 10515 bytes
 .../cldm_decoder_45_30_8.cpython-37.pyc       |  Bin 0 -> 10319 bytes
 .../cldm_decoder_45_36_15.cpython-310.pyc     |  Bin 0 -> 10517 bytes
 .../cldm_decoder_45_36_15.cpython-37.pyc      |  Bin 0 -> 10321 bytes
 .../cldm_decoder_lr.cpython-310.pyc           |  Bin 0 -> 10508 bytes
 model/__pycache__/cldm_lr.cpython-310.pyc     |  Bin 0 -> 10270 bytes
 .../cldm_lr_psample2.cpython-310.pyc          |  Bin 0 -> 10342 bytes
 .../cldm_lr_psample3.cpython-310.pyc          |  Bin 0 -> 10342 bytes
 .../cldm_lr_qsample.cpython-310.pyc           |  Bin 0 -> 10347 bytes
 .../cldm_lr_qsample.cpython-37.pyc            |  Bin 0 -> 10144 bytes
 .../cldm_lr_qsample2.cpython-310.pyc          |  Bin 0 -> 10342 bytes
 .../cldm_lr_qsample2.cpython-37.pyc           |  Bin 0 -> 10146 bytes
 .../cldm_lr_qsample_0_8.cpython-310.pyc       |  Bin 0 -> 10348 bytes
 .../cldm_lr_qsample_0_8.cpython-37.pyc        |  Bin 0 -> 10152 bytes
 .../cldm_lr_qsample_lr.cpython-310.pyc        |  Bin 0 -> 10346 bytes
 model/__pycache__/cond_fn.cpython-310.pyc     |  Bin 0 -> 2949 bytes
 model/__pycache__/cond_fn.cpython-37.pyc      |  Bin 0 -> 2913 bytes
 model/__pycache__/mixins.cpython-310.pyc      |  Bin 0 -> 640 bytes
 model/__pycache__/mixins.cpython-37.pyc       |  Bin 0 -> 625 bytes
 model/__pycache__/p_sampler.cpython-310.pyc   |  Bin 0 -> 16149 bytes
 model/__pycache__/q_sampler.cpython-310.pyc   |  Bin 0 -> 21269 bytes
 model/__pycache__/q_sampler.cpython-37.pyc    |  Bin 0 -> 32803 bytes
 .../spaced_sampler.cpython-310.pyc            |  Bin 0 -> 16422 bytes
 .../__pycache__/spaced_sampler.cpython-37.pyc |  Bin 0 -> 16581 bytes
 model/__pycache__/swinir.cpython-310.pyc      |  Bin 0 -> 31251 bytes
 model/callbacks.py                            |   74 +
 model/ccsr_stage1.py                          |  406 ++++
 model/ccsr_stage2.py                          |  407 ++++
 model/cond_fn.py                              |   78 +
 model/mixins.py                               |    9 +
 model/q_sampler.py                            | 1043 +++++++++
 model/spaced_sampler.py                       |  670 ++++++
 preset/test_datasets/19.jpg                   |  Bin 0 -> 13552 bytes
 preset/test_datasets/49.jpg                   |  Bin 0 -> 2790 bytes
 requirements.txt                              |   19 +
 scripts/make_file_list.py                     |   36 +
 scripts/make_stage2_init_weight.py            |   80 +
 scripts/test.sh                               |    0
 sh/1_pip.sh                                   |    3 +
 sh/2_inference.sh                             |   14 +
 sh/3_create_weight.sh                         |   11 +
 sh/4_generate_data_file.sh                    |   10 +
 sh/5_train_stage1.sh                          |    3 +
 sh/6_train_stage2.sh                          |    3 +
 train.py                                      |   32 +
 utils/__pycache__/common.cpython-310.pyc      |  Bin 0 -> 2069 bytes
 utils/__pycache__/common.cpython-37.pyc       |  Bin 0 -> 2060 bytes
 utils/__pycache__/degradation.cpython-310.pyc |  Bin 0 -> 20296 bytes
 utils/__pycache__/degradation.cpython-37.pyc  |  Bin 0 -> 22425 bytes
 utils/__pycache__/file.cpython-310.pyc        |  Bin 0 -> 2708 bytes
 utils/__pycache__/file.cpython-37.pyc         |  Bin 0 -> 2625 bytes
 utils/__pycache__/metrics.cpython-310.pyc     |  Bin 0 -> 2652 bytes
 utils/common.py                               |   51 +
 utils/degradation.py                          |  765 ++++++
 utils/face_restoration_helper.py              |  517 +++++
 utils/file.py                                 |  103 +
 utils/image/__init__.py                       |   26 +
 .../__pycache__/__init__.cpython-310.pyc      |  Bin 0 -> 591 bytes
 .../image/__pycache__/__init__.cpython-37.pyc |  Bin 0 -> 624 bytes
 .../__pycache__/align_color.cpython-310.pyc   |  Bin 0 -> 3716 bytes
 .../__pycache__/align_color.cpython-37.pyc    |  Bin 0 -> 3843 bytes
 .../image/__pycache__/common.cpython-310.pyc  |  Bin 0 -> 7506 bytes
 utils/image/__pycache__/common.cpython-37.pyc |  Bin 0 -> 7517 bytes
 .../__pycache__/diffjpeg.cpython-310.pyc      |  Bin 0 -> 14822 bytes
 .../image/__pycache__/diffjpeg.cpython-37.pyc |  Bin 0 -> 16804 bytes
 .../__pycache__/usm_sharp.cpython-310.pyc     |  Bin 0 -> 1235 bytes
 .../__pycache__/usm_sharp.cpython-37.pyc      |  Bin 0 -> 1220 bytes
 utils/image/align_color.py                    |  119 +
 utils/image/common.py                         |  235 ++
 utils/image/diffjpeg.py                       |  492 ++++
 utils/image/usm_sharp.py                      |   29 +
 utils/metrics.py                              |   66 +
 utils/realesrgan/realesrganer.py              |  339 +++
 utils/realesrgan/rrdbnet.py                   |  183 ++
 220 files changed, 20910 insertions(+), 37 deletions(-)
 create mode 100644 cal_iqa/cal_iqa.py
 create mode 100644 cal_iqa/iqa_G-STD.py
 create mode 100644 cal_iqa/iqa_L-STD.py
 create mode 100644 configs/dataset/general_deg_stablesr_realesrgan_train.yaml
 create mode 100644 configs/dataset/general_deg_stablesr_realesrgan_val.yaml
 create mode 100644 configs/model/ccsr_stage1.yaml
 create mode 100644 configs/model/ccsr_stage2.yaml
 create mode 100644 configs/train_ccsr_stage1.yaml
 create mode 100644 configs/train_ccsr_stage2.yaml
 create mode 100644 dataset/__pycache__/batch_transform.cpython-310.pyc
 create mode 100644 dataset/__pycache__/batch_transform.cpython-37.pyc
 create mode 100644 dataset/__pycache__/bicubic_torchvision.cpython-310.pyc
 create mode 100644 dataset/__pycache__/data_module.cpython-310.pyc
 create mode 100644 dataset/__pycache__/data_module.cpython-37.pyc
 create mode 100644 dataset/__pycache__/projectdata.cpython-310.pyc
 create mode 100644 dataset/__pycache__/projectdata.cpython-37.pyc
 create mode 100644 dataset/__pycache__/realesrgan.cpython-310.pyc
 create mode 100644 dataset/__pycache__/realesrgan.cpython-37.pyc
 create mode 100644 dataset/batch_transform.py
 create mode 100644 dataset/bicubic_torchvision.py
 create mode 100644 dataset/codeformer.py
 create mode 100644 dataset/data_module.py
 create mode 100644 dataset/projectdata.py
 create mode 100644 dataset/realesrgan.py
 create mode 100644 inference_ccsr.py
 create mode 100644 ldm/__pycache__/util.cpython-310.pyc
 create mode 100644 ldm/__pycache__/util.cpython-37.pyc
 create mode 100644 ldm/__pycache__/xformers_state.cpython-310.pyc
 create mode 100644 ldm/__pycache__/xformers_state.cpython-37.pyc
 create mode 100644 ldm/data/__init__.py
 create mode 100644 ldm/data/util.py
 create mode 100644 ldm/models/__pycache__/autoencoder.cpython-310.pyc
 create mode 100644 ldm/models/__pycache__/autoencoder.cpython-37.pyc
 create mode 100644 ldm/models/autoencoder.py
 create mode 100644 ldm/models/diffusion/__init__.py
 create mode 100644 ldm/models/diffusion/__pycache__/__init__.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/__init__.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddim.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddim.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_ccsr_stage1.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_ccsr_stage2.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_psample2.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_psample3.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample2.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample2.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_0_8.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_0_8.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder24.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder30.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_22_15.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_22_15.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_22.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_22.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_8.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_8.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_36_15.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_36_15.cpython-37.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_lr.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_qsample_lr.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/__pycache__/ddpm_truncted_diffusion_finetune_x0_qsample.cpython-310.pyc
 create mode 100644 ldm/models/diffusion/ddim.py
 create mode 100644 ldm/models/diffusion/ddpm.py
 create mode 100644 ldm/models/diffusion/ddpm_ccsr_stage1.py
 create mode 100644 ldm/models/diffusion/ddpm_ccsr_stage2.py
 create mode 100644 ldm/models/diffusion/dpm_solver/__init__.py
 create mode 100644 ldm/models/diffusion/dpm_solver/dpm_solver.py
 create mode 100644 ldm/models/diffusion/dpm_solver/sampler.py
 create mode 100644 ldm/models/diffusion/plms.py
 create mode 100644 ldm/models/diffusion/sampling_util.py
 create mode 100644 ldm/modules/__pycache__/attention.cpython-310.pyc
 create mode 100644 ldm/modules/__pycache__/attention.cpython-37.pyc
 create mode 100644 ldm/modules/__pycache__/ema.cpython-310.pyc
 create mode 100644 ldm/modules/__pycache__/ema.cpython-37.pyc
 create mode 100644 ldm/modules/attention.py
 create mode 100644 ldm/modules/diffusionmodules/__init__.py
 create mode 100644 ldm/modules/diffusionmodules/__pycache__/__init__.cpython-310.pyc
 create mode 100644 ldm/modules/diffusionmodules/__pycache__/__init__.cpython-37.pyc
 create mode 100644 ldm/modules/diffusionmodules/__pycache__/model.cpython-310.pyc
 create mode 100644 ldm/modules/diffusionmodules/__pycache__/model.cpython-37.pyc
 create mode 100644 ldm/modules/diffusionmodules/__pycache__/openaimodel.cpython-310.pyc
 create mode 100644 ldm/modules/diffusionmodules/__pycache__/openaimodel.cpython-37.pyc
 create mode 100644 ldm/modules/diffusionmodules/__pycache__/util.cpython-310.pyc
 create mode 100644 ldm/modules/diffusionmodules/__pycache__/util.cpython-37.pyc
 create mode 100644 ldm/modules/diffusionmodules/model.py
 create mode 100644 ldm/modules/diffusionmodules/openaimodel.py
 create mode 100644 ldm/modules/diffusionmodules/upscaling.py
 create mode 100644 ldm/modules/diffusionmodules/util.py
 create mode 100644 ldm/modules/distributions/__init__.py
 create mode 100644 ldm/modules/distributions/__pycache__/__init__.cpython-310.pyc
 create mode 100644 ldm/modules/distributions/__pycache__/__init__.cpython-37.pyc
 create mode 100644 ldm/modules/distributions/__pycache__/distributions.cpython-310.pyc
 create mode 100644 ldm/modules/distributions/__pycache__/distributions.cpython-37.pyc
 create mode 100644 ldm/modules/distributions/distributions.py
 create mode 100644 ldm/modules/ema.py
 create mode 100644 ldm/modules/encoders/__init__.py
 create mode 100644 ldm/modules/encoders/__pycache__/__init__.cpython-310.pyc
 create mode 100644 ldm/modules/encoders/__pycache__/__init__.cpython-37.pyc
 create mode 100644 ldm/modules/encoders/__pycache__/modules.cpython-310.pyc
 create mode 100644 ldm/modules/encoders/__pycache__/modules.cpython-37.pyc
 create mode 100644 ldm/modules/encoders/modules.py
 create mode 100644 ldm/modules/losses/__init__.py
 create mode 100644 ldm/modules/losses/__pycache__/__init__.cpython-310.pyc
 create mode 100644 ldm/modules/losses/__pycache__/__init__.cpython-37.pyc
 create mode 100644 ldm/modules/losses/__pycache__/__init__.cpython-38.pyc
 create mode 100644 ldm/modules/losses/__pycache__/contperceptual.cpython-310.pyc
 create mode 100644 ldm/modules/losses/__pycache__/contperceptual.cpython-37.pyc
 create mode 100644 ldm/modules/losses/__pycache__/contperceptual.cpython-38.pyc
 create mode 100644 ldm/modules/losses/contperceptual.py
 create mode 100644 ldm/modules/losses/vqperceptual.py
 create mode 100644 ldm/modules/midas/__init__.py
 create mode 100644 ldm/modules/midas/api.py
 create mode 100644 ldm/modules/midas/midas/__init__.py
 create mode 100644 ldm/modules/midas/midas/base_model.py
 create mode 100644 ldm/modules/midas/midas/blocks.py
 create mode 100644 ldm/modules/midas/midas/dpt_depth.py
 create mode 100644 ldm/modules/midas/midas/midas_net.py
 create mode 100644 ldm/modules/midas/midas/midas_net_custom.py
 create mode 100644 ldm/modules/midas/midas/transforms.py
 create mode 100644 ldm/modules/midas/midas/vit.py
 create mode 100644 ldm/modules/midas/utils.py
 create mode 100644 ldm/util.py
 create mode 100644 ldm/xformers_state.py
 create mode 100644 model/__pycache__/callbacks.cpython-310.pyc
 create mode 100644 model/__pycache__/callbacks.cpython-37.pyc
 create mode 100644 model/__pycache__/ccsr_stage1.cpython-310.pyc
 create mode 100644 model/__pycache__/ccsr_stage2.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_decoder.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_decoder.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_decoder_24.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_decoder_30.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_decoder_45_22_15.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_decoder_45_22_15.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_decoder_45_30_22.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_decoder_45_30_22.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_decoder_45_30_8.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_decoder_45_30_8.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_decoder_45_36_15.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_decoder_45_36_15.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_decoder_lr.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_lr.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_lr_psample2.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_lr_psample3.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_lr_qsample.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_lr_qsample.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_lr_qsample2.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_lr_qsample2.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_lr_qsample_0_8.cpython-310.pyc
 create mode 100644 model/__pycache__/cldm_lr_qsample_0_8.cpython-37.pyc
 create mode 100644 model/__pycache__/cldm_lr_qsample_lr.cpython-310.pyc
 create mode 100644 model/__pycache__/cond_fn.cpython-310.pyc
 create mode 100644 model/__pycache__/cond_fn.cpython-37.pyc
 create mode 100644 model/__pycache__/mixins.cpython-310.pyc
 create mode 100644 model/__pycache__/mixins.cpython-37.pyc
 create mode 100644 model/__pycache__/p_sampler.cpython-310.pyc
 create mode 100644 model/__pycache__/q_sampler.cpython-310.pyc
 create mode 100644 model/__pycache__/q_sampler.cpython-37.pyc
 create mode 100644 model/__pycache__/spaced_sampler.cpython-310.pyc
 create mode 100644 model/__pycache__/spaced_sampler.cpython-37.pyc
 create mode 100644 model/__pycache__/swinir.cpython-310.pyc
 create mode 100644 model/callbacks.py
 create mode 100644 model/ccsr_stage1.py
 create mode 100644 model/ccsr_stage2.py
 create mode 100644 model/cond_fn.py
 create mode 100644 model/mixins.py
 create mode 100644 model/q_sampler.py
 create mode 100644 model/spaced_sampler.py
 create mode 100644 preset/test_datasets/19.jpg
 create mode 100644 preset/test_datasets/49.jpg
 create mode 100644 requirements.txt
 create mode 100644 scripts/make_file_list.py
 create mode 100644 scripts/make_stage2_init_weight.py
 create mode 100644 scripts/test.sh
 create mode 100644 sh/1_pip.sh
 create mode 100644 sh/2_inference.sh
 create mode 100644 sh/3_create_weight.sh
 create mode 100644 sh/4_generate_data_file.sh
 create mode 100644 sh/5_train_stage1.sh
 create mode 100644 sh/6_train_stage2.sh
 create mode 100644 train.py
 create mode 100644 utils/__pycache__/common.cpython-310.pyc
 create mode 100644 utils/__pycache__/common.cpython-37.pyc
 create mode 100644 utils/__pycache__/degradation.cpython-310.pyc
 create mode 100644 utils/__pycache__/degradation.cpython-37.pyc
 create mode 100644 utils/__pycache__/file.cpython-310.pyc
 create mode 100644 utils/__pycache__/file.cpython-37.pyc
 create mode 100644 utils/__pycache__/metrics.cpython-310.pyc
 create mode 100644 utils/common.py
 create mode 100644 utils/degradation.py
 create mode 100644 utils/face_restoration_helper.py
 create mode 100644 utils/file.py
 create mode 100644 utils/image/__init__.py
 create mode 100644 utils/image/__pycache__/__init__.cpython-310.pyc
 create mode 100644 utils/image/__pycache__/__init__.cpython-37.pyc
 create mode 100644 utils/image/__pycache__/align_color.cpython-310.pyc
 create mode 100644 utils/image/__pycache__/align_color.cpython-37.pyc
 create mode 100644 utils/image/__pycache__/common.cpython-310.pyc
 create mode 100644 utils/image/__pycache__/common.cpython-37.pyc
 create mode 100644 utils/image/__pycache__/diffjpeg.cpython-310.pyc
 create mode 100644 utils/image/__pycache__/diffjpeg.cpython-37.pyc
 create mode 100644 utils/image/__pycache__/usm_sharp.cpython-310.pyc
 create mode 100644 utils/image/__pycache__/usm_sharp.cpython-37.pyc
 create mode 100644 utils/image/align_color.py
 create mode 100644 utils/image/common.py
 create mode 100644 utils/image/diffjpeg.py
 create mode 100644 utils/image/usm_sharp.py
 create mode 100644 utils/metrics.py
 create mode 100644 utils/realesrgan/realesrganer.py
 create mode 100644 utils/realesrgan/rrdbnet.py

diff --git a/.idea/vcs.xml b/.idea/vcs.xml
index 94a25f7..f6eb092 100644
--- a/.idea/vcs.xml
+++ b/.idea/vcs.xml
@@ -2,5 +2,6 @@
 <project version="4">
   <component name="VcsDirectoryMappings">
     <mapping directory="$PROJECT_DIR$" vcs="Git" />
+    <mapping directory="$PROJECT_DIR$/src/taming-transformers" vcs="Git" />
   </component>
 </project>
\ No newline at end of file
diff --git a/.idea/workspace.xml b/.idea/workspace.xml
index 58203d5..e17ae08 100644
--- a/.idea/workspace.xml
+++ b/.idea/workspace.xml
@@ -14,16 +14,17 @@
     <option name="stateVersion" value="1" />
   </component>
   <component name="ProjectId" id="2Zu39E1mU2U1EW9sGWi9CYEFFC5" />
+  <component name="ProjectLevelVcsManager" settingsEditedManually="true" />
   <component name="ProjectViewState">
     <option name="hideEmptyMiddlePackages" value="true" />
     <option name="showLibraryContents" value="true" />
   </component>
-  <component name="PropertiesComponent"><![CDATA[{
-  "keyToString": {
-    "RunOnceActivity.OpenProjectViewOnStart": "true",
-    "RunOnceActivity.ShowReadmeOnStart": "true"
+  <component name="PropertiesComponent">{
+  &quot;keyToString&quot;: {
+    &quot;RunOnceActivity.OpenProjectViewOnStart&quot;: &quot;true&quot;,
+    &quot;RunOnceActivity.ShowReadmeOnStart&quot;: &quot;true&quot;
   }
-}]]></component>
+}</component>
   <component name="SpellCheckerSettings" RuntimeDictionaries="0" Folders="0" CustomDictionaries="0" DefaultDictionary="application-level" UseSingleDictionary="true" transferred="true" />
   <component name="TaskManager">
     <task active="true" id="Default" summary="Default task">
diff --git a/README.md b/README.md
index e90e669..dc73229 100644
--- a/README.md
+++ b/README.md
@@ -28,7 +28,7 @@
 ## 🌟 Overview Framework
 ![ccsr](figs/framework.png)
 
-## 👀 Visual Results
+## 😍 Visual Results
 ### Comparisons on Real-World SR
 ![ccsr](figs/realworld.png)
 
@@ -37,7 +37,7 @@
 For more comparisons, please refer to our paper for details.
 
 ## 📝 Quantitative comparisons
-We propose new stability metrics, namely global standard deviation (G-STD) and local standard deviation (L-STD), to measure the image-level and pixel-level variations of the SR results for diffusion-based methods.
+We propose new stability metrics, namely global standard deviation (G-STD) and local standard deviation (L-STD), to respectively measure the image-level and pixel-level variations of the SR results of diffusion-based methods.
 
 More details about G-STD and L-STD can be found in our paper.
 
@@ -52,16 +52,17 @@ cd CCSR
 conda create -n ccsr python=3.9
 conda activate ccsr
 pip install -r requirements.txt
+pip install -e git+https://github.com/CompVis/taming-transformers.git@master#egg=taming-transformers
 ```
 ## 🍭 Quick Inference
 #### Step 1: Download the pretrained models
 - Download the pretrained SD-2.1base models from [HuggingFace](https://huggingface.co/stabilityai/stable-diffusion-2-1-base).
 - Download the CCSR models from:
 
-| Model Name           | Description                                  | GoogleDrive                                                                           | OneDive                                                                         |
-|:---------------------|:---------------------------------------------|:--------------------------------------------------------------------------------------|:--------------------------------------------------------------------------------|
-| real-world_ccsr.ckpt | CCSR model for real-world image restoration. | [download](https://drive.google.com/drive/folders/1jM1mxDryPk9CTuFTvYcraP2XIVzbPiw_?usp=drive_link) | [download](https://pan.baidu.com/s/1uvSvJgcoL_Knj0h22-9TvA?pwd=v3v6)            |
-| bicubic_ccsr.ckpt    | CCSR model for bicubic image restoration.    | download                                                                              | download                                                                        |
+| Model Name           | Description                                  | GoogleDrive                                                                           | OneDive                                                                       |
+|:---------------------|:---------------------------------------------|:--------------------------------------------------------------------------------------|:------------------------------------------------------------------------------|
+| real-world_ccsr.ckpt | CCSR model for real-world image restoration. | [download](https://drive.google.com/drive/folders/1jM1mxDryPk9CTuFTvYcraP2XIVzbPiw_?usp=drive_link) | download                                                 |
+| bicubic_ccsr.ckpt    | CCSR model for bicubic image restoration.    | download                                                                              | download                                                                      |
 
 
 #### Step 2: Prepare testing data
@@ -70,21 +71,22 @@ You can put the testing images in the `preset/test_datasets`.
 #### Step 3: Running testing command
 ```
 python inference_ccsr.py \
---input preset/datasets/test_datasets \
---config CCSR-main/configs/model/ccsr_stage2.yaml \
---ckpt [ccsr_realworld_ckpt_path] \
+--input preset/test_datasets \
+--config configs/model/ccsr_stage2.yaml \
+--ckpt weights/real-world_ccsr.ckpt \
 --steps 45 \
 --sr_scale 4 \
---t_max: 0.6667 \
---t_min: 0.3333 \
---output experiments/output \
+--t_max 0.6667 \
+--t_min 0.3333 \
+--color_fix_type adain \
+--output experiments/test \
 --device cuda \
---repeat_times 1
+--repeat_times 1 
 ```
 You can obtain `N` different SR results by setting `repeat_time` as `N` to test the stability of CCSR. The data folder should be like this:
 
 ```
- experiments/output
+ experiments/test
  ├── sample0   # the first group of SR results 
  └── sample1   # the second group of SR results 
    ...
@@ -97,7 +99,7 @@ You can obtain `N` different SR results by setting `repeat_time` as `N` to test
    Fill in the required information in [cal_iqa.py](cal_iqa/cal_iqa.py) and run, then you can obtain the evaluation results in the folder like this:
    ```
     log_path
-    ├── log_name_npy  # save the IQA values of each restored group to the npy files
+    ├── log_name_npy  # save the IQA values of each restored group as the npy files
     └── log_name.log   # log recode
    ```
 
@@ -117,28 +119,27 @@ You can obtain `N` different SR results by setting `repeat_time` as `N` to test
 1. Generate file list of training set and validation set.
 
     ```shell
-    python scripts/make_file_list.py \
-    --img_folder [hq_dir_path] \
-    --val_size [validation_set_size] \
-    --save_folder [save_dir_path] \
-    --follow_links
+   python scripts/make_file_list.py \
+   --img_folder [hq_dir_path] \
+   --val_size [validation_set_size] \
+   --save_folder [save_dir_path] \
+   --follow_links
     ```
     
     This script will collect all image files in `img_folder` and split them into training set and validation set automatically. You will get two file lists in `save_folder`, each line in a file list contains an absolute path of an image file:
     
     ```
-    save_folder
+    save_dir_path
     ├── train.list # training file list
     └── val.list   # validation file list
     ```
 
 2. Configure training set and validation set.
 
-    For general image restoration, fill in the following configuration files with appropriate values.
+    For real-world image restoration, fill in the following configuration files with appropriate values.
 
     - [training set](configs/dataset/general_deg_stablesr_realesrgan_train.yaml) and [validation set](configs/dataset/general_deg_stablesr_realesrgan_val.yaml) for **Real-ESRGAN** degradation.
-    - [training set](configs/dataset/general_deg_bicubic_train.yaml) and [validation set](configs/dataset/general_deg_bicubic_val.yaml) for **Bicubic** degradation.
-
+ 
 #### Step2: Train Stage1 Model
 1. Download pretrained [Stable Diffusion v2.1](https://huggingface.co/stabilityai/stable-diffusion-2-1-base) to provide generative capabilities.
 
@@ -149,10 +150,10 @@ You can obtain `N` different SR results by setting `repeat_time` as `N` to test
 2. Create the initial model weights.
 
     ```shell
-    python scripts/make_stage2_init_weight.py \
-    --cldm_config configs/model/ccsr_stage1.yaml \
-    --sd_weight [sd_v2.1_ckpt_path] \
-    --output [init_weight_output_path]
+   python scripts/make_stage2_init_weight.py \
+   --cldm_config configs/model/ccsr_stage1.yaml \
+   --sd_weight [sd_v2.1_ckpt_path] \
+   --output weights/init_weight_ccsr.ckpt
     ```
    
 3. Configure training-related information.
@@ -174,10 +175,6 @@ You can obtain `N` different SR results by setting `repeat_time` as `N` to test
    ```shell
     python train.py --config configs/train_ccsr_stage2.yaml
     ```
-### License
-This project is released under the [Apache 2.0 license](LICENSE).
-
-### Acknowledgement
 
 ### Citations
 If our code helps your research or work, please consider citing our paper.
@@ -191,6 +188,13 @@ The following are BibTeX references:
   year={2024}
 }
 ```
+
+### License
+This project is released under the [Apache 2.0 license](LICENSE).
+
+### Acknowledgement
+This project is based on [ControlNet](https://github.com/lllyasviel/ControlNet), [BasicSR](https://github.com/XPixelGroup/BasicSR) and [DiffBIR](https://github.com/XPixelGroup/DiffBIR). Some codes are brought from [StableSR](https://github.com/IceClear/StableSR).Thanks for their awesome works. 
+
 ### Contact
 If you have any questions, please contact: ling-chen.sun@connect.polyu.hk
 
diff --git a/cal_iqa/cal_iqa.py b/cal_iqa/cal_iqa.py
new file mode 100644
index 0000000..bc6d6b4
--- /dev/null
+++ b/cal_iqa/cal_iqa.py
@@ -0,0 +1,236 @@
+# evaluate the restored images with IQA
+# PSNR, SSIM, LPIPS are given as example, you can add more IQA in this file
+
+import cv2
+import argparse, os, sys, glob
+import logging
+from datetime import datetime
+
+import pyiqa
+
+from torch.utils import data as data
+import glob
+import numpy as np
+import math
+import random
+import torch
+
+def get_timestamp():
+    return datetime.now().strftime('%y%m%d-%H%M%S')
+
+def setup_logger(logger_name, root, phase, level=logging.INFO, screen=False, tofile=False):
+    '''set up logger'''
+    lg = logging.getLogger(logger_name)
+    formatter = logging.Formatter('%(asctime)s.%(msecs)03d - %(levelname)s: %(message)s',
+                                  datefmt='%y-%m-%d %H:%M:%S')
+    lg.setLevel(level)
+    if tofile:
+        log_file = os.path.join(root, phase + '_{}.log'.format(get_timestamp()))
+        fh = logging.FileHandler(log_file, mode='w')
+        fh.setFormatter(formatter)
+        lg.addHandler(fh)
+    if screen:
+        sh = logging.StreamHandler()
+        sh.setFormatter(formatter)
+        lg.addHandler(sh)
+
+def dict2str(opt, indent_l=1):
+    '''dict to string for logger'''
+    msg = ''
+    for k in opt:
+        if isinstance(v, dict):
+            msg += ' ' * (indent_l * 2) + k + ':[\n'
+            msg += dict2str(v, indent_l + 1)
+            msg += ' ' * (indent_l * 2) + ']\n'
+        else:
+            msg += ' ' * (indent_l * 2) + k + ': ' + str(v) + '\n'
+    return msg
+
+def img2tensor(imgs, bgr2rgb=True, float32=True):
+    """from BasicSR
+    Numpy array to tensor.
+
+    Args:
+        imgs (list[ndarray] | ndarray): Input images.
+        bgr2rgb (bool): Whether to change bgr to rgb.
+        float32 (bool): Whether to change to float32.
+
+    Returns:
+        list[tensor] | tensor: Tensor images. If returned results only have
+            one element, just return tensor.
+    """
+
+    def _totensor(img, bgr2rgb, float32):
+        if img.shape[2] == 3 and bgr2rgb:
+            if img.dtype == 'float64':
+                img = img.astype('float32')
+            img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+        img = torch.from_numpy(img.transpose(2, 0, 1))
+        if float32:
+            img = img.float()
+        return img
+
+    if isinstance(imgs, list):
+        return [_totensor(img, bgr2rgb, float32) for img in imgs]
+    else:
+        return _totensor(imgs, bgr2rgb, float32)
+
+
+def main():
+	parser = argparse.ArgumentParser()
+
+	parser.add_argument(
+		"--init-imgs",
+		nargs="+",
+		help="path to the input image",
+		default=['****/sample0',
+		'****/sample1',
+		'****/sample2',
+		'****/sample3',
+		'****/sample4',
+		'****/sample5',
+		'****/sample6',
+		'****/sample7',
+		'****/sample8',
+		'****/sample9',
+		],
+	)
+
+	parser.add_argument(
+		"--init-imgs-names",
+		nargs="+",
+		help="name of the input image",
+		default=['****-0', '****-1', '****-2', '****-3', '****-4', '****-5', '****-6', '****-7', '****-8', '****-9'
+		],
+		)
+
+	parser.add_argument(
+		"--gt-imgs",
+		nargs="+",
+		help="path to the gt image, you need to add the paths of gt folders corresponding to init-imgs",
+		default=['****', '****', '****', '****', '****','****',
+		 '****', '****', '****', '****', '****','****',
+		],
+		
+	)
+	
+	parser.add_argument(
+		"--log",
+		type=str,
+		nargs="?",
+		help="path to the log",
+		default='/home/notebook/data/group/SunLingchen/code/CCSR/CCSR-main/experiments')
+
+	parser.add_argument(
+		"--log-name",
+		type=str,
+		nargs="?",
+		help="name of your log",
+		default='test',
+	)
+
+	parser.add_argument(
+		"--num_img",
+		type=int,
+		nargs="?",
+		help="the number of images evaluated in the folder; 0: all the images are evaludated.",
+		default=0,
+	)
+
+	opt = parser.parse_args()
+	device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
+
+	os.makedirs(opt.log, exist_ok=True)
+	# init logger
+	setup_logger('base', opt.log, 'test_' + opt.log_name, level=logging.INFO,
+                  screen=True, tofile=True)
+	logger = logging.getLogger('base')
+	logger.info(opt)
+
+	# init metrics: you can add more metrics here
+
+	iqa_ssim = pyiqa.create_metric('ssim', test_y_channel=True, color_space='ycbcr').to(device)
+	iqa_psnr = pyiqa.create_metric('psnr', test_y_channel=True, color_space='ycbcr').to(device)
+	iqa_lpips = pyiqa.create_metric('lpips', device=device)
+	
+
+
+	for dir_idx in range(len(opt.init_imgs)):
+
+		gt_dir = opt.gt_imgs[dir_idx]
+		img_gt_list = sorted(glob.glob(os.path.join(gt_dir, '*.png')))
+
+		img_sr_dir = opt.init_imgs[dir_idx]
+		img_sr_list = sorted(glob.glob(os.path.join(img_sr_dir, '*.png')))
+		print(opt.init_imgs_names[dir_idx])
+		print(f'GT IMAGES LEN IS {len(img_gt_list)}, SR IMAGES LEN IS {len(img_sr_list)}')
+
+		assert len(img_gt_list) == len(img_sr_list)
+
+
+	for dir_idx in range(len(opt.init_imgs)):
+
+		# record metrics
+		metrics = {}
+		metrics['psnr'], metrics['ssim'], metrics['lpips'] = \
+			 [], [], []
+
+		gt_dir = opt.gt_imgs[dir_idx]
+		img_gt_list = sorted(glob.glob(os.path.join(gt_dir, '*.png')))
+
+		img_sr_dir = opt.init_imgs[dir_idx]
+		img_sr_list = sorted(glob.glob(os.path.join(img_sr_dir, '*.png')))
+
+		if opt.num_img != 0:
+			img_gt_list = img_gt_list[0:opt.num_img]
+			img_sr_list = img_sr_list[0:opt.num_img]
+			
+		PSNR_all, SSIM_all, lpips_all = 0.0, 0.0, 0.0
+		logger.info('\nTesting [{:s}]...'.format(opt.init_imgs_names[dir_idx]))
+
+		for img_idx in range(len(img_sr_list)):
+			img_sr_name = os.path.basename(img_sr_list[img_idx])
+			print(f'Processing {img_sr_name} ...')
+			print(img_sr_list[img_idx])
+			
+			input_sr_img = cv2.imread(img_sr_list[img_idx], cv2.IMREAD_COLOR)
+			sr = img2tensor(input_sr_img, bgr2rgb=True, float32=True).unsqueeze(0).cuda().contiguous()
+				
+			input_gt_img = cv2.imread(img_gt_list[img_idx], cv2.IMREAD_COLOR)
+			hr = img2tensor(input_gt_img, bgr2rgb=True, float32=True).unsqueeze(0).cuda().contiguous()
+
+			if sr.shape != hr.shape:
+				continue
+
+			# PSNR: convert the ycbcr to calculate
+			hr = hr[..., 4:-4, 4:-4]/255.
+			sr = sr[..., 4:-4, 4:-4]/255.
+
+			PSNR_now = iqa_psnr(sr, hr).item()
+			PSNR_all += PSNR_now
+			metrics['psnr'].append(PSNR_now)
+
+			# SSIM
+			ssim_now = iqa_ssim(sr, hr).item()
+			SSIM_all += ssim_now
+			metrics['ssim'].append(ssim_now)
+
+			# lpips
+			lpips_now = iqa_lpips(sr, hr).item()
+			lpips_all += lpips_now
+			metrics['lpips'].append(lpips_now)
+
+			logger.info('{:20s}_{} - PSNR: {:.6f} dB; SSIM: {:.6f};  LPIPS: {:.6f}'.format(opt.init_imgs_names[dir_idx], img_sr_name, PSNR_now, ssim_now, lpips_now))
+
+		PSNR_all = round(PSNR_all/len(img_sr_list) , 4)
+		SSIM_all = round(SSIM_all/len(img_sr_list) , 4)
+		lpips_all = round(lpips_all/len(img_sr_list) , 4)
+
+		logger.info('{:20s}_all - PSNR: {:.6f} dB; SSIM: {:.6f};  LPIPS: {:.6f}'.format(opt.init_imgs_names[dir_idx], PSNR_all, SSIM_all, lpips_all))
+
+		# save metrics
+		npy_path = os.path.join(opt.log, 'test_' + opt.log_name + '_npy')
+		os.makedirs(npy_path, exist_ok=True)
+		np.save(npy_path + '/' + opt.init_imgs_names[dir_idx]+'.npy', metrics)
+
+main()
\ No newline at end of file
diff --git a/cal_iqa/iqa_G-STD.py b/cal_iqa/iqa_G-STD.py
new file mode 100644
index 0000000..bf3fda8
--- /dev/null
+++ b/cal_iqa/iqa_G-STD.py
@@ -0,0 +1,111 @@
+# evaluate the G-STD of restored N images
+
+
+import os
+import numpy as np
+import glob
+from datetime import datetime
+import logging
+
+def get_timestamp():
+    return datetime.now().strftime('%y%m%d-%H%M%S')
+
+def setup_logger(logger_name, root, phase, level=logging.INFO, screen=False, tofile=False):
+    '''set up logger'''
+    lg = logging.getLogger(logger_name)
+    formatter = logging.Formatter('%(asctime)s.%(msecs)03d - %(levelname)s: %(message)s',
+                                  datefmt='%y-%m-%d %H:%M:%S')
+    lg.setLevel(level)
+    if tofile:
+        log_file = os.path.join(root, phase + '_{}.log'.format(get_timestamp()))
+        fh = logging.FileHandler(log_file, mode='w')
+        fh.setFormatter(formatter)
+        lg.addHandler(fh)
+    if screen:
+        sh = logging.StreamHandler()
+        sh.setFormatter(formatter)
+        lg.addHandler(sh)
+
+def img2tensor(imgs, bgr2rgb=True, float32=True):
+    """from BasicSR
+    Numpy array to tensor.
+
+    Args:
+        imgs (list[ndarray] | ndarray): Input images.
+        bgr2rgb (bool): Whether to change bgr to rgb.
+        float32 (bool): Whether to change to float32.
+
+    Returns:
+        list[tensor] | tensor: Tensor images. If returned results only have
+            one element, just return tensor.
+    """
+
+    def _totensor(img, bgr2rgb, float32):
+        if img.shape[2] == 3 and bgr2rgb:
+            if img.dtype == 'float64':
+                img = img.astype('float32')
+            img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+        img = torch.from_numpy(img.transpose(2, 0, 1))
+        if float32:
+            img = img.float()
+        return img
+
+    if isinstance(imgs, list):
+        return [_totensor(img, bgr2rgb, float32) for img in imgs]
+    else:
+        return _totensor(imgs, bgr2rgb, float32)
+
+# log_name
+name_log = 'DIV2K-valid'
+path_log = '****/G-STD'
+os.makedirs(path_log, exist_ok=True)
+
+# load metrics
+npy_file_path = '****'
+npy_file_lists = sorted(glob.glob(os.path.join(npy_file_path, '*npy')))
+l_sample = len(npy_file_lists)
+a = np.load(npy_file_lists[0], allow_pickle=True).item()
+l_file = len(a['psnr'])
+
+# init logger
+setup_logger('base', path_log, 'test_' + name_log, level=logging.INFO, screen=True, tofile=True)
+logger = logging.getLogger('base')
+logger.info(name_log)
+
+# init the metrics: you can add other metrics here
+metric_psnr = np.zeros([l_file, l_sample])
+metric_ssim = np.zeros([l_file, l_sample])
+metric_lpips = np.zeros([l_file, l_sample])
+
+i = 0
+for npy_file in npy_file_lists:
+    # npy_file_list = sorted(glob.glob(os.path.join(npy_files, '*')))
+    a = np.load(npy_file, allow_pickle=True).item()
+    metric_psnr[:, i] = np.array(a['psnr'])
+    metric_ssim[:, i] = np.array(a['ssim'])
+    metric_lpips[:, i] = np.array(a['lpips'])
+    i = i + 1
+        
+# calculate the mean of the metrics
+mean_psnr = np.mean(metric_psnr, axis=1)
+mean_ssim = np.mean(metric_ssim, axis=1)
+mean_lpips = np.mean(metric_lpips, axis=1)
+
+mean_mean_psnr = np.mean(mean_psnr)
+mean_mean_ssim = np.mean(mean_ssim)
+mean_mean_lpips = np.mean(mean_lpips)
+
+
+# calculate the std of the metrics
+std_psnr = np.std(metric_psnr, axis=1)
+std_ssim = np.std(metric_ssim, axis=1)
+std_lpips = np.std(metric_lpips, axis=1)
+
+mean_std_psnr = np.mean(std_psnr)
+mean_std_ssim = np.mean(std_ssim)
+mean_std_lpips = np.mean(std_lpips)
+
+
+logger.info('mean_mean - PSNR: {:.6f} dB; SSIM: {:.6f};  LPIPS: {:.6f}'.format(mean_mean_psnr, mean_mean_ssim, mean_mean_lpips))
+
+logger.info('G_STD - PSNR: {:.6f} dB; SSIM: {:.6f};  LPIPS: {:.6f}'.format(mean_std_psnr, mean_std_ssim, mean_std_lpips))
\ No newline at end of file
diff --git a/cal_iqa/iqa_L-STD.py b/cal_iqa/iqa_L-STD.py
new file mode 100644
index 0000000..ab56658
--- /dev/null
+++ b/cal_iqa/iqa_L-STD.py
@@ -0,0 +1,190 @@
+# evaluate the L-STD of restored N images
+
+
+import cv2
+import argparse, os, sys, glob
+import logging
+from datetime import datetime
+
+from torch.utils import data as data
+import glob
+import numpy as np
+import math
+import random
+import torch
+
+
+def img2tensor(imgs, bgr2rgb=True, float32=True):
+    """from BasicSR
+    Numpy array to tensor.
+
+    Args:
+        imgs (list[ndarray] | ndarray): Input images.
+        bgr2rgb (bool): Whether to change bgr to rgb.
+        float32 (bool): Whether to change to float32.
+
+    Returns:
+        list[tensor] | tensor: Tensor images. If returned results only have
+            one element, just return tensor.
+    """
+
+    def _totensor(img, bgr2rgb, float32):
+        if img.shape[2] == 3 and bgr2rgb:
+            if img.dtype == 'float64':
+                img = img.astype('float32')
+            img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+        img = torch.from_numpy(img.transpose(2, 0, 1))
+        if float32:
+            img = img.float()
+        return img
+
+    if isinstance(imgs, list):
+        return [_totensor(img, bgr2rgb, float32) for img in imgs]
+    else:
+        return _totensor(imgs, bgr2rgb, float32)
+
+
+def get_timestamp():
+    return datetime.now().strftime('%y%m%d-%H%M%S')
+
+def setup_logger(logger_name, root, phase, level=logging.INFO, screen=False, tofile=False):
+    '''set up logger'''
+    lg = logging.getLogger(logger_name)
+    formatter = logging.Formatter('%(asctime)s.%(msecs)03d - %(levelname)s: %(message)s',
+                                  datefmt='%y-%m-%d %H:%M:%S')
+    lg.setLevel(level)
+    if tofile:
+        log_file = os.path.join(root, phase + '_{}.log'.format(get_timestamp()))
+        fh = logging.FileHandler(log_file, mode='w')
+        fh.setFormatter(formatter)
+        lg.addHandler(fh)
+    if screen:
+        sh = logging.StreamHandler()
+        sh.setFormatter(formatter)
+        lg.addHandler(sh)
+
+def dict2str(opt, indent_l=1):
+    '''dict to string for logger'''
+    msg = ''
+    for k in opt:
+        if isinstance(v, dict):
+            msg += ' ' * (indent_l * 2) + k + ':[\n'
+            msg += dict2str(v, indent_l + 1)
+            msg += ' ' * (indent_l * 2) + ']\n'
+        else:
+            msg += ' ' * (indent_l * 2) + k + ': ' + str(v) + '\n'
+    return msg
+
+def calc_psnr(sr, hr):
+    sr, hr = sr.double(), hr.double()
+    diff = (sr - hr) / 255.00
+    mse  = diff.pow(2).mean()
+    psnr = -10 * math.log10(mse)                    
+    return float(psnr)
+
+def rgb_to_ycbcr(image: torch.Tensor) -> torch.Tensor:
+    r"""Convert an RGB image to YCbCr.
+
+    Args:
+        image (torch.Tensor): RGB Image to be converted to YCbCr.
+
+    Returns:
+        torch.Tensor: YCbCr version of the image.
+    """
+
+    if not torch.is_tensor(image):
+        raise TypeError("Input type is not a torch.Tensor. Got {}".format(type(image)))
+
+    if len(image.shape) < 3 or image.shape[-3] != 3:
+        raise ValueError("Input size must have a shape of (*, 3, H, W). Got {}".format(image.shape))
+
+    image = image / 255. ## image in range (0, 1)
+    r: torch.Tensor = image[..., 0, :, :]
+    g: torch.Tensor = image[..., 1, :, :]
+    b: torch.Tensor = image[..., 2, :, :]
+
+    y: torch.Tensor = 65.481 * r + 128.553 * g + 24.966 * b + 16.0
+    cb: torch.Tensor = -37.797 * r + -74.203 * g + 112.0 * b + 128.0
+    cr: torch.Tensor = 112.0 * r + -93.786 * g + -18.214 * b + 128.0
+
+    return torch.stack((y, cb, cr), -3)
+
+def main():
+	parser = argparse.ArgumentParser()
+
+	parser.add_argument(
+		"--init-imgs",
+		nargs="+",
+		help="path to the input image",
+		default=['****/sample0',
+		'****/sample1',
+		'****/sample2',
+		'****/sample3',
+		'****/sample4',
+		'****/sample5',
+		'****/sample6',
+		'****/sample7',
+		'****/sample8',
+		'****/sample9',
+		]
+	)
+
+	
+
+	opt = parser.parse_args()
+	device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
+
+	name_log = 'ours'
+	path_log = '****/L-STD'
+	os.makedirs(path_log, exist_ok=True)
+
+	# init logger
+	setup_logger('base', path_log, 'test_' + name_log, level=logging.INFO, screen=True, tofile=True)
+	logger = logging.getLogger('base')
+	logger.info(name_log)
+
+	num_samples = len(opt.init_imgs)
+
+	# combine all the sampled data
+	
+	img_sr_dirs = []
+	for dir_idx in range(num_samples):
+		img_sr_dir = opt.init_imgs[dir_idx]
+		img_sr_dirs.append(img_sr_dir)
+
+	img_sr_list = sorted(glob.glob(os.path.join(img_sr_dir, '*.png')))
+	num_imgs = len(img_sr_list)
+	input_sr_img = cv2.imread(img_sr_list[0], cv2.IMREAD_COLOR)
+	sr = img2tensor(input_sr_img, bgr2rgb=True, float32=True).unsqueeze(0).cuda().contiguous()
+
+	sr = sr[..., 4:-4, 4:-4]/255. #[0,1]
+	B, C, H, W = sr.shape
+	sr = sr[0]
+	
+	std_alls = []
+	
+	for img_idx in range(len(img_sr_list)):
+		d = torch.zeros([num_samples,C,H,W]) 
+		for dir_idx in range(num_samples): 
+			img_sr_dir = opt.init_imgs[dir_idx]
+			img_sr_list = sorted(glob.glob(os.path.join(img_sr_dir, '*.png')))
+			
+			input_sr_img = cv2.imread(img_sr_list[img_idx], cv2.IMREAD_COLOR)
+			img_sr_name = os.path.basename(img_sr_list[img_idx])
+
+			sr = img2tensor(input_sr_img, bgr2rgb=True, float32=True).unsqueeze(0).cuda().contiguous()
+
+			sr = sr[..., 4:-4, 4:-4]/255. #[0,1]
+			sr = sr[0]
+			d[dir_idx,...] = sr
+
+		d = np.array(d)
+		stds = np.std(d, axis=0)
+		stds = np.mean(stds)
+		logger.info('_{} - L-STD: {:.6f}'.format(img_sr_name, stds))
+		std_alls.append(stds)
+	std_all = np.mean(std_alls)
+	print(std_all)
+	logger.info('_all - L-STD: {:.6f}; '.format(std_all))
+
+main()
\ No newline at end of file
diff --git a/configs/dataset/general_deg_stablesr_realesrgan_train.yaml b/configs/dataset/general_deg_stablesr_realesrgan_train.yaml
new file mode 100644
index 0000000..ba0a4e3
--- /dev/null
+++ b/configs/dataset/general_deg_stablesr_realesrgan_train.yaml
@@ -0,0 +1,63 @@
+dataset:
+  target: dataset.realesrgan.RealESRGANDataset
+  params:
+    # Path to the file list.
+    file_list: preset/train_datasets/train.list
+    out_size: 512
+    crop_type: center
+
+    use_hflip: true
+    use_rot: false
+   
+    blur_kernel_size: 21
+    kernel_list: ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso']
+    kernel_prob: [0.45, 0.25, 0.12, 0.03, 0.12, 0.03]
+    sinc_prob: 0.1
+    blur_sigma: [0.2, 1.5]
+    betag_range: [0.5, 2.0]
+    betap_range: [1, 1.5]
+
+    blur_kernel_size2: 11
+    kernel_list2: ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso']
+    kernel_prob2: [0.45, 0.25, 0.12, 0.03, 0.12, 0.03]
+    sinc_prob2: 0.1
+    blur_sigma2: [0.2, 1.0]
+    betag_range2: [0.5, 2.0]
+    betap_range2: [1, 1.5]
+
+    final_sinc_prob: 0.8
+
+data_loader:
+  batch_size: 16
+  shuffle: true
+  num_workers: 16
+  prefetch_factor: 2
+  drop_last: true
+
+batch_transform:
+  target: dataset.batch_transform.RealESRGANBatchTransform
+  params:
+    use_sharpener: false
+    resize_hq: false
+    # Queue size of training pool, this should be multiples of batch_size.
+    queue_size: 192
+
+    # the first degradation process
+    resize_prob: [0.2, 0.7, 0.1]  # up, down, keep
+    resize_range: [0.3, 1.5]
+    gaussian_noise_prob: 0.5
+    noise_range: [1, 15]
+    poisson_scale_range: [0.05, 2.0]
+    gray_noise_prob: 0.4
+    jpeg_range: [60, 95]
+
+    # the second degradation process
+    stage2_scale: 4
+    second_blur_prob: 0.5
+    resize_prob2: [0.3, 0.4, 0.3]  # up, down, keep
+    resize_range2: [0.6, 1.2]
+    gaussian_noise_prob2: 0.5
+    noise_range2: [1, 12]
+    poisson_scale_range2: [0.05, 1.0]
+    gray_noise_prob2: 0.4
+    jpeg_range2: [60, 100]
diff --git a/configs/dataset/general_deg_stablesr_realesrgan_val.yaml b/configs/dataset/general_deg_stablesr_realesrgan_val.yaml
new file mode 100644
index 0000000..92ae1f8
--- /dev/null
+++ b/configs/dataset/general_deg_stablesr_realesrgan_val.yaml
@@ -0,0 +1,64 @@
+dataset:
+  target: dataset.realesrgan.RealESRGANDataset
+  params:
+    # Path to the file list.
+    file_list: preset/train_datasets/val.list
+
+    out_size: 512
+    crop_type: center
+
+    use_hflip: true
+    use_rot: false
+
+    blur_kernel_size: 21
+    kernel_list: ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso']
+    kernel_prob: [0.45, 0.25, 0.12, 0.03, 0.12, 0.03]
+    sinc_prob: 0.1
+    blur_sigma: [0.2, 1.5]
+    betag_range: [0.5, 2.0]
+    betap_range: [1, 1.5]
+
+    blur_kernel_size2: 11
+    kernel_list2: ['iso', 'aniso', 'generalized_iso', 'generalized_aniso', 'plateau_iso', 'plateau_aniso']
+    kernel_prob2: [0.45, 0.25, 0.12, 0.03, 0.12, 0.03]
+    sinc_prob2: 0.1
+    blur_sigma2: [0.2, 1.0]
+    betag_range2: [0.5, 2.0]
+    betap_range2: [1, 1.5]
+
+    final_sinc_prob: 0.8
+
+data_loader:
+  batch_size: 1
+  shuffle: false
+  num_workers: 16
+  prefetch_factor: 2
+  drop_last: true
+
+batch_transform:
+  target: dataset.batch_transform.RealESRGANBatchTransform
+  params:
+    use_sharpener: false
+    resize_hq: false
+    # Queue size of training pool, this should be multiples of batch_size.
+    queue_size: 256
+    
+    # the first degradation process
+    resize_prob: [0.2, 0.7, 0.1]  # up, down, keep
+    resize_range: [0.3, 1.5]
+    gaussian_noise_prob: 0.5
+    noise_range: [1, 15]
+    poisson_scale_range: [0.05, 2.0]
+    gray_noise_prob: 0.4
+    jpeg_range: [60, 95]
+
+    # the second degradation process
+    stage2_scale: 4
+    second_blur_prob: 0.5
+    resize_prob2: [0.3, 0.4, 0.3]  # up, down, keep
+    resize_range2: [0.6, 1.2]
+    gaussian_noise_prob2: 0.5
+    noise_range2: [1, 12]
+    poisson_scale_range2: [0.05, 1.0]
+    gray_noise_prob2: 0.4
+    jpeg_range2: [60, 100]
\ No newline at end of file
diff --git a/configs/model/ccsr_stage1.yaml b/configs/model/ccsr_stage1.yaml
new file mode 100644
index 0000000..c097546
--- /dev/null
+++ b/configs/model/ccsr_stage1.yaml
@@ -0,0 +1,108 @@
+target: model.ccsr_stage1.ControlLDM
+params:
+  linear_start: 0.00085
+  linear_end: 0.0120
+  num_timesteps_cond: 1
+  log_every_t: 200
+  timesteps: 1000
+  t_max: 0.6667
+  t_min: 0.3333
+  first_stage_key: "jpg"
+  cond_stage_key: "txt"
+  control_key: "hint"
+  image_size: 64
+  channels: 4
+  cond_stage_trainable: false
+  conditioning_key: crossattn
+  monitor: val/loss_simple_ema
+  scale_factor: 0.18215
+  use_ema: False
+  
+  sd_locked: True
+  only_mid_control: False
+  # Learning rate.
+  learning_rate: 1e-4
+  
+  control_stage_config:
+    target: model.ccsr_stage1.ControlNet
+    params:
+      use_checkpoint: True
+      image_size: 32 # unused
+      in_channels: 4
+      hint_channels: 4
+      model_channels: 320
+      attention_resolutions: [ 4, 2, 1 ]
+      num_res_blocks: 2
+      channel_mult: [ 1, 2, 4, 4 ]
+      num_head_channels: 64 # need to fix for flash-attn
+      use_spatial_transformer: True
+      use_linear_in_transformer: True
+      transformer_depth: 1
+      context_dim: 1024
+      legacy: False
+
+  unet_config:
+    target: model.ccsr_stage1.ControlledUnetModel
+    params:
+      use_checkpoint: True
+      image_size: 32 # unused
+      in_channels: 4
+      out_channels: 4
+      model_channels: 320
+      attention_resolutions: [ 4, 2, 1 ]
+      num_res_blocks: 2
+      channel_mult: [ 1, 2, 4, 4 ]
+      num_head_channels: 64 # need to fix for flash-attn
+      use_spatial_transformer: True
+      use_linear_in_transformer: True
+      transformer_depth: 1
+      context_dim: 1024
+      legacy: False
+
+  first_stage_config:
+    target: ldm.models.autoencoder.AutoencoderKL
+    params:
+      embed_dim: 4
+      monitor: val/rec_loss
+      ddconfig:
+        #attn_type: "vanilla-xformers"
+        double_z: true
+        z_channels: 4
+        resolution: 256
+        in_channels: 3
+        out_ch: 3
+        ch: 128
+        ch_mult:
+        - 1
+        - 2
+        - 4
+        - 4
+        num_res_blocks: 2
+        attn_resolutions: []
+        dropout: 0.0
+      lossconfig:
+        target: torch.nn.Identity
+
+  cond_stage_config:
+    target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder
+    params:
+      freeze: True
+      layer: "penultimate"
+
+  # preprocess_config:
+  #   target: model.swinir.SwinIR
+  #   params:
+  #     img_size: 64
+  #     patch_size: 1
+  #     in_chans: 3
+  #     embed_dim: 180
+  #     depths: [6, 6, 6, 6, 6, 6, 6, 6]
+  #     num_heads: [6, 6, 6, 6, 6, 6, 6, 6]
+  #     window_size: 8
+  #     mlp_ratio: 2
+  #     sf: 8
+  #     img_range: 1.0
+  #     upsampler: "nearest+conv"
+  #     resi_connection: "1conv"
+  #     unshuffle: True
+  #     unshuffle_scale: 8
diff --git a/configs/model/ccsr_stage2.yaml b/configs/model/ccsr_stage2.yaml
new file mode 100644
index 0000000..51a2923
--- /dev/null
+++ b/configs/model/ccsr_stage2.yaml
@@ -0,0 +1,118 @@
+target: model.ccsr_stage2.ControlLDM
+params:
+  linear_start: 0.00085
+  linear_end: 0.0120
+  num_timesteps_cond: 1
+  log_every_t: 200
+  timesteps: 1000
+  t_max: 0.6667
+  t_min: 0.3333
+  first_stage_key: "jpg"
+  cond_stage_key: "txt"
+  control_key: "hint"
+  image_size: 64
+  channels: 4
+  cond_stage_trainable: false
+  conditioning_key: crossattn
+  monitor: val/loss_simple_ema
+  scale_factor: 0.18215
+  use_ema: False
+  
+  sd_locked: True
+  only_mid_control: False
+  # Learning rate.
+  learning_rate: 5e-6
+  
+  control_stage_config:
+    target: model.ccsr_stage2.ControlNet
+    params:
+      use_checkpoint: True
+      image_size: 32 # unused
+      in_channels: 4
+      hint_channels: 4
+      model_channels: 320
+      attention_resolutions: [ 4, 2, 1 ]
+      num_res_blocks: 2
+      channel_mult: [ 1, 2, 4, 4 ]
+      num_head_channels: 64 # need to fix for flash-attn
+      use_spatial_transformer: True
+      use_linear_in_transformer: True
+      transformer_depth: 1
+      context_dim: 1024
+      legacy: False
+
+  unet_config:
+    target: model.ccsr_stage2.ControlledUnetModel
+    params:
+      use_checkpoint: True
+      image_size: 32 # unused
+      in_channels: 4
+      out_channels: 4
+      model_channels: 320
+      attention_resolutions: [ 4, 2, 1 ]
+      num_res_blocks: 2
+      channel_mult: [ 1, 2, 4, 4 ]
+      num_head_channels: 64 # need to fix for flash-attn
+      use_spatial_transformer: True
+      use_linear_in_transformer: True
+      transformer_depth: 1
+      context_dim: 1024
+      legacy: False
+
+  first_stage_config:
+    target: ldm.models.autoencoder.AutoencoderKL
+    params:
+      embed_dim: 4
+      monitor: val/rec_loss
+      ddconfig:
+        #attn_type: "vanilla-xformers"
+        double_z: true
+        z_channels: 4
+        resolution: 256
+        in_channels: 3
+        out_ch: 3
+        ch: 128
+        ch_mult:
+        - 1
+        - 2
+        - 4
+        - 4
+        num_res_blocks: 2
+        attn_resolutions: []
+        dropout: 0.0
+      lossconfig:
+        target: torch.nn.Identity
+
+  cond_stage_config:
+    target: ldm.modules.encoders.modules.FrozenOpenCLIPEmbedder
+    params:
+      freeze: True
+      layer: "penultimate"
+  
+  lossconfig:
+    target: ldm.modules.losses.LPIPSWithDiscriminator
+    params:
+      disc_start: 1.0
+      kl_weight: 0
+      perceptual_weight: 1.0
+      disc_weight: 0.5
+      disc_factor: 1.0
+      
+
+  # preprocess_config:
+  #   target: model.swinir.SwinIR
+  #   params:
+  #     img_size: 64
+  #     patch_size: 1
+  #     in_chans: 3
+  #     embed_dim: 180
+  #     depths: [6, 6, 6, 6, 6, 6, 6, 6]
+  #     num_heads: [6, 6, 6, 6, 6, 6, 6, 6]
+  #     window_size: 8
+  #     mlp_ratio: 2
+  #     sf: 8
+  #     img_range: 1.0
+  #     upsampler: "nearest+conv"
+  #     resi_connection: "1conv"
+  #     unshuffle: True
+  #     unshuffle_scale: 8
\ No newline at end of file
diff --git a/configs/train_ccsr_stage1.yaml b/configs/train_ccsr_stage1.yaml
new file mode 100644
index 0000000..4987365
--- /dev/null
+++ b/configs/train_ccsr_stage1.yaml
@@ -0,0 +1,47 @@
+data:
+  target: dataset.data_module.BIRDataModule
+  params:
+    # Path to training set configuration file.
+    train_config: configs/dataset/general_deg_stablesr_realesrgan_train.yaml
+    # Path to validation set configuration file.
+    val_config: configs/dataset/general_deg_stablesr_realesrgan_val.yaml
+
+model:
+  # You can set learning rate in the following configuration file.
+  config: configs/model/ccsr_stage1.yaml
+  # Path to the checkpoints or weights you want to resume. At the begining, 
+  # this should be set to the initial weights created by scripts/make_stage2_init_weight.py.
+  resume: weights/init_weight_ccsr.ckpt
+
+lightning:
+  seed: 231
+  
+  trainer:
+    accelerator: ddp
+    precision: 32
+    # Indices of GPUs used for training.
+    gpus: [0,1,2,3,]
+    # Path to save logs and checkpoints.
+    default_root_dir: experiments/test_ccsr_stage1
+    # Max number of training steps (batches).
+    max_steps: 25001
+    # Validation frequency in terms of training steps.
+    val_check_interval: 390
+    log_every_n_steps: 50
+    # Accumulate gradients from multiple batches so as to increase batch size.
+    accumulate_grad_batches: 3
+  
+  callbacks:
+    - target: model.callbacks.ImageLogger
+      params:
+        # Log frequency of image logger.
+        log_every_n_steps: 150000
+        max_images_each_step: 4
+        log_images_kwargs: ~
+
+    - target: model.callbacks.ModelCheckpoint
+      params:
+        # Frequency of saving checkpoints.
+        every_n_train_steps: 500
+        save_top_k: -1
+        filename: "{step}"
diff --git a/configs/train_ccsr_stage2.yaml b/configs/train_ccsr_stage2.yaml
new file mode 100644
index 0000000..9671fa1
--- /dev/null
+++ b/configs/train_ccsr_stage2.yaml
@@ -0,0 +1,49 @@
+data:
+  target: dataset.data_module.BIRDataModule
+  params:
+    # Path to training set configuration file.
+    train_config: configs/dataset/general_deg_stablesr_realesrgan_train.yaml
+    # Path to validation set configuration file.
+    val_config: configs/dataset/general_deg_stablesr_realesrgan_val.yaml
+
+model:
+  # You can set learning rate in the following configuration file.
+  config: configs/model/ccsr_stage2.yaml
+  # Path to the checkpoints or weights you want to resume. At the begining, 
+  # this should be set to the weights obtained from ccsr_stage1
+  resume: /ccsr_stage1.ckpt
+
+
+lightning:
+  seed: 231
+  
+  trainer:
+    accelerator: ddp
+    precision: 32
+    # Indices of GPUs used for training.
+    gpus: [0,1,2,3,]
+    # Path to save logs and checkpoints.
+    default_root_dir: experiments/test_ccsr_stage2
+    # Max number of training steps (batches).
+    max_steps: 60
+    # Validation frequency in terms of training steps.
+    val_check_interval: 390
+    log_every_n_steps: 50
+    # 50
+    # Accumulate gradients from multiple batches so as to increase batch size.
+    accumulate_grad_batches: 8
+  
+  callbacks:
+    - target: model.callbacks.ImageLogger
+      params:
+        # Log frequency of image logger.
+        log_every_n_steps: 150000
+        max_images_each_step: 4
+        log_images_kwargs: ~
+
+    - target: model.callbacks.ModelCheckpoint
+      params:
+        # Frequency of saving checkpoints.
+        every_n_train_steps: 20
+        save_top_k: -1
+        filename: "{step}"
diff --git a/dataset/__pycache__/batch_transform.cpython-310.pyc b/dataset/__pycache__/batch_transform.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6bfee369184e28fc27a045af1f0593ceee053bc0
GIT binary patch
literal 8226
zcmbVR>vP*yb_Xs%5Ck8h9;PK(c91v^GwVo>nr)k|o0lDXo!NDFB4?*%vqd3ZP$EqN
z^aU8(D%jmb&Sv`XZoh9kBkec$LqD`L{S*2}-05W6>4%Q{+%|n?yK#T#f|Mv(w$qYx
z5AOS%dvMPaQMv4B_<j1y4+H0Gn)Wx;nEaV&ynzz`F#y-N-qm_~L+`N$>lqEBXEw}U
zu953m4NIqS*3CEa>S;G@JdJLl=QN6=u~MUqKC@ftO*N)^)kd{9-I!Kmx$aDFwlUkA
zYs~4|WsO@re@El_ZPqx3mdy)j6@p{iMx%zd!;5GaRr@&FC0<6mtlIOOt!eejebB4b
zEor<N-j;UsK_I%3&!u^>-ArWey|5jH()>m{PNcmS+!zF5Gq_sUrPCUO%>;vfw{A%L
zy|r(yZTg}w?ThVJ>z!}CeOcPAb~g#c>cu*fOTrI%)bo7Ky$yd5$8A6K!l)ewUO$m1
zCPw?{jHCU*CXqM!%c1cGO8g8!pfz-^HCSM9Juo>7%$CUwZr(As^+t~8a6n5|-tv>?
z<~1ORTaoC=+&Z<TC4yui!Y0E$;~&*^lnW>^y;82NY4`N2_q2OVnsLx=i7E7onKAhM
zot4d~7p#O)60Aqj^_6DCgOz?Da0b7-vi9n^SAKEz)vspV7cQ*5yTbj%kAq}I5#c3c
zs+RlUzUMXlZrAgwgq~<I+1Bxc)Gv?cT*b+-AoP2I=Sj!&dJ!LVsb2ED8w0<ajR=EC
zrM*Orz#ipU?HKh{0I0+kfUPk$udBX?c)xeaPhsMi_u#w%&KqOSH_Q>=Ygc#>Chg?*
z!4qU0q$56_^{Kw5iD^vBD4pG_sUs-D=|sm-GG6IhPaTbMf@e8F^Tj-Y1p<%f%p#@`
zCUqBVt;PiPJj9nX=%imvmuH_Z4dM&#BZ?W|HU?IZUx3Jtp7|ipZIr^O?Tns9Jad>=
z!dN-L473U6T5LgE&<OI%XztYLS;gFGl$lX`cJ!RXllD5swbq$Ww3g0m{P>+xqYyYd
z#a)fh^MyMaWTiA}F9zkH0&GraDyR@9h3(ixZ#Af7y=v6tC-{<*w&_naQ)|pX;!eu>
zcLTrs_S(Cb-@JP8-T!q3402rjt|TwSZW2Xq+>LIc;`T+<4C2`JT_qWA)N<QBe<O(9
zLC6E)?r-VxU_<zjBWNGB7jAygb)(R|@U8dUGwZiqk}=X^fvBH#Z*I1mn{Lm)?KY!0
zanCKUy1(td?8ZSe3b_kW&G0O%t=1y<X5a?5g61Fz++;JDM7BpGP+bNpzXko~5~6mv
z@i5_PL+CyCdLY7}8@nwL^(JV`Fv{`(S}-Q%8H9n51rfyUt-#y7AswjO0mjgj#Zfod
zv@T1d29(D}Ag92M+hc7?!?TA~EOiz$b(<}j9jn<G3McTgPN&lkHb(Vo#vX6II}l!m
zGfwa}Rz0<1j{+K6mE}o#R%P`NL924^2#Qu^X@VroIfABDS==XTRVbw~m&uxfR41im
zpu~R%kZ9Z5j=l>yOwCmDSegSgQVY;b^MJY32DH)wU_NyK?X(D3NK1fDQcO!pDJi$u
zHj6K+R%M%suO(AyX<O&|b*fi6<Hj8=EnS4@Z0qsGcVMfNX`1mP?T^^D{*j)RvD-{q
z0h~>z0O!&w;IVWXu$ImM9#3Zh=TjYUA=Nqyf+vgH8d@hpgR5uC&{|3jz>~=-p5xYC
zn*9`6rF#DAeNtXcvRC6Db&!FKQ9dIj{UP@Gjx*E~E!7h}VRWu<gF81E?V;4`JwrO}
zxE<pC!nVs?3l}<(MmtPk^Vsq^so$Eg8l(+n;rs^f>aD(b?Fu2HH^9EPq;0*jFQf<M
z9vZkEwiC}=KHrUS<>J?t_qN!YpS)Ra(+^!(Q+GXZFTu~KaQG8}AVcD>cZ0esbMXK|
zOSUiDQn)}bo+j`Nfzt$@CGZ@9=Lx((fV8yuIRdW`_$q-sfHb;6C=FO%X(qR^cb!IF
zB=8jiX9>JSV41)Qfs+IZ03^P>eeo5<bnz4c@)xB=*B3j^3A(D;vi&J8^?g~&pG?&~
zT^8UTbp1H?yb1gM%Rob{iQfWnY8JC}%RtHgXq6ayU^DhbiB)y;LCG+GW?6^lIJ#P=
zSM{x@pZJ;&TuNdS@%aMEXMc(7ip#Eq^nG)v582Qdnh7pC*CBFW$C#1oH8{xES?ofv
zaQQpNk04+v2^ajjkMzXcfmmZ6gJ_t$<_^TWqwSJtK*T#XDFE!98}0q**#%Al1|`s;
zVnDe69D73@bc|hnNW11!Yk@?O_AQL+P#v8-R_Y%!h-B^{Y?%w%PH9&gsufS11!V+Y
zVE*IxJWu2HXdY4h2;7)QN={Qs&L%%>a(~pbv+BpSgH&hdXqgA`e98Gviv5U>gzui^
z)62h-xNX>X*iU%4kW<*lzTXyZKZ?38yl}T2Hbvk=?QCRLwvE6ehOdZ$u_ax0lm#Vk
zL1~k*`K2}8cCVde2>9MoxD4^_5J;1DbJA|dlbgDifIYb7_xj!7tQ)ImHenB2=72P#
zL6<94wCR5UKVYnf+|1ayWC)3rmfuVw5ih&fHle3*Hf&2i1Q0~GaX8|`Lp<2S1<z;1
ziih2e{gX{Hh4?`~X=m<1R|qg%+bXK2OWuOJb=r+L75z}7*arF?%F~5wl$}jYx_YlH
zS@#Cw8gel5n||ykiBQT*=5Qp$NV%;*M6qOOT?P)xHj>5&Eoy~~fPMmP3>Sh!bt1ZI
zcB3%3S5OL1>g!T(N_|u6H>K{us_<J4r+dTW6iN<)hO?23>xYfZ+<3sl)?N^*1wQEz
zQsgrQM}L+r=`;Eg)L2O`87TVm?0NkRI|Eha=;w@cq^w?gw6YozF}c4AP7cH+6nBiZ
zS9}VlE`=MgpbeopItWbw^&OV7T^%k5v>Mgan6cNZbNJCiXf<x&2|dQ4BZj$@?U?Yu
zVbnjy`1@>V4f6?lQ)8PA?Zn#2Zy>P$F|=C8-eugznnGIO1!$>nCWVf*<8YGZB6k+F
zT{d*WX^a*z`hGU5#;}SzYLszj=wR0pb}ju_2bIO7+^IlQIBAiW?;60Vc2&A)>bfRA
zOsb?uI@7$u;lgxgc=c{>XotmQR`t)RK8*2c>QC*RV<;xKcd>h?hSB3c(1xXCerF-I
zk(zLJfdjn;R+sq9T{`7H#zkQKGo0T}O?Cbgn7Ndkg!A<mR0Ds?!^-Gn<+Oq`O@(9|
zmWETwDL8mjC*X{smd|R84XeZH16+LS5Ep-v?WVZsKE%ZbBV1a?CEh=Yi@;n>ow<yL
z>8#db;Oy)n&OUuz`vYC9jClLZh_|N?@%CAq>@<yKydCq2&#Be%k?*mZPETMYPUeV{
zN4%Hkp7dVAbN`=J|0k-ZxO3eAC;oJF3#~NYVu({Ser0#2HmB`84;iuf@m+)0PH01t
zw->PL({#N4Q?;If`#3$;SNQy0bI3-$<6TF+%Ar1^-0wQ1^+gpZXr21bS9Z0Xv%r(%
z3wYD!&P%Z1gzF5?&Eeh8Ye&vt$uf4FfkZFjhMc%-z%5_8Dj8z8d=~V*Jgy*;fw60P
zS+_U`t#Y6IW+czppf61x-9(=4dJy!LHW$B16G*QMQhAtV`mmSMUT?$b1ip~@^>%ZB
zg6vu;i{dgvBuxs3uU)E}4Pe~p`J;%=Y<AmyB)uYp)3Ge55RIZYHE%x(6xRu7n?Q#U
z>oQNtBv|G;6Yo&>Ccp*?et-44XyNg<StEMm{sYwR8T<2O!6(PP+xFx86r)Kaxs@Q`
zFgpDWX>KCM5CYxxmCQ4K6LoQw0C}p}HbvAxw#pyMd~-8uBTb-ue#yeV%n!mgW%ZOB
zEv;-^>X&2@n+KvFb$vWkOi-ddRbZgvgfcZLM399rxann<&K_;&WpTqdXc!g^UP2Z_
zsPI7r5@Zvy7-HWs-;Wej;IJ1i$edz(799KvExbbD*9cH7BJ*TAk#Ud?k`hUKV-O8u
z+zF&$>$6$ppwf=NP7B^A@ErobK}&OyXnU~K^tdUFXpl6DgFam+kAfM5OZ0NpTWhFj
z#OupKuN`KAe526wZ)H^#A2sYzFt)i@yO|NYpE%r0y?p}w@t*+Lh$c$9gJLsVxAi&Z
zm^IXDsMmCpRbYs#jG<-e_7^P2bQ6DuzxC63yF%k8RvQlHS!_|SnUiZxeNnd_%$P^7
zuTE9i9B|JoEb~T1$6xh0#+<&$W>}5YRP18wi^el(KQ?{=vygxV*AM0iW#036)I@gG
zM4Gp|A3SEUB*ol<!YrX~Og@Y}N>OERn8aKfF}x(}S`fyO5R_+C=~!7F#cR!g&`!8O
z6f6nF=@Y0?iRpFQ8u(^0#A?i7KVv5QA1lXxYFX@amS_LXZ1!JPf&HiDu>Y`%?B7|5
z{adZfKC>$9U#%(jFRaRb!lv0jTQlsRtXcMt)*Sl>cC7OE*6WDwp1Y_E!qY@K@FPK?
zU;KvvkI!x*bBG-0flQ@^%%h_+kH~gTWF8CLIgoiQsvM(rDD!Ad^d8PU(#H(S^Y3RK
zEArS|nN~cQeh^Eb{BZD1w>Qfu(c2sUs7^18v{ifu;BZDMLCBey36il0SI?@J1t;Zn
zl857xWr$8E_9#PSjxt1NRfecQGxjn>6t1dpO#DJpRC%A$DDP8JW~!|6J{6Vg$z4?z
z<3L!=KpasUqllLYsATyn)qbA9A&Gqzeb2$7h+jfw>*XUE_1HYkw9+J#37YzpI8T_=
z%~$+X`te|TMZ88+UMKJdfs+I%Cm@u%-^+&7b4TV#)V4|*)OnOh5WaaFMSTNu@)qv-
zLkWzW1*JeK3rF8N_4vacU@K9UM^Px>CNeJ(_&R}i2^?Z8De+ZeYXg<77axJ?ONAPQ
znp6X!j=uo-QlSRhG2kL0BRL8qD4k-`-Rb0%OKegmG^@iY&JC@^LRy0|qZS+{e5^rM
z(M0G$y?IKg9O$)o3Td9_;fOiP5i7tE%fk_~)g3KTAmAZ$x>FvxQl*^=T&XFzQV2MB
z5w27ft`xcbMb6+#72!%9HG?64%CMWxOK_zSj=+_wTouaVk+UQ`0_1I66A@K=1i*Jq
zgamF7Xc7<vNGR$vhrJ5rzo@*Qx>dhLGgAVjwZ(S{e2>5pCR%(*10NCC2B>H8ZT2lo
z5lOm2d`#dy0^cU^eFDnVAG$R3;p$iE0@LjuThBRmO0QWpSSh-;d)Dfam-i8-O8gFC
z{apg;sH7t@D@E6O-%5$!qo(pfNa2a!CvcYl-NEc@;SZ>`PrH%{C4oQ006s`(68NxD
zs=+8RtMa+EI5I3WzrG+XxFpD#WMAD3+FYSlBr2?m#1Cjd8B$xNA(q2X*wjZaOy<Ht
zuYX(ILLHt?((cB~>O)bcQyiK{K4W7#8=Lr8BuX@>{Jsj+=v$sLd4kBBIHx)>om1N9
iv(Kup(M^l#v!ScO!(s4r;N?{HO0Jr(TH1H#ZvHR83bCgE

literal 0
HcmV?d00001

diff --git a/dataset/__pycache__/batch_transform.cpython-37.pyc b/dataset/__pycache__/batch_transform.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c2ebb1778bc13ca65e21eb4043c7d028059a135a
GIT binary patch
literal 12024
zcmeHNOK%)kcCKf4^}9*7D3X$;k{`0#Ba<@YycLfm>oIa{jz^LQMq>t*V&6-$Th&#q
zTh*q>RL?l32A+)qCyM~VI6xD~CVycz*=FrkP{?9{tdvE5fEP)=bE{vZNJ(~(O`vMj
z>HB`1d+#~lIk&ErN_hp(-~8wiKlSH|^6wNGJq7|-@P@wyVG2_{rK2{~j@HmRdPDCR
z4WnZ=%uc3}Q7K>ZvW=|#w;C4ybuZV+HwweNVxxpK!z*_xjY_B5sCH%=GcwQgW;=6@
zxlXN7Q<Yl^%dqStg=P1&#tFnMmP0JZPweT9dBpRqfOtX1Pa<AqCB#cIzJPd{RS>Vp
z_#)HpD)s6Uj9;l|68);blUTuLTzG-Y661QS871Z?ek<@3<HJ@MCDvX3pvV0tU#qK0
zexv6%BV@W>T~Dk}?tXN4%N5<ky58E@_~7Fkw-Rfk<wabqUaxD(Y2o@T=r}H8&ZgT7
z!<Oqie$WcJ(~Xi-Q@LFvhQTDWNy8gGW2W0k;YvegN<-s1H@LynuQjGI{gJV!HcVz<
zYGzWt=0?q}duTq~2t)@1G1n<2$p{|xgx}PLEz`e>_kFw}^}-Z1psxLsf;vXXy$w-8
zs+bv(Z$Dbu3Oann4<f!E1ou~%8@Vf+BItEj?)Ll-TmEKqi~B3hfbkV#$+g>eWSNjh
zD{>4@G#b`&7mRnDrt5i*Q>78eVQIT34-S8MRAvp!g$lmg;f|B!9j6npo=4%L<2>lO
zUYa9x8ZEUYYNK&aW+3J$tqel1oV2BBs!Tb+bZX@a6*&zu#dLkb@`f>F`220g{iqe~
z92|LFOW8-0JD%sUDtRu4h#Ipfe2DAmm}xR6@uqB2ch4T(=@h$6ae^9~Cqgx!!;}RS
zp$f7COA{17z!J)&C;V5C5lfibkUNo^Jj1h#%8+Jx_CP$tP2^>{CDVzT8``3>s1Q9+
z^2nMDqU7t^m&%69bY?s%HgY^aDC{fDWSK___I@#q^Aazkk9@nr%ha-LeUet?<ur}b
zSr%u9L<yapnJ*PXY0P3N^2x#-?s_-w-nn&k?cjpCWt^70jpugsW@tx2V256?jey-1
zL6e7}?b>pi+rft2>bRRcw0l0|!k#o~GrlQY92`gy*%r1j_w2y8-~ad%`@;H;O^1r4
z5f}A~_V!k*xn+0U9lIHXk$q`-)&7P3jvex5;4{06re#0NvZ=d)z0Ga@kT-h~x1%jS
z-Pw4I=;{`_ayKBSZ0ZQd_UVDk1|h%f`&{_k3+)XNbf(6Z_E@har5@y2NMextTqHTc
z!`3c$wjLyT2+khz5KRihWH4$yDGno$1)DsnfFC;}xk%Y_hlM327E-~PR+$?K%n<T}
zJ85FR-Q}CZa5ZI+v+nhTllB=#SdUdl))<d~vR0GQ^mta2>Y)LxCbc6*w3-yBhJ-Rl
zjA=C~Oa`?oq$Hq&Ob`8&&?QL|{s%~;>?s3vA0jz6V#U#76I72gphlbpHDe1j6X!s)
zaURr)3!u5U2%3)yaWN`Jr44OQ3x6$R<vmS&7FFWnp32nw6s}5jp~S`OI5vAK%FQq%
znxzuIQT{>OQ@>E-5?Y&!%b>Nm0(v5@g3iY?peN&5(1myobTL*zmtv*8B>p8jwWlC<
z+RrfgA8Uxc66>I6qO){RA5&>t>X_`>6OzzP(mwQ1#KarYOA<F$eqQUVkrJzs8fi46
zpMozBw4sKA^3AuxmX9@sW=_lvY~v`=TYd!1r7d4d)Q5*;{<pcyo7kTZyW+ju)Cu)y
zD7)`Y4(7qL4msNLTan`|U-klQvhcm-aRa*x(|gHoy1osKXs>hoW?gNR85AEEq!HY8
zkJry7X4u0)CFPYiDPBS6;tY|qL~J7Gh`dVVH6pJQd4tF@krg7Bh-5($-Q#|uLv<$Q
zt6|6m36mQ_1R^n_9rSgcO1??t0+BkAw}||J$VDPcL~<Z>fIE`|EKX4dt$`pMm}F=J
zLlo6g)g-G;ihQS;JbzSxSUKo>&kaMzNmo!R?BB;oh&SOMfhhBNEu&_1yy=f(T3MY}
zYg(0359X+{G)nHCd+uV7HzYBJIQ>4}@ZYdavB~A({F%{L`&wV`8xb}*Q*qS3LY^M0
z^UzWEweU)0Fl_p^{u@Bsm<|>6%NJ^74seiBPQ#Hf_Kg7!bX(b{BY{KR&e6d~+vc$C
z<4flm9V{ICHof|N?JHGyXdNP;t?#RS>cxsPi*yL7<@_)VanZI=SN)ZS!)P9)8^)qC
zDD5kK8Wr{M>{8UOEdT5_FB8{t)b6ueML800iWF}--<xh7;dHL*N#)>}(^+g72YY$J
z@`V`f5g(Zeou6@O;crKF3tAmo6Cf2Q6<V?DwuIdc0?!7Lwp)Hva2H}{Gu5pvm<=Id
zA2LSzb2%^dCSe3APew|Y>UgbAD@r?H<Ej7(;f{|^qgHcTTZhw)+BeZVf9Q6)9=~XZ
z@;@yw?k$BK&7kKoDU7z<&!GKBlE_XKoK4D)hO*%{qd<hq_Ps4gD$EVd5^_KX!44)z
ze0Yin<Gz4*hOBrRWK3o@%@pDXJ&Bb<0#67M{ksJjwFx2EyRX^dmK;9>DVl+NhxoJs
zh|;;qLTmMMqInNsoe-$VZn>cwMIsefBpVVF^Ma{k?s5@?iH6vpQL?rPOD^pjp<Ej3
zMi9!t4h#~rlbB5}@OeEa<y@k!Cu%cMw-R+bQ5~oj_AtqN4;)6;V2>yH&1e)pq-m1c
zD8oV14)<jRmjnq3V<k5WiS)X9TAjt4QOkN!#ZNo0p4Tp@7a+?l{gQe~-F^FbX*C?g
zbo)md8zei&QtVxTjTkUsQRzd1w4q`^)qxgk`zjy@q#K1~o<2@h8IW=x(v9i(hdg7D
z6n!(+21bg$e}(*?Y5hz;8zD8;_q4thWd>Q6C2YM(lF-`MSRS=<agJpHLcWf2?aTm1
zJDpWlSXB14e%^l-xfXJNo94<q)M5phYp_B;k9Lb>+_Ul~NxDK*YL_81@^OJx9_#2=
zwkxGg<-P*BXhl_$DeW0nWdJYjSvLFF>|6ePG$+$*G7Wibj?!ama00Kv^84z(+MY-5
z$-h?m#b{x$7+d6K*heo&)ls|1YL98IlRlSZpFa3t#fId=DU>}Oy|PC#`tK+VPL=xQ
z;S5W08FQ=nB(YETE72K1y~-)TGK8{eNK^XN{>&kMo;}D<wEdrHi;|=En4_<!9Q{fY
zqaKyEO)K$yo?-;mIA2T0GLwcjH1PSvAwHkOoHfxM^7+*vpI<w~XWD76QC`aD5zpAX
zto`jV>spOxrurnVn#9#3R_TrBt&+d~eQJHjaaCC1zW$Xic8B{i6K6LxIG<9UrF-aP
zO&OfW$;q>YeVv^=rSt)(+i#-QH}R<cH?qEl6~O#azseRL8+~n<I%;*8Dq;2o3Cim@
z!*5CRL21_qKiF3W7txQ&mayK&;B9Dk>gxhCYuEt`%8@0sXc;Ze;+&ttJ~{nZhjx5r
zEzw}jWm7}&olyX`4OC&%Nt4CfkTp-po=LP`H!)bS4G-FV&bw05KOu3S7%uc@Vy(BJ
zgt#k`?0T!&!wb+}Of9)oF(n2W#rJO3jRyMI?6^a#&S-kAF5FlFjOsASNi&TsI9YCD
z35s=eCjNp5iCuuoEcu6^sUanQNXZV!CSE+h`(YaRU2x^ee<NPkCq<H+ODNuJx#1Hs
z(-J*;7{T_?+uhB?*n+tssB(QJMbm$Rupq=J{*;L9selkhEFFPKwz(Cw;3tp<K%)8G
zB-`^_<fN0<K$1!G6ZK|NK;v9=1J5NZoti5WIY;CykrENY#7WNQ+fJ(QtYLFbYD8S6
zZ0I!rhSZ>twuH1QNG+rW#RQd4%misQjBSO)lpIejgX>iBJt99LLWW6_B?StXLz0JA
zB5G~+f?kL%0AFf-Zpx!|hbr79a*xO-RMiYb%YnY8zwJa1dQqd$>(cIV$dZBWMC&RS
ztRYPir`yPNT7GKBBj`R%gVI=PSVPNfYaH562Rrs*PP}PfqW|z8K@{DBA}*?Vs9{UR
zPpcVuZ2_TqdNo6>s#R@P)etM|8vYIZ5Yrc6c9ab$biIt)RVeQzV*#OgdSyLBwe=%v
z=Brv&tpNxusAaXP6D_AHdLCxXk~XW&qXjZ>EPY9PRePrK0v5rE3SN_oZsPDuvW~-o
zCOo7DytUrMeoW0tGPng<RzmKRtRFb!nM&7Ho@6})Ldx9Zei#Ump_1gZNJ>K!t;wm|
z3H+(6;xZMICfg!HWG9AXwOScN@=e2hGT>bLPoeokDYI`?&u3IYwVQay4;f_igntWi
zwDS>;Ke!;L9gEU!XuhD^Fmr_4Fim?gx8aFvsW3a}M{r;I3@84~q0i7v#pT3{@RwNA
z)Hz0bk{<e}El5%~d<1gXITpc+j*JLiWLV&z%2)=@uA5O7*ei8*-Hc%G;%)@)w4t->
z(<ldLSDs3Yon2&sOC&1%s3=`nrJ)O}Bz0?9y09wJ@no(^H8=JWXc!ocG^F>1+*XIY
zInYP%psrXYa!5q|9@5@W(J=J_SJgX5?yqP1sb-|`oOV#vtKwbiOCEd4_9NU&;t#0E
z9}*#(O8haAvA;-2H8)<ox_P8yh}s^>1357%6sX$;n$rRTN*11(k>5s&y3|v%d*<0Q
z9kc(Xljd2joR>}-c>yx^v`qUW#;%N)P8xf7WtGZi_-JhDG5P-fnG|nH@OuS^eBzst
zFhg9UGzpT)F(a-Mxk2P6ky}K51Tyr)h})F#0TFzQpokBNd_-i8$d8G9OyrR8{0Y)l
zNzi;Jm&;qn2%Z=9w~#<0m_*YH{|n@W{4Lr*heZfYHT0#BZ`mNN(>A3yz@Y1(Gz@=$
z+0R6ob{4wHj5Dwg4SBU<!2d$2S@JF)NVNvJ7~UAzoq6;EpA78IEbLB8YPSOUK^*vw
z$xBOKpmuRkhUHj6NqB9f<yeK~NW!GRw4vpAR0%DbL5mh<T!iIVl9pq6O-Rg7wxReb
zk)MOyn>MioM+zBHu*8BZa@Prlh}a;~AVRsQvPI$6IAQj%%_Vo7ydsf$EufNJB7|-Q
zsb~=qkvs070V8^p@fneAkkQ_&P=cJmLkjH>ne4({ipq8l?LETCpVHntgTUCf&}!<L
z3@|;={1mD`R@EH?=|xPv_KC=;68cEde3n9EG%tQZQ91>}Ao5EhpA#9wc=1;h9glsa
ztfij*0$KPHM+s*Un1J_rb+#tqIpsI;jvs3n3Tb>#!*dD5jVT}wpgyot52C!k&!usm
z=LKHm*>;H+#`pJ=v@(acaFl-F{=Opb@2m3u{=~<3e*D3W_wT&~;?KLhHx2{wDps#a
z8-^bGC+w-?4G9uXVQ@mt<VDtm9qAG^8$;A=NV#H4)QqcOtcbc+&rYGvV}Yb`6p$Qk
zOSw<S81i%4n`I11$>j5NoJPp;$fX}+<d?`1OZM`5@9C?pm)CoNH^#)@|9bEL3v&E>
z5-;BkIUXZ;UKBZ!RvshAZx9(H$I%TO-Pm2k@jdFsE=7>|GH!?<@f|^Av`GdVF+3!M
z5Lp&?2s+&)#X7Tq21)zoaMOht&>(%6Kpzs(=Q{XOfWEDhXfQu00u7c>QfCEL#61GA
zA4=n1pXoq@McgACRYJo@NA#tG%1S_kWr+qWLo_Hc)Bgn|tR4vo|2IG|bJ%+=$38)W
zWArrfD*rR>9r6Vc@*Db<wVE|=9K?Oc?3dy9u=SZkuH3`8@7X|4qQcKHgcn15dnma>
z!zT4VL3=~2N4fQ-Bm;LmJ{?FuA|*(W0!!JI#FP?%wkhK(NMgy0_+{T0gbf9)LSp*8
zPIpIm2*b-CwY+dye!VEaLz15x3A!(7%%pi!UrMI8AM(zF)V};URXT|UK~X_toIu$$
nL|`GYE~j4!zlZ6iUMCyweGAThIQiA8S<Y57)n6*x)vf;lmR&zR

literal 0
HcmV?d00001

diff --git a/dataset/__pycache__/bicubic_torchvision.cpython-310.pyc b/dataset/__pycache__/bicubic_torchvision.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..535fa3103944a6db3cec07708d41b08dc821116c
GIT binary patch
literal 2275
zcmZuzTaQ~c5Vn1;Il1n=yU-E|cnCrpagh)Op{lG}wW_wzRze&JS*OQNa<b==^>LQ&
zDtTdB^&jv8NXcXWlAn-x+Mhsyma((D-BKKRJRXlPGxN=y6!m(5pgCJ1|Ez(~A9`}J
z*)X{Vz5E#nLky=V!y(Q{NU)xh)Cw&#+M#VmCv?o{hMpPy(1+1VD_IZ*W{;g#vszd)
zypz_mM%c)jVbjdrw3RJ|OWAU`oUMc_M&_lf*;=@U(M`mBR(XtAg$Kj+V=LTX<UZ;I
zC-64ZA<EuKVyXQ5e1FXInD2J5vRNb}7@b>L)aRhH5ho%Zr;(s(@#82eOS;EJ&QrB2
zqMQ{O&5}o)oeIm2rMzg|&v-6X^Hd=g#fU~isAddeF6fy6u4$x2#Aq)`IZcyN#`<9w
zTkUjGroeC%VunJ(Fu0I#n_1jpHjEB;nF}2p*sDE7J#5Ubg&y~3l{sPo=y^QgHus0s
z9@#`an^jruv3-og8mu@y!s@K?*beJYk&VIzIH{=`*OPesFo}1-BPEwdmxDI6ciRu{
z?%ciFju>mlMW!t^a#cCELixp5(lR;Xs$y)Fha+IeC8vX4nv9gESy4#k3NFVYkKGG4
zYu-BT<t{K1NjyQvWJcx)BWX?WG0dqwAwZpp1=O9Oq08_AD&M&c7}K*~&~M}z|AZ$t
zta}p&DBubVvvwg!?jjWM024%I=S9wsR+9W6N)s>)Le|!=zS{0olv9pJTnHTt<x`sE
ziKJ9i_0$<fa!?A$1SFc8v@}@)(s7kt^0X)FTJk>%&#<ac8mAgk^ci~oZFf*)yqg!2
zKP-y9t_EVaFN*P~dw-nYN%DR?;CVML81L>Rz25a(_lypNrVCNhk)pB?@!%jSlOo?9
z9f}ROdJTH1uW$GPcHy^$-OT_WZM+`c?K8WSljMv)iV18vYaTGVe+|7{#|VLg(FAXz
zDPVDGNzB02JLreaseM$JWM)ms9M1rsb36rvPOa}Ca5*Hi{z+%%GG~r4`X239Csw%)
zA#*3>JYZ<(F>jBEtJ2qD8&;0V)XCexdCUj-4H;aJugdDIHi7e|&cqqkp>$N90K)Y?
z-Fsw@l;LdzKyX+9tB6&fSULw-?JYEG$m*~OdtKR@Xb>$yUp}^AzA{^#xUcV+t+5K^
zi8Hrm>vIGaZp_it1+4~a{;CU0Yxl;9hOV;i-Ms$b1RRBq)_M_SH*3qHt#iGd=(^Gt
ztFR_kfShPYZ=bw`)&~xu--~NP?^L=%k-@&Q<RjU+Bra*019<Io;b`72Mm$$uT;vB_
zNM*|;<ABwa^U(r}70_BBPuFZk@{w|&4n>EcHHaW3-KaFF>Gj}g@{^B+uXmh7SUCny
zD(DHQbNM(M9jc0i7nGx-<jRf<5IAOU#}k`+?-d}SH`u2-FO*k~<CvGF!c=*qB&A8#
z4_zR1urJ=$`i5KvPE{%GbD2n<QMv<y@>3vzW#bwqxP@(l-3DI90cqd>L$|KrE2L%F
zFp`%xAuotSu>0a@>kVX|9~^7Po;A&D=NnXO4@r$0C@v65ybI2^Ql1;Pcksfx|2)~F
z6f`Iu!roG!1ZlPY#^pu@$|+QiJjytwDxfqg*cfU)RW(ZY$5DE^0w=N}2C-F2p{xYH
zA~qC*)MT>w2(D0GUeLaXn9vt0_W{q#LWs{{BEHa)`e{vxd746%diCTjt)g9R3YiXa
z`5q7-*_eR$z<n0^2e`k%jeilBJSSe`Pq%voufK10bS^0u&;oVHTrCXH?F{0)0P_cM
zm~sqC#E1GC{oJr>AhbP}Nm>g1=RjHC+`6*>@RE^kGp_B4Bor;7i@(sNMpYM!CS4Y|
zG;r2&3WLR({@^Uiuy^{6G7r6A^DB7yUoKq!m5zF;bE$=VY<|QCjMZ?9tO8xP{sp0>
BJ=OpK

literal 0
HcmV?d00001

diff --git a/dataset/__pycache__/data_module.cpython-310.pyc b/dataset/__pycache__/data_module.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9858ce974e4245176fb56b55d6c9b7994cc22333
GIT binary patch
literal 2727
zcmZ`5U2hXdaCh&+=Zo#waS}dCDbUi^v`tV2QbASK@X-nhjFd!uI8{0w-gR=}KHR-^
z)kNl{LE??4zM)DW^(B94U;EU*&;seqo(;AUC*9o4?Ck7(@8$KnPvCic{8U^5>~AEd
z9~+48;VljTaKdRqQX0^d1uV4!iz1IDcIpI9>ISZ%t)vooz_*iX>IZ(h5G)wJlho3B
zP&c@nG}6Uj(cqP&nYMyfx)dx?@+09MuRbQcD!je+p%pB1Ym0RJf8c!5v9z_3?Q3Uy
zI7o!9+=~W-INRwk-MszN#@+DN#`ebD`y01zZ*J+<_U6W&?lh|la{jGIMR)Uv3#luJ
z7eaNYuHH|@PINQR`aqnBvqD9giX$b$zRc6G2MaMC)3_0--tM-HvZ9~MRIlFQ0tD6m
zEWd}wPCu0~QTzdjAOYnhU{%UjI1?7P;I%pFyH7};a)-N*>w(QH!V#8m`|K2iQdEQ|
zs)+Y(6CgYAxo`GS;4JXkV>?&?xmus`I&T0~2dV+oB5wk<$XmkYOQH!}x9{>cUj}aL
z39(7A1OZ;v^&59KF(UVJK1@Vc*OZLnZ0x)Ce~pp}?n<GCGV57WE+Rg8iz`4Va!8)j
z7labUo?9dOf{sXNkC<FBG&8ir$rh9ltQhQwb~~<iis1m#<Aq_I#VQPC71U`vfvm{}
zpmGral!9G|r2m$41ZfPzUJ?~W7#@*-e^}qmQ?Z`qN<7H(hwHtZi*-<3<XM!gZ(X~5
z<(sdsT`k!+Z*Fa_^RgVwH%v_w)&~2s4X5F;g$uxUYV<5USe=R0+Qd@rmX*?DAUb)l
zg)lNV;Vmu$P-H|oJw*=bBQ~PXDTi3dmf=Ev4{62}J0y>+5gS>32D!1ix-vFSFFwEu
z3?VmV33e=r9jvV#p=1rPjxAT<BTpmv0KpjqC*Ocmfaw}WGaM^D0h|K&)~IJSj%~_b
zd-lQUSs$l2K#EkfBWB!0m*2R#2?RR%kjaZ9@*EPp2L<pmgGA$N0d30<$1C=D#eo&q
z6a+JpZ)jB+(L?$>#Xwm7San#p+wtUCIK-r1*UM6!#Au}?=OSgwMlWC;PXyaY%!%oC
zH&=I3m@g@`w&1s=%w-447Ek*K!CWXE7$L&}u9z5lbeX#JByBQ}9kgd-Xe4_SL?;j8
z4!ou~FlCBxi$kmI(HJ4)P!fO9r*w-P*<F)Qy@2M2m=7V1F;D|IVB~oOSgJ>~G<VL@
z=G!opyE`CTVEMQdl78$__xRxKtjY6R-f#Wqpa(35b{-O0!8|HD_Jz#*B0mFRF#BWp
zl(u6}OSES`yp2i1N|V<>q(C2iFnP-CGqYyTZqklPvTk7HN<CrBe|u+uEr4E=i9J9@
zFuKs6Jt`NVQlV8}Ko$(gD%$32%agF;lq<?Lc?0=pa3_{>fMKnZy$WEq!92jAvVmtm
zBY*m)uHEjz>1~|0LY{-pQsVkK-vk~mx9jH)>NtF3u)QRGUx{Lj*ZqMMJnoIjCBKMZ
zX6}c*UD11J#5+}8F}iT1PL#?E;Gmi5W3_s5ZlnAJC2E^PxZYA>ngH~a=jy3LA&I6l
zz65dk6@s~<JCSE$)K@~zG@6iCLAJm{Z0a$eHmOe;W0=xom)Wmf%YNnB_Dk2X|Fe$|
zF3zU+JvwCz^mW>Jb!1U0!chC+c#;EN3&UTAQ8He!3MI{KmpGs$jxUMxOL_>f63PMd
zDPlPK<s}5ztnv#0+KMw}ywLUWwbK(x0@u~s0IZ@g&RbarfD^XBVeAp+*ss9b*UVzC
znB91Jp}OQX{DyP3VkA2&+Ex4bYcpjRnyJqZ_Ej$70!ZSWU6p~v+EB$wtYT5D!4^@J
z7NWtr^^UM+gG4ut2D7hN!;UCB5dGFwxH5Mlyv52}w}#q_wO*d4d1h{!-~?`gNwzjO
zILnx!le&F(V%;r}Eq^GCV^gs)Q8IUt_icIEeuu-bXaZO!Fg>9pYP80hbk+JF;X<k<

literal 0
HcmV?d00001

diff --git a/dataset/__pycache__/data_module.cpython-37.pyc b/dataset/__pycache__/data_module.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6b67885a9a18678a43190ddee38f138653011f92
GIT binary patch
literal 2661
zcmZ`*OK%%D5GJ_~tyT}qkJwF+woZ_wQ4dmjC{P4NQ729d1C9zfMh}Yui;YOxYwtth
zQbjE&m)dA<z4j0nu#f#qy!O<;&{Jo)l4Gd}fgR3}LvrSu;c%Zc8a{#MJUkPh1NS!y
zvt|SHEquio2q&B-B&7jOS-?^&uqckP#7>>SN!`FTvXxW<5611Jn)-pC)`FTDcanPA
z2pWcWlV-XUEE&F%w9@5ZIb8`>DEXdnk5|tKuL|#I_0$U1xb=dx{eR$m(zdj<m7Qp3
ze>6;lt~`&1!#F!=Gu?Xj<JNAty|ur!`@`1uv%MF3d4F$fr+0m*2YUW?q@vwC;zH^Q
z@`X@ss;fVw;vjmOX9Ey6;;c|nrs7D6a3J$E?88Ki$26Zrs(-jIqpTR@GS%xlT!5lF
znIG??v9sn(nc_EK1PLf70pqK}5*C7YZoMG`$}Mi6H3C~WFzyc6O(H776IJB6bI!QS
zD`$4#!y3vxkgMDWxhCqoc20SnH$ZBLrjeR_38W?Aik5JB>)hhYd<909-w>MwD*)%3
zZamr9L;s)We3XbDAX`^5inFP&+W$F9X1ptf8p*70O+CSW_7xov6gedq^pa4b*o8Hr
zmvllxd&1;BBeP4IF-~C2;RH^ny|$~JVl;#xcwrc4u?j<3g^jeGKx||aSh)lQ%(4rS
z46cnpu)?sPL`4yXf02K8x`%lxx>>Ho%RE2sa`35pAoJ0%`(l*sLU!~IMb_=-Ty(eN
z!2sMgIuLi$hp7ReGdz(ia3G#sxIl>G(>fim&%xB084Ic4mPsL7pxbA-0vcWyMCK`c
z#Umh!Oem)}$tiu!CiH@Gz(U?NqY&9sn%Rn-lGoORO{@U}Y^<KHpbuc5UV4ey7D8Og
zbZc7@Z>qKqgpzgO+O}MSPJV>sCX!o7*00D0aJr6|hSN=Pq;!zht=nvTYu>@@s^Fvp
z7vVt6`-r|?`M3uHdU(v_cN1~}F+PIm|HL5H_*o#^@~i2LJ)Lo2#(mY}%;*^@^HgO*
zPwB64J4zI+L9E)W*K2$7HXLFyKsU-9odI;C6c=DJnWJl%z%#{F6c*I<YA;tiDHM_v
z?y%s`q|9ZTm56?fM=s##Kuku%PhqA&@Co(k8g=P9YqIg`Jd8%OO-tY?Dnr8_i-AcF
z43r`RsHg~8jFfRmi9guy)Bs@OscYy#h*1#|!e4<^5-l)rzyT)w3wE~NfiztR3kpn4
z0_KiCnKyM&x&K~sAHrfV5z76sCz}`}MW@D-4;e4-!y+*H4s@mE=1RZc=;kb3rO8-T
z3J(aOKDE*8+w*46t3#b|On7w@0W0?iWByAS0ZdnH*w^4Q5y*g<Kv2-FBP!P+N#TlK
z0{UjeDjt~LmXBb@DQA>x;syF|p>&sez+sw`qY7}g;XL3Vtzpc5OaAn)(%S8fZ|!51
z6><}9mJ*LQeS<%gsmGh+1{U2I94JNKSE88m4Sy&FkNZ<`#V;bLnkOL?xBjtFZ&!82
z?1e>jwyE3%CrvG%ZmX9T7RtM*QQsTEGnNW-2VkG_Tz&0Oh@;6hOa$3RvXFHfc@{<k
zCFI;a6Y@*YEwHQ-YBLX_*`hvUbbNn4o*!6^Z^bG;5?PdrFw}lHE!e=<!|>H8N~SYb
zp`<C{5;I+5m68~5iCH8u>SZ0dha6UT`2Y!CNckBMZN-_AUm(|*J}`ZeB=9g@ISTjf
z2C8EXl?@<R28?f><I)vp(`)%nr&%#`?Hk%vC-{4E%@#CG-x;2$TtZ<_;)6q#fktPf
z;v`nFC_1o26s1*Yq+b4?vS!0Xx6BS^UD3gmFE1GSs;lrI9z^&&m2YYtcP~2qJWcb=
zydc2_E`cJ|S*VF6PUx0i-JMyt4Z7uTf^qB}9wleq81L)y9sLHYhk;~`wV*(@Xr0w*
G%lZ!_4w`8I

literal 0
HcmV?d00001

diff --git a/dataset/__pycache__/projectdata.cpython-310.pyc b/dataset/__pycache__/projectdata.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c553acdddde605b91073a9625d8f6c5edf6263ed
GIT binary patch
literal 3722
zcmaJE%W@n^v8wtpJv~p!vLs7uBUsijUXX10z$1)pY;8D9EVjd5BA|n%X?NAk^td0R
z>K?DGXmPO=d)f#Ht{g^3U-$yPf@7UFf(we+!|n+%3rl9rj4atePgGV`W>#fYR#s+C
z)@Zl{zR$Pz#dw~Ozu@5HXTsnnwET|%IN>xV3H4~gJVtTMV#70ZZ+fQgEzi=u?b*7o
zcop3{o}>G!=jy)Z)u1=xdeZP3dMz_<CM~a}aVu^obKYDs@6GG69iK@SyoKbfcQ#q{
z7B#IBFD1*~G9`BicX)N5@TzDG&OJB070w=!j(Y@qCLN~Cttc#%^GIxuL>h{F9jZ(o
z6an<s-6ZG<kXZ^N8IIyW`f>KrFvxTNiIAy?)q)ICo+W+~?Fc?4<{d*-f>AFKX`xzE
zfly{cKaf(jLLdw2&j_HV8D|0ayHPCsILeC<H+J#^)9Ww)6?B4lloO9}3Z`Yk<c6@g
z3C$8Vw?&0l0Cxbd3WvMG>CS&lx|G*=eZTHig*&bt65im=ee#@ob>8Cbed0AfCMNNk
zV3s-6eDLu8Pu{!z!4}vc7sc**cMZO`*FLzvb^nKJ0q1LBmf+bmT~+nIhe6R-PBtq1
zJlYk?%EKTQ%I!-ZIB?ulxsT(K1a2=0R88Yk6IQh)if2Qis-s-^{cao$RRvL*6`}UP
zsR0_$JJ1Yh`CkDP<T)ABLqbUj23a8`@mXmAG<syug;p!+7Wr^xuU@)^IW`MxV2>+h
zowGxBNWTX%W$hH*n9^(8XG#~=ampINYDosp*wuQ-pgP4&;M7VBV7;^fHaJq694Res
z^U6Lc+T4M;4PIRU|9ft1fzg3G$K4sXnCG~99r(2c;Bx0!>I|=g>;flnEsc9e_}F%x
zOT&JaiuFso)k}FOg-F*g^}c-Z#h-ur_}4djFMl)mQ}O%$jgW%u>|=nS!5%@&+W>lC
zrVq_MTGHRKZFmYVh+G2l3#QomWkn-B(%|LYjeeGhjWjF7<1Bl!f$w6YC$rIT<IyO6
zKT3OHU!)sh#>K`~)a~B7`%ue(-)szJHV|QfQ`U#ilzUD4c9smUe;;>(+iH{DcTaoU
zy6;D6RQUe-wKxMG%C7^0Zvy8X;jxZz;u{A<<LCn}0HP0Q;{iEYC6hyT=n;C-<ZR2w
z7#>`%B6thH51{FdEfD!NB%;F3S#amkP)Oy>Hm6+h>aeeUUA90*nY}oBtXP&SQ$GcB
zkVUC-;%)t`bLGHGz=s0-Jd=gQdysPo@Ns+fG>gFB^{csZYP6{=h#)<o+Q)Auh;^7y
zMuhR@)w7yVUBP?8A3{8eI}Ok#;;zJZBbx|tv#%+aXOM<=lV=fp3&FP$yaV6`RaM^)
zV~lzJ#CqCXUk4tVF-J+a<}#a_%%l!orYqEjkqbZUC%+5yB6ZpB(*L;aQ}>BR$4kDA
z%j2aL;IISwoWeVx_aGV@u}JkcJ5+7}|CPh)t5g-=2MOQ*5iWo$f#$_GG;5I%*|QHX
z1@&pgp8*Hdl9HYydki<AF{5B(qcjfb*yQw(?m<Wy+og$EX>tZh3)0qs(PifdH&@BH
z0wKrIaySoy&UBuo=UKq9O6&j31MZ;8E!YESnPL?PH<c$$`h`1y00bH#{0!==ulmX-
zCaeo`E0Fv^GDkkIBBtkHGIlWtft<-}C3{NbftEq(gX48)GLv0E2s?mt{gUlDd)1;b
zZo-UKi2<&?jJ9emlS8)W9&<3To#fO?6N4%SyW<`tCX^*|z%bZq?yNt>MVo?^Etrd|
zw5KyrR^ys)f%Z9War=;Q_X&}o!>vs3WHS52=r!5jQrVhFADp^J)QPiRqgqu0y5>Rq
z8Mwvz3A{1I7l7V8fw#b}XJP-%veji}y=?G0-V=Dj;>3eAHuZ_|(&Q9ykLEFM7R%#v
zC<Sc=JZlwpX7R=+j5jfHfp@)iNMOa+rOf{#g_bmV3leAh4<=^*d&+ot=hphsSMc@p
zoA@-lsY9SV4?}o(th)p4nOuaSyZ~S#Mn1jat!-Z3yn5yG<`sV!2WbJZsMjxK2Bpz$
zsC=$&UhXi@*4h8()BYD&XazeFluLT8SDF;*n2N>QiuJdZQS21cA|nkF(K&x2Kv}Q_
zWoAQ>s!EupPlYTLONYvWLMwQt>>$@=MWqV`R&f<VY2FW@%rL^Io642~OO;9ZwkK7Z
zs$dUcRj%U+zk$nFMrqW|WTNb_pGBdN-$mBD$g1B7MtL3u=`AR2)Lf9m9*U?p%0~HQ
z17?V&j8$YZ>?^k`!5Zl(fn-=MKv6!-ax9gKAbg@My~j>l2Q+0xDHl6l4XVwDBG;M2
ztM|dAr$+@2qon7#Q$fTn!2P~wqd4?Upy@>AXEJ(Y{E=M2lhA?J)R*J$_*u6L>3cSQ
zHXFCFT!YkeEHER?BwmlgHQ(=vA}U1U`#%OH`2zr$723?D7ib%>3y>3B$Qc*tqA^eB
zbz6b_u|?NFs!5x4o-V?V(fahaNL|CFb+b+{vWv9Ic30kr>&HqvEzH>HG`_D~-%m0=
zf)d+Tb>H6}1@V*v*@9;wkWDN!<af{>1|aewVkXo*vBYdA@8P(TW`0iwTwX=e&yduZ
z=EqRP@hnH`y5cg5=xkXB4AX3mB!`%knT)cp7-Rot2KyI`zGN2rC$rg0R$>2O4*NT+
zHveY71J*Jm&hD%#8=M&`E{Q2iVnwIS1j=rCN25iQ2>CX$zKP&HBwE_r<V_^#-Pg1$
zu8)c+&L!4N%6M@1eW~BkoTjdGfyWDqC`+Y{6DufIn=m8%KP0gPmsJFqEj@R_!}3Pf
sP`EPv>!3G&!Yz)AE#+LBT*h_G^*Kh8HgUiaj(^P27v5$I&@USQ0Yqc8j{pDw

literal 0
HcmV?d00001

diff --git a/dataset/__pycache__/projectdata.cpython-37.pyc b/dataset/__pycache__/projectdata.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..08ed181b6300359844556bf0b90522279c651588
GIT binary patch
literal 3669
zcmaJE%WfRUv3uS-`{at^QxcfiP`q+NN>qS6ph%8HLy)0Rf@C1cAjV`g-Mc%>c`SQo
zZIT$~LJB?^NRX=!%cC#(f*>D}W1kEJFhC#&=PDRTs(N=RQc5z5sjjZ7p6+^dRqg$H
z-6imRyLHURZ9@Kmli6bc@&Ua0UjQ(|XiO67(M0n!igPX2Jzb54XQ<KiOf_1brAFJc
z)#!MR8Y`Zw#;R9^QIBg$-K(o_8F3?NdQF9!aVweg=90G8R&y)9l+1hc$z|_yvfwQ!
znjJ4DOWqPC4+(Qv<(RMvuMd`wbZ>=e&q>EUg)@_mCXKBqETr?C?~ZsH^6d_l1`CP+
zM)Oe;^f<^YhLH$IaUlFS`)nBGxxd3j%40b%f|O;6pG13{O^JC&mv%7fB|I%;b1D#u
zZ0H9<$Yuy+F8mn*wrRv!!2E6$b3cyqBE*Bu9%X!(<l6u^@hBsn#;B)rgPZWPxUHbW
z9j1L*`J8kq)0uHx^D5jOS5F8tnRQH#s8?e)bB>8u|C|`aYk&zVvhnoUlaD@p@bMPt
zoAct}s=EfyJ8K_5*?RKxwScj;FiTJmMTZ3yb;@B-^re%H3O|nyxHR)Hh`Dt8!Uqm0
zmrD0VJQBd|C4sCed}^?Y(!1Cna#<PW-0yeeXeez&MOK6}b?cz51Fw#%{}Vt#j>wpv
z5K2l=e}$C9*Ge6r-Xn)Dyw$xsCEX&Q{P?g|x`i<|3UgqM?Xt$S6YYfl1f<I91-d?^
zS9jn0H7Q;AmQz*%R!TB(#;)3n3@TI108X_u0oF<jV4WeQ!I08q7PF5@(P9p)t+UEJ
zMB9<R1!f2C9CK&fqRsH@HQ-n0fy<mTsY|Q|vh$39W9r+dc(o3k8^eB<^7R`Bl^b~|
zI8WDa^uGJ*Uw{AAi{Ibtz52u8uf?DH_d*J?_-OE=!JfmLw*d6OOrJQ1w4{I1c8%xc
zD<T$w3<0gJ-;^{`GX-8f+UREq-$=89zsRzk4ZHyxJ&}!u8_!4S<0$QgeV%TF8RHvU
zQMY^l(K96jQL`}=*?@-yE?FP$zjE)WD9)1M-Jjw~@K_c(c>iLot^0nIMuqRM--$Db
zpZqQ$82z~J6jeIKg%^}VqbTL(0a41;P&r0c$wa{pJ;ex`YAKcwU!D;r+OFVq6~NEn
zot{lxLn7|kxeU=f8ge0>+4-akK^^v0sEa1ZNTU~LFQk^`(oi?S7-Uf@op@K>;aoa!
z5%8V>KhH!V&=q110VGi3)zU12a95Y+)`i)oG$CE|xNMzWnIKj<LFy4Elh?OuLUlRo
zaeoNOC+^f?Hy(EdULDaufQNl!b8!i2Xg+Zn!F2>bK=3XA$VC<34`WPl{yB540S|4N
zzXyO+UCpJIW>AB=bctTkENTJfsu6LEUZdBwgT?>h@H3a9!O_PL;LXu*1;p!s9#Oag
zbQ_{nk9n$&*r8$__>&`H2n%$hw0$2WeE%0PQLZG~!NOa+Dv=O3!Uvs!d#O#o1uASw
zO1exAHQZcbdZCT=Qa_<%gV7Uu2)Sfzl?Gy^!8E8$P@7)rU2T~#W0j0;$Tg0V!*!5t
zrt3_#&IFEGn*V1Va0eA;!Wm$fDP}{Kv3E4_x56Dj-hmw<>kMkEuSfd}1HKD#D^TG;
z@*Cu{3Sw#vR$Lcz56Bs;T53DiAE{_4DUe6}?AtTB$w?r4y#(EUryV+nm7+ezHO&%p
zTWbmJ)m$Pc+M#>K!JKxUQ!Ne5r<mu?Mv$0L7RgHu^Q_{|#tU4uD_GiuwfL3RbOp*P
zT=7lVeGZG;iN@R=0@{27zL|Pvvi{ucO=(TMJCS~Q!JD`%u6l!ttJ|>qCGceJ9A2N|
z^FVK$!<%5&%W(ci+3ad%t*o;e`U;|9VG>0Oo5sa>aiRtI(m2D7VrjgLQt+-obgja<
zOjiFwV-2iY5MftN2z>EPDdWFLp(PF0gu2=K5{f`)Zd>Zl9^PL+{Wm;4yc6W>>;skl
z#1)uedFbxJyDt`CDy{;Uh>@@Fd25?DH*ep%xp~VU#z9&@9_sZAkwFji09u^en>Ra}
zXQ{&fo9Xyl?5Kjh2>K%Rt=BtA(lMkK?@Fz|EA?WpnD!TGknqmcbIHksBS<3~@>JSk
zmVU}bA+>ZUP3WkCeQ5=`>L~0kbW+7_NTYc_fDS_sKi!m;;Mk>1Qnxi}$E1xTq*U=E
z)bM-wd3%&b-Ap9X3j0|Ua`8iC-9T3DVKB<`C`j)^S0m?w9L`Wgy-_yGCkHS>>|)F!
z6JcMvT>;ieN6BzsRtnfCA7(jrNktIuNK>7$(^3gdno-L5o>zso@)^%n)$nS4FzLlv
zfzv4IdG1sYaT9RA?^!4gV*_ZaNcowF-kN`=>Y$_puc4gd@A+A`3)Ooz-=EE!*r`F;
zITM%>W)g2^;i~WVco7vm@%>+dxm2r4R7Ge3yhas-n&DE1Uegz7TfJ?%0>3S~rnPAU
zer>uy7ql9k{1$YOHvreJV=Y;EE32Pv*=b^}#*p!S>H2<>u@Q9AzO4EF?kI?-9H<p6
z3xRB4mm%Iqzv_U9r-&KQ^u%HfQ1KY&cAEJ;5ioHJN!WA9`m{2JJdS78QFT*PY@m>8
zENXyZiOsQGI=Z22^y065rDX_|>a0o&f)*Mpfu%`c<0g#+I&1NvLW?Nj;`_*Y2LaYA
zX{x9acaebS^r|Z4)<;DY=K|X!sXu-6Sg89lr>N_Uqar~OWvQ@mp^akI2{Xb!I08Fu
zQ9*!}(Q_v}Y+*zdh3)D819k9o?tfj0ojVh=-^H4qV-{(Fv(J7SV(82kou@VZKmW9<
ALI3~&

literal 0
HcmV?d00001

diff --git a/dataset/__pycache__/realesrgan.cpython-310.pyc b/dataset/__pycache__/realesrgan.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..b556ade137e2d1e1b0672c4f4934d4aefaccc53b
GIT binary patch
literal 4046
zcmb_fTW=f372cV>k=!Lk>PpF$%qC4zrKyxC37Qy&k=Bllq^(ofE{ZM?Y}TA1we()-
z*`*r`%a_VQUYcHB`qC7j`j$V?-_hrJD$tiIy#>7nX_NGv*_Eu=ae)F|V!pZ0o;`EU
zcMj8Lv!TIr_tPJ7dP>uNi=C5?0iBniWIqF78Y7Vw6OY8yqXhdj(mh?(hG(eS^h{M-
zo~`PdSA$xQ>T$zss69sH#IEP6wiz|!me-2sygAjkqITTzI`O<WA1`<dYRrxn<0Ws2
zXqPosWA(cltMkTad0+Qd7`>)#HV#2HZIeplV%X1R?Ha!|;Ypug*(B0nK^{PDz7hvR
z4r^BWq3BPdK=@Jm{c(_G{!K0t9?3-!BrJ{nIDC_{YB1Z>r5#L$F;8+iS1ss^bnFL0
z$hkfYbKxHiz-wAj8ZdtlM%<6WEbpUWC!Z&)PWEePaLpr3^C%;rLdp%MbCVfRo7`d+
z6dU?AZgXQ5KiTFrs2f%7RJB{xO<rRySU(4ST;Hz7J5@bj)eBX<2<w-iEHiEJ<lEYS
zuoc$5>w2KtnRB4ARXA&zyW9X(HwSc88yIYjow{r66R!m`=76%(Z2hj`&AqJ|n%9Pl
z*pQv8JcwStcJ=btu3QAgXFT6qZ}i}Kw0Hfri?5yS1&sCjX^aY0<FbCNELlq@xu1o5
zT-KFJ^E+c`PBQNA4x(@@ZET7(m+c$TL{wK)Zg9#=R7#geov}!7$c9>*g~K?I&JCUi
zLm%#O$ji>S>eRC^>HBI!=l;WYq<iA@dp%se1IM5&>U@|w@q(&Xy<tZ-j$g4O=LTUC
zM6m1VrIz6NL?r#C6UKsVbUz(RCINCSCxy08XY@cLTCNvlANsygP(ZWL0j)wCS&ZD)
zvNNxOb7Jr3+Arun`2i_l&bJF2uvXLn>qQ-~Q8WOZ!U1#(7qD400b4~2aITmGY!_|7
zPSF9JFXjOkiUq*MVi9nuFaVbe6L5v;%($!NU1l<CpI+3Igsb`5>=d(^eL!cY4>VY}
z&On0EI;)=p5<wTf0-k1-{v@r^(WH^23Ex`?liNWQg4uxI_0T5H_WA_War97Vpq{Ql
z$+`eTc*E<pDfm`F7J;%>w4c&jNEGxC*B;{L_mD5>*3**Q7EeH5!ACD|@1`-|PSTv;
zNYk6!{gm<TF&9}1*R_4^#iyTpVdur?%i#+bu3g<mb^#;b7GU-~6T=|c8sCwPbISSC
zczpgfI6K3)cD22&2b^Qe_roO2eShm*l=g!tI}Z)y4cvN&7k}s=pecyQ0pJsmuf;aD
zH>8<OKsw<<Pu7mk61YX0;~?M7L<?8XA<QEzAS@!NBbTwYf`EDxs|afd5DGML8ett_
z1K|up5220F0gz_tOGge;_RKsL{atD9agk=y8~`Wg(j8ACaNnFKSt>S{q>=GxAewm6
zJsFP={y$3l9+7q5??*`WetFl(!r~bKR6@1`&{}N}3w~?FqEMHGLXE9^KOOta`n=vE
zdutEBO=*S+<8Suy-j5#qQoNW8P_hqT2J9c~b5)xH|4;RtF!0ig+Uu)RW3QRhnO@KX
zGJ}|KK&B8Cruuim=MxB^_{+_i#moajwC`!RoI=mGz}u~Y9{ajB0&lxX#o4@uemSb|
z)2W&Cpk=cfj9<zdC&m|ZXXX}g-qbA2Q4_*-{Via8+=pdAZ_><rRRaz&Sp#M@=Dek2
zOknOQZPv=2(H!ix@-}kJ2v}k?zpq1oVFvs#T7o$n)-24H(f3CyH#PCg!e&jVe_W0O
zb{XU!SRc@7Epbkax9*MStH4|~BW&&+eOgbt`5K&e3d(8Zwb443d}GvgG{hnD-x(-9
zkhuL$eOhOo!tTP?t!dK+eDfE80fGMpw8R00ch?^x^&=qd$CcFYpOEf<mD;JKeiY>T
zqLTVCXgzkH)Q`j16P47?15!H=NL}em>Gw-FwVx1i<-zZ`U(1y?J4%~wzBzlcaN(}p
z1R}}k%TpYAs<1wL<Fl`@I-3XE(`Q=;8d&J|fd)5?d+V^bu38SX9I%BtL=~4Uu*G*x
zVBe)Hm&8+G{ZhYr`Rmseo$W0R0?3tEFHd`zIrT!!(t6@iSSB7pcno0c@Kbn((9*Qu
zq%R$!D;=5ukZ@UJ$hYzoT27LNm>*zp7h5nZkmsegX-c|#OWI`&f99wGkz0k|D*My(
zH}lP>1#*FS76H$fMmpw+wEJmtn~Pi;c^Gpb-H7vXNvd^Rs`v|OS<+-IO-K!cJFsdu
zfb_)*GL^>IiUptD5jEU#kLU19A@J>{VV{fVuwze>aFB{vQL?1tP*}LdM))j!OCtu3
z7sn(nbx=l0(+?%_9v<^NfOIY)FJD)DEMCBo#y~*+n@nN|Vf7ponT^woOQWBT?})F$
zip{2Y5%-@1@a)}NKE@bnXOn)PXPG3vw8vrOhw;#}0DMsUQ7(pQoo7RE9R_}1T9wvo
zczZY{F5$VC5sss=K$@10?+<w%<~;WO@4!IzLx6^E5SLKWCWa#C7MUjv+9C}CMc*JB
zw5_{(o78BREYmij)PR)~3huy`^`Yx@iCRPc%Nn;!=5?1W!?`Ym1q05epBj{Y($FDb
zfE3mzd+VPU2##g`Osawwi9taH0koJL2@%pba1U|?G7>%Fi88`XQmXnsZP>mKD}4V~
zuveuO4YIfTdGA-NHanOVBtgu5Up9O{PT2(hnfTK6{aceDs%GF&miA#-&vGI45dP~J
zkp4tMhqtJUSHZG08#_t;p#VF>?~qnyKLM`n<4CiLwS68%NAXsD#93UTA}^*Z=mFUq
z05uJ+12jrMp*sB-TK}dd{TBq@k7$j4Xw~UItp@#v<<P%dF8!O;q<^I?`WM=!A6Omw
zXKSAR$y%U)v=-?ftR?z;x=i1vE0E`PTfehj+91nssN`s~D=l!tFc~Vb%MU|qDw*aR
z(2}MyeDN(DMF;U5mBwsM@-WH-CYVzH_AB2kiJ_xLw-`rN1bLVy_q<d(<)q@W(%%(L
z;C*|Iat{CWlzEz6{ZCSIo-lr;GvxubbLG9AN9$1uZd<Dn^`kW~OP93iBH*(AAL1zM
As{jB1

literal 0
HcmV?d00001

diff --git a/dataset/__pycache__/realesrgan.cpython-37.pyc b/dataset/__pycache__/realesrgan.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..bb85401a7b860b2e9306d9ae67331bdec4cb56bc
GIT binary patch
literal 3978
zcmb_fTaO$^6|TO|^vv$gUcI}%WReiaYr<~gKm^9J9OJ}91ShhC0xd;NyQ^ll$9)-B
z_XK;@BO%sNo+5#Sc!`l_UyzV^=80cYPe|~xKY_f!cWP$W>%^8oqDS?0-RtzJbG~!t
z`rMq)@cirIBVnFq>>t!PdTdm#p_G3BaK??4Wkz6RW?&lBHd8CG^xF<>{dNLJzumyo
z??%wT+e(|6ANYEWowl-e(AIS)oy+Ef`D`Ir&}}#EWQ)OKwiGO7%fYhl^U{@UHCQ#+
zO~xC%`H1nR@CR%AR&a`&ci5JH2-(<{sq7m`T&cz#aepN8Slrq&l+B|m!rOT-i~0g{
zPQ{6gM`<L(wD{F9D$DSmkhw_Jii~nzWMP&(5Pa5IZduBUM*U3Wm0Flhh-EPhBPrEF
zjBX*rqYfN1pB51hdr2z7G%2f?1Uq`xW|i_!s0bDqoCPK~0!!G!L2>bOgeU9`25dYN
z8~?=&Qn&ed^>fxUxW(;9?EvbUv<?{7wsDR>nCqDvj607_?sD&uy>A5b!r_g_25)lz
zksU02&TJNRpxTyNye*>iojbR0zJ2QkR9T8@@2ua&^K$p24{m&Lxf^lbjf;#_r2Ds4
zRr6Q@s!@!puuS%ZYHBrA4~M9ZN)hh%(qyPSs>-5LoxABs&eT3#(3)P1R-`&=3}taw
z`Fd`d^s`8{?ushvhq#-*m^Oy9Mzc)vIMfRkzrOh;)jqNNlNQasgl$L`y+4Ld958Fm
zPPnA}<0CGqg<g_JDOMdF>Pk_KWFD{1?w#J`P8Le302Ql@n(dnt^MDzwvT9==?a;1G
zpi^5ww`K#E8~5DuYUOeC58N-=AI*K^H%5)Iuu*%!X59e#braaCePFw80q5#AaK4@c
zF4XhDPQ3tJtUJJ^dJ(u>F9BETW#DSP0$i(Y;HlaHu4@Zo)dqLCyKmlLTEflh^yCcp
zxOZSq&K@w#JI5ix;2dvW0V$A6-h$`2Q~yxz)Nz!Z7rEG5PV!HqG=b5;kGf<Jm%Fh6
zb<ks?kdE%4l<R;#cK@U^hOgAd3h3t)`<;2;#tG&j%{`=re<t5Iw_jJrr}9;_HGK8s
zo!uf6J9$xwyG3zthlpUOFN@J|=gug9Kgs*?uE=-df{UFSNw0VPz1w;YEPO}8=0z#{
zQNBHV_?3S}`*M*DuYQ2-OFGzPd)wdejO{Q?@}vsG?JH>!M`?K#6`}!JdPw(v*dicH
zkmmvT3Zh=QL-iMwQ;vp0$~Ic6akQ7DCCV8_)ov;0Y4!qj%)echLRLwUzawQK<r0l7
z6RZ&EBZzV2DFRZZ+#uK_0LwFZhTts0If54mx&$49MSyaqZgu1<)0R^eGTv3ro{&YU
zoE|8zQtja=g(p`cFAE6@uuGBl<Q(mEEc26%|Bupwbemxqr^Mvp^!hK-sc!&C3FRdK
z+n6_e!^LmY1Wt>IH`Si}EZw%&tR-V_^ZB<6YjcvEiwE*WY#oz1&_f?fcl9HblDr%q
z0ju3$W3c?#dT_lmI6U_T`{b>${cy7~Csu797!#XY2gcap_SpK6ea!B!8(;(QfirQr
zb6^<k7wG#r%Ezp>N)8@yYx9@|*r36^d#0?bCVA|@-#5ojehsw-Z=nD8RqI6mg{nQ7
ztFi0Yshz<*B1Q8v!~$AJrDxtV%fD3~*u~*KMj3B?X6b0rKFuZzReR9ETDMvx_8EXV
z2Fv>v+AEV))YmZPVb03r6tTi!{T}<RAwR4=KBsFJr(?hkj!c8z2j;kuubt?hf6`xV
zg4aB5@P#j|aWj9tI*mQgpqwSv8=Rw3y)f8lF~UBvV;AK`NZt9OId1aB+FQpLZnCkD
z?|upV=-2qhDfUd9-F%MZFGF5a`p1(0^Mr)|tK_Yj<llm9uV~57qjvsllD~?+3p2@E
z-;kVG{b|WtQ;D>mx3si>`R2EfhEHmGrqM;M(ROul^6k2fJ8S23o3D+j=R39g^vzGU
zkrgb#60OP30fW81e!y_!w6=-0O<ilD)`Cqm5qH{rnXf!{z|pI>-j%Px8<chX=Jk&>
z<L#~XB4kIrTNPc(mAVP#WnK9)rpcEGUIA<${u@spHOBs4TstJ!I&=W=c6W#jt13`y
z<wZi-1BHIMjZsO_K(Q@HnY;IuH;wRb990n4b#R{Ldsg+J+ImA0J;?78(EiFUh9Xy9
zT;!jEU6oxWnE)@RLJZ}bG>gbrHVHJZ5q&E&A1VhKVe}AFcOzstZdB?_$J6Zi<|Wyn
z701M?KNs<Fw@6|kzejs`qde&qGSgJ8%wZy3TH+BrP3y|e5DVlnxho4AnbP-j$-PV4
zyaiCLcZsh*(mXAHKs|m>BJIsb+3=xiR?ucSEJ~s5xEMZ^-^YxtIeD4ZUjYQ(?){Kr
zk@Ct>9E-A4MyR}Dl7>mv4_rV9-5;fEl=%f7;&49-W980N-=Hg`A^9%tdz0Wea7&_Z
z)e6JDsFF%#VfZt2l)nZr%QtLe-Y|_0-bCCh#**oy2L)s2%yU-9YU{prbIs_OCg$0g
zZ|fDd(ZJ|B&A^K1%&=+3lF>HSuy5ORE!#McGknW6_Rc;lC>+bVWlRIYQ8Ghs3Jy9T
zkfBsWib(R{8pbU|5i1qBHbAV@8vdFRyfDO!F#Idl>a2qqdmGOZLC@M+q@*E_G7*N#
z55uhBBm55uRXYstkD_!of=zi5quVMgsjLM5Y-}V*sYJwL`o4ASoLY;Q7hzw*yy!cW
zJ2RVzX2#SrHw(S7NYkV6OCNy``XNoyp_uX+GJPG28<uUGxDoxs{X7Z3=F=_ASi0aE
zXMU;kp{;f0!lRJ2X;G%al3`4l<0q&oN87OcDfN=A2Q8hcY>%oWEhVLu%KGrV_ot4&
zsC&1$pwEjak9y*8sx=+d+%$Dv?eTPnfj=Fizc|y>%$xl$(sG`#V||(0m>XB7cX5?0
Vq9n)d7!CTFgl$+^2bSEn{sVd2*P8$U

literal 0
HcmV?d00001

diff --git a/dataset/batch_transform.py b/dataset/batch_transform.py
new file mode 100644
index 0000000..ef1a70e
--- /dev/null
+++ b/dataset/batch_transform.py
@@ -0,0 +1,374 @@
+from typing import Any, overload, Dict, Union, List, Sequence
+import random
+
+import torch
+from torch.nn import functional as F
+import numpy as np
+
+from utils.image import USMSharp, DiffJPEG, filter2D
+from utils.degradation import (
+    random_add_gaussian_noise_pt, random_add_poisson_noise_pt
+)
+
+
+class BatchTransform:
+    
+    @overload
+    def __call__(self, batch: Any) -> Any:
+        ...
+
+
+class IdentityBatchTransform(BatchTransform):
+    
+    def __call__(self, batch: Any) -> Any:
+        return batch
+
+
+class RealESRGANBatchTransform(BatchTransform):
+    """
+    It's too slow to process a batch of images under RealESRGAN degradation
+    model on CPU (by dataloader), which may cost 0.2 ~ 1 second per image.
+    So we execute the degradation process on GPU after loading a batch of images
+    and kernels from dataloader.
+    """
+    def __init__(
+        self,
+        use_sharpener: bool,
+        resize_hq: bool,
+        queue_size: int,
+        resize_prob: Sequence[float],
+        resize_range: Sequence[float],
+        gray_noise_prob: float,
+        gaussian_noise_prob: float,
+        noise_range: Sequence[float],
+        poisson_scale_range: Sequence[float],
+        jpeg_range: Sequence[int],
+        second_blur_prob: float,
+        stage2_scale: Union[float, Sequence[Union[float, int]]],
+        resize_prob2: Sequence[float],
+        resize_range2: Sequence[float],
+        gray_noise_prob2: float,
+        gaussian_noise_prob2: float,
+        noise_range2: Sequence[float],
+        poisson_scale_range2: Sequence[float],
+        jpeg_range2: Sequence[int]
+    ) -> "RealESRGANBatchTransform":
+        super().__init__()
+        # resize settings for the first degradation process
+        self.resize_prob = resize_prob
+        self.resize_range = resize_range
+        
+        # noise settings for the first degradation process
+        self.gray_noise_prob = gray_noise_prob
+        self.gaussian_noise_prob = gaussian_noise_prob
+        self.noise_range = noise_range
+        self.poisson_scale_range = poisson_scale_range
+        self.jpeg_range = jpeg_range
+        
+        self.second_blur_prob = second_blur_prob
+        self.stage2_scale = stage2_scale
+        assert (
+            isinstance(stage2_scale, (float, int)) or (
+                isinstance(stage2_scale, Sequence) and len(stage2_scale) == 2 and 
+                all(isinstance(x, (float, int)) for x in stage2_scale)
+            )
+        ), f"stage2_scale can not be {type(stage2_scale)}"
+        
+        # resize settings for the second degradation process
+        self.resize_prob2 = resize_prob2
+        self.resize_range2 = resize_range2
+        
+        # noise settings for the second degradation process
+        self.gray_noise_prob2 = gray_noise_prob2
+        self.gaussian_noise_prob2 = gaussian_noise_prob2
+        self.noise_range2 = noise_range2
+        self.poisson_scale_range2 = poisson_scale_range2
+        self.jpeg_range2 = jpeg_range2
+        
+        self.use_sharpener = use_sharpener
+        if self.use_sharpener:
+            self.usm_sharpener = USMSharp()
+        else:
+            self.usm_sharpener = None
+        self.resize_hq = resize_hq
+        self.queue_size = queue_size
+        self.jpeger = DiffJPEG(differentiable=False)
+
+    @torch.no_grad()
+    def _dequeue_and_enqueue(self):
+        """It is the training pair pool for increasing the diversity in a batch.
+
+        Batch processing limits the diversity of synthetic degradations in a batch. For example, samples in a
+        batch could not have different resize scaling factors. Therefore, we employ this training pair pool
+        to increase the degradation diversity in a batch.
+        """
+        # initialize
+        b, c, h, w = self.lq.size()
+        if not hasattr(self, "queue_lr"):
+            # TODO: Being multiple of batch_size seems not necessary for queue_size
+            assert self.queue_size % b == 0, f"queue size {self.queue_size} should be divisible by batch size {b}"
+            self.queue_lr = torch.zeros(self.queue_size, c, h, w).to(self.lq)
+            _, c, h, w = self.gt.size()
+            self.queue_gt = torch.zeros(self.queue_size, c, h, w).to(self.lq)
+            self.queue_ptr = 0
+        if self.queue_ptr == self.queue_size:  # the pool is full
+            # do dequeue and enqueue
+            # shuffle
+            idx = torch.randperm(self.queue_size)
+            self.queue_lr = self.queue_lr[idx]
+            self.queue_gt = self.queue_gt[idx]
+            # get first b samples
+            lq_dequeue = self.queue_lr[0:b, :, :, :].clone()
+            gt_dequeue = self.queue_gt[0:b, :, :, :].clone()
+            # update the queue
+            self.queue_lr[0:b, :, :, :] = self.lq.clone()
+            self.queue_gt[0:b, :, :, :] = self.gt.clone()
+
+            self.lq = lq_dequeue
+            self.gt = gt_dequeue
+        else:
+            # only do enqueue
+            self.queue_lr[self.queue_ptr:self.queue_ptr + b, :, :, :] = self.lq.clone()
+            self.queue_gt[self.queue_ptr:self.queue_ptr + b, :, :, :] = self.gt.clone()
+            self.queue_ptr = self.queue_ptr + b
+
+    @torch.no_grad()
+    def __call__(self, batch: Dict[str, Union[torch.Tensor, str]]) -> Dict[str, Union[torch.Tensor, List[str]]]:
+        # training data synthesis
+        hq = batch["hq"]
+        if self.use_sharpener:
+            self.usm_sharpener.to(hq)
+            hq = self.usm_sharpener(hq)
+        self.jpeger.to(hq)
+        
+        kernel1 = batch["kernel1"]
+        kernel2 = batch["kernel2"]
+        sinc_kernel = batch["sinc_kernel"]
+
+        ori_h, ori_w = hq.size()[2:4]
+
+        # ----------------------- The first degradation process ----------------------- #
+        # blur
+        out = filter2D(hq, kernel1)
+        # random resize
+        updown_type = random.choices(["up", "down", "keep"], self.resize_prob)[0]
+        if updown_type == "up":
+            scale = np.random.uniform(1, self.resize_range[1])
+        elif updown_type == "down":
+            scale = np.random.uniform(self.resize_range[0], 1)
+        else:
+            scale = 1
+        mode = random.choice(["area", "bilinear", "bicubic"])
+        out = F.interpolate(out, scale_factor=scale, mode=mode)
+        # add noise
+        if np.random.uniform() < self.gaussian_noise_prob:
+            out = random_add_gaussian_noise_pt(
+                out, sigma_range=self.noise_range, clip=True,
+                rounds=False, gray_prob=self.gray_noise_prob
+            )
+        else:
+            out = random_add_poisson_noise_pt(
+                out,
+                scale_range=self.poisson_scale_range,
+                gray_prob=self.gray_noise_prob,
+                clip=True,
+                rounds=False
+            )
+        # JPEG compression
+        jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.jpeg_range)
+        # clamp to [0, 1], otherwise JPEGer will result in unpleasant artifacts
+        out = torch.clamp(out, 0, 1)
+        out = self.jpeger(out, quality=jpeg_p)
+
+        # ----------------------- The second degradation process ----------------------- #
+        # blur
+        if np.random.uniform() < self.second_blur_prob:
+            out = filter2D(out, kernel2)
+        
+        # select scale of second degradation stage
+        if isinstance(self.stage2_scale, Sequence):
+            min_scale, max_scale = self.stage2_scale
+            stage2_scale = np.random.uniform(min_scale, max_scale)
+        else:
+            stage2_scale = self.stage2_scale
+        stage2_h, stage2_w = int(ori_h / stage2_scale), int(ori_w / stage2_scale)
+        # print(f"stage2 scale = {stage2_scale}")
+        
+        # random resize
+        updown_type = random.choices(["up", "down", "keep"], self.resize_prob2)[0]
+        if updown_type == "up":
+            scale = np.random.uniform(1, self.resize_range2[1])
+        elif updown_type == "down":
+            scale = np.random.uniform(self.resize_range2[0], 1)
+        else:
+            scale = 1
+        mode = random.choice(["area", "bilinear", "bicubic"])
+        out = F.interpolate(
+            out, size=(int(stage2_h * scale), int(stage2_w * scale)), mode=mode
+        )
+        # add noise
+        if np.random.uniform() < self.gaussian_noise_prob2:
+            out = random_add_gaussian_noise_pt(
+                out, sigma_range=self.noise_range2, clip=True,
+                rounds=False, gray_prob=self.gray_noise_prob2
+            )
+        else:
+            out = random_add_poisson_noise_pt(
+                out,
+                scale_range=self.poisson_scale_range2,
+                gray_prob=self.gray_noise_prob2,
+                clip=True,
+                rounds=False
+            )
+
+        # JPEG compression + the final sinc filter
+        # We also need to resize images to desired sizes. We group [resize back + sinc filter] together
+        # as one operation.
+        # We consider two orders:
+        #   1. [resize back + sinc filter] + JPEG compression
+        #   2. JPEG compression + [resize back + sinc filter]
+        # Empirically, we find other combinations (sinc + JPEG + Resize) will introduce twisted lines.
+        if np.random.uniform() < 0.5:
+            # resize back + the final sinc filter
+            mode = random.choice(["area", "bilinear", "bicubic"])
+            out = F.interpolate(out, size=(stage2_h, stage2_w), mode=mode)
+            out = filter2D(out, sinc_kernel)
+            # JPEG compression
+            jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.jpeg_range2)
+            out = torch.clamp(out, 0, 1)
+            out = self.jpeger(out, quality=jpeg_p)
+        else:
+            # JPEG compression
+            jpeg_p = out.new_zeros(out.size(0)).uniform_(*self.jpeg_range2)
+            out = torch.clamp(out, 0, 1)
+            out = self.jpeger(out, quality=jpeg_p)
+            # resize back + the final sinc filter
+            mode = random.choice(["area", "bilinear", "bicubic"])
+            out = F.interpolate(out, size=(stage2_h, stage2_w), mode=mode)
+            out = filter2D(out, sinc_kernel)
+
+        # resize back to gt_size since We are doing restoration task
+        if stage2_scale != 1:
+            out = F.interpolate(out, size=(ori_h, ori_w), mode="bicubic")
+        # clamp and round
+        lq = torch.clamp((out * 255.0).round(), 0, 255) / 255.
+        
+        if self.resize_hq and stage2_scale != 1:
+            # resize hq
+            hq = F.interpolate(hq, size=(stage2_h, stage2_w), mode="bicubic", antialias=True)
+            hq = F.interpolate(hq, size=(ori_h, ori_w), mode="bicubic", antialias=True)
+        
+        self.gt = hq
+        self.lq = lq
+        self._dequeue_and_enqueue()
+        
+        # [0, 1], float32, rgb, nhwc
+        lq = self.lq.float().permute(0, 2, 3, 1).contiguous()
+        # [-1, 1], float32, rgb, nhwc
+        hq = (self.gt * 2 - 1).float().permute(0, 2, 3, 1).contiguous()
+        
+        return dict(jpg=hq, hint=lq, txt=batch["txt"])
+
+
+class BicubicBatchTransform(BatchTransform):
+    """
+    It's too slow to process a batch of images under RealESRGAN degradation
+    model on CPU (by dataloader), which may cost 0.2 ~ 1 second per image.
+    So we execute the degradation process on GPU after loading a batch of images
+    and kernels from dataloader.
+    """
+    def __init__(
+        self,
+        use_sharpener: bool,
+        resize_hq: bool,
+        queue_size: int,
+        scale: Union[float, Sequence[Union[float, int]]],
+    ) -> "BicubicBatchTransform":
+        super().__init__()
+        
+        self.scale = scale
+        assert (
+            isinstance(scale, (float, int)) or (
+                isinstance(scale, Sequence) and len(scale) == 2 and 
+                all(isinstance(x, (float, int)) for x in scale)
+            )
+        ), f"scale can not be {type(scale)}"
+        
+        self.use_sharpener = use_sharpener
+        if self.use_sharpener:
+            self.usm_sharpener = USMSharp()
+        else:
+            self.usm_sharpener = None
+        self.resize_hq = resize_hq
+        self.queue_size = queue_size
+
+
+    @torch.no_grad()
+    def _dequeue_and_enqueue(self):
+        """It is the training pair pool for increasing the diversity in a batch.
+
+        Batch processing limits the diversity of synthetic degradations in a batch. For example, samples in a
+        batch could not have different resize scaling factors. Therefore, we employ this training pair pool
+        to increase the degradation diversity in a batch.
+        """
+        # initialize
+        b, c, h, w = self.lq.size()
+        if not hasattr(self, "queue_lr"):
+            # TODO: Being multiple of batch_size seems not necessary for queue_size
+            assert self.queue_size % b == 0, f"queue size {self.queue_size} should be divisible by batch size {b}"
+            self.queue_lr = torch.zeros(self.queue_size, c, h, w).to(self.lq)
+            _, c, h, w = self.gt.size()
+            self.queue_gt = torch.zeros(self.queue_size, c, h, w).to(self.lq)
+            self.queue_ptr = 0
+        if self.queue_ptr == self.queue_size:  # the pool is full
+            # do dequeue and enqueue
+            # shuffle
+            idx = torch.randperm(self.queue_size)
+            self.queue_lr = self.queue_lr[idx]
+            self.queue_gt = self.queue_gt[idx]
+            # get first b samples
+            lq_dequeue = self.queue_lr[0:b, :, :, :].clone()
+            gt_dequeue = self.queue_gt[0:b, :, :, :].clone()
+            # update the queue
+            self.queue_lr[0:b, :, :, :] = self.lq.clone()
+            self.queue_gt[0:b, :, :, :] = self.gt.clone()
+
+            self.lq = lq_dequeue
+            self.gt = gt_dequeue
+        else:
+            # only do enqueue
+            self.queue_lr[self.queue_ptr:self.queue_ptr + b, :, :, :] = self.lq.clone()
+            self.queue_gt[self.queue_ptr:self.queue_ptr + b, :, :, :] = self.gt.clone()
+            self.queue_ptr = self.queue_ptr + b
+
+    @torch.no_grad()
+    def __call__(self, batch: Dict[str, Union[torch.Tensor, str]]) -> Dict[str, Union[torch.Tensor, List[str]]]:
+        # training data synthesis
+        hq = batch["hq"]
+        if self.use_sharpener:
+            self.usm_sharpener.to(hq)
+            hq = self.usm_sharpener(hq)
+        
+        ori_h, ori_w = hq.size()[2:4]
+        h, w = int(ori_h / self.scale), int(ori_w / self.scale)
+        
+        # if self.resize_hq and self.scale != 1:
+        #     # resize hq
+        #     lq = F.interpolate(hq, size=(h, w), mode="bicubic", antialias=True)
+            
+        out = F.interpolate(hq, size=(h, w), mode="bicubic", antialias=True)
+        out = F.interpolate(out, size=(ori_h, ori_w), mode="bicubic", antialias=True)
+
+        # clamp and round
+        lq = torch.clamp((out * 255.0).round(), 0, 255) / 255.
+        
+        self.gt = hq
+        self.lq = lq
+        self._dequeue_and_enqueue()
+        
+        # [0, 1], float32, rgb, nhwc
+        lq = self.lq.float().permute(0, 2, 3, 1).contiguous()
+        # [-1, 1], float32, rgb, nhwc
+        hq = (self.gt * 2 - 1).float().permute(0, 2, 3, 1).contiguous()
+        
+        return dict(jpg=hq, hint=lq, txt=batch["txt"])
\ No newline at end of file
diff --git a/dataset/bicubic_torchvision.py b/dataset/bicubic_torchvision.py
new file mode 100644
index 0000000..c54f513
--- /dev/null
+++ b/dataset/bicubic_torchvision.py
@@ -0,0 +1,107 @@
+from typing import Dict, Sequence
+import math
+import random
+import time
+
+import numpy as np
+import torch
+from torch.utils import data
+from PIL import Image
+
+from utils.degradation import circular_lowpass_kernel, random_mixed_kernels
+from utils.image import augment, random_crop_arr, center_crop_arr
+from utils.file import load_file_list
+
+
+class BicubicDataset(data.Dataset):
+    """
+    # TODO: add comment
+    """
+
+    def __init__(
+        self,
+        file_list: str,
+        out_size: int,
+        crop_type: str,
+        use_hflip: bool,
+        use_rot: bool
+    ) -> "BicubicDataset":
+        super(BicubicDataset, self).__init__()
+        self.paths = load_file_list(file_list)
+        self.out_size = out_size
+        self.crop_type = crop_type
+        assert self.crop_type in ["center", "random", "none"], f"invalid crop type: {self.crop_type}"
+
+        # self.blur_kernel_size = blur_kernel_size
+        # self.kernel_list = kernel_list
+        # # a list for each kernel probability
+        # self.kernel_prob = kernel_prob
+        # self.blur_sigma = blur_sigma
+        # # betag used in generalized Gaussian blur kernels
+        # self.betag_range = betag_range
+        # # betap used in plateau blur kernels
+        # self.betap_range = betap_range
+        # # the probability for sinc filters
+        # self.sinc_prob = sinc_prob
+
+        # self.blur_kernel_size2 = blur_kernel_size2
+        # self.kernel_list2 = kernel_list2
+        # self.kernel_prob2 = kernel_prob2
+        # self.blur_sigma2 = blur_sigma2
+        # self.betag_range2 = betag_range2
+        # self.betap_range2 = betap_range2
+        # self.sinc_prob2 = sinc_prob2
+        
+        # # a final sinc filter
+        # self.final_sinc_prob = final_sinc_prob
+        
+        self.use_hflip = use_hflip
+        self.use_rot = use_rot
+        
+        # kernel size ranges from 7 to 21
+        # self.kernel_range = [2 * v + 1 for v in range(3, 11)]
+        # # TODO: kernel range is now hard-coded, should be in the configure file
+        # # convolving with pulse tensor brings no blurry effect
+        # self.pulse_tensor = torch.zeros(21, 21).float()
+        # self.pulse_tensor[10, 10] = 1
+
+    @torch.no_grad()
+    def __getitem__(self, index: int) -> Dict[str, torch.Tensor]:
+        # -------------------------------- Load hq images -------------------------------- #
+        hq_path = self.paths[index]
+        success = False
+        for _ in range(3):
+            try:
+                pil_img = Image.open(hq_path).convert("RGB")
+                success = True
+                break
+            except:
+                time.sleep(1)
+        assert success, f"failed to load image {hq_path}"
+        
+        if self.crop_type == "random":
+            pil_img = random_crop_arr(pil_img, self.out_size)
+        elif self.crop_type == "center":
+            pil_img = center_crop_arr(pil_img, self.out_size)
+        # self.crop_type is "none"
+        else:
+            pil_img = np.array(pil_img)
+            assert pil_img.shape[:2] == (self.out_size, self.out_size)
+        # hwc, rgb to bgr, [0, 255] to [0, 1], float32
+        img_hq = (pil_img[..., ::-1] / 255.0).astype(np.float32)
+
+        # -------------------- Do augmentation for training: flip, rotation -------------------- #
+        img_hq = augment(img_hq, self.use_hflip, self.use_rot)
+        
+        # [0, 1], BGR to RGB, HWC to CHW
+        img_hq = torch.from_numpy(
+            img_hq[..., ::-1].transpose(2, 0, 1).copy()
+        ).float()
+
+        return {
+            "hq": img_hq,
+            'txt': ""
+        }
+
+    def __len__(self) -> int:
+        return len(self.paths)
diff --git a/dataset/codeformer.py b/dataset/codeformer.py
new file mode 100644
index 0000000..f002d86
--- /dev/null
+++ b/dataset/codeformer.py
@@ -0,0 +1,109 @@
+from typing import Sequence, Dict, Union
+import math
+import time
+
+import numpy as np
+import cv2
+from PIL import Image
+import torch.utils.data as data
+
+from utils.file import load_file_list
+from utils.image import center_crop_arr, augment, random_crop_arr
+from utils.degradation import (
+    random_mixed_kernels, random_add_gaussian_noise, random_add_jpg_compression
+)
+
+
+class CodeformerDataset(data.Dataset):
+    
+    def __init__(
+        self,
+        file_list: str,
+        out_size: int,
+        crop_type: str,
+        use_hflip: bool,
+        blur_kernel_size: int,
+        kernel_list: Sequence[str],
+        kernel_prob: Sequence[float],
+        blur_sigma: Sequence[float],
+        downsample_range: Sequence[float],
+        noise_range: Sequence[float],
+        jpeg_range: Sequence[int]
+    ) -> "CodeformerDataset":
+        super(CodeformerDataset, self).__init__()
+        self.file_list = file_list
+        self.paths = load_file_list(file_list)
+        self.out_size = out_size
+        self.crop_type = crop_type
+        assert self.crop_type in ["none", "center", "random"]
+        self.use_hflip = use_hflip
+        # degradation configurations
+        self.blur_kernel_size = blur_kernel_size
+        self.kernel_list = kernel_list
+        self.kernel_prob = kernel_prob
+        self.blur_sigma = blur_sigma
+        self.downsample_range = downsample_range
+        self.noise_range = noise_range
+        self.jpeg_range = jpeg_range
+
+    def __getitem__(self, index: int) -> Dict[str, Union[np.ndarray, str]]:
+        # load gt image
+        # Shape: (h, w, c); channel order: BGR; image range: [0, 1], float32.
+        gt_path = self.paths[index]
+        success = False
+        for _ in range(3):
+            try:
+                pil_img = Image.open(gt_path).convert("RGB")
+                success = True
+                break
+            except:
+                time.sleep(1)
+        assert success, f"failed to load image {gt_path}"
+        
+        if self.crop_type == "center":
+            pil_img_gt = center_crop_arr(pil_img, self.out_size)
+        elif self.crop_type == "random":
+            pil_img_gt = random_crop_arr(pil_img, self.out_size)
+        else:
+            pil_img_gt = np.array(pil_img)
+            assert pil_img_gt.shape[:2] == (self.out_size, self.out_size)
+        img_gt = (pil_img_gt[..., ::-1] / 255.0).astype(np.float32)
+        
+        # random horizontal flip
+        img_gt = augment(img_gt, hflip=self.use_hflip, rotation=False, return_status=False)
+        h, w, _ = img_gt.shape
+        
+        # ------------------------ generate lq image ------------------------ #
+        # blur
+        kernel = random_mixed_kernels(
+            self.kernel_list,
+            self.kernel_prob,
+            self.blur_kernel_size,
+            self.blur_sigma,
+            self.blur_sigma,
+            [-math.pi, math.pi],
+            noise_range=None
+        )
+        img_lq = cv2.filter2D(img_gt, -1, kernel)
+        # downsample
+        scale = np.random.uniform(self.downsample_range[0], self.downsample_range[1])
+        img_lq = cv2.resize(img_lq, (int(w // scale), int(h // scale)), interpolation=cv2.INTER_LINEAR)
+        # noise
+        if self.noise_range is not None:
+            img_lq = random_add_gaussian_noise(img_lq, self.noise_range)
+        # jpeg compression
+        if self.jpeg_range is not None:
+            img_lq = random_add_jpg_compression(img_lq, self.jpeg_range)
+        
+        # resize to original size
+        img_lq = cv2.resize(img_lq, (w, h), interpolation=cv2.INTER_LINEAR)
+        
+        # BGR to RGB, [-1, 1]
+        target = (img_gt[..., ::-1] * 2 - 1).astype(np.float32)
+        # BGR to RGB, [0, 1]
+        source = img_lq[..., ::-1].astype(np.float32)
+        
+        return dict(jpg=target, txt="", hint=source)
+
+    def __len__(self) -> int:
+        return len(self.paths)
diff --git a/dataset/data_module.py b/dataset/data_module.py
new file mode 100644
index 0000000..d4dcde3
--- /dev/null
+++ b/dataset/data_module.py
@@ -0,0 +1,68 @@
+from typing import Any, Tuple, Mapping
+
+from pytorch_lightning.utilities.types import EVAL_DATALOADERS, TRAIN_DATALOADERS
+import pytorch_lightning as pl
+from torch.utils.data import DataLoader, Dataset
+from omegaconf import OmegaConf
+
+from utils.common import instantiate_from_config
+from dataset.batch_transform import BatchTransform, IdentityBatchTransform
+
+
+class BIRDataModule(pl.LightningDataModule):
+    
+    def __init__(
+        self,
+        train_config: str,
+        val_config: str=None
+    ) -> "BIRDataModule":
+        super().__init__()
+        self.train_config = OmegaConf.load(train_config)
+        self.val_config = OmegaConf.load(val_config) if val_config else None
+
+    def load_dataset(self, config: Mapping[str, Any]) -> Tuple[Dataset, BatchTransform]:
+        dataset = instantiate_from_config(config["dataset"])
+        batch_transform = (
+            instantiate_from_config(config["batch_transform"])
+            if config.get("batch_transform") else IdentityBatchTransform()
+        )
+        return dataset, batch_transform
+
+    def setup(self, stage: str) -> None:
+        if stage == "fit":
+            self.train_dataset, self.train_batch_transform = self.load_dataset(self.train_config)
+            if self.val_config:
+                self.val_dataset, self.val_batch_transform = self.load_dataset(self.val_config)
+            else:
+                self.val_dataset, self.val_batch_transform = None, None
+        else:
+            raise NotImplementedError(stage)
+
+    def train_dataloader(self) -> TRAIN_DATALOADERS:
+        return DataLoader(
+            dataset=self.train_dataset, **self.train_config["data_loader"]
+        )
+
+    def val_dataloader(self) -> EVAL_DATALOADERS:
+        if self.val_dataset is None:
+            return None
+        return DataLoader(
+            dataset=self.val_dataset, **self.val_config["data_loader"]
+        )
+
+    def on_after_batch_transfer(self, batch: Any, dataloader_idx: int) -> Any:
+        self.trainer: pl.Trainer
+        
+        if self.trainer.training:
+            return self.train_batch_transform(batch)
+        elif self.trainer.validating or self.trainer.sanity_checking:
+            return self.val_batch_transform(batch)
+        else:
+            raise RuntimeError(
+                "Trainer state: \n"
+                f"training: {self.trainer.training}\n"
+                f"validating: {self.trainer.validating}\n"
+                f"testing: {self.trainer.testing}\n"
+                f"predicting: {self.trainer.predicting}\n"
+                f"sanity_checking: {self.trainer.sanity_checking}"
+            )
diff --git a/dataset/projectdata.py b/dataset/projectdata.py
new file mode 100644
index 0000000..3e139be
--- /dev/null
+++ b/dataset/projectdata.py
@@ -0,0 +1,121 @@
+from typing import Dict, Sequence
+import math
+import random
+import time
+import glob
+import os
+import cv2
+
+import numpy as np
+import torch
+from torch.utils import data
+from PIL import Image
+
+from utils.degradation import circular_lowpass_kernel, random_mixed_kernels
+from utils.image import augment, random_crop_arr, center_crop_arr
+from utils.file import load_file_list
+
+
+class PROJECTDataset(data.Dataset):
+    """
+    # TODO: add comment
+    """
+
+    def __init__(self, 
+                file_path: list,
+                out_size: int,
+                scale: int,
+                hr_pattern: str,
+                blur_sigma: float,
+                blur_kernel: int,
+                crop_type: str,
+                use_hflip: bool,
+                use_rot: bool,
+                ):
+        super().__init__()
+        
+        self.patch_size = out_size // 4
+        self.scale = scale
+        
+        lq_files = []
+        hr_files = []
+        for path in file_path:
+            lq_files += glob.glob(os.path.join(path, "*phone.*"))
+            hr_files += glob.glob(os.path.join(path, "*screen.*"))
+
+        self.lq_files = lq_files
+        self.hr_files = hr_files
+        self.lq_files.sort()
+        self.hr_files.sort()
+
+        self.hr_pattern = hr_pattern
+        self.blur_sigma = blur_sigma
+        self.blur_kernel = blur_kernel
+        self.noise_scale1 = 0.002
+        self.noise_scale2 = 0.003
+        self.blur_kernal_list = [ i for i in range(1, self.blur_kernel * 2, 2)]
+        self.blur_sigma_list = [ i / 10 for i in range(1, int(2 * self.blur_sigma * 10), 1)]
+
+        self.edge_pixel = 0
+
+    @torch.no_grad()
+    def __len__(self):
+        return len(self.lq_files)
+
+    def __getitem__(self, index):
+        lowRes_file = self.lq_files[index]
+        highRes_file = self.hr_files[index]
+
+        hightRes_img = Image.open(highRes_file).convert("RGB")
+        hightRes_img = np.array(hightRes_img)
+        hightRes_img = (hightRes_img[..., ::-1] / 255.0).astype(np.float32)
+        lowRes_img = Image.open(lowRes_file).convert("RGB")
+        lowRes_img = np.array(lowRes_img)
+        lowRes_img = (lowRes_img[..., ::-1] / 255.0).astype(np.float32)
+        
+        low_h = lowRes_img.shape[0]
+        high_h = hightRes_img.shape[0]
+        if low_h == high_h:
+            lowRes_img = cv2.resize(lowRes_img, None, fx=1 / self.scale, fy=1 / self.scale)
+
+        # scale为4数据集用作2x训练
+        if high_h // low_h == 4 and self.scale == 2:
+            hightRes_img = cv2.resize(hightRes_img, None, fx=1 / 2, fy=1 / 2)
+        
+        # scale为2数据集用作4x训练
+        if "20231021_plant_lightroom_Crop320" in lowRes_file and self.scale == 4:
+            lowRes_img = cv2.resize(lowRes_img, None, fx=1 / 2, fy=1 / 2)
+
+
+        img_size_ori_h, img_size_ori_w = lowRes_img.shape[:2]
+        i = random.randint(self.edge_pixel, img_size_ori_h - self.patch_size - self.edge_pixel)
+        j = random.randint(self.edge_pixel, img_size_ori_w - self.patch_size - self.edge_pixel)
+        hightRes_img = hightRes_img[i * self.scale : i * self.scale + self.patch_size * self.scale, j * self.scale : j * self.scale + self.patch_size * self.scale]
+        lowRes_img = lowRes_img[i : i + self.patch_size, j : j + self.patch_size]
+
+        if random.uniform(0, 1) < 0.5:
+            hightRes_img = hightRes_img[::-1]
+            lowRes_img = lowRes_img[::-1]
+        if random.uniform(0, 1) < 0.5:
+            hightRes_img = hightRes_img[:,::-1]
+            lowRes_img = lowRes_img[:,::-1]
+        if 1:
+            sigma_x_offset = random.choice(self.blur_sigma_list)
+            sigma_y_offset = random.choice(self.blur_sigma_list)
+            kernal_x_offset = random.choice(self.blur_kernal_list)
+            kernal_y_offset = random.choice(self.blur_kernal_list)
+            lowRes_img = cv2.GaussianBlur(lowRes_img, (kernal_x_offset, kernal_y_offset), sigmaX=sigma_x_offset, sigmaY=sigma_y_offset)
+            # lowRes_img = lowRes_img * np.random.normal(loc=1, scale=self.noise_scale1, size=lowRes_img.shape)  +  np.random.normal(loc=0, scale=self.noise_scale2, size=lowRes_img.shape)
+
+        hightRes_img = np.ascontiguousarray(hightRes_img)
+        lowRes_img = np.ascontiguousarray(lowRes_img)        
+
+        # lowRes_img = lowRes_img[:,:,None]
+        # hightRes_img = hightRes_img[:,:,None]
+        lowRes_img = lowRes_img.clip(0, 1)
+        hightRes_img = hightRes_img.clip(0, 1)
+
+        lowRes_img = torch.from_numpy(np.ascontiguousarray(np.transpose(np.stack(lowRes_img, axis=0), (2, 0, 1)))).float()
+        hightRes_img = torch.from_numpy(np.ascontiguousarray(np.transpose(np.stack(hightRes_img, axis=0), (2, 0, 1)))).float()
+
+        return {'lq': lowRes_img, 'hq': hightRes_img, 'txt': '', 'file_name': lowRes_file}
\ No newline at end of file
diff --git a/dataset/realesrgan.py b/dataset/realesrgan.py
new file mode 100644
index 0000000..2a4a812
--- /dev/null
+++ b/dataset/realesrgan.py
@@ -0,0 +1,183 @@
+from typing import Dict, Sequence
+import math
+import random
+import time
+
+import numpy as np
+import torch
+from torch.utils import data
+from PIL import Image
+
+from utils.degradation import circular_lowpass_kernel, random_mixed_kernels
+from utils.image import augment, random_crop_arr, center_crop_arr
+from utils.file import load_file_list
+
+
+class RealESRGANDataset(data.Dataset):
+    """
+    # TODO: add comment
+    """
+
+    def __init__(
+        self,
+        file_list: str,
+        out_size: int,
+        crop_type: str,
+        use_hflip: bool,
+        use_rot: bool,
+        # blur kernel settings of the first degradation stage
+        blur_kernel_size: int,
+        kernel_list: Sequence[str],
+        kernel_prob: Sequence[float],
+        blur_sigma: Sequence[float],
+        betag_range: Sequence[float],
+        betap_range: Sequence[float],
+        sinc_prob: float,
+        # blur kernel settings of the second degradation stage
+        blur_kernel_size2: int,
+        kernel_list2: Sequence[str],
+        kernel_prob2: Sequence[float],
+        blur_sigma2: Sequence[float],
+        betag_range2: Sequence[float],
+        betap_range2: Sequence[float],
+        sinc_prob2: float,
+        final_sinc_prob: float
+    ) -> "RealESRGANDataset":
+        super(RealESRGANDataset, self).__init__()
+        self.paths = load_file_list(file_list)
+        self.out_size = out_size
+        self.crop_type = crop_type
+        assert self.crop_type in ["center", "random", "none"], f"invalid crop type: {self.crop_type}"
+
+        self.blur_kernel_size = blur_kernel_size
+        self.kernel_list = kernel_list
+        # a list for each kernel probability
+        self.kernel_prob = kernel_prob
+        self.blur_sigma = blur_sigma
+        # betag used in generalized Gaussian blur kernels
+        self.betag_range = betag_range
+        # betap used in plateau blur kernels
+        self.betap_range = betap_range
+        # the probability for sinc filters
+        self.sinc_prob = sinc_prob
+
+        self.blur_kernel_size2 = blur_kernel_size2
+        self.kernel_list2 = kernel_list2
+        self.kernel_prob2 = kernel_prob2
+        self.blur_sigma2 = blur_sigma2
+        self.betag_range2 = betag_range2
+        self.betap_range2 = betap_range2
+        self.sinc_prob2 = sinc_prob2
+        
+        # a final sinc filter
+        self.final_sinc_prob = final_sinc_prob
+        
+        self.use_hflip = use_hflip
+        self.use_rot = use_rot
+        
+        # kernel size ranges from 7 to 21
+        self.kernel_range = [2 * v + 1 for v in range(3, 11)]
+        # TODO: kernel range is now hard-coded, should be in the configure file
+        # convolving with pulse tensor brings no blurry effect
+        self.pulse_tensor = torch.zeros(21, 21).float()
+        self.pulse_tensor[10, 10] = 1
+
+    @torch.no_grad()
+    def __getitem__(self, index: int) -> Dict[str, torch.Tensor]:
+        # -------------------------------- Load hq images -------------------------------- #
+        hq_path = self.paths[index]
+        success = False
+        for _ in range(3):
+            try:
+                pil_img = Image.open(hq_path).convert("RGB")
+                success = True
+                break
+            except:
+                time.sleep(1)
+        assert success, f"failed to load image {hq_path}"
+        
+        if self.crop_type == "random":
+            pil_img = random_crop_arr(pil_img, self.out_size)
+        elif self.crop_type == "center":
+            pil_img = center_crop_arr(pil_img, self.out_size)
+        # self.crop_type is "none"
+        else:
+            pil_img = np.array(pil_img)
+            assert pil_img.shape[:2] == (self.out_size, self.out_size)
+        # hwc, rgb to bgr, [0, 255] to [0, 1], float32
+        img_hq = (pil_img[..., ::-1] / 255.0).astype(np.float32)
+
+        # -------------------- Do augmentation for training: flip, rotation -------------------- #
+        img_hq = augment(img_hq, self.use_hflip, self.use_rot)
+        
+        # ------------------------ Generate kernels (used in the first degradation) ------------------------ #
+        kernel_size = random.choice(self.kernel_range)
+        if np.random.uniform() < self.sinc_prob:
+            # this sinc filter setting is for kernels ranging from [7, 21]
+            if kernel_size < 13:
+                omega_c = np.random.uniform(np.pi / 3, np.pi)
+            else:
+                omega_c = np.random.uniform(np.pi / 5, np.pi)
+            kernel = circular_lowpass_kernel(omega_c, kernel_size, pad_to=False)
+        else:
+            kernel = random_mixed_kernels(
+                self.kernel_list,
+                self.kernel_prob,
+                kernel_size,
+                self.blur_sigma,
+                self.blur_sigma, [-math.pi, math.pi],
+                self.betag_range,
+                self.betap_range,
+                noise_range=None
+            )
+        # pad kernel
+        pad_size = (21 - kernel_size) // 2
+        kernel = np.pad(kernel, ((pad_size, pad_size), (pad_size, pad_size)))
+
+        # ------------------------ Generate kernels (used in the second degradation) ------------------------ #
+        kernel_size = random.choice(self.kernel_range)
+        if np.random.uniform() < self.sinc_prob2:
+            if kernel_size < 13:
+                omega_c = np.random.uniform(np.pi / 3, np.pi)
+            else:
+                omega_c = np.random.uniform(np.pi / 5, np.pi)
+            kernel2 = circular_lowpass_kernel(omega_c, kernel_size, pad_to=False)
+        else:
+            kernel2 = random_mixed_kernels(
+                self.kernel_list2,
+                self.kernel_prob2,
+                kernel_size,
+                self.blur_sigma2,
+                self.blur_sigma2, [-math.pi, math.pi],
+                self.betag_range2,
+                self.betap_range2,
+                noise_range=None
+            )
+
+        # pad kernel
+        pad_size = (21 - kernel_size) // 2
+        kernel2 = np.pad(kernel2, ((pad_size, pad_size), (pad_size, pad_size)))
+
+        # ------------------------------------- the final sinc kernel ------------------------------------- #
+        if np.random.uniform() < self.final_sinc_prob:
+            kernel_size = random.choice(self.kernel_range)
+            omega_c = np.random.uniform(np.pi / 3, np.pi)
+            sinc_kernel = circular_lowpass_kernel(omega_c, kernel_size, pad_to=21)
+            sinc_kernel = torch.FloatTensor(sinc_kernel)
+        else:
+            sinc_kernel = self.pulse_tensor
+
+        # [0, 1], BGR to RGB, HWC to CHW
+        img_hq = torch.from_numpy(
+            img_hq[..., ::-1].transpose(2, 0, 1).copy()
+        ).float()
+        kernel = torch.FloatTensor(kernel)
+        kernel2 = torch.FloatTensor(kernel2)
+
+        return {
+            "hq": img_hq, "kernel1": kernel, "kernel2": kernel2,
+            "sinc_kernel": sinc_kernel, "txt": ""
+        }
+
+    def __len__(self) -> int:
+        return len(self.paths)
diff --git a/figs/logo.png b/figs/logo.png
index 79c46f3fefbfc5678c44d2bd1e7d3bb8c254b6a0..aca7662a8d41a7949ff751c848cbf5f70dfae85e 100644
GIT binary patch
literal 30467
zcmXt=1yq!Mu*R1XkX$-MV5Ovy2I-|sVnMn=K|)fxVQB%STdAd_I|M1|?vhx#JMa75
zdoJg&Ea$);-g*B!^UOTIA>y5yJl=Dv=O7RWPf-D?2?C)q0na~UJp&%s=0DH@U(j8o
z6}7Q|4<9T`IPg8TlY*WL2t+UR-zSP}5|1eeL<drYN@;&F-f#8HFq(lO?+omYT!v{%
zcBg+PAhgpy4aCxPCs}mMBtTqHu>%>B(O;V8>;ySvfx4G6{b&{<rKw*2TwkNSy(qkG
zEx5x|>7K;hDZr-x*M%l+PK?VVR7Y$OPot+<{4?!H%%<YvaTaO3B*Lm#ubpOl?I(T_
zUpvxRztMa+cUM1C6No1;Ya0B&4}0i;@jp4ueekK%F4IZA&*<&#B@M9b6@35QaCEu2
zjQ_!IedPas`b`FJpo3FCAxOIfR<6zZLKZw@%RezU_c0|E#nj9!(77=KHK4}nd|Sig
z{`#}9&-E`;nW9oMvbkoRlGNp8<F3~@I5_AaWVIk@2BISn^7k3_r(aTk!?^V*hq|!?
zinUqmG>g#(Uoa^Wal5cU0&Ay1rpOnR7L&q-gZm%`h^EKzR4^TUU8r1LU45pvr{{$L
z`rivIl$`i@@N%_H9Cp{JzTJW5rTZ`?h`WacqydwPYE5J)9vK;#LHSeMDv18KcKwCp
z%*lE$Hpu;G+4r~b>6}gW)Z?QchCqvXNwiM0z}5-QJk{*4lc|4sVY8-+Mf7rE;o-py
zil}z>DFd6QpI?xtG)h7-Ago2US*d6s9Vk2%27M6_0=FG*^n|35#;4L@S#KecHa6DQ
zTUoxBbGK_<;faGsORMveQ?}nRpFR6T5hj2Y#I2anN0De(TkBFT2^9v>^QP-~6Kk+T
z_#tV!Q^Md>`n->jw6ZB%7x}oCWpQzFI08(~Y(o(6__2e&U87D3rA3+=S8N^}4oAr`
zuHZ^z>*Ai&FDo#k5C|=V_oedM7?Q}A>y-3wZ2bH*d!8nvVDAgullfL}K~*)i@s19V
zo}H(=ds1nwAv;;bO{tA&_K4ttn1n<{qoEpdYl|;=z*8pb%14Kl)V`AUm)YJxXBhWK
zT2#p&KYkRvXP@WB?P5@b*x3iB;vn>&jed;Y(VkL+alCPPDOrS&gW-V);}Hm=_*Z!p
zl15PL)SDW3216dPVnRknc6K&m%J$5{!s7Fugn4jD55zl4L7RmH0`4!YJCL+>1n<!0
zK?2@_F(BqB+)m}KvOk}h7MMALnBgBFvFpXWHpA@EF)@s@DT|zfs_MG_{t^Uuc$9W!
z{5|g>RZz=Jmp~UztROn4VM<69XqLhP8Uo_xO@#A(diCKXgr{0-s)>h5XQ~N4X=7t!
zVQ2qVR5bn5mu}9~{OO{7n=i4~@H}{7K|v9Ji+;rn9{4kfEN~o&!*Uf9{s<wXYo~ug
zzGb?6dlUrG)m9cEC@eA%Oo+6fb!yuDiU|zGGkg6L>JkUzU+=@pbUQTVR!n)KT3$&J
zZa^6vqhA->GT;Bx2FG(kc0uoyD&#?>UQ(SKx<%&%Q|IdPP&Jqx$QnfjvCCsEzMJ7K
z3u9(tW=0D9>4)X-(0m@oprN6O=P96rSV49P<IBTZQyWVWvLFn|a~OXBIRkQ*dPh2G
z-W#O@H$V|`o8)p>gz&R-a5&Uvl@t8a|Mz2Lyxc4YaPnf;JM`0u<Kd|;Jt+#XGXxcn
zOO|F_vI{2zml`{m5>$#n=j*n9wV1~3LemihpVW36rr=Fx=x6s--+sY6()+o^X`?Tm
zLcIP%SCq3E+1#-H0tvQa0v70nsJjZaah%TXMN|K#GgGWZ*I*L7^!nJA4kBec@ry?d
zO3#alpyIAZ8J*R3TE8I$_YSIQjxYx_AvkLug^*<M56k)nQR@wDW_#Z&=~af+T>}4Z
zI-7~4lA^fk+p!_$e;}5ilNdSZ?LwQ{{6Pc@6xI$tjm2)7Xdu|s`H?gfUy`QdMF2K(
zaiAx~aniQdk;+2N0iIYQJ+TbR<HmK7BK(%p*d;t;Uv>bFcd$Mri6;V+f;K=Q$BC&J
zqlB->YwHoy+HOgy%-R<~^kkrIuOY--ZP))6dqUh5+4V;?%_#Zarf<lU$bzgOtiL7d
zWg&lwHL#ABaKAdR!260=>&1Hm>KSh^l;H{>hU45hq1wF+#4zC)B`6W%ETWh&VMZ}L
zxp&dX7xziVhZ8R*Hg?{*@r8hNL2-vE?0XLZD8x)bF@fSQoWMdw5}FH#BB<iY(}k6g
zHjqdb(##zKyQaWT_k8<hyMm<9pBu=Xu-3g@K?*zr^^)Cha21pkwFJx9R3teW*=}zv
zjj})%a`V4PT8*KS%4+=9H!$Dg1n&s+oEO^?fZc?VN<Aqp1?K2}(grma#`n|61ivfN
z0B)C3L^X&B)B=T$pndu1?ChLZQQ>$vnky3{wDl}GuAmhsRC-uAxqjN+@&GS~*pGgo
z>!=~f^$jma;%wEvD%1^<hT(+Tp$s`~aXDl}E^frXk|*ow?KRrkq-MwCNBp3|q2fjU
zWZW-63~{r0YG}TDT9m^xNFav{VIa)3P8y|E7b4+rTrA?_CkuvTqsG6|@Qy_EQCf@V
zBYHDkdQnse$Ikcn_ZNxP-+~J(<~q9Wu8s&IyUls|)S!&8AuU!FAr30j>}D%XdLnw#
zm&1)Oc;$~rKEPeqa}a?DfwxQowft*_1jw7L^{`Cpk#e-JMYc2@s5gt1uOVsLQ(6$`
zzrr5#`l8<%`VFZHmRM{*5U10;@3j<Tz_`dwdmbPT-r?;z{~%IXtZs`3bBm{tV!gb-
zy88E%00j(=z$#&-w!jKNBNRp28W|Z~6c@k7G)19C+&O=a1(C+9+1U#Qh{Fynwj978
z*)7YK!LJ}uOqI%&f&mJ!xhLk3)V<xM&WpgnKy*JMyQ7#H<{@r76BCn^GZNXKrHIVI
zMOmoPj=x9IRno?q#1+VfoCPf)sp>Tt%qzIR<^OS4NbzXwTpV_hMFMl%cTqEk?CMMz
zLw2jDFHyhfL1Vuw<wu}l;tzge5EERxJ8G9;|E^?(G_2sU4oJn+u9aS%u!1X~z~Mqu
zli#}+X9=GnZ2O}I>1Z6iq~S&U;URN4LoT1n5K*6&=xXgI&aj_83HTAg%J%fQwn=!i
zQJpg2vD6l7?j9c5!cMClatYq)85RL<N)x_VITYwpm;o<AzR;QoxEG2m3Jb(2M|;Yg
zGArI)q5WRTk68?F)e7c2EPTZe@!Fg$*3zP;&m2IFi;bnSv;Q`N#E~L&McFpn@`D{P
z=NVr_>47EHCKL&D`<4d{wWfHXu9LuTf<0qX<1$6uY^}RC+Zmz+z>oxjpq!t66QeTF
zR_LUvICBxS&6ZWO-2#7-`G?@*Wemnn)hvmSo$hs}TM(a$GPF>30|pVvL#&`5?{;=}
zYB#+vvpJ$TSOG#s7R3meFGVb*dQ$iWkN);UQ|QPO){GA`bei59>%*|s{-a@wEQpCn
zTWcU)5Iu7QUk_Pk!fkiEL=!>Ar184YjTE@b;P9{-qoU@*mvXK_ud9@IwfdXVxrDAB
z9!UPoS<ol*EtFFnZ460D8;pL)Om<zPmP7dYZ<GK@SQKJj9_ljqHXl(vJwN|D7;{NZ
z0hy3%Tl-f=M#hXp7UTMO5aQkC-50sBK1Tm0H;v8g==|zihReg?Tbt-qr3A}5L(Z8s
z=H`JbWnICj7Q=ka*A)mW3l$8>xY{~(s8_$Yq7w9*GI_1sYz(zAxaQMe&O+6|ZeyN{
zvrxwD7n#A?@(_?z)0f?jjt<IFi=+3zpKi!Wj#^Z-O9(yy{lbt8P5m+HjwdyCtj%w{
z%*vYnFPPW3yyt<uir+KvGk)38gi_b=ZYu}gQFLhi=a-U-umR{~|6nHcd`KzZnH@-w
zfWmO0{<m>oYGWo1JfGI5CVd{706!-jW3m{e3s8lOt3s`Q3S)q%Dv>>C&X<|<Ia<Q?
zs*Qy!pKUbaqdMXdos}XQ3}-4Qulvr?`_I|rn91qV9v{W6#n&@LGe?j%YMCOWXhl{l
z{=re5|LK|PYF^BoQiRyx+m1m4&?n6)C6S1YspcDUm^u5*vJoPbhRFS@$enC*!#U`k
z-c(?QmK#pUEraE4L;Q{-rOI?J_>O2p96X1*8jTnRGWs+iO-ZPU-%t2K5^T2AU<c1e
zF>^*JCSHBaCmHXSFk%^2l$6NkpvFn8TO4hA&)*Y4cGP6B()``E^rSfzj>}7ZVTMg%
z;-$|;7Im_{T~xXr`uDw^3v`kZl8*8TLW{}<hYPPD)AP;hbnfcrC)<aF>+Bzm$dLH_
zJ!v<xX<jD3LJG^w_RiB#xrZC+NOdqEwl#5Zc%L>a=_h#NS@p}%eh8Q1i(aGY`A644
zD($2EZAe~Tfv9y2`?~{aM2)iQ$WSE>6-4yC=C`UVkWl*bP0&+GfDqUtdh>flmG@<d
z4w-gscB=X_DenHkGfaY)(Sj{Z_SJpL2@HzKwRPF>XAvKiyOtF&ir};E%$+*y;HA|f
z$9kMP{Kz>?(DONOviw&50tm`(<)0OE&rP--v9p$VjO&OSOG5O-O=r)Sdkc2XvB?7z
z6KGGAwPAY7gFef@3<kW=JGhh(Hsov~%?j8ze~Vqp6K7r>`!l}r8+ya@v(mz0h~Msc
z^0_*yBqTw-A__4bM3?f6TZ^5He6>Vuni^1<LW&7=xLv5c$^9d(Wu0&C(irvrT0<a%
zi`z9zMR`OQj;tF8oCrh@h5gqy+WZ^8m|rG#K{cvRZQqXb&#8*ZbMfNzSRm6BUA6i#
ztl$6k<9)8;+fz*$?dyY6zI+)AFJgy?nNi?#pAiNO?UvY!Oxo;WzAvCNV{B5iABl;r
zcOafmLqnpl-?5Pg4C$<u8FV|eF7>PWlT7;`H8dp--m`);Z>H<pY6H3l7N03koJVe+
z-Bh`h&*(Q(cF^%Us<m1Isrt{ysu!z9)d!{|KGoH}f<nt8xGy6+G%mB;v3~y+ShHXF
z;~@HPqer6d;z>l&IVd%bc+7Q|UoY;mU~Eid;n+EzpbVeFLg3Yfvql=coE*FPCbs|t
z2E~qRnp{R8VzQ*mzrksybH<go56LDnF~6UxMWHlM6q8|-5cW}BirQehp{km{V-K-3
zc3tD^KYtxujICxJE6rT^-fpLTsN*P26kbHo@fs}537MF(bu-#Zb9Hs)!~5<hw|J~=
zxUV-jI|bH|Zc8=GLkv@_Lm4!(;V-&$RU6NZG3l^FiVTY1X>hu|R>F{#*JylIs9*O7
zteqF20)@n*YFb3vef^@=KZV_oYk11qT%63IDx5+3Q;03rWyhbFw;Jw5h|UTruddcY
zOzkdu)!c=?SMLmwlb5##Qe8)Ku014A3KR^fib!PxMM8)m)eB7&ORI}*46|-BgMu3=
z&U1TgW|E!_`XdjkM+>36@_&4_U~%M8-IO3|#b4xMS4S(f+%DqD18xz0Kt3XD(JeKB
zHTT)vg}b`fFOd04cQ<D=u010^#Kktdyj7wgYs+mSTz+Z}{qnD&$??rugnpR-PJ4RW
zF?(y>>cia?=Jb3XPr(2MFX*Y%LBnKo93$to!e~{*s4LG|+^-eNuA|uW2f5X|cat%;
zKGo}wW;XSgtaxt4b7WowgXUDcQoH;gy91(8QbF(>$p|dpRTLr6uyk%ph)T<!)4wj$
zt<E7EBtBV+da!2g+V?CZK}z|Z3ciJVSU_*7$@<+K8HY^Kn!>P9fk04({Z5gizZ`96
zL~at6FknV>c~s!3Zdx=Awt&R4SW!%x3s>+}GJ1MsM*s|0e@!(ba?s@+X!TtEXzbFU
z^Zc@>0i*~y#5Fr*u>42*qcDsAa5h>gbNuZaFueutS3mNpWU86)urQ+Q<9n|~ANGDP
zFw>A?tWVve^`+iQFk&7vXFp@$Y)(#`9JUMkHIbS%6V>=RYTZm(1$kpzkrI@*{_y#n
z_-Sa`*eh!(YFZ!Xsl)i5w(;>p*lrLVOQ_kbee@ivDcgPLw7j?nb75P{FMAZPDD(#p
z2G<E|{$TW`j&F@`L{hvZxXJG~j<!xZ<LB?=mzS4a<qXOBWCR-x%h63zE4zj3`7gem
zcSuyIa_N>l^c`gcRvX7#uk&8_`*WiI6Sb&zD@>WWdcNY|<DbU%EN~QWVuGUp8S8aC
zzs}nGB%;uES}R$wCIc6SRB4bH<x+0kT01)fu~y1VEnK^6IWHc$dcKN-l6F}ZaB9%)
zN~}<*y*1LII--|VEX5>k>NT6{6y;nzGM4n_vx%B(b+_!F-tiv;Jwd8*dG(l!&7vsF
z8SgzpHq!lj`)9HV&Ko3AVH^nr5<NXB!m!M!ydnV+k+e@637J-9IEh;fy@Q+FxZWtZ
zs1#~X`}BDUaOj5JB3F|FCB{mz`<{J7Yq3EibqA?Ij(&e&)ACMUTo&%0zxdkpLIJ|Q
zNWbj#_}D2H{<kMZ{0mQ2OsZhlLUX+6pW~?ce;$3NvV(05i;wy|f=T_Z))gkNl?Ks{
zN<t-s3;GFErwYnD1^1hBZd#};L{QF4cAc6cW=n0n^+V>6Q+vv=i%cmR-ri$h|1FqX
zK@rIcX}bMF$l)Hni|Fec?mhmT?1raS#p_|M;iX$<fyZ`Z^G%`HnW&f%Ganzr!}g~~
zeSiO_Z?%^&@Q&6L=cTX=83>Qf!`FbL<wtgd#GMyIK=2^(SBnh$5M=W8c}23K4?8B>
z8_~&6D{F(7lw-y1$c@JK=LyJ+Bf|w8Q=q|RgiK^+W?sa69oCz!;?4y|N)t0P$<=L%
zNG1rC=2Q>E3|@8)#XD+aBF%i!MAPky2v7S6RblT3+jrVcjUqPw_bU_k?fu3JtDnSC
zZ%$iOp*7#(VK6t4WVf*Hhl@z7l^oI^tAY)tcL$sewIGu6u<OBb?53meuXn4j`iq&X
zW~L!!c?=h^pDeZxKmuLK?&LMy15>9hkq80rFz&OK>tLQjcV@!|QT*j;jMjw!G|<dY
zR*k{XDyI4E_KYH#U*bQTI2uv+%=>|I5fMQ_a$c7odlw}=9)_G?jpHq$h8r~-2S-<k
z;q!?;WV71q*KUgkL82Z}ODDqB$GY<DMAy$-c#l2hc+t?%+`g!<cs;!2(<|XLr=p>$
z4Gs<t)hyOF+mf<7z=zD!Lu7hV96+)cc*&ei3a!H!25qnKk6L99V(5+ZgVr}8ByD@{
zmD%3+QQ;<o#s<B7XW{T7QLG?(km&1!a7ic<KHAMCfm1x5Nrc~C@p8FY`5?A~S;~a>
zdC1)%T&ex!kI75%_}03f)#k${Hb8UJOwXfOFZU1P-NC*_68*&f8*r8I&g+tpmy2)=
zH$bcOe%Hu6>ZmOIsCj5Ejzn^S!qsTl4Yx{cwSIqU)VKdnw*P5A^Ndy2Z{Cr&!M=Bw
zV6fh$Rjn9w^0V$I6PF5{%Q4M<v68QES^TS0sGHEoz&+UfhPqRFu*~wAy)%>B@#JX(
z&bfqA**L{L$druuZ?|-WL+BCqc&>d?9-a_u`$VY6Maj;9akCD0a8#_najpk(k^@%F
zaE@&)6`<?M*O;rNKnoEcu-aT4^`MX%aMK*q;>D%VtZrbjAr3*mCZW*E_t8R-51qF?
zp461>l*4jMW6BoD9ykn$CJf1B_(q>@bp~-AeV)j@l&wwl=ms3UtA29mY9qx#|D3wq
zC2}?HLFMD#fJH~saV$sR@bECxrgP))AP&ej6aRPXXQz=`tS&=E+!TlJ$91_)4mU%l
zYfs`sKJ2(%x}~*+^78U_do`$zV7tMMAz1_z>f4iI25SvW_2_I4Zk|kb;CDAxmPYU;
zgb4aF#4fvuMI=1yIc|%<^xF!aJ8{E00-B@6T07l2;Td)40(@HzBntl!02i6Krw_)N
z6$M=zeg3lGG1KLtW6C<WH$!Ey-OTK_J1RDJQrr2Tc%q=_Ln(T17Y_fB?&1L2Ia9JB
zFf`DPkb=5^)bzHJS9Izho=RW(oyYz;G_lY7i6Zkd`+7%xIgQ4|k-M4*yi{tPE@slX
zOO>NleCPddkpQ=dk|_@#Y50vv(u8>-&#qVhG=-1N5p+&8rUP60J!mkMjc*`LU4F0n
zH{YZCcbgx-@66duaEPA{!OMFi5!Vl50#>M+-k4JF$oxD0>3~E*xcQe342s6wxt{4x
zGYg5^c%=E|dvnXS*GQBWVbQi{Ij!DT&u)C$N)hRxPpYz;p-{#DhEe?n8MUu)em}^X
zUgdchg>t7zUXuhI4I3K}KsEkA|9lg8rTncz4^o7+8!*WG#hC647lF-3u%(72GHkF`
zt@2lCiMDhu`?0jl{|v(-Hq7k9+K7(+Oev|^xDaPEv~BVe#SR1`M;;mpB{BRE4}p38
z!0|Y6+qwk#81?)hmufhfR5`{6J#lz{q#5-h`q*AuXLclY?L024m=I;M%k$hXs=9tL
zY1V+l=;_1`b(!x*E4!)n^g7t2-`Jpg6zD!MzIqJ18Q9f>K6oju$~Y=VB&5Dgoj9tP
z^A^Ba?$mV-e=SrHjl}X&l9~V&fK<>#@HX`zwFs7gcR{;fiR{DshW76u9MtrlPw{Bk
z{R9%X9X^-yRN<wypQ5odU4$)Eu0707Pkql`V8;C190-skSxwHlC4ipefw<6apXYET
zj<PY4w~XI(mcOQY$~udLgpN<D-y3>}A&q+@Tc0l$cQktzKMLA^MZ5BmCwg4`Lg?Rn
zb}^TDyn;OI&6q*4lS<+97=})=a4kk8$wvApuby#g%I2<$t}iYHuX4=UC8xjF$c6}t
zDT`n-vVzH*NJ!g4HaOi5ec$_ch8zK<#%Uml!|J;0st#=y-}LMK4b@FlL;`;9J*Ls&
z^2^+1wuYZ5Alt9=xD)5>i!B;XC<X5|ntoyrQhL$`)H0~NV<)qx2xGK9B#lS84qnhF
zHbMdIMnrtDE~IE_cJN~%c=`5DjZKE0pXia7(U(Z+3G3C<2O`h9)9c)cId85o>7DOS
zH^j~StMyc?UC+KG<|e(F#_2F+o!B1&Ju-RQ$H>&`4?#6ZIvv^bL?lgkZ}-da>#ZCe
zFu_>Wei$9u@&A+=?gu+teC(a}R}C+hHzn!1xy(>IkQ21qkG(&zMIVeIHz<LA$h@je
z0eDCH3i(1@ENPU4UYX8EAD_4015M_&IK;Lm-_D801VExFxV6m+c)}Am%R&tmP@it4
z0A!BV2({;luu^R6A7}j!pPy|}CzvAB4V`rN-is2RZ0*ldh5sj;zbWOv@lifK;DFR5
z%-;nm4hdpiwJ~EK#gv=UM(Oo!%&44>pUD}Ep?b)T2t1&IS~%+Jm+9B4m<>=f7rM+y
zdEdM-&tv%;p^%z-H<f%$F8gzB34OJd@wjXqxlFOZQ4x3GalE<Ha7PgE2+n!(Ra5pC
zxS!oT(IiwjiTZNc5s1d6XXk%KhJQ$<*cohOd)T+!O!Du%%012)4{)?tlLx}!($<^^
zg08d?%oAQw8uv}Jq^lN$PgdW(>BuXB){ItH#vEop#Ix0U(EW4-rnqOJ%V|61mUs1)
z5&X@z|83ySs_x^>$+~|=Yu)nuq0SJ%>G+Xu_sXX_5l)r&)2UK+_j1;OMlr}2Wy0#|
zMMhDEweKyP)am_-=G#}$n7?0TU9mP-zM@w&f6Ht}Rk!Y)=AJx`&R<Knhfb0(9v}2`
zrEJ_5Ff^=PC_05VqKbLZFo;WCFEFd#ce<bqF(>@IH2yR@6#Lhkp}}g*G@ZqjV=n!2
zf6gL-AtH!wn<(A^M6<7MF|-wk9dK)4?!8VE;k%IkmQ5ACBH)--FX741^L{jHZXh{v
zZf@>pbF=rO-B<rFzk4Je)>8Z|34Q!8+t?N9R{CPg++t}Z)SBH7sy<>E1^Q$~6c;zO
zSDk*ZXDB$>a71~ms+U=o3dc82+)Zj5d3!(k`vG%BtM+;fB|n=nzfNC;JLR_j?TKSu
z+oN0f=P#{W_^9t&|F22Yhs>8CWH!8Y_!e(eH4;2&KpjM5qW_#m%FG>%Z)25vg*`oc
zR%gVMUDVqw7j?K&8iJh;*PL|z_|Uyx^e#r06^;9k#FlBKLHHtp0b%}%sMEBTW<dQ#
z!gLK}>w(h)iC-zIQ&Yqd7}?@954_O8MMWExc{l=JO2c94M34mw<0z`@F1mt9pKO^!
z)jxs8r28F3*tEXEyQ$%t3A;TWlt-o`j?YIC!3)Og`_yEuT~%L4UXML3^VaWNecbOY
z?-AA#5f*-3t8edjwl&0T9)o=3uF1Y<=5OjEJm_CugjzN|#jCeGfAQB?Z{qywiWjVR
zeR+SQox%+sOJovv*8SvQ*IU9th%x5Z)Kge7vblBs#Ir?kN#9AmyJd<$8Q*I2I%TMQ
z!Fzxf_1FYZjl?Ay|2uyG?9)&N#ike@nSE-W+AWF<-ESnJK?F=|z7MtnRuyIF;0*2U
z5x>CA_P3<dO`w*r@6^mByl)I_H1i4m-1xSxE>EAhgX7z`z4up>tJm!a{}X;dJtLIW
z;PRQ0)B4tt&F^fip^38DF8aEq($lA7Q~LAdZ{L4{E;#MJf3B1pV=A}zlVmUPL67f<
zmS;Xk*dL@GFNnpxJs;Y?w$0ZsOQ@YD(uhqJh1J3t)y{Qbt%BI@#t0Vhm`V6Q)0K71
z>>=ak@#wfXWnj2s@?{MJxFUXCcRqO<T8?1x<kouZ3FM5KKZJ^k>dk*9aYDF%fk8H7
z27M+c3OOlLsnFtCU)RiHb;GY;lKGku?IE}EufR`*1T)S*ave{<XejJtTV&!SGBTg-
zr2-Sv`eiRWz`7!p9n0ad76s$+xKvZ2A#a{`3IF<=i1{p7kupN61N;&Ut4f|K!)fmz
zBa@Tap<jPWNVkb@=Jjtaan{>VSU0$R68@SiUu^-zi;|KDXdxPX#8*_I!O)supR=0h
z_xN`$5Z6TKoi3cQ1<pbi@E76nwEt%TR!Ov<y4jx#UP#<Vm^|JU7gGcj*#a|>mGQ+L
zK~EF0COgIzQOMgY8iS!15`G!*zDR|ii0;WyM*~@G&_3DW68S>i(lws;u^_|gR$K4O
zM}?Jj0uwilguM0jbq64Mzoe!%xE{0-=Y5J{-uxxWR3#`tH}w=xV?Z!z%db^3C2gS+
z?=!nlKrI52#BAV*3m(USuE74S{tnh=C&L2`432#sZ+rsx`0vu$W2;$wS?Nd1Zu%6l
zR1!b6#j}>1n>(t%kc&v~7o4O*9o@@d`|5#_<q>~du77NACD6TJt1N|%&HE_BeyN2P
z=X*cpnzDQ0)j8;9cfcO-@jCbq)n>^%n3oIDqH77SV771lCVj`s#mmMUkE(UMm~+WC
z7ThjQ@EkC0*ZMq;HIq5BypUP!WAma_n8C0BDxDgB{EF$h_|n7K)2)h#f(D5&SdY6$
z<yEa9FH-jB4x_ia*kZNT(B(P*fc!D@_uiNHPXXaox;h-q9!CX2v8sgV^Z^+>On^~r
z?GX0)x%6+*_2ZbNkBZb>p^{ARZ~S*WJ+J(m-t{A4auc*Io5%FRp2(4Vo`5*^wr*mX
z=}hioof1RNWISH^4T6u+APNZWc(rr^R*j%zqOpq^YCuFE;%b>3C1k|e*{5*L|4U4l
zU6mkLOdEB_Q!0%SS1FU3^nbRMbEDx8RGX69#RTySc}1?j8;(}*8}=S9J}{U!y^Z>|
z&6{-c@Lm4paMJnCf7bG0BQnbqln8Z2ZL1T5K#1Z^<JNPVGwj(!pM1E-ddq((UMWe1
zO4DO?{o2UbQ*XY;N7-&n!CES-J2Hy!)lx`CW>BL%YHIsAO;})Nj~I^21^s|$R840E
z9Q_{quy1|z`8QrYqV=cc%ro<I3%j9KT3N8Xh9#VZFdif+^Nml8$<b}ie(V(A#(P@C
z(_`}_n9G{WQ2py}vXpBqx5jF9`b)8I53yyQR1p;$XG)PEX~-LhG>990_fvIdr?$Fc
zxhI8<tt~4H$@4}A#lKK(R6+#eAx|}H;b5C=Lhdz>lyNeRfjYZzrAas-KR=u~F<;*Y
zm4R-@xlTRP3zxp7v>iou)yvg5Yd{D|$K(q{mNJ#$M4%3UvoI>tF`&y+1bbq141kO=
zoSK4Fd>OE>#{RzHeyX|Sr8U`8tccmz3UVnQ>-K;H4%9am(%d<6l#Z8`!7IiDa-9RC
z-xSd!P-%}im>tDef_MZ3w*CnjgSfF*t-zkC@K`jVC9aOPKl?*1mU>0I6_>tS^Sy%=
z=$~IC!YQl;^t|zfCp;66W6CVZr32b)_+3`Mm)A7DU|Cyy7u&~IQa+&7b$53s9BY9_
z7oHTE(c@trRku*8_H?fA(JB7rm@Z0zp$)yw05-5YV!e*t5UQ(a+}~!ch{ZNPSn945
z+RNZ}5js83Y5nruR2gJk9ZIuv$Y)@T1!~}^OkY_Gx*N0-($fxjMc3ATR2km(wKYj9
z+iSP@%SxNC7UWckA`F*~DKFJy=HQCM1+xJwQ0T|GmYw}o29TPa0T^PKVnP;b&=<aU
z`}y*HAfxQIQRUL5T%{xT!9h7E9Ar{kD&ZuKG9To3GhkA_<a@PbV_Q2vpg~dzp9HD2
zv-P*y$r`cAT7&<#|HZGsJaoM*5b)@Kw_1IX`g30Jax{|4#)yMNsCL_5<s>8cM&aoP
z!K3}&u}I+aP*iFQY1k02%Ncuh5ASkv4V%jLNcAgSO0Sb{@~+3nM^qnt!*!N#gvG?B
zK^uc)-Lhx$B(93GV7t&MI+YKmrX55YsqouXF~;<WA{@yO6*B)|>Jaah!+k3-BjJ6-
z)P~G~d*EAuM_K@SG-(o9;%R+M(=EBkf4~1MS3+pNkFGDpiQRk5t?@s@!@~<bro2wq
zRPSd3#h~3J)m-fv{@RQ&UmlH3KR0JNw&UI02={m)QuDR74n1w&yIiw41N|%E*yKS(
zt3~1|BNB`Pg5zI2cIN(=TdS2OL`NVXkXZeMP#o#s?)mm*S6Kw4o+xT<9!~qC2tr1X
zq-B<FJP(MAZGFUhpr5?l&2z8zXMOz)w~N9@KkS?P=q;5w8N-i-Qq6=}e`AeQ{~Kjj
zzGo-Y2!%fbop&O8_>@-*BdmG345-GUbm+>g%00}aIeHGm2-ZIJ$g;k9Hs$K$&Eb={
z`8|pV_TcX=vJ?jx`2)i56#<tjVQg&dh8s5%G3-lydc@xnG-7c4JZPKvzexOtCa^?4
zN6gDWg9lsmU~I<@%Z#%9c(L9qX|vv3;f2t^k@Ihf{!`5@p?PLUjR5@>VkYDf$1Cxz
z^`jnKd_t^#r*i*TKUb$jks^#UGLK^U$99Ix-@&Hd)Li$G63s4&`K#IA#rxM6Dn~Zr
zkGKJ}1F;qIh$+@WtM!=s1lfoRx!*Lg&BK|SBaB}+(r9MhmFcKPC3}FPW?ti5U$=g(
zm~W=|(~@yvJv0{9Sn*n$Gg-z$1v3Dcyvf4u;6<sr3BFQs&TIdrQp={F1y*H6yIkk4
z4YY;@EFW_K0IB~!EWz5^diY!axuxTKvY$^iM-#N^!}t}9<BF`Pqy2+uz)L|n-#-)I
zP(gYp91pQwL}No?JCaAQ2p*!87h44m9`4UZvaz9DjePIX^ZrQqe5VRDA_llGJ-na}
zVWcfW-(|Raa9<mS`P+ByE?FAZHhPsB6{aKD5G(9=FEq91_UhxD(D<L3o@bVXi)IT#
z{kj3<%`e4VSw;Owf#H!YK%W7|-^XZ5y1cUEJ(%7E^txv#zGhBEs}k|&a&s&1S$o*%
z<HH>qNET_x#p_6)=LfQa;Z^=QZ2bAck0n;_xxGM@FfFPEhF{gySh$3HM3gof4K1w)
zaQ)a(#$GP$zx@wA&(}t*!)GY?xVcRMv;$j`me(=R(a?RkYbi&!=-Nn;o2ZN-4lMv{
zm;F3`S)SIgI6QBzzLUko87oY|Klf@3^GQUsEouAcNEfD>()i3Y4NkV&WjfDRz+tAz
z{19VX+XRtn`dw>vb-46rFh<ThCbnYe&&-hhRx<awd84UrNKL*Ja6&k(mvM8(I`jTz
zh$2JwIc<4+zn&)`<7zwH*i?XXfd&V<81t=CyVBw!YIKn9`mw5&eZ|Gm4WU%#q_#J$
z=XS(?EoqL<b;5iE+p$|@Q1)<a1;3n}94;?1wNV<H4s#2DH~kS@7LgAj3kf2a?uuW0
zo)Do`fZH$X{nGndu!>wKplVHdzc5Grf*~`zFSeW?pjH?a6Mh<3jFjw>@Q$Dqg+p0_
zm4I9Muf$=E?LXI@Rc#ntPHlL_r$&Mw&EpCA3)Tooaw}G!CuZ`(Q?qzUNm=RZ>+5Nu
zA`7kidZf^b=O+VG!O+$E2Kp)<UI!HnsIbPICY~wb&Xvv`-PRi~L>Oq-${Q#JS95*Q
z6<eJ6kvzTJ8YubVeMxJa__e{D9b5i?;~79TkHnf#r0V7p)bpL|my9mAdIz&aDoGKZ
zqs%U1t<{JxmidI^9I__r=>?_qv1hb2-;D)_S10QzdAxpnRpdi2>PHW?e4}x-cK&mm
zU%k~>9sVH;J`C@p1mQ5+OAze!!>9BOcJ}rL0Q^E={hGS;rI)w2YMBmBrVB>Kl5193
zpV>qb_gV>?N`rn`n8%8*N2O-*IWWVE<`LC0t5WohaGEGlhawR3xhgatDsXB`E70^u
z)|yC|UcGu{4KzodcJ|}}s)LL2P#T4S@~UR!!VAYny@SK)%1sEN)iJ@j*y-_+RfQ2J
z#O5+;)zaNZ5o)b?k_8A|!#(8MyQZRzh$09CZE&%7Fn>C{`2yogsM#)kwyexzK$f20
ztfd`!66~Ma>Mook6aI(dQ3GTTi^8((M0oDq+~$``ozGTU046xjaO1?F4^fA!-Y{t{
z@|Dh<RrFli2=WfViEe<IL1TP;JYHax7u{tDR68JhC7l|qlNXC_6lGGym*)cuQLNSH
z$HadDuX9)e=Hd^>x#`p>`>4P4V4}%Hj8JX=VBW5SZyixNoH&PPb0-&nE?87P904~J
zPO@fXj)u_JSXsWxsOD?Z#&L}Jp)9gmLi#M`^6lwpg>^3%GmW~>J-YS^OGoG(X_1fA
zY}=#K{zipNks)VzsB~`h<T7Mj5$b6Kv8<m;9d1rEl?86izxOJ&NAbbQB*u2h<^ETr
zI1oP6)lPGnV{fMT*6+QvE!HrMnV^%C&t_Ztgy>ozHBbtjlqlmoC;)}}&}wVRjN%Zt
z%ZVE+D9ih>vCwC=BXALrVXqjR#zIYl?SNquqeZmU*4GCf{3roPI(mM<&cBy@zfZE{
zC6<lXtCYh_rfeI45qnT=9TEEPEjlds$UX<y5_Y!j=yr*fzH9`dhkoonhNpfiHj5&A
z;rxhT$g+2am|XC7Dqn*$P0Iw*6fy3Qjqp+%@YKn>>K<DHmzAQ*f5$%;za0fymv*9a
zvB%?TujAEaEv>C3P^%VcGyErd9x>r<8pT8AviwUw+cH57<y^q(CPtT1gR$+qM0mlA
zSRlyRh3Bb`9&4>^#T_5N^fb|WA753^l%xB`iCtTXeK}5wpghZb=S8hu57v+h)J5@q
zX6aXQRT(Fd=({8$1qh@>mPl)Dw@&@L-4_#;`>MzvkcQ2@(rW0WS=qs}W7`+K#G%q=
z&pY<f2;ROEb>2|G?T7)u-BC5cXuNW=?@FP27fsk*+$8a+H?hM7CxecG_H5CwLfRV>
z&NhJvzvE2q;>N{6%wx9)Qlb9}^@+=mM3Ajc;F5VNY4cle&BdoJr50YETD{+vu<V_t
zp2_5XO%^pxp+yuAz+L3=YSs?t?0Wm-aW`SA^TyAfZr*jvB<E6D=>7aIo~`Piw7Rz6
zp&ttHltto93-WEd2x8=p-f1jDgIj+8Q0cveSe&um$or%k)l0`h3F;bd0fD3!LQbai
z*(2Gg$7K`We7d~PxrXt0R&E779}D0l{j~eD<mkV#LCnsLi?6|{R4-p<RZmajzYuzt
zotdc$psQ9gKMiSB)YLTP<-<d{z_B3oVd1TFx?aDrIx=U7Az0l4c<=m|LYp)=qZA0h
z_U0z^B3~a{qib{(5pwWxs4gTj)w9mPa!qGV&}PM7kz@pB+_v*}uk&`TiHHe6cbFpc
z3=<dtjK%LlwJvE<Ojq(o)TqJ6>Nm^7W|lndZN1&j<TMOAVbD2fRE%}NwDM3n_qy`8
zj*W;1&Pt1P@BPTd+uKq9L3{@%)v>}XFV?oZ!)6=!cEg+Y_rY1QlH@}ivw7A<yv0{p
z0wf(*O*;&E#sDG9Ed^!%!6H(I(2^n?1jd@7fZH<vaJ51!udyKj{mUN;NLm+tR+Z(w
z)@|T&?RigN7f00=Yym(KBYz%hYHCVH6rTszj^=p)3Gki;z|T>)6_5zIW>1|d>kB;a
z6v`mFQlU6^_8t5L*;3o3ULK65&-M3$L8a(DyFU0el|>V*RvjjqB0QeNe(>z5Jk-0(
z{a+kXN;^AC{)E}9JO{@q7)Kfe7{dr#f);Preso~UQ_p<Ksm%64u;MiI5MNKQylCnV
z&@(M%iFuPL<&XbFw7u=R)n+)K-uKnR4O~e?cf2Mof1(VayC@|~ZOlVF4Eo5y)QQJ_
z@@xUVYczZi7axC^lbwAsj|D+1<U}|8MKb^|*BQ2baKJy@2q4D65`i-xl-X|8dcV61
zWJR1>VTbCJz$OS8dQyTA?r&TO?!hc*(8&^n(V(HWKex$*aza~Z5rW%A6;`OYvl0I9
zc!)+tu=X+Q?8j3rax8r27NyMob3e}AW`TZL6&(68c}Dz{V>|ggY4I>9S1KiUtHT+b
zIvpU+cfW(VRh>Ld7ZV;9a5+zJf(K}ThUR9`@oJt0PLL;9xYxir8^`0q6a&bYy+mv)
z-eu}9Ta_%k7i`~KUPLfw!3kanaa%+Si~krydIdy1CkD;txgQ_CclYISH6aJ1(FG7-
zkfFH#_OAzcZ4{xQ!9;|Y2v$%rH?EDWWqDy?p-!>(E^t0iHa0ASs)b*7>|WpNy!tZ;
zSRUH^1lieRgIqhrtaTHu26R*-{Sww;)gvIMm!d#;w2HMG{*SK@qXWl3=e;V~1@T@*
zC}2E_ekYEXBi}7ZhQhuB{Vondq<fWkYId<+Z<{8{Z_V?a!Vn>CUu=bLIk}lM;NjN2
zXWxR=Ca_MtNtzXtm)>d%(RKg!HkJ19={$JEs&|GWgwIfuo$O$LpIugCBg^mhqsR5}
znq8~p1fV3)tT~Hp1wF64Q++Z#<6A{T@>+-YKYAWi5e8Bu#C^K;2%lF&80Np5|27Ih
zWjG*?OLw7>KRz_{|E(+NvVJeFxFYelX<)*{eqrUPAL;#PnUr}Wi|-^}d-BcwViMXr
zZQ6f38dJiMG}sid=wQAzf$(B_I7$U90$t8za)6|lIPO6E42Y8PkqYo(<FZ5-3@APM
zUaby!EYE}J`bunzIBc@<SZhgp=#K@T_2LibZr(}o@mX5i*kFSqyYrY7Apih90i5ib
zm5mJsps)Mi9f6SlkUabgS#sYEg6IRrJWdz~v4+~jFV$O`0WU>!kPE_V{tfNZafast
zX#b9?{V0!x1_6{~!K$vk@A=<0VTJl-X$K4S<RDxM3kRN_#b>7PV4@&)R<NJ4xntu*
zXmZOtl-u`!T~Og>r37qzO#CSef^P)Vao6gXSx2X4r>7fy`4_>w+aO1552Qa)k3;rK
z$F+MZNYnTGMj<O(VP-VYGPS(NB71?V?lw`;;~tR69&h%Kse*y*0fe>)ILN>*0fVCS
zaHCh5PP(Z3{%q&<^_OTn%Ok7N`{hUF!^duglUH-p!{L$_6)b+IT%8-CJt-y<+H2$l
zy@Nd9ZiGG>w>tC;1XMF6m^zx!4}=z!j|1W|+=yXk&k?(R8>Z0M5*-~q`=0{Pf`zAg
za)Fo7bJSrjfRC$qK0hPkGIpv@djA{qY0G6Lvn~Z*TARzmgd3Ej(a3z^wN#8KLWG#j
z{-PE|qW;&{_v<edw?)`+@m>{OZIJcXb71ii*M0P<hqKV<Q3X(fh^-0>B{T<o{Zoq$
z$jF}GD*@s2d-`6=u5WKh{xT6jmZs-3yy(Y&EE^DgEFdcU`J!!l&LEA;kTa6Y-J>DS
zDtv}2M)ohogcP*3Cnbw7NfgI4?>)QP{z|+57F7f>6$QmF#RMn-4-;v`!M~;6yl_-(
z|F9!x<ItdeQ;P7Pomdmkq%pC0$1he9oJ!rFEe5qLt(};jk1%R>SJf_=nuF1)xdPw`
z;OF?-+VvKC|9K>={~SOK;)Wki9bbr&?~l@k)^cT=W~#O;qX(qU{ltkbvi(h=r86aD
zqz&vf#KNs*Et}R9E2pzGrv6(IiuiB{=&Rz^#|r?h|3+7lvDECL^`e&vE67?Ff7t8Q
z=&Siu-yuz*?>0w3P0B+g#qiMVwo+pt4M8C2MrkcJZ!aPZL<pw8mU94fbn5*WxzXSv
zb`EZi<gyHr0+s$V$$6qWlvOy(Gj3)Gpz`+PJK7lQp5D+=Io#iIhQaSWEB0>V@ieMZ
z_5g!4!T1cS&C{a|BFS(;DYUY{YvZV^z~(yX5I7p2@%M^g#;71m6GZ~Blp1mobU%le
zKF16Oih%}f)4Y-e$aHT2xv;vcR!$O@8yYvGGt10M_SN0zQ|%Qz%7J%sxzm;NO`N2k
zPiQ!|@e!Y_B8jZgzmYwPXTPw;t^MVou_DM1$l=U4Z{9#=q$rNeuY(B#ZPb0v6Weh|
zCHj}0=%2wc+bWX0uO_7K2g&EeBzmdP0<58taC8vl8cib6HNKd5EC?}7X#s>e4L)47
zyTcNx7S;Ey)}6^9=&gXjG}d>cA>;=bu`HOkn!g7MAp&c>azq1-(oc>JhgmKk3-%=+
zJD8jBObiXmW|Rq&gFhp7K@5mpIM#-DLC7w(2U(xD9F<~&jurv-n;_^&m_3=cTNs=P
z?RP%`!x*(7^LeYaFa~8LvQy}q^EUtDGkUl4hbNg9X~0>;@8Y)H(ri-rjspGmdFSdm
zsiV_nkFrq#oM-x>uggD~#zcmli1FCmQk2(`d__b!#KXYAy1e`I+yBT_kl|FQt)*p1
zFdY`G7TR_NeU<znzO`OdYmUi%VNREm0OP^ui0g$2Ktjl{Gb*_gj=2Ma05ByajR`^=
zwkF+boDCc;N8j%o>7amNp(`#C+;-~K-}~J-VD33UEFo*v96Vz;ZrZUDDl+Lmx_?t`
zw?S_wMBIk5Fyy^$F+J3t5(%IJ5AmFpURV1tMQm%^VLMm8&+uLi#I&85n7D{JU9Y3J
z<KV4!8{V9)|0#%2LMWkhc-kJ(^^z0PV*W~Op*v#E`$WrNW*w@2PY$#Hq^)vG+))v%
z1LM%Fz)L#*%I%Ve;0LE#nVh)&!;3g*DsiBN160fMJOdZ=SI489{BWi!U4e$n&Ja+_
zWSNOh_7`pGwDWndbze@O82gE`bXn*ybhvy|b8K1`fz~E=if_G44*^c78ztm<L=h*1
z7;#(1pjhzVyVu9(-gate0uVj8cU8mqrx8{dr=rQScR##|ybbnNl0*M}uy<DtcN%6^
zatBeEZoRobBYtxh<Z>9ovqb@z7B&tr_6*UN60WzCXxX<su!rFN+W)BAvcL6<HKy+E
zeTUTk)C?y6@tfLvVmnt3@e1@SQIOt2{I;Plc;fWVVX1{$@o&KBzh7Gn0-_n90(qzB
zLv3sId*xD)=6q_T_-nGzAKEFP9D@eSQDeJ+_JVm-@DYgnTRE-p1!}PNh`Xt(BYA}|
z2uy$Z(o>FA4vw|ktx-*j?bFAE@{0~~2IZ555*uVP9U+k7vnYOd&bo{m58lr=#L}lP
zdp^ptpS^0m2Wqxm8Apf5{XRq93(b1J^{fO)z1Z48Ogk63^`m7kF|9;AjmazazdVHT
z9Ehg>b+477r^gO^8Ml@dEwutHWKqeB^ADPzhcN)uni*(5hEcf7E|>R<-eOg6?w2HC
zN=XO^xfhjno$1i?0meO5Zy6c|-%fwQ#>n?w?Wv4g)^K3=4A?$`qYn;F+18d}KY8We
z@e(}dqOuXCq7biOHSfuMKBSQ_9%)A&a5m-3H-2qZf9UMnby+|KM@)uOil#JD8y`Eq
z%A<yF>wrBu0v1wkdg!p~++AJISDd==GP}CE8qZDt+KIOt$NitbF9$`nP%(z3-!OuH
zVJ}V>ukmac@Cq-;w)H9q-j6W~Uq33Kw2(6>WQP*^;YWj@gNqc^3axBu&nY3!Aal^!
zvDFqce47;Fm0NV*c{SB3Dy}^sF^}Tc{FKa&_MPkfXGi_@<wa5#RWYWI+w`Ves6Hcs
znj)!h?m3@<oa(EaQ45s4`$$LJJQ&(jXpCR|KQ*4mvK!=xULymmkCMo*SN6G|F7n?c
zYyINq_lP*$2CBY?t1CJ}zh>G!(vi_mP4QYBPsd5K<{q{=7VhsBQ+H7a(2CWLD{Vd$
zVMfD^zzqwA*Ab5~sj3SwLeSb5WJKnym#y8#&iaCB)33LgMX~NJ=IjRYe@=e(Hot7a
zZ$Y9jmW)t~oz_H{q~>;qV-~FR_8g{mUzLBjRG`npKh5nkhAml^i4`MyQ8Io_XYC#y
zW{Cs4J&R*M<oBqT6J6*b)>|OS*l}LxKg{`g@}tw0#Qdv_;GKZ0DkMK%_Xp)HeYP@}
zLCsM;F2x_`9+9c)QGHM~&(npy-<dM8{@kYX<7$t}WZ;Jj;~6&zvC4QYXP)w6ZomM}
zV4Z>7%$lMU2iY%Lgbs3;JAP{Wkot8T&5m!awoxHJdx`9{{N5f~lL{B1%429XSv^zS
zg}fmE#B)h1h0<|WJfU%bd#5^+)LzIKJ;{NqwfFkbfb7)$B-N0U;|r%$g05fxGVDSe
z>rg1(`7ghd`6dbbo=O9rg1H5zS|Bq3JG!E2Buo+!w=Iz_bnKX|;T*^Pr8`R;HJ8?1
ztG%nhJEL1MfmnXF^*%a2*@RKgwaV`HfuSbEdwfKMAg6E8X!q{q{P$6Ldp@E<-;TM?
zaYbBAM5M_HFJZ94003hENbeog89FJ1J#Rujk;P5#gSOo4u<_{wlefV_k!_d38%+QC
z3$`1mdfQB>2OgHcKj>u(32h1%k@<`L1J@@un%m+MVPv^}c|Lb8ATcvsp8UNXUm7D}
zOLzV=^LM^0ydQ<+ksp3pdd3pJA$&uT@U=)Wg(`=hBt9h9Gr;)!GpIQ=Xa1~C=1;?d
zSv4bl?!`N{(QsoY<=3r}Tw2u9GL)QWBn1qse_}r=SAN1nW?UKVUG?6*)bkRfp;1m^
z%pN)0j&Hr1j(<3Gxit1ZsBIhI?XgU0l!A%<sp6fEFxvX)Cav6?bM5ZlEBZw6>-BQ=
zsPHmV?F+MK!!?O6*Lts}xT#uyBILE*{!s85vy(l%exB^2CXik`n1L0P6WF;bKoF1P
z;oOF111p;fCcUk=IY@6Obu>YLXJV2)WL^Dw<;n%C!chlkSirTrV*w}q0TLo!Tz(WC
zaJb5KTf3>CJnHMZ^@?XlZAZZ9U<TKalvaxkU0O;tmoTMs6NQszY;|;9?&@vlZ{oOA
z(=WTT+c%o$SysSu_gj3sT)s#JUNEdJ3v>(^4Ah55N3;u#E;f|$%zK6fnk{jiGLe`a
z&$J@1_}LIme<>M#NzB{_(Rs<i<&25TGaq8DcNdOPqRB#fy?ng&bR`9;|J7aRAixh8
z9TmgG>?-H~&jL6uoM(Q~%_Qcn2vZcvtb6fHTF1GU2At|1<055hdzpk~frPvj({gmI
zNK|3c>bHEu49O(PN^1sFF(xwGWSW1YuuwvHrG$i;n5ZpI)3HQbiwGQ|4#QC$938p5
zvuu>}H9PeeiV*$RJ_e8`Nl&+Zd7FU4=z4b4hW7Rq#lnf#AdU!Vo-!<%SVJ(B`Mlrs
z<n<W`mY792%~tMAoVSPP*RJ5UHVe^TIY)H{9-nT@@p0c?t-?L=;@ilDVpP}B`4T=h
zy&rY0oW+s%rCI2zLXvMt+^`Yv6KrpjQbN!$<kvL~t<~TPX{ppfbRQ(&>#RqpFuAP8
ziXIBxkHz`bgt#mSrLr%F|3dtVdVkW!hbW;FXMN;CiH3&j=e>J7LJx_ZmPph*P3qi8
z4@qE@l%rb@k5R!@=3Wc`J(Oq}(Q@X~ywURJ!5GDzlDHv>E&cg(pI@I691j|v?0=lJ
zi7?wN<0b<Ch9L4E9;;@DA`~L}oktmM;u}VG)!*%tcYTjO#Wz|fml%VVQv<QFv3vn!
z|I^u3|3&ph+o3~ZNJ+t=OFE?)Is{?pl<qER=@vm6Bn5_28l+QU=u&EEq*FTH{l51f
zJbnT`!yV?HbN60*?X}l|0GVoQFtZfIEX<be;V^MRvR}E!N<`Z}>RU4&W`>5D9lAsJ
z+ijfse6i_%ISt9r0&pTza9AacjYgajS}5Z{!?UsTiqUC!rQ4u?iS=SSHyYaMW#n)P
zt^*|HCq3zlDhehOc($SllHhCqU+B3Bs#||IP2gx|Yh&2ofl!%@l$>ax72klPNA)i`
zX946iDMF(|1Q&#Xm;|m`Jthdx`t(wnipF<iM$bG=E2>Ia62xyDwS(7(N&<rB=-(ef
zN#AP<AokpK8rS8OAu7>kbyhgHcXx_o-s7=5{%TNNIl|U(P4-Zf0%g%PSOsj0<N*rf
zP*~55n(xB_xkySMT@8;lnmCNv_J62?ghdoN*e7_Y;e65+P<vdN=1~AiLthB0x!V6}
zD*ftIC#EeF$*(SyV-2Yr%w&{z2@<0sL;<z1@{D|a3}XIxQX@}lB>*cf(dw-`8i}>T
z=jYa&)lC#<%6y?@=)RjwG&#W~e-+2%V9I|}WS%Vjgq(;9Fnw#E!||$hlX&B<Qx3{z
zqv6SVEyP8=bKhOz-wE5QeV*c5MLd(sGbj+N^dl-u*Ac|2J*!J3dK6gcK-o>fhTj3-
zY%hpBh-WeT-1-vxg8xp~rNg7n86_3j>dn~Yd~THW{Vfd11(~ifv7yfRWm0H!bAI11
z=$qb=<NB&iuZnyOE%im(u(AY70a0Q9dtv@s-PtV}aKqQ>T{Ylg9I$r{c}8A0=K)4n
zfQt9O!(FUI{3Hr{is5?rkacz4`Q!<?EsVJjjS0SQ|KuF3{wu(wdEr?Sugcd|qE-Vp
zPYqGMYF;lssFl95@Mie-GY{|L&4$cWZ3VKAJN^fEotD+=9+{svbcEPp7!^|<)uTK_
zO42ofJS&~jg`J&}i!(E6^hL~$&XqA<yq?0noVmtebT(XuMV95lH98nb4MY+AtEIyW
z<2v2GmWD_giEjQ%O3A9)+A@I}T_!GzMH`*t$@IBXM%*skzLi|GN@KnFbQ@#X>i%P6
z(;tdyT=R;kQ^rG39d@lmGFz(#6on`;b!x(yDZnJ>>^!;C37fwP`wHdSr`@%1s?PcR
z7}UNI+Zh%>Vjs+f?Tb?0;U!PjVs4A?H=A8&;tkz8;Z{HSedT=eTjnPPXllaq6Eg%;
z8o`MGbTlKSpE>kypFtyiHI&MxBTlxPJhsFsZPP3H7^3NuLvDn-M5ktFVQ1p5MLO*G
z)Rx4;Z$u<UA|I)T2wFX3u~qti?=miRIR#(6-kuSW)XU`VLmzlIS@O0bj?qp2lb=VS
z1fhl`zL<hz0hRuZnc>QRfozK5>)>;%MBYPCI+bNQrp&~!g^)C)Q#mqZCNzc_y18pK
z9IL9=w<PdTn%ujZjn|z8o+A`UY7~;UaK<AU$N+&}5g>@89Rj4-&xf8J2i&ds-Jym9
zZL-#mj%7o~X<+(3f%gzHHHt8a(cu!DVXdGzacA(#)%nudc<_%?tJCFngf8+W$A)u!
zn9`4JIoxIMlcT%*+7e&>$>X7+Q6b9pC|ew}d&H-XzrG)lLDOtc2hE&s95Wq^TeB}<
zjds<v3`O5DzV^O$ytuq1j;65jf%fz9D>K1<*N$1rt|FmfcLTsS>C)#+GFx6bswO^v
z1s+S#g7Co+bl$5-WV`de>-w^{L)q6vbMk_(U%$o?+vD<XNr6O=!pqu&<Ji^@(SD1l
z8=2RrIFZwBC;)+y-xpeoE-U=k>L_@C)ndCShUR_bu}PckC@;&-?$2D8y^fBKrHKi1
zoms)c-x;cfR${Qn<(->GYDGNGxSKJX`uzPi%6ac|keGZ)PE0!JCz>5NB@S}Nd#(;8
zp+b*#`KLYa!)_LIQJNfn*5LJf05gpA^@%*lOuS`nr4;0Xap4pCDn2<+odP;y-j?xM
z{mPW!GI%5{<WV&}h%`>YCd=O#N>H))ZO7kILPBE8bJ9_^@OcICy#vn{ds*0oeqzPf
z2t^_)?>5KzS17Ab$<0okVDiK7h4mT+-fw=H&zq(bc0HZa9ZRKVCGOy!zsm$}EinAT
zm%k%A2;|q4ndu{{*E2KMn5n0HZ_Q$wU(uJ>){YxGN;o<f$B@R9r)FiK*kMhe($h-R
z+Yc`36hp&TC#so_|EOBjW(97M&<{%`G>;0A;;`GNCCX`|e@^&d6lvrb=SO@nB6LGI
zl=B=s)B$HGD#fndtAl4N=%=}6gK2v+iC-4sU{SP=zFGPWzpr9@7>f2D+qrXk*|zO3
z6?%2twV&T;<whOmg5^d92`>98r#3YDP7D)<*>#ZK*)a|Qr*MX%^6vnI=)~F9$k4c8
ztuiV4IOH9eig|kOf%!MP1T)o@rrif;hgWo)A2E}q-u*jA3ia%;G~D}v(f=MDX<S@f
zkd8O{=Q?f=@SuQfY;bGoiDRrc8scpD;^1*2w9g(EQPsz@y|&7xchlfPL|W8DT%|nE
zifN_mBVd_j1F4|tH)9HJ*FsXK0nR*SD<vTC{!?P}C^MN1)4X@yaNvHLUlG}`s4X*0
zw5;G`h-m>?1WSF@)UFpm?HRYi<F|i|QcRbu!ED9z(Xs2rBC>`^;Kj}AEy7#o@I@#2
zC$1H*SDJq2BMhJ>_KW;7;iUk3P@9Roi{*Cd>Gyhh4)^`)(*ENoan^?mWMXi!6<a0T
zz)}s~*!Yhplz6NV-}!K%5t=d-A#2eXSA~*A&tpIP-!uk85;e2t_RzX>{ea$Ke>k%K
z))N%1<NeuUgThiAd39B7)O<9t(CD}id0(%@Wjix%<L35}6(QJ6r7~f(YF8Zr+ncFU
zmyQcbXyyVj*Yg<9JAcyD2*%*AWhxScx246LVV8bO=2Zdb=b}Nfl~@xyn>O3Osl)kI
zK=)NlB!OZpB8wAR$+6*@X%1(|uBm3h_d_7TcKV$kz`i5LjkGG%wDX}UL+c)<$DTvg
z??ad%a<yl5w(5z#xZ<Lj3{eGTT^G9_Tf?ZCy1SYWD6Yr-Hd@ic1b#S=T%9cnJzWK+
za$>pR7rX}fLnM8Kc+fjFm*P$J&RmRWPjDk<<5LGI7Jqa*JkNUtl+*wSPUIb&{J9@*
zN9Ul&A~#FH1d7<C=mXHrHd5x~v)`SQCP7zeEbE&!3wL~?YTiLk{y|be<oLC#>nU@+
z-rJ%0tM=Y5aeX#mt^-Ezry1LwDiE4vBuZn&-sbyZ)$|fHgG3$!aT~oC5`_WEo75{6
zff*LhXsX3>$CNw0B{2v?8eN#?qj;;+y1mso&T>qaD>2d}Dy)m4vH3jOl&WULwSyN%
zgRtFnJiV;X-0&a3h0Z8LO|UV}>YxJtuxQKVWZ6QeOF!H0>+qrgqc}D?YEM}uEBY0m
zS0o)*beJdK2=JU}#3CxYFI2EQP!qDBDZfY-aPL!Rx?4Eb&-RPiVkv$5%c|Z0)9ouf
zU|e_xU`0viUU-O9eeMGlz-RE(hX%iZ6+b9$ID^2UrjFo1yw#n7JB{f5psPLN&z+ap
zBa#Hd>{x5CG?v1%+MHij1T|(|k4hbqQ@i}!QNK=3cp4iTJ}P_~%dFE%VQ-9ger+RK
zgPC5t?dwDvEBvKY^Ua%NgC9QCz+X;v9`95&9)iTO{A1zdAOQ@Sk5)F+&IhMue67mW
z;5zPC?`z%q-o?O*o^lsEZ&GcGR%zv<Sox{xIg>C!HN8R@3m(L+VI;l1Pbx-O<SHlh
z|Ni~^6BUIy@4c)BMi;eId(H1#u}2JS4F{dsk{B9KZ<0r#0Tl?LlIl|1eSW>dm97=E
zojp9ZloS@dKI$zCVc3#Fxi2HYtD=*1s@46h_u&LL^!56NW6y^?j*;tWjWdFf+~LLb
zhj>~?F3*zNIJ-kdt3YA6vKnmw1iGo|GGvB_X0%~qxBT6n8^G5~ZpUvUJ9DKg3-wVy
zF+l<;`IoN8b;6(dnCe{Ayz$?P?t0VXTKw-f$W0_oN~s)fTfJ;no9uOqh_{#m7G}#M
zOMJYt<&=5Z%zr&iwXwrr$^T_bd6~bmPNNJIH)(@YYxWgxBfAG^N`zg^w^tpv`?p}{
zZcE)k=TF4n)?cPrQ@)tcUv>Zm`{@_ZA(V^BM*(@o+Sm2-H>-d`y=4%-1FF3UW$NKm
zz#5E1x(u${{*Da?ql`V9U<lC+PtWo{7_6&5gM4uj$doZG`TItZ?flZuGjD^0lFbb3
z#G96)CEyxdqqVIkdxCqiYaLASZjUs0kv8pi+ic@R_Ut5I2$A|dvL?#iCKanDL5Gb{
z=Hj|yuA24lO<0BV9a88UdS_52>xxc%*gDVYKkJ?Bs%(Uw`l&74khe~p0E0mrnx;8h
z1~lml-v^Z}WIgXgm)=4jCDx71b#i@HzWcYFvgo?jZ`uG8Sa}1G26L)%rNf-zV%*<7
z1(*?v0Nepr5$yv@$1~{`<(C&}X(pxXbZJW?8Ej^}B!sEb!}P(hjR!QW^&n4~R~ZaL
z?n+SR$Tao_j(JI>6?uQ4uz)}dXsTv^8`J=6XVe;m6-U`4Qw@Y!vX4ny{LUjfnzTnF
zQ_<iun><9za~(k&e(jZG6|o6v(ID}XjxTL@t5aW=3uIkEf71<xL_p@4+(cL5u4iJe
zkd58@&qqP;YzRJd^VC1<)=eM=DZP7pzC}8*;-@M|y3+hl8AKg*+;(h<`he~LUb6(7
zZ0D$+uiHEgq<0zMI&&3kHQGZ|<oB;uZf7+P5jhs9pcE}WXH~A57A+iT;!a+*LwYJD
zMXT|vio%Q%CgO=bNRiKf`YtqC^$bhno9@kDqVz>?bNIR%tjv~5mPEbP1V2vE?KC?-
zCi~qPuir`19zV@|SlFiSPFZrDzvo=>OXsPO+weE-rI-e+a7<yZA;l$$F&>CqOdhsT
zx)34hOacUApeA(rc?N>ZK2z2Sp$x$NRlNo$<bAQ9Szeah4>L;8HE}4n#Gj|64{I~D
zcn{i`)d|TGWc_v~vR+c5MjJu#DHZDuRG2-~xT*9?u#_lW>&$OO%zY)YmtpG2kRsPr
zT<(EmmTjrTD?Pd9-`2ISJg04)PaU!JV|w+K73F$HNJDkN=>ID7s6_YRnOK<y@>UDb
zt$rOQWt!+7sqv1Tae{%x9ZJ-00O*YYP(m9RdN6?OnQxQLEu6>xMTHpR6v)BqsaCXK
za6d!C7OnuG^eeK(K*z~L=ePWp#ZWS|n?oZ>a@Q}8RgEf8l|n4@PgMoFeVdDu%<e)m
zz*T`%fpm6tePiTU52@?FdVmdPo+t9|py|s%52?s$R3UA)U&N4;{KOF4O#i@T^yaR1
zE7!aVEmRQt-pKLq`vJdzyy+;Hr?tMmV_bs3O;_h#!r<Y@fTWm%ctSCtz?Ag63aTgW
zssm?cW){$6vr{o+YOmb=a(L!<woAz@T~3^f<aH~Pb|qik^$QW)Ov?c%6mH|nCbZD!
z=rRnwpvx_2!<=D2KCewy%%-Km>(j;#jRgPmOgq*>E<_bvF=&3&a6^IV+NRw;Ka1)h
zGft>TTsEbF0Ke9mosuFye@V3nx=CHS{w(f3;e6KEyKY^FMHYr!t_wQv=-&35&*c*k
zC`sBV=T-PRCI}M15uQEch0WGGceVl<U*Az;-dj6qLqI#zT|4U-l}L~Q@lC8S!)h<^
zhsr=k+QAmbjBd`Y1U{^dqbunADXw{5)VaK@Y^chp`PDDP=myPZm-&{k)z-#*?A4O$
z$fb&rwLgZ9ek3s45b)pjvEoyPJNy$V%4RFo5p)b&zrat$jdV7WyqmcV<H$O&)#{r~
z!n$+;=6DymI<E>Q6+vznPiG*;1YS0Ld8RL3CcXe5KK=<f2Ye51^z?w=o*;w<-f8h=
zC1A>cdFtu72I(Q{mCCy3+X)N_$Nh`jv6f5MKpu-J$L}V#ae>W}Lx%0r!op7aDBqyv
zlbd}cE6Rx<OXRZBRioBnLT@j&zDxK05{s8hvGDtMITclrWTXKgGjt$5i2UFv*ljx+
zcy)1+Ir8cU;#8cw6pjyc86wC4bkhLgY|ofvO-G-s*tI%eE_5;zUt5_WRhb&cCc^&S
z`4YHU$bq}PT%lBz4utV(s86LsG@tE@&9mCzUp>>@-0$ZDBH$~F=s`GRD19YGw{gWe
zS0!5S<P#KiOxg){t0<X46HP(;wE#gQ9-Z(YLJ<N!mLkJc)4M&g<_i1uP^o>Zo0SHB
zpo7g?Zl}S&RyDraJiqAys;-pV*|A?hWM-5FU4sINPGglop*qV{xhUoXwHwa1nZFL=
zb=W_HQyOBdV#{X@f1pd}#B36!%e|@C)A={O@es}7u~BN9aqwv;rgiM?=bAA#Y^Z^o
zt;<k1)<%$R93aQr-J7jMgZ+ceVu52~Y|<*n*<&p>4>FH+*tZ0Pg;{8Ze%Y1FOp!b5
z9)dhbOF@n$a-Q%TA!w=N$WfdvhF2oaXI|R~>0ztCP4^E*+r|RNG!eik1MPo2*PP*x
zh^!1p{TtER6#5ks(uBZ4w#Om5^t|V~sqSB;)uR(52LW^pkohV5Z3#RlpV{Jgf5B+4
zaW#2<MB#uoeO+iL!@`WB2y&`H&T&_IeJX+@W=5A(Qv2jx!C(<$j2kg%TFeSR8$^Z5
z^0IV#x9qrOBk4Qv<*^*#!3h*-;%7e@-p|7HaQHJ|jY<{^l_Ky>o>|hXB1BhWbKXm4
z{7+EG8r;DUnqAfCAGdT3I)g7D5caTt;IIDqCo@%4ZD-V|t4!8AdraSE0p+0_eVEp5
z@lq1kmzb+O<+?vr<naUD-pj<YI4)nZ1ljnJejMP8ouE!YtkX-8dKT36ha%N>+Xr?C
zP7#ls7h;3uks|`@ZpOk9@k4pkXIAOz)H;HE8ok_=Lum-H={N_4Y&q2!JIc^k5vJeq
zdJ*hzVF%yOE?)E$k6@JH#WA_9RlLUHa}KHZ)_Sgo4jMm?>HYml1<nI1FmRfUiVZ`w
zS)dkt3TnJrjk*iQ{)<H6pTLg$$+lxEm=f6Yd^y(jKG&YE*Lww$P@#+|^AYo-2aT<3
zp{9t4kdRMFZEdaoRlb_7DPHQ_obK$&h4H&>UKkgV)<<m3bA^F9L~QPba;D8keMjxi
zV_`!_PtZb9tuDv_HVRTw>uAb46HA>S)q7+L8#Oa(lEBea7YrwTEAZ~U=WC${$M$12
zesOV5AZ5mE`x^t^UZjkgMnN>#%-I_B-vpx7#)N@S)2?F`+C*V#$cAUfMPCKUA^1Sw
zQ#jb;j<0Dx#frnZcve61w{5C<-tZl_WQ~dfeG6Nc<+<#Hj{$3c;sgLn{Vmm4LQ6sE
z<qpq!9mTNLyq&Ca%8#_4x;<LT1k^wvHY*V4`@RxjzwP)}r9|Nr+0JQND-eJZG&gc7
zn8^&AoEIb-9`nu6;ZMC5r!$mm#KhdM0_M3lR$0t^RHO((<9N4uJW&f9f(8w9ga+}5
zCktO6wOF<vNr(=+3;%4j0c7m&0j-oYOjQP)fXZ6o@Ns<sztrj7y<59;QbB(H3!`7{
zKdtyiK?^YNR$dz}A``0we@vq4uRN0UWpGOT^Cgb_PxhKtA3-N3L{EgOD_i5-ueobx
zSy9}eWL+Ub6im`0NEz@o<SlWE7K9z{@bM!RzE-+i8L@u0{C|;r#eIl&W%x4(fBug3
z^sd*BMxRJto9Y<jpNMgA%I6q3O1x5buVdzhwVry;F(=-87KCT*+47FTj7=J=m6Tk_
zv2kq}|A-sVKdGLng%k<FOvVt7nK$APLk%b}SyV77xD3!1uo22kOUeBqIr@d{LCyA{
z@4VvTYF;|L-V3(HM6)K^@jnas3x`q{JGZb*&1-d5d;H7<!Pw8^J5FFs$5G^&2@&hD
ziO?Bbb4sFWrlLlWZrsmLfX3aJmGo1}*etK(V5`DtRJVOm{tm`q=DP;1L^Rf%f~F3I
zmm48OuDK@IAOMhNd5Jm8_p6po|FY8>rn1CEEUBzXc<REaL5vrqE`b06v>Et)R5&d!
zzhn%m8!U!<9~I;J%4zX_!w)H56iq;jTldFF3ziIsoQwD--Crn?#bonFc)pz;*F*Vt
zdFi1u)DL)-n(9y$q{P&LfI@K)!3EF8_-L`s2vUSHZ$xPbBFE3v7aKbR^0dQ8T0VfJ
zzV$a-9V74U+G{n`|GR}UBFD-wXZj*?kTdBEfdx<lSSuX<o)n(>X)uA<=EWhgL#<2v
z8r{@f*xPpR?^+e9;XfT|BzQp5V&!8z7VABJi^h{J=XrOH$8zIzD#wci>%s$fm4)i+
zYcToulX_K;^KTNxE8w4Ht8ifIyq*l93XaZ(u&dWe=U&SJe92M205QXd*-E_(9d9+*
z5GW08yF6wCLC+LoXZI6DGvO)^6CFgR=K@C8u(T{@fhb$56~k?^L<tK(1sE2I*?6+f
zxX)Q)8BluhRr-<&aQcPQb4N0snE|>FAKA@#-L1=QXQT19X}tk3Y*vei($=Pnv%GpN
z@pcL*(~p)~J)3rN-oJl;{Ymu|Z!EMIn0xW0VPS&c&n?hIewKd*O`Dzosgr2E?US}q
zg+{P%zvj03qUTd{q*Ko<>B7u<(xs|+FHGv=OdEeSn?;n0MF`Rm%hLere<0bRG!SdO
zjy{#+g3iQz3zTO15eMr_=#8o4-TsMTtsy|72UAN88y{LE=SbQBt8;58gFaf7z=>U0
zB((G0L7>CiRK8GQI^&RI!iY^Yy!`JO^JoFWH?jHFVNdnkqFZzr=-I>+IkGLT0)~$K
zkTV^LORGmb5}=8rqNk@Pnxd3|DpCX|8Q|gZ7{s>!Dv@e`0JQ)90PQ~<Aa!qU&nzqq
z|8bfFDyDh&+=xaY0Ba;|STW(bAOKAvlq!}r{HNT9pY)R(EhikBUJgE)50xr5v-PN+
zk=n)7T_50`&DP@>A;IKvNhqP@@$&98Ww}a>3b21Ma1GXCV<QvTw0X7q)#c^h(ejUx
z%*dP0%}=Tc`)zaXH1k(M;}oeD+n6AOA$LQ8nU}vXWF~$hD%dc8DgR#>6}Rc=w3^R1
z6B@5GwhY7v5n=PMtkgMAKiHnAIP%H@C>yUEaHe^kw85w&h+lvs{fxi4!XXE~UEI_#
z*j~^*InGb?IreW%u~=ZxAO9f172-%zwDhjHW^P$_!TO&RatLwfL^TKRIuxq+#^WMf
zi^3s(c6}XE+C5TXlT^pGQO{;gCrpdx#@n`{35jVZdBz5riXvi4ivN8*lIRhqVkVTl
zE&vc|O%>b{>J^hUG^m*)h%g_<Xn6P2n{I!9|21ksyW@NfXiefVHnZV(KtKRn&M2Tf
z61WLm5lPx0PT_zk!_*aznyMO`<mKKd*GJ6@0{tO7#@YmSJYpCO^tce}IiG&$nQd;k
zzF`xUAh-Jus5>5g+jpGa`T6OKk8}9v3uCT=L%+KhWs_6Bv#R<YS*n_GcKdH1aKll2
zu<I1O6?-w1cMOCeYHI4dpA^bqtwHx(X8hW}9@v%<**c9x(x3WP1uKVIS(3H0Trux5
z&}ZCXlLi^6f-k78btE8sY6(V0P56PkYR|StQHBD1ShBnRtW%6<HVVCj*yGECjj9?k
z@X@l2Kmw<ePGUe!T#+r8YK;b>;e}$EN*n<NbLk@<o8;8xV>Q2Cf=Q^3gUwB+RJ9}(
z$Pe|Z<v2>DU$C08UdCl{yY@e#ge)`|vtVTXVRiy_`fLv;crE1jS#$6vW&oIdkjgTl
zvJh+(4i{>xSp=i^bjlXGX~GBrs7y*1c+xXW)%+K*_d*XPARtJ8{rdF~?4LbMLdV91
zO(xC-e<hq*FvGxc<d4%OJyBxkp|dN;%}YQ%Tt>;7zjONYC$@k7kh;41AcbY@&)d#p
zXpT1E-wU0ws~&YIe=}LN7UZb{OT~6v$W@6)@1De0ft7+bR`fk;4v0-@>4de>{ScVj
z=?7>zopi^q7k_sIG*_8*bs`YO5WXrngIvPeb&CT%{jADm@5}K=cI*#@USbkHJ9)n8
zz%=B;_wT&FZ0e^a+dtoo?VLwtwelgMo9Qa05mD=klo2_a)}1-vHvMT3F^w)^BY)m8
z)4Nn4-OiIrFQ2;e%ZJj5e#{VbUA)clL!yR<7V{bbYQ{8;vc@w5+7~a<4J2P`@GAgz
z;^O0;)r~k2rXLj`4!nb;j5AcjkSS^iq}Ib(j?{D(MR6xnwjvw{Myt$2*Y4fxFQega
zvi1B69#LAt+SS0~bAGMSDoecga<e8Wf5mNIH3BK>2g@yH&A+K_2B=p+yZT@@GYCW;
z_TL!+T~x-*R~YM(%|A%**Hv9^?&KkH@1|)Y@~=okF)h{5qSu|?0Hm{vAHY}dB(N^F
zQ!yt?TxKFCCwXO{eF$+zNThp_mL9KVAERQRAivZo;w}(LCj_LU$q3#-UlTp!c^JK#
z9s}g);6OkIECN{vrzk>We@4!Qe(QBj<^3unH}7KDx2~42aEhK;r@4z`d@bD)iULOa
z`;Y0zCB&V(TDRu5q&#qzp|$ZetlG9j_Jp$lxNnj~i7dpi77!}-%I$bKNn?4}tMHB^
zdvY}M2?~Cv1j<n>K6xTw7$Wn=#nCZGKtVwKfjLw3u+0T65@M$gq7pz9!2No~BUV!4
zC}!Aps3&p6<kf)W9zUNpDj)S`NRb%QT^6X)-uZ3gzm|Taf;<Vj^ptz9PN61G=W!+P
zS76~rfPGl{Z-4$S;PL<m6AP=fu@S}~haqxPBtFU^BqE|zU0uC2J8NClICmAqlBM<@
zFoxU`(V&thiG0t`1z#vCoYPe33s-!%*S>#Yz8!a6NNBDG`*K9j|5(R+oFt&xxD6&*
z`Sc^?eB<=eGwg4c#^Z9Nm}1#+|F^MP3)Bd2AJW^s@crWlBBBnC6*I1#c+pyKn#gg0
zsl+lM+|WN(Hk8#8dJd{;lmmirUBJ~*_N(sa#ARJF`I!I3L}T{>GLpFkBG~d9Xmzv_
zR!*A7>G!fCEjLMn+eUbkp<}e*MQ?&wlJ<-yzdWehohC_$H9CekX+uW^Ru;=jW>Bt-
z52KNY=oLwVXa9{)s@J5I6M=pXUG+fvm>LLWl$JNO@o2|x4j$Xvr*)N78#z9Ek;me5
zxInh#zmrFT2>faI<4cy2qW<Oc8O}t2SsIOtN5Ru7ZxC~;mViQ2&$+bRe+~!o*8|g4
z1JvoLnl}?=MXeg(6s&IRZRHJ5=5u^$8dy-$Z|jWoyb)u~Juh`2b&$35b~#TYIY0Rk
zYm~kv;1{N_#2w5G1f#zoiro;|fC}K6BfOVSkB*Mk0sX`<u$OuU6RFXNh>%y-!2G({
zEQb&PJ`WEMbxuD32Oty{_EZ@*4l8&b|M#Hpx3NJS+)SF48Ms*^73|T7N_m-52_}0i
z_u8XF@-9B5foNV96^;Pm<*VJ?>jbAQq!&Bs^<Da5jt|#|{T50&Qk4M3@#iHf1aMNL
zlABoGSULIupZnwo>Z2RV_(zFel{J25ML>`>*?zKUeES&Uf5rjvfcCqUhi&rJb3oLf
z!R(*e8=IRmh@x0IqZ(Hu5S13F(u8!9#=};8iZnxiToWj1bf1Km-Nus%#mVv<=6F63
zFgeeCtGaLD{fb>c!^K!M`M?vYwU(Ym8==SdN*1ku_FGGuoJAg5F}zA!Lmi}PP@beH
z!YQDXzqQDu7vpJP{PxHTpBMyQZN~G5jWl6Y>J<WZ65QUtzI2%100-VL%;Kh+h~GTB
zcn3GqyU#({@>U26{;-MCBL<yxvIv<9vgL?j<$jq=K|xXE<QHdwl^E<%t4@>{rcWa*
zh`NzgasJZs3Rxc-l;C%1F5}Js11Vn@2CH#>3(W5{`jo<bC@rTZ>tA5GMPw1)bcXzN
z&(*?}aFQot6HU?vSmdz%HbDGQQwdnH7#bP^cH&uTR738}5`}{lZ=jzbpX%m>cs#=u
z@_BL5=2EHVxIGxgD|(dih!a(ynykbkpUhvh+4uVE`iAOJ)pWK4xcijtbWr}C-B{*u
zAGD#B`7;-+RT@5*PS2}H+T?tLn79q7!3iDb)13CMmiwkx88DFj?%Tz=%kCl;wxWPI
zD=&!0QyE5R<QN@m*8mLIn;n>@`U|W5xAC2pTHKcmUvJ($u1Fl*pJbu9!~v^BS-L$Y
z{UgNI3oMNm$`=~iTjX8@A4Zf+;!sR7lLj?|M6gjwxHPnf@;p3Izfv=fla^VUn1A`h
znwFegNrob6jrR1Wj?k&P0(xce0TmmVBp#v}5~V=zS&<QvGGW+mJq;@>BAc1yWW^&%
za-$eMs@!Y1`RevIruX$>oX>|z9*vT9l$@m)c4XMz&LS~}-mTAWDO0C_ys?q>!A1cj
z<h4dtqOGjj9^OVvO^(I3E=V=DTw5*~<7I9i4uq<CE=jrEUBIE5c3M-{ToyKxv~i7~
z0a@JI3pCGBSccQPff#9q{^JMMPZ)C;i%Mb`MP_o`i3Va&c;F7~j%aRPM;xA2inY$n
z;6V{Qlf&2585yHS#zi&kS2NiDvO1=7-<{XYevPIT7QS?LbR+>0OBjKqJun3e%cC5T
zt<23qBZQYOG%Zu`2LZJC{=3T3g0F|Ljo9Zw^QR#vAGw}z*X{)H={X>xHX0~ySXfxk
z7qi{)&Yd$sDf9jAxkNTASE^}U8yU-=w%9q;HP+aX9VgA=rvi~j-=NkOL`2p$Fjx#W
zZ**j%fbm4=gouo_J3P>w9r;FK%eMvkzST8DVp@1yDZJ8WHy@~@t>1eFFK#EliXxc<
z;B+ZKgJ7?IgdaWNe|T!c)G!NNsxSwenAAOgpliM`iTFdT0wo^(&R35qBP}gb8<Q|<
zr@^s6^4SM8J1z`T{%|cCR)2Pn6RZ%|Vqudr9ZO`6JpoiPc!A6-%@ezahg(*ByKo0N
ztISQ9AcqB<@#WGtRD4zPG$PZCG!cX!+e`M_ib6d#P-qqhCX1t;IR(rc+HYZ#+<WI=
zS;#Wu;HS=?r9~D(2pXbdgz7*^RqJH&m-W3&S4zBY)OL)aFB+p<ofd^o?AniM%<G0*
zc$vW|^oFUkduo@3d!_s1jI>v-v@?%bE;Jql4_wl<@mG5BeO6y+MKWqnMyEpRH<041
z*`MHIhy#>w(0%}CmtxTc@;nyD$5FPw+{RmObj%1DqV-Jx#uZ<BEDsIo2mitLv<O_S
z(1$MtrMS}FegAibswhlllW**uNlBHLu&!p3tD#`M27x&1`qXg#f$eaZtm?8=8Vj<U
z38vp=BfD<qyVL!SW#xT=Ck}vj!3e_uw=X!sVufwoMC9VRM2>0WMlUjl==ghh+ztZu
zwr`>5?~5))KHp%{8<h5Y4qvtlb{g#Sz%thLx#Gslr}C;_3ys@zef00wcQheDQ>(|8
z$icyVgFpun4`TXbCz8t8B>=K%<)`Oy334E8si@Q!6LD7&Y+m7u!$=M!iq1Rgul|Zf
zm&4*%h=WQ8qHc=Nl=io$QAGj&h}WLGbM@>Ob<yj&AUb``-?Z-=TL5WjVR`w`QJ=pu
z;#*yNn?vOzsWQFbfJZKl-HwTTWT%B)Uyu}e&)v1ahZUn(BUi#t!a-f6f%`9w9<xE#
zL^v@@FZ?%C_Q-265C|!A&+w$g#LuvQJVX-SnSZ6I$oPfSK&$V+>%GAH5Q20}awyUv
zFPtu|bEA~zCN4XqM5%WAH_B&ZDZiP@gF0a3x!4?N#p|b>6n^lhz6`K7SE&1Q(8)U&
zO{~&!DpS@Qx*KU7Fk6;zf5+oB)giig?u^pL=$38_o*32@G{Gxc$jIHGr$C?1@mxgS
zB*%MoqJ*}!wXwTP+T;HVBc>wZkh3e`0l{GMf5!llkRJTd`mP(4c@cwBXd&CJwZU+?
z0Mf8aQ1Xi7W~f>gRDd!pOJ29t1;9}LzWk;a|B}5s8I^b3HRL$GY&~d)7%-5nbC`uo
z_wBdw;>WjOD%hN8`jdA|dV<187rz~@4&DVb!@sO*616kad~4-ft<=3+dUTUrxzc6_
zd`Obgjf{+*?^E&qa(Y!;HUK=nTd$Ef8sBx;B_6N`rCxcwb#EY9XJawb(pZ`x%QR>`
zP9~Q_p&tpHzJ81RaXnV#J90iemZ;ialj#i5T2f_mUEi$lwh6h{)o}qcsPr%Qn=B-N
zDM4QVOp^S#HXV}>zm}8Pg%x{rSQ*ZT#QZ4uu3k8!!**FTKk(*X=k#=q4;ur5HB(SW
z%4KAml50`u&=hfw4+2mNRBKo$I&6oGdx&>jz@gIbHD1v^1%ADav5imGpHiMkWMU^x
zM09m3?yCBY+8uWI1~5<%!@F%CA@e67pUCaTo8;)3D_g6PVGnl!Rl{eHbm*I&midOK
z+1KOt#poi2u>7o{EPwPuVR*kzhjnk9^=VUmO8T8}P>+T2%WlH;AbuuQn*nwC`iI1=
z3l7hLA}J>C;K{fDsMqmi02_)X;4BjN<}WvDAFiXWLmf^r&SFmKZsN3W?PG0pQ%gPK
zRKgGaC7Y{tf0W2g?(T7#&8o@JW8q)d95kta(LbP`A^lJ5WK;iVChG0&e)``c;391C
zw^CliQV(x9lGEt<Z&!7<Ul$*kmM=ANep3}g&nq)xUbTw(d7Sw-k!!Izm&5xla0zY$
z!2d}GTqQVYjB1`0Wxz)!KPy_u*Y)`rPUS`+ZUiX;1eF^I@oZ-P#{*(+Glt8(g!Qb`
z`79jBNvxga$ffU*c(mkr*@wAWvo8Sh>qa=n{Cx)kbY?b(szJ?d5m-G?m!YpSvi@J+
zT;P|qCp>^rYGbcvKfnVU5gq|U4@TqqynQ$3W8~q}p#Xb@cxu0o%ICZuIUUP-8$tTT
z-2Y5i;Zjni?un;?RY0Ppo8fCjuRY<!@4}|mf?Q{ED}L$vo{}6t4X{#o0U-r`laQ;_
zUEFaV%73@#(ck752etX5<cg>^GyO&v|JkNx$s0+P$DTe$Cj2))Q(Me6;>Bp+Q5k)%
z^WSnZ{iseN49jWK=l_87t+N^W>&j;-M8n!fGbE&I-`{lZy!|u$Jy9qbwl$Ij?qKq*
zqseLPxirDY7w>b!3yJ0^pF1Q_L!gYLHjXz`7<2GF^w94mZ%>Eo44Mr8W{s|nIQ&ac
zQ_PnRT}j8n>?ArC9r{l=nlrLR1#67Y;-)WxJ;&Itd!D|WF@*IcM%qq!KS})=Gl6QM
zmw8{1uk5#Qf8Ab|7i--iW2oa3;%uJnxbt}bfd1NY#_|b<+<m*>&Mv(smz=8g<R|Mo
zayKscJbk9gX1t0|(Bq@;qri%1k_;NqD9%YYJu?Ykco(k$1x-q~>dR{S#8hPs-H=bd
zp^|fz(wmhGc>n&OG{ZDox%4gId7Pc4hx)hR+dJ2F>y68W&)*!`H!$E>6nAtxXZc_1
z4}#)0vB6X?q)DATNMmmqhlk-MNqT(#ad0t<w7AO8{2Zv-f5D!nx^AN9oL>vrA16<5
z#s>3_m*|e+XrR^3&Q7SOr{@?&eL|L=BtOr1$9J6|oBWSe97|Fshkkc|#(MCtO!0eM
zc;N72ueUmq6LmqWqo>;gy?-1sK_H1X3v*|*>$`f-IMH~``%g;3RzS~xN!DFCALVbw
zk7m=+glNB{4jo%6vcTk*zFR(c;-tr~hYEm$=7dv!0RsiEiMW`GG;Cf~ksCft72UO^
z`d7t$JbmB{51&B8QYOUbRk);<(1CEFffK!)@aQYvn%51kDZp}IsdmpBY-c&upK-DD
z9#BFQZ$9I$q#?iG#GfXKP5;+<sCQXzE7@;rrN>_Q!nM*kf77&=ErZ%o`ptRdgIZb0
zUNzbw^XePoQ}5GGN)N{QRpH6GYW}t+Q`THRG7;6^U61+A{F6U64xf?utIW3lDE)pf
z+~B=$P-^bi<mUF!u_&!i=V}&%eL^8~?0mC)Lsz<7YdoWpI6Jkt=m_MGCE$>TNlo%j
zP=TA`<ijz>@5=R!${sx&Y1W&9y5i03R$lR>WOG>qq<70pKV)3B5euL`O&YP{EQEjk
z5`9ss7xy3=I1vk89gTF(kpjD#vDzV{bHjJbR^*|siC3c$b>oNeg5~dpj(%*X1zMhT
z7WQrb0WA|B8;ShXR=LPux46;VsR{e;=bBf%N%bK0+AXHmYMZjMC!Y)WpXHwTkl?^7
zmrxCsVOG}5t1DHYY{QUoD=D~id(pIQuC3|>V&wWLuC2&Lgq;#R1^L~84VU<0?=@dz
z671@FtPi}LaC;h=NGL*CvOwl)q%=_@#fIj$7=~}urPQs6auky+N`}hTY;P$sXeH^t
zHx`zlo{5Fh#<}9F%kaAldd!xT?(2QDU?{8vw?Y~{){PqaXNG=0?X1s=llfFh%{VI!
zNLFob&bEBC&(EH3B|Pk0ht?AjrNx~1&aHM7o>=5fM*#BT8`rj1k7OiV$-XZLBZEb{
zoK<Ie*=+e4s4>s1Gm}AHp6$ma<-zShpIj<Vn46qD0+6=!`Qi)2uW^I!bN!HiqT=qU
z)Lst>bT|^tiHS1M=!clGqyUOURt70E>=TlxqjG9SL7DLV=1b~ztaVyjN=wMC3}hu8
z*|{6*XPY;@(H#}wL-e<6U@<v<9f<3~A*?Iynp&**&Mn8ELzf8MY}=nGpdm8(KVDts
zOT8vNHss9xrzc7V)I&k<pB7inQ3>LnNHSYh{u(oYnbyXM*3h`r&jtpl(!9R@bVXkD
zuOy<)w}w5SQwR^<?NAe4n+Gm{B|mnBOw0YF-_F*O3*3GuYZHS?NOWOIgZ}ipE@qm4
zhZ^>PRm6@*lG3*~yt6*~EK7;&GWE@}=!t0Rs(R>BP*DUS*-c|zN9(tkL}GIke>HX;
zk6wdK32ct&F%DEwrPbuj?kDrhu<GQ2yZzi33wLauZoA`8#*sq2mV=SBwzNH|D%T%b
zgZy6Fv*nH+AY1;a=sxPU2X$(un1x?k>OOS2u4pCGiIT9U5iA9b&udT`F~4^l0ousG
zA}}im;cW)}2D*z<7UrU0)0cevQg)(#s=+&RDZi6B#mI9l-mZ1u-igWFx@0rScVy8+
zqBH;2%U)lkM6DwCCa*omwfC`fsreSMFQ)vE#ta|O=mVp|LWUEYf|+nlEBWJrMI67=
z5LR7l&<vz5G1HtHZ(i!!GV&(;owc(X6_6o&%7Z<G-i%wKwwie3oZ~xqN6sH0Owb^<
z&dHsW9uoL!fbErjUrF!dk0S$Bh56JI+SOK7lH4Xy!CSNHAz}n!xG&O|vff4KXSH~S
zto;X;p6dqiBionLE)ApRvGOXJysSb^D$|hhFPL$@z_{>Cs~9CjL^^?+W9?k>{^`eW
zrq1WM|GP!Ex<Wm;DXLL%^iMzXbTqm>r9VBVGqZKCTLqLRShrYIzp`^cw{kCA^Lw~X
zdHFZ>;J51FAQDET!>^xDm984Lg#TTJsgXN!eXGBPK-w};tq~AB{IvxIJ^Sd&H!Yk)
z{ef;~BVtT^vVBDvUW+;HfBI+$4hubHK<z}Cb&VMi#ywuVuWZxX1Pm_+uSysB732PS
zUoBZqQZQ?4vC_8PaqFU6&9T`DD*b>}<S1uD*#@9&An+_NGi~?@y;87Ze?Zz_ET0cu
zvln4fQWPP}Vzw+|Jl$9XKX$@4Ro<E?t1Q#+MtOvT-!guJyd&zR2<44B3g3GHe$u5!
z@+V#84bZH0f`NBx*<U#~x2LT%E$qg{_9hx3#(vwY#Pgd-33Or+LxXva4UNj$oyWcZ
zgx+PfWu~TooUx;CQY7ZR2ON%8HkPG;g{jZ$3ya<V??>QVtN;7`|N9&M^(QRXn(g=K
UtR4QqDYhVGc@1cdtoeuk0VJ0g!2kdN

literal 18213
zcmV*3Kz6^0P)<h;3K|Lk000e1NJLTq00A8U002h_0ssI2wQ2+j000U@X+uL$Nkc;*
zP;zf(X>4Tx0C=38Q+HUC_ZB|i_hk=OLfG)Jmu!ImA|tE_$Pihg5Rw34gb)%y#f69p
zRumNxoJdu~g4GI0orvO~D7a@qiilc^Ra`jkAKa(4eR}Wh?fcjJyyu+f{LXpL4}cL8
zCXwc%Y5+M>g*-agACFH+#L2yY0u@N$1RxOR%fe>`#Q*^C19^CUbg)1C0k3ZW0swH;
zE+i7i;s1lWP$pLZAdvvzA`<5d0gzGv$SzdK6adH=0I*ZDWC{S3003-xd_p1ssto|_
z^hrJi0NAOM+!p}Yq8zCR0F40vnJ7mj0zkU}U{!%qECRs70HCZuA}$2Lt^t5qwlYTo
zfV~9(c8*w(4?ti5fSE!p%m5%b0suoE6U_r4Oaq`W(!b!TUvP!ENC5!A%azTSOVTqG
zxRuZvck=My;vwR~Y_URN7by^C3FIQ2mzyIKNaq7g&I|wm8u`(|{y0C7=jP<$=4R(?
z@ASo@{%i1WB0eGU-~POe0t5gMPS5Y!U*+Z218~Oyuywy{sapWrRsd+<`CT*H37}dE
z(0cicc{uz)9-g64$UGe!3JVMEC1RnyFyo6p|1;rl;ER6t{6HT5+j{T-ahgDxt-zy$
z{c&M#cCJ#6=gR~_F>d$gBmT#QfBlXr(c(0*Tr3re@mPttP$EsodAU-NL?OwQ;u7h9
zGVvdl{RxwI4FIf$Pry#L2er#=z<%xl0*ek<(slqqe)BDi8VivC5N9+pdG`PSlfU_o
zKq~<N&6lL(006w`7+k277fi+o002awfhw>;2Moa!tiTSO!5zH77Xo1hL_iEAz&sE_
z2IPPo3ZWR5K^auQI@koYumc*P5t`u;w81er4d>tzT!HIw7Y1M$p28Tsh6w~g$Osc*
zAv%Z=Vvg7%&IlKojszlMNHmgwq#)^t6j36@$a16tsX}UzT}UJHEpik&ja)$bklV;0
zGK&0)yhkyVfwEBp)B<%txu_o+ipHRG(R4HqU4WLNYtb6C9zB4zqNmYI=yh}eeTt4_
zfYC7yW{lZkT#ScBV2M~7CdU?I<ybXVk2PZ*ST}YR8^E4n?+7FUi+~gC2wsE`!fb+&
zkVjZdSVO2K>?5=ix(HVZgM=}{CnA%mPqZa^68Xe<Vmh&qSVpWS?jar_o+Vx<4ijIK
zNF)x)lH^VbAtjJ9NefA9NZUv)q*J6m(hzB!OeX7)ZOPu`2(o~zAeWK1kPnbglKaWS
z<hK+$#faie38ExYq?8g$HDy2L1f`!cLYbhdQO&8I)Cj7GI-goeZJ>5gFH?u96Et<2
zCC!@_L(8Nsqt(!wX=iEoXfNq>x(VHb9z~bXm(pwK2kGbOgY<U{4TcSa$4Fu*8EYAP
z8K)Sx884YkrUlcNnaETy*D@QKXP6I|Z&g?-_9}c8k;)R4I+a$HewF8{R8@0TKh=4v
z3skFB5362QeWpfLvryxy3Dg#=)u|m-yQwy=&Qf<$k5JE1U!%TX{et>q4YG!XMxcgB
zqf}$J#u<$v7REAV@mNCEa#jQDENhreVq3EL>`ZnA`x|yIdrVV9bE;;nW|3x{=5fsd
z4#u(I@HyF>O3oq94bFQl11&!-vDRv>X03j$H`;pIzS?5#a_tuF>)P*iaGgM%ES>c_
zZ94aL3A#4AQM!e?+jY<CKGS3CdFcuD%JmNE-O)$&ZS<q{7wYfU@6jJOFf<4@kQr<-
zIAie4kYng;m}$7t@Py&05zA=0k;G`D(Mh8xxF+t0XX7<^7d~dJZyaK*G~Q+0Ydm3M
zX)@cS#H7XKzA4R=Yno=d(X`Wa%*@Cv+^pEF$?T3f)tqadVZPbC+x(4%rA3^@N{cp&
z$Clcbe9HxvO_ukpm{vYkc~<pS*Q`m_T<a|BZPr(8P#ag944XQe%eJVko2|rln{D3|
z;uMc5(kb;*ZrU;I{Ok(sn(PMcIrd@pCH8Ih&mGJh5*^k%bUS=<bal*jY;e5mq~SEf
zsl=(n=~rhPXQ6YQ^EDTyOOVSFmv)yIQ*Eb;r*5Bm%a!FC?z+;o)Agg9yPMpt*=^L_
z%ss_@tNTqZn;Xep!#(do^zips;&II5ou`|p!t;>lFJ5+DSzi0S9#6BJCZ5(XZOGfi
zTj0IRdtf>~J!SgN=>tB-J_4V5pNGDtz9Qc}z9W9tewls;{GR(e`pf-~_`l(K@)q$<
z1z-We0p$U`ff|9c18V~x1epY-2Q>wa1-k|>3_cY?3<(WcA99m#z!&lx`C~KOXDpi0
z70L*m6<QnmFw7=Q9@Y_#hR+D!5Pol_`Aq4|wg`yeM{J0=A88qx7x{e@DJn9mF6vRV
zQ*?23_bk?|<XQV?y^isZsf@Wh+iJFQc4w?=Y*K7v?3=iNxT?5;c!&5!@s|>G6C?@k
ziR8rC#65}Qa{}jVnlqf_npBo_W3J`gqPZ95>CVfZcRX1&S&)1<g_shTvOnd6AVN?t
z7*FM=ZcQB%@`Rg(Pes0>jiOPpx423?lIEROmG(H@JAFg?XogQlb;dIZPf{y+kr|S?
zBlAsGMAqJ{&)IR=Ejg5&l$@hd4QZCNE7vf$D7Q~$D=U)?<ay?8${Ul1%J<|W`E&Ez
z6>Nn}(WA6du22pZOfRS_cv~1-c(_QtNLti0-)8>m`6CO07JR*suu!<Lv)H(JS@GZ^
zzeT$iBa2fPcP=qqQo3Y#Y4Fm0%V^88mi<uTSW;E;YI)r9j#7itrKLkFf>$(^sg%jf
zZm#rNxnmV!m1I@#YM0epR(~oNm0zrItf;Q|utvD%;#W>z)qM4NZQ9!2O1H}G>qzUQ
z>u#*~S--DJy=p<#(1!30tsC);<r`mZO5Sv#dTRBK&9u$R%>y-IHSJr>wyfLop*ExT
zdYyk=%U1oZtGB+{Cfe4&-FJKQ4uc&PJKpb<?7X!rcvow^MSb;d((Z!Yj~Zedy1(Xr
z-MB}0PsN^(d!>5^_C@dOYIJXG+^@gCvI%WcHjN%gI&kHifN$EH?V5MBa9S!3!a?Q1
zC*P)gd*e{(q0YnH!_D8Bf4B7r>qvPk(mKC&tSzH$pgp0z@92!9ogH2sN4~fJe(y2k
zV|B+hk5`_cohUu=`Q(C=<ELb&o}ErVJ=B@pdG}2GnQL89UA<>R&z?UQbnZ;IU-!xL
z-sg{9@Vs#JBKKn3CAUkhJ+3`ResKNaNUvLO>t*-L?N>ambo5Q@JJIjcfBI^`)pOVQ
z*DhV3dA;w(>>IakCfyvkCA#(acJ}QTcM9%I++BK)c(44v+WqPW`VZ=VwEnSWz-{38
zV8CF{!&wjS4he^z{*?dIhvCvk%tzHDMk9@nogW_?4H~`jWX_Y}r?RIL&&qyQ|9R_k
ztLNYS;`>X_Sp3-V3;B!Bzpi<y^K$R2sjoW6BgY@S&UroYru?nW+kNl2@4DZ|y&st<
z{6PLt^U?Za$EVOw_de%*{`@cZg!B8C!G=}<000SaNLh0L01|Zo01|Zp3?coT001BW
zNkl<Zc%1CL>5pyMbsx6&c&2;qd~2>ZPdzugIZGrZkrGLn7G+A-Y%2~DCxC(=pAsO+
z2mKvEfFM8;IDq3AFcM^t99y<2SrbJy$?kq*O}DD<e9wILoDa9zq9wB5d-bYbGeQ48
zT-1lU`>fwR`?uHLd#$}Ta?Zg=_y`~2T!BAq13tn>`2Qh3($z=!2<MNFboCKF!ujJP
zU44X)aQ^s6S0CXcoImJ?@>FZBR$3cvjM3UyXAvNT5havR!a1jed?<#U0oGcjlrhFy
ztF^J#I_C(%7-P;krIZkY5c;sq{0Pq#&sA5>IjyzQx+sdQD3UA{i71L(8I3Wfj0Qmv
zg?_i!?Q}a~5d1iIN9UZ@Mha2ab(Up$mM^m`FLH?-CYTZucz!o-^}3x-tL^z7#y_T0
z1LxdoP%zdQV^<$wtph*~001FGD8U#LLMUTDw!_4#cRA;*wbnUn%zJ-Z`+hLVA%qd8
zgiuNe!IaVqdN1ht#III^x)gFYosMTSDP^hUcrtF{SSzKS!5AZg3C0*ZKk(N#*RNi=
z5;qz@hOcntY^$nDl4L%cXGvO=r4XV}!dhz>p^O$q(TwAm`JAzxjm@3iov0E0SV_xT
ztF@L&X|1ceE~_#xin1!TkU~nNt#b}r2LOcNAPBhU1z`|2n{gZmLBJUMQ0Q1|rIaiE
zP(mz#v(8$pv@*s>DW#G{QCe$jQ5&s|(LzXT?dt28`55Dn^H#GN20<7FK@fPJM=3qu
zo_687vepVAi=vp%X9oudRaJAJ&u8<jEZWUh4Hf|~Z2{mtUd|$fw$?Y|X6xpiJ7E}J
zsCiF?)l1E?d_JE~X0zooF}l)PqaPR<4R!!P8JALEZ7HQe5We{G%Nv_pVHkes!^Fxv
zNg<1>D#~&(TVz>YRh5v^YHghZumGU7a?Y8RvDgtp5JiO077ZBVyH_uF+MPxeoqKou
zV9TYHS)R`q^E6AP6r6Jp`C4l!WmQ#L8)J+GW3|;z8)KZa&N*$hb8h8)3HA{}9zv8c
zhB5cNX1Cqz_dDHA7=+~fRv#Crt5q*7%j)p(a5kTdx(4L(JTI%-^E`yH(OOds0G<W+
z-iYEbLI{CXa(ioM`^uG_on6M*1zGsySozyL&qt@@$#j~gStV;{9a!NUVCxV<4mkkO
z2&{7$W1o9fnl%~?0Kk~M{QB$LTicxT4~xQTnk0o-Ch2^>NQ!(qog$62)()HlhXAa#
z)@o}krWgP$b^u_}>JtDU&<JBf0&jDDeRq4O*Xwc4pO3Vpl=J1{!J|h<!$V9kqf|*5
z;)WE`I*SofS_>&KMnM=ztpH#(My%cofCB(2tmkp+Kx<6_xaWaKD4~Px%`3Z?nsJk}
z^PzJg>#NnESmgP5G#;LuEEWret=1YFZH?_Hhm2?{1b}vnnh4pE2wST|VvL?51)Mod
zofu7}>9;zx)BX^L`S-bu#^d9YlPt@Wkk(qPwAPv<fe9H%ObBV87DTE+jfr-^IiN~<
zMv>+HhGETkcC_DVHfg)_{HISV3RPJqi{*4Y8P3OfUewq!#(d-q(mJogC~{fZiW^WH
zXDupg#yJ>EFv19Ft+iIZ?-hs?MTYv#(ZPX~0wFXQ3@Cjbt8^jdyZ7&nr&Efdt_rO+
zr7Y)2TEQ@pQpUa?hC!WWSi0Ccf(bFkT8p$Llz=uy8ybY=JZZIBSy6eBZ>y@Rv?lnl
z%Ih?_di~a5FgO=!o&PUlt*z@iP1BR%@%VIH$Vv)fC^FVyZ2*9Aj;!TMQ>`1B1OOzh
zLr5!wFb;elV?+^K3@0X_v*QyUg`Ve~yPZ$SMF=sUOou1O^ZBAyQYocsVXfr|bI+sJ
zSfgC7!P&kdn0ka_U<m38oHLC0gtk1E&lV$8cGuSZF!Fu>Ia8>sI$116!{f<xD(bq#
z4gj>VwaAnPMycofj4|Z<V69ImrN|?!&WnIir}c8NjG{1%!dfat5evhmwa(bdWa4{%
ztJQiwy0XSv>!hf`NauiHLNIa|D5V$>CRizjF>XY$(W>cr>0(I%dkB2Sq>xyOcB?&2
z^Fa_+i=-0-TAMV_X*)8`I&fK*zH|2-t@ZZS_Onls&iNNH#^hNxo=y*sj!tJ|2iC`w
za^j~2rF6}CwyxXb1t+wF7^cjk9AQAr0U(C0v#yal2M%ZEQoEbZYW0`*_2u0UMOVg{
zGA|Ae_scBbn+%OH)Pb?4Ei}e>(=nUW+f{*$We5WR5bd0jAwmeD2GUYvrJPom=hRt9
z?=7JEGxw{P;5o|57?T&p@#*pY@F2@GE2Oi|s|*0_B1&kmg#juNmI77twj)YN{9f3@
zG0KcJlr!YD<aB9<^Ptl%Cuy*@b_|)ZFltl4zP&EtxvYUXXFI(?J)MV@^*9etXSS*b
z8b}zCfK%&?){smnXU<wFWuw&u<WwyH01<)!(&3}as0m0SYyk<$oagzYY~lOf9xlN1
z>$aYnhs`K%b=uFk);iBQi7_TwCWre6_wL`DPA7oC7+aKuQZfn~MoK-I2O9dVb{GUo
z$}CARK^SAe1PH-^5k>%j0Qg{i{OFOWM0)R0xk%2@$S3ELEtC6izcZgs4-XHO(o!iQ
zgmW$mLt>3Pozc8zm|%j9*0QQaT`Q%WwFo0E6+)OYYK$qfqAE+@_v5U{M$_j>VX<87
z9qxbYt#8j43n8VmPFrK0^Mjz(YI&IWr?X(1G?WG*2<HHfaqfAZ*XoE`P|m?YJ{ifn
z4qB}`%_!$8Nu*M}R^0P~;$T=N>GPrEd0u~gt-rSBb1tXT&_X+i+EFvo){dsM6gaQ^
zFk-&%07Y?Jmqj{XDyfV$)A3l=b+g?9Yk@LRmOvRe=KwG!QtEElpm{;HF;)HO-h;ZT
zo^jvdeD*Dzb4sb@a(Qs~{(LgYMG1f?1|daDAWqQrsdC14YvZg#&N}BLwGO~D9IRP*
z(pkemBTQn0fe_FLtVL^EAk}>zUi+Qje0C$Bmdk_t_p>aUO-Gz_Bx;Os*HK&7YbR;w
z`+Y?Lky0%YLLLwVfDp3AYA~xv5cfS!JpiyqTVt$82&H>AeDy#7`_F-eF=m-8_xAVR
ze(-Li(U3(hm5giS1;MtaxT?E{a|aNBA(RuwWl<1DMOl$3<Q^X{<{OtTIjIl#_jcE}
z0b^H5BLry;+<0UdCWKVrc;I>48}#cx#X8OBKuQX6dU&+_pa1t}5cZZ5oU2Gh8C60x
z;-;u8%9s|)8dDZkZ!iGLO;r-^sj^^=Rz4Z~K|q_G>7#f1yO&*^Rb>%wZR^pA#?)`M
z7uZc^bN0DU^zYml>~521nBSg*zgjJ*j*pM;-n%zGInhdKZOXEcLVDc8LNpNJlv0YF
zU2R7j03d{L#z9GNj&mA@Q4mI+=K;cCtW^rFtv!6}FTrWg!Th`Tp54wT=W=>{^3Hd^
zH=Rx=qY<T4DTx4koTIYvt&PJFApmOtY}klA-@}Bgq6C;AgfS+#(d}^VIisCX2xGz-
zSSPD`J!((??Cs~kB1_ZZ@$j8@-}OA67r79EG2Uvou~zPM?23{xPBGy@XpNRd!5cBA
zH0%!uC0UXT)&{4Chsk{2?sR~A`FsJ+l4jeT9D_Bi-$%_RCRkPV{PfiP@mm+oXa$5Y
z!5BM?F$QD68UUb_d4X@N7PUYKmqi)GaTo-E5NPeJ1?S45bk<VNh#!p~+#9^~vaQQJ
zT?Xxz9*sz&i4g)ytFrEQyVdb%ad4!RdKOckb$zw!h{NG<JeeHq?}LW`L8k`HgWb6a
z>Y`sDa5+c;m>>@j!9WpMN1Sn5IlqH2I_hYQaZ@M&*f0_tjO3z`j<4N*alF5`dF@92
zfBy-*bn_WaeG)EZS?wPjjFb7eNS4X0(TJi_S&ObOHC|4(!2?PFq)~Fg9ig}(Jflsm
zH2|o95JC!00l+$CoDP%$fQA7~2nL2RanmgP))3x&;X;W>DQC0!{kPv~1tBbRoo82q
zCadk%bmaK~tDKSrcL;eLcoPhr>vAoO-kP6Wd;JBo)eE~F!RWQk&G#FO0Lm<@i?ZH-
zWX7lJ?L$#j+q8jwpH({fy+885{1X?-Xa!?*UX}HFq-z<60Dwa1fOJAjYmAB7WUz69
z0N=R6uWlWuY5U5qANb&TTg?~%8UU-$>G5G*7Ut+MJ3ip|7Yt>wac(wlOB4Fx<m9wN
zh<@!Q2!b=4`b@gA))qxEJQ^MjkFqS|+$(fxouibSvh)k{-kvh$fB>8WtDLjeS%eTm
zfH83ntA<xVuB|aX^*z5hItasH>++TH(NWCp+LfzEr=y)qyU$|h6LQJ(e0Y2^o6V1=
zL!(W%(=CeJ8q2J0DXV?Yr4lLWxWkwtvf6wm7$YwTFv1oq8HJgX8YAUgv)i@@qc{ke
z25T*LfVn4x6(`dRUFNS|%F)pgrL4$vBE`nWMljD+o?8GZ1qZ}9N6LWG*=*Xpa%t`A
z27l?sXEz5Q&}<mTVI0H8dbwOq>VHA+KXl5~dEqy^7s=+|fiWgoE=d^HS!wKhy~|m{
zJr2$VAG^&y{pu$=t<^vGe{?(jtz7_s8+QPp{N^{SKl&r&H~@qZb!FM@b&d`W`5*rA
zt6%-OGn;yLU0G|(qMVE;!@a|7ndDd#f^R^Wq{%A^i%bTyd9<-r*^$Qs%xRV+jc{P<
znyg(~%x3dT>)qRzj!8Wlk5#U!s){UfMU5L?p_QACxknMUuF8Q(yAz+;&L`wjRn^JK
z>A}&%EX$Y?$e7k?z7Ymjj;5VPTrH-pX0woss6T+TLhW`q8u3o2bVU#}4*UJ?oof%A
zIv5^PPO3`0k_d5{UWu9}OQtR#4EktZyCQe~Hhb~Pg>unKEthkts#(3@K7U=saCg*A
zs<7RzU0oNovVagGJA(;w^h=)$uUvkr_PsbBeC3Px{<!V_#XUcaoxo)0=~_4&oO42o
zN-N_*J&0gFvrcDz)f#lO<FtL_YWUU9ouS#_GoP7nY#jaf|7*J$_bP22>ZhWI0{aK=
z*pI!5H+pBY_KfQ*=Ui3SlgZ?r@4Yjh&OF8eKx;i6jn{&RXhR4IyFHhuK@@4N>Z*zd
z18|Od-lQsezxesr|GlsEZePyxq87C^w$pBLp<OM&A;PF1_`c^A!=rMMY;SGRS8t!q
z&L`v|>UualK0ZFlvaA-hl8O=%_@0=IskM2UF^|`I-nenoNm0&cF3)(cFJ?2o-eX_-
z$rpe78_k<nDp3PM)9DNojMAbXHmdOm05&!^X7fdrWxzdz5P9`td!dwazL=L)2@W{p
z;^>4q=MVsZQrh#l*J^p|ef`-tuKWkTweyss0sult7{`Olmz;CeGBrYYuV45S9loEI
zS(jyA6qJ&n+4LI`Wo&#pVjd$e-#By2QL`DpbgLH98H)&7rT$<_)wkX`bDL-MS4ye*
zbb4@fm{&O_uGZydCOwQ_tvxoDlcQm8&<AiynzkDo*diNx$FpU=(Yy8&pSt$?i>pQq
z#evVUbM0LFd2zF(qE@~^fVeJXon@Wtm!wd4+w_eqJ7;G2SZJl^v-$Dl^x*U;aEdv%
zDF`m~rAZcg5k*qeHSs+}$n@U5PNy5adi&tufOY)AuY9HTsW;yHK?n*NjVRtEzL}((
zLQtbg>>FdV;nG7~Z*^x{r<?KGwapK<=NX}lPRb&a;)RTQi>j@0<as>6<zmt3w3U=c
zAHRJ2OF#3BPuaS0?fzCgU(B{Pdclp2bFut`VT>v3N?XH?^OZGP8e=elka*sD&uo3b
zyzhgcwJxgaw27CSK1>SV<Gr~8j9fs)!GQ;7TsNM@U#;?b9^St{IvIIhfHAh#GK6p`
zbtObyb$Y!lO|_`o+q<GH(&f@x7`%Gt&To99^?gwRfbV--TifgF>!IgW^MwPalm<e9
zki~r7xpd|9<kUHL=gZI92|k!hmZgV#`?LAnI;V~Ce2*i9i-HlVv;hD=^bsaBZb%_8
zW8=|y>tnB7{RhAPzNi2o2!i$Xb%IGVh(e&+nB^j2jo3LSm25N{LMZ^a`Q@LyI4&zV
z>sEQ(p65eRg<${y*bj`eS(5s%zwqp$S|Q1jwVT(vU;d2uTsGOYG0AdajAh)jMwdn5
ztTW1ljp%4N47bjxZLRfWJT8j@`<^uh5d!W#f2B{xXSVjK{nct&e6YX2Oi}_`6ltqg
zA!BbWb&=&t*C7vaZ5xb{X3s2_M|FAQg%`&!U%K+iH~;#xDP0@%DM7*AxqujDrlBZ!
z9#mDQ**D8fZLj6Gu6FuooZ~(Tm#QcZ_xBHv4p`{7!*Fx1$+WsMuDrkpC6rQ#Qc65q
zXf>H`-G0T0+WY#aTCczMH@R$WZer{B@BLABG-`Jn5C+A`sdKI$Y@?E@25Y=>!|$~&
zh|6CAA%Y3PbZcA(1%c3MS<jNBx3*s1+6lh)h4XsOoBzdsd|pjG39L0)o-3mQ?QmLS
zPP|scOM<ZPURXa{o5q+l%VtTM)s^pg&H)C1RSGy#fc2TJeQJNDl$y<FdwY9~5a*my
zk|AWP%34b>Zf$NUsq!Qv{k~0-swi&Uyjj0;<8O)z05HZu5Do_GF3GvaL0jM+fU`of
zxLM}~4*cCOo@Fumq+Fyt86MB4Q>~O%N(kX&Vv__TWQ{?9C?)`4ltdn<L8D9-_H!TG
z`WvEJaq)a_=klfH!O5k|m(uYFj7Av6m#%<yt}LC;Hoy4Ei{tWxmT-@A#yn*Z!mHzb
zy`4*CnzB3BpMwDx1g*8P)&W?p2`1e0jFNR#<>S%d#?>=4xC*P>yLZ1R3S%tB6l1W~
z0m1-ajls{5%Jg*pYBd8qc(`9{C7f=Qb%*j-#<`XP)xrzH>HMhCXf)$3n=KEj{MyU6
z_22$nckm+}4#wE`{e7$&h%nC);wo%q!}CLS%Fv}RpWER2L%F1BI!~4-C&PZPSL~m}
zzJKi`L6#t7K`G}DLVy(paFcAhe(PrASAN>>bv}R#=J@K>o8{4Q>~-zDB2hyo=`1PM
zcXr#qaN#q+6(}YRFZ3}}$Y|;aSjtH{8F5DA>(683?jqovQ&LJLthKSQu^|TPBClIs
ztdUl0PrcW_64j&q{nO>#fayvS6?z~ja~3@8j8R(f?X5E#{1pDmIaig{!QNq(=9u8R
zt|{fVDxFpsVbp3vRr!H0q>x!QpU<wn_A>g)XX76&s`s(^l`jN-fC({13sHwbxJ;I}
ze(UeFHqYqapNva69*>4chjAPqpPu%5J*G4mix@@R2ZW7QN-F0pP`ZBmHvRffet@Xn
z2mSnOej^5FjSyZGXUk;c&K>+SA6whH*zJn<!8mKR+MILj+$u?pa?kfYKa8f6u~O>)
zixrNw)@4;yWhq3Bt<_3d=O|-UUa)?z(Rn&u8Dp{}eYC%SaBu_&ilV>>sY)P(oO3`h
z02pn^`r4TdehPo3wHZ$)v*n_cbqv;2@@3`u5VFM*g@F`P-f9Q48pVx0vAFcoZS|E;
zHTzGrOVhk_?UbYAKmB&SSd^t+yS~<b;kI}4${9I55f?q3%@4*W6}D=bb%J1bZypmy
z0+!BaajT2M$jx#d1U1Gy2<<QZWH@*tF8t1wghKYm-(DOI%N8_lz3RR2(%SWF7sutV
zfic$YcH8ZCmcBjM*q96-b^C;KPYLnp?%R2m_18B7KVY1F=!ch|9IJB@1R(~{Q^-S1
z*~0TYQ2;0##7`RDT%EP8^8Dbz-obP%gske)C%8k1na()!ZCaWRtHCY%aq!C7qdQNk
zE9YEM<au5cMPaQqR<{~a`tC`F@OIn)jIk!*41fh8uidy#e*UMtCll2Q8`rLGTszAw
z`3!K*m1Q{|kMgoG)<m3PXNgb{kUGs;S8uqY96x%{Z*SvP%Pkkt-}`y`1YY-j3~pQ*
z+_-XK=064)<0y*O*4EnVYm<|cUVmVnQBoUYj`#P{G+So*pg)M>I0{3?D8~55Utq=<
zA%rmoW9(`P{Jry0jB~R6L1`lIV|9$bDvSASI@%x3r!(bq>#TE*aaKJ#zP`Py-+D+p
zU7*VMK_JxFJ$shVlj=$-HCrqui`gQbG0IwLy#{34P;d@>&{?WHTx3~m<J#!qgXT|u
zwZHL1Y(Eret!MN3GEJP2RNGA&ipiv7Fk}+9o7M4=<-E1MNt&I-bVh&mi}b8P_YVt{
z(*9r#Jip&r1E{B`lR>vfdGN{{q^|e%{`};aw3?gi8|_Xf@I2r18U3*qn6ai7HB~0W
zlu1Pi)yO%pUZ=17kJ(~f9Z(QL%;$?yIzAbnE|)2y(5)><d0i7O^h<I7<ei6&R?kl7
zLnbfZyuI=B&$P+(M7na$Ng?Ko#nI6bp#)<TN+XN{jDb>DW|UFh@0ubn#^arDe3joi
zd+__B;hd}MdNLW8Rb{jyl<0XH1_3|~07a58&Up}OrA~*(ozH!|^YWb!+vKMP#yAQ?
z@$yyk`+sh@Uf<l*MFk)c=i+1}8|zh7DIt<H3tCaY{q^3U*=U4*;CUYV@hf=foYP8I
zWo5Ls#;jIB$RX!!o~4Y_#*g^o_p5-3`D}hL9F}EylneoY5pJxl>pBPmCN#>k{o(BL
zojdZadxWveSFcyCU~Tj4n-=hdx-!OOdA?XK9GFe)S9$&d_D_dLz&Kz}H+(|ra*;|Y
zk~dyj`>8h`-{P~u7^9SO4mf8|d#=+MvrN(?No6fMh4m=8yi|-*Z~`$F+6XM#Pv+}`
zwb_l0YhV8S1yT4xdxVq-fOCG1^?(Q={!hMsG8|Ta{`UITCY2TdHVV70Xnt=Ptn~{-
z^7&~|v`UQA(UegZgh8*@Yj-+f5cs}N2>Gyznf*0bYlRfjD55Qqnt;RBV&uSD?R%}u
z+begZwRR2=MhIbyR&GkBX*NC`<wc$sRgxr8gKJ~_S|fz7&<HJxn{z2@fjmzg9I$KK
zAmzgVzx>Nzeny+0NLNy-(Rf@{C1+gJH3z4vG7JO26zUq7Z;dI+axmy`{oM2Dh-sQ0
zA0Gq2*49?D*?g+*D5adu=T%i&V;JW$%NeCgYilh50LoOFtX&%9MgHoSf90Y$d>_V`
zBuz%sDM-D(vEFJw>+$a&1STZ-*o&P+RLcdfN{q3!hy_8v7s^^_kQ7i_3&LnE1f@hM
zxmYYQC7pJ=-EPN?Ml+6sFk~NgEzZ?hBx@Z4AQ&NxoU?$Db8fjzHU{0}ljG@ZmL`SP
z3Ngkw4G?w?N>!CbS)}=@wy5X(7HVq^=N`obA1*ni00?31H#*7bajV-)tMb<0`Q>M~
z`SH54)@D_4v>eN-v|9F;VmF9Pbs9&3l63{ODWToztX~`0JJ<Z?kMhaq17l2<We10c
z!{N|qUDx9F?OUr_Vow5VZC;jXS?F3)gx3nga~ze@X;a1+BOTO~cl-B7r!Rc{Yu*z#
zduA?ao*f^b+<*8WkYI)KeLsw%b9_Yig&PNd`3_B&&A0^!>tZ2ferd>nkef|Ui+X5T
zxYnF;m!{baSDZ1?@)yg+{^)ox7<iO5<9MStXti3an%EzTx~wr}RT^u22h;)r1kPbV
z&KhrRkRF}fJAANM%%yMyW9d^uNXH^$Ol5@81~clMvr=G8u1p;TGME%iOgN!BsY)k2
zpG#5(+wDJd^H2WHZ}9jWUcuvarL|7-OgpQz0tX%D9Uq@wMNU?=l6A8?sFenQVUoP`
zxsN})K_38P%wn-PIXT(i+gDm4gyxI6Hs;A&&{}7Cu~;TX8}4(Q7ry7Yq(T^R#-)^?
zlq!pxAAdFe_zy{{I7b*`7Rhq|=;+~t2hO<&*m5>g@-m#KtN;Km{p#1!|MOqf|JNV+
zy)Io%H4V6`R)y9|MRBJ(o!Jm~LcjJ15L_Cwm@izwPsbx<UAx^L4Nuyw*5#euC~kyd
zNZIq=pf}daRn1xlPW@GGkpqa<dddD#R%BXh%6x)wWsNeXs5Jm+qyZs}iIkF2#yu~Z
zmI08S7t~cH>Iwk@3|bj&On7PM^S|{==VkNbb)}TblLYd*RVt71%|+U`7^nm=Q51_L
z;dr1<m;L|r+-&-xFvd(~)8oD2@zJpm8Nr0Ay47jYr-^cErI+)Wt?QsFd!DyjmH;L4
zG$5pvr5H&FcxP?)+O5Y2fS&=zn8k8=c=z7u@JLVv;Brwny%xEcfNk^3pBr|g<nBZ7
zevY(idy&iq7mRrxF^3VhYYWRueglVy;8R)CR(wz!i%=Fam@ei@g4WCVVr_kGeQTrD
zY)4`EJl9>u*s`jX(u2x2N|#{87$<^r?$3K65UZ$3WPYu|7t6&fI4-J^aAvJ3kwJh0
zg($$~lgihH%qKx3Ooj9~;{pSK1D{dv<R@;v@`dv}^8A=zXLXotv6xp?#TheYi72Ct
z^DyLbGfxt)-36ui&1>t|E}f%Uf6WFd<vdwFJUlo(KCQ}<5^A0E!?0W|@BN!^K1Dl1
zBuQedVU*^}#e21@5x}Y%Dgf{H@*=-{>!OZ(t<u{U%f-8Q-#yyfx5hYd&VlE9R>{%7
z{q_Ytcl#4>T>s5q?*H89cr%vC!t3<_0JJ8QS!=wwS=Ti<qmpE;-A2oV7A4IKBLp&r
zP*#@3a=AP@8vf;X{_<#em}R+nnrXr_VwHtsi~+5kvjFJ(ugeCF2J$=r=ytncEk_8(
zi@w1@En*7|AQ2#(7bq*aPyi73y6IxsYPYId0HrF=Nw*dL+RvV^sNk{w${15sl`2ao
zWP|ux<VI@e0829-9k1Vcr99l1yxD*AqVfr>wN+g$ljQFA-p=!^bjAb_ElNzt?U@eC
znzG{LzxyYf|KJ;sZORx^6onEJjE$+M<{4hq{lmmL?yYIb0010(Nkl<Z#}r_KMq+yD
zmoDm9%Bn=>a=ttm4Nt}s;W-D;D3u=sZ@!%buEcS0^xz-u{3pM5K?eArdBgwA8z=wv
zPe%9dckULAxM{Olz~a$-zH#&R_`w6;_rrPCi#pxX0f3nn2%*!!8IR4H+?2I!hKb$N
zWqs-D)wtPw4qj_jtXC_Ati!dW7Pv*!BBUK5XmIHg@_F!xyV-K~zx^Mdef@PI3J+7E
zN`x>I1Q2o<S*tSyRvEoeJ_;eH9nzZ)yvA2P(f#!67h(2ebfvY<^FnEjFtSzcQL0o$
z8AWlkSy#!!d(Hm3C^Wfq<syu;))r-Pe0p+lbeJbgXKXG?Yu$Dffe-^sr?o6+i?}SG
zq&2N|p5;bstuzQl2m$Qi2Zlcch%*M4FXWV-wKgyE<!pKPdv{0q-19tRt+h64G>YXi
zq@kY8PG+N>mtVe!r~e=}KlO>tPkq8XIIRBsyY{UqW^`?B{q)|u8@FDVJh&G(oAcpO
zJM1E#TWgu`wIeRI&dLh;zEC<@EKZV?F^(>i=WyYRwN@!@jBx-2;nk~JJ=8f%I14*1
z=HLF#|L~7r{=_Hh<<e_)W=DrXGX!L)vH)NhK?n#X1Sm7WTb*(|k3aeH))zm{TNhQ|
z;qm?|&r%|F2)LP7z5zrq^}J+nuerVJvARyP{F5({_Bq?PD$<`N>EYhqd_G-H#tU0D
z<K`9TS5>unG^5ta)3Vd)^sjCu$HOOS(^{KXCDsNh*~%2t6s!d23<igQ2?P#TAKPV}
z3*nOI#limm>2Q)|ne$1i>h)34YPYZK&!Lv8OjtAAjy8(v;q$pbg<RilT;FZ{-LD+I
z`!>Blbd(JL<U51)wRAk`UcYs4dM~08Y%phav#51dzEBZB#kj?29L+YmV((zl2m{Y!
z7krb$YUb>WC4fTZxUoK9tei7Iv}>&O0KjiVxZ$hAnJ5Y<#{Em%rd}9pkTcF$jVZ#2
zIfe+n-(~IF8(aVI>->V3h>y{gwYC-lIq*CW>QrMu2&^%2+*<BG>fCxcfA`(p*Iqea
zn^q5<&*#UF_VP4MvLzv8V{L1AJnZo%HM$c;WmS=O%Vnu>GJgGuQX<|T`3Rv%RTwic
zs~gAgd$*Y&)IRT%6sy=vmSuO}y_+OiI?toHk=sH`)$8}<bm{6kib9467z@_m>ld|5
z{ZZJt{Q}&60RY6Ge-Fk9`HQ#IWVF4#GhfUo;9A$1kXF0xKshP7$6YO8kt}Bmw>cOL
z)-GCC)><VcS|#QqY{4Rg5JnCh^8f15+G9rR)}>3uqqpsBCaH1=QG*dim|*9eu@<ax
zYyG?|hx>aQ+uKq6bdk@e!sB&ilxEt3HEYT#DUpw>s)*YCmc!ZcF$=@+@_81YFvgTc
zd3=04o=i>_V{1&z$ZWCfFVZ)A{k88NtZi)Q@yr%E-rY@Yz8nNMp6tA!Hl|V%g!GIZ
z2t<gr)_`*wAs`4WfW>Ps-a2P4WqEo!+TTA|X1Oui#w3IC@>F<~ygW{wG>NJ)r!vG@
zk|*u(*3W+S91VCXcyGK20Nz(VbF}}E9giRWt8czZA}f`u4Q)j(aFUp=r7q7K8X<tA
z@7zb%FFaw*T36Qs8AmNN3}_A1A^_wZ;I+R>?Ed86{l?z!|I59-{r=9zqN+oK5XL@W
zM6ibtIB>NE2kmJJjDPp;_m)L^?dF|UBfg-Z+~ak%x;$4`CFk5~!-7C7rN(0#_|}=u
zOXpT%^1b`Ait^$8N3+FzI-RnZ*FxC3?)N(JqHsytYc;0Bqc-syy+OX1HCt`}3!h=l
zCz`ogYge^+k+Ya!0C4WTYy4KbOZWo|=shhLrBs?`j~+an&KG5wTfkatjjRLb117mB
z=S#|1+>S+EJ0Nj0_CNIrzx6z7MSL)JE?<Mo*VaGt>Fj%Vdq=0@55HxkP*QM@yRw2X
z)S_k{hde+2!#}zH)t|j!b5^IwoRW%As-y&rz^yK|`gh_ISHAd#D*ynX(=7kz|5Dt0
z;EZv`7~}weDgXdtnK(GDuQnF*>9@c8*0oEQH`g{e=jT2re4=~{rmY9W&1WHJ<#Haj
zTU9<|L4YZ%d(E>K1^qspbK2;tsHT(2$-yvNrim&X7*K*}eTjx{l-$0%2-nx$V%gCM
zTC^_HVV5^=ZQpq5N%zDNLfToZr00<DP`OOJmanVYT8Alx$b+qO$XkE!zO15{jK{;1
zlO#=4C9JjBC}!;q>6p;#<BW0^wi`h3-soU+dutgnyt94v<(Hn_fOACr;w^abR`he9
zKX~iA<+t9M9i6-o0zmC+1G>6Qxc&B{<kjQWGn}h^EQ~QqN{$FKPA&7O*>uIy7%Qv}
z?)>mwXxfb1zxcJ2@4kzE?@!C=B5L}}(P?{oTTdpT7TquyOw8%<c(|1>q*caj4AvUq
zIp0$Icz-3O<i=oQc;JH{V5zeCB3|E6vuRCP@5-}QqFvR*%<}yBU^tmfL|Q1NSj4P#
z&RWix881W6^F!aJ39=Rdq!eBdT>jLXz0W*FGQY7#N)h^jD=S^qgc2~CdfZU5TrAdu
z9)IQut+iIv^<p+39v?3k%Uo1a2rVQ4gxvEu*TdsMuP@ZBv$mGrz0VlCcIS>La{ua;
z?VtVWXMFkv!CKqd-G)jkaOh+xs+vbFaIUIr%Gl&PZ(pdcoOMd7x~do@m=G;0iU|!O
zqLo$h1HQQV(rq&sr2o@Dt0t3Hujlf_fuodYDTUS7I-Ox%@85sGv99WBdu_AXY_jv4
z6Fx>)gb>g3JW6q0gOb3x@B3g)Ih_Q}n1!K<!!wu6eQ)AdRkO)>cr-jcKBk0lk0A_X
zw&(@H?m{<|?N%HL=K`v0fqX)vcoISL&aJJd5fuQeZgg*Ye!fU3CzJ-VTv%fq2`Hwx
z6RU>zEL@b7qN*P~dYCSk(>jwvfYwUuwzMAOgGq(6>YC>6;RqT*vcDIv4ZynnblH9R
z#qFQH@O<U7!CG4u#bh>nXLyu3AvzxR^lgomQb-wV?Bt#MH^2PFi!{L@>lk2!DEB<8
zlm?4btvKb6I~PaRx?BI*Z{PoeKS&P_``>!d?e!Vslau4Y+E(&#|6^A+d8_-)?>!Xr
z>NF|!)@Eza_dM@h9M@wy62gdck5Ec6F~*Q0hbHs=0Idxc5AJKuyq&W|dsZnFMNv%V
zv(w3Fc{-`8dKH<FQW|3$L8!A7%@t`jv-!dr8%7=`#ERNgW%G^Kw!Zk(9g$$-b5H96
zjLnmz(Hn5z2Ly_eMgt!n9NfNo7T@cfQ&QD+H9H+0?;lBIX49$UNNJ6fW|X$#hDZ}$
zXu_DAFHjV@71c)5&S%XR@2p?^A%HVr6&@&xVthK@-`^K?%^A;2yXO1(aYhIU;-)j2
zUU2-%IWX2(XE6dW76&15csZSV{Vw*O=(fQdpa1;fx4#|U8zD-U)5&0GM=r9s*~~4B
z_Ya#L-&W<`z54{AT!^*xwJ3_7eV_0#y21!!N|aJQMkM08tf5o@FrVEUtgkPdPrb-~
zufRZ6mu0cHw^vV=Wmzf@s7IJ_O3UpMTrF;nDvwd5Ev|Ld<czZ#3`V%Rw#lyF*!t|J
z&p?%69Qd9r%9igp`ouY{lvowOX|1h!ji=6dT~#)&s%p7d+`oGl6C`9UWsM06iaG@7
zI6{<J9}XkmYYG5Up-a8RblS@OxV3Shv(3Ntx$TdibClv-@ZRLSEFSIcosLI~<<bv)
zLWs9u4d%--YqlC?SvhOz?nTx1dCwAKWLa|{AZ3+igjTrCXe;tBtqW_O@Ve67*Iu*h
zJMaFt|7f?dW2a?db??d*djGy#%s;;Msl|8SHM^bBx8EXHw~J)4edTH#$Io~l`D1iN
z8S8dBGjdRs6(Wps;}{@-_08UNG)lqktg3-M0p6SSm1UM?kM{P`Bvnc_3`XF<L0RSj
zA)GTjNdim*-$RVMx^%LZN_s(1ESKG_E%e44{g+;R{KI}22qm3%XM~ZnPKvrJ%P{hI
zyJym*N|Kx)YwZ)x0p8~_8&4PWxfFGtX0@!1F^pk^FvFNJhNme(;B%(5a>^l$s<IGL
zqRu8%+1l5Bs{P83-C^X_PG(USN5kQ0I-SfWjC+3I1AwUN7?Y~5JDx}5I5|FUM2+@E
zF0SCceIT18Nt1d(+yI1d!<X~<Ti^bR&;0r?K6yKMqj&Q^`i=Cz`Ns@Uv(;SOe-QcJ
z?u}de_{d|7i@MYAy)zycSzc<hHdtF*Uys7z%yYuWt*<a{#Bmjc`DpA6SmX$yz}RfL
z80_w3#>%4PQTPGUT6ICCl$3HZnGA<Rsg#nc7PWJ3inIcCt@gC_cczsms;1G_DF+5X
zg3(qh=VKfPd%|x0%HQ^TXWz0Tm^7j&i$a@~G~^sRqovB1$RcZuzq7vL@&P~Ps?RB<
zs<J#8jjFoJl59Gi<yJ&tD2+puv?@a>-aV;kk!&Fjn1Cs@Qb~vyC6Vy`Xzfor)u;dA
zul3sPGjln2thO<gQdyqoMYeyqpDYt6B{#-frbO#D^BJJ66Q!gKytb=xx3`AEFuHod
zfoul|pcyq9m|C6GMPYf(8bN>k%G0()2c7PJ^&fuofBC0(-kCa}gwXQz6!kpf(Hpa>
z9^U`>&Q=ki2XY_sA}#Zst?g!`aW=>GIDbW{@B3l1>Eakfu4@HI6L3)@D)YR?c>JAr
zS})x2JdZNADm>zxv(_48j4@i9qR6YNn$2cunifSNl`4xO2t%U{CfETu=aAMe$$K=c
zq)}2apIf1<tQn!0dtt~H*!|3Z`CmURx4+3H2!ckdDVKRZ8@IYWU8bU{qh_bh%QQ_#
z-~C?e_N^cc==Y1L{2&*lbzPQumY<GJrBaJzDWuTCG~&3bwYAPU8%426X1;@_2IPCL
z&;SMzpo}r<001x|yMv$p?XN#0mmdXVOjTCaTFU6^VE<$Lgtb;GrL|6$%Vm;G=9Bq+
zZnU<}y1E90JcpWbObQ{=9AnIVA9=nx8j|g9-SpP4K95?TWmy8En3AyF&Y@n+7ZpC^
zNwv@Z(l0Ll$?vrfr}@#N#?J0?c|;hEqnLT#q!zpunn#n-2#iYe{Oa~jx7!V$KI!Rk
z>nn^KQB-z&2XDRYQp)44Opdt4R@Fh+7%)`6d+!k!y;gS+G&%Pi0_&U*LWsJks-h^8
zBCG3KDy5W0&H-T0_lz;bfz$Sl!`Y>%F+Y6R8w>yyo|QJBMk|f6--s6oEf&kpFZ|5T
z>*uftSrq`Iz0PPj+yMeoi^(KxwriCttnQ%Z<S)K^-$A$E4I2&4IUums2`NM^%A&~g
zJTHp6u4^Tw)EZfEuI+J=CEI{aUF~R$$J6d;*=WW{)IgcRk}xJ}<u@9UvQC+If8otH
z-~1sZwa*3C+BD1WKX_DCWwY5_>vbB9m~%!cMHs(dZfWJ3j4{?&t(6e<a+xfai$#`}
zWm!vQt!2t0=hjIi>UztvXraBsX|vTrr~+eCw1Dk!ezW)TFaN>?ad{sOS&da-#HI&U
z3xpsf9tJaihAVtK-}utWyAQ}eNsYBF$pLw~PHC%2l4Ub)Tp#7E6(5T!jF%7RS+Tvj
zdF^tu5kHl*9>1~7IS+fCOE+)A-glgJyzQ+@S1VBqOX|`1c*{E;pR6m7ajp<J0Bwvm
zMp>h@lGdyWtQupj)&R&O%ny9)WL?#}ELe`Fy{(;WGL}`@zj{^9CX6wIP?Z<n#@fwa
z`CI;_bDP0)#=0BpqkH$c?`6rd(Q2bc6HJaV7FlZQ>ZISEjK+>)LI|P+A!LlvMq8z|
zHp*xO$YQ_*5g-6A&r;4g(gxQxDGjBJQYxiHWyPBWV+<I3VVKV6e$;IK`@i(&<qNEO
zE2SKdC*#RPNU=zg>G5&YXtd%+6gPq>LKsoTR=15vrOK?xv#b_HRaTXhN@-z@T_s9d
zu+{`YkWLm2&kG8XCh4v9jdHQ@iURpT7)8?B>p$~6wlP;#VB$EQpN=-PWpT@ajap5F
zKPY(hSZv<DZvI|0`XB#Uq%{eA8pq{y+}qtXQZS5BDK<KT(=-uze)#Z_^yrPtSEBGK
zGNK=^D?&)z@9lNNsjK?!hUbh?21lNfax>S)*yTy801jz_2?4~wfnD8nvZ}%F03(F+
z3S-=}lye@PX1JX7Bxal`CtA(cax`xD`%byV-Tl_a7Kr+!MS3s1y!O>|4wVu@;!eA@
zwnq1pxVvVjv*n|Aqnq17cl~61+K3xGFNp7_s<zf@4`YnAwc40~Fa!_+LV&Pz&e~PZ
z?x#d(jx%3s>T|0VfV5FytoASO7HLudio@`&dU@qbpWghr3lIAmWAbGRm9$byC6`!x
zdBT^BF-jc*1c4W5rE9G)Ar>3}RCTGea@sm)eGLRbM*v{dk&dR>ZVhgpf~+?~zg`ZN
ztS275bg6&0SiJuCe~ovZ$NCUq0v;C>)!3nik358G1CEfQe#Vx`&PL~VeqH|afAh|t
z{`nmU!}VQpGNqhf+S)LAKK`?}Uy2*wu?G1;a(t{dHsNV>MSk~pfA@p`10h5Ri5iV~
zZ=6o&K@?h1Yl<CMiWp|>h-pHph&%_5QV(P75C8%M00=Pz2;l~0Myj9?s4Ukl3%KW5
zFveITF`;1`XNv{L#P@wu3lavcU;c9Eljncn3?YOmZFJg;-~ZPE^N{CxltD;zS!G)E
z`u&HZrj#O&IcF`#2w_YKA(&xmtu?ihQc;Ag)Et0YxK?2TV5*W5s+7cp0K;`9l$6f7
zxZ9m&*{$FF#@ZVfTP<USvMi4ehiQ@-s|lf)Qvf8jlu8(-q?GA0De_EcRf$?kRSRL9
zHr9bL&N&8%F&-o0q__@VKAsMOsK`=e4et+hl`-x$x`WDRSAXLhynXQ(4gtV>JN{*H
zI&M(1AyBqhu5O>C>mAf<d(UPD<J`Y}BSP5SKLjaRyN#?~E*E*4ZCtsAtWM)l*Mj!j
zn>#z6_e9~c6`m-;0At*2wewrMT30&B2%$xpGs;~J;M^xl0xp}knP~*~T*`dUbAT}k
zgu>t&QX?Uol>`6^764&fBS3(FVgTNdMKO7Jys^7IG^*2Wp1iXA+D~2JH6j?}uo?3&
ze(vu7^RKT+gs)vk^RZgYcedAyljRps6Nbq+v^Mf5Au<>h*nk66OG?Rx;Q?W5GY1x2
zQh}5y%Duoh9B|w7atD;zMct6N3Q5%N9NgM^;Y+`Haa{f;YLEr0ZwlrX*`|g9Y|-Hk
z<D5u@(8LKw2r#0pX4)C8BX9_zh6c`gxXg*nnpwp;b9TP#^V8#po0qS+Qcu74y>4#<
z%>Ucoz4gX%)d2wipWB?d?bW+kdpCAAaRQ-E3Zw~wHc{^i4-_g89{~6Wd<eb(67M`9
zffNa(wje;XL5QLNksudxBe9dzjuWqEFS9dq=G=JL&{T*D+g&@Tn%~2`0h|V7`R&de
zY5qFWod3lvpJ*RD``TWvh*H{UG)1*SBS1KD9OmR!j)&N4Ka*3rMdhUzqI&J>TfaJ<
zE19Y4-2VL9b(@UI=H@ezL=F@Em2uyezPXKs|L^^30%J@Xwb8{ZVX30x7<rU2iZcs&
z4hSRPr2)0$)Pw=I*3C4NG7yFvThZV##t_07VGGt8tBuvh8f}dk4^zKfDlHsLA3h9^
zA1<Fg-FbG`b5AkGx)Qi64@ea2UJrW?bsR%cnxtG|pwtPtk5Gm*mC|6ttf@MjBOJs^
ziOh+GmXa6*gg7B(egFiN5Xokr5v-#TuGdJX;ePE*>%^`X|9+G)M&!8=!h1Aw4K%qA
znH<IgqI6CSV^lFh9VV>R+BleE01nnF#YTx4j%q@4gndE~LRu-`4+zF-96M5bfsY7t
zzkPQ0E2s8yMF4;?MhG*?)~;Qx_EJ9xz#>8!3b?6+d`bgyy;ie+sEe-MA&jRtuDjiX
z$^B(kt{`3lA*{~>iT?J%o%YeKPEy`RLL94C?5ii2fA-5xr{l%~q{&OURzwlQNahZR
zV{K)93}C@puy#^~F-hoGMqo@0!^EK)u%3Jc(5l>C$g}J+^gj2!H|kq{=ueI}>Nl6>
z7lZPndv|N81x)h0{c^LBi3n)~BTNWcU=#rCz*wu1PA$M>`n6&RBQ)?aMwV-gbk4>W
zti>};+G_o|x%T<jUYt4n6ehTF&JVOZ9w2meohE|lq((^DbqS$lBZoW!KLnj;kZY}-
z%(5VWd;kc6gAhUh$|#9)1;%oP&1lnWcSakJ?zd-Pwt4i;ukQVd001Er_&)C*NID1T
z=Ca5I%MBvbr9LBDOt;vQpfd;G|IW2@zvy@vm&?(r_DdY38p)yAOqo^G+eYQa_7Y9I
z&z%_jb2YhhuOu}{sk!B@O9{?J2GRgyR%wJ0AOMUI{JX}<TE}(4f(EUWGGH*qoN|P{
zex9|y{_>fnr|=dYd-=s%Km2if-@fGf3a`)bW&=d-H<}=_Or(?cx0DhCSW81G1+W?b
zz*-CdlOah6BczQ%i~#_ymeld?D<|LB`?X9ELQ0j2<2W1Ug5$c7W*CFCHjT#>#M3zO
z2nB03HjFXIBxnPaf;B47jnRMsVazZAf<YS;mZJ3sjpt8Rp8tZkC!Z)j&g9U{`jgih
zp&+#5INf+uw$pE$b@W@`NZ<eKopWy==yv0%$B|d6R}phJ*4Jfg8+#+#!IZJ{KmEzc
zd-*Whq%L(hk3~W$4K%lL4Bq4v^2ya=3;;rDzg$kGNRy=6YCU-HAlyG!JM$&?nZ4Vi
z{;R+L{2lkoy?V83H%CgT080SSlY<KM;RR4CNmzn3vr?kK2O)Bq^IA;^pp4yU^TVgV
zTwUxwtyO%KwRZgnKPr{WgWJoM`GvLB+jH$X+fM)<Z@ALI0Kg;y0OdRBI4PH^IvEkq
z&-;;?tqLJF+BJV}cJ9=vp!D?XEGMGX)!w;x#^ZQ@qt)w0q*-75?rt@%8uWS(e{-G;
z$KiT}5RyXsVJP2xsrlKh)<^Fot^i>0*1Pl{%Nj`~C2~j{kHfxTmFnZN%*Q7C6#@Vp
zmZSAGw_0Y7lP}JfzqrH~Iy<=UZW#Xl;{A93)OJIqBsVDlm`O*$NfWe5nuNt^)aQ)j
zFaWK9GDL8)*(<;Dl0DLK7e8qc+H?)yesB1fOZ8?mx_Ps9^w8?9o3p`_Mx)BX06--M
z00_O&uwUYyND{YHk?|0pT0-6V(z82E!rWE8OIL!mN0-n4aqjri%&Vu}?F=Q`fz8X;
z{5u~cS8t%O9KQYn`%HK1zujS60l;`Luvczf{q=iCc!ifrX%unSwIZJUEdVG?0AS1r
z07M>PbFKPwOZ&Rpc(^^2#p9I=7kjr?j%;KAfUSim5d{E%xRa$RY1c=i;SCWTKXW=Q
zhlh^+Kc=Ej-2L4j@NX~Fn$3Q1z1nQ3@fZLkK>#or`tD{J0B$qta1KXWUbBW<wS_}_
zH&^o0ZRc_Iq2Bu!!;u)R+^2&vNuR_O9NLdKH|-{$pX2kpoAo>`*`*uu?waY3)WaSg
z3jn~AbGU<Z!W#eR!MqyQb~Jr`@9F+9xqQp%MZ^26Y%l}>g%JRN=hJ;NE_U#2Lod#I
z`*%J?xA1YM5Lbml;Xge6sbPh}r%w@Ag~Id|aaAZxUlCV@!t@n!RVYkf5m$x6^c8Vc
cC`{kK05NwFlI87Lt^fc407*qoM6N<$g1PyvbN~PV

diff --git a/inference_ccsr.py b/inference_ccsr.py
new file mode 100644
index 0000000..2d61fa0
--- /dev/null
+++ b/inference_ccsr.py
@@ -0,0 +1,239 @@
+from typing import List, Tuple, Optional
+import os
+import math
+from argparse import ArgumentParser, Namespace
+
+import numpy as np
+import torch
+import einops
+from torch.nn import functional as F
+import pytorch_lightning as pl
+from PIL import Image
+from omegaconf import OmegaConf
+
+from ldm.xformers_state import disable_xformers
+from model.q_sampler import SpacedSampler
+from model.ccsr_stage1 import ControlLDM
+from model.cond_fn import MSEGuidance
+from utils.image import auto_resize, pad
+from utils.common import instantiate_from_config, load_state_dict
+from utils.file import list_image_files, get_file_name_parts
+
+
+@torch.no_grad()
+def process(
+    model: ControlLDM,
+    control_imgs: List[np.ndarray],
+    steps: int,
+    t_max: float,
+    t_min: float,
+    strength: float,
+    color_fix_type: str,
+    disable_preprocess_model: bool,
+    cond_fn: Optional[MSEGuidance],
+    tiled: bool,
+    tile_size: int,
+    tile_stride: int
+) -> Tuple[List[np.ndarray], List[np.ndarray]]:
+    """
+    Apply DiffBIR model on a list of low-quality images.
+    
+    Args:
+        model (ControlLDM): Model.
+        control_imgs (List[np.ndarray]): A list of low-quality images (HWC, RGB, range in [0, 255]).
+        steps (int): Sampling steps.
+        strength (float): Control strength. Set to 1.0 during training.
+        color_fix_type (str): Type of color correction for samples.
+        disable_preprocess_model (bool): If specified, preprocess model (SwinIR) will not be used.
+        cond_fn (Guidance | None): Guidance function that returns gradient to guide the predicted x_0.
+        tiled (bool): If specified, a patch-based sampling strategy will be used for sampling.
+        tile_size (int): Size of patch.
+        tile_stride (int): Stride of sliding patch.
+    
+    Returns:
+        preds (List[np.ndarray]): Restoration results (HWC, RGB, range in [0, 255]).
+        stage1_preds (List[np.ndarray]): Outputs of preprocess model (HWC, RGB, range in [0, 255]). 
+            If `disable_preprocess_model` is specified, then preprocess model's outputs is the same 
+            as low-quality inputs.
+    """
+    n_samples = len(control_imgs)
+    sampler = SpacedSampler(model, var_type="fixed_small")
+    control = torch.tensor(np.stack(control_imgs) / 255.0, dtype=torch.float32, device=model.device).clamp_(0, 1)
+    control = einops.rearrange(control, "n h w c -> n c h w").contiguous()
+    
+    if not disable_preprocess_model:
+        control = model.preprocess_model(control)
+    model.control_scales = [strength] * 13
+    
+    if cond_fn is not None:
+        cond_fn.load_target(2 * control - 1)
+    
+    height, width = control.size(-2), control.size(-1)
+    shape = (n_samples, 4, height // 8, width // 8)
+    x_T = torch.randn(shape, device=model.device, dtype=torch.float32)
+    init_latent = model.encode_first_stage(control)
+    init_latent = model.get_first_stage_encoding(init_latent)
+    if not tiled:
+        # samples = sampler.sample_ccsr_stage1(
+        #     steps=steps, t_max=t_max, shape=shape, cond_img=control,
+        #     positive_prompt="", negative_prompt="", x_T=x_T,
+        #     cfg_scale=1.0, cond_fn=cond_fn,
+        #     color_fix_type=color_fix_type
+        # )
+        samples = sampler.sample_ccsr(
+            steps=steps, t_max=t_max, t_min=t_min, shape=shape, cond_img=control,
+            positive_prompt="", negative_prompt="", x_T=x_T,
+            cfg_scale=1.0, cond_fn=cond_fn,
+            color_fix_type=color_fix_type
+        )
+    else:
+        samples = sampler.sample_with_mixdiff_ccsr(
+            tile_size=tile_size, tile_stride=tile_stride,
+            steps=steps, t_max=t_max, t_min=t_min, shape=shape, cond_img=control,
+            positive_prompt="", negative_prompt="", x_T=x_T,
+            cfg_scale=1.0, cond_fn=cond_fn,
+            color_fix_type=color_fix_type
+        )
+    x_samples = samples.clamp(0, 1)
+    x_samples = (einops.rearrange(x_samples, "b c h w -> b h w c") * 255).cpu().numpy().clip(0, 255).astype(np.uint8)
+    control = (einops.rearrange(control, "b c h w -> b h w c") * 255).cpu().numpy().clip(0, 255).astype(np.uint8)
+    
+    preds = [x_samples[i] for i in range(n_samples)]
+    stage1_preds = [control[i] for i in range(n_samples)]
+    
+    return preds, stage1_preds
+
+
+def parse_args() -> Namespace:
+    parser = ArgumentParser()
+    
+    # TODO: add help info for these options
+    parser.add_argument("--ckpt", type=str, help="full checkpoint path", default='/home/notebook/data/group/SunLingchen/code/CCSR/CCSR_weights/step=59.ckpt')
+    parser.add_argument("--config", type=str, help="model config path", default='configs/model/ccsr_stage2.yaml')
+    
+    parser.add_argument("--input", type=str, default='inputs/real47')
+    parser.add_argument("--steps", type=int, default=45)
+    parser.add_argument("--sr_scale", type=float, default=4)
+    parser.add_argument("--repeat_times", type=int, default=1)
+    parser.add_argument("--disable_preprocess_model", action="store_true")
+    
+    # patch-based sampling
+    parser.add_argument("--tiled", action="store_true")
+    parser.add_argument("--tile_size", type=int, default=512)
+    parser.add_argument("--tile_stride", type=int, default=256)
+    
+    parser.add_argument("--color_fix_type", type=str, default="adain", choices=["wavelet", "adain", "none"])
+    parser.add_argument("--output", type=str,default="experiments/output")
+    parser.add_argument("--t_max", type=float, default=0.6667)
+    parser.add_argument("--t_min", type=float, default=0.3333)
+    parser.add_argument("--show_lq", action="store_true")
+    parser.add_argument("--skip_if_exist", action="store_true")
+    
+    parser.add_argument("--seed", type=int, default=233)
+    parser.add_argument("--device", type=str, default="cuda", choices=["cpu", "cuda", "mps"])
+    
+    return parser.parse_args()
+
+def check_device(device):
+    if device == "cuda":
+        # check if CUDA is available
+        if not torch.cuda.is_available():
+            print("CUDA not available because the current PyTorch install was not "
+                "built with CUDA enabled.")
+            device = "cpu"
+    else:
+        # xformers only support CUDA. Disable xformers when using cpu or mps.
+        disable_xformers()
+        if device == "mps":
+            # check if MPS is available
+            if not torch.backends.mps.is_available():
+                if not torch.backends.mps.is_built():
+                    print("MPS not available because the current PyTorch install was not "
+                        "built with MPS enabled.")
+                    device = "cpu"
+                else:
+                    print("MPS not available because the current MacOS version is not 12.3+ "
+                        "and/or you do not have an MPS-enabled device on this machine.")
+                    device = "cpu"
+    print(f'using device {device}')
+    return device
+
+def main() -> None:
+    args = parse_args()
+    pl.seed_everything(args.seed)
+    
+    args.device = check_device(args.device)
+    
+    model: ControlLDM = instantiate_from_config(OmegaConf.load(args.config))
+    load_state_dict(model, torch.load(args.ckpt, map_location="cpu"), strict=True)
+    # reload preprocess model if specified
+    args.reload_swinir = False
+    args.disable_preprocess_model = True
+    if args.reload_swinir:
+        if not hasattr(model, "preprocess_model"):
+            raise ValueError(f"model don't have a preprocess model.")
+        print(f"reload swinir model from {args.swinir_ckpt}")
+        load_state_dict(model.preprocess_model, torch.load(args.swinir_ckpt, map_location="cpu"), strict=True)
+    model.freeze()
+    model.to(args.device)
+    
+    assert os.path.isdir(args.input)
+    args.input_list = [args.input]
+    for file_path in list_image_files(args.input_list, follow_links=True):
+        lq = Image.open(file_path).convert("RGB")
+        if args.sr_scale != 1:
+            lq = lq.resize(
+                tuple(math.ceil(x * args.sr_scale) for x in lq.size),
+                Image.BICUBIC
+            )
+        if not args.tiled:
+            lq_resized = auto_resize(lq, 512)
+        else:
+            lq_resized = auto_resize(lq, args.tile_size)
+        x = pad(np.array(lq_resized), scale=64)
+        
+        for i in range(args.repeat_times):
+            save_path = os.path.join(args.output, os.path.relpath(file_path, args.input))
+            parent_path, stem, _ = get_file_name_parts(save_path)
+            save_path_now = os.path.join(parent_path, 'sample'+str(i))
+            
+            save_path = os.path.join(save_path_now, f"{stem}.png")
+            if os.path.exists(save_path):
+                if args.skip_if_exist:
+                    print(f"skip {save_path}")
+                    continue
+                else:
+                    raise RuntimeError(f"{save_path} already exist")
+            # os.makedirs(parent_path, exist_ok=True)
+            os.makedirs(save_path_now, exist_ok=True)
+            
+            # initialize latent image guidance
+            cond_fn = None
+            
+            preds, stage1_preds = process(
+                model, [x], steps=args.steps,
+                t_max=args.t_max,  t_min=args.t_min,
+                strength=1,
+                color_fix_type=args.color_fix_type,
+                disable_preprocess_model=args.disable_preprocess_model,
+                cond_fn=cond_fn,
+                tiled=args.tiled, tile_size=args.tile_size, tile_stride=args.tile_stride
+            )
+            pred, stage1_pred = preds[0], stage1_preds[0]
+            
+            # remove padding
+            pred = pred[:lq_resized.height, :lq_resized.width, :]
+            
+            if args.show_lq:
+                pred = np.array(Image.fromarray(pred).resize(lq.size, Image.LANCZOS))
+                stage1_pred = np.array(Image.fromarray(stage1_pred).resize(lq.size, Image.LANCZOS))
+                lq = np.array(lq)
+                images = [lq, pred] if args.disable_preprocess_model else [lq, stage1_pred, pred]
+                Image.fromarray(np.concatenate(images, axis=1)).save(save_path)
+            else:
+                Image.fromarray(pred).resize(lq.size, Image.LANCZOS).save(save_path)
+                # pred.save(save_path)
+            print(f"save to {save_path}")
+
+if __name__ == "__main__":
+    main()
diff --git a/ldm/__pycache__/util.cpython-310.pyc b/ldm/__pycache__/util.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..dd9a8cd9619587b0fba514a9f81d35aff092a541
GIT binary patch
literal 6165
zcmai2&5s<%b?@qrnVzrx;_QkmQF2>}5sxG9k~A$*3`2^dNJ)vg5ky6dCt+H>Q$4f2
z+tWQ;UBg|DdJ@}9n=%k8K(0=JT_8ZR7sHoa@(1LU<d9qHl#c^VAc+&efPo+|^Ly2^
z++C7RW>D2tufE=U_3FJ}RoZA&4g9|M`13G4Wf=cTm7_l!mA8<xzeL6j&SIlod|8{J
zZpN%{woO*hthNoB6+8V>yQFD$T<*JVw_j;j`qg$7w3m2^m+!H5jk~;Z&uG_qmDliX
z@H%hcJH@B?G``b(hR@<V!%y)OyonLBe4a0WG6%{F{3IwR-Zl6lfAOBdUkuIl=8n~#
z=j^7@difE!HC9?oIq5(~eH5xu)*U7t8Kp_fQqKGRpckr&{=6fCtx55En#f0_k&Z<(
zOnw!7DG@FE4`kB#%-msPOR^nfY)dn@cGx?{ZU6g*wAO9Tawvvd53I2xo!wII?6Gl4
zmUrDfV_cRMZlm2<G<fMHqiO80?=xQBt=>28v-@V((rR&~$sJndGh-1wK4Ux9JNRNA
zcT~$u>os1@U0%z}y#Byu#w{ah*oLg%{g@f_mkr+7Va)iZk+3a8PVG(;54$r^t;47H
zEI#v+!DmkzV<)$FXQ_=Dn>F;C%WbS-%M-iJqLkZE=iG|2K799W<!r4*G9;~b(l`~5
zz6@0ehU^j6G!(0LsB-W#QE-sHhLrs>vTqx?al0{Qg;Y7qA=eAWcg;oPV~pQ4xpC9j
zXDj<g%T(slY2{=x5OUvA=4zzOq@&F3j+n+o|NU#rYiU1RPEr}Jrs>9VC*|StAQV}e
z1o86CH&35CfA-C<7v!r~Z+^HO^ZxQsM)A_%uBu$@g-N(Q5SJE+GqP~=Tdx|UFUD!l
zm)p`0GC%6~mM+F=Cy294OHcZ@Of9X>l(QNKoei*RccVB)4$~R6t_F$p!h{&|GAW{@
z=XF!zVd@^rOWw_mXfUACBT~;J(pL}3F#gGZ>!B_f|9QDJudFyslpRNzRE`LeUMPr9
zWhLR3Af+mstX7n$SdY_y`#kIh!&oXSN~E&a(+J~bxGmM(ZzmD)b4_z~O^8$|X8_hD
zlsZsOAVdJINGy{L(kxV^U@!<1uJzy9P>xJRXHAu4m}IGFRg}54CgwE{Su_e2TW#A=
zTc#}J%I3t^i`4AIQ4(f)$HW{6g7mGXj;px_nzjztDzQ47VO8duP4*fy%>~r!%zS*x
zs)Jf(HWRO-_c0?Q1=6v+O4GcCOv1<{TQ_&i%q3qSt=w3*cNjM}%ygd0+``ebH(=QN
z&dT*iq*npgXFQ^L_8syRQ3BITI-ytsNtHg*l8aZU;$&;VAlzq)Z5NV}K^BrZQPvLz
zWa%00OC1cKV&W|ne=*A!(E5L}eCi0x=HReqZJ0L=@dkzqnnPToO!IQcP#KbtpZf~h
zp0M;4Ca$3V3`=B>T8l0Ua8Sk&+7nZ#6w^L$fl{~)3(ZcWQ%2)mCf>%UL*;`Xl_}}D
zNrfct4ilvu$6dk+pxi)?T7Ix14jRoPXh(3utUE!h7=K2*iOHWsFD;%6jn6T04fSWl
zqaD#K&{jjISQ<W-9hk=s{5dxtQ1GxuKdgHAtw}k^E-WwiBDpqPUFxL$<rL>4h?aZ9
zh==^EJnD9b83M}kYMibvpM9fq_H=N*8?3(BSv~j0S>F9Re|?qn*E{F=nKPa9-E&_%
zz53?abCcoz;qWAZU;3qQoPS-f`%$nFdU7rF`eBfG>77t`K^%KYntXLNkexM;M|~J1
z*3Fjm0BybS$`%@0j<RW-_yWWf9%ZC+@$1N1Oe|CRDJ!W^f4>_Ck_<`j<0(hHhw}Af
z8%di%u19O}A_@|4fSdQA&0V%<P<%p6G;=0SAtD;<h=e}?7|2a-HxciCVDr+S;UF?2
zhvl})E628oEQA)wD4j&6gGG4^``9)1jsys8kraZGVLyyvC^P7+*x)It#0wy6PU(D`
z-&!z6U+AR6MEZk31pVwe-hJeFYeDAq9{l@X?f&d<espPMEuHO-s@@H6G6cG{vSD9U
zbv=}bH-bW-s`!2<4zkSmmFxRS&<}l|Kt>;4H&$ir;41mmx=dvdPZ)TLUsXTK9T|T{
z=WqcQ{R)!!5S3#JBga=p(lXL!5$d?ZM3b<`hAFBPO;LuQ5HnQ9ZbAFsVK{Q=OQ?dE
zBHBAb8rF`nXZ|_+ob1260$(49UMSmhz8__NHxilj;Sqa~tKHNzK&dq05y2^l<d42|
zZF>-QWXQdZ@UHhsG5!-zre5JoWfaJ8NtFs!YT;Y9a(^?tTLh#TEa#`I>wZ_HeXJ#v
z1qz}`)REZj5?1R*y{DD-!jZ{~kww4vp^uDQm&?c}6v<<v4bm~2CVM_+)6j>u>OJ%K
z%&|GPV9ia*cFo+>xQOjqgo}W%Yzi{4ZD{xb;XkrzdquIOgVR_Z03JYPu$q#{25}@+
zCF&1SA>(LOHHxAS6B)(`xgw+>uF4QPkW$c~0zQaOz@jQm`dOVgYgda3K*}>#|1mn>
zM9K)m&OmgEn@uK&-a$I{qmx0ujLa=G2+z08id899Tt0K(=rUg6)q9n;rOP#5zvs5?
z5P)_M;8CM=9QbzmzF`|}7vj&U<`o|FZ*9S0Ub}H6Ncb%vo$TTN?Elw4+*tkK^1j*o
zyT7I1zg_PA>iva({QIB%<nqG{e{}j^KEHeI(!)=iKYaI}NB1sYfB4E@{P?}4$3M9&
zUIt%!%RBm#LcHQfh=J<bUHljYSiZTzvKlPh+(mg0Y<-7;YJ<7}ss$=`Z9=uUy$F^q
zV}prTxw8S>kkF0-kE5S%J<_-Qr0;p%_w7kvSNBEhlfDjCyMb9MXaS<0tX7_^bPJ<d
z=1)df(WhMWfv<%9<aX}lrM#TGd4<;=xZ^6F$z2;Lxw=Q!t2z|i7@gJ%Msx_Imj!pi
z%!8smI2aucytJzW$Dw~(D!?U_G+KN=K^Tp=7l(mJ2$%{S$qU}6<HM$KkVSEtc<4CP
zO$FYieHeY|VEigfv~Y>!0zGgNB|O~rPSaf4`v3pFXY{(S989?tM!hu&2-^wn9xgt5
z>7YIA2TwP)TxG@pvcwIbCpcI2qtmD|ynIjKOjVwYE{cP6D-^1xm86|@wu0kU{g`Af
zoKXpoy&p*NJCGZIZ-|=(Irb=tBpiE9*;yEO#Ybor8<fSA^(jjzODP*rwn>>lrrZfu
z$@ZDJK~x=t2@s1r$ZjKLuOf5pDY&aC_9B1`B`Pl<xvYfKT>AN(HG27ZhivJ9y<_f;
zjtIFoeN-q^NdrcOm`W}TA=<>H)kVlQcP$F#WN4ce3KlJEMT>F0Ws4!ksk-lHp+pRj
zhykizD6c2dA@p8F%S61ECAO$LUEO4;9Nk0Q1?h;uBdQ|l2uIL<irJ4xU;G7P9WJ|a
zOg7RT#r~@_Qxo`sYw*F@kcpS2h9rMngOqedH*cZU#-ZP_n31?=05P_p0s4OvX`+dT
zW2=YjJd%S{LMkJ<NEM_iQVs3_*VS<yU)&-PA{yhV!exx7@x`TnJcDm(5pd}Cx5u;j
z>^d|%G{gs+@8PO#=d)V#psqXOw!}*axryANmUU~gE9HC^^UO&%pF@y#5y1bHqs%K8
zv4888d=`0?R}psRc<ni~`b8tJi$8g$Zv*t%-P)e9TSsc_85lnWS_!LTCURxaMOoMC
zlbWk*Q<GX<{B>T{l*WVtt{gm3@aZ>TWSFa_TPsED=6iB#eR_T7iF2#wwY;7;a?;t%
zqJbQeBA>$yc;k7{fE%98r};EVi<X(1al$xh?3m*d5>FGkyF(T`2fa7P^ZESliTuPK
zTW{t~k_e+toHY3S1G3_uZh9JKE{uNkK`7`vXI|jpV%Lj9PrnI8qkzy0Zt3Bmz|kIl
zS41znpiwo|=+}wUY$!tS%XkckgVi{EO^+ws)4JeYdD5<ZvIs-+ov{6^fbfI>k4;BY
zA~Jd+ltYmey|BUr%2R7`@(QPN=taoejDDqX|2)O?c>M6-ng~#M8O~!CdgSh+Fp*jD
zLSd6=08fp6*iXe>zwm0RjJFRzxYJYBNx{!H)l@&&_7CgrMsdO?6z$auU#C4B`MT?^
zvUU-L3qDhP&OX&ixtKTV<C%aUdk_!Sg4TR-tFGZH1#SAowD=SgDq9!n5)Y)5`Grub
z{vl91ZamsGi!oY@i+FVuiZt`%Xd_h90}*DSxDy`8SmFo^c(ic3fK_-x!6qi5fRiv(
z%|Cbt#HG>+09dw^)yHdzHHrqfJPl)C*#YN%YoYLE<%tF0IR^VA{(vT#(dx;3f;w6P
za&-aElsnN_rfP@U%-T(gLC=h-7XB1>icC!(-vErCG-404BBb0S*}qHKp;`O}RY@uB
zMrSDe!%*kw)!I7v>Ect^S^Oaek+04u{@7&~e%NKF9#<_Jf#&Gn#Ul*i=(<^DrtU?Q
zs#&*O)-;=@t@9~!iu^e88EeLLO%r{O{!0j8B~VH~JvlXpJB|6+wnmH3_4(RtUC{+!
zg7Ax2qbx#pU93})@JJyJVf<D_tl*=n{sc`^fb#v#VGtKrF&$xrjyg-!v7DrZ)sYHR
zV{&=xgfT8MN3^7)mX6RN`u#3aM)AfqOfB5h&+GFg2iTfgakZw)Sc_eU)%-q6+Qtim
zqnqmBu4sX(%DAx5JxRZ8DC>jwzoFmt^pl=`%+oJi;w_qiUTKPFkZV+;JD$D<>RXIP
z^s4e(^tz4HB{UaDn12E<lzXwbE?lA;W9A`q4cDYCt>c}oYPpMU34hxyRb97StuDC#
E3m91fF#rGn

literal 0
HcmV?d00001

diff --git a/ldm/__pycache__/util.cpython-37.pyc b/ldm/__pycache__/util.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..4b1bee37bc64cad12198a31ced536cab780c7944
GIT binary patch
literal 6140
zcmai2O^h7Jb?)k)nV#9%|KW;DQcOFE6pbzKQZy}53_)5FxiS%Hqfm;;Od_;;rg~<3
zr>A?ix`x~x_aK3nF=ZfhbPQqy*gYk>2ttBfgPd}5a>y-pN}PktAU_Tez<?1T`Cj$x
zEdQL$pkCL1y?XWPy|3Pzv$ItLzuSMY9EN8M<DaQ;^s|xq22%ECAl%?AHk!qkH5tlg
z%=%{2WJR6Tv{7fpPQTPF={h?u_uZ!3uQV(DYO{*E)4asX_gHg=yS#GGXx4a@&)_@D
zYkU^pIX=hd@tx-jd=cLTeugjeIz}w=6MO|VOZ<8M96yPg<+lyK%1_-h_^Hs`uJ2pT
z6P&$oG|oN*x5j3JDJLDssE<rF$~wcOEu%DPSjxH454xeM=+89~+?`}M(nLNaX|yex
z0l&&5Q91iB5NUjD?z6Ea*}gHhcYi3&+}dZ?j1T|%dxo^OZO(Ftg<B7-u_K*>QtlkG
zaY>dB+(TnrmKAQJ-B~qw>9kQd_SyFsFCSFz8~53LvtwzYxYFbft@E+5iXI=ced`*&
zn8)2)$xGWayqddwCNJ~a1DhGQjO2`M$Xc%|K4%8~<t(4wXUzC5BQb9qa_(TBxHwpV
zXdOO(Xz_*9246gBjGf#*Sfn;)?3}zsJ(qGD>)3MnpkCy18v<S0RMyS6-&D@stw@F>
z*LE7G;^9jWh+rTOv8th1wM_;3nMe-O*O0Pb1j&sLPmNizr<~>3)pN%0o2$mRFmlJ_
z23BF4pBfEQnQP~ilgU6xoRry$l$o@Zx!V?VnC53!*0<7rxSpgkY^7;$od+^l?}~Id
zSbu+*yb~qe_Ewmzw^JUjUyC}OH*efrk9mK6D5H37u=lgd<!+dSy904$g;*m#ZDaJM
zIPLm!SNcKbNB!>F<v48zadu^G+Ph)ujcTHtRvff@;MML#aSRH*88zMt66u8rE$wAe
zfEBNk3J+6vkze!P??r<F<sQ<i4@p#AB*XYe{~JZd_{r7A31!7$qU<=zq;f=%bVEVc
zP+3WMSCCAVO)4u&l&!^Sz<nNef?+I`6(v&H+i8UHGTfDF={J*z__?mRx-LX2lrsQp
z5>g!~ClDgoQx%D2vO$`KsuT<cVZ!wRw0p{tsc3Jhk_?k96^)8A?{0}@%|jNALd9B5
z8)D0p1*&XLe4V6bCytUZ(?`~@#3HK15)m3J>eOTIagooUPAiIGxva$I*aEAXE~~Sb
zO_QylJZGK(t>Q08sivAnVx-&o7ESOnh=i_5wr%d4*#emZY30VYz0bJWGt)0qo?Ezo
zc8@gM+1z+Y(iNvb9@0Dwhm1v(Km(I@D87O!Rr)~hU!10blWhe9x&&-j@2?<wPfnEe
zgTXJOS4O8}uo|<OxP;uZ*`=2MliiDtuxk#EOIFV;m=_l?g=SuS9l3@hUZo1+TD(St
zSQZyS8ce)SdAdc9uw2Go$ryea)sHg$t0+IlH0i9Cv5O3Bn9-}~JFDmVIg^*dU1)K3
zj^?KE1{0S-+d7X7eW&xDMS+B|&qOVUJ&>>x$e#j5DJR3uHgOGU2BA;NRMwp!R*XNj
zsmX;P;Y>E=(n(xJ`LRvWDd|nY@nPw0EDGbxj#OqJrk<M*XlZNoS=GaDOUglZX??vL
z$*o~)t)2GQQ{0#!TJH`c9`dhp7@c7TKeOJ7)7JX=*V^aL1s6L(>-Bc)!fWSw=j;5{
z7U!?FFY;GjX<zJI_}aPF>*p^_hWp3ElLYqam)^Pfs$Tblpci^_EA;wdka+2xP<TNc
zdr6vnwH3(rmdB$$loab`YkGiI3TUZ?2I#a+;{<I}JWu3Ny`Wq3qzEh2-|xhMBqc3Q
z&k=7TzcD>NIs`IET7q;e5(t5t_aMmwc4&~(f~z!hCN9EF8ryJ>KL9?+O>WoW1b<-j
z(x2dBG9xE#E^khchIE1^&nS_dhWD5oLmv)|!^b@a`J!al4`V3CJUS{icw#5w1gbTM
zG%IZVFSOHPBK<)if`0Z4OFweFtswKd5B}{h4}SVLpI;eSYv((os`sup83Jiq*|4vw
zx*W<-Wa0%{yW;!pILI>JSFZ0TK|l0;LK}Tp;8>M$W-YR@HJQpFo-ptPx2pCicO>qN
z?wbJ<XtT3Syo3ByOl10|8hi}}0w2Ayc!?lL&lGiXtH{Gz2#@kOAISSV3|9+%30!pO
ztO#yc`^KU9ebV&OChUD6x}j_?`F@o7ok(QThdu0KKh3(X1C~k?9uc0xR{YTm*LMeD
zTZY{0g?rvd#rTgrnR<mKl~EwWHB~A^*szuRYvEquj~1|!pSHIBj!64hMkot4a3)b(
z;&@9~s1tRcl+$yM%v+2s`rUvSGO}?7bMdz)Nqm5ssp68-KBl-3f>zx_^LNa#Ikup(
zFG+S_=B7qN?7$*41Y~8CXMqAks1FGFk>c8$ime@8xAFkU0Fr`|ltea&BdID;e~=0p
zM=dp5WPNDFFow4kA^C4r#)bna1q~{YgJ^<9Rhsm(+Ht0;MFIHaDHVSYo!>{wG?Itw
zv<pW`^=+i-PwNs&8Qs><APnC$`NDmp!+3>P?^T+X&d>1LJ-2Cx0I-LEiW-ySE;q~f
zaf6yJHoc_kZ}6ah>n`lz^>^P05`GKNCi~*QKmE_Yf4BAC)lbdt-~Nz(|9Z9imHR9I
z@OMA`kE>r?`h#;n`u^VaD_?xO{@L6AG`e?n<BPL@`r~h|{SVp-9)EIGA9P!91XTnJ
zD6g%-kCB0jn?07TK$Xn{<PX8tcNl7I)V_>b3$-|_Nv*~0Rj_OsJtnSkr)Olh6FgVY
zuY`WO^^v~iY2ShFdvDU$)qT-A?dxE*KgKK-v;aR(Rx3|d`V6C4_E(e9RrDzreP9Nm
zHMyNTc_}aFZeHOt58QE;uHS(TTwFb*$5agho*kXj0*28Kq?ZME!pwu9Jy;QK<-N3{
z{lT#jTAP4K!x*jJNZ=VG?!{ps5<;Q^_wSPT?eTF_ILM+nO+0iQ>L!hCr$BV{!qFTp
zC}&~&$kut_CQ5j?>z$+dwBG;!eP7XQpFNuLZWwj9Byejx*gIxmbo!`0><3RaHe6-K
zK(E9N@FuudwMVB>Wtj7>z-_8bM;F;adN&kmMhi+?<!lodt@<t6cVU7`!0G)!ikna{
z;Je`f{b9kQB$BY;Gs@1wxFa^vDo8$Ji%3Ldn@EpHOr%dF0a5M*f@FY3Z&R(tU4#;a
z0QnFp^FR#S1+1lXiY@6xc>=rygR5-x{Ijmv+7WY8lSUVWOq#wZ1q_Bt!DW)!+C2sD
zYvPf*0B>v_SmdQi(RS*$0OX)-)~4RXM#C06m_*flKMN&HzJ$qFwL*0Il?6M5mdWM-
z?a4*!*Mba{qkBBP;UW6arl<I_#ztTIIomp3c5}KnLI4X$Z_ylQ;KsQ|6`b`d;zg;k
z#vj&zBIrqOVn^&ZjGM-7;}f>~@5R19VITfzY<2NqM{<x#NM$4!se)8Rnt?gMgLGWO
z7Y_tDg4yw0VIs!!_~Ow%Ucj6y$d^`uZZNAmUd$J_A<v#G{%GgBcuw2-B1xQTw;l24
z644Ga4!J`u+m>!Kw#)e<W?PbOz66Bn_RLS1xWG~3mCG<r+m(C~w92b+GE02s8FjVG
zMqa}x@2TFiXjwd%IW!JxNVA6qW|%`=2^?T9GGgGbtV^{?$<?K~Nr~nac~#fVPU^s!
zgKz|&e+?>z*=BTWrD)yxE(H&^=eHN8rmdRK<h6V@CkZaB8lc!Q=o02YJm<kIjPPPU
z&*y2+XjzysmW`9fzByjry&w^W$lZO?*cWjQ_3?@P#KCgDe8{%zd7bu!amyzSe&PYN
z_!GACFSm@HbGHf{xib3vy-?7N&b+|GORgJ-o{j@WqksSl9_Qhp0LdPHZ-{PoNke8T
z(XSn+*-(VuONar)K`Rbl*5e84G%k5>Oxv|(7GX#h6gMR+@Hv6K$EKqx;U7H_%ArV#
zURYs*)2X$%eT7*$wj*S3M!!-x0iI%d1U)<$Cjt{*hC7*s9vQqSOk`FBCT!XoK+|kL
z?5ASSFYKBsBjDi&ce<)N$@tlhn(GI<{&BfETio(VP4TAG)=t|vvUVGdvNjWi5q_ok
zoG(n`w3Bi%Z`4P0fUtQG54M8FiQ*YOgLe|7=@Zl94klE#&Ne7p0n#isrD`_;(D9tn
z##!v5wRm|~N1;eFKaP5#njeTT3&oxAXpbch$ADOc(*dMH<OGM9gn}@?n(>cfe|RlA
z0bt8rW%Usxu}0AVFQZ}XD?8xaZ>$vdtUS>GL}74FVvi<S(BjE_0z7&LWbOi<DR&~V
zOwAliGi%n#3q3WaTG&)PA2K!n_yJ)2q!DMB6;9=MY45*B<X9~T=xd=gXWK*JA3Hlo
zuh!J=PiLRN%wmL*cagFIh*2hNTxAO;!Nte_E&>-o&*l=^O(a6nHM45LDVcS%&Pqg0
z#AFbaL9E7FFqaUkF`xLC;J`{4aS~CS7tPV?GcCSWSXTsKNVJHwiExk&5mL58gq$b6
z{wv~}$f>G7LD1x!e1B&c#Kmh%`&Oa-jSk$DleEw_+Pa#ZywBQUj5o`%bdEfh_Rb%n
z=WiorH4xXNfa=_G%_%WKE$CX-pN$6FfV$j9p@_C<O{2HyU{Ab=T2;mih2BIuR-vr-
zZoH%8cshiqgLgWPrK3ZFqAf*WgF+PgnXR9C`q?4~#;eM&Qz#qPN@zxoxi50WyK*lV
iZ-guKGR!ET;^JIzmS&CBt?H^<!rw-6t7W%Zb^jO5Gzh%_

literal 0
HcmV?d00001

diff --git a/ldm/__pycache__/xformers_state.cpython-310.pyc b/ldm/__pycache__/xformers_state.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..2a2b79bf00235b25a1841e40c73f631d9a4dedb3
GIT binary patch
literal 982
zcmZ`&!EVz)5Zzrnsq3aqfl_farz#SP5dw)KgeWDIDiwp8QXoQ#T+fzTb-dAf-9pq8
z>6srud_aH6S592GaN`0qwh~eSyV}{=*;&t<H)FN9=rSDVzdp(Dg0Zi3m>qRIJjRv1
zMKNjs1D>!RD|e2&2*oRwx}4DjD-_ZKd(1g|!=NVZLjiS|JGXl^SON#;QLCfo!UAe@
z(168r)^mV&S*!UCBd}m^Yji76-atW~NN?@^q0$4Xv$eMOLaRthh|_o8X>5*EZoJsE
zi@KI3*J(trX9vf_Ca%mzVHif>L5mkFueYA>KI`mu``g|A#{S0k&Xb+a1=HkJik}_-
zNDZW)DkBe+I`$(4(jOuQm4=DmUB7+z{+;!Em3ecsyXz+~@FzI?nF)<-4?h&w<E%fm
zh9_a1ga?TvwlgG}*+LVyf5E0Zg8gF|g+aQBOr)6urC<f4D^It(8;IGPByw}>LKL;3
zj#Hy8eDo)kN~@u*ScW*G_)|zr6oZm>xFa;d`Y#kchqn}Zxf5KDyr&(!`e(qVOfSW|
zMve<48Ym*F&Sa2Iz$xwQ0Se%Tov|@L!54;)sk1Qo$_C<C=v5=8Rtu~%_L+SVWB!r1
z?4Yot9KynihWV)ZGCfY!Y3hMIi6iN??83^7E+LdT7w&$T<g%l+(%2mJ%5Ns2lDw5L
zH)_gUS>imqL1>A$26s8FC84k4u_ZLsFlf#9oJswAjfvJ)!>p>(JUSINc34t2LJ6k=
kMqinMP<CdC!+BW!7LSOrY!!u|^5Lv=n-Ey8?baH#-<L<v-2eap

literal 0
HcmV?d00001

diff --git a/ldm/__pycache__/xformers_state.cpython-37.pyc b/ldm/__pycache__/xformers_state.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..5857c9a68d380091c0e116b01918c32fe53ccfac
GIT binary patch
literal 996
zcmb7D&2AGh5VpOWbX$^bQG?Xu!lgt)**ijr{-h#O2`DKg<PvRn+-8-H9eKSit$HGT
z3*Z4b^EP|s#Dxdo#MnzpbKt;O9?y)%p3k4hyY>1SK^c78<Ubi9-;J}7DjF|vW@A*B
zTR?!OWJv5vab*#EN5mE-W`Gkh?tnX|l)NW`4G4Fk@|i*vR&Ly31*r8rSVgb;$6JGY
z=&e8v>NjNQ0qv8f{{yR$?$N>3ldcS=67rM>+aFG)nsSwGw}LlH#+*YU&Vq|Xk7cfd
zM7Nr>s4A{=C1SIGAry^0oSBPC6Fi@GF|{!`c-?!|?)68V{%GfTr*pV_*#1hC$+GI#
z%Wy2GJQPy%6DcPlL^=x3l+0&gKNp8sJsxur#u9kApPZiVc6wn7({KTFl<7$G*6gx)
zkYuCVXmlPWX>^iuLwkmd6T*}sdI`0OThtq9Xyo0nihNTrH*mUF_xC&f9fTY#?(y)T
z$%@KMB|<9)H~qayX;m;*Y(SEk+P83gj9O+0^_cR|`#%^P?d~#)qb5@;ru;uKtaGsh
zVRH$>s)1nM%!padIeR#>r>KBxa!uxRQdJF2=a?WKl@VYQrXFi%W4?C4xh9{;7dEFK
zOU~SG;l?>cg%i*6tM#{HBIJb#fS)HZ51MY_WLhax-qKL49Y<--+e%4gI;+gPL#pMo
z5#?Iml4)N&&$bL&`>AM+dS)`Uj<M!y;Vs%=SX)>vIm=423~hyry_VX3urNwMX7`=a
fwnuG!?^*dho*SmJ1}cUX&8SkB`i}4V6~FQuWHi=T

literal 0
HcmV?d00001

diff --git a/ldm/data/__init__.py b/ldm/data/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ldm/data/util.py b/ldm/data/util.py
new file mode 100644
index 0000000..5b60ceb
--- /dev/null
+++ b/ldm/data/util.py
@@ -0,0 +1,24 @@
+import torch
+
+from ldm.modules.midas.api import load_midas_transform
+
+
+class AddMiDaS(object):
+    def __init__(self, model_type):
+        super().__init__()
+        self.transform = load_midas_transform(model_type)
+
+    def pt2np(self, x):
+        x = ((x + 1.0) * .5).detach().cpu().numpy()
+        return x
+
+    def np2pt(self, x):
+        x = torch.from_numpy(x) * 2 - 1.
+        return x
+
+    def __call__(self, sample):
+        # sample['jpg'] is tensor hwc in [-1, 1] at this point
+        x = self.pt2np(sample['jpg'])
+        x = self.transform({"image": x})["image"]
+        sample['midas_in'] = x
+        return sample
\ No newline at end of file
diff --git a/ldm/models/__pycache__/autoencoder.cpython-310.pyc b/ldm/models/__pycache__/autoencoder.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8045f7e2b90e22b4306e48e6f6c1aea31d905252
GIT binary patch
literal 7866
zcmb7J+ixS+d7n9R<yDkc(n{V-6vtVcK`SL&7l;ivwzF%wj%}s2Hg?M<9gKK}6sh41
z_nhJGiXOFbwWv{`n<79TS|m=Xpbr%a<WI<3-qQ4;K!MIfQJ@cn(LNLiP#^`I?e9B7
zQlwUHQ4;5yb3bRk`}a*;sT4K*9zOm~@b911w11_>?9W8w6%_fW09@mGs5SDZ-e9V2
z^h~DdTF+|OI*qW<>A4M8)s3*wD>jO~Qlq5WW?1f38WmNy!fJ1!v7qX9SnDk|7Ip1)
zjXT_ZsBt%7-KArraS1&IUPMnZ(7I-K8GS2U-_`1+k8vYyr><w!W*jBKgQVw2emf9#
zmN~CQ&6o!wb8ZK!f>~E@cl>r7`QhvSK+29E-R{UlboK^GCyp?Fxf996kCKj`1YS$T
zJrCQqI&BQuZ*`K_dVZ4*oc%fDhWs%qf!5Hu2Bhe1Ac5=ZZ?v2bwU*9JZapkCjKDmm
zmd%}qdcz8AeuKNb@K8V28xAk>(nGD`@-naBS>RQ^fM=1{_#&Pq{w%-5m$9zQSNJM=
zD*Q5E!?ViQ`BQi<@aOndeho8f{5t;(dKUQx|16$M!KKmiiN@>vY5qBN+6sS$e;y;N
zA8Mx7xD0fDA*;MPNa7$z+?#LhWLBpKbZ$3GGp|hb0*ZkmU(q!!(T=r|egXzai>c<Z
z)CM#XlUF)gs{ghA9UW~evFF-O;->7FMhYqLw64h;U(^0bPm0~rsLUDqjCE}pSWFE-
z^Tfc+N<Nwz`RKT}iniu0qz=~xy7;}MmO96JvUtGc*J;O88!d4wDUL328*@yoGEW%p
z>Z<vBsf*RiX@Od(t)yCaMSM3cpuWnTv|!>z4>a+By9b(lpLWzWb<@kKhE;_}3^Uh~
zbzV%KO0Hn_+UV+ub_{xcPvfPJjAP7CO?B1}FnTRDQ;U}$;RamWRi$3t0m<XRUKn`8
zj~UR~C=R{mz8^(FC_e^<wajhCVJtes;A2928%6#8Er0yy%ivIp2TlKQSbH-F`kkn~
z@!H$3%8j_Sk?ETlEp^&aECTN!IFzSa{d#7}K_AlLdR`~$B%YU*!@w7j7sl;-zQ_uS
z@-$u0iJ3Fz3PIW+b8r?3m$0d2hOv~HmBgaCpRp*)>|1ek?<UWRcL$Ir5~6$ALO+&?
zH`7rFdV2x)c&C?{$zeapj82qf3nFMkC<Ec`4O%T+&=P*cqs-~WxMmD-9>~B8dVYrW
zzNg+<kPv2t9&Qw>A*(NFFCz>hTn^i1PQIJWs_JcX**@qe^-^ZaAZ%qW=eevmiu5*l
zU3^Au+{8q$?<f0W5pA&w(5Oua;zgmq2ReyoXzT`o6@XJcD|lWr^riGXv4o!geq(Dt
z?gd*>oCJGue6U3r+yV_GX{4>)7ryxXmv6rC>v{LBTf6UUg}k?=UP^BHXL7mOKNQQ@
zuZkkS2v98Bx~<pnFY0Cey1t5*i`E*Ttrr=MUe&7@aoLh?J|3>lN%`i41DZtW$&XM<
z#Jz<g`v8g7)k#u0d&EdKkBkJeIWmaXkIdA(t$l0t$VwTWi$`{9jm*?O(Z!9#;^qkL
z6G#Lk!APy&(C)6>U4$$+slhF7lPqAClR{!Xf=<XKqwelxMzcQv@0NRhAAH&LNlj-(
zsA#BX-f1SojoU#OB*g7t)Q$I#Hi)nrXA?HJvD_vdkIB_{0tpTZcs{G{il^{i#JxTj
zUr<$&4t0P?<oyl^{v37NtyuJY@Fval&UUXm;#bwODQ{#<a`9#A(+@J%IZ%Yp!A<mv
zUjcX?MG_A!fm$y9MbO69*|*szRdY!nUY)1TJij(cJxqR7C)3?Tkv{_XKm%4aaSc!(
zK`#+?P6+GNHq*Qg3k5xSEwKpST+cB++~C@~(BSXoJG`h?K#xy%$XC5IUd7JnG|^(}
z<Zpb+x_9T--Ph<}H+F{WFK+At(an8OC5cIAZQKt!?fpa!ue`W1C0HVS-~nVF4+8U_
z$NP$B0iel<s#&q`3%?gYb;!(V#{EMN-HLvW$O&I(IlVBdoP~4L$u$Z@zaP9z7^SuB
z+cv8yI#qQ>@`RSFLyA5znf*(u*L3y?W5dhyR3qJ_c50IPp8QC_sC*qot^p{d!Dx<P
zdWcXHWS;6fxa{;5>&BSTe9@MfeQ=+i=Wo9hN1noyx90<)UJ&seI)|j#28%)PH|8(-
z<bpZ&vapbDdy9720Kh$A&`<OuVl6#os3|3oD}Y^1Y-1AcdKnLKjR0*?xAM#D4>En9
z^ms3b5}6gq9t5Hji}NpH=ceWztd%Z65fb9+raoMrzkWV#?#dGc)0KaoBY1pe@&lm1
z)aBXus;5j`$3#IBPsb?MhnWGR^2D8TgZDeMFCmof^yH25NuPQnL3CG062C>Y-zGq&
zAqS)(vZ^8>*+pexzeS_p064D}=BM46)7^BzDdA|>DdCKii88b{ytN}UGErkx!?=-}
zz>Sq!snMF?r)~=3ASh73*AIitBx5A*psQ}@G_^WWe~>h4`2_Ebq;F%AAPS0I0_R@R
zfkFij^F55oG5{eHL}o<j;mSOj$8+c2kem>DOU0V6ppt?el}&<$>B7=*{eX!D*ca7H
z@5v3=F8JUlW*5F`m+X!{vhnstYQXn(c0>jH)Y;~aAdwa1<_glTuppIQ(2K>Pm)ij-
zpE~-XclLd_Ee=e0Z;%8TgT7G~Ws&qtr#*-Va=aw7TCg1n5ULD@r6g2*g^u_dfphpN
zw1Wh08BlwGQAxLOA;B(8IGFC7s+rDh5>_U^iI=W0^fUD4JV5&Sgb~G4^H~2TJF;M_
zff-=y1&lHBPT{}9qwb=NOhgR8+#?4wog;Y4z?%s#+8Y1uM+o@PYr~t)dmZ4_LZ7n>
zO2B7kBKSY{EgV4sV&4M1=_mIsp4+!D_L+-0nc#JjX*gwY+kE(!p?^p|qwI&BB(wd1
ze7QSyGjsbQZbLDO%!Y%D2r+Y@{h)yQnTtx$kDL43istYHJF-dZP8~rEs{B2%j#^e~
zhw+{t=Ig42_?bfuY@E3O$~CE%6sd(z>k7F>?(hd$CGIDk9vpe$QPWmbmgM$=Y2c^~
zn&jk2p#-R<4xMc2`GFvQ$STtwy6So6sVI!VSgLCZERatFRLP358iT*_c$F=Kk)|z!
zTIMf?SLOw8Zl5{vBehPNSuqze@R+$6Sc}*O>gCT6$%Cv)FqPA29GUOIzM1$R8yJJo
zeMyto@5AcR|92^TBtbNGGLxKmh@C_Fq}6$VXujs%^TQ5T9$@}nui@CSX;DZPchMop
zQ*D%|_`yEs(X()FqYtq{RsqV;KubW5$!fne+3?ytGS02av`?3|5jOi@q-`o?95(4p
z%2MO9HYaIgS%kb@JWI*V$f5`tew_7@0U4au*EJR6YVf?>+yg|dIQ9i?pxGYFkAW8x
z95+_XKP+s45#eh~DMED4a&eB}Lz*fHBm|~xDhAXf_7wLB%xg3;MBgJ6<@FFNl2cWD
zTr&u67-UXueRzEy=jUlY=SvX}Qc*nkeN>3sj`W0(%JQ&8)y=$qSyL@*u4PYJ<k55;
zaT7@ity}08&m7~FjY{Y%cPoUvq}pBR)>2AOkXkh0|KvS1W07{7j=;F4W}CG&GN*NO
zXSn#{##<C&lN$>LMUY*`sQUfO%k97WDgFNAW$-!+p~3co#FzCY$ohDVClK|hFeUe|
zUd2k~x#XUfa%3p*R4%Xh5^5CNNAQRd+uMG>4}UP%Vx$3adTy&MrH?Z6Rqpq)iqa=g
zPmvE>Hj%DBz+R22A0PyVgDpJ@s2eucK_N9<dS0Z4mh%vLfSkir<;m}1E4cwc5h;AS
z$7K{7Q7IWWQ@0=27{M|dK0VJ07i~7j4rH^bD0U!&38KN)b*H|Jp^!z1l<1xp4`3bg
zSfZ}yiuxSGM%dzSP^1d7o`}lkb^jB(fRy8#w0aG9qWl;<E9HOUR-`2D6s(6j`CN9&
z<R8Nx+sYo3m9=3FkrCtUz!W!A=d2~48)vkzUachgc2IM$J2GmcLh6o+6!V?v$TAr(
z!4BhT;%Ug&ab8K_q){2_`PWE2kIy7cGJ9q_EzF#`H$Ag}laLo1pH!SV$-)T*oMgX*
z2=YwHM=n(S2XiVOnE>u!wgsVZcXBVQ0!(F2?n6qYR~n^pi!4Fo(mCTuok|DQE15}o
zfEhde`?QzMgL|E3Aik*@N$iRCp7L9Bo-eAD68QKa5bt419%w0pGHbO<aTFfrZkTL1
zWK-ze-=qaG0Wy4<`2d$wYOvu_Rb99;b%fQDdDhC1U}}3Ry(a$_TS!U@*o>5-iBi><
z;06~}_J*F5o;B33{=#+P0Ito;#@Tjrkq0p@<!O~fkg_I;AWu&^Dt`$@$=NZJ*FjOn
zSz?S#j?5}3Z-Vz!o)p{#P9jT^I9(S$Hk2c`9<d|juhG(p1J;nGMXf}tFK=H#0)Gvv
zIHeRlvN&ZKv?=CO##@lPuUu_Kdr04(pku5DwO0gL;(A&6T$!`)OFv13Apb>?PJE7P
zL{^~)t#4;G3G_{__RS2aE&v}cLoU@sqw}_i?9fk8ROC?wTAzV6dIrcpA7G9dXB%_t
z`W;*~r5fAbuCIv?P|b>-hmdH1ye-nF9wj%!T;Zw&U9PN&x`JpSh{41g1a=5e93|c*
z&>*l!;68!(36Q}Q<eLc6L^<C&QS7w^l2zZOE_#iuGR`qJgD@QD8S~er#GtrIfOHW_
zn0y|<)tJSQ$(a1B|5d%gO7IM;uIbvYRkYlK>$odL8$A^p=DD`6c4%Vp+;0+W$giUN
z?866V-`CaW4uhKyiw%|Fd#E+c{F_b#Za#g<Y1q8P%Xs3`h=~s#V3bO>^acl)bdtmE
zj*!W2LSG(Ek^Q@O7eWZz*YcaeROgefrz;R<7f|Ff0CA`XLyzwfh->q7$|NcA1B{DH
z1Zd_s5)-QTb?Q}et+1xVRiTb%{ymDMlPHHWyz=B7&+?~qm&uP-&>bn{|2Z{C8uHkm
z_CL)d!94mqJ6|1rc0TdRKhOrWv#Af)FF5^zEXeHH;;*pu>|8OT=DB0;pzj~4{{-N|
zLoP^c&>>Xl1!r?vNEG7Zm;OV&1#e!d0Ja0CerHE~gqatepSkL*0AiwlqP-}tSNW=o
zu|1hPr|bT|Z%h*R&vYI2#!p><)^XOMUQm)Q{(>g|B?01z83FzwHI)F9Z!?40GXW&$
z^k*na-?&KoLevb`s9AzW>Z`eMFZK`Nb0KdRQu>p;#EkXB#wFFc8AY3|L8M|s`0H1&
zE-TJ-{JH9YCZunmVP{V%oW|4mfUub>D7ncyt=2#yww!cOGAC3Yck*vK&(6<2`!Gst
zXId($hd>v5<GX2?e*#*VT})=up&Ko{HF=k}m6}wVB(uj2dixoKP1!zkU&`<J6%rto
XDtD3ppmceSmGv^plA-<K+E@Pvt_Wd}

literal 0
HcmV?d00001

diff --git a/ldm/models/__pycache__/autoencoder.cpython-37.pyc b/ldm/models/__pycache__/autoencoder.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..4a9d4146f9436ada52c8b74ccd297a6611b603fe
GIT binary patch
literal 7766
zcmb_h+m9sIS+9FlS6}9`v$He1>&<wxiPH|62^Ip&a`0_;P4Ie>cuj0I4n<9$>YnPU
zu4<oCz20eR6vV7#B-kY4_L5lYc>^Kw7mz%K7X;w}35j|@NW5Ty5RZr_e&0FOJ=47l
z;lbUiQ{OptZuMP$-*>)xz1j3N{C@OnzZ(79vzqoVRGIxO5U=7AzmI}xOi#3qy6YW7
z*3F@1Xu3AEJC06IjKm#!9Z%9`QXBdme^~F-W!*{|!)B)`X*+2RS2`<_c9PZMT4zny
z-qe`Oyhj@IB4e<AYIZK6rN(@;_>nfS1{-L*%=G(OyZ$M5r0uu$venJfJbIW9!!+zi
zyls^3owS>=h?nl|ND}CE{dOGovouWJ497ymVR}0jIgby<d7PzqekD#t9;SI5=26h&
z*)YJgy||A@&b>IlGYq@5;OysCg!nOtNbBfK14Hya7{GM-H+$~KT2E&dvmezuW@MdG
z&0+2%y<<lX`y%sL?U8<}cU<PP`XjC5u?B16USln`g1gUF*&6P3_8hy!HZZQiF0)Ov
zG}#rlg?o!_v#YqTurIOe>>2b}WjEO8(6YvM*t58=N0+7>XBum>o9y#)wae@l`vRV9
zeymwq=L%T)eA#?$oM(|@+FSSbOFJHdm3O=K>MiMB!)4+Uzo=_kuAORA{S1OH{6Y(i
z!a-@~R!=MR@9LoK+?msE?iI!<wblyoW?K`tbN_erygq15o6JC)xvg!0bA^f0Iy2F;
zrJfe1dRn!vpsoe0h0C;~8vl!Yt#D8EeErA}e?l`B+Vm2$^ZImyIp||ylyzockDl!R
zmxYJXmx~(JKy4P<VDtMrzgkf33UiB^g;zY%_`fpmNE82&VyxWhR-s{3?XiJ=+xb=I
z=hyP<7`;7x=1e<<41b`p`X}b8j<qbB9kafG=Qj$guvz0V=3yE*-Q0(ev++R^1(Qz=
zFtg)Ng6?6Mrcok31z)w&>t;#D<4N=>F}sgT{wm@>Ux7Z_Eb4~GlhwDPXcVXYojY&8
zCU&ykPO0zWX+7?z8IOXa=vcHrSK4Acf+cuC5T|h-1Z5+MLY@Xm)_)N4vL+=>ojI*n
zx|J4ivi#D;N@Pgfp_-LsLX@r7gorp<%^l7&UfOxayN9Kbrls>nmOglamHvlgSP~h~
zgK}k*i9EQ_P>Y5K5ery6EUo-_6qRP2=H&{H`Y_0d2M6O`4;!|5n6k8VhZ(kwW!$ld
zf@m0)7#{|b1DuSotPQclL_V@doaQo<D8*(mUFoXXuxt7D%C3(_dAnX(B1(Ft$C$G7
zj!$o=XnsRZ+(pM=6y}F~6?MLWqO&>^Pmm_z0i?!1PtU$U#U&IFKrINmNhm}R{F(OO
z_x296VYHWKd32CvM|&*H!@WMw#-qLa<MbZH*FB8VJ>v4-?YP%_{cG>+C2Y7SUrOwS
z=XSX}I_B$Gp@mEAqR<+??%?0R-^X=B-$c#RSM@EFj^R_OU)5Xs4bYR#1smU;=|h*q
zPCsHMiTMp&;vtG$8|Y-O%y?{&g`Su>EOcs;#-CUZUo5QK+PA-QVn2MoFhJcraSD5C
z70#K?cXOLrQ`FC38?Xtpu)nH(xci>=;jQ;H*o9k|%w`VR1$w##?B^38L}ex3era|`
zW3aL@3`fw^Zb<4@`T%HvGmE?VWaoC2L^)|cRCnjMPIgGNJLesC?_#(^2#?OKcOn5D
zikRxv_V_m5kMucYye&zx7P$b=#YZtj{v0(sy^Ie-Xeady&S!7C{O9DbxptI>VEmV;
zO+PA)_=tZ6v?_j{s4t`V1zdtesOjrcW<KQBlK;u|Md>VRYnPycesm{i-Nhvypa3s5
z{vt|!3Tz_ToDsjNZWW4#X98BfnA^l<rYk;Ae5QR1==~mWNzO5nV}RoGIn=mOHO^4u
z-d8wg?oYN~+PM!#L3`Owo)HG^d=$m~!(2?Ry|goDc+Nw-3&d1C1m|vH-~5XxfK4Qo
z(jSF997ZsHQM%o1bR3{r^55l3aBV~J8&9RHU*fL91Q_oqdWAT%i9$0RSV99f(9#Xv
zmo-nHTv_BZVTl~qCGe#mnFC#4$0f+*C3{xE7`PW=t^{&@A6uHgNZYK`A>@5gI!7PT
zeR+@TSsF-g1P3A57DOrAr!@#@H5lq-XK`y!cAu+(jS;lRH*kp^6xbac@0or=N}v}8
zC~2%@sD%Z`gZ-NPGJ5jsRL}%%TWwl@Sn7v_v%@IOMOh=fj(D8$<ri?2Yj_7^g@;0e
z_7cWSHWoLoy3IE`W2oHh7KY6?Oa2-NBe!j+O&a_e^x!1G8B_I1X#(z_*o|^c@1v(6
z*iyS$-izw=S?z_tjdeIlZFx5>DOqxB(qVoZT`ulw$+{OizdMIudXKqWUqEHfDzh;7
z7Qh!F*NK&SpbVn$+$=0G&@Sx4>{U#(_&tmT`h~+$5|tMD5>EQuc9h8NI311i&Z_DV
zoJ;&Ibo$?4(ghkqwb=oBi3SSdxdq$uaGzXW)Pkz-{M|^+=kDcIR0`;+bSiKU1GovM
z9~t}x9D{77H)JN955ng&Yk<&lKn_EnI(TohFrhB)K5t^4wz0d<$$B|Oshr>u(BB+J
z!;BvX3c!InUqU+o=P*RX;UXd$j`OHA05sAStP!xp{c$!H)sWKe0rc~>Rk2JOE&nPl
z@hem;;(J-^M>!%au=NN}g*;J7?D622mr-Zlw=0bw18D0MU`r$UDF`Kb0_8J<<Xl*%
z`Ul2|4JQk(fKwdL3<@t1dLt+v;F?+p^ue{qE_%8r2zJ3U3xTj*{hh}+qoLJ7K&)C_
z@YF_|dmmCj&};C&{u%SuP7v^8-WmepC+GE-=B?dVc<Cz<Ay_1TfFqiZ2q_Pe@R&k9
zF-qdRbi#;2uy@;5>5X{S2Tt<RL7a+1UFiZ2flH&(0}+j~?%`c2Zv=3$=+ZdI_yaIb
zhH3l?sIuNqvV$;D<66Y@3yYeVxb#rSc%ogGV&=b2<7x_^WA?CYW}`eFA{yrbRUIi|
zLD4D{2GPa1OOc7-(r|uLE;^es2qPX*OLN{rTU}P1rsN1bq}(RQ06`F~!EH4R&sasM
zu&Hn8mZA-!C)XBDZDEGF!4byKRf<B7N~H`a6e$a!<a?V^9i%DBIL|z>C?fc#_Kx;G
z6aq$O?w1zD)-WlTaH1DKg!EQ}2VoK;mc`4fec#3cl}C|G<O4MD7)7Tse}Bxg{Q4_P
z6Mc*km3N`2_NqR4X7Od0`j^_%ZD@*l{tp|P+YW;hI=7v|yrM1GPi0}SqNn$g)-bi{
z=!)RV{=|f3&FO7T9zHb$!k!AtK>1aG=770QWfdk~3h~5mV9z|Dg5Y3M+k=)N`hALO
z^B(o;Q*lTI@%f?(EAEpr@q~&Y6>}ZsDOE`~c}B$|lyQN!$GBwlBlQr0)Fw1>)7XJX
z0QEN(MY606^H_>hkS>X8KLi0TYA1S5JU7%qLeiF^uV}JnFVvh_jbfXD%Nddh+MqV@
z&x6^*nAXwO7&M8ud26sTSS={6Kv-xZ5>hSHV~u8;Ke2$ER^RAr(sNq-leL$2?$PO(
z0#;}(4#;ghmA`*}rT@2oL%)B21qNUwfUScl4@G+&240N`IL;^XK%ru-S20q?GAinl
zp$#1pWpK;)K+(B8MeK`nxf6~?hzS*R5$s`g<(zB@LZ$T@3x`Pk5QqSi(h$yTW^x~4
zu1+hAa9l+wEdo0Hb{vcY3UxfXKMi{})Nyx=Y{4Ao#1Al4C3?qr5rOGCpol#DDvAE)
zqUfJC!Cd_0l<AV<Cyxjr!J!SRFvXOt=^){eJ_xcg9Dq6_v~`8WOKM4S_)}a$9)|Sg
z6VoERKLHA){KCR+D078yNk*R7522w`K*>N;AmtQ>_%FCyN4i^bq7IxLa#qYZvN$W;
z^P2cZ?#P^zOJ`n;)TVCXf^tE5d2Q+!-n339vNIi7AoFFoS==q$P4Qh)SdNvZP2lhc
zgu~UE1SuES{Ap3Uu;x$ZYu2z5@>bPK{)LrntWd*B&dc(ASD`sh<IKeq8^Gi3t6+)Z
zr_xnnppYohsaG`u{m!K&KSzxcQQEE2qI38KmrId&>9FWQ+>Q9RP~jBpmrkAqyni6$
zPo>Vj%<Mo#k0Sm}>UWC@=|C=e;d+)P$10=}9hZCvEq<5!O3b4uth63thZ5O39+7g_
z&3Qm8hb-z5A`eU@vtHuQF@<;@1sxXQZxPHk^p>$MQI3)*gm$x*Ef@k0J6nqea6Z}m
zkU{!6cf2_&r1liEiB_&7?Fk%X#;GA*hVT&W<`(=1(wqZ_Sx_^Xw1gr-X~=uz-T+Bc
z%-}Qov2lU~GHN;rdkZO3Pz}N}RlkPp`4;fCp!_(JC=C(y=};%#G^c=F2Fz0QFtXo9
zLj`Wys}+I#VpBy>Qh4rRD8fAF{8<`IkutwYg%qz8xPG^E$bes9a)HtWFrf$$B`7N<
zF)e$TM`-&iTtXcpTHx|6z{{;Cj*?1u==7?;iye<KrfuB4+uq_6kfk34INOYoO+{KW
zpd@3Wz*Hv66cB$I9XW|aY8Zc$sC!h<frx*Dig&4yvEzs+5}9;@{5wPuXO!Z)X%_T3
zQbxZ^P4p6Fv&!{#qa+D}^Y_GN3EsKP{dl0r(gbOuXB)^z%>Jv_TXjQkv^>jmJlnJV
zn&*0#J;!f(P0vJbZ=2KTF0T2{swnX~i1SYm%=n=$KRB4odgOOxf(>bYOMOx4AY!NQ
zC>@8@Sp#=`oUri80h-b**T2S~CUJgzH|8S0pHuFivTqMQg#E#z^}=o>QJv77mM4C$
z;8F<^sAmAzjV}T?l%6L@O#Vmc!q=!+M^PQ6RO=mTC1aj(N?NI87xn!+T<4jQYftX)
zJby`hp&#vw_D6D+L!}w$`%);gwWT)B*FKk*5)|ph-_vk8r#{(!%DPM0jn80J+VDU8
zj8&KN7_=${R`Ol1f*^?Dn+R+Wz0<#M_7J^S8~}?+68|qAl%7o0<HYli7*gdhD3dmM
znh<7v=h*idu!F4epJ?q$2v?sXgetkvu1Or=-$Qr)eJV(IE|~WpQ<Wy<e?W!QF)C(C
zC-v<Q(LhITd1&|ab;~pHw>h=9H<jtHjgAp*A#0aVij(5Q(ikP3OR{k{O?P|aR2~Jx
zgp)i<|3bs>$p)Y#eWgs|0|`f+n@IETD)<t+45|OIz-emMKslR4ev(mNVxC*<eg5f^
z#$Kq&)E>^Ym>XX<6ZPR{<>FxSqAu-d<=*UF-j*OLk*0L23cdX~!xTmAD!rH09)FpP
YO#*-izlel3P6!PHA0y25?`W_85B-r=EC2ui

literal 0
HcmV?d00001

diff --git a/ldm/models/autoencoder.py b/ldm/models/autoencoder.py
new file mode 100644
index 0000000..0b605f8
--- /dev/null
+++ b/ldm/models/autoencoder.py
@@ -0,0 +1,224 @@
+import torch
+import pytorch_lightning as pl
+import torch.nn.functional as F
+from contextlib import contextmanager
+
+from ldm.modules.diffusionmodules.model import Encoder, Decoder
+from ldm.modules.distributions.distributions import DiagonalGaussianDistribution
+
+from ldm.util import instantiate_from_config
+from ldm.modules.ema import LitEma
+
+
+class AutoencoderKL(pl.LightningModule):
+    def __init__(self,
+                 ddconfig,
+                 lossconfig,
+                 embed_dim,
+                 ckpt_path=None,
+                 ignore_keys=[],
+                 image_key="image",
+                 colorize_nlabels=None,
+                 monitor=None,
+                 ema_decay=None,
+                 learn_logvar=False
+                 ):
+        super().__init__()
+        self.learn_logvar = learn_logvar
+        self.image_key = image_key
+        self.encoder = Encoder(**ddconfig)
+        self.decoder = Decoder(**ddconfig)
+        self.loss = instantiate_from_config(lossconfig)
+        assert ddconfig["double_z"]
+        self.quant_conv = torch.nn.Conv2d(2*ddconfig["z_channels"], 2*embed_dim, 1)
+        self.post_quant_conv = torch.nn.Conv2d(embed_dim, ddconfig["z_channels"], 1)
+        self.embed_dim = embed_dim
+        if colorize_nlabels is not None:
+            assert type(colorize_nlabels)==int
+            self.register_buffer("colorize", torch.randn(3, colorize_nlabels, 1, 1))
+        if monitor is not None:
+            self.monitor = monitor
+
+        self.use_ema = ema_decay is not None
+        if self.use_ema:
+            self.ema_decay = ema_decay
+            assert 0. < ema_decay < 1.
+            self.model_ema = LitEma(self, decay=ema_decay)
+            print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.")
+
+        if ckpt_path is not None:
+            self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys)
+
+    def init_from_ckpt(self, path, ignore_keys=list()):
+        sd = torch.load(path, map_location="cpu")["state_dict"]
+        keys = list(sd.keys())
+        for k in keys:
+            for ik in ignore_keys:
+                if k.startswith(ik):
+                    print("Deleting key {} from state_dict.".format(k))
+                    del sd[k]
+        self.load_state_dict(sd, strict=False)
+        print(f"Restored from {path}")
+
+    @contextmanager
+    def ema_scope(self, context=None):
+        if self.use_ema:
+            self.model_ema.store(self.parameters())
+            self.model_ema.copy_to(self)
+            if context is not None:
+                print(f"{context}: Switched to EMA weights")
+        try:
+            yield None
+        finally:
+            if self.use_ema:
+                self.model_ema.restore(self.parameters())
+                if context is not None:
+                    print(f"{context}: Restored training weights")
+
+    def on_train_batch_end(self, *args, **kwargs):
+        if self.use_ema:
+            self.model_ema(self)
+
+    def encode(self, x):
+        h = self.encoder(x)
+        moments = self.quant_conv(h)
+        posterior = DiagonalGaussianDistribution(moments)
+        return posterior
+
+    def decode(self, z):
+        z = self.post_quant_conv(z)
+        dec = self.decoder(z)
+        return dec
+    
+    def trained_decode(self, z):
+        z = self.post_quant_conv(z)
+        dec = self.decoder(z)
+        return dec
+
+    def forward(self, input, sample_posterior=True):
+        posterior = self.encode(input)
+        if sample_posterior:
+            z = posterior.sample()
+        else:
+            z = posterior.mode()
+        dec = self.decode(z)
+        return dec, posterior
+
+    def get_input(self, batch, k):
+        x = batch[k]
+        if len(x.shape) == 3:
+            x = x[..., None]
+        x = x.permute(0, 3, 1, 2).to(memory_format=torch.contiguous_format).float()
+        return x
+
+    def training_step(self, batch, batch_idx, optimizer_idx):
+        inputs = self.get_input(batch, self.image_key)
+        reconstructions, posterior = self(inputs)
+
+        if optimizer_idx == 0:
+            # train encoder+decoder+logvar
+            aeloss, log_dict_ae = self.loss(inputs, reconstructions, posterior, optimizer_idx, self.global_step,
+                                            last_layer=self.get_last_layer(), split="train")
+            self.log("aeloss", aeloss, prog_bar=True, logger=True, on_step=True, on_epoch=True)
+            self.log_dict(log_dict_ae, prog_bar=False, logger=True, on_step=True, on_epoch=False)
+            return aeloss
+
+        if optimizer_idx == 1:
+            # train the discriminator
+            discloss, log_dict_disc = self.loss(inputs, reconstructions, posterior, optimizer_idx, self.global_step,
+                                                last_layer=self.get_last_layer(), split="train")
+
+            self.log("discloss", discloss, prog_bar=True, logger=True, on_step=True, on_epoch=True)
+            self.log_dict(log_dict_disc, prog_bar=False, logger=True, on_step=True, on_epoch=False)
+            return discloss
+
+    def validation_step(self, batch, batch_idx):
+        log_dict = self._validation_step(batch, batch_idx)
+        with self.ema_scope():
+            log_dict_ema = self._validation_step(batch, batch_idx, postfix="_ema")
+        return log_dict
+
+    def _validation_step(self, batch, batch_idx, postfix=""):
+        inputs = self.get_input(batch, self.image_key)
+        reconstructions, posterior = self(inputs)
+        aeloss, log_dict_ae = self.loss(inputs, reconstructions, posterior, 0, self.global_step,
+                                        last_layer=self.get_last_layer(), split="val"+postfix)
+
+        discloss, log_dict_disc = self.loss(inputs, reconstructions, posterior, 1, self.global_step,
+                                            last_layer=self.get_last_layer(), split="val"+postfix)
+
+        self.log(f"val{postfix}/rec_loss", log_dict_ae[f"val{postfix}/rec_loss"])
+        self.log_dict(log_dict_ae)
+        self.log_dict(log_dict_disc)
+        return self.log_dict
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        ae_params_list = list(self.encoder.parameters()) + list(self.decoder.parameters()) + list(
+            self.quant_conv.parameters()) + list(self.post_quant_conv.parameters())
+        if self.learn_logvar:
+            print(f"{self.__class__.__name__}: Learning logvar")
+            ae_params_list.append(self.loss.logvar)
+        opt_ae = torch.optim.Adam(ae_params_list,
+                                  lr=lr, betas=(0.5, 0.9))
+        opt_disc = torch.optim.Adam(self.loss.discriminator.parameters(),
+                                    lr=lr, betas=(0.5, 0.9))
+        return [opt_ae, opt_disc], []
+
+    def get_last_layer(self):
+        return self.decoder.conv_out.weight
+
+    @torch.no_grad()
+    def log_images(self, batch, only_inputs=False, log_ema=False, **kwargs):
+        log = dict()
+        x = self.get_input(batch, self.image_key)
+        x = x.to(self.device)
+        if not only_inputs:
+            xrec, posterior = self(x)
+            if x.shape[1] > 3:
+                # colorize with random projection
+                assert xrec.shape[1] > 3
+                x = self.to_rgb(x)
+                xrec = self.to_rgb(xrec)
+            log["samples"] = self.decode(torch.randn_like(posterior.sample()))
+            log["reconstructions"] = xrec
+            if log_ema or self.use_ema:
+                with self.ema_scope():
+                    xrec_ema, posterior_ema = self(x)
+                    if x.shape[1] > 3:
+                        # colorize with random projection
+                        assert xrec_ema.shape[1] > 3
+                        xrec_ema = self.to_rgb(xrec_ema)
+                    log["samples_ema"] = self.decode(torch.randn_like(posterior_ema.sample()))
+                    log["reconstructions_ema"] = xrec_ema
+        log["inputs"] = x
+        return log
+
+    def to_rgb(self, x):
+        assert self.image_key == "segmentation"
+        if not hasattr(self, "colorize"):
+            self.register_buffer("colorize", torch.randn(3, x.shape[1], 1, 1).to(x))
+        x = F.conv2d(x, weight=self.colorize)
+        x = 2.*(x-x.min())/(x.max()-x.min()) - 1.
+        return x
+
+
+class IdentityFirstStage(torch.nn.Module):
+    def __init__(self, *args, vq_interface=False, **kwargs):
+        self.vq_interface = vq_interface
+        super().__init__()
+
+    def encode(self, x, *args, **kwargs):
+        return x
+
+    def decode(self, x, *args, **kwargs):
+        return x
+
+    def quantize(self, x, *args, **kwargs):
+        if self.vq_interface:
+            return x, None, [None, None, None]
+        return x
+
+    def forward(self, x, *args, **kwargs):
+        return x
+
diff --git a/ldm/models/diffusion/__init__.py b/ldm/models/diffusion/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ldm/models/diffusion/__pycache__/__init__.cpython-310.pyc b/ldm/models/diffusion/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..cef85efd89ec3becef69d590a447db46a6bd9b40
GIT binary patch
literal 173
zcmd1j<>g`kf|mb%sUZ3>h(HF6K#l_t7qb9~6oz01O-8?!3`HPe1o10CKO;XkRX;Dk
zBsD2NKU+UJKP6SaAhoDCKQA#yKiJa1#KOqZ6e8^G92}&dlai~S3sjU-te=vZmR4Gv
knV+X0AD@|*SrQ+wS5SG2!v-i_nv-e=vZ<H}NU$&f0PJun=Kufz

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/__init__.cpython-37.pyc b/ldm/models/diffusion/__pycache__/__init__.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..24deab1a4f939b016ea61eedf10db811b3ac62eb
GIT binary patch
literal 170
zcmZ?b<>g`kf{4w+sUZ3>h=2h`Aj1KOi&=m~3PUi1CZpd<h9ZzKg7_7zpOK%Ns-Kr%
zlA4sCpRJ#gSdyrpUX)*2pdVbC=aZS2o}7`Ir=OgklB(~LnU?0{8Kj?+lB=H!6v!#o
nPXP**7H8(?>Bq-s=4F<|$LkeT-r}$UN|)xO+JP+l48#lozLzV_

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddim.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddim.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..b5d2fce8af3e6e52d3884444ef1b621ec90a64e1
GIT binary patch
literal 9347
zcma)CU5q5xRj#^KU0q#W{WsI|KfBvI{*l_TcQ+;`v03ayyIFgUJx;uvn6$hu^>kIw
zbWi{7t!jIAxCO?u(QXJEoLE9Y5~(KwMf<=55=ej)i5DJugO>|JLP&rk9)N&>O}xu@
zZgtP}Ol*{H)#-cB{i}QH-tT<pRQGc^N5kjlgKq@Cb3xPokuuXC6PeHA3SWZowVLMZ
zKI`Z;UH!6}fnTFzcFmfpQ<>RGb*-A!P1n+0yJmMYwT!MkulXt8x~chA;OrT-tk14!
z)%0%u%I9AG!b_LepRc^K{?b>@ypIvI^{Ot-=(^vnn({=qz8!eJ-|Bi{y*uc%dYj&$
z&g<PE3V0}I4~wH#HwdF(5K5=lZ-s%^X>A8`HrR=Hy%Bk>Uex!ZpcnReg9e%Y&;;od
z{yjvXVPaa1`75`yCMN0|H#3-`>8EaLHPc`9Grn_EucZQOl-}1e+40xSq*W?6{<V?I
zXL3oYf5I>L#hdz`UbDyT+>DEI)!IMlm;4!3nmH&<N>2H+ep!_`hvWGbf6kv*+H9ir
zPx}l0qSCli_163)e_3gBw=`3$<+0=s$?V08mtVd@iy3ffVMT*FZy0I;5<$zsaAk3Y
zYY>sPr;YS|P1l4jG>-|rZQ$1w3^L_wSG2qOdexGta5zApwujYjMV=>9-M$}mWUe>t
zdWX+v#`gzZ@2JiW8(V=t>;!n`W*Bsu+{M7W03ogIx9WT|ysOKM=QTR@F!a2;+P#;~
zZuPst*<L>iHv0YTvki=O7UPHgUcGbn%G2jQ^_eH1ekxI4xNzmwvmL*CR^3o|)^9bN
z!?4xwou#dHW^jYkWYvUiBplt+ZQap#%TsH0W-_fKi$x2U!gT>x_$)-E3AU$4dYf%C
zz93lJi1ZPT$iBX9P<~`i^80$sqLiRTWqSsW4i>V?)}`4P`gJU)8QmBJG8JLRZb|E^
z+Pl*7gEw1^K(eSW(_s+Rqli}xUc|`K?BQHUlT^G6MIz>?sKA5G7LEt^Hipe+z)Mhu
zG}XLbRNvi|>8T#8Ez`qZtJ&w>e?|-Sx!RyfPd>D88C<j&;rH}p5rlqQYwN=F^*iQm
z>>A{Zwi%~Hs-5!Ll18Q5EL?t3^Nl_I9bFwaI%alk-PT5yZwdUOO?pq?XFBRbmll>t
zP37%z9eYZb5muBLIU+U6?rZxvbO)`Sq{PoAWs65kT~+FSOlj_D+m)kj^T$j5!eQ@G
zE*THAVZy+Q6z`a?8*w_e<BXbBNu&?%aYon&=~<CENSA3&xHp^nC9N=`JTwLL3Z42j
z7qyjfinJ@*qE<%jvKBjPoH^l8OSRGF*KyQJ3R(SD*x()@b*t~o89JCsJ#}Ie9Zv0`
zF8yY}H~L|~@wm0!Pt-eut$JA5s=pajqJCu~sQ5t>*y2~PZz@5(u~nJ0ZBUEJhtjld
zW^silA%r&4_7&WWScDC;ua9W!q}tZLensQ7&8lgcYIOR&fKEe%uK5bId=WyXo1K0=
z`ne|){8*;iy3}{>>gtG0oe2AUr{3N0>(8#DPDprn64RJI1j@EdaOuqXPQOv_gwG<8
zw3Dd~3?9mSGIg&p><%~<aQ0~48}Q&wIU8Q*k$1F2o={56!|S$s!|+(~_^kcgb8@xQ
z-}FANnp{%7@SxEe9P2@@92Rvyd93h#T6t`NjJhe@JZ5t2u^X5le+1QJR+WUU%`PUK
zWRkhaqAD3rOigt;sY~-ce$R4ksvi#<!|U_bW(#<&%z-lB)dD92#PcLRO+s0MC+Id(
z;Z_|8e~Pq^lQ=_K)*HZPU_5Cw2199JwK#1lwIQn61RImN(LPqA)C%fO)Xop0)@~qe
zrI^f^tMP2&3<SUYY3lD862AaZTT7PW=oAhJiW(u=YL8J9g}}KZ+i3y%@G}q=nVJVS
zThv{>%!+zhUp7wa^j*_!!_-aPz3&<>O6O5?M$hZ)f%Twh>@I!mR$afuKZ%~-_aPcI
z(BxA=!ou}SxWc#b`~gMUQ0D<`7Oc`9<4@t233FuZYkQz~OWKkaFQRlYVkkW=7D;!g
z2^Ym;8vqa8l2{T97Vc9Kz+EhT8=x-Ox3%w@d%CXm&Z!Zok1q=IqIPXIK4__0z4B45
zuAlFfg_-oPY6Ys+SaWm)p|vg8(Tphi<{j8lVE|y4<CS<7BxPBwh}EVUpFsPRIDwXy
z$V~1HEz`mjCz`CK!Mvp2(|uc{Av3~;bc6$$6<J8v&q3w|$%1epiy{wM5(UT^QG%Qm
zGmvF53pppskn>^=azV_$r+-6#oy8~PQ{tpJ)g%JExSqgg?zKAaf!jV)dA=W2c28Ft
z{hr^7Kv19zD;vY8vQdv3TTg`8s+Cq)fypPv!7q}aBf49;SdZ!zbwCv4()OrKwj?##
zwH~ivdsn0l9P4b<8{5)tbvM0(T<Q9--lMbN`GKO)zBF;(w<X&-Cl}Q57=l+H99gx^
zFM-`Dy6yA+pg)XcHVV2pqxERW16c?geI9raz$cs}Sv*L339FOFj(1gN3E=`*lp9_o
zSBJfW=>r)yhb_O}YXn}1I}POW(b9vOSSXmY8=yd~h8Jz|Al&MAeCZAwn`k*Em$r0T
zVXGHL)S)!}RwI&T2gFF4+rf>H6M>Vd0f)E1FHt(Draf84Y9#>^{fczd^YG|-B(2=(
zJK-;&ajn<`WrEG{!r=Naz=-@Zsj~;`ymO8d25>r%oV+Xk3W;?RUm)>C67+cat0b<F
zxJu$nB)&}IH4;SUz#F6P02sItO8tV=Uy%B1z(vBgw3<K=${Z#Z@NVF@>QN9LCy7>4
z?dQ>1*nzN@9pKR_h_H=c_O`=p;18wP{d{U!FEWP#we&nQfl4N0_f3O+XqxON>PoR6
zn-;rgrrH0RHv1nl!~WZJ*av2o{g>&o|1@*#M`oV=hgq=y-7f9UOygN{BMl<4lg~p)
z5X+@)AAUgKrM1Bz+bk{%rXXle7=SatEzt%X`?AmOuz0=)&K1uKOO-t=OkP3D6tsYh
z*y^pR5}jWXW_T$|M_`mx(#}wyU{l7Ps+r`=1dGO^7B5g8UlR*o1(S^}F~6zP<(uDO
zfHsqVIdY<`uM2Ar&OpzM+!6RA2>%kwJYn@rF&-gaj`9GdWdfz|f(PqbR1ix5qGjK{
zW5g>_QLOCgfXrV5bgafF0A{OTdvp&cG0&2iYwF^JC?h@NXYQB)7h8e)_eF+&!{2^U
zdk0=yTN{B*ic|Y6u84}T%Nn2!yiv@TwFM39GpkztSp?{U`%)H5d&JKFM75q3tC(dG
zz08Suz$^inDbm6fIgu9yQ4}RH<7e;W;?v?Z?sGm~6Dj^TA|+O7M#ABLCmgYc9P!qZ
zB0I17)ce06y-ag~tMmbki@T(a=EeLzwORCYDz`?t>u+J)MFpAp1TsPBh}`j~AgWBJ
zcgVws^O*(*@x7gM_f}pV_6R*I3J9-^g9nwjTG3YJo^wD(!(I2VLcs4bph6Hf$nA>i
zox4npB2zy3V#=&m9hqwN{9s4U{4^T!7wMip>JT40wcAvW&RBIJ;WvuTs2M35o8tAJ
z-#a8h(!lhj0iKwkjSW%>S2yq_)!YpDbLdN&%}#5;U#Gr_(8^2z!b4aLtZNQC9bjD_
zFlk>;f++#GQqYsR!8ogEzqIOuLD2K74~?Bl!ZkH$f<Z4+?*wl$Ga>Y0GVMbYBbZCT
z>StIA1z<FDsc&GK-OZX44maT1dbkU|MFkFx<f)0mzR-h7Se&$5c;<C7EP4xAhwu-N
zr~!*Yl)<=>v63e2Qm9t@7~=73sJx3SBzSdf8$egoO=hzh5U$4nc6qi6(pAK75l}a;
z+mHl-Hp8WW570OMTKDWUl`b;014!EJC`mhFN*WZ}n0y{Zf}C+;t>Kf9n+OWDweRVW
z`Zx7gwRiP5^tccg<I-(LI2O(cmrf8ri8IMY2F|2$TiZ9{nclmRIZ6o~Sj5Qpi)U5d
zILMbHc-2uFRBRh5Myc=_I^{@d@!V|<So7G|bgFe52pHL@`T8yGmVS#h4IhE9n|73m
z;7?NN9=z*|kYet(eqOu&9F29GQAtw1$@UM*dS}$g)Ze$zE5=geBOmUn@l~%$kJpMo
zq`6Tx=}*j4e<Z0tg$Hl<(&Mp|E}k2&!>lNaxqV~g`ltycS^yd^0H+|y7Y5SJP|bGE
zPhs_bpjzc=^~4NTZwXdm1{m!qEd1MXO$$i%SBO#|4gn-}+6B^s?y5C!rdpcF{ywN3
z+FAS`RGXr*E~R)Gnjb37jMB`;E71H%Y06472ZCzxpC}DkDy_W$nrgNeMZve}+cbgr
z=8`5#Kz9R}f3$36s_fLUvJ+Efm1AYAQ)Q>s=(HlnyE?y&d7TzZWyB7|$<fk2Y+$_B
zv#HK#`Cz5iL>hJ9P%T$Rs{$zf28-9QLqUmY^-$W3PV8Y{!3v&2UlpItqEG6PdXFCx
zmFOhxa@da4o>A7wCdCiKYB}2`T7MmPe5ze(pBALc!jeBMPN~*wsQb`@CLYGqd_+7V
z)|6dv*AuUYfbU3zffQ+6D#Aei7!H?qtrY7Kq72Dq6$cxZQnbmw$*zN+E4Zk#B(h*M
z>JcmrJ9Ccz270+`NWCHT?SzUEQ%U?h=REPrM!!3F_Aj9fT?k8K%)IY_NICa+pO_{(
zN1%5)#xUta>TgN|QSIFV0-co>MQytQES;|`A8*sRry(fHF{rL3K&|Wu4J@+&^Ue)V
zg+0|Y_`B5Nmq~nuM2!SNAb*?0I}ix2Yz=$cvUm_^Z|&B}7m<r(T)~+=+>qDg{jRs`
z!A^N^N*5!vnm53-!S|K(oOsLRpevV|pQF@y62C}7In*(wJ_{k2Cq09jBXDq}?^3hH
z!w869-$bNDMU?Fcg_g7H{pj-Hs(hYv_!=qdlJ$DhX~J8eTd#LydhDT8A5B6*RT}&~
zs;Nhb)o^(*2<nli92!2P@}Je7e^vG4C2WBfBj|XrIs-UFCQ%avb0KSH15#}twPL80
zcr$oYqME?O2k?Nbad0I&6<4XvePpyqTH~mUQTL*nRafDrT3RWgC>?JUDLYuR{1*3i
zBAHTimDY}Gmzc+I(Znfc!G9g%k8vq)U^!KUSu7$(;lR96fSk-+p1Jo;lcE^zeRj{u
zSkM&_y<iXW20R4aMH@1x<VVaPP2W`*Rr=<QdGu$Y%wd+kh@68`GOR}p=aECwpa|gP
zvw%bvmvRd3tB3|iM7^!`NWwW#jzI?Sp4ryJk|MK4?0|$njB>aO!chc1iyUGh<DA<o
zA&(aulhO$9dY}QZPBbv9uF&J$rWWUM704h82d)9$7TsYmpm*b9WVN$^H+2V~=tQ2~
zb`IJ+A{1s^g4;nRXoK{l0*G`;FlE3$A<_h(;g&Lr)=0G+6>twF)XFu1PAXjzrGs=@
zFiP(mzO~O__GV$`X5q*{9+<tXD8h`mBv-W23}841z%2`<r#zevm4?%k22dxOas7#g
zdJ{8fhuDUg0{|nm&Hoy+L%4leOK5$Prxx29(s?)@IYsFUNEIp7n?b98qE^0zT9p2X
zTEgc4Dr`|C7vlN_%pE52J8&-0_Qzs=R7Qj*-%@k{W%J4iIOBLq{__YS_{9D%cYzvJ
zoe!`}I441|G5%X565pYg*#Xm}w?}?$X4*eaIHqEt)wx3|o<J<c2UYkn!8t|Rt>pb;
zf_G$g5@1y*M`)aQEWb*r-zD*TBorQhmQo6Pt>M52Ux!;zB;89S2&8H$;=>JvO0tZQ
z-H`VZm*>EnKXfrv^kJ5QojkzH9tu5-33KK!0y+Vg9=#&{LmC1opfbd{Ygx4v<24|I
z$NzxT<Cm)h*EI)NH36{fPUO86v^KY3nrgF?R05@hExt5>hsXH0w33DY5Iu))K+GF1
zVhJV)G*Rv{QWOC#>lI)EC^>Q`LCo0$rrY<6#v*g{M^(;51eGv|O3Vi?aKuFqWN=X%
zaSIcu1XMW^V@QTR@*az50@)4)CuA9g!816Z8s1s~P7p1~M8viM7JvZKrQrKh5LM$V
zC_xAkdNV>85%CATZSR4Zm%+Ai6eoF;Sm@+MMuwloD2Z*PEu@TLvZKgUN=LbY>Qy$U
z$gSFY)%4Wfpofx}sz0LCVRwp=C_zvzOJUS+;0;ZL=vXrI#{riiRUr7bvAB>}P<kC`
zKG;1m%~F1<ZG*g($!7)$2bWs4zd!<v)+Ta=^#Ld;R%^sI#U+HH;Fd~Fyywzm1X%&J
zAYHK&SqL0Eh=Um8I7k}2&jduIz<U9uv5VMDR>eY|0L10uJU}OhSO_45Ah_*d#Z_N`
zTX6bz7CP5Y18fOGbAUbu$SL4im?DMfN*cYm0OB;ihFC^U!OS_~A}pIpV1#!u8Qz1~
z&0h@0fYKsB;e9G0{7{%uH5GwKp2JnrehH$QPrPyq|5qRvJx>LnCh;pIDAsuBt8+nF
zwFeYlEK^oN1sUP$jC$lXd&lbl9T1(|gI1;$E8@{qOmIl;==}4WByN$oP2y`1wbTw`
zyA4G^_}57rk@!s#-zTB=ixO(n;Po2e;(6LBYbc*LakSJPp^zVpJB}kfU3?w9q+oBG
z$BtOi9Oz5i1tIGSjd2>{YISxUpGzvUo-p@u`szA=o~jW{qe`{3ehC$kI#f@$qAJd%
zOUWybqXpH}rKD$6VVCf_GwBnl0{usUUY6rO3XVsd68}6v+VwmT%T7|F{t}S9Yobta
z;sm67eQ%T3ecmF~H4^lE_@9#aa}qy*kh$?66&gXO<9SodN@peER)jTlYQhZ&8}H$5
z%dxCeEHw!y{C~gh53LJ~9oBH~Sx--L?uXV3ly)CjXI$MXP`dnp(|A=^rajN^H*inN
zVx?dX@aG2#9wiZ@)A*CbnPfe}GY9{Ym?(x(tCPH1Q?KX<DithAtH04kfL{$VHQU3T
YWS>v^cothJ@rg{$MmX6;w4`kOAMjj9h5!Hn

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddim.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddim.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9adbd777f9156c20c6e24a2727045ac38d798726
GIT binary patch
literal 9245
zcma)CTWn*;c|JED9$rO}6m{KQuI#v>FDo~3oiyD#vAvePjV-6%t?e*&4k*r{M2R=e
zIb3ViIYDA=>P_M;Hhl^Z8>LbpK#QU%+B_6}=tI!H_^D4tG3Y~q1}OX%q%T2$0{#9O
zlA>fAHHq_c=AZkVf4=`Rd?uGm8~D8b_od)R&l<*mQD*XSkoXFo@I?@2)D31bt7Fzp
z^=s8_{MsF->(m{S%A8K3>(<?FvYzaEb+4PMr%dB{gC&@I$6#)dKDO%_X5BDq$)o&@
zr(gQoi`TcFufDwX;@7Wyh!Kpfnkk*=fOTt*ywGjz2R>u%t{*nKgHF4*;}06V(G8-2
zhq8299JRYa7zKk+rhEN%82Fv`ejrQ1VZ<BF$Zz+ez8?j>u+N(`$mCNVC&KrT2n;OB
zs9S9H9ixRsGW$*n%X3)bj!}2m8cVVC9kZSY+)?txU>DdNE8H=U&AK<{Tx9dCs5q%H
z=MpQivf`x2oGM#j6~)PnIhWZYTT-0tn6u87*^1)i-Z31bp2ucAAT!slU4Q8YZ9~AN
zi@gXMylIaYKyS&ylfe_NgG9!$F)~jK(-5XGd`p;nHhvvpfhL%NF3qi)D-+>xfc`um
zd(@76UnaVJ7Ib8;H|+YSyIN#y(Dl#qys)_&uwf^_n{vXS)8aHb&x1&J|E&h!3Gpsd
zzTfOL!qE5s()i8G8@v5(u+i&B!FIpDzrh+&V`GQ+hl7n9!`_Q{zv!~J(Zm!suC-gO
zXRg1p(P7;UHK%Zcfi(=<{oV%2#FfEqPI9Z(>>**8Y11`5^Qb(tRaYh}o9-3uSpv_q
zc*3VZB12fmW@PSJ`xd_<tUWt2N093ibKj=?$eH9%%-D(&g7(xpwjuAdkJgrSnnTvW
zZaUHJK_C+m#B5i(H<iFimj$=l%|KdFUnavKYD5vQ*?bNoOQ#26mJYFa5e!XAR5Ztf
zoi=2e``g1-E8z2BhI*NKvw-i>>g+Pl)t1R&uifhN?ti11`rK^NqUl2$mcm0D5&p!~
z8z9WP#-1r0X5Mq|LST@y_nbH(5_<_|EgMw2Z-q}pF0+r#w@r0M=xlhgd)F8xnJe&%
zHr}y$VwtE9UQ)OsF_ZVkby73DlyIZ;C?gW1?1^!L^E7Rp)+KCQwsf{Mt4ed9P?|g2
zcJ*xA{P|KgciMZD*K@&YEMa3u3iq5h>^K>FaZ0UfUL>b;EDCR$E{W7MU8Xf*ZXWeZ
zTwzCZ;LM@d)RLiooh4&+oFeY3v1F7{yKKa1HO_)aQ%fael`WjIc?E^69X7cS4BYKA
zS)?<m*fWP!^Kf8KdC6M=-|mM2$LltZ9%*z2yN$5A+qe}}qkeTesIs61NMTjTOf_gU
zcdL`OO=>|O`X!Me=?pJ|2xDZND3lpl5rpN$9Ff>0_S_Rw<4aN|nw@?xpfeDmQ@)I5
zyaFPVtxmrYeg1PAG?u^uSeS=3Qyq+%bKqU=G`icY@zfgX{0|;Xe^&675|}{Im8+e8
zv(X8kLPEEaiERuW%Di5(-yC)a92-|UoA(DixFt*B0gwE%CGvt|VikV3-5ZAIig##6
zhVjcU%e78_$N!{ia#{7lgJyeht_Qh#TGajRbA=z$uH)6G)TA(ZtmNG5wy{3`0IJE1
zDhb;=T`X8<^jc&=m5djrmb#+r(t6L&S+39Y<3k|)K5y@|dyS6L07}o+3Y>Hf&k=co
zhywpdXc~!dw*eT;KOI8eBEa=?ocB4uK(!tw@+j4^dIP8pj4$oxU?^?u9w+fq;!(?J
zeAH_tc~+y==hV#92?(P0Q6N3VnCt*o<9np+34-~P)ZZ6~d=aF+uD9mwlBSl08bJx(
zL<kiQ=gtV&BKip*2Qi>(@=!KQX4WbLjm!3mc~O1WZO?ST$*OM!^@?W2JX-!lq21aX
zgY+-I{2@cnO&WwgRXk*(d=*dl6eIy)WbCK-2hdp1Nyip%;nxz*$UZTSVUaBx%SOC(
zSV8$xWTE`3SR!t`bnw~Nz}o}7BeyJ;g$3RAK?G<Q%Re@lEvz3K?>on)Y4n)tom&6o
z@O9x_Gv0hCUYfL3?QVTsyMtS=8NJU6M~^Yn|Cn=j6k+VS(4r|(V9q^gQegvPSK`%p
z4Hn9ZSQTq6C%%Aj65;|{x*|21A6h1bBQA)9ND5D+MFwgnC$b_h=ES@xijpXc1yNz%
zz2X~Id@;TxE{aPnvN%&)8vb&xHFyuk#+Rzk_oM33<!ZCvV{MXDFvIHhFsg1hqUP>n
zAw;a&4y#btPvaf%XNb^IJE~rT=~7jKt{{_MZgnC`)a1y0v<k7TZq^FY10;2}8_j*0
zZFhJ4X>R^t*yxcM`z%mqE0Ye6?!L4RAD4?tvWKu`r>9RzC`>bD;WFMI^oNnmL_rrO
zTq7FtK+c8DJ`a2lAPvrdEKHNWHp!%Y=--qX!m0o|;kF;iwPA0%0XXA3!!~R5nt>l;
zqJdmFTRN?Yt%5?j4Fjg#^rKxKguDF?li6W&2Q9~zp(oSru-ywI>QFkY-HfEufmtJ+
z{or=Ue-C>p69WzdonNDLPOW;fk+rIZ0rQ4Tt2g4)8_}(@voqn(qj9~^gTVvz_ru^|
z7+^&H0<kmGZ9aUQ{|-3(4~WpI;#)*sBJwhkuMwda%>R(cD@1M(xk==Yh`dUq08)3F
zQFj3J+76}ptTbPc=4(JSLZzfyKoH6t78dYsz}k%{2+x~%uC7KX6>fqUD;`YPvY9oq
z`1Q<+mBw%0T(R+31>~}3+44+FeVy^I&P{06Ix5a0jUJ*o@pNcyH2G)ogju)@I8Lza
z_N=%pECmh=!Uo8A{9#h*IE-ax-Lv9K&k=T95w0qGK{)&?Xqf;P2RU|o*HlT0zbl;Z
z-6$Es(xZ~S6!i%!$3BMRG0B$+tn4KtUL+nf#Nu~hVaBeg?3na0=RFIM<M6+Z(ou$)
z!aat8-t(gD2zDKe@MV<!lW=<&MUN1#M0r5P3PHvDuun}RniI<ah84ho9k1Ru4_Bgs
zSUonc8TZXMvjB#*_yT}t4YnA~<RVr(FBV!RfTfIdk)`fAfDVrU_rA$r7b%?zpNvXx
z!=>6YMrFX)r4uWzimLF+21}E*Cn{xQ(ZEiX#x1`cEr=4_iv_WK3=8_c86K)0N@5Lb
zE}*B1r~nEHrW`<87GOqDRv<8AnR~hTvbc=-R^oM$z)b!Tz_&&#6lvr>006EdPZs7y
zk*OF=kMRrSR%m5#X?_JIkX<%Li=uKu?Uq<h<<=>8@GB^Wida$rnb!ab{YM5F{~Soo
zlF37Ipy4GZVL34W@bRP7SB5>p%Bq6B8)M(R`c^yItsbSPrek=NJ*^OMbmg!hY?6Z%
zH99p*jUp3_yr!ugRZGi6yT^h<S^OyY@#{3Nk2|#IPU$YyYt#Ifv}K|U549>~ekORM
z$9kvci?p#sX~V+P=;FZ~M2sO|l4|Y*{P)q9bXuMEfPaJfCU0A&0+=R*y}-8Cu+ssy
z^?N(~DwTN$I%pBVv<Z4LHyCG?wJzPpU=Z|J?SZi$NMNQ$*1$u-7KLdtHL=4(z4lY+
zB5czL^(pp4p%yJ)n%h`wcc-2XhuiQ*eawQls31)v`D(RLD)ff5(Ive$UVekLiP;9g
zA<V+N)WC{DltH19f|3rDQK+{39L(`IQ27W?NN{R+glwK!upA&+5s3D%nYZ%h8qx(T
zZ&kn{+{@!z#&5;+Q0m~T5suor20dwO1$BznS<C8-{%DeuO`nJGkne2c2{%Cz5243w
znLomj{H7Vt#f5nOu0`+`J|wbqjQE#uIIWS5!)f0&PVBgN-|RgaIirLy0a6w@YjH{C
z?P<Oo!3~a*u%7-MDT`7ekG!`{q>Oms@aK08SX4K^XHvbp0LLhWy8rNw@s9b9)v_4^
zPIo+lJzaVXhxi)cQY_pxuNnv6qS5YJ)J~W0SSQmm*k4SIjvfIGSI{#?RI>ml{58$u
z@EP5=?)S}g0M)`Mt9um{>Xj(<O5kwcJaEx#emtt;#|z_KQFhjeJ<2iE20$$WvKIkg
zpybE`1dCL255P>Y$NxmNo1@(nMeOD>lt&Rzo*o0`&&Rc01(JUic`QkR29e$)kc6G8
z-$N@QGJi@cz(t?`s@lveHB^jO6z>;`S5mxkysCJ=R=fqptH8!|?^`^pc#Dd+1Y6VD
zTNZQ7qi@TBYH*efF%LChLphu+Tb(Jpbgt~eOj-3@+1gCmWi>kOjg7hd6l=RImdgkM
zh>N3@6R6C1y{FbaTAl9Lx=5m~Uh~@Mf&hSj(~8$2xdo9|yJ+$Mj4mD<Z^IW5m(W+0
zStayIJyP%S1ELyTQnEjfz>HlsNO8pvLb;{)o%=SFzOz@|yS%q9h?jxle^6Xft=Cca
zfki_+i1+##@foqM6h?MyirXI&B2I$^lu6<t7b+y|-GQY6aWf+X4731?sSs0&MDB;y
zfpbI0z@QuyPTS2!1ifKhd7K}kmzpijrZo4pJ*FaNG{SGvSIMYs_Pc|p{)I-Q;a%(K
zu~|##3;@pt3MMV3c}v=eAs@{lKv`{5q_rDB@iDCkwNf0f7gIsFVgURm6IQFs3{o-!
zVDH@aRm4&)fq#!Gyhh}8BBUJn*NKEgZWH-q5QIv0hrNASm`1qUM-B2x<Pxb}Sk6A?
z=C^pi>mT{hUj8ka#W3yGZP?qe`jzvmy<Kv!mCMVoQ0jMye2EB|n*0_K3Vh0yNzX9Q
z5ePWb_k63(!w4#%v4hBm3JQA@Lt2)$`qA~%4bl!i#TVh$_hh;SpMj>{=*Z;QPpLhm
z13SM(ga3ePnvqr`Sso06M&v8EhWDua(`xx|seXKIN6=;j9Up3F0Jq2?qXhw3P#Ef<
zN`C6&O4_Hv6@_XN)$IU&kUI{RWM<+d^#v82`ACSw)uDvC-SDHj+fd=7dQvgLm_Oer
zQi`$evNrb*BbiWZmF}Txr(+Jb+7<pTtzUT?B-otv=Fjki6c{iP6$_CDJO$+!xQgbj
z70W@Cfix$@B#P#P(3mTvO`#nfr9+nxqflRK4Vo5TcogW%a+QvC?IpBMqh<MAoZ`%x
zX{Kln_)vU;gk3oiMVvL5<7wcT(IX0HLOBvC;CyP|4A&zYC~n7T)OrEsh+&AdGWavd
z;VxpF%l6ihCwE|68YM>_I55052ii6(%s97W#Cbe(FzPvJGm(Znf$N%kmNM?&hzn72
z4>k2XVA2zg=JxX7&LP_1#Pe`gpp#%cgFh<3;GP#2>G$5NB1t$J{u|91<In3FNEJ~l
z*8)hX^t_m#rjx>=^oh;f6AOB<1ie>+`vW@FgBekPHpvoQHAW?%aSpg!7K%@KxGE|Q
zS1Ad+)|_$u35R+UMYLO>2nTRDryPQxVSQ<lSTQEXKUW-TwGR!BTpq4UPMQ95NEIlB
zNDVm?|3Ph;i<&z3F*Sw9zmGUefn1J*@55Qhk(PW94hed&@ZKw<1(p)|wz3#dT2Y!I
zJ&v{HpF?ngk*{#Q3nYPSu?NW++y1<R^i+fJyU2~5j(X}4mML8Xv3haViPko$vRG@C
zQ~O=Ru)@|ys27ERU!asyBd)&u)0jxgL4&VCLXu<+;BQgtheZC2hyrfHNsZ6$aDe-&
zpth)i?5kAp1CV-xjOeC9E?MRQ0x>=9Firjb8&p>v>gTCJi2|BDz@-iaDaKZI>NLtZ
z0iYh;U;YJJ5K^cN8S3?n+K=%@kXGdH6MKB~NswMo1JC?1$Yl;A|E-|CvkOgDFHKS!
zZV7*ww1JuDjCSej-Tz<o9FnJD*cB^lIzSrSJ^@+ENYQu2timEMQW>J3ODJ2izNjqx
zi$D~_D?rnxV`PCtgidFp6MB&8%~TTOkQPzEQb$7AoQy~V_fP;9#2Jba8RP&FfC2Fm
zh#F8()Z<U01R+K6ohS((nXWv&V;JvcnBzE(le|NwslJ{_@y9TR){!J8N-1g$N-9&P
zm6NFMlDSLer0Z+RnfFf{uT|DhDRtVNGEJuTiJT9keiJt?5n^R}&5t6S8&U<sA@KIF
zfi4`PG@$+H!mQcyk(NzzWavY!(|FWw{WB86<pr{@urB~4W!T!WM==axE5P+o>fpx8
zj8nn`^uq3n)40*IL>iG3dmK4Q!t`?h3kjIM831nfu!VR|h8&SM>j1c1oCo0K5IF&G
z#B)(<FO9ucy#Z|VNN2#yvLv8ZK`(&M0?^FiEjS{9h)fc_WC6oTzKw`SPC?5ypd~vK
zr2W(}hWxc)j4*A5FqT3XzK;zVgNZ)fgf%rs*&O<nO;2SluibkW|1BV2KSxcUAQBN#
z7Tz}~rI2GtDJ3onK2%IWAqVN|T2V<y-8=L<01JdQ1-SL3GNgDk6HHWw4#5GBiF}*L
zcZj?VQcoNrCfrmu3%^5L^0@hTiTpW{bMl4zHNwiPBve(D8xE`ioH&S+@}9}hzKGLj
z;TLD|+$q$J0nH>?h;`HD9=`-~vsT)|XH#Xiw)k_Cf~~EaTYQsp#6p$Y{MIHaB6Z51
zZADemo12@nU5wk-5}UebRIxU3&#C!H&C&l7=!P8sOGx`<gyP=@NN0T?mS{(psQ&=y
z%O#2+YrIeP`u+}YFiy%v1t63q#s7w~e@o==KxA(G=Y(d^>G=NXreZ7Tx}7T%j6jKw
zS@<f5XXahcO<!^oxDWk*zuCuJ>$Fz(^X|As_Bl$S)XdINiqjaitW5eo>o+kI5-(VZ
zxU0OaL{o>GF5|xtSM)A~SEhf4m@tM>yQ6Q=9JJ<yqYAa8+uz>9ZGal&bd5SiE0o?<
T-3MJU86mBeVly5jXvzLR^Hw$L

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9c011eeab5fd424cd2d9ca8bcb03008fc5b30058
GIT binary patch
literal 53110
zcmb@v34mnRS>Idx*84I&J$ubQEzPFUF3Ykl$s@~_+OjnAQj%Qm?pxi{J>Atkx2k7G
zO*P0K8Ou1b4J15*E$AU(83;S!Wr6UL7eY473nYZx1Y(wem4t*~Ld;-m^80`1-rBlH
z81kySPMv%1Ip>~xwr~H=a$jE|#wYcePgnli6Y<#Z=%W3T=Hl&qYyVF?7V~1Im>2gF
zi}6z2eiKW{QZnw+Ql+$|WlC9olZ&~fd?|1DQj3M9o>I?JZ>iU=(~Ese{iXh;fzp6o
zXBG#ShDt+WJ;SA8%4ZixmPSe=mX=!_UD{UKX7T*u*wXgWc8eDl$Cq}Lc38Y;ap%&m
z(ypc5rQJ(=N_&>}mi8{~EA3k<mWoUJOZ%4&lnz+F-o=AUhf0Sm-nV#o=}76w($UgU
zyY62+wlq<iSh}rrn_Ukq-oA9abbRSV>BQ2>(n(7jTs*aOx^&v&LyLDTohh9mKJ1Nn
zqu#d9CrfvFw|m>Y@t5PJySyFV&X;4QyB~{ryS&{m$GqK@?7}^jGgnikdr8~l?Imq*
z<=(5w(tTX-^NL&-m4}r3N!jlmAmu>iKCU0&`k;4+>qB<^7OoF_N4P#xc?<U+<oc+0
zjO$}|{Z_6gyxX|G&8{Ev63@gYj!&&HjD^edi(YYQ{z`q-uN0SO3+L<gmD&TRPR-5N
z&#z9OoLOEvwYWMn@A*^n)!HeqQeB>}%~$76c=NNftF`&%>WP&X>&yPk`BT%2%hRXs
zz4!Dz)86dNJ+pV8xp(&N`@B2voW1X^>3gT|K69pW&mEO}DtF#-&t21ZwrVOjYbr-I
zHJ`Hi-pQ30w`$_@N~JnAe`<bd#b3Tu@mj5Ot*YFA*IoDBf9AeBr>F0`_nxVlJMOyo
zUhi(tqqX->PfvSiy!-DzeP*W8tf2MaQ*%?Rwc7ktwUu;>HaxPtwDSJ>+Nt`~k~*mF
zPgQHP%l=ZuuN6+5DDpcqUwB;$YH}hT<er#Xn)ariczPlc<o(K&PxiS=knt-km8m+3
z{WHtedgV%eX{tIkSMh_M>gwWRm`3^R%9LN9pIRiEs$cY$_{}d(U8t1j{CSUye(JkW
zUaR=a<>l()i=-5un6K3zS+35`&rRfl{>A0Fa{WrZJXI^tFU<v+$`x+cf~;4WomyS2
z2kH6R($q?jWw@7U0JT@>gxSTZdeA$wyjrc7X~EP|E!fV)s83aCOubT`^_Q2*v{Da=
z2bm}4>+f8e(koW!iK)f%g~ed^qw`a9%hjpH$HJa{ln49s)2nr66Iq8JXZ-8)^%u|1
z`?dNrbv?ZA9jo=_N_A%0tN8DEVj>v~+q+Iz>Qm*~%=wD9x>yND=q-P0re2<})|bom
zO0~A^2L(;oN_lbqLWKtPJo@P4Pd+oXw6a+7xg304ulB_993A}R{FV7?kegBKXL!jT
zV*YZiHdU`zgZPU)GZWs)x6+eS-qN|5ga$SID9*ZyJ;Jw^B51@|Z1J_}_g<>_m;L#A
zrMSdfES{vP^VQ;ImRYfWe!f<mU9HZj2a5IOqF%08(~7iQrz)4G7OhyJ$j4i*)QZ*R
zdU58w=1g&_`r^{EUpaX_M(?F+mBra0rDynuc)SHS-hb-+@>1m#g(}m_%NI_0Q}wA+
zbN=$`%Bg2ot4}ar>BQ<O_3x=i9(m?z3r;N2!c&Xh(kW{(t3O)vr#x?EiDeM<d-Jua
z>BWjywuT<oOREP`N&gr>GwOioqqsnb^~`P}jaVZ-9;?R|;*EqCZzSjApN@}`wp~4z
zm<%#r<<k62g_la(4ilu9)80fXNceZ~JrksB=ciT}Na||6tACWsU}&{Euf@mYFS}}P
zC)wJzJMNz%KEt=;vp2k|tucO;7+@_{qFr9XOM0ofn3wi4Ue?Qbd9UF05bO2&ynZj`
z4e&R}?-1WKe*=6iHoNx|vDvr)!pr$mG9S<Pd+!!#03f6)Y41JWn794q_|<qRQ^{`R
zuEi?(t2|{k;f;GcEWP0E^mYL@dc57<9)5ehz1}{4`@EvJpWlA(fOnAJ0RX`u<pcy$
zr6E_F_k;lf)ju4jJZWG+DI;OZq<7pqVb2<+Jtw_Wv~8Pr+Pi~V$GrElG2ThacJAHf
z-A!!VyT`kidpo@Qy!-jx={?}Rh2LG?gWg;D1+3{!yobEEy&U5?d%U-M?;w4z_ptW}
zzx%@bPkCp($LzU9?_J*G+}Zz$SSnUJz{tNZNIm-KQ%}DBr6l8+S)8x3IR=TvGm~f6
za;$?>i!91>Y)@LcuWPNnPR7{l9>MRQcx3l~`R#jlzU}wFVf05I`>nOV`nKQy>h=He
zD^E_p^tRuR#yX7r;oE{_Wu-P#jMDkt&o{?c_0<AuuE(y%HsaTKNMpMfzZ9$Wy*uV5
zuEswWZ)|VG%87ci5nD(tq(2w)6ie`%dM$M|4jRDhE@vA3gxN+PVXjdi%r|-n3yogF
zp2iNs-g;jn-Wa=@Al_depq75kv{w^6Gf&zeDT98gKBPNW<D?ANlU}wyQXeH{cw^f&
z`uS>{C+A*E>dB39FSe?DW85F}D~)mD+Z*IAk2gjLcQm#U?rh|g&tGlixZ2g&O}M+U
z*MF(Lr;)xIuf6o{7;l-jJg-sH-bT8SX=J_ps~~ygQ`%td+m(y<6!hfwHa7P8KSFws
zwMaRC(CSl*_gR0&7K)AGMzWD;3^WECLtgKz8R|{Ziv5k9ga_&e7Y;Rc@sz`?p}toW
z>qX)R{C{0PvT&5xF?y;0RrT7jYNip}nD7P~#cS%dJuAOizwN5}|90}-?*Eti@r4r`
zC$A~xl*itp@5JEbMACmN105u;oDSlb)=tl?`aXbX@x`J5Cfo1I$}%9fI#)cvL^;4V
z>lF`NIUPnW9a!sop!oD^wW=KK<ZFW`R{V-L?>2GmnCE~nS$6&D<yvKJ@I95vin6}*
z$#>L1w2DFeBs*Kr{2Y+6Qoc}mu@)5c;IeJDH_A^}YL&V@zOpnWps*r%U^k1GEAw;b
z>$Q``bAlZFuhuHFtBb|UY~n?~Qe(T+qXdj;c99BB*gLVw&s<np20os=@o4e()~EP^
zcRW4$_~c`}(L0}c=ecu4i<8ejS*EeYXP<iX9p~QpO!3@%i=}ry{oamZ6PciAZvIlG
zS{5j-1$~_mIbWY&qBV41(CZd*8C1cq2ZiveQuP42%V3n1a*ggN%+4<^mRaSi6(HIC
znqYRHU%9wC&s%}oOnK}G%X62e{E5?S|Eui&L9PtIpRbq8!JuIzm3qaWUz-vv4bm%g
zL_Nrt%QK5pwHkTIU7cF0C^z8TQ+wtruFr$MnZ@}Pz=fd<UeF^HslouhSgr>Hv%*Hg
z?2J}WunnlJM)C?k;%a5F7K}3z<rQ8?j|bl|oQ1Kd1%r=9(DMWS)QTFOw!!p=H3_pS
z$2j?;y2{%fT?fg<N;OC=f+7Xk>DAd;u*slLnIp8v4+fjj5c3LpE~yRb2QZ*fCd1sk
zHD+;ox$NdnFtA9|g$G3m*`;NAk^DV|TG64i^L4*Z&FQ!Pb*SBim3lD5U{vbm>grN?
zbp=$SQd80#ZLIjsb_IED)aCk%D-{NxJ-k|WBj?|yI#bNo#h_1hv?fWSS`G3~39Zy1
z`tMc7fo9F_^@EhwSkSvV1O6%;irx(p)fG@@R_O9l(DStYf;iS08D(4Xt8)|kePJi1
zp4BQ%cNnVwE>ir*6jb%7Jc~nUZT-Ceu=0W<PkDwumsQH2SI)A64=E@qcuv6w6?{O!
z;|kub;5`b?DwtC+tKd9=n|C%pXt@@?|FrH^6g;nBnV_V}$27O8?Jp?#5`jnsb-iZB
zw$><c8I6Q>+V!}9PDKm>JgV6H2=3rp6PPUQAB<<>JNdT%^~U?{x;MTzUP#y-3v(6@
z#&hw(cq(4tx0l}>G5&P#W+s_Q3{!SA-kazpg?u;H`UJ#JMr(5>%X6boKM}DS9^qU2
zS@6X=Sl`vSpKZj~v(<bZd=SWx62$T1ucp_tUZRm5kFDqGiG?I^3w#WCkgq2;QryXV
zU`-!S?1)`UtQQ*Dh4l0B^<+KM02;F6XAP$0s7uf{zmWElJAecW*$au<+kxg8LGyZ{
z-m}olx6jK`GcbCi-^&53vsdE=Q*<rZoo~A?a1GpR-tTGj(4v7x{u=MnOzm~4YUg;Y
z<4&L58QkJdf2+){&N2h-cf7$?jh^s?q48KFI|_!<D2&JKmuK$M+YTvxdu$!NgEtwX
z<-Kif9Cj_C^}~%`Z{%8{qn)Gb2Ya)^HSlxS-d$iF&3D|^=yA`qT)C}sEsS_$7ZRV1
z`(Qx}qu%zBm<I;6unl~_r;)p=di4flYWJ?#dSCT^u136ZD|5ixp_=PG8{5^pR{zfE
zN^eO1{g!{1<$oLZz<U1t=Uj_ZGF9K<?KYgQ&l_ZJ^m)k#)o1<Q9^zo!!g#0=tYl%=
zwfM&FYcbyAdEqmLNfd=ibglabnb|%FB$Qu}nu4-^{U)Ea+ou*+&QH~f=cg_~+ASBS
zD@Dj{fS*@{SSePfW*`PbkV(n3XS-+G!Pb6PxID8eE{7M4bY8Q|RW1c1wTnJ9y*;Lx
zbB`r117qP#tk$}3?-rc9<=q`t*ssh08oTP<)w;EGN7v2Q_4dJV6AS^=eYhwym0)ix
zO%IOJfzb1SMQ`rl!{63>?eo9*mSD7%(RdFt%az$XwoE-^nB424JpG3VexEO&FE+9M
zxBGA7dTnrN)d=Shx2t$Srb~;{AqcFUrOyjM=9(Z`wLY=Oz@NW^qW)!qARG04Mi9K}
z`6rc{1jX<#DpP9~XuETOE)y~$va>*w`kiO|y2_-~bpZF}IfD#Ac8OWCw7TSA!Bv(1
zD+*p#=FB<A%g8fGHDhH2Ba_SZ#|6HZXlTWIr|$!~VcArts{XgDOmb>*aiZT~o`XV<
zsX7CAtyUV0(!+v`8(Mpw|8b?{#6uBd&oxAqb`jbB5^0~}Tbm{jm?$I?fP-A3A5f4>
z6bzh<#|IPq0TjlPI{`X70WYJ;OgwQj2bdU4?oH&#ty1HJy~&&a)6L$*A;3&8@!ZW^
zvX|03<7;CEQkwWcg!XLWzX-O>s9irp@KLShn7@xOz5%_kklaXJV^@VzPzrEiA!%2v
z=GC~r56B^ED*hUDUOcw8-;2E;`tJj-mbY4&Y)37w+}px(30E$nGxSStFos)}EOnN=
z9Mj)Ka&m3Q1I1_H4~e~{%|f_%i1@GV0#P?KU$;UeD7rb3b2!RJczJMS#<1dlSn)y=
zJ%-IUv%CUYyX-%sJ6TW(<(o*@gwb8&J$4OvkQbs+qiE&-rC9A4LE<IB{(kU*{&<2=
zSGo9bJoU9yBJ+mggYm@I5{b2b8@y%(9ju{kRFvcUh-%Fwg%H3LGPwfaIOQeliAHK8
zDN^@R%pYlxvLm*hYNf8H8|g=5FO02c8VP=f*0YVwMyip$7WYrrVaIG_8W{mK20hit
zyft=l-$jv@SvK;w)^j4bUyjxD8vv(`9ubNgy-X2m5@`+E(YN`Ymjw~Yy_)1{zxO=I
z2&hNCk-8N3%NO6zGx9qCARrP2!rnTl#ML+<Ko_**q7lt@<>fD={Eq{4L1Ki$07%0a
z`95wWGYEu$TJ6C`VPoJLpw894-tX%33KtTP+JlWg09MZqT1<<2*9TxdfGj9wEGD4V
z7<e#t@q4PD3QOfLq<Q!Iy2|Zc9|%(lVNIj!gY7jzs`q%@8;i}Vwba^x@kGu4_~PHM
z$2LGLbQiEX)KyPE^$fL2r)iT{xED?fWxqs8As(*ea97E`&XPQJxH0%3AoyasF%0U*
z-^UYb&H6C!HY_Gd*wX^jcQuALM!a5Hw8h=pVcuYPkNVVlmA1UUk)*F5V{VKx%V6-m
zny6~IZ);sM+9T_u%&Gy~C<b}jh%sP>h>coo_%#qmkOsTs^;}5Nl8clWX^ea><~9hI
zH@x>z?u|A^qkGZjF|lnjNY1RRUXKU8@ZzDLU`&gwEkL!xY!@MPqxfiLv0{7zNQvS%
ztUIc|mEq)!AaV3zZNX=Yu`99CL~G75+h54=PVx0bBmS+4i^zsr%O0+G;^gTdC6YOa
zms>X+m$mSq2vcN4D-ZpqDii=7jjbKErQS(2o_tWPU#vYuBr3CZ;Ox9(aV#T*3imc`
zb=At{=<+%&ZLd<B5f8XNwdjAANB9CdhF<v}C3cqn$w-7iZ}c~}MPq*4$T$xBjXq>t
zh;d+wh>de&un0kSWu-C$>@14+vNrH!c>g_IJy0mFjTD}3WeF1}#zNv%grP5}7o=3$
z2E?z|#Lag*YDUcGsk)=wrj#)c>))?0FT*5A!!AJ9FsNSjoESy)E`+^H3F&-aTcSfT
z(ld*|{vZvBI|pY(ctDVraHkgJSF5#)tCb3DhykNlTSIT#{=_!7r>B~9t*>o1NYp(4
z=M;%w2om!bO1Ww%(D8p<sc8ciL7tH*8=D;wPUUiYv@L^+#^;0hLa9(K!=8X_Ui6xA
zN`m3I4h{;fuJ4*p{so@=D|~AoBj^Wd5UrWxuVA!k;-$pbaw%OI`fxLmbg=|zNB>uI
znS|Ye3XShh6hJFfdN@7?s?iTr8jc@PEIAA+(a)6-lTk{Hf_mf<iI;XJpl)v-0(sFN
z5bWmK$QBcMMoTFAWQb@vsYh!M6F|udbg`5c5?7NA=o)BIunAI=P-kH}piB8u&Fi$I
zcW0B5d`#_S#hRq2$mg+29RU%90;?vHk`*r=yHc(f8^tT-(-S9~Ss$=*AdxBH*B`KA
z&rKGu6d&Px^5n@EiVhbQ`_Xt@i<KJe!|KW^yu|KW!tyX^rYpWmG^Ia9<LDKS88K13
zW8!4-T@}A_ym%TJCo)girwFfL`n6T*SCB7PA1J<~SYHKcr7Ss*7bD#Nc(I)lm^fYp
zdDSDsH#u4>+7xV6=!?LbNc%ha@xNJtFmZ>yzM@;15RnOM2;y~QVmvcIgy86qQg8b+
zyJliAaItJON0b|lgjh}MaWp#$MpKBdjoRGk%GQC^XktVHuP-0tV)Nv%6fwcwbV-iJ
z#@U*$$0x6++`D~}V(#6(h1f*Oy;hRZ_kW#)AZ_E=-TLv66YVB~6Uth~J-o)3^%}!0
zh{Vmr&9y!D8n?=S%lGmBgBtc51W`*{qwoorDt4Ry9~G#jMu2s<bLSS%T`RLgmoFWR
zI(#@4wshaES{jzPWm^qT?dp4ax~uP5suDcfJNdt(7XPk-L(SkQv7p(r@U&gOhEKnp
z_WHj~maZP{&ezdOQHjwEsfi01wvrhUZ69GPqm+cPaiZ-bOkGb~E^Q)K4lLN1XWCBe
zeOL~*Qa;Ti+-&-cVn$)UN^Bx)^UD9XN^K5V(BDabxAJ%QO3%d>zxfUgoH%^Z6dO$0
zdKnDYOS4X6n09nWo>I;35W^Hi9L``P0cQ`%&;kSj$gbpGEEpCI+`m_o>uE#KP&4A*
zu0{${An6}$q&MIadTGa_Vy8{31ah_H^U3whdUic0X;vd=J8pK;_y^Y$Zl->bC%3Z^
z{_y8*ySbk1>L2zW(N&IQ|1Dx@2rQyoQ1-vX#YEn%)%X=xtJuR4V}E%SZvK@rOcfz>
zj@hs6#(FpyYj%-1=1m4kcc%}mbZ$`(sN*F}=w*8v=9}WLH$Da(9&T1$1!mOO(WjS4
zv+o6dK@8&yvG0WpTZn&a{ArNs569Qj>bqwDC*jVjKXat!k%f|<K22Rst{313A~(MP
zhxQs<+5+U+T--~E(p`W=@e<<meJ{0sItGyhdE})V`O(-$_8RGpf?J9c8FAMBfVQ-H
zu!Z`;<2E$#Te%O(Yy1C%^8S&6T?9eeST080nxepz5|iM46TNP}ch(*BI^~<PZIwZM
z+W!evWtIM+iVnMx>KK$S>PByMxoWC5OH;KA-Q$#7Dc4Nn<g5lryosHOF))ndFg!vo
zd6c6Ax6n4N6=Ar=1`+z5HLy-UdvQnp!VO1nf>n?Z0X!F9&(u@(bRDiad9zpHV3=Gw
z3%^{1>QNgzc;WETAy#MOj25KzLcvSFs`b^Q^+8P<;odQ@7KqS??JDcRe`kF?23g(L
zNQ+-`vDirXlj70OLNYYcU@o~yL&_N$_1TSrDcT6<uKPbl0_)5AQnPC!?&KvJLfBMS
z`EfUIBW8H6S2g@M$idbGH?Da8FDR4o+5Swi73EBU7tgt>ObGKGBr+QgD}$y)FzQeb
z=>%2K#qy-3RCAB|itZ<+{M6TZ_g7WAYvb>W#)vh!yz*WR<!-WpIx>`*coC#B1>b5n
zt1a=8#3QNr+OUII;qY6**9R%oG0a+Y&4qV>3k@@atfAJRXsb)HvnOXk7g=!9-AkZB
z<|FHdP1GiZN-H7lcAdiX6pSO2PFlvcY+Wywag+1I(d5j)#}g|jb8`)CDMUE#e#b_g
z3fr9IxAp)5@F!`?E-dYrb5|1}y_$k-0<VDhUQNEjvbGf7!KE~Sz^3SvoSdP8j*qGx
z;rnGR7MKI^S70fGpo~dmQ=}kiA0kEcx@@SJbOycb?F>}s9%M5{#qOb5UhT^Xsw%|R
z91?Lc9FlgT)a9lB1iQKsOQAF&^5q_5D9sm@(;kR2Ys`_vxv1`nI@C}tHe;hEoJ1_T
zzh#F$LXGX{iuP2|5rF+BBG9BPgetbsm8f~Pr&-Iaja{Rio<n+gX+h2XZJxDOupzd@
zjs|ti%JbyZtTgaDB5ZZs)MA*QNq-2=TmrsSs*!T6PA|bmahf}a>Pacbz<Z;FfpBuT
zv;|8uL0;08tY<ycPCgcA(?{9o4*yP`3Eu%)0;W7tPjA%}E7ta{RxebSFIS-$pj{*%
zDIWcXrlfiOD2u<UqRFZ0n*Z;J*>v%Jr4=F_yQ#$>w<N7V>hQmZThX#f2_X$KkWX}p
z71s_!rmk+xsmF9fI<&QMYA66|ZQ17wUr9kBq%xylO{H>gggbnczoU~Qm}|>OMBnCR
zq;238(Tzzj$==L{01ul*M0cj=V<37gFE2%h?{6fLo$j^kG}jZa#JH9YP}0kAeGes5
z)Wlr6rkr_LPTJ>_k>(KEJ~4)c-Cr7RuO=6==)CoOZye1WSfJd?`A>U&)wtKco(Bpb
zOQqz%Yl%iqP<tM!9L!aVlNSgMLllV85SuXPAUlvgDmO^Ep0JGa!Q?c=7OX6+c}O)!
zwK^CxF?{Q-!biP4Tnq)pTC37x+7SFcZoH(M{RH7oAf{TaR;g{?2_kf_wbG?6WO$(y
zCJj9mDJq28CdIX3lP<jo#om-K!JM*rM0JL)d{&j*Ljcn%>CG>}*Ao9eNTT#|qt}$%
zLdA?5d3S|?>PFh46Mb%TLRCK2hI$U_QnT3KshB1`(wj&%BI8svO!vS<L#B@qVL-NL
zHeXRlsgR!+{r{#)oCb%ujQ+T4M!h~(DhT~ynas{#k%kN^9!`%W$k_$0ZR>ou5l5Z;
zO_aCYZI}hZaxn>!SU3f)hZDOK{fTkdFT=29`V*rtWt!-b8>9|Fk9MdGUnz;y#14?m
zzz`pH8zwy{5(o}V7SN;AM%s2LO>m&;9B@g<=RC{dZ=!SxL`^!>#s4Z5K@Ay*$qK|Y
zWqZOV9Nxf0kNKka!d=5qRK&bKT0WTzm-Xuq)~-z>3r~={-gm1TQdD;Uf!JEF?ZDx!
z4$MGVv^5w|kXTbB42T~LrABfA$%O6413~~T)iw#iUhXFA?7X0T@r?{`P73dsp0tTK
zkvKVdV|coFzIeGfQ#|odahe}4n2>!-6*P%oEIZ~U{V>H-9crzH1PO#h)-^*C5Y5l6
zE+b72Ih}t@hKN6rHtfUymnv=WsH-RP+76BH^f%QP+6RS3iL(c!jEAzKW~NQ!uAZ@8
zarn}PdWO+wQ$)bm0PTgJt<4f85+SFKPr)9@!%TH28o0ZaU=7sN_Kzv(YHAOhCtw1c
zl*-EA;(kqYuAkYPi?2D7wk<=)@D=n{Q9}bFG%*Giey1UK3?U*A!{e21R~km}c;t|*
zH4jX@9!JClhA&}^33f)^j?biqwr!-~C^DI)E$ZKG6fQX%8N?X-FM>cK+hkX!D@5&T
z>gUi}c7M57K^944uP2Ps3kbBNg&AAO`xhu*xF(ta5t7$S%ak+g^`UR$>WFe@nOgmm
zL43vEPyYn57K@&(?|zcbT~7r$R2JsS(`fQae^2%dSv0fNsX>lNWo3EhJgS74kf1cD
z<g-RffC~7f%%WcgnzLm|1<H{DQ;Mw-Ky#m8MoZkkPY-7gnlrJ8|2buN2Wh^iT!rNo
z6f4lxgUjA)Y*MrdQGc&Z6dbXjsSxBH);#ZBadTI#x3}_*Bxp^NhjpWCru0QyuC@Os
z$WhZ;5$jr0JB5^coOuHi`;GpTDOKGpAnlg=l$g}1L`W}mRMRn|wB>0=X)PLyoH1hR
z+Za3C0A1zJuvNFGe=^oW%wJ*tU5$SdX|@O3Iu*N$W=l19N33?_GMk$IKE#)C7H>uT
z6XT(Ta|Oh`GtIPLGLs$hC861BrqK7AcuwrGu@~~|lIDW%gMcvQzulxv6R5$P%A@)b
zwEoBV36h9YuE$G#21^8anmHnglM+qh^ab^>p0_i$cFZ<WK<GT`h)Bq62-;wJv(VXd
z{)3c*Zqj`5uT$KAR0)o5@@+?WOt(b)mj+wys4lmf@*ZU{*f!Ah%zwrm<9_u6+@uDv
z`;E~gbiU&jcV+JAISs152G7)U3`5M9Dw<JJLa!pZpyTN!TVj4E=t;Vnd_s-GLT%Fo
z9j0Gwa<L9Y9HPMTD{8KGv`}epyY=Jdd>uP*iv0T84t$&l@JMp<1QZ@HDWv{@N)`hu
zi4_Ek!a<?b<c+av1alntsX733V|=<8MwNbytBEFPN>%;k%M9BT&-2w!zIY=Mz{xry
zwHEYOWzAG8qs3FEK+qd%^O{*y_lg}U3&fBuWaUaQ`D^MC+qwZp4yj)ow_e#I^XpoT
zw%^UD6Qj=n5#8zGTf0JqvLpDURwW&z)dir+!Y|9RkFy*ASOBiJ>!bnktm&qpwty``
zLdpSfZ?dex#DlSV_A1(tVCujPyOT$p%&u)P;%+c+;nanYgLBrd@*eUh=zt6Lo_a4l
zoWdOYh6jj*_XEEU<pOEK@U)G75QYMIOp{T%kPCyTHlXcj+@W4@{GkQ{bvQtxDTW*9
z`_ZZ_I&&%iyGa@GNcqf=?IWWMy{L%<P^fRcf78D6>5hG;k9vBfdq=JPR%<U;Y5(ul
zw{48odH0R&QUQVglyE&~2+&Ww0v&979OC+y;cjV<dxotBxHRC<x*+3(T!w<*$lPJm
z4(Rk12FTyV00n(b>6Ye0DfiUkvPq`TO(gs*85|Sk8D;H>rW6@FR&e{?_vJCFw!JT_
zy{~=WYnz3SQy<MRSY^<}=oD;k(W?o07GtrsgEsvH*-YiUomOe?YZK%Cs$Qmv;si;=
za}IU8Q|&j-+qAO$f&ycseNZv2WJe=^r(%pNyN&{*GQ>~w->*yCDszfmCJn;{G1O{E
z(C=_^<cWbcKpLM!GNqpMRU+sKnQ)SoyRFHPuIE%+Ck5hTejD@E&;IiYtXIUF3b+4O
zFZ>3HwfksF29iPQWvHCRGpLI7Cw2o)C3Da>@lt;xg;p{<F^XjUfZ4r(;^72X=0I1!
zb_}MvMPJ+G9w5Rxh^@yRTg#~;3E5mop^l_qxUcBfChD1m>_Scq53%aPEiW<2APZ%f
z$OnC<3NOuXA5Ns{5I9oq`*?<_ewzljL$lL{wFK#RcvDLsh!{MoD99KNQwx&hXq!+5
zlSZ9(JLV$g7S@wwLK&hh?p_8wM9SQ2JFE+~%6~mG>7P-P?ogn9_U~7qu5j=taSqz{
zEb9Z5Q92msXy6Bk+J1O$a^j%>EO~=MS=JTUXJT^ETP|N*omzAT7^ahMI)uJ31~<Ws
zy2t`+W$WR3=Ynqjdj-Fx;6Ew&Jq2GP2vXQHE;^pSafQVMwMpmZuhGUv7K>o>Pbsi9
zV&MN>x_nZ>qyn4g;@kO)1VLYDH8NAdCZ%2X<(u?$TdJpto#I=2nIIQqrolP=%l-BL
zR-%Lg!_CBfz41&xigOR;dUKgvE|=-gXL?D?DIM8PCYLT`3In-)b~l|FM<qX(OJ%Z*
z>0w_jn&I2=QUB^w`wH*XGHt{tF_BN^Q~5N%iF}4w%6q?Ur5HM5w3yQ7{c^07u7oCv
zcCW|lg~toTFXi+7`62HEmg>wL3ufRblSLUf%4G3_X8R6G=PchbS~P2Sj28FF%Dpt`
zY}-pi&ZZrU{||XbyrUQ!hDM6ks**k9m^bk<K&~_znh`$d-R9kHW`x_k=bR~HX)L@a
zV@3c0#*5CJaomg<WxDu$Xw0}XG-lit8Z++pUNCb;%ocfO!uzoIU}Z06m2dSPdO01L
zE~+)$E1O}X`T#3^nLLvEHua+S5i@nvo&C)_l;F;k_o(+yYwdyXPQvP%_8#-zW$6dQ
zwm<H@+uD98ynX=#%qJ|z;buvVS=yWNp7P#jcaC^Zd(Y6HN5eAbyl1VBV`f-60WkD}
zfhX8=s`aKN(po_#zD2W*n)L5th6q3AXGSD>^pQ+a=#9{t+BY(TP}z@_#m|S!Vup(t
z9FSrrkf_Xqs$7WIwwd;#j0~lGoALK}NtlNz(3XTJ2IVnrsI_CB5Scy<IWnFMIQlKl
z>?1y4@w~;+ZzNA)PI-pv>8m1Vdp$&l^YKQXcsQUNFheo-kfBEo7AfW)YrXXXydX?F
z7W%|3fn3JO@&mCd5M6(s%rN^ugut;@HDk+xDn#zapx0L)0tJz&LBfcs5fNDy1EH}P
ztkd!~OhhoVG*i#*79Vf)n-cL4Q_BwYCWlnslM(26W56eMVQ1}{X@BnW224W`v&)4&
zW~j2aG3ftXeU~?=bj$fO)`OT{I+Mu=`=1z^bgaa8F&_SY1qTR%QEY8$7%H)wdy)BB
z+5ZOJUne;0f1{!gD`+TSAdy_wP|Cw>AG7&o|D6A*ZoHFVq8Mz4u`^S_%8T8vSto$Y
zc#y$xhvk-Kh$@?Y5%S9mk}JzA{!^;(eR@(4b)ddzrgniqRAhT+B#Dg3&E-Lq3aa36
zeaR|HInSd6P6bL?2$v=N6O{65fQKZY=3gd@snxgem@cmbE!(|dtleEfs~N#?D=wCn
z7<DFBEcH)MFQad>e4$bm{~#EVvNZZtFol{Xai+QE=&@(?#?q{GVErSCEfP!&8-7$8
zZZ{+hUrOw>=*3-<|DxVWi_KqG&`|JEi>X3;<27B1yzpBjUvs$IM%%#EAr&*2rO_A=
zHt>ETnLeNjOt)M@=3mkmaDQW@>MsG$B4gYOl)52lolZY@kbA1648jEOQfTZuqcCdu
z+S+*6M$@tbqq)rwjfiFW8jkfe5mWnyJ~GB_LJO7@$+y{B$|${Tpy;xZakkcOIWJ_d
zp}jABOlUUCCZ*-Lko|b#K_izZQf{$8IX=k){3jHAOo1ALut*4piIN>k%-VyCWWzD<
zp<@Y;NG7V7tJYSg;4*1gKzR;R3emB*NigL6Qn=v^{Zc%xzSjP^7b)>*eC<GY-<Z8+
zN0AOGk$OnQMfBGG5drUqwILx}8XKuB3+O9i&q+Yc?+c_PYh^;1-^lRyVCRN49CmK7
zgl*f&VAqbd9XYhoBDzEB0>y}$v?NSDjTA99(v4op`ZxOMT{hQ`=$@?86G&-LMc){_
zhV7NMIO>%oDitX(N6(42tnS|lG;Y-Y25Z*d4kmma9YeXaKy|b@vx+golI)n-n<AB6
zc9avX6hsc}0aMiq4Cuud*LEE|c(C~FGqPHK=G>zvuwj1W9Z%?rbXzNHdm>|MxyLx!
zzUfKdcw)%cV!{B}*?>EKq+=g`j2mvf*~+#Jtx0A1P1-6LcJ4{Anr+$%2jk)|v?Vmn
zf#utxj<uIF2EsH5FC*CbCV9(97puvM!pR2SI;7eQPeZIp(N4jTdz{}=yb9uLe&<rh
z^tY4aWBb;m=HobZ97D4%hmx)Ig|zljh$TMyKi3X*&(N)^xg~YbVd|;LD^`e2UQrRw
zr2<n=d?C|PyXIKVn!1A8!XlDB#&%L2YVo|#c`&F4qt1q4iMnhx2HR@qvD3!{y<-Xa
zpU_GT_1exV^J5AOW<IXiyA`}gL1(8KMN(EDi3>EAltC-9s+(*3yL;u$4Ut4V>aDf8
zeOahg#3gGaHel28rp|ptqRd+_BqX6iIFMckr)nh2nTyyqpb46oJj+Xnn;N9#1tu73
zpH8NaWS&<1qNeHhk<XtbxE?Bs%4H2u(}AKd=&r$m7BG-V?mRF1Wg&c$zc@JzuY0un
zb=%=)TO)4G8vK5$<ke!eheQptFAddE4++s5Xy$TKugE=V?j7QuI>LF=NIPE^hFV!J
zX^x4-Natm(^ODB#N@CN?bdjVY>^$s#%RLwG32TmDNY)-TG)hlN*4}N%SF^6D4eUq3
zu%S<FFIFzE9e&nK1}qVkW;_L;nOs{gV#2A(=8{hOm*~W`{JF>uYi*Y@;CG<|7Yp6~
zoJEXb5Z!i@@6f0T4m*2o5zIE&{$JCLW4i1dSpWahjjlm_iL_<S*aH+0w03IqQf|*B
zTSM9lb3{YDcI3^5)~zDFtN#G?c7P(S<R-vA&&8aquoyxCo>vnL!60BG<v7-j^oG_l
zphObNR}F8zn5%Q2fy~r103d;x6a$H%0hVW$7^-R(%cFLdfi+2LLjtCZSSUZu6VV14
z4iVGky0{6*A;HStt8RHp!TSgh>Zqx{xL3ijhnl4aj@#OpTidq7F^%hhsod1eg*M)D
zs77YYC_TEgCweeGBFq(4^n8;!7{jWRSo8dIYR9t*ItMD)X*}dM{-MJl#dm&L<2S}r
zq&$MgGn@cfP&Rqcw{?%%t*h!-k=lY%e6<bL9wpFl)gjxk#Ks;H$eyT<SQ^SYq(bVD
zQUHR--I_zoBcau3`XO0qA~76TC+^w=fTXeS6mr(K9c^ktn3D--rHb3Jg1pDk(HH{z
zdRxb;F(m9YU49P%4$V;OF_;V5CREG+j5_nPn|sv%7ScYeuG9YIItxDpi9&MiSa*lr
z3Kn;C7gaTP?b;d_+Q5v4op7u_QM~>EbNmAn8m9#Z$bp)xC`8zUh+OL#aun%YHYy>d
z(uzW9$L{OO)uVc^CTeGFLlW~|Ohi&_o3m2p0?TzWy2Qf$gDl45U<$r%8K9|{Nr-Ur
zzgIo8p}_j|2<;BCysz>`*lK*Hdo~(&V(dsr87=$BDaGfaxmYlXrFM{1rhq!rFb18e
z_|{#!RYe`rOX09-zBUuzPXt^ZxIo7w)IIpwLNjTDt%A)&I!dTH*nOlW8>F~Ywv;fZ
zHXii4K_ZZVBd2sau*^-=9O#MANbIAE?d2N0!B%+&r5&@Ox-?(8Y|L*=>~Odj8jQrx
z2l2~(QKbycO5&KWyWev7v^g&~dq1dt(neDRvp1IVCb}oe+R+m#((H+sxX_ylM#@4+
z;T)SK?@Ws82_}#<ed&ot5^o3Y%_He7`d!i+P;bzi=mdmPTbG_uI$!h(8X4;;JY0}N
zp~|`KHMjP2g<|*#t%%M)P&{vLhPrQDegK;~uYSIj9$EJ|AAfAe_$Oma(;^pkZBmQk
z^s)~Ed%Rhyco`pm=P}ExwlcpJ8=|RtnAI%xFpIa_MQ)wETQ^;8GctSko0YZcR%yf~
zwb{p@2e}KC$_iRmH8&+}B4k=z0{=Rh{2x{D5d}X)5NQwpuo9&4Q5s%czASN~d@iw0
z{3@<#endBV^_Jzu%4{9)D=;*5<#LHHQOe`YKt`I_ZFRu`q>@lCW!EJ?7Hwa}tb`cj
z%q7;??(XfYy-0_&QL{;F%huHf5oU{lWGK5#T{iDdOHh(Vqoy&?#RdWzjv1s`h3}0d
zD-X{oT(h1=xsTE&l%LZ!@lrd$Dy>%TQft;~HSIw!9oEWRl3rmWt6DeJ12C5d5<Qg_
znw=~Sn5!iQKlHs2B}`oj6Kq6Fh5CG@JniES(JzfOizv}qi$;kQYbAm(|F{Ov)_Y2&
zO2f64MLZgr6fiP#w6PaxG;Cngum_1JUgUelBxgb1Jr8QM)bm96aPh*r2DVQ?rkVBU
zG_t#Sf_N+Pc5oz7fY0JEwA35pm@+cbd%8!uqYmqbj`0>dR%34CJ(^^^v8i}I<*ZuJ
z(h-A|v1TuUl-PlGgqJ`vCG&K4(J-CgFdbUo6m~D6cO=E#OMoZxi3mYxpTc+@pF8X<
zIfR+LVzjRJSxOpDhyXo+BecI{&@f`%eb}8+`E`<QJyEh_a%QikH*);KR(0m-SK}Ml
ziLkThV6~<&^BxGVyULEnW;GWFJ{zz0TWeGPpM|acv#+DIhMdw0=5^RAVzBkyJ8D(v
z4sgfCxTE(T>VC^GpWZkd<>Z}*tWVnC**zz$?dy1V<;NO>RkyhhHq{mNMfH=`12QcC
zm>WUG{sO&`-Wl}<v5wx*9oHM~&aHYw`J&#ioLlyWa@o9VDE8;i1m(&r?-sqWsjg;k
zB<eljrw9*lF)%QTn)yV2?c+`85ZT2(0PvU-mMW)h=$F)v7iW$a&mS*do@nK50yE2u
ztJf8C*XN30vl^CSN5bKo8CX1b`i0{dk+-hzc%fBa3m3t0X-`|OAm`05I&*$}`Fbdt
z;)GWwzFsBh!9Wb7TxY)VS=PGK+_c>W`q5NTP26vYvrmQ~^K5k%lj|ULmeBv(+W$=C
zZw#uBk;B0{;nyK2+9^K%jB@^6f*`Rv8zelw(>465&G=0K=>Meh{1ic{N6!d1&>*uK
zK2nMPPpgi;2nDbgEA_TzKhwQe4j4kN`dmG~q}@^!N`HDBl_~j*8IBhdhY;N#3a`c!
zDMb63H%=yxCw9kg1JPHW?$i`Ko3(vg?W)!s_q3TjPl`U`uLy*)nZ|0uqgz}mz#X)x
zNp*(gYODck*_0vk09L>$8<S*4c>ExS6bv;Q*r=v;E7cj40j%tI5zp7YD^%{hldFPM
zJ^Agqko3=yF3%?K02~8Q*L$nmw7f}q7vxxPwZGXOR9~9U?XBWeOfJ0^dSU&h{8e0u
z_SXBMDk>HmfDXApiw32xgQF`s{~{ihx3ol~7w;GD{XUB9E_y#)mM#qWf1MJ;qFhkl
z>Blz+?nCuPv~ZAxhNCRxU8%;ve3IHnkmOO`+B4+;NM}j;5b@HVM+pc}&^Lz+5Brdx
zClAiWXvc74Xk#18eprYv(yyHVi!Mb<N?d3BFK|8P)Q%cMEN__x;-+Q_t<L3Y30IC_
zm-sluwyP|0O_0R3HLfFiOte{sN~$b^=gap3#+BtO?&n3&#DA{2{Vxb6wmZh6@phYR
zUH_B1FDldjlmavOH1U}UUQ9sqA!d!^=@<!S3v*g&KcEuN6Dw_##^Ne2J6Q2xuDDT{
zpT*hsHZ%;{=q~dDil)J9l#@elvU<kDm>6m4)|h5$urt)CX=7@X-KK#KoMThWIBv7N
zQ%bsoK&@#E*YpUfJO0n}xDsY|GSM$P_Hv8h#y#qc|NVNAW>1w;GpZ~pE7sH%j99xl
zsOrKvYpHgQckf5XRWc><n1_)XN!#Bs8Z%nC@HphKVU(Cjlu+i8ZVEPZN$lw;W7MaW
zmUK$;CUB-5Tdi)x?mGBY!tV4Hs`TZL5MZ)+5XwAuEwp8-V@$<0;v>)jbQHyNhGvIc
zT0oHpk_?Bw5SH>N6URO&v60t(6Lu7EtS9#lmK%`?GZlVqcJg&S(9t7CZp7VoeOg`k
zAh)jf8fns`M&}@L{6A3{W;!e@f*=?1wMxU<B*klz+inK5?!Raps-VEog&O$ADUn1$
z`9jXGF-uClk$wG)L{*<x{fG1<3yvzLsdkTss-G+Lq}X2N&8eT*W<k|vM%pcV+y6Iu
z*sm%0bp^knp!3<m?#=a;<w3=Ce;#f&icxQC)bZ|-f3r$%&Ds&Wh+#ImyJ=g<cA)Mz
zcFJ`p<a0o>v*ETUJfZdhv0}iQO%lD$qM(kJ2-KwbO8QVM60i}9Vlxv4wp!EW79|EN
z6d(~<`Q1y~LOzAlQ#K`Eu_@dFHTCqarrt-Hr_@wE(;0=Y?dtBrTUjuO)vK?mB!4Fn
z_z2?Ih*`C@iH!gn3&%j-L*RXQRFm}4c!#i4c~7ViHhh83g8vX<j{qKWMq?Bkz1fFo
z2(HCOKfL|{d1_-}CauJvs8#Hop;ANhom6PPgOrySjhXq5DRd&yGu-I7#?zD?4QVoz
z$kthFBR6imZPnkV%vTlI{y(5tkAgo|-M15q1WK(=Z;9T2olIEBbhFt^=j>Sxfe5z)
z%w|!5lB3HJ&F=Wx_U_>bwGTSX^EC{bGLP`B%@V+}#l{v0pQ(;n93|pfe<O(zxJl5&
z41$~+^Ac#?$fz5)2=n1Rb>nOTg3hd;U=jdf(W$|$ej`=81HL;1KU_<Z>M+yx^E7|2
zf_)0;1<k@PP}l5E?AheuCF-c7>OSk4lweIz2*Hf@qj&4xe^;HIlQ!tVyJF-B`y=YP
zQ8J-t0QD><v-k_dQViNTv3V-DmE5#9HCb-QsE&7LGD}f2^0d)FLE2K$S9htq=mlqB
z`<r@cCsS+r$0+q^do5pq)8%hs@r=tx&KWR7o<dy4i_vP74*=wBoA}hXD%i1wbTEBN
zw9ZH|h*@dj8!s6W4n`txII4RvWlSB_jDi};YYE&o;3`9M7^JX_xD1LQd$V;|IP##j
zk@qs!;_88k?9^KrsX{UQJQg}6!nHO&OW*SZPKEp>^Is}AMdmQaIlsn47rxnvX{k^g
zmv11QXyM^+s9DwYe7$x8RzK&pEVZ2bm{W7BbY`l06m}p7(}dPL)PrbkzV2MKdz_V0
z_2h8KTpfi1iB?4s-G@5TpX(3FDTud~jp)FsqFGCbY<zC=1-&m~*i}y8I0olZFMzuu
zc|77Z!kqYHlf0u<m)n~@#VENA>Mb^#@c)g=k2^|Pbeie4JKXh|==zNRZOT?~=!0!$
z+JfC?X8YK`*X1?^LkgspZK~uEE6X&iv(W-9WiO(9B_|_4LB`SmS5l^q92}K$7Z#a$
z3~BdjEV$(I^oqBtxm%!kA-9|DOT^dO&hG82mAzvJ6M~@?NEH|_(JIzj%CtIRa+}Wp
ziGQUj58$&cUkWH}_gqUHkip#9SdRWaFU6jgHlsH52HE9g?3}O4<!{Qvv9tCut_STp
z$2EXvBTdbS3#chy#Xa;@2b9E^%$tcqawF#qo&is|kp@uZ5PkK&hHj(uJ3~`sz?RJM
z!1tiK$ccx?dLN()Dh7P8$GKQu81TQjSwkN+^tRPN&C&>5C@c)NrOHLK<pQG=Fl=iV
z6l9M#fDyK6GLB4u`P4TWQ&~}NVLN3|-50h|<wum~Uey)lkU@2CGe=f-k^g@fRKLu4
zI)my>63WQzx6KYK^vGZ!+DW$eYkoe&oG=$WW?6mK->t+w3Jwsg^*BT5xv3=_<O>-w
ziH?6z84eLljJDZD|E^LGD|lKt2Ne_7qh%cJA5ogs)1vciff*XUOlEuzqI1)2Gab6q
zwRQE|szZRy|DpoHIh-_tfG>u0m|2ser*D?h&VG7=OeL8pvgokHGJNee)no9lS207%
zg=t*XQoD<)wafq%&_#lNk+pOtu@m<(a<P=*7f^!b^f0i|dEVWue;18~bGoD@9DO3*
z$0H=v{*VAi0x};GLRt{Ig=9Y^Cq0JL9pnz{6{WH<L+f&aHDd2-mCoXlppiF*C}^EO
z(dYr0?Xmf6W&C?7Es;YKV<I7SeO0X^6hBy_Z)@~$joq@X%#{CbG=6oQPDTo5Z~#6?
zPU#Hv!HuBgdOu;mxDuqJL9AL4Y9&b>Tpxm?kcRgmpQCsd6+^I*^=sjEpIzHMbN`f+
zH<jN}@0*4H>-2$}k^fmS(QN*IAwURgb4$~zq@kXlMbQ;`P{sc(<*`ZjUllVVB2a~2
zQn`?LZ{dtKZ9c#)>5>}N-K6lObT|}If32e^aDd^c^V7kAJyLQgC~uT@b)G;NEj#0L
z<mX5_oC&|IZGlk^&C<rtfnErDXYhfev34%K{)b|nIG9;f{GjIGac-uVTk`6Rj~no=
z0#~L73e7HxT=UP-=1}=taNfGyPWb|VGkpZnXFHJ|zBUEN=(jC>Pch^+*dS+2OMrxA
zO7b0rB6$4OJS?e<zn?Q_&@h7DK(n(kh*V^VQp4+`)ihU0cU3pJ1+Ht;cvivd)cUqY
z0qvla|IWIR{Y)pNi16Qrm9_cL5aEv>s{+g`4kWBgo`G?Jmu~Z8HOBLonQ&vMf)BGn
z_nI#&Jlt=OHAXmDAkEo1ar2dw^S|j9`9|&jmUXaQ7bcm5TTCPWziO`ifr9_0;J+*Q
zLxQ!zXWS_^MGrnU&ecy^=ljCv9E*ZT%>S~2Z&hF%i++c47>@Brqz1!}R@lanlEZ8J
z00((U@U=_XZPWA>+eSs49miAP3ORb|e=2=Sfo-e*jvVmV4l7}+clckr(fJ}lPrb4P
zUNyzs=L|S=%T<n^(R@i~#izOLM0MGa59TKLpDO*2c(z;8U)1Fd1zq#~&q(`4&3C=i
z&J=dq_6HB@tEseT4eh_;Uk&59E0s9JjNc9aYO9?#beJ?mMbSsSr5VoyV70(DXbjdw
zD~plJ2Dm<i<s-DRzc>6HekE3ALzN6|o%o#m05Gp*Va91kmY(9yCu3%CfRUEm>aZ$c
zT$wkp5m>|h>m#UVjlKdCi<)1kXI_b8H=6Jt|3vH)%tL7jyqtR_E^En8#;c=tA0pCs
z<~(wV9p0^`RmUslp!$<>7`du%7M9ph0~WKfa_jZH&d`82Fl1iGvYhCIbL?uSJ#~Fs
zy;mn_)ce@(zqyfzI2mH~<r~{j99vr-;~n}N!{~mG={>j_V8x`?bxy>f|F;Q;7KU+m
zopdG?T#q!igNpHn3z82C{oD3RyfI!+Zj8MGJI?EQHA#D83)>qbq^q8Bz!`05jDbW+
z7R4EEuNDAOBkMatCiw3-Kv(-!$F+bhooj4I2(Xh=sgPiUy5R|}F=F0oMri3^Xl47~
zi4S=LW@G!ujZw@$b{OkA<=<eZ8Z)iO3^F=N<0*F)pmy;UVz%NrjmpsoUuAxDm?4V8
zBw9~U`pahtfUvO__DQd(9^<8$AmRel=mf>1VXF4Zbwmrhc6T*2q3u0hw^$R+kK&q}
z_87+-i>iu9V67K)@U(RKF|xvzkYOx+RPbk1U~aa0V`~q069sbrfckcgv!X(%#wOB+
z7&Z-ANpqgKt|uLjL7a4qS*KR~$9V{BIeWCBU#;gq!L1v5zA&3iED<l>)M)~1_y48a
z(a}8`1<@|nJ>x{oYj#ef5Oq`NIr}%bX<G{o{V~Ojb^cYF=QbSct)Jw^*P`AUkNw44
z+4P`&W9+F`#}*%}pqk=z5r0b0euE&$oPTi|p`xMO=U~eS?`VS+f1;-SF9od))leRh
zs}noj(MA_lMg!q?{F{mycxZ8fKB6i#Q~j7?zo6JP#f*2hqRY50zfD1tTeKy&O<S>R
z=64XA*yUh`jkq!K-cG7hew&uUrr=)Ci)6PfeK;lI>Q{akjjGt{{WOCDeCVDFvc?%G
zpL4JIE<N}w3X*DNO4)L9+dT!IWiCMA@74S-l3prYIW0M;j$kDZ50^~Ejj?RE4UrGF
zZ8`0KkD)gnUMejU8y$av;=O^E6ar2OM+AOyH;*Ny0-8t+-;_H(ovt94y}e09B04O?
z=<hMWQVL)sRg%J2b7`>`Q6i$86Qkg!=J40G2xdf}M;~Ei8q7!d*1nhbZS;lkN<=6O
zCJgl)bQ*wIH-9xD1nM{VGYI^VHp3w7NX{Oz9B_XZILnDWSMcK`j3?q!_61dcu#WO?
zJrDedyjG$JZ1AHk6{H>ktY8_~8NanBZI3R<f2Gqw;e4lN91`UPyFUzcnO`5FeLd&~
z#MyKEiT6S*q}Jt2w$J|*VL#9&2W+|Q_I-RFi~?JR#0UhoaAG2)cf7F;*kZ<jItS52
zKcf&8ah<NX0HnavL7KOAt^6UF4#Pm1QJ}{d&;d*ngGEkcY783m7^EeAA@uke;+!aI
z(Bto$sB<tBeEvhh3xgg#{yzm`QOaOP83F1qi=Z)#aS-gVuMLLyUt{3R>8A6>bB`{y
z9rzhdPPsfl*wNYwS}T6LTMAc|*Os{9c(&qg7uaf@+p7FOS>hpD*jnNV8t50))VHf4
zS;cI18j_pWrLFFqVm%5z*lcj;0`_}#_g8hdPcbcCUl>?70w}}ry(y;qBzaRAr=(?P
zo+V`L#x-kduhy9zNK@omMsnhTxi!cdeM3jpM5_-cnlFi5+DgT%!<ncV(9ehfiavtC
zy4JV$%>*{#c>{P@qRd%>`p%}-3n6-UcRc}mhZeohL8^w{L2*Y2oS}ERF9q*Yh846S
ze_iMuixg}DE*EGEDx*eX<>3wInPOWw`_V2Tu;~A(=G2>@75)wNd=ozUf2PYWz$i4L
z#2ENrBeHg9_rB#2>o>(3{5MpsFa_VX;Z79cL#jK@;ID0BR^7VpTVeu!Ler@AMoC?E
zy^%ou^LlFP>w9VwvID8R7Qq(tA34l(>U#+Qx+OB2&*%yny=4o_@TNb%PDlKiXwf>a
z3C&DZ|9!kmZ8wk0$#|>{Qpr?Q0(%imYhzoFTn9hcVQQfe^gK1Pp&gL?Mb;tV3>QI@
zn3Y;-2ELI99AxsdNVdEL%4TVCnn2!jru3M%jZg}%&eb<F{PW4J@-T~SS2=qQ>w>PH
z(HkS@(F?ZqtbCY6e!`jsU=hlFV?m+0n{u8_PztsLcgDEu%x5$mg-vwrIUE~e*~EyK
zw94#vv}Nev3xGNxAAm1QkZj>=vaz*zax=@ltr~EP=0`GLeGQ8L9uX&z#W^c7X1dx5
z5C@<wvF`{|*g|SM?4D>rPR&BKNlH$s>M@KP&?qJKwubbBVN#Qhu(6p#6jsjEmL4Gn
zECw=?t0{W2nlOOqkrgV|s4(Q~ddh5Fry!GN{)>jtWk=6ba$yanybeWAQ!3u=o|daw
z<>|gJOIEfitV@lt`-|%Gukzk-#Fo9~wf;MF3WoDJMZ_;@TnF63(N^nFT+`qA0Tull
z3R)~;Q}%hks@p5Aj1KFr4m}>n3N#dWO4Ty7O~JiZA?KpVsNS@a{9n;Sj7l&dr5V^p
zDvYviT-4M!LV6BNPwFsIn2{o}Lb2imnFggco*cUgv^m+m`u?2Oen2hww*;@B;UP1F
z1kvXN5iL{$L_ZA)gc}3nTU4<j0<*L8p*dP9;rv?}bJy$;SfX(0zy&kUr*&M?`nL74
z_3g}{Z420MLwz&uh$>sx&ga!O_#Y{JK8^9qNNe_l3UohGHw$IX>y+1VIB|$zMwx%x
z8{5g7eK1x%M{2K^LuqcA*`()i2$%hGu+*%$hTIQRxa0Mp0o@B0*Q!Bkd2j@%;k(v$
zZ(WDi|7JDdT4E^0%zW{$%e{oW=~eTWW1Q-}9Ydk-jate9SGtyp+z`A4dCwZ)U3=jU
zjMRtm?8Q0VDB?NKUOP<bj$@_3V(&991ehjRBQ}$@FX*IfPPrU3yzx)!WA*Kvd%06e
zbDZ&+^oDsm_!jUYa6Exh;x0r+JIS+Kyoq%kfVPJ-$agk&c_ZZ9-5B4%fAGe>#_nsT
zc6ad~khj<vC+~QDKj(^%GG=>-?eVr*>_B5rW896WHwL$KPh$`7vYmI4Uh-doA2NP5
z{+UtO_pyb8;K&J>@t_pmj#txc+k4mdb@&6z-@zFJwKE+a!PJ_)JfEX(sCg$f7pVCO
zT6w6kkM#YG{iN?AJ*OwuMPnSPAHAvr)1s?m?uyoM%I9ZbRCcu{?(Jcg-l{!oo7)rh
zXR}><d48X_&yM-}HQF}8oA!hCAK*!vzePKh^!pnJ;CPLC`<Z(OKCipDyrD%0yIX`X
z<A8VQ)hul~xPHh+<RH&F+&Ij;a%7oz<a4Rl5)XEIadf@+uR2<r|7^nl1iWXr;`FN(
zp<h-+GMPxOA6Y-TevDDRjTRqm9K-G-oGa@S-qFSc>ThuU8b`fjtd&CJC=?7aQjUcw
z$D)+kBz+L(r_{tOYkYlzH@uBE+}ZJl@-eOXZ9eTNbGyCi-t|M9W<#QOf5-dSY=~Oe
z?ET|<GCgwwj}QAA6Py#Q=brpOdhY7i@!V59w}&47dpzwlPaC(lyxnNQ<Ba?XeP_kR
zy#ujSV*R%D+qZbbAL@J(XWri~M@6ir<Bj9I)m=O#%Ts>6esbef9bWLp>1&`!9Br5g
zR~|b7_kX$R1@CU}9=zaX#n%nJ;N3fk3s(ON+U9>*!M{=<txv~v{2;NQ-wq=||J?am
z_y2<ud!@4zSp~w;`UREwCkj3bP%v-8ukRoRL-0u@2;c@i=U11e@XT*Ix7#{YO{1!u
z?BPSINHExmgWs#zQwn6l>pw>j<em|N1gr3Rx(!tu6rw#uz+oOpX$`%4p6qM8BDa0T
z_l1VR9>?RE+`t1@ZsKl!vj)GYx<pl$MjaM~cJS%Qu9rh6L>b@c!wjG%%ggGCDG!(3
zGi81tp!eye4^YF6BT+-$d2^<AJU=^+YRjx&sTAi{aZHLi-1Y1&8xid8L@1&nh#2wE
zQ`*^iw|!uO;i%;^vvV$c`9|{OnOP_75oABPHrVW%$EzzieX2uXboGaIg}T+z)}pd8
zoH7ixCetaElITY_)`pt(y%$>r-71Ylc_n}eGvC<O%-6=LtTC62X&`LKPwSDNBNhy_
z^UL1My`uku3bt@1BltetOfxR4c*1#57QaknGg2hwmpAdJc!%>xC-k|?v!XK<wC1<?
zn1(6}H?)Q09QTNd4|4CRylAKM`ogRH0i|U)TGyYOcIc*|iN;zn7KU&|$7TIa#ah^>
zaZh{niUSJ78a3@*|GRZ5wBJ0U`orqY?@`)@<x?=OyW;8j!YBQIqL^F&wb3}2l{Tf!
zKSylhM4Q6-v{FR{J7aBIcsOK&?9uA;bi;VCNS1BbgP|3?XGX$n^REggR7qXKTMI^9
z&iUEd)tcP)GUhe^FfF^~YH-_1AAVxwj#u(`1cXMdUZ>qbPB+vxXDx1cwL}*Fzg9C1
z*S3u%yXKU_S}Gg?N#^?=e=E)Q$jaCDm!NO0{e^-&2g|vYVTS$~&&&ip(fnr_nvzw}
z*UVK`nsKInSPl7G3XF}FB!)-o78k!3`pTB`Oq<{~t!+kE$onlK2L3Z|)SrPVCH8Rm
zH~7`TwCF~sZ7+TMLj0tKiTH6y1tj%}1Pm(Wzyl;qEm<YDTxCk<zCe9)!EQAZN2-5w
z@67^sicV{P3=NbFUaNH{5$}Ct@6BOaAyPyxI8|2ZP^n)lq=(~s|1^_70F&&D5h==}
z+J>zK@=MX4qBn8p&Zt)WW79&}!J1X|CKa>R+SRmdepu`+CE1_m0tk$PzB9P*@Y%T!
z%EL63QSk~-dTt~vU(V%|q?H_b|2UoyoX?Vkjt5>q(F=b^IFv(H2~82vwy1o4(EOJS
zUW(No5)ZjPL~fk9VXciCi#WJ-l(!lwPPNBhJ3iENUjFk^*M%wuFEmH>=BB?exptV7
z66I$kRC{F8HGd-pK2bAtzCZV`?Qc5rbIwKD^V3C_uw9?kL<2L4trefCxU*JQSBg)V
zi0z4|i_sC<KIE{@S;5n~6TfjlIZoIOPGe}3Z{|bRA;W8VqR9i<IC!)=^CJ3Jb9&A+
zUG8K%-=~-<RzIWIIR(!uuq{#=R&E1(uVThwKCcSq6?jUkC@{Vvn*qx@YCz<t`(fo=
zP;f!PJG7bEYO%mp%OW>=nG$y5g!6i;gLuT)m1rciElKnyHDgm>OWiDFAf7k*J_6F>
zL8sH1YKJPW8YnPfT$TkQ&cMXYJMl|u<MqS_dMQo~7wjL)+G`w1A(Ih2&}8b+2%N_y
z`V>W6FhBH4xWYPCIwzy5ujiUVyWSP^GT^w}H%^DFdlo^m@PTp<&vqlmdAQC9*>p$^
zhk+Gnk~jlY*5k(0H!x;E=AG)omM(`p2VFYj8L*T1^@{%=)zMvje4Iju)W?E#jzK<*
zwI2Pri3R9j-b6m~GdVeF?K*qbU!r1=sR#IxJK0Y)Un6t&>?D5^2{Kjnbn#jz(q~CV
z6NVqsg`XLoLsI`avWNACX<SZD^qieMH)*$2&iSh)MR_@UtjYd3=WkOVyoDMk&z(Da
z&bh*zJU4Nho4YG&!bO50TgLO?43=8&)6+kp^nzm6uR^N*Z&k3Qv~N=EB?ShQzeTYh
zQ1GJ)eq6!dR`8D%+@s(%1;44_lL}<Q;T+fgu44a5!Iu<BP0+6?_#-{Y#HI$-#rk!y
z+K>T*`GOth*4EdU`E^~k2J@R#fvuWVZb)QO`<o=?V(jPX2*O0a{r@xfKmySv5Mz*o
zci+h6;ZJVETWwFSFPF{r7X~wTDevIT%w52to=gE2RT8hYJ-MM=s?eV}Z1?uL`g-5U
z?2T&f{aWVUw)mN5jv=6fJQa?&?#8}}GZe;$<GxxvLt{HW%t`yyPV*Caylv@2l#()E
zufRa8q~IrKyeuAG)8;8vRu6gq1&Vd}>=1&L;;Y~A$g`_h4_$-H!^%vEC5?9ud|FAn
z%p7{Afe);1Th&RR!Y)At`K7`qA;gC5of03Ho5TgKzCv@013blv>JAdSQE)PR5g-13
z1_#{zDdn)~_YJ&1PSCpQD~nojj>B+F5#d+PVOWE*p=TZ3?sC(da+;S<ISbS6Zj(Me
z>68|_=DAGy1g_Ek-q8d>AMmlg02RZHRuxwZ<;j-KNQe7^JQXyzOYPBg5Nrm##P-4J
zU|W-$6Xz(x?#PPekc{ygalmA9duXYLOhCDDNYs(rX_CT%tlg<i7Z$PZ=S&<eFFRI0
zf?7GfH)j;~4C#=6c&lf$*I}i3-lo#hIDgr(LTmE|`nPE9)-_QohZwrml00>7gmrz!
z)zx~7R{J-VD^TvIcv}u@F}?WcJ;mU(+T}C%V{r+I_iK1<lif@2o9|fkUw=@kCLDgZ
zCRoBH6qTwO=W8AsyMVk(g`Bo|@`P^MzV}0_?LE5u9^H4xpE^~)UsvJ13YrwN|EtO_
zggi*hY-uxfkEYwB6tYto_7r|sZ12?0!tV=mr?C7^jTZhuX~YH#|3iXT9rvB;Dg24B
zU1Hh74HMv|GH-B(L$nB@i5i+&A+rPwWV^O_x3s3!<Wp+sE%!@p6erbE&FLJTK6Yo_
z=D}shr@13ZSQqU`+BUr8=bfe&yV1)9b|&pch(Rl9S&@6gwrl0_!!6I(S(l3M=e~TO
zH{uO?LodtMbk6oJ$!JG|5cEE~Vw{8Kd*jUH>B)%%9_W`j$6ZVXvk4w>YL$zevbm^)
zCPuBDPA{chPh@LN_=OD7Swtz+(9#mW8P6->_iL)MAX%iQx1gO5yp|9NL8w8-jA?CJ
z!VQCM?JfnaRpxf?Db1nx5m4eV&kP1zEq2(L-^|kH`qp6BA@VywrXA};bh_<WNJ)&q
z!1-P17@)g%n)_A<ZWrzNkEq&j*F^ZM1aNT-HT`RPiGfz*Zt0G3ORs%T^M+G2&_`Gh
zTMO-4H#KX^E;`$>ql&R0F}wz>_Adz<>DoXzP!hs0TnJ)Pupl|Sw>sRsgKSk|8oQq~
zL5&@Zyq-t@U&tsdT-dZJ^LNk#Sd-Weef~xd%9%3qOmn(e4v_*n{8*tt81V#o6F@j)
zYC7*i(3de~>|S>6EZ;sa`+2nXQTUlAS01xZ+0YE2Wza7tFs71;*YDY!tPdOglIU`%
zth_n>vikn#6^v<ioL1}(f*^J7y^p>ZUBS-Xg_H7Zt{FaM_W3qr)I*M?VmsDds>QJ4
zXQL_SM0a9-P7JoOK-8=L|EKI~gF9wQqkT&*Ly*3LRQGa_SmTtuwiOoS(NQyDVNE;)
z&7E5q(4W`bE%Mw{VkhPupj$WPfLYerlK2fG3)HC==yyTR$*I7msUxwUnYFc{xUc_H
zy@`<K4t=lNNf2Z=Ip6cl`bJ-yoDWTT4e0uy<a|i|oDl~=R+G!7R!+LeEjb_ZB6$3c
zKi!n{Mb}Q!Hv-pOlD<)XCFzs)Y=+<M{6ZUVpbzE_A3*LmF3LwV5E5dvvE*f#oXG0X
zs+T_4&ms$sp}*CGZ1B8d8-qGA7Okyq)Ygk^@NrUyOtSY$sBpDJay<PGG-ZQBVG4J=
zLF64nfX-G8$E=27WP{hYuE865)ABjwc-xTr{Y2EJ`lwz1EQ^PyKnV_cqr8>mTL<fT
z`guoVxG4qPSs&y1NDYw!-jC!?vb*t63iz9LwcXn$DIldY^&Rz{8{{!L;23hi?Ys%s
z$o-J`AqU*cqTNNFeGR07$o*mqMM(i0yS;I8A`{%$PpgpbU2}54Un1{8lNs))A8PFI
zcCZcXCAQbwX|cnNy^S4iw7p%>uY0*h+Q!>R4){D8<G8pF<$y;}VM?I4COP0<lLPKs
zFK&{H#A^?C%0*~RDDR`@ebg*@-}AKcXrstG9B4?IRwTX0+i!BXKjq#rp41EachKhT
zp;uz`-t(V~qcU~Cdg|3Wa-&zILY<e4(0{ppo4dxcW1Di+zS-c~LAI_#uf!ijK6!DG
z(g&^d_ccrJT)5q2jISxTtD%*@nae%nABO5#hrGit=aASopQy2gZfHHnTGqOz7mv`3
z2Myx3^&&cfhrFXELp!{F#K!F~J#rLz-jH{U{+PI>wD6BR`dHG!<2*$YJ15IR{>0pn
z9P6aMCd<0rWLd}7CpL{)qPDxEzSx2!h{wGX%!u0>x6zuDjE|hRoU$$5cJmKv<7G=l
zfls^Zb}maP_C<efOrbW%^4r_?&)SXni6(|ZF~~1FD+wfn?NU$>trC-bIc5QetRmBL
z%4UWev=D?oihf#~EYAdfM(Sppw$T=|u}$+VGGr8_@W=HMI~07DsvI9?r!IFX_(f&e
zt=Jv~hI#+_qA!21GVfFHj4F_P%P~8eD0D|ZkkK$TAtvVCJ5o^(=OeQ`R8ens17P*X
z)ZSkuxSnu2C<f?See{#u`scz?OnvR{TL7A^dkDXWc21vvP%m*v!4U;V6&zDAp`a=K
z@o(dD;?&n6cL`Dp%NV%bRF_Qa4StA#1E*G(R$gq=E&m(3ZwyBprEk=w_*5-PjN^s;
zg3>N3_@IK60>cl!Q?WZ$%rs#^b~hc3WxSg2)(yigMBKJXV2tx(Vwfr2cuhg4+-2hS
zHY{TcL5qJvePlA0Q;MBdaEAg(N&GVk?o@D6ZIB4Wze|^&Q|z-!>ymzaNOxaQ@SFw8
za<^heqkU1u?$PDF3T)=xr`Y`po+WcBvpQ3*FRl0wDCI2*enN#FR4l35-m2I`3Tz%t
zDb^HUe4`eehN~Z=Dv2Nx@v)n7^(C=HE`A$?oBmqDi{rv}?5`!X*o(%kJKyW;5n4z*
zkwNHCh#xWBiLyNJ7@=r_Zh=L&7y>5~q2u>A*%+=vL&~vt8!Pd{dTrrZlWmN|!j@E~
z2mv?YjCE)@tFh~DNp#AE{|D8RqdXuN0}IS$a2}f3=GbGSZ}`eF!_AHlCG1n{Bte_U
zK&3ozF&JWw92FVbr=y>)#=IQ%Vj@9BfZ`g$jyvk*@S<LC*c<Q$jRY+i3EIQ{J`#jJ
zP0r3&v0JTHnl$LzrwhgC^YBz{zE*?^$HD9DY{dtKEzY_EQ+Vd9t<Ib`XAe!mb5U6<
zEA{hdOm5q%>D;n>#L5$LRDmQW!q2G&Zxvy!M(=WF{Bm&;hYeQzdSdN_QEeyH1}B4b
z*M>Nkrx%xJF4Rt*=Fc(I!|akhb{A<car$(~f0yumvU&Cm<aMosbQ8|4{l@6)mxy4l
zk3`CjLPCdwI8v#rI`k+n90n#BZV(#Tkn(m^Hl$ivP7=gx#Dzk11d{2?p;PIfA;xiC
zay1>E0%F(Z6ro_gHxq5gEZ|A!8)%3BFu~dc4mg~vji?KYI`XDCRh+|+i-V)iS7t7(
zuou@SdVKLA-FDn0^Mk&qoHz?Le~wuh<mHnWx426+&fF0xq*H+eLd+HibLRFmc|hw^
zdoB<}I=HIF*Qo`iVXLe85LN0g=<WptC)6U_7@M!O6<_m9T1n&7-`j(WGWeR&-%j=&
zm`gV!3hQk5bZ#;Y&dz8QqmLRaROAu9H7#QKO+r>ET8YzEyd)a8DJN%yo*02Vm$nqT
zuu-bq^?<8B2kS1u?~oLr(XvI~)<3PYRnUjcT!ZZcE{GNMTVogZ;kE_cTSog5gflVf
zy<|sgL7Lmy2s<IpMc7M=Qg@$|RI-sr{;B*0ezUJRDW#lS=p>lj#dlDvW{$_q;TH}M
zPlL+A6@c`@piU^sZ{D=Obm$fo+{ImuWbGyH^zvQ>Y%mj5wXLspO-pbJ!1H?>&~hBM
zh5L-FiuDugwb;OGIH5seYInRrY_qSEf2h&ZJYmLi+MU0{9gO^=JK{GPmwjZ?@B;sb
z83q4G6#OW`+5qbt%YE$-%*o=>wOvQe)3j!i60w$NqnV$>o8V%Rqa2RSoS&-V>a#Y%
zCWBuW=_$zN49Ct)c=Mk5xn}};#U&l8s~Su^TR}>yt;Rj=<XQh`)PldG;O{E<dkX%(
z0_I0-t^dh+j3X)h9@fbN#kHYltCcG&82IBb$0ebtY)Hw^s?4;4pHuJ;6bSkCe_nyD
z6)g?_AL(+QSWrMVh||NxdAyJQGo_l^$`mo!B!`vo&lNnZKvUNL7YeL)LkE9}%OGE^
zTn_!u{kU#_LBYRK@M{VT)LUSnUaR${met#dNH;(-SAlVf4losXX*8*G4y3&RpY~=U
zk$7nkT?7o=UXnM@!p%dbQ(#=&o6&g{o0p$S8d`^N3H+Y(n~=9dlUaSI-S$0vYeLp;
zt;xJDH<xl#(azyWM>@=C=Wf$_m6I%D(o!duNUZO`zB%0(gAUf6NWVF{c3h7!G#?$8
zV?;X3JfT~^IR_8pL!?W;8Lj|NlQwWZ<YDxB1=j8wy2+{+X21QS)!e4v9NzKz;9>Mb
z&$nt252M!`poZ_-x(07BtWjsaZ!Vu>J)+TzCUe*(G@Et(BT$`#(i5hoL-tlfCeZo6
zc^*8B)mZJHl4AtUf$bZliib4(@x&~yRISu7rKJCSsy+(mV<%d&k^b|px@_;TpEl0N
z?uNJ(+u?j5=&Ai@yBhOGLe9r<eZ0P7V`pQxaXv=jd~D<WDFZhowt!ai#vV8yyUDY+
zv5UUlL+U>9J<w<#Bj;XprHj^5bdJTxzWDFRyWhAZJL?A;JH74jJN6OVhi0>42OIEE
z+=!#qObXW{rl%~<#}YK}E~x1pMzg~?fse2a{}H=Ck2Qa&3%$p9AjS3lo9O9y?U_z`
znjVaJ9V6aedQ7~cCE9zqv7fg&*f_|u#T)5E<5<t9|2K~CB<UjabZHQali^wbuVBdA
zPwOUb-8wXpk9Y@+w}76o^~_P8G|`ygNe6k(q0PL7#>g`<)8ziij@FYpakxeIe}Sdy
zc(&U9oLH#I0dP8We-Wx053mu|;QT>)e<3;>>bJ!|-iqelf`pwo`gKsMe+g=|g}xYV
zIj!#b1qFXe3iVBiQ&bxa0NkzNcysde50vF3@WlU15~5q^;z^Sh?9X(4|6){V4l2~-
z=0=ZFM=49nEM_&`G-(%P5@EF$u~zM`(5Fs^ZZCdqbbz$r)0>f>qlp7ey}36dL3Oe+
z%z5<K3L<iifE5v%GV;^i6`@x6Y9WJQ>MN<j@Smg#kr&a;P_D&Z!8C1FBVbfTm+ndv
z>~-p{^mdb+kE?>;(qO(%v&!a|z;zo1`sT!Eg$jOKu)Rnr(-6rfl8FR5R9`_G<mTY^
zOsLtBh>w2FSMC`yc6=yppPDFzmMqi}qF%2_JS7d~1mcR5g7#S&B7t@(sHK5o>NMgu
z321*L0e#$&jxFl*myGBWQB%RsEC`yvS{04Cf_fKz2_vQDP@C?UkUS~gF}*BlLKLVx
z)N*Rc6%aFSHMQew?qrk7$+?qYg0%~(HY_>?a#+WkbMeyrRM95w2`Ohd|A2pa+O^28
zB^6TKj`2JkQ5se3`|5D;o90KIE$YN*%ritH0U)rT!;dO(Kolik`FBNQWxM#<E%|M-
zCBF>@+n*`oE>g@h+Dt6fHtK5&$?Rn<K}nZ)qDX`uNhp==9<e5YHTfnY&jOQ~He)6Z
zodQ*a?j<y{-hou~?(`=bt5d{@Koe7EsGFkRQVszn$XpQJ^$667PR9hVlE`xb9|ZC%
zfVdMG6ar3pJc#>efq#*Bv=3)xIZB(Ob&OT<Rn?FizJ8p$KTqj_g+XwhKD_mC%&`Qj
zI)4XuXOzK5fN&$+ngLmHkB&=p;@}=fD7Ov%J(Hfih+P>kgs5KO^Ew(i?SB+XZ@`I7
zXYmoND-)RZOs?J2JX6=^s5_^w);yA{=A1$?gT;Mm&HHm*!qwN`xOwxYV`6`bp>kgL
zs*LT#m{aaFLcb+rHR8i2z_hY!5^gDNHEybD9@<3lm2Vq^81$HO>B{mV0!`<1-8A67
zLm8xA@+LA_XWn8Q&n*Crzf7x9G`ZkENu-pTo;iQnf1YSJKR_?@lIEu9h5n?7mBE|+
zNp0dHPXr!Ze9#CK9h<46T_O&_ruo~H!FDj~wH%vE@*ytjb}ATCjY$iGCt*JHlAJOI
z(}2mM2?GJaK~6X*Lq*ii76ZLafb1nftfC{YwDw!Far}VV0Iew1_Ving2-ZP|)yFM{
ziSduS)~%}eF7?eYdB=EvfY?nezOgyY+-&hlH}ydq{7U#2)fj81N9_70j;MhrgG?F$
z|Lf(5>K#|2wuL|m)x=C5uIP@jxUc)S218HqsCF>q^%!XR-?g3JYg=U)$J-^Q$<iii
zSLW2M%b2)EG+T=Uk>Nm8EZ8#W&`~CsA8T4kn{tv&7c0ZxSBiKu@Ir7mBKSx62k4DA
z!tvfaFT5}6=li_p<eWB%YZ-3j^yEE1-jnk_@AG}1-%64YL6CzbY|{e}B+cl42%7Rk
zpLNq$E^SpBF(v4NgP{;Tl)rYXwcV<9o0UMrS){}?>|6=goSqVV+aiQWaX6!yyQWmY
z+!G4FB}P8!RE4OPV2m(#g`+`D?7_sKoWzPL<xB|y3paYPw=X-ApEY%>^A7a$?h9Hv
zs#--He~tQz=wzk6qaDbT3!yQAh?))%k~;WagT-fz<QYt%-uWPCdC2=31p?Dt-S>OB
z0&&7{9})xv02qHIT$T(LCStI#N+wGKg+2tlG6Vut2LvA&f(2mX6ic$|^8^b_0SnZ?
z2-rW$b@TkwaVlw$c{X!jzF^0{NYDaQ#9L4<kou71F#w1;gy;3QKKUr+m+MnFg9|oh
z0@%37^(W-3gYX}RjXbb{+1>jFpeXEvje@~OVWoHZD6kP~;u`^K7(#p{s+p*6*1CLN
z#m97+2Jb1gPEpIbp(4GipVqeQt}I~AGHm`tP^Hbh)f;<1{PpW{_1aZu?KB@n!<^X%
zU`|C{RkevscJDLogN3aQ2xPpj+B8-CWQ77`+Tz_FVa#+)Br^S1-|y!T7~|d6-32m!
zCYU^Q5$@8@DA4&)^Wg=aDgqs;qMh-}nZ(eJ0Y8!)QhRd~$^~(HKL~}f3&ZBna3GqX
zLrp?k8}6_Zt_>p)f|Ow%R?Umml=L-ds;SaT-9>;2qJ>LBw8?j{KEzufTGET7V<u;V
z5iN1Qr$n$sF92QAehIqLLDr%(49#`#uHN3)JckXw|A@W9(A(J{%tmxqrx`wcsJ)Na
zaMCac`&7q2*5(s!4lp9!@fW@L99_=fO=SThyjB7j!mL@tUS<r@#=V|yEHLxl(<YKP
ztus@H-r>OgupgmQE}0@&D^&!UbZj&--QdCE$dpNB5_KmG1p-pwE{dH_1xM>SL$1(W
zH0@fgEnA;7l**W$?l|v0@6YYC=nsgVy_<$UMgH|=9ibI&YbwT0g)De}ab5zBLX}#m
z%#oNlq21cLRc!A7U?|beeWO_EHk1|JFwK?6vf!kZn&gx4pXjl_FnWV80X>q0n1#{!
zZYpN^GW!F7max@$se_ZF*$k|}aQm#-H?&JVX!A%l%UItw5Wrq<ZV6v?o^i7$-FwA;
zn7PI}5indq39wW=r@hf#U&n8jA65hL6m(;rgQkDC)vwuduQ8+{JsJ{zF2Y#VMu%d{
ztON9@ets97xD}Sm4yRS8h1TtbHMs*z^uTcSf$|VOtlo7!$vrW|h+>Uc^%MT!A)H_8
zbN@{t<)iwizD$q*7fuJ&rzFC?1}5Tfidij3qSBmd_pm3C4=-wOwWCnJOb3U-_LAYe
z&XFNZ=CO>WhEsYOiFq}pSPi^jbGtb!RU%|}cgu0N4!DfdAs+Z02##~U8^^?zW3`Y)
zEzUQDxrGkNOAVFNLS>!=LznkuDAbV@k6^$L@|emFih19T7W1Az3SM{39UL-$cD)ma
z>`M$t@5yShR-E5iC~lPavq(WamnBC{e65v=8?+tskL@h*S#*-%o1!3t<`aj8W|Kf{
zI+$C`JqSJq*EX4KiWnRlhktmv&w0A^fq5hZ|EZZb0KqaxA>nBTDrgXe9)-CD$V#@Z
z5LTrG&T*E>fk8Eoku!8S42$LU+1@jJ8H8c2;N!AjO!^Cp-@K&Btl?zcWefR8=6hE|
zb%5FEXdZ@ZtMD1(bOF8617>>+_)tZ117ar(`>_98eCLjyK{0Y(B*+<0Jd!l`8ISv$
zM#K5H;G;v3X6~PK*1TNKftotQYtc{Xs_RMj_tWI*tZkIHw_PBk*r1igAR_&a9l{7o
z5nm;+hK!qQr`-7#W?RPFFpM%a0aYc~Hn%DN=oA?mA6In?RbF$m>It_z9+H*bI>{M2
ztyZ(`3BbIcw2@@8Hf<f;rs#QVyWH(;a)w;Q{0jrmV@JWFTW$PGR0gt2xwyK875W*_
zv{8|0<z;nsx>?_{^PM}J0gqNNmR-Pkw@Ehggd!HHu$I>qn4-H)?vL+^rx)vAh`088
zm3=#U27PlyMox0c+R5M&?BV4~x6=x0uDIj^@}cn0?Z0W@RTK8Dci<^yi1)ZQPw1YH
z#qQa6!zo)3|7jg}WfjWkTmc*JjIR4x6gIEDi){S->*262+i&zP=}=cEy`{Ys5rQ<)
zrU$Skm!qHv`fN}5=cli@hkcRD$J#g9B*)TeXTDaj5xH8K^UhE(IX(|BP@95}Se#oh
iaq4dr|EbyJ*j@cEyzTzF?aX}PY1*=U@`c6O%l`tvNJEhT

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..5070094d349634051294bfbbdc0dfe14e0bbfa4d
GIT binary patch
literal 54432
zcmdSC34A2iec#!4HyVxm7|h*s4~BDbc#5KUj)tN*geY=kQjn+?fNEfX0UDU<#&D)p
zP-G1yQ;I2ZZD-eOQ~qH5@hY*SD30S#R;z3hEAiT~*VpQkINnO)$WmgNa^lEG9Iw}F
z@AvnA)zyt-Bw4oizziN<y?XWP)jR(8|NkEA=_w}oS@^zpSAXYXGVv8%w0^Q=9^tqC
zQ^`caOH3xbq?cMuPA2VdYAHRLPP($pWY)@ZlX?E87Yj>WlU;T%vshf}p6p&KO_uCB
zyV$eTJK4L`H`!;`xyAmafysfWox#aLp3g51Ee%Z$Sy^FmcxhyE#PVH>qf0v`cUZo-
zIJUHNa;N3H7k4e~p4`2(XL8Td-pRd7`zH4-?VsGgRGus^9hf|@ba3*Z)hjI?S~@&=
z*z!G#N0yFG9$h*%dCabR7mqKEPmV90m^@+EeT#Q2ot!+mbZYX{(&@?5R@T3GX6fwY
zS<4SB-nn#c@*Md=Z^#?=MqW=(-sRol?eNB4PEOwK?euoNoS3}liG;V?+w*e5+f&Ui
z+*>_&Gc$P~WqZASl<lkDcQZYCKiB)cGS_9*q2x`J9PkcOa<F<o*AH-g$UDsSVY_}a
z*GIggTpz8znfniNeat(~^>Mp?3)kb`39e7r^+R6jxy1O%sTHQN_|p8MS6-UG-dOdk
z<>lGp<wj$r{=k_tbMuYMtJ9}vmY2>fuFlMR{?vS}e#WcTmgnp9wYgK?{Os&%eSW!i
zYUM^_*`K+5W_odX`pkXzoxOM3o1M9L_MUV1&E9jrch_CB_uoBz-}F7_&Q<TdvwCm!
zt~>9&d-|?+OO?2#N~@*#EnDxMUb(Sd7nfJ6wW;|t^GhrK^0lhh?w#vZ^-Xu*egB)z
z-GA5g^!@kUJ2i9X-S^$+-Q#)m_D$2%)80AnO>a7TZl)SH(0=imxvAB9eSWIeF1k%0
z9$#Ks`HuPena0$T2B_gr)#|g${!-Pi7f+ok^LJ*x_=Yyq#CS3+T%1~(_NFdAJDv)=
z{OXiX^|@-8^Q$Y>sRo6;Gt0F`^?GAzsx~!O^~3Jk>f&Nl#`F1=DZep4wMaJ8xauwO
zw`*zYO0_cQ&wE_-(%zLyv+6HbmTQYQC@Ef?uQwiFuFcNRjTge+#pSt5<9eeqRj<r1
z&4sz@b#B+gyjPu_T3u{}+4=g?)Jm9Vx|iqxtydX@*~O_wSejX0tu-q2U}~uz?qFdw
zrfPJiQLW7S%S#n{sh1?f+{O9ETbHKvi8V%IYO!)<G2C;0erj&HHnsRfG_vP;u|Ge(
z+F&(Nb?`~%zcJsqabe!CH=b+g<vowCHkPZknPso)KXq|D9S+*JPFEXKmHN!(s<*mW
z4Tl&le`=;tnXfgLD~)QczU+rZE!b*hasEn`4t1YD|K!upO)ae~R(&q}pVX&aT%Kcq
zpPs)yUkeK}>irBK*-g%0uGgm;jarzz!8>!&ovwCyddge6G?UV#@}s}(sl?;_)-xnQ
zf{m7JPQT+?)qlyKZ&b@me8J*rx;bAfzr+?RH!jcD%d@Ms8GU`Zv0T=NmFwD#R_jdl
z+SH;wRxI=5Em!O1+H#{jb6M-9JXO1~wCq<;|7e1d%haojvtdT>@DKBPOaAyBXD%-<
zRnPEHb$WUE${BB}F?D9nUtV1~^W19fB5Rcatew%=o;lC1e(cF-&n$XNXRNDi`PN21
z<9RDfto^Xpo3BqzFIK&Z_3((k)`lSMA16PfFV~O$3TN1G>>(2*g5+4DkyuCuDK8nM
z=aZjE4pX*6!<3o`b6)k@{7jWE$r?xrGc00nJQJq;JNZ2qX6u)yR+u^3YJZ@AjLUFf
zwKlKa#%iy)X6~Ta^(*P0AwR=!{If60)vqyO)dV;!F-f1il$Z80a|ti&<-ELC@VdOB
z*G;bE^?1Er#_Qu>KYs`K&GN60U(3zz`>w=nQqce9uF16bcHsaZf2NxCp7KV$9WN(u
zCMR>%{CeS5qS|$nH_WEIF>j}p7rkBHZtz35x5wMd-;%e_+t1$~uk0P*Z?AXIJH+3<
zYQJ~*^`t@j%;bP;&AVvOUhNM?B~Kf?SIJORGU1){PT9MLy?1!0y)*P}#5?QVNvosY
zGeE$*DB0n?)4SWdhuoNVuXi8!c6#@FZ{qJR?*Z@4{N3$6=)Hx%;4giN_mKDS%L(4I
z*L%c!l=6MvW8UNZ-5=e5*1O<6Vec(_Z}Xny&Vlb@VN4!m#-9r_=g&X$^c%mGX8v-E
z^EIGgm|8qHaiLjY(Vtml_gw-GX^+04jr9f<6K{AVUp@8sp5OlRy}KU%>Ia5D`^0ZF
zzwhC%{>hL2?iZh){?Nm7tvQY|Uw=4ES6AvY<*57se+&Got@_t!wUM})SWn*K1;Gw4
zc_mlh)kt}%o5>F+gB?Mll5S*z#6or<_k#&9O)kyf%%?Ipli+$*bR{43k`{s<(ypLL
zS`4~LyMq#GDcDKc)94M7!RXBt`MyR!t!UlkUQO}NF3JWd8SsCyF{nE?lavfKGG4wh
z+!&!`Xnph+V|+8on+u;x>&?NKm$;%Q>EXf-|L21-@?*g;>CRw?bXUNWmEA!>_57~|
z1+Ml4dr0>N`(8`>nZ~{#do%f3vYu(A_|mNEQGTBG_6OM@7v#OJSE1*uwn_)6dscPn
zSy6rH>}{~0XD@T7+j{3}zt37$|I0SIiG>5fV2}<{L0`}x40xqibF|OsRt^TcNDnm*
zFB}PW^RA=pq@Gt(Yi06>X!&!EV++U0k27k$uWA&>YmDyt39m0G-_j`VeJ#22ON~2j
zYDP{{^CVCGX5-Yt>Gd<WRCd-AjOm{kPy26X;=|PSvtjaD^X$y3?}Kd?Z<K{4fx9a!
z%iz`8T=^i2<{<FZD<8amHp*N(*z9?r{OoG2rW%0oX8)-bzv|69=q-qO3AB<2+)ppp
ztIhtWs?`-$ee2VY)}gM-Ve&N4t$ThBbXcuisotoEMZLIUz_wX=wpy<??Df^9DPe;Z
zAp*NueyKV?ceznNUA`og!T;5Ib#`^J{1VW+>{si6Oub4Nm~NM8;FNt6(0%60$}&jt
zbn|@qj`mOa`A46fcyi(izUZybz4g*1vgL_)Jzb%*<##=E{?SWseXe}z9p%ZlKKqUh
zkB#TT?z#DE)mlaPw;uLvM8jQ;`6YV8@P#F}sVmS2ej_YKSJj#ao?V8FtXAp_M{#z3
zak0XVU#)_|=9@y)J%07->O5ZsJu~G29hT>=P5I+@f&x~-0AZm5mY;7_Dq-)V^?H>l
zTduv;_d)pmMwV0?Re!!YrQ$HV!oW1bu1aNQajIVD8J@09Emg_&gVR0rZ?5X*Anci0
zoL>Qj7~SB7-J+AKOz4eDBkY?MNfK3O_QIk8r;0kwCxMb{)x~-^#>`Y!_&~iLqQ?jq
z=B6I@3&9%M{JcN4q7TX1<obhJlzBbJy!per>ash!4%3U(T9{dcP73qWtFyC^lwpr5
zw`d<eJI+SrSJ-_`eb6{CD#I*}xq0i%;`DOGt);MUk*<pvY8B*{mKjy*cN=}h0ME`h
z{2q0u*T&n?cUM*#;Q*6UZB%NjOO@3XXo_lGMRWAA>c{;Gy9C55jT<Xf76b6TR&g`w
zpHQ0_*6(82qc+-$CRMA2UC)TN)Fk>(sbXK;vitlnqdgavR%ak$MNBceVXC$Q{mhPC
zUJARPwZBl04Q58wR{YxB_yJ$kO4z+xqw9`L_0Ln{Kd$78?sc(QMB6ql`;X`vLV3zF
z8oHvV{8`m}uafUj^1PB4l)OvHB_(fFazV)xN*+@ptmV%rsgk&LXX}HW3lRA4(7kCT
z?^d#?T3UQ8dt2(h@R?s%G5mstUMpjyJquj6X2J$-Jn6qvj~rHVOv%$Eck)};YZ3>F
z$^K+6xr^V<f2CxvU6+#kQbpxbb~jaUwGzEa_gpDiNR}v5TBK$vrJ93W52s40Uj8<F
zgx*iLcIr%CZ{??#Ur8A>=Jo$MB@7&Z-F$H5TD~^efI0+AWQ2vh<g3}Wyq5~{V~Mpw
zBejqQqd}p88@d{)^$d5qJZPzpq;@85rPhi;ej)qb<XXCs3&4}W{k-9q0&NL9cP(VS
z^iJ@>LjFps{-H+E%L&gnx*Mg19)5ehJS~I0*ZaHzSUi6-Y1l>AU0io{Tno#C2jlzQ
zK{q|>54vvgEpchdm8zd(i4Avp?9RY8cY52;?B4iHU*{dKzulrInqqJ)5#)yxv{)QV
z*k9hcTVFe<@*Rma;mkg7fS#8+`Z(x%Lhpxyk~efKwV|KG8VCEb;w`9k*WcaHBk?zm
z1l{hPR;#dGt%YH4^h)Zrln;fnFyiePN_e7NMxp+@gThU<t1sB0e(z4K_0)W>hP^R+
z=DfF4EjLQ*V;Ws+e^={DUr7DOtp0AR|3U6Sb$#>q+)DCfp|Q)`W7J-c*U#GM@zM`!
z%=)~&<e|ew1<|7Dl0CPQ>w9k{_>RvtiXVn&fTk#mrr5aSwNv@|`olitB@AGgnS!}~
z<InhM-Z8bfa(Sv=zC3je4sW?UT`j|Lg9^Pe>`S>iH3Lr=AyA&odImeQ4%7C!hbuFy
z5_EXs(8g<ET=iNwRKM!O*4u01ntQEq8H|f?Vzs{c_8wum+uq%25Bt>_5aXtHceigX
z-MQ)J8~XZiFh)Y4cppK^Of}rsF4K!!<>2RekfJwt=&|o9-Fp3J-y9CNE1LLWX1O|h
z=eDKijIMhFXp-<BB#GwWOEEw2i2oLfoBd0x#y>}tUeyCHU0a-v5MccRV_pO=*M-Mw
zjq!bk{QRA2wm}l+TVtOS{;qlcX_clSF#MV-wO7J~E(P-(uFFLNiTo@GrE%9e|EeC$
zX#BwM%X5Yt!u%3TWodQEp@R?W>A$YzrYh$yIhjTmGgUW{MmRLF+;~zLe2LyxJx2l|
zx~Wao{MYnMdTMcTyw}j4Lqvw@4Cl4G`h;GgmxVbuzxF==Wt9{pOOb%jbwrJB;=TO^
z%GAsHvm^;&3h+UJf4$&>0{B80WGvO6>IdWWr$@m#`nNAV%>Dk<K5FjbcV9|4r3^kf
z%x@`qkY`Ig(Hu2A5|aQeN@s)IMF>UcwH}bLRaph3$#s~8h4gyn7Jw8sK_#Gqg|uC<
zfnkmw1X+l+N`4BaFPUf-y~KCG3_fq{9qnARr}7)xaZg^0o=mwXU*HJ<LVBXP^MUen
zNOvTd(jFH<9Z~Pi-B8;`tLs)oxwLMM7aU>nBA)=E%cxiV_b6YCX<7tyGs`Q`qRaj>
zx|0Y0tKN9ZQ8wH)so|z63cExd)Oob}&v>jpN|G>iEkxa$Oo753@^1FpV8_*CY0d~w
zh;P>VsUJ@ZYa_=o**`cl<E0y^AhVtpr+Ovl9}OtknOK9nY?ZEMuiqbJ&nI3yyOz6t
zB1n-txRwuc>zN>bE9pPhK$5ba3v$9f#uw(^lDPWV^NFkXKd*_)19jIM1@X-<CmLPr
zaN6tL;_cQ;OfId7LtZPqmb#v6^lZKF<)Iu3ucpr@UQ53C<@Z8AKrM6ynJan!7p{H|
zDxqs<Vxbq@UnDIx!2UOrq~J*?g|9dIpc3rL>$;MGx7-7!gz^yG0rrb>V4q}tkLtqv
zck|rmg5rArE%1|Tf2}uaqo|sG<NC?QK+pr#(Y=%2GakFv`j94o<yA785H<?>9;C-t
zpM9Pb`CzLLSw`dgu7Ag-=g+V8U4JktD@JXeSnKa>6VAKa<KAduR(+@apno=H@A$Od
zarJwiXJ*%-2~-OVIdJ{#rWW_n;y}BdEdBI~_aRWRR{qGI7bW4IAB?}U)uUY-pQjgt
zK|jx?Y^DdH=lJ)Ll=`<ec>OTnJh;#s&xWw?$AiK3VXs8*26^f->!T+a+^cc6k$y07
z^>dV^UrW`0l9e;UT10yHYO02$W?{5_%_<mL8)n`0835?#O+zMS86Y=oxxr6C1A_nU
zj@QlHr~Lv?3<X0!m~h+P)s5~Q=H75H+`4A~VPqmq&#bJ1SxQLMVSo@Xi!Uw0b|Nwt
zC(|sSuP#=N@rM^Ee_+kA@$CwyXM{}k!_G6w>T~?9i&30o54>37Tas(3Ao<^<u4bO=
zfY8XTQ>V{{8S%Seva$&-{BiL?ag}I!Rv!9wH7NX<OEixe5ZK7^oqkX~U#vewru9to
z;Dvc7uUJNn6u~%wyju08)+HQ%#;ew6B-Cw8E&8A275>MS=#%^pkqGPNqy%6jdSmct
zl8k1~$;q2NXsJ-dAmNZqrrBSH8@s+zodN5WB}i%ZJssVDimL~T<>pZFUF|AS;rNb-
z^AuO-S9o`r(bF~=exoj#yo0PcNt34<j_rD%Ds~~j_Z!R0aQInxeY6Gr8dc9JLd2}X
zmAjINN%sXXjY9B64e;0^s5{KUq|PDh5E>7&($UnzuGL!o>T0!$AfeAV&i33J4j3PC
z06r6Qrv9U9GFA8dpHL=wB}~m<2@ADIFXR7hd!u2Fu#4%bnD87;O!cMCY)1u`P0okO
z1@tBr1Pm~~i(Z`PNeC1-%3-lR{+kw+f0;Lbk>9!qg9N-_KS%~H(f;*<ZbTb6E>d(t
zDme(I2L(YT3@3Y2`?!jzhvYud5d1q!E5p&%F6f9+wF#fOi~EQ9)xYM@Hmi0<`$a!F
zelz?U{(6iIT&HOwg5*N#W;%cefp3Hipfn99#a>c57)<pt@pabmrn6~KeQIhSYuFSs
zWqzKhHqhrl(6c>S3a#?h<JT*Va!|frIXiwjuKIwDHHA#G-*~_td(TAqdiimFPoF;h
zV%d?45*nH$YOz{J=vP}=MFP0FmFRiIDAQG6PsBQ#vdQ7f$BpPO-#LD|{I;rJJy|}B
zIui9HLQix}i0V3;^sA_XYY&tkEjL!7D|wchC(A7w{bae5Vjn+QhQ8D*qc1sDFWWL}
zHyATR<5_<f!|s1XiKu2rWPTU9@mxf`M6HC$25Kzc8=?q^c!pBvdpE7Pu<vTc_=8q$
z3<wf+ZN5Zas|Ol#*w)6TYB6!6g^}gg7RPZiTNj6wNO<j*OL{mlCTcG^F`jYX^$L%<
z@A?=yN{{d-Of&obuPB$bdEDIdu}FUG7J}2CTE)Hk1obPC8ca6#TK8{P_b>H3+Vk^K
zULJji|1ZhgcNiPB`73s9)3as;KwY_ZDE^KN-?9I8-x1aRmcNAY-!vm^U{fTvXXM-Y
zl6XFlU%T0edUi+r9b4yPbKUqW#Qw$WO59Pju2@Gcpc8#Xt0ZNDnikNBO4qVhOQ6S|
zTgy4|m;oWaH|JL1tC}O*S)WkO7{~7>H=eh(=>M`x<B1G=H!|1l`kQCB`)XT~e5Zy%
z%$AnrsBv`7{%CK-?IC)UAT%C^E-vC>kN_2`9hqN{LS`rJ8eGJclrIer+lwuRT<_KN
zTGr?_?1#8_Z;)Bz^Zg4!b{+Yjmvz!7;Br<^pi4`?H@%iy%dZury9x?6H2DdW3cws`
z2l~INq(eCH^?%X^PU%gf1GEzB<@oS_MGp1<B8CGM|6g-4-sN_8^15qTU)vJcFRvn{
zzg|IHA{x&L@dcc0e8bUrkR%C@MIW1oyl-VAKzl&LKFC`VhDZlNmKpw)Qll-<w&{Tx
zjkkVE{KjK%%yDvov0q5eCchsp<bBDtti~-K`!q6Fjb(w-t_FCK+DK8sghix$=*O=h
zcfAGhTgV1k^mG|9!3*#&UP{utA4+{90T%`T<Yj}d;lz6W7NtQk+MGE_%l<80YL8o+
z7G&9oE{rDSOAo^j+iEs_|Cja9Zjvx-5)@;7%^bjpvoyruc*(8vjqQgeXH!!#uo@<(
z{m-Z=YxGNMZ_v%rhUw5!2}`x*npw~+P1Ub#o}t1@rEZ2K7uq{w6uP80b)qHtZq<ls
z5Md<k1w%7GLWWUi^{+9?Uea;PNVu_#VEw09@7cK|ymGCPX=EG7!l|9ViIBqd(0Qcb
zaI-H?*xVuEMmi2ZJDX&-;Kdh;UiMY(t#0iN+R_Y{Mxk%uOFwE?c@L>Nd+RZH@SY$m
zk%;*106tr?_gSpOf-ESmFcHZyZOocQv!4+mj+v7GS-!&myCmdoNWW8gxBqQ65k7rb
zb?oK|az_4oHKUEg0uoE;=Bnp^UX^M}zNTbJH8YUQb55sX+LiC6kVSDs6|^$KVMndV
z;-`wGl_wjdx_i}Ebw4fBrk;&=e@;&yj+0}m*VCFRHs$ilJ2azXluMQ+@klOt5D8T=
zxrcm)lIEa8W6=t*hu-L?emv9Kaq+G@$c1K_dDd)eRs`oVfb;Y$lqEY(7I`V`!u;Q%
zE?cCnR{C8PWF5FEPER3JGIgU>?5MWsgK}<dzE59bbP(d5oM>b$pwP*<yD@wgnW5f_
zR<1ul0=h|?Jqx?~<-*Mrl(AMH`~MZF<D2PM*wa=*?_Eg%LN@f9o|vJ5_(x3+@oP8(
z!9wyCgbI#sN;y`nD3u+0K`gXvS{MXFRQ5$SJB0^<yco+pG|T7wzM78ch^x^Abh4(d
zCL`8Mz0DM7Cq-i~Y2yXGuJvm7)>i9P4WihG1nMnl7^l{1-Q6;Xwz)fKNP$o>7-04o
z*T*c!CK~c1>a$k1-`P`xXNJ72yPKvuJBKdt+JZW%c{FmdXtQiL;{I(|k`4(k7&!0`
zi>e(nBN!HE+Mh;bmqKinsXe4+>!kn^la!xtq!%&^S%h@d7wWz`?MfD`#1wNdOsG`y
z9>yhz!4O4^=O!@fxzoRkt|5ki<AB*mgXyiB(L{6qYVAsG`K20s1ssmF9pz&mh_%LR
z$Jp;z^k{l&y6*pP<ZRXW%PK3j4Ckg6!@`ox{Aj~}om;IPlM!JW=3ufI40~Kag-V+S
zv7oWi4H?CnSS7)JFR|OjwPxT@M^lAVH1(~z@zAz#tnHsNEo|LC+WD`Dt4w=o;4v@&
z(KNe<Va(1aAn4dvUdEtt8ZD@V*J&@y^|P-exRy;z+RJgh$degbVnN+f%|RqiCY)|r
z2QlmkCIpBt>$O+Y3wbQquwR=?BLhQ1%H4whX|J!A^7_}hfc@x2d3NAash}WiJ&%SB
z`K#rr3+6^r3LXh0FDy8W4u)?%H^_6{(KD)t<kLu4=&-2eA+_9U)!~)#;Xfyo@c)Zu
zzH^V-1Zt!De@OBn2HpR!WLg}CO=h*e4P=NYz4p$3vl<au=rlwF&$Mh0A_J0gbI?>q
zH(>8$wGkpKw;DyUIGNVH<XGEjZ+;0`mt^~48mpFO$t>9-ON(ZgyF%yG%vyH5$APLp
zQbTQ;^Z(GLma>ojG2xbOp`|ehbI#_%j0sH9V+IB-0%#*>6E!N@=Y}s4Uu2RYsiV(4
zCw$DO*h!*(*ekR1*JX8p&4x1!2@7^XZ#N9S1CWwsVtiBoYqU@o#1nw*hbYyq$wj~Y
zsX-(<y{R!Mqe1jNy{X~WK59K5D{=&)1ZdyEZ(YMF9fQ$5036moG!pa;(irHY%z9QJ
ziYw?(=pJy>53##GA>@+YK(WC77~PQ<qKOrC24#&X`A@va@8ugJpkeeWI$Ixuo!pg|
z^w&|Vn_xyB2_bjAAJXS%Xe5z7m~55|^hUR0dO)`Y9Sk`<L`GZ-6dw|&AiaR*!Z@IP
zqWrAX;Dqolcd<akZXxq(ay`cvQ^FT!CwyCOsnZk9!Rhkl@=N8J@~MZ))BNFrW!SS+
z#WMIt#fdQ)1ChxNx2Pjw3bm09%Yaly^K+}qsD&e$=U1rE5@2MF=J3CurwtKp8ig)F
zLlaE>h{i#nFkwWKM0{XNr%#G2ZJ29~ij9b)kc=l}#({~b?E)xWq}qa&Sc8Zfm9Qn{
z<M82ZZ(7qessihpma<<|vZ<5Z2&F&-2wAEtpW%Ll&U=CI5%=0rA)d4%-4pR-T@?Ru
zy}@!zKy)t{DaX{I`Y^&><#wg%gK$T)Sn%)mAiPn?BvFDv)Jq{_YM$c`+_Q{fWP=RC
z#>|{p*p%2?S2M_=#JGU+s{lcsKxYY1W-!FwYQYyUNp}AVuVw+(Fa{W?0NA6_BC0MK
zS0)y^{9ot!;w>=(D3!bt{Zq}n*MsSdYok>=&k*-cgvk}3XA@xr)&$@HcKFsLE@6vZ
z!17?OGL5OOjPm4}kjLa&0~;2|R9BW~E@K~vGYFP%D!yRc12ll1@GGEkz*L<^9m#LY
z@+B@|urDuT{_T(H<-naeQ;qoFud4eg^QTm+xV(aW1O|6V+e;=z#Y7Ne_WF3yF$P)=
zVV9$n=e_H0<*N7gRlbq`?S&$2y?M3tw7{(O|L;+wt_3RA+exo8@Ofhf%aCiM>+Ut>
z>1ZnA3pP9+uO<PvE`GK9>NOJP1u$Uq>IedikUx1;v(_&ESfWjazrxD9nf#b(rr()(
zHt~YySHcnrL`*0Wtp!N(jb)~W%1y*Nq!e#gcZ@KXP<EcCUL0H#g~D=%8-<qtUehE2
zNit@Ws2#=b`!W86X;dZxF@_t2N#Yu2Xoglvg{D>dC5@imhgRvhfksf`JobWUz-)Ls
zFm(LT1s0mWTQjpqiA15kB<(OpS{(kTY0sBnaRUfo=n}@;PxiMvR$Gqy#v(qUDu%rJ
zHg)6YxZ{8qng(Vxxd*v%BsJbtH6BLIq<#%ujgnzL5<W)U>&7gJjIxmjY^B?peW%Vz
zyT#hC4oO<Qc?}LSJhqP5YU24YHTj~tD3F-|xi)v$U~a7y16bA6uR*VqWI#1ik*8qX
zKsOOn2g0!&2uH#l#10O@WG0%UwH7*Yn5E`$OLJ_x9A#C0oU8E|>tt&F@=JWJ?P`sU
zuO;8YC3(kp*2CVKoQ~=hENv=05SAi)Tk~GpJQPC}VGpWB0$FJve^p~4_$*u_5*<K3
zW+Spq<u?SxjYgXhb)la=e$}&Xe(N<FlNZ3pwDlMqZ7Xn49_d#eAgvvtK}vO8rwupf
z&6os>1v~*#vIRgElV|g#9!xayH?dfR1_v$Jov!QG(G&xa4FxPfQ;S+IMiQA%*j1N@
z<_Q|VCW}O5GQ~N-g9rXYmV^8aYXTXs@V@n4D1jn%`eYF&i^T;j4{5_mdrDC60|EMR
zBsw)iAcFyRdGsugZJX>DPV=Qh9%Y|oL-F)5lP_)|1(NAm>y`0K1Y&x4s#_Lu*aY-i
z8zru?w6V|_Ss!iiz3V$<^??K_<wi@i4EJAn1qF$LFSPG3BcT%bdM=^T@jgePg*m6v
zGD%T0cc(2h_UTud7FN+FI<_a)YuTbHJhQlL%I8bKC5Z)`Oymr+8Fsf86w4;wuL48b
z5eD|OUk&W_w!xkQdF}NV!8=^V=uXmMqQiQT@g33#$mgpC-VWLN6W%hb@J`+%hHT?w
zzOM#iA|_0un{#B@Fzq?nmGsgb0x^?)spF3I{Ts>|(P~VQ@eSXrvU6&{VB;U?vO$8c
zfkb0<^yGIsMe%rF2bxSypqA2L`g$Vlj>Kuwh&!lcWYs-tcjH|6T`JR1`}Zl)sI*4N
z;HxndZXEtC73z0WK^8|+2KNJ63+WuFvmY=kRj!o5y|4y3Ou+UabQ3}?Co|Ht^lci|
zcs4a4`Zf3_p%u$Gv6gg#Drbl!GIAk<F_QiwOvMB?)yOU67YdSKNbD9tI}^Q$6d3OU
ztI1AXX4yW{M6(faY}pU<4ztTP^J+&tXN?*OvyXaHOV77-WojtQ8EsMz)6{6MD2IH8
zo#{Eg9hEjUlh#sl<g<o};bBmkJMHtgtB+vCKc_Cqq}#t=$(xjDD2!6R1dn=wCH#97
zkVU$PbBffJcWK5xN$I7D@k9QzWWr)a?hklT;uBG-RIaX0EjkYgGa5GwKVKAsTV5yh
zNFLNG=L|POmv!@BD3P+n|1~B5yAr8O!VI2=i~g5&{WT>LYTA<XZS@+jn^v#HhWUd^
zY=;=O$0Ix8pHy;6iLLUtDEBx?*b}*R%v5nA>9B8MHNI1Cv8{TR+!=n&I<SyQVQuj3
z{p)`-a-sWD$x<)oat{?sg<PS)Z&z=L`6}>Bc_~-S7IMYD;(oiEEsXVd6*7f9caHe#
z$qc{oPhXU){{z0J?TLsF;ddD9#0wFR!i=CZ-gDV%<U?rpkn13w1*uPVdCyyk^FJ(_
z_o4g=<#{N7!WYZ~J&8x5`JdxaXg=qYea_c>vfp`_PYyWGa{TV!?H%=w;R)FCAe^-J
zk9*@UgTW?;BmchlcqhC&%)f8Md$02`oE(kr$;%Lkk5{4dGaNH7L-`cG7<n1)io6VW
zM_z_|y!V-(q5KIe)qUQBuP42?cn`gtjr|GVZ+?dQV!Y$c198ec4pnj>Dw+1qdvCQq
z9gIq5yeGW3S;?WO<Vo-C*0;mawTGAHMXPZnDyiZX`A#c2>OJc{$0#0)?p*TTWzQZr
zPsVYu*=*Q%5m-@c#O^)K;*wl^&EaUm7x24qNl1;KuRlCvyPWJGzoJOA2I{Xc|5(E(
zDw3=t>o9LYZ22jX-XMiFIuyp0Wc@@VgN#p}eFguaAm=~erIF=jpf*w_F?>dVu>OoE
z`lDxEo~-hX=(RkniF}{syDZ;t`Qn`F3^Z~#VGXg}LxVS;40`^%)F&hhL^_Fwqw}Fa
z;!-Kz7;AJRk-=+1mU$k+V@ms;Pt@4D`sd+kQ0aSjqLzY5Db~MaUXJ}Wk;VgFPh${I
z16lIJhEXysCdhIl$TXEho>>{i0|Wm?^RXPW{LY{^@=|<1Pwm30aX>A2^1s{}^g(e@
zzOWk`%!?YK#vTt}N^JM`Qo7H)LG}mz)Owja{kjv?ZrXU__vri-jbtB~aB9i(%#43P
z$w88E7-yC`en7xuujSoU@o(t<8p#F!gUUXl<O52WQq+ib%;s>7$2EP~zx4kXC%*&X
z4x~XdRXn@^tnmkf7foCl4sP(;vI?y)oxTBQ=Q)$N^vd#ze^F!cw8o;FhA@9LU%OCf
zu;qMbypFueE$~6B6hg$Y5|1*HVZrlwg3tmOL0_&cDM~??df}lysQZK~NtjW6n@ZXg
zo1pEy7mjuYEogTm9Bk($43l`z^p9ch^z<^OJj++AHOU^rAsJp{x`j|F_K7pYH^=Bb
zqc0v-ViMgKm76CSA9R|$!A?)Y=%XZvYkjyY@+b67&nbCU$vc(2TZyeM`(i08{il_*
zdC7P|w%20|9g0wMI8pP_Cz?Y9K%xii1kE8!JV`k)Er|Y&QmTI`Bp#}R{u8oHESHr-
zsEPSfzC<0@irQ$7Z32mHM=&~TKfFAV=hrAW^}<Zb0d*$iO$n?}B4s!Lu!2p@I(9JY
zIa7bzT{dw37PkDN_SPKZmJz^M$bTgDpyRzWZX?4?Jjo0Ew<-BfCF%saC6O|wY<9Y5
zk=1feWg_N1j5Coe$y*iA)%wa5vMEgnB-IgBOH*SrW#{5jxQUG7Qaq_KHog=4#d5NF
zaPzpBtL27AV#Sh1M~{k=u78$<Z%g<;M(PB{$}tkKj6Wtdvi)Vm>MBP_j;j9w)I+$!
z0S<GXIjxH^T`w4u+Xcwapz8z0EOf(0N-0$eGUNdD&LqGXO~C*Abx%&~DbzZcw671`
z!tqMrmv*JR%Ct<LTkkP2iUE$sd5&LxzW#9Y5oqem7%3`*1@@)onN{2umgK$+jBFYA
z0z6K&OHf3x)2C`x1k{T+n!67jI#hnwb8=RG?$Y^FxGq2b=tW&oZZM#^x8*~va2TgM
zfv+r+*@4<J`1v~5Zzr*I0><}n!@-dWx@&g84tB-tnjbT1;h>90f`@GE6C937QqfTm
zyZ$P+MFZb1=k<dR5Mf2|=}q&gE$b}FEq%$b2(Uqj4jBQ%eMo{SCM+Cquk+icSYfic
zWmn_qyHRdrKx#ruPz-M(bKJH#N*cDUw)vQtjKiCk<aUi1vthv+zCrYb8Le3T=&!^s
zwsJ-PxG)Q>J4u?%x${=eP_Wjmhzbj?;GbokWfU&J)=w(hy;(O63$$ZfFdV60#!nv?
z^bPyQpJw=4<N25>3$!>7-?u7vLCF&&TL#WpoQlRty-uV@;Y5+L9oRfFxA#I4FRHKA
zE9+v%TJp9awT@($FO&d@{aVRfCtSlqYCU^Px`g5y#9olD<gOl-*adPjHF1HDkdz|K
zDEvz}(7B|{TFD}<_=x@!{tr{ne>+J_%9dkHK#~sg{Iu>G&~HQj=Xvb1KDEeACqOG`
z864aEv7Jb=qYK-M7=MJe_^d>|Db^ZLRBm8B646;O;Fa{%b?(TT?>u*yav2k#dCt0+
z7Dnk&NA_eAQ!;FM8@9B`&C->w>yPkETFQ)Q*zoeLbYES7?mTMmOxHi!C_x+RE$RCA
zDBo^PJ&1r(IB0ZS$0t=P%_HwJ?*c2tD4K`>*f!Uf%Q$Olk-4JN{!<KIv+GjJb*s5s
z6^OU6fhdgP{9Hh{Ylz#(fx2evIn9SLrb2i&$-ZFZb`xq8R7<GPP$eTKwwuxyxVx+=
zEz_#x!_LGWXKW-bl8|bGORQ<R*fo!SqltDqO5gAAqSY7|X<ukp+xlK4Gl%*W=OCoY
z!ac91n3OV@BjcpsL3UlM9c&_{=Bq~7Ume6bh#E|9u8~{EtWa2o=|z8mKs8S;50+A{
z3nn0k*pqx9y%dInNi#M>?=Tn<rppZpDo|cFtKW{W+CR<bp*7Nn`4XUogC0hk9;A93
zq_+XSBSb|n$Tx3l=8CV~?&RB%O`6HWgr>v2tu7KtLQVRuo|@vrWM*Ni>G@|ht9L5d
zxJScXCdclOVmdNYvhSBQ&!R08z;9V3$)UnIS3q{1*gUtlZz$f!`j8C2Hnr!;Xi^(+
zp!jR!LJ4t|svWg541g$Qw^1U`sc3((i&NUgW_yyCsS{I%BjoNaNK3u~&d8@Zax6Bh
zupku9g_YQ3MTMASh;bG6g|(cdNF(q6fD&0n5cLdGAw$;CcB038nLznhwvMGQ1FqLJ
zXimE3)CIZJaJqSX^I+W$f5$^bGi@k;jSGQ1W5p+&<WY<}&Wr}>87v$dse#F?m|Ucj
zXmVk)Q~pH;7Vx%~yo#$kra*SLcG$I{HgBft?{5?UVA9q~eo2V|M!=il2tu}5Kr~jc
zSb{K*TRBvPZy*L57RL<nQ~n1vB7qVc%cB}9kvCzUkJSS$dNw(;c@Y{pW8zTQ9KnTF
z#EKu(Xb#edtfa<_m?+|nW9D(&j_G!d%xKx^M-;Avi?;l}gN*D1!HEb6G%^I`!qjOU
zC<2_3(Gx}!b{{3pIwh_YXcN_x%MOO&P#I{|(Th4$TE+~c35FqZKKl?kKuU%Zw1~E?
za~S@ZW7f6#>Px;h1de+I<BPmgl9$8eOMaPRN0Fro+P8Vc3WUD7tPp&EOCw~YM6#SL
zOOU;VA8J*N$3fNOarh7y`c5G>#4AA9vBVrulrN^ljRPx3=E96akR}X*`|g%57;`b%
zB4{iaORNkc9kDCVshnSI6@r|N6QM7tp;XT~s9R|MqhdMwiCmU0KTy7Gk%%_mc<BK=
z^}NRAc6rN{$i@BxN)yfu_@#X<O5C&q<>_S~@OUzQs{9hM|1RU~S8G>(3l2+Dji{>m
z|KTQYx69l*eUEOs)@I}e@5hZfA&x}dzQ$9EB29Q$xKgdIV5e2L1<_gtxwhoM|A;os
z$Ccc#<f9}l%j54-fuwce;NtR2(qJku6Z^xj5wGTBx>3@1Ru-$X4MMVneRkz)NmhY3
zwnw}na}(PH_EChw>r~vJ{M{I4<;$S&VRjn)<DShhtMidq!Kt;RePw%UoeXQmFf}}1
zt|7m8XL2azrt#P~`{F<WcgOnCnj$1enq3F?$TfRunEP0MVkX`)R@@0BwpO`It9fhH
zOcA|o)G8}UMv6gRt!`-t46nc>dMmp#KN0p>>?NkY6^|$?n7SSn*nEb?#(cFh?Gq>w
zIZXVB3ZuA2Q6Z1D3!$iA)5O_M&*-Ueu)eZLh$B-$w!9{7-nI1&@o$~-)WsWN>f&|N
z(}i8`eekVe_r>Vt5|C}0)E;4#xavRBtd8*pXHQZ}9#!-lBiTw=$w8j&-8{n^p0_cG
z=UG5lb8Yk73q+MO#(Oi)6$VQm35A(sAc&dI?E+Dw1(WMLkTWXhZ&cnb*Bc1*-@<~)
z{(_IFC;kN6fDDfDgy31o%qD`;CkUQmSJ_zt;_0n&&6lbh!Es4~V?@qKYux+Y+oYhT
zX8u-I5{Lz<uFc{7W^x_p5+Hd2v2X_W@V@BU`rdwaI5DgB*Y{egw##}dRrVjazAE=k
zdTVqlE0eVx^@<!}7~S`2C+H4%#^t!9ZywnEl~Fx?alTcPZywn4&F(!>Yv01Rt3ECq
z?6j?8u%)edERwiMX|^&r#{a3CKjpr~vB>UfjRm<4W1%~4EZm*jjfLv9#=>fDI~J;C
zD=tv(o38`a>Qdcp#$robt+A+O8r_hfs5%Hw&^IgV^Fs3@F*ay9-97-am=htYxNR7)
z)K8XYPL?m9#BQ=(H%4AonUJq57QxRwf<tWd6flUGZ|0Kmp0h8W#1*}Ledmkq_S%#O
zHgw$`wUBDhFS>vL#PszrP$d|z{D-}2*o`k49>31vLs%@b`Ybtv0iP0wvX;q(xp&oO
z<*uE%K<d9PU^lKBCbIfY<qWuqdXB`<&T8^gs<}=QrdDUel*jLMooH$^evIk-?^m7w
zUUj<ljtE*}h8(?8h5irdeoqVAyKgFWR6n;FF!mX_t@d0yzbMf72>nZBd(&eI83Wqu
zO_fuJL4Jo*a^#gEaVEJZc{+8H^yD^WC(@ZzbN_ZA)Vku{G$Z35{Yc6pJPG7X)PI~0
zXv?X<X>g%ZIr#?_W-LIeY}S)qU{7!+kO<lw%3g2`87js(gvc@@T^ZzHWe};WWXw-+
z&H96pUFZVWMOlgRxBE)k{|w~{n(|TbE|_zzboIw=N9X+SgVEwD?h__9@w-bk-mH*b
z3nl;aLB{``Mo&;`^ujm%lI8l~GD3RQFGHZUo~ys-@~!7pMhhi^X4p3ZKT#LPUr9Jq
zh5YnAo*EQ)f?-haT3=&`a=wgS_Or(bmW5I-6ZFleY3(>FM_RBxZS$<cka*eGF)#uP
z^vt1dBt!=;f4qSpWAtY*7+4=+*AI%$MK)OA?(42hhESB`DESN}qt3V~7+{CXr4WVC
z6lS3~ToTX`f|5vwID%EAvkB8gxmM6k)5TWciK>f`^$OyED`w@o`*Q=z@b7Dl|0D@P
zB1~gtQr}qM>wCH{UecdcVzL<1!<hEPv@~K~oW#c%F594H%yUyuyqmk>h|D5aiTlB(
zhgc;P!~869w?{Bz=wPrcHYnx{FY%mWb5qqb8O8We+kD6Wf*RZv8SQikEh-LfU}xvV
z5Hn5=tr(TkY@k((J?G-Mkh<f4TCc+=Payy*PVn4js|lJq<9}Em5|31vnNeeD`MhSX
z<MO&yuBt&yvb&s`DJ@xrq);-N9VIXIk&{p5Qe!DO4h$ow8n#?XmO#`UOlR2h%B!xP
zaON?}HFs{e(~YEyC08Qn^lLQI`^n(qcme)9aSQ8hsGXw?{IR%Bb1j^24zB2vEbK4p
zaLoIRi&O}pB%V#F^)7MlruQfkU{9eVtoAz8jd{1_C*IHtjZ|xK7Z}}9w4Js79Oe9c
z{b3BF#h%36=p1a0{~x&J%)KN)3=1uhS2!r>DdCdBdNacf|AvXA!Xk4SS@fHLM9K%%
zYl#xCsGU;FWq(F$tZRDhVZF(cW6Ejm-K&}F<to99l(5x8+Ju4ZRwdR}$=Uy_!|#7V
zAM|nh{cR4v|4-Gx#<BPRnaX}ziO_}rzbI!)*q<TSIlBJO>ee=c8t&OTP!$E8F;l9?
z;?ZWkcZ~MQ&1?J{HL|^^N}?|ov<Vd24wza|CLoGrK#m~Tv<?LKE~;l>XV$^ix!<Y?
zi~Atds-5r@zx6T+z+P;qd3k_R%<x_8B|={iBH$bgwN4)oCwuWtIc9d!W|3%6D5cJC
z>Z=UKo>5=*V!0y$A$D&bz3qHGB-z&QJNT`C4+*jn<ScX)$gr*Jk+3(mghn3((6-PC
zXj_rkQQqahoewR_oW*PqyZwODlCV2^Rg;Pv7u}0C3`q=Tyac#Zj5?rC+MXB?4~~O9
zyf<!58FOnwm(AyCMT@T6Vjl}pXEM@0jZ2Lhj!HBd(9A{K7)$a$ho=lNUDXrv6?U|u
z+2omR=~q-wHhC?jT6^@{)&{>qCESTNOApPA?_HXUK`JDY2U*c#b7eR%h&y$2$L4v7
z3@+l(1)2|@cW`@_3?gV8k)dl!wUd^|EW5rZNaH4N+CxdYfafQ@6b4apdM9MWd~{Fa
ziIW%(V#q(BBpSeR|9wHG{yy{!nVkr%C512r5pu!<yv^UIWWN$dKtXCYA@z7XaraZK
zoYo+jZLEz-Mp!j0MhI1aR;I-Mzf+qVS8mu%Jjqrd@b_!foP@uaN+BLlV59<}62YFS
zt?Rkt$qhg^%WEFz*L?H%W>&Q9UR#kz0*uYA6w7~Cx?5s<GZB?ji2sTbAua!7BrW*%
zFI8Y)*DVxMxsF87-(uoACT~LL8qx}+;woO%w+SR9z$RM-@V9#`hA0~ZTaz?n%qPsd
z92-nf4%3RLlusSjJp@l?=4(z=LHbsT*cQYRkpc=eG*d4pEZF>f12L7N%&m8MSlHW8
z<fq=ubj3X36H$C8;)xTg^{I=*X!4gVY^$O^S?nPf@f%-e<btQBWju0Hp_X)>he!CP
zxT)#+M*S4}3r?z8YR5{l$nMtY%v9|d0#A<QiM*X?2R-9_!^NrhI3=d$DSnehj*7G_
z?S@*ME_PVoG`y6v!0&jr#raN^&Gkf#@_Qy;)c2x`Ugi9e<H%<90YraN04Y8rs!0qx
zDRJ6ub#T27Xw7Nj^#d22WngM|hIuUJ89PF|>-t>l`dq8(BPcj*5gKqa<97pY2G{-t
zP5C2A29z|Ed_PG`aQDk9_<bd|V)IuqJyR@}&rn3vDEWN1gRmeLUS$0-xtlj<(G^#w
zSG-lN-y)AE3dQ6eAQ_7DU7I0UyLt@FdU-+{lE)MEkFu=TdKt5+MVN0f7NilE2Nu}S
z(146^KmlA!>5;=r+=Q$CLtX~(mNi$vjQNDiN4%?s=>)5t$kuY;=elgy1+GCp>seYx
zDMCwKHP-&kM5L?B!P;V5{Azl=(DD-pX%R^q@9hG5z1~k@gDOkg$b%c!Rj!PXx*#-a
z5;3#Uvrf!v?U|t4MVVfZHD-U@MjvgII@+LZi}zey=<g_n;dHeir^VJvJRJ26R(il2
z#Cu)HlQSd`yY&oX#snD-`x^CBe&#(<FZKM8>fEQcS~Z5%kGMwOd`d_Dl0DI1W`dn3
z`W8)W%ip=f<t++qu?3eq(~g{w#e*yu3u9#N>9_Ga-{fT8?VQZ#rj|HOK(rSx5HZ~8
zsBWyrHty)d9q#DAs1D?nh%WNYPG9n)wokfW&|STvLsM=$)oABMZV>(vd#25<>QPK$
z+o1oe+Rz4dA)16giIoY*vKVo)=G2Qzg*m7@xAl@L$x9U2fv21igFR|Tcoh=K8a}2r
zMZetCuDzQp=c3*VH`8l)Rk#)J6Fg7$Aa6o}3C}r3U@OU)a94-Z#9e)h>aI1Dyyp^a
z&KNu((~>4VPQLzil0a@-klj!pP=k{5fNC(VHmXsYVKzTjdME9ulmfwjWCD(~9a@F(
zNzf|YLDruLx*?UjZBtl%UlJr_?$qlxP|;HxwYC&Mq8W4}ixGXJEuQgT!@wI$3XmFO
zgt_=<sEO%((1T0|sm)q1X|H5Ely@OHm8?H+A)xx#29U{Q{VyWv86Ze&NfOl{?|(VE
z?y+mTr*K<FI~8Q7@e7)VH}m<B^6hEWLXLp?H;lYh^ljPMiksjWTMF~D7!{*pt2+M@
zTZI3Ho4&Cy#+dzUy)Y7wwCPlPza~j0yC&C&S!P*y4z$?m_K_b1CvozRvtggTQmQ><
z8(42kycA_CE;0ai2q=(TWCb}ygBBfmXBZ|wkIsCCkVIcnSO2<_-%zrV_A|$rzoiv=
zlAGeRwF-M#g)+947d@+#6)K)k7JH+{9dfmeoO4qcu&=5?{mk^Zzc}_fuSZ*=hnXnR
zPy8n1cwj6ETMQ6|ELpIT)O3pIg<nAO)h+~oIsak<D$DFW`cXCw@WkNSP;F08fTV7h
z4A+WWrQLnbQPP|GICm(m$Y5(DK@mHxjQ@QNPFIppku!^mCTJPKE#9J?NYFy0v0~6e
zq_MpD9N<ynW#dR=qx9uBS)B3~7z87fEZW$JB9H__3q36Ru2+*5$x6b+!fk2}Ta9g7
ziI#P7f{<_9EGl2lZT_z-`FBdbq~te9n*GnYvxLeXssn=BpKj4J0#4?RZUC98Do{SE
zK-{y)ETOl9f%`HQ!ol-ZKp-XwNIIW)PHGkst*r#`|5WwzdM7eD|0`Vi|6a-alo&|<
zr@Rsg=Mfcb_eH;HH}oCxXNBF3>Jnt$6#Ih1_$-iGrS*>Zlg(?h9bApL0H(c&7XNRl
z*{`Ts!2t*8-%xJTUicbiKdZfPf=nVa%9?ksS_;FXU15+LNL^r;oM2BJ27>R|7Dh)=
zu0$E?XE(pv1vCyzgfJuMuN!b?MFWV}2Jm#Xfbh}Z!>=T2k@|9N@D(@;!Z@(z<wVHo
zQz@xcGeoGyYgm3M1>b@pv11S$TKu*l#Ou9lLzu}8zk&!6HNxi`xmS`HV5j`=`L4uw
zu`W@HXI?J6l9Xrc$C9<<b|0qJ|Ey7ligg~{trgYAD;9D5V@U+*YVQ{iNDc%D`Xx&D
zf2+~eC^otgS`1i}xIAZU5%IiM>MUIwX_R!1Nu$SnbbEs?IHdu0BMOS^y}^jZN6)N{
z64<FX7{m}?h$YJvC!m13bZUouXcQEGU}4ZJUO|LkpJ)sjH8~vYfHE9;CAlCqt&Chp
zUr7dIjr97CR}hSP-LIzUdtzZM7@}Nl?F1##pI{V<PD(-I6unvm*$u7jjMWmq>d+<c
z`3LSji2n=04%9QdI5!Lf0z80-WETus+@T?Q-0!@qdE)oD8u0q$RZXw|BpAkTXs4+X
zq%JK2+eXboRf+{*31_EqN?THK&`0r`9tsRHaBf(q;AKX8RzH%5<@puXuU{ZBf8vL*
z@=EY|EDy;l>PKC^HF>o$#RF**{_d_OEJ#Kn7YL!&dWHUKz1k~EWCGTxTLs3dx2PR$
zWWa`TI<}#XQDBjg{!bov^yp)oxfDh6z|OwTSzZzLL;2W2oW@?ulua!u02^ea$v~vj
zz;k(O1qA<NdXvN&<n8@N&9ysm2e<xM8(PDKN%fn)?QX=QqL#!<*{H~r!eXHJ%LSx0
zIxY@6!VEV;Q8z!VZf+T-ze5ZESKPzIpTCFgc<7_gw8ymkL=}S|1J(@9XL-q%K#5`Q
z@{MU7*t*%*{SvYy5iM;P?Ee;h_uo(wL)-{saL|nK$M|k{29<Hgrgahs^?zH5G0Sc7
zmDw~&lH&hIC7)Narl!B5+>(<2q{IY+v6#!u2Dg<Heod9Fzu!xaju>{ag>YK89wKoj
ziPJKi6=DfXDB&xzQdAM)o)u%$P{TGKY70xTlrM#O6KPZ~xo<tM*1xLcKPwT*9TpU)
zeF|#LV!a@ltouKw7Z$IdmEKyX=2C~3OZU}U9_^K@ofD31yElGQi{u6FC$br^j;yTF
zN!zg{gJ2@zCz<jTQ^%p-iX3brSE?dXJDK&!IC2zfuC2?GZfh961L`h;UMqnl;mEhN
zIc9p*2Yy7%X&N8rxBe4+WY80#G%?DE!?3|IL<K>IXB0^4oFevtq<RfS^@E~hYtoMd
zb-(~_L6YkQ4&4G23&p$_F&TN;WWxzuZ(yU{=mN#G0+C{`Z78Or6rKTXxzHWt9DnOS
zpu>_lSc8zT&O(TTxZ*tA&*ydj`o9kbu^|69YeV#~g!M)e@Y_qi2W}*@rh}t;0n!{4
zxzM-J?-elc|2=D0VuNAO4q9r$b|fJy#rh|M5zvmgmg*o&)6b2-w<L8)<^l)=kw=y<
zx1Q4OBqEX_P|z@FWfY_Vfr=9>2WSQThF1FNO;3zgu5rboun}7M5E|(nL4T~F{)7;T
zp_Oj`r$A~vE2MJur$8yvcz;&>laZYL0P&s;$mMq7RXaHMjzFvae`L<L01u!|3_CV~
zh%vWqKFoS?Frf`13~X!%7$#KNw()HN!*;rZO)w#%INLDcchv0H>Ayh-|34{b+u6v@
z|Ex;`2fw4-?<#p$-0O`y-v2$_{W;zJeak8NFG>sxwNcUUQe|I;t#Xk}BBwp<jvDAs
zWpHHF!A07^>|6epl4W+QbkzOv&UO2{y=OBCMXK1&WN$-2Sqs51)2)~CBQ&nOe(N74
zv29ZDcOjvZsDO>D-`UKX(R^&Ag+>ax5f=&nv50OonPCUrsQa?af1eQ+9aQBex{<97
z!GQ!1$_K+=<LkPRe|NDEHwB2w(=$V`|6iqSCYeb`+1#}mYC6jQTP9`v-)6CVi=@oE
zwSFQ}W{ZUH_V4|xq+Q+{cPX*e{bNg)VqygnW78fy&y+igBi?>!y@c>wq~SIJVdRCj
zQ!C`8?Q27#-@Hu+OiY`5SWR@&jl#a>@9|E3lxVv|Mz}M^tnAqoI21y1bo*J4CGJI@
z&GGvbzjov$Jkmo_c**@J+GXO_QnNBq&LMmgV#M5I7B!-m!i+I1lTNsi0y6VG)J$Xe
zicCaY+w1AhyE0rgX)>xq(K-Y$OZyv#1ohb-?sSaiA8^?US{K(vSgQD;MWq%W%?0AY
zqTCAH(e-Ws&D%&6&n7U(SvFQpx{Vo)Vn1~xRT3oQsI@c_rPNlPMaM&sIrgK`)31>i
zV&I*cn>Y(-?iop^jNJ!cjPFlW0+gzsvwPxW2_TGtoJ{31S3iOC3r5J4UJ7JNju?`1
zM4qi0HKI}>=}(y{9KuNCK7xe#Sc7LRJ(qyT#jyt07&ICUz2%p#zr9hy=#5f~u5}Q%
zhh9NA0v_UUqX_7&_o<Z7ikx*-8&#dn_vH``EW~kHhF>hrzb8B*(h9yoR5Sv=WpBCJ
zd#8?Rae>jua8o#m;7Ybf1lnC(95VsCsWucPwL?b!oGzbJgZpI1v%!6EgDIE`0&66?
z32~TF>7tFo)`sJF9z|#{!#4jPsa0_~4jpQHxz%C&>QA_FRVzZ}3T`fYE77nDVu*I0
z6j`I;z(MRS(a)aVyvM#tU+>bZWqb3+Io>d1E04@6ezhG9<-HBoE$D;!2HDk6+Cq8S
z@#AZQOr`~4vmp3VW+nBHMVERt_ui}y>slLG8(rJMLK;~h4gf*}rXR&```QI&8$pDV
zAxzmQf0=GAqSzYqvUybL=+~70L)<B%${J=N?g(~JJO5y!_7jwrpfGd(msnJK6Nkgu
zUyd-thCqvwIGU|HUN_eBC1}ic3oF(_&lQOk9=dV=_N{om->3~O&`hGyd!<Ml9RHwy
zE}#k_pZkIxI=tO~rqyQ-m3JkV*-irpM-=R>k1sDF=6I+vNH98tE?9TEV01gLzigJA
z9C8NDea!?Z9Fzuqi+e6&lONZC>>R+#5wj*Z`R>MOV+Y5t?vfoQ2gs&T#-;U?9kjX&
ztJmGA;&xGIkHjl$IxKZ>u#;AHdqdRR6O66zqi6dAj!+eyjDY34@de$nUJk~nKh`+F
zk@myP<z8}oy%EbD4E6?NZl1kSl=ypty?oUUzDmciWf1I)-AsORA6gBLiHELFp`(DP
z@pitN1!(VE+usq!PVe5wn>dMyHg?fQk&*cU`gAzhPx*o10Oh+WFR1Ot5!yW3ICeAj
zTDso4I_|D`F9*hcvb^QJNpCOfblW$##wqT>KHk*h?Y9HvK1?sh#goB%9OS)Pt7XPL
z?cX09L_9a_9RQFWd|h`5(8RkR-286zHV3`KujYCCp|!&{(ua80k>Chlb%Zw^{lUzq
zQV({<4()m^<9`kzx!Xtjt5vSQtnxI%t+k_T$JUNBM<?jpvEVrVKRQ;I*<Ks>js@fH
zn7-hccbxTK435Duk^4w0I3AT8Z<WlZ`P!&HPmRyAbJxcCmJ@snF*!TtPl3y|PA!}w
z&)i|(v~TV3=DBS3B_7d}dIw*7ipU@PgK-Xw)*DX$CBESdZ|Ht4D>Z_5mUfR?&+jmf
z@+9AIO21farMx@QpQYALtliPs>xIL?9SVWPN<JB!q^Eb&UY_>OH%_mgX&_==KYI(}
zgyS$%(W=*O%=+izK#KQx_Yz1kFL7}cNb$Z2#Y+90!0!K~<R2>ev=V+3{@tWuuboAM
zPk{@$@BfSnOEOJr`E!EjKC5SB0N>^`-`GhLV74dqrvnp)-IrIFrZ}>}Y^g^!m=L#G
zG>OzVi|F`mdu{vo=o1RhiOxZ0i34X$K>7bw-|$DepT@tn9iy<hyA?08{LaX0+~e3q
zQ)YM|)NNYuvk>}DS~cv*BMipRwp_V!;$vX=A+AXe4kC5A-_}k|d4zwTsqlv}+s|!3
zPb<x%t<Jjh49yySes&%cpjp3KEzhkI6c@)xi1OQYB;2!+NN7ERUX=J!;jWE$JEs><
zH=_vam6_Q&SH03qpFTGmeE@wF&OY7jkB8{V+6n>88e6+#L!#kz+_@NqcJtU0=_&cZ
z=|*!PZtorFf^>`HsiUq)I}}xJj>Ppk<fnDU&%_R*a+<f3th%TGVPB`f?yorLXBPyF
zV*wLe2_zX5xwx0xv7h62x(JG*sy9Ufb&jU?!EK4F5sjh}ZPU6?P?8^ng{P`F?6hZJ
zG?w3|vK&V>`*YKd@HB$4rh6tf5!L9_gnvW1wpWA6dH;*b{z%F9Dv{;4e_e?<0Sj^9
z|FJGVpu{L#Q@e{j^Cg4yMOpg)P`N);uFWzwRQ4v7{RFx3Qytz55H(l{C=n;*wk#pO
z;Cr-hL2=R6Ly7K|&cF%*Ut4;53kZyyRz;iUz8(&_n)9=>t91pMR`6iIz&5<?9&)g(
zhnPf)7cEl*^qt0{ewxmO1>I2JTz~<iuuXaTQ|hKs-3F5KO=oh~HhTyuveOp`>l61!
z9^wX>!k%X52ZddnfavyzMTNj4YbNY&?HD$%GhFM5YiUC}`FGn_%XoK*!)4<iYk2BW
z<aPyn75NWb|9%a-F$QnpD?Tp9;O8ky<YZ<lvlyp%m5G}eN9T;;&hb3WT`6WylJ*<J
zAO>9lTonPVKUqd_=EAy4<gI_O@Z42yj5KomfW;8gA))waS~b+FLd3c<;%ri~i{}m^
zIzK1og!>0g9N@y5L|4{ZMNQ@WE$e#}Ja&s=UVC@O0vR@Q+dJeZxd4mf&w>9WqJa%@
zWFLahFq?18;-b^S1Naxp3a*}1!(4UwFAy5Tg}zC(qH7T;%fcl>OUcSxk!O))`#)}h
zaQd(0>Oo@wmX;8S3j`;`mp`jBba-k=!a8Vi&Ice~f5Gc|UG~nf!cc~GEP@>fQJiQV
z;m}S6tcgrU0m2rz#xOD_nl2vgf#!iYNTrKX(^;S4!JLYv0dWfjLDDQgS9K@+uCA0X
znilZlv*p$i?7p$S94X>M3`j%Bf%b&_cj~Bg*cI19;UdR;1vBab-*M7yZRQ4N+CHTB
zm@vs;-VZbU9T9WhX_rb21THFP07@bq2hAQ<1CJ<~QrWbU2_-{H*zxS$Rxeu7;_c%e
z)iaMNIZUlhA`}ZPTN0QzP8n&FJdUKnk+fF}TqSVIX>Ecci*#!YEf6qKm^Q}vjjfo~
zu~@RQz=d;1a?4IKlKR;Oiw~D*vtWk&#{v5mXP(Hxh)71c2229xT3USAFtePnis1@J
zYFW$-tG(`9O6_`g!plJrbKhhe@&j9hY$K52y&9&ElMq)Y=X}&#M8P95d4rJeE2?MM
z=!9~=t;BGZVG>zW*pW=m{zMU|uATD?qe)GC9ZY^iPZ-G5kah?q6oc@vhFW+{LJ0}r
z5kG7r;B)wGysH&abz;JQJ2QCULL3WuNfTOAvaDo<gm>@ak0K%KJKe|TE?k)4-*}2=
z^_#Q+$=M4OW9h@63m2?u|3C03MZK4(9=$Fq<8orW`@+Pf3A?4|T&QJAR9A-dy)M@B
z7ii>?|EHwx*-Mu$T)JfKL%pA{2+aO@-F+)bn6D7gat7zWr`6qQl^2z>fxn>3MI{R=
zdtSL0lo)ERDreFr3tjP=E?-yjca-c<Vz~XED<{7mcb>waa%Ob*%gX&LCA~_XQ1WHH
z=xa(0w@d2lFt(8dhWUj*Y|r?bP5*7WjHelJVjI~o&6C_JQO6V?2%xeFC)$Y9-ZO=4
zDmglkNcQFn`NBxCROl}Bkn1h<7w^tdum3J^On0%^pDAPu-NgZ7EB6+UP`*!hiK?*2
z{UvAht?W#6udlzCr^n(NA1sXd>eCEg75{(<?We94-ipcGHrXIHbH19X;yLg-cG%eH
zWGAy03s_++3cj2Y;+QTXpwyiHMn~)2#0l#bonmvQBxojj3!%wm9%n&nvtS4|xNYn-
zc;Vhg{==9x&~#=Z9IhMeln%QRvm8W=i#%2i@)pwx`iIChi_W@0EXgJgg>GI{U0Y($
z@cl&8+thMJ+>DDtM*x!+CFP>eeL&9&%p&I2#tsKm!=k!ctXPGgQ=Qh-+gh;s*9*^R
zw`)l&dXStt5$5lVvC{KC{w!5L4i_qauzl1uUMh}z(M3rNK(*l+FHB?5a3i_0LmYI3
z1M;`pGj>OQL2pwehad^wOz((1JrVq?HX*B`+LFH@J9I|#3z_<?z%XKB-7lE7T!%OS
zNSURA9m|lb{f<4O!`VK!-5WYvuxEK&{H$b+NF7behG^RK<-gL%LJ4fs*X>*4sRA1E
zTy56c>aw;Ta%~~Y+vi$qe}(nI`X4_j=>;+zd(z`;IVs583tHbP>1Hl=8_+UGYkfeI
z@^KPoIl~P2T7BgI2bD`P-MRjlL@O_IbJJS1wQAxD9@c%BPFrAS1WQg(VBvTC)4DX=
z{<<o;BX}K3`4v@)7l|?4{y$NdM45-FnQeX3?$t89pC`2pMRF>>fEJ^y!$oQ$7w50M
z)|At)S#0oEwud=v4!11Y$c2sAXCQ$5Cdlq~n^8LkalZkW6bw17{t0$KapWCTQD<8}
zJyE89iA)Pf37&ZA*PVSC%8i$c0IQsrCwzIPnpJE9Z_of4E*}=iJs*`0dHvqN%L=1j
zFi<9i@NgKZ*xt2bB8K>DiGcGg+foMry%dQP<Obg73_}+=(sfZz=0;)*@M)v!uej8@
z6axxqz$RkBTYPZ{Z67N^*$}!wX#ht7LHWJ_3Jv&yP4gWv>sigN7DG7F8Ku%XWU&kH
z)@r$jB+`|H{q4>=(#(&mv_mIv53t^?nQ#D0UIPZOq6ib4hb9JFF{#&r{*UY9wcJo!
zwE|p%2l&80e_T)t85#Dqzv!C<1@#*mFQc)4nl7}Fv*Z8nWt^tkA1NBlB7#Pa$6G#f
z+abRY180dOWR|2kum1u`kOd66IhQ(znMb9QK}5>Q(*(V1-C-R_;bZsHrt+~9z1O;M
zq7bEv*cj1p#=>xPgD|BB#4#f1#zIxDxLMA~E1-zLHwm90Jd${W0$iZvF@vPn&>ZH>
zYP|#!$n%SF(d)QBVA}N4)aoMqv3#2Pu#D)HPS<QwiIzWGkb7v*YdYQnL|e*9TR+XH
zZZCe_&d;C4H6(bHx%7_n?*Mad1X#ohHI!^rngx=u^{Fv*;u?d5BbvvrsO<jM0(T0+
zpHWx(JH#trP^HhSx_apjRMRrKtpp>?UPop85^*Ux1F~Z)hg}%ynM$%Q`GgkTZ6iHg
z)*{k(N#DE+%Qr;~5c6-`QmQ|h!CH=kwO}O`{bI|)l6uh`5<3XzH~u-zeN}DR;R(Gi
zm$-w!+LNYJJi#stdOCE9xEW~5*ZQSXM8nA9u_c6xUoeuzSCHv;Tc?N)3eAI)yT@9^
z*0s|r4k7oKR&kiW(kjZdILF@|{Dm`L$2QI#(}9L@OpKRUFjUf5r71c#HPIVl&qlw{
z-F+DKGZFx=8$IL~m0R!E*~r+hjnG~RJ>>UO%5e#%MI1#kpx#JlsK0%&CUPJu;f~jj
zCT0L!-EQNOwJ~UIys&*6-q1H}VGs$#h-n)aT0LtFM<pL-qwy|S(gAOn?~_jLg+>=6
zzB3q%wT-*bHp+|~4W(%tcL&lBj>X!>pSSBB-bkcv{B&byW7j&hOxrk$ws8mF#5LMQ
z^onR3_p!-$Q)ho5ZInvO(lrKqyfJE`d0an0&(I&<YH1r$_*4IoX(x9!4hK8Eoq&#g
z<o0>HEO#W>7wmNN@9l=6-p4hXKE6t(;y;PW{a$oFPTP1CQ>_$Me9|`VGi~GkwQ{Gf
zhj&L>McUX;8wG14v3@Kl)8m7IG<jvpyS)RZ1N>F`dz>~)(3yv9%^tq~d7b$8;!k`m
ziRu4AXtT6G+CangiY&{!q>J=(jT7z~r<f7ds23Y|fJP4i+77>xWDTwG>>+#heEjUL
zg_EYqyjOMI^HJ@`;ySMVW@K)5*gNua0UdDsHq9j5ihD<?vyInMSAP{d$3E{UqlFNc
zx#<`!tRWA3$4sYpWbLTU(GkYu82ZHl?>M6|zOA0}DD6pCc+}cEWrlL6Y59zPP2+Zl
zY21#ljd#w%g0z$;y;IEQiQoipIL(MFX3!Y};ew0)0|KLDoUY9OGr`IXP|k0TDix@!
zJkkLT>wla)70V9w9M8j?0rrIP!0o1+6Z6ke!06ln`rAo8{*cO_09bA?VA}!?+v3U&
zWkO_&?=%8>J=o~8`jm~@J3pygTeWw7N?)8-@^L!xSF5A*GwQ^)>N%%eJEM-@M-mI0
z+B&m%m}$e9Xvy@<Cah)eA1ypKgtPZ=(>09Ak5SvqV91-m{tGHNti)K+zeDclg|E~+
z$5Y;O8=SQLBoN@!4!HbFQ0$pz!tc_I6iG}o_ZL@*`8~8Xe)e0?H-(vnWqcTaTT@}&
z$=_r^I2dkqY2`*om@AY1X{B${MYT0hPWtm#r+G42Q=o3XPu(zH@CTIJsB{{?qk|0E
zM$6>)Yp8)u95trgkdk2~QW5zhN=B8$`X7IXE=^x@MUQP#{k&OsA6D`fCDH`>W6Bv{
z`m5^UPF?O&Vk>^Pa(k5ANoC^b%~TpoEB;=U>{Id?J+xmrb8ILpcR-2O?}i0!Ygq!g
z4{9xIIEbESN{|wM2QAf2A$h{MSE+AGNvFO!NnV8hx1+eB6;ANe^G9h9J85XWenOcJ
z8h?~QDjunBxYt(9w9Zrb27pei^92OXu_f6h{}Fxnqa+g&n_A`)hLg`~h|X)QB{p@;
zwHTQWL8;NwCY<~Ry!JVsc|l`mY)f*uprhLQM{*`i^mTVegWCEL6RV&4+ezAdH+=R0
zBEjtIN#n1@bNA@DwwnpBfTN%I?c`*>N_Y|R;9j5CZ`^j#xb1Fm(oh%$#JMnE!vVQg
zjoIzyCyM3P&tp^d`Fa^K1L2QnXRAI$ba~c2Fhv~8`s&PO3ltJ7;>)U9U1?lCXIklY
zOP7`vL0KWfTMbBKBo2XQh;$OQTD4Ggiq||run&9uM^nvH#&DiiADo8OT^s3MnO<C;
zxl%uUmVZv85LK6Eb@L;gPn<pLs+%!_@Mr6^C}ausn7I_q54G=v3D?OG-T{4_9dd?k
z4EYp(cRJ3jOLP=;KW6SGW{^lhUh}bF0Rcp!f?o?M{wvTD7?Aq!C&!_BL|?}}+00Ds
z+UM+VECm&zHkE`8WyXldWG5{>LMNKzgluuaSy}^E*1=Hasq!5DbsPnExjJ)Y1vuMa
zaZBRrKy56A4SQP8*?{|RXWfNeic?LX=%qR*Q;7w`_t!#}m=hG<=$&N>5h$UMgNv8R
zwXBQPmj9F*9JB`GH>qX+Y2DQkF>NtO`@6Omi2Rne%NTD+bW12E<yh-q5jQ3rA|eLw
z!sfLc%SSHEXcG0KK8tpGoZq?-wL)*9Bo&*+0YzRK`{ImKuws&V6xo5y$FTfum2$}X
z&jg$<`{F-d%U#D<jod+5bR$c`!kA$ESucbAZ|)Xi7YIoF+24}5`WOrW!ruj^8Z$eI
z+hy>1zfR4&{s>deHl>}rkswI=pv9s4AD3D-=t2RkCyM;df6A$8wKxd>xmLLPacfyi
z=!*@^NKM|rJ7Ka2``}?92V=yu{OvNsIF7#vU?Z>I@7#mxVhgcl_KNt+rU>qf<5T;Z
z{tyU)UX%iuO3u>j3%YMA*GsNsxxP;k#|Xu--SPTiFOv8sXEEFCjt^<Fy3xH4a*yB+
zt$Vh#hbDXx0nGm2;*bBMBu$K^7OS%DA;|5G@(0%3`er2(%u=qRr<;910RlC&svK+X
zK4#I{wdhnxa0I~7{2bvq7t5TSa(w3URE_Yt^>IKEk%MHbp`b<_^EB?wdv-hoYt<_+
z=?rYOU`p&Nnq&ba_qNjtiSpy>VDGb>$wr6>&-JXCbEa0!9-_(P<;vqzwAJkAtd{Ea
z6#_a?Q=rAj|7P#gQKhFSdZ64KcvriYD?+?-LZk8blo+8Tn->4?D|tqVtOwi{)MoSl
zM_pdjwkx7(CD`cVJb{0WYWc^yW#TwdGKhYTsNkO{c|^&hN`6X-z2XbX{ZkSSysN$x
z#d`ePD*S0Bzo_Jwl!!X$AaMl*{!DxIZZc@pK+#fP3vHX`B|V&UqHZ*9c6SeGdl$LG
z{OVupsF<z$c)mvH^!CwrLe~yce^swCNsaKSel%!;+E4LYe<2OTzOD85n&QiDmegch
zcHh`Ak7)BNUK=@?_mjg?bti7^*<cjzU3XeG;jL>Yy&1)>)JbcGWaZ6!+BV_Skxh63
zS<FvzM>gR|HF%$_rn}&y|658+Xh?GYA7ag|meG;eUo62pY{H{EUJsI(UStvN79@!&
zd3`2{kwy46ZFv1ro3y=63xi0Euy|%WzrEEjHn}UA24wZimYV(!*w+mpXHw9MChtWS
zqkR2!Y7HU#Ffc^9<YI##Aub+q`t{@>Ps(E4Zww>D*@fkC%VK;tG8}nU;w}$ZyoSt0
zQkNabaL|>}^1rw1QEw=c;rvu%tg#bk`#mPZ8AgUP!f5ama-GBiR^aP<k>Tv2&c0wb
zBdtR^_e+9<_4g<>_hDOJw%%g39a)UyKS})qCiB_VI2i2mb|AUgPi{Y!;L05ekO;X|
zfE73;To0MGw`4fKh6G?YQiS6sML-6Glw@Qs>Fs%)zz&h^_+FFdl-CY)$`Kf|megj*
z+s7D5&h=~b?nrQez8?w>@fJyWdH^W0O=qsxb?US%x_OW6uqBak??Ub|;2oeJ7jEAV
z?75G62Te|b4YrNLG2S#DjPs^Lyyx&1Imr+d$&t2P<8y3CCt)TNg)x*hut_-n;<AGd
z+QS7S_5L4|CUmHS|9p~#zamzB{Mfg^V}Bcbbek@FN%Lk1=P$xr-`1e>PiP)96hlE4
zX!RTO%)hSQ3<$w}o7}SrV>YtTvY3Xp{We(V0<5!HLKqWG8)2T3SzMwvGt!%Iqu9cI
zXeR3)WPh}59`_NLw8b(@+3RrSR(5$TdGK4Xbwu5<!P-%*H{ol*y}Hw8l3O8lw_@40
z><KlKP7-yz(I(3xW00Ta?u4@}a#{F{O)T<BHSqJAzNZBfY(<Gg=-`;Y8MFL0n)tFP
zicy{`WHK3Z6-f^F_3w8Kb0X=h&IXp5`u1af*Be;ok}=CtRCC68)hVcZtPE93r-Itb
zpon$ed54O+x22+f(lNYkp7|GzRTWcKCA=~OqrX}cXS{+%9+5a(rst6j<Z%(5GM8il
zS<$%oW(8)3cDxRq$gS90qV7&Bs-B)Z4P{!tqGqE<MPN1v*IvFhKUKD6c}iw9E+Xc)
z&vBQ3G-U?$ghJze1YI>=-k;P!Lr>X{23dJivY#Z=QmYb4A&Makz7Y@1uc8`bTXbAX
z_m+A+(^jvC{hjX-LuzDe%Yf4O$@ge6^&2c8*`7~)N&8FM|3r2e4zYDC!{*6~nZk+N
z=#cdqI*nBtCr84?$vnnbdufHyCF`H~!i_p(XRZq$_DDld<RNLgufN6<IY^j~!bRyA
zT*x<Rkr#-aq@WaNkV_CI;=dHa;uE0Smd3e<KziZ>37p6M+yX)Kf)dw#9N|C+i-mqg
zL;`X5>cn;FsCAeYq4@d@ed~||p0sDXQJRq_B*X+HO_6XE-7@T@Ho9yG1vJxM@pYZk
zo`rxbz@7FvZS*V=vUO#8=<bQ;z45Wxwm{t(mG$^oy}AoQCWeUB++OszaT)psI_o$7
z^w0k6&zNJm0(IK_Ac@?kqWvTU6^pGhHmkBtE@8{YfR5Jbwo2`B-8ikpme;a!e@YT|
zoBi<0@*>=(JA}d0()(Uj_^8(3?N#5-JL-z6Z-aUKORQr|gs%8^DwCO>x%`s<5ZSFO
zTGGT1X+<CBtqG1YP>4Qv6u}<jt6X^NJ*c@I_awn*OA@kW&Bv<lSj{1UNW6m|;G&U`
zbvcWZ-5TJEj*B(CkphqDj3S+x$r)}0&%*%NP7{R6PC<ndi!nayP9WkN%xlXI@CB;*
z2K#R}HQ=8kbg8}LU7X#>w$pHHEQ$UB4N`&H3F5R#3f+?XZW)9Q{Fd_Htxl-JP6oV1
z!fWWuFqg66kk64#>o-GM`nnbMsYI6w*-To-rO0@9`noftj<$YokE*&21u3*f8&x^<
zh3<a~s*=2SBdY2NTiysJjeT-yYOX`Xd#6KJEi9wR&P&U+<(bvRsYrfXMz{gkSL{D0
zg-wQqC7j0U7))$H0$N6f1hkJLB!J(2kRqZ%MiC8=4Y8oi*i_DnEG3Wv@Q!7Z#-duY
zz|cx<7fUCd9cZ@+lAW0+r`qW|<nF>*c3_AIIp`48b~$KKwxHJLH^CKuQxn85a*9zX
zq~@W>^?aMr+74d~@Sv8Q)^A-)1IYXr#23O8DAr+c>(3enDHsNkfIt}J#|X%3_#?&^
z-@OgCkW3H@Bm@eEH=Gm50yafpkbbO(9i~8CU`Cuh+#jG7C?c<Gj_s_-%#?vwHkKju
zAWA18ioiE!@6SPM9PwD%L6}3Y2U~v**@10!Smym;nbHD>?9fK<1}xJDmcfhxEb|a}
zrGEpK={GFXKheDB&x2)Jx`w&iAQ{6R=3^S>r{`x@`MrQGlp&U*%w5>SLGW`r*jvXb
z*3CfyER$hI!+dmX;~e$$^T^zv{OIp~@#*Pj9*#;lLM4iw*@>4%z7@PAIh>=2+rDwK
z;d8WcjEMXfRi>R4W~Ml)P$`GBwDHhD2OsqXwHcdWeG7QVf1g@6MAQ}SHD1xm@}Aox
zrq5{mn3yj{O!8Kdt*S(RVq^-^%JExBALQ@<UE8_rHjzYOI2RL~1hqpZLJT)G2q`fV
zo*@E;1%eGi0zp=+&`3iPC)nWFR@okb6~vwmEO>!oH7@|qz=k!F*|R`G;t}}%Q(f)0
zY!YRL#7encUEOYXSJi+1I=2&9kO~>#M-8@NpgPF!(nz&grpZPSH9bht+nHYkQ;Zar
z>lgvKHo3l$sVsI1{Ti!z`gQIRG581T=-0&NFOwR2+oHVED|x3+=PiXHOL2+2legJ;
zp*#Fi#bt_+-_upNjr~r-9{#>v?;Wsv8th07HS7_W)8|~()*}iUjetA1rTvdp)YNDs
z`hC!o5Ymhx94zoU#=TA>yN$ZT85A6Mt2(hML4O_qKH!{hbXi)4%}?egq65HR>{LG*
z{_fFmfT?*_mN2!%q!{(4k{FjGgT`P|ocFa>H(NerFgR^$&KVG<&^t9?XEDTiyMG$I
z#r&nDOK^zY;d!D}l8z6+iI>E*<muKxmjHF6L94N~5x1HkB-#bD*ojv=50#L>$6!oQ
z(|p@j)&tP1FQ<qpzMS-BLUnJCgP2nuF+FjT=H@R_vKTj5Y14O2(_$&jTi4oCX2E<=
z@ou16QZAzL>7#1*)GZdxxw>lClo9rm%(B|EP)yoWz*T~-R;9DKS=%6JA=(e01n>$E
z%<CFVqF&*8o$0(~D;oN&p>0zw|4^Y_eX(u0&q|{Dq|RaM^|)mE$cA1~&|Td5X{AuS
z|Mt$_;#IP@kB<LFTem5KXu7+N^XD)I#+VEf8|RC={J_1~I^WqxxkHKdo)ECaWRF+i
zJJ!qSK_gTbk&=?JCHK@B^qKz>DVwb}VH;boO>$8i(^#tB0xBx@kr?iK_5_9-ZAX@#
zC}PI;XDe{FFUS};rQ+W5g_#{W;~&{oM@T=jgqq<KJhqI4F;Cb<Nf>!a7+ee5NaTt$
zfzc9B64%vd1vHw&9QVeLIqsFcNFN>ZRA6+S-)!$BfQ+I<-Lq@)dc4%U5Lb%4Et3w^
z$9GZ{AJ>a<g}RB}MDqgw%R$r_GUU$69t(SuJ&b+X(A}mhx!3CAeR;%2zYS-TH;e`S
zfAZ^%mQu-M2N~n5hL3=@Pk!-(VIfmehE$5)FV@Q&%AKGW*55{wp~6XO_{92WXyz{s
zOBw@|SVMnM(h!fy(%}pNPLgHWY(MI0BXinho08rzi=;GV2ZCDwH9S}0rx7NrxXT=p
z0r;G2Fdxf3$S{6(<QlD!-2Y<BROj3u^^_v+6+n*-Q#51tOsAdp702%tuR9~s>OH+Y
z`ttQksb2TqyEkcKNpyOLn@?>QI%|@f*BMcC`P%)XzP-7am#~fvuGP+Qq%u`j>m<tc
zVI%bHd@e#3L~GSGC&+bN99ry6f|axzjathIf8ARZVsG~k+ZX&@l_@}})2`BDT1PAn
zU61VrPB&WE6e(fu8d){V<|^BpMih9sD)Q;%9)R;hdCS^6$z@(euRCOMGrawx_@kRw
z@e2jMH5`F~L8r#~v0HKrs$I<1rr#>7h5j|@lTS&cCGy^p<PfYJN~@iABgxg|!)C~Y
zrgwVqq=`jU_-@?;r$9XJn2O^%;@8YMQxmL`g{M%A!6bJ|iIr=>%bnD5?=+nlcK)Uo
zg*)M~-nHN8epe$o!kVL*S-sghzO9g$oM`<gV_<MrM=a`O`K$h7ujyG@eB>N%+<|#W
xouX}V5MC2Yl&X_I-JwZ^n8;7gOyzS6&&Tw7Fz4TWxio){&$Ij&=9cF!eFwHp0jdB1

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_ccsr_stage1.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_ccsr_stage1.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8b0214479c1661c5af59c97fb834564e717b7f57
GIT binary patch
literal 54491
zcmb@v34C1FecwC#hQVMZ2=2q3khp7?wOETO%aUkYqAYtXc?{qVK!Csiyfff}3=%n%
zV!4fMC+U(p)(hyQNtwp4b)5ROS(^5Jnve5do3wo|&2nFwI!og!ZQA6yNn_R)-}m=F
z_s(DdQg)ID5(np=?VfwK|NcJ<y}gAPKfm?)zgc_WgYnq!>0#|pnumAt+xTbkSj>x6
zVqV-!%*88l`<s|gR+4d-ma3#JEmO(zH#wJ^&sXx6mzpcg_f&f3i<P21r{{X-OO?`m
zU!~8UGjsj(1C@cWox#c=^|NzB^Fx&(OUum-&u^-1vUq-OWPWpHv&9Q@qw`xTTP)r)
zw{?D7W!wDr%J%skl^ydtD?8_RRd&snE9Lp!mEH4uDtoM4ac=MYzREs}_s;E~KTtU^
zf3R}Uo=bCw=Eo{y^S4!Qv**6K+vg8g4$mK{9GO2_IcjPBbI0b7SB_hJVD7~H$;wIM
zgWixg>}~pNvT}!aySLdJeI;JG)7#>0eI-`8>xr1R&D;J;%-dee&fZ-+c_mf3hqN8u
zPSSSP?zxhz+{^PWugr5<Wk|V?l-=GQQufsD<@tV|_j>zy-e=G6;Ca7yfae3XcaZ-8
z&j-CjJRh>>2YDXzZsYkjdw$4EJQo`~Jif>@7B0@rdFA<;OU)&}R$iDYoNYE28}}bO
zHa*ikyEJiha$)}1+|uNX=a0|S8^^p_ePO0CQ=dNK%}h-#HD(s-M;0$P7yQYy$0p_$
zCXU^6&+)q_ys62%r|vp=&(vM_dUxD0b?==M_e|V%@?`Dq6Scc*cbvHU&WSrZEmhks
zRil>LZ`qi4bn)`dy11}dtB=ndo0(tq7cSJiPVZc=YWLlF=e_rxy!VcYiF@z4dwlZ5
zo%h`1-Q{`o_P&XU3GbwL-+jkVPS)BDbY6UHdVHzTm>I8kl5Ww5#}?)nKQPlc)*PSL
z05$#bdShzApRf6i!jU6o{!Y#m-qeOVJr)mgPmj+}c;in$JC+FYer?>R_;fAE__f8_
zc$37^<U+k!yVRT?ua8gH{Gg}4G&dKfQ9rvl?l)(~=ZL17=e>FU=I6)H)vD9}jK@QX
z_Rdw8YyLuYp+0w+l)}?9jpkzu^{JWZv0PA^TbQmkFEy*<jq1$&bdafCBD)b}z1r0H
z(p)o0&ot)87lSO*Jx>Q{y~ZF+&5bvM;^e|oy;-FP<MWMRGYg|RUZ*q7T6M}_n6J`H
zy(AuFo}OvGYkpjxSZ5^0=c?!Cg6)sbj88Ar$LF31NA_`E?9WUrHCas*9ek4cZ_YF?
zpPKO-&F7kWdG8}j&4pTha>1+lPdz=B3<m96Cu+^{YGd+j&0CtQ1w)LMKR(&4&eWR=
z)n={USnz{_7Hq9LH*>B=hk71={K;pY8=qgCtNA?kKdDc9dSRLYerD#<Og+d=s`rz8
zWDhZaq0ty`HtRwBGVjcUIr&a{a@?CgGnvq&hChn4uVRn!+ei_#Vr;hf^2B>D)clM7
zOtV&==PTxp($$%I`6An_+&nweC{HcbCp7})=0aH?S8ixWTB&2T3*&QEtx)F2Tc|b4
z^@V16@~qZOdAxpke!;IDy%uBiQjOZ&RFKj;{C&LMg6khVc6MRDc8p53iG_u8$25?~
z7Fn5^hhxv(cl^$KPu_QzOMdLJ=bk+_=gl9pezNDIoqo*o7Ma}1hF@jEr)wwJ2tmo4
zX^c<I)x4^8b-zAaV~|Svhxjw8A&7ny7Z9<r*-oStYsE)n&Dd<bmGI)N<V^gN@nO<7
zYs?a-gN#?ZFf&=>v(k3P1SwXvH<k(#{t13h2I<DxaXw>=wmM(yALKC@SgOxx`?371
zu9@3Ow!UqT`^Sh+^4s-O4xbv|$qd$F0Jd0#et8Km>7~51mzj=vSuf}1y@J={6}?_!
zC9lux_cGo9{|5Ox#BY{=1N>TSs{HZTR9qn8m3$?ckLUZnr@awxGhiWAOMB0Fqu!QR
z;#cC8Of9>byBe$Iuke<sgtyh(X6XfQySD@2(c|s(cJa69mA&2k?e+F}d-+@P_IdmH
z+sE<WPdy#|smg$B%{y&JzuF%RQ{D?$I&3Wug(>gzj(W!|WtjdP_fF8aO&p4M(CUcy
ze(z52E>bp=cei&Bu~F|{?>_Rjc=vnn;O|!N0q;TnZu1`U9_BB=O<&@@(|hEV7;oO;
zJ?cG1`cCh0?_K=e73M$dz1w@z-dpzG<2^;r?vKY(vC1B1|G6Ob_~Y+;=FR_<jK!v(
z`lalHKlJ>6e0X|h;g4VaKhFHShl9-AOr1kBNX(r)eQG(!{x~+rW<A5Pr7iuY_TZaT
zjlJm+eD%m<+yC?L-o5qVuYSky|N6x5EdPy%zxs`9fAW7lGx74nUk&Fx7DoQ~;UHOC
zY)tNn()qiW-yFYcuki=8+>Bj`t;VnNlGYY4ej(QAeNW6wT#0{oytSnjt0tPsR%|vk
zoBpAgr&xl&sW(zr;-Ckt`D&)sPnd1>5$0M&!hEZju+S<I_O!MU7Ms1Tcx&`Zf_SM(
z$=Q-t;OhzA*+UvF%=Y`0=74gp#7P-!CcSKPs5wl^;OeHUjQo{2Z_d4u)SFvdz1Wh<
zjgUX$*IHYNZ*K8a9c^tQ+|t@ixV2SKIe)2D;AvZHC*k&1*?+mYqm{i9Z@m1T7+;yS
zGH=k*&Q`XSYvsNC>mYxWQ(AxH2ULpw6!hk`eQfRaKTmp(^++ZEp0%eQ@3QfX&6Zok
zt#m8V8fXo+hP>kI8QNtOtGin%!ac1F;ojDE!hOyCvj<u`c;i8KSnunJl|97w`9bs0
z>=?1z7_rjp8o}G@)LOmW>uc?~suA3^`1{SnS2Pz#D0jsFljhOcW2?unD&>U7>7(DV
z{?lVg|3M}_NL)G|#4ju#pIq{NfYaRNvVbZ_^y1<IV7NYA-ov8V!%^&&_gp$2MlS4G
z?!CYK>{7k15}fwS{YMu4nm6MPdY!LlfI?Xy!^A?Pw%q?zt+uG5?|SBu2FO@Bh#%$b
z>zSDbde*AvYL^>9K`*Y_(R{u7Y^_mi+Uslc;{qIuf*F=szF3=?KHF>@EuRr|;r~*j
zHnlWYzIe7)FZ;Cy2dZ8rfK9i{G;qYe31~WbZgBw^di46^<=Z<y<rf}#_VkmdpWut$
z_1wG8oFQ61{rodkI$M7JeUCqK=3UQ~&%C!>dDpY=?W#7G33{eyF4XE(f$K)l+YPnz
z&6#<6!|(+~x2XkP{AN%HpK5gvu)F})S*tb}j>6Q;++39%zf=R-%`6M9_xiQ-OEY{G
z=+C&v>98<;VcZ`RfLH=P1i30m!A!GS4f+jhsWof<%<{P4ZIE7MAeuqGTAiF5Z!{=F
z>H7G5O{D?(o`ZI7%!A&^xtT=(ilG%=&?7Xe#sprjHiNz?VJu;BW-BPzVN_Ko`2?VK
zy*AefMwyA~A|IsJgA*BE!(23i{>LL&`+`5dsE(&?GW|g<!mR2sPyVo;@|L6LAURj7
z2dO#Gs31GBG&KcW4|-KNLXG^Oza0&+v!LgK`k-+DOB!Y|OwU+n<|Y=ZZtVnpb97y}
zQk0OLUtkm|-(%<(13ER+^n2Bvl8vuJ|IRHog8?R^)~wc-=BrDK9;jhMNz?SP2Ko@8
zZ0Tloe*6;9y$a48<aLBso0k`BECo*YdezONf1BQrV)f1iy=t+uWD<3%y-$d!rqqA0
z3ih?%;|3#0X~zY{rAaVeVO@-Mkf<+$Y_ne%=7XMR?JuZilX+6HMZZ2hw$~RnQ|Vc%
z^C1qW_1{g3Ustf8;0Xnb1VNrnBb2y#)_;`e3i$Q7XQ+2o_5B%@tSb1Bf{KC<EBK&-
z7Zg0H;5`bSQgBMaw1O!GX9?VTv~|Nb=s@tFRbEZOiwe#YRJ15r{<h@(StVa05SgJV
z(TdsBnH?UZ8M5KK7WdDniXo2&6?;Fy34R;uLSb)zJQLr_@7jOGc*&lN@tyHP!g4Ik
zS=b-X#rxx_c!9q~{^p4BPkA>oNy0=XF-ZO4crj5Vm9jUMdj<ZFM*DU$%e$kW5)rW~
z9^<$1b6}fIaK<ZfKii70Wb64R*d!1m1r*76@z>KUSufGbj>cAU&BSc7l@m4xtjISL
zt0{8wUd&6sn%EM%npi2cva{(I<15K#rUldl8e|RX<Y-G!IzOBCl3Rchv)OZr#yf%f
z8A1JKq1iKA<hR$$(lYRQwdCc1(?CpvH+s(VoL}=?;2HSX&hKgU(4)Rq{wm+nPA$4r
z^>Z}VmD6iE{Wr-eb?R*EuG6<R$LsI3=n0<~7>%{E!(cM4!f4F?^3HAg+5x3+j;(;J
z@FfHEytt;1gRUp^ey~;ahOQ>M`Z=s|urDiI1-p0s-3E5j{>DwM9`{ZwmAhG~*&%P_
zT;kJlAFO0{*xNi5^T1YSH$g4*v~pL}uD)PI{oWQ^>8;<#(~vi6b@q5$)N-?Db+bm-
z+TR*I=?f{p+sbdV@(+_Y%Ja8;&(%0JQ_U^jcEkO8y?)k4ua|s4V^;EZ5C>-$W<-nN
zTeI7)##gsrjqx2X3hyy2qbw|A-QItImF<g+1h%Edq0HaB!O!yT<8zB=#~bCd;};+e
z7s?a0GUPgd(JMphlxyRY5R)M|rDoc*vo!7CZpl@yPA-W{;srz9&zyO+3&Bw1ybn*n
zUehkQ!x9&Ozns!bjrG~v1p{xHyTvN|wMoF{x^}m9vgS{$%Y0K`9}KoZ6ae3c3o}^@
zc6QSA;wT;HJp;(}ruRPj{l%-F{qj44;Z8y0T}&?2rcT^6^`v2WZ;Ed9A0qfFzx;fq
zO&z?`f0*ax{`n;%yhFUN<^iWJ%uR&gv2ltqFNmrZl&d$#b{J^%w@}r;Ob}$FvCjy)
z*FFEJQj?$`enW*it3bz{1ACc}k&&GOqBQR~>0eNtl!gw#zc6jkA;<zoXXcmY9Xxnd
zwLhugdsH}c#_>Y(Oj5&GA;Hk;h31n2?elcB=Do}Jf$Fe)>f?3)Z>mmmd~R;6WDw9n
zq9@dxfxk{K4NmE0LB>t3z0d!+QgY(Kh%x9oqDI$|=>9xupXRqQK_F03NF)FWxkL$2
zkV_N{q>RS<6Z``xj3l=Lc(wv!h5;Ih8#%y4e{yFcM`_g>B`hX$0!}xIiG6^YBJtdf
zT(U^*t?}g%11W9nAi_d6@NWcLCe^Pm5qwO0Ip*&ojITmV%qCY;S2<OoB9sDLm`&Of
zyLl<@?*ej&>WaSsJs6KI@AhIJfKGhDweq0V$#%8k>OCCROSpOw4Whr425Y!w%}RI8
zi!uEhOP*fda)0?b_)EgIbXW*a4~c^1Z6NVR0w^m)iK5K0oWoN-%EyC0Glw<*BZ?Q=
zs55NB$%REw-Ua_T<zzu8RBkL`3r4xdyIeQnL0-s4gQ~S(qFUn+LE>e>{t`GrDV`wI
zQ!YLjPkl3$$h@U^e?0NcL}IyQlh<ybi?OtchI0HGLXd$>H<psgB>=}MFWF4AQmaXk
z#1~@zP>Yl;v6WOObtT<OKOTE&WF^x|@ONM(+sdq_TG^{{|9CU)rB*YojDQ-Go@!+t
zjGf<g9wL1<%R&BNGbh6Ql~^;s3UFHO5fQmsWQov{NN~`O-i>)4WV4rhJ;~et;6;!T
zP>*~obs_Fo&wqe-<hKAoKqLx;#U`l4l{g_l7qsKN5!d$Q<<F)3R{^>pF+yPgq+yJ5
zuNr9$0wJK*c%W5S?Yj!7bFHtGTzg*OTmq83zts!C>e)h%=}~c|4^{-of>K6e0&1<k
z2V&=cxc=9}TKRKnzWv^H^>(iGg(-!wrQwzSwJpK;=<&!KiA|}uwAzAYMa%!<{LeOH
zs~{H21*{INYo|mz1D)Dw`s5Yvfg?lR|3pn89<Ju#x|+S+HF@h`tN#H&@cDFW5Y&%<
zuO`%+l|jC3P|TEYqy?t$Yz?l$grG+^$!+ZC3kG*+Ol?%@%LiIX#`+1?#xSd_SL}zn
zw)>{eGqXLkGR&&#vxB0aw+$JiW`NkR#RlI1aRh0w9IxkGik_UO#!zeMhhpx4aAm{1
zkC8Xr8jkX!!((jI=^#0|xO6QZ6yfDVO~Kd}$y<QJhFLGN=z96_+FZ?e2#^!y?^tmh
zflh&=lY+$24|@v^U5rzS{VO_jn$`YNhHr|mBwF$BOPoix)Y<lM%@aqD2Pu)$LA=_@
za6H<=10q_HF)cpydumVscsRCv(6)LvL3#86^?t7L5Rs_P@}5&Oj{UKK5Gy>}bkx;r
z7o*2(u*AJuV^SRC=J=fd{k+0|U4cH!zd~?|@ySR8!Dy7)$D%Ptu4f#F{(3JmGQ>bI
zOvK{3-d~2OyR=xF1a_9ims#$6Cd_||r~3=#<)Om!og!i4*homqia7KI^@5aY+l2Vd
zhNJ=RM9qkGJ>GQG+qeqm5gqu=g#{Q0X;=x6%l#Tv&xv6~{X+D+l#mDD>qvAcMtX7%
z*dL@Jc&Fi-2oDI-5)L(j{8GJfeyLW2MbT$8YiH{1*dN>E&h%89w)G!Ve-aJP{{=<j
z=Yqt{xk|1c3WWThQfk_OMJ3P7RE<fG2&i^(ZPW?`R=`E$GeLZ|Qm9v9T|hd|dF?ng
z!EoFF2ZheiuUk+4Io|y1{5HOupajw&dNaqrg3+gmmlNO2rSxRz!;M7J#S)+$rElai
z3Cn>Bjc-pBKr2*xFg^mRQ35Ir#t$f#90Zjp@g&4#m>R>N9=Sx~<*f-Q+#CBqUi1$L
zc4K+yCJT8|TPXU;5Ycwhh&CQ2fRZ(7NGm>@xRPu^*FcMcO^}*|I%A6|9nht6srGZ)
z(YsSg$xX(0vb0T7ROaW2S`z^igaW%JlCqW0AG%a+mRsdZ)#GDF+ePoUc_5J`;5YBL
zY9BsbzEpmU-=jy5zEpO&s92N6|C*~cU@g`cm*6w5ZzZe`yJn*1t43RnR5os3`H&G4
z<r8B^%kQrFwZrA($Vicy!dgZ61w(L6lYR~Pbp8JFBjx52NGo+IdAJ<m{)fwJDS@%W
zWsq0BGW?Q*jj}DlPJ_M(tg*Del|TM#3WSL}?DelHD-$9zVGBXLiA;@m28bFQ9a1T-
zedoHB81$X58qE=v9$`uv+K<E8VKACPe0kW`#=2r%SdA7&nqMuBuN@-TxHv2&87+fW
zayT~1(R?j_`dZ3;+h3=e`?l{THkNXqm1Oq)-y|VO+dQuC{b<Ogb_>A?b*&(Z6S1%9
zGX~iZi5rO<%RB5dZdLx4-{b##b?mnZqMmkU;S(-(ze;>XfqH5L*!q5My~%r*tDMl)
z3x}csA54Wk-F2&;hBa>4SHn}+jXfj1ZtU5r5?k6g`M<9o|3?M;+QC6$L3?E3ce`;7
z-;Nqd%>O-#tQ*nw<+^$)Dlu9i4RHd)Ua}&h<0I^4l#(zuPIP>PsViwKr9;H(fdw1$
zOvj0R58J_ZiikD0n!ZmlqcA^3Y%FW*%Kr_eeno*Tu2MGv-YLI+RC>;Lc+q!j;!3m;
zEwTQT?U(*=zqH#lrfJu3<f+vj4lzwZ#NiCK5^(mA9nC@zfb2>>#yQHSf&2G*awTmD
z8fr@9ZEK|<1(N>WR(jQBzK%!5Nt;#;<ZsCrlPj5(>`D$f6ezr%xY^UjKe(1~EA@9Y
z0BhI?fBXw}++0hp8z0Ue(N&IQ|6O8e7|fxYQ1##BVJz?VYW$LGRV?C&>A$c9r~gtF
zrizd`CkW7SV<Q}lw1-Ii^EQKI{h;?Pb{|m>Yv2cYOOfMgklz&lit!QX@bIu&S7B0P
z9sLxEw8viH7sN0=8~a?GF#dh<XF;Yv5?@Jc?Aqg>ggdM8%#oT$W=eU+GzEMszz;;~
zehv=pRgSb-$g}CVmlCBr3yI<-#OM1Qt$s2Fkpy|<rCa&o*lPAF>8*m>ienjZ*8VGf
z>5O0p^@GoCXy6Z!56Nri|Cq{tO~E#TAZ;ucqi#*jU|fkw@V>F4TkqX%2Sum6Q?;Wq
zh)?)mR8!XIpQ!4fo2jlz(P9aT^@X~r>dcQf&aI!P++ww18Y@p}g2bEHniv7YI1Ixh
z<dR1vx^N2})7lY+TdWdc+}Q&wjI$Sa<S*QCv@X~M2@$~4@s&(7)l4_xno~A=1rCPE
zue0#WMW`OMxq}xDA01+K3OR5K(t5VwrC-<n>e2q7CCzYg1gr%j^ig|4BVshh{(1tk
zy0?`UzvO(mmGDoCM?VG0&`N{3<W3t>&dg}ct`|&EM>u!W|3@USzicdDAwCv&G80W9
zY^qoJ<1%j}W_YewH~crq!PW#fu6h2KRj8rh8wz}tOo10qyQWMK^MfR^8uqJzmP9b@
zP!H)8)zIbgq~+9bull<3lTx1Q?auvG)n0e-_eOKXo?KWwqp93ZF;GXQG7~R@RHoot
zZD+S7UY3X?6<;275G$O1tN3O=mAa-`o36d__VA!-W|B44niL&%DNgp}6zC!wPP%^y
zG|+sw(y)ovNTLw53n$XI2Zf1o7)K_#w1R7jt^2TyTb!a2ZE+&Z;fut_TSnUo5l)z&
zqS@$2arHULZ{vOf;7`)jWZ2rT<gO$@dbI>O1YQI2y^?&5ZEY!hgG&*vV`@Ul(~~sN
z^`mBo`2CtT3(SG|Yp@hTP{ty%B~p;=4*?^3VK!API)h&Jbq1<)4{#WxqW91_KJ9A>
znkvMW9TIUqoRYOfsVmF)l_<F$vHS#SA@b#3V<^q<sHDBHmsdEFI2W}&s{u7si^JHc
z2`3SY^1C2kG@!QE57DdvboMsqrRWI2zNN4X!W7UIp^6=JCF-7?Y1T9Aqjj${DCub=
zh!+;r-QVLi%LSWaOYG`U*Qz{ANv%o)zeB=SM@@Z(^_lbs;LIi9OQl*V$LjPF92CdN
z+1E@;i3Z*qH4KDP!llhxnhEkChE%he_0kf;aOk5Hbi%)bcfxmomVhab<kMR+rHkcV
zOZ9X0g^P74251+_N6H7kqb+w{ImqTWd2VuiqTvTLV#~#Es<}d>qc=Vm<mRPSNE`lD
zvZ8I15<(heAfFf#tF9A<LhFVxrx8=el4g0777Bn`+xGdwpQoS@Qkmgzq*A%J!V^9!
z-!;e)%+-2ti0Iq6jdTpWCb}`{B{`cp5a406i5SlGObkSi?d7Ey@ZGH>vQxBucuw;?
z_F9Z*=?W#i49|B{Get|RrK>8LhvlSmJ{f5kp%oNkS~&fsANP84HjB1g&*$Q3@W28k
zFXun&_15EFX(bO7K$c3)zBdxBoS^m$QaPBb7N;x_9EK<mr6o3D&Ovq{eN?ZXdOcws
zm4nG?h%H!I*z$l{j#_mvW^C}*M}=R|k`_JU_b6ubWxKMy_WcqWuPPH&_2?uJQ?1dc
zH8!3E5xUpe>Cz-JywC}g2HqE`ErgmV<>f(>E?tIVZ%de9PT4x5IYU>zUya;N0Mjbz
z&CJ8s68}C(q7ZYvXi9LQ(#G|?dqO~UJ#EpkUUxY4>mxd-C))NgTgCo1#kA;=-b9KL
z8K=Hs`UfT&GM$771F|!-{fR<Kh5TIh|C<_d`W)gi`e^2JC=yf(LVwsMQ!|&OJ%hrB
z(=iEh_CRmf48A*vqd|U!`bLZmvO(A`CLt0Fr@-@IVtb;L7=`^Z2wSF<7=|g+MvvSi
zbqRWOLW%wIUVa;52S{dMh!3X?iyjmS1P3My=uv7lZ6}l#IM8$&xFqCrhVAg9sH_4}
zlMZ$9zez(-Lk4290x@mbp0EjrH=uwP^A8h*r-q@Zh<SZ9fHLPU>aRms*Bu&Jc!K15
zKcZns(cB&cV#`H4fy1mW%s@qSG#F42Jt7hY#1Dp2D>;i~!p`G9A%K=@hlF4+xydFc
zFKAzUHN%&a!Z)T*+rk@596fz~aH4#+e6c)PKJrj`f<HX4AbaO)Xd+*(I_4$gFwR>Y
zYORh034}y8Gy@V4%}g&XAWaTAoqtM!h(D1w?8E;R)i!vvZY1(L4vp`0O=F>RP-v7m
zdqB!~C@X3g+A!}L85<RcFRf~1n0*dK1bi*fUijHMEKwp6a%$%Sh{HX+%+z<Hi4(2r
zI;f?MMmfQ{uJ*ur0w%yosV#ni{D#(CiPf8nFFTTUO@Xe{y+>aab@aE0(8U;7_#HaM
zp_EdHK!nFj0_mPKjR>lcL&p3q9&~Cmj);o|D`AWYc7_du2dqk;xs`^an4T6@oAB>~
zp0$##EMkn^=RqKmZGu5uO)wZa&TK9Iyy-Sezd84M+9aTy_T=Qk^aAP9*Nn~P{W)4F
zToqM-5XmdjH=Wd3uNMtdwU9-N+Lg|+xJsvk_@XcVQV{Df={fxFC+W<!RFFeyVY)hj
z&abrlWD$`?vs=R&<cQQ37bef5XowLCDs@UeWyA!?fM3bX`BmUKht?&=(GddUinR!!
zy3a14W$xdnmvahDn_$F$Lj@is&G%HQu&{_~1)6;D*^9;~Mb~g)X|XZZ<H!Xqh9K|o
z<{9r&(A%z1<<Tu>eT;piZ!|(@)jXz*bxWr=I)m+VzC?+JHjjAOTI(sq-lHrexY}>^
zr%co8Mgj4+bgIO)P9;JDqN|y%6GCfIOH7-y@f8B5jJ@Eo6T%(fo<SGvnWeUh+ViKd
ziRv!-iRd{j`H8q;^Pm>5<X&UZUWtDq_VGAMx|}Jb#4RQMLd=8vor+ya#$)x|iCE*2
ziy#5|_aS~+EMn0}iC(|af=jfJwysZm4MnUCY4PqfQ?|*6=d@tyBqQ4X5q>X_V043+
zT^#PQm+~v{FFXhhFdz2X514rC8k+m2hpGAESoi{ef+TVrFik`#Ks13om8c$fb829$
z@PLv{*!Aytvr!aVW6Ou^{04f@pe2cL&L(>eI?=9l%kKJO4ERqguod(Sv5j3h(dml+
zK1%$W%D<oB)EWOl)kxX~@c%wv<G)J@jtB1BA@^>T{vuhG!Oj=f7u-kQB+b2TEaAUb
z6`xgLgV49`Lw|)F!+)jXAutXgXI*dq#&CMr)_2HyT34-WEVO#{Ymi)1$I1bX(^}OU
z5@H!iL>)IX*->zCBCcde@T;506s+5cgG`I<BF-O?#i;K5mO8He6RIn2wrSj0t_$|0
zD6e0GJ%@<^iKQWsfVO4Vg>*0Q*kZtAF~MNYIQW}7eSM@JL0boS>kjl@ADt+NQKcW^
zX{-&DQ+0pgBC|Wr`+SX)FRoAoYO{_!Zv>^f?7$jT^p~ns2#TSOvl*1FACRG{fGfp9
zCbmSfU(tvZXxgcFikw*<wNbfA;Wu^c*f~F`L5zO-h$u&}uW^Y6WocoD3j<`6nOMof
zDa-=F*pGl}K=zvFq@gca)5<|%18a+f)FOa-S&oRr1CWVV&;ynILMv%Gc?9V|D}#=A
z0igt=3L)?6ls)A=ByUha=bAmuBHXLOG_;Bbasby0P9<s;g223OwFIhEpp5BoN(*$h
zAEgfTJdH0_gt{4Mp-KSnOVre03r$CQ1-xyg{QrfNA&-<#4cO5)%+!mH(!txiQrd7-
zezNPR?4_L^X&usP$yzP)l=gq7xoLHz$+xd=mhuanw}cxxL;b$+8XN#Ss8LjS1-_n+
z<>xr-fX99ApbRojyl2G6^~?!dc0lE?GeQ1#CMf7_%k{J#D!KQ~EtpLH44Xek0mnZ2
zkc#$1ONxSBJGkrc9ieGEe6u=y*B-oUR-w)~x?wQdHmgn-JJ$*{W@qb2Y<aIOKS6X;
z#9vFLwU5=YQU9XqwvoFaiD1$pt#_*b#&eudkuNJ~DEOcP?PSLt_>5wP)??Ki^Nqgx
zL8ZN@M>|pTid`lRa}Kc*8|V(4a1!*fzBLF${4_E-jij#{K~Knjl$L-ynhYNvS8Lt0
zmoJHzuW|M#6xgVEiaF@h8HKNs*tnOjWFVGw;-YL9&!8k%N^A#jaObetkflTleQZu+
zROm|J6Fb?`g9)~E-@0+_noJFge(jLEp9uRPwi0&?J*R{vG<!CMB9{KbcSf@~(ag+d
zXLDlph)Ee9d5P0Z@?-ROEFbimNr3d;eR!~@J>iJhALkvW_-^{^j=)G8@)e{X@y6$0
zh*(={D99M@)(Dc6ShJvPkcOR>KSngwHuBc!WkDH<A--k?giEUI%Uf&+ZdU$U=CprO
zU6RhcuW|P8Q=p-6@F#Hws{R!FBTI*fo}q&;5w-L1%xMe{o~LY3sLJL9%T^3giq-1*
zrSUmumSP(DrakEkWpNA4k(_{JS-`l_Ij2lf4*u^d_yYxBRq!T3kirsl&PhAIp+vEN
zZPB^)%Q1u)M1e?E-{h0FM}#K(?@{oa0<p5)Iv1bR7t$T{hPE`5H7s@3YW;jnFVax^
z#|e({+jxy27h|QtGyco}EB%K=74?l9iF=FjObNBYhjPVSCYQ@)O8HEYw4BnBY-Muk
zLZ;A{+hw`w%qT_)y}48-%e?NlUQN=Wt{;uDej0zyw{^@yG3HF<llfFW&EG^mLoDTe
zz>ZU}Rg67TI>29vRnoQ4wA1o19EHOT)UV|8rTl>Rf~Dd$fEooe6P2l_3`Aw>`9bp>
z5Xacl{03m`X}$t5_S_{O0hNB|8K5%YJObb^;X~d5?;u8+p>e16s$%ar<c+-opsNgr
z=9eG#Zu4$8^UF=%hrPqz5o(Quc`^tE7+~}%bI=$jpUxa~i}zw^4!YHhL1pymj6t`1
zFPS+g=AOJW;eEt=ptcip)CawXUP(tLpXv>HRWl9M7~tbTrlO=itWoqnYNn#f+1<{e
z207zqJgS^MVNSx@n(&_R-fds8H|+b9-g~U?`@-j!yr;aUt;GIzP0d-_oAlo2z29;U
zc+Yyz@tp_5I%mA+t&KxwkU9oH^n$*pId|&KwvE<uVO}QOW;J%&zmpXLD*XU|CdH75
zek4&9iX+s<42v;Xh*c%cfX8I!otRsYVuqp^41lhji#Ilz?q<$N`CYAyzr#zyI!%GP
zB#b?KNQY|UkSAoO7c&-!ez+GUi?jNO_gOq|akM)rQ<zqnfoA%ONcp0N;B+S5>J<kI
zltX5fS=hHSNrk~s%`X?51vp_Cw#@d5R|K;FQ`8q?bxtk)^JKVM`VivQMjbRKR_&`3
zuJ(Jq%>mF58Cn1mh1S5m*BfH1AT<_)aa`Smfegl|W)QpC;-jsSsZjrYN^U`GbU^hz
znZ%B^`g~Gnw>GYt?&~(M&-5cP0-fDqW;#1t{r=B4w|V_aw~}A75yU{%8J0#Az}UcP
z$4GrQ^WpDSu!kTR#?q*PsTL=?7a8VN{T1bZ7r`n2WknxV@SO@c3Xt12P)Wq;0Zs=N
z{4@UJ%6J#SSUK1XTWhk0{TnA>yG?+XabSbNE-N-G5H&V&84Az~l8XzA{`=J6`}L+C
z+CX{Jj1U8XsmLnP$PyXgo5P4G6?DPj`-}9mlJh)j;F6(|g?O3AyF(?f4tU5x8vY7J
z%&4G)%dGS1(6MR^M%Fre=yW3(?8L>;6Pwbcs+H2j!~)tz3+HNe@lJvvsfwdr21~4M
zENFUrjwX9fUp%b9_|qR%Y>r@T&~T*6;93*Q@T0^`j6U2Y`QN2)61&*HqTss~d|ZLe
zoqh4k_NV}FE+N_1p6)fX{V{dHkST*(nvE^O2tGh0(+gCAV9h0D04Hq;_iuz${Y${0
z$r!5!m3Fw*&Z0pEZ&EGHPy;lJe<|!0oLL;Ld~<np-9gi_f}^MHA371s@@sh3vqVhk
z9fH@`)(LG`QY77%^)xC!tEl;`W}HR3+s?DutJp3GCljj8wn=F_&SqatJYXdASjue{
zsK;k`f&a9ER}`oth{}X?m>}Ju#q3u&Xf_=)9vY_b$7H;VA#Y=G91feN1+-^BwGf<p
z8^mJHtB0G;&_wh(jkV6toyf?C<I8*2kBwQPc2$`a_8I+Xgj8LGZ{vR_;QO$*B;-hg
zs*_y>V@2%45|8uuB~p@&Dj}?YG#GYZ<%caGR(>$A?AXa*{g167C3MgtenjR3)rgX~
z<ZC^x6fv>Niqdpg?PYX1TtBKj+3+WjQT1KrpuCFZnT|Nxl^iY=sc%Q`iH@xGf5*@_
z()b&!S$-$j@L4o0<q!nL+4AHPrWo_G@aAla6qY$rj&xEGxNruH*K4rf=PoaA+q-vf
z`T6H$&;8t)$B$s${n#T<>xpzb9hP@QX6AA(a&+yXCvEGo0biR5(`aWM?|89}b^00^
z?u4<OZ3kML((>E%RWRtByI@b;wxAA1#ba2L&^DBoUyvrYZq5`6!zg^TU;&-vD<e&`
zHq#0RCFAIQqi+pgLws1#Pr-nDo!?PA3*yWE$26$zt&UN2H`m9`tq*EFj#BUtdWAWd
zR??Qz-bWpn`0$@D?_0k@Z`RB$DTFRVS1n$#Lu~Pil5mbESbE|OnU3C7$8Ois71S0M
zk@PV&l;Tu}?}f&PK|KT)HU;yvWxFxh)HsU=1B~;#wvhii^A?TeDHVP~fx*lt6?>0@
zrxbJ#no%V)8Y6LnMv^k;g%fmRdH4EJd3!S~ZDD=24!5rfwTd`pt%Nwxe5rG?k*M<3
zvkA%P5RIf)z^PivYUVta9q7R(PM_i<hNvE-<U=PISi77|8_T@Z_`9`Cf1GmuX@YB^
znyeiD05u&b`Zvn8#nE0NzDa}m*slrU6Why4q4?Z`>pyob++5R%TeAm$j3)WCSmPm4
z#GFe5O_XCo^ai@Qn$#zfC%wge<Y^$B&ylqA_+hA(m6G0^*o+Kb)&?(WT(2Y+%`6v5
zX~Mz7>bcr;{_e2l__<``aYLi@mSp2ShJ3Z#iu%BL6bu^rwC2OA)#d%qo56%7qR@?7
z7WA?k3uO#gwb)$JQNP95FXzuhR%pxHQ~-|@UAS22H0cx~9E0d~ntWEXCOGUY)<rbi
zWcxp)j6-_to>>14WvrXT?;-87R_q=s2wFR%0jc)qlAS3n!XVKUFCTcjsdc+Z-|D}E
zcDq24cCvP~*0kBuPs;|4DHPy&J;4<812$5QYu!q(YA*vyB%giV@aFTmCizTcrkMc%
z3CyIJNUYglduEAc0Z@wNQM${(o+RGNLc3(dM)?x2MJzBFBBl#XaTJikf|Wh5Vflc9
z7YN1*ra9;TAQgf^4<%C%9Jg~YcaCj`V;biHL&5RMb8Gm^p%R@jr1a{_j_AdB&oD<`
z(fdv8VN9z^V%hU6>c@u_bWc>U)%eJ3c!&;z6z}<yn!gd=BGno6z2OANepLUA4}IJE
zIlFaJT{}`oP>Nrhvd4*Nx|)z}*r8($3S>{zhb#@5Dyfh<q!fVQakuBtaY<-5ns!W9
zdRR;cwwc>D03g{bICZ4uO$Xb0Am)C<+34c5tsvif3^ZoYzP{G6YYYkdVLe_Yz#SY)
zO$Kv8+k|TQe^Z0`w>FNb|7p@bqoLFJ<pv8c42eQ=`Ox|SyA>?%8ZMe@pW2OO9&~^i
z4Lj!8f1-H(J?0(>Ml>!P_D}*PUr~s#2oX3pGn6PZxExeMN~IHq%97>l$+e?)uOu2L
z?LZO(UyMXj?3lA!=03~~GMd!F{evu~^I!_T9T}jh7>S5*^8cPj=EoG+m>!_tL6+}T
z*$7*WPp)5$hMgEo5>iInK61<PccZmfF!8KTkW{9CGS?vHtf~0ThxTR-Z4lLp5oIg!
z14N|P2(W|5*6`q?485#XjtUMFX=tJ3VfhGTS4nZH94TQ*9Xx2YgG3<zcm6QY2ViZY
z<iSXUCURF4+r=|@gYEJRDobWTc44M=(HP(u7~-xlG?$5=4dNI5vT7NcmBg*#`f<zQ
z?dPnV6#kUPNoQgi%--0_8|a>>XxB)nN_!+;=0RU77%BTGh0`3Cd^0I-Bv?Swb7mx3
zNqjxHFXzyd4#ewdw@HUWqrqsVM#1&hq-T`QFJsfn*ht~ugAxi=&mFJ1<)1H<!=KQG
z>g@gHv*v(lea6N6u_*MKXFKVUZIknm$Z?E+H;y!Ia$(mdxhPL8_#m){+qKFU@sM~H
zqr`fr@Pk-TjW@%hW~Yc#0Lw0ub@VP}y4EIT^zXMDTccT~8JFay%%lgobG6zcI%W;G
zBy1sMI$Q$(A860~w1O!G|BxWko&JIneu_Y{;foR}${Q8?#INI~=ZnfH>MN^rwW%il
zXJBaR$(0gcqLRmzgG^MhT3ZJPkW50oRNavLlj!&=XC=rOWi7GCwy!_F)>i3~IBGSC
zKD2$cN`%#7AQ{RI{TcR_(~*?4(Wq(6lhLe!hGPY3RpEIg$<D(s4A1PRVe(N~h4K^C
z2mOuq4SMcsm0ViQTC1iz>7~O~SxeHYY-QEzhIRnv^1q_DvO}|{D}CmKipdW>Pecjh
zm%;>_(Mq8?Q>#w+IFs}%L+vU`bhfflBGo#HAk1IV<k@~tsa9pMu{ejXCzAw5MxHkJ
z0*!_ZY?$^S@$_YWFPZc#$h-GJjaGV|4qq-_*t&`B6_9Bc{jz3uJ8uwgMZO*mV3aAY
zi7>R(Tca3qGSfTO&vaKCHV$3$Eq1Kt+~#{Y$$Vo`@?y%_^q}P=1}k&TSpq4sCAJm#
zjdV)J`JAF*I)B4-Xnj*xzJOAS)Rr#*o+u|G1i^g@(|tS+akAtPXcon2UFo%yG+rbD
zdH_dgf61X?#JYUgol<yq(rvv_@?%P7uclXX{DrORjPtL=SFw`eWXJp7)f7hVec|)E
zy2G(4t;N1i$Ll5QZOZ?jVQ>HEx6xZePU!{fI_woO*!u1pwJVeZ+;K5-^xXsNzcMVR
zFV03K`Q`x|leORM-V?U=ZG5}RV~fJB+c*Xr+KR@a{s|ianbm){n?c3?0;7@M8jS|A
zuF+7A8x5CptI<%oXf&+kO-DneY~8gK`_?N#rSd9!lhN4FR(muO%^vVmga<fS=$k^B
zek{NIY8yI4R>AiJJf?-E%AFlLCXK`8$;0KdhszhoI%V6y%nIWuchMXfx++-ghP616
zaD`{)B_BTi(qT;KJI^Ox>a^FvMR5Asv!+y#^JeCpGXp#!d#Kss!dM2*UM=XsJPwm&
zXT0+L>~*KNYNrkK<55za$lnm>CXYep`T7(F=RxWeq5rcw|4i&}460XD!Opv@#6nKA
zQ~UfmmHY*QAh9$RBs_j68u*r*^xFW?|HmrxPY5bKdPjJG2AQSsl}hw~Ms4&)D1d!f
zrMRZ}$@OPtpCROG&$aXGIxQckf2H(j(tPGX2bB~15Z~{U2OXY96Dh>|nYWH64=1+A
zZv)Z4S!xQN&GN3BovPLy_qNGAZ;F1zUl9o9FpV`<`H&8m3UCK4YEqp6Ia+JMS_W-G
z`5(XvaO7Z;%m{xc#E^oaN&_3!bdIGygF=AS{S5JZ<1=hpskYw1Qvpg@YBxRSlKvUe
z<=5pAz%c-IrC8sj?M=$NA;*gKQoBE>zqH?5tm9Hm4$WqZuzpki5)N03%@R~a&0>Af
zA?N5(zf^W`^Cj13#H0F_mS`36{o%gvW618J@57nv?12Aws4*zY1r?qWo>7KK<%{TH
zKN}4<VaU5ut-hHgtq&o|qrUZL!2f)AO?go9(w{{U2vE>FjSLU#m|vg_uGQ$rU~6D?
z6We}Jh%eHwod0E)B1I*hGybpeJmOT3S_5ov8AjryXB>^zg?a@imtdE8f5ft_DsfGa
z#4$LIGkQ$4S%pff&4K63^8}`)Rs1#T&t=fWf1|biZwbaWJI12%cH3-S|5M5rmFa(<
z0+Z>O2u2>3oq*;;tQyDDF%rylaVC`Zc@=q)SY?xxCYNwn!j1=X#fiks6t2oQp=Gd!
z?lS+VXdC<p_2kZ+qMq?E#)dljHKv9eYz?())-W}y?$AIF&ao*b9k*G2J0)F0pw>18
zZ2QVoj{l3iu7Z)F928U?d$~h!<2ZKG|4DsFd!#C<Ni~*~4R7iaCcNu8s2ai^_EO^v
z-(EtYS288>m<N#>$=1O!8Z+9@@HphIVwjjolu+oAehL<ZNvsK}W7MbBmUK$;CUB-L
zH{0EY-F5M+gx%>^sM6OyN`S%UUMTa})zFfwiE$avNuCAv%~%(j9dc<FWgbW}+#bV&
zl0O|Zf3GA~^P=fZ*ipc(pqyJ+X+$QhRQR>o({Jj9jvg^`Bkqpt6B<H^)~^+fG-*?#
z(~vm+|4<#to_uixgIvVdstoFo6t786&zaDsf7v)xL4m0Y?GKDoB8h^^g`8hwmQ;$7
zcY#TXsy?Uo_vuX*98^q8?VXyc5>Mz#alFcxRf*U}LDfb^+AT-h|9|wd-%{|~3Vu&P
z_q&7b8{4bOe~jt>Jl*aTv%aQNhu6>i+ck1)){fXkOtaD5ZOclI0}a2iQ?5B7p97NJ
z9k(;#1L_|TD+a9DAko_>3hJW&#X?Q-E9ryrY&i%;v6%q}N3AUQ!Ab!S@|0pUlQZnP
zm)=D(jk}5C@2>9EF0E0kvz=;6`Rn>Dou`WWEI2o4J85~_`mwx~AxD8a>a^)6|Gz{u
zbYfH(At+lTU?NaY415U;Gcsa__!Nn#vY5Z&lS-uMt;#}D&gLNnaC8cC9hNdNFl6cL
zcrRn$iAz?%7(-6#&Vp3(xABBA_XPpp%OLhjJHu4-N|cDr_O*ajQLhB2QD+5a(vtxM
z34?NiC&O^70ppWA=M*71jsJ+yJ}MpBN5NJy^EJjXW=2w`JUTY)<i>`G{gSc_IqSd=
z+th!JO3vr*5yd({<cb~*O#Gl?pHR@Rpr}9wMb6Bs;Ld-sr0g)p=@7np-Q>50|3UnW
zd&zH73`?c;3_sI=t7;91OL!oD7`zI1wRR#)2ViS_Pkedv`ZW>iLUg%Gh!*f;w9}X(
zfPsv?Gx$cPK4Njynj5875`%n`=ZQ51aXI28&;gQxJuWI{!aR-pR08tSEYVmsz<V@&
zxFDgGYMej{0Q&(Rsu%~%!EWB>?^Li$0mCFU|8;=3J;2x&%G*w~XH3D=Mkgi67ZgIE
zr_=GYM*NS}X7|nsdhm1^IV_*o=t;Q)tp{j%IsC<UBUXD*@`;W6X-&-yCuN%fc!=is
zW;VZ6M<cH%EmXEG6%Bfqx{XnA$H0HkM>`3h{}tk9Ygg&fVPCxj-_75^a2l1pqO<9U
zytcT4-x(bV^2dPGZv)r-Ru#LB)-D!Qg&xsk&Ij<sgh5ziJHgiE3x|~li^vpYI|f?V
zs>!tmA_u$^!oCz^GX<jwhLXJ5t>(SV)wo7rEITf-RGVIVB6JRg3v~R)KJ_#Xp8R?9
zB`cRv=Df%`DaTwIF5Zy|DRdl`pC;W-;^F(L-PFWPvvCB$19$JtcN{I5^LcA@a=d;J
z<|LO4g?3c5gP?Dw>0If1+<Q~^<owCpPKDx;PD2qDigMJq^iAbd*4NaHsM7JW*>8x5
z{qX6R^t}j$m$*&k5PVd90Is9teTvTrOX3?((w9zK9g5T)WACulgr9cYubil4(XwXL
zPPpfj(ep|FZ>U(o9dCBpeMT#<p}+{u-%@Oof&m4W6-c=~Vyyj!5^~yP+JKepc~rsV
z@a2no>f=f3-I24YO77eoD~~B%zZ(lKxjM1veYe)GKvN;No#RWS;PTe>$5*HLq*iJ4
zBkw5Mfi%H|7{GB>oldA<WY1+y*A#rv{5S}PNrxYk4TBXxA>de!X9+3tn4rtx0*5<Z
znzJoqMiVIy=i2@e4A6^pxnYKV#`9gREYJP+oagybT1wM0E=XvpSjT1e6<tjMjRh@)
zH%PDN_2g;}BlP0uBt61#12vr<1aiG^B+!S<wT;e!i!J<tf`tMJDG4K_*}GbjAq7}!
z3HG0^-{|xITDyf3E%dEvftK;+DEQjHCRJ{;trXB)xoi3bU0d|}F^t#u4C?Kz{^6MF
zk`nb(ChYw)j)qhHA(gpDZAB#p)sJ?Gf^1dWBl@qR5pB=OP&g;LGDLHdY7KPFiQEDt
zp(f?KQaYf6-Z0Qz;!QWFo5K2XpcHQn0fmOGK9`3ur_kJ1UoHcN?P<j7%T0H-HH3l6
z`WeyOX1Rc5ctg$>%h`OBI$Tc@&f*B)KDuVKqx8+BZ@o$SDCt|)q+4BAax>SsD92_x
zYG=#Fu^rp$gkzJ@K8XHmABYR`?RpXTekwRG0viO=k*)h03&79~iD67mp<#~>f`|0N
z2NgU_aLV7IM0tVm_Yy4kIP3rE@p*2H5b9}yJb#}G>?imuf5gv3AL0KYt^bvq`)5&e
z%Q@HcOUIW7qL`EPH^&#2hh37nSn8yPt`4|p0e6)m#vC5_NBY<U3Z7NkfMQn_+oo8c
z*g?f?csf|89neNpo~H=@L(w2x=l{SgD#o@s#L^Du|Ez{|I2#sONjIx>*Z-<ge@Cgp
zo^ZnokvSLg8q8iCdld79;5;N?SPL`30TC<?jTE_EX&L+Ala<hgRLJ49#7f`VMyr>O
zhlTH`jC5s~V+M7Ks}a}q!g;G|4`pSvEA!IvP7c5g+g92ivPkz7RQki*c6kSTdg;Bh
z_&7}L(@b(hJRP}s?ibPo$3mE3<i#PA*os>#xqi#=7qkkG4l)RJ9>G(XAq~Q{#B3>H
ze~E8R$S3${_)mVrH|ozuo65P26VQu(BHra=Bs7G-;s!+!Nd#p9+!4+qm=v1wMmOyx
z2ca@bp(92)<)n8Y@r>Zv6b@Bdc@t>#V0t&!>VX*TF^-<q@$aFwM21NW0_BXQw=ReS
z4FxoNxYffmUKb38rTq5*q0$l#NkO!g@jpmO>HPG<O{C^ZiLfNDBI#HFs8xj0NmBb)
z1~|jh@Gj*s8NbM4=(n<dBYf_)XUj9^bvgNh`Y)O#J4>{e-O=)&6%)_J|1Sh6QP^(M
zMyhBT&rG4Hjy$*KjEk<S;2$Y4CcszJ;G3!!GT++7m)j#RkR@Gk<C?WOG*TKKh`2YM
zOLlPI<G5?bgFbtu<j8Q&P`cNRC}FhfTxV2LBJG?55_t*$t8;{jP$jN2L{VcB4_<$w
zu6|O1DDri9l37LlTdmc@WTsew@=1?}E+~}(Po~EW&rKXPty0HZgG+Z4&iOj0dA<<F
z$zFo!XETu=er*YkFm5;XoX3>gWJAudET#Nu{s&onMV4qDv?1f~=4LH41VMt(%x?7~
z)f}MK;L314%~R4nHBIh_<Lxwlq41fvvI&0-N!*CPt7&XImXlJ%xZ8w}2=o3U#vNYx
z3NW9!j<qRyBrFLJ{}F93d?UPnfdw~$DaRnFOwl~?aX$yZb%=XW(%cXhH&1Ul|GREc
zZrJi~+6JUc;WTUTCd<hGQ?0c>Q}90&$V$uqI>B=PbMAhwvWFast09is;QJ4#K!+Rg
z-x+)V2NW2`vE-0+!^*!#YB2bCjUyT@7kK3_a4C@l*Fu1uB1BU$%^Ejn@)ize$9VWN
zrH?BxK=L0b0grsY5^nYlf3A$~4+(mjwRtF!an?RJV7lWrdPn;soh7c;wiC5yLq4`!
z;MbLYUEgci!tdzue=6vv(*19c_QzW90)1OkIF+z}@Q}ZeN{c$zDJ}l<AZ`a!iG8g2
z?eL#(cG8CK-bBVkKN>BqcwPXjROCW`L&6&&yBvTk10Z^^qx_xW-@(^nbq-V`fr(Gc
ziwWx*^en+WZkC?HUHd0M&Z&o?rJRDXD`20QFT){3JEfH&)GLQyg9%N`FEul-#j*WP
z_)mUZjwqzVv&Iocec1A0yZFu=kApXIQ&CT=jn~Y*{wLxHV$|LgENzf`lr1HSX*Tn^
z0SKx8fcZ?!f>`6)zn)o}y0WQR)crrrUXJ_kY2_j32iSf2)+Q7`msdvkhEi(~rkxPz
zG*5l(nAD2U=YIbm5f01_;;cXEOrUrkYHfyS;S0qyToE$8={0T)Y$jJnUV~^5`9ptW
zvztM_<GR~#b`)?%A6g@j8<Hh+gXQZ5fYi{+mQaZGdk)Yw{%^+^M68r+ZAOT<mD>W*
zK*1ObKaQ;-^R+cZPy0hl`Ts_I!0R(R`TyP;#(ZmwiJnsabxx`g(^}3TW0y7!^{xQa
z&c8;?c0Bj&IjZ`btdB_+Yx_qWV$q#L{C@2e0T4F!(k|%*He-AgJQDHE8VrKs(KOYI
za(rWSD3DhZsa#EsbKrpjX4h-<O07+MjpJ=cO+`#2?H6=twRQPpgt%=X69bfJU|)(Z
z8b>>}{OEczTCQs_zi4m^UI^7VMAl&VZDVxOoFBn!Nyn9&;Ds2Zk1zVq@DkYS_G-g5
zJMVuVS$p+<VK$jqB3{0syHJ+z`%CwuYk2fkqV{cgMv0iu|D0wa8m7?i|8J9NM++VO
zg5p0(&-@e(xC6&V>!-;$5RKMo>@VKSw*UX@Bk$`BZ25^As%1`>_#f%rHwc2v*~=3s
z1sLsg21cgvjy162E%on)g3f_zD38Fy*j8uf-%uS*ggfzLDsJGR!#AB$npUchz%_If
z@Z*XZ6WUjXv9gU%-R7@4j<wMp9k#WOV%N<dBo>hvHseQ?^-cn(ygDI;Yr(ysh-OAr
z`m9RAQMtUx+JS$w@6!qj@Bn@$$Qm!Ldd7X`yVd&F6{OVjw2I|)>A*O6mia4!E8g(m
zB)w9AF2Ign7f@4%mrEw<<~XO>)<g%JZo2IIn0n)Sr_wT!)8#!F&KPJ(p`ak)rod0`
z#-XHCxD$!N8*(P9yJ6&TyqH9!sjGgBV>JR;N&$?d3Rd_=E-g_hN`TaJVo;oH?f<qm
z!DNXZM?b>IG?|8xeU9&K^@i|Df_<3Gu!tOV>O*_TylsULsATY`ANV6}q<$Fq+?-}5
z_EVFa=s0r)KaRqxB`#%sQ1u6zDDOA(z>mo1t})AiA8S%U>QSH&ECV~^Z{ry|qYLuB
z>vYJtovs~+M0v^b2Z1g#FmaOjQtSh|47Ezcix3N`6)|#p{l89F0@~z&Ef?LnkDzB5
z*fJp2Ij{u=4@|0fYZI`=jCpn2tBG!hAu8hB*9z<baum305059-%b$QLGzgR#26~JD
z9l$g(IOaY%F|mN;{q&?agdSfap7n|bJ$}|y75YQPg?}!1VbG(;|0N(6wG4Jtfx7=%
z1dU;g{a}Z^YhZ}~7L#W#gPkX_yY;B!z!&o2Bu_wkB^cV++3}m*R(Mrew#9`H`L^Tj
z6u8+ww_W*H+2SEu*xBMq?Y}{F^__ZspJKK<t+xYuwB6mW*pPw`wmaOtf&F3S{;F~}
zDW=Wqk0@r)!vZ^$Gh{rWoq3wFsf<&yv|Agwn^9^9+KW52&+Iy_GS4z{71!A9LH6id
zx@arfeYgOAS>)2qR5ZI>A)6D7NfAKNk07v~_1pL!0$cDgA15~~G$-p}8&0KG3L$!T
zS2F>6hZT&^1=5Dzg>YXEZ9WkPy;HsnG>#coum<^Chu*PC!4}{LgSMdZ%_i05#_8Ad
zEU`>C*1#yXkrO}8)7zmHenNx(+f;EQpVZ?zz$i503UlD6$y~l;{ki23>$k-k{EV6v
zrr_H#+?ijY2;Z%?^J;qo*YVbE-x3q><!EQk7xn0I-4_YezpA$uzrDA%Av=(I-6pun
z`bTp4VGXE6X6t2`TF>bE8NFp2%kZXe-KHb{j%fsK@Y>MKOl_9=md18omy_}58b~Em
zZz<e}U|Jrz`OJ06fE}h5Dlwd;MGmw*aN1d~rZz6bT?9>HN^0R5_(mdd(5ab17VITZ
z7B$H_$h%xKvPSJ7l%&vIRKN=VZ1QGhSjBd#oH~Pb(7KV)7b6+f2d){}yoNFqp|>l*
z6qWPYf<i=c+)`m`6e#sug88fW-C&vpn6Qa;XAYMhSuu&OzE=aghxh8p(6v1PbwEA<
zUzS|i!Z&3jZ}H?tAyucDE<F*cYUtOX_#Y5)N)Nb}#=8(h?L_W<(3aR|g((~%jV+cZ
zT5uOI3>z6BV5)uy<D{gSFWuIXem|lM(h<+MONheCncCrF#DK*>MoP7Xwbl{_Og*w9
z$C?}gOjB=}YU-|c(#!|h0JbRTX-kn&Qz_43(c6@YZ_a1U-L-VN*X7F+T_Rq5k!ZAk
zS%F9@EW#GNh2_!--QnZ>#u4#9s(I~m8%Ia2Lvb%E_IXvkrl7+VH*X3uJPLM7tE$U7
zvr7-paoE4CcE%MjwNf&$t*Bkdtxj@^VOoX$Z|EgPB^Z#>3UtfFpwf1Zjn8Ol93VXh
zrYChaDZI%LK%rQ1f=vA~`$&%50NNZ~zx%#LZ(mXm{yo8)Cwa+aKSA_!gork(0ivIV
z1i}r8@h$48<AB*o9aR^@C7h2sWA2(A5nB{49k^iT#k4MaTiLWSva*>Kv}qRq3sB$8
zJHq|Wv-4ZG3I0b4zilJ@Wu~=yLIv2*W6}x#joXHGX96*VmBXyR%`FMZvJb@SXGkr2
zIh0-(SWS8lR|wi)uEm=Y*O2@$g&eO33x*<CT&D%8)xr^=h0iQ+zj+&8>Frv;HOxRN
z)+}BAJ8};rC!Y2E#aOeiwHZUd&qY1udOkf%#dQGQf;`~$@vTL;14GS0yz`B=P=*b?
z^KCJuSuVu`i@n#l5MY{Mjo3Uleg&QcqDI|f3`gb9nj_84+}gWUO0V2dob(3yI`|gw
zB5)joTGcj$t6M3vUA&1EUEjE)wS_jec|(-kjxZ7L=c~I~+pn6)_544jY`HZ`+0o|i
z)~Gkkob4dC!`o!BJ*^$BQ8%C72;9;gtsQ*JX1+ys1>XxlWb{h>Q^T<DW3zj~krOcE
zK`Fc~uctY-cdqQ}^68g9!3`;mlU-i@)SI2WpNqO^c`Gd!X!&V+xv#a0^xdu9q;DfV
zr#Ci5V;pE6yrSzKqo+geiQaHW@uxOz=uO<)!79B~f0j4)CmPRozjpHeUT>FOQuu51
zZHzB1f%Wg9H(I}CyHokcT6^Gl4STy;dwV{sTwGSsqrK~UgfL@|x9{~VecHRS&t_yV
z@7mwm&$n{nqIckjQg0+4;6_ysHNcc}9_vQ$-*)vj|LKJPad^*e$LX(jg#NN4lF3AJ
z<-p3pl|#((ZS?qH>kxKZ;aXW4^A5JgP+x`X*E;AOVy_fh2cckyk#Z<ZITWQ#B^iUT
zJhjHA*yAf>eBo_;Aws-0D?lFdwZ6?SC3SAMFWtGaZ^QQ_8uxX5kFAENhwafntT!_<
zNATFPt2M?g=X&qaztVe`zK!=D<GnqM@XzqJ<GgLuzVddX1rPJRNAx=-F764$Qi+w@
zR&Kw^7yd-|n`Za5ZkMAmcGKb3VZQ24-jd}lztKFpdaMaAc=h;IP$Vv9O@up-lYsnx
z-S%X8mv=XwEVDxVLr<3XoW^~$|EoIY|DA$gSMZAp_(h$NFeuqoO4yIMTO$15Rbo+E
zZjn_l9Ibz;I{!++_p73)$~U(Vzz}>!2?Dr5&)KE<aqbl{o%u~&s^U>oPWDh$LxRCh
zuK&DZA5iduf^mW%_nZ(UScTWpYf!aOA=+|5#8pJdr47BBp!o8($n|FV{h?vM$K{?T
zH}C+!ZQRXox8NUATcRo}!w!p*d^fVo=gJ&W#@Bl><7vopx_)Hb<HCW-Du2d6kMz;E
z(ZcluQAgcvnPxUHGc|)U*py$Zm8X~RZHzhVwd^fB5p3^9D55He81b-G+1j1Ec4C6T
zsOOVY)2?{+dh+PWDJSd^WIwvx-yWJL>x(#TYeHbG8xI=_4XdNAMP+07Xc%fkrjM#6
z(T_5g2ionu7h5A`RfeLn62OFouWxFXTf?bru$GKzU^I!`OZrze-F<8MWpC#q)fZ8=
z2Aej5&uga{>Lh~iBq)ntBeD@GlDgR2_(VP7d`b#^UgsUvnfhAyKcd|4RA8v0a6{WT
z&aOjLe2{ypcG>P0_k~ya@`1<gC<|48dcvWbh9(+o#aI}^6&;t=^y50%r*Th<`b6WD
ziZyE5_x_J6P0V}qYV2>-X#QQL{g?uiMUN_1JUw6dr2k9ys91-3ysWfo)%XQsV@Dje
zdbbkJDwt6qLdb1A9I8R~X!p%3O&ds(Wt;Y3U=a_vk?`7lB*O_+(z@ww1VgUm%+%CU
zL!Q5x^M-$fp51acxMQUkKTqa)+pf}f`s;t6eg`>asBg~OgW=WxRi759n}%!K!IE8u
zW|N}pZz}V5RFx!C5W!zkyFaqhxAP_FU0(Y^L7r<d-Oeya9T>k&20hXGXB(PYV$j<z
zRaKgCroNz#{FDOiBjYsVO~1p%Z-jpS<p9@+0bAC#qR&z5iy{X8kp7l3Fr~yE4*&Ik
zqdzUW(OJvLUZxN~YGDFRZ4nEpL;?ntO5nv6rk1RVJML~}GgF{FT_9mC6Gy6lW9N+m
zcE--?W&}Nz3@-MSlZY4J+IeG;UWgQt8(q~^I#lX63+ci5&aY?kdtj2CG$KW1)Y_o+
zKz?P`rszxDtyyZdq&FsH$8KH`HYjCw_O8x>_D`0W_K(rK-yi}6#=yoo%vj@hcYT^v
z%v1hJs1a05X0s?rhIiMu5>_|o>Pq@5nwe6a6xGd3b0G9gl+YFEvlyj;4d7QOiZ^i~
ziF6Yx#His^_>3})J?qVV{<3*6>c0?cRK<mEN=*@Wj;J-G2qPYFa}WwRGdraFDsc9n
z^YWjSio9J`$(1bb9yw@au|m)mSl-WlvGO+*Djaioo8KV=!zlPVKft?}cefpSIwwPG
z%QMs$@L@5h1-pe{x%^zs-8j9pSbo|>bx%KAZiK$-e8_4%?Buy^!in*MlvpDM?=iT`
zx62_jkpl*Kq{;&!xqi7mc^PAfdA(<XsdGjSKd#snr9G$E83oTPu%lMGX=dNnaeM67
zRd_*xr`(!?Aq8wUj)$lNKtChLA5+Oif=a=SMimV@|GW~iGLg{JliH`Y6)mu>D9q_D
zfJ;LpvvZ5AO{IIvP}~YPp*%~6CTNLbQX3=nO&lGvWj4A<1EAw*tK08#myW#p-)82_
z^5E>y-0~J%rZL(~tfKzn^y48iFu=dcy*;ul!%-F1G_yG_E!6TwLmmVZ^%b5lqL*sR
zu-fans?eUdai0XJH2KC*krCh=xE!vMg!Q=z<<fBHtH6}I4ON2UXaYZjhqtLDWbj&}
zmS;4WO=XOb!ObHJ^$RCGgQ=48Uh@B=%C8&kJE`=XMq41&@!<zCK7rS<fz#+f>ln9O
z+)3F}XHK8Cew{kyH)t4~P1^tRJT3XA`;5$~Q>XbimY`5mZx@q#EPaY(6mIxKy3jyF
zr${QDp?KJCn8xGjv7S?>d7ZLU&$+fHMG0Kc*yggF@potp-a(6}&zw1R#(B&=eP--7
zx3n*)3l|B3Y!wHklNfZq-}<U_F=HHha7mAUTfw~2URCUS6d3IPv|`4@`f0_!q~K>2
z{GST$QSc)QeqX_-6x^raxPm`Y@Lv`DiGsrlE-Uz&UZiu?->feXhuNV7hI1HdArNIA
zuzk&$zpTg3WPVI_ZU1}+8Ir;@ew3tK46u=o09o|6^ryM|6G%t_O#R$W{#Gsz8*~#+
zt$T1+oz0aB{h2#ew*N-vPQYDHrT`}_iF4~xZXlN`loF3v-Y(Z(@vY3xsO91}Gxw~C
zpKO;H00_yU>1Zc6_Fdd7vvV-+tH+acw(EyAX+MqQ{D~Z6choK_NqM=b!KbXHU|VxR
z3yz@E=1^JYNO}Jys&%n!g~z4%)!*=5NL+2kud<8SnF(>eapHsPJ1Mc5-(PgjwYzOs
zCjlMX1Ulrt4%em7G<LK~N?mQ!Fn9zE?V$GX7AFzhOYC~VsS1e3!R-@!Alvc3>wk>z
z$5-OI_NpR^oPR<5*F<QXDbN~X96I2_J2#o;yVsln%cr2>&9)+Jf!MOpLjQ<1y4JZY
zeuWiama(M`p<eJYimLf-hp?wR##~)qPVzL+J}w2l7$CSA=t}j%QQ=raRUiL0G7GaD
z8E@>9&yH3cE=_WCXwZx3L8b9q)>Yamtb#Q(?$Dtt4;tj>OzJH&Lu(&-u<pm`yhFDL
ze&lBFSlfox=6xG#%OvC#$8l}U7=Cb*-rl?=YUPl2cX}f8k=3xRPr0@_U(xCRhI$3+
zeUY!_J}Wb)jJ{KRRBK(XwYh>4S@douz#XzuD)hC~P_z9BrJ4--SuL=HODHQ<E6&$C
zG_C|?l?v5v>*S0w4W9f%wf3AIuPWc&5$!Yz4VT}k7TScl|C=f<ygx`x-qfJ6rseiH
zmCRAzj>7MYyPn!w_^Kdx3iayLaN&=XMy$W^Cz9_LS(2%q!k-DLCYCK+H#u=C^A^{=
zM4KR5sG<EMqEWy=cHI&0PWB0P*^b^@o|if(UQ$oBrgO1ke0$a%9z1sav`-|-1fvs4
z$A*{utWyZ%G<v1L$)wXLR!P;;GLH2I?bOQSt{)A(O4g;~BEC1@>kWDR-oPtzJfE|(
zOQPW6AOyY7FB*fQ{oVLHdiL~Kg3BrvE{4nqS^M`n?aeu^BA$~kfk~#sest35qjc(t
zI&KKRkiQBxQ3`FhwDif1wU+QpTB>YFHmM=lI{D<nEmlPk5h|*&V%|p!Ze*`$;y11J
zJZil%@09Xm>wkzK6wL(vogO=E%x@Q2<01QFYP|czdQB6f6ANLC(J}bQ4E=|!AD#BO
z)rH&X+cgf4sqMc*pZ(7eTyu_F&6lUH)ryAGovw#ecDKqJk}R!-n>zQb@P(hIgZ>Xw
zb~!YE-_Wg_4$-Nu6IJ|+9KY&l<6jZ9(v7}wq9p5Kx{&vz;B<1sd3~^*g9umBAj?mh
zJjia3U&&);AY>FyG2Gykxkc##tVueE*+8oY9Z;w-&d|6_lVlht!>6lp-XNzVk_Jf|
z=-tQ@Gp6%h<mArs+v{aNi*W#YMH7_D<AqvAZG9*+lq4uN%}*`|nBvwwr5Ux90|84o
z_UJEaP<Lu|oKWl}L6AE0-pAjIN@Dlv!gZpy)(oHe5JTctj7G@G4ei9bORX4I{8F^!
z?CdgD(OCZ)PRp;T;;*T=`rvNj(`?_8nGvKfAw0eqB$l~Kbj=P6@~Fa@OtK+Xh1Sk3
z{A~Pft=%&3O<_e+g1+y+AwTVM;vm6dKO#q6fCI?_wI|;r8<viwi)L@S2E~2TSWH5i
zyVS~VCqetf#ECDm>RY{Q#EB^6n>evw;zWdt&Q=5<tHotYD<_rbjyMsq6ru#j25$=#
zqh}{j9D+G7f#NWKB~X--Zic^``3nKPirSgGqXhBds3;%NKnS8y`jRtpN+J?Od0z|d
z=MYOnrh7ezAk90#YQL`4MmcN~trZbLK1u$72@pR4sji+#T&TZ&Z4u-^m_m-%k62{@
z(AjC>khL&~2=dy^TkwY7wtfzA;U<KOUx@nD#1fI`pJVgz7AV01Z<w!=7;SGe&p2<v
z$}ti)Zf%b6euR?<8}CCvC=uakC~W*~d)n-6lCY85ndX+})>X=wxN!t=<7U2uXT*yL
z77;h@WYcb=%&r!~N5qS<*|LO<t?k|@B@sQY?xt4=5wALN<1bTouZbqNH21Z(cw0CI
zb`sm^ZME3`*3Q-zH{0Ge=+~V*Bh=$-ByM~W#dL1M2*r&D(3HY<)WnTDP29L^rMyA>
z5^p@(Eq<Xlp;(cYchR!MiZ9a3gRL^(u%{)VU77SAZ?}p2zE0jD-c$tr+iUA~-)k{O
z@5N8V(ahRoBlUU{@ziV5)Xqy3>A%vv%{^m4vPmTx-(zrXFGttD*WwQ#20cGV?Y&m}
z@3w1ioxR;enQy4HYoSxVUCO=VpNCpt`@H?H<PZ$E->A8TZs@$nde(Vn6b~?pdkx~Q
z8AVhN_jw0RRJVWSfX&-}M&ux3!vXIQ<1uziq2zz>8e<715AzlY2A#+ZF&Aq?;<BUq
zHIdovCNev;GPYsP5{>O$?Zsv#sO0V*R>W<s+vv?v=0`rBj@glJr};;9@N%R&VVA#)
z$4ZKG(O(`>sKc@P&Nb&}<9hr^8$+R*m>4lCfn;!83JRjUVgfeDEZ`<;L}$L~v>9&D
zK@dt*{dpa-s-A0WwQV>?J3P%b3T~0%tE6#$Ss$@mL0kIf@6qF41;4B!`xM)+z}Oq#
zTEgZZP~n3LKBNXDMsv)LHVWM}4iq#@O)NFL1R&Cc57#5BJk*454+CJ;G~uP*axLLh
zT@285#^|TX`nqrw(~`UE7J%mFBZPZYyHV7?T_15+!BGXr6dYG@LP1-|<e%hm?AW&<
z-U(8(3)s<RRegdU`aem)-E>Ryi<j5<V7Zm(rz7xjypS)eiH3szzXd(Uc`<3vxSIcA
z1;(Lii+9FuUxQ`bM4sc{rH?dG&OM6VtKdEb62|!VD|m;3V_F}QwD=F`@eRfPmeSS>
zS(IB-AVk-Mx&DKSwdF1TLwbBzfvvoED)xwi57S;Hvou+4&M*3pD&;W+U(~xES1hF$
zzDu!F3Tz$CC}unav7kP$4fhjNC}F!LNk<|+azj1<CH=_Z`A?+VS{cb<VLSG(BOBR?
zVsCUM>$jGTND7icwo!;5Fx-i{yzda9XoB@}k@eCJyuXC^L%hwwa19z#{{Ppo5~uXp
z!n01VVI&rAN@a==Si?FroYmNMw<J10K#h-UB!_uHE(R8u%iv=;v&pf?hTrm4Vv@|R
zA8Oc7qniY6-xinhz{OyQIdW8FXs>RFyAtzqn5BsX6#<IR7P}R#$Q@@zZ_w-W`i%rF
z7zx_L`92hc&W29Sz@}fS*V;7b@+S-B=;zV##!RCO6^<Y1si~R|3R|9X6~=Mk*;tx9
zYhGB|^6j#U))t#*PnvkR)6$s*xssN{^r!&|RD_??3~pq?R%_;lPmZfk<4MJ;UrQ_>
zF{<sT`rt&J?%5FM>crf_<hjPt<NR|B^{}`^oa?Kst#SN#$bXlN|8)C`EXeCd7wIOP
zJH@Zj*Dn)kWmge$*^OzaiV#Q8bwzih<%Pq*1j|5qXG6-{QQ44cRTMyt*NC64=%zGN
z%R_Y&@fljUS5V-350VE{@3d$CA#jC)$gZnMutlYZyh(Sn*rRl2c?{1u&Ie93gk{}~
zRUR)-V=u=2a%XFk=N37Oo7^=1F>U?{s&&Zxpf{>#L+>vnIH&V+-HrF``36^Wi4@Z1
zVFDp0_r+>@ZJHb<_NvWu1d*z*n(-G2DudQm`z31Bmq5uEP+!9vSWy#?D)^X!4uBJ1
zb2D5H1DVAhZ1~}8M*p^Q?!a8S5m8vD)-UHa)8N#kW-<CvhlPqf#&1J{3*5BBkTY5d
zCvaX8#om+?T|!R`L7qz~3zgd_RerR<RX+^tF2Uac=~<)9i(0RLTvsrmcAL2h+Xq|_
zvEzfW^SjI=8?)?RfN(|(O1l^A20lo0yFz6v#JLE2DQhbC!xCM#@`zMbzQEt?8%~rd
zZ!@|uDtG=7TGh(&SULQK!^7L4a_~4H^)sjwYVtR4N@nuQ0tI(|TPxXknVcfuCBG_(
zI;>iRtnFudg6{#|Uu;3kaVKA?)pJF$60xGi`rg1h6k=7&@%k|Vzef21`N(<0E}C&A
zE$1JTgD-(7r@fMgP8(j}f1X+J|AB&kNU$uI6gUXb3Bh(OA6(vc&>YfhB`Fc3d=8qK
zX`GJEmAUcd(B#?iIv#-=V;nL#T#`D3d_i$L)tEQq*>xB!K(9Qn`?uAC$t!CJQFYX~
zw;es@|6}#wXB7Obf}d0H^9ooWvE|Y;GuU8K`6-Tw`^(D%&(~|07V#s1Co7kPKD03{
z{;BFrD3Csg|IZW%`St&~0^2Lv8veh~<9T910ofo<QRilGQv5HK`Vj@=1h7f=D=wyy
z+a6l7{x2)A)(su}*E|OKdhKH9rt+th{cjZfnu1?fU^}X_t+ZQ@YFoXNh*S{V#We!L
zx~Mks@^Dhup-4#rKJAS{BJpxRDi7Gvy)0M2g&X@!b-}o}H==88Hg3O}I=<gKcl8PU
zp7Gm|w@YDKW2e*hDSjK$%DJ_|^oqRU${WU74o5mtl|~5|Z$t11@Loqu%J9SziIpvw
zQKwrY(80=y)TpCp$MqON!P0R#hNK$J8`i5)=ip&{h;*q@!xi9dQaa9uJPdqCv3HkI
zfmXXP`|U5v(`(eI!yK;{9!3dzzSDwu7)7s-7Cv+H7QFtjMOwZ|{T#av#bFet!#<%v
zt>@1JN&2OZOiu^ws|HM<^S|*vco<8u#xGG~2+o0>8>EVdH27*_%DhWbYmiz}vp&`w
zhV!u%W!#AKv8^c+Kg_g^^Rc}pZpCIe9|(FHe`rr5-cZQ-7;KI<x2$e$Z8y%xFr1G~
zd_Q&IhQwx3o?hJn=VLo%cDA-LwmV4OCB6rW(<7AJiAr|adWveZ_}J%vld`*wOR}}O
zr?u7F48LO+v0W%oE4H@<55>(m%G0FqJY?$7;(W|Q^KOHh-eNR6oD=v6n{ZRH{j)d?
z2vx3k7!RbpvU>wP9dA6>O;0m|5wBy&+sTNDS2Rz5_qTTQHG5lodAE2Yy(l*8{fz(W
z0p28)XWlMFWpOfG58xFHc)RJ{*sXhq!t()dkMR~zN4AkU$eYGmW4viE@7cGJx6m4T
zE@lejKh@QHQpfgp=>A`2t2&-7kx(oobywp6I90vB2vv=v=m=|YZezW_5S<P6+u<MI
zjOM-x2|IT1+n`kc64Yo1efbfMhn@6)6$<r7)e7g`1}#Jpr>Hg<0JvL)-k$vYGZnd8
zkmWB)i2kG+=%z-cunYG0m!Lv(P@yI_H+qydDp^vdu-;iuldgwM^x_#ortf7)oT}bV
z+~-ImNMz>Qk)OkfJ&>$UJ@4&EP+h_fa~^fMf{0urU`2$cjQn)DBGd}sC}a>!{duY|
z_-Cm?<ZO5&l<Tlpuz;&;28^m$r@PVyd)>M##SJ8<8u&xag>YZDz67q<P@r#5e9qIr
zfJU%PEz=OmCX$H+I#hp-Hpq?s&6!ZMBM~3|rmxbI6zuwG6QUB*bY!9bUv1~|8`o9G
zamO<qk3IIx*iGG}gfJCVV<R#pQHu~l=^{`Osj-S$BxE9?vB&Wv<A?6dB+bxLERYb7
z6bT_D5D$q3QdayA>{w8VF1i^00+#Hfn$P!p?#mwAX{Z=$uJ4@tIOm@GJLh|T@4$$*
zr>*gnHk3r=Ye5S-<1$nNc`c~RSYj$^c}@dbO*Ei40_&J@pZ_q^C#GhNaMUc&qW+F}
z%stGz2pF2!M#si>3tHsKE)Vi@(V{p|1+pf&(c3J{IFuUUcS8cCwdQ)0m0<UdZW~|P
zVsY3b=zHVtW~*VMy)5I55V-Sy=Up4jTC?;nk=v+`2(jjhqk%eH<30JPvx|JqV_sz^
zQ2>Gkg*X;+F~?EDRVsEm7Awz6o}KA$3pvKFp088P9cj2clbARAK5fWxU;9*ggy;H;
z2#D|4a~Kj~M-ppgkG`>#!CE>+<v9g0)8?^g0zL(<2m?)cW<3L?*xl(*Jl0_qCkjnG
zs^M-LdP+s;bc9dwUDx4Glth<xl~kTvggQ|80@R)8piposu)>+>HSjM{kIpD+C--wc
z1`I(7iS-Uy8MdnAfVjogwXHg9o*9Aykquo6Ri)`7sv6f2ia={4qNjnZWKke<0{g5z
zUgWxGkl#b}oQv9(_k|L*J@tEKP=6}ghtnetw$!G#Nsp#qR+v_n247B7Z<|I#I?HaF
zxvv{yPeH+wzGQU&))D3Ezdt@WI0!=Q2YQ9J7rD#3U7QcLpGS18W<E>+D>~hv^Ri|d
zGcK97)!_@8I>m(2)80Ji4J)=i-P`S;&<wd5YyoGJdYPA;qLU3?G`~@8#{f-qk5*%7
zawqyWJIlqD)s1^mo84$%m1m5Lgl~TpZK}ytV){w%S}kZ27keUj2>W$2D9Wy@3{Z)L
zV1#~C9qb`ty@q2tB(L$K=j4Hq8bphLXWXIs3M3;$XaKQ@U|=A~_eLBvOhlc6kmc<-
zi@gF1t0Xg?#(t(7CqS=X(Az5YoPN?9!3oKOx2f*4B-!%gBYmS={9S!BM~mlqen9L1
z$8ua-O9z>pbfgb*_%$BgS7Y4H4tozrC89cRWHQm%nDP@yMD>h+O4=5K5~hizNu;9Z
z=g&kseliTb&I9FPs9iC%C>BEoK>?P~O^-m3B%_BRXl9t0##>%_W4pDU6M|j|FqEK&
z0!wf2?(DAiJFQs4*&q`=)4Afh`P_%#+cus`a`!oTa@UZGC-;K9d~wB(Cw4?jOpFKz
z<@E+NvBx9^<rJ1Ilna(!aRz4hzVt}Lip8zLI&fGEKakRq{4SgEmr28lOjgo6%7HwY
z5PA~GsObnHse+I6viOb>qk|zd_<ZaeeN2Cy1Ol^Mo!Y%9avl@zrv!lj31%)A4GR-F
zSa_3wnFfUv0={t!1Snk)HZX`*na0VfR*{(k9tJbO0;zK1=^x>`E&p_!ggb<Sp8T_H
z!KNM*C;=++B?uRYut<y=0K^92HT`c-KTDSO_6)}0f{nQtHeTZTr)8@H^Ph)}8nA)d
zJ^CF$Q9lG5b;Cw|Y4FOEz(yj8Uyo742=OaX&BV&)YM<|G*qGj=AbL@yGo*Z9S1xa*
zS?oF*FbkLluWdLHCv&v5`ud}X|M~Fd${Vl7+D=Z0j5%i?gE=i#wW5tN#~||~%_g2m
zQjm$>Rhfp$u-GC`p|)7JXBab^V~IY{?LJZytgSZJdMd_5KhWI;GQ-Ju@K8mvOFbun
z&c8GszDji&=tvYDlwUz5hI9=05$8~Pba6r<CoUWYp(MTgabxIsAex{;4MMLC_t;6+
zh8YM!%1j<k`p0TY*+7s~lgyj+tN;^4%Pa}e7PoMA$(KO1#QjIcjN6Ne7Wd+FB3Qf^
zfG$bD1YPAgIMPKr`K{<D>b+mn4!`{2Q|`jV+Sw51H~M3p=GXdD?fr}mCmjc24|M$J
z+WbPBecd`kd)@JoI=)Kz9M)8J5}{jhU`T><9oLyTL?bz$ZY(g1eyEM_?u{$6=5wC|
z|4ix;I#r7$e8Nf%kV(f*MkZ;b^T;$=m@KpdGZZsOvAHOEI^`v9SB+eWxoA0*x~$Tk
zHcCxePd7#1*M=7lY4k@}&%rIDPlHd&>j<qdTf0oqKoaRO=QkF_;K-9{7m+y}6DM?c
zx7QloJpc?Ly3IElt^T^w6&TY(7CrkeEv<?_Nj{>-bYUb{qL?0ub<D!(%sZt#pCbJO
zfL8Eoe6fd-qjd((L*D(Irf+DMwDGn`wPq~U4aBfFxOf8zGHu)(NcP_HgV|~}uz<-5
zN`R#tTiuQR?c1F-;>#iG4@vxNIcWL2SHITfeur0@(4&#?mm-YsYNJD^b4FA2LA=5W
zLFlyVv<Kg#!KKq*R|=F;?WZ60rxt$K`DQk((Ji&)4-9}D#hO_42+z=nV3+#xW2$^o
z`_!*f<Nph%<0N%r;l2qbVsDCCtx7!9s#1KhCzcN@>gdg${Qo9<7z}oA80U446k$@2
zO*#UdqS%PdTU8|2*b0`!3)rlp%wiG@cdK%@9=MG8Ar|;O2+qkFP;pgREmRPTld1t`
z?n;jssYd0pP??sX=!)J?m^xzb;W7Ab4OQ7uHt&xovw2^661?u)92_#ge7l#2Y?;yy
zURY_YHWv1-G&UCbca^-EAsCN}__?~+*r05DIKFp<@72Hweoqu+Xny(=L$e4m@^G+O
z%y0EGgKG~a4-vz$XZTOg_XSIr6qrwh;0qdgBM>Zg6dayLpqvH~=pB@SUAY@5qn)@v
zY;rlvd|;erC$<$CnNXX@!?1>EI_-UtAEPj=6?~o-%yECk7(%aWFza=)?lR+Nz;utl
zr&qPF&VE4~k8>TprbpHSU=DX+KyUFeqdf<F#V~#2ZYS(R|8MutyK2KdZN8-{%}RJq
zd?Ie{eXcF3X=iDgxjCFPs|V$Ztz4c5HFZYRr=G=ELruE>uuM46)s5!PPH<$rqm%CI
z!0g_AA3_OAu5ZQ3hKxJQ=fe5+!fj%0m^h)@hCrDbH+RSq8w448KOs3EqP))LN+jHh
z)OR?_5(A>QyW8nTG<q!hn>ON1mS<hT-AyTtoBiG<XUP9)cxj|PJ_;7Cb+Ic^w%8T2
zO1AA*NKUZrwe^-*tEi%@%boVN&yS{6lYvJo7zZ&3)@|aABJt^V0b8!=3RKbk4)@1)
zB~r%@FXc<qTBYC4+Mr({`_DX=EYHW6pbu}h`n}ya8B&O0a85UU{_ss>tD2;4^+9xA
z?Rrj|&*+}#bN5WmaGH)4a6!jIW{M^<SHLFvtgib|6t<wfm)HzzuP3kdrk^qTk`Bo+
z(7rgfhO#8uasV&60tI<AWE+i(VfkyJvDdgfWP~V=mCMtNb0M)51=<gag6T60%mpfw
o^AU@4I}?{aM(|%cTO4~&fAy{KKWyi2*1temg>S8X_3SJE1y~@e^8f$<

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_ccsr_stage2.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_ccsr_stage2.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..bddc9fb2c6f9d97ed0733a6487c53622239207b0
GIT binary patch
literal 58092
zcmeFad3<HbS>Jv4P5aVn^?rMA&GbI|Fso;#$Kx4mhMBgF<r!Z~I#+6`r7QKhSG~wb
zVNAE}w!;{+2Qvd|vxGJnFneqRf$#<hAtr1g2Lc2VoFS0FXLt$e*zWiHd+OXxQrjeX
z-_QF`u5|0xsZ&*_PA$(~Pc3wH6=HlI`L2(a|L*JJu|L&C>nF{{BYdm>G9HV0v9XvJ
z_Y!mQvAF#v=96Q|xJyfor7bNpmgP4&mz&Rz<?UW-t}x#*)-m5X)@j%2xvu%{vF`bv
zu^zk5%=OOqjrB$K^pEvZK07xsKQK06X}P(<`AuV+ES{emn%_LO+2Vz{;rT6NTP)r&
zw{?Eo*tYrYW83F<jP01;Ikt0t*VwN4;#hHh_t@_FJ!5+;U+3K3`F&&iEZ#M@fBwMO
zf%$`D2kp9h?$G?m*vR~CW4GCL&)n_vhsO@j9~nC`e{}4qrS;Aon?F8w+~R$6C+1I%
zowRuW+#U0Gj@?Ooz#H^7c|)%z$L{hDd&AzAEAg?ry{+E1E3vVAo{V|hy&YF#-i~s1
z_TKWz%c-&ZNZaY{B5haszRSt6`?)T9ySd)2JfwUTDSN!Vr0gx<&-DXb@ALL^z2B}M
z<obYjkn4lx2f6<c*N40ju1D<pVXklUZs+=TyMDw=ybv2XGQP-=7B0-pdBypei?t=c
zTwItcoT=3otM?x}Ha$~2vovvZa$)}1+|uNX=a0`+s>i%?Wnrc|Q<*;E%}h-#Rc96|
zM;0&D7W~OG$0p_$CXU^6&+)q_ys62%r|vp=&(vM_dUxD0b?==M_e|V%@?`n$6Xm<h
zcbvHU&WSskHI*7QmFhJ$p0fVl(Zx#}HF06FTp6D^HZ#BIFPtxX&DObAmG8Ur&U^1W
zdG8$)6ZhV8_xR+AJMX#2yUX)v?R^sy6W&ShzWa`!oGdpgXg>JZ^!QS>Ix}8rCf%Y9
zk1fnEzI~>8tTsNc4yyU%mFm=jKVSB%g(F9b{7%jkUf1Fp9f^mzXU69zyzytAA4!CH
zzdY`feYzZG{PJRXyhdX8<U*xZzF3<duZ&NZ{jj65G&dKeQ9ipk?$>6<=ZL0i=e&7-
z^Yi0p%cW_5#^a)!`p%Y?%l<-Xp)z-gl)^JJ)!JhVm8qHOkzCk4w=i9*U96SHtEHLw
z=`d5i$n9#F^~zJ@OLMg_JyV?@UktMh_dE@t_A;F?H8)-hJ0};GDzy?V7@x0(o0%B3
z@d}Npl}l6p!hDHV>LKwk^UO@`iTQE8VuhX<pDUf63%5T$Gd{gg8J~MH>e<J6us<`g
zRAV-gwf`x`zcy35bZW-0)?TRT;a!g|)fUQ?$px?MKmE)|GVHf^ohaAFOV!CUWp8P&
z91hT1{`h3AG*hW9lxpQlb-@n{ny}^4+|1cB4eEIO@u%MU!ub5+T-oQc_bI*FGYivn
z@LOjt&QwV2sV<IBmP@sn`Es>ZUaW?>Nxje{@7+PnU#M2cYqd%kzeF=K(VcuVJvr{p
zpPo!;oTHE8EVbBUe5)yfAjbNOFHbyszU*J{XKLl*JTEqPl-AEwiWgXU#oC#fYH@0*
zGO11}))tC--(ppZ)N&mwpC6yIVud0fZ=qZ*Ru*c-$upWp#qr9e`31jx^m>dAO;yWt
zQ(;Qa@b~k03vPVPu`>(v<zo~oPb@5)J*K`rw#a1F&>VZ=zT<b^d-A@!T=HX&z3}|8
zIdA@$wG%K<U;M{BZ;|nytokM9z;yW}Ya{ITW~$>8b7ik&O+BdBRv)C2{s=#l>Vo=5
zap4&2o9#q`SP&nM)nc>pAmPP><V^e{@j=oytIranVa6+;pP4N4T4@72VTy_Fjikba
ze}eDHFkL+}&THI8UCp=k4{;gxEmdZ;B3Th7SIrTUt!>-m{&C`yeA_<7=&JfEBUp}s
z_hMtT%S(7kFXg4Z%yi7ldO0ue6}%3w)9WJE?e%!QUdHR=ub<xmzFGeI_*!hL_`%pz
zTzKM2ek_@f=X<^9y&-Qim?9NjIlPfBXS^2-cf>Ep$Fk+zYW`ZRT)50LrxJ!ql-}X+
z$5^Ms8)ID#Ta0x(EHT#Ou)|od!wO@44jX`H50?k1=ZJUoN@{G-)$X14j(NwgsNPLc
z%G<mX-bqUtic;S0-Ra$BDVu5E-QGR4eAv6!yAK?<#e0W$zxP$7Y~|hq-h;%pc@KFH
zb8ox%i1#SJJG?h}kMX<Hd)#}1-(69?r@SYx#CUSid$acx>AStRcu(`YC%XSN-l+Gi
zJ$JA7ocGn-+4sR%DmJ#Cx$(|0_4wn@z4i6~l#IoupZ=BXgWvhK|M>9q%)(#2|5s1{
z&ck75Zl(g14HI)GM^7#1SUtz)Sjwk?bAkb{YlXf}(b(%A!RL=Ww*5c<$=zEY{``9f
zfB4CNzx*u^fBp;C|N1xII`Q(upO0olEQ<Wq!(kFIF}bIn&aWn4j<4#gUS;mrVwYp9
z@oPLJ*y6>{$Esa#iFt|3@vn;qTY^|AQA-A~+0<<MJ7b<=34T+rr7p*zCs+}sOwdc1
z4SEQ3K__88=prlx-Gm*%Hp0$YR}c?|FDHn1*T^~Bt@-&%f@gM+Mh&yQ{#dO~cP__C
z>8~ZdY;B-6NJ{_erfc;4<v35yy_VFIgRNd{N%@AjKjfE#t;9D6T$P4{O@v#5&4gQn
zg7W!GL4m7n!A`>MLD7G?wj;=1j#poPON_V7TAtUaX=jiPazWn9zXIh*KBe_mzgf9x
zPeD&^ZDX+8|0L-h)*|Kn0jp0f-evt6n=J-|K{`kTeL;UP;B~%|p<a5iv^z)<?(w=-
z_SE*)_K{=Hbo?^CzOuKrfA&D_AaA}`@66R9ulp7C+=$n+vTr)JvY+pP+HJJ$_RI0d
zV=oQAm{>Wuaws?$>^&dz|1Q|Sdf4j?4h8$R#fD?o)LVyE_5}x44+jTzb?%R~_OHZ(
z{g2bDEBmSENNxY>fnfi&cx@oqe>v_A*vMpf=Fwm~;j!BB*%QGIa-I}ycqOqiLj2fA
z<NiqPj@dhj-^F-tvR2+zq1@`--cT@dP2;=kqw&Q%Yxi8%T)CH=_fqOjyw!cH_g_=m
zS9w5BeMdHrjwJnun3iGU;_)zke);(1lJA3W=PngRCxBuW7Z;!`D$~V1OzS<M60f-D
z;_)bQe$R5({l({(Di!4bTP^n<S@g@^j6+ny52rz*SuoqgLbbfy`*gXysH{)C_0cNy
zWigB&1$KALOhXWqOJ~cMs$oG7E*WaMQF^{yE!XVv<@s?TzD40YyIH(Yo|!&Vs~#<$
z7Ix%+sal>|nk!y7Q?3+!(Dj0^M~McZ*+nWiV($cjFnM-y0kYuejmL|(H$TO9Jo^0T
zQ=?DvMo+x(#Oc#Si=%IQYl+4d-}c<&kDh+wh2rUFi(^kb|7=^akxbYzJ#)TXDTxB9
zhF$H1NWMutbULsj?C96RLUdKGc#stf5YpvR^>rbk>+vw{)h>a!9^~qEkqeUw>L0_O
zT)e=yx|Lv>aeXPj5)b0D23b}U*J9t8I0t2KSP$AH<&D?I!{p@RQkbqR%`aY}-xEiV
zhbc97Eov-0Fb5R|wZHh#N2&b{J#TAldAC67(GnEZOs!Np`oP@6<oI0mAtLqMBMJXW
z^7>B^@PNY9%-mcF=(<#f?wDB?(beUb&n?X`KG3G)9*}rp`uw;*@&KrMi5CuY_4>nJ
zBPq+ZvOlvtF3KlNFEXUHFkdQ7&W%^ArIK=2#^=k#dLh0%M|Zf{6Lw9`%`Af2jpp;h
z4$-(}rsAbiE$o>RIT>YV-h~AN{3SJ#!GLb6l;^79FmtN3$k6EV5UfUYGOMa#@8fl<
z=N<m|q8gsIY3cX#>S0#pm~sA~uJU$A*I{z5TnSU^uP{5YG&Kc<6?Q3eoo@BR-bOSc
zox_gvYJ>U#5_XVjJw0QMnVVQBxm6JM%+Yiavh{@Q`~ssy{tly!8L+9Dn%|}7bldnj
z+WhQdE$m}3%C!=+qO`QgG$~h=G))`J5OQ^TI$bNxk6(n&D#6f%NlN%TRa>I67^cOr
zTnaOg*tN0;60OwkI2radCcbEJdMID!8D*a*gx%wFi)UaqCYg%<0@FZ!$U4}(?!oo4
z!p|kGiLi4H)LkjLiR0g{Mx<Eyb77ZWq47{!RtfXZi7n9_^53P5J&pQZ@3SrdR$=GT
zB>ayEZUz^&3oy4ZA9g%%zc3xNIHxR&er0;(HvfRC>R77KYmU3I3GXkH?7vsRn-zRO
zx$?joF-NsC{$sj^(;D}T2`j0*zoeXJ6_gcBDVSCu!o#0b@D>G6D|kkMEwHL$zJeNo
zoBTFOX}{p8|8Ctor{Ii&4=R^d0Bg<GuYX?2A5k(4r^a8?b5nCFa9N)Y{`<%_X4ACU
zyZx*{_(ct4eH6#?mGT7#1jjn-Ihi$!VSFa+YRhV)aXs$8N3Ae6L^Qeo62TpOt5*mL
zdwb)V_*TBHf1UAeyY7tdj29Ai$HJV2z42VUH=c?Y`0eC3M~pw+yO~L5;;BS0HKy#Z
zpS%5*H}P_Jd^ny-WD-gnjCUrw6IRd7<*qiEH<_hf^-nhu$;&jZ)z2DT7eGJw*&x1>
zt>kO)EKnXPsF92pe<i(=^%6mLIJT0jC1#UBPCNs6Jzq<#rnr;$VqW_Fi7l~fiIqZ-
zolU<OUrE+70n`^bJ8RTSj=Dtc<!94gatknjHhVTveFW+;BkHkMsCCSC^6m1n)C`@s
z+U@0__n^FtGSGFN>wL>~fotfj#{G_<gBJA!`D?sOBem0|s-2>6TJCh&o!$-ZbT`Xv
zYcJE&ddKT+)(909o9!Ep1=&F;;-D}bvtORMO>f($^v$sqXkOl=kCu0~w4~p)gx2>5
zo!-E;L|Z!t)erV&g=_GTuD#o!#~SarDd=#|v|PE3a?K8SLuV7vQ&6|FgWl$Wm<QE6
zy9u7OBgkD=y?TQowR>A^rK@ruR|DR#mD%HMQO&iE)y?W%tAA_#N^eO1-Ijlw<$sua
z!(4yG=Uj_ZGF98+Z8sjG%j;!sba}}K)Mwq^4&ulZMPXATbo=bKYw^|X*J8ZKi=v;6
z7A%flPp$<uqCR{P9uOp{aYRC|-{f=Sb_cwRXU5MXfLJI_l#9rTAYZ&940W+QK8eU8
zl9>3T6c5uz8TpqLA7}^tz|`{jaG-k5kDy5tDeSPs1;`>$&QjF@o96B9Lg=@=yTuCo
z<w<CO_9uqhnz!apthwnwr#ALC@GaDfkGyZP9PVtU>B04MNRb(Em-qR-Z~CUrYp?$D
zSA~Pkj3y<UTqsYS*f8~^5#O(iyz(C=_&i@ef81azAMqdI`bO{klF4%-avT)fId^_;
zBEt36Qw&c55~M1WU8#-iFg)*Xp{Tz?5N0DZ>kF*<$CQ)=CHo)JWfQ#!xH){9iR5M3
zDbN+T-2Zx&NvYo;VHTzh^@Ldnf|>cHd53I1uF}^P*f3{KJCR$S5vrQREgTqKs68bj
zWu8`+y(fGh`UJ^hWxV45m>!TEpPRehZRpjZIm4qOCR<u*NKg-hXfPtep67p$QgRaY
zN&f5_qDt4$w*Dp3R{6d$K_H}2NF+cGxkNXBKbI&Nav6^ICinxD3?;XMFt&nh20<T*
zo4I)EW^ZyQ=wZ-Gfi`Y-CUZhlH#-yiKsueob2oFzPD*c$FAo_4X^<Utdg}jRlK+EA
z{@=$WwV`XsOaFhzB>roxE#qAcd-&g_%T)!}6c|eAFq9Bc_k-;a+NF+dS(7SEQNLv%
ze9e91GRF4Wty|%RsQjKUhdZK1?P{WlHAo`Lu)CRI9UgI695xwlB7_Fka2(OT=ADgj
zN>s@Y(rMw=X4WQFi5jxey^~R4iFf=TRv--H|A>MgRq%*{PZ5Out;oa`JLUg4ch+GN
z|0h+bfkphE;&N=DjqP&f$NF1E9Z_0uQ{WLbu7Pp<pH{^`qd=&}|0fDQt-zR^&nRXr
z&Hts?uPXTG3O=jgXBGUMf}dCL>k9s<f?rTz+(nbcct2^sr}SSUSl+%dbau^Kr?(Nc
zKgqYM_kN$KtC+uwFusbka5lM`y5^*RN`X?JO`0qa`f4fe?}C<>JT?9rGRHW=6fgF6
zB*5=*wLECGWZP<S<sOd8C0x0>>`uSrj`{DrWy!Jjk{4q78%d5XZ@Itt0(vEJ_9EXT
z--?83%i9osndnTnB7V4jb0p^&)j3`sX$^R`?9VD*Xb5W17)~xMB4AnYOS+SVpHx2N
zLdG+5*VLG61_j9g+-8-c<zC3ms(8V~%i_km;m5k;2|~x4!LNNWmB@Tad|Gcj@x?@9
zx!Z=YQNd)2!u8K4B00W~@vUZ(61}5JGIIqOq?DJeC4$syQo?QoxdQ<yTVgAzX6i~h
zNF(qaTFC?ne*0FkL1r}-WUrwis-?ZuY9`2t%|Ps$3NjDI&h0vf_<c4DY5ia=Cz1P=
zSS`N_yRh0J;p=KAQ-qo%XoeB&T7S<&AnxT}N%FKmL(H6q8O#T%^Krj)?(IAyzXg&W
z)~-O<S%bMlV2u*savVnRoQcit%FCZk`R|7ff~^!&32PI@$oGB|n#0lw0aYIe3adTW
zU_M;!E8VU>uW&Yju)R0vf`#bVLW^lp=SmMs2G}>H48_D|1U(PL&V6^~A4a9}XVbj<
zy=%(tT<M8Y3Q<jiE4{5Xq0;E^xHlA=QfsL-=$%SX^DmwIOf9wwtEam#I(=*E>875(
zX6ZC-@(TB$YoY9aqNG?sSF(Ri$*%U2JheaQeE_EDTsr878Rzf)3AJXWpLgq*nk4FJ
zF-~^|{i~=KXwe3DtNVF_{vGO5>s8wF_8>`LKgrw}WR`VF!BEk1-_*Qjv<FrOnN>Xo
zD0+F?fT>{mhz(k-|20^7*f_i6b(~Gnl5><82nN10<^Y7t8{K;!_XdN(`n@{v7}+%H
zKhKhYZ|FokgD4G!wgjvNB-7~aB@??*e7rnYHrYG!q~d#5oFu)O;b@yknze=QEfPsf
z>S-eB4DS?QNd)m0VYID%^bV)V6s9D-4dbOX0KtDdH*Xgx5N|XXTRv!uyj^g)kpOym
z&#4)wFj&BnC4y=Ka!97?m)B8(d*$k+bjG#uIfop@_Bmo@m6$lHjO-BThwcVIG*!Zl
zjMK^A=)zD2TM$$XQi|N@Eh5jmxLBTqvnfiGvE1|4=>F4O-Crm!4;0?k%n~Jz3`G)5
z$p!ri&kj>6ZA0PLs?z^ESelWtdc5WYck{}a$5_CxEi53YP9w5LEZeK@^qeh*gq{eO
zT}mWt_5}<b1DKwigGLI|$U3Ky&WlS6)3Ok$hWVvR_1sdqj3~Uv#FfoKH-JB~$wBi}
zL#*jPuJ$CVp8p>diJuJ<GiSqGC9(?fg>aGHOBNgE8JUu)<*@@nUf2?C$zW0cV^-TN
zb{i#x<j76uyhfa|aG<V>!$Pz3*G{Di)Uiub>FbDe!_-MKnB%WtlElQzi7)0-x-$0e
zW+Lfg3E0B!FXS=_yMs(HzCBTZ?NjOg_z<jMH~6VPen7EgKdfFiS7ICoDKQ8`nM)*I
z-kLxbd2=7krT)NPH<t%Cn97q{KJ`zAh?bHetv3;=3DZlO7@tjCP6kMskPJdekeWnR
z7Nvu>lrPn|PCM!2R8qDQ<2&gBvkocpd9qx?s06VM5`((2P4V2Ji=|pIC|)cbA355{
zdcTbXiA(~&cE1&S*J$x#@iD$fj~;!g=wt;_=b5Huu3SYOS6N&_H@CKys5}JYMA=t~
zhV4kv%xj8=Owv<4F><u{=CWTtTs)4c6sA+CBQbcv0;i=)zl@n$<^JNM#o7`=1<I21
zaIsEeA1=0%)FX$B2p9Co=uHk*i#GF`6&k}blJ=!g^S@VtBqdI|@cqO_G7%9I)ey#O
z*wOIJ5CeipNYvT-%r!GH>^WC5(!HMh-I|qJk%QSmsF^~1dC=y@nrv+pjV4B#uO`PE
zqLbE74ogYar$LY$L@E6OuzvJ<%Dvk^rkH!TA0S4m(Pl|T-v_y{P;DI7wthI$?7E5I
zY`ZMuEM?x%YxJ`q5;qe!mv`7}+$#Sq-^WK+AM>*WBT-9#xzW;JQ49;xwe&n<Ij3-5
z*DOcF%dFAwB3R!0Wg8Vy{OaENr?sn_O{;febAo&y1)@&$3woS7(Xm&?c37aHf`?n*
z*5wlDqjdf}Wo$0=r=m{Wb*uJ9C2rY`Mn12Z7EFRQ(*h_WD=O=5AJf~IFVg##gO2v!
z4SDqRf1u3PukC}qHeXvS#qKpGuLzK+l}y$;WQ$r^Pf3`(zYf`=)Rna55`43A5a*_4
z5u|g|C*vmi5Ag^$(Lbq}2`4{6?0VLwye~7K=8%Qm?JRaP|Jq*ZIM-CjUZIXRx~a52
z`+HNilzXG4+^F-GJ>g^6+}`eAYqa}cDrW5Nzff#YIoC|8zLv(gS^SD}VHF*Vgpcbo
zm|n51J$Y)ck4$xn3s@EG?1BW^DC}ltVVPhDWE%vw0koj8dL_A%HWC*5ZtiUhQZP$N
ze{YapHETGhR{<ZTRRY_i<crCb%u0487i47hV|Xb$Y9jFK3AboE)bbWd_+S0J0cA83
z?L!5W7U$_i0@#rw3&tW~ezD{aaWRs25G{Vu)jCPmx)Oe23C;V(5_$oVyG}7LplpL3
z4mCzg>g9%ve{GldEVcvkZ{wlWexA|^5bx(J1!HG?2x)Qz($^H2RA1LWokSXaFCqof
z89R=T%bFwpjq&GUH{KmzNvrP~{hvgmsQ%26n#W#@{Pbz+a&o1B4hwt4vuG8s0r+QO
zV5j3=O0s}im^d#X-PFga^&>IZNEk>j9pndNtJ!O$2L%TeMl!Mt+M&@fV5S~4SyrSC
za&gI$5DFNUyiGZ<s*m~Gh=plW92oy^2AD>#CZUf<I^EK1uRH8?md+(ZC}DiU|F^2o
zDlJke7QV8HZ66d(mawz3P%*pa`SI%6wd0gqELF|C_v2)(7N~M-VhB3yaAG(fX?E*5
z+9)s~Gpz`tz*dRS@2r6p`q_&+0Rb9Z>`_?-35f`%<13k3s+O*y=_PMgc4cO}mqo`b
zR{o%k9r{~zzp(RD*uVt{A7%?)`W3CO4y_Mr(g=4BLEFKmzsat!6&;STzMe$z&=sW3
zo~#%o{88zZr?Bt|($IIg(fY`!&u$dVs#&~A%@0XneOX^VOME2m%nHQujCA`S=f~Z=
zotW_hUd8x?Fo(Pq9-{2|f1yktQgFM5_^NWI;2)-4Ri5r#Az(I$mar)i4mtwO!6euc
znS@u}qYi24lCm`KGC^qLt_c3Mqu5m+BZhxr@dq`OvJHm7W+*f9BBXZ;?aOvnTjFJ@
z{8RDeeuq4x;kSaX_fn{Bn6>B{3vUk>8fFGrL#;s(s7rxll2Z`XEI4hHNMNq-pQH_}
z55Y-itZ|nT(hlkrCdLu1nI(~BY{|A?1uQUfAmb+Ir}ajfoERYRM$$$tW15bsmAl^t
zsOv;YlKj>C2|$HOGazPZU&&ogz#3=@0tBwY9$ZddWm#Ja@8D8I5t;#Ya#ZWO?W1Z3
z_`ad`AU=s-Md%V?RVI;5k%G))u)o2G-d+YX4uj6k7t1{WFlIl)L%+zYne|v9w(RJc
zbJ38rGVd-g{U<cKmi+!c(jW(ITG?X^AQ_^t2R^_foI!W4UiUAkL$!v*pBc~PtT^iT
zH|$Vb>lzlLEm+&YOL1FJAeO>3h*F@OL{K*=r+V`Yrdi9Zjn=$or=+JbvN`|H)m)>S
z3O2-+*w&ylZP3>mbcTp#rJ>&e(e%S+;K%$-`hCbQG4vyza>8IQ0iZa}oqe^W42Y3V
zQo>L;Ib7PTrJ3<b(v_@by|fGm0Qwk2pYZSGnFwa!xJKAU0>!#y=A6sBmMUi}3l}Qz
zRd9GR#VQ_rZ^JBW<sggy11g#vpQ!o?y{D#&|6!#S>bn=l=fd2)wqQ_)$?WUPCMDW9
z%)n&QC01MzhD>X^F{d8W4eddz4pVsnRBOvVU--Ked{ink_=Qv|_oWEJ*Yme^avkSt
zzE{N>uU|$216RdeC%q)F8GwM4pG8D>re|W%oGdRdMThSWlGyuT>CAPS>yfK5u4N~k
z^fFxEP018BF_*3>XC9%P;C!-fU5!O_j9~%#Yv;f#$=NLX2s%C<$9fqdG52!*^Ilga
z?sc!^!2;O-P_pN>M358Kp23m?@w3Ir3kFAI3`PkgNX<FS4yLb{>!n;rR7Uv_0~?JE
zRTkCUr<&`vIvg|7e``?T->0D#*XG}^z{bl~rNy+Q{0F)52fFzXK?DgTcC1#*)%B2|
zPVO~Vx+$?tfAt8}E%!ZFx9W>5LyOD(W*~71xkkfC;uMft(~#;+x@StTdnk`^FX_$9
zBma|zBTQnTy4-0S03svG<-EJX#^OfWq9a`nI6bVFnAIzpHtG>wn&#;(is@<CAT;a<
zojIkM=$V0oFSETm1=J4AM3jp7x#Vj$hI#dk$zlCBtLCSaSP)IhGMSpWC|f=<IGfKf
zXBV`#rSsMAAq0@?<lm>f@rC^?5SEMCIY`h|;JQDt9fP-FgjD?queuY1h^`vskt^S(
z3KMk7@D(*teSkn(6(f9rHcWa*BnTWtQIJQe)wDq<O>nU3G<ZqG=M2l?d)dqbMol`>
z#Xn3FN3ssZWQAfH7Aj2w^Mv~1oyzEtt&ymhCiz(8X3k#FuOnF30F5lNQ|@{<)D1$R
zdob%*?lcG--D;x@ltrMyV6Y~iSQrRD#BM=y7Tt}(;~o)!mTEvkxR<-h8ju&VFTR@L
z%}L=M)1$skyu{Jb8~qc-GsO$V$>NcRiWB^B!G!FZFSAGMQppK<=?78Fj<i-o!USqG
z!@hk|q0LM$EnpB8Nrd;QLn0|o+Nclz4V5;0w5BKW0*5A@izVm_!9kHxl7Gq;GSa3s
zGOZhT^^En3qnB3IGmJh!5fx|v*^9(qz!Jkp5vK+hARNjXxH4ty1LrBh8mOu5zfHlK
zrgk8o11BKZDKF;KSk1X^W^XRO?6~We3~kW8LvK}Y=pPWFi7}|~I|Rkylu{^(M8``<
z>#j76sPfR%#{4ZFd}=L@5|9Zib(N{U291ITt!k%NkVfvBo)%Y|@b7}3wVXi~MbGYY
z5Rhnpp&+g$=!_gNTa!O;yS}vZEB8tocnz0;Ehex%N-vPEy(+QUygx?`g=^vpP_ucR
zv`vsY>vgfaMm1#FU*>Y>m|WeXVSLeljvfpnO~1e$RM{DlPG3&}1$+R<1pA$2ZLjU4
zS@yiCTf-cY^5Vke8MZL6V;2*5C7&{80%AZV{1W&ap!E?db&P;b&2QuquKUaa`|tc|
zJsc=B?LgM2mDMC~i^^44Si~F<%X;(?ohDAk_J3h%u{zS>=sHb?Fz@K*8Si4))hJNn
z(oJT)jJ>2O1q>j%S#wS|)=ZtQIt;eg>C)u6jh0B_ueqMWcx#x6gzW!Iy(zOCzFEMC
zOE$ui?WYnE1JPDZ8-&nY)D+X=to|_pL&jS0SP9VzaMzFvcFj~<jbDp>!)+Sk)|?-z
zUq?AV6gO%f(&9?)DwFnd{6n!1#@Pe}Od%z1De-s2Jah!9*kz15E4dS~>Z2FHmil`q
zUnYx0G}>OMZ+*fg0`$>q)2?DXye=)?o@NH3`3O#Ha<!@c=v^aNjc?e>1;9PpjpD&*
zK>4uNzS(x80O(S-FGAx5STG%65`7nxCaR_g<taz$xEoVLV}%E#k+l6Nhk2U+4oWW{
zGUyHNo?*`d%1ayS7Gk1N<d%)~mo;_X+n7Hq#MU?DM6)6O%j7sxZ^5b4{#lhs+VuB-
zi$?i1B{)5SZy@dh2`0?I%BJytl?C^T8>G3XjU@c9Rl)ZuurBCX^P>ATr$znBOi@T2
zG|sHv{)NHxpv~?`@U*5_TVH7I>T7sh1IN6H`7#?f4pL-gU1RLTs>!ASh*Jb6-THl(
zn)#vvjcc>N!hVLu))3G~Ofg$L?oh+EdNMQ_b=YPb#`XEyFpscK1M?i_0z51WiUgc3
zYcAq?!NwK?8%qR+XvSgR)aZ?&N*&iav|DkA_r~x<F^Vev5LY7&l$@&g3l|t&19$2t
zUoxROqRl$yyc%{_v}L|p!T`5Kfv_{$ab$b%*LKK2NvM@<ku+U9f$mX{6sX#5RuD6@
zJZ!zPLFU&5bWCMEsZOkadWh(bFkkf|BcVN}23qJKfdde97CB)S{KaYnQG>8su9HT*
zWX);`gBJFgCS*1O&dUNI5)Z^`*~{!)(9X>uX?OA%8i1<|H{Jz)5{4>7lB-j8mG{t2
z!vCDDb<{eMT@|L`R6IxnWM0TAF*_0d<!P(k5K0B|*scTZTbS*|=!KmLCKc<1x9JNo
zwO}Wuc&YvX+k9FDwhdDLcabvSk@5}r458#8LoY5$V7F_fdmW(sNE@K+qMi=f$x~~$
z)!NBb+W&NI)9O%-cVFEs;~?a23D<K*_<ias{J()|3^}eK)e}&D0Z0cn?r~r;%s3UO
zu^u-vCv4h*lO_4`pC$;q8oE!LHo52K7R++tG>iWp<&u=i|FE)l)Ta~^r)>o@bDE%c
zhzkR~S#1Pt1-mV?P%w^W7>2gRD(GTxZJf$XRx?!Bl?k8QwzyUvtpQd?hW%Gmwn5s3
zNtCyaWSvmcg>lWIp+$V?DNQTc$p(H}F(c}~N+pcDx}r2wVHiZcS2xUxy~9L>RV*w{
zI6H%po)+>UK8oH_J?X1N*bxaIWlP~elM$QWq}srckzCt^HK%o{e)hjbf%S?c`VR0K
z)8%^MAuZH<X<i0aNe~x<@puM9@9xBQs0IgzC4}rwq!OL5Qrj^D?}nP#$&&6*u&jG*
zSvUK&Z7|g>`aZ_jM0?EU*h<`q^PG{Ii0s)EhHCmn>dao4L@hI$oy|$mBN1f;@)DyA
zvM8gGeAs3C<+L$OTkqV67z_J39%MViY&Vu;FVaSOh3QAV@%eYu1zV~p%oycX4U^<(
znNWtE2HiF|Hs_Waq+8WfcZtD}R5Js)rHxt3TdWH<%6~mG>ffOzom6n2g8LPGl>*ZW
zCr)FSbBgtGHw_|sng)JEBPG`U^ytU|f1F5IC}{^MPC(ce+F2@{TN<Bpn?!AYne71Z
z->*v33T&1cYn25rYg4N0q6?NowM)U|cPr>q(62yxCV!ZK9kf{BIMv7PN|Y?v=AfI-
zkLc2*1L9l#mlW8N5zXvhQgB7V`xV%<e@(IPB?!Bsjjoerw$ZlQNWH4I*~&dZOyBA^
z5#(Y_JOs%9>wn#Ul$gij<!0hu7k0)o-Pl?_l<Ul8_{;Ilcjq&mq~~-Gg;yq*E@TQl
zxm|WYof&5DT~{uZ$ui*keYI_pZ`()xu1{5LUDJ;S-Y*jQWImNo^P9+Lh^4$SgR>B^
zc*jTyvR{df;R&OhUCmu{Fc?o82)-c)#`3hS&wH1Z==TOFQ82$6F}z~Pqeg73(|j<o
znbCXE+vaV@8wLBLyq(<X@pgGdetW&$-X4DYyuIE&e*3-s-T{6GyqC;lMjYQ5_H1k2
zyX{%>rvxU%M~2<;O89y@;(TNfKk6OhIh&(<$GsC*z!rGOIO*L%y<5zCMhyQL?}>b8
zY>RwnY&Xvt@{b|E8S;-YZk{vZW4m~E!kh3ODHrkB@u>HvE9o)(L#ap9D(+31FAep{
zo<{m(>Q&G4<X1y?_BQTNf;(mNv!OftqB{wzYs!1td&Wxacj?|+y-|CI1JU)k_pJAv
z<v7?VsWD7@(|8bho839&z1@2UZ!r>;`5Nz?R>y7Tv*UI!=b5nQ830tJ*4SFTT$q;+
z9NX+Y>fgnTf#!aYpUEywpXf8duV^9BK-H_P8}o@Fr3bW*dB8y8$-YVRr4e*XWOFuN
z-DF0v7{$Wz?+UX14ljv>H3hexFe&aKLEq{jFGfl?J2a6EKto~vN}Sn990OM3d5ia1
zyfCdi{k8OEar|8#J1uA8LAL~5&?d-y*~cj#579o(<<43Gff;)?XS*a)LZ*PXj(5Z=
z06YD8^6k<6PK<Y}74xUjQ$b>|+Us@I`k+_X$LMtIs{;u6^@P|e)K3)Kg#1BiP-&$h
zc7fup1KJI59Twjb^q5upk5XnU#<u;c$&>$(EkUnO>g=}a_nSHKcCW`wPUVYYr|m`E
z74-SPSljOPD&2B^#`-cgyKA+0P3?vbGM2G#)Ja~SVr2Y13ic9&gKY7yvi%t(;MI55
zm;7(m{ck5Y<^Ll^g$Dg^Rlty<=&j;x0GCp@bXxFF`%mb`DT0yR;b!DuIBAkgEbYg3
zb%O6qWE=LkZ%<H$dSw%rU>dzJxwx?Czo-gd(vv#Kf<3Hx5eS9Z>TU&$laVhYbH`Or
zg~D(w!p8-&bDl>D9A04wf2K4q-&uJzz(a9T^?!scZi9mmj%|8a<1wr0mLMEzb*0s8
zMmXG>?qSG)Cnpp0hW*W4k_}4YX$sn~dtzb%^R9)n<%+~G;ehP&G3P_N*6;)18^SxQ
zcFA_%To*V|_na;-5{wKO!4>wmdQXU6S8~aEmqbaBPF0Nh)f@h&^yWXM;3pOQgo2+{
zU?Xh(@?*OEQ3XwYvN0)I=7b@?{c4k8K#kd=NR6){B8-%Q6V4^HPh3VA?r(@x{Ux-|
z+ay5j{YD6`KX)gCI$344cU<^0#dbcof1G;0xIDZDG&S5{w2^Q$BPJs!BjKEZ6Na0x
zUnUPvXo({$m)U?^(qPuKiYe1-#<}@%0AMzIjfE?6PgFc*Q(CsO+4m<NaEmJC78Q)?
zJ9(0?xAw)s+cH51XEW=n5=>!#f)15?5e8?*!^9WCoBWNi6~|ng*}y<i9iRXjd2gM*
z&G}YygB$sR_#ySSAns0-?}PE>J!|{MTu!tVnG|7L|EPylJju8Eiv+wMLLEYYAPl3n
zw9r??-X*;|zb}!Jtd<DTDPS70gDvlDjAzR`QZWNh8C)u{@tzz4&1ghXX2FnRpD)E-
zN01^W8C<7KC|0}ZT>$ZWbWe_v5-7iVt^r!Fv3X9Qkb0%yOVuss>(8l!wza<_7&6h(
zZqwWHBM{nWFmRO*Htd#*lS}M^nJ=>FfMIouaUjo;W(t}dV8wW)j9h>2((<;wd-oRK
z_JUl{ym0#QBRHga?9pd*MY=(e<sEg;IP&Lqv=unY+<m0a*K)!KigW1WguG5B`}102
zw&HA^8{lg&aejk!3;UgmJsgKLT&0A=5?{0=G(4in6<ckbjkjM<FNEwH#w{MMPLjMO
z$_&k}raTOva5$rpWjTOL0$#CZVV`@P-!!cY<I8^gV#oVQyFkgH*rzlfhk3*yY$tQb
z!(@h~)sHnc@xlLldEeR@x=}T^WIo#V>T2@l_}b*Xmx%M`#nhA7$?OoXIXS+jE)&e?
zj-=1Nds$?)V8=wXkkbzLT%s;pjp3&1897p^7&2>kVDW#MUaI%ylgj*N1%_MSqS(_4
zo>9==X~yvUqWVZerlDjfW+n)kZZ7X$+bipLDO0NP5!J4@HsB^Ha9s!&BqZAAO`Yem
zL`hnfgcN`1Tn3yt9Q9z5O-y@OYVMc5DrL^~BM)-FlAWYS<#EywTG%5~Jn?S$mIRcX
zKXsaS7pWJf<n1u*Yn_E=9cKOn{bx11UnieGMo_mdkZVSvA>(_0Pj~-Tfnkv*D)}^J
ze@z|L%S}n+ge0_KAeTseVfo<N?rO!uEv;#jT()(Z*x6fln&F9A)0fh1*+#didj^c0
zN{rU~z}`}l#wPN&x0g_<I_aZqsf<G=t?LK>->7$f&w^WZ!e_brd+G$S?sS5-y7Ohe
z9Qtw{(At)^9%vJ5d_Q$EIbzj^B(4Cm_tmgliU>)B8l|M(lzTEM-N!vmCFcq+?Hth=
z5o@_*(kK})(=Th&FKHslBo0!*7E++mkLeYRFoI;M<J{d*&GECz>f=VR>M6<UTa3(Y
z)KzZ-5IyWSLbYXROQq%gZ?io=mdK6-yalqcpt?{TuNH;zT+&hhhv~)T{OS4@%H?g!
zfE%(lsyFiVc?#{CQAGxie^;Y3q0xDdf^h|!_x|tb#)x7z$4z>pIetB|j<T_}3BJ+X
z{tTu6y@quUMMMI*Z7|xbl1pl)n_>)K2GXx$w|d~q4YtGWdU=0_dfTwOfMbIOI>QCd
z^JKs+BVYtnj7cw;IOTK<L3&ke8Q9qwQ;KCdm#cA~k<HXH;B;Z`6eG*tOoY@~Vp(v$
zVtH7q48mmMK^E|lk<jb+c_K=j{)i`7h+v4w-Vdk)Kd3-l;z+@)9L;Vq?03_*6}2@n
zp(F1_S83bp_~cn%V9;p;B7;X0AL`NJj{1Y~JmyxX!LZ0s#YBu@V%hWmq1y5F3fh5C
zxYeZoEfQZxqDxcow;Df*d1dCrHUR`R@>roi(;D2icFb;FRofEP>og>fk8`2ns=+(s
z=Z3w_5KD>5fTf}TCKY~@lmbLi+$}=vkrG;-X33P5%^btQPUCIsaJqKjIMcD^O$Qq`
zJLbB~c?H26aY0U{>1e#J_<CC>O*6Xov$~YQ4StU0A5UZql(5kkpHpZ4#`+%h{{?9Y
z-q9eM>#Phgt&m&~(QZZFv~?Gw&_smGTnM6>AYsG_YbAE@_m~?`)CRbO+CvTuvc;w&
z(8V~QmLW%x&INLb&X@fj=4EzYSFRq_dpS`(X%I^47KwdHF}P=?%n7dRWUSuBP=#6i
zszH4D21B6y+4dx!)&Gy`nLkxveR@!xnB{$yx6S~>C)dtKBlt{!46mqVU-vTBr4Ch1
z1yiL9f}}D94Db5coSlkq+_f83v`(!ql=&Fn>e~rqg9`0{q(j|<r(JC5R$-t3CbFi(
zAjj^b-&!Tbr2<l-oB}-TK!Dc3R>S$*!gj+b4M(nvjPX|$yp^7TTCi1~!Mx0NMV_B2
zUodqAo)huI9c=)Mp9$j^{M{;Llxq_AcWe7Chr{SI^1i)Q{Un%Jgjq5rz&cj4p0%wf
zRHV@pFLR+c71Gp>KZVmsdU$72Tu(58WRp%$1W7!AxHkuA%3kFakf7{c)Eo3>Y8dvZ
zB|W2bzVuCyv7W+z7&#QGoCB}9<)1ASqffLe@XY<iGv<bK?TrifvrE>iooS}mck4OV
z-+*KM69dw;$i=XjhNU>M;KLIfZj>rs!1ebTHriI2nIB|N;CL;{YI|>SLu9v$+&X%f
zZo1kg<z~olRMuhzxJrLY&o;`Shq<%m@*?(tn2wR`CPJn~KlK+R`+u$AA1U}tg1Vji
ze^J7JC6Ffnf)v2=Ny<9$D|j6IE8XbSTbAa^Q#IV_g*|rVa!DkF=L2tA8?M=hxdsnF
z-N6u+T$k*wgRf#%YNui55^HSxT6o`Dq)k7p*(8?R*3~KzW{aU@m}P98SX<7XRk~pl
zP_Wq^D=64vW{_qTp3Rc1Jp9ga&3YQ-KIX+R(_*9%X<#YEo&?p(U24r*t!C%yrK4Jz
zOR|6svZ{4mJ(7)Pv0mk=tWe0W9&_Ny;Mbj6MhW8=qXZk#uuz*RmnM8%q9fR96j5U2
z95YI!SThl3;=gI|Y`v#cD(tT=&fzE1^tSbl%Qp5xjYbWuANIsEm%_v|7fsh0=H2sP
z@WYO0qK8Wixn^Lygk&07k7;DL^8^X3L}njgo3R8@qR>)b8fI%UBfVqoNVnBt{m?ev
z+Ki+zxA7k25y+WiFQ(l7D<s<J3uOlYEHQJ<t>ABTiE=do6phmPjnWaHm}&WW42Wc6
zejfBhKJj#zS)_18fk#6iOAhsOr=;C0U6zu@7dl7}<cK&$Rs%@G-F+m@GER542zsKl
z{p8GEORwhm#Xp{NHE}t<%8n)=J6@u%rI5MzMAvJ|4#uW57kfS$uXI~$Q~v*mTKhl#
z9<4RvlvXgWqgD|^x8mMWt3r3o+*2{`=)L>azGak8Z=7}UsRZxbXMNK8&h9ySZ^gdy
zyDJy_hFEp$`(RyN^}eWl$a+9dBEHUzpkn_My^-Ep?+s#Yy`ej<H{6|D^@j4*d&6>W
z*c-}a^Da>AEBA(S<&}4X-dI;xqc;+@4(QWvTyijrrq~cMl3#v*13T3Bx!(`+m=={P
zSHM^VRSy>@4;RlIE?yXE=4~J|%gk{Pi{{SS6=4s0R0@cM8)4fD^seJC9mXw0^ZLX~
z&H9?O2uBxmwB$maFf->|b8y6mhe<TvX|-kFD~BCyWn`1R+h`;OxU(@fXahrej1*^}
zVTALCxePOJt4zs-Lh2NupM;KeMj9sbGa0hMJJHu8{f?*>f#au?GfNOAmZrjl$9JNN
zf89yHfdKuR$|#%SutU#?0BD$5iXN#%Kd$>-brQh6SZ7Q2lWSpRj}hdm&(-s=prtsJ
z?(}fdy!62b6%+d~w%8}1eOwJEQW#@ozH~HsII%r`8-)HwsVT%@%eyv$s@5F$w8=bA
zs(-|T3x(2xSoPz)ZBy_Ex`P)rg<2nmzNq?~iBymWv4R`{Oj6w8y@wcvQW!oVnKt`y
z89QOVVP!u?JYW47i&jd>JGd&qDa%Nz<1GGCNSBwQM?uFR)RoT4CM|DL-V8g|S?O-H
z2V<qib2}@zY?Z^**-oVCDSwHB0Xl2la1~{X^}vUmrA56mu)>X~T&ohVm$$S;(1|A-
z_kQm~tyzCR4jq{7^S^@<{o-7ZSasv)XMj}Rh!*y;&~QVF$~6`A%p|EDIo1m0tv!AI
zC)-QPACQ-R6|*fCcSq1Qom}Z-FVjCI7tUL0Nq^9{x`})JB7IRk=KNoADKaeMI^%zi
z>mg?<7WA>ewT%@;+&K2fZ1=@QIP|4NV|b*LMl4Kn<_+#(I!wb{f=?>XLFvn%3ev|C
z-mdk7t8{;xR={?Gk<Ct8YXWmyH@7h#{9o1V7oX|>ngWx;m}blLS*G_YE1!vZj0rQ_
zsfyD6TqVvB3pdHcbqNPNta)fxT%*iPaSG2S%s^T=FY}X+ImjPUPABz{)pNTqoqEm*
z+TT!xTO%u|RuPs18`v#6LB*sK;mhl}bVO(~8yhzpp2l^@mld_Pgv#kg$;s^f2FuL>
z0+ashdXYv?g{etZmeh{x)I~O2uN6zH3#(esvI6eL{939y3GVw5H)>~!lfY)QqLFgQ
z1<N2YQ`2F9C3_<FdnVcYNg1*AQr$VFpJGr=TQ*wXM&GqbtVG}G`xxIh-b29l+P(1R
zv1`%3yBb>oxke`kL4aMU1o?37Fif);U%`}dKnGGN`G>+z`*K1Dv#el4O^yPNi{-Gz
za-)Z0rs9huJNmjF=r|JNJmU^vm(+!_aJb%S%t?bEorcNr59qEGdcK5&VJ?y^mBW64
zNlBY@cn<?w^Dmi<Dl9N`(QXlwl}Jsae07<j#3k6|M{X%5rT_bPs=u9NwW;<48mexZ
zhHV=FR=(M~iLKZEtrw-;vbPaD@UYtym|^!x#oC`8ZeL$tNnVl6RO=@ijbc8uH0to$
zk^gd)+*-IJ5jcZroOfd%Ea1Rov=XRXcbe0Uw%Qm>_&K$Ymx_UH)~QL?>(@4v06b0c
zsw7+hLh)?2-G;sY_*)R~<#bM1^rjr2?zZAh(YQLhB13dJ<@d@L%@i_BOxIkCZN)yd
z!GKSb_KRw}Fdlotw3@c9?bFunH)-<KKOIEGe2Ai`{tbb;P$COs31JmP@Pq#(+lOcu
z9%O}NYho6amraZKS`$-xtGqc@WL`{xu?|r#)3L*#zJgai)}QHr*+L`di)SEK`K>-_
za)05%XX(r?naE%;iAOzh#AY!+-Xbh*trDKb3^^GTXbE^|wO`I1wVmHeAObaopJdR;
zrl+$(I@;HkiT1Uz=gXu^^dz%eef%`R$Y8XOUa5bnzzDS_Cb8xHOUf?EqCcis6P+0T
zFof}&y7A8xnANP=&+gNuZI6pK*2uKHInub^;w%C;4;m&GVf?gv5*Sq5%+^W(&0_JF
zH5DZ6I}kq%q19;&sw^ku_iA5!dGp$-5m}$K`MHo%pblou$4Ctx!lcZZ%8<peO0RYY
zNxUg&P{dJ6MKt6ku*{No1pJH4MEBJF@K)4dycH6xAcNf%=Y9pL>Iu{W$RiM{N|L}p
z>`_<lQczUDn`)QF8a&+SWj2P&p;TShYWBg_J1HT;un^%u0n2}&9^a*44b<x3K#Tf0
zHov3Z8|0Z-mcZO2dj_=)+8G1vG+MW;T1u`1oLXj&@!iPB4`?R9;+~$M?_>|G{p;K4
z1qTsww6bo{TTqMbl{y6bOBWHu`RgQ5!`f^rdvB}<qWuak<Db<bgPdt>q!_$a!8S<S
zCX<pP%Q!`NZEfo?Q8$1BIVU5{o5h(8Zc9gVcU=Sn(LvyR6M3d2oj^eV32qW=oR8*Z
zuEo^@BiV5+kOl+yWaR&h^JH*X|MWBXF!SfldA+=%nSU;K;tPA-5$he9kkQOxxgyh{
zdmir18dXiq)T&3&L~zRFeAAbhok?VsPL5X&BCX^&&1e%G^`HuzsksxDJWhD5c=B&%
zp4B3KNwcClTZ*CDS9HJTb{w>nt+T1)McaNOuJ&D{FX?^Jpf7Pq;2{L7dI7wn$-x@0
z5#_}7qtr3Yx|%GigJlmeUNEa^`OY~JW-)1}*G{<WllAM9=4lv3o548aLw}lb4gwn8
zoYq9#q_jQ-D+)eMP)A+ZOT^r62%D|hbC~7HTaVns;7?7S0Ge)kkmSxW^BB^#tFhpc
zOB0LUpKI<GC|<~I2YiV)T;94Cd^NMTVPrwChE(ke7t6gSl?O41K-AXfSCG*QiGYBT
zSaJ@<c>~OeL_x^d3v%%qB;ZE5-zm4AN$sfN7nS%~2tg-fLCdk`P8BB)IF*GBVs0nL
zr3GtD$S2&nm-Qib=U)|19C^Hm9pd7pqMg-abG>C+fE3urx!z$PnVrkOHRYQVBbuP~
z^B$`A7sp#X0!H<+V36@^ZMmm&5gpIDuA&JayP!#(EdzP6SXOm&@(JNPcYT_}Pb2MT
zcIv#@s&mGvnr(_VdZBoWPord!W2iVZGinycGEJjskaDtq)Mrb}n(GK&i$jh~JfzkY
z-{tnTN$_*Hc;c{gh{y?KoKmJX*(-2PM%A^srGMQjdN%{dCRnWuTx969R2!of)w57K
z2OU&R+iA<=XUN>#9WDlNk<IczcBvc894PEr(}`gcL$kGOVz9Kq38{al)gkMbM7{{3
zI*<@A)_=2*SZ_kLQSml`iIA>V3suIdFGGoF6|e>N*I5scZe}@;?#i6jNV$ot8ZbQP
z;y_|W;u2weGzKEY@KVK9!3eMpGYpWaAA%3e0C0bd&_LGIUF^+9Tp}M=SX7(q<3^5d
zTmi?Fha8={-x1yCJe@AnbM=q|9R=;|uBSns>h#23e*26}%{1E1c|O+mUhXFmSa2?*
zwY@Ls!v-A3!?ri>S;Qi8>&rPXNFC#xPDpKB^|38CwX%uWAm=A`2LtHj*j3I+h=~<^
zhd_q*liqxLT}5wC4V+cO{ULb^ffLxgGQ6@S*kbMG^q*#%A*;94pshG@W^=Fw%^sp5
z`f6*ib#)jB)qiez-Tt`WZrvY;1$%6E3oI-?LRRG1dh6=8fYW25_;!nL#~X^Q!*}>^
zc%4vcT`QdGD9<NIn>ovHcI)i+*`2eyR^%zY2;H{b{95e}c5y;#_bYM{TwECrcC79R
zc1C_piZf22Qw(;PhpWF&9lKlVkdKqyt9u*1QFgm}M*9Buy_ocY1DqF=b;cHTUmGoZ
zkRvxpM)~_&v_2aw!6d7q?W1L4c0TthUsl2W3XHIPpb<MnHuI;lv)&T_C^tvq&JuW=
zGD?8pjD()kt&IN-{OIg6=vSo2&Yw@!zo@*6x>c~lyynI)p+ZZW8qEJ{(wDQxJhLnx
z+2{13pHK<O036ey4RLOFlpqtQu%Slt->2L23YuKZdv$3<fU!JlxSg-l-Kzv)Ugx+u
z&r;f3<t!DXaCIgi!gG%I{%ys6OTqI5Cc8RclB6l)6l>1LP4DKvpddjovd4+EjVUt1
z;sLc^goSZHko~3F_=110Qah9?%D?V<E|Ru82X#Hp9bJ^h*_r4dIfc+!4Rn}WY1@uo
z9qVbi(DrBB02*ejnv>yBTRrCg#hP%3p8hfQNXpzaBx+UvZZ)f@8zO!<u$rS6I$e3U
z6<NoxMe55jDF^BXNaiMi?R(}p#(z%Lh~*#4xUp@SU^chJKXx*=%$~l0MQtxcK6bPq
z0uHu8^;}5oGs20p)Y8KUU37PdbeWPR{%-r9ckh-UFExNr4n9$$gThT&5b9Jl#fOQn
zx2LbK9^0@mDFUtjsSDm8Bcb|N1o&GLMvyE78c1SsASu2Kjn&`F9k2>U;zP#l=fp`!
zw+Jhrn8I^YkT;I8gI&HOK?l&M!^F{6#=nQs(su*T4VHIR1OZ|D;X@w|IuK#w-sQ5+
zuub`I3A)gSquG?fIKt<zA*T%1fWN?G_8t><OC(Nu4y+?>Y1ZI%?+O@#=buGDu5)TT
zrHx__QZ>5nvTM6%UNUoXI`^OkPQGXyAbD7^t(p$D+O_Vn0J0VNZf@b&eSr<eY;eIC
zq3jl^^g8}_6*2kMQB??mTHn^u6cjX_K5D0k349yEMQtMJs|y~Rr)%OT%n62%b7s08
zDI*8~4pyj=yGPNI^U;Ix7GenJ=}0a~;6NM`ZDf5mA{yOfzph5_R-@ZFY};}99h#?y
z$&u2@eF%kd1<33waAo^C5EsiZPP0^dLG*=`S*P|{<DJtNbJMoPY%>XxYS|Q2mu={x
zlOeakh82j6!OhVrldqF0nnxd!@pp54F({Y`l=E*3dQr*uQL2Asu#)BqjfbvkrZi_;
zRvHI%IOkju>Q8c_!xJ@=I5V9bF_2i<1i@o&#w2sb-6!M??mn|MDbqy)K~gl_eGc*b
z1ty%fgY^&ESrjnV9A39M77RcFr6E4zhH!KK*WM!EpxxiF4p@=GD09$Mh9+4w&?-dd
zYje3jp+F)4Ut5FyGQo213+{;FqK6WdW7LjX=lio7@CL5*_v-RTRFMhGyB#hu4ss8-
z!v4q0fHzFEU<uyA;h?g**rx2pbADK{CzLO%rvMs#la%dOt%j*gur9|QsCOi3-S8a_
zDtr5jgdMf=JY4fQbH6Ob<_xaw>d$Drq}%jwP+LvL`;KXDt8Rh|ko(i>5L?oQcy3eL
zn)!Y^X+NX+t~$1+@Lpqos6W1tN@FC#>}TFH`}^^4ngT|r;@eT3Yy@c|zmj6P>mT)2
z7hh2!)oxl7^j1YmON$2pT<OD60uoezXY|{DHCB<A4T)=rPs?2r^I99yIgZ-WQ{4Gb
z4DFw^nJ46ZkW~R8Z%$JO&`Wl&3}72PcojmOnqR7AuEudQk?^1Tpgi7~!CuqjO=ZyT
zBO>(QVXOmyw|=*hRvlN3oc~Zf+S)h;)7%%J-!v^^t(MpEn3$~enRC``AObvD$+V_I
zj(6&C%vu-V{sTcC(NiC*FCT2eYI%8Oh<E4?`XS=w+%L^l4=X0Mq9Y)C{huW4o9*Y^
zlw|c<BxfE7HX{t+4JGG>K!tF>%CYRV<m%8>NL;Vul_c$r&2A0`NLM|>pflPK3}G4~
zLx@hV>y?7B5L+S*?4LSBSN$#b3^by-U^5zMh<ZCVvs;IZ1_S2QZ-AEeMou~YEAc+B
z$6Rv!O)$6$!(dv|lz#*9+0bPcYmk9tmnj(KE`!w0T_t8Kex$d>OWPzfvGHN57?&q3
z{l+N*Fl_9lT{3E|#ds+uh-AD~IzjRJFjYF`VXMgly1N=0=P3y5;kL)<k(!(K7^jx0
zS5+5qYQ10pF7U(8J^Z*O)S#IEg4JUlE;?h&Z(1wzk&i66<SGYwM_3IY(n88N{9cge
z9F|^BItj9rxcI&pU-Vb>=*WAA(U!J^?fEZr>p$!H1#)L%iFom*jt*SD?|-=+ZQY|$
zNb|Ms875-KujDie^=^ueU@^wRy2`m<a$4JSRM65}vb_B-_1+qe{m-|uaU%GQq34<%
zTYR#NU8F&2b4zTTOPD!xX#$HDW0JJN#SlglSDaPT<`gsms*yZG4<lQBj~^#dG;O*;
z{9_t?!O0dGzZq}lly+Qq|3b0<pqSx;tGYDZiz!xSb!iIxf2J%G3YvNrfnwLpE94v5
z=5U6MxG_18lIkq4C$w8Y?-F)mR#cKPwvsqELhh;!;BWMPn!F(gGMo;xCig9!cCY!A
znsrD|HxA)WWy|Rl7~Ab5hQWWgdg^{Xv2gLYtZ-^G^0}&~OP}usc}bZxFT+h6&ijAE
z*qe$bmDcW99d_wB08G1~;iZ@t{>j}ul+*@<M56zu+!pJcE*${TnM8Syk8Qbll0UTq
zWz0isyzqrwTHFC9tdw&qe7J%Af3H<ADbBP05lN=ee2lM|;{;t1!jgOk3F}ICgr0h^
zeKWs)5emWs05X<d5Rh!?deJs<uCV3UPf5;cmYfKLbQHY;acwn$R)3(z){|Nu1XB0z
z%{CN6AT6m-_3|hHGJ{~_xB6BC(gpeXb{3Gxl^St)l$Y#&Kj>v<Wq|f|n2cy3=qBC?
zw~$(qyr|3nhlJgrn;gjHLWEp;gF%o>AA}xqU`WK+EM_>tCXkEy0{}H8&Fo+huENQk
zKov-g(!8~6<tLCc;YT1C1bqyF9w0U4vjFtbYv`jF@ZS}okDnr*^*RlGe8$Xdkck=J
z^9w>4hCVv{Ujb!N%8*A1<DJimqcM`P7xJ*H1&R1aG>t<>&;MZs_b8Ag!+)<P=hMVc
zv^QZvbIEUZYvJQ6U~9a&<UgdejTX8s%m1A<9?^x(HGWb}{aQ8kQ9U$J%$BD?u799Q
zTiz>*N&4wpZ4GW;!T$Snw|(*Ymz8GdqlrWA;*mW-`RXW7Ol6#ju$@!R$>lPc(OTTe
z8cf;ovqi48@4dMOS)*Ul!L;?&hqLjQ#V>7SVBF?(-5l;riUX>Dgo1UgZ}kHNHsNt?
zjSyF?IZIIAsnkj#qVMjiC7|!v!{zUe=sQ^MItBNkIP{(FYkSc#BMUGouisrm-?2y`
z7YZ^oM<ncctURi}JX0*wPBzes4eG?NaP{Tr3V%VJ-awGPybAbhFyn)a!PAU^?{jnc
zj<s;hQPy9UZ180dVx)p^V7NKHq7mMry3ecbb@Ihq*L_P;z`vkpHQ(r>F4w$~aQ$2L
z)JuQAr#4VKsCvyJ*kJx6e4OS1&I$gX@`BDL)ljwOGnT5xaM{8#n(Zss>A1faRhe~O
z1DlzY@J+l+RR;Ar?FMhbRPz2HV=-||%R?KFT${l{fl4xbpXOUVLjXYABlWuWCPp({
z98F?Mn;J3*j>O?$4>pD2f+xf8v_wr%?{Wf*0lNW01WfSkNC?;hk$N?`Q66ToK^68q
zC;c@&qc=v`qZe%H*?ZNMg*9EtF?XoD!h%94j&$Zfoi4#n#_(NdepsC;a$*h4;TSs0
zCOzwEb?hFVE0CdG(I9nDJ_uj?5VM6Z%4vqhlj}9A&1%{-s_I^S4U7L75xPbA^LT(K
zGrHQ@hxEW)vb!)!0fbc1b?E#IfCX9wbR{z6NmUN9wGJIDskfmY)exa=l`Tmlhgd9(
zhA^6aj2O6>fgo4IvcPI$`(20JO5k1s>neN7RIL+pS<+0W)5lIKb{WVFS3@Z;h4rT?
z6)(Wg%O$_^ti3PS6o4t*MHo#vtOY6KAVgvd-okSC38@gBk4hr`H)>pa92F~2>s*t)
zqS&8nwEj$i)ntg}r|gou;Z9iuwrw_Uvmtd(%Z&y$0SNS-^aKY~u@RpI^QL1KvHpHF
z(gZq&q-@PuAPUm|Fn1?3HB^|5@~k+~i;}dNL8r;$gqeD^X(>5$6Kr#I?dtmqt^J7J
z>vkUW`bi!#*-KFW93i5GYKZ73Xh+cFo8Y2?2^o}~jHz`<XTtfzH0iGGf@6tl5hl<;
zrKO=;*|ajWvY8pQX%>epu;0!{X7k$l$lQbgB!!R2A$}QY&7R17`;%xQ5x{xc+R09+
z*#hRCgUmm4Z{*EB5UZRfwbRQn`xcl@dJYC>_RA5aQxY3;KT6?_*TMdqPN=wM4KfNx
zBtQ)xTi(8L9bWgBs{xyYz7#X_(sww^gm=a&`3tdHPq3M7svobnl%rC0Ed%X7gbVVv
z*~7bbA|4p1_2UV2IKY%UaGpT7n8`N>ltRVcYhnn*98isH;jjKHgc4{ty<Vdm{}x^3
z=GEa~s|>b1jOx68-cH)cE#%sY-f|mS$*ttsF5!e}8FvI*sAHQqK+f&K@G9pltnLc7
zUo#Eixt}9%F&HNAaBX)m><vP#?;y6r+hnml!H!_qji)yR)xRUy!MkkcU9>Ca>k)_y
zUygsnAarbOb}zcw1k!j&3UA9RX~6c*m0fN6(fkR{Vyd2O)sR|icJh1<PNe3o)LfwE
zXGq@{>>_=4u$%O4r04X+ns|%@wS$*+^lts?kh`Ka+iA_Fb*+heJ4{cyaeJ26x2N8p
zjdtzi`CZ;FJ23WlY1;^I+6~peht_ER7EN3G{lOkYUW49l=H8xHbr&yqv}o_z7NO19
z<L!GzPAc}U?6W@D%d_?e`*~Lmzx57$XX>@Y1DtN{Ve*%9ZmM1H{p+^Y=0BS7KZx+z
ztvLN^Md+6mkxV9%D+g8%t{h^NZ==NrgG1~(jONP9h<7j;!SoujUvSVn#9Ao?2jO6d
zk#Z<XIaE)XO40{WeoBo@>8R%s-tacw5JgJM4A5B|n&0LFmom58o9<lMx9)us)%)7s
zM{{#$y@ie5KddLyGe>aYv?~}HcF#TfzxCXuzsGZr@!Sr2_|v@Qah`@wmAAazY!D6;
zJEHHD#JDHWOC?rrTe*FMH~d8VlV<k?x65}StLboXn76u<r(}7`@70d39;+b?UOj#d
z5{W~y6Vb{85^(=>4X?R(d3WPAH~TobBComkjN+)?FACA@R<Kurb{zOX{+QEWcem@W
z{S({{J7tnt-;<8)ZIjAuR?y%oU*AF%NCe-hnuKt}jx$U1<D6h*`#m<bZF8ts<&+N}
z72XjJb{hN-DE2`Gav<S<6G51JK?D-A!t2Ng^jsri->=vQ^sFR1?3><Ym$%j3{1(4j
zu5D}|^y6hy9C#q$4QlYeU3HlZY0yzo>`XcCJmaXi>HltYu?elJegBms;~qzwOqTe8
zfiCK$KSm8V4%8d!Or~th%godamWxw<xm=uH!jC$ev9D)u*@$p^J3&z|f{qc_eBsvi
zyR8Gm-dZCgtEI`QX_vipBYE`Xl+*SIvmagVZFJ33l|}qM)?hH!^oMnYy4CU4;<B4n
zWB#a8(*5WLr^G}am5a}!Z_%xApxz8=V4}=7HZ}6Kh$^ehB@-Iha4zeSf1u&+X_c3K
zId{zdw`z(sX(h>q{cmfe8R;a3?*=4`|LsK96GgIZ{4)NRPdL}LBA?fIVs;`BYyPb6
znjq7Win`RoalZJ(#fQ14%a@+;eRQ;<tNb3NWoip0e|o}^n?@#@XvJi$q7|KD&5Svl
z)TfD0WoPR|rB{?so0R-NQjpYTUcps8|4)==4B_u8CYiM_nV!*;&NBXr@-?|f*_il{
zEr@N2jT~{*s;7jHDfmtWVuakn!xJNHk5=ESO0xxoYS~SMzD4FnU3+c**pY-PY0dCf
z!vU9bW@>6l28Wi{e~p&iay2-x(uLzU^YV{291WpytG`9N!<=rYZO&JT(bdvf_=ohm
zMr#{j$u2vaU#E<hS}-KZY(#LN)@YA>OgJy$UCXU66y`am*eyD9GsD)+$*`k7|5=94
zhPJDbOH0$qmYbV8O^hW`v-Suz^dr^iL=idLwh3<2+Gg}?<b5wQy87q5YIg>ql*Gf)
zU+))s)8<VJ=D=73ScRi}mE*9bq!I}PRNDED$7qCFG9PKW*4BQn0`=*r8mpN&D*c-~
zZx+}k?{<t0VYQgyFb~~H#5=#V^JYJ-aE2&y(CyOUQomS8_s4hs_e_2dLb8*_r08j?
zt>0Q8uSl&adJ}g(pK9&a6BF{nU_L$8Sr9kZZkuqXZ4bm3=p!&No5P&D7i}<u%aurB
zZJPE&r2La`C)k+GX0eltzlN?1e4enfIagNdSMki`YNn`cUZw+)qpbv%!`Mig^j22H
zQM`s1TGX3xA;t}-qHC;S>{?Im@t4gZSMT{)wInHYt&ddvz~Vy!OBl5WyGLyCX9sjL
zj{M|$`B$}{!%pbs@F#a-B|Nf3A(#s+@8|q(xzCEM;{o30M$6DJHowm2`0nN14WGcy
zXI5)|y81k3#owhh^k#zP;tOT-q+eWGEIwnpx@Vp*RwL(?K5RAqq4Jz-<Mep`OImlP
zOiG6NMmAI?@*W~rb9oTtxg|{VFR^>)yY(0oUm0lq9Antxyz4!>99J-*;DZVb4!%pV
z+ZC*;f@=yEb$L#K)OHdm0v+lN0Qng=eoi?*P7oGcXOyt=^1nw3S#9Uh)n^QHs193>
zwi3UW9ADH*6vym5uxm}ZBM~J!)(Vspd$)TMhp9=e)6^Gne9DShZ|@5}$AezGi}$wu
z9BRLik;!t@Ic^K3o86MiREKMcRm@-9UJiH+yhmKa;hi=`;&v@lgA?I&RLx#?b3k8W
zzQPqAAY|GysQNmtDYWZt99#n>&3%)n$Odo@QVu^~rb@5@jr(s?*Ze4rF)|5hqiOpL
z7usl;rqD2@;jv#(_75w&*^9_+;Q5oD;Zv!2FZy?F(AEE#v~9e(^Oy&i*a?7={l$7I
zqeH7BoTYOY4>@&ubky2)>XiR|j2HCU3_tR4ujac+nK^Z8l)sS#nZBR9Ce9s6pCTD+
z8-7SP<%q17Na{XK_NZPXuejhd(s62($LW^JIT!k*D1l>O8xoe&{@3djt+LanPn~uS
zxA8o9*v;!7P!p~bgxL~ajVIZN{F2)G%StaOX8rmJUH)qYmz4Ht#XhUxGYSm({|m)T
ziuKou{Wk?kJ$q0=S(ipV+^Nf7)#bE;mlQmtv_}-YS;43R87}%is9=u*Q_>m<Avv=n
z42<%Sz{(e*GAH}KM$i8tUE1QY;8nHRmeD7<F;9-_CrHZ0Kq{F!I;;P7|4r_J1bR}?
zQ*R>P{iR$U|M&UACVb|15X<Je3%!{;mAm(5=1$~d9hm|$vLt@<yK{ZH6!!;XUyE_1
z?(cH7c77?dvtEDa7c=*?#7{PI^nsANa1}G$yc_#kPP5wCkLZSWOL^Az0UX$;dYqrS
z-~FZy##mDRTFS^Q%PB<I91DeC`Ly|G*FL(uKSQxL@vUe&k%Rh;PNT&SZ~R&e#N?b)
z$sHA59c9sG&co9n7?!#%@g!(uo6v}Pw@i(OgH?mG(pZ-o{LCo{OZV^qa~|%rgZsF=
zQE+wx;&=S-V2FU2FDjc&#~<hYaiqGYzLFRu=S37xOm&jZjA)HWj$FxweY$B5z|EsL
zaL>`uHZN?3*yPZY4rt+P-fQo<C<OKsZE0Yuclhk)EBg&^(WA{>h;9CQL>D(;=NxUP
zgOF@sIMs!81sEs4Ud}U;-O;|teeymN#F5w}H%A-OQ9~#<9{AdFJL^>_i0U1J!14vd
z{hVpQwe!O2M>ni<dCdEgJws=gzI&r*wANvzdEUCxSxR4V(%0&Y(Fz;1cH^2TrM(f&
zmgK2xHLB|yTwTq#XtsY{xdP>WoVVo^Jv&FW{+^PlTJ7@G&%w9orEP*4(3ORuO*L$8
ze^c|x^wIxZ6D;8pb}Lmg&euFN2?cqT3I}hz*01QM!NI>&ZNIEb^Qz)brg!FrP)n_w
zQpKeE2Nj412osYV?!-N!>Gn8}GMA4#3jaZp_SDwG=Y__x$d0E53xBCJV!egGmaeyx
zDVgdh{H+i^v25Xn>5NmEFLBINeG$|rYP4||H7RI;V>tDK4Uq4amJvS&g>MNp1t|8b
zrJB>ZSZ92D)&UPL+dd75G)8hAB1w7XC0}*cVL+oR1t624QEV(#PHS_w*KbfO&uJi-
z>y2eyDkn~K<-5EAuh;9lqH`?RW0Z=GNi94WMhd^TEt;^R@!ojvdVX{y!67*d7b3xg
zyjJu$Q_eXKK%bLOj0O=eNT-())Dx#%y_-OuS6D<TOyAPd_?eh3;dg7QvLIQcMr;f6
z$r)j+gi<21T4TmEf!0qlBf?FsZiTH?=4bWSH8cJdL8PV$dz&qGq?zBy(&9jW1BmsU
z21XDIBWdGl@b(&+M6d0g2Ha{R^z`oPhsXH7aVG(4vbrZ;U_bA-kt1k%;ssQh(`VlP
zdSdj{b?4F9JlpF0zxoI`fILFYYh2y>zppi80N4V9!k?p4zJ_;tL(&W>eCxh!mm@C_
z>)vX^9zN9ukb9|1K-n<cZx9L6)t+dar8Z);Q5~fanr1lCxxaA-ZLkzec0XyVCGJcA
z%a#QZwg{aOET_yrPzP91ibl2|1RdCu!q@?V<2uGghoIrX9wv<^$oq~IOHz!!8P#OQ
z?8iF+{8_$TUiMWs8epO{Nv=E|?&Zm+2jx(PZDnYk&5pDu=$c|7xYTG<fjl6ZwABQy
zuV{jm)W7f1G)$d-_VH&iaookbH>HSBiZ)%1IxFjzO*-|GQ$rfi(`s>q_HWl`ty4If
z<mnzw;@%eN%^NDcM>*9HcaEs0z%2zNVfrEp=?nPL<sjsil^W);uQNSoRpJj#C}hSP
z`JE|g4HS89DzOzeCJ=8obsU78nlOK$<B%&L7hq+;LCP1;x;Z3`rER!wA)R0U6}^dw
z@itR6^@-M72jA+8te&8&MMa5~zo{sDrJ_V(={BH%tToGQj^$)W-BeMc>Owu?MC%PD
zW&PSIDF+bvOG!D%uauOs{LS#YnO~UlRZQ93QA(&Rhs92bu|m0xA(>7*ASY@@b_8g~
zeHJw~tiRWRTGPB_toB}I!!rhFo2ac5wdPaY?=z+3hhYEJ5~(2d+tW~M_C+b&@p@6!
z^nrq#H5{@U`cZ3M-?#>E;LDcJp@Q6m!tzt~Hr3cv$@OPhMLY%evd<gjt)!aUTg%hW
zTY~<E0&{Ck_S`5uQDEMO(oq{7ha&~%AKKMsZ<7?5l+M())NB)^sW697VQ%J4D1*w9
zEs&@%cd~l7k!M$cLKBsxHa()i47Pj2<V2mhx|>#^n7rmxn7>Zmy{7KmQrj17@wNaD
zb`sm^ZME3`U}vz!jkdQ9rg$gUC<=KSsW4x}+I<JA9jC%Pfb~@ZlRK#}cbW=w*Gh4n
zN+(`@Q@cus)<mjGYTiZ7QdPc4D-Q-m-eFH5MPQNi4sW-q5dS;(4)LT;$jH4mZ}(k|
z(R(j`D2}z)9_y)BYN))f%KSSo^`?KNcAL9qSIQ>ksD8lk<6dCczN_&EP_3Suqx4=Y
z{e6wnTW4=K_2_HL?P_S|Z{%{%_=U)_Y@fIPN)Ban<B1wu_>bmutYyt>dhr0gxYsau
zOD|&1xX(Li>c9Of2W;H-(<28_MfQ1z=#P<GDmwqBt&gSXJj_$1Y;<ZkRA<Z$so;+4
zYihXLO$~QwWn|r$C92!o>Wj@v`N=V3%!u29+i1;E#z#KSj=`%ZobK%R1XF><PJ!ls
zM9?(_T=bWR6bd+&9%+G})f@364N8Y%uyO1J=lcR*b7BmpH>I5uLd12xhwtp<)*$5v
zDf)f^SryMUEaldL(WacV#b~a+ZPw7mH|Qlks-U6K^S@D--=x3{XuesoZ&6?r{8!f6
z`QNI{-=^Th!d2g*KybT>>Du~%j7ISds2^I*>!bO|ERW3V8{NPd$gi2%bzV<6^BF^Q
zO~OT%cfUX%Im^4dZh>ev?jalo+j+?Tcj_fxRq)*kRuz1Yg6~z(P+a-n$K}YezlX{w
zOwBH^Ij^STM(RJW*U$`T@k4YgWl~dx<m4*<->HsFiu0!$H#1=RtlHPEFdDhNg~r%G
zL*)N}UeT08pHS=v73gduQ}6hnRP2WpG;}xqkLdE#dem>|$vYGrQkto4{;gvFPQmXh
zkQT@PQN@hYlLp8ClrDcvfz7)gSL`Pgd_8w@mN{9f%`f^tsg$2mV4KE%S}`;D{~5*p
zi2|ENf1;Stz!GOIX}L*E)y;M==}r>yp_}rFDSb)~Pnco=<zcU`bvZ0r$NrkymYtaE
z*AJBb%Gws`RWfKx3h@I*IZ>A99U>HGAq`Bcre&@E1m{Q55j0;0EL?{Z)d3kT0?I$s
zYiqTRwumPS8}h2+4O#?}Mz*T2ZY}WpHA+0Io|MC;TntxDxePwHGn<@{ZSYIJa!hiw
z?URnKs_l$y<4DVt2PKAB%`vrNX1jEZ<>i={W2c)KS+TLWEV3gUJ2`l<)9d$oyk28u
z3&zNH0Nn?|$k*1X86^Kpm2!iVUH(X+SpU3fygE}Y!pq|+Woim<B$Zlm$`u&Lp>1_(
z@{BnlYG}-h%35BmojGX==Vnc(7vx%92ME+FkcvjsIo04rDXO(>PATP0WfVt3R{VNm
z`G_%cN7V+WJayMbHkT&m7ADVDj~?gG?b(mAOKG~cNNb7X$D=L&vKkm|oYn`EUTx#x
zL~%E`^q08^va6`g>=;LkPl%&ByQ~8n^P*r7rDfpXvl0L9cy&0o5+<dDQef%$M$}*>
zHhh4wF$P5;hh6VwOv2uRGQ$i;?b?3`N}(X$>sq4az$Z>_`)1l<CmSx0;26refU0+4
zF>=*2%@#Bcy*yK%Ji7=auHhg?(o6@88^VIHt6t8!-v3YZMwpi?d%VfdS2+t%jF8ST
z6aukbL~O)vO_Mi*F16<ZLEU&*)%dTgu70bl@eoz&|7YF(3kB~|iwrO}UTH0#CF$rN
zX(b6^c6Q*`6X|IEPp$;){P&2zI<<B>H$)GoCN+xnj~blgE7?P}Mk8>bf>_TuCC+{G
zFbWP*P8kX}F#wk?GcSyI>#1@h2c`Nh1b_*C`((L|c`$~;{&Ai1h9Pg}8qy3XK@^z}
z#?I|Bzl4mke;$q+#VYll$A>L~Q>|9&*b0v>{$8^}cmJQ-&h0m@s|w@xjAM^y?D6;#
zr@7RbG-;fvs7oS>pj5qxwjye4AtgvqDj|EuvB%hB=ghb*qX`Xyz90nRk*ci<JfI*I
zAs&D~pbrQM1mRLTP;u);?Gvp~C?a(DeQTfFT<lybY<c^feJ*>SbN1ffUTb}Ot<TCB
z+8D-2tMk+RA9~gqM-_=qG3iog@8zy)IVD;S|K(7if$JI6sOSftf&{HA`9Ey5H9QP9
zkl)VkZzQYdIWod?rCB|isG#u0Y#Z)V6Wke4{zwB!jzFQ=M&`@f%Cdz+8n(us#W@;=
zT02t8v8wz6=Z`lsVXQkl(~kU<Be*9Bj!0`|y86j^&t?U`<tuo<SMm}`U2ZY(IiL}O
z?wFsg@1L~*4QffsWSu_)O?ep?ylZ(vA{{HPEL8B2T`j~tI&D-4fLAD$OBNo623X3k
zFVu=Fx`UZOm$1ufsBxvuJ>b2f8vI$wUzGe+$={UFKKkm}hs!MM<l+YzBJRl7$3IoU
zXA-UnaOCE)U|5^p;_te~<~*-z>op~xQ}VhJ>nrLSF09sDY&FxE8*vf5R>o=fKXlTn
z63s{oqi*)`wldUOz5A5Bqr~oRlfi$oOMuX&$9<=*f6(DeO5Rh_QeseVi9x-3>m_xo
zd)Scg4ViKU<1#?7#^L-_QsIF_szBR|>!!r{93%=>=AD;o>~!mh(JD+A+zP@xb$7pV
zg;)sYl8RxdX7GF33z4^tf_)1W(P;Z1e^v2(wx?j9o20$U!A4xau`3{7Lo42`{wn8<
zY>8gaRw8k3FDtWmHKvh*btE8W2m4OvF%4zY={P1uwx)zmVs;7*#%IVEF&mu#rHRHo
z>}xPaN@@CT9Sxc8i^;(LL&uI0v;8BbQ8XA?r1|I$(qN2~#<;`Nz3))U`FG^bJ+4pD
zAE6>c!S?F}C0qM1Fv{daaHgi?_Eh6$7WxV0p~2YdtNx5LCea(%xIwNoNE1&b&X|u?
zuA1N~5w~xvO`-SL2fa5SZtt&Y(I6|o8N)t`rb4=kJ?K3!7ggV~z3CEesZ1W9sNGQ8
zOQ_%jruUda@3EWb=NcSB^{qn3e*Q-E9tSvQ#+*XQA=E)>J)mMwbLI>r@4VF%vb40}
zXJ6*rL#89ySG%dPue1kk$3eCZLdn+F;RYHM_r;-Olf(X`5v-;6SV!XBk1V~{BzE*B
zXc2bf`QpGc_#yB~+&7vAB!BKumo(j9J=HEv^9BP=$7E@SHzrNdI`zG|afp_3xN(@W
zr5PE8Dy{N)|IZ(#B$22oUDRjkF<cGM6pWV+QM<zSwSz)^v~-hc79cp=o0+AgLZd)Q
zhbiYsw`QR+d5U1~Xik3AR(o;_H%C&xv1m<gtKrYfEBOYSCjUF~-?hlpu~~z2qFlNr
zDI3|>>SERCC2{u<U<<QXA*Wu4G#X7`j96eJ{riznx8*fTI4#_A)#9hPiQWYz*C~j;
zsn6Rkjfw)y?C(0H&=gXrS>H__<sQu;a^#t-Qzq>|CJ9!%VYpSlMw4^ozZpF8h-4&u
z+DZJJO5B9V>Ii>3iJ(jw`4s^#Eg{zgSP7v6CVsl3CUjm-4`SYWcOX4+X&@cA-tI=Z
zNV&qAx;OL%OseQ$#DuUn-^Pd;=@dC%)D1qSL~~y^lMr0TB+#9O&)2#3WnsH|i?L0H
z63IjYTGqQTOIo=-0Y4<sKlNWvXBRoK?Pr(2SCuL(vZy*iv@~LNQ`uC9LtE;o<zWzr
z+fpr$DJC8c<95_+U`IXaM8`<_{Dld95^9!k&dvnQ+p0*$+=MZP`_F)F?bG*;X_6<V
zAI;0M3X(wO)}3HqPcSj#P)O^qx`1FybIWtg1goFdZT(B}e2KTwCu;e}$_sgGw6}`l
z;e2;qzwDZ4)cPWC7cW|rhl6C3ioJLB!kOQMKYH1cPRtT~f{nlq2o@xLf8YZ#i4v(&
zu3zE3?2$G*vH~X~D{wOxFH^#uAH>!~Jn!GsAtrtGu<usX1!+Wm`yeI3M)7UConI^z
zu;#lhpOUcjoYgd4nmR)|Lry^|g7S&XtY;t>DxZEOW3^j4v1_tM8|fymrzF;)*)vxq
zcYPe`L~**ASIP3ZiYzUM2v|L_Kw;;UtA>8>0q`%dl8)kL4Owf|;*&rv#miQMC?%^|
zVxm0C)nltU<~*bL85(pBKNYqPM{Qh#a|6wdaL5L-jy5vH5n-FT$5F1sg{J7KoQw69
z?}aV3lzt|G2d`b;myq<vobB`s;R3a1gw@P^{lhiIbTx&D+Ng{7S`8yvR-HR5TClV)
zQM*6wwcA(Ue6Q7NITg0)tc`W@Hs7`|?dYT(W>ynEOgJ~x(#Pp(aMfJgEnKjM@>T8L
z&qr)#jLy2bv4%O*MJh2$-B)yiNGdy6WZk+P(|Gm(8t)x-lmvcGey)vy#p252-k;P;
zJKCm{Z;z{Os*c&Dgq2(?o75mK@kH?8wsR(+$Z?<|q6HS%uJ#*RVB1vn>W*QTyo;Th
z1=FOd(P&ZdB-9;QNi1Ho{1GutR$y7;AL3dK)~SW$WO|!mvX^9Hh4*;s-DiD2GLPe*
z+CI=FJ*SQ7+kFw1H67LUk;26E#~trRxA>;s%>?I8^ZbBVi>2x;mK|(GdeW|U5W}y8
z_b*jOW0dzA$%QVBsKJvVlS#YZr%Xq)+!>VgjLSjYd{BbsI5><}bnkS(r{hJc-S!j0
z&_iU$VW^ZbXpwh^2!b3eVVUlNAZbS1A!xK2knK%hd2)SWIi>_X?qDcD5BXJJ->7U9
zw{RluG@SXRD$Y9136axN_BT`OtbM9t&fk7;pW`w7!Cf>KE$;URc2r9YjHm^*j!qHX
zAo-Y*BsuR!Y&9Q-aBDC2_9w|`W=!4cIEI9U)U!xSM=;1}#8+72D4*d*L_`||qSn;W
zkW|2P0RV`=knOJ@_E%f?cwgIrfLz0cAaEy7z<~QcLBK(R36}+fg`pTMoWcjEfkFrY
z_xFGRh0lTtgKCvwoEU={krY_70M!M80yGNIYQz*X{-BN%fNB|^%HI*A3}4CQ02T2Z
zlnXd*#On<JVjS}^{V$E|>D!N(4!>NmG2z3;ZCpPo$_~PR95%*)4UF#I(*Q-T4K{KH
z8@c)V?biYufhK<0M-4-WAE?y~EtZR0{5=3u)z&l4DtB~Yt-M@8x2v$DRqr9b7ZaFy
z{xygE;HJibU%q_d?TZgDe)4W->x?jG>I#@6g_fJn8cW_lCdr8BCbl8Sc>mGhBQ||=
zV1c-uO5w~$7&9IdiA?_$vVN`tW4z0{yMfHG-+3q^*rc58K<B&a4|h<j2y~>1cE&Gf
z5<@!%{77;bygV}`Cl&|VLFmX1eW*7OQK&HmZEm>9PB1r&K!~Qy;9)^9nN2As4w`DB
z-4jm+V4~47O44WtR<S<Bb7-`L{z1pgLK<tdELgrl4VDEFK$o;%g07SgU1|@fy_CM!
z52=TJ{N^rug#l;UL72`M2tR9DTNx5g>IGq09UoDmSkRqluRFe|7oVlbiNq8$QC>Q)
zm1PluyRTjYQVh}hy?)(TVCJ=yMDnJMdZrH5JoQ^aIYOuGz#yy42J;}3juk1s?I%R0
zPm>)-rs3ppvgCvzUqJFnr`YKfr@NFj<O)cqDVJ+`nbL@%)Ues<(!Bez=1iMCzf1J2
zKWgZc=O>hPgjSH&Zp9BFfhEC${4ohQS{t>2${dM_6RI2Q1Qyr?z)+$ac_Y8DwXA3z
zhG{OUq^*XUF9y%57oheSPK-o7@Y^HqlbIOxHwNQ!itG;nTGD3Yg&JH(vl-ChnAIQD
z@J-ttl-m`vlVBMOZ38~+)n`tkL8gx7x^(YEap1v>d+S8NU<M_?QpCCH%GQ}PYfE^n
zL)GtsS{ixK^zSzNHCwLehCzEYB+T)Ry<e$y=um8$)dLtWs-3tQmh4bYDPd~@FV|;x
z{#+qYZp(!Aey#E^D*26)KhgnWE!G;V9^n-#;r)({`iHJi<W<sCL!7>k68~Q~?FVg?
z2=@sv5!xwcwJa`HvkG{}mP9_d*xo53EU;CBE?7NjIInYL2$Oki*nHZ_FOI~#qMXC}
zWY}yY3w0G^7K5O>Tb8@kz-7pX@c3&qIL`UD2(eja3mMen>><n@ui<vpP&p-3rY0D=
zysrmB9p3y{gLrNXQ(32&_uTek-bbzlueXd;PBXvtOf3%CV$;@-Ear>(W1GkGD+T_Y
zAdaka@1rbsi-r6Od5LCX^EiJe%)_zwl3I|7gOgVc&2)r5&$@%rVt%aM2(GO&Sr;)l
zHVNSU%YBaN5(4wK5d5loUKa$*97Qiw57atD6nZVzAp*pTIH$F@a%L>2@g@fzyFz#A
z4Z|Ap!eQ?N>~zDhX7F)YFed$BvgaGJN>iO#pObl)P2?S5>d<@oR2_qjjt<f<O${>v
zkr>b`T%r0g;2UtmH;ed$ec1mOR8X`GBNF7Uo!>G>Gx?>yhU2%u(P7<Zww1~l%jFQL
zsWZIq=~?gftR%ISW(rU4#g(~A#VwCKr5c^nf$_~R*ddId%PTdx-liF^J#vFP-`4fc
z!8HulwVLBR#VsvYh??xY0_gj=P?f0i*2;^XaLbblOA#R|IFa92t9n$r&wE3OM!ZMH
zt$@eMT9rSyRV#Cb2>Z?1uKL(fu!#6Za9N0+yhx1Ik}XywB;NYc@`6MwFQcndYo&EN
z-y2a$20WU<m|FzLZj)@}2}S&cWV!-V^wt{phr8nG#WrW-xnZePHC6}hx!_mRx@6kF
z1bg`0!d7j=j|JwuB8WrLJk)+upH&l-t#{z<RlW8pkw@?5-q<}u8qS<~A2^`nE`Y=w
zI#<BPyHQy#w#D38-+O@P&FF43YUQ>f<AZk6L&3*>*q+=wu9IG%(bzg4<4Bo)+Oz{U
z5p(S-n!?$ZP_dc1(>=mnT;5k*BpK*SrJPaQtx?vtGarofA7g}&NkI#n_)Vz&%CRK|
hfx)SPz9-#p?w(-F{hzoycT3-sxeR|}xf4^j{|~6Nsxklo

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_psample2.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_psample2.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..812b3fb45c7389ad12c93071be013d929339ceb0
GIT binary patch
literal 58248
zcmcG%34C1FecwC#hQSVk;6B_5iMw`1mMn>)B3p)SiMH&q<S~Fd009C6@Xmk>e9=fD
zZOcg*XM0VP*bV45X_(Y$^3tTQS)HY?ZnM4Y>GUN{Ui<1Q{q!}BlfH(v#eRSPb7uu8
zyX~jI2M6b#d(OFM{h$ASmU?@05q|#h$Iq6JUyerpwI14kl03Yfzv>sFk%$*5M!cvO
zTZk5;_CK~5FUF%TEm2HbTB?}le|#acm@Q`QU1A}(*i-CT%op?aoLuN#>?`&y_80r@
zIkhmbI9MDE>lrEzQ9iveyf|DOwzSN`$l{jb7K>*WMi;jhw^}^6Ft)g@xXt1{3)>fW
z6n8A{Ebd&~Rou0>ySRICPjSy;p;%blTim<2uei_h<rnrZ9w;8Lc<;i&#Y4qIi-(Jc
z?YVE^$l`c$eDR**J@(wcaPQ*L;?c!p#bb-di^nZ(VBy5#$>K?i4=$WqJY76Ze8?O2
zM!YSbju-Fqj(J<XvFp*|{oXcj`}Iii0dI%5lm8Ds8}W8|yRS#Q-R1QBL*>)g62*r}
z+v62TE0iC;7B4=+^ImTs&-;{zlt)R~?;RlJK=~1#zmDgF-XWe3+4EyOANG#$e5Cvs
z@88DrxOWfF_t^8}Jm2db<@u;RKjFn*jEo<jTxL3Rm**C|!s6VO+KOK;EY0N3*J{hv
zM^BuXovWQ+nL0kbw0L4+WqQu@C+8~F6JEKpG*_Lg%pUXRW@c8Zb4!(D%U5el{`C11
zQwvK|Cmwp}<bzY*%=CjZ51f8z=7ER3`|g{0`2MMfrXDzby8Pg&@`L63PCa=4)P1d*
zO3j)|jhdQY+4%1G^3~0nxU^iZOwOH{TU_>+E|nSoM(bRw%8%TC|HF@*e)ztrsfQnW
zaB}+8{SQ6lJ>Yq?_K~TnDetuR$Rj6DPnVk&v_5=dc5<a!otvz*lJ3%mr<WF&-#=G9
zQJY-U0M-1-N_A$*Uo898+_7T?{!h>4-q6OH7?0L7&rdE+d6Un-FdnOC{qm$w_Steh
z<(HSslQk0irk5(U@|D`+WMy);?ALoLD+>!@8s*c=lYVV(a)D@~cF|kpe|B;5Lb){S
z&v`ubQQw8qYS~{ZEman-l9GFVu3CF~sWLM+JD#ccEiBEJYFBEd$!ckCakic+U*UDN
zp7zQ!lPe3gdUCG1IJsO;Gu?|cfZEFp!py>Ct)8D=TB+1Zv|w_vTHngTs7+RAOs!m+
z@s}1$v{E05)>F^V)t*_LR2M6Z#N<Nh!a{xLnYqc?rOM>Ov*E~|;luvi)Jl!jMAo6_
znE%>b?dsV%zgl~-rjPeNxl&sySEiS|vj5KK$K&-O>+4jxHd(4ppD%kW3+4JSqvcOd
z*GhAh+ES@ju2h%&dQJ<rTw0jBP^LjWXU;tLt`{d4mlw)Dj|0!CYtJvuGQjVeyE0d)
zXQtKqX*$_M%wMWjCu_AzJ$jXIrouPbR(gEWTRb-%)1-z!inFgGPxDtz5Cjo6TXc2m
z-IvP#Wq+<#E-ccEh2u1Ju2Q(nHY?Q5&s7UEE0t-DK%ur&Q0EF&?MTaYqI_v`!HVSy
z{CG>{YN4`JD@>o)nkh_Ht}ZV5<>R*^j9#KzUYMyT^bP+2pSR%l2Tq({S}dQSP<d);
z>B0$bvNm~Q)?Zp#KJns8<$2aCgIGDC@jY>dz4O#_FPvEL7Ef4H+4jxt<ax`BrRA!P
z;Au8Oz0aGgPEIY9y^=Ncpt`LwNW}dk{OcH-J`&RW+1NbKDAgiCv=+S<^PdW$>oG5S
zGwMHFi<6e{qSrWNJ{H>+xfWfE)skL3h+K=EiM%qp7Oy3EP6Ux4el3128pJNd{1dfQ
zds>VscrkKuTP^J+gUI(qywqpN`HdVYUgkRAT+evfo6$3>I~qi4iGZ(dkE~Iri;qUu
z_`+IhHnNtkW%ZUXzw_nTT6`@NB!YO5yb$pxy?7<+CALP^vOzY;kmB>bT#)5^JwfLC
zBA-d<n?c5lUx@lwz4%(Lk_ob2d@Q20vB-|dTF%R__3)K^Eg59&EB$M|L4G|Kq-s4u
z&$Z~*$jxXp67&SQhazh|7e5)KgFLZxkP5Q=PhEU}Efu7%MOkCLljm4(wnCl;ps>&;
z3kz&5pn9P;w^**$%F9))u9a%Jpnb~1DJ=POvvcglLUW1>fY9l3wR+TQnw`5;W=Wne
z7b+`@Q)R!ivdiVdWp=(Qt!c?xt>ud=lM4%1k5%VZ%U)r52{5%(sakDn2~{;e8h+|{
z&fcC~@(blFwzQ5G<}9mUK3px_ck*Og2PP|?RZ=ax?P9Q~T9{<BeCX84qo*D?divxm
zRv({oO`j1UDV*x8hWr9M9&M!gQ)S+_zH+qji6y>9L90)n=qmC+dyz)VW^6$9xs4@G
zpA0+uoEr*PUxEEUX?0g-`6tLu#9+fd&zoD^XYFZp<_b-=a%8fG*|2M11>FkElgrE#
zH8RiMQ6||d=ma33`Qf1HdkYIom&;z*oJS3qYm!UNweo1!ytkH!TS|bh_Sv+i2tIB6
z3Flb@?af_c%nK{a$1Q*O-LPH2UJjP(qlGeHkxsN0uv>lAqpY3QNCA&)0MW)UZ2YdJ
zfWj<J9;=qs!x}@^tZJ^}%>n5()ar*iRj>xUMNbME+px9A3s2g#EtjV`hzrw`)$&o>
zTsCJD`wExoL%A^VR@K?Wx3*Tt9(9dkiK>N-Iex3kTGJm6u3rYNt30X^D;pf>p29}0
z8@6N!CtN-Cgcjk%8#>o~1*@aelNCFG9H3#>PS;pstGjFzt8VE6UO0)%oULxTmZw#=
z@neSs4=?gp9U++IsC*^07S*9}E&A=Ti(sjCdT~(4sffcZcKl>LdZl%)`G+Z2&pjqE
z$22cL{v6L=Qq3cg)&14w$?0-yIUj#)VQG4Dq53$HM*iE;qpRD(zJx2Yu|<GYtNE}y
zLMRGfq+0tK&)0K8f`#==m+Nu%Vm)WOwyG`b-%GA~e4$*aCmm83XY}j2_fIaYl%Mf^
zkdkC|d10<r&z70_vQ}k1HMtB5>rsLo;EngydwJDrd$F%`hZK9ga&^w<)U^d$kIoe{
zg104>XFX|nP(5o+D$RLU>RC2>$<|h}x1%D^1xgirX26@sAaGbuPp|kYRqPQ=uefsc
zXsLVH`@6>X9=ct9J3($V7afWA@i!8U-HF9=(ZrpRcs`nsCiop7H4}~TK5-`#Pelh}
zO3g=yqB+v_=RZRw)54t^KZ=VAeVV`OP67`9=osk3JjcHm4dQdrZ;g(SwpGYhY@(j>
z%9rM*%Pg6s5hwKoGvSRV>M{Qmf2Wy&^8%A+sjItRgDaIe;qBnfC0EU)KC4x=GwMIj
zziIwDes+gX)o){kmm^Tdks|H#VqV-!%tpMVmlEpXWxbr&LoDz0dVOBP>*sfX|AYJ`
z`R(V=Vl%rx5t)gKro5gl#<S6EpEn`u5}Gm*K8<=?c}kX3-n+dqZ`*bDOfg-~tY>dV
z%DHQNb0+3(_jXu%kGIp?1^t=#c6)pH-|H2;z5MU<_IdmH-|rpp4)TA%JLDba{~(m)
z5amThCW^zZR_{GVMd~{vVaj`<w2s>swuC7!cqhG6mNH7aPJ8#!#;xA{-UHM+=Dp}W
z=siTrHr_q#Jwj}|_o(-EyxZYD=Dm&oJH5xfC-}e1d%O1}|96M=KIJ`q9dxa@$2;Ra
zLwdnG>pjc=z2W<F-aEW^+IRPP&wKCU&HhhB5|QEo*2DYii8E*3^R74kU7UTHT9~Uq
zZPjB7rzg&?W<ab@EP&;ogWeG>@P?50H^>-y!z1|Ov8Q+b`~T;`?N5C1qa#21?4PcF
z*Arj-@~!{$+wYos^$Gu!-n)ta^C#+YkgMr}d$ds4Z-zhBRh?q>)FRg+>(QHhAlT|f
zFGZ@o?}&IYLlL(Ikx~rAH!`1?Pl7%w7UO?HC?vcE)_W-x^bw|mUcyX}Bg}#pMoPIL
zPuLS|Bh1%&gJ>{%Ek?Yr)=w>cT6;HQd^1bh04W20u{NkTpw5xfP%ZAIYs0k>Qij&I
z++>WeMfq|@2s5R|yvU03jq-lfF9&19w+7@cjRnJm+k!2G+d*i_=dT1Ao^}K~33mp&
z{a0(dg5<Sm_0@MoK!lT)=QGr_J4k}SroHS9I7j4D+CcR?m5cV|^yT(827CO^kq$ac
zZsq(Tt4}T7W8)c_F9buN<*}eY7zhTv{EZZ&*1J}y?XB%2wJ-~j0kXf>>s#9gy1k#j
z1GW9sb-<ABF;M%1Ylnh^!QM*||Njm4uOIaKgG0f-9g#7IfFD}h7aUkW7#tKje(~4Y
zwSy26`;CN9J%?(C*N+7IZ$@jw!G1`WAscfDiqd$no$#L8z4J$d9ppU5z8=02TRTkr
zo^Ol#qqXDnCy1Y9rbcdP=1x{pL1g`uw<S1yQ!}^g+oH>RYp1Vi72HS8`zUpyc6k2&
z^#^V$?LiNyr@!&hiSfAqK^9y+c7>Dj(&|ZIxeqmM=pR9KW!VsN5Q=>)*nNf_?z?g_
zj9l8c+WTnXg_VkMDh}_$>cFvOPMJA}YKs~?$7!5~l$~0tmRASfSuQUt>oe~H72?FK
zM~?$@d*)^#_sgXV<*U_tPRM7;2+7-}7s?!khH;e_Cxv4yi^#Q?h0Eo++4HsP@xnQg
z!w!F)Sy>PkRVnz8FOUrlL<rP^3XU0G3*|a}VR?x&`uOcLg?n2+g_oXuVdA-oXX(*1
zFFteb9MQtW``%Tev4!`&=ggDmo_Vow?%jprGcUZmqu6+=-eYu6N&JOsy|<IikZmH!
z^}GWn9Fd^x^<4O*b38M>1cp#9Rej+%{s$E3fV?{jw4Jwx0%f5|qw@wT)?+s#?TAnu
zObp7s9-m%bsVBh@mamS-{G*htC)CKUqs%|fo1f>;P@?UT)xCm)$4f9s=4z$VW(X10
zO;NxX@M9)(Gjm`L0LhiI|CExV^<EzW57eUyr)kmyo-WN^n)JsX;z(T)_K|5+Umq}<
zqFgKcbE}})OW;GxOd)trsWiPXS*?~z%3YaUEE5|5oA)%5vt_qh>b=tobITmT#=-IG
zJ>poES$bDXwR*qNbYXVZ85o(scS-0BgAZe*QeLRm$Dn~q%S?qn5BtejNUWl2ec()k
zr}2_6+zYhHmZ3jHH|uGYV}<!bOksVAV`I+`=s6+$v7Vk<nVEsPSMOEk1{cb&4>Y48
zBdp$YNo~+LfcuR!sNYg~VQQ)5R(-u6{0#2pLb;KUUR+|5$lqg}F6L`yuIBgZ^L;jb
zLfuO&{91jG$tc%KLKs(;Ss&%9l4faR*>AP$A&MGi<v(hbOfD>+hxa$lBJ!6!|Lc^v
zWka!2ZE~p|t1Q=(;!s|#XLVSXYFC%bEPl?>O3BT?|E$U<*slxz+Z2seD)sDp#39rI
z@RyX<->lvZ9jh70QqQkU!-^C;jG2K03Z}NSSnqkk{=ZlgdYn;~Wxp~ze#94xso1kp
zq01~87>%}AeIaH3x9c-mjSAdJQ1K5dFRac<&$yr^mGWO#&Tmj~Rl%x)k1DvL;Ie|U
zf*A#~3T#V!T(OTS_ymDluePpewLqQk>)mw)fr9TKC~DENJ#5GMpH%YSA`r({Q^pF@
zm^WsR$Hoj6dz(;3Q&6-u)=Vl6v3+3OY6=@OSj<++mmS0^<{LRnuyFmy)sWtf%r+ml
zqJB*+I-uaNf{O&|aP@Bx<n|6kQ_=1Gwg2Wp-1VG~?vCbS_Qt}Dg#)1P1EBLl<Av@E
zq32ic?j+)Rqu(LYhxnh1#$N4<jzv?kR7`0j(R{2gW;NVd?d^b7(`j1N_z@pbBCDtQ
z`_j)EY9*4<PY2PpbR}DZC;%Bq2+8oGH<D{<FBYW7B5RpiY(6gDo0s4K&eme<3EpHq
zkY#ZiZ^qVgL3%#<augn83hts1pEUId;ew9{!j+we6S+-DNcuvo`gRbwln}UDuGTZ3
z=dagGQ!{AUdY_j8J%cl6N4TDau4UVvbK>f?J;U##Mg0Ms$+pzIOI16CinP7ywKoHs
zyy<I|+0j|1zx|Cj(5lfBYB&gIGCcxP8sx^Hpk2*7)Z0O&!|4Roq$h*4Jm1!mA=eUG
zKNRG>;hV9Jc8+KqtY^8K5NVBC+sbbVdfYcHS7x(Z^TXcgg&2qri0}M}w{<w;fi%x=
zf#`$RdQJ7J2cv2?eAV8{BRmazV^(ILw@o$Ide*mUbglmFjVI&7_Ez><{vDS83EqL|
ze)aF%jPj*KZJW2#(D+_&fVI)<#UIm{^?AF9LvRVnrAE-e`5iZ->pO2o=*P=KVGU_3
zOx%ih102H_#4WjncxMvH%o}(3xqYt#+J*Czmk<pY*Aj6Jc!pPiVl9*>rx7NEtX@hc
zJp<HPLt_k*4|PI)11(#-(m?nwxfj)oeh6fm=Y+l3YcJp@mXLI<R2>}Z%u>){{gyd)
zStUNav*pby*wcEwcxsajhE3H+o98M>ijR~|3?ELceXVpgr1?%5#j<1-lS^OR|I~Nn
zZ+`lpeqDX6mDyxm(@W)<Q&!DpZ%*rtA=z&Lc_Mz1;EVk6^CwS)k;vQq5Al3^U~$C+
zH6hwv_UZ!*mlmeNGoEvFa_KDdlmq>!>bQo<wad<Ue;dX9PZ5Y6(>kI3cPiZ&OoJT^
zB2`r!Xh%marLTpOTC)X3-BFzi<zwj?PRH7Pr~SXFk0mtf@UE6-?WC)x7g>CZD~s-I
zyQR`UqTq*>Id#s7*s{!X)g)~7;fbZ%b3$Vl=}XxQxm|PB%4Eg=hbj|?6LqW4&Tx0C
z+OaKcvug{~lMeY#xz*|4Q0d`KmTEmCVV@+>^`X{8Hf9_wrES2w@#Al$(%;~(ssmmp
zSS}Xhq{_s^yvo41()l(9i;7>)wbA%?&cp4Td?PTfVs|pIum<A0Ik85p6ero8d_1Ew
z^G-f?fD<rJJOi66Px<Z9)$MksHL;)uqNu64>jCHAq;2Dc3IDVFcZ7rhga3CF{D6WV
zRPgr{{E7mNl>dJc)Q4>kb&grx?Yp~e&HB!_+PL*?tzFr*C+j;}uQqI0+pPW<)TZ67
zIofPz`kzy#kyfVGj@&f$wCzOy$JA%-ME}PX{O=011O1;+@XrX&`aemyVSD-iNEw9u
zwzifbux-n+xoP}=tb9MMK)c2N83jM9z|g{<Q_PS+n+J8*O@>G?{}&W|UcoOaP@DW;
zQt-<P3<tB|upYmKAk8%YuV0ufYZm6{<_oh!dMjurtV=t#`cnk2>6DB3dkB%<62{jP
zHyyh~DPT<Vag$zh&aFiKJ>XPgEk!>AD<%pt>P6lUGwdZ-%iFA$bVn_&+!JBBm@C&{
zcIZF3BmM{PTC&(#@^VDK<MD~rZI2dSgjFxa0Ue`a2!z~+)g5pPjMt!7Ax+YFIi7K3
z!>e>2au=Xe_Jt_}z#D`P3KY{z%TTCGepPSMVE4)gU477=C9@KLzH`D61%UfkDO!F)
z(|&}!u~&r^_kkDp!7k5$C5L$PR}!(*n~D!aV_%8IR{L!7niWhZDBSqjLL|fA)BIIa
zad9ExnVaSoh-$)%*J43pJubc<oUh@4lx+yYTd8ZwAZcRvR3P#DphVB>i6D)r{-i|i
z>!~0mbPqmNB1pY0a&gZ^ak<i<IB%<E#C^OTsbvvpN7j49ty<5sM5sx8MiBDejqkiP
zD0=2boUi>`_;Oj0@@$Z}6!l9N-_JL)+d!>Av2%p^8c6oFC?V)52>C_hAlj3cy^!!f
z2099QFXSE+GmMe%W5z!OwHCTpeJseW_um9*bhWSbx%#{uiUpsm4FtWQay?Q+phfw$
ze)zDU(MlPO2;B?%AB$Z4{>nFprLq^2^!?$ka=X|1!<1ZD)5zLDdrj~ydpzEaMrPDn
zY7GWvV$}RA7k{P}SqJ6TTM*8{u6p{YXRuW|Nt-;B1>zCP{yRzvC3hu<x=QwTmgK8L
z!N6l6rx%mK5JVBbAB(9qYeV#HNF3vEq=mrV9}KMzdwE*4$=m8ddN8z0V``&HTizeU
z8S7_R8zZbT<R3R;6>aw|t!HL?cx{AL)o*}efUgZ3H+qoRh{c9J14#usZEw7u3kh0s
zkrKnfFbWq2SX|!l-D|uX2}T<48o*<G%Y;A2mH_$6Ln6b(kRTWimE?G+F^Mm6yKttw
zP&Q^I)P3QjYfc8+%5c1cnaAEj;IKiSNuN+)3EJhEw)bJwI=-FQU;Y^ic7X(A-W34`
z{s~^L9=1*1$@$rgfw#Kv?40AUE}_^LLNx(7xC4#HTc`;kKb)3utTwsePET<sO}J}F
zh0xNJ1cHo1UlSmjxcGL;iFa@JqBn?Y96VX+7~CExz+Jh5FkC`*v3^(k-xa=pCr^*&
z3ai7p_qDQwiQ}Uo??_x4|20}sPpGtizZzsfG$nD|WX<tUj02ZNkIJttEy3bT!t6tF
zYe3`ZIh{0doM6kjl#om33kup9Pl*N4AOAu7lyKL2QhIXLdUmB!y|_{?!$Rygo>6PA
z4bG2maX>uL<SqFnRT8Uu{udRAUZ}_BF4QxXP&dy1zm=M_)1jDULP{ofL>mrPO<UA5
z)GdQWjb&4h&KGl)lC(HV6m7;S3tsL<ww`N^dH4GHCcZtP!{i%@^nrkiC6nPdXDpo9
ztFf<S5_&RJ_D(GBVlhzQzAtA|F?$0`Cb~101I1P8q39@RZyyKIQ1p;u@gcB>KAwam
zk5FO+<T(?Iy}BK#Ztl(j5NZ9w?76c#yvYil)~;#%q=;w-*;(}z5i|jGTCscPW7kj%
z8;QV}0f8V@N>pJwr<L+0n$Jnc{F#YM7CgC|A!sU1Jc}BH)GAO~YpNU;E*`m3suhC5
zmD0&^r|?kVC)7;2;+XqR7v@6~g)4=p`8$66_$viRK}qak9O?y>rjVH|uOP!{elV=1
z_SJN0ju<6fI5mE}@bxH-94(wgVH5>Yr?rD-N?Vl;t&S%PwG~LM!V%>>T4<0wM~(Jw
zJdGcfrjHsK_5_70TY0Stji?w;`r8?D|DP)`D%0qQU*mB+72+;o4fSXZ1w+1BM||pt
z?qa?jneSSO_5O<`L*pB{Z)jC&KaQkFKw@&y)e&18UD-Mi87+(?e_9-05_+_8aaf9k
zfNr_OM-bk<2#BA!m2iFgb&9#ZiRc?oxNgOneg6&#NK~XN+THrGPy*%_g41xdjNd~y
zzNBspu^~`&xU;&;x^cJsckPc4gDB!B2pTPI&BCuzKdC|<eT143YpJ7c4k*n6(G2zN
zn|yb*1ZXZ@`hE)9fDa|YmhQP*OT!X(ZL6WCU1QHkca1$;RjL@)CqJiV_bBLUadTu5
zMY?efzy9}>8N-1{*NArK>u9COwB`zVfrqe_tcV8q2wT}mi5X9`0Y1XiwWQ?|5V3M=
zDbPYV$pR<VAGU+-l>bY^Z#(6;6#IgL-z2z|cC+=S5}H$1?CYdyTlu?3rRQQxTJ$-_
z;7ek6H<s8y!uHERxL=xe-gPAWK8>Z;R-@fNs&-ks|AAsZqTuf<z!GD_6tQ)i-w>PP
zr<4aXjYxCLcDy#}Nk?0<)ZCbr>I@H%04VANF}Uq;*5@HbK(nQOz`kYg!jZobUrQQl
zhzSYrb_5BCkhs4;NUp=V_mYmI4!}>U1nLd(m*Z=xwKTf9C=~=51N`X;;|$!2xeX&?
zx}8<`=f7Z356)}nGyw)h(K#|cLm{-j7f?wp`2#$RXWi5uQx<^?>u*CoxU_;$@Jb1m
zrVvRdGZfr&jE^=}<y*qa<xW`n3w*XZ#HaGWN$E&AZK+Y1Q{fbJ6`0m&H-7R&nxii~
z00}d)<<CmO9R2p_3n20zj;<v&dd(S#BQw#6W=PFS8+#t9K#;hG{xV#8RGKazH@OLH
zorg4>MFUz?_B`Z~7ZdmWv();n=qjTM(u4MOb|kW%zDatJb1-2%B_)N<5p?_^dJ435
zaj|hAw{4l8CGPLwt-pg<J!u?EqpVGb%MijinCEzY)7i;8lZukT?0R&{A=_3dc0&<N
z2~4`%F)3Ou^?YTiVg?k8lhq5|yEL<0a$4}eO4e$QhHa0Hg6bTNjYUJ*RU=0SlB4rN
zd%}>Mbs~&AyI_rR_M(pTMn;878~Y$8;(Io_mXgMM4Otd>(^6zH1)(%jED_0vZSIgz
zA>D#do{2JBkoEI9FL^^dtVcV9nl!`tQP3v{;HT^<?I9Iohdm2P-y0-NwWScm{0W11
zs9y$2P^ip=kp;|*#_V>^%qN8P*Ze;sfgNUJ`9s9VqfUH!kK%of9ceG0CBd{Iy^0~?
z^$bD=ko2-=Xt?9J9nchir*bAh&}Us$CX*F^j@59Gf3_s*BMt|WrAHYH8&9?;Ro9Na
zn!315T6#O*KCaSTCs1!=j-UXSmVZ}Mxsz<*uuNqtS^&FEz|7srZi~Gtab_aAI^>Sd
zaQdy_8v_*Tm}YId=EmE{gQl5D)>La!1ngu;5T5~WWy8rLD~4T=e=ludf2>l(38k?)
zivX3>_j+z>5~h_YIatQFY+W5oxyAXf)gvuVM6C2k%<WZdU{C^a?>j(sf_!STaPsP-
z1RR8M(?euyU(Z~_`~lIb+8%oy4E$RBb+)yo&<B?ifC`unW_)6r3OasN?J$2|QhQ)b
zL|=!U5uT|mB3mLk3IEY4z*@(qibZGe*?MPKPUbPdF@{r*|9iS-dM&xgs>5b3hEvi`
z7Q4KRpAOM(4EI+_3$Z}=83XBdRM-a};1f>8IMb;6&orQh^8t*FtZ{Pc#`{eN)b{!j
z>$MgD)9x@-@Q)@&(j;+&v$pW2M)M3_TFb1BU9+8>MVa8ze^he~Ys%RaTVh9pI#%U*
z8bospAi()!MUI)$5$iMV55g6U!4pda3CCvjWM*@cHwS8QX(__@ql6ve<Zx;8mS*Dh
zxGO0G9EnB&`e<C9^54TZ>j~hhjImIH@K#K-Y<15{<w9lYas|=_B1%%r!r_lLRZZ3o
zv-!VWMdOoGRbMKawp{%0Qd+KIIW@UZ&n(K;i#m+c-q<z?VaD|o`cVvt6&HjdQ`azN
zG-7)3SDNK9D$ham+P2T;{)<$X6RD9eClZ-ALkQo<-!aIIGuO6}sLzetNUPv=5wLMD
z4r~S>z&~dbF`UUcta^A4ro({m4dQ4t?6&75&*QIS`%{rJX535h{2(P0)C8`2Q#rG+
zy9DRs4HZt*E+b3}&|k{OH{$bY)RB8W8%0GD_ABo){tI3&c1(R(F{RegaG+%WXJSD{
zr}iAOau~Z7Cod;B%ves8KrG6PJJ~tu8|4Nl*AtdeJ{YWqVS|^2H4m!hMy=q~0((Pu
z2Ngbswc%#CPqEgnw3q;b{~#}tF(MBU2_b<Pdev&Vx)BmI5Wdz<m%5}A9es${>fn1C
zto%@Et*|;|GTf`M4w`aY*kiVis7_~{<Dk7r47OPuTM0O8k{#6JsLtKan}%_y-*r3d
zo{&o3PFi%l*8!)8)x#F@`G_7rugafROp6{-XH!ecX?mF|l9MHUoU9ED$Y5r(i$Y3;
z{9N_1nMU_WV`Cgn|4G&S6hSd3^oMOSGj~Obc4TnsX!VRe(Au`acYwGt$O#7DsKOyO
z2;0TPrs5^#cpi%FM5Sj8KF|>Sp}yD%+@U6V<R+;@$Rr4r;_u!3RUadeyuuJ4pbd*2
z6bS?eZV%{DVm)aPN(-FRbe3~T$mbl};qPJ(z|x94)WttY7su1?h)L^+0W)woM+?Ek
z-aF!-Q$}~#8j6bYg^#je>cVCHcPM`s&`86a<gK?~!;qlG`;Zc^<_!XeuR1UTWf5qw
zLqRx+h!_w*I9ow{9;t{CMEyblE!BX8&R*VP`iKHj;zo2mMbAm0kI4yBF&&Q`pSV3V
zRXAU`T$nB#d%Q5kKOR_+y^Ce+L$8(`7nN}k#_UjQHKZOxKxL=zpoCL%vnxyJbc8I|
zT^f*(QIa(5!~aW_w)3cKB(egB#sC}FSO^XZjS`Dd+FqfYs+noSylZ4^RQ3@G$wX!!
zponlW0PTgBE?|jjmjH(yk{}KxO`J6O8WZ<nC0GNswEg!g=xS;YEHKUlSUTn9Q);Z%
zTpz196J2$*bz6oG=-xv{o&VNQv3FUF5wP(41jV705{O@f$4mO@o-~aJ$&i6Y{B2l)
zrfN~dYb;m^f=sA1Vi-Kfs;Ju_32QSsE4nu3KLAB*IfFDJl)V>0Ad#zrLEMZn7#U!;
z7Jn9#68D?Akpx~tC7^8vY!6d&h|^$1kdpNms3CV#Q~?4kFHhS9sncF>kn75wVR7|M
z)T7J(I~c)w2;>Cr0De;>ox7E&XHXuUElpuiAw^)>si!fU(XiGtM9Rxc=w<ZG;u-+$
zTP2@0Vgh8qFQyj!66ZNUOFEH`5U{2BeLO;UpI^d+$Dh>4fkLwmWc{qN&TC4qDOYZ3
z8HFd5eqs3MjdzY(GG@)y@g7GmXff2Y4sVtzRBy9@Y@FO;R%fh}#&<V>=vK`Iz35sx
zy$u*_-FZ-pM`I$Uy4HFEU63&r672Oi2NI?gc_)WXNa{`sX89=KF(e>5s_B3bT8mm@
z+MLxtA#i&E^Jq=FLl=M=9_WHSv((n3HzS_{f_3KnWaBx^`N^nZ^Pm>jGq1B~uSGu@
z`9u_Z1Yim&QA>%w6!BogCnDG2N>(zbBGo4^g9PaJgZ!~r#H5iuli4(d3oeFr$A+}m
zQIG7-_j<H5%~asBA)MAgPBZ4sA4b<m5~CZ$bpyCZUNH@M4?+XXhrRZlW-53K(<oEX
z)_egLj`FV_N0J1li8LoX@|2@=(#@$IW4Xs9>oeoQ_whA<fnuvi40>~R&!J9@SkR`r
z4Vh>bxoczns+LZzxqeoNZEVP?Rzv(N<amFh1!vFsWtE8A^7sFMM)`|MaIARWK-@A3
z#>_w+xY@tTlI!9oY3^&|G5@j(UR7X2(BIY339V^iztVcrA<i+*s^0nKk)%xe;BPo~
zeOIxLvC!JppPl2HIMxkJV69ZGAfc5FS*&AD##^ctPSzH8`}h57=1U5?_TdoIVtWYa
zBcd3qi<i}K?H;TOC2`tn)3`BT$H|i*zy6-)uc}4EIV|0T7?dq%aY*%Y8e5FhSS&DD
zGw$?FOxzx=G|pOg>{i^tdwXoE5Jr`Lgs1W5k({XbOP85l19uuHUp%44L7R5Od9~hG
zk#CNwtU5~+sOLjt2y^n$Js`s+9j#;wdFe7z_=rX%N7YUdTg1%jn2pLNncoo5F;sS1
zgV^}#C!#kx`Knr_GGsH*!T{N1#_*C5M{^pl4bTM|C>gUgp5unRq)i_PRS}HY%&!Ay
zUK#)qdn{5*Uqd$=3Izm6sv2+|MVX*0&AIvj=aWvTT*$dPYft7y0s7}ct*4fUca@uk
zQo$!g9G(|^N)$SD{_?f;J`klGc}%Zbn%VOMsFR?3ZCtTDw9R0E5(7qEqNRocw2x^O
zr)`k%|28SZ9x0!K<4wsCre0K(z;5qa-v&VWtsQ`}mwI}n2~4ehR%@Q8r2o^kE$gE-
z`o6wZYB%uRVs7LN@%!=Dq5lm`qatw~uAYGMi$FR~<9-Jw>nSH?G}7aC>Xa=z&g8w?
zC1(igy-hoBt%qXfJqt^w6L5~re^R-`W%6HD)}F?aB4fu6X63X%?<fxjdebsaY6rV*
zt57hGX4naBn^n-o;MzA)nel2yG0V2)r}Nw_cG_vQCRiOG^S@hVo2Xqqj(pUitS_tS
zXZ7|q1(M=BRNgp++R2VLV7hvS)ZecXMqP=I=wH*LLDW-<eTp<J*2GAZY4j;4${+7<
zLm#3Ohz2#1zDm@4Lgu4sDHv!nWb<{^)=7E!?@^k@+5e;h8<lTX%$+^0QJB?6eVC@C
zAe97hWh9-#Xu2=96Rg3(VY8QrC1QC<shwz*VnQ9;&6XaDv90^N#<gQIH7xowASc!y
zt2u(Hz3pG8{v{-PK4FX?IQz4hQnQOw^Xd7F7(HT9h9ECC!6XYa8qd~y%?3yYFtTEI
zGh!s{@9+(?fH6xGM_wci^{OYI^d=WyYB07`Q9WgtTeTi1N85t3K^k#JKG+GCny6d#
zhK5T7hPawCMv-aB>NXpK&GO$$P53&B{0Rl`Q}BKTFDWqLSnM1${aN<M2^vK791Xm#
znUZShxry<E{tA(Lt|a3$ybWQomM@j$>crWvneB}Kt*Y-^6o`;@tL#~oOLH#EUd|2C
zH|b?ewM)e0;}R|6_bQN($sZ!Xo)7I7C-pd>M6rQw`MH&B+<~W+CX&_vfCAeslZpw)
z_E!~rRDrGd?@;V}2<p9|`Q3CG+rD<Q{&h9WHtoBJmHDfxFPR8y4&L$q*l*vT#%9qa
zxf6TD!iV$GR3FNVk7x3kR3^h;wlAB?lbX>Rq+F>?GMCErXZG0pWNHj!qTWm*m1dR?
z`fAxUe;q#>Z2eUK3;k=^-(xu+%f_>bY?A-6Y>HUId%po!IReTS69VVgBeJ3|r`Iz#
z9pLTp@^HJs0g74LHt4-%C7dmO&aCjU>+)oOFN^vQn2)z8HucWO8)hH!@P<wO9(i@c
zrrx}{VN-8D-0*DoL9?=tVpnh9v{n`E8%MnH>l}O7*V`NKL*6~!z0SU#_)+f|7W1Rw
zyW`%8>l_VO*UREw*7dTsk6>H>vf10qwqDluvaOe`y=?1WF>Cv1aTni=c^~#3EAPh2
z|83sm*OSFPt=90aWcK$OgF-X?35}xn5l^=Gdb79rhT3>DX}0-#gXcPpCaL<)l=rOn
z^;TlPOZT4h-eGM&5I(=+z0-T%avW@y)SM-~X}k`+*WMiRUhrO|7l*?#=e+k>9Y@U0
zf1KmdtM@+-6spvkrk$&~MYwPUGd7*@-^&UCvHk%6rUg41Ka#Er;Sr)!{Tr+xOo}2U
zNjBg&nJqmAQKU$E6T_qj6y-v+y2W%zW%n)>=#;<9xq||ciFsnx9ubJG9`S_Q^kQfR
z0{~7&pT${y#QQCtwK!U)<jKt{&tNTiO=NvuZlva-L9e)3ARw@XF-HTj!5j@{Ln*tO
zujSy1VR|;-E4~qo1ML1^ic}C3=+~2HfW8kRdaYK>{=dHhTVQ>_>#Yrfl*kkcAS5IQ
zPQJbnSqH_j7)<2S7EIW%{0A+==HKFDL7%B|{|Gs^p}jh&@}B$>j0OEZsq@>bH%$k3
zhu3d<ns@@3-(`k$yMqD$=W08=0i|2cpRy6e8-TOoZ_oqdgA<Oa>M<YwUIqII>LVBy
zSFzv(zIzRO-je@mz5m+;XZ`P2^ur21qkxG-hFitu3vTJ~PqXBo^FN{&lLX_1`c_z6
zm<!3Ro(xG{og8Jxjja!Ln%5{pqq3>1(1Tt*zPz;TFRQ|f1o(?Hzj0=}T-Tx1FlII)
zMK03JrC%cz<iTP8-%UG<8PB5xwyec8q|74z_b}(gp4USrQuY5XS)9dz&Jwdi=<>QJ
z2Uj!-3+Vc2yF0#CGwMUFxR`rlS(=2k*f%w`gbLIWzFx#fsSitK9eZ}zV@(@l-%hM=
z(ATR9MikgV|Egl&OfWuV7*lbm-7-7ulo*SRj=LoP2i2z^QSieGZYlUt1vYoq<G-iJ
zA5hRD0-Mv_HrwlJf}vJ+c4;=gNBDtG_*5^a3gl}hCc{<vnQ_0Pr0O>YIh!)J4JN5@
zuARvbZ;$}1q#3D_D}h0eGiRliudI%B0h*>weFxS>BU=9MoJJDTKY)m_uw&Y=u(u`Y
z2eg#XLt01oXg%fJf4J>DpT3C=g)lTB-jq#fJI<#+7JJMH=<$TxEYOlqX&4q2{E!01
zO)_&^G2f|#w<!=pEawDF$DCL7A0rLs8cRmLl5A2yfetD)uJ<;o&7lpwF1pDKy)OQa
zMq1EwH`4Nv=<2@ikumf2jv~_uvNwLxL{wZvaP^l6=pA?&Api|9D(enLir9xFDChqx
zq{OQwLWBV5GVH<%*Sz~7`!aVawG@t4umvWEz$_w6q)$+gsH#iS))OR%iD{OX&ck{y
zV+&~gh~CLvN(^aL|4qQ;O)N<T;HXy;xkN*Wyz!j|!0P@#j_#30-%gs<w}TnWmzmsU
zq03vCUcn$_Q4TGDObz8};K{L83ZfSvz+|NiBYxrP>W=;U_ZQyxqTE-!c<#(GoLM~m
z<nwwW-C)D&u7+0)Ie$Ff4)Ub=JwE7bE8$PUIRkNAS;s{E2fT3b#rCy<vnJK$H)*Z<
zkaNC?6M`0`856&uEuraMLGG7olhfy&O(7PB_|U`~7LKw!qz_tMEd~}2B=Kk=ZEJWO
z;>e1Yst>x)`7PzMdUVxaSIO3D9O>fh7}WYhtw%Ahj-cw8fq^BxDeZhTl!=f0)#`!n
z6}nk9ccm6O?DVvFo19uvZ_dXLOHX_vQyaYL81GuTI<tjC#C=SGr9#!>fuZ|hXC5L9
zn*t*t1S0EOs^{^bh#RhsE#&_M?;2w{rC||H>>MEoZ*TKmu&&4Epc!!@!&T!7jbfN6
z>Iq)i-tJM^V5PW$Yo;yxG=E>ZpP<1t3u59$(^KcqC|07^^D#;4VCE#(z^H<FDRmM1
z5>#_z6KCnfFvaT$`8cf)wl62s%kq!#)_+dR^mi1i5Znp{QgX;D{||N${r`Gv{cLsi
z?^EnA=<Jt-?uiBF<WY3@aCdjxPtCSQZ0K}?n$+p)4-(L|NcC}1$H1k*8Z7w`xq+50
z#nnaLNr&+O?=%q3y;IUTS2D!Pa!G?uEJp?}ZG#s#4p<zI11uNGYr?_9D?zE}qB;Dr
zug0rq42jZL;?;K;>eZ~P(FWjAeaMifwvLrbs|Vj_#v7K1qdSlXRK=@Hg~=*z5V7d6
zq~reIVGvid=Nh(ft2>kd2Rt2^Sm@ODEW#W+&kdUVl4k7^<K+BIh=8`q_J2_?Hknxe
zm-V7+5<f@UUu(tgqrA>)XU8CGgG{_NrFobon&Q<%Z#A`U7pbqlj_Qu1NIO|O+NSqB
z5xhi6Iao@jus@iB0gjD?<6sBLb?s%25=m-r7}k6-gBw?lAVhpA4j>&f2__PoI#{4-
zVrdR2#j>d1rC?PO57JOBDX~=km@gtC7z)wSrJA@4$Ya6D0u9TW0+A=<In$wYE?b5?
z)J;7w+!kPNfo+Fj3I&k4;^g!NUwhW^nnQKES5>P%UEJ09FfMn@(Q@PaCIT@ARx!2;
z2SX*kQ9<WK)wdf@xs9Lb@JI2X`}l5kl&?sM1{HMp0Wz=Fukobs>7KK@SJkm2)t%;!
zJi~*g%RCB07?}c7tTJqANLTSQ4570YtK>Mrqi)Zk?Gkgoftk#7fJPS6fraRf4F`~U
zzBR?9)h&mc8X@Mh#~HhWY@MbBzPjl+H5z#T=M{XO06*Onx5E}}%}&9e(J21x#-a3o
zinQ$-GXX9)N~!2bEEiur(mhglJA^yNiS5%;I#}g_-hw|ux{f;@pa@-mpShre8;viO
zedIvTSA-#aLS)WhRWZN7r~;vci%KmFttES}Cs&W^y%wvUHb^8czPO5{7<{u*=AO}w
zF6z`m`Ri#M5rHE31}nf$F<243<o~)xrZuLAG>B>XtGo>~H9FnB3=J(Yz9bZk_Itxg
z&%wqb%$bB%pd*pWq0cphHEklg`Ot1w(FSR4l9oNqpOG9=ZRC`Lw$||A0}|D&bzlTg
zL`quddDweovbeksQ-LL6PC*=0+rb`C0dS6LPzhjBqUXVIfWa&H0R_+U48maBJB8Mg
z{KqFQ&6O|v+7dVk!;gGux)VKLk6!i*DrNXp9B)qD<Ceii^m%!YKdo^RFf4%58*h07
zxzotnF%l}$9En$XP)~JC$yQ6@EKpKzzU3KHBf$casxu=I#Bti<dJeRd8pI70+@waK
z(O@(aV<7o$=_#f2$JhiZ8z~&`kVB!$Ik=ix{kdEr{0S|w&Ocf>Z;s}=UtE3^vqrCW
zzLnmvvvRKMfx~#s1eVFkh|p>iUlgX6d@xt%EZe?x8Q1mav5c&=GQSOLuE|=M)r=)^
zt6{GTygL4XUb@<*<u=1_Ho;A{IM6oZ65o8iHcaM1xx9>;S=B8GTL`H(vB&Rf<NON+
zHx&FHK|_7|_m%Jm1QHNmmRwO@_t+;sCXxOh>P23?EG?90YIvi9kEth@OYDeZw(Q|(
z%`E7;4gkcHkS!%QB#$(}S0OEF#u#gfJ+`wOe6<(p5Ibr$i7>Q%wN8Z9VrMZV9V#^J
zEvF_ak)sjOSVyB=1M$WR(yGEGP@J8I!zP~DPb0iXYZa1D<en5Pw;>O(T6s&YX{*)L
zC%t4?D{D!Ll|fpyZm36Y*wZLQ@>O<ddZO5GF6)^5(8Wt5Ve(3tU^80G)#l2jDIfQ`
zesQ>2M2XH|w~<J(Rw9`4ztrT}eov@Wamf4#JDsbB$*Ik~0Dz$Z8>YP;d;ThaS4@Of
z&${nHh8BCC4?ivzSl7h%>X2z>eO5EOlP`#+A|EJ+Vma6=4mC@>IfhXwGrg;OraS7e
zap;(DabY#*Hs2$B0?uaSWt{NKum?pa@l}~?U<owDw#ar4ZbVe_4FD7k)A=8!L*$#r
z@+I_Aq`7>F<B5EtK9Jrg@Em|EFd$0?sb*gM)wNzrN#doKLyyA|qF=&j_^<9gTu$jc
zI}x|OC?PUA(>Igr8UDjfb-n?vMb|Nu0kY%l|7HRk`~L8`tL#W*Mr*PE+oF{|Yi+{+
zuVHKd>(|g)!%Ard>pE-|F}V4zkJ=S_!@1*PyivagyL%btQ;*Y)ob-9n#-zQ^?mJ;^
zUqj!OA9ECT-NrH4P*-CtDxb6wkX8RTx*1gLzhg9#+Z&@ntYb9v#*K!1bGOk@zQ$-+
z&P_){xoq79ihcE!pj=ty-DEU2)YTk~Sgi*IZe#}dcj%u%&wf0+`myHe(6CT`l*40I
z=&77vqG*Db{OO~G^G6Gp$6I-uhnZ#O1{3arXcobGI4lK3!f~cqn0)BuD@U>8Z#|!S
zrBz=G7s07^Pg|~f#+zGkz7)8k2EXxg=@@zz@4-S2^Jr(kqEpSOt{Sv~a5P=q?<U!u
zWZy-so_b$p2IKd7;w+*6CxU-xmBTnwKctw!yU!B~8PHDi^G}uY-w@PeD>L<&$KO;H
zr<Buv^8ob!NO}I40RQK*%@2jpsg>}PO7#Dg>ga8t0M@Z$zAgLdZdloG2)XKW^}Hl#
zd4~4&CCB3CO9}$05IcYj|A0J8@iZ1oAmdNHc|3kJwljJUi2i1&379pjdp3ir)*Sb>
zY3UI)e#BVO5lRap)pa`3Vo!10L4=wxXAu2SEc~42O_1ea<!}TrNmPWp6=Dd#(5Hca
zYHG*Qp+PIa%6>2LZ1sEDw4x>CoHz%GEZv))3vvG(>2fRcB*!ra>RP_CMcbQ{uZIrH
zSNfXmK?kP!-FyXKhw?NypNHq0@K<nZov-ylRg^8(4;^xW77a*m2mgEWNl3g=-qK=0
z9@itTf3G33YxEE2+4F<`_fTR;lnZ)1eK=;pk+ph63kTR}`0qo^l?eLh;?zEj0FUz4
zo<aX}oh9YK#Y=t~Z6J<<-dRL<7|8qrdGL8iJBEV6^(}1sAtAm9y)ynkbt%$T;yLC2
z2G65T|0oz_d&@Qw4?>eDwJuePxWom!<TeqE+e(ts)Z^UPg!hgflW3NplFAF<`SOB+
z8E)x{`zN2kf2Fnkp9#jdI*y_-b#2$!{`pD07nSM%BL&7mGs%m|TYf}oQmJ;#93#PO
zW4@-e-`3lJSaFN=CRgy6!j1=X#ihmE40lLvLBXJn?lM2nC>s1T%ITgdvO0gW<HK#t
z;J>88`u0$@rj46XazF!RILDuuc5G&O`IQs{Ia<>iu<2P>Z~Q;wb4Bb7bp=Mr@s<51
z!Ht9LY5%v>k>*Gh6Vs|JE;HW56|8u>_|2wY@y*)J=V)plI=vDliMc$4ut??(j>Cw@
zxR{I}<T_)7m<f{5>XDKPMuc$;3MpgMr<ImaN@6B(rfr+;Zd3N^U{nda)8Etled!|v
z*lg~HGLPI0jk#)Alktqy2y_5tMKPP9*&&za(ddCB!?irjC%N-M`S)5(mj<k%m`jcv
z{yybs!*U}lVWq;TO;5a`4?23p$c?B2*B{gnN}_%%Z=^|+8l8p2@%QMhL{7d~f%Qy-
zp;a6bkQ9?iw^1>nHUFxyrs_GSE;K<fPKiVb$``VHjagF6H#{9oOGfnt)qg;(v*55|
zT56xyRLSfQg(<+RTxs<Y+bE&h$U(d7X!{X3DSlKzTtP-b=eO%SH`Z6uy&$Il^J~pU
zG3#xOI@&$+Z&k_NSvz7EF^NWZH;pR+2gZpNJLQ&>?zuy<v*EsA!tbknbSlDGvq6-%
zQ3ljO|ED$9{OKGwW-WkF6r0&_0BU8%4_3<YfHxOMGdV-YkMQA!FhiQn+{IAw%2)iB
zZk_Gar+C#h*v>jYZPqzAWIJhfNB3CX&6Y!ov1+t5!>WHyprI3^!U#c`8-ea}>WOm?
zs;<+|5qp*|z8zxZEptLpVp)OjVMD^h`b_pKBOY9*vFpSm%gl2EuSES>P)hz+&xVQ-
zIU`T8w&N_(shlTAWWG1xq-wOuEI|4YH{tad9oLwomT1CC3sxltMF>NP$S*>C7EczT
zfmAXykU}la_(_axXy0*AgM5Ynbjl>gA~*g2Pl4fQts}(t_4|~)b$S@@-_D1h*NY!h
zV49crD<)$iXJVCefM2{QgTnYe!ezIc_4g~6rXC|Y{8w|R)y#LP?AfqY>gM;E?%Ti8
z5)k9{5a<(*H;o*XEMNYK_C;5>b}xxg1ERxOLSy|tO+D2a0vN|wJcDkeDx(%hwYl2I
zK+0SnBuShpXv<MAh5@Ba>~T{u7rxWL&%~fE%@z%g1;;&lJ={MNB&w(2{=<EMg(|)Q
zoV~q#&EKtHj{-U`RsXJYw>iv=h#dbkwu~v6+UO*7@YQqSnJ3_QNrO42pmVd=dvJQ+
zxR>ESX!J(-CJLsI@^G3t%0|o;H|8oEwo_Zl4S=%A|2sl;^qIXc_0fg{mH@v^mWmd=
zOWnaJI4F>$l?|;>gr?5HMzKe*zH|k)o4-LdbW9eDQW{309V)%MjBhp|f;=)H?c2aK
zzgxi$$lAeyD$*aL3!$uK8(@rW00dl3dN`tYaEMGvwq>G)t&W^)AZoxjBHT;yH52fe
z;3vtK-g?$UMO`B>o}T<Vy4)n!o(=s}abE&{o!|L9zNY*|^YJY2Rpv*@-DrTpG)%l>
zQ@S<ks9Z+r)<_RGRL!cU=4#br$Q`&7b+P61$?n*&O1b^&FuX}_<qIvTs0RVx9QTFk
z);N#*NGqQFP?_VbP%P4_s6m8E`^XR7T;0u0)&}QT4We|iU^W~gNIx|3iu#K%c!gUo
zkHAJ%2k>Hr&A45@=5pehPg0jwJ6Z&(S$o>;JdmbIIbb<eOrvSdsGV}pryI|w%~LG$
zGy`o0=}c|h0H5K>F)hR`%)LLTK+58#;NIY?y{Uwg&DEN|2*{Glmouan`+1Uzcf;>h
zF>_&omB*BJ@5Y=<E=?_aF^<rxPNrODC*Vtz;Oh2n@YTvbtyS9i*-u0VA<I8fy+p{W
z6NLIF?76gQn1T+PCkJskrN@t%21k*uwUjU?!?QT^S&YzSbb-?yF9~c*nc+mj!|A<$
z3={NxMefnzp7H!Zkmgy=*U_fs`6+5jQZok*HRUU~(!ZwLe<WVZ$`_^Ab0fZ<!3;hB
zSqYCY-atL42SHr#XV9*c-sqeok3dEPype*0a`cfCK1i*1y-&szAgO(D|8&DqzyD3m
z8v3ZAzpVyp#+M}Z%n!7s>R7Z~9Oio4)-K4}yf=Vpy!yk9LU6<eMj|RpN~4`JVjl?G
zsPe<g^N{Ll<QP&rnmKZ^R&9>xuQf)rIVZ#6oak02%}IhAjN0c!jsfBjlQKY&=I0Ri
z1jW~eF<sfBD^u~9*$WXFncwPd0lJQ%nK>dC_q^E>Y-x<%`gZ!jT}|^lxVI@2Z1G0T
z$<OH8*0r%<j2gBD!}B{qa>CJOY{qK4NZ%cd24hHAw%kOyVQm{V?4gZGZ!2an@wM%g
z-x2KCwET{+{Pti+v;59r+j=43J8Qd|qcz58?P{B!ZR>k!D^3qsk-8~Ly_b{W5qsKW
zy)W>b4u&zY>0VEJ?M-2mH~X4j+{PF8wtcZNZ`(=V*OqR1T*}snx5Ks4TpG0IdgsQu
z9^d6;b<>S~+4!%0DW=TV>epZhSm68`Y!Hw)tnb^b5<|x*hP^qBnRf^<>Ev==vgP35
ztpAJ>&nkE)!D^55M=(3dEt#RQhTc7|;9Ug&)17#<(FuQ8bsE08g&1UsK~G^Ws~Ojd
z3k!|*n75|Y5tnQaPMZK9WM;py8v0S-LYMJ<Z4MNMMUhSDL%#ZLb`qag>}jP%6ueu(
zlG0kptigUGSHD9q&MWwYf`6vq;|k1F<0)cz9MRo7=;tn9_5TmOvSYz;WJ8LF^Z`3k
zOc&ty^r(ZR9=ot$ngQec9sX<mA5@Dq2!5|(eG2jl(kdt94j;AVqRMeX%rqI38FSX)
zTsXWyrlM@5IcYG8G(zb`rKN1-P8UNTS0Vq^5{v!qm6mj0Sf!Y@%d-&#Zqv7Qi23_e
zPuKuwm)Iz&gO0DQl#5$kRqbb;dok!<C{yU1M@KKVxEB_RqaEK@n{b=WEe1O_q@&%U
z0oTj1j+a+Xw%*{uf_h!ohFkGw26yB>pne@zzs?aG&{?~{lkTG#5UvQHMhI}jeMKs^
z9e-YOzL(-Z*d(qjWDD!=2S{LnH3Y8~i>+8Lxt$tDk#k}rOJ1objiG=j$E(|8+lh}w
zcan<xiw)BB#$Gmk>}j3-jh_Yw_h}Lwg(~Yts8q;{94IvWRS-(d1dM*%&l^PR=){g1
z1(^|&Dj}7Q^%*?N1X&a7^k9WJ9`rzx_ZVN(%J>gaT7uI!wvqBm(_7I|fZiI1`V&D9
z&$yqkqdeiigM(TZR;HvM8>IXXkW(s8y*HyA3fM*x_KCMjI_3|m714a0)Pc1@j<O^i
zT)C~rIkm(ZLE5i|&%O3+@63;1MozJYv>iUk7&s7O5N$*=DbAB`z$EM2&ex{5)jBtW
zmOfJLvfH<xSH@8l`6UHORX9Mb!Gdg3qHS*dP=q($VUte<K~*1Y@UU7Jli|n6@fc6m
z`|XpGro%tP;y|}Vh0&7pAyQ0ei#bm$a+v^f>*zCWwt5;87^eqSxBtGNMsHVp4=84V
zt=%1@eOPPvDCr4Sq;9OhRT;d-98adi4(CyPMXg%LVT4z?iTt3;d!XUX)4f#N_}NON
zhd*0_8o^Cn2r}h1*-&*XOX<FvD@GPyo+X+EdrJ9xxx*100$ecT-xCZV+a09T(Ar2P
z$y3}t=>m3^cR%+=#^pQ^oF#{HW5R!?X3R^LlTyUK+=8<VbAu%ICGH4wu;97kor}|{
z0ecu6X*Astj`IB_7ThRyEJI*)d2>m~^+8)B;61%8<?M~9@HE=?4R^^mV(&Mt0}3uT
zfjZMB%gEoQwYFQq9t8yjdkIztUUauH7CfYJ+)8uY2H*EI;VrJrPCfoO6X5iH`rN@{
z#EF(cedtV?qrkZ@e~Am(B);Ea*%e5L=0C1{X?=wQz;R)AtJVkf(hfcO@Ig{=Pzjs$
zp`h%Y9jW)!%8SsIldS!+BxJbGEj7N;>||(^ta{aCyn~nsFkIj|Tl{?*5Zlsr;O$ph
z*Lpud+M3q8hGlyK?;7?CC;iKbB-#cBwZ(-W!vA6-c7PSX6E6H_kT&#tCu*+oqtVie
z7j{<F@f!?OCBhQU4**;n1eX+hZg=>9==DeifGSOZHR7{!=fb*{!7Mj-T6%&vpM*xC
z9J+w=j>fKliD*t8hY@h~tqr4HI)XP0)%;2=^?DRD_?Z9PC*+YsnoMmTIVvOe9`2Lx
zY!Lyx4SyY#r0RIx?l|~l6mgX5n}M+pjf1YK1X8tHRu`0^`!HzEBh%0q+^<kcwWqFa
zspWM&S*;gv|M4ITS7VUfmkqX{_qw_^N+0@yA()q<G?F~^vttr#x`b=M|4G8Z`5|sh
zh&%f)o`-|2&~EfltW#(VJpI2Oh0=(xkG>AoA}WyfLbZU!M+4O}#&Jd)f>Ee0Nz}O|
z`9_XIY8a}+iSqunJLsyv;~1m}*D}FYM48*UV--CdEYWai84MeSK1@poLZ<=$k@%q3
zZ_Y0NG8lo@*=7Q`gnt`IHEK%RDKrhFQ$)FI9BLO|CuTc-d?3`k6Kiyuzvd51#Z*>w
zN`Fbqn-ezj${uM))*^I@1tRugl|fLvF-?`cJOCLv&dIBZ)LTuB^JIelX~$>uNv%!$
zjN|Y(s%r4gv|msm*4E{p@gr;t*=g}075q56=SymDKC=2$H#b2(e6T*L&Rt>$+jW}&
zkv77jX_rr$bEtSL?wFIWQ6skbTr#-AM-VR9rwyZReg6uteoWsNW|NA<qJ=xUDQWeQ
z|KWCY43ByxwxbQt7!kWgHKSQ*3{!ZAYJ!FsXrZAxIh_FTC>6Awt;c!s6OGXti~RSu
zvUR!q=zCfNTX?pNf}K;P{v&<+Fpa6F&R?BE<I+gBbMSL7>rLydIHaZ>R?q^dhVlpw
zj&FAtRvNSFH5GhT!4Y-Rj)xZe_=eKNDDuY@dr+}oQ_MK>->F9<T#RYoWHUSVz0pW5
z9=<@aYo?qjjPGz~hRwK>50L6qc&DUquD;as=%<vV)vF|2$IG3r0sPJSrxjGkC;Pd2
z+E|LEbFQ17TDezWKdRuEvSoA)#3Xo@yD%XJF4fNmi#Z4bv^i>XatW!AOET={_@oV~
z8o6cDW&dfW-kAT1r0fHAwHJ0ncC;i=(~yW)$4}<Yk+?M2W3iz-@~Wx}dgKB=A4jkY
z`Y#VOa`%*@j5&~u=DwUsN??pCB;}ki7;j?-zotzvEjec6M;MtV(=f8nG7dp+cwR}p
z0nRj>GIu)lqeG<=Xj^y!_1XC|!1*JctO4X0T<mB$4pNefGJ&}|KaRt(CN5hAQ1!=Z
z*e=wvoF5J6!^Sh@{Af!BsdqxGSHRAwqxvp`(K)#xcA9itp4p5;qP$}7hd5p4)`n?c
z4;oHUU~V7rJj6m`O-{jk{ck4h<Fv_ewp?~_9~UkooGl2q<Vnc5TBvgbTR2-#@aJsd
zUU3s(k3dvJbzAv7Ck2jMlJwfO@`vFr4ROkhaC(e#I)G_nZOs*%;^lFY573g{@bvg8
z;%P5$r^nBj!p1<TzVVAXFYNT_@xQ={MJYQwN}RgCDT2l@#sRRy-nKKuKcoe0K)_Uc
zUt}%~9Qg7k=7euJX;P|zg4T}T>bAnK=#_2p){ZwQu-QJhUHLz=#Y42PwZ-pM^B+}H
zUsgjNQp|Rzwe}G`+U|Z>v9D8brP<)l4eURrw~y=X+Z5C0^&eMoKxr1(5hm{!znfrZ
zPLQ5RIaO1;D3+^ZrF@~ixSKtg&}Dn58fS3E-M$Ceqi^a>WB1_;`Bjlin<=Vx_=Glp
zAk!j%8b3OL^{l_@#|dn~!@YqmD$1NKsK1%SS}sKI9;n4Y?=S-L_lD?Qc<#%$&nKgx
zcX}`TloN&(v>|_8=pCCBYyq}7XbXz(Y*Lg=5h`U_VyRB7fl+KCCs<1WtKR~x@bA%}
zH_u1^UOjdlj6x#@IB5N2yj;Dn8*Vwo`Yo{r|G26Zrr;YGZq2Vygnre1N_B5wcHX`2
zyJ7<VX??5Jqx<yO)gvACL;C9dU)xukCp#x~*Cz0&%I((X`bU8KQKhabXr0Bc=~3$$
zjmLwE+r~1y>8sc2h(FVdu)%AdX2w=b(-%|B&dBz)?WB^OyVQ<EFs+VmK64$SaEGad
zsvKg50?_tJa4X}saD|JYiOonkJ_X-M1P+=-Gl;~!7%H=-VF2>(RJXGR2;nb5vcn({
zdGqP`W_eh}235|U!$zuWWYl8>#L_%(8(FzzLW0*dl;dQo=TJXgL@<eT*c1w|Y_J6G
zbHKXDjpeHpH)Ns<<Zu(1Ws?AJo5pn?H40+r9wp9nPJGUL*|DW_Uy*+Ri^n&L>RQ!w
zD9fa%PJecY|63xAijLls0e+SkX{UJ54{3>Qdzb<Qscy4(A_WnNpamjL%|zu07Fuz$
z;kqY~kpsdJ(q&-N$RPqNW4f156XPgm9>~=cSzA7=xO!wm&yCiIp=$cdOf8SrC28i3
zZ4iSTl+~rpshN}ywZ_+!ijUG4<S<-$y5GwVUxIGtS3~|63Pe$17PsUrt@fSLrAp3&
z9ufZ&yut~W?HfT_|1|~Qt>Cv+^fwf=`1e1e$4k7e@0My=hiz(yGNS)^Rbd#gu|E<#
zhE`HWyl~By9D<l?rT-PxYPv^uM%k9Lzz(ISff|RkG!9Wm21HLfebV}qHH1QO;&d<#
z$o3^ZdWXa2c=zV}Dy_Y)7R30_8ybn}0fNSlh&}C7JBR)u?Z5?#u`McCK!VswCs&vA
z#hkZ3<L#O;6+2XWFm@yLa#9!Wt!-ID49E)FGLP>Gh;QZ{6^GWd^We7y_D2E_cBA}f
zrnP!PE#S{#&<gvFyU=xa1Tln&h#+AOY$b2{u}I|{sd+Dh=I;`#N#EfbNBhs!gfn6q
z@;*%Ajn{)2M;;`uRfBYkVF*ye_pa{TybiDLt!lt6&R~L-dG&kb8b{7dE7{ADT7R$=
zYsb$vTFMoPdX}E=Agl%X0_>-+d6)ykwITczjs?SL6FWbJ+e~wqiwQwuA2ucgh$cuQ
z7T?w1fF*&*Qg^JwQ2Ecb(b`t-6yGk*U+zzjdqeaNwgs#RoaP{Q-GRt+J9!Y0GIF{?
zb62p9I(B%&<lGsIA&8Bv?~&+K$mhkMCvPDbBkx#kZ!qSKFlW1n?eex*Y+tY|7<2Q9
zsF!!aF8Z>SzR0-ZTVRKbU5kDSCOTi*4~iUv7Y|0^ZM%^KYVTg#)8XAPdy1Q4s;4{r
z8>%(C`93$NQS)|c&QbI8wDLf(hxEO{Ueb4vp3xU;A~6ot4qwxin~kR<?upiLgZign
z<af0u>g{5c-mN{W8{5+u&t|)J^Zj0Lk6je``?PJGp7w$C@8e5azXiMT{Re}6FuX>*
zy{x@`pVnL4b<v{z-7P|vvClhjBTbw3uN|-%+0VBQ1_$XYmtJ~@zAy2a*kjyk>!Ayo
za87jH=>1klYqQ@L^FIOW+3h&}*N)JCc0@cLi?1D8JG^#;S-yuB9}bRSycVvNwQ=uo
zFpdr{Ouyi;cZ9u?3l2lT5F_PCm~y0%G81PE!u*sPpJ9)$jnl(>=po{TwlUXD0b1YY
zp_4NAT2FVc9oW#HSoM*P{@7}0w6Hn)NA+b!<`~{;_5|Z&?z_kTr@p)LHGKC3-|b<9
zf0|yN<ZEca(aU>{6g*1onEqzO#65*nDz<jd+P$0f@P|9UG=CttS5DU0O-F;H^y+@T
zlIAOaSUbLcq6RB?{p3wBBreB|g*y*O!2929`t*Fjdk~+VY2p2$PtS)YaHH-2wZQys
z1z%S1XA1b^Vrjy9pIs@15skYo!%uS_)brByYnV2}(E4+g`A-U(MCBXXr~-cAQ>sY^
zZoTLH$|6ErXE?B>Ltnm8l@mRDTovgIcG?7iVrvTihJx1#>X{dXAi*iTm2AV*gn0RH
zR|Df1YPWn}uI^|!AuYT&bZg*o!K#T3JPzO%uGg%=e>c?_Qe7NzSQI7>CmUV~xH?FL
z@$Fu0nW{1$uN<57xOic@#6NTkOp)@v)NuPyqoMBJQnQ_yo0&r;ZpJT{3$rVDdB&#q
zR{E}usPF7VC>lkOFydXTxV`gj`^40T8ZDomnRVGqx8ui8&p26+PWI!g1I?j%uCk27
zx*GIF*Lc`aXjmO-Eg}Q!Ps32Fvj0>mNq+Q#J1jzfuZ4FbZPBaZa3ineFJb1}TblXW
z7?oAl(mOSYMv^?BPySO)cYiy(?5&)|`ro7_(!!OD-uvcenxRgj_kIDC#s3x}8<8R@
z61;_X*i+82rqJguKe5i<*_z*CV;ZU`+|V|Tb2=s>zMgq!`RX&i5BpGfmEW(lRBfr`
z&rUgX)6hiYtQZGFxT0gSnwDS-`!wdMsW=;>RGd*WLGb^%x-H(l`P}fI)M);y(hRRP
zQS_MJilyfZpLCkmPbk(R9!<UETUG3f<QqTch@B5B;Zq8}O@ZhjxAAZ$2ic?D_w7nE
z7A%5goA&zPGHatDyEbppFhZ5oHNDmPu**3&GqX~a4|L|d>Yt-!cijyRtn}jL%6yR9
zb>>ER*?iSAdZD&CmluXto3QbYnrXPU0haVCG@I0YO(`bL%Q{J>AcD83W_#qS!1*xl
zU2X4BJ<C0%ZfBUA5-b3x>phM2&o=bWFpBlwX0DRbj4}25YRDfbFe+Y*W^NT|G4ZS6
zeFt&{Yzy3$wXNvyAn)5n4E!0D_od)Ti8~zr4t#kaDZ0^_49RFH7d>uaEP51D0YQB%
z27gL9@MjB8OQzK=$G<Wh%2A&#(y*F|Bh<gM`%VsHYiD{iirPwws|NHY7R|r8`_2%p
z5Gf*u!YZqDsMN3Il0(tmf1S$igGYASh!o{fZ9~=qdDUr8QBU0MVXC!HUyR9^-&{j%
zP)%*^-45PNnqkrYF?#pQL^y%5GICx+WT^y|D~>{UniTC5{%NQYG)(5xsEJ11HQzzZ
z%4S?y313AsldF}YvRNq(gnpS~x`uup%RI0Fyec)~HQZ+++=L1-YB&)-qYGot`f|U&
zYW|xBE=8&(F`;Wx*2I-1%H8O}hy`35f&xy>59{6zE<MP2*-y(*z-|xaY8t!iNHnrI
zAt(#19(=U$q`WPKYS#d7^SWfmaM>?7W7NH?dz;Qoor|UR{0#La9C>`F7VMJ*tA!WK
z?(XlE<-+qOseAr~LN)Z%=R;QGUntA18%~N>Plz>Q=RH=3*=9aOCh|fde_UBkNN&Wg
zOkW)z@xM#od5x)a1|Pqp*bghMs#r~dxp*>A`yF~TKxmqF&+74iQ|7;^;F=z<D;QP)
zPyre@8o=pyP^lM`^ScR(IX4<5l>Pj_rG&K1Mf9{RNNHQq0^5oZc#&VyRuqtR4x+WG
zbYB~KT;V2^kL}P7FP4vMV<f(UlSj79Mh9^m=(ri{^g!KVDWLW@nK|=3IAc7wyv3HO
zj@4r8Xumis0f-E|5ZvVcB1tQ7)0L`0ig0<OmM?Y!Aed;c@PsD;X|{~0zMh*3?Rf|H
zYJf`f-WV#fgj@iZ!%@_W8n=50nCO&q*kCH|hAP2uG<lz$hc+Q5E40(n&a#>Yv!yX2
z_iC3;dv>Nu$a}@lZ8F*)BJEE!+B#Al8-55A9C#fY7>({|9p|=>_mcPQxrqsD*V(iF
z_cCSRY=Ri_!L1&+Zlunho#1yoMyBuMt#Ne6lV?dr--dsr3k@`Miln}CWDn~N(|DX1
z?>Rfc=k!YDoRe=-l)y!mO)kqh|BSkLiW(=*ojZHZc?+I6H-3*>+MiJq{x(59UBWr*
zG*+z__4S`ndQLGL*YDTk&nfs8rTu-yKBvIW{-09pw-x*g1^-gPpD6f`3Q7uOzvySx
zl%G^=Qo#oln3CIwUTjrxPHF23qzmL9P+-Wzam5TJFr33s3msAB0o>P|`8TbE+Gks5
zT`}7~5-I*oa#a5=Ntp=8MzV2`HU9VgRp!wc0#Xj90j`UGGn0i4x&@csJ-F&lXZmsj
zsr!|8;7;m(j=P>z4o+Ac$KHLJ!Av687n`tmdt80_H&eSCHRr#QdZ;aax|w5;LrBhC
z$69YA-@yGkyN6)k(Bf$t+wsGiw4drp{xzI}x3n&bak=^^!>24K;97H04NkC==1f{f
zSy_LLVjV17;c*H6^gp~G6Gx-bn=q5unK3cG@nnSKKk2ZU&tr7gwYzOs$2mH7=;)AJ
zK-{)M(->%#kh;{QVep3-T7&K5D^4J|pV;l3(-aVm<G+W=0Z1+@hb_M!qW^e`?5eLM
zqR4qy#EDM>jWY#W^RgRyA;U8{FU@bUIYO34#m}hfbp<W3sxSR}ZFH@3nLP_Dz&K@F
z^MrcI$26?$H=W^5v`pGMT)kwepb0J)XgdQ0Hv>U+FB}!X8Y=%d`H^v)y^#sa0XZ%S
zqHt;ATSKd6BoE4s*S3z_PG1$Qp?aTyuH1<5K4U^}SzB8Dh=XM#+4_cVQ~dB|-)OJH
zO7p!9rDYg%-Emy2bA}&m(%Q{yqErT9cdI2bqFE2?`jo4y)r(g9H<Zg!?kDIi_j;LC
zX``Rwqgw59E6!D%h@$suW(~+b%jo-BYG~R1zEVvDeN_uA<`N1@)r#}A4t??~uTr7f
zZJm5nFAeVfo@!gw<LC8$!)@8SNo^74{+r4!yuTis-qft~FSOjwP{`Z_?#lfean}>u
zb6?cSoj|iXF_QZaN+ULq`%e<@=2?=7p4@*HQcWzKyKQ3PMCwg0zG-ZN#zGCPK#`1c
z45Yh2yxUr@smTU<?+PykDE>?>)tb)W6=`SM0S_KKewq+T;_(JV64>zKpLY6SK%?t9
zAd{d`q?jltWh3kj8Pv*hRRIcm#k5PsJ$`St*BkZ*yus^ox}P!FB}wo|Jv@5fw`>fG
zX1{U#^uok=jB7iVE{Dtrxjg81%9{&Z$-E$k3QfdakWQTv)Dv}F6+tYg9c-cm%5F)i
zlN)O-=J#o-vLV@|hF}Zw$&X#6gdif+Q)9)nfR<3l`j&RTtJW&>&G`*lL$4BqqM7<Y
ztHll*^P5@Pyo0aPye^2fh4vf7LKtIo3?4#5Co0{e(}Y_cxLpI{KdridK`Zw21h<^q
zTJvP8>%|(w>7eTo<$X?he@C&3f?w97K~oFtq<xJZ?xKPIKjGzS=mTIwvo;-~vmFps
z{0pfBX~L?%Mi3;c{ozDO*28om?@7Sv#GO`UsQCs7uB1WsK5p`$(A^eR20}*R6vGWp
zm~WUKjx|XKu^I?^PyvM+1BOP;I1Wnz8S&kW@&$Qbku*rs%&$kDm@<{`JdiuhU$2+`
zG`0b#6-|*Viw|y@_VuI7&?j96Q~cy&gc)uhRGLvs84$3TV~_rsM)4V~jt?qUB&a9O
zz5C3&(MarsE?lc>Yt8Vfs|<--F&ZH!H#CU#ifY-V;Cf@pISHY7tm6Z1oR+^-!T+k9
zYJ<CxP_unkW=1`E1>y1KdTf;|Roix0J&Pus$t0^{RcP(p#o5RKt=$6OO<+LM2Yo+q
zM?UK1$wGp~0Yr|v3J8(~YEKSWHY^=U7tQLn4aI%q3+joG<_@j0dr8pA-osz@WmbLA
z+a^v#AK%1@0}>}9Ty&-)9I{$mwzM+Rcy5Uk5lbOTaBT3VK(X=c1d78j=Os`a;lBim
z($P)ve=GkXpx4nlbN90#UK|tUBN_-nG&)~$#ZFE{g6QsRq5Ui(XUKG~2NC3X#nuOO
zy|^qOsV$EP@;OomO@R1GNOiSD;zIrJZ;BuX!xY|l1Bg`yIXYW49I+aP5JBGByasRh
zEz4&R7j8kg_~VT>)kf_3XW2Y_1xj$x8=+ScqwTL{8Ru=-dNzcO+iTLiMmUME@e$cv
z@-!9-8~@mz%>L8FS*hB#+V*wwn7DBiapP8c!ZYGU1dE6pcO!1xL7qJU!bil5k@<pz
zjloWDjGTxb*Z0yYgorntxbfG?yWd2U+iC}bZQeG(z;0r@z3mn|h;8OJH{0G0=-1sm
zBh;fe5;wk#UOKmcgyP0SC`w^LYvRV;CT`rbR@fkZiB_NL6u;1#P^?JJd#G7r#g}R2
z;h;bt_5~8!6-e*#_L{iwuX%TbFXci1_S?EW@Op&Nd-;=56tni(NZqI*o_bx1+F6Mr
z{p+=R+%xtgTa=^vaXZ)c1G)~p9(@ck=*0y}@3+!F&@8=u{$3Mhenz=n4XymmT<#mc
z7%G7s@D5(jAQ*0bQF9C3(E5(Gto6(&9%2;t+lkvYifA4l@D7`(?%>)Xo413E$YI2W
zgWeIwWBjf{$-nFvV+kdX@)ZdN$LuSJxmX(#mmSxiiOlXbk=c>8@eOkptM2TmFETGd
zCHEGwBJK(9p*6>uA33@@VIbY0`6mQ;0jW;d<)7uTm;f&NtD_1997}I+gP+yg(PK>v
zg<_CjkqCuAGJsq<1yNox0h?nMME(0nb1H3J=kJ3QeTRUoif5WiZ5zO7i>KM9-_|g<
z)k)z$tKjPu{Fpwbrufh4@f`~Og75h6RP1>LhIxN=37h{eW!7z+Eisy7b~I&M9pgYo
z!_<cPZ&x5R6yd}5$SMyN;hV$2v3gWZm3qsqm{WDJgRX_6NJ;MF!cj~~?t!};G@FkQ
zo@4E1RsRKb;zb2^$*SFR`jQ?$pr9#a@;}Jq_=&GUyi-riFJVS^Ld9QG59SoiEBNYy
zpO(1CaYFvriFu6gVzM4nHu`zBuT!`)es3F&@hPp7PHB#>OHG|1=fjGX6nsR1L^1xP
zf++>26!-`FfY4WSQsGdOZV`*#ev1O#%4nio-&3q9ZSl)`oKf(q8Z@iec?I9dV==Wd
zU8*hO#zHCc3eM=O7Zj7k#9vTsQGu<ZPbp^Tt{6~{YP<a~1^O`DlAI$J9lawrh?0F|
z@ZKZBtzANLRJe})wj?9F(X4M=P5ad)Ba(rnkZk0lhYWL~EZ;jKqg|7uh@sSyige35
z@PHHEjqw&>;T9C6?mlQ^B;HWBg=I~&u@Q5d)0oO?=%_Ic4P(_<-JM0ZMu`!P<Om<g
zM8E+vDcmflwm9b4$eX@$O!Kni=Pm?j^Paqf2POtP%n_p^LVI;L-nEFA!6r=vsK`(J
z)^N95BAeHrH{|tu14e-6i~#Kcx)0Ywhf8PYVA8Kt{#$M58XLz|#c_Mrv3J+@di{vg
zIEk}q(l{=}O+w2{rIePs4K!Azgi5%j#ok>%QolO0PRnYmrbHkCq6k`4RFc*y<q?TT
zDM%pEZ+t-@zVHbaJ|NzOPX$4Xl-B(I=g#boy=yl~g)Lvt+?ly^XXf5}?*E+sJ*QAb
zpryxC+3@F<34hklg28cxJvCMEn80SI+yxW(F!h&nGv<k=YTeH2)WTA6=7<@GYbBjn
zls9epj1LQtIYqN`ioummP-?+E5z2S>7~W*;@(Z!jLBnh#s)I9hx_z5A-#<CGn0vq<
zImEA1rw6CYz`6dC=oN<!1=@F6_{XZj!2s9(8sMfGcME@pu1~Xpr)f-F7Ap>-2-}#s
z&L}EeQZpD-V6EWpwg7iKj15q`ANr?h8gW$?#)vag9>h(IXLwIO1AXg5*d2`AY5U$y
z%oI`rUFTzwNHheLI!iq)kXmUN??BEEQ1}Y7T8f{Y$WF6ZjL3R3h1>&6jKoFatxF5(
z2IDHouXKjj*{k=S<ZV=v@?MSu@p+%1WCB7gUNt{CX7y#sbTm)CB0E)|&ys{ZUlrqh
znPj}j%BtQ(g?dlv=vS27ts2?DSbd~MTFvLxlVpeQXlFq`TFvlxfN{qnNeB39bbWKK
zstiVR`ikL?Dy)gft^D~#DuJhZ)NzKC@Pn6+L+fpEhD*>yFW_91EC{z@u6&6xQ@sbp
zU5x+TB3VP_h1BaEQurxIx2@+<`Y=bs==i>-2Y2IB3ezWF`xLO5Ey_J*u?GR>=QZmc
z0GcO(^`dI(=shxAmXjD%b$*KfZReaJQyzB|EGzNg2Pjo7Cr``aKl&Z+2FAhNfXHWH
z6RzZc(kN!~O#^;=a7Q`rpXNvh&n53Hu?5wwv$ju7a6G{MJIWwAq7kOc?Ps)=W~;-t
zy3XMt3!|zX$!A#qeS!14<tNCQf8CjO<i{MrkwADv+9uPn_l~()!LRuWUPZ}oNlNm2
zfp-9n5Ol}vP-(}I`Q%qiQYI_%8E9sw@w+;gCECq_+|0xRPLcg#1{s`Ai3}lsSVZO;
z&d=s86bKD4pPg46a+P4#$^s@-4K?m=BctB$Rf9h$`J<9QDfzP!+DB6<eQcHmm|Xl}
zhKToOOWhw?C_KD`69K$=xh&YwCL{i$YaUbbS0&dg@s<2diS-qA4euY?l^v&&!ZL_^
z*tuE!G{2%;n<8mOf|BGKW&BggeM;_E@-HP;x=jZE&2A;RP<SZt)cHFdepQJK`QGbF
zOw?O4q+RvapQu~i%!UXEE@+R0umbkRPWQzX3PuzOw6v?KSnPBL!UK!wPRko|YIUCx
z7fgq{8iw?#?|!ojvxFd%a}b81qKV%VURCn0p-W#&MKs#V-06#xvoT#7-%9oRjV&Qj
z8Vc}M^;dcDV@p(cwqmhIwzGzMOSvButRo?bI^1_UkACQtPRG$JVl?+zPohqs!T2co
zB2l9gkas4oNd_7W{Aba3OAw${E~<X}4|O_1q7IJaJJDdILG!f|q`~OOcTvLAjZ4U9
zf)XjY!SxAx9kgNS(m|b|Piy}>NRo`m$kepko~ql-IX~ilXfT$W{GV|~FM0zTH^`L+
zspnI%DRZ#NRXtoKlJ#}PKJ*>~P;o=O$Bv@BR<TB#G3-5PDx|B}hTa1+kN<bu>(BR!
zWX;t*#jA?jR|d*EP4Cf%-eW7z&o$^mn#{Y^F7zHdIcKoEgLk`&+}+Z8K%2%N7Fvse
zm8{hiVzbe#f5W+ZOh+<Myt+J)--fnhH(R@*Pit#$84Zg2;!vl_VZYbN(9(O%gLrp<
zrMDZFMsI=^VJki=cAmv!KtQ<OWg3v|BYW!5bhCeaEt=*HhMJDv{2*^knxc8?drf%{
zPqVkYmwQVy(h03u_vihu?B`A*Jac!^l%>aTH9%9)o!>+4SQE^9XjnVw&inIMn`Qwr
zvb~uh?lfE;=1zOL&%S!iLb>;NlhMUrTvK~;hp(xj{{Koe57HNH_;W@fkzEb{JEGp@
zU}}72hqDIfIX8b<C>!i+bunPw2yr(6u){-_fK#skjn<|w#saXB{;eR?pX;qM-qmps
z1x_J0m;l&_Oc;Tmwz~6Uk}OvUL>mzPMAv2ZcLgXk0SYy%x#3aD7;htIinTlIG3h$V
zBw$sMWD+&Q5%mV~og+SwVCH7<voCfvkkyg%HiMuFMvrP9vRq0a*8r?QXp6y5cho@V
z#Z)V1sn=RkJ^yJ*h5nZ7rCd$9!t%E#^aTu6tYfZJ#ok)xO2>N0`LGJeC^7DUwJ(Y5
z2m;+4e13^*pV1r4a+UEz+G6oo3<lL}a6wix+gbx&N36N;bx&vKIC0HS6^PzNcFhv%
zfM~wM?47cO5~97}Y@nT%he04}1GPM+m<rb(v45sR`{z*y9c$?Gmj?O-Y6^JdW`gD|
zF9>5ULGHq7YRHTZsCEa;1F8ckmNhH{Dqr5jcY2tK8HcJAeBZ@@DvV5zFcb72P_e<K
z6HE?^_`T1bnw`j6qdh2UhI3we<=8dLs3o2%?%}u}UUCf;d(Y~HtG|aodf6@f8OD?{
zDm4Eg3*v!8=#~hg1Xap?JbaaH(q`AJxAB_wwvvh7Q{awBWHyo*Hd~<%5%jgfP1o6F
z-_b?`q7gL>KqG=h64+$df3YgS8r#I+nW9TmXZYRV6i^ZLOfa*afm~R3`W41nYv9D3
z$#QDYO;%4SFQw+?RJiL_(1`*FGp~}t^CW%_<huZ4Cl)BooARpA?2RJ-LgQ$sJeN|Q
zg`|K6ia)F}fhkU=34?JjS9hJvFz4ySBM@=mWlmM3GF+;04SonTH^Oxql9fOb1dm{w
zC0_fvZY$b*8a?M?bme<tLd~bnDt7!9?=g^Gmou795g|@{W-e_XE4{NCbUYHu)2~J?
z^qu1=TClV)wRV4EooV&uH&$0yoeEpiXY)YieU)!J-0zs5>nyAWKCA)QsA59Ln1#3N
z>YA-}RJ)ppd)82%)YeTJm)nhby0kclIn#xwF#enINJTE$#3Jj~(7eP+PK|v~g|2E>
z3={9HI!X+WBxN?nTPAZe4|$(cD_zlfHQ*xU+l^|Qdf+q;{$zDc$2Euxo=7~n?Ti5w
zMHW^Ntk42m*M6%O*fpwpb;oL#yosHn#ZRQE(P&Zd#LQ1$oOp~V4QMQCFaQWb-r)od
zf~ZAdWO^H8vKMD!6$S@X->+H5aZ+p_=$vwGOyB5>ums6EZ%twHl3vS-lG?jb5qIm|
z^pM}r^CQGoSyj%$*4EXUp0w^AMC4b@dq$P9dVZ0u3w0V%6Hg|YtkFEHkEf5~9^kyT
ztVWht1&35iJ>%QMx&=ZBVq$9&t?2gtW;?#I7`l@pqGBlDZlXnw7-}NOi6t!4bs|Wb
z(X}F|vl0^HO<#F*equVJ1U>A;P$)g*Tzh_TVKKKnHxXz!vjxA%a4sj`PEUz^n`fPp
z7W-^GxXV@(4DNBQ@ns!;FtDRqVqio$s5Ni2Cbk&FXgTpQQ_6X~R|S)-YVI6wu-(+H
zjvH9%ayw}0i2qj`@y7{4i%wQ+HpoZhi9l$BK&>^cP<o^x>GS6Fu_OR+l0z3N?G08&
zH+Wwg34v~|mYPki+P{$}V8DHgLcoax1DC@l7TO|W;W*whO(;}Fz-<jgfFdTrfI+p&
zFiu#xc99fVU(ktIAX-o`{G(hq-k**Wa|n0X$>+obOMO004x%EOgK`1)ia4l2fEXRT
zOaJp7+lWw~?}Q32vC$KVjqA9+S4<tie^hLAAvQ3&dru=MGHb*}#>7TutaQWWh>cJa
zKOIO7lMpZPDJpG~v$<vdMq!xVLWZ|rxt)Z&pH>`mMVj=@12GerJNRsswjiFRan+|U
zUU=nK$0qN3zq553L(ybT-<!xBGqF9U`<cu!A@ePbHa3t{g^c%{@+5sKEfa)lR0_Af
zCS$rIkjV63Ro2fX$QbWyDqTXRk_<Wzw+J`6&qhLLM*ZQv)T$<Qq>6UNFJ}@%JBIiX
za%jCc*d~V)JJ$-KFtmN6KGawcaZ_XX+1yZ_Kb#xZgb+=c!NcMnGMiFB6g1TYGABYU
zf{8}UC`qGjImuF&Xbz2**azsCS@dF!mc_+asll>#0nsJxmqb@0@QgG{Lpb4mTW{}0
z>R}(hvc+Cu$aOXmW?%GzPP5PYU2T1jgp(SHuordw1ts5CvZA8bO0>`D#iz;dflFme
zM0l+%UkE*OHF}v5iPr3WQw2+yd0$u3NIHFk1NVjZk#tJ8w6Y?sHH&1@v5h5@EQnFb
zl#D0ic?X69fE18LvC}Cub3Sd7D<q31T(0G{=Q~VFCCyHk;@x*u2G`j0>!4@pUXwmq
zeyUhU(hAbrK|BT$OOP=?dq4n3>zEc%nG-XD6a2;bLe^hGfT2V;@<w)Id0HVGOs2U&
zg0^yLEGK*t{-`~+@MrDuZZ?AUh@)dBM$P5cC{B_60f82`+4yh~ilf;KEC##3Tf;YP
zw^c4u%ua%3tZEww#9nFeC>mtyI9rnLeZn5hII|7{hBGJ$EQJ~KXO^d?<_b8ML)Gu%
z7#eTT^zSzN6{XXdhGBa&N%%wMelk3?IWl@m?ZnNnoCiA1I&J;l#nRB`pQ|b;MFFV(
z=zU5=-pXH%!Rnn+b$&{(i?rAwouWs0j!HOI>h*7O%gbe_evmu<zvOffr%w>>4rC&1
zQ_N~<oS~)_<%}&sJ~-6g@uF7$C5zAn{i7!5b&d>SGLI$A{hWN*2<GMF#x>x;5_|&2
zD#k1ZL04Lu(u&AsD2MRxi!?Y3P=Sdn&1|6^wK(A_0CR_nI7l_AoRCzeCK$TB$3swu
z<2}|3KGKD$Y(30-W@DK5?#q$a%SHyLnID`gMn$%Ww55HM*<AL((&6mPF#nDaI@G!1
zQ5L(o;p_}~u}W;|Fn>oJ68yAUkcsBrOBT&)h@N!^X(GJ0DXEET>r9WTSQE#$>*!m@
zeU9l;70ero;0x+`bs|{iC=i}{pq31x&|6I)cD2|*LH&gOVB~U^;lLo&P{0+b8&F3Y
zi(z%q>aZt9R=pV33_gklBj^v4J>M``FsaYUyvru?CNS-dKd*PXqLcP1v3{<jztAIV
z7T&{RV1(ZAoAmaG;6oKj0j!)b*@yjqStXw%&sNBJH6UkWqJ7ItMRPx(x=*NTeUxcv
zjwPDu)kM3oT=pO}b%u9V&w8`Flgh6o@CBNi8Ch6x%NS3pMqki@?xj!IA&j7eJC~De
zn(^HDRqlLSgF6D(FkFJ_c@(c;m9q;3%5~aveIFN}5LMpX?4&2T<>}qAOcoCi#l^)r
z-=oq^-pfjaOvbzI2B&6aG#*(l&T@uUKUIe6>SITdMFk&Ti2~+M5?C{DiwcR3HeZ;Y
z5VZ2zb#-DcKX2!I9lDbV9?f9Pg@I$Y2^)EmA_nBfyL1Jn=;b-ehr8nG#a4!*xz)SM
zzFn(>_FM#{el8jB4=%wTJ~FXfTnqw2IS&TnP*nD;Ei_=&g!k4v@Z^`o8&I-CWp0m@
zS*78On5Tf9I_{!RjG%Kx*m%2i-K%P52efsY65;ns*PY?FJ!TK;9n&GL4YcJrMu{?Y
zv?&8@-sPH+b%|Osq{0$6x*AI@)#QJiq@^j5=%BMp9LrjuZD%d$Xg<JzpggTVVuo((
oiCf=5`@bdeuD+J0N8NAchHy*&d&V=@Ha(hY=dUYsr0<6R0eN4;g#Z8m

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_psample3.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_psample3.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..27dcbf2da311bd5601acefc634253693edcb4f3b
GIT binary patch
literal 58248
zcmcG%33z4aUEhEAO;@`{quKB5BYSr5juWpL<HVT+dmIzR&b6d-rIALu(wuuWi#RGy
z#!hTtlCU&|ETBngiYcL`P};K7KWWpl6k1B@=|-EjV0gM=;Ay8x{QrD^@42fqPTD?|
z9=&?cyPbFay}$h|_4XEG{LFmn@yhYb@z_`N(EgL=;UWHNUyjFOUTh-f#l6Hrd?IfD
z6N|}-WZb2tCeoIcnaJ`#xsY4TPvk8xwNO~>ndn(8P897qz0kYZH_^A)KhbZ`nT3JH
z!HL1BouP>#>Sq^*7l$W?EiJb&vbbepi^cN`ql;T7wpzTfFt)gDVw=T#7Pc?$nAown
zb7JS>u8CcXyC-%p?wQ!LSehs;?w#1XxNl;gl`AgnUpz2zz~a3N2Nw@b99ld)aoC>w
z7LF{APmC|#HgTIh_b=SOcy!|E;<1Tii^nI9TiU?FiN%u>CoMj>aBA`N#A)I~-mo{~
zZTW0+;tubax78baDL!$hx6RxBQf%TbZ-=*&|93wg^LBZ=Uy6CVE7|#bDyOfdChjF|
zk5?kCRJr$Ba^gOo_j>zy-lsC8+)v7W?*J(WD);gHHl7c9hj>0@&kyi?*gL}Wk;(()
zzn$lC?>3%qv*&m4e7kp)=cD%gpqF?$Hhz3^ndK~8o?Gxri*r}%D}JT4G*dWVuP@i`
zKXGDqu6}-H>iG21;)#Wo={e7zoU7JOc$MnXTy3s8d(4}gnOUjLEme;#U#&0s)8|i2
zEi6r)xaXdecTagU(|6C@b^4x}yYBVwxMSwtJE!iMy6g1m%H5|bcUSH>b@!c9ceGk6
zH(M$<T57&!W8U%QtDALkX}MCJoI5eMxa==osxbeJ-nm{??z{8Od+$4a?;TT9_ug~&
z<n*aK@43gj%k${%eN$6Y-f8c?`%a#ot~48Hz4*lJ<VvkJH(6~Z-J%bVE-fy<f39|-
zKDnp~s{51G+RT!_Sn+FxW5-JTpPnncu7h=UJRatroLrppCZBw2JQ3#o%A`;6*-Du4
zE6bJ1I*EPLOVxViN_}y%IyqbM!=CEO!a|fr{p|9jU!R*?AeyRQ^cMM_U!1&9DbM<I
z9uIxAccHvm@t4X=)rG616rP-`)gN7|&dkk@=fb{)rP*@*O1(T;E6**?hMCG0vTI@1
ztISNUEY!pFTy1f3In1)$i*$h2D@?-7!el)xPA{!g>t%W{xmXLgvN7tDRXS6zlxO^<
z#WKCrOX6YX$+`MF7bi8uDl;*;P`<Db?tE-+a(1aYx$t;2vybs&e{O1}&TgXU&=ahG
zeXf4>%$#4VKV8?$dmmn@FIB43OJ2o)_mktvaLC3wRjE&wYt!c|-pWEH9A>us$?1A|
zu3BFz*DKZ9k{=edVJqc@xeFCK)brS5PrT>p$;IV`iqGS~6B^o+OS4Syd*-gpRm0r0
zdOyt|dx-f<wc2F8UJc_{d1ofd$+yyzliuRF>4X+F`ca&76?>GwT8bcuaoFOkQ}4Z0
z@h|&x^-5`xQ7jy%t8>-TWsX^?etxc2npvq%YX(a7rIH3$s_8^psS}k;lM7a@P~yj1
zs?<u=rFv=ly!K3KvU+uK$*&xLEynDnYL$hVFr|0+2Y9^&H^1$~`K86m2`W{lmX<D@
z@Fwe%CuaSnmE{vpuT-C8zcPu{6Pn)>k8yS$dE%)P3*O=h>ng{-d7M0Nd9l1)vl+aD
zgAn$4bG6B-g^E|Ut{&8|H3zAre}sP>bJIscmOq=DCz+*sEQr_R*Ao6CL3};o#c#y@
zd+SNkQeONTm(0tFZLw?dwM0GbC4<<t#AC7NN7s_|6wj$379_7Fuf>DJg@k{io@q}@
zumn%XE^e!5y>t-!Ow7xCj*=hYO7U_p@y_*}m%kBzOl`-5SUnZ+w(YSs+H~>J*cxwG
z%gn~svh}=jdGouUOROc=azQFc2I&hif6_}<<6dfOY%L$;gB&S7?<)j(-q#c4J`?*~
zO79GEUh+cRzv?B|3e{YY_mX2VrH#dQ#MTO4ajl296zl0AXK(3W>kW$Qg&<S!33{%@
zx5jS7<FTM8DBKfU>$&)9kPV8&vOy-u^FMR({q;<cy%uMW^-i8+zu69X6oA4;n=CDG
zw1DcR`rKlrR<A7Aw7XVnm6FaW8>h77&&|$p7E8?~E&)QPE4A8DYiV}wQiUyfzEY~L
zEKXJY&c-fRN|!nLYP7B`Ypqr;u1qd0Ts>BsTdjDd<t4z>QnhBSsV6kmd};Km;|0q-
zv*ed5S8Qt?EzMa`zjC-%y5r=@wgF65J!_;^amU4AQLQw|Vfo;xlSfb8b@cSf=dC?n
z<+?s2KvFu@*$m|cc0Bq>_opi4x88EJ@rothMn!8+ujs0BS9_I4&t_~w^}3BUPM?eh
z`-Gbc*ItS9KWS}OXZa_{PQ+ltKF^z5+-Lo14CV@5wt5t@j@h(pVg=nw%ahBj6D_jN
z-cc6WE9nLxp!?CJ>3s_eOP4EN)Sdecm}`;C&AoDe*Sfd1h}%klulCint_VJD{E60C
z6Yb4iV$Mq|%g3#J^xmjnz+Ntv+Wn;pV39$zHn7`$wWI8v)=UA9>j2TlG;DmYrGUaL
zP9CdOG{QPl*KBI8>dgV^HPsr2234{Syv0aLn%k(i$4d{}vMpDpxrj^CleNlGJ6yJA
zXZMvZGloj(>{~Tw3*S0g9cR>aiY=-hHrDv98fz_oG`W5Sw61!;W~^dxpnC}$t!_Ay
z5u9-Cgeh&pv#;x3^A)U)PES_t26BLgQ#)N}i>>amS**FO3wYrsu5h=y?OK^u-Nuhy
z5<EQ3Uu}e7maFpl%vxNR!nOGKBrbxb+U><founZyx5V+2Vf;$#Uh@xAFDyJDFvl`4
zzvBs>-%!gVvDN*x<;m$vYdar*U}0%`a-sGPB8~Dl<40GwMPrF}X5)wet5%EAaD-5l
z{(xrfXS^5|gaixgnXZIM&SF@wQ(M!K^>3$Cm|Un-!?Z){lFWWsc>m<WO68rt4^on@
zEicT~!+eF6uV`0>naO2PSdSW<0B^i6>?Nz!_pq<?goM3br8eht>)M75<8xt7@V3nM
z4AX`Og?Z~zdCt2M<~i(T+go98M?;_s)CzlMz?&!_a2RH%SA470BbZ)|>cz|5)85}T
zzqc{$+Cv0|(L#JA-pAiaJn?2CQHZDB97z`A#dwO}0aA1E1o^2qbIDA6AfeP^d?;QZ
zU4Q<&X=GZsQ{zW*QK66WSKCRz<sTmdeVFI^_u@fvF8=ZO2x(h|Y$eWy8Lx6_Zo0yj
zNgHtzrdSDYJQXJVQ~aG~1<nghK1o~M;~HG4&IxY^Z!WuLCiPnFs-1EFDgI6K*YUGE
zda8Yr9bSn+8OJ8*mzVI8UTQYxrM-+$4=?W(ydGjjuh;AIQeHp51N<N4FU@a1e-@kB
z{ju0gTr}lN`H5scp6~O{in@fROhr$l-d3K{m5ldZZ_L~F5@%*2Tgk2GZ^SBvYrJzN
z;cfSJSbC4Q)7u68S@d>$d-&h$mAt+D@ALL~`}yDR9q<nFf51EB9p?Wal;jZgMMb72
zhFz=P`;3azdq$#^Z-vr2Zg1EUr99=G^iEmIDE&I^-9aC>dUtwv(dwA@w0F064=LNo
zyVtvq*mm!J?``Dm@E-8q&i|d>JG=+^zsq~bdzk;bqjn$h9(@UPZDNo2nD<W7OWqmp
zasKa(^3Qqi^4@Ll-RC{&y@#CrAB&}869?E2&xEPR9(&(=UjM@+=Q6V}SB2UN6AP!$
zo>|R-Sf5w`%RdLbBU<2fA@8qKF!s7f@CU~p-TBA={_gD${=v78{KVt`ZuR>g{DZH)
z_TT@>d!}A|&_AVoxA6b%!7vGOHC=L#77F{#@u#+GQ|z94>{@I+euEbTTfO+DSgrS6
zF)v{#;?^KmPJsBv=2P=&&_~4*{7(smgtx$cFK2>2!fen>m<tModC<aGxeycydxC9*
z#d>cL4@R#gi1*d|X{Aql?|OoF=1ChMWx$`P4=M-LIaVI3C%tTaxIRM4(E64e%<;82
zZ_Wu}rq-AjTT!`D@<;tjFh+c9K<V;WFif~D*h07+gqCvtN|57eN3fG{XRzCUvA!!v
zUyIjXd{+!aIBjJ<M@zedGze_g%U_3cL^-7m)V@!p=ubg!Ztr8T$NwVfpu?0_$se`$
z)Z;xipRxH;Fa%nj2>OG8V9+aG&oFDfYo+?$`aV)ivk(~|`+L2<wSA!5`}sRi-%ncy
z4Cx*NwLiFaC^#7Gy%h8RW3Yeypw}N93ij=YjX4DT(AvJ>!1}@9pwRJ)zrm>;gpk;8
zB!t>IR6o3aB-no=ULOwjL%IyvoI_BQ$Aj&Jx7BZ-KN{?y<T1|m@b$#nVdA%aGVYJo
zkI$bVev*|MxvrHvS<M8o^;6!K;P4Ht+^$c?m-p6BU(+tQgOYbp>qPzV{GID}-B8-y
z9#Btz<D+NClYW^E7bdQ7Q(js<2`u-arVafgsI4p;LJmT)j}5!eu)}><PDYVS`&N7J
zFFmzV6;8$FU0NMDw#+Rv=TL1?gXg%7vyie=OSQ`Cz`HAzWfgtrdq9P_F~j(AU~bRc
zEaZNre4%o+78ZnjmW`0SS$?X*RcIJjWpPqC#<GZ9%Pd{4%*~#!*N&IYi5zzL>&(i6
zu&8RuhkSu-XdptM7Bq0o@LDL>=?lwC+|kExK32ND^;3HG;it|%arSXW^v<W>dF~w1
z(%EO;Q>L?}XWsYN!{^@lbm`oCOB3&W>b)J+#xr4$(LH7H7iwW|Cz~PPM3BRx114ON
zpzL8GdeS|fn_dD#sFZ83>+bSDq(B$stx=%u<QfW;hbE2B8>m=M+=#U!LP;<&DEBZq
zy}S~p!4Q_Ojwk%1)C^PVWY<;ZA1CJ*`7@Mgdu(;D;NbBx43fEexx5)dM04k<;Lj7V
zl7*Q$Fb9C-O2t2;q<GltL*RjW)ZjEtdcf1A*-MlD_&r>yE5bf<jpoAvqbVx&ia)ms
zn!N-*w9FEM_ms=i3zM~4xvbLF$;Aq>0kC;bGdWvvyCv+MUYJ|v3N{Xo7xsu_RblI0
zE!V?-qv@jJ>@zSjf$y@RGdVFvs+EOWI0g+=US=uude~3KLSh%y!hy#cJdJ04;a;Fk
zwhjFuh8bp6j~(U<F@^Odj*UIvrRS9J$1pp!GBX2nFYHy}1{cZ?2b$4{5f=7bQXe!A
z;C>?w>bG23m|7~kT_5&?pTWIcs5BC?i%TpL<$H|N#d^)m)%{+*zR#9VsC$`>Uk?Xa
zj7q&MgmGn={ZXkYX_h`#{8qp2p{ik4{{7a-<ihfKcz@GuB7e#A-=@SZ8>*G-lS^Tu
zx*Vp(p}ZRAby=3{SC=bne(up~*{#3-9@S5AUKjkgE1IZQ!~FZiA=C!&FDb3R*}R)N
zb~BJAEUrw$iWEDHm4O2arna;g_B>_(pRNl%&Z)?<U!5I4;)}%;_N-JHGFt{lqit4y
zj%5D_l$Y16z?}pY*Bvfa=cH#`(6VazpHRt<DtJ-BhZTH8!FMS53kohMSWvL2z>dUq
z#lBm?#|hkiwS7ge1?v2(%KI(_*A)DqN@>$_JnY2zpH%V>D;bWjmW&;yId7~UkBt?a
z=xst7Ey0BCv1U>@#PNZ3t0io#;6%P!x$Gd;M6pq_3=7wPNN?=zC~WKTTHOD3_2_^C
zQBwXDf?57*nuNmMfp{jqoxk?qB8a=5i}Bs@Lc($^%vm@9`aS?UFEn20z7TqTmG@>U
zsT}<dkv_!#LOk(eUwkZ{Nn{dA8;KVaeF<yf&DGuxST&ucM~xrx5hb#El)pEA-cT!%
zjD9wVuVt(GIz$1;KuSo47r&lf%X*0*I~H5Z)f4kc@!q@?7jV9wSWl6Y_du4#X}pnG
zD+Jm3^mB1|j2XC#LVU8cBZLb+A_!N09!}&oAtBidiP}RTa2X+R^+LU8zQ|v%m!)OU
zvh_YM2YLo)&aQAh3th{%Jr~5)YkP*@N00ggIFoIuMVG363KeO~>9w4JO>+8Lb#`>t
z>2J^R23jq8LJbGuOlC(wN`t}}6tru3hekW7bU2-$nv7(So)_DCGUR$f?}vh-H+&<}
z(a#ahgN>|k10t=_YFqs+L63W<mC9{aYJS)oy^sL$0r8z5@wN`fJdo!3Ef9V1TCb^H
zjbK#$hOgRNy^p71Z_MiK^R}twde8b+&91e-z42sx*xu@1E5F0aKS&;k?l*qVjW};g
z)wg*&4UO;h2G|?DUh)CWS)aFyI0ToFTv`MToZoRHzP|HDjBz|C6xNWo(%ILN-2lh%
z1t^i&YlwFyk<7gQCO<cCcR;&ze)1Bc0pnUCt^v>RN>Hq&%H%Y{goxEk&9rBLI&Wx<
zLGqzasBfTUYgZZw-zAq(yXZ$irg=|T#$L;SpIAoHwNi6%sIy2(m-Sng*kz6Q@XnSu
zYhX_+d-2pJ1q_=CN1OL5NQ#e?P7EJztbMI?b)=amjAB``ipixv*#F4)7H@p^AG|Fb
zYZW#b*Yr|l=9D$FS<Y$Y7?S-ukSFH9KoF6EzyDwqi9O^`l6-Suam55R5!zkx!hwZL
z3scb@&pkT1bcS^*fPU0;UBl$sWp}*4O-+A_K<t>-4efuc(v871*ufxDQxgO2=%}Ui
zwNO&)wxFoHsxy&%EIY&PSij@6|2e%ZrCEn}wKQutU6@^D^DV9{y1VUfsrKJi@G~l$
zIp;)ddDgjR61H&o>{9&+p)re$rQ${0uDM!uvg-e`>LlSrz1C-UxVu&D+7`Cibp*n+
zL%uU^clzJ2+QXY{)i5VvpCr)XP-`I@D-M>@HsRg)@n59T-{h}0MW7q3kVtS-<q~3E
z<zQUtej9^D#V_~TXmUIE;dXAm5g1pAH*>JC29mqEu|}*GH`$xTWKMVHo5jQdZone(
z9Bi&4^|!}Yx7(f8#DW@#qL${C7o7h^`ZiuV>;DA*9U&pW;Qw_6KdInvD)=b{zokGk
z<^L0caM;dJ=bVM^+}-MEhCAQt;D*~;r?TrzhC5nW8xE`;R{wX@r`@eJ+U#Wde^-S@
zT7_CCa?{k)b`t%c*O*^W@Cyn=^VkXWe@U@_MsUXe6~Yb2%m4pWK*(?FXc+?AwjG;?
z#{ZJa{e1;GE&e}H@MQ&t7XCxU3<<P#&~V*i{3FGFO~J1#_zeZ>ldrw!|6>J)gS8N`
z?;|bCGXKBdm~3k{=IG`dvqO3-XeX>|n!lO=z=w6q#r!>lNN)*~>!}-#U7{2)run2v
zFS+Md;{G0RDzTR0pMw<>hZyx@?}r)otZU`%)=IXc6<6=Us9wU=YcM<XpVBe^1GlU>
z(OL6yOuysFv#Z<gFFg&bUWx;{M#T__xDTs4;1(FKL0J(^(#RaoIkMqn3?6b9pi}W*
zR=m)pbWxy~URs7iUGi6zlLfn1Ip}JmTgWx7j;;kq6aemDqiW?rE&CDbCSDX)+y`FV
z2fI86mK@>DUrQx2Usrq}p7>fKvD#;g*KA-qMdilN79u(RgeuoENpT_JnVaSoh-%79
z))PT$Jt@8)oUh@4lx+yYTd8a5AZ=pzOd#?5phVB>sUVA}{-i|i>zN=UbPqmND#*M&
zc5%-|ak;XfIB&1##C?1zR?j2Qj;;5ITeV(fi_ntzj3DH_8}qy@D0=RClDGXQ_;PuW
z@_dlG6!*&)-_JYp+d!>Au?vL7I!N}lI3egL2>C_hAlj3czmW1@1|0>x7jh4Z8O12~
zvhfc=t%dH@9taBS{Wm}wUF&Oou05}SV!;>d13@pST#pnH=uvU4A3iK-v{FW6Lid9H
z2Vxh0xcXgDt^9>FW52hn-tM*jD5Vg!G_p3(-V%Jv9*?}y*o=BhtHHobf|h^v;?LD%
z>!94q1>qd*YNwBO23xh$^vOe6AQ_|XA5v2&xvM$URkOFVCT|@I1|9%8y_gP$Ad2{X
zIicRH4KcPMag3vx76N-`Ftk4G73t9?xwV6gU}%@-)Mk~wygx`X*N?L|M%ZP@KdvXL
zI__Iq&#d<F+6cR<-vGq`ZyPpl^dPYjiw%7ak_vR%a=e}kDSC2|8pFXb3Ks@gT-hk^
z!{m(wBaOTU@EG56)?emGfP57pkzry;5DbS(ay-<S#Fw~PdaSZgF=i#yed*iRoD8;A
z;CKf!kF$lqVS_xAF`>W`w97N??8B%%O6f2q_Lu)IqFo@tn0FHZ1OEh>tB387cXEC<
zW8kgsJ2U4vtV<~NMNmyZ4(>qX@io+hkRMJ<I98uraJT1e>H@EGNQKbSj0A$rLthgh
znz;C8#))@t_M$h4Y8*US=@{G`D8XI1f-qb{cCmg}``;7gzniD~3#HZJ!ZWQRQR4V$
z#5)p~#(#wlgeleb?^It5h-M^?o2)zj$z3XuM~}*{FD=31OT+9#ace;H=sBG<ahzbw
zxs-@Y=L-tj9Z!t~&>#P9dzEn4FfBc~T9{v{)-JA8Do8~8jc3$aYlHLSTO1HiHF-;Z
zMU5nCp8tj-@e5&M?n0QWM!IqSA1XC%w?mj`LCPj}L>mrPO<T0BfJGh4CO(h$T3K2g
zWvVvg)CDhhGaD9KbKbpwzK3_8)n)P#B7GpBV#(zAEf@<Y@nYg@xs;v^m3=djbg={|
zaNk#RnS|xQl8Nt36hLuRdni5%+S|uPG!#FiSaJyLp^qma$s^Pl0eQ|P5-)B?s#|#T
z0Eo1HVfMVaI=smap4O>p{A7se1le8n2oW>^bX#$H<`dUY3mb{Sm;r$xRZ3J*I=7X|
zrJB!a$NZT|N)|l1n<;22O+1bogw!ffT5GBtmM$K-Qm&VR(v|Ybai{Q5;wREfx#F1n
zP8a5bXG>Q~kMei?`0?jUj)Icd!#LCnC`}<VSzbYg(R^XlO8pz@(i|~Lx^!y%c<EbE
z8aY}ziNYufq)uxG&6Kt#8(JL?m+C8!TBRc@d9>6Zd5#+G-FO;5Dor1CG8zdARkrh5
z4H{7~p7tey@PAi<QJF?Z{2q_vnFx1@S_tEH6byN1i1^eI-H=&rWOmn14EryZ4UKP<
zzOG%V^Ei?n0f{NZS4V7bbQSABWVA8T{AqK%afD#w=CBkA0o`^<jv%~y8W4Z>wUis%
zzo42MTS#m?<%X4H_5D7&ghWNUqTRh8izHxfBRCCLEBJ%dc|*e(;y|G2@aF0+8^*25
z-*P-YI_@!lgrL#W)++pS+9y@Wqd%%*>Zzk`4k*on!^Fbvo4j|m3}`N2`eEwWgb$^n
zp6<C-Poo;Q?5m-sU31S&cg;OVRjL>^CV#tnyhA}(kDD`#DALVq^!C51!YK8}x@NSy
zTt_cOrZsnn)HI@AvLhPcBkE-%C1E_x2Kb0l*V0x>K*Z{;Wk3tzBnzC_csLGrQvQ{u
z-%iTER?Nt~e@X1MtXr*K4WhYZVP7Xr+bZ8ZD?Jxm(xNZY#v5XHH@4V7%FfF`bY7Zm
z-f|{<!M~1v|7^40Kc|@SbN-HEdsMP(i8#8=cZf~#Gb)1_MXY&bJF-oB($SYZEjL!B
zHp2rX0E#+60&Y8;^?67U&}^w6aBexfaOAHi*V2X>VnRaRjvxgQlJxfn>2)~wUfOZg
z0r+XvK)oUPTyib5mPI!grGg-5fIoZII0LUG+=0Pbq=QxW-+sxU9-P<CWdaPothq2U
zevC?JeJ`MrTK4zyFrIfydrU>_AWBfukPj}cAQZe(hNUS)(#Z@3_Z;J+jaB)UuyVN*
zR{j#Ntw|<X1Wrmv%4tiD!kmhhpsT{PX1noIB+{IH;Q>gPku85-66W~##Ge9@|4@7_
zt=Vg?KoXgWW;91?UfS66NCkq_HT0L^(xcLJ0lCQyVCy`j;Vc@^qO#{9kGzDq@1LjD
zkE5%MDo797*ZGmydiDnCLBYX<@r;xdI%m-FhZrf)+Qr4jf!wy^^(1lsD7pR)Vqw}i
zmPT2d4woT>Nifgx;-;%pbS4#LgV|wx$|2j<C^kMZObJZ7+p#FxE@82{R5b&N#mU-*
z?t_tAE;}vwU!!PEe39*mQBa+uiLrPjyK0o^Kyq|n=u8-rvrdG0=M=0l&tBY--pHs>
zY2zFuM10T2*D}(0uOrK%Y*vaarXZ9>iX|fXu&o^uDx_Nw$}@3R3$lK`;H9tYg!Slz
z(2`cTI12g%0sM$PWj&-~oUq3s>3f5;skW4Ygn!oH9qN}s8Wbvb*2n@@Mss$vVCECT
z`s@DdByhrPE`OT%c-)CkMN5qLId-IFzJ~<UhV-h2h=)0Z3?S(h&(Ltkal1=P{Ara;
zfuPU2rc5R){v5mEApdMjgd+|IlBGul3mZ?iCpFg(Y_VW7%cP~ZGxs6Y?gDSUjWvP-
zTw4AeEv3{8!C_g-OuPhkn}V6Ulhc-XQR2*0e09iOoze1J#n%U@)UnJubj^dej|VL?
zi>#&Aq6pZ@kRUk&-pYZKMOFg4ApZgS!1?$vRh&>-{2-u`#vT@?CSh8cl7kg&E7moz
zjN6=F(TKD;5wS8NF}GK-fk6qx<#&MU6y?-s;pDaZ3AhN8riaMUekpei^9Mw$>U-i9
zFz{>1S2)&|!Wdji04iWQn8~x#G|=&*W{3HEL;Zm<5q|}CMs%mLiEN7$B>YFG0Bapv
zDmI<LXB(YiIk^V_#~4n%<Bu7d>9rJMs}7sF7%fRVS?tO(f4W4wF<jwR5f<oPV<6qG
z3VY!hUg1=XbB(tDNE2!}AHdki8Yic2<Zn8mcGlZ)R>4#eG~gdijHF582xo2KO^xmu
zytJNKAG=mNJ&Q8IrT?n#8rD>>CAP$l4t4Cx^Ne1*(k|d(q0M8abj1El`h#!<Q9dG`
za%@&lW;Q3uIZ#hZOA)>wHS7|ngiD*ZG!w5UT}>I_NHhx2N8|F8|9;*H%LAf&9D6l1
z%T`RYY<15{^+I*&auw1AB1%%r(&2A!s+z1F=J1<pM{;ti=3_DHgjoJ()LfxqIW@Tu
z<`!k^MH|LxZycMHFyk<TeiTz;)dgWF)HRJc&6qO&OshOb;{}LbJNEg)|B&i(Dl_uc
zR4VuN2*NkYcT94Tdd*`b>}%sO5*T<z1Z>hv0-FH{@Xt9!OlNuys~+;fbeQnHK@yFI
z-S(X3dHfY@f2vZ(OnMof@1|yome@--R5A~{OK?8fP~k-FGRCq1{iS?-JvpC69l7W8
zaa1H>zmk{ppYnRKW9q|-DYK4-12y|Umk4sYwdatP!`QVrWx2s&#&V+sVo~PY&CX5V
zs5d~po~Vw>!C*BE8@w!Pc~C7kS_P*T*c-YvsPG@vQj0M1A5+jel@=2~@E<25ognfp
zL?TEahF+~!scnP=4TP_C(#vY(1qG7?tAp=r=qyD_Yo*m8li^;4b<mXK!XC4IM02|9
z90%<K#9*5xv6X<cCfPxlM0M_F(KL)B{jQsN_k>jPX4<0Ty$(1%p%K2U)(rW4mmYsz
zkLMKArbpD-)RJ<VUZ#rVWJ%X4+Q5JeW;TZ?q*TbyRsV<@aVlKmaQaWF<)@Wc5c<P0
znVGvHoh5X<oH|;Vvj=+HHu(+^Hzqm7<Qr8u!~x;BnAlXjqyo=FiJhqQjKK#Qf<M%k
z7=b&~M33AebqJYsaKs0CFMqWM2qdpC#0O}@rUyj=!GYTYdX!pE8-&sZ=Qf??UJ~*-
z$8q=(>;c$XNr$@lcQeHCth-{ex?;c#9L~`~FyE%Rm{CD@*&2$9@`aDGVCKSQ{dXvT
z7tqMUoFv!VuW3lp<9$epSBnOLqpS|hKt*&F+NJOgA|hfy{NQW_$$6wA29NuN09vX6
z3EjQqV)}>zQtEnqJ;TUJVT|dszBb-?;`rH{LsO;mrOT!1(y@1xrufGL8?tw?f_>=K
zvg4vM55kxoYORih2?SJj`wmJtH8;Dmgic4qay_mIi5Mkm!#;dz$UFSKYbNpnhsFRK
z*IWn=3XKwrQQBUSoT^!9!@6r`Y*zLWiO57&AE1bEF#zp_mo8w5YL@_q!37Y9i|Wsd
zRB0@n?Al!iwYB{z1zlb3fd$5$086K`d`g|wp6g@x=Hjc4wr(rX0o{8jsQceKDi#2n
zF$NZXhoCr=QVQ{l@OVi--IJCPAsI4IZfC4OGxa#)H8!jSK_*lhF$|t-Rn%>ehP9cV
z6<wR~?}DPWl0g;`%HE40kjPb`4sIlvj2tjqn?H|9iTll6PXn)^63{jSwnwQ2#AztR
zlalurXrXXJQ~?4kuSnknsk2^hQ0OY1V{`SL4dcuH2bsYz0&)U(0KYm(=Uz*NIh03d
z%TpLsND)|e>RHTYG_7HdNM(5my^Nk&TmztetK>6AOn?lihF|7B2WUwr(h&l-H9yZI
zbocosOnCg7UJewRHKB|Xw+C9%?^LP6(lQE9*f7EHFB<P0wPehiYvVnRT+n6+^A2y8
zDO7K>f^3}JX4YVAkj8g6farG3nlie!PHzJS+i>pI=Fyypsjj`ALKkF=jRbrB>jNp%
zihQ$xPDtj>49xOTz+*%}bTrceA+#5@#dJ7pzfa)K0_M?LbcZegH9XJ-duFSxV^{JC
zsM*evuQr~elCQ=Mn+LUcDfbGS_FDYa*vI17BLGuKiCaqi*_a0#J{7x`jK`|EQ?c5^
zmq7yb`vLygEMn5gp2=*Qq79e8x?@AyE2u|ymwP4NnPw_*`3O#HBBz=2<`1)LxRlWi
z;<^FcW6zt0ya%BH=EGV0J~I^r(4|a8Tk8c_ILf~;i6jY36KM`Nnyx&RD4%p|YS&od
z0VSL9;4{3<uTgFFh(T}e?m5({5ewQ<w;>bFD!1&c|50t7k2UwttHd^T<W#F8{&!R2
z{f!=+Ip<$ijihaV|0{ILe^CjJ74I8}yGDXBGbV;w<Et*YA#Rf9-Zq}_ze^RbE3hf(
z?;7Y??P+1Z(t6S*&Na@i-ucy$v`qToZ#Z^+SGA70(B9Rb-Q!v~_D#%}Vysb+LMt1x
zSjU`9wp1&etS#xz?|F4|PC?f>9Aa7Q3;}&a6k~PqoI0-4gH@p<PFrmmH<s(Tc~X?u
z-=qB1v}w49rJIm|vgIz0s9tVki*Xx^1qN%z-M*=_H%F_DyVhO1Rd?~;9GfadQKcW@
zX}ozQr>g$aWmea~o#x3GPpEOxW*u=}3;U|_%~6w8XPF9NF*1fQCm-DtGF;ZxO0kHS
zE+d5(G$RF?c8b^{W>&{+RyHa8x`2+MveTNx#!o*H<>=<CU123;$Y!902@*Kqg3iJx
z%yNHm8o8*su-l%KhP-4=9|%<ujM>bu19x5)0FihgR?l8THya8C1W2kHa2-XNpexP2
zdKdSTZm2@UxjJJ{=0ySe=R&=wUW9j5n1xcoCqxpS7ko+-I&}Z?w)H*`r2=J4uUeYf
z^8=`ppnGjxu_Cn1V1N<>MqQ$%h61#Y=@qwakn(?slwpsQPYfF18)4~1MG5TouJvsI
zlppT^l)bdmBTZmh?Xy;kJf;1gt#4T$tuyxZtx~&z@0M^gXNccVz5@MkU>X&Pm*DCN
zC_fFP<2LShU^2`&DWj1dH#4Vf+i@q0^W{H7fOT-gxW%?j?tKeOrW0_E!~d{KiOb}F
zzl!!WwiFwu;{>yFTA+872LruX87H-a-L_pQ7)Lk0OxZ5D_MJ3lyqZzWvTgh6J~xY<
zb{ee-R>#NuA5tq#)GkaSA9W~eRb78lxvwa=Tfy5E7^hGt*^pWPONtp%|4!8~>Pmb>
z|1atBD~dg$*r$kLu_i{MOruXZQT}*;8~PAGi)c_YX*I&0i1}z*3I>{tJM&q!)=7E!
zlS<P(`#+?>K-jus?(S*L!mJMJy>ul5sU(OiBk2rA(|w7ZU=0oqo4rILl_)|=?L?~-
z6Y9inj`UE1W8L31uN{lo0p!Hm;~c~=wYT%@)W3ve&!<oU(|@?mX!|GXnfdH|PK+M0
zC?k-UILjgnGaAo_y=DU>0~lGcyA?4K_TOrLe^9|C1x8+^4fP7s4||h~&o&raYADPY
z=2i=nlxW*fHc2DS$Ok*Yauao{UDtGpzz|nc#wao^S>0w+uvz)nGH3l~)RA)vKB(Y1
z1<xxO(O44apy|(WJ|5zU=s7z014IoTo;y2!&<`jZ7RoYC!`l!RYsGR|u1=i&n%U0y
ze^Kpyp8^rGZkO5Ev)s$Fmvd9}J<1$X>k={fI~D9wuvfu;0_^$FZgEl%<BN+8>^7|J
zWaAE;Q<_Lte_nx|7NN`j%L+cO;1ddL$D2&`CwL5dBlElI3buXiX8k{|ZfVl}rwJ<j
z)xMh`7h})CJN`fW?fZ9$S#(L>Ox$PTy~TK@59P&o<chgWF2`TKFP|xrno|x^u1qdn
z$Q1f>dn`Yl8N-;UH<!v}S>=P)vuQfj@uSJsPwjs&zLxzxmh*{xGM~z)`Jc#Vh^4&u
z8*r5)p!`Hi;QXbStmrG*_1p~ycze7e+-`7yi9CH9^q#dE&KAF5R`{`r9#8i7vZ((y
z^YIqPrr!B@!|X#I-mt0PBd>1Q)SEXqZ0gO28=mbxU{>~V?CR~E)~gA7#}RM*C9XZ}
z>n+Fopm&>hyR)w+e$+dL#r$ZLcicPi5?2G(^|H8^b-k?ZW7yU|XZH58t(UdEZ0luf
zFWdU(&DuUbv5R*mybpN~RCZ(K|90;kFQq5;w0c8c+3fE%2c>5EgPKL}1y8p4%Guk@
zp*3<Q%{E^-c&^iIlB)Mid5?SFVm0==bngl8UDo#l(ev}(yS*o^#KC4wty$We#_Pbh
zTFxQwDeq}UaX6}T&U?n%IAV7G<6Mtk*#9I@s9JBDcCHo{;lh>7*z~ObEIS0m`a}Gi
z7VK#JNV+P7M~F`CFR+6!DT<XP*?`|<w)7Z8ks|3$0+Sw4lne3N7Skb>-MdtvGyX2;
z4hlpj;fYmyL?E_y#1m@Mi=i0|01%}<i?jQP_gg$~akNY+Q<znm!Fu|d$oit(NX^BA
zUU9QPKwt}Fjs{|bIU3A{a(=Z~FTfSU^lZLYd?OeK*!@2nt7aX&=E*Za-v<!A)~aUz
z-(Q6-us-1R)(1gKWD1oqnsFFTzTOa92gR`%Oyu$wOxUpe2Q9<q-{NCIpQ&>HHA-$n
zdv#FtJ^3XV3;KOh=eO5xm=5j^uix}E@dPlx%M9su2Lt{u)OUCTO1F|fV>5_10B6PD
zpa;eW&pM{+JnP}_Rj`jB9Ko=-h6N|^-D}wMmi-@5{*Mxz@n2KaQ}6`^EF?1A8ZKXO
zONW1&CI6g1ql{UC@lv=I78mA1a;qmpQr9L|nQ>#op-%G}6=*azbrpKh3zN%B%l>6G
zctvmOp$JN(X1g5f(rOqp8<8RxY39<ekqYwQu>T*YpJC4PsDUjj>f`6ji}JshR|h;~
zA~pZ-Qp8yt=q@qLrpxP|99+>UETF^Dc6WTOZiGXvxR`rlS(=13?3<ceLIr9GUoYaL
zgu_x<$DSSbSkuPXw;SuDdi$z^5e0V9e@wCOB^V!aGRmQL%j{@SVk|ZW?vnhU)0lo)
zfixWaUsUk-71-L@h<{#>KdYcc1U8quZME&%I-q{o-KEv|LE#7QCz9#qR)KuYC1kiN
zKQr!klvMpDAZIhiw!tJ7&b2f7AqNStYMPNMrBWF5ICEB7`P%AO7ocg{)OTQAbfV?&
z&S@l}&Ik}O7Is1h7WTFz{eYG-dPwW&9<67b`ww@V=d(Akp%8{9#GAS)9mo0X%ZUe!
zfF4h|!vZaNMl(=X@Y4#IH_6Ox$9!4|h9nCimU9A@W6rDj;;_S9W6Q`_k}V1-&_Sgp
zm2b1!oZ86iqFc<!>*7DvOlPRD8)^ASe05*<%$WIlN0sRm#T!3aBC0MTxc2u57+uVN
zmJomj7?pJgGezu!5|s1*c~X+KG9f|$bQyMGg=^mZkbRlElzIk7E7$^4LSPmVCekM;
zNL1A&Y3m76#Kbf!O6Os{m$?PBzMwq0OGzND>c0V)yn!XD037W~BA03?kvHDc09f7s
z$I(5~?AuMV`Vg40e3{8z7P`Ep=@kq@7Uj?a$kb4t2A&*ir676%0!&sbFya@kuI|{s
ze}Czjr{%ul>2r@A!<ogS4?n3V(hWAO?rM0|kn_jm?I2H@-{XV6juQS9oHG!|mGuMC
z{4Xo8(`M(|z*&>(@|(0)IOLpf;)I|DX~x8FXiI2%SCIRq`sDO^XH$rUAwD$mhJ~Xn
z59xz;SBrs#14%quNZT47hd8pw$PNeH>-?7TSr}jSI}bXZ6goLT2DN@q`%%oRBd9v&
zU|>maN+%x;W#S`$wtAp@hi=x)EvbbLJ3Vb)u|{n3ih6TCe%N~A6PeoJjf~xP+Pb>4
zg+nBLOo63B)#8Dn`(bw;A`DvsBOwGL!!5P*cu>R*SH}_Zf0Mk%T%PB}&i#Sim>2Z8
zpkPr!=cE~NBEwbV3XNizDC!Ab+1~D1*<hu(forBwKg!=5cM>$XW<f%{Xh!P%870b$
zdOjgZ9n75c8W>fOEN3oaUxI3G;_Mj)F--L^B_F5ZVEcA5y)6F-xxU#!7)`mt<7<&X
zN)B1&|G_Szq;_&Ayz+mQv_E06ZxpE|7MPPqG2Fx5!)?Dc+d8pf(1KkWbWQgZLyOhk
zA?g^oG+2ivA0ao;(B-6tNS<^U50IycaPFPb&bg8yR#r+HbYeL&d0Cshq;bHKcpPB6
zNL~|79$pE`Jr~X4kG(Znd(4n1y(L+Dm!V$Gwi<l^9)&}OJhct1TwXo+j2UlOB987r
z9#9ppEtMu~xIx6C!;+5se}_q2&7W)7!maL50UYpjU}BL|+cOAr>^?VW@~^cdYs{1L
zGa&-n7Tf<<%GhLK{ePp3u0{M3Y5!9@b|3Y1S35ffSsUb%ttBnOEYT9L9(t>#b*D&U
z^;c-O<0{fg)`_;|)uF;ml$3*|WD5I(B^cn^NI4F6kY3kW<|>h-_PSxs7jw9A<qAT?
zm*E1^HIrf?v8jUvnkAOyf>JDx`dtQACGj8&<&qIg<&StHB7&g^EnTXMyMR0voa~dD
zmhVyUy#(V0)1h-NTZTN;O+7H&7GQ3HZHHkB1(3Po<n#q!XV&qWBXv4sPU+R*uEvXT
zxnqu&8}By}h%vCj#H#22MfKxT3OW}m+-^MOHh!YRAH|36<Gr;}-XbL$RM6oE$h=y=
z#*@CSd(Cd$RL6<baGEFbF&?yB=1~~J$P}0o)nQ9Rx{9A+2%WV=wZIJ?cV`Z5mxS{T
z%wlGxk;QUgA-ZG31!SIYO)+V8%i*R*i23Ys#_k|nr|E&OVLDEY;a>kpkH1cUpKhw#
zWec`ux8Uz;7XSIisq}xFwC$QR0WLR7nfOSekX${|JyW;3ggfSm<71c9qx`SwY%*H#
zM@ZLk#{(3h>+drcba12brLvC_==q8;ginaf8LTSdmzY%`lyFh0g`u@%`Fe8gsNHLc
z+G&GC;^K>&NQ%KXt7YyP-Rz=HEtEgZ;)n<o!8cd|c8bA@=q3M;G&8L^Jw(65EaO$#
z2AUe5?%sxmmKa|W3P$I>;iTtaV-prkLMzab$`sJ&8p4`372kYnH*08vv^Gu8Y-&Yv
zfWmVjNH<Q?gAYhlvrt@@>a>)!(DSf-#3s1Bj#7aoQAt4@RNKKGPyuj`YETJaQ=;d=
zbi^EK{F91(8_ysNcDyraEy;g;>e5{0vach7lQ8_qN2WXR^I`n5UsA1OmWMZ|?s?1M
zBKo|%$Dh_b2^f|@>5aF%f!t{n?U)HwY0kupJZPl4rev$7a26;jH{bG%shMB{N!6K|
z2$DE$aU%y>N)6&V3T{%P&}=Z9sWFiJw)BkB`D1Q^jLj5|cPODy^&DKyt^Pux6#Yb&
zSm*CAoi|5w-5Hnf$E?w-pKql%?5v#Ydf+f#Gl6AtG9t9v#22NhB_GVyIm@;;UB-3&
zc`PHVt-^1|nrpHi6*Xf?+-g{MiLB#yDbuw!Ew>qdvkPvqf1ar{PY)8`d_V^#ccD^Q
zM$N3|wuEhjOq<x_zw6-qxq{Ct_!EML`t%!0_)`LWNa6n(xuU%8aZY?pBK<#8Mp2_I
zFH~mgc%usY?a7rAI|5U64@YZeLDzKwAfAM5DZ44TPrFFqtCW>AV~oAT8Qa+nzS^sF
zh#j??L>St+S|`G8vAY<O4iy^CmQ$0I$kB*stfNt`fp}vFX;<MAD9OpgVH3}srxEhe
zT7~2jxhDn7Z822c)hfBPnzdF<ebP%ut+JP-SQ%v1>V|gYhCPc?ByZ(}f@bxb%Q_an
z;l?dWn7k4t*ouaQ`dp<v<>Ou#c1yF05+iqTQ6kk^iD1fK)#BNCPpMWoWPXI5&Q-(Y
z)Ye`Az|eq=%bs}hYM6NPiiyy|yn7#HXxQ^)^m4Jlx)!!qmrS$h<67CByg@7#`9L|8
zD8OEEs9EakV;GgP(!07>x}y!7hmQ3gqCu^>t@jA8fU_BU4k!FF>_O2<d{x#OSON{P
zEw-JD8xfU!0{}&%bpA)_5c#IDd<nf2X)a&ldZL`D52W`gJO|(k49Jp0s#z3&b*<M@
z(s=3R(&KW3=$9}W{;SJ}%PE~_C*sx{B}ArV_C|U=$A7q~&Nsld_&SC%Kz5w{-$-F&
z-yc19)g6h=XfO7EGG6Vo-lqKj74`OieG|Pktdw4`ucKZOgPZThs8gXF?j08+N8=vs
z9%WQcBhEHTGUh>>llC#Y_e8CI6Ju9-%uzUX8|Pp{TaCG>zG^cdtNxF;6;$khV>Z&;
z8?!;IV>XoIX2a#&YBp4^F&kEL)7eld+joIt-*_jeR9<B_nT-u?HD@DH??HhZnF0PC
z`e)FyAJ4D8+`Jtc7RvW?dCUqumGes!P4JRGeYAA`XzB8Jt8DW!v%-AAgl|DKt6)7G
z)dC{nIMXalK6vu^quBAco=-jBYOjTh;8eS(trV>6xdrD-flt)nH(o9sBhTVJSjb@>
z?d(@{t2xzGgEkP3ri*KGe?y#2xj*wvbw<YTsWXIr4m{S${*5zbd`^RR!mlF+wA1{|
zt7MTNOsvd=36H<28cr#v{pJPeOL5sxt4xpH5dqLJvl6{hiGD`;y$uw=4XoH!{B$?0
z>^Fp5?YVZI6|{Ve{`IBDlIBYa0;rTYfDHeDJWKI3mPjGv&wTxO@@Qgb{5BB%%~De^
zYgYGc234&)?rqc3BWnDJvC`zNu5)*^*i&405TPc_8AN{+3qPlM6XdyAxf}sZ5*6WY
zg&4vw^l9Lqn%c2+XwV9<x}PSVuYH<BD_TO%i3^a((!J@qko3=yF1Ioda~*S`t`)0W
zbi7IV7U-~IwXfM9bYPnAEmrY$C{Kg)MR>j`e+8%3#d;r9Ma5$M&><J-(SY=J@V_UY
zgv1;5EiDlgaXsS3_hCeKjq%|;dw$UWL23+%azT%$563Jxvet;`;Q$8>|9yzLQbGS*
zlGcY2;8EZDGw6S@v!)!lc<Ilg4a8N@JB#QJ1DRi<3_cI($51f1zJ+5yB*YgsV9x&s
zE=Af(JZJph<ayNT9|eOPZ`np70-8jr6&q4q;(}fBH4%*4%97HANxs;G_l_QuXqKUp
zDhuHG@`8aGZuyG)C!fG2?d`n;<69j^@uE<NCO_BzEj|9Wf`6jmWp;{_yqLV@mzDOS
z(u|p7B$yq{SCsZ&m3xg?xJ7!CEBH&{#Dlrw(qe9gZ%A!H!Jv)qGC$BL8vF<9={r*t
zb^d0@hufIJG39QLRBPI}8D$4FP=<5-iD}1X#wV=ta*(4ntpS^!b(P}}>EvQ(s83*&
z9begR65Ke*p7#HX2GX3VFg2~lk}~5>UBQaCn{}!w{4Sm5bBw(YonDEO#9SUiSR``?
z$6>@{TtY?=a-A_k%mhhj^+-ttBf=yGh14<X(`rj7B{35?)3(h{w<&vdFsg*z>F-hg
z-gtoko6Y@D=CK=*F;^XHGM<qdfexUoC}uM>JLJ+l8a<F?xR!_cBzHb2|6WVzqXBCu
z=2D`7zfU>Zu+qp%*r_mTk!VzTkE2J7+=x4HeL+(wiTZ0rBTbsr=qx0Te^R-ogC$mA
zm}@Y!!XW`kF`4u=Di*ZvUp3ZLSYYWQ69nUwNR*&*5zE(@B`A8MyDw$lY00R*qL&;{
z?<_DGptjmCYpG=Rhr$%#Rj#!9h;5WmZRDWca<={BI#(wYoKzt7@2+=;J2$ph)^|Zn
z|L516onqD7I(4*r<=?82TeEgVZ?T9*cQ=hI0SCs36+7iMC*5<GWY?q#Cj74Y$Dm@|
zH5){E8)ZNp^uKtgDgNHgpD}9zgreBYh67M5Gk&mAt_O@C9nECASH6&mhA?B9INxHZ
zdX+0q?N|H?cT%4wt822Ib%6S;dv3^1(&~=xxxAGvM}a!(v?(e7HzJxkF)EA@l(`Y;
zF1MaI_n_*!{T#7p<&xVWM&7a{1SOUg_#QSSJZjHmzcS*%bsDEmJhI$8H}Fc_p9Q7l
zf9*`97*R0t6l*)~65Yy0O2p=S18%BDpUeWJ4{;M-kI`{WNNR~Dth8WNV^D-Jgoyki
zBxmtt5gACOBLgYa;%;Ny4x5u1k(o-F5-I4ENsJ8_yQwV0&stZAo$FbxN$d8o>%#7b
zf1!+DQ(*dPrh|E0k23^tun(%m=Gx*M376e&)tkb8jOg%R&81d1-=%$2#8#=B-)FjS
z|5jT-jMqb;PdMH*a#XT>`6t>JU)|chB_a)o4rd9C{rf2Gn5zmH$5=drZe*&X7Du(Y
z*2hH3TpuJ!oGEC_Q7?f3rA+K`Q!y9iY2s%RP?u(l2FHTy9=#sEKNF;Cr{Mm>eSn23
zz5$%Qy_${P3ic>qq*C?ox_6t?%#6tKPvgi;Qrt|_xWTyi!a{WC2{?YI^2QZ(9`>*Y
zr}vHTGW=)F-U#nR!4y&+PBT~8h`Hj%TxG*?YOA>cP^M|c=I;m*#?0B5`e?%eOMu@d
zOGS&`rS4!B92D3>FB@8+qs<}p2-cUcz;^RDsD_ToV$m*wrehda@ar0oAc8zHAnn`0
zG{05F4#?WUftsK-qYI&|WgB3OZ2$yZO-4APJUB$AB-=94!d6GlH4ru68xiiM_?jvB
zOz@NBOK&~zp`xxC7|%}XI5o+&$0I*gd@ljN&hLH_UsL{~`FNK1D)Xb{zG#5KG)%l>
zQ~GMuQMruLS0g>#P&J#Hnyc52A$Q=LsEaM1Pxg%sYm~2F9fmi_SNS4KD%wH7H^=wF
z^wl_z?~ztL`Jpn$S&>+z)lh>7mG+Szxw*QPo2(7au^L3_WXWteM38>)?DHBg!r&FY
za(M(css?};D{RK?<7=)Yp7|toY4yW_Fau_HYxA69y&z4~a=>yb%%W+{tetYtryI|w
z%~LGU$UvJxx-U`B0lp^{8&X?a6dP3VDFt69Xz<lc-}~X_Zp~f<WXa{ryC{R-E18qG
z+)u-|U0~<2q}`{n;F8Nz%U*&jw5FS>klP9P5+%60y&HVBicf2oHh%UKX%c*wI0WoE
zL8vcr=CY<?3OZ<>9K_|69zSLqTt)h<rGz;-p2eBZV}vfF3!LtFX<%E%3@1_^PVfC=
zn4lM{a*q!8jOV+8EYEVjjy5gNkI+(@mbq|fsaVC8{xyC5N8+`-d{KHm*OTix%+QOU
zm+%PV4b*db5XAL<4(&?mjm|mp2xK(C8!1RAS05$egVcN1`(#W3lG+FNPhU9d_rJ5*
zLLV*kx3xga_>!cZ`GK}nU5i$V%Uro_{erA5dIOloYdm~W2#(mmNKAD}Y4lS@>;q9B
zRexAz?onHf5<}`ovqVAGs?8bw^~Q`g*JL<a6MdCQYm(v%M(t}N#{fx)NqNqc=I0RK
z2}-UFW4f|MpG?JLW-ml!Y<{b^1?W14X6A@o+>^5<*wUE2_3ey-Z#B*D;JZz^V2d|u
zPJTw$wyuo@W3;d>7@pq=k`v7~b2C=oMf&bwG#Ep|vgHQK4QtzIVGn&wds{JsNv>_D
z{*GYBruBD3^|uE*n)P=E+ty0~?^)Z`oUJitYggO)Y+K(;UvYZCj?|Z;G<rE19<irA
zHu@6J*<ctGo9_Lz*K$gm<m_v{aT{;k+xEuBx@{+YUt7ABaVcA4-VWDGb7|08>zy0d
zdVH6Y)lE0fW#d05(p;c?vwjVRfCbL4!3F_&!}`9>Dlu}5V%VG0nE9Xp(+j-RdC8t6
zcE*3d5}#G@9KmXj^G7f{$yYL+v4-e7|9J%;BKW@^#Qlvy_y^Rcf#^fTAWIB-3a?wu
zxn5jYX!OUtHLZ@gWP5Ph1o$8``;FDej{+Zb8Q<6DK;aQlWM#eRR$dhR7nFNRv(iFl
z4fY$kYDOzd%Kf;4f2QENf-zM)Lky22`t}a`xyx7mUsaY}3x*>bQoLU!?Me|!YT`G$
zNW#R01=9=|-|z5W8~;P<u_nQPK(V(gc$<Q)N)vL2k6Lq4<v1Z`nvBVeIcsn(96S}0
zjWl;0Mv+D+RY;r2*vy@th<sc{{8vky=x=YdtnY<YC$jeOYy^R~>D{`-{5#Z6)B$Ig
z*r=(Cj<>B;Cbqh!+V47-G3YW>D00p-!Bvbc?uCVk(T?}4Pq@wID+W6@q@&%U30LM=
zN9L82tsFd9(5@TWa4X&{;I7;=8keyMM28IMu3cc1`X0@Ia7FkuLVz3YD>8}g`16wU
zy$t`sCUIpUTUhsffD{&3L-1;`*h&=A+i77GIVU!<l$CnY7z&7Tyt+NHo%mRMC#kr<
z*dR@BoMq=dPj`Rgr@_H}l!Tg5sIqQ^N`<^AfI`Dx1);=D!05;Q<RDr{Cw9~*$efT=
z38{3g&)``m$eUQF2P?$!pa+t?$M~97$G?Z#5}YQnjg(iK-m0zw^wzl49}IeU#{Gm{
z<thJNT-5qtWk&k3LB{_!N=oIa_ePvc0ozEzKJivb$NWL9BAQQ<I<Pj#RhEW>E4S4+
zr<PbF$ojSDx!0a8&;0o1<P_^+9fvY=;6TU=iXGHSiu2?fFv<IN@^$F#xXjI<rH@p*
z;?AuE1pZ-FF;>WFHMp0@1`D!DiMF}*LlNG1hfO{e1XVcL;9<2sOoksL$74Jh_S-8Z
zO^1Jo#erD3QMBxQh@b@oo$EZY$Ylb^t)tI$*upF%FisC@?)?3ZI(@r3eLyh_Z12h$
z-iNezkCL8ZN9v0exGIC!Sm4Q&*x@{iuc%$?IE?TrH<2H7c@H$adAgTo8$Vl#$n@8?
zpy;wqT?n$|w%Aa0Y)k3Bnkz;&Uy&`E2Ybr+d-;YVI0U$0&c7`fK(;$bt)aD%YMQ5{
zd(sEk+1~wpH!>;bf#56!lp9n2JL|^0WIHKE?8_}U%P==cVqfBpumB65PrUQtbXve3
z21goAcZ8$7e~Aq@iXF=k7+uj^67u<=tugSPUbb@ndR%xKefy4ElpC@9P1}HiE1X50
zX_IZ_pVMA5-vVad|EwOrjbL@)Y4<h8l7}>ouhJa1$@ed5!R?kL0*^$Gzo3}$UHja{
zV#JBIK{)hSg{#21FMpN~vPpcuL&e{tKteSCy(*X0TetxH<7)7Pg1eMym!5o}ASvi%
z%CwGxe@ewW2NL$wD~r&TlkEM9BxJbGEjQlL9HcW!wu?kv#yg0K0HY1AyTzZ-guI}x
z*oAke(z^EhU8JpPziV2yr|_;}zi`sOno6T>U{G6J_#yl+rV<C(@jKzdZw6^2zjvbM
z8b6vX?Ra5lwLW?i4AdmT63!0*TpI+J6nk!W^nd7;SQUUOO@KAxvvTLczLvo(U+%Q@
z6gjU#qfif0RNm1z6)+LaspBvLuD-Qlv`a_uhM|_9uV-F~V+Nn_pZJ(Oa!8Y@%_B#3
z#PZ=j`OX#*z}xWGQBA9jSL_=Huf`Easl6E(`_MS(no1y5ujloFGISpX&3R-N`hxE(
zR5R_VYg_6?eV(k|3%LJikcX==$mz=mThM!5T^nT#eZdgSOHmqWp87d4sWp9sYry|o
zgoE=#d@&*E?7w&(4z@zOF+#CUp)v6E|4JN6Be_2M3RH`zK>7>S0u~<+)Xo^!8GQ&w
zp}Hhd=PSwA3tUpeP#sQ`_Z4^1)&8ktkRn{m1zQniZs!}T=;2_AhC|C>*f8{AdO8p}
z4fwCb2fcoCcKN5l2(->N6TqeXn?R~jQ`*j;X&{{<>RsbfyZ8z*JMrTKk>;KF3e)^G
zf7mLfvZ7o14Pl1du(9X&NIS9~V^C}mu@7rZg5r&3sutw|$jEVSUM-|@wKUF?3HqlU
zuhA>DH|;e=B0;mM2JcMg1r=g}AO0CX!j6#L7LU`wPjY(90>2nreWaV4ARj(hpVa0q
zaf0o0n*fnE!lG%HPnvV6_*&92Ctu-(*yi(*!B=?+!UcP^VYIFHe>YiwL+>w8IulF8
zOK<8+Nvrq$ANQkUdbA2+JKFTf^25HOn$s#YW<2_a>IfY%&_YLbb2_ntM+0ql>v1xE
zsxez*vH$&EwmvRD`o7l0mL9L5VCPh+<u=v7m(GNl^H-<PxHOXO9Q@qBpq$oS@ga4s
zte^!@4doFW9N+F-9vGKi3*jLCeHx@)4=whwX#t8+<lnE{yA^v~F}n*stw$qVjA{R}
z9>1W#*!M;wJ)y$J!xt!a-FzEG#&@_o!&clVo+(nD3h$Kk%{7*=h<-|0TD?lbb-di^
zzMvOuHa_j35TESl!mP0r%jeuM=T+gf-u`w4?@+OvJ_9ico@I`k5DwJ*cj$!$2m@RS
z)aT?9QZJWe*sbw-&8Ch;xMkCA|5=vanE$D?>;v^_FYJcwYDuA{ArY^xpWK^AlG0#L
zB!=FUS5<wWM=s!tNd&u~|MEa1cTWZCm;=dp;j6i{1jeXBQqKv4@iunwn>qy3l4CZ0
zgpp}64I|Ui2EEaJCG`e4({Re%?bMGBm2RMI(GApR_s;<Lk94vIkYn&+M=NoVntUh|
zn5+BaI2>!@vQ+?8f1r--LOsv@(QrO&JVWk}wp5UMC&YRl?2I;Q?=cu%kQ-vBNyo=C
zn{h~#=PiGT+huNTnEv&k;S>kv_7N{aETq=t6uj5}F2X);n;dt`We4|h;WEPAf^ZAx
zGvacIR~HYqaJQHxv%V{C0_+ipinzX3KF>{o<CZj|cD?*DxJyIaG9%m`qudT)npj)&
ziB0kHxXB0TNpEy}{0#A|SG3#X=S*Q^AX4A>72Owhd-V8U=EkCy-5q6a-G3~C#xTYK
zu*2TAJH)?78`yw=srH)OP2j+<YI8nA4Bki!6tqtKR(BMBi^X?G+;BWQ@dgDpJLh&P
z|8I_Xgci1r_yg*yudYt1BTI_e=`_G~QI9VuV_7i?lU=VJOj6zE0ru<4eMp5@6w~4L
zFDtOy!ved)9;Dju=wzOyY%1eaP3=Rmd^%Rj7dnf(IfE&EY!6lA9L~7g&md>?>&j`I
zK71j+C~|2tMb!?U(B=<hS_Dwzr$|K4`m24Ez&1RtUto)hGUo{DZzi=?h|s&c>Iu+0
zjDY;T5qcNh`||DcY8>=V`La(rVOT*M^4EplaY(@yV2gvcp!m)qMadMQQl2fA>BJhC
z#U^s%PxACuXoWweNpIedzT{wj*TpC_Vt|X*uaLQVM>pJZi1k}y4gQRp6{g@D7;dev
zP=tQ9J*T!eFgtJE_AN01e_roujp%|NyGEp|en@X!_~zc)yxF;_yAHu7`yT=BM|gq%
zVFj(b*lud=XEYuUD$5R*;Z5JTO-KCAX$Ecbnzxy;6|;=R6ti=(eQmp`WalomBN0ri
zqnoc>hbX*2$1?oAo4?w50sz`R32tTF7VU5mG>I80$7kRhiNHa#Xa<qEmq2CKGz>uA
zo$7Yp03rM(NOl+mB5yvM+^h_{*r3XpbJ$3A&5TBjfLNO6Z8Q6brZUneXjK@eP*_0y
zbP>TM(qYpj!Lq>;_?`pSMQ$#?UvnvBq6_5kB`_-{0p2#v>pof(#L#z?xYN1ux$kAi
zmMwfu{sAnW+$gGRHPfLiquJG;UE=?V2(zNA_hf*dC1%<w9`r+6V%r|206}WoEKj5$
zA`!Geq^X&z9>GE@X*OK91u}9#I6}G%Y#Jp*VC78r@=;=3#jFFRnj&i}hZR?kZ0Px-
zHDai`-ZE1!qIF4{d1D*IAO~f2X>)2N<wLFUHl^aD^eH(ESDEg7+2KpjjV769dza3m
zC@L6YOWx9I-zj}m$$8Kt;(vxLoN(E>5u|mFlYU4s(=YjBHDWE<y<!@RYvhKzrCQcu
zo7$m_=v*or25jsPE)}U-*hw_hvtSNEOtsR#O`ZHNdecLCu^l-J>{41I_pr9cAu@9y
zdeZ5W)}O2)6p9n4i)lc%FUiq2xonPiAHHwU+aIuD(*En3iRl4?#*c_Som0Dqewu#Z
zg2mVtRV*Mu?4*;ckMkv*w?5<TnlTk8RA(@8J@Z^zAKY8pvW6It9kgX0-xCnutUD?W
zt!L-KZwu^?6dvqG`Oivg_e5I2U&Npl_8Z?q*S8~xAw)z333Fg8WwQ^&s^>^8dO0+I
zm)K2u51(<g|9qNoModHUqZD$y9?Uq3AaSi0q+1L_fEGTzx^we3yuP<;0k=4VDR$=7
zA9T)fJ`k(sFURWr!B(stKi}vnpGefR^n3?lEyx#OKVvPz92l+-;iqscKqENvQ@G7E
zhxsreNbJ4FgaFY5X~g2Y_M5OI5LxOQ>o8RQygpjr$~VQgOY@iSrzgE3MhDvhRs>FS
z5WDU`<hh+P2uPVZeL{0ru#Gl$c*B(38H^!_jjivI=v2t(#b2asDHx;dSbc9W=8dps
zyNK=bwpeUmuqzmI>xrnBykHk&*~(aCT=CtoL&mPfKf(RToA!euC*Z|{QFz;~r-9nL
z*Y<RHH_V^nOEI<69sUi~o87#hFQ?J+c3Ljb@{{!PK(L4Oy}@46caWab8|xx54%H7|
z(<e6@Pe<Gnz2OV$pFjhst2c3P7rXRU{aM}EpT>MP`?Z_*_j-HmLy^Br-^LkfA4vZ`
z-lY9ovM;`WA=n4QYsA~j-rM(C<>Ic39_{b$5weVZ-ht~``m}%TfUU@W-gPiI$XNO4
zrFZBvsm~=I;H$PCx{xX7MAyyUZ+G-I|H*{^F<8&;#Oc3Ig#L3PlF3AJ?a<ocwIi(Z
zZS?qXa0KJEXs@h|dxwK@ba-L<1&6&OoRvaw7y^bEDMzA|BaM`qBy$jzr`GrkXMAm(
z5#GiKw|9(CUkcFvHV>WDx!p#(d+or6@g!>Zb&SV$L!*by**~f`Gc(8VR<kD<A9L?L
z{(tq}m2cv`CwOlUGyJoR@+5B?vr*n|q~K9j{+Rw|#Kb*?R4TD{+uH4$jPS=g-!y+9
zxLr=xI88@`qm1fK-jd}lzgIuLexeR5c>UxJFeE;Xn}|*xkbwMu-1O;rmv=WlJ+s35
zBcGo4oW+f{e^3|AAq7Vi3@HHe<3f=z?6XfwVMOD;mf_z;c370IU&FK+hSqk~*{PsO
zRKC8ACg2C2QA@gT!=CdiiwJF<;lP#-efdUHPW13;-8;I2oi@QI75g3q->cxW1Yz!J
zAxLlvUrV=PYC^pHt9n-wL+zID%heqXC#0osjocb|d|=hY1|Ao1C+cP`_;MTSI7lN7
zi^9a=WWy@~pAHgXe6tr@rkc#htH&liKD;np<{$L4DN@Q$<jq5kj=Jxbn(f5g%p59l
zGk&E~nq9%mGd8`iWpCMuaAzk%(Wrui5$|H*_Rie)g~3GA(8yYOdS=!YFW*caKRx4Q
zJyk|<e088XHBVHRaadP}zUZ0{n+i><BdtYbx0*(;PPHWYQ3l_zi2S{l-ix$FS$qmU
z>W1VmQQ@0gn&sLUl{NN~@eGV4S=1~4y_UPbon7`;&SL!^))r~uN=EOkH`5Gt6213J
zpe+6kA{&t+DH6Peci2<Tv8K@HE<dr(-r2hUvU0zmz)(fuhIVkA(=ie8VeZ|PtMBxE
z*oVTaeEI+5D=|xDe|E~Dn}#MDXXQIoK)9k~vVKXi7WQe(Q)7=Bqg0$xvv%?SqlPWs
zz4_emzg@HWhe|WN)<n_bSo>n>`NAiC(;~T|*hiH1VFjO3@RwEX4PxWR9JVS=dH+Wg
z{1pYFgWSQxnH*$~PTv=lW-M3)%WfkKF0(fpvTO4e4I@-ZUCUbwhh52bPr^0-ll1JC
zr@?`hUc6kH4|4mQxl`i*3jGdq%23~&%L~J+g~a;DG+e{A4X|Waq1mMF+o$eJC>WAt
z3L<!mYW7F23Y-t)-qrR2g?YYH>UO)iDZv78I_zoee~zJ3&g^ZL($RFB<u~-gKUH8{
zEb*yiY|@mD)S~Yl$Q7_{aNE|lqyH*pA7V$>{)iFvW#CDPI~@HEe03l#y3v^o$!Mq$
zKW<?neiTvxL46_te@Z3rXA4hDrqwOSzcL&u(4Ib|VJ#C!sQ>2fHwzeBJJX|4)K)Tl
zYCt)Oc=7AI-yEVBB1Pm-Sap>SmHM?pdMLj8e`fOg;E|m+B1L7?+K}}?UUk}2G!pmq
zFtysJHzs7vZ>}LWsHV2gZU=8B%e3hH7`^*dBHX}O89A>ZvQ&b~6-S{vO^Ws@|1{JH
z8Yc5u)I{U%Gv7hN>gHTs313AsQ>vAsx_K!MM1GkP`V9R%mU&<UcvWh|>$uNExCs?v
z)Nm?#Mi<7O_2zzm)%-UNT#D7oVnWxYtcfd2l)KS`5ev9J1O=R#AJ%s}_~=2-%YRme
z0`~P#K22laIuea6P6)~Zs|W8dJuGiak=ixD+q^E>HC*va&Li^P)xAw;rq0DudwHh%
z5{^86R2%jwg4NQ~75DA$mF3ctCaHV!sZuTS)aOH1<6kJxS2vs#FHDIwV)s2(hxukX
zL?-e=A%9$XZb-g}U7fx<KH|To_Za)iVC%nS5!;+SeY+k51qNzOjm`k!C8Zgo`p4A3
zUsLd%O8YJalG=%v40vdCfZOk&QrA@S#|b71ZZ^s&`}to~LRRJ?div`E9j&{@j^a;H
z;_vAw3dlMK(K=N6UK@H`(IJ$N?Z^%<QB3M!q`roeM~=)!2XS2JxEbs8K;2;}p#JMN
zvdukD#2Mqc?Jc%UZLFSHNBhND2|#4vh2RF?FOqp3Zn`pcND)5XsO^iL00<`9D?H&z
zK$<NhYOm*pLVMo9cQrtz$v1|IEFl-b<!}^bqJ-P%01KUQ4jW9x-B2YMj>k2nb|1DR
z2X<RlmGM)W%+FBUX^hCd+NIN;-Ki4tUh(hSWVSy|+V5+&b)`Bs{17HM@H#dy8r{`8
z&euMkrR<q=XU|%{&Ybc8I<E(3Tj8I4aBBo^UuMpnIm_>Of<k{oeH2G`Jbi{_^lkV@
zy2(dQQH7+wa}<x-H57{ne#U#woaJ@OQaxkpD@6%>sItjrLDr}tT4U$VojK>c1)n`P
zew*9ce^XufDS|Lt#yRUWR;^dm*WXlnK{1=xpVs5=Dfm96{i<SLQebz#ncW%_>(7<;
zRRuY{cAElOnL6a*UOoPn9%~Bb6_}FSW6F3^fs`2izop=)g1ZzL@*oDYLkSG$Fw{a<
zl>ZY7v}XR#DAqbUKdzqI`T5&qe2NmaA0a6h<Jw3!F0#h|zCX*|pFlv$#WcX@;=i8D
z!v@`gOYa_Bb!T&Zg@MeSDm(CI=1#7=o=gEwSQ5wHeYwG0s?e7>Yk7NId&RG3b~jos
zel2rPTl{ph#2}ZDoVkv*a%11Y_jPs;#eMa7n$C9ouqW-Oc9MS$r{FEEi;1LMeN^C6
zR#I@S`A`i`u+!#DT1HuU|0Sw*uxyhgr1;bS==+#B8jZ`3BPTN<#y6gfaQr77HuHIm
z?z&F5o$4f4#|~W`atnyt)^Ma|pjATZa+8L^A7W$;wvV?sf#7~(Hw#WvKs1j3-7F43
z^0G?U_WK3KkEh74_R1oPoOeZ>_{_|#)buGEQ=m1(IPyXk4l2{7i&ZL*iiS7aj<5}4
z+d^;pV;ywubD2F0E5JBqTl0o`*2grg;y0b)o^6@5b+~%T(?AnkF3@)-2yO;~>Rvc1
zfHhS9aq=VMILna<%K<qq3F2^Rl3OFIW+V?Pjn}r0(oSC$tf6*?fUew#ke@T5x2!F#
zeZ;}Ck!-y~UsL?hX76Zk!)o)s4Yg$$^OEDZ*5(X9*rd0cw?wTR!tPd2WJI$bwe<<t
zR%;Zk{%@#Ppx#e0TE6RLR;7(`ijQio%dI${;zSg^S3Adm?6)!dzP1`#w*OwKCW8LE
z+F%KnP*SRPoUeUoTnWl56{_9#$<Ha%;NG98wV&1FuPcAU*Rpr3g(hL{A5}@={b6Ey
zQ?t%L(RO=`O6DeTSK&X1yPn!!_ygVCDKx87BZdF0G-3mV|1R-vku90(Dg3#RYGT>K
zO%oHRGGFJzH;qHk*r<^eD3VdGfovCucSp<c9Rt0$gqH#of25vjPv`K8v@`2~2ag>;
zO^7r>aswg>Y<S7fI(;yp(Mts&lb}&-B2`JtM%WuNsFmka1t{oEWL+xm@q6>V-mo{|
z4Zb9&`#FPMk_3;0(bfCRvN0%{<HqsRQ)kB$e70lha>R^~%Y%NWyt%+9nHS`&p^3N)
z(rHkFdgdKp1hJfUaEMYUyQQU0Zmcyvv#zbmf#i@Hf-T4=KX$P)f`~{@jUCehTEE0*
z3AeQSUA11BZ_Z!V&iECANHh};w0azYsAiEi@8Fw2tS$84AQr+Hqhs(88aYwvo}DJ#
z>cH(ZcFn`1{Q03a_`f80&AF{LPp103SYtXJbUmW76Dn&+@`@h+jUEpuW`Tj!zsCr7
z(Lw+B$y|+m0Bq>irc-pL1EPw5k>gJtt?BFq=~{oZP?GhqT*!M;a5{0PRUK;PAi<S1
z$nujW4~pDvVPzm>6izYR;FS4>>ET+FbP%h7pa&IDs4-w@+>GO}6p#_$jW}<R=M_nV
zB+dL5<cS$m`7Q#vv;6gX+0SAdfLhTMrSkaTmT6x<x(t2NWiZ80K8!HK*9Vnm)KU%v
zEaBLr#u<H4yW@i9Q<q1WI``hk-it<JCv@Sns<zh*pL&rg@omRwh8*X{Al8yvF|7DY
zjV<RSgyOM|543SwxaGl*R7ibrUnJCO-;$XTrmrA8zKqXQKB?Mv!oob7a3+(iiB+M!
za|>r92efxfbU1|pNgwq6z?<?>FHaT{EDj)Y)Te+TS)lghfMvthk#y0lZrf1Y*T1YO
z5YpVCRdzcGy4g*f_#C@F=xq}xqK|Ll!~uyD5iUAY5iVJ6F56l;X*{>YiHM~TB{(*C
zQ=r&*b^^s=nDY`Sj__XsMd|2f_`j9^5YX#row@I`AYL33<s%vhK{PsFa>Y(bM1tt<
zYoq-<B4@~SuLlw2dBxTT^m%bvK+;+f5#$r34w?Y*RY-O9MB+mI?{A7A2cr~nyaB{2
zgIt}h7LHg8Lx><>+q?yD_$}+_5EpJixcHNeKGjF;`R6%2yah^d&>LY?5~J;}=b7hi
z*m^dEjoa(eyhb>Qu<<_GT=Fy)2^;^7J(>NdiL)~GZT0QzlreGRDB{MgjD%;ziwG7G
zH||E<xPvl#0)&r<7i04!2^)i*-WVkjJ+AMiR|pYrIC0}|P<FqGCb!iO1lzoAfPvk_
zc6-|`b`aalZEm%_9ni14c}A$mXe4fY4!v}~0uqTE51}Z91+9r2cbm9z&su4N_$6L@
zq*MGtZz8cGE$^Xai4~utmxqHAW7ro+XjdY=$J=Y-zW+(y5#Ce;{o8N*_P{GKX79OI
z<0xkBvzfYHM?Ce46t(jbMfxw*Z*$Msk8Dwi+DGkP+YjhE@Jjpv#Gn@!sJ-86|CMI#
z?en*rDD!hF?OJG+Z<ccJ_?1Wr?0|Rhr5u9c<{P!P&<(BkSkGF|%;F(talhTTZL^5x
z;Q{ZkiRuon9kO*h$c!9DY&hs0VLrxhDU|%vjyaZ4@+fbSV9<%o5Oc9NBrZFyKNFeV
zZX&ZIYvUW%EK%Fp(Ozs`f=a$y#E!TvxQ*T%XMN=8>V$!GgXUirOa-_)VVD15LDv*;
z(O(@^DBxIrs11JBZpM!_F%+r^Fvq9_k^#8XEr{}p3D_L7AnxBmnp0`(x__n8_AUWg
zRnIk*+BSgE7EiNHzpY_ztDD0Ah=OYhn$kD_B|W~ZK(aUgqY6Hzz%cJ`EMfCsQQ=n=
ze7`QAA5vh<jwTA-F%J|pOl_F|_6dZBB7C$T+2xTUd~+JOR*!0O9wK-x;Z$AhqHEzO
zQj+_qaFhkwU%l%V7tQ7~gdbV^va0`a4dN3DzE{EbDG-an|4Ry*LMHzQcpN|RO^A2G
z)cg`=boy+bV@WP+7#9`1tl$d@ZYB6>iF+I;<o}(R$M`NL>oH}cf23Y@3U|hDZ^JP*
zk>>cH(GX3L^Et(SNWl*)kSNAqS8zjtDFuE-FZjG3?UMiJ+A+tJ))KM&L*@R4f?rc0
ziHrZ273)(>(iZ<m^!Qg4*uMLMVt-Y^r^v-W;&i#bxa|L^QeIQ=F1_K$6qCfn|7(i<
zxB}Znzpa>|yJA4yujBS(ROrKSOLC4xeDqDZL6qzxhxZ;4ZtW71qr!FUw<Q_bjb?r0
z)3o1MG9npB2FXSte#kH<>hiuLGTJpciWo{QsYti10}nXSw=v!VEW8E<sc#>&F%o}M
z!`5j%+r~yLY)WH_3}|B<8pdkOx?7T*zetS{&7{2c<YM4}xeRWWGg};UY~<^{N=%d4
z@pB6TwE3O9lm{jTJIoQIB0_uhZM<tSFNaN<2vCuq_^sjFZmE1xf8LPS?+q9MS}+2%
z2k1T=Mh=(G%)z8zsaBd4=<3G{rN+-AleM{82^t(%*fTQ~9|X2E<0?$z!?d<Cecn9L
zG^N`m6|F4S&!0Bo@c-Z1xyHtERdL+j-PpTpd%b?dNu0#lG-;fz8aF8|psKV*g^1AB
zDkLrRA&TwY*iQV)>^d#0tr}F2KmsI$1X7h&ZPhQ4-~;_qBm_bNL?z%0;??B~NTn?W
z`UF}ajrsl0o!OavI89P)?cLcsGk5OH+<VXcpYy-xge4tWmN#wrj1LNsIYqN`iounR
zUuw}j5z2S>B;I7~^fR%_KErGiszXFWHeu7|`=%F`XO5R9_Ve%5>Hgs|a5kS3J!AiV
zU;8c#|71Nl7~s0p0Nm2-PVg~weToe{O=IG+SaA?V*v8EDu%g1HG=o6})&=hF_Hnnv
z*Z{TrpnsaC5m#kFj5s6ZLEOZ6hWF%$p>KU0yMvKC?YDOWGljH3*AuZwBpL!rJw-jd
z0!by0cOd5nD0qZ9Eyd4G<z`tdMr6IY;>__CM&dH@)}@7XgK-_?*ZP9<?9qF_qgB>Y
z@?MSu@x>BB$pnNHHcbM=tiCLnj^@c%WS{Eub&`PRt75$0tE@pQt9}s`>eaOO4@&B;
z#KK!@cIqgt=2PlPo7w2?VL?Ay&ES8OamOM_`}pc$b91h%3=YodEe3z8uqGlm^C^{i
zEO@F%9cM@hKX`>WwBCd>T!JQs0Oz7)LAVWa<x7m2>OCm#VtfyXWDS)UQm?mP;in+o
zc0Ga8hnXNo#}9R!*nv+eOrN~%lfY)SDEFkr9t4<M_?S_kc?wuBs;2hdBg17ig+W!v
zr}^%F)EP46aYw<jk|#b)scJa|S`OdnceolD2X_M^pMgy{lkb#K%;cK}{B~k%HC{T!
zo?h-t-dSQxs#{<9o0{Nwfa~{GL2^VR%v5_G)>ekCUfb$_6c1S#Rc%ip%lhv#96umG
zLC*Z^j<h}BV-Jo5f<4kUnT~yE(z98?^SlM`1tl+%ROI&p?*JMh=#IJZ%GPo7$*-2A
zOjhJG(9F-`cXc60w41#%b5lz=MK0wTWN<nqExP<+5t%Drm@il;5E@`1x2QPeD#5Ik
zMNFs~YTVT(4tjr64PH_5s*=}~ysm`y(NW3VHqQb~PQHa9;<{XA;Etu@11mTYz?+xL
zf(>mdi#K%6mz2Dz<QgRvCI3)jeMMcvdrLo`QzuDd8N@y8!aRPO-%+kjku)PgNpiU|
z&M8rk_lio^lvwFD8RSLKYnO`m`<^;q((br=dykS{CA*bpiHUmk*6*uZ-Nc3n2rg)k
zgs=kk#!d~#6$(Za3AD6p=~(Pk7QzFI=uXKSa(ZpI5f@B{yB37>Y3+XVOY?*vlXDP;
zp|XkJBVJwdHqfQ7pduP=Z{brCCue=SG`^Ku^&1-kqBIoXjq0!R;K!Dz@NC6m4{l)%
z^@i#QC|G*}5_Ry~={!cDS2`WXkciP-r<p{ZM1ye``65xH6OeZ%&Pn+i4E$%&cPkK}
zRW7Q2`-VClAyNB#3VmoWGNAde1ZgmO3;mSvNc$2BS-(U|ZgYN;UI%R$y0l*>=+pZB
z9VAIsWMpbOV0SfO=A37^9vY0*j?%LnF@)a0#tm|%K^pvG?5H`|<g7u?63P0i@-TXj
zQK-0q-eYT7>-$)vZF-MwRp}}=q4&VdQ+nBcjTDANvgYi;^5*iE$49H%Oz$y_-eV*8
z&pGHqI?TJ(cJv<GIA*N6m1n!1+#S+-K%2%N7FvtZ$8%Owh|NZ`evV^znvP_&d|7q0
zunBF)4z_kcpVro{DjF2`#-UD=!|x#@Lrd?m2;$uemfm8B9lZ%!gpK&9*mfF^0Y2e+
zyJ<jj5AJM1)19TmjcA%D7-%|%3S&GmX^Ix9@8#8<+|91)F0L)jNFTIjU7zRw_#Un#
z!ZTMFO<8&jR|7N!1BIQ`F28>5pgZp=TxOaD$jJ6&#<^0yn&(QpxX$iY%|dnPaEH;w
zpKho<x%taO)NhQ}G5Vr|{)T8pb~XI(h<X=;sUx!n=Q+1<Q79YiYjv?|v_sr&0BnBz
z0&wajpwV#p^1Pmhjr8vYp{~n86gY+0U;<!0GGPRM-XWoKTFE5>(N}ak`}E4|XjF7v
zW`CD}LX)6Svzi+orHooPIY$ZR+l)z@B$I$uO_J#=)Wi|>#_*jZK9FGM`QYbp>@py$
zBj=qDf+`q2s(HwAX@Oh=umYh8gP(4%fzGq(F3eJIC(?ueN~8mS%dJu_RIaf6?YDXZ
zhANtvD|N9q*}z=sZHAoBsDK-l7<a(hm&A1hfu0|H{)TfO(-X{bmhnTnWARuF2G!ee
zLDsUHx_n+otaJDsPe-p0L~o~v&=Tr^Xrb5aow9`zqP^&BpnaByK_F@awLGSn3fCU7
ze`W&v=OG6jL-hGe1APKDMLcpdLGxCZgfUkjci}WOU`G2?yS?TC)d3XC$_s(YmpAd9
z9$;d|t~v$3<YGV-CuS#@36_qlSpU>1CWmGG-seuvPvxx9?h`e`IWN6*?3!cLG9q}d
zu7{UgL&e^Udf?1&>VF&9O7Jm^DP>e({zDeTeTUGA2%-d4%6&Y1l}*xSht}J8XuYjv
zqt_I;BNCa7B!<miQu7MY*SEQ7WOAr>_r7i#5RIr~6dDmM65l4<{Ko14Yw|n>&oo_{
zI(K;8;1p01^h_|b?txrbclsB`YBX?S&SW_?=q9JTl$TQTaw^<)Gw4JCgqc^#;CT!`
z2l8Eju@egv=1qB3U|9w7FEEbw$#W^?Sx5?Kp!mb85}4vxhA<fSa(4f*EOVYdJOU91
zUglIqD#N83=irAxb0b`*Az29|LGTE+S>m;a^ERTrr_pmPMpxb!Ce%Xuv|`6^@IDXH
z>vu-eqeO_)FEf|+Ojh1s4>}$R<yopnEi5_5Q?y`dU&3~O#A`II{^i`-+L}{gYx-;+
zsJuV&Zu29K`PpP)HSl2#z(y4{!nU^7Teag8O01!Tl*1llo~|q}V9s>mDUAPSJW`QM
z&SQ~vYiLa4*+$TKf1=jVOpbd$&_-f<X6}CP7iy(VZBxp(>(w@O!x<d>$?BSkYY-Pa
zk$7<1Spz7FEUX|{fd#gu{njn84XS!|$9k8%fgfc&1!1&^snMd~iJ70iIPn-!8qiqO
zU;q$=yu%3^1W}8^$n-YGWG~LdDhLj!z8_k~aZ+qA=+DZvF@3!^q9sUr9buP!RzI7b
zjf%Ka&t{OFBiuhiY>idrENtys3-zQe&mbbdV%`g?jMekkY@KP*h?;mZ$z*8PsFW9=
z5!F3j8Pv@eN)QvfQfNiDjC6Y1zrGk!gN%xyLXU|SIbsM!kP}N-rdvdiG^33osIL|f
z<4s?AXmM&bq69tQ#84nT<Xn4kd1-lObz#caaOR4oGQ+u?d^<g5XDzwNI;9fk{PhQS
z*=qd3J+3vrti$&Qc2r9Yj3@`S=8e|G7K0cqCq8LPIh%zC(WmX{%ZgO%F?Fls1`-KS
z_aZGF@&CdRzeWh!Fdva80-+58VQXq?NV>pGAOLWZLl>&-@>fQ;d0y)YfdPt1@=>cU
zUH^X(0!}0txXhba=#Gem!+6Uyp->kAA8jK76fp?~460Rzal*>=h@`;!f<D9o(SrQp
zALY97{<NQ%L%73EJt-zw`m1qr5EaoJlnc05#6b-L#OUDt`Y!ZtB0_zk4=T9C#-J}Y
zuHyV5F?9g{QL)jF*ud!SJ%XUfHi(U^iH+=J<=Tr88-XT%)|VP4A!Plmbx+UFtnxVs
z!}K5--X7)l5$=9gam*EIvgCb=_r(O}F1~8remqO#s?VN1^ZKuDo4)-<XX`YEqRE`$
zcab?}Vl%<BM4`?>&KnwSY#^x%8SiD~N&3_hQ-o?%3b#F!F#{1u)X4g|02$-`Ri&Gd
zN%@_JO9Y!-XFZ`ar~YsqwF-rfRMF1(<xFB|#}Ge44qa!*y5(?UTcZ#PLfg0NL+u3-
z7d3{T%?+h34d#ZS5TYqFcv##+W>X4?f~J~4=0vDPFwtlkC26#YV=Q%v=Fn(~eSnUc
zMK9K9SzNqX4VJYFh%RZrB)XEmXC$?UBi>tjdOxHd_VQ~R>=6cBXA@!eM&GQz)_-g3
zKO`K~PK3Rq{r^>RR!P%9q!PcZ2cO}RgK(*Ai3pFC<qLslu67SIBGEd%*Ho~CnfHp4
z5O3PW#&t8Vvg0$sbtIiKi7r-zb>)yu+PA)B`ZT9UB~vP%iWeLh@&S@h7R64d(9DI5
zNv?n_nsm9A*Hh><DU~uiU7Ba#UmI(%=eI!5%DpCia{Se?j-(Z&wS9ODB$gnuAh%Zl
zN9&lDQJE7nf&)s+i^W`N1p$T<-N+lcsnuD9Y%rPT0twp6rO6rLli*M7u|6@nlOKM2
z#L+Pmqt0qq6sO4kfIy4eY&>6v;%GJli@{#nq2Zgh+a(t%W+%ZiR<{lKVy`lG2n{lI
zoUcgtK4mv%oLL6}gBg?rmcopc=2nj$T`1yQ4pqO4WB60ObJM@u?AL6$KO!$^k0uG%
zbkL8LXjf#K)zlt!oo}g~xEYr7K&M%!t=~Ia89)E)std{sV6b{WQBBrkr9Md|94qyj
zce&)nvQs}omqW_FN3hmjPWy5C1mQl0OoVNUSuKMz)QqB>u_ed{huS+_*6P1h8M<KU
zkjZ%+BSV<XV<~e#Cm%L~c{6h3>T_TTJ^^DDV-|y;D=kB5W#lrHL-_Y)8k_~Fz{Hhd
zw$OuGoNyI@xdUY!q?%MtN-9$m3|-#W0#JwJJ=P08*pI2K8Rk8<KFs^z#mMVbBZJe-
z_Z=-qMfP!CNM-kQZYH;P<v?yO&)3z24t1`0l*P}Pd~S}sSS_}4fX~$q3I0qi$VBtd
z1&e0Q)jaDC(nNS-okkW?TW7K^V&d2Y_j|{Ej_Fbt%<GEaGwOLQB3R}q5T1IVmJFiM
zTT3E#wb(#G{e=Et<Z_hZKtI%w&lPDIP$$}p;RZZ-CC{W)3~L4-#exy^hsmCI=r5Sm
z>tx<#z%Rsnf7h#O8f>)p=elL7VFp<YjL^%!OK*<|zJwdTS!^fShyC9ce?Cp2wnEMe
zf!xLOTh3A9H&ivpZ(#`<OEfcU$sS|5>_lqn2=BYP*PGmx)K+Rqe1Xo)O)M?BWsE2E
zRKBSl11q1iT^K<LcP=N{G~<Q4H@o9)4ekV7!(a)jxi>^jT%2DbP_A3SuXp6)6Qasn
zn4k6}w>&*NmdWA)qP)DkQ1YmBhxbn<LMC?)*ac3`%V<2YTAt?!t$wPFx75e>B8!S8
zcqIy$J56BCf-Nc}I@)4!c1qC76FnR9d129x_j+|D6Fi#1m<t2PZWA{0Bt`rMU^)X+
z^y&iT!(H+8U~A*i-1=2z-wx}bUoHaDh)x;tPr)8OF|}G=_5(sW4+fic#LmV-eO66y
zZ9M}|eo4GhC0ljImPna(8qS1y3fQLoF8ag-I#+~^w_WG$((B)=E!lIuPwKG#TZ4Cd
zo872)Si7E4gZzTFM2RwWv?&8@-sPInZtV?|aTS)l-qrX9PVXpvmL$=UO!m^*C5~k+
z(6+M{^mguLKv166A2CC>^~7E0(Ee{oeqcD!@sRt^UK?!bd+_e;6&(*{d-(KcuO7bk
EfBmK?2mk;8

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..2d2e6823835a6cf1764c4e63dd9f620d6cf4ebf4
GIT binary patch
literal 54428
zcmb@v33O%GS?7D_NmsK<rJ8RIc9ra!Jv)x)l;gxMCw4hb6gxMP&Xr0k=}L9(RSn__
zOv)iPp%Ng_>j5#Syt?TaNWTZ9X}Y1AdP=WFH{HB`J)G{x>rV3shGqdTKtqwZ-tYhI
zbM7RS6QEV4t$p^{d!K#w@Xg<MmU?>&F@92?{Y>RoUx>&4yDr*)(p)^uZ|zs&v6vT|
zh<R}@u@Ik#+uy`uav~XbWvPj@m1QQf{7o+87V{H%yO&xhEcQ(FEEXq<cAZ}6UF@6a
zTkN0cx9iNpz~bP<VA#*l#1QSX3&V@U6T?=PTNqi~GO@+-`GwKNtrJ@<UsxDh+%~bz
z@;wXN7k5nTSll_Wb8*+iuEpIGyBGIN>{%>Llot0+>|NY9vCrxi7xph6m^fhh-i2Eh
zZ=JYx@!-TkyY5>!v^YL7zIfZjZFb$iaQouniNlLWCXOr~oj7V`0}IC%k53%8{NTcg
z#gh{!$q#wM-iWv5^T~-jyxYC4-q<Vgi95Y*-u72w6L&ok^LBVUUx|4;E7|$GD<`j{
zChnnZm$#d;-IaT;Bq#3WdXHD)x}-Xk+(*e?ZyzQ5D)(}IKiB)c16&`l>vwT|i+3y6
zw^rW8{Rg-{=pEwvkX^r<>v8Wku5YvJ2ff5|vGK!`%PeEz;@pB)TAaI7U-2uYrJ2In
zdVRTe|FL7UbM><;Q%9$l7LP5gOwW1#<Xp9O%&SzF=4x}**(2WE%*;w{ZmD`?`Eq^9
zpFVqRYGG;W*gf|gzkAA?nZA4Gu9Nr7+;y*a#~m~G-Z^#8)Lkb}R_;Dgxw~@5iM#Kd
zx}()ox!F@W>Z$pZjrWc&U*2qpOUsq&<lM2j#btl#Ld9#1&W);a-<@~fd*8`>@0gmp
z_nx~ar%&8@&pqB<p2ukKo0^*PPI~v<cl_jZrP)F2!N+DNS8BDn$!e?UCS!PPX>s{O
zbG2jj$wf_2-Jh)1W|sWLieD=nIa1>9^jzUB9jMde@gVoq<l>Y!`P8%Hi6HM+CVi^U
zR)UOQS*}dhDeRkGs@5x)>Wh=r$=Qk@^i)?C7Q!;xXO}1a`rPCK*;M_!x5(f8;^etX
zdDfrvxagz5bLG{Fzf@kTE?lOh@YGzb{@7A=W^Q&o7xXPG&6ewz>gCB=d2Vqw$W$(I
zyB1`<%FN`-LOn>&)fOk0gDlIv$N=cQ!X(TrOxA<q^wLVTUS<T7i?v`Y8>2p1Wia(h
zdB$H_EHg?yBpzg*nybHOaZ;~XWhN#U%I6k>osZ8=&Ms9a7oG@b_HiEU&rPk=*-ca(
zdXn|8&($xVn)7S*=jwWR?;|VqrAl>r$*cJ9e`-7#4B5L*RqB)F+Vt6qx3W+PhM6sY
za=KogtJasw^-8t2<Oc<9*h+a}?p%cd^*sLglTSZ4xwyPg@wps$Qm^*Z(kv7F^xUPn
zYLJ`O=%;zf9&-Lttu|S&SA+Ovo|y^n<Xh#*NpJDYbV7?7{wUA6iao|}Ek)9ZaoFOk
zQ_oze_!s@TdZo0;TPz%9sB_iQMUGjies-=_npvq%YX(a7rIKE*RMUyHTE{9ECKs$(
zp~R23RH>D!OZC$9S?!t9WcBjml3zLcW{lZO)hY`!K}yf?5Ab+Pu7B*<*`>wGF&b5-
zmX^*P^Cs()$7cPdmE~j4tyG_4zcPu{W18P%k3II>vz8oLWQ4~Syv1WSVorZ_=8t*a
z@?!aX%_frL5cGL-waKZ4idVJ)-=de-Or(<jA^uEjBBCGVg-UFOc9Lnt8u77MJvJY2
zB)oVdIT!y-e1x*Cnz_X3Amddo%uQE#xwJu=AjRJH##2GUKf&+GAYD5<xy*vnSL>bq
zgIorKE7dt2LJmRMb#pt#HnyE{{}}mcemj14hgY>#*0B-;y~QRNmzVI8UTQYxrM--o
z^>SX`D|kKRie9hR=cT-U{tfVVkl!@_`uVlo%<fOcX5vB&ujD6^`FOt1d!JAQh#^%;
zd++x~y{)gruf!)ZmF#-%TC9@4!c%4v-k7(|$_w6hZwKh3$J^=c;&0L0?d{=juUGQ+
z^0&|1=k4cjKL}xfc0vZJi9y$!_mm+6^*<DrJZ<PeCBtFKY45Oi#GW<6c#e9<7~2-_
zxOakHN4;kNkatkBm3w!3caa<O?)L8C-Zt-E?>_!+_wM)J#ory?1Kzv&3wqO=cn^9H
zy%OU&yS#_JM=0OzJ?cHi-#y{|4|u1%C+xW;@4eoW+}Zo7SSmKLkCp#mkb3;_4?O+W
z?<QHt%)(q1AQ>bUPM$usn&UhiTi|e>0Zi!#zoj$x78PS}c_d#w^4QMb|H|FlANuOI
zjr_eQ{@dygJoMGCzxl_%`SjGw4}CRS>oD_24+Y7}a&5X4mGgHmzd3%@UoD{LdhAMU
zJ${XcG`4#23$a@7`(j?=O8gV?#@0rxoTw)ovH8?|`bT1(atZ#XUQb<#Lk_UJ%b7+W
zX|~ZznrjqD^Nk+TLZe99)7VB@toJtJjnOL!@_qGwdg;?nyPDvcdCCST8Sp3SgSvAi
zPRUR`>1FG~^$|*j*0)?^p0C7ta_;q{p4=GoVk@dQ%KcHl(ikJZwL$IjSYw!UTVo69
z_C`+i{FO$Is~wG<q&pkC{g>;z8tE(X+RN{Y@s??;^Ey54ZloKTM%K$;h1jE>$_8qm
zQ!U0*(39K8*x2KLk@6lJk!pUy`qPN_*nGz3OO2sMvXN-?HwGGmUh!&%ep8HMZ(}>@
zzWV<81C1R#<rdCR@72UwiTpnQ-_>uOKS=Hnv($H0vv#POX~fpYz5Yh&nr3a+@^95|
zyQ2BOoqD(Xzh6H*e`NjWHI*FmfGzqRA2>ap^xw@w2Z>9^gZPEj<I^j?4+2`aToU30
z3@<M)foiL>rG0FaeSll9wC~dKFmqwwYVZA}XIH9K)c~Te4jftbE8d&~=7OAOz%*H)
z|I|{gvO4hoN@ZD9-}Ce%H3+U!5I+iJ>zSJcLsrV?Dwk_PK@Tn)(7j%Mwo<Fq?eUex
zNg;=2;RU-{x>%W;JzK9GEu9et;r~jlGPAN!x(F~Y`IQ>rQjZc+X4oY<IAZSvu%AA+
zyab*+dj0Xz?X92E3y(Z|`pMHz@J8=>?mcJDkS(2l{^>G<Ej|B%#~(TKp65zuo+(Yd
z=h<gEnvG|Ip4qtzm1<cixfb+xV&;5(ZjsS2eL>M3<T50KUk?i5Ri)~IdY7OqE9Dx~
zQJ9%qSSWMKS1Mq%xmDrrUcYjFWsbLk#F_Mf2urgUCjIf_fd3Vse~>GK_~+{7axh?&
zNu^%#=T;|$Q-ky}6HyQH<?{5xWUWRWYF8&0E2<4z_cWf_iktJGcY0xN8FXQ!gBSFO
zRI0Fmm&^5_e@0YESe?}h3I>468YHg(My^&CYQY#QQC{YS^mynVqg_~wS}^c<ghgNQ
zCzmzwv@NDTq)nJrJJ!h`(N*5==sHL)RH{L00a7W*POZ$$Kvf34svMC&elXC?h7?%P
zb3tR!JU|JJuo-6OY%mK`OJ%org8l`DE_x^`$Sy81i`4Hi5{n6)nXCJ~8cv_huOso!
zE!TrV7Nb%xS63FxE6b1+m70oX8Dqt7jw{FuP?zhMmn$qjFuYoJE9c*)K2z-1g`ih`
zv^GhiS`G3a5P7LZ^q*11{$|ha^@EhoSWsM<hK?02#q0)&>M|rWCv<5s=y}%uLO|A8
z8C6^MtFz;KeNiVt&q|e{J4)4mFD3pHN~*e-=WvL;t)KND)it#8q-SJwS*`pz)hsLd
zh>{5<A64>UB`+v>Qpx+2ykE&FC9_Iql$<4T`_A?UBiF(ApVhsJk{6XMsg^b$+uXLc
zKd<5oDu!jy)N5yKX{`d6(Ms5)y&3n<sEHAR2bKFE$q9aILX(BP1My6JJHPG!it#?X
zF2;Aq3kkbpY0lDtcrHE=PsI!TE%G-<j(@s$Ba_S|hG;tyFD8nVQ18ZSuaNlB=xk1B
zd2aO6M@Axt$M~)NBJ^S%s_#nN&o<&~*=oKHJqTt<3FCP2tLe3@muO_iVr#j2Vm=Ap
zf*u1O<m-v`6nF9-RMRID+hW%eYlTL3KK){REm_Yrz=lBltl^X#eF^*K=hI$t8<=1|
zdoEFX7;K&qHm?`zJ@Z9=d%Y|@gQM5`yd1bXdnIl-Mc2aJ`L^o<*WkV8{hme-BkFJD
zukkL;(xNNXILBfgcY5v4z$SP4T5Wc8w&`!b;|;WW^h75Nj>Q_;5h#>KVJv2UdFBqi
z?V!rH#@3)ac#}a!UThoVkQ)i3A8Hi6;cJPGagJyn?9B?-pwHcScR+PC-*HQ$$34?(
z<u<D|KkSX3OMGwKhYFe>@wN`fJSeF7EztWtjocOWt2Y?cxOc?XdaL(wHSCR9n|<Ck
z^<3{+->TWQ{<lY0dPC~(wfZ}({zKe@>iOo+xfZ8os=m$JX*6B0H^AQL^^y;0&icGv
z<e|7l@z5hw$^4FM@%5e8V!X$TqGybfD2bBjI`<E-vwa9iIKUt^35WgG4SrT{pIlfz
zJ6S88oxA{xw^W*{lwi9-eqIS?rBs=mhA9kDCN0yRLC>_qt$nU>d3r_C4lfw)yaviu
zE(F81^FF-2J*HW6mlZC7W06p-)Vgo)6rQ{3-EG#`uS|m)yZYVHy0v(s>*iZ}`(UVv
zhCu2*(v;~+u)9^J2S?>#=sD1$H@pAQ?=N2a{J(ftFw&}MGKlG=%FKyPOHUdl_m((M
z|3Q+k^2^VEZBqRY`wwxwI<UB640K4_RXi}$g@vgQ1=db6=LIlxO_;1&AKzu@&)-H<
z{~}3{jpjZh3|{s8qbf~8V)*A(skI9P?i`}agaV1|4A`W8$4S4gHYrUV$bD(nFhh`C
zWS1<iEIM59U26SVC7)8|%o!)s$g@Z_6KMp)r<dwa3Vkm!(2DmS-v@Iex~Wc9{lBF)
z$;pL<@jk<O4hub@?hN6zMrk-o4+}DGY3+Ibt18J!mLdV48;CmXVzT{3%0A6+ZHh!_
zqL4^{4swY;P(dzHFmy5&A4u>Iq%fM?4$|2UdKpP(;)xqM(8NG;cOpk^wHhNWCUZhe
zH;Rb^pqV21+>KnaNbBwK)low!O?n_Adp5{ngj=RHuAd|MPMzhLzlSuw4!<y;Tu)sC
zs=_I#1iUbxv@1^YO5EQA<`6d(e;qzA9$Ve(#Xbc8_k!!?-PR`C(Ti*MP}nZv+C_YZ
z{!$x?;ifGoI$K_h>EC$r^y;?zOV1%El9)@tLbP~D_^<AOP&YDPw?ZN)x;dV6G|I<$
zd1z$Tu;PDA`9hOCM%*{Ov<z9h<Ugl7Sx5=h8&5bQiMu9)>{{?3FG8b6)5<@gS?v%>
z;$`9fKInnIc!E?{x%f~#^^H^_^S1H>@x(V0iPb(^yk-X-qM<Exl;hV3f=p5b0YV|u
zFMy6yUb3EOq}G#SbuYyH;RYq!Vr!{Z=~}vxemwTl=vt<c;P2pCwvky+HL};@{_#5E
znDtB}Bc#Tnry7}e$IkCLFV->(Ab)o~Cx-i#SUtZEa$4^ZqqtsVi_nu;Ysik?jrY7P
zgh=jclBfO7ix4A_9{EP<LfkK({}9i}Zv%lqNEAqmbx4UTaZ->jWXE}9n(fNVpG*0l
z1nEM=h=c)=hB@ke(pY8)2qCrF1C7FZ|20sZ>wT@y_2(7NC1AA&8oeN_o^6bn5f#_^
z5j{XGsAM!Iq}J$vAa?%8s-F&9<<F&g_j|kA?Oy8-OA296BWnZgJ;AE?c-$L}&1kgr
z+CcC`&;RuNFVth}5Ei-%S{>}_r;mOHTdmWK$t&E0q=mM>OG^<RuH{fy%ihkGJawos
z@Bk?Ie7Z3N>BqlMCN!G0A>M6BLXvQ%g{JRp46P4)MMkvA-P$d@!O$+vsm&^5`A{Rt
zTtC6y7-5$o;Jccr>bP%dU9;N5Ya{HceghN(JZ;znFoWbqEI0Hzgd;?Q-SK+Pr5MS1
zS`0Uae<bDrgsU6g`%dnSG)AI(5%3t_aym#(FR#2A4~od-;hzvti>)odwIXyEBXqs=
zcx9ntaspV1(zmTSuD?~`=(I3#^uyUg&K3hIael>X&a&HI%J5F{wL~NSeTnnvh+4-U
z>38Dj@gOCZIf$2AH=LBU@PHUobVbV#{+2ov0v?I29<-z0$uu5)K%-x%JxC^Mv%2rp
zoD*>@p^OS)n?PN)axuDm6VbL;sZC1;T%TO<KgT2d&nVGr`LB?iVtz7GB`_O(O|WRf
zkLwvHX}{i!t_yVzLJ^5^t`C%8=q@c+roo*h$zE3bpAPT8pR4-|rPblW^Q|gj;rM9C
zyoxdOh4q4zTHAv7^_rym4x(lxe4ea3&TUc^^QZv)`qC0Yf;8d+bPfZWRnMtY#P7n`
zyONO4_XQFiiIJXO0QU!JSln48BccO>w3IxxAiq+ronNU`5JU7EzuH=Q1N-A!986C&
z`C4CKHb~Sw|5ucWp9>Om=Ym`{RP6XatJ1U~iy+U+lugWzDyMR>J=<2nW#e-}d>+M1
z8Sw;c^McpR(-I2DO>j_XO?}sX^3U<)-{iOU36egD2JxCX{uPWjO}w1=MlPi*BOh)g
zk}j8k?CAS?E|ai3aG~*?i2`JWS`WoXAvOBIN<;Bml}ipmO7w9h!eoRNBaj}sMB?S`
z3Ao!E2OwVb4-9r=b$FAFJgp-X{bb1KIB7;}kCMR23UzUm<`Y+v4fq;(QK$(@lW=EY
zIjBqZQqAkM<9BD0(tJ$rW^0?qr^L?_l{$(d7zIvEq%A9*KXj>FFEvV+%E!l#Hmlxm
z>p&q}z^~tL%|3d%bgA?hzekTAeW~PVQHdW-#<ftXAwH}wuOLh8?j>xGfM%-Vt3^`_
zR5FQP>5wrKr4!>vOYg1tmBXdu=sMATB0fclh0w39OTU7ixq5%;ky3pHqLsGPJY0%s
z|HGwrPGJ0S3F1|c4BzBntz=uU)uAs2Ydr05=a2uI5>eugdKE=Fo(Tz=u!kUCM>obZ
z15^khA5v_8X4g&(`p=h*=ZI=!nGmb#JdR{XplAy5)e+kpUDZ0M8f}b7;q^7qN*gza
zl}HHgwo7s(HU?;ZGk*HblzX@Dp_zNPtK`O0?zNJvzAuOyq-`C$M?V%yqTNPt%2}(p
zi`V#?USo&@k+_k#vAWA%<7V}5`ab@D*T8<8BpPXJ6~5b*O5EoEjuMU37_jbfZr|j&
zt7RZ``NIBa!iQ4fNcY@qq+yGjj@9VYuDNHXyXKywD#fF{lmB}f@$V}+&`b`J3z{>F
zOxw+C`1IQuum2WRx@NSyUdJfKB}O}>CMjSzN_Ip9KEhE(B?%MbMBpPVT}xXn0TF8l
z6>P#YffIWlj)R?)@8c0}H+{cy#$kRhx$&&+EC2siX>-YfzD@?bRlj>ydd|1x%_p>Q
zlJG@aY#?RlWgt8+%|1<F+A$q@S~aI*h6@OBB!i6vl07s-^DqPuyV84cU^p~L|E?z2
z(ng?RZp6JEjTEdv(%;`muOlV&(oRMNq)n>@dbQ+>$+gT{b}c7uRwHK+H+$OT2X7|a
zPW`H8;Xx8*`1zw>HgNN1vTJ^TKOf_^W7&U~9F_tLsK(3wA969CcV{(z$@MDna3t7Y
zT0xqBsf<uX#GDiM3*6WY2cyj?lEl0zAnBg;{^d>(b-yNlh^G_*Pec5s_*aaN!iR^z
zs;j}Y<~sT*l4;Jp&@Y5xd_MLDY}kDK`{K_+On)rCme$-g=Rb*bR`Z#oG><Nn`pjwS
zN^-4$JP^J4IV7~#0BQ5EXR~oHB~Eu97R5_Q&i4krekKN!1bgJA8~KsgdiEOSje<Lh
z;~7cT{)Dl#X0S#2A>%eO@CUdL%WLp|Ty_6K$qterZ6X)rZp~0&QiVzAzVV{l@11=I
zMQ4OlHc%PFr~EIeE9>-ss_Bqhsg6a_9tn!orK*|SEKb(Wb+1!yx$F!kPHBN8o7kQh
zg~B+Dz$28BM>RTV3xR2!2%|05$uRGnfi>pYi#zrgX*hNhoPvZH;Mw?Erk<*&>qyP1
zo4tYr!}QWw<mF;i58B!x3rCI)vpN%JwP3C13tsxF&R37l2R&(pi=$92FrkmyRn|lP
z&iQ%*wz{{Gmb~PAsgdwcOGZBf%g{(ex#Uh8QO?R}&aM~C;6^le-IpYW^JR1SN%G@y
zr!UbGBBuH-{<xdBkuy5is~Y_q<Pd8@8&^F4%c^8@Hq^W^e_1tC(8aT^D^tS!5QXf9
zTU1qBA{cR`hiro?*kpOKR;sy2-O&A{jG%fu@BRn1?gIS1(He0kmzJN=QtqT0q$5k2
ziI*TMQ^>7$a@rCvOFfc`uMRnk6)wLud~1M49m}jk*F1RpxX?1Q$XaSGia=co$ex^m
zT;#yXhA)97nU7&5Vxl%J6qc2uiS+J4VQLb=k!dHbVq3MY7t6TK`7ynbHfIJoo<upB
z8>?7Np~7+ZI{<Ym9J5H_+WjQppQIVPaI{~^T}eRnY6}7cUWM?zl6;k8Z6&;eD`|j%
z&Cn-#dYTS8e$?$SzhBdM5Dvs&MWhg-GB%NIk%F{+s1&j9vZZ3v8TPWbJ4m}A_W-~c
zQ@aOec(t!3C@REO9T9OpT#|OC)YWDF`l#7WSbjHUA@SuNV<e3s3VR?%sxikB=c2wV
znouLP0LI2mIE`3zf71!Ivwnz&8A9i5124r#fcC9~Z4j0~u834@kt@;g45r!0Y>Zv2
zot{N|cwt_{eT&De7Ho;Fuwy_SyYehGwJQz%4vShHGqV`>XVM=;GM7Lum1?A%sMAXT
zD2{XIKs_nr7-Vm>FceM=S2k~DrpQaWmi4TM*~us30DX*oPWX55Oymym5(wqddU`8n
zSh2cirFyQqbg>G@0PiCGNa^6WH8ss^2RZyzHBC-V)%^d5oNX81S6LylwVPZBa*MJK
zq!0fMZbio?C4w}_z&<f0)?5&VN?p^K(~RkcY-wv_^iTlR+Of|U{yYVTkjjjFJ(bG6
z9YXl1e#az7IM<Gmn7)n2NMPVq@r_9@32X)+Aj9SmF`enT7=#|j%S$ogdmBk~r@QSs
z&Gq=JF|K6;l=L!O-%ZODJ+YUrsb(ILli++ZvK+$NC&sb>{bk{HH94Qf=B?+AIF>nx
zK)ILmpY?jHaj$PJ4;Da|O3VJ&6OEj(_8eL{gsYaPE*KnvC>W(7F=5VOb})U^Zh&?@
zVH?#$$Z3QvR9V>bpn8sabvR~x=;olpAJCTmZLPOhHMd@NDsANi5d0BtysVp;kw=g~
zLbY10Qrid#B66>F(q%1VbfHrw4SpapR0yq2N~=SrUAhd%-qbK5oU(mHcfxf}eswoF
zgjPv!ZV|bb<o7`mqnGPNGj0nFGp^^|6$+~BY0HlHI^Z;_muQimSi;9_7yEma)22sz
z6PZS2oQa0n9++y#>=7ag$Y5sk6-AVa__^%=sXB2M9Fj8nd(?A@1k)X8QjW>Y+$9+)
zVB+EINP?VQFxs}scYrvW<nN%pF=Im<5RQv!h$O-(a6OdRndnQ5A$}P`EYp`5K`7HC
zkK7`4D0&2;GW<TnZ%yI==?skU0ot(XA(0?(5VAlXrPk91p|rulrnBHB5ubA$hrf-{
zDHt{BNEaVfuec#YF<GIQrfyHvgrggBnu|MB(IHzSQBm{ySovhmUDRJkuyz5BEHXjv
zdOxaZNYULs6k@AIgTUdf4$44P1R4w}JV-_?41^y7rABfd&4j_@ei49HYCuA`m%GV2
zkQcHqzMkRDDd8Q{r+sa_@x;;7*N3J`XG<4L)1@O1mZtc_1sk$=v4TbM<+2lAG7po~
zccirj5+qO(+0+b5K{PkJvV=A{lyv^3x(?+NX`?>;-%xAAM_n_K7dSMz)8E!y2o8#j
zl4K7|84q<u%}N{AT{B~|;^?Jy%?zs#P(;Dkfb2z{EntZeiHK8!3lI(u@-Q>qi5BkN
zDzE`+Yx~EPbPcr!$rCsMNlIn;r?_9!p6g@x=Hjc4rEROw0o{A_R?$HJ3mJwOg9^Vx
zP#jJvg$hJ;yfl#RO3R3%3O!`Z-{!%m*5jzS*sxN@m||zdD0tASc-uxAiDG(ITy4U?
z3x3vWHnONO_MV4;M7IeAaV^1Q<bc`Q{CVt_-M`$`v}r(r_T0&Zr3F-4vcrtc=lunG
zC|na)fD*|oGB!c#tk>Hpbk)wWx%y5A@n!!IGZchCPT&sU_XtI2-b@8Kj233gQ`qv#
zf=`|eS!}a4twD}VWqE1(EXIU5lVCQd;#0;<Kn$pbUk0B8v@X!8V+1CZTP1<(KD&gS
zxc{&o4iuU-#fbl-s_-ObzNcD+rDaSju+~G)UNk`|_JkP1*T#DsyP(Yw<Q?5S=UobV
zn+?iby3MSYv6u9XM`-Pu_vuF0*6EF4u)WStQ=_KCBU!fgdJ47o7#j(x_S*v~v$DET
zK>aOyDoL$ViI9Ql=%xcgXfJAu>2TKmD+x=+S@1Xs;R$fpkPCLrR$IrM`Lp;ob=G_>
zx(;i;7B^}h(&Clet8CgU@z-LXierciOra!hCGi(x9@6hr><Un`nmZAzJ#rBuK>t3%
zFPlXo8X3)-;e5E^5*QV4D0>w{tnPZR#yiW**d`ysX>G1)X0-Vu`d%o(_y!5P0PeAu
z@@rBf!)QSHaMnI&>Ma0W%4{vQUVw$4;ZKl6j{~KN3I&WNl&2cy<8Dn2jTIhHu_?R$
z4No?XVtZ`$kil=T_Z*gxDCcak+sKJ#qni%be@>g{j1t>HA0oGLC?{G&@jpV1f2;Zv
zB&W{!kE=z}4uJntyp8{~3Y-kwHz4-`)&3H<CWcyXSY2{2xk;IO+IYhMuo`|;iA_R(
z*Ngs3?il?m6Az(r5IOsL=hsKlBeuUo(NkBmj=9k8)vsZ4Egd^2=F14rct{bJkw(->
zGm|X?2dCmnx)XethIzM=u9G;#ve+pC{)jEcbmu>4;5t8{x#Ct^#*Otla8HW*`Ze5h
zm<*U$76J))TTWfb_kxct2R@b%4B?E!zp2yLN2?LGb%?j>Q1A7zsZy9#`5~^xn@BlT
z^_MQPx|2N5*F5=>3Pq?k>)7*J&{vf|SFMcwQke!pF|=bg$FS}R87>RCQY{o>OC|fC
zH6sPOcBY+TXI95-RyL{pmcWh?v(uWy=%=5I?g;nQF43VpC=9qTK>`S%=q!@LEEtUQ
z2&x8Uw_PWVe94+s4h9=|Q6yv*0oKa`A`%b8>e(yUfx@Fe0@<BBigd7*VaL0`P{L7#
zP<D07uJRt5H#nek^`3eW=~ZDCUd4krK<b6060-_nV4k+#2dPw`j@fX^3Uqz|qYmsm
zO)gf1yBTa?N`UN3+|*D5OGicpzHOxZAEjj2qvW%L2Kq)=dht;LyuE9E8$jh}IzVMF
z{q)G{kY4+&*CJPG{~y)2tdG`t_w}tZenIk<a5HD5-%r1a1i%0_h6=AB*ArNN4p;|1
z?stGP$T;<$F(216Cv4k+l_mS~_pv}hZ&R;l+a~vcg(cI;pW*QLQNf9irc|{j+EQ$s
zjuYGgd`IcpfNxg7w;k}d?LxsghG96`4y$o!f@^6yGuS#BTitKlPZ-^d?c2GuCRiOG
z^H=mTP2w&{qL_3<>$^05lQ~YQ%9oXx0MTTHI>}Bt@FU6@SwEx}#$QR2=u3#>AS!Ok
zp%Gw@2@7l34xDfr^zr^S0wR7Iot$RUSBs!06hF#Jz=0+sH%$Q3$$R-Y7sq_fv%jFk
zW~Hi}!#=H9_zH!!dl_B^W=RkiW4m|;Be}lBPUr>)hb4yWOQf*J1{z~R*9SeZn<G7x
z;8^!}&1=VEYFhMbKu*Fv&OvM~?gV<y2uo!4d<sJ>{YCDKWpSdOna|GWB<hinG6Z>v
z(=4)Rqw##uYfb>NdH0dQn)QTZV!yyM%;?>0*ByhAHsUKtKjKX;z7UDF)KQQz+N~BO
zsnNEfY?4NtT|bU8<tFh~%dw%1#gJSx1HmOT_SJ1R1)J4>GjrNMsUe+EqIvf3Q=+MG
z_$P4&uKpC~qn81ZJ;MMcWHfkq=Jfb}|9R>Lg|hrg@KD9sq*yMWUzuERt|w-VZ`PB(
zNEWxj9LotxmY0j0opZYR6(zr`<bNpn10{b*5~T3xTyWZsgdQ&OuWdTFe@!Z|S7l;V
z{d<(y8F^T_XO+CD<RvAx&sUUdkOaM<zsz(6Z=80kpHJxNc2tj(JH~JA6_Q+xorcW#
zulKL-za=Ix-MEpsw;0d#VKVq&u9(Z@a=A=jK2xMDr*gDfnOwS%DfH*|*xhty3}=Mi
zTq=`gO>glvqG^6Rel)-Osr@<c)pF{@9W#+n=2Q7Je-rr(xs>-I1Esil;;xwz@O~vW
zk*<W!nRc(oD<a(n<4@%CefdG}1uJz9o&|I5l(VMXI_0eSVe|fv<F0AG|F~<Kzd!Dp
zd*tUoG2p!WCkCBYKQRwJ;@#>U#LY5v&$LlZ*fS1!<FA0^CPqRB%8z=tdAFMb<reRw
z&WUqkG`uG_P7ncZo6eDQ%-lHTwE1G_#<@Lo<J=Ltaqje9GDl7vHhE^k`<VAYWj79`
z@Ae*iB^^0!YBby{n@gwWfH(zm_N4S7&7${lbMn-kz0Etc;LfD?xc43#?Y{6%!up!>
zp77pl<@>|2Kk2>C#(p5YehJspr>w>;&6ZlTv^VX2!26)xxz&5tdye@$7`8d%J#T#+
zGMCbEkf9g!KLyOG)|)<9s|7hZm&_~dw0|c%1TtDuq-hBwq918fMc#<4seL;;2xo;@
zS(*&wOXjkPs|6+I{E4#wq{_K?ZHw7$%Dqx%#~FW@mqcuuf^12663reGK&>6}M9B2w
z!UDq&S=48Fb|3kE%jYeRWhZqCv#K*#PhSxmU-VF#&cz$Ol3qb>AO^<K1u_OVQ-llU
z{A#gYKnjLa%Y3ioL<kCSFMT0a%{u<flLKnsM^Lras^(_eUquYCKH&A%2O&Y^WC2VO
zNrU8GPl&BU&{z(kad`{QF*ul-lj~N?k2U(teEKJ;xebfaLACef{yNs^_bHv<Ub|*?
zuRFYcvyH^zbbgn)=<IF``2V=R!y8b!)%-b|L7YyVvuH&9j}M-9Le%%N9{ye>`$&Qj
zyo_qNXaUW=$nmc1H+27Hl2iV7D*LFCZ&$)XqO+}Gj!57F0tYPluu!`39+L4=uoaQk
zbOrx5pkK335SK|`gP{(eHLDPHHgy^9&kK^vOUwQT)Zqv9q#pXfaMPR&1EHwM>(E#d
zxzAhBho}@%!O{Awj5El29xVu~fXVjR@}eR%<TU^f9Z1dp6jjWjphd@Y#bs!DEe50Q
z0UKJw2!>jD3G*bPG>vM|H#N0{rP0#4N>y@`U|442Se79UYdQm(&7R}Op3@tTDEXL@
zk1JUq86R>w$)R=^%J8KmNQ_?G75T5}oj#-Fs*>+f@&ih2?d*-eTbHjYX|a6G<!)PT
zLsthh4#QbmjgbUL`9owfy<ioX)m%c3Z?cqd|3)d*zXZaWjEQM5XGc=)ycxKI##b%n
z3MYJ*!dJn$!_mt(R>!&kP0It0;WmF5L@dj%(OA!tF@tybBNJICbYLlwW*^W}#@uHe
z6QA{r^D=kGc|Lm$zXj1_BC|O*DILf8>?acs7`r^4a)$-Z@o665Kc(c8N;DAEWFk0B
ziS9^Zk-zfAWjW?NtV@xN$vGEC-P-ab(lsp$B+o6hLTT=8P>H$79Bw(o$Q-9N*MgtB
z(T|VBSNC<#jd_}OG?^CV8U1L6)Lcw&?bk?nKm0670cikI`BX4h<UT6ZIDcQFBv~ty
zBKF66VHX~L_yOYKhv3S<P6m&D{0ylfK#S56eG?ob2IA7K^)yoCBql4$%3-~i*#)?M
zT=(RcpFls=e+@u+4X-nSIQo?iE)|(>N6(2sR`=g=Y>hPkhHF+Ih7vxDb)|xOU^H8r
zUcn7xQ6AjDrpQ<sh;pP=g0ck|Fj=i2dSAG_x?}(T{iWxhlfU+JXC6O-_x59tJf$nj
z4LYpuirmN*AmnH}(37S0_@J-DgzK~Oes^+MCpP_C+;9-aPPPHGCYR+md8=T^1!%!v
zxalDsj7h%GR?u{WR#cBVes0bU3a2RItl;6C<SipBw5Gs{@FK)3kx?@;4as1|I|YO8
zaem9>EQqiApU|W>k2+4zof02|ThD1fN?Cgdo5CCdD_KhE>|+W{e&kPA4|MO)&APcM
zchKSFs?D3@*EX-X2p3F(ttYvV+2>tzqIPXvVQo<nNgp>u8BMihURZk=)<bDwOCYZ+
zJB`7X+F2qL;C$b4g#6caQbUWrQ>y%g62qBKD)&Am?^n_}X~vOwnj=YpMw4>oMGACd
zb#M2qyt50Ij<DWZ!0l@yts*H|BO&QCZ|VYPB+9(?d_uZ8)FSCMXsSlCoH>tI2R5*Y
z)2Da|iF1RL;@|{>?c2#Lu`I@ne^A@>3)J&Zle`(KpcJ$Rtm#nEZ|bgXj^+;Wzn^A{
zyzJLR@JZz5G*G<m!S2^>$D3_~xVdQX$LW$+i`5<!Hw;`FtYa7xk~i?o<)mJbd$L(P
zz&%ZbizAYDQ9O*avRbm4lZcVY%i826P2!cro0;t*4NW+Ccs!ST&fgvO96y(=J#J)_
zo|3G+&xo&PU(pzVN5PPhPi-$&F0bD5yg5r)A;#RqNWdn$wp7A-Rh!Kf9reG9NnFjJ
zi9FC&cc=mpCpu`cF!0eS)HjCF4Vp-o<_ixyPjxZOw%Gnp>c$~mb}p>{McwFH#LrPy
z*N)vs17U0DFd(!3T(Y&KMT8?-;?-N<X=&Xl(!2VP(r*VU(n)S2?6X|V%KwZd6ymv>
zU<n358!0ETZlu?BmO&-b&0aOS`FyU<eHJoP&wzl0W>PF9o@|Ibv*fZMDCP1P++`3=
zl5b?;T{04(`~#kdN?<4?Oqc4CCZK_ZD*KqGrL5%RB;y6Mn)4;S3WhukOg(7a7GQ3H
zZAW98)Bz{K$?0=#vgOc#Zc4rC(ZR0h!9>TfU|iAjP32(%t01xJ`4S}go|4Xm3bvaJ
zxlMlPC`iejU(@<YXehH8Y`u{L$ahr#Oa^^h_nO_jtBw;X5R~FqV5s&uiI%Gl+lCK1
zo}gg%M0ME8(5F%gt3ycv3?6r94*QmbPNP}IWMzZJa^RP_V*>(`uYxm0THSK6X#-;6
zHJpDgf!7L(pT|Vw0`2Q<ov6l$uytL&K|)A344Mq*Lbi$2^8bz|^T#*NsJ~9x4{GWJ
zzuaUIfgw>yt{&>1u$$rHj_IPSCe*I2av=a_JnXm={fXoC_gTmzgwTX**hdWve8nLm
z9z@w(&rqYp<N~Nfl*%3qb0xd4E7y<uy^^S%G=L=Gy@ZIA7?`tG7W~UiGFH^0{evuS
z^H2)Dfegr097M!8`Ts~W^Ycn<PH$!0L6-Mb-H2L^Pj~M|qfSg52`i&xABE#skM?50
zRI`F0sZ0Swt|44mQ}NBGcC(H)sA<!T%yyz+o@^RHb}-qR9wLlkleG@00GP-+3j+_k
zkG5o;5?2aH32O@QV9^c{f$rbM!oU`Qy@`PbGZDJSy`tPLTthe5DbHZ8WZq*J<|-F`
zoeZ1{2{{<L%EZqG@r!;*t&Ge{5}L4k-f~3qIjcZ}U(`GaCYGS=O{Bbm?}@5*%!HaW
zXX0fp^rpg*@{LkD3$Wy!DRDEw29k|4Gto#A=fS->KvVV)SFvo9{e)(N*-VW=>$jC>
zRL(DR)5zFN5!-_rO4ZJR*WBtq?)H2-dw=Py1z767aq)h<3BCH+R(a&t<f0q`j)}|$
zNYfz~b!}RU($ta<0eiUFs&tVkhi7p}thOq@8}F&fdRWzb6bT4mw@chQdY5jx-lpa7
z?>9SZvntbyOKWp0iyq|8RVvHam(|>su#J#uNeTR4)tUL{N@T+5{~AeTH~mjk@J~qs
zbj25?PE@oh&WT?oD9=CBjiTPNyil2`6L%))w<}jmatZvzJ%TEk$Jzk~<+jm5a?(n|
zy_DUQd?^B7rL2?~W9%i)*v@Y7)!w8-+o;_n{?N|VIvIA0p=3BaY-c!I&OTC_M&qV&
zO~$GQ9*!NPT}9N9BqxuUFkEw<M!1i;Dx9CVK3P_7gQ~P%xl6BE>(%Tgy>!?sdr20R
zjjVdz&=1I5u~+m|PAJq?zXh3M@uR>?VZr33u)tO{DAeaF<td+lkjPz{O;i{La0&})
z)+&Tx{&g*$o%fVl1w*yv1!6gw7BF(~w6zxiFlu1qvL~Lp93-B)WSX-e@16%Y8uUCB
zK3uY}u7&Ltl4(}`HLdJUo*>zZ;ym1nL#CuAqR>)rkKx40O7H4k>5e{Z9y-=r;#jS@
zt@lWh^~Rgz#gy~s!NN%bR@NF=0xPjCwjKP9W=f9vK+&+AzhODNzL_dtz#v5?%NIaT
z)Dshe(msXTKGBANEIE{!MG0EhdaWc)3`me3$PwOOI%ov3?mps989O`8ww@^6F*UQ-
z((5_?B35;d`B&oWc*p?RbBJ0~IJozR*IjK#Vl&!{{ofm}_StAt{$GZp{mXA*v__mV
z3ifq4DsqVR-8<@3=ni<t<+!8w9_)V0u%6yH8`b2U2W?K;-`PDU?Co24ch$!ag;Te2
z4mR`^%|-P!n*q7hf5NSxa({){NN<m3gIvdK=#HBWcjsoap?cA5Sj|mmL$z$*HI)12
zJ3+Pbs=LW-Z0M^w8;N=k^eM^%f-CgTU|>I<U;ShgJ47DA_k%oUMWrfS9rh))!=>rN
zrL%`i7sp$5o5;*66YO@`0u{O@c<Y9(fJlV8GuM)j9)IaD?(?nd6EC&;YtbTvZ0%{Q
zg-m;H!38oPnz4t8En$e|-0W3?9$e#aM|O@YpX01Mn^l80@Q-GMYwCU@oP8<;ndhrB
zayCz$BK7~H;Ge1dO+fWYYB*f?2DwlY?My!ZlWP7Kk|41%6C^x-r)or!oA#Rs(EoR;
z^M9#MkDd_%&>*uCK2nALzt{cVhy-viR&1+&vKv<R8$quATtB}lX!$tf>r0O%EnW_M
zP$_W$)%^iQ&*5q;kwUefdHZPcaAIftHVFOAN>j*eR`+ZMRc$!#Y11+vihd+t5efyE
z#%f|IT2d;|9lWS%bp|mU#kJ2F+%)naR*)lrNjf89osdHdh8YcFRI@jh=?umI*7k?V
z=W9RAp_M-D4z3Dt$}+j>IhXX$P_CFRkARLrsB6XQ79DR&-U~ZctoAj>gXv52xy35s
z)D+BYzKG~I<*yJ#wOH?itEgD6A3o$9BN~vo4k5l2_KbYg-pUe<B5^<5`+X<6yXgH0
z);d4v{}3&P#JOO;(??Vk0;lSY7~ud1jSykzyHbt*xg@<0qsgPajc3sRVrNT5Pw~>9
z#|Q{i&^wC`56_rirVe4%7{^d!aD5BMen^Bb+OM4dFI<U?l(^3L|C;MjXLi&W<ao;|
z5?#<F)~q;>62ubflIV_jwUwo=36cc&CTK>FsW!`SNtFfYd_|qWt+af}{kaU8_($5?
ze@rsI)d?0&w%Ziz`ah=o+7kXXC8pCc6^tofOhNMzc8!zi7z<_xb4q2utQIel3%1B$
za)n?eoOmc#0!Ylw5N>=676xs6m&HHD(%@%lr;y}S^-P8_KHRdcF%#TiduUP9Ce$c9
zpn(mX6H`n(X|v*TN}Yj1t?35XjD@K?{#W!k91InlpzOrUEry%Gu+#o8=|!3|6{M!s
zSyKMHsY|%;c1ut-g^FKPJHxy8VazL?l4Q(7XpQ9O-~^2soo8em3RN*e&NNCG^T;*@
z@4+OVgtRg4(^^YAC4Cb#)3(h{w^4T;@+whx`W314wU3kFthpb~Ja#Sg;;Q3V#&wcw
zp?!1Ig=dFdn#Yg_mW)uw$e<KU2dm#JiS@jAdQ)~32q~z*7FHXT2|E>eZT9qAdZ6P+
zjNOPkaDA_)@BwbUSv1z9$&Jp!;`rZC8|j{WNd$vjB-aXt1SBPEQs8nHwC-Ov2~|*F
z=|bNFlaxrKpn9R?*Muclcw@9LE8c0Ts(wcOAJCI55hr3>?Oj@`KCZBn0=z1YRUf&H
zimHu*w42Vh|F#}>LrEOC?q`*BK0DaCvA?on$C&NU6U{-f>TQEM+`aPe)XB|7I}#VM
z%*J;&y($3*CXkgl<xQvLb4apt;06;usPXYqF>uWWjowC8PzV1n5o(HGX&+2x3m_E7
zX3iUcT6yn7m4Y7ddUQOKnOyk~F|bIdA^i}s?y6qpQkzMgoz#-1y=%-5Q}(>ZEIc=4
zCuwy@_gvo0iK9Rp4f+`k_>V|5brMt<BPc&3$X&3W1o%=IX5_%Gfm^-gc9@ZOtO-Mj
zO9kOgjR+6>GbOORcZgfYsgsN>HxC9TJl!lLC4XzDLJNt4v8TA%flGvyi`0nC_cp*(
z(U{EjqYteUfmJbWO-Ol(0j!K))nZVLFpP-8AtYys91(g(r9;mstl?yA@2I)El}l=p
z_mMzOomYX;V*gS(qt99>#Lo4ds@}pLrUEeh@CDuYAtj?qhLp&Q$oW<k9N?EM%Ahcz
zhzQQ>R=uh9$9t{}6w#mh<dS(^;iS|p?=$1K32gxh32%jbBDkQjqf+WC9?-t{>elWp
z5t>4DL`jJD?_=~+n;}7djH@%`My5Jyc`TZ1eN3eM^dXWYnS!?*^%9r>$+w>96m#J{
zP5ev(?$VslkXV53G4SE*28~qh1WEv;4~S4DH$bwtR}-;Y$sQ$4mMr|c@NRROnGprr
ziH?lfmi`t^z+fQ0pb+9b0mm0K?|-H~I}c9KLy*fT9{E|#-UyX2dVrNztY5-5;;;uJ
zpV)Yu+FEV^luh~HA-dz8Is392jRKxDu-3LxjOShH4rakYf&Z5lkum7+XvFqPJ%aV+
zONib44Hltea#b`$&<w=lDqfBtf<iH%^xGgbzgfc$$l4)+nqWl67s6S~CBOvR00^X-
zyy1xMAt5r`*p_b=Zfc6EfyRNrgh($X*GwU2LY|~RcI$aBb1kkJ7|%}XU^UsbC&IW;
zL^vmO?E9Z0wv)eTVPq98%HkHe802_LBg8v0C3}v;3eltwjCh27YIZd>SFasG@4)AC
z7F#ivEVjILIz3rEh-{MY1cfeC^n;#nuI?h`dwja4>M8D%MVtz?Bdv}iCKTJKZyJ`$
zS*y3TjhNEOlDTe(hyCd3m-N2qgje_=$|1z4dI6$FDex4p5!NIuo|G@Gz8nZMU}hlf
zUACJD!%pOt6G0XOYi8|)yFM9RpY(rF)d~)@8Kg67aRYqDXMRN+af^~cB^Q<aC`lw&
zGc)f#9WvXk+4F!b#d`S?b%--1Kk`<P7v#1J>^zpV`!p6@ad~RlThrbZW-8=%0=~ov
zu5RxJU#;rX+NIIY4l+7{bP=ol4rrWRCkXY=ICEJuHH93s5Dt=Z%6uPh4NwtLUcl8t
z6r4-soq4>^<!eE#9WM=R%b34J$|J6}e+1w2VpS2Ck)CmVS0l^yfL-Ugew3cl^bEqG
zr(%_8-B<Lr30N#xAp${qJy(<KIsDIyZ%BEB&jwaHJ?P_lUr%5TnQQu+!%Q{^1qu-g
z>Z2xdkb3WWpL{4FQu~np>5Cfu{&zQf=%a`JwjSu2K#ue?KhRbxv}m<J=DOQ9F4)?l
zH-JyP-iI$(AQ2lFiK#6m(KzM3J`j4ZtNpO*+@rpt8bcaKvqnK~s?8bw^=L+$Ycd?J
ziM|4&HA(TsjP^BAM1UmBq{3Fp1a!z7Va24~bZfdLY_Hf!alQhQZj4xaz68RW!g5=C
zMGF|Qt5IvO2;JGnFup0>E26c{Hb&{C`_4AIvvrd;+net`%Da!X&2}^=TPfeZN%<J%
z+uF*lt*f~;=I+=^NBwNuIJe{5oo;M8ItS5TodXF$zSRH%S83n^2yD<xM{e$I&VOM9
z5~G%!J;NaZf(P}$cPn{_<dnZ(g$I<3ldSeQ$N$;MMLq~2%G30C{%xvoJIUYpCH{W&
z623^<zj1fJ5Ouejb0fcWe04C&IYmFz+3JWZvgk^!(lE*a-&Y_~rFbz%4GxQKI;=N-
zR%PbR@P=}GRi+pR{t@MDdRi2x0chhXm#9MMPz=brLLZnj#rO_KUK(&7)UyPf{a)qz
zloXYGRb`@`h`$Plxey8(%uyUi6bpjjLL;0|CHWqMLU1)SKI9~26B&EmlM`X=s8GCV
zg%kblot7^h4=dj>G1_s%A~J}+<=cjBqi|uWniDxO)^YRF@zxzs8&0kCKWv%K1(+BJ
z?>6f@%rmUnJ*Ww|gWRWC;+h0HitRigk_mBz=)fp^Lng7Ecvgz}mf<f%6`>sD3F<<D
zr|>@-LS%`*Qo{a{-=0t~@Uifp0)}rio{f%_ixnr77yU$n%f~2ijG27-V9ubb3s8=T
z6``XrlQ*_$KX=e5V-Y%Pj8jgunRGM4W-~-mY2;0t(Sy(3c%uhSw8zAG*2cew*3uUy
z@dQ*9mfouH4OUy=vWFTyToZ1=Fj&feANWaM0n5lbw2|>YOida5^dd^6<ys$UpG1n3
zWB#vR(MTsL9atL#f~OH&DpWE7ktNM<Wc^xr-D}r&&!X1l6a;EW+teV5PO<|mf2^DY
zF8=>Vf(3=0B^@K%#&a`RsiVxTIIp5NRPk#{OahQphkvMcq2#T}fVmU$g1W&F+(c$g
z@r*2n2P4r<>kB%>?Qx>D<3YbYQc7gRW?0-yav5gJE@lSCFR*tm@`yqNK+`$4#0aG<
zoF7j9HUEEVsGn8x=Sn(x7ITOCb?w!|+)S|p6^x$HT<|FcuFQxVft$o>+NDmghKOzx
zv3XatJYVGEbT3Ktvz3e-U2O}Bh21n{9!qYE4Ku^Gl;x)d9%S<s*`j%fhK#?LPi0{q
z2nmE)c4GjQ<{+(x)<&vnu9EJmZc0ZYZl?(dMX<cJEd*jn5<mPsb(7h#om3*J-4=pF
zSlAy)?FheDK=RD@RO?bk;u+)-IHDPTZ<OaRvEfGX;~0XFDOw1=Eegle80Ld0X+HTC
zw-9eR|2u9{Z^Z6z+6RnF;WT@2lWpYh*IqlI<Q65jDmh59I`EwP_*Th73B@-dj@sn=
zj|n8Tgb~-7d;bTNn7Faek?2N~{{^MN(Bl<AG$t+x%3t97L(*D{{B`mQU%8$1$fqxP
z3V_)OAHJ^gNhO9x5{w4{`7J8g>>b|Ljm{Sddg_%$_>f8VKA*63;5K?j^Cg`NuJ*Rm
zvu8sgw%gz_z3R9o#ApRWMCb!H?00}n?Dw_bh4{9oh$dnG5Fmd&l@@O<C@tah5b*|6
zi39BTod};dgS26sH!(5Mk7i3do(CW&Wi*X}nlv{ebpe2DgOGWWO6?B+4!s(y0#Ky@
zUn4)O@FnbPh_eJAX|wVacV2@uryU-aiU!81Kz3$942RL{^sNnJTRB4X2lf0?J@aZD
zr|*RS<fjyYLIyl-5h$u7b|0aO?|kt9yit6LYFd4~YH{^ni=%~6e>2FlA?>lWlpdyD
z&+F4asQd>lSYj5knwb98Ond3tmU>Yi{;Br@?!T*%hnXMb^yM2{u=-qG8|5AP8biqJ
zL_Vjv>gU9y)<irH_<x^taDIp={YmHl#Px7vD=Z6dD2d^kNa-!F#v5bx<of8VusdRY
z7;kKTE2Ou5XmEZEbjBDOqc9s%B=hOXs|AqM@Y=RegZ1AXqO1L%PA-T}Dc9JF25&nb
z)5H7(Pb>mBHij+8)-WR-2wmj=ko=(6Zw~T*(ip*aYn$nvQvP)y)u@>)XHc-ql!kU!
z*l*`wC1)p|kLWp`TI}>Rzs(=Eidkw1OMmSY2^co^(jFNF)?>UB8$?pG8k3-Wv`p2a
zBHkDm3g*>9>aLc?Mee}*vg0v&r1qvg#tF8gt|F0<@B;?4I=cKZHr!B#Q8puVuqVY3
zHQr)u_0eu7S}|+zy{PetyAZ1ZMB3>4rWZP8E`Z>hNhgz=qDMT@Czt&X@epL{_GqIt
zThEuj`))m7luag<h?j2YBPgr){k6x@F+F-K@%lDB@_Vpg|2eHfG)-Z^|KH}Op;89=
zGs>GZ{SG>C2-arn?{i~sG+Se_zxpVff&Z_MexNn6r6(%bl{rJ=pVPC`%u0|sdwB|r
z0OOp_AjTBk(S|FIXlO^3v;eA+JVFoS+uax0Ohz-MhQ=0T=+~JBw4|nH{3iK`pVi$j
zEB6nT`%C4Pm6#CQWYr(j<+m#_skLz(Ezz|=u^Xn0uOkM-R{W^iJWS#&SEpocExZ>L
zG0P~+m{mnYC|CGc1NfW0pLS3{=<hQ@)+A}=GwwCttKKc3LqS96QMH`DHZTdDWr2#2
ziP!voJ+J^@fCIh0AWa<}E`_LD<9jrlTIgWQrrW;2(wodXm6m^;zS@J=jG>kk)(O&V
z3jO4697@W5JCPW=p(vvID2yVF7nA5S^^G5sSB-*}QXnJQffc@<OG{6R1t9I54wUFx
zw|q;7U|K9=^dpK)i)j?u8_YwaH^f)c>LZXvK;*DfKc+(#W-G)%eTIJqz&|oY8bEx{
zC)2FPEwtnlbiiEUkD~}`$;(n7QvHECmizTQ_#+CoYoapnM_VaGz0=*i1a(FqwWke6
z7Zi5a8IbW=x@I00<t4j61a_H2gp(wQVn5hrxY0+x2(yq{lNhJh|1@bI*d_;Vx#-|N
z0ZT@}E$DOzwGxLyl!z+c*aB`bPhNe#)%3O_Fcooq6m}j=L70*>Z|z3;6NrR{z%nCX
zk5RA#lqTNBd`M0rEHL>1Bk2vX$Ip@HL&%0be!=V%212WaUlqPE?9t<Y8H`0M!yRR?
z?r(^pF^X{j>ae#BhxjRNU;_f5V;L=6oq+@Y-P)Y{xQnK;1qxaxeyckQ-=$VY0k%%O
zL4nQAxt+?t$q^69!qySbYxs|7s1Iu(4=QJ;(}>)Mb!n&jA>|%b^5N!yI}fn`nC|`u
z-F;j+9bW&Gk^?HU#DH><M?R{Pd7AQ6##va}ry3LTROOm8wOi-RzN1y*S{|;F8M`yc
z8GTz{*osacv439{yR?~oW=9NUi+(XJ1}ORw2G+HHYg!81@JLmVSc)^}2<mqxwN?np
zySwTM$U9tMe7^r|<XwpO70KqcIOLt~%RA$kQ3Y+pUl)1DA%$8%8VuQjy*Gyx+e&mP
zdA3-llWJfVo5YEKoU3;tEBq6h^xvk5oAi^q>_UtpBfgzA@bBd2>K)y1%MsS^NHzF(
zsasJBzJcM^`iewYQ{VTh?+r4?oA-TFO2EHQ&uYEV{krUWBcb}2_0)HLYfo)rb})6<
zA=qU9qqqF1CR7TuZVjgPGlqV~Z`r{zy6Kzu>6pK}HG?*JO>Aa<HY>bKZ6}Y*$@8-f
zQ_0U;);408R!28qxeg7mqtrqRhO_htK--6$o&9Py<J#e3Xc9BB3C|!m5`%+5%?z4g
zFM*|~Db^w06_b%YYJd>=5-d9c0<kxrPi|I+U2IV0)EQiZx@Jaij9yeP*fz5YK!oC|
zb1D&uD(Y-ODGE8XqGm&ZVZIgc`2xIZ++<E^GDS^v!5qGJWYwg&`jjSiAI}xY(D!;k
z>Y#iOzTCL7g>T3~-tx(fDymjD9Y!J<y3?;=@!uiCv<UYeZxB|9sdftYet1iqv%(TU
zNNt<l6EC<29EO7o7%){mglAIHe3x!($haTH1?4E`n>EC#<;>>rF>>Hy7J^z$Rju`e
z_fn5s$Z;h{{ZiLcX6pKQJ7pGpY!F8jjI?FNsHIfMu;^(jB`D{!ijb{3-S_2&E(Pyt
z=biTln(`zG60s$3X|?Z!KHlR3#*y*oxJAq<BVq+={V63YO3bd}7u3Eb690@YKg8W&
zw@j)!Tr)e2@XQ6REx0C2D-#3TKSp~NEFy-P6#CcIt8od2q_hLuu+r0jjUUw3xRvr8
zl%8zaWbGy|0HqSe2{R4I=Oa0K18j4&`}BR2(avZDF&^~RNggsiKob2NA)|w8i0DhO
zK*S(1xkVKRKqx!=tUX_ZOSs^4CfqeAB95pIVd84$#k9Wiwzg$$bZsj;Xv;jo7vR2G
zcU1eWYZtI?3-XT?0oz9T%Svnagx0TL#Ge!S8=o20#}de)svKeeZEZ+JmVF>rJws{H
z%VF`l#BS1a_;#TE<$HKDk{WV9Ea8sVgZn}eDz4RoY-*7R(8CX}?%cc&ukW3DATG>c
zik*2GwnqOx607Df#_Iizt$6vp5sj4Z^66Ukt%JxG6#A~8cP%0v7_JWymT#<qC2Sa$
zZ<|@n@-;lD*n3S10i_Amh{JR3Un7%1(WuWCBT@OY`e=PCpX=Q&i&s8Pob-lxJLDF~
zB8VDuExx`3)#`TY?38R`P2X+Y)!0TKJG^0P?nISH*z@&0jh)x%0r|u)QMc3>qwZLJ
zZ)40GVa;}t+vRPs+`h)H#+X}AZxm_iuEs9jWh?I@r-JW69x`?%{#lSZPudTSoIn^4
zN#Sj~ng(p|Ufa_V%rAd}PovaMc7*fOXm;~_zR*R_+v&ML&rdPR1C2eD?``a*d<W$@
zJ+Uqx<JS7YEBel3balvGF&aKj{MjuVMicjTu}g0@p4E-xiRQC8uH8Jp*V|)XC;az}
zZJamlgX-VMXtaMz_A%ujZ|p<jHRA1M@9q1%?h>zx5$*3D5z35x-hr!G#<YL!fUU@W
zo^?y(7T%RFEqb^9Nb2>(1AI!=!v-+rqQ<(}`?npV&3|vg{}i%kcjEL{CqjQY5y@mC
zxpwQ?!L>uI@@<UxVB-)@Tj5?=8}|-2#<5*R>eo2v9pbDM8VBKE$Wd}AEIAaF%p{qE
zus*HEXE@_)<GkT*yy5naH&keR?QaW6Nt@g4O?R&y*zi7y+I=1GW4j?5VRQBm>&eW_
z5yG|XX^iuEb3OOy-{`q3-@<c`@!TF}_#g4K<2-H5-tu<i1rM|GNAx=*Deei>Qi-+O
z)^6YA4S%BZN%IF9w=2pRr|EFxFmH7yPs#F>|5!h|eyolxc>VY_NF=_<ng~xGkbwKY
z(F|dEmv=WIEVCl~!w{DDoF;Cx|EB`;f2QQmm3&1BzcGK7H0ZN$DB(QfJ`dq%zz0E5
zHg1tiFA}ZaQ=9*xq{&sjwT&(i2tKWzgm8nNvnz{}d{D#;=C^d%iAP;I-NRXRBpmFt
z`X5uStmNZL=1GFwb0Uxs6~38nBh}u^*!(9oz$6aJtx%7vJEEAIr4NQ){T^TSG`)cb
z0&Wp|%^v&-^<^TY5l2NyzZ*H_^Q|0l#@BoC;i<`Ox_V^N<4Xh6W&XfG@6k)YL=V?*
zjRxvIlW9H!b2D?8g3b7qN@;e5pvL&JzL~x0AcCEp1Vz*Y6(ixbg6*Am+ZP7cL?a_>
z<>{GOSG{~adGzFrQ}zh6A6*@2PR*0mWg@oKVKBPp!=^&h>Ue8$*{!azCsQk_esp7X
zu-V@;I2!2|pG*&jAq7lW`TCY-y*5c@jlE<-1LH{^&?A3Z%iZ5DFMFrhsQ#k1NQ+i7
zhOgQzGlpIa-w{X_e~HXSqDaPK?+^_2gbOAq^0_N4s`K-;;kV?NMk<Onw1eZKZi$Nz
za__HPwvUSYqO1IVm1X!MxIa7P$jvXS*hDKP!Vs<Kq^v)pT#NcN>FJPOaa4&!qh@;V
ze@T}T-di|h|NWZHf1om>Ykx{PiL8Ce^nB5izF8PQp<Ii5yr{e9Rry!QjURE;szneI
zl<;dx#0a^AM>I9q9-Y3r%IpB4S#}#?aGB7zk@DJtBO?h_QP=X;f?-#4Zf0hsrjXwX
zr0jp3k=^t(IIz-7kS7bdZQsy##_K=DxPzQ-XlyRjgVEKdZTyCYX|%Qhmh37#o6KB)
zOYi+vHKoW5L<pAD9FJV|4ZZ}utL-lo<oO<^+wB&m1JAeVpeNe@97AW6*xRh7qv=G;
z|3m|kQ@EWYUy^1ySv1unwJ^ZHBH!8uw{2}ZdWE`wPt3sYG2XrmLMe%d!+!%`A4rRD
zbgnXTmMO%KTAGL-hE+gQpGY8}QVqhlBGi&gaVyTP9A*mir!PlX&*ahS-`IVlfRnLv
zxf#VyB|~g|-ATlYZ|}Y_#3;myD1NTmsvIu$8-?^xeD_~u^7{~yoirvzb=2FCjX*(V
z+Dr5%?sHk{wNFn>$cf!TB5W|pY@OW>;Y^lk(fKic_v>W9z<AfVXc=t*ce~4EV4m_%
z!i`{JGM~jfGW=+LBVlcGuC27M;+d({DpA|KEC<4{L<xQSd>)T9r~!fsMfp1MBhhZc
zg%~%S3a>GSv1>iK-(R)xMFSUNwX&qpbr~rV%Mqhy3}GY#t`EThXXc0Xp$ek>=e+#q
zWgu@~spMNM?gMi0$P$HMEwFkEABt75q0rnI;B7&N3=L!6>jD7pUESM^=IJ5~wby5=
zFA%`uf;Q|PlGW0475Ay>mF3b?rmB1D*-|YGROiE16J96JXC|B)FGxu=V)!0!yL_`A
zIuk`-P&leQ7?STUSEnzJkN6k$oGF&hIXZknxvMIBPPsEmo>yX^R+ecF5I&=vNvbcY
z@?|BS$|_2vw(|ihfQM)RU_WEW-=~_ZBohTU8)fY0{BKi1R&Ely`VK)#JBpUrQEX7-
zILM_Ymf6Ka)}hh|$}rpt51~RzhyG`YVp0bq^$nsNa%48fM+2b~Wvesb^0dYr{V3lQ
z=FD^9oY36%mRP1XR!^*B{^G3TVKVT(zs3i9<W@$Ms!SbLgs&B9`(hstfr<GFS9s9N
zv}Hv7^;}bG*E{%-1f(?gO`swVzy)YI;w0V8v)zL7wQv`xz)ZW1R6^ot3O~b#wjib~
zG;C?O?20C{Wiq1ZVi!(&hEt{Gz2p}*neBPXPHVP>Qk@)r2-g#29UCN#4z-T+d5b%#
zd+N;T(>AVCr~EaR44Umx{wU=1XPU2(Id$qZ|Hczk`gZP`NOwGaiee0I_(QqKKqIFp
z>N`XAu-~wZ%hTgMr%v-Y-BLRj)0z?$@MXrPl;w>7j9&3w^mzKrnNw$6xZKlc#&2_5
z`(+K`J4u3Undqd`cyoSGPk&wI1?6mBKcUONt7K7S->=-~lo;+`SI&f3|6I9WSMnc~
zyrtwpB|obquc7>yat|xHN6E0twkX-9L{^Lb6(uqT@)OD#@gNDaBMFS=Fw#OO%0gfJ
zS~LF@UAB(SC)9zRpHFf_<Y(<~Q<RH=HqsFyi~jcgY3}|68d4C`0G}m)JC{ccx`l|=
zJ;bWc=K2Z)nLAZ?;6~<7&|OcafFvwQWb3}%U@leYOFUxt_PG9vZ)bK#Jr}=`xu-3E
zvRPvgB&2AjW39Wf@8E+nyNBYwMm)`6JAT-c_ES5~pC~GJ%j{wzsSp<x<dl^ZVr#z4
zLKNt<MN^hDQr>@wW*s71(Qzq$^*8(=Byl$5*J7X(7tco#e29A|BQ^{Ai^aK4x1H)F
zsAGpvhho<ew=@{38EBQ3y4>Vp2nQH?LG9xyP9wOV-1UMp6%dc(zsTYMBtM`Uw*5ZM
z`x7X!tG}|CA{Shc;588$X9l#U;18p31%tY2fqN|yumTExySla=(E_V_(noaAwa?}5
zE2;pejBQN}^@5K}RK;&bgFW4H<?0A=lBa_vxLjcDOc2rxY^8dUr~uY5)F-%&oWksm
zTsICVU`Hd4lqR_~^yWqNpxOj2>!|GvR-qbdcL?Yzd<OS(ruCMWq4keGSiXU+XXx{R
zAKUC1?R{8lp0}a3970}k64%<C(FdE1cJrQSl|$Rz8i|}o*2BI&>-uWFMQi*U+7)Q`
zv%D=IvNCVV=shJzwcZt1n{Q8|i{7i{H6SaaLSI`AGuxk3sp+6!(*{epf|5$L<9zKy
zlS)umrEu-GPd=@i2KRnVy?wVX-_U*cacE~zXtexp_0VL@{jaOK=>8xvy{S9n$F<!a
zr;$a;+g12&N!L@`3ttuHPGMf18Y%pN%E%2A{#g3mB3m-mQ}{Cx)#S2;>!v48W!~mH
zFVP{0Hfreoh-wrxknIBT?r2SE$Od|E3NHmHKCF>yPv?jRv@`2~2bUc`O^76&U<8q*
zF!Pe1cLrfVqgM(*CPAavM5>aOZ>%?DP%BTYek}ARvaXc)@V)t7Z`d2~247J;`kcWo
zse(s>5cNL4YyyhrdlT^J+0)|*zEZJtF%(9~)xY0aZ!Yi+;spf~XcF#%bb2X4J#oi1
z(H9C<!68av?Ut53xrx>={nA$DKypZpz!v0F{I*ybMMP+*#*S$LEs>7FmiDltHYy8E
z`Mugh-$xRvW`cp%h#fWNH><RT$NmJvb12OanIP5`EsP)*${6Eg2#^^D59ywrCfw?v
z?IMW(nEHNGJN73?-gHr0Es&?arxi`7gRX~E_n_(;k$hU0U(}^RQ%h`s-_IL<ngRO%
zJ2zKD-}eo}+H{IebwE_fFLL~9!D_!w(n#0(!-bNrhvh=wlS0zTr_HNF%{!=Yr46$C
zNz(_}XXDrMxEY8TMN*73IAt+WdO&N^4&pY@=)ndQZVVV2H>WUM1?0STEzT1ZaYWi6
zX*1u8J~3l9-$fvImfv14`+1xLuq&FPR-O>la%t<wn4wRKVzd0@%K$TcZck;#E#)A<
z(4lAozoO&=+8y^QcOOZRI`hoq&tQ_+30?S(sO>eQr#`}zxE-Sza*`W^SP!Tdqly(~
z(zcw_5K6{6KF}s<`3*JvpQ^4gxKHqDwQnlS2-25O9$zG6DBlomJ7GZ{Q#jK})+DOX
z-nmJTjqlgqE%DqGUL<|+_X9T+q+Jmlq*xq4<)|;;z_P&YDe%aKtt0KCd7HM8xNm(~
zZz7_(!>sIf3Up3Po%kZVzR}yJPQ)Nz3%WKSbt1||=PLq|)#kFTm6OSHOPz>X3RQv=
zgEtk5(X~@34kMhGLUDw@QYgwuH^bkp{DpyD$L!2~pak{em^dHtKq#Uy`cfovYN8Uv
zcwZas7g0IGrh7f8AkQkdKA`VuV;r`H-ioLopQLoq6o{|Es%s=t7wT_+Qw2E~mT<=#
zK&>(e>TLCJ$a)w;1^MRYJ$S?K*gl84a0|-CpN_^<#}AR~U*z!c6ga^_Z-lp!8f||)
z&pdC#$1zeiZm-Mq8s#L)#`{nZN<}ypDjR>xuC{txq->;hroOGdeVsa{ZX8A3xRp2I
z8ucQIMbwSEQ8(_O&YlL!N7Re4`I3~4jh)^YHBmjT?`2dd5wAIQ<G-Trep5|us~>1=
z^R@v7c9YxfZMWPljopoHZneD~@UOeMMybc!NZt4%hUt6;BUCruilr2uqo!`$ZR*B7
zYo!h9mw4^bPW20;3Dt`9yoa8pR(z3B9&D6&hkXqx?MjsQczaFV_ZQqd#FL7UfBS9U
z9(Xmz?7jF}9LubIHd9yYsHa|)rFLGbNdJ}kZSEQek}axH`!2(4`vF}CUX4G18ua`E
zt@m5&zte2Jeg1Y+WxlT3u7_6rW-a%OUk$Cm4tTe`l0z}ve4^GCzM=IT8(Hg`S-h25
z+;14SZ5A;-Jm4KPRoyLXx7xbh!i*e5Z8+#1Vm`)is+9bbjyaZ6@-R=4VsOl!f|`rH
zA$8eN{hG?`c2k)hS{vW6W{KL)j{aivQdA!Hj<6$cYuv_Yj<P-q@N~>Tx<T{*O@J4W
z>Xcpn2e~BfP<g>$9aSpeSbn$-e%7wXk2EP1n!$d>q(_0s09*<SV!UDsHYY6L6V#~A
ze6wlm!oRX<`*{IbHP1Dz+BSgEmQ1tFz%6on6{hf?Rq~vY&(W2W!<^CO^Gg1J=lCB|
z?gb@Ad4F>aoBv@|{)mz(bs$d32|Jo3bjLhU(I_>E)a(lYktKY%AKB%hC46%lK&wBa
z@yfjA&4e>`F+|s*QDi0eC!$fzO75<kAezl*i1?}YDN+ApdWo{Vi4qZ5{*01YB~2xh
zf0oPfW8Z>$CrHgN;YWA3nosEszDUAH=~fn(FSo^EF;Ukhs-Kp^$H_trC7xG5|Em=}
zCVBaJt=y!#Usuwp-Wk8Wjmp?WpW~m?OPVTYQMsy;B_&eE_{&PpE4f=wk*39$K-GPA
z@W)lwrDXXb-F;ojk0_DO#n%aF>05k-oA+0g*v`A4+(jiGm5B^EU9K-K`<GO5S;?35
z)K%rAGx5JoxrP$kM{g)+1g}I;U(kX32^#d_yCqFWB0hRU0RW}_$Pu?jtXsQ|<gln6
z``6Nq?8dM+`WEXq*NsREl0mmoh~H|o6K#3kAyV-K-FlI3Z3nSm!jD6|17LU)9#X;o
z+eC>!s@E2sb-GQESlE=y6eG|kIy9Qq#C11iIzK~;PiQ7bct9=&6`0EqU^uhIiN{9X
z_Elq=n;k#2u%B8d3)=itT*`wMLm=kZQL&-D`aIl~n3uyXO)RJwP$IV2C((+0;H>Bk
zdHvpiv7iNGL3@Dj!$BBo=+qoy`ju*>$%C$brcjE09-XYs)k<*T1c9EJsrZnvr5V>?
zlIWeamFcq<gQcn8E~#o|xqkMfsfSxVomo;GX+=wqI*>v|^f}!SL?-ODVlns>wfZ!1
zRIK@%iPa;<wH?(MoT}4Z8{u4@T3DJsS37!~e@>ttR+oyiyGeVC<Htk!yKvL#=J&B+
zuWKEwn`rJ7zs6s`Os0`tN6BTMOT$!zJc_O>`Y>8vGz>zp44ij1<h&i14Xakh0MulS
z1nG)ClV)alm~Nsz!wUBb23+q$^I+zkcI`h1txyo#buH2MDHgt|)L<O`qa>^2(|qY-
zsk(5vIfW&C7OON_n#Es?kIS8{OrKi@7T5X2xa5!y95>ngptob{{WqD7Ag{RHgw9^9
z@m(&lLi%!;P>AV$@tSTgQ<TJBjprOmWa_JK{GU)?L)KUGA?nosDc${|lDjk_1B}gA
z+DxuFgp|WjX0Zn!e&m|bzwN*sgiALf4(n9+c5Vs{PEBhSqaO`eq{w6Z)&#qW(T+gQ
zcqJmhc}Wa=Q%-dWKQRn@E~6|=Zlh8~(Sla}D5ARre+Ok}jWI7~z5a21e+jeO%r(S5
zP-&<g-yJ)@hiFPDKUwVyFwW%Y_kw+X57ykipRyh1T#UVpHFfu+Qe8Iks8m(Iz~Aia
zPL-)xGx}Oo?))S4s-5GpbNGvdho`~i5OF}}XK*L9<Zs@L%oLRc4(|MpMzZ!YcZ$51
zVy7gk8e4DcnvoEDfaez*@N#^RudmT_MY%q5Ma%WSPHZUDs&>a4zy<tG>JKVH&g=G-
z8CTQp{5*GvB@o?_yvd~O!>5fd@c$mG;Qt>=ewt)eAt?wDAPB*6EFE0kanPdCYbU7?
zr+fg-+$<4C7fO8Y<<Rum$tvN1YvTYJ!Y#=hLXn{OG}X8_=h?R~*nnPXQ6Jt`52mlI
zphOj@aZfvX%KwKN!7nKJMJ4}O$uB8kf5cY%o}R-8lg94{MBHCm9elo8xwK3i0peJ>
zBJ80}Xz|NxGo|ELl>Dj^5x@SgDY3Jnqv8KEUH&Awpnz_Wz^DszL@1Vc$dO2s<PekG
zqJn>=<WVKsvi`qT@+(S=9Q-%r_#|KDVi=?HXLaW{mHcNV|3!(RdP@w|+iBIYdYGFs
zL2zHK5fauHY7;MyB=s#68A%|gy-`RcULL^Y0YAEz6$iL*<A9kim=yO$^i7+M$8WAm
z;P=*t`h<SZ_)W~)VKA+^6SRFlzcpFq+}vP#O`&iV4WnJck&aBIF#;ym5HbQ{*O8Mk
zJh?<-Z5wXX>BcC0u<k@=)X}w*dW>RV>7*RPGL7a5-DcD|WEdZzTxQfr1$df_j`N`m
zgUC^w-BnDW)i1(+`-|~(n;CU@$LmFg(FZ@@>OnG$qSsFkKfHMl-ayzRJ#W%J$Em|`
z7=!6>Oc+q>`io$a0huE+(m{KxK~w1b1D=NrV<lGmGBt*g92nf7R5GNYPbOw8v?Q&D
zXeBf2WAzavAKNj;jU*pC>T>bJP1__NI~$T#Y(?^cqNnzIb~Wk^hmwz>`dEG2`u4_7
zlYERI`Pjny(*|isY#!t3^<79lc2Z||V+V7)i_$%kdtf*{O3mGvWS4BDm^PcC^?#!7
zUXzk+ukUMY_qHPM*h6j)2Gq*!Zy-Z)D~|CrC0q}iIkY4ni}1WV;HI}3H;v>3Il>mg
zRP6jbK?6dQ>s=-TDXs0@z)#0(&vo+C%wQzz81{BEW0DmuGTvJndwHAvjr}}ZvXNd4
zoArF=fBjaTB$H>JE<<HWGTaD|6%2ZN8Qu8JM~A`lt=>M9EntppGjot9jW@=5(te(E
zV54lIG5lQ2490(^WAv1c-_qjy|CFQZWVU2NwUEqRO#<Ld_5LbcbsN>-V#a!ZB|aPO
zw<SN`jOX5jg&jZmEpV!T4Q{kWzL*`rl&0rz!l8aivkJW1V1+2=6w?Mn05>xyY-2yo
z!E=E&{$G<3{Ur_LPc)FG$)5~%q5l3FTxbq1)b!@YkJ3kwrDTTh*>>}!U6@IX)q91R
zzK4-GQ@!1U&yhut*vxlgKSvV#U|F4c-aE0N`uaM;dCcVsVsedv6%(2=_S4-JqgMEO
zA%kM-&r^k=KT8#&NW&YkTuZ!y2izC50>)Kz*{(FvUZ?F!v76=mlsXttA`O|NB!sTp
zIM8=yKL0Lli<-d_t;|9sn@A=S*iijB)*v?qwq`=hjzoOq8@_5!Q?cWx$%u+mXz4;7
zBkC1R<0)+@Un#FREoiTmp%Q4<f?63QroLF-rUC7XG@ws9*0IHX{wrhp#MD%XMGb-G
zuT;fjE@R$BxX{QpIyAOBWSoy<P^@TN9H^pL^JSw;5N6zJ8sXR6*N-YkXOBV&*3PNh
zu<0bkVV#KYr3-VDCEK(|WSrr`cK-Ew*Ah_cVP=<*S3HmSu*Mbp|4<XI^B(<Zvc;Vk
zk9m?zqyPjLB*Re1sWy%hu2O;1u~_;4Yde?MxT-jgCz&*vbdt#=t?9#Br&yXqY-)-X
z6|oDoSZRXLu1YtVnLN#lGZV#7sm;D3R1r!kC?eFQE)-FG)rD9eb>ZUv2QI~ho6`9C
z{?2_RlSvzj33KPpxsP-1J->6l=l9+vd3K_|%_REUppdRpd^I_^Y$h?@>`U5^;=Wd>
zbe(4Q0}&9Pro9*vVMh{bW!Lvu%wUafqVkl~i8d490zL(<2m?)cW<3L?*xl)0Jl0wj
zCkjnGs^M;mdP>FTwD_LlyPk(TQ3hSsRZ@A*5b8kT3s85WgF?ZnzzWUo1o#)JN4pfW
zllxgR1`I(7iFKZo3^REWK%C?1?wJB>o-Tp{kqTW3Rpsa+sv6f2ia={4qNjnZq);Gj
z0{b|59pbtNkl#b}oQv9({z8db*?v{&)3>>caC+TdZ8|}2H2pG#X>N4+$vE$JDwkz7
zPTN=Yv8SM5NnaAWKeJA``q!P+)m1OVHmSAQi+oDo4)%H5Ph<2G9jh@PMgS%CxPgw@
z78>J{Xj?6O!H`ZdLUG#LM<)h3vprp0m_?!K6Em0voSD>r&k398WW5)SNjx_I8uuBk
z#?a(dcS$>~rSjD4?kc<M<pFw_Pehw)a-Nue(!1s}n#9GP2p;@?!59>!)>ZnaNJ3CY
zzp)N>jj&$BF&>i7@uNntAk-EuB3cAIEf%UTLncCm1`vw~1_pv;Z^S{vMAY&Rvb=3!
zv6o?C6{W?~*iUrhgy*#m`dp<pr#I>moRDnr9^>h~z~q|xWJMdp&sctYL+_@S@;;s)
z5L?Bu9M{&i)kIEO_YP9{)#AQTW2~JYvv<8tB5Lqt$Yeybq81s2Z4OF$#w$_VLQujq
zv8@BC=$<~7!W?f5L#KG4Gz?X81}%!kkU)@!C3Mqu5G2XyS_tY2BGY)2SDu_3n@kBo
zhdm5M=%KLEa|`ne<)zuNP{LUx4L#Gj;<|Y`Wpj|7!&9k>Lcj6kt|1jp?iqRc;))+m
z?1+|_7!eN2>kVpRk4X&5$&8v%&Rcec>6h*8OOF)fOx)_N1IdA>XOWbS%y-F*uN<sN
zIglq4LYo8<YFgoDDMC^Ox71mDnuyWC5L$jD^o`!&eKi7s9<G*{O;lag?81b5pCI5N
z!I;ZIgN61KESx4_rh!5X0WWR<0m>DG4Gf}Hrg0Lg<z%LShd~#xK%Sg%`bW5KmVY`<
zwjBaNcYG*Yu<e&JlmHd!5`+sxSR_Ub0AdE=-TJO{?jl8dr3+(l!A5Tg8%MZ)hir9V
z{?oA04Q!xxcUJ(4!W!5p7;F?qmyc})HX=#<WQZDu5SQtSpuIF*UgC2C8`BXA+#!{A
zk?(y{iM*9#v1+k^S->2k+k*BmlcQOyPu{-%_a7srSB`saXM#C9?}9mFs;Z=oVUB^!
z4Tim6*v25^zEYWnO3*q+l0t2<Zci|#C&dz({42Kmxd)7KU+eA$G9BUIp^9jidK!Vw
zpBfL((5eJ<B#QRRFRv0qItKiRb7;H0zg;0G_N@h>D7X7YV`yU_YNtXCLR%XyvJ<Th
z6A*%wnLI4xAFC;)0zpzuCU0`H0!$Dsvm`{@I)k%Ix&)#n>OV4O++K`maWCF2g2j6Q
z=#unH(3K5?BOQbg&bXiS_THl%*7?CUdxep;vmt=>=#O!?{;a*5Y&dBn2>VUPf6?Yw
zZB}&awTSkwUVNVNUaYCiNrcym149&?YonK$LbPV5Ops1s=DyX&?rs~a%#iat4&03D
z5jy2t+wci%D}qcq))<+jkxnC1N2ViF@yt-jAcf|l=;@S{xRN*Iip)i`zSPQcl}<yc
z4%5?Z=iPS)``2jn>sZg_bA~?B*u}h#&<eA)!vqZ^iyl}|91w#ePo@P#=IDwzp}H_P
zUaT$xU<lF8e4{wFG^u0-hG{;9p7}0~mc^e$f1<}$K1PpH-G}r@tYa2N&F9<Fe2VlB
z09wXY<AXJf98G6nNyw`YYx;(E$r*1Jsiql=bps*nE$=^x1erEYFH815W)Egovz`Tv
zR!{;gCD^J?Elo_!juT%FQNPdNXO@E|f49}I>2lwpOCx$TB)mm|yR3~4ZO$2O;~j)6
zEFXkUt4{OayS?1Md3VJ?c@H+M?p?K{kzHzmMtHl_C+<?^R_#+?_#bgP%upv5?j<l0
zdsEbEdE%kwmEVg!v3yujyVEuK|8>+b7_6Q&oYy&0gh@TtVG-yQ#YSviS&>{rD_GJl
zV6%!ci%HPmEzjL*;4(&sSm4(nI7`idiYw1*A%|F;L=7-=!!=^08Y*Xn%CrPUmwP*6
z>WICE$KadYsLIx}dDj}Vc~5Nxub0dm95O#VQA<O%Olg-7mWt)#z~XRmYLKs^B+c}}
zcvQsC@?dd_vX-D_ahT6hAFcm=QIN5NJMI~pwN`UB9Ly}{ihd?=ZG*{%h{3UO0Qb-L
zc}tfVm>WXyE{(i82$nhu4o@RcPJ;;aR<ppa+zph@PTU_hxtwKuV3=elv=ynFP=_{#
zVNH4QwD%Z4>S0(b_%tt=;{H(C(}zv&WwyrsvF<X)&j8cj_&4gR6`f?r)UQ1qeOiyK
z1wb$Ez<}Q1T}FEf_*(t+jk}$&5B<Nnf4-$QT+}9E<ZMpd5;yldcWhSEynPEcIyh<O
zSF<^@aybBM>I`>*dIpdCnsomlOE}Q-)X@CAcVzrXC&_<4=vjQj4xt1k(YNAcL&mdb
zcl-0r3wH=>!^jEMY^ao`aeAH<v0jj&{_&al5arEImz;3R>D}QdOALtG!oqCT(dZ`k
zgErz!&h}UZuT4v7JhW7s<_!5i1^spHv7=zocon-6rHU<)QnF%hh2#aB8=o8#Yvppf
zIy+mLv-4f2YBJ!_3dVvMc<VOtMoxTs=mMrIP(?4za(`@BoL+3upDvASm3}*EgMRtk
zKYhBSFT4bO_|Vu=Z6QpB<YO2pnn7@2?M*|gny7BQ1Gh)*+N;fjx?)f2p0OFuki`nv
zr{g{|#Sk)Az{WkK>;4gi4QTI(&gecBb?b;dqkCM3<QQnaIA#rHl4z3y*pkarkd209
zqj3>rpY@HE)WK(jXl=@7I~nIfVkru=?-d1|%>&E@DwFdOi*s`(Zo7lvzinr0(*^w(
W#{IY7_Kp-Dr7Xv%yKr>pvHt)v*2QH2

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..14224bf2a75c664e64f4cf1403ced86cdbe34623
GIT binary patch
literal 55828
zcmdSC3!Eg^UEka9?&;}yKX!KZ)%)IA?Mo|Jl4V)eW4*F$X)H@C%d{=GXQp~)XLhD%
zwz_9E+te6euWZR+8O4DRK<3U80yqKRgb(C84gqeQ03pT%lMvD*xduBJnFJgz_l9r@
z&iD5}Roy+0wT*%M=$Sr!>eQ)Ir_SU5{{QQto}PS+&(GZPeC2<8CLa5O4x&$*%maLD
zKNpY1yjUsb#l6I0ycD<J#8R@9jJvW_DQ#t$QkLK3Vs5Fc)Me*Vi}|JQQuk7!RIuap
zV$V`<sduTb)Mv+;#r~y%(m>eGU}=!+vx`GZL!}`r%PkHsjg&?#-?ccpw5zns^7+NF
zrQN07mhWC1U)odJv$VIgcWGZ~-_ril{-pz@153qHap_>`;L@SeA*)wdJiK(Ibj0#K
zi$|A^m5wbPFCDkz-o+D36Qzlzlcke(+_!k!(y7v^rPHO;OJ_=FtgL_W?9#c?Im-_$
z-oA9cbe{a6H{=a_Bd;V&cX+pXyS%aI<E1;j-QM`~vC?ZEiFtdxz0b$Iy_M|3U6u1U
zQl+~o+vn}4Y=7nM8_Ci=93SwC92ZrGlGjpl&^tuQp~^iR-^=k~?+C|7?D%yYAN7uL
ze5~?1&fmxJaqk4jC+zt398Y*BIX-E}_j`$_ViTvPR+z^8bMuQ{acO?7zUo(s%d`2*
z_4-Qf-m_=t=IfVNr_anRFP&XnotgLisrhQ{tXHWn&)4RwbEm!e+1b_F{Brg5%JurP
zKXdu)^y2dL*}LyPch|HxJ9F3UYtG+2`<i>aJMNgh=g#T7r(bjaeC4j&D|c1yxc#m>
zr|)RDRBp9Yj#_HnW&61^E7y1G;_^zRIyHZGerd&DzE<(ty>q>)y!OsJ?|JR{d+wN?
zzUS_{re<!x^X|L7*LWVieeLx0w0GWn?Q74SpQ*GOXy5qk+|+8VHa}Hu7u}){4=*pR
zynViQwm!9_0jm2`)!OW`zf|#S`O~M1{LakhU)5%soQOAbk4-I2dsB}+Igw~~`IRZ3
z>T{K5#;>eYrs@>-&Ma5!m9_fPRCQ{u;y1gitBZ?a8P{i5ru_Q+)FRnb{i?UbZ`ab)
zl}dTepZ7TErM)ZVM#Wz)FIN|@Q<8sdzE*#DxjH*PH<4@hE-ufN>udG$RINO}G}p{j
z);L{jX1&Vn)aqirnVzpLO|3MuO!pEUp!EuaFuOQaZx&{jSF80hJ(yamHFvQv>QhxZ
zQ?Ha~{pF=Hz0^(O&CFx-^*1g}=@F}p#MEN>%3^cxh54zu<?7VpBjLzi;Ku&^^lF{e
zMAgAZng9BH{rbguzgB;$uABEfxLRMXRA-jGivO0!CX&rTd)DbneX3lWxm@v97c0#n
zM$4a?sh8)g_2qKCQmrlf&Ab+DrMx(Qr9y|gFI;%^@u#MiRu(Hhhy9Q0(H>i#V}Ku@
zUz@Ksb2IAw3=i2&&R?$8rt0--Gk%?WX2LUF?egT5w{&SHp-JVVU-ne&VZOB#Nf2YB
z#T(Oay;kv`^XKc8;u24=c!qAySBuZF#ftUI^R?paYIR0WU#u?|^<c%CcB9ogTe&v1
zXqV-Se7xmKtyo>I7iTVOy%eXa*O!+4%9)SH7`arfvN+pJ=^p+OZg0t-JahK)@>1n2
z7geU0m#>`lrs`8?=ltc>m9tN+Rv%-nGJw^y8r!oM*wqg``sCR~Z|SUcl`S7_^s}C~
zvQ)lWv(aM@G<&`I+SK%7#VcD+kLsyyD3bmO@-rF%ee^5bVZ*YQOb`p=W3hT{As!^W
zc#xcre<VIk*)9!NVzQa>D%a*`Dm+ozKua^lV)iCd&4hnD-}B9M?ef$LGe}$Q2lbD0
z*c@1`&TH4P>dUT~+bDMZiu-5D&+u)1_J_IJH72eS1DC~0^vO$jNiQ`Q^U_|%%X&Gl
z%gcM+<O*Jo*XyOcKK}ao9pIbhuaB?gX7|4kn~e(#Jl|DHdT$mk00X2dY40uGsJH9+
z_>FieQ^{`RZpJEIH@L%W!W;8;TY26a_x6A%y1l*LK7I?{e(wOkJzmi}$ZxNA$UDq$
zU!~tW@=DyGeyTL!TJs(=sIT@1!;;4h?yF=dESdC9d8h4O!`@rHGu~PHHsYQ0Zl~2z
z?+GB`9hB_y-sauuy@uSFcb9iJ=XQJdc(3Jm+`HF%9lv|L`@Gll3m(&xc=vm6cs|B`
z_IVF@4^qD0d&qm3-vi<KC%uc_BX-}S_a^UA&K!J!g;6@hj6c;(UAXYX<FEd1lKIOl
z&R2nk&BWsQ$%~B~i~sB*yYLe5NPG2FZLn9V7<<(t`O@i!_x|2*-!=Y*FMaFqCm#8&
z#&^BpOMm+D|MI!Vr{DL6xoD2V%vavfOjcHEGsUp{Aip`jYOD5TTCK-!#5UqLxk0eY
zi(kpq#_I_$aU=fzc(5yom6P>U5L-wuWWGP<CCMfEO?@nNBM#1IMVGTdFKI65A?*tC
zr1_wmv^yw}7J}WRJ@wuo9*o{dkngMa(~8zj=A{Jp?4oRdk^%oG>w`LTBTmUsJ>_NV
z!}SqLhBiiTGR8OJ+&TBLr0yJyd9f?Hk{&MX@_#89BR>`llkN_NNXG-NEbj?&s^@<>
z$Z@nc*h{)E*#C0ePu2Ga=^OEv<F!;h!IP#{kMaw&cOXaynIP+Ry#!rnwN*Mm-E*o-
z&+_U+XK#Z8Tzi=_-PSu-``y;E`d_rsjV&Au27_df2>OElV8AQ9l%ah_w|poVCp}z0
zvT!um!@Z8NlX_lCtQW~2rsYr8k1w1cKf$Q=zNAr{s4}`6C%wL)cvGXe@8$T)XY04!
z(2Sg-<|(fFo%-p8GaF}bs_dL67}GyFk@R23#5WUb=bG_rjdL@rz7N(}yj~Qx1pcn9
zEQ4RGbHzg}nnS=@uXt$fT$s6bsL^w8@yXR{RW$(RjsDXse#M)2P+XAn66hrh*q>gm
zRT}+osZ>@}^^K1|ScA$cHsfc2Zr$^9pv6l0O67X3nb(cW26P+cCo8o|-ELo5ni5u6
z5kjz&#pf#XbC>J2GsR0n9sIA>DzmGL#pi(HMZZ!5Wa?JJ#B{qz1E=kofbug}R+d4I
zXBroZx3xdTXC8cV^3lmhc%nBx^~Ot=$QCD`e!NU)i%&ms;lWF9e5!cqt;NzCpM2}4
z%O*0-?z#DEm1<ddxYq31jE=kN^Go!G;cFJ$rY=J#`1NK!JgQVZ@a;0>WTjkVIP$ae
zi;HD;{AvYMHs27!?(r*ESLb;u=$a`H=&(F@ZOWgx9Tcz%259EWVEg%cx!ml1uvV)u
zWy{q!`aX!i-^i0nz2eU|rc~TauP`w6W>>j9vp7|&aSc~jr<N+@`oZm<`Zrf`bI|OW
zS)5-1g&6(dHM>PGRhZE0<$AMkR-{Q-o!M*V4LFt6X&wpGT&*nDnq$mNd4&hm?IC`Q
zfMITG&3++TBcGq~r&jbJX`5VsP>V9F>zFrxSVvuUM#s(MVx`(lEkZ9fv(u}yvyhg}
z9#xKLARfDw4au-(_cirF<G`p4vpDAFtuu?$%VoEgnth9OT|`k-kX>44RH@%>bQS|V
zJ74#E)SX@%Z%5}{S*bS%n4C(zTwPr%udYB_RB9@kqmLE8)vsokfOxrneWk);0KQkt
zZYKScYBR<9U2OKKjrO8RRIAOdCq!dv68*QRVqdFe_xR0}_FS{DIs-8)qKeUNCaNpY
z(d^jerDpe&_6rqRXJ%Aw#jnmy9P~x4l)6`|bls7w{sl_>hm~B>t-9DOqH*h&{RcQM
zK`>8wMoX7<l|QSR?^5!1CC?~%hmxn2TvGBzB^Q-EqU0eZ!dm`}k_w4iceXz0xd4Ix
zR-Kzx@-0dhNlIFLEPGq(zVMk}BN5-Aq1Vb7Y0m<O(M;H&O~n1T>5?N#jw^Yb<aWL_
z-6nQ0AMcN6;^TZf{|fP5J1)fcC-TZA>}(?EYQ=iv?z%!e7cWqzG*8V!LNy0D9!?Yz
zz5F(Mgx=3YJ9Q?jyYlJfD`|tqy!O)xVc-Dl=7S^Gv(>>m)FD_RB`oB{UrMiMy+n{5
zi>>GCiG?H>4GIn1&{a=tq&U;%K}&rgu{(A%v7QgI3+Z>o*OT>30G<TyXAQsPXiM0+
zYa#6=cY_ZWvR4wd_to=WMtHv7T`w&3@a^@ov<&v%=<{-5@$8MbVHX{Daop8$EG!Qm
zY@P28y6I7W&~=k%X_XdSsroq<+jOSK&J65urnh~~p3T?vb)ND1+bw#cDF(-4L3TJs
zi}|sb{c_JedfGvi?~1JpXZCpm^t{l~$3fQ<dOs8tyrG+kP5m6!IM|crZ$hoR{_cSu
zX+7ge(CzMNwQ@VvS{U|5uOwbh_)sVdBi^o|m?z3*6zab_$lXx8dV*c*_nz2#Pu1sW
z*c-EJE_l1ua=oxIrqQ+b$D<=XA@v`!`g^SYdpQTy_4V&_GtQN{`nb2(sJ$MqpS98B
zCGXRi^?CcqLx+nBqD9dqdvC@!_T7x}9G|S`-w)3KO;Hq0v3bX9r}FvA8+^!17{O+0
z3MTv2ukvZ!Hnq5Nd8$^tJar9DZ@D;KDZ+V!3cVuiOR+LF18*22P_9gS20POZ)AqWH
z%QLGIczDgB&BwsF%C+WD?WzxZZ?|dH+-HT$U|fV0tF^7C_X^A1^6YNA*ssih7`L>$
zr+sSa_AMu0)zdczTSy2L?;}u|sWkVu%XH(Y9Q-^FQuO8yKlIVU%~yWmb<N>+MH55J
zELUc4-?8+((RHr^O=A9iB;g!<zQqqb;J=>YM*q^P@zEirSMk70*A}Nk1X#PsnCHRE
zHQ}*pePX{MKYzEHt&=pf(b#8%zpI{qMx{vz48N*M?UgX8L%}?U>oTEWB0CF0so!zl
zzp4vU8b9#+@|+=uW_F3Cvb4J7(80Iq>KB!~Pn9#5oNS|unW~v+qd7FWTz^y;e2Lyx
zJVyc{!l_PG{g-r2a%ypLqSw%#Lqvw@4Cl4G`iO3!n>90Te(iq#|3pbCCz*-_eXb*F
zbPMn8FHrU!d}~jV#Dpop2RZ(F!38<+g)qoiqCe3O#_3Ovf^+n@KRL|#{=|N2j`Q80
z5KbwAPmb^{#1C<8fh!uLhDTZ?Kt$<mlD`O{2))(<61FO<fHb}Vv#^ldNZkaG!X~H$
zRIre=BQ`M1(S0Bbu~zYq!Suyrjl36oJIvrS#@^A+HM=UisU3IaweZS>yYd}e0YFGj
zHg?}zd<rR#1XSANBB(>^y|D*s+h}#23MrT9<V4O97SHks5W0+d#eb*r`4&xED`jVv
zSD;0g{U>xL3;tKViG)ot%hO2@w@gv9OVmM)ODmt|vf3y~%+R$Eb#FWY3U|o6(QAX<
zs@{_3jBthcX2aSilET`^aZCpQ&P;j9dLl?|B*m#-$@s?tN_NNA;Vz@n_4L|3LHa`M
z*>mfewUa@D+@bYskl9ED*_(0yp*oV3jZBac_A$ON^ZMA;hn|UDz2_NCTo$OiQO}8Q
zem++3+JMvE=oW9cQDAauO&s!i?&ZW<rrxvtyqAS?$i0-j5PLcP>~Fsd`T=U8D@a|*
z`k%S_NvMRb-LZvUaDSe(PzU?ph?9aRp%lJS?}JLPBd_a93f^)rm=elEbO+ck%z=I4
zwY{nf@88XJpA7OF{Wrl+uKo4iu#LQG`t`L_^?{%Vs-t^1y=Oee*ZYtrfaO&(8WT1O
z`tGC0SD$=_6ggqkhcu({U2AXOa{Yz%zP0<pvV7R)$@TuuHsQRxJ<g5BX4QAv5Bg^l
zc8?#{J+6NA8D@3^nn1O{kOOPywzRmP76;nxr0J)ZzZ-#ywerVyy(kHH{b1`UqaKZK
zzMftT2K`){u$dl&p5yNW3H5J%aP0`sJh;%?nhjy!Zx04HhP?v48|13XtdE{xaG%E6
zM*6<k)lX8Ed^u73NmkAXYZ2+)ONlCynuXE!F{@x`eVBFEX8@p|I}MqXWq{nU<pw_n
z4G8|XGhR1ypY(HFF%%4af6Q%rS2sL&gmc5eaCFW9!pLMZIkU0~W+@<9hXF#&EWR`k
z+llC0oJ^y5p|V&p#vfjw_^s=Xjc->tGb3cG4?E8!t55M;6QelC9(cCEv&7dELHw^J
zuBM*qfY8XV6KBpfQ{s1<@$we9@F)5E#8sm0S-Jl|s6pY+OssL-fWT&s@63Jb`C{#U
zGSM}SLl@_rykZ%3QV8P&@~V~RqC+_Rlvk<ENT^$%TJ%56E&Lah=#l(&5@FqplmU!H
zZwow{B%_gW^72Lx+A0(>NH`>uY4jK2#@1FUGhn@<1SyTa$HVh);ppCcu`!f?x?LqK
zoY)m|p5p5KGWTw#bhS-}U$03v?;vYN(&VYSW4qq1id_iu{rd7U9DW*JAFV;ZM%8l)
z5izT9<*p=T(tQC;qY!*i16;NU>TafCQs<C$2#q(>($&<OU8~jF)zwM`K|-H#ob9<c
z956BB0DP*&nfec^$wbZbe_WaPm1bi8N;6jt^)vnt+Z_#al)9LnvI)=8$W)%|%vyz}
zRdCt(d^5gK%2&$>7+`!Cy;hzpAyC{XH}mcB-?FIu%iQ^Md}|^MV(^0fAQ`ws`|AbW
zh&FIsr09f1d=N|z3W7=)j`t?^a}-h!@%^G9_&Y}{!{O05bi}CIgwGu3{1LwTYYgqM
zYG<@x^vUo|@iqMQ5E-~m(?kUEg~W|y01pD+2pK?W5>AS}q;fEr>ZMx8X~&z+CPDRy
zsr{^BQ_vLoJW{En*MXpCdqj$@;?)yt<$5tFu9eSCoM~0P*T$McrrEFGYnQ!qvba`!
znD3c0XPzxOa#2D<lSD06Y6$(RE2~HVx3&^qj~HdT;_Hf*?xtvRxZ(*T`ir+uoGHGk
z;#W=;&!Lt?O^MJGofD$EjwbyID&gwA#RrS^Rp?5trRJ$(M5CW7c2evUr;5;*x@Gtz
z$7@AfX6*(;Li{Ulm`MBM47>k+C8C-gk@-BiiA+engsn8=b<|nhw}~Ra@eHLx=Y6-V
zxMttgvhfE|Z43!wHEq6FR;vdZa@f|!mTE2HMhhd&*A~YKGTRr2l}LE)mP>LtHYRE>
zJ~@$c&-DQ=bI<iba+DtBr<r8-{a;irZS%Ob=VPJ#*ewL7N41Lk^a$!#EHM~w?6dCQ
zs_tLvbF}B@1~(6%!~YC<dk$lxwm!x94n1p>0o3Jdhg;8);yDi7>N&#N-|&+#{##~*
z4Qz_U_KbWpPtuwX<kxOC!k*pMdXDY$v9)gNDa8J@)|I%Ua9y#EBA^pKMO2b7K}`g7
z!qWA$)e`8j>((<)JZ3<M=gqhk`0bh_+gakq9KZG<auZowi~cXDv^9~X-p$N)yZ+YM
z?Y`QUB;T%K5VNIaIcgkTqd(kRt@aQ-N)Xx_hAs|T!!XMMR68=iAc4$I+BLX{D+ym3
z9<~=-47uJ*$@R3+YuFKS?yexU&g1(RgY*XSKQHa1Pr&80u0WTTd{=Tkvz}egNp}_G
zY-qBRCKZ4=(hl^0SxJX*;444d22RN>qXV=O>*e_H&yvFoU=bB~+5c4zCc512j<30v
z^|X<|et8us{aP7uiD*10#20X~@okQ_21%0emgr;akoT=@259$c*rg<h86q77S*G|a
zBt|39w&j8ujki7pzOAu0<~Y8<*e}Fq<9`z_<lXW0w8pJ9_DN)}8p|A|U3KsxwUMHN
z3G+z#(2rk1?s^m8w~!9f=;=~of*0Ulyo988KaluH3@!@($x8=a!?BI*O-h4&xH&VD
zmi-!CYL8o+7G&AHz^QOjKK};zVO!0n@Ba;5w1=dbHVKNczGe<!#90zzaH8PW`R4YU
z1!r4RHn7@^Py0WjrmWFlS9^nQjy6q)mP)fwU9OrX&C*ou%GMdmt(2Wf#KrcG7=<qB
zO`MD*-%*Vg4I+%By<lkOhsZGMtp0UI*^4`F83{L*5v>0N>peXehgYuFQ}uKmSva+`
zHxN>o9y*H@9B%g6Nt-()+(^gaXJ_Ng7QFaE-b=rvz16L~L0g*P!YK3&eCY@6DC;3r
zXKy_O58e}`B@z+89l&Qx_CAZXSda$A<t9Tprj1!6Z+0{y#4%U$Kh9J5e}{y;4e5K8
zcl+OF6XBDIvFzkYP8#{^RgE@o=8#xIH&;CWr_@kY$>)_Uk)ZNguFg4~ifLC~rjSK(
zR28%`n!}D-k>yVXiz`pINHy1wf3NdNnK<=qKKp51eWaBfSG}HSs@RmvD{s||j!`aI
zmc%2O_#q@z`S@P)DM}iH4vmE?z%F{VpZcwt){blKx<edjrkQ8Wwq`|eE(JJG&O%wT
z<7Anaz&_0XB6Zm!ZMD*uRFHPyCO<ueP|4JdR<WbnmIunXwfSy6iB_G2JsHu+SVEzb
zac5ibIYkBaHp#bkFA3--Y4$Aa>gRJe5>UokeeC}ip^k4PUt~{P3B7kE0SMX9Z*p>m
z23j9AImFj+27-n7iwG4Q-IQ>wSY9eS^nzGw*|abSzrY1<ifndr_W^k^n!A6N$N4|h
zbVx^B4JV+JHFY%^v0gQ@h2p$HSwzv;P1<-t0~p=f-8E{xqCph<5JSBM4dc{W(b;W-
z_%Ib~CrAuuvzbGR5zKBSOnO)X#j=Hl+?irAcxK4Uy1Qkn({tzouPvyPnnxoS^ES(N
zqSe1mOVT0X1p^2E5mB{cW(33HO#0J^>=KBrQq}vlY`p|PqD1+HdU7GPkVZ&HeWC8F
z)2?K}N=z{a!-PsD>tS4i7z|OwL}n7Bp4<I9=o(@OI1ZS7G??D18BH_}tX8j7m!GS`
zSHR&&+fh9Jtu3wb`f>KV>1C5s(>4E}(}=Aae_74tBLljr#b$0v=6<x{e=nz^9g`Aa
z+RVUYF&K8afC`ni3}Q}Wr4uraHL*&J{a#?Vi)&56p^heUiE!$px~-v&aI6-4tw;;o
z_m6h|i{dJiUJ`f=3_vu^?qL|y^DziI_LY}1Xq-d~D&cj~OLP3>i!qL6(~|Tu94~TZ
zik2V+Z>r`X5+@T*N7g|MdyEMI;>&vNrQ|{u3pVW6=90+3kdSgV=l`(RS50{R>s`Qp
z^rBol@UcXY6Skg5!-o9T^3(-$BPj)s1d<o#97YGjN7oH<U3Ykm>LK|w5*9ivY<Wm6
zN3A-%GBNz;gcAPmY34iks7;_Ys{fxz-p8Q(|Cvn0Vc29=YCAxNkkV`Kd>I58S?DxG
z15ZS@2cZE;u`y^WqwBEuEwvFME4Lb7(A6JSau*3}JL%0YA?uQCznR3UrBN_Tw$Rd|
z(dCZNIW^Ljo#=6(>i5+~o96rj9cn52|4oTjelr(o44N5db795=rsy#PgNOjy4BAAE
ziuSqg|0^}(Ofn>O^#7|`{$r9-UephJWp;i|CNJ1*IKz-;&JO79rondrQnE~pZ))GB
zJtK%G0ND>ws$G)_zx|0pBs#r`F({)!^gX?a;b<R4*S8co0#O3AvW2K=IHhASx(9&6
z`iDk>zCjuTeU#cr3q)}Q{R!OzZu$Xs*GGh0lA9<N*dL=ivO+X1MV;seM=5+kIh4E+
z&@lQGovn|-PUgyU`gPRm7MPJmLdaR~2lV(U8i^$j#Tx|!z2T`AJ)qNq4u%}=CnK%}
ziVulXkX%4>VI0tYQGQlxa6)*OvsfTvw~%@%zLDXHDd7pzlWw^s&P+B2r;C@1&lP8i
zr|&OL^TPqluxF`)h4A&V6Js(4B9k3Gt&TJksEuq`2Bb2YpIco<EgZ@`|1A|F0Y=(r
z4*#chwIQM{qtGR2Xo9I9(l`hd#*Ao^h!1S(^hvEso90@hVk6=xB;yI0abO~9y8ucT
zskUGx)*zxrC2UFfIFLBon-IqP-72uIX({{Y?n2_hdTL9x8=(}403l0d<;T=bht7L}
z@FDlwQK1FayJ;>Lb@7X2*6J*`7)1AiEj)}DQGJMTmrl_gY5GuNp;^rNcY3g-C}iR&
z!C0bF$e5bvgpusvQZceY8ewC4PAqH!_SP0NfHsQ~<08th90Yj`oh5X{%>=`d17vH-
zcVU$5{&Fv+O|uEO=S(gv&7<s+d1Y*&%l|wr<Zp@<K&|8z=%E00*6RuKTWaSR<lf0<
ze8vA$9=90+Hi0<6T`vQ732W>ewg+?NX^eGcmM0&DEJoKF+GdVSWo3EhG8Tfkg<$)p
z;)}*VKn3`v%%Wcgkprk|H0p?cTbR%3ewmre%NT(BJ={r<XU>!({&%bDu%>lNwerg=
zSV&-UhrGRDa#V~2F>9|)bUW5S3!>TOXytitt=ZGMpv<9L%6b@kNdMDny}feqG1|6#
zdLp=MkMsAbQPYYQFYd(IDHy*oQ)VORM)}=q+SJiR$S7>Oyaf}q2D|ubJJvo!!ZZOF
zJl21>?%gpof*rHSHn3Cv&NpjXJ8OO@Iu2`oC~mYn^u+VI7g?@1;vb5=0Kbp!R3&jM
ziA%x#w%C)gcW^b7eabA}&FVc|PzkJpb-bzU#n=vI@y%tX_3Cmk*m%H}@+!eo3dj%X
zkwu%&bMt2h*F_q!k9c^Ph`(+sEI?Swj3Kp)+4g^%pJozW4JXi$fT4uKRHJ;(QFGeq
zD6~{;3ax*sdD0zGXq_;q3^ty}R1xKyO=Jgg(7NcG>#8qlt;3M6r@j;e+k1L@yQjYB
z#9wW9@giHx?^iPeN(PnS`zB^k+rhudgZMbTMZ$6i_QrMUM>tg)Y>!EG*$vAMWv&AZ
z=K)=KP>BJ^zAZ!Z&p8uX3CL(e%9?(l`Ce1T3`a7jEf=-MPAghp!{ei5SYbSp*00u@
z=u2B8!1H9=<iY8=l5P(J(_%gVY};NOWb$l30fEE}V}tW=)LG$@7*t1NmreEdS}n|@
zn)*J>w|0sI94y;`1RO29FXVQ?#FhgSOX`I@#$n#nWMi}%;aP`rs}AWl#-@v5R^=x+
znrNZpRMlU8j;A$Xr?K(1^&=#kbqsl}*;|!YR;`Q)Qke^yh0qY$+`HW3(SC4~HdGc`
zrCKPdmRj~VH5PfQ2nmaK0mF^ii0n}LRbe+HUS|v;a@5CHJrnM$RcTD_5e8rw9Gkqv
zdKR%@7VM=Rt3gV194C!v$(of7rX09PBxEE3#>)aa68FXG*&CR0!kj?(*qN@iHB{)}
zDZ`Bm;3i?Gd?>2=h#hr#sN`UNsxs9?Se2iHN%0^I5ON`y#LPk%nEP$?LL%j<(<f6w
znQ|^*l1Up*3|4@x83-^OK-5(=%sLogut?9qw}A{h&ag;_Jj%Xv%&_M$lP@+(Ah>6}
zw*zSQa8<WV1u>TBwKfVIrD<cKKC&@d=Xp1F$s7jJTEdN%h&;}}<3%(_2BI-bcplN6
z0P<5Yl}_|Iu-VKwU7vx^M&@=~Xy9N;wOB>Jz)xGI-Rx;8{InWMxhEEvO)GzijsB{t
zeo@I$X0zEHEhs9s_N#!Tc7%at?N<ZKy=}1UKyrKih446pat=Ya)}=wVG;Kz}_pq%$
zp>s2r?_|naaCc(N_tnW3b=ORyv~=_;F2o_JE`@!Eq|k(d>bR2$d{ntFC^2r)7%GVj
z{S#{7rgC3aQYXO)Lo&k}1_ZY|4f;f12lWu2L|>=Dv@4q3p`1~s0uCw}y(yj0mbvi1
zsyk??{ZmRbD$JXaFa~Lj({khR+f=CCNd-ygq$TeMyylV_NQizwt#sT1yuI*D8EoVB
zB8?PbQH-agg6i8cs;$}7faq)RP0~CzL2NzlBzVq3OZ4?Z3M(!BB4x%lI8o0mWEXN0
z?MNOOLOYX5iX@uoYWA1`zwFh0#H?mK;rP`b;2vhMZg%I6t4JHo)l5I=O)WhWiQLps
zGh-xMt(l}odqp{jHtcNjae67YXtr8TOUdyS8YYH^L22x^$KR<w5-0z>x+Gh7{~jf;
zRidFV7UB}j`$d-U?@~Y(u`6CX(&J8O#y(8xrOAoI{*z>y`Lg^%a5KeOq);wjU7cEV
zt|MkmZ>EpFC>poC9G?mPmDh?Jq02h?FO+;%$*(B+mrA5ZYNl|_T=aiK$Ddd7B_*~<
zZT0?v4vqg$s%u##wnL00*sH@sN)9WrReoN%dq|o+p})&a1@D>;!yi`T+tjR9qkoR%
zEMGI5%%u{TVSICc{jWn{c6TCP=*5ok{#+rK$>sQV^%j_~9AC-{nS45z$@k?C*x7V$
ztiLyx%4Ioo)Os>QPg);gB7JKAfTwA@XyRP?c7vU`XX0v^5_HC`FkK0qEA1QtFT~a$
z^-`DjjFmXY%)Gf};#grWnR2vzhq>s-akVrDeOxWgDZkX`obgNj&IP|T;9TzsOz<t<
zG4D7ol96j=$=W~RO*{_<D-DN^kMHzOdbgS5<B0bz=X_Zj4bRE_5{QqxrE|a>Gxtk5
zTRt1QUyg_FmwQ6@%e~&a%>h%6mgUNR@4i>!-s`>lpHH_OE#G4fn0jKu2bgQ-l(}N6
z<X~7b?OpKRXni^qmdtpMcyF?j!(qvz-kYs&N5W$d_tnR&#?i2(g1hM3tmK&Yr1unK
ze>^;M$$Q$aJz*}M6JWF1X5VALifX;(_teNQ$p_jzu_k>1zl)c&==pr*4KudO$qw=r
zMWQuOdy)CaO(0g5$^jvexl!WeK#3G83C#4NFs{UFC+jJMpK^K3`S%AI|6VVNkTV6f
zkuXu}69R;_Cp^&~JsV<f<TTQ2c~%qoKFfDmzTfisIn^1cXKuhTW7LRhb3PvQd{5#d
z5|Sde#q|@~1LqP%QRV#gv3fV6D%?qAvgjdIr?l^xSQUC+e;zIv<-Tu;RTHpX`P%2r
z{j<L+(s;n@sSo0UAs2^)v3SEs*>#uL#t6b*<q-6iM{!NT@zb17$1J})=ndUB-$UJT
z%smIxf+xq+-9aA|2jvTUFiw3;BUIn(;be+Y<339Fo14smpr2aLbEaQs!rBcRPaHy>
z1E-Pf1CvIQIIpRLOpkwvq&bY2PYnkt;IS9Ey_Nknoo|p_^uJZv2b5e_!jz(4tzl0{
z>;QcCm;FosfAM}i*xZGfY^H+i8-TU-z~Dub<2DC3xwBb?=t-xq!*_bl7BIQ8yy8Em
zv3Oi#(M?0xg_;v$Q)n>qN;F<a?)DZLA}WOtaje8$jASY2d0av41+2p_mzN0RQ0h`I
zJoG9x|3Rvlw?Ugq+7i5>?bX;E?TqTs?nZO4otKnQa!*stmU^eBm$7kLzEY`5B-0#{
zMLITZNZVS@f@Tru7`-R-#KTHV)GbqYf1YGw(5SD{V5f^^_)wBiMi1_a{D<{SZ&LDz
z66xUmw<xjIWlwxThYu-f^OCIv*<OzkIuxPiaH8g8STu(SfLIUM37SJDijwwVa}oZH
zQmVfMk{{JUF$`HImdnZ^)Wm!#U!aa_MQt?3wt&R8ha8=?4>ylx`5FbMUYM;tpw6V<
z34s+#WF-IqEMqLRfq~FQ#_R&@EE_m`6DJ2zd+UyIO9@~sWIvF&&+*<Vw~=8c9_0r9
zo0Pm)i8_JOO{9!z;hj=mWVO6@nTUB0t5pPUa^uCNx3)5c$W9XiNp)1!lGNBr*}1?S
zZX(0L9g`Yk<2y0zEXEs$wvLPWpl-TkMs%${8XaAl<Xig$3D1Td0x3WjMo~FN!gA3s
zBUV#6+5>DL?nQ5gk09VM<C!<V7}JfMF}YoU{1i%1P|QL%Y^1bxg&;*vGB~r9Fh&#b
z{~n!_H-7?s7q<Qz12^$V6ZoZFY4uW(EqHVv1EW|bX`JWy^7+ae8V^8IU&eY<F+MQ7
zEzYds*RdqOZeV0&^$hSh-7Y~x!%m;7R*>8;UT^F<eE4wj>8Iqy{nVukr}622_`%0?
zM7hC$#=gk8T=7EAbOK+QY_kKkWpIFYKJiYF>*TJFal*loNxEwW(hhdD*fqb!q%{X!
zz!qGGTh7wWF-a;q3R*tV3i?sU>CL%A;T%QW6`Y-uJZfaNCULGWK^Kunh+!h@Xt)na
za>ax-2i)!awvAdd-q^OQ@qpeeH!>jgl$M|v-dO5{ZE>_$Y+G&fF)<lOwl2w?8Zl<W
zqF8*5uoN>|vHIv&Vi#MvqJLcc3D%t?P3HA_(+TdiZbejB)CvC_>nx=>3buX{;qFbl
zX;`2g+k(xJ+GPS5;G4f`-}rB1_@eQA$SNzjpyZ88E-HD1WZS?Qi{ojW)azJs6iyT=
z+rh0Pb8Gi3?O8qT!+dLE$0B)KkdQo^CzJq*{aQ|~5yxU7v5~$hB};xCVlPORGgpsE
z>;gHNn7qhCNJ`O6DLzhfpmRx?8JC5i@&9kC_{NsLnOr1g%P=M&Nr!oUR#gn>TXI~7
zk1|S2del5ePJouvGC02VVLOpzM;CS$G5!#3@mR51L##C*saVIXCZw}qz{|<28=R4u
z;|0zz<+8FsC7*V|K8(_%jtuQ2rexT%Hf%|gn<dL#YY%WuQaYY+*l>R?cVAtA?mTAq
zOx8YFFF+gXF3H+Qm2bDE9)v)tIcRiT$0L=?jiXPS+l3WkEl#KejJ|8jMZ8$G$XwAG
z{{aTC(RC^EX>06J1p-lQA_~JmNEcCh8{#%{V45rar!)b^m<r+9B>Pi3aYBciC)6mY
zNT@KQN=8iVG^Hp@Vs%Yvk!B<xcD@IAsUvBTgj5q;U`@;SuyO2bO|;uldVZgvb#3cb
z?F;Q{Ti=Uh=Fm&yHHB2!*>*80MKDLoNxy^ihE_Y+MB3h$jIh5th}RS~nBGi1vw^*(
zunyCU7aaoCEV(RLO1UnWfDB?!@<En8fkkE9tnkn_42Fd1a$SN7w7QMTHzTa}kMMXC
zdFQ?IG7r=o^sr|2Al2I-y$$dkAu4)7PK#4BSA6YuC*KZ@^-Lc2aY{@3qAn6*LrwZo
zPfaajGP6>m;rYijtH+gW-lNTNlVf*CF&&vH+4qMu&!R13z;Bsq$<xGnoj`V-+&Z_n
zZm6}7^&ly}HnkVXXj1ENpg6MQTMBWNs2;O2w7HbB+bEGsS-8L0#R=_VGs?*Z*nlx*
zc!=)VhP32F;jEt;Bgb3zCKjE;`SudLEw31S3^6{`zOa^)6lvuB?@{u-B!q>-hRTpN
zw4LZN|64SSrR`(sU!v^0HE70n+fV`A6T`{IiLHZmEBxIWDw=6S`D+{q<QXeI=_HS0
z-0|`>NY7y5@ktGAiN)k1okYbAlb!G%V_*Sq>+!3^0Ki7d&Q_1OHq_>gMD0EG8~{w}
zZpkkxF~A6TGaNxoJBt{{3KmPy%;IkkRpA?mfriCnMf{Y1MI$n=#K!WNhDzj3Gt0y3
zf{31t&um?UM$VWx6gEe2AqwTOq0t<q5t)yT88MN^P08#AcO2858rh^FmT=LQ-`mN^
zfD)XDfIuTdv@&eKHh?0)8Cip2O=9QK18-2`N`W?EP5ClmF%Ol2xq*wKf%ySrhBXPp
z5PIdwArFx9C;CY{VQvPCDDzmmHeY$p*M`8Ok!XdX+fMv)Gya@kq}WkpNunBV9kCoS
zfi5dX;g4&CjFgBM<3$OwxA8+!)z&zudTSis$AO+xhz;=yRMJ>%4k*eKQ{u*fl_NW9
z#vw=&AHqF1uv7*YFJVU~lMRgpW0@Kwu2)BSM&*1NnIL21L`)EBDAjci>gF2%IA07u
zp)c3vdyAJXMAOz2&)tg)qgTJ&E{}YyTyR97G;!X5U)tBA#7#R;oL=?;kEdE!6`vzG
z;$^%btL@6K$CGQS9#%ESNa6?B=_04jyhbNoYcq-u;I|rcLYxmXl;(gbMUf`FnY&V{
ztYE-ab8KL^3^HxWfv*$(b4pGtsgXn`*>x4Hl1TCVoHUq<SjGPEs|517rV|A{XL+$Q
zTPIcxLRKBQT9Q?ix(NOtk68TYwt#&Uq3}9oHz+^Zf?3%j=zEx*2LHHsE6nP=q^020
zS`z1IduoFWYsD}%JYS}+fC0`HQ_4-_vGIb&j0f(H^`kXKtdb<V4(^d-_R=utF^R=~
zKC<H69rItYRyj+nS!>m7G`)1#Dl19Wn?Y8sZfge&uZS(WE4wp0S?aT(R7`ysry?qt
zS_=zoK1=!fe5E|?6G_r94Ye*&p>xfR3c0LZ2t|Ee6K4oKrK?JVwUtF;KbZ<LawoNU
z*VZ@0zirB!iN~(<T{DxqW|zAkd~2!uvGC>+kZqaN9$}SM)qkQ{9petpuBH$_rm#Xr
zvK25Nggo1~b%r-xZ)4D!X8~c&was&Hl6l5U<y|S~BZLu^gu=`*5Ckq`cWfL)jTTHX
z0f3xgIlp0fw|tTy(0>=382beuQBV8{wgD+31`wNQAvGHd3Lhb=j2)$Ci4>^2DrSLH
z-3X2g5*#CPMq1<U@9riAH8rz0(~>|eNOf(I0B*!L@VWt#=MW30@SE=okFD?RYlmaA
zT7P{nC#vJtQ>n85$n{mZuhUzjQ(2j;<*--eh)?OBPdh<pz%wq#89npB)~5{X>4~#Z
zO`dsR+cUfSgspu8&#w9i*x6~@$6#Aqt+9yXKc?AA5i#J;-25r`4UR>6JQ@pfo5n(C
z+*r6Xw;Bu8i^jrg?l=~zWh*XF?(44u)#_5+9mZl?ThUllQ}u30P*fenZRneoS$?kZ
zffhE1ypit(S<HzLRX{roSZb$=GpCA|Phn8nuG>OhR+*T%D;C<&U4o}@coi^+V0h-M
z^3HS5p2A1JeSG_~?e^M~2d0JH9krS{Z+_85N+9U5hvhEOise}CRhr#6+2IQ9JYj^z
zLUY)HGZ^qGaVTq<Tr=}@brwJNX6hoT|6T#R3Dq!>mB~B}xQTiW#n8^2RCdk&`$?LK
z)!Amk<2zj=%-oFM!gT((s!l*s>ef9%Xo-DxcuN)f>uRGX!uIw|r9wyb^IHL9pOM>Y
z&$aV8fyM{uUo73598)YE&|Yt%m^cFRJCcxxvn;Gr@xAdgiBqJfb|^cUOeGozb^@W+
z6?dnZF5QKXq%187>)+!6+Hxvz8eFJUPW~YbQBg}dL!O`u><P{U5<#0o*$a*#L&Z3U
z*kNY9EQ1`Z3_^94yc~!+TDy;ZDy7{5$9Xtq8S`{sN%}uVxuVH@5WEZKTrXVxd$yx9
z{&&G>an$M)CN`~m7pmM@vCb9>{-=VJ|NHfxpiu9HZ}_a``rtB}^r~NmK<hnM|Ip>5
z>s3Yz1)_=AGXg(R7q*^~IIoI@>U&%@DDDKqpx*Vq`Vi$j8NKXhj}b)-rCciLn@`f(
z2~>`>V13%*TE%Da(yw4(1QzI-L)}R1k~mRLVjDzXqd$Ye04l()U{G`}>bV?eUvXtJ
zgrX!v$&XPo>WrI$0d}~2EeQxTg|+N*wM3*#2ueag;vrX-&Ze0p47j3^nl82sPgGfi
ztXFgkd{oP8?&ms`;YYQ`|2hfLGah4EI1_}Hz}J6D=fz9<Pb)E5jOk$>(xGW-#Jo6(
zk1<@fLCu&){x%lzurwm0%~b-Tu<0RIiBT~>OJMR5%osWtEDM^7Im0a1DR?<mJzp}4
ziJ`Xnj+yv2$3vr?4xvTa!42%}oET!p$)OdzQ<@F5YAu(-R=`Z1@uh<<;jF0G2xTXD
zZnM=ylLZXwL0ThKO3kRTq@00MYxn|hm#a#2o`EiLYNn)Q6_P^9czBe&)JIM}l}U^x
z<WVt<oNCx|1z7@7cQBb^&nvIGy26>qDA(A%(@r;%ZboFBKKj+yXr%X$!T0kb{CDgo
zX6;Zr$La(-;213uv;a1cl0c(RvUtd-!!hr(Sd5B|gmv+a#739So8BW&<Uz&Mu-Y3?
zH|F-2oqSa{G*T^MGBLWNXeT#@Im-DEJECChNsAkugU#`OlT*&zO9I4ZE)sb)2L(MP
zTvFU~X1MNOH<45`&s>JC4kjRx@<H_?QKDodn3UrwAk*R*sj(`W)+4%;B_<};+IyX5
zu2;QdMhe(!A#EZ?$#x~ycFEcQtHbYqOb_$|{r)D0-<P|~ma+H$XO(RzVOX$*Qtm+|
zKS<I!y8ho%=?;V1+`E0C%8JHgrc{r#Mw|8CG1{lLuJNzc$j+iFiN09SCQxX5rfNl*
zfGCs!If7uzIuP7@n|cOzW*uyw`|XObmPkVaW9@{u@HJ-wfW6pIbAJJ)DBJ;LH>glS
zXdEMmjasMsOBmImn_dtlz3NrgI!qG{9Q`qkZ(Y2y9D7=w7IK%v7NXHTTZeHcZy4!B
z)a$qMt$mq9qh_jTG0k$%f}jV@OMd|~B6ROqSgV&Dht)Dh_*OMxlkrI*)T^->VJ%a5
z%lV1WaDWzbUR;3Wfb{q*WHZ0D4=WTW_vkjJyC6#$GjHsn1|i)7P;%5K^PTBK(bh#z
z`q85dC@!Gc5Q}gDopa6wkVwuFV*`h>v^kVvQK&8BXw|#Ov3W#7ZZpLN3sqIoWU~K2
zIl;CLS@{=qn9}NMqsd8?nL5O<;d^xUaV5X66K2c$Z<R|@hEVgU^1r4;lT7F);uX8;
zPwS$L3e#AUOo77d<@(gJSqkIuwN;F1_Jk#^g;83I#gD~LLQDz?34tgc;5g>=Th~Tt
z{m}|Ap*2Ap4vEZ?K^~5~I7C*Vddl+Hh}ZT8NrD-eDpO)Nkjp7AfpwNb6A;*AK0K!x
zz?Ba-K7=ulHFc1R|L!1FdpAlCJg5jsfJnr&RR}E)ayNg!k^@S3aCt{;p)gwG%xEa0
zQKSa_sxJ9Ol4eRMzL^iPyP&?jH~fFC3Y)ilvzw@&QMmTY8abzK$R`rWDippdM?gzZ
zd1CvX>bP<f251R3PjEk;83-UF+$cUwfaSZD;)dW#_eh6e_NwFR;Rlr1D*GV02-to`
zrFyzjxA0kc4e_GCO{z8~pHD+}tq?Y@;&nZn;8P6Lwp|-=tIJv-R+1M1n1}h+N+iY!
zBT0~#g-PuJ7szdS)M1@N3T8IPZC5w^{uG%6HDyXuZxY%RvNU97iUqjQ<zen`Ly?_&
z9n;lfP#+1Sa1r&LfVyvaj7VDkk_FdQY%Yrs<>IpAagLDk^t3ElPAMRk-oD}yWUJNG
z^nAT`8f6Oa6<TUXw6gg7*67Ss^*GX3UZ50uq|pva&-uE0n}EkVeX5?KY*~D-Q03EZ
zC}O)Y3j4Z2z??fk$F&g~K2<b77;)|IoP1W#i(-70x5b=5P^<?aMwh~T@fcxEA`MET
z)NadxH37H~tj&?N4=j0Fv1@KGWij7p><GZG<MYw+`Kamx047_62Khe9)eaCEqiyET
zBPttEQdc62_(+QQ3o3Y3iLKb|RiKw5=KKgnr9O`2I}pY8D&?*$vi_Ldt(!CNip$e0
z-n!PWFjg#=kM9MNp~WBH3d!2lXS81R*~3>GlFMVYKL)R}_5fCJS1+>(vu5fG(6X2q
zlIY5D4QCj|Vi9rz5#~6SE~X2we%TKt{Wp1O0A0qMQFyBdpzmGy^cSjFbg=Go94`i0
zj*E8O#qkGeDNV~na-pR{m0fTn7OD`t6j{^jekr+;b8n@Y6Tgj14BqzLs2lq}hC!?h
zfJ2XaIO>2%tWe<a!_-6uRPWj7mCFhgdM`51KF-KUbiCC@KW+4Nv_acM4X34r{*F@k
zU{?z?uQsx9sA1pW@G-b1+CV?$yz>KTVE-`IRiN<x)?J3VOMlo)T|cBccdM=FSw_^4
zR*k%TfTK~RzZ=oGwq|BDoEi01Gn3*iBb_s&cne9mX<~~rPpM$g+r_*MwXPovuOHU+
zyhkM+4BPd*ONE&o5i!Q~qjtTHc&`uFE0TFO7{VQB`@G1WB^!*=&h~SPQJxKU?RZ^#
zE_O2pV;!R&jm<da<2#h^p?r5oxn1jO?uxlHHs?`0d$*6Vq1Sy*Cpg1KXnnLRBp0%s
zB^}(>U@l)TOv6jG;KDR))5b@^06Ls9!>BLYh%9FfA`fN%J|(YH@_Le61ZCI~fuY+)
zV3?a);#C`>zq!E+YC?iU=T`6ZUfW|e{N;|uPer2vSuen60m}YzZ}E?!TQqX+sn^am
z2BMs^oPa=V47(zWkku{?LsjtB3!;zqs)I(5{<w&v|E@!^0{$N;k?gb`sKWm<9cpMh
zC|^Sizp1iC5@Nz)%eN(dg}KN~>~R#XA%(xId)e6hh;l!w<cF15I2@611l5IuUkoKX
z=KYV0kwvm_@huRKA-Zws3(ryGi4rQy*yElrh4H^ak+Kz*`Z^meubm4k-%%Rfbi%@F
zh{Wc#lO4Tqp~hMRSsL4Pa_wCE3}_WkX4)UJ%;pIw^@nF$^$p{Jq-|`DYb?bRd<T<U
zdlyGCIuNSsMJCZ}1gVHpe8})0+l$l~GXx|@W5n%~w3uid$cu>3jufe%bGpuc6|t52
z`a{-if5hzArlyZ}Ed9B}y=M&Fl8Ge850kHbg(Q%RCd?wleI6nfnJ(HR94U;Adx$e=
z?Xk!`W$%m15dlr^9%$BJ1&ARBE6^RJ{mGykwz%75s8-*k#pu_l*A4Y3IjbNu26SN1
zM$pZ%*azXwl>ag~5vwU^My5W<AS(s-qHF6xe#@2Xy`;U8<5J#*mPC5Jti{3WUmpNt
zr2Wq!)g2&GbwS#bAnX5jc-&*hc1|$|4R+>all5a-rUFf!yW0TP9EE(oX86IpZ`;gv
z*d&!pTITb!I1r#tuQ*>a+jO7fq;KL16T<u|hmk0>#Ur{Mq)C!_zR56K;xSos4@9D~
z_Ir<r(Bx#D=bC+XOR4dZW8$fym)C^ZvI~z=N{It?;bjyP0;<z-Pd}ntW?@c=>QM84
zU0r=q$-hyunYA(}xX)-MpW>vXYFewktW_D-YhAOJS;vz9$U1k_BtabQpwrkA3-L>8
zP@kC|_e+og=izKiRMu`{PxulfAq<MhlCZ@9&&86JVXlRhWMLLqzFm+p8UL|5OpMuf
z^`lrG;EKWZq3Yfs2glkj8Ls9zO1ksBOip*|<K=5fg-2W;AqGeCO3MGPI<J^B2TYad
z1vewa2C<M&5~>nzF(330!Xaz^MR<>S#0?3<EsoNc-(hjeGiVSpt6(7)M-+xB7+UCI
z*>}AZw{U(^9OQ0MbJ%L^*h;vp^OMBA++k7qKc%HEH$DG<Rq{zC|B$57|CD>7TG2yI
z$eTURgiD__>f(vC#VQoMz^sH72>uw_p7c7l!uW{aph9!-LIv=S`4L);XLzHO><UD-
zJDJO}a(|?HS=|$$?wjnSsq%L#F%%>rCfbb6qLlVF@&BD_Zhow0cfGO%7d6Gc;LW5C
zR7dw{J+gUdw<Amv8pSk~;o|=fYW7na9wS@~iT&>?+p-rvPucHjF9>PIQlqSS=a#8h
zWZD%5-+|dVcF9Th#1Y{9-W_3d80t@qmp&RP?E)GnW|J74{+a=2Bh?A_4>@XY#Rz|o
zycny7rV8tWFT$`B%Z4=%37Oy(d<mIOqzKbbgaG-V<$Q|>$Bsb;XhAH8P^I;*4`C8J
z{37ZC<aVE`XI_lsY=K_s1%)Mno$@-ul2lLFd1OZZCyd<!Xh%Uys(H2XqP=$ELvfU5
zYVR{BA7Df<jzw+e|8Bjjp09Ufm@#0XGPAG%y!N45=qz0ysTcGzr+SaEcD+Ftf`|cj
zBg)9N-eAOD5|LUTCAwX2Fo+CNh$YPtuYm$}iQt#N+%<j&76!fi71VXL$@-A7Xv4uS
z*sYNl<C21|ivbvYF&>Q7lN-BUL<u6+jlRbg#$Xxj6_X3QL5cJy7==xe0RX|;Udn^)
zhSqntRFJ>!&?Wcz2kt(o`f|Z8RI%f{JP~6dy!{C084OwQwIO=k?;<F0#UFAs;Pokj
z0=@pzU>G04-KI8_X+$2_Hfq-BQX&FNV0?I0Wn3?V7$ru+P+*XOmnS*_o0ELbu!!v=
z*>{$&@dX#jSnPoNvFMEPcq|Wz0&B-yKAODhn1UkN^n%MZy{f4R3*uq;7ve}px6rTF
ztKFhNCSZ-aRbY~^h$<3j0&JKd!8X)!3M@R_0Yq7j?|5jd5J16<@TsZsLc<XI0Z=-a
zkXD2O%BGeSg%WnbM2uxxCrZK83JCsh>rRpylDGRC;o9y*kyHEWgbRrfZ@l)K?nY}=
zH1HBt*r>>C%3cJ~FL$qKbll4zkQKTS3cD$(@%CZ*QCc|I>S659-^2DxAx58Qk7@Cd
z3dVm1tQne*bCYe+Vw;)E*QfRN>aB4tE+MKF(b9ohewM!bKc}PxaYKy3+qPbyF5JV@
zx}>T^Gvv_1Zz*Ruq%9pjt!8pM@!Lv%N{PZ~I)t&L!!IZ?dHg$-yRJmkm=n);sj{&|
zCZqqnPFR0UA04tWwh+$f)cqvR{CHYM^1_$R0tPo_StP57nD+`aY^Y&}2epNzfby4`
zS(6x+FS%#Epw|DrlK-gWckDiTS;Z98n#H$5_+Rrs&J9a>xDNs_=oRA4oNg{XceFg(
zD_1+GIkMy4$ZL_jgY&U;3alfK3hYl%8weLUeYDK>@`)4BZ+YJCB)7>tJcUg7WC%S9
zHP_Z#OOHLA*emKTh7LQ1pzi25v^i#a)dxNz<}{5D^R4{^4;l1?C{4B+NSl#)8KQ!q
z!wV85bzWghK~lYjqWVEmvMcLHK`~$eH|M{ZD|z!Epjas8T}YwH%l`sSV6BcnM!gFZ
z6GdY-`5!2zqZFROsi{5%IY%3{?=v-3o*-<YH6U*zY~|s8KBe<(UknEE7WsF`q>@C^
z>;v@-1-;~Z;6_sGl1udhq<LHJLf=9^vcqKU4@33WKrjs2L4~FGN|2S3r^SO2(2lv0
z>utWKiywh+iR(?l3m^~z+oXAN>nZJ)kaG=zf`&mWqaY0kR2<%Ui=ZTPp!|M%)6+sL
z*Er(M*dbbZAJ!+kg8r5@%14Aq46StgKL=9dS|OFIKL<*Yb;~EjKN-o{4-oI!gj{YF
zW4c2^(-Cdj|1@*H4R`=;TCigah!}I*=G#~=4knmr;<t2IRk9OcnA8c|#<v6vkJAk&
z@A!RH-U%jz6lWVId{H%jK^F=-`2SQn+s;N!|GW+j9Q-roeo=`*H-luGcf9}4b@tP0
z<(HJxZukF%k|Qd!M7!ETj69TLt31Y0ETcW_-nY`9Na3`sx3|gegni3z0TcZ^yq3JA
zqMfU=oqIMFQ&gIr%yTzI)wXaeGu^r=AE9yO^{q*9W!og@k0Z8~s6YZ(zO$)yqxsl`
z3-ttaBkoZCLm}N5k`9WV^r1L(qt43<=iNqFbWoLB=tj0S1P5YqC?Aac+1i-bq6O|U
zqp3{@HcQX`KTX-Zmzi{w%^h2zrlb77VN%BbCoGn4kd#^1`Vj$wC+Idw?yc{;X=wkF
zX_xo4x)l2cy3`_8AThqkb9C&%3p(7q{ZJ|k;kj_@^V(Bw1GX=5B8t0f``VD`*KgAS
z6DjL=#=hp;wW@~_Z5K-kce<SiW@XPNz@d!(=+3j=(#}gLCq?XA_}Y7a;gRk~;lZ9U
zcLnV-achZLxg%r{z6mj6+BJ)o-b-MzoR;}E+z63Jk!_&R!pI$&h`6>_lAUL@iv-Z+
zP9$hXM8l-Nc}UPs@8wL#XnvW)$fluH7vZpiDCd=;97Q*6Rt<QMSOKr$z<tXN;86{r
zD4s1~ju(1aHK`6?&?xp(M^Yt0GTt|tM52_~t`clt)Uwpk=;>?Z#qW}Fb|=JCAxIV@
z>1-YL!x!V{7nT5}YUk~o_*kN-V}T+yUFzybaCF1rfYM8W+(r;XQjW+os!=^dSq_1y
z%w7>jBKOf8Wm^hzYw4*NJT6XjxWS?3s_QPFU3+uAfPpKe7D?|AE>69IaKwDYxlXTv
zi0-FS;>+^Fzv`&!Y(1~w2h1Zu3c4Ijcv++T*OYhMA5MnL-g2Y&cD<|01wkakN#P^{
zso5S8Xm=#I?H03bSyV%6)FEE84SkAw&Hb|3+vF;|$?Vf97>q=Jx0+=}TY)gO;dn=x
zqOq6}tN(p!-wdD(9UjpImKgq1%+2rCicpc>@F3et(Xa~Kit!66vPQ#!gP6l${%~gN
z9{W0dozVSdXZPxP?l5C3kIZSl+Kz_u-URpy`e436c2#U5pgd&*-L*c*WLm&QdlQKH
zsL8L5MVELf^RBesShzm2KDxe(g*382%nyVHOh0xN?PC{&aRd=gijah({4(8IL@mqf
z%jP|)_g^RcAK*+L{qQggahK^Mv-icSKS5~$3Nzz>o<*fQ@rF$M<vos==x9+AN;o>>
zb>k^hFm0$bnrdOiTIjhVvBLe=@7cK(ulH-Up#_?X)qAhxY2#iOLdgYBBJO=(u<J&g
z@+YD`*Zaeg%WS6sgd+;<$on&c0%DH)>w^TIMCgJgw+lSEd+m9%+U8A<(A?Kdkiy#{
zp>J`;M{M%<5Ul7e)Vycg1ScP^kJfkb4(f5)T%!+9dV@SK;um|{^f)GOd(dx>Q)jQl
zE9-h=^uAy>t?WS;o?PFH9(O-II}q?5Y0=3DSU!y3_pXg%Fh>2c`oUn#8)h!|k=y5u
zSng1;FBo(4?2RH8-WTlSsdn*HdKaZM++#Q5--+mfyB&tEPoSfKsPT5clm=+;Uq8?h
zvQqcn&z*Q39c_%$MxK%RKKgVdI6(Qq;2`CDD9@?w`cc|ERzH3t@p7^j9i4DT+?Tgz
ze<yN{E%%Lk`&g$t-Z>hlRuA@brylQsy;bx5^kPCh8NA0K+SFPtGVV$Lp5PGTxnb`h
zfb7sKI!iD%?tS0Zd!x5G<Q;h_%iRyJAF+`>%)O2VM|rBF-09f&r#_ar&tB~tMn3L(
zIpu#6A-UT}`qe7eFRMHmOT^ZXtsh@M!5s0%+TeI_0$)Gr?V0WM3Ga9?;oemm9QRJJ
z{`0|cI3{u*NCYRsk`qzMY?7x9>vPq_EIW67f@e9&vz*-YEP8u`)~SU~<(k{<nf9+A
z**cd|Us@x2O84MtPZQ<mKrk`p?r`QW@eXIXL-)&RsS&($w0q2Yew%TWrx=;j`p%-e
z^KM6fmRLWzep_d+7mftCDcl_^`BZR<p594&S=zf$KeKVRj)-~V+)aoR-U*xtSG{&)
z+CSe4n)({=E`p|JMfQh5Q}3Qs@Z`TCu=_nqzE_D%_MK%DgG=lfdm$J8L+-sJW|`b9
z$nr1p-G=u4h_3lgB_9S*n=QbryUBss9@S66Z(+h__vO{4Dc)OSeiI{`+y$Z*O(OO6
zqPKq25gPp4^jx{8MCTy0#96yVK>5F{XZWnnC-G%($0Kd*iDJVR-xfxQ@ObyEDKk9C
z+#Oo*|E<~*L0lSk<dGECQP>6?b{QBx#4zf?3!)}h`|9Z_k9T0ql=)%I4shCU(m>-_
z)LHjZSMxlXpPk3TZq~0<igT-k=f(*dqWo4JY3|)jBt(~>7bTWlbA0pJ&gpFqM!lbz
zopaU8jpUj0v*82KN8#)<jsDgUJz8BMSYds8cWg*BypB5;pN`Y45u7!-&FU)YrF5b(
z&}#3k=z?^rG!)g9b||de7-`k(ke}8VKNCBM%4yt6vKlwZYO}9XVE0!X;P*5HjAH>~
z+X*Du>$w=WJF%Y=x4USmqN=w<{dKYzd*HUj)rdw>iMDB-Ly_bM&D>im*X@<}zGy7J
zPh}b2SMSeFJHpcl#;VSl*aWIClm~xQxi%g81s(pL5)<7>erp~^{+#AGX_b}yiViO;
zF$&kz?qbh;$sm1EmcAK%e@3}B%UIXhGb;OWaucUJJeAI>^qdlLLT<|v&JVsv`}TI7
z(AJX@-4;r7V1?k$k)Gaye<P<=(U!TdHHTcy`PtdknnK>QAZotigWqxwIat<1{4on4
zZ|{qDJ_R@FTr;N=>YEEDVifkTsn@Tln?`jTNXj-~+GK6~i#qwsx|AZbl_7*zt3Qec
z;et~4G&&!s*~M#M-CnW?FnC|iG`pi6!{#;hYO|+RtE@7UhPC}Nwew6IE`OeuhNl+3
zyh4GUMg9ZVzgvAY#^5}cd_s)DFHscB$ckIWGfwd;^B6IX&hg5<%=nB6!V>XQunqmj
zFo;1{ux3U3>yH-^oVgIr5_#(n7M`=pjgdN+tsE?dm<|cWN0X`{Hbmi|T^TWTDH-Rw
zLx|4Li#g%^Arl9<5JusV^;Qv*c?qHQJq#4R&F;0mJ6i&oEKjW6VNC9yk^zh3yM?bb
zq5=7TS>UPdWiseb`5wFn;nx?k*q_GTThRk}8p?9+TB(c0C~qp!wO#(V6D-FC!%9T)
z_Mj*TX2n^2ABm%9!BnBX2wnx%kP43#)kMd-bD#f57B;K@N~ZQR^#Pb-*?1Dl6#MbC
zUPVSLLomG=sCizq!aFT;Ue_yfNU#^F&Jpv$B+;<NCcwx~HjeTJWCghjExQ5c7VOF}
zJC@WghV{Y5!B${f7ul+_KEsU7GD~bkYw^D_sEy)N75BRI)s^C7rcHe8$zm;xhv*v%
z&HI^rhze;Mn<>dYoLVc{>}u6Raid@!3V_xH-sH96)tT#<=-r|Fn9j>!=MQkljwoAi
z)8Ug!43sV^XMj$kA*Us~Qw_XE$&|{bl}subQX+X@GZ*y&q<B=N_HcJ`O)2jNrHs9x
zf42&<@=Vdu|EVUm-#XZE#i%|f93o(^z+7@&lpt9uqCxV`9a6ha=U~W_GJ+tQ&JJPu
z!1Dxl+Z;B$<%?i)O73B6dx4DeaCEC%qL|vbI%^!)BlAOm%OIx1P2MjhpGAVBW$N%E
zScPfz;)%eT!}x|HVl&7vXISlZ-&AVHdw4?%bT{Wsgd*3{MW{n!qu#~6-E8u@dl&X1
zqCFl6Gv?}O1}ld2PU?!^Qle3E2<yLTFxzG>3Tk)lyl040+UPY12Addz%v%Qi^ISWr
zK^OXz>_c*Xq#QfQ5gi(x;59CHQuX4c$w~hWTz&CkE4=!<)RAYEyjw|`gnRe#qu}Z<
zsKZJoG8ZpS@;8y-T7Badb3J{LVhoD-xp>i<_J5j7DeApM_3(CK8HbY--4`e6g-+=@
z7rmVl704Z?*M)Bv4R^`^9I3nZ(xr=+E?N692`AkKc&*MNP5>{J36eX5Kk?)0?%PzJ
zSI!1rZYBOZm1xyibU^>jIy4Hv9G*?6^_<GCD|xSymy{Hg%qsaICFUaCqg+<Wk1P3v
zlE1GcsRVX5))5KbNQf`#zF$`ID?0nDN`$5?!n?28^k1)}HO)X9+vmr1&HbDjVD+1R
z1vytb<|HVswD)W-orsSP#Nxf#TsAk7FXXy&J>+_G{rNjH)a$<kRMwr(_os5{TyK7W
zaP7VM2PxmLvjl3{>wd{ueJeW~p6l!H<?69kjrZpETc2j=T<asOqmRLXD9n4?GK6sN
z*-EN{|IaG~@x~M=T}oScaYe?-`Eo64Noqx>sX6_IuQ4U0Yy2jiVsj=W&n6%g(GF#W
zXK}AF=+_3fjhzH}+*QxMAL|L4PEEq?7-*HMyxih!2tXP7ogLyXrh4=blWXLi5ra6N
zEsPM|d`xw1iM^3IAxP(zmdj$ATwqILB1Ht7Sq-<14v?m6V-t#=8Wt$p!cQxrr3@3I
zskgOY^RF9zi*~z~v}}XrCyvfwcMEAf<KvZ9@mmpy<zTv>y2i4#Lg{o-(vpBzc>u;s
zR$^%D5V8R1aEl;#(#!Gc6Lv=4eQ#25m>`beCb=tg9!I~R+C;$%YfH?7h?3d?xm0ad
zU|4}#IG;16y!<__eN@$Y(@UoM`*w|9g!##x?$FtSUCZ5C*Gi;`Xwrmvpn6lw|I2zB
za$$$Q?%Wbr<xs?Ds<YM>&N>oqy0#D*?su)VpThbe3sR!QhgV8Sf~d%@^mtm{$!317
z(Q`_OnrXTGXqlrNKd4DDXzI)&eXTxbd_(1*W=cBOACqY1MNV#6i?&uxD#FD&4;nYC
zmao!=gL*aP{!EFX{_j;K_quau`)NddYmpdV?|;9livegRW_I*4eUFymJv60dD9TnL
zOSBkesxFEa@jJirJylL$GYsKZW+fR+8@Dam&=HkziXedO7Rc^)n-N0>alZzb6byNX
z`X|@{)sb~jMV*a4Jz=IMlY$6H37&Y#SDYCeI+W-004rXW#v2h*m9+eoy+H$HUA&<}
z?}o^Rr9)o7H}JgP%#brsCcX1;Gt}fgy<(D#*3%MH>q)jHZ(mt{E|gj*vP7S=KV9VY
z*^7!>VU)Z8pEjy~#SGV^e^87RHW3TnUReM=?_-Q8Q$<uxCWkg4h+0aZ(10J<G}{5Q
z9*Gu1bEGryt#!!4p(a5xeu^ZD2-x55tRv0*R+V-%_MHLN4b6lDSaM-7fE5O<**Y{W
zu+<{<BIy6H9$w20?N}5u7d*fN{`o@<TNGNoaxNNv2bNFJ`j)V?k?*bM;?@yq(~Q%1
zv>#G5m<1UQV;O9F$Q_6LVhcD+@{!}KId^XnOaq48oJ(uO%%eF<A&JfKw%{YJGpqyY
zmF#@dG)tV9k{4eMQMyQ{kwB*`vQIY%Q|d<iFM@6iVd3-u$8q!L!>>@Dhd1NgK{0dC
zB?++NFUc#n%s9RP63Ft!hUpc&EU=*Zr_|~qqPqMM`!K-3Ngwl9nZEJ@j#*wpsWRil
zau96^C)AZZcV{W@PN73sTsgs`)TOsxcq^E5Gr%JJu%Tq5(%!C_wDqYmbVT+^<?@<G
zBgV7Q0(W}IA5>TRJES;2t?GYM)zwS)>Pjt>TM9~==`}Rd&ovVb-nZMam787I0-ENt
zCNYQ>-7SO1d{m1_&n30?I0kkKLLwQ1aZ8E*cnV&LclE((Dxl7`g(WSfxfOR1&aVo(
zi|XXV`&fEiE_NHg+LNZbJjpH#dOB2>cr%#lvR|r86qzg@TS8b82u8B_a<UL_t1eM>
zq2O@h{g(1FI(Ev-A=C>}UJmms<)ti`GyLx27xsMv6Fm3k9~7BmV!XtHq4~!6O#%0*
ziFy*VIhG<A`u%9R5fgabs5?KW+(y4#7cr9?p}hji%D+h|Z(uQH<tX9<^+u{m{r0sK
zo&#YCXS{wCJ_F$Db{m(hjX`VU9Xq$-4SmfP1`$(?m{M~g>REj_EO|d0jeEh84tT>n
zpHz46sCO~qyK!WVl$zrxHD#@iYSfgPdjhE;$689wU$Wy}-bkp_{NeiU`uGO5OsP4F
zQd2Ii9HYcUeTh<YKT6F#)Hx7H>7~-5RGPtFZ;YBKI5!T`Gt`tfBc&#G1=K%mO3>Z)
zBf)NOH=tubx&7X_<&FmXgWYcay*)70`#DCT$WzH8{U@>P--l{xKXf|^NG!r|6gH*i
zep6~5STA;}l(=`Oyrhi-w2`wmVjIVUB0WA7NMTr{yxTiys>WZZzb9z30G)Z**6fkB
zPw7Re&;G=R;&>Pwf;LO~qjeN%FUo+uODa!4Q$OjB@x2*Qjat5b8))<}pzX+uaVfmH
z_OM-hp>^%}!YNaj-le+k`mpvxtvas#MrhS`#5?+Y4pnpOZkkEB6?cz9XB#gkuKqga
zlzrYYMhpIN^JoE1k9fyTb$E3Bn9b2q#^X3@%mMENqcO3gI`kOrNhNvA+B<FbcxPz&
ztiGn$yUi4PC)OuAXJJ9g&{N)NX7gllk~^GX#1)3@tbuUBMZZg6lnj)n4u6fLlmg26
zjZvinb>#;-pkeJ#;-`(U5g5{STn}>w*b~MBx0?!3%s;PPN3{pg-%0B6pQ!vvfaNAT
zxozOEEfww1S%fD4PVw-s=mOa2BkKEiD)}%CJE_jU)~W4ELjTuQ?cXR7@%~q<3iN+N
zo!C(e=yZ5yD|L!WbM%D|Ci5-TTx*y$#3G}H_?qci<W>+aJT`=LE9ek9gpLKD<`Twd
zS&wTN@dyP@w*I5!{z&*r&GSCq*W3ap?K}y@1GYE%{w3%SO#$NnJ9Y8jlb8bVFRpp>
z{{wAJocjjUTFunLGOm)cwRED|A7nsy*W~Kb%Jq(rK#MxBmA*yg)mC&l>Ca!Cg3Dx0
zrsaCKx?#NF_b6wAsg`bQ;<gSl=&X9ZS()Ykmdb8d@&zSdRPrB{{HBuMQSv1vNQuIi
zDE_W;rj{uwXFK{8<xInrX9fAaN(xG(aPnVOBFe?Ng@0Lxzo*1j{C`sJKPx#-Qp&8(
zl<P|?{_m^g50w0fuKbE}|53UBqTC-U(fZxAppWak1aLp9wfub?5s=e#Ed|^gA`MZF
zI7;GPr6nrFotEenc@g^GjDCn#cqyo^KW1&>zRkUkB?d#SQA<Na4av%#{wI1x%hv#O
zVx2EyygH$Wb%N7A4QfBhq{OC{xrBYD=?z+9Q^#D3k?9bW7SxEw=4**0<6QH8jh(SA
z@!_1_{@6d_M1;Q1&S+4hkC<3}YHucK^WE^-1Be9EuecYTi|6jq8z67Qyd3_7;<v?Z
z6C_AHxYy_P8@HV|Zo3<tG}H{kT3wt+Ke1Y^wAk&&NAkt!^UzdnzE(ucKt%P~*@_Pl
zU7U3nOc5HlwmNg!!u_=L{Y6!+tkf@`H|2M`rAx~S%C5H#L=8xxBo2XQh*uQ0TCwPs
z3IRV!$R4}=<B7&;V>r*K4^ENmj*WCLPcJUdT&bNo$Dh+Egw>_M-Fivq73a>m>Nf4d
zpV15GkR{Yw%%y04Ap;X8tdSvLIVwJTZz3i*<P*5}=?#fpqNAYuv8FdMgXF4m&X1Kt
z!2$w^SQ+;jQru*qC9qTV-$Rb~bn-$6yr9jR)sFo-)CvIt2q2q?!-g_rysk|5_a7i>
zOb~d<g;<IPuBdm-6{m`Gxbg9J$;*|QD=Wa+I*VHpR|jfaQrKosbe#>jzl(L(>{94^
zV$Uztc-fs;FujgX_{enGIK6k4De9<%LJlroCKnkot1W+wq%>#^w(g{s{c)Y$qa>|v
z8PIDz*UkcwK_oiP2E1EBF{#O-zdYVj_+5ky-o>qJxg{UDIHO6_M|~FU^f2Gr`{@F2
zIzaGfY#O>%FNqm*%IR#e{5*#2KvrrP7)Pb+>9wB^cwzOkKeL`$!zPa0AsLP%OTv(u
z_j?em4>RM;O~ft`kmy`rAG`Vx3<1L51*RHnK8f39S9*_L#JTnWQ_VJ|ojZ{rNcy0~
zq4UxV6aN97vaZPUoBf#6<!W&dQFuLf_1mpwEuqiVu~s#C1NVf<BD#c!9UhDk*YewC
z_IHZ<0~>ku9_NKr6I+PMv_<Dh5HB46s`fGcA>IVNC<HK-yq>c!=)R#`FS&x{`aVW9
zGPKKf#_NZ@h~omC#yXT(OimN->W1gu%Q<35MCWX24^8?%$aFCKe~Taghe#UOWGz-?
zLPXr`>&0(fck7##NPJv*q@HOgLJ9F8qAJH5dyd-+AhhUINDKzR(fk|{e;13qjOoP8
z<*6zWxoZ=EBI19^Bt)Ufcyru@H}BaSXIQIVacQbPb6G8zI=6zNSpdo1?TjMq{9Sdh
z_eoy!NF)l+^{kO`7F~@VUQmHEEtgNxR->QSKULONh#NsoffghG8@-Q*mEJ<py~W1B
z)9qTW2=R)^V17(PVq)AMSMC!^o>20?D6uW5&F24v4tJAl=Fzkgzj$$;SmH*t{6n2G
zjkqWoL_bGW@F^t^D0xuHKT=}1Fv|5GbJ*;vR-OyP$NsQR|E!XKrsU_9h&phIkp%_>
zv{$7sMWY6amI_;F+cYo9;kXlZqj9sddok=7CwGLe{-U=}ZQsZ9Ro=$Xew(Jy^(9}J
z`b)Z%Nos^o_0gaSYD=kKdq)z8eM=+rb%m_nDyd0F=H-}5k7)BN>LNLL5s<?mb~oPh
zgua4%*O|!lJUVvLn^DYHowR01rr_MCZF)W(nx5y7#rz~^WO|NNgZs%iy9+-0Z%|4g
zF+jloh0(ZLMn_`57@T*Qo`+|=9wafn$RgS;ND@=<`b-id!}A^5@cP3xX?uqj29X$H
zSj~2RbJQ<3vMZSeWcABroBj^i(+wbJQf!R|_eB<?eC-u#4I%q5FhsfJVuK$bx}~C|
za^)ab$}rup4<o}F$G|y~;p{<%BL_~r@&Sw2khw_evI`jwx-we+q8*QVL!k`k=jvni
z-T3tHH5tw@GMo`cgR7A1#1=3%-`Ix?XD@a32YVQ4z4!NkBsdtEk5Y3#ChSG)Eym%Y
zVLEP_)IVr4pYi&kVBFh<<mLdm0~nkucQ`;I<W>R3=9F+eWJcwZ;rs>?fIUbNPM8z{
z84yyEk-50H_Z8wtgeL0yOqx?%KiDZpV9X+^&5*aBF_N6?H|X8b;2?cJ9318@lJfKb
zP-FtnTyN+__A(sjJ~H=~M8@3<xyOKakbYdebw4m0Kjs}WISHoTHV((R(?l@Aoep!K
zBirO8Lr^3~+j5P+AzW?}W-?(g(w2ndFD^SEc6&s7xcJ82|6|gG4t4OKPqOe=#Hvpm
z{|0#MZ-S3*(`B=oOG7w+5#D-EtpI(Ic`5l79bQ%<wi+`OLqXjt_}V<PY(M?qR`N}9
z&&#@YGYc)lY-rnWf_2WpI-4bgG10V9%2G0m=haq5dJAq8Tlgm7#`iKBk!j_A!V5<%
zbHpy^cVL&t;)lKglS<Sbn~W;OdJ`oF+^aJwOvjuVWfY}*JC<$Bo=`K5il`7vjBa_8
zWsxz+`#C%5EQ?$gK4S}u{NFY3e@DsVS^>7AL?U!>%wLaL9^ozpjoK*J<x;7Xd8EXL
z`}z+!hB=09A$4Xr-TJ^>+sE9yH?hnGQ$|Zs-PKZ1_gERKluiY;l@YI!D5sq&>fT62
z{itJj+dT6x8>=d&tU|<g2u6RkD$aNXi#+1#Mi%a&iR1|powBB60a?+6_-4g=hjv_p
zPUKX}DWm3IfL1v(cLvI|c16vGmx{ow^9G3GwfU){Ez8p~qj7O<zj=<k`0<3<?Gsp#
z`w=?XczOSj23k8>9}Tkd#uuL=6RA}R*pMV@c&fz%!w2gX3fLB%kkUO;uczAT^=5zP
zeZ-I&*$TKdAYuIEVQ${$7<c9qZ)&CRrFX+{h|!x~woX=yDV)584!NH-OQ#XC@iJPt
zIGM*7YcH)Zx@7%{FWjs%b{4?!VGq=GMHZ5#d+lYe$Uwq;5H3n@)`onO7I}dXO^SVi
z2Dt!XBK}Jn<Bx!9BaL$p@%qFE5}%LrnFV6=1qF`#cq<H%G8Xz3cnQSai@}A3gVOu?
zi1^oU=v(hi;Yz!<8>Ja}B2P?0(&UMH(Jkv=YNN}BP(U;7<zGplhfYJlDfoMz(?-wo
zW<(vC9=dz7aaUchwlx`!7N~peZLRgD&YFvKCWeUB++Osz@J#v|I_p>e?5kh>DswE4
zqfVP2B#}mA9H5-oWi6|Wt*UHeifq{!&~aEufXEoCymF@0p7<ZNol9>URTRe^+ew``
zvE3?t5{g?&lOWY?LM!o7iv&Wd)I=3x6(NCRr;dvgN7puwD2Ol6MUlE<1v}`955a;3
z8(6ULN9YP9BqYRw-~Zg1@mS-eO)0iK_B=ekp1Jq_&*Oj2(fKP{Sodn{)1NH*d7~Ha
zv|DJKzOIYW7w_u;v4ak9efuCRv)hh99`^$M7>dxxPM(~#?0RkUQ#ZrMWRI3L@u7P3
z0>7G~I*Kv@`syZgrdRnS^>etnz4j!-XJ`pY^!dou9l1GV5Q#?c>#TH1RE0k&t<k!t
zB5fH;l`;V6vPkSI!%JC6a`Zs}wgRO{*=eLuQa(oIGc`pVa9=|nV41_bp>)4n6B0=V
zHX@nBlNU(<@}n-2Ogbsj!O038+I<rjA%<^hcTSB^gS`znVc|9SGUT#e^M-oO6Ipom
z@4i->rfE`cJev`@6dUhpr#&?_G<@t2Re6ITW$+k4m4`2U{|Ts)y>}R@X8K{!gIQy5
zZdDsG6Yr9Ts}N<BpM9&n-LCao)xdsR#<;<>uM~sc3JVTy5u2vqt|1UmH!={=I0`_3
z@cz9ti4(F)oPceJ4rOFh#YC#+9fnbZfy}BI9?c@cY#&)CJr6W*hGb9k>=*qC_8-b)
zf)D{XtWlR4YqKnrY-d&y7t}_6)N_)ZMmO~r-5_4rDNZ0E?Y<a9xR1D6qwqq>1l8r#
ze>-Sa&j2rk6gbvVa69iAg5(TAWFQcNJS6X{;YWlP_eP+FY=SV5Fen&q_(+kvaEd^X
z66|4*6u1kFiaEpnqnrhas6N?XIFk)2!!~ebcpF9!l7W&53A{1BKeZXD8asWGY>M?6
zxc=1bLT&XZ^ARXBO@49sQWz!TC{qMwpfP|l*TI$25XzJcWlF2N*A9a+p|7E_56BpL
zn3uYLWW8DI@p>07lmW{{?p=ST+G;ko={c&<+A&u)DD$^=9SJJIHdxW)t;Rp!|MBa}
z`t2LR7D5$*6rFKgI{qZMBs-j!h$l6#7!WOLLbACg0wVX4wrOPbvsH?pYDwLv0S+CF
z=~2TJtS7)BcTs0gvFJ|*V~w8(Kk_^K1JiFBK4#{NfJqT6;;Q<n6?ytlIC$kU=Xsxt
zgVcos23r!Sj_~(2Qf-#m;X^2DdXS>GGrtI?7%7<R7y;RCmg>cl)R;i&*Rba4*V((o
zIt_25UsGYDPPO?xOD0RN<efg9v#O|=;tbxV;?<$=(E{ryVaZ^3i7PnhB<$w#`;Gp%
zUDJ?9YO0RE#k%TtpDX%wnFR-#s2fdM-ut_({a3VD(m_FFgK3uY{Dwj3U%C@QniRsp
zEYCAs>nMTTsOc<*f@4kzP@gpb&U4P6)vPI6mCIz)g>m4o?6iF}{0)<eac+98EYQ>f
zN-^$D)kUut3>rg9an5hGZTb4N!QiC1Ij0aziv6=gJd1?RyZeE`TbW0sbO{dO9bO_S
zCKWybmhyr$Ed|20@g<<TaX_cNwNdWufRK0>jImR$_8KZH;!`~asA<t*tF;j5)t4j0
zl)fDFWio>Aje(dG9x*ZILDZG!=0hqhn{Ke$rSF=jg`SO;alJcX3}#ix0AuOFxE155
zx_kB}Msr%%+%;!}{Uu{odo2`7djz<u`PbR(-MQ1+AU7i34_{Z*NT0W9Wz5rOK5ubF
z!#-<hdyNflLJOmjM*TS1X9d3FT!tVW(Ft?q`B?)Po<I0*MNnMhr|c0*v2U~XO!>~v
z{=8)Gr{*}9^fxsIQcQ*c$NB1G{@`J_&bR(YxkI%&KO--S$sYH?ci7ALK?}Gp3aWF+
zC&isk;m`bzgxq}Uk{z=1z$6zPV;)Owx4ex?u_O)mC-#IIE)`h55nGx%EWz1|ob3@Z
z29c|Tw>)BIN6z?Gyy^ndPo9i6tf9x&kuZ*t-BA)oP7(%f!5fKOaX!>&$y-T$>lZ~l
znn@k^o&9y(R}Uk7^ej?=(RJxg_aFgeluERFVZB@{FYGLpHy3$ZroymKBc)BO))vc~
zY)kjkJ4?KlLs26^<c`Z8vxk#CB)vKtx=l$TUukt9kJtdP;cW7TA=ZCaem!X^k~|KO
zF-~guC}MP={1OL)AyZO@T#7y}*2^0z<e(?)Z&Q*{g%=(()RK}$IGLEFp`kx4X-JQW
ztHUt`7;*+z(8mIgy4%Q{HrbjKpv!(31@|>w2uVPKc(QRVKIV_b06y!kGE*u%$cRq)
zkv3XWa-W9F^a|%5Q{xJ>4WK6+vM^<QrgKjF7RTOSyx~ldR`2QO@R!#%E8E*XdUuvX
ztP~b09X_{R_^cb7T_v=q%eU4}`Qt4Dy#hNnjJ$R}5_@8!xlQeD?`6=v^W{x3h_;&R
zPLS(#)d&$MCrY>7Zgm`uO}QVmkoLBA)PCTTrl<gwUbo2+llsPdqCK`3INj*LDN+sY
zbqe*?EmW5JgIgPos+3Qs@Bo4z>RWcalU(MpQ&-5+b9nxu^hbA03$aYsiuwiwgI<gC
z!&`E*Z0*m-w?@Az*216-T64#BBALk7P6S_p<xr{iy6vFin$OO0N+&%xeA0kX75=ua
zfjh0!ozdbso%3|;oDl`9Vp%ZewBMJXsbIMVyxjBp-UCf1hMkvL?H6x_w{^w-qr0Mm
z6r5{;X2yCmJHA;+Oir}^6C)U0&^P9FvfP`#vDbAkEe&&)Fz!(EkUK@6#XaOnqDbYT
nC?OZKmQf~)TxMz_m!17GrRPS;zx(CNv6orT^D35>OV|Dd6s1Z(

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample2.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample2.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..59f96e82bb4992075b77cdb835f94d2940e24eac
GIT binary patch
literal 54422
zcmbrn31H>dUEh26O;@`{qtR@4_K`iicgJz!CF3};mBb#$$&H;GN&hR2G}4vkf3Ic{
zSKwqE;xsr3(DEq6MUw}847>oLlv0-VwLDtdQh0?f^wQD*Ex|xv;RPri63_d5f9L<d
zn>0=U&FJX-&-Oq6v;FqpS?=x4MflnM%imo7C(lPCzoUotpCk`&=ePQQMI#X}QjB;}
zFSZyhM(uBGDPD|6U0R};w6s()&ENQ9W+_|DT3%u?x71VYS;`ml_MBYoUFs|LE%g`s
z?K!nLuryd44B8ng4pBe7IJ`7m9JaK~;>gmr;x>zC7e|-27q?qHw>Y-6qqxK3J&WT@
zJBvG)b`^Im?Jn+K+Ed)Kw70l-sZcB|?JMqE+F#sn<?@RMmJSvVTD*7h(9+@J;iV(R
zBlg_4cywu^II(m`@eX_LU%YeaSn=4>@#68N6U7sjHn4bd=~VHQ#RnHpFP$l#AwJ{{
zdn4Yq&&7*(d3So-y|I_0#k;*7-uTOr;yq79yq(^zmm}V;a(dz3@|kOi;(es;_V$pr
zr+nYFc=3Lo_j(1M3o1j(1ElQp_LH)|d_T_*@_fKM$n!yaehbfsyu&;nF29BRhj>2X
z9p(9`J-?Ob3GWV`@37~Gz1XvniDOeMOk?iy{GwM_n!i$8^~;6j+1&YBZKe9)$&+*Q
zwezdfCuWwHPA;y_%zOUSe5HEQD_551tMiq)<KF!2>}qv>xpI8vYHit{Ie&6`ae4aW
zefOQZciNksxp(%SGxyEjbH8`jU9<P!J$>KwJ!j68?>$|<w|v*>d+(mUtJPAe*-|NN
zsri=8c_&t`Zq>!*m2zcj{^b19ioblR?6rF5dR2bl?z``Q;LQDZO;6u{-@Q{ar|-V+
zKJOmSqqh%CPfvSiyaygQb!MjAY@qexlXFw6)$06IrImD>K0LO(wDSJ>>dD&Fk_M>e
zPgSb3%l=Z?ujY;)FYtF}KKHsd)Z|38k$GxrY1*55>Y0gHBkPx^e2UMN8!5lMQl6@j
z*f+CWsg<wPmZmCGb7jBLQ(0YH4AQ8dUYYW1^HYmN6Sa%p5`VKxQy0pmIe*^cp^x@1
zl-A1ra%s7;c$JjgQ}fl@W6PD<`MHTqqi=C}u2j2HD@|2P^GkD$RQU?o)kfMY&rYo_
z)*8wA>eAFoBh7R#(E(a7GYGSbQ?*8ZW_h(zE760grD|h43!^qwp)<8|Y1Us}D$z^5
zB-%(lHD7zj(v&{2!bnUlmM$zdc0E2nHMd-uT6`iH*~fXYKR>-%V>MB9=t<_kHeb7X
zcHXbno~`NSy^pNcmdllyWv}eN`>Ba|W5~XBx?GzoRcFqZz178XW0=wMr)FxU`ATiM
zR4Z4i%YGxL1zRpH&R;0gp`OPdfAZ;Pr<PV0%RY|-PwLa2TApKopPs)mUuk4!)cYAe
zvWJ+zT&+&kYL!OxD(_4MIoVcve9BuoHxtvO20x0kuOg4}TTKwuBW$+l+Vp!amHo^9
ze63tq;wu(U(AD`$;WFE-P&+?gEzGV~W;6nY+Hyf3SEy=7TB(!eOH+$hEmz>jTP{}%
zmE~Gt=DgNSVXAU<Y1uEIxDjFW64mnJY$Ks}_y>8t1vfu%^8E5r`6QLf)62^jPI^<d
zsgra5^6JXTXICpvv0fR(%1Mpy$;a8*k3RX#$whDJq;-{TA8z)Op0~18x>&UlJi|t4
z^m+5usp-YCSF)}i(r0T75^?`1e`YiU;g8}1A~rU=h}0wX=vbr{S%}tSUbG&ckA6Bj
zLfUqXS!}YA^2(RyXUcq5($1Jhf>rHJBpNaQG{0vW$?Ey36{eE5T3_oQ;juBeTAA1O
zWBHd{Gk20~ecKiFPZFQux8r9|@Kjx62FnouTck+8yqFjF5_1tR>7~51m+`V*&g&tT
z_j<iPFX8p`Z-Bpp{3iL=&#%R1_k1EU8x=@+Ia`cpquD<1T>=h(ghV;%z1ti0w!a*`
z7A>aA>5a_wNI83rx6H=8F>i;Z=e%)mCxD~J+vV-%Z{FME?d5N;SMc`nx6j+}9pG<2
zpkRP{0t1QSpli*0%D{ly9|}^QHb9`1;UHzwJLVm?ca6}W6W&Ssw#_@`ou<`M?>!ui
zcagH4yt}=7h>dyodiRmH!@J*mfWPD3gWg;CyVHBfdn<ndZ2A)KVef4(M|jU}@9o|r
zr0?+_^&aEz-XQ<I-dXPndvC#er}rc|`#up#M2h>F`S&#vk3at2r(gfQIP;iVoUd>^
zHe!osCeN;A*a;^W*_P)xptN~k*KT{Af|1uff-fI`Y}X(B_Pyh8`|>x9{Kyl(we~mP
z_T{hM_~T!FditfeeL0-#Ao53VYsAYd)tN$=&fk6fX82Wm)dnrsBG)1t(d)dVzTJyn
zid1{w74c%%q92Rax7Q=3SS?<UEF>0^-y881i}5${TH;z1M1a*@O4a)a)Ae4$Og%@K
zt@jY->UqMR`VPW;t+yVnk6w!r@2mCGN}pERt1;f0C2fF|0l!!qRL-?1DMPimm#z)h
zMo1aj*mj+9z82-pnb+cabA8N<tg757`J;ZhK1O_dozkVT`Y_>+`ZmJxdPe2^)p~}f
zo%LOWyXt%VmukD~$!pQ-OYe&Cl}Rh}8ZGUqC+n$t+RMHQnnyXM4OG8VrRYyiZ*K2n
zeXsuo(tE5&D*1!fo_f63#xt@|s1McS^;o^XK2RU@@~@_7H$gA<)yE0<*A6TktncJ4
zhuA~Cug2C3#P|FEwsv^o2(hD#Qs1i@wWF0(J+d+3_16p6HEO$8ezSJRHI4tBl)Kaa
zgW9o$;~OWgE9Ioe*`nWxfys%u|5hft5xa7#5xul_YG&2<0X>UX3j&!Oz$+`u0NKi1
zVLuCHKL@Q>*nj0z5V^E}t@pvgGpm(~N^q914IE$b%ig>@#&vX_1IDB|_otVu<+Xu#
zm&+?E`i`d`se)=18qpJ+Y(4XHz{YavLiuX7k<*JycIw_NJyWihYxer`(v-l$iXehz
z7A}|P=g!xvCkp2TJNUm^Ezhnl7A|v)7yNRS<5I5@IHubL8aQs>#4$f}VPzQzdE(~d
zg*#h6h36l6X7b6&C-|awJo}Dw=ZF?2pL@DQXA950_wh&0z2n)!x%U)`?|9}t9n~gM
zjh?yrOXW&Qz_{Az?S#qM+WZo|VfY$(w~<RA3x2JU3!cgq51_jYPFXHh8IIiS{NiGX
zUA|fdmd&pTYWMo(i>vc|70At$$BD2!cWKI>IK}b5%IV+8lmPkjwNj}uU^q#+R`%!D
zrUXqJ$rT2o*2tDhGmBHzDrG2LnOZ8VGyvUGf9A?=%p1Kki}Ndh3qu;bMvsuBG81^U
zRBQCl3LgoIGh2<E9Y7^@l1~5@SIUdk#uzhETH%BAdN3ZtT9}J!V?dDAaOUUzsTFlR
zX_M&>X%VJXk9qP(^pv$6JvZWu<w_&52$IxDPp{6-f=@PjRX9X^{Kh~t8sJ`yo=fV3
z#sM5?gvBs7Z=G43UM{({)97EM>%xS>g!Ixfqe%H4L#`Om+4-8^tM2sK_&Vh7!b+_%
z$YhjjrON73X>|o;qFhzd9DOYN&3-kqI;cyvt1D$DpEJBtax>@Op*9n&*TqJ!+Gs73
zSf$d)zE_B)CeeS73idZ!cAwu!Xpc4Wt21D*!lD@6My#>|^2`ohUTX9_V}C&%Ys`#_
zt@xF>iG9AXlVZ<mg|0gs)qf``{u2r+dR3OqA;h+J-hWhO!IGyuL!e8l<<F~RNx=sd
z6cv0(!3PvPui!}q?^5t?1!on^DVSAop1`d;TOagX8{dCMd1VDJC|D*aYVonmZE5=p
zO1?xOVnIW%m9ecg3p|E1VS{!f>Yq~;LjjK{_CA8s{8j}fbNdFOspvSr?f>%8K6}nb
z_e66s%ds$H;XpJK9f&5PIsWGPn<2(O<=slfQ?ViHjzsgZJSmjBwbm;jej?nPGilx%
z{`3(Mv*9s*t3M0ASOe?37WLEh=z6-6t$_~$84`jxUi8)EdfJQC(_@kKOf9w$2X29n
z0S~gZ*hYe!tOwTg@z{>Y_1Jo@o?b}45M7VgQgxsqCw|&sN`|%seX|QmFTMjvu#mnG
ztG*p*o)R>#<!U_(d47AnG%W+8H~PE`usVG$YA{95g5BA+=N!+#y=Hz-y@ww4*R$98
zmS$?+rK+D}k&c{R%Nf`rr>|9KXJ?)M_8f1Z)uJanVQ?%`Pmh41)N^AI`^!6b>T3s;
zzCE%I-ocj)((`;<ABS8|=>1SV?+ss%b@X#Y<6vKwyAFQt`nwaXqxp^7>OJnARw}bq
zsfA&0^g`^jQ6DU5VZ_@$9Pz-Q7Pf)!_tZ1j)ULi@RQ=u=S?{epz|*idW_9*^JJfQm
zXJflY*V-QspY(;4-)H4_TKTt;2iEiT-*Y`m%|va7x65$4UT=W4(d)$@(wOymyNQEw
z3*(_hu#$zH*P|P|u1ENe7lh9kCQ%S3(Y5a%VrBavkkEdO#1ypk>$mt>yK`!B<@{8&
zaDM6%#NBdXx?F(Z2Kadeh?PQlY6fyJ0GZTGdUkpy9c=A$l}j_L;&XV7;m&7Hx$>pP
zaP^`ORd26pmfUTL%fMK86RXwk>|KI$x6R#QmHqMzps}mnovo~;(_NXb>+2gsO)vyd
z_u-?=lpA|mX?k&(4uqZuEP8VX9{rB|_0RqDw=_mt1&#YKvs|7%y=Ced!{lBU<>^06
z@MV7a`BD?>f4l!Sp4SGJR*iHHaJ#YxWV*CC9e}{<S;jmEWUdO5RcaHv4gC2#sOn!P
zXr#ljPYHroJpY7J<DeM+MHOnT0v&e_(4_)aM0yrzQoHMnUsIigh7RDqJZF%hkzQhz
zEUhj%Sa40XzfHl*Dx5m!xEWa{scOuO#_;5F?MZ>}B|2L6-r@T|ZkRTesfz!dsuQ1D
zT%71LnCGC-6Kc)?UaOY|qx7;y%1y1k&;PhmGUB3$wdXpbM!Se?e~GkD@mrlH5SYlt
zVt|88tPfC-iRBEOj70}x`~xVA#>W9V<A9fucq$sZl>tl)#P`H9lvb@V!hAd<z;r7g
zI|!J`6VKer#Pif1kFJdxNNM5&A=<Nv10&cnqkjD)!AG^1BmQ2(=mzw{LVP1}ol_M`
zK`Fq6g}6Pjn^&X$ULc34spxCadC|z)J}>fq=)dP(D{r+r>5f)hy|)GRVy<3DXXr1b
z!5D5^v)Eblazy_o;*)DT9xOZyhe!-A9TvjH1H^xACy2VC`N|5ApfGbH<8YLZ@bTct
z%wgI8u;RHUdJLm)W_bm)cG-VcIcZP{m79p!f>Ew<AG;>Jkrkp*rE2--sa8Eo5PM0m
zzYlz%FB&7%QzkkTO?)j8OTD4^Ks5HXSZuA&Ca>8*2Ww~>4Q2R!Os%HkLI_|AnPdTQ
zobckcSUs^37pZ$G;t$tJ*%4V!v{KiT_2lD`7f07q^%#E#*VFaXMxvg+9`#SvV8?8v
z>L~#=COuJ4y)|-i??sW8X%6zY)-ocvUyjtW8vv(`9ubNgd6oz*iL?go=-r&>r9nh8
zuf}=X@4f&s0_u^iCoV<((#7}lj_eKq2#7?EFkb_exE3V@=z?}!G@{v_yzGU9|8amW
zNQ_Vz0BI1T+{cY%27wSzt3FiEZS-FU)VbEz`&@fo?m`Swd!XJ6!0OpSkLgi<y&u*C
z$bwQvBLZsm{)ZwLzpwJCpjP%ml5fAitKOdV{vah6v^26l(B2ZHdXGomXk=EsrPVr&
zCtCi$FaAs|vH@bDT)^sJS37;QGuWz~q)%S%J~%DZ{XJ?5@o+VVx@z`z*5s{2^?`=~
z!55SDAy7a5eLSY#tPk;RLt>HyBP}p}cYSDM*vr$SEpn@e_=2I`8dDop`ttsIoUwj_
zwK2jfgTeP|tfK9{t@X@o53i4~s`~Aq7~pNg#()_lHe#`%*FYRW8Z5`_xsaeI7pXB^
zAO7BmJ0M)yAn&8(jnqfNyzuat*f!aS&#bK8h&J+Y<Ds8mO^d9}LAAne7a?@B@OXK#
zY<vPpiNZInJF34`;KYm|arndDg3lJ=RAQ%z)|_LtznJ2iqU*7G^xI<>kq))CJ$&!j
ziBpY)NajYg)XH#N*4#rPOpy|;Jp7w#Pyl!&vUbF_dMD9%;vw~ZvHCEPu+G~4v-6I{
zv5XihINNm8Rmzvc#~U!Uy>fL%JmA{YqW@W5;S1;(dgXtV*jdIWB_RT%(bqf{jrnmi
z<#_Bjdy#S>$bl&$HqOm~0tDTamGTU*vmoBfTL05Q{=0d4FjrU`&OO&E5+qKH2E?lf
zLtjv@kx*@$5WiLxKi{3GDKVd?YL0T7Qo$_Df4{c843i)Uy8vm!fJW7Gf)vra5cV!5
zp!0nli4Mg`&MX4^8%apqIXEN20~$$*cdCu-YNdK{wOob`(Qou>YwGRTpV;Qk^hA@c
z^>xfPVpY%oIYpuu8nO8cjZ7sF>G(gU)T9B6VwRaH8Jis;PWf_s)Cx4LfQv@w8_|Vg
zu2O<M0olChHRIF-!*K)L$hC&PYd!fFc=NCFTm2Y8A4r2}%?$r?Mw`Z7ihV7U(37DL
zw_<S@i-C6ZeKnJcSq@ZabXP0~TA|uQ(NR#1KA_T2^sr*_AyA1vo`jf;P-6ttBNL0g
zG#-Pxy>$@eMgM?cx7LQYSjaQlLg7z}h_;hPwE8Fkl&nA(TWKM7EnbJNfffauAT<tk
z7Ni5ZR4&ncPC9ybHZIA>)E;)M$$AR>JW;M8B7#t0*MxGi!o{OkO0`10aHVu=;zYCP
zgEkK&vIP9vgI4WBlZ7jV$M`*Q;>3#uhl`5+XgsdPauxPrWn~p!Vs|S+eHb*;WnVR#
zVxWR?^a@9fm?)f{I8k_K*)Ja}oI=Wp)D!k8;wzYbZB6=RB+Qiu3y&0Pt01k^rR1?f
zi2ENav{M2T#|j{?dS&n>N2&!|f~^L95m*yRf1E%5w<-`O?y%Qal$8n)nV^M6w1!lS
zcQz0qI69=5Z+~akN^JCBEE&xamPRKbQq_JONsoZh<f3aMwl=zobzn7G7@@%HD@S?Q
zyf`dHOmMec;v<nUj^-QD$r}mxZJ(r?`?kM9Y$D-4E6(iuzfM9UY4h0K`>}u%?G}O)
z%UZ$Re8yMw8AEJ{*sa*DwcYj^w<~|!@A3bWI`*Fl!k)Hf;S(-Z>^A?uC{Rz00PF7O
z_!jS7D{(@XE*%I5d?*q0bnoqY8q~OLUky*~8hb{%YwX#o5<J>B`M;wc|5pVEo52xc
zjpoS0({|$;y!}r4>;E=Ix<<6STt_cOC59`cDlTBqOIAd9d<4A=Q)0%(36GB;bv<dN
zbck3zuwY}J={T|PVLRAP`82O^tLZa}8HM>Ov5B;;EB`+zwK-+QzD@$XRla*vdM>v3
z&8IbS;_!t_Y#?F#Wgyru%{Gl;+A$niYBh&L3{wzsID_>VoINB%3lIb#yOMjcVc0Zq
z|6Yx+Ck;VE(TKdA^#r6q+&@rHZonn<l8#5kNt;v+<ZAI3;_IpP^m<0pta`>y-1MaJ
z4{pTVO8pXVZf7I>(Lc81=0?10d^mqZS2>dXw}_!5u!wBD<bRomiLBeJ(JQW1v4=y(
z{_-l^{3|7xDnjNQvtP%Jjc{YMIYi=^HyI?|gWkW=c|<*^ftN5L&+#<GZ-RgM=qPk}
za9DL!n9*2=KY1d}u^0FSF^n!mJ`WkT5dHS(Ga%C+j;<#)cFpmR!=2T5W=PE<3#B|`
znz$BU&%qBwZhip{?RAc{1<13xsFx6>y8wye#l+|PJgt5@0+9rH<R$Cbk;q2+I_dSC
z+lmt@an}BjzO+WLh5EtcHZ<_t$%o{%^M68R|5(9Jf=1F<E=Jv&qQI0A<KTT0dAHs>
z+iv8Y`c27>%0_hB|Du|*M*oMZ4!N1?n3OLmBVSprn6k~%RP{plJY`l&PGjP%CP=)A
z@z^LB#xWQk0hc^1(Scj&nAVOk++u?W<IWyfXPmvLBY)wBqc_1Wh=~B6i>{|?iCVG-
z*POEHYj7}3E}e#7E<*K)%^kdO_~;O;vr%RX(t081C12J4>e2q7CCzYt6s!dz^ig|C
zd+^`cUr#_*_tul*ms~8=WB#Og^s|r*^(2@}X3~&yW=3OnGiRzc!ntey-y?zjWn-z;
zH4%045=|j&s%!jlnRgH~JlCrj{@cjF)&w^$d;Tw|kn!1WDz>7M3Gm`M*OUoizKcXw
z!yy&Wl4y)L)I(Z9Wwfz8=_ys+tNy3*<5GX>?aci()$Tg@d&4<mPcE;#M^m|rVxW#p
zWhz<#sZ7AP+Qn{*y(IBSBDyx@AXYH_R`K-#Ds@b=HeGY$?dL($%p_~7H7Ppk5}fSu
zS<po`oV531=#cr$0c@f+DO5rUNq6exrl(*WnRL<$wiWC8u#{Vz9}X913O=4#IjLJ~
z=u08Oarqqwbt33<oWIov34lLwQ+8o%znr-i1L@Tg<PdlT#P?eK6}Gjd@C`1d4g@wu
zpZMeq4Rrjd*<pUaqRj$xAo>a{g#eVXh-``EB<(|_h-Q~f6^qWGmwlaq>dZqN#wgi6
zJj<tjB}P-Z$eKeUE(TN5PL#T`jGtgvH)8n-(gNhmy~a?QFR7%xu!mPTk~kB#UDkja
zs>NY!)P$3Wh51_!=wr0lj;`oT6&(TCZz2Ls%0j4O3tb7jXJ?xA%=*|h+sQekhnE)A
z-QVV2YdM=@OYG=S$ErL}Nv%o)zr(^-$4o7T^%?gE;mpP0OC{<F$LjQA92BR>IarHJ
zIR@SvH4KDP!lf-(nhEmau4XOmp?31ID2G1EKBxV=cqe=ZXbG6|NIkt(Q><9qyIQ$W
zS-xC>Vt{s$e57#Xo0^j5^&@QlimJw^rmOycBxcLS_m!3lwd|%A8<{2P1=5E9ePo5(
zCLx5hk%D|;NUXX}7z%X_V@4yU3~A9;$7mr3sI_gM&HZ@-3L%ji`D!AOc_TRC!}1-2
z9Ku}NMk4w)ZzCN8uZV7pdvVTY4g`4EY$Ap;IUfPhV|#fC27F&Vj_h=gJtujdcqPKK
zbb#VsisyT&nV==s(sh;0!gA6%9}hK$(DsQiEu8++aC<erkVfaN=krlCb6|mzm+_zR
zdMi<{Z#@eXK$c3){?}sljG*>BQaPBb7N;x_9EK<mr7kvM#zA%<eOPaRdOblMm4nG?
zh%H!I(DI;K4qJ6FW@6~}M}?o!k`_JUqiWv1E1L&`kIs4IC1v&!1Sf%*YSn7Fy7?ps
z(Y@AAm$s1Mg-)0>_})-aA<#A{tPPoT=_(X^Q^Evu%GMFh8M<Qpt9xk)rd8aVUxKeC
z{(U2k(#y@fDYpfR88@@;2?5p3q(vus-Qm=yk7%Ku=*~xM75iwwN8GB-gi?V<%Bg6W
z?tzJhOdlb{fb7g{exi_4AwO6B|4t7B4GwV`ee~=j{un_qC-jGHGCO}o8ZxMOI6aa^
z#vbVHrXw*N<Zq$Aoo+*H5Vng+h{VFl@jMjU73+(Q!G0NnEz=hpfhp5OkK8162zqov
zrTBdhzg4jVBr`C?htq~d4~hhW1Cs^xD6x^W6G{slXgUX667o6EcKCjjPJyU#hr0NG
zK|@eO24d0zF-_TC3&BJj#Hm6MoEnCrBIfnc@=0B|tiKLn?K(8l@C3>A{*Hzr!Hf1I
z5L?UJ2^?f~U<N9pqrrf}!$d^FfcU{ss>c_QOxStcF9gt1?T`@cB{yE<<OS`EZlw5f
zQuxN?q_2fH5j!z?b7;D7zHqrPQ#k%`VVXZYupoPv%IFebEji{T<1j^ehgz#6jTk~A
z8=650i00>3mysq1oX)SP>3~0xH0;CwCDk@~)HM=W9f!tu`hJat&OxD3;_Lw_qk*ib
zS!mO|Yh-Lx9KN)nkzw{Z6cO;%L3`n6>##(LM98U~3m^{n@G?`~2`BDuC0GZww0(>}
zHsO6eaGrn(a8k-E|0nrXt+_r{Zzj6tNZPgn9jAMbzAEgf*d=r^0v3LkPH`xu1R@aO
z@sdEgCru-QD&&w6e}@O1T8kp$V!=unV}hL#!{7m{qHXI*IEu+RQMG&<^sJSvrx9c9
zy9fe_Y!m9>dW^xyaAs@qXVF`B|1z&8O#;ekPfjLC%^}i~9%f`A>o3wm?z*S~gh*bV
zzUib+d%g8sSLqCkt8cOqUGWQyP$SY}(sTITPtv&?iADycg}Kr+y1dfhlTAY!-E0kO
zBSWOTvOIGhWkSqIP@7ZoStBMu2K-`b(JukdIkYY@j*bwRQf!R?s{8yhdgA_ldO4@i
zoC!w!?^S`fljeIWm0MmxwE}HD`0ROOl%h|F3Vd~<$B_$K42`VAo9DeNjoxO35|3^%
z>tpOAeWMXttL719bglB<@C>%k`64B%+C1WAYpo{`dylb@;A+1ykT5N)TRFtv(x(#F
zIuQ#9h>m7DP6(|<EirA*>X!(ZGWLSUP6&2@dj?&wXO`MV^m^no9A2FzKN&s;B|jN8
zY#!9&<;*KA+H28IMm`Zm5tlQCl&GadpO1KOzY~#bD9BVYrz6!zE`tQ<-v{|+v4}+@
zrFm1F4;EYurQ%I#ub_z4UG9}=XPPP7WP@{Bu(U(4CH!6>!RQ7tyExn<FJ{*zMuyM;
z^I@-jr-`?2pnq?=mYOe)g&*ZlBaR#gOcN0b5KSOYB}%8<oEklmdq~M9?D~J0OQR^p
zBWp+P{04f@qX~&{&L+DJooH6NZFl`AwRqmuTtV+9wz(^(TV3%dDe-SqUUcZ$bN*XY
zBW@eOzrok|k14_Nz<oR9#CtGyMscY1g_ULZkz1s>w@t+Sr&aO23TzPiyFT<^kYo6-
zR6GR60pzUfU0)qZj@bGRSWjKmI>th)SHA|yHFd0<h%Y5PqalS@h7wW7&5XAc9Gr+N
z?so7ob@RA_uAMl<wAe1<{1I7<>dt>q$F+X~b;a#AjhoALz@7x<^=q)_7!e?`Gz4PM
zw(Poq?gbuO40tRi7|a<5e-o28M=K$;b%3|xK<~}5=|T`y`ca-Hnm{>G@s}?%yHmW+
z*Esp&3WcCH?a1?Lqpu=Mu4)PWr4khy`9P1^Ov1VcWVj^YO0j^6Es^YRXhd=}?NmEO
z&a92usBBUAbsal)&d+EN!=HX4$`S0VUZFu*P}t$Z0O>#gM5o~trh#DWM?f_oyX`q{
z=u6tPa!}a7iXtYp2%uh?BO>-tq?W#h9w;;lD3Il35u^jH3_9Kegc6L(1-z@X_LTLI
zyg>n7sP)wHaIbQ6&?+9t0bDOQm8ew+0`s<wK2W6`WlV=tTA&L9D0QIcX?(Ff)XiWW
zRRVZlqNaxGXgbm>;B7tO{~c0>JyJe1Xh+`&Q!hG72XF6s-=?GT(;Y`;FYWY5>yTFa
ztkpbEN&ly7+crjPeEY_BDZjvZi@A|A)bES0zyYv>8byVd;p^#GewMQic--#}%0|kG
z_l)?snL2ID4yY{Nmp{$~HF}$JJ*|gg=Dmx{CX+wM<_8r(&aU8j73~R^6a_nWaL3_0
zO4D}urgiwXAG~d=P-h(7Fc@u{Ri}%cYe^cjvvoAGcEFaOAi62rw^M1&V|8N8Usc^E
za@U9>m~=?%8Fk%wj?*gg3kr;Z_yNVVlO1>9yA=auMf{usqp!Y8X+=HSiJDgIGHIA>
zh?Q7Hci^;>pilI-ArR3?WO5owUo{#%0sB!}0`6!sbhD(^I%zNeu+lWn{znwps7x#7
zpigTQew)PV{d6S-v7{3hWxHq!CAq%XF7O6-4vP)h7fYaz&1sAZT_5<w9=7yQjBVZD
zHLe|#sbSHt9dcssu@55aQOD48N?1a(7ZNC9=`Va|G>c=k)Ixe8BUX=?l);e~n`Dw7
zp}!N^Mz0wGNax*$2W#3Bj)?sM-eF4brn~M4jHDr7jpQTV)Y9`IYfB9^Qii)#8*xgs
zEhrnL5vSLWp-icXyj3$SC?he%*Gz$MNsWDNhYi72<!_`W{WI#)X$2Z*{{aOW3I~5;
z=b-A(vOhqfrG0Xa4t|KJormWpCl2_}QMQpQ$*u(3RE$mXrP9UKsYPdcV%qqoJ?RT&
zaSP0moPcCmxwz4}pv>P;@LLN0n}Xj}@MVHV0$a{SC+)bYM6rKu(Yf_&TmfSUh*b4Y
zD6l;uG}(Wrg7+zSMuDyKdBw!gYV-#7GBahYaoV+hzC|yxt$K>sNq(y@6J#Q+G<e2;
zwSRrT6)U2;aVvIzKAP%7W$@ulK9kC1GO507Do<KQ=}5IwnPe`N>(A`9++=DDV}#yJ
zB9&%N5BciR48I*e8ejcX|D5k?8FgZg8Oz4AiENU;v22Q1!h63RrI>hPu9?u`{c@z3
zEC<G!me=Fu;cf%*i`i^ncF=p?Qk{Wk&P+RHtSPfj8EbyP?Ea&eYns(R=9*^dkGbYv
zS@;(RoLzr$(Ao3j?cjsnVebfLmVtSu^{Qy^IO<Kj43I011O}8J^6v2NGy}?Q-iMqK
zXK^&hlNl#~0JBYJ$T?<aoHE*cAu!_{56n1s24<YQycf-o6N63O8S_5uJyhO<LFrq)
zhhI*HMw{vlc_lOH)EMBQK*pY=zD=X(eZ-7Bm9wvzLk)7KyvMzFSa0_SIWcQ%+Izx#
zr==eV`u?Q%F6;Zj;Q2*NPoJ_9hnh7tXGw3yd$0FC%Q@^l<2}oG9trB4^PaOdj+#m7
z1i;X1^gqR!Q>is=u-0-*GQc*guu1=JRtRXcxJWZ%Mub0-rV70gT2uXIRuIMtk&+}C
z@R!VF6H^OP%=i;y0Z^3-(dstS+mv~w)Q(gBZZ8hoGy&QY^Te7xsspup)Dt4piwO$^
zKWI^(#aVsC`z@ZeIGUZ5$<3+EU@dt~WPIL3XgVLQ_lkQ3x&a#)Ll@8(%uHb}l(K92
zS`IE4MlB1y;uFCrz`XSNNClvzf1V6b`#y-MwOTPV)BXx<fQ<pKw>Ah0A|s2K(TKxv
z?)8Sq1_+JCU>cXUVH|^jsTsL$xA<7S&(x=Xn36lt7#&o7Pv);<^?sk!h4Jcj(|g_N
z^_y-a2B!<V%|vHUeZc>@+D>mk=~nV5Z3Hnob;hD0_CGN=>6obRWIp_T3icB;MzAuf
zVxq-q?uCYTCI6e0zfN%0|7JxWRZv&JL?W}TqK=5m0$c_x`{(?}mGKUOi9%yLEUlR`
z_HCSg%{BpC#(ixJb=a&~fv~aZt5APlBfhe{;=fl7zE5xJp$!x_&B(AJ5EWV-8c8Da
zd2{y=rh+OsT)#p;iy6<O1`ZX9X$Y4kd@~fY>VSs~r0RbYMa-a}g~xPxWN29}Hb&cB
zHMF|X7;44E%oB^!B&x-}>FH%Ojg~KzE8?3phNUKsW*O|TrZJ%D>^XYuS$(l|N*!4L
zh+>Nb6GMg{6^GhQD1(m@BQgANm*lVOn?9!CqY7S9@Noq;clO2Fihf-|i{xugciU_m
zxH_nQ7|haaj0hWeKao@~Pz7Q&6O-YaG$q`>QBw6U26HB5Y#P+r;Z!?o26B-2s-{fg
z1n&~qDmZgETKU@ASl2<*vVfzz%^x}uN%Lzs)-yy*;T`(OSk^IZSW+a}=V&RR?z4f4
z&qm5wnY-=0kiL%Hg77h+*=(DHw&Ozj<FSW~T%Jg{%>w24G%xU<Qt&Ya>Ih;oAsi+|
zcPKHZ0^Bp3j(HF5Qg~xB&c#r-x-tcKP16F(bBJ09&Am+`G3S%RO=sYfV^U+S^K%dK
z@sa4-{_e3cOVf@jGr~N>AB~Wzi|DQXQv$vZ6q%4Cje}G+6^s?J4@or6-xo=VS4)Jj
z{n1|7jfEd}fLQp!xUyp>g+)JhhLq4ji_j8z6BHu~;*zcP)Dy(SCd*69VWXGP<#7Fo
z@?@7ELq64iorCf^R%bfmXjd}0M5wwQz9&4gy8lj~YozfvShMzaFyZrPSIV6SO0$KT
zRm?D!WWmkZ6e=roq8x9fAZ+0bn5vXvy)Rx}+j-!?fx>go%3k}~bB`a#di$|Qp3)QP
zb~>!>4$a8r8stR#p(jo2i9uhR3Dak1{qFd%j&1s_WVjQ?cD5a8O)ASjqg>HejUnfn
z1$*J9g>+*~{DroJrXjSvdepFUb7oK&Md4uu3+Fgr8ET<58CJL#!BdHpn&D}P2P@jC
zG3Z|Bw^Ys=(KWwwt7G)s$?>ss>shVGG1@qaPGJUyl{BTa_fZ8VKJuq)2fJ73R?Xa&
zI_NNR)#7dPW<^CfXA&$u@r6u3?>h7~i$_aWP+M3;+{esNN>eSK7up^M^$=Rv6v*nz
zc4K2(^*kO5Fuw2DLjI?inQ$!6s_+vE3}!y5*t-<GTS4cb8Aal0jKl>Rjmwl5F3_#D
zechw-<|bI$!rHPr+`b~zD&&&YW8yyZrOt6iti)F@#3Y+TERtLYr>e(Gsf$>3paUD5
zJj+Ljo!dyr15RVGeL0yXmid_R_i7>j0OkBif*XMdO741qnhq5GigFDOw1B}isx9%c
zUlGD5mY0)2@wrF3Kerujwsqq6tiiualYCmF`mm^B&ZWT`iZKCt1I=8D>l4Y7&f-Dx
zG!V{@NYZ)nFx1LQNoP(hMg}i!gBLfBR~&0*mWw1b!Qf%>T<W=aZ_sk|LcIF8p;3BE
zy!tLfzM5@?ec(K53>o^=_F<*c+M(ynSi%xf=Eg??I@#6b0>-OaY%b}9zs4Z0WzU5c
zXlpxF0FM(LxLDxw=q%zJgXngee2->LaM)R@i(t0N_W!mrj_R><V*T$`M%N_1Oj=DV
zc0Uyat)0Pu)cP~=)|BR9j%bS44!_ydx?QAi^&g<!4p5|>+yvO?d6<*^8B-|0^J<JK
z7yxV}9LKtz+|XVIlt?!Fs^QHSGd1#=$W$!_01}u<Fp*fY!SYNKO9P-3%c5|Xf;CCJ
zo`!ZwiG}hLyb+PWP=J^&*ThXg0t;65tcK;Bg69Y(a;7!si+j}=@=!4Kz;Rm#bL-f4
zIHqwOFcO@axzNU24ixB&8KqYjcZV;=JBGR83g2%c4`WyrV{4xO0rlg93OXmMF>XBM
zHvXZ*AjNn7q~>pww@7UUU2ix6vK`ew<3ZoiJ!iLXs$)m$2ukp)W2pK#fu^el*@g`|
zmY_iPSY_DKkf)LgsY6N*2p)BN4*iyxcB5&=q@{z!bYPdca}xlPt%6fUTHAJ{sRLr}
zHJp7eF4uDMoX0?80`2Q-9jnHWu-EkXeFQjjL!rrFE@+!jE&n?;n16Hgi29!<?Qd!5
zbbh(P!UIDr7hgNtJz%$k#T~;%Q_WMmy2gVJFr#579P3XMufN}19>IjhUBiA#px`SC
z5%wU$=30sp1qPRcN=T{n!BAJSd_B2#)b6!d^^6@zV&031NQxbER?D1!xj{yYTDX5B
zjoCbyf^SC#XetIGBAooctC9JB1vaLK>31W|_o{4&twv|MSEFGk#*T!P(Y6oWacqQZ
zF=wJ#ogj%+4n?jZOj#4rt%r82hBk?5ll071;`@n!%L5mr6sO_A!x%bQ8ypoJCeqGA
z!Nc;AmTZvXQaMtBk~(<MXa|Wv_V0XQpbNm-M8SiR2+ZR?s+jCOz#D9rr%+ci>#<Ao
z<;%wW#<&n?gMq0`^n4?F*)OP;p;>X95_XSU25&y+<zny)8Yi8J1u%PKDQ}{C!lE4`
zp(@Rhc!>vnsbHjRqZH0@Sn|!JxRGE1NynLysK@d1;J%zgQ~D3DqS+?>35^D$nHU4t
zZ%a=ponOYLp0bg`w+AH@s-8PuGiyJWD+E7*{nYsf3+K(jQg_DX2eBsfYUf+&p<R>n
zZ^&_sXEu&BZE|7PCbcL`FZ&>{$C|YYm+^0S9)rY6tMFT~o|>u!Ma@PLhX9scAnU|E
z%5<&G$l%{^HrA$9rWu#iW)G9z$XqCwSI{r3x+P%?A=Tm%_+Oxq|Kkcg1wTR%>P`Qs
z5`K(8a^cGoC(7Ft`^2x{l;?}e$m=Uhi{;rGerI54>dBQ7U!s`Bjf2cnu~_SZ14t#I
zUP^9Aej+@+3TX*3##l@2v0dHAS9_HXX`@z?=tJ9A8$?(w29lxd(4Ap#IsHgU8jYI9
zG#RZLXgF4oRux`H;_N(p!tl&~8X+HbRVY7EebTJl0aj_Pl1r;;Yt{57y=2fTYe^cF
z^|V^u)DFO0zE|{Cc4&IC*l$jznEb%&M3^vjB}lLtE#_+T<<hi|D@eaM+^nKRXD1sb
zQmvH;!u%7OJlpRH)hZ5ER~GT*WKzJ;z|-blpwX~_P1D|pJ$04eD~`I$y7xhi7JHrw
zUM^l(*TnV;$TW-oQ_buy-XPwJ{5%|v<>0e83@!1-7)G4T^zQDN?r6itp<}+qj@6vo
ze2>JLZ>&jPNH}{QG@Qg>Wv)3(ASHH0#)02Rrev7UDH^2nH%N!pH&x|JD5OYb`4ZrX
zaw0+y+9xpE$J-DmO9o+PUX0fDUQ0>h0}`MIaD?`k3>rqP%ZJ@5WoIYZ)*B@|reykh
zawEfE*s9Jj|5|hd3mH!K46N1!2JZdAb64Gw$gI|4|7W9>KI?75|BIlvfAJ0U){s+r
z!MYB5MGUsS`$p{w<p6hFj2wOUVE0!B<@Ckruq59+Xk*g;&F(!xYu~`Pt2}ln?7Gck
zu&J$ZEGnP05s*p!$J`7m_Lmrq<ajt5#5zVpIc_vu&h17+<-*ahl3R|3O4+)rEB5tQ
zf=Xpoc8k&2)K+seVznOdQ-lXNSLmNb!G0pU_VFfk2rYsi1bEB|OO?Ai^h>J83NyzF
z=Z_UGPqfN5fteM?&FzXgDs)w_)(vWLBH`-JOiMm=>cwN2&$pgWzu0Q8g^S>_wWqCA
zBje34I!6ZhG4@cg#R0L5o4s<Q2h%vrk)7enXW8pcXVp#{=ton+HF3Wo&OQYispl%Q
z7@Ic|X9@j(sPoT6{$~{X7$qF66Mh|VqMgd;Pb>fD2pX}~*+$Icce;uvxf#C+0R5j-
znSVe~?9n@d0~DR=;FU`Be@boihA4o2STWyL{7m;**>4EB+H>vvs!mH$D1FJXxcSS0
z4l2YBBDy~)?>Rh;#S)12Q*WGzAB*jZ-T|V&RcZpB&D!3rPF3rUd)rKwH-$gquLy*4
zm`17_d`OE+1-OG2HL1>^oUGMhErT|p_zz$OIC3yaW`wU3Vo1SIqk)ZT`o>b7K^eg6
zeh=|%^?TT~l84>JQw~a5DmOhB;{G|(<<sR6z%c-IJzv?T?M=!%A;<ESzGi<=eQCZo
zU%{Q4oS7};Vf`ljRh+2iYkg1^WsCJghg_gX15($)#h2Wk5fAHITCASO?}z)ok0QGZ
zzYk}v3xodOrpAyc7u0w9@QQ-tRDBUW9AKm2A`E#~qTW9rr}beZdDOT54EkT_tSRp)
zUh;D&0Rakn=aAuH8S{@RgIhKFF;pMi*v7UW65<OtFysGomm(!4o>TrW^E~R*j_QMK
zZy816pl1qg*5yhOCzfEBcz49Att4?xBaT~doM!ZxXtM;BR9*znm)8l*N=sMVpR1sW
zzpS<WD+CkU9b?gWyS8g=|GcVvQJMZ{6qvn%iC|3dVged-gZNS98VP0_bDEXp|FF_t
zAXeNag~?T%m9XQ%TyY>VKZ~RBZD<&@(Ou^M6itKc)RQxEih9Pwm>6#9)|d)zV?5BP
zX=7@X+@XOEoMThWIBv82a!R^{K&@#8*z|>|9RKHeT@eFAIVUJN_Hv8h#xd-S|3mta
z=13J2Giod@``yG9%y+vvs2W1~FRGs7+xt-Fl}t%I<{_j;vU6~Z#+3FmJPx_47$Ig7
zC6sxjn}YRV97{s#81-qjC7qJI37l!iR=eA<yAFPpusi(<Rr<<D2r$+>0A(Jz9$0bJ
zFeBp`@e$|%I*Q^sL$gCJEuhE)Nro$9cu?}CgVyi0*hW_QChW-JQc#X9tTZAMRx143
z^yKS$p`%BP+=#m4`lN>NA+m1djWlUeqjQir{-3H2-0p}kj$k7b^0kUXIwZwwlFM=?
zwB}zm4pk$^)CIN&#wn3RLFEF@uQ5xC`Ove#j6_vB*t!m3TWW97RQ1sV^rSdm<;SXz
z*k(c1W=7g=N8A68dRarke^T&Uo8R5owYj~Le8-sX&lAl~G3#xeI@Ue&Z`R1|Svz7E
zG0jGIH?1l;4vZlycFGMW<a0pspr*%yBZ}!vc%S+Q#EJlGHc9k0i-J1nf3Z-tN+f+S
zo-GHVC^j?R;HZ`LK3FN>0jozxGdZKKdwFpv(~x`!Uw0L+e5FmL&UWe;S=|#YeW&Nt
zXTiB4+evFXyT|f&MjScnsMGIJhyRE`LnlUs5rVQa0wx0W#J~qt7xZ()o|TJ_LyWv>
zNeD_zDsXRVNO;hm34vw3gWocCop@xK1t2i)>E=Kw`CC03Xh`IYJjKioSR$yLr$l6-
zw+^HV`y?p6(1+9s$Ev8d#w5H%0anVdsxc@+7(ztu5aM%qjtDHHl7VFu+HjxZyB#(s
zjLdALOm`G?%CspB7yD<*GW@IsLTq2(sX1vu4+9qlAHGW&pHR@JAgACfYUK`MFtQIR
zE?$(K!gw0OIj@`brqmznxe}*{{?w+H&FTsxrEY$oDZdr91jHmf4Els~K_f>c)R#Y?
z{n54U-Af`+h3N2-5U$_HXs0?$0RI?MXV8sQWz^zmG*|l=NZILwB#AQxZ8_@2Pyv!{
zJ>DtigFFrVYz*qsjL_g%0Pa!n;dX<1qIw!30Ne*ysNx&I+1saq*rQ;t0)|K${#|gl
zIn0cR9PNZ#=CH>0*QwG-2=Fy>0nF3k_>}VgU$xn}*&97LxeWc0pV#P(Pza?5NO}4C
z#cd-7dl2%m&D*K1=B9(P$^Sb_b9^&<Us|J~!;?DN+Lns)yi485D7aJLH>nXSgZ>xw
zSf5nv(OF-*0^7~s#BUmtsiHII2pzV#f>*;6L9Q4O`fXyG->zcE$=bnzD$*IF3!$uK
z5@3vN4hXoKeBp@l;1HQ^Y|AzaGc|eDK;nR7Lb#XWYbM|`!B3JSyN#@ux*pXCOr)pY
z!b~;EwI>3<P<S}Ub?m#J!nc#ZWNu{TEz0~BIUnR$O2foEJ}rHYV{*}?t3^EAJ~f+~
zp08DpBX{89oTZkJCG#zBjm}I}j=-DbhM>TNigu9m&DWf#e2+_ODxUm4na8O>I?`$=
zBtp@R`nqnZoVI#f-H<4qDwyepXxI-;zNqg-CcMg3C`VzV>I3i^CC5{IMo<#BcoM#}
z+Hyyj9cF)_#q$;dNYkV|ubeKXQLtvzPP^wb;qw{)J5(&^jy5~#zFU>;;4?h=H}tqo
z!Jq=^h?`z}$X7Ep?^L(}i|LCTS@QMrMG6-Cd6H&#=x3^!xv<E}V@kVsW6mX)rdPam
ztzAK;TxJ)?mngxt@$TcRReVOPH2m2~L_3fsBGumojI-)=Lj3}JE^Vr&po8YZL0nF$
z?_;e2D8kDNuv&nEeF?uai}kr|E%3GDB{|ztW-pQO@T=_~$M!s5kq2hDXFT6iPxCxr
z&sm-yrKKb-18`_5U%^}VHQhD=i3KTyBS^33)%ZpR`}6$gB|O4r11+5%<Z->P#ZZUL
zH0{kHChNEY1qlW8Q4&5#t#_kOHWVPKeQ^JDqej2~H<~T<(L#S)3$%<QN7`8!XiF7X
zv{C?b<+k+;vNrDxU=y$J;RXvhVgn-))g>kDr>xfp0_%0vA6A+B)K*wxNd0J*$jMB#
zIikN5j%agEhJ!iLEg+hc1UF{1&xt$&#33f-wo)pfL*588Ch4Y|(``Y0`A&**3rMm)
zV)eNRggJ%ew)*lGFk(-mR$m^v)AeC&Q@UqFbDOS@(n@#E4$IlTMV;~Hw~zAeV{M}y
zj>&e?$G1oyBYj6(y47_hw?|x#&2-q#j?H5`G45nzGvPi6|7stI3G$6P5O|de&Vj%t
zxpZje-e&w4cpx!s$>}o`bPznO7n(bPw-Gz*?^NP01^WordYs|^+|&{mLJ0FTIi9~?
z1r8AWwLjwThachp9<BeioBNrtxwVYz`IS>^gJH}G`oYfDMqHA4S8AmOUJkfl0Z)~p
z#T+*Hhx*uq3Z79~zhYwHwR{}-hm>Z+)519IfHs<PnZmdZMS-l#^@0Ce<V@^z=%pRb
zzpsXbO8CF4*q0UjHw9l(ns6uluR>ui28;$X6vq(7oFF*Y2uCSYl<hGGIi`k2hfF9f
zWuJSd82F9~*qfGE>~C+hbmde~_^#q;M}~Q15Pr+8hHbrYZmOCCSsd%gymG3Q18Bpj
zmG*}%(m4Uefgrb8-a(!o4S*4SLdQSF6j#O2k#FY#p-iwVga?N18&a`x{8`E8TZ+FR
zRk(7HC8%=+p1}TS2$m)GN-_IOd^{#+;A6o*ISk*dJR5E)(?^~W$P0f$#^qxqIKoV}
zd=O_~)j2Rn*oxp$sL31IbbuTr%4md+8sU@?ZYJ4`pxG>*RO(riX7pflH&O3_674Z|
zp4IX1qqgLQaV!Dlg{8M5c!Sm!u<UL19-eWxU=S?fzYF-JTfkD%4y~vB4^UDHKfSOL
zskz=q*e8}E>8Ss!RV31JQU}%tIl+@KF6Am2hsfgQ*VBGAc<!}l%QLTa899Oa1I?11
zB-+XDV0m3JF<kuLCqRS3_L8<yQOkIK7A<vzxn<i3Zd3h%N*f2@H`U;mRWIPYH3=}c
zL!Ku~ir~gFYqDpgF+3QuZdy0!;J3%I)=oA0?UfQD!!|?XUe}-m(US9-QA~)ubDl@!
zA^@DuktITuxQP%gjTu~b{SS5Z+Z2c#@4}MI9O{>}R*#XHU<Jw<J+8T+Q*u0+5;qJt
zvD36l9b*j^-4<-~E^m2;F3$AQZ1}UCNDsfZ1V<USEnVg@<u=(6Gb~GKewyP!7GItv
zngwY{`TMvk3-v%yAk?z!1Bf&SsWr4dQc3a@cTY7FI^uCVi9;xy<*jeS5knmR;qRy!
z&yMA!6mjjg;T*!;{)lUb`@I~TXKqrhNf?P`kcZ=lru)57-oMO(8^w-e2t+1tF8H{Z
z1JF9mg(yj`{EC{3w~YTSw<$Mb`CGOD;gXxw8f2gxH`NYWEhGO=wATJq!JjGkp9=n*
zU~S-8cllPqLkPv45GQQ#{RdT`#f<p>)W)|fFm_{~L(vT<|6@`cLywm^qET^yQT{yl
z4@qh*^w()Y_==6tB9~tB77k{|eE1WkPbn}k@}DUIgS?X$(b@|BpQ&W$hctR><t6Bl
zDb_xhSGwaid`I&moe8eiwv)4`117dx;D4_4|E2FWT;WUhsGyTZ_htF_2U_m}eB%kc
zN!ULa$X`t)MVsrC7V~)se}jqGL00@On9p0Cw1J<uK^h8wG+J8myZ}C_h=qZwBsW5J
zIRMuOLG#3w+7tX8dL>ffK$QS|o%o#Gm$0ru&SG5BX6XrXJ_%}0JuEEc4UAm@@64PS
z4kOp;TOUTZas=-WYWc-l>Xj%)-!cEmPsjs>6nNS^P*g@NAEt}%Z1FgFL;n<&q}q7J
z{OW%)iWEle&BD(HwMWxZa+q2ztE+zy`45`2#58C%KK(1H_SE%lwY)C=sr7Q)e``Gp
zF+a%e%htD{^|`h_$~W}Yhv3->eNOV!&yGo~3wa*!|32a1!Vq5i<Ietx=i&NxNEW_O
z9K&^?(%W8%*2ik`jnP*icSQWq-^jvtP;a|vaA6E^Mjz^<5E~LCbM@q_Ie^sg`i?+?
z^*avGRe#O#1(7Lb>f4dvjdPhE>L*xY;lQyzY)-a@>FGdVBL6>#4|@G(Apa-z5p1`1
znA|Dh-{hnkHI?NQ0(Pm=Q12Q*?cyuMY{zqno};P%g7q=OVr~A2JuIqIQ2Hxp34pMX
z7xzjjuomH?SRmq>RT%`u!)dDI<?+U-P#~`+Qn{KM=eYyz%Z}IRm0Fwjnmmzuv#F3}
zr2T>dt&Si57#VI`$S|8c4eU+OMPp}2)*kI9qUEy&+lwk!+yzjLL!=GAZ(5;~<{Sv#
zh&!I#G%v&weQL#jH!p#wZm%{>v-SQ-vi9ix!faBJShR3Um!PaY@K^3f$MEQ@MC;q|
zj1e(s{~66fIO2iB|Nl&;9W8Y9M-~4edgjMzz#TX?T0c(4zHqe0B7gZ_HXZ-p9DQ$V
zU<*%_(JOO`#D7ok{u)6eb^hu!8UaQ*or8^O0HXyfzNY@Yp`dl38p<Q^Ffs0KWV@)=
zG!gE^e?u_?4=t{#rzW&g{aXrtLBS6w?JJ6{C@?0r@v1+p$GQUJS{v2TVqKdNwsrHn
zC=wDFHsi(?e><s8b9GwU*7}x49<_{;lv$O8hjO`(wF7^v@6!ru;QISqBW;|t(mD5;
z?^KWervi2-v!-Gh-8L`<o@I`T;E7lL-y^-4gD${;zQ*0rl;PzPh`KpWYqmAfjcr>l
z`y5kmJnuwO_Hnwk2dfzaEeW&}B-s@B$=o^`m;QDvHgro~M0F{QJdNk$$TM}vkMXNU
z0ZR#hk@UcFU(F;XCq)C0dQJw4cdbL;&?cA>$r%1@C!)zTjO_D#Z@o8wSCZ<(kcC0y
zpi@7pLxMm%0tnP+@Mi${BUPjU*!NtSW+e_$lPl;ra|J(6z^ElIO?^=HhiYi<*RsHm
z(AloB%77njsUY=EcJm_G8EsUbwlg{>x4TY(jBDwdaY&RGEq@5;G7k$UjuXXxpv!Q*
zk9ZzpA+attPOtwd!akr)2H0}ho%=W}83DE+)4|nB6bw-;s%U*1u*EESb$zSJZATy~
zqPi4z0Z4(Hk|ba4dig`Jgoc1JBS4Q)paYmD*2Y{UCl(fve1M+x2GHXtiE|OML64s?
zJ%xclYvG>=UKsS~@qZDBMJ<CJC7|xFh@dfyaRBVFw+)8)UuE*lSFm#-cCQ|F9QeIj
zoa4+WlFHUm(Ax3a-B!4!vbM#A5Bav^?G)H*pWCkd>um7=Eo^P^C>`_%)z!D_^}?8(
z-IX0&1A4UG-LKe?f)6x1+_{1MVdegsa%Hh<o7dl_m_ZK<>`+ed%15;`Cn=jqISos@
zsxc-@Ri4>Xd$iB&9<2h;vTzm8*zG~~=o`9WE8Knf{(VX0(pLJJ9X^oF`^Ah1pzuc!
zSkL;cek*}3cqR{yr6_Z@pnhi)>$w2EyQdZdy~6~?=l*9y?*h0lPd1;7g5D`#))^-a
zD`-Rhy3ji|DcAzsV9*xyz1gJbRw7HuvcyuISOcTjL{9t&Pj7}+_)!gd6F&MeJ$3;`
zp%LHA9QX+`*Y4^*w;W>qrdWfYRI|bqd^?6)^D7i#U2V&a$=W8K<L%qNEhgY+^{&<z
z<@DI~MFRCN>8(BA*jt;B9Z21^2@E=Qu7Bi~AL0ehCv^)JPb*F98AU&%w`^k>-t_g`
zbj06rji3!)6PlTw%_`qg-KBY#<!2kDlAX7-ZA37wjcz@29TH%NsRbGg=V_4xZ9jZ=
z)~o4^YlVxTiOotUJO$rK1P%%{vq*xy7@DFcSO<BRPe#_L9fa_gAlYFMh`jk+e5*37
zVmnpNp2IY#Yh?7r$VK&mZ6ll3PzGZ3W(8QH@;aMSh(M0(DNKg~jebku`U0$K++aSh
z!4x*pb>?u}krk8R>Ju8+{k&I4hVJVDr~~o=_%h>4=e{Nbd5gz4i>O-7bSR0WXimQd
z#ebIw!y?#ws*YPBhS~|-`=Ko{&I(dELaI9~Pqg4(U>F86K)^)hD3(cavt7EQF6Dj%
z7o;PcZ<Y|HmNA{f$A|%onFvZXMYYxv)=NDyA;**)@k>o_nXT#acGArG*dT@|C}~TJ
zQBx_GVd2}9ic`*K<RM#Sy7Og*E&(rohcw#%N`Xi!%)yqu<+Z-ky1d6Zj3eTIn0$Or
z*~ZaP>rmWP#eP^-{|^N%miWi?_<l9FM=DhvrkNc|c#ailDDadTV``;hU|UhUkn5b}
z5yMmp{eQ1ke@KA=DXl=aObjaR<k<LIS{jE*&w%MkmrdGkvI0;jR-7QyfNVbEqql%I
zC%Sjv*Xiwt^zz>zc>N47nHeAme~uH;Mm0e66G>K3zi)ht3I>2+c6O~jH^Rl7b2?-0
znh_CORGTpNYU+igZh2eZwm!PPofWig0p|-)-^@Fr{noQ{Sho%SM*@d!qx@y2wR!^W
z*Dqkt3IC02hILs2F+`OktiSDbiOA9qMJne=&3hR%UYA)-dJlI8+F$O&n-$lP{2+xK
zuLtvmJXl<-1?kkn5uk<dS=+UB8(!a=wSZrk!2~Pw>fe@c7<uokWG_c*{q^lw`F%d@
zDfjZ}S^BMm@D}9yuAgts!yOo|4dIq=td1sZ;FfQPY0PpP9$4)C#)SaW1Z%|Lx%$iS
zBoH*}`eHaLe^wi<ZRfh)acR7AHF4Y<;_KjBz>C0Z%=PHTPDHEYl-VWT#JcWn++E*6
z8#}#WO723Gh}-jxz4cw!X#xJkKcZ}*K1SKG+P?ajH^Q9lCbrw#X0iSC-SsgypWZ0k
z(%tple9LyeMMeeR20vu%TJ$rp26@u~aO4=wcu)#&$E!(>?LF&zJDmAtPjfX&^-PC5
zKlNr0@8^atS{|q694$XZFAvuDlD@CLkMy0SXY|IJXpF<PBiD4#WB7E`J<%I3C;rT~
zO}&YFyIG~T>(AQe{)FS%?AIRN-|OwQ+X??YeVgD*`@s75(;Kbdf?cNkckBD%c#U}b
zSbO_Fr(FD1(W3+1JwlkV-#hqfnm!#^KWH;@fOj3LAL3iNY0*3Uy@}Ui4{?>MhYn!E
zd5v|W_is9SoBeFe{{*~ex8w9zJ3@ci5%G8|zJ7T9$of%c`3`z~q<$2mtzfOJPk2Y_
z6X>qO^{XH8j<Q#B^&?O)#7H?Bq#O-XX5)-OP@Y;7v+VKp3BK?Sz7TSwZ3W0RzSg%n
zq@>QB_N9B)4{rLNSoML9@3GYo_OLnn$Mj}K<~Z(J_SPr3-dyiJ@z;9q>NoJ-lf1Ww
z5&mi3c8a$ls^ly0G+OW&vE%xk6&Lq3VyW2r9qV^)@r6Is`KE<~^*iNdjNNpsevGfW
zo42HS%dgi?Y@DpY3*I<&9TbThS!2P@<0K&eSDG#?@A2-%g=JcZf8fIMzDfK>`@d>1
zM!~;R@be1zMZKJ`(PwukVLallhw#6o#JqIeLX%!NT5?kB|4Rj*RYg&iuj^807=lkL
zK>)YWbAEMciVH<dVSZbOo_N@llRdnsh6ID1RKL^y-j`>`M&?-|NU#cTB->E6?Epo8
zjDS0ckV+eR^%07%?F@Zx7Ty<F^?TgvX>tP(0Nljg{ALUO5w#_(x;WynD9Lw2qkQhl
z5oLU{7aN|c%%&^Hr#x;Nm?`lG0(!4LdV&^i9u7O|uE{i;f%(~aRKaHba=9?Kic@23
zS#P9o+lj`mPJ|+?f`}3KTE+3s-1doS423<PnVoaROE=>u&dfStk0ASrwSnf)JXu-6
zV_OXZqiZ~DC^W2&wicC*)uUmkRoOkNmP9|wSQ~7%_Z|$5lvNxK%Sr$f6u!BwS+0#!
zS!FF5)4*twyk7Zzn(qE~e%YJ(M)jvC-iA#Z!8hGZGt@~0-*HeDUj*D{q)5tQZ{iH~
zv~wmY^tsC|s<ZR8?zi}uhAIj-w2kAuZi$L-WZqrAYL|-p!mIp#rKPwL+@G6v=%%5G
z##%8JhHyp4W&Li&TG%J(#3rw**OeyrT+8KzzeA(>zbnn~T9ZYKW$lZn=L?_o|CwT=
zinXZ6%StP&+RqW2IPS33yOl7jU`~MuA-D1HrUu!g-FIGT+CY*l+q5?ZS8#nB3a`yM
zGMrE)bxm)zG3-jt&(5w^<?@?3ulmR6*==`&J63ve@?<Wz?GA0HyncgzH!{jl-<+!l
z!>c8-@PA$1G+f&bmh>7lo77x?U724}Rgz3W1ZPRj{>Vh%&X-2-TKfkzvfPL1c7}QB
z!18UT(G#wJwxKB{HhP<-N=h@%)Q_qoG77hSWSnN)=(o7|)xg2OJl`5IV9VN8wET+x
zh=_sTr@wtEm{MX72mc1XI*=6I=uBl~ER&0#urL-q2C0CgJ{E&Pr4qPrg{dWz;+CIV
z8O-EpPd7(c%fylD-`aC4hmo-}xfw-IC53N&<;0@-H}>2bq8B1X<Udz+l@68qwOn#2
zy61nTvio6@oiQRsWz^b`^*~N#+Eer;?z$|s+NU?hWW;VR5jLr0w)U>}cJn7qO#8>^
z-LDb>0%Kj{yk)dG-0e=2f_cI}12uw*$wC_S$l%iYdd%u(TwO_DMKe>Xm7=;?X$}N#
ziDJ6@d;yCzumPM3h4C8xBh7XI<#k|tB6vm_#-8=&et*r}7Y$sBR7>JQ*QBI4V4iMJ
zgb@$8HUtHnS{T+v6?pm2c-hZMLEdhu<SrI>fgCilSRrT&tR3Q_SUDRC)Qvg3&FPSV
zVbpt_1K@pY`<mW7orj_J@(lGQ99S%9!EPs5D?D3vS52?36rM6s-BZsLs)3_AAF>+v
zI$5rnaALehLaY&k_gLFyo8^$1$oqoaQDuRU+`C+vxjHf8U(kD|nL20a@B@l{RB6vD
zCSx#j3G5smNi)ri+gkpP{Y4dCQs61KtYBCHhYH6-*a4v5A*DW|k`;nt&W%P1{W*VG
z32B*0=;@t0DQzoSU|Uh%eNQFXiXxevKV)qxT~LPNR<H@>QaZ3di{<0m7>Tdp<&Z72
z**_Wp9WPs*4wpML<kf#cBU{`HM4S<tTi#;JRL5$u4b)$pc05D|*7w)But#QPc&SR&
zAVs*XP|Fwncore*D?DLAFV&V2wbyf9p*`>9A_-7w@{OS)3&2HiIs7D<Ea6jhfQe2y
zM+K(ZZKx6)$FGp(SSNQVW}*XwmIlk7(qQ_^FqIK`7rS)EGngtV?-l=dG$UQ3eHWD`
zHQEBHjt@VC=?T1!O`JvtS|_;P;%>^GJvTXN{W^QrzevO2Y!C29E~oVcZe6C%o}J|1
zM2tehh>fK?kvvN>3OD>AU1*@8QzZ4Bqj=D6kjCTWM9<ktUZ*V8b3Uy}Q35wJHn}Y4
z{9PJ@x6tC`xpQaFId{2}=O*rOOS`5ntP(WRCA^c)V9ohH>#NemjB)6}B|ZKv1xrf1
zrr6604EBFoF=Jx=xMDw{;2$dZMFsaLxUS&06}+n8J_RQf{Go#XuHcUp+@avIf<M-a
zOgd}GgE-6%B`}=BPz!-5bA9b=&ipHSY)$63r~%tQSILlkr~3UQWg>u$WC+N@zkPq2
zc`$~A6u>mVwd8MPvamt7;nBJWU)AYMUv40Ex5^IOO5F{(>q+I{gvIe}-Ip26ByxSR
zM=WozYcKysYERg5{%fiG+Tv%LB?bXP@@6{L%8h&r7s~7ziu&sD44v)xVNKdk^%Q?X
zuh=cMi(*_ZF3RvJ%L&-l+{}U(=%jg5mN8P+f01e(EZaB<34ZlAxDXOQo6+lFf9%Yd
zINx~i!S9`v*v#oK8t2;GwyWcSj-3J>@?D4DQfL}GS|z0}HE9^!0R~o3`+19#2p%AI
zGv`zVMC15JnH&zu_o{>~zi;9Dag^BAUP(leb1sPUnh=dM1zHpE2i~`E%}u5`?ln)q
zawu3=)3zdPf!MOpn?9(Gu5~VRUtt9pW$b7|sONo5qRM{L8|-Aul&izVNtOnh$EAP*
z0|Yk%Np&wA6^=C&^>J<^qcF>n>Bd1h?5IcK(!{q1*1U)wR2rvc9i^SZDp*7HE*-jZ
zpFw`cq~5YJwDyq)%Qmp}4qYGk;jP}$-iFoYeVb~_AmnAoajni9ey~Mvw{D4A8Km8<
zp2&D)BWUY0uC3NrwEDlPUXFTykFVt-E3>8yzf*ivYh8Y|xqA{>^gd0m9kRO_eP2rr
zHQOIls>z_=rv(;s2?eEU#rax?I)hbKsZi~<PM%e!!IK|RYtQKM`;_l4hjv<#hRg3E
zwMCfw|3Srt_cvlQTbeU|M9b}QDw&tO-MRlN?s{T8_hmuu1nSj^k=%b*8nJ=gA4|TQ
zXGtb{a(^bInpiq_)8xd7)EnIM5^jQUp$68Eh(-Ye>8>N*ZLMi_*^b`Zo|if(KA@gz
zO=s{1v@7in4<0*ynkSNs=)x08$A%aGoKpznG<rG5$)wXLQcRSSvW@kI?9|HQs~-)$
zV%nwRAHFx+>kWGY-r&pfN1w5?OQPVBMgV%BTQLSj^Sg0)^vvW$j9V&}F9*yBnfmuT
z?af8*AYPOsff-G*est35qjc(tI<5-8kh2OlQ37qZr1Z&+wHEXHv{c!UY*It8b@Iu7
zTcm^_B2ZLg#k3Bsk80w#wYwd)UYTpk=e34DKoE##8Uw8!J8aBv7HM;jy~T-jO%tON
z3t^1WF*wK!oQHIePV?OA!0o~l|1sL||2M7J?<Tn6ytbMnPu-^#4yQX^kE-lmmHh$5
zq_wc6XU`Avg`c8>{%e$73vAyvb!*EZI@@ugihq&eR})tKC4zdg+8<1mWIaq5@}2~o
zPOdhu3^j8Q;Yu20`Eip6*|qWOS<DQCjKV2~8=NqoC_R8RNe3|-sP~`)3N^+V8a1OZ
zOa)}Tc0I})<Z(pOAW1Xdi99i7I^TIt?lix>Uixzw2cTCpO{pv{sAbaDk1|7_1jVNL
z$;|+>T(_q*qn0utU@`yKG@d`NV7FGsNySbPG!p0D^Z0vEN$fmbxJT61n&DIC!^2jL
zM#yn)?8LfDtr%APLb&9dgit)ziGemwi$>M|w<@kaxGVTH+qY$AG?G^k9$#+6*0@8o
zZHG0osKS{{vMN@E*3NC5Y<!2-Zh`kEup;S$z8|<HC++g!Ai?4QB1heP1IYrlC&we3
zmX4&0W^LMr;=cZi`Vt|{9cpEFlAwLk!*BHkR(-v<O`M2Az9w{iK;lG%i_TUAAgje?
zODiLl=ax7Tu@s^N#|Cc-6vJmHP#lIiFM;9+e<e_ql5UE>+xZIty@A@9yPyQ|;+QBO
z(Le~IQTmc6a!Mi+M0sBe?Pn1=L#BH@h#=1^wlSdlv{4S*Mr(OQkWZ32XadAfLaM7L
z5*O-ke^UfG7^INn4Iow-1a!7qIBG2nA%eWIbqn6`o7T@DF5HH2@rz-fYS<z2{IhHx
z-U1~!=#B7I5~CfcWf|uk*f@s5#_`%H??*U^u<-!|gc1>s1;WPPw5RRfHVGT4ovQ7q
zjc-uK#Eqkf8@KZ%JR@F2u!y*E58}q1l-XNH_=tEhvQUt)vA)Y2qa>oojeYbAA>wr>
zZv5AjJz%2A9kqk?9o`O(fjz|bc;gm3RNqtI;bz<03H`c<XM}oujl_*FpqS1z7=gI)
zFq%?Wj+(e}kBJ-it`|0mU!v7VJH;>bCJ-yq@?Kh&Sn&mVd8A(88}`>Fv@4L_<LxtX
z-~S@-C~wMx{vEJ&d+?PAqxZrmqiAOBw~>0ahIr}~X=-OBiu7Nu-Qk`wAlar8)o(Sp
zc7UVn;49IG5QAP^r1k-;{R7R~;|q71DDyRyb}h8ZH%qy9{F6Wn?4Wn(<qU%1<{LG)
z&<(BkSkGF|jN)NN@qj_xwoyd&@Su0ZM0JPO58J#QVnmK0HXQVhG9DAR6-xd|#~4c}
zd5pJ6FgRv!LCnS4khtuGeobU{r-{stu1{>5vsiUkM|+V42`Z0y$5|0~)bF4-Czu~O
zcsglEx}D}F9lRW=PT1w|;jx(DT=dsQ73y#-y}j-Htlo?spHcbnM{z2FWN>l`3ZlGX
z0yf7i;0kI)XHKW>Ai0j_HIha;Z9k<$R@F02t+q|aXp5)Wrr;Kuy$VwJd-V|o1)tR`
zj1RL<kNXu!_U0c@?4Sa}yuZGL%|E2VhZQ`p20o&|m>qld4#~DU#(`(U)Q0(MHvoj1
z@WFayl?R&e&0zqnnkM|OQRPO=sk#`Tvl2E&KTg)42uCq3xqEH{Xto|9e52Y`qW&HF
zh&vS=S8zhXNd>1AG=)t5X&xs|egop2Mq*(ZJGzvrPt!yHhX}ZoZgpwpYMURH+lhW!
z0w2c<`2{s`QNafk0Q~}r=DQU;qk%VtTS)!Cq@c$*FDC7oQd6%f=oIfv+}Va@Y$4C_
z@770}DCb_q?o)8T0tsXM2NXQ0!1MwoY4P8p$M03_Z>qz8sMxP7W<r*aC|6e?_|pm`
zbMYTitSN8t->S!l71+vqn__QQ@IeX}Q>!zj+R}>uh*BO^@I}4hF~wqP?QzB4p}^MB
zdBtooi3Rn8+Hjh#K5Vxn>4-%~Z^;3mq#qgl)`)a#mysM3wqyTVvXMO~_J(({etp@9
zq#!9|8@cFV!=0$h`;HQdCg_%nbW1z%{SsUb@g@hu4QNO?|8HX@mi5`%uaj+z#N3ut
zrU-#H)}i67#;&_9(fJ%TKBkeBznV-0EHIP8!EkDuV~>rz;j6?9nH@jWu%But3EI3W
zF5!WT!4PxgsL0S>T@QCH;$<*P6A3B;6pt--C0d>f&hp-n*Y6D&37Rt!w1@M3xDmJ-
zIy(=Wezj6=(x7Xf&K1I+N2jXu)dEyFPM~LJ%RVS<Vb)cc!h2_Rb>_VJU}?&?3o2S(
zshvM#;^9_H=a%J1THeyb1|(1seoiwukqKHYn-4yDt)9e>idDZ6TRU!4+X?l-i8|f0
zA<m`g#pRg`)f1=q=NRfiafvv)tF+fRbt>S$OU6IhydMkly4pdy3Fp>%WAyb)MDVeU
zkjt)1Lsf)0g05@27%eLt1}0bv$~zrU-j2$KR4btXYP?3AbcNTXnOYvIn~2ZQ!o7w9
z*SnBBn0lu@`wxRF<V1E|XWNQO4|tRATCqpz%-X~ZH(e}O7Oysku%K(P3R8tS?8UfT
z?tFRX!U|_`jVs2*hjhnrlgw}Qb_~5=VO2J=^4pE;?4>IAa)}hu&0zu|Cilf^x;;%^
z5@nxhmvDuuzG}u_CMXVBTg{iKQU9WHB~@zU4SbWDcvQg_T!JCE90oG;J=pNW*9`xS
zbMC-gx)oAbXS<hklWA~vMza|HsKY`<9^<zvsR2IPVaOS+gox3Lqu85pqD$zBVaRhS
zWubB#rpk*Jxax;s-NpDjC_QVGc~R^2PwD<8)NWJPVf%okA$ELg<l<huDIxr1wl6_A
z6QkWrcKsftx!s>K4skBRUdo!v{SfUW>sdsqDxc$T`ZXuYl&=}x7L~d92(4=6c&r@$
z!r|d<P&s%Ukop<a2{rkfH6=57Wr2dbxU(LwzC=!*?~?D7SVev7Z9UTyd=K#cd>vYj
z3;Fu$J=YZLBbK*V|7-Y$Lab^z-T)@xHz+?S4>_;dEi<m9<@^*m_!0<nnk#vD((nTR
zhnWTc3krUOU`;M5a1fvqg6&v1vbOVxd85}#QX)qA95nNDcpP0UaNWz%ne$T>+yPf7
zIAn0QBy|XRg5qkb32)xByD(UQUSUZWZ>t59SC$c?>ZoyVJ8{<kN%i2T6#PR4Kds<r
z6tF%bYkg18V}nWMcXLEMSXdi;u2R0Tf*%3=Sh*zhq3=-fpH(oeK>8s5KT;s%*Z;=~
zY_Di*`2SRoE5sT(WP><HU7W{5@xM^&hZRf_z$Q7Q_`g)}r~)ln|6eJv)(su}*E}||
zmGb4lN9B(x`&ShFI|ct<fq{Ap4AfsBZ9?1X?L?%4;BKuE5Y`Q~v6n{Tx(h{067Xqn
z<zlgy22gpxj_xJ-0nXhzXsQdw#l02YX|s9z%~x>z-nyty;P;&0guERJ(;7RSw(sV*
zDy^K`D@?D;6|THtv~xI;p{g`W!1x-1M}Y4-Vp4`D7K^R#z>GRsAB7H9PN+s5K0B_*
zC<>O2%P}m~Xx`ARMxB9&@j=q1Mh#bhw@K+Z8}Kmj9L3&ULj_vx!tA%dC{MSkQ3p9*
zFFcGs==oL);$h^yep>jRty}O0f);6ci~1RM9g4#!Ob2~Jfm+XB0Fn$y9hsgE+E)#l
zK<6LuK6n_bk?Jo}Vi?YWog1Wzhcxu@*sQshq}C9%q-K4xHUj5k9A(^)^Rcrg6F<zf
zjq|aqE^ftkI3Eals=sefquy}9`53B=)pl%**LNA`V+78}Hol)aa6=*sC{J(fhV!wD
zGJEPf8Qa~Y?iJqy#pzK>?m;EHU_C{(*%Yn+BW3p)mt?%Qzdr75hu^W6*j^N<6+2Le
zhvH@&<!MrQ9yWDoaXyxyd3Qoh?=Wf_&Ix>kZMdn}^*Njd1S;3NjR#U#-?xdLj#i)T
zq^B9dkk>Kn?P0{kD_WwzhwA(IngjI%yj#4HUKE@4e#U>}FmIB|GjErovN##82k;68
zy?yj<;`Y5m;rXz)-*^kCBiqOv;Y}0u3Ep&o_Z-~JTc{5|8!?6PAMfZrsS}4<bpNli
zRUOY($Db1mN!`^r08Ul!FGE%1B|5|!oX=SAFGXiV{kHhWThZKGkgyX+z5z=0uRx8q
z(3k5Pk7*6hFDUpcQmCI-E1Y)`Uy|20EkqHgs5Te?xLbwZoc#P#6}g8x{$G(0-9qJ0
z3cFx`e+4Qu0~Km=bE8LTqnIXT7VDjEnzRctiLiR7Ak()o8ctPj5AJiM5hOD6&B)J@
z*nUV>r=Is_B&cp*hdGbBTuwx;5wIdc6GnczToG!yujWz+rv5yU8~U?EF7!0K8OpWT
zD_FoSX$Fj{=+a$jg1t`Nm3%kJ`Cc{fx+e2|S`X6-1g_gC&^ISO7il1`5iC&4G(^&|
zcr1nv)t{pca%*6FD$wkRMMu8otMm*7JARsks3?V&EYuOAUfv|0l7@0idD%%pdo2x-
zK)V#w(m*kFV|kkdv@eu^KIuru7WMhBjp!3mQ^prH2%5iI5skTmdKd0OL*3{=+3u)O
zK8`@KqzO@=@?y=+Mpr=0$ol_lJDc99sxXeXouQpF?aWZHyofkN1JuAEM2s2&F({FQ
zHYT7Zm6%LBr7yI!IWq$?0TE4%5h2EfCTL7_ho8cQOBZU4i~9{+x^hFPzyEXYTRYPN
z5~rElJLkTgd(Lyt|2*%p5q{lgKbopcS6B(wFX*=6rBxP(4I;jmFU?lVMzjZHoZ-WE
z{&(KB%&b+Sb&1@@^N0^?T(P^N2G@9xK5A@nC&ptQW+zerf(4l{6mmAjQNmR!a5@$%
zTP4p<^tYKre;bt2&lF!x4lbKXj5ljhhZOhKf#>>O1jM&%2Zlu0k%U^=)jbw7SYzv`
zJU2qj)ESc}_!PJz3^d`H^$wI`cc(w`SSwYWC^Ye?hPx^2Eft^B;(LnkIstd047#kV
zr1G32)PcempzcHmg@RLo6}sI?@Gnx24k~6R_p@XS7=jWK>nbT3<_aW$ILFn+xe{xh
zL4pF23SA0S<>(@+8rKktKx-qSr-7`bP#|pr`#5>+;kqY~-$V4Ai`td`LWx@2_^r~X
zZ*Vu@^om|>dY;^9`eh2!{8;P7IPZ2Umt{Rp+gJCor=Va-UlO`MyGptG*Zt+?WiQ0O
zNmuz8jZOM?WZ2t&8l#`+SdIBG0+`V020A7_)#BSF+E$aMn&qJpiqp|8bYhS<+ta1R
zc@&yHF@ssaeWep*Ub2o()_c*I#Ipm?xJ%R;Lz4^cEIZ@biOHGEPMZGJ@&LWekD^Ug
zxj;-m>0JvMP2yrt1P^|{WDJT@>ni<IBq3;{-&hB`LRhci7!S!~{AgI_Ku8UuMZnWz
zq53jpB1C8av4~(`AV~H`95hTsE&m|P+a4Bs85UMiT0D*YL^n=&UhAOCDz!O%tscP%
z$p-I!-rh@+O@6#;-0BuTt8Ip;@i6ZXh%Mt-j%#c0av~>fYl9Si^|<#{8LQ_7_U^Vx
zL=Bz{nT%-6nDQYcqI$<WQQbmN!ZfkBAF1f};clnnYs1iKUMLMiwY))#VlgBT<Y5Wj
zbQ=UoGP)9i27|~n-sF|17OK-JA?P6wLlJr?tn|X-#l^|yd^MDCmPtd;bgsB=UQP+V
zE#Rpncb~N<cMYj<a?i-i7gzjnVn?*Z#E5WEUT;tndrV?bPG-!6awdd`(l1-<OOF)f
zP2B3O1B0Uffs~HScgc)DP7YRNvXb6W4&=#%&?bR|npP2#9`Ktwi*FY(Iv7H&XF}iT
z4(+uT2yEi&*zV;h#L0yFm>}RG!I;YtgN42nESw==rh!5X0k3y}0Obn81_se8(>Mv$
z@-kDv!(b3tAWu#>{UcmA%Re0_+YW)C`+txv*v4BKN`Q)V3Bm;;EE1yz05OB`qW;$g
zwvwW~Hi$8}U}GqRje}gjS++Vb|7qAL0vo8^-5r3Uv;sCt1{<ZZ){#enjYtwd9ioOI
z#22EPzKPk%Cf}3Tn2u85_Na7_eDBjr<gFZwb&CZowu^2H`oc_(X01Mb@9y8fk50UK
z+-p0{M$s^5%L6c{sz*&|W0+$g^AW?|FKlCwaaUEQp%P@PBq`Ju>-Gd=Hl<i1lYjj|
z9lh<5ajolOjJu}03uJ=+aPaVmXqV@#1v-CfJiJ1!63~$-+AF`jN(|{3@FUKl_uj5P
zg`C*A5`?1M?w!U^XCUh1ff|IiHe6ySS{o)H1SvClSjInAQ%VJbq?%0L<Yon!AX;Wg
zh&DTivrD=Jq9y7-GG^RfjA(H$-X?;@djaT@^h?l{3xgx=q><0MPt<x}QV;9=V1qSb
zWbJGaW<C0`PBVP?RC~&=%t@Uf>~kIeOq-k9teS{)$A8rD1s*wsHI+Gu(5yHxM8UZ_
zHD(IYy4?-kSYYO^Ym@Ms_AnENe#n7eqUQ*m3fW$K!g|Xfla8&8OwvfFk*Pn^pQ(9f
zC}faAb5ZnkN=jTS7;;7CqB&n`W%=5Gp;W)=={C~##bDP8jeZ;J**a(FQ|1%%IzlVV
z)(#LfkSu!4`Q?3LaOBCfh{zmW5hv6a7pBVfB>)T|x|wg3tIcU8D=<v+DfG;DX>3yb
zN%Rpt#tWmj`4Q41v5r|7bzkjG^C{9l0B9LojgK@iax|TRB_Xdrt?3)uC1<=@q?%?d
z)(wQP*V=Uo2{Lt@ZAtcCwHGt1S<eDSD<}b$5^U9Hn&;2YPZ3`ZQNPdNcTRn7@^@SP
znlASp$|8C+B)qMY#<kI*^*N&rv_ZJS@<Hgd>NF3&d#%0ecUKIQnq=aS+EWX^6MPdJ
zR`;%I@{Ojpgkp_Y^$H~_;q6jic)%kc)jst#p7?*^beN$|EZiGlBKD@J)e6KzEhxVi
zdt&*pqIPE*^8f2^U@%xeWjL>MqzIFGtluKgDT<BQyh%lJ4Xt2FyMWCq$}A>9f42g6
zYk<ob9b$psfZ!}O11hcptA#vbaS}Da%pGbFBh^qjCsd{;D7xJ3h^Zs?9v*|&im1v~
zvw7dG&E~!ID0tm8b8yJ~!1+cRvSmu!dUm2bS>CsFs5~>mzr!TW^uc&k#Lvl*@(g7?
zLC?}5zK1<0_?{@ppn3B{L$e6c*>Etkn6LFSfomH~Hbe}LZH9k*zRz2_#K62J1aH;I
zYlC2^qu}r~0_8M_KyNt*?8@Ch>FmV)VUx>Q#s`K;c0yZ`wh6V;8HP1P<7sa{KiXkf
zEBG`onBxAjA%woF!L06N-DQkl!t;HguId1OFYA@H02sm@7|<Jez-UhaU)E3GxZ4T)
z(Eq=szHF!tO>GiJ&gR5N;^uB~$8J^4+qYn&gOg@qIhQvpm))SI&Tw_T>v8`i-9N|?
z4s>#+a`B>fWSrAU@}I}?{fZqz2}+`G#mR<@=YwtjeDlJsU~L#Vp_=)C(lpLqBt@(j
zWT=09W<ErD^Rp99xaHJ#ILZ<OqOrI*Uw2fx%l)K{IFn$LJ>b%;l*W~2W0o@{g%9?&
z)yIy4MN@U`N|Y*gf|Qaqb1Nh-*uvCwRjifE>+0NmZNbiW1A3AHk5(`i#K2p(i8pfM
z)6E5Jyr?TsMK|ZUKej7Q4IAuDm&Q+(emkjye)-%#!(1{x9A1Jxyi#p87Q<9XK8C?I
z-E{ZLn}$|3(X-VCPLWC64sD*)J-4Us8Jpo$ELOly9ru|jD#%;`8~2p1`&AUSPkRU1
z1jW-)w~pF7y61IBj)CTjW7be6i8eWaEx9}eCD3Pkdv&wiQD50(T<$YMWV>>?0miwI
zSc(Gedqu%O_dezVmC5;t#kn~X_ufbF-@7H-bzOg@DgWPZhek`!QI_XhEFIo*<X@rB
Bk<kDE

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample2.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample2.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9c761d9169cc72de72bb6753b7364ca696e17eaf
GIT binary patch
literal 55829
zcmdSC37lltUEf=KRaaN<%k=bYHT(2vHjQLSmStJXcx2g9Tb4$aDO)ag_pR>fp6=?N
zTh*hU;<oYg$d(M2Q5*;XWL^&;fD`ab_&}cHkid%*wipvkLP(M18SG$W5^#9D7s4Yr
z-{1e-TUEV`Z4A7ny6e=r=bn4+xo7$B|Nnicrzaodv+uL-t^9A##A9F3LG($Jd4O;2
zXXCM$7c0fQxR+Rrm*V!DSW1?XaaWcqrL8Pe%JQ3B%q?}5y6jwPF~8JZ>Ru|83U-`c
z>{;q9^)B_5`s_Hf*uOMT8VK7NEDdsfc5!HFs5E3{xy9k7k<y6eyB0^6c9nKnKEF7&
zw7az1^4*K$OM6OtmiCtRF6}GrTiRdRzjUB<V5wLtE*&f#Tsl-bWc3P*hnJ3&j#$2D
z@#xaA(y^uErQ>$oyLe)0qBOB|vUJjp`xb9oI#oKgbh>nU=}hU2mGv*4T{>4fXZeA}
z+n3Il&XXVXhP+{K<dtOU4(~Q^mpAr&ymY6x+Z%sAR(j1NF>jBz_xYH&w~}4Bt8)HE
zs&qGH`@H>>?XTQ@BU!qK;{#ri<D%+N@>)s`dWR@ERJn)adpSPr9pU(h9lws_quw!&
zk5yjB`TICN?w#QHgdM-0;|cF1$0zOhelPJ<Y~s|^3e%W>Zhp}#F3qpiSN%$Hc{YE!
zUSFx*d-m+yeEst3^qHCErL&8xGxMH5HD9fr^(xin`PzJS?zA^QJG)w&U#^~Bxn5uP
zXD*+eUR<6&d-vVv?wa;yXYQJP&H1}$UvrOl#~rix+&O*s^lQ$auiSNe<*v#dx8HT=
z^d0S%%B_~lQA@46Y(IBq<@!!tTwbYEr{>SjFRl2?*D7AScdl2J*WP*OJ+D1~&mGg#
z_uPHg)XeR7-hH?C8qcG*ubrNr_Rf2+eeJpPGnG~Y?Hiw+n_8{a=BKLdqFeOg;pL^3
zx6jwk)~A*<Ky`nrTAN+=mnwcOfBJNh-<kRRtJ+MH6Y*y5v8knLZ|bopClbvrzcS@h
zeXi2X_?4B)RGq@!ndNG|vQ}T3s!q*S{APD`b#XB)<NEB%lwY5pS|ppQU-g#w?OK|;
zQYp{*^BxDiw0EW4sQAm}<?7;fO7f4**Xj>1S7+zvCUVW*#pStjeXU-es+H%L=9-zx
z8mDW`tXG+xT3xI+)AO~Zsg-7y>0Y7(v|eElW*4XG&BDy`YPDXb2UAP6<}MaSeX2@l
z>Xq`Wzr0kYm%2&3nR#r!{>G&#Jz|xSm|84fS#0jTFh4c7T%B5cBplfb+}NL=Uahm5
zs5<y4^IxB@U%xo-*XmEzb@QGFSL@4_>ddlN@!#^;M6x+(&pKVHPnBykmn+`tVx>96
zX!%n!_40hRzFe+Xs<mamnb(4?lo#i(ROnFmg$s{9{?ydc%3{Umu>Vm#+GERe4DjRg
zYxC7+ZbrSI;UT-p`OCH1RJ~qp#;<eFOn9cNU7nosmM+aCG^u>_%btoo%(s>z31V!t
zcw_pl*DC&V{(QYsT;d59&(O{JYVkR?Sh0S2zE+%Ft<LD_i}mHA9;{f?ZnRowE7ztL
z?XrB4kGEW@6|2kj;>=~Om*Q0Q`qHvrIrH%tBbTaG7H69&-NQe^?JfC}cbvVvyi_^M
zMV0C0<tt~ssruB}Ie&R|<?K_d)yG(?3}E%F#`f$5cJ)J#K6!T0TRLl9Wy?n!{jBG$
zES0a;Z1m2v37Wm$d~IravEr4jt4H<NHWo?$1o;^afj;^b^00B)OD2c~@v&Gvwh#{z
zUOY(7$3GGurfiqSD>2#3c$I7OGZh{wZLp=8Vl{gcsb<2zo$vW(x^{VLg(;-1_LKU@
zIcyHBR_C?tSoUSt%xx6Ae#QN><Y)M{KKsL5?HV&ziGj>wCHmwgyrh?!i+O1;<7K^^
z*X8BCZgK^$$LsY{ULSw`{0{I<^Vi4Ma<ltih|R`@2A=OKCA~Kb8Gr&(m9+O3Z`9lM
zeEdedl&NGlayMg@t{dE8HsOtVyRAI$jeC1Q6y4rlZy&z}Z@+hd-yW~%9ptyyJLDba
zx3ALg9eE{gU_VtFaIJZd8Q53*gJH?z2KZGn6qZbSr@Yg4uVL@4-Wl&KeH-!4dAHN*
zsP_b*@D56Ld2jRX^j<@5%)86Gn{&Ipd%V~3JMP`<y^h~K-hJNd`2~^bNxb{LH#{HX
zKKr}}yay@Y?>*!_%<qBl{FB~A?-9Fi(R-8kC}$47z``gUV#c3prY>A~;_+91H_7~E
z7U!#g!)9Xf{N%+(j@5s5k!^ShfTYd(s&?3`RE)jqk$ma&!+U@4x9=K%!<W8w_!E!(
zR^z+g@TEWf_<#A_<J0eZ!(24SVdg7uXeKKwwV7gAevscBU$s^HGOgBQH)0#{o7^DS
z<;AaLYUA~Um$(sse>~U~#LCHfDu^wl7c$=;^OEF}{H8vZx)BHIv!ct{pqDfk^pJK1
zdD48)P1+q4NDIMk(w=&65D!LgB*^#G`)Nh%Ci7B)dv;MaK*@mrll4KJxe=#isGjn&
z_2K#mB|{sdHyPs_aqgV^SW<Tm#=O`ST}clYcKN>)jFBG;hDmn^L!{#YSC;n#Io0#O
z9OO9K8|)?B7wmsI?x*VegY=E~%kf&Op5RH-sz><++B*=WgG`Y1x?X~?v)U>hpzb-<
zrDu8dp|iKa0j|BwnQrTytNm_kS^Y2C=*AWf27^H|NCbUBe=y(`Udqruqgy@{jFTR&
zA6YmW?BQO=*hxJvCDx1N57Y7|>&F*Pke^`GdSB8gPE;A)jgwwqP`s&8-1l;P<+JtM
zZfHhMQS%g6{Z9S#!kLY;H&u4d6O8GfoJjhwW8#~MwR6q*wZ^%bRo@5gEM6}PT>^kt
zR+d4q)w$v!7R@05tyes>b}r0ZJJjg8xA^30wW=Dx@<#vZ6~E%mJ1{QLc?tZI1@2ET
z*D8(vw^S-Cs`|#qAFM%U6`S!hK)3GsIq+hoe5G=|*39e1WrMnn@{^TXrEa&cEKLb5
ztOz66$>MXB`MJyW+L_`d;ST;+Yn9p6#o}{-@uFX;0Wx(fp<=pSq=D1+On~{BD=W)j
z$1{x!#oOAS;xi9EIr-@1BRtU?pL*k^OJs|aPd{F!v&E;MxbWbmH$GLo^wwhOjZeOH
z(`6HxX7}9uwMw-tL|kk3Y{ti3_4y@w!|*i=Zc~>b6#RNKA0Abz9td|Cda_ckF&z2X
z`NhRDJASnSE}L%%WB2%#tE=-o6@<-{2Xt7TyEf%d+ztv@1p_p5WzhY6y<BegK3J<&
zn6l;S8+{+l-*5CurC#yp8&fK7rdJr4db6uso>`o#)wqVMt5ZuAa{VB8PyL&#xH)L{
z%q-5YfI^Ib@S5Esm?})@^>V%0H!IpCtj_E;^9G#C>NJl8Zmw1qYt1ocro6%f>h@4S
zM!_&QwPwFCt<lfV_){x-khD#%Kd42S)pg99KdhrJJEP-fa<NivrWPTXn%U{q*;#1I
zW{)aIL=cbN%7%1Uv-_I*pmAVShFKhQ^VXTg>E*IpOU=GTx-P0HD#$J^Gpf|@HbRR5
zo}I7zJ?c)cjkhE8uB_CX158e(Uaqb#l~-3FE-Ez@&C$n--|APhOF+C_zrIpoF#zAI
zWjB-lNwt|`{Vq0p)JA*JB&yYB*ApT#HHrRPRI#tsvU~hyN_(zZSe=2I6;;LPHWSqq
z2x)fg@=~+=N&AJ2tTQvJw&GXkCJy?dR!ZHgRl4rzRsRAd{=-VH=vG~977@Ai%l-o#
zm!O!ZJR_#dy2_tb&37qzyOL*=yhF*;N-imRqmqkC9#Qg;5@9WWMoERltvg#E^jv_z
zf2+<-EBO{BizFp2K9;>Lbzk_*uaSsx(9mmTjI?Kg!)PXK&?e&k+jPkhCC8OKPI5cn
znr;(2n2-0zGx2f0oqvURuN@cS`xANP5_UF`bG2f<ad%xIo{JYKQ<|q{A)%Us91kZ7
ziC%siJwoqiqMbUE)m{1Y@|CzjV_y5|gfMUbcJslJ>)GmH9qJG)krEd2;xDDwvtA;|
zj>Xn<^~6FFj0S}UZs@8fHd37F@}Q+Ykk}o&nOM&U*@g7G;_JzJCIC+Y_p^pya<nDv
z+_jMQlDokN3)w4)+WYExFC#o(@2(dXdieHwSy~2rZ}fRNuz2=H+^~y|yEyLZI2M)%
z54O&C2i^3jKj^y2v$RSJu2lUTi)}j7V`m0-IMds{X3yqp`Z~{e{p}V#(G-JYu^>Af
zqs9DK%znA&9zE@#%6G-qg){rS0eW8O=;NU43B4Z*3f|Do#HM}@YaHy!@;9N@U4QpL
zkF=g~B<Obcv|72HYAp<VqgN6yCwwTBg%NMpP|OqMG79zI9prAPT|L1r^?Ofjy{GDP
zH0+JpH5a_yYPnw67}Mxl`{U7(o{;(vS^Yg$|Gk`p>iYWkxf$omTz%ZzYt&wk*U#GM
z@sjsx%=)~2<e|ew1<|7DlD#+M8~bj?c#coj^Y4defTk#lrr5mWwNv?g<qbaMC7fV0
zH3gUb>R0(RZkt+Mxja=XUY@!JtG8U7t`uRtL4{rs_N7>vnt?eC5hz!tJ%gQThiQA=
z#pRh*Nj$vf(B@-cT;*DGsCLzdzqi}8YVNbbWiT#Miq+cI(|d*GZh3aMUF=t8K#W`3
z-P1m`bo-W*uj=WWgDoTkiuaMI%v75D+hw|OR1SWg2Pt}UhadWA;pQtp|GMUIyQ0Y<
zW|k|nx9?bb-srklfhIBkK9X<_KHuU89`IjJaif1})fnlJ(yMsjrE81RAp)#jWX$v6
z<(lwVwLY=mke|O>&DKep*=X!D!rxWTKcmtl1cqN#rS?jg)S+OW!*!WZGLfAHq15j<
z?_br0DUBcaeR<B1Lo>U?QdwGEa_HdOboGl$-lxi$OHR1a#Z1*qw$U7#T&_PV48BBf
zE1n~Pkl|FPs{Tv5CONgZIMHip&mkhibcXZVU42Bi(9N0|H@|j2|9_&Sl#@_Jl0MfF
zHM)iO_7^Do4!*S~Nn*kj;Da20z2Jfz_(B+DEYY9n2jlc7N5MJz+n*ffe1Bp;HOKkx
zPY9<J!6!%f7UGAvw!jsQQNtrG5+I^<HVI&aP=sD<0SQ}`RX`fwfLT~bZlrDkNMRFH
z0xDQY+7TNV=IB0<g;=Zj$6)&6u}0pDy&Y!o8DsBg=bBxW-PDe|@>+Oh!d>|et^gn;
zCmXx(Ek1>qM-nRSaS_xZ_1@S6wQaPzPKA_9baEo+2#aTV1PEP5z2d)9`Fx9}t(CGf
z%PY{L%l;EOlLh~)-bBKtnC0ojhg+to*(K_r#-)`{b6IVaBxdMZh`KkP0EIi`-RQN!
zZdGqdb4IvAe6wNg6G>rh<T$1T0B5GWWIYk2Hj?60uVnmV0VTU*>u{G*>3VwYo*;c8
z_UyU!%-YEyLGI9cHppzGg6z$>|4<!C%0?#02>Tdcn0bBd>O;@OuHN&ECN2xq-Kgip
zH$NY%cWuCFZ*+^d+bA%(v?dODJ@;~AEmQB=e%{MMIpkhSUWmOMfA+WE1^ocE&=sVv
zWc|-v{UlUE*Y4OtFStKXTBw8lZ^TK#lTZp@srNx8*pb(DB?WJ}7fcD|A-V(X7v{h|
z@!DS1h4=5~x=#lAjsBb9C)fUZZ`ej&HU0Y9sro?B1J%*Jo8B`X<LiA$6TtE+8I1`W
z1%3C?<Eu|TLyDj<>O-2*_^!3LZ@K=$df(c8VOc(G^W=JeXPa=|-5%#gW3%cz?Fap{
z3A@J+>mFA>`V2F>0ZpJ<V90^Bb6Z;6Pm2TXcGC3I%ioPa#aj7eyIz!pyMD0slu?hy
zH(yUL27`XCP1sBiLeKH{frR?EKDc&-XC7SWZOw+T@3#ko8^d0K-VJiqW!6VeFt|_S
zY$JVN?CK{eOTL__{Uj@AgtdtD?xjQ(NzKA&`<PWQv_8zb>oWk*&z*)$$}&K1*m8p(
zg9Zfu+ZnH$xlj5zt{4i2zCY%+y{j9ZJHol)U^qHw0AXaZnVeZ!1+x?ot-}BzXBJ<Y
zhwVgmE>5OVyii%J7~>BwQ2f?)$HuoSoS6|a)rXyDlGUg9t%*^bV-GxA;9279i6H*h
z5?51CbwFqY*oiaenkn(S&3JhWT=<jxec~!n_pIFiAJm}mXC~G-Za`o&$9LvF^?b2*
zKbh#7#-WSzPF}H$J}HE80(sTSbI~Cje#)!VW+c?DPc8Z%<`(`7O7uwnI*G7uM%n;I
zqPGPeO_I^bIDvVi2Xz%186+H%$u#<laARvLl^L*JQG%34-{ax=w{UcCzStPbKi#ep
z7EbI6IZttQewlkWQ@YwF!>`vQoOh5lBWdzf-LYNoR>dwP`F?$Q84f=UuaDZGU!&?d
zjfj|4xN=t#GU>hmrcnsKr~xip1a&vlFsXCMI)uiXX(?-J&92pI?doc!f*_&KIL`Lm
z8xEKlaR5Hm;!OPq)nuaP`9H2q{7N%1f2EnLh6)=0hwYArIZ9njPuYa$sAMY7b!M$X
z(<-=Ze7+f9DCMhV1Pm~~i(V_wl@KUyl$-hX_-|QM{$=j`IleU!1~GWSevk}YqW$%P
zZbTb6E>d(tB0dPF2L(YT499yD`#B1!hxmTc5d58^mErJc96DlDZNg`cbN&cl{WXSm
zShX|SFZyKoruZ8EdWZ~Mr)eUB_(I}FGJpqxZ-fk>GzllgUQ#(2O!ZQ&<Fw;VXOp1%
z#MFM)uxV(Dd>*OPQS3m_vppisR`KeIwQ{`}6xYh<CeE~~-fLq`A=B*F@3qU`IaypQ
zKFs&bnKRE89l0o>p-G|^D>a0E)s<BwfLmJ$uSbkBUGa5AOL<c?Ib89C5&gy6C(ab#
zRPigPis#TvqNhaYiP8yCT}P9C1)Xs9-r|GB`YLoK*HZITF{05=6+0>RiBm=BOWiVj
zlH;|aEwgrmAtC;iH%z4cafaQ0zY<Z+j>vqT+(ag%Ucy$I@jCh}?%PBW;CO~oq4U05
zR$R01YT5XMs5X`av6?ntEUVQ64LNLUV@tIbaifKi=4*@N1exuN!%8H)cFQF>92*n0
z7oVI+x##)-m$~QqAUR5p^3zN*`~EK|m$rG_+Vince(V;4Q>0qOeR>4-E0!3HH}+Zg
zZ&mj%^*P$}bAy|Q&*6WDygi4pQCpv4e21Pj$^h!}wZpCFNbwv8ZuJ~t?Qi%=82>FZ
z!Ui@)VtYounI~z@2l8t-8)46GYdy#I`Pf>w^%P?NTI)*OQMj&HM-k8opCT$rn4l&C
zI$`O0+G+{(*mdg}Cmu5(#Pep{3jB7>k?kyTV~$_@5V?u0twsOmRoa@!QtxKwx?O+k
z>~>#mOOkKbFo@aGvK%#zuF)UvtyX)89wi8E4MP_Ptznqu0ID6CUywj%C+!+s#Fd0E
z4G-IkErwk0rQ~|r=rs(9ICoc&TIccoi$Qt=`Jb0|(kI|@T34V;OTH_)o>|YX=cKy|
zayB&CNs|h|9BBvozpSJ~IPjI9X#=O^meB!PiS=@P_-Dys2e60^yzKuf2NPXxcgNRU
z%X->KV86VIlzy#@xI{Fb6XFXv+4wd`TZ1G?cuVxLb;$cxHUqSKHSAIn#0-%Rf-F<~
z6%wNnXxnnZjK*7^0^ioy8*?0AVC)y-v+=(P7xM1-dRpVw8v7(NSB+(k(ylsqk=jU6
z!Gw9FeCWroAa}h9@LNa+Y4mg{F~JM)FJ3~@yB|n=BnB4+|Kz2EuHo25_9mr4KHQuc
zNy~nXF15$4O$)MYUf@(XDW87>{IIQN)A#>|F4{xVOq&G7SYI;-Fybr;F*s3h>wI(j
z&4M$oDH~XA#;5%sQB&6FudBU5H%FVMLrbMus4iE{lxAtFc4g}f<yOkhCgNgyM~p(3
z^d?S5lJBTSiv|%!(q1q$^Fw49byoj6qwK{Uw~T}v%Lvwgg7u!Bi^D5d>#2IWjx3zo
z*&7HcOb?w!3Jy2>?4->d5^kj9@UydVW(!_?A@8MM(%$OU-k>ebaA6es2EO!zc9ivy
zs<XEqf(P#j(h`Y?-wxojC3~O6Tr5a~;&PLr9Mi_EkvBsc5#rb@`5)&g{J%p&-iGwO
z%Deq<vx)G@#8`InBqxph^{Pf2H*-iVp_?n7|5IwHs^s%ZmPk-}Em!B9PQ|n<FH^{(
zII0R-8O>owt;qDJg2|O9W2Bnv$G_M4q->meHlO{pu0GOAj;mfzG*xWM<(0Q;M#m_Z
zEKA~%O#BcMs(gGe`4lCML5Ie|6<`;=+E4w~Ol!xrcHJQkG}FwpW?Qo&IF|yPCugB7
z*>N(>OJE@8f04Ruk+xdtODae^aFd^&La1cwMyuFSZOa2?+}eD%o<yro!k&z1WK5yZ
z$+)vE_?)7GdYj~1yO#uXlQer4cJ=eQ8wn_5tv>eui%`cmk}tBSt%TmYk^qEk=r=hz
zLj$dknjGS5I0L~#{6&Ncj&4dgRxB@-9eP1bwQO1#grDaEH$^r(x%+^;Sk2u(%j5hX
zYC5DNu7(rP$(p*Fj99N4*+OyNpe&+j>?UozpaG0-?d}@2UeO?meTbpnf`)Nwt?2Bw
zL425swG$+Uv)Rlc#Rz7%5+*$?fnwQ0L+(tm7(6rNW!>E})#*8Of!7w)NzJ2?i+P)6
zJJIUjrX}f+@PdH@|A?sCF*AZ;aVGs~M0N?pR;lX!TDD#SAW@?HLOr>VT1X?LqrOo0
z)oE9<U?rxQgJD9YlJzhyK@5f{Vj?q%QP1uE9dr#b1RMv<J{nAK)r=+@2Ue?Bs>{z+
z;Va;9r0pmk|JIh)c>Os0-So1_sp*>ki!@@Z#$Q%*`N)E9YO$GHlD!{o_}|N^Xvd^P
zm^L#oSqz3<E}%lCErXcTSm}hUV@<3QW4{;J?c!QfaHyk+Tq2zMsBUX$BOI&6UMteV
z_Wh%s|Dw3cq?ZI90|O9EvwIlE^n470j(z2&3>qiVf=YOu^wJzZ`C^P?*|a3R49AOH
znW7~~!JDc%h{VZ+(~)%$!yaQofcUardnvh)#exm{wYelRFeId$&G|p<^;HvI|9Tg&
zAH68o4ty*T<b<u~(Xb(ZwLEpf+(=5nBZ1_FIfv1~@X>XHT-P05qk2d_jf8~`3tJvi
z%TcQiuS^X8IiZCAdz$&qJ!%uEjq3j=lJ_y_{(mMDaTqq4mD&!FA*A%$J6{GtMix2^
z(ZCat?LlZjQfv&G%IG@meM@bG$jYt87j*T9mE1+b+D>}&OUSw;+ixbZYH1YAk}b5f
zXmq(FbWV-5WhZ(ZsQP`i(WW{7K!;k&{(n=VmEX)o8iQuW*<6@0fhl^-z#t-kHiI@%
zqoRGT`~OOfIFk%X9sU2Rmj9Tflo$2GUYVU=lgSG<8_qDKnX?0WyJ_$pfRro~<D1&I
zY0n7a2|)Hklxo*x!f$_K5Q$E2VhqY?5PeT?VmR7I(e*7wjzE+Et!yD`8cyjLjP3#8
zu>PTupl^`IKp&+x(gIN&L4QK`fSZ1R-SrV6m*ggj1@_12j;s((OHn8K!BGldP!1(;
z1T>63MQ7_{u#>s+oPHg(x&>xrkq~m$`vE<Eibi6|L-9tzKyP@eMGxq-po1ZY`^kuF
zf#O5r6eJhWTo?zmUzDGf8k`W`<t!G6*e#@9if?3iVoG?z^rTyEi8GUp!Rg}V;&a8B
z;_3T~)BJG2GVEEZU?P0I?8KOifyiV>Ppc!%1ZpE2mI0}Z=I2(IQ45DM&wopWNPv+x
zn#2ETU2TYH%P4dS8k%70hcpfXg)t+VB;o^GI(<^B(x$oAsMv@&3dwjvW*nG^+Ae_7
zMXD`Wi8Y9*Q3+cTJ{}~__9le!ezyv&Yg)=ay1S5gu%6md?M5gCB0$JeS@|(_)1mWT
zAbiNZc2sDA^=_KWMP2+NnYB90Ee6rOU<(i9MN}Um+@(`=N18sASZEe={+%8yDGHf5
zN-&nF6f&mfIbkF_xKxa6kVe>;o)Zh3fW5WF44}=T#JGs^D+fUyLuUycaWlbi<N(=P
z@?97uyT9B^Y13>1?m3eSOY<nZWL_Ct=<+{L3;CO31yCz_1$rnzo%MQx{Fd4|2Dx{#
z8DH`Ll*es`fK4C{aM#PgUBVhWhwZ^!c^YF~ndQkzA&b$qhPIg_Q(0M_xr~J%ej(Vt
zsraJt4^RPqDYNL8LF53c8jU)l-xlU`x?g7I@-hbCeh+sN<e4+&i2vQHI;?4(Qmy>*
z3KkNW+#zo-m>d-&LG0RV6WxwA(1K`oIa+z%TWj{TE+}*8ma-nk9@77`T5qo$oQ$?D
zpPmTr+T;9vYSgr1#fv*}b_&LC%#_&(x>0`jnl^Pb5i$y!E^omEt-&t7+K#o)kT6Za
z1&{R~u6uV3jbO(tvJLE%zw^zS*3OzAijKpYABr384n6UF?nRdCjrfORFTn4kJ5@>C
zO5##*zb*D;>>XSUWuG#Oce8pA7gPePU>$ENdoi{{S$uPuX}!7}3^pFHrMyb;lmhZY
zdSub&^W6N|!F7>F>?0l?CgQJ~3JVaHGGj>XVz&L?=BJrNSHlT3Bw#3^Fx4oZbJU!6
zItncnn?mbfYMyjQ6j~<?Dua#ZF;ztQW)s;#9JDU_=DO-jTI(>R>!~lr!1kWr-tMU{
zI`LQAUA)NF^83}yfRaHaIKPP*)OPT1@*qB5Z;`OvfxU5^`Vmf*2HRs&U3SB=Lz(LU
z!+Ag#9#mogvTw`K{32&UD*+j8NLkYlG~a8=nBhp~wB@4K*l9)UYj}K=3@eOB()!g}
z6MboG1bCion>;u@SJLfaU|P%vfNk50gG`?7Cm@iRVQg^zjXEn_5`*e!?6RreUaN(9
zR8!xF`PNR6fP-Z_kbt9Q_l4XpnAmb)VoAM_$2iQJnrw_#BRuO+Zq*^(#@KW*%&PnZ
zM-we{oT~cE&+)Vd>@+sMwtj?UvyLIJHG8Xa%c_+zK`L`Wvk)30n}3&EJlYR#(uT@H
zt5gdm)l$p;rp6*q6(M2qE?~GZ8<8C<zbfoz#OsV9M2`CSs%OG|wJMFtJ;DGCgJYAI
zSkEFB%!0kNV>L*Lj^m^eEm^a&!IT3RiG++Kz<60eN8-L%J$nOFPM8x2A3M{vwuTBF
zJY~3X0o)|)ln+H!AF-n@50xCuPgSOx2&?jQFex5{0YWYWlbBfu19QKPUPz=ob^2r~
zC{xY_OfqT1iNOl6H3I>L1Bkk+hFJ##3>N7b_%@JX#~BvskVn~fjv4kGX7a^G2?Y17
z_jUly9<J(^sUXG@z1Bv7qcm+S)JHZ(>pbtqE}6q1T1&Xm5|PLGcf5$^$Urn^3C|<C
z6F`0{rqYQ%2R54-r|UED*~r{(3k@7BsTQl~=lN;Nw3|IGg`ZYKDfh(UvT5ZnvC&^u
z)h{YJ%4{~fqXk99)_xU`)Q&K)to>?Wxwj3L9Y}7kzYrdWP|hLf*19yvmZr@J_#U?P
zCv<M+@|{ds3+_&g`Mx^YqVAeWl$MTO#fLZ~)upiSkQACwP#t$NfsZQp1trEU8bc+K
zp?^XR+*Iz%O6nvyVMu0J!+_v+r$L|S>!2Rulj!R-n07_8JCrlZRKP(cqc^1!+A<gZ
zS9J#swSP*9MumAZ62>5{aawL1ewzxlJE<V)oV4WqfY)3y0}0U&sFjXefVUT(DT8g?
zUZjyCEQ;}zR8W0eMzuAY8W4RAzDb(LCWx)aodnNWXo<dFNMWU=U!=^~1}Ey7h3rC3
zq8-U2Luh9*Ns&YoUCkad;FrC+kC@erCmg@}1Kh(5*3ItRaTRH!xti$*y{V;VB9WUK
zYG#aNt2L9<Xs;*-(T1H(K29&?7R^@6X(>6rLc_%HFer`P_V_#1N8;q4SC?e#?%$*2
zwMsM;#zI_zdB4aK{#^>lB6h`VM|#`|&De)2y)-#-*ng5tGhdcp2!5tGixkS`tE*Fs
z&UM7B>CN=f7e(Wim*X?Rzw%mfBXn6O|AmszD)|*9|5AzcNX-<!nT!5!==k$WzNEw!
zsjc2W(4q1FNp&r&#CC|01bcOONXcO(w#v^dcMnOkC-irjso-7HVfe#pe4CopYV^;M
zoaJj~lets^GmLNUum5!j%<fLa3%%G8-k&SvGPxYzuHFLkmE%i!A(Kz%GWovz0Xv({
zjrI5DQn@T=j#^J<=t=7%Or%fkAMiA77fqZi-)^uI_e@+ZQ-aR86{ahpbETa_;Dy*4
zq+aUsp0N_=n3*@XOdKoBB~y-;?=T<zIIfoFp^vMjdF7Y-oHu@{-}&H|2AuCbkqN%V
zJLVn7MKW@&ELr;}you+*V5Q;E@$sGBN$)mud>rxK<(w}|qv1KZUjp%Qw{#AeW9EJ-
zXUk_p_sj9n{c=y}e!16sw>e<S(Xw3G@7?!G+<U!u|MTgVqvd<d0aH&*`~Y*!oHAET
zl^hI9ro9W^8?8@=!jc*95${b_ayTq`)O)k_?MQg+;lBEq)i@fKRB#u4o0S~%p7fq#
z?2m_ME_qMewI|HQa{_EO+w6M`SW&IF{GJ;5B{@NxC)T7d;CJzo7CoP@ykW+6IoUzJ
zqDZs`YA-VXxCz9{QaK<5GB-+`94L`OC4ref6vmZ!?PNWL@KY|2Isg73<KOEg5pt%W
zHWDUEeL{e+_Jk+;qh~|Rjhse$EzfEq-)H$Q%lBJ8Kc_ka^~?=8W{et9ZO+Gop6^L~
zL_$);wzz&md*EDxD5{*lK34BWRE0Z<Ocp()>Xi096RSec>(9dlqulo`v1$UgD_{G(
zxqtRoMH&xyJ@r9cFy!KpFcxnZDZB0x+ZaLEs~m#f@+ht;IDVS*>6qnr2fd;D=6k3+
zj=AT6TJYqUx;yBD;-GwC55}pFX@u%~J)BH2YTQTZeshyK5cE^)dCv6fOjx^N<B3D4
zbKo?RePGf^66ZB_km>Ocku-<#@~Pn<1w8g5x3{vtrt=Mwi~hGN`+$<`N|;jgt2OKi
z2_As+{<44R|1aK;2b;SPlg(6ceFLzz9vHl6a@^+NCU-Wg5IyPib@)!t*#ah4mRJ18
zG!~C*EV^k3yHImNYzhrVUWvx*$lcyTLqw$zB94`~i;*nlJdY~~zJPW3<?<4797<j4
zg@<0H=08Xk^EPNxNn4^fw7nXeqn%+L+TCalw)2t_O73Zj*;4QH^fES1%U3E@iDa5X
zvPj3K4QX4;S<oy39i#Vzo_JV^iMnO#?$47<3>x)S8tinj3?E7o%ILvek^ius=}k%=
zQ6e3@{}v^-y6lNB=<p#WZC<jqAlvIPLWd&M98T1H42$Lv0TAl}J3({EL{ZWnY%apT
zQA+ifK=PwHD25@+#By0VgqoNy<qOntt*DL0*cOo3_K>5q_TlESEMKGG)C;q<2h^GL
zJ0Y+_iL3+wfMtw@HZTy{$e3M#on-@OZ{p-2YH!^!ZYcqbh3p3s_c`7><u)?R#G~B6
zf0L5;Dp4m;x`~uAExc38i>#K{E)y~DVYQ0DO>Vrn^ww6U5ZP%$AgPY3T9O)DDLWUr
z!%buuxnojeY<wq%oyB<L(AIG=AJk2k%!sblN28-llYDERAmQ1tLm&m{!YC@oNLViV
zWyESKM|*$`#J%Xv@DT(YW<2xe7h}4SGbXnSke@;c3W{0ihK-cAt`MZiNd{-O62@o(
z{@<f>^5##V@50u9W8fwpX#&5rE3IBCvIURsV_+1^B#rYNUp`-XL*oHx>dRPfD#!<h
zx5b%N{5qE8*A0w}teycLr`siHXxQmf)e4gP#p{hdhYuevKK+!uxSzUo;WR$o4?p;r
zjwm-6(AXC_mn&e%nNHv<lWlgOwhRu?&L`dpa-H1uF-|x*GD&yMK-$5s7Q5!Rn6&1g
zi`ateaLZY`IVMR(M?uRcT0uYRIK4S{D4e4Ryn?fHl1Giq)+EmLCFmmZ2tiC_9S!#(
zNv@c%=777M-?mX}#v9vqH6GBL<wgdip3)K&!y8MTuq}@EifyZHJ|-sP$krvfQzOP~
zSQLw|5td>`D^?%<O6+1QSM-mIKf$__q{+NqZ#u!f)~$#Ni#p+-W1XcGN5R%lBHX=c
zHw_E4V_UE}QoBq51AOy0?Hm7X3|};!4_Rd;7nHnF$weiPkZc<`V{tr<lX@LXj>3r|
zWjnZaWNz)Ar9G>seVA`e>{ujk3lfrN^Mn#0v0uxnHR4z-BsS7FrDVykL+k~~a^~tW
ziCrKk6O$Kt2uUfLDFw)B4s<RlGvl%lH2(ih72nv>H<OE`Y#GJ`B<V2E&!~z4eM^q(
z@KHu-NspT6$O+JLS_a3rK5QqF?C8SIBE}!0EgmaYYlyW5Bo*tJ)r52w40t(tb%Qf9
zbG*PArd(DQsN~Zw*oRSi)RCc`#FPwM)`l%<a<gQ)YwZE9NlM2P4jb;z<?gEs(4EKZ
zp2^w=>jh|I-6dK3sPgUB)PoQxH3yB3>v*JcxpDMqbGxuYti=hHfYEntxri657MUwL
z<3GUQHM%ZEK5dOXsz4x$O+;ZB3F#tAZ$sQh4oq{U|CA=c7*ioUn`D1VCr;>a^Mo1&
z6$uq)RLO{mou(9JNvy6ZEz*qS!_M~rFLfj>l8|bG3#@7R9yX4Bt%-I!O3&{TwXSX5
zs(qndZR>lH%p7`Yyrz&UJKHWMr3mIoIq7$h-q30Xn@HRHk`eY-2l1Ms2Gg6VXEw06
z6xLyS@uEYZnkAP7ODWd{6OcjdNj}K3C$OlDn-w0~hQW|9U9L+|fmXLs`DTRG{t+H;
zBJaFcUgm+CgC5qb9;A93q_+XSBSb|n$Z2tE=8CV~?&RB{v7X7pK2B+AU(`iHY^X^;
z>Zz$^OlDR}G(7*9X7#v|&3m*tZgT7nDW)ScCHwx6=2^654EQZmEqR(auM^0wlUwKZ
z)(y4xu^uGF*QWLY8BJ;(4irard`ls&64hf?hBlW{b{i#fDGT=(yEvg;Y(_cR02?r-
z3=h#g+mM!=D4g|EW8`?t-o&DFINx5Px8)Utk0Hi~+85Sxk|K?~|2;~+mxQ=**iad=
zhPD$u=6{QZv9x_G{Y#X6w+7AlZW}6qdtx}*II(rGZiT;FLq#)fD1VIufjncyC!OR`
zj5}VQ2I(0rJU*#`EwPwfq?4$)VX_ncV+<_dZ9RUK7y#Hv+1ct5*M{1>k*K|=o&$hM
z-7WbgB?cG)Z-yfXYG)ziSixcmnpynqp(=a>G0?DhtcaiTuV_T(mDpGw(@=@LX=Zs?
zT@cZ;@tLiQ(8w7Rhr;FvE<~X`HZ+=pG$Qk{F(W4OxG9<4;ErRuQzM%+#1byr@_RcO
z8Bl@~5fEr(h*pLT*alDpI3sH?tV!%Vdf*L8Tq)2dtSMh6EassyFgI{fG%!D4%&;b5
z7(%Z+IphIS{zN}%C(O-Y5oI20*XAqF`PvY8G!m{bblZtvZpNSUixfMGEJ;|yts|Br
zCeURCDg1GbkdYGcV!S9p_BMVfs@fU{Rd0>M`#8{Z3b7$xfl3;S%>hMuVoKaNuySNa
z%{T-};zPLS2A0a;;w9|pWU`^LU@TK(#P#YZ&#0U)BNJq7oQMfR4W+uyLET*AALonV
zC-mjId~flxg=pG(;<<ZqVf5;k+vSmum5YuDlqSv_@Jstzl(=aJiqp$J;PF)Js^W7*
zN4$&|WVK!S^>}hk)x)ah7)ks9J6+_|nb+u~Yi&mH0sK~DPKfhihSD4`r6|&bH*;4i
zl@$!wYK{#ImqDg2Iq-GDe@@A1B{h=BB)hJHRT3$lpOXet5v$lAew9Ez*L0$w=PWN)
zX6wYNLCC5jS4*;rQWwD=<PnSi+!nBpA{1Vy>;~m0TQDnI1bq*))8HTXZiQK$m$Vd|
zT1(;_ZBK2GVXYXZhUd%F6)?crVoJGbJT_j?nDM~fv3|6sh*gqg*TFq<%w8JiJSMT&
z&qr3AyJP+<)+%RdHEXS!ji#3lTV*B5dNatX)oty7;T5q(cV%~GCrf=6l!~bj<5WZi
zQ)^*?&1WfJpRbgseL_k4rJ>d(Ds-;7Q6ZPL3!$j5YvK%nr*u_mu(q;D>?cz}M((6G
z@7nr?__s}YGx6AUzH4Su*X(llgKsT$KNj9x0<tZW+9Ryes`^hgt7F{3+0_){#}rn`
zNVWpzgOF$Yw$AXT>un5L^DH2&xwd)kO)}4Tsk|%Ye1tHfl2DjA27<t4?2e6tsL_Hc
zCIFB#Eax{Y@0L#z1p4n{6Jx*NBkGAi!8Ra8!~kLwEu>~+LE$4rm9eArERh0rSH&!l
zsvE&^L4spM&PZ$A{oUQ9pr&T_W?B-61*xtr62Oi423|Kn@*HB}6n^u4;j#6-eeG~;
zR_m|t<wSMddMZ`+AGy9N_jP(}bSf*8wH)?}9PugL^JypO40y)nIHPAC*!q-VJw0(Y
zs>w4CY<p&RpRl!W;Mr9l0XsWw`xtC%t2GvJ{Kqs~DIx~^nVUc5zQM6bk4IxcZqr!k
zj2jDg=2l~&deK-|%^k-=wQR)&%6<KHpjut3yTe#)YbzRyYO3B135u$NxD9=?GRx03
zKG4DjkvH<aAd5K>q6%n-0ZZ*vapqL<@+k~T+jU#W%PJETcf~>*x=ZjB4zB_R5e(0K
zRo;2-*;DxFw~udsw%uNv^1!sPyQ5Y!=glv=NC`we_ORR~T(KOhy-KqiCp%oBohOX2
zSZEGga0UZDB@SgRlWS(4uFm4e-b`I2_1`OCH=!CPvND;c0XI?4p%~hklgh5ye?Lhx
zu{zsKczma8gqfT1TbR!OR@DheO5M6g2raSC4sWSKe_d_#MA+V*sZ{8wets)p>@#v(
z?YVY7C(!r+{fni0lVggd1KR6N6ca~4en%4WaF&I2D!w;<CUJ`N)DC4Qlc_}Gz)m34
zy5jCM)1|xck(8w+Vf}kNKwC})PJ;`T%E>>3Au4JqXUG$Dfjz;QKq6>!D0{&%WT+VD
z5IfAQmt~NHl|iVkl9vNfM{D=7Po=b5;5ZMbEMuPTD@p&yC|5L@4}y2Woa=?Ff6sPw
z#{VuDEsk1!!o;R^??RP3E7sXU!T(f{@_)bH6BO#b@C~1}TpwIUlV0`95NN&U>L0p%
zbiK-Gp+Gbddq&_V>cZAj66aO1P<@Z92F0CV7}UGoS0AFBC!?4B>@lK<p_EGnee+3L
zJAulP7OYP@T&wsjUiuXbjKBgtbEq4MT@okCNo<4YYxHL@7(fNs6%2~bMLn0}>?^KJ
zhESAbDETo;MxAj}Fu)F%uO$J2rm&V>u9k>&2|-B+NIc}q(%CeV!~s_{Qq#qj;fX4X
zkoAggfsbl=&HY@5GW@94_+KX>dd6cc3ul7R68QR0>AZMJ|7j&Ai!nXSLpn4qjhGiF
z@iB(WHmDi%$lt~y9+pOAw7E(^6gEA?DlsbNXNgQcf*C^xgJnTeF=v?NIz=z1s^?2a
zF)`FO-!T*4=6Gnd(;>7dJGg<JofAXMI61U}cS^H?R;}ey*b11bGrn}tC7cx%8=>q3
z&uzAvXtIDoJxFV$N~sw&mXtGaY7JlD?Q&I#&NI*@PR*2*tU^*K84r(=m-@)br!tAL
zggh#Sky8y@t{_Vw>JBDT?0Mx?S64Xm808wfciQPj(#?pB(?`Gh8jbWGGWdR8g#V7+
z#H<}^=UAO!2OOhCf)>CAQW9wNNfr+obvWjI7K>4_k+3elk=W?cdDDC3i9D#78diG)
z>c-suvXigshDNGIOeRKm6z$~3Fh@BbVn-B=J!x^HbFex7Z*t0+dr5%U%ta!v=AfXb
zgi8v1&J5T6>n4(F=9$aT)xiWLQa-3&BubQw1e0<+1!P)0BQ;h<(|SaAvc$y1T6?e4
z%=M~w%t!%SEu>AvDA}&W+AcZ!e|7l%kLiJ4px@u*@cVLi*)sP2|E#hNB@7F;P|7{1
z<OfMQN7w&bD&1jFn|rqpR9Vq@%#`Yp)@ZZdJ4XA|);0dM8rfM?CD9iP+5`%1&s41_
z6A*<mAV&~vSqFl9Z&S~}&a8v&bH7~?))Hw*V62_+7QW_80I(MuYVI$f6oos0>;@Go
z2#sR|u~F-Ee+i>Hbkhr>q*uMlT8C+(fuldB@vVzjmSaz=(?afY*g`bAXX`NT<P9U8
zh<g22zO^rtXw*y<Ev8xSSrGJ~dFd};MuhGi3v2a~<FH!B2;ZtEY%)G6gnBhLBdlc#
zZ#h2^8V=B6&Wj6>9FQKLg>2@x_F;wM<R0C|bQfePW9E%L)F7l=07{PfWWF<fDB8N{
zNk4j&0mTJ08)6YIpmWZ-020YrVr<}0mNthnEDE({9IYyPlg%R%a+@hGSg5LsCX@XK
z$_ci0$jZN<!<1H68%<8C%+w)<4d0`)k1P3ooiJP0f2&-QGK88(mH#y*nq)#Z5wF-y
ze_9u1RG7w+WC|2sFW0A*%~BYLudQNCvnMQREl2<wiyw=hgqRc(5&}^?z;Vp!x2}!Q
z`lA(KLTiFH91@u&gFGB}afqx!^_1nY5wGnHk_0m_Ri?ylAeU2K0_!Y=CLplKe0WYX
zfGZzvd<bJ8Yw92s|J^~V_HL9Ocu*0N0Fj7ks}Ncq<Zk|cB?px7;PQ^xLSeMVnbA;0
zqeu<<RbBE6B+Zmid@~<ncR_u5Z}|UO6*h1AW;ankqj2q)HF8efkWVC#RVaK_j)0b+
z^2GK%)p6w}4A2s6p5T5wGY~*VxKVtV0LynP#SOug?vW0`>{Z9r!w)F2RrW!05wQJ?
zO7(Q5ZsD`?8sbHNn^bK~KA(o{S|My)#p`-D!KWCgZM!z$R+qIvtRyc2Fc0&sl}L;e
zMv@>e3zOOdE|A;usKYvk6wGXn+pcc-{V6gDYRZ(R-XydsWNFCE6bo>p%fsB?h9W!l
zI;N||pgs~t;Uelg0d?Q<7?HI6B@3>r*jyGN%Ee{J;~XL7>1kQAoKiq6y?(_b$X2VV
z>G^u?G|CiSEVR^)Xl3#Dt<jmO>T#s6yg@1SNTVH;p7VA0HUW>9`cyqd*|PXvp~|P-
zP{ej)6!vw4fH`-7j%y<}e5z=EFyh+ZIr*%f7sdE0uZuZ>pjZz;j4p-w;xWRSL>iPv
zsoj<XYXWc~SeqkjA6W9VV%OYW%3{9H*b#tV$LFKt^HJ3Y08F+B4f1`Is~sRTM%&Dv
zM^rYTq^?93@sSkq=T-2k5?is^t3WSB%=r<DN_`y3cOZ)GRmxpiWc@L@TQ_Ik6_=-1
zymhT#VXRm#AKwckLyJGY6_T~9&uG2svxl!XB$vl(e+*t{?E$Rdu3lynX3f+Wpk*;J
zB+-@Q8qP3^#UkVcBFu3tT}&5V{jwiQ`fu{m0J@AhqwrP_K;OIY=`U2V=wRLFI9?30
z92f1li{lT{Qks^D<U&h@D!brDEL0(ODYB;5{Zeux=iW*)Cw?237`*MfQ8)H|41-u1
z0EZs;aMS^jSfRk-hpCASsNS>DE0+~0^j>71eVmbz=y<D*e%k2kXoI$i8cs_K{T-$7
z!LAl)UTtLIP{Y2#;bU-3w1IxgdFKbx!2V&Zt3cuXt-B0!m;SJqx_(G??p9mTvy7-8
ztr~gx07s)re>b9WZOzPRI5X<2W+ug3MmlFk@fMPB)5I2Mo>IY}w~KijYF$4TUO%ks
zd5=mu7`E$qmkKjGB4Ui|N9}qY@m?RUS0wXnFoZkM_IZ&#OEws#o$cooqdXhz+VQ&f
zT<m5H#yUnl8k=#-$9E{-L;3EGa=X^m+!b?YY|f*0_HG|zL$CXsPH={e(E4asNG@bM
zOFFo%!Cbyxn1+{V!G&qqrj3t+0dzQJhEZR(5n0X}L>|iieM(-Z<n<)C2+FV}0z<cp
zz%Vzp#H%($e{+Ku)Px8-=T`6ZUfW|e{N;|uPer2vSuen60m}YzZ}E?!TQqX+sn^am
z2BMs^oPa=V47(zWkku{?LsjtB3!;zqs)I(5{<w&v|E@!^0{$N;k?gb`sKWm<9cpMh
zC|^Sizp1iC5@Nz)%eN(dg}KN~>~R#XA%(xId)e6hh;l!w<cF15I2@611l5IuUkoKX
z=KYV0kwvm_@huRKA-Zws3(ryGi4rQy*yElrh4H^ak+Kz*`Z^meubm4k-%%Rfbi%@F
zh{Wc-lO4Tqp~hMRSsL4Pa_wCE3}_WkX4)UJ%;pIw^@nF$^$p{Jq-|`DYb?bRd<T<U
zdlyGCIuNSsMJCZ}1gVHpe8})0+l$l~GXx|@W5n%~w3uid$cu>3jufe%bGpuc6|t52
z`a{-if5hzArlyZ}Ed9B}y=M&Fl8Ge850kHbg(Q%RCd?wleI6nfnJ(HR94U;Adx$e=
z?Xk!`W$%m15dlr^9%$BJ1&ARBE6^RJ{mGykwz%75s8-*k#pu_l*A4Y3IjbNu26SN1
zM$pZ%*azXwl>ag~5vwU^My5W<AS(s-qHF6xe#@2Xy`;U8<5J#*mPC5Jti{3WUmpNt
zr2Wq!)g2&GbwS#bAnX5jc-&*hc1|$|4R+>all5a-rUFf!yW0TP9EE(oX86IpZ`;gv
z*d&!pTITb!I1r#tuQ*>a+jO7fq;KL16T<u|hmk0>#Ur{Mq)C!_zR56K;xSos4@9D~
z_Ir<r(Bx#D=bC+XOR4dZW8$fym)C^ZvI~z=N{It?;bjyP0;<z-Pd}ntW?@c=>QM84
zU0r=q$-hyunYA(}xX)-MpW>vXYFewktW_D-YhAOJS;vz9$U1k_BtabQpwrkA3-L>8
zP@kC|_e+og=izKiRMu`{PxulfAq<MhlCZ@9&&86JVXlRhWMLLqzFm+p8UL|5OpMuf
z^`lrG;EKWZq3Yfs2glkj8Ls9zO1ksBOip*|<K=5fg-2W;AqGeCO3MGPI<J^B2TYad
z1vewa2C<M&5~>nzF(330!Xaz^MR<>S#0?3<EsoNc-(hjeGiVSpt6(7)M-+xB7+UCI
z*>}AZw{U(^9OQ0MbJ%L^*h;vp^OMBA++k7qKc%HEH$DG<Rq{zC|B$57|CD>7TG2yI
z$eTURgiD__>f(vC#VQoMz^sH72>uw_p7c7l!uW{aph9!-LIv=S`4L);XLzHO><UD-
zJDJO}a(|?HS=|$$?wjnSsq%L#F%%>rCfbb6qLlVF@&BD_Zhow0cfGO%7d6Gc;MJrK
zR7dw{J+gUdw<Amv8pSk~;o|=fYW7na9wS@~iT&>?+p-rvPucHjF9>PIQlqSS=a#8h
zWZD%5-+|dVcF9Th#1Y{9-W_3d80t@qmp&RP?E)GnW|J74{+a=2Bh?A_4>@XY#Rz|o
zycny7rV8tWFT$`B%Z4=%37Oy(d<mIOqzKbbgaG-V<$Q|>$Bsb;XhAH8P^I;*4`C8J
z{37ZC<aVE`XI_lsY=K_s1%)Mno$@-ul2lLFd1OZZCyd<!Xh%Uys(H2XqP=$ELvfU5
zYVR{BA7Df<jzw+e|8Bjjp09Ufm@#0XGPAG%y!N45=qz0ysTcGzr+SaEcD+Ftf`|cj
zBg)9N-eAOD5|LUTCAwX2Fo+CNh$YPtuYm$}iQt#N+%<j&76!fi71VXL$@-A7Xv4uS
z*sYNl<C21|ivbvYF&>Q7lN-BUL<u6+jlRbg#$Xxj6_X3QL5cJy7==xe0RX|;Udn^)
zhSqntRFJ>!&?Wcz2kt(o`f|Z8RI%f{JP~6dy!{C084OwQwIO=k?;<F0#UFAs;Pokj
z0=@pzU>G04-KI8_X+$2_Hfq-BQX&FNV0?I0Wn3?V7$ru+P+*XOmnS*_o0ELbu!!v=
z*>{$&@dX#jSnPoNvFMEPcq|Wz0&B-yKAODhn1UkN^n%MZy{f4R3*uq;7ve}px6rTF
ztKFhNCSZ-aRbY~^h$<3j0&JKd!8X)!3M@R_0Yq7j?|5jd5J16<@TsZsLc<XI0Z=-a
zkXD2O%BGeSg%WnbM2uxxCrZK83JCsh>rRpylDGRC;o9y*kyHEWgbRrfZ@l)K?nY}=
zH1HBt*r>>C%3cJ~FL$qKbll4zkQKTS3cD$(@%CZ*QCc|I>S659-^2DxAx58Qk7@Cd
z3dVm1tQne*bCYe+Vw;)E*QfRR>aB4tE+MKF(b9oheulpLKdYn#aYKy3+qPbyF5JV@
zx}>T^Gvv_1Zz*Ruq%9pjt!8pM@!Lv%N{PZ~I)t&L!!IZ?dHg$-yRJmkm=n);sj{&|
zCZqqnPFR0UA04tWwh+$f)cqvR{CHYM^1_$R0tPo_StP57nD+`aY^Y&}2epNzfby4`
zS(6x+FS%#Epw|DrlK-gWckDiTS;Z98n#H$5_+Rrs&J9a>xDNs_=pEwCoNg{XceFg(
zD_1+GIkMy4$ZL_jgY&U;3alfK3hYl%8weLUeYDK>@`)4BZ+YJCB)7>tJcUg7WC%S9
zHP_Z#OOHLA*emKTh7LQ1pzi25v^i#a)dxNz<}{5D^R4{^4;l1?C{4B+NSl#)8KQ!q
z!wV85bzWghK~lYjqWVEmvMcLHK`~$eH|M{ZD|z!Epjas8T}YwH%l`sSV6BcnM!gFZ
z6GdY-`5!2zqZFROsi{5%IY%3{?=v-3o*-<YH6U*zY~|s8KBe<(UknEE7WsF`q>@C^
z>;v@-1-;~Z;6_sGl1udhq<LHJLf=9^vcqKU4@33WKrjs2L4~FGN|2S3r^SO2(2lv0
z>utWKiywh+iR(?l3m^~z+oXAN>nZJ)kaG=zf`&mWqaY0kR2<%Ui=ZTPp!|M%)6+sL
z*Er(M*dbbZAJ!+kg8r5@%14Aq46StgKMPXhS|OFIKMP8cb;~EjKN-o{4-oI!gj{YF
zW4c2^(-Cdj|1@*H4R`=;TCigah!}I*=G#~=4knmr;<t2IRk9OcnA8c|#<v6vkJAk&
z@A!RH-U%jz6lWVId{H%jUKa{F`2S2f+s;N!|C|mD9Q<?TenE*rH-luGcf9|LI{Rt0
z@=MBTxBLG>$q|)VqFrqvMjlGBRUYFgmeC$|?_23lq;Oi++uLM!!oKCVfQfz{UQ1q5
z(azP`&OMuoDJsoQ=DC}qYFoIKnQq;bkI=aC`qrelvTc&{#}V5~R3HH?-`Ui<(R^&e
zg?a+I5qBv6p^$D2Ne4ww`cNFYQRn4_^KK(7I;hGmbR%0Ef&;NQln+M!Y;DYI(E@jw
z(bOgco26&}pQdcy%S<}T=8mmU(^3B4Fe&5zQx?lNNXo2h{fGd;6Lgy-_ty8_G_-%o
zw9ETiU5b4JU1||4kQiU&IXd>>1s!hQekhfN@Lag{dF`pT0o#{25yjoLeQike>$mBE
ziInv_V_);_TGc~|wu_~NJKfF$v$AIs;7~??bmv)bY3C)BlOpyleC@rz@JRQg@L<oF
zyMlI^xV6Nr+z~Pe--H-3?V3eP?<Fu<PRo27ZiL9A$Tm=DVdRcXL|oe|$<DLdMFQw@
zClWLxqG8hCJS1qR_j0CVG{4MYWYf^9i*Q&$l=Dhaj-s13s|LJBtbo^W;J)Ps@Tdk*
z6wekg#|yoznpB4`XcYUYBdL-g8Sk4+B2h|gR|&Q+YFX-N^z=3I;&;h7yAxun5G0F{
zbhZxr;fwL}3rm1fwexmPd@NDau|ScUE_L-IIJ)6*K<TAGZX<{xDM#cP)u^7KEQdf;
zX0He%k^5+lvMmL<we(aB9v3G%+~81i)peK8uD!Wlz`&JKi==l57pGo9IAXryT&LGS
zME6rE@nw18Uv*S<ww_n;1Llz+1zip%ysT0FYsx$B4=2NAZ@JNXyWUmhf*_LNq;L{}
z)NGFkv^x^qc8l4zEUF<j>JYElhCW5T=6>1iZE_XfWcKM43`U~ATg@_~tw5OCaJ-{T
z(OArg)&D-VZw63?4v**pOAP-h=H_>6MW{$`c#v(SXjlbq#rTC3S)<{=LCj$=e>k&s
zkA0oKPU!x!vwQVCcbKu2N9HtNZAU|SZvy-UeK6l3yDByjP@b}Z?phyYGA-buy$Qs8
z)a2L3qD#D#c~@F*EL<O1A6?(YLK;~h<_AIprXRbC_OT1XID!Z#MM%O?ewl79qL$_L
zW%HiY`>zxJ4{#=ret4LLxXbjB+52MEpP;k=g_-d`&!W<uctfWB@*YP_bhIc5B^;gc
zy781Lm^M@zO|`IME%aQGSmFNb_w3w?*ZZ~F&;m`x>b+O;v~jNsq2z)m5%<0?*mWaL
z`4dr}>-}NLWwz4*!Vv{_<o%gJ0Wrt@^+5tpB6Pu$+XbH7z4p9WZS$r_Xzpt!Na1af
z(6_kaBR2Va2v+nKYTmPLf|C!|N9((I2lcpYuF;1ly+NKA@r%7}dK{CtJ?OW`sk2w&
zm36%_dS9@cR`#F^Pp<DpkGr3q9SC@jwCH36EFZ@2d)G!W7^D7J{a`TW4KtVf$nEn+
zEO#i_7mT@i_C^s4?+f<vRJ(X8y^B&B?y(#3??m*#-3~+7C(uzq)OfpJN&~d_uOH|L
zS*d&P=T5wijyA?=BhScuAALF!9H9JQaFFsnl;_lT{U~i7s~^9ScsW^%j!w8E?#o-V
zzZ1E}mixxNeXP?R?;MR&s|WkJQ;&DR-m3Y2dNCoM4Bq1qZECF+8TX`rPjCqF+^}~L
zKz8UAoh29>_r7oIz0uno@{YWe<?e^qkJv~b=3YmGqde77?sV+?Qy)v*XRr1RBOiCY
zobo@3klgJf{c4r#msOsOC1UHx){n2BV2*fWZE!p|fv=zR_RRMBgm*lcaPKM&j(aCq
z|M}oJ922<@B!UxR$%&|BHp$b5^|@+dmYusk!Lyv?Sx#<x7QMYe>(oN0a?Ne_O#9c5
zY@N%fFRc+hrF-zSr-|}&Aeb0)cR2Hxc!#sxq5I{u)Ck@=+C64Hzs)$xQ;f`MeP_|#
zdAFlKORS$<zpb;^3rB+66z-0dd@49aPw%9?EbU#WpV>HDN5s5w?k2<u?*vYSt6sY?
z?VoQ2O?{1b7eQ0ABKyOjsdrB*c=F#6*!><Q->XC>`_8h7!6kN#y^st4A@|-9vrKLl
zWce5QZbSQiMAv+$k`Du@%@*L*-Q>V*kLoAkw=iL|`||426z{Duzlo7e?gCMZCXxDj
z(ObXi2o3&idam44qH~a0;;h{wp!{FfGkjL(llZc?<B>M@M6qFuZwn(tc)WYolo=jm
z?hY;Z|5j~@ATA9%@<<BnC~N}`y9^8;Vi@({1yPf$ef9K|$2%})%KR{92RQ9FX`pc|
z>a2UIt9hQx&(33EH|tj_#kp0&bK?XJQGTnAH1}>M5~54cixNw&IllR9=kzuQqu$TV
z&bjL4M)J)0+3*4Aqj2__Mt^IF9<8nrtgyblJ2oU5UdNq_PseH22+o?^W_6YHQaaHX
zXtnoNbU`{*8j9*lI}}!KjI`=?$WLpGpNSnr<uq<3S&f@ywb|Dxu=^_x@Ov5p#<760
z?F5qS^<0eGo!HNb+g-F&QPo?b{yN!<J#bs%YDA-`MBB8^p-A$BX6`MO>-NffUo@8A
zr?L$1tM})o9pPyNV^!x&Yy#C6%7Z_uT$>L4f)0OAiHU9`zcmjde@=6pw8~0;MTeJ_
z7=>$Ucd=)_WRSinOW%yXKcifmWvuJ$8I}Dwxrx&qo=Rs`dQOQrA-81-=Lg@TeS5o3
zXzNLdZVRP3utIR>NKbFUzme0bXv^H!nnSMU{Os&%O(E}D5H(-%!Ed>T94zZ0{+I=j
zxA(<6pMslou9?#b_00tnF$(+F)azH&O{2OEBxM^gZL&801)cn5T}qMJ$`Hb<)gMKJ
za6zei8l4Z+?BX@BZZBB`7`!iMn%&WkVe^`Lwb|3ERaTiv!`gnC+Ic1pmp@NS!&3`i
zUZKFwBL9Kw->p6xV{o2JJ|V{7mne#5WW_Dx8K-!ad5joG=Xm8_W_-p3VTt%D*oJ;%
z7{s6}ShJ%2^~Z|{&Rhs*iM;g(3(r~Q#z>vZRt^?JOoxQxqe;~e8=`Q~u8f$vl#FxT
zAw=ir#hh^dkck6a2&3@GdaDS@yoAvD9tMiuX7}3Ooh^Y(mM7NkFedj;$$-W2-NIKI
z(SZEFEb!F!G8y!zd=K7(@aqd%>`&wFt>^(f4P`lZt<=S0lsA>=+Ajaw36|r6VI`t?
zdr%Yvv*IkikHpclV5(4G1h0Z>NQK9WYNBJ^xzGP23!Bw{B~$w+^#Pb-*?1Dl6#MbC
zUPVSLLomG=sCizq!aFT;Ue_yfNU#^F&Jpv$B+;<NCcwx~HjeTJWCghjExQ5c7VOF}
zJC@WghV{Y5!B${f7ul+_KEsU7GD~bkYw^D_sEy)N75BRI)s^C7rcHe8$zm;xhv*v%
z&HI^rhze;Mn<>dYoLVc{>}u6Raid@!3V_xH-sH96)tT#<=-r|Fn9j>!=MQkljwoAi
z)8Ug!43sV^XMj$kA*Us~Qw_XE$&|{bl}subQX+X@GZ*y&q<B=N_HcJ`O)2jNrHs9x
zf42&<@=Vdu|EVUm-#XZE#i%|f93o(^z+7@&lpt9uqCxV`9a6ha=U~W_GJ+tQ&JJPu
z!1Dxl+Z;B$<%?i)O73B6dx4DeaCEC%qL|vbI%^!)BlAOm%OIx1P2MjhpGAVBW$N%E
zScPfz;)%eT!}x|HVl&7vXISlZ-&AVHdw4?%bT{Wsgd*3{MW{n!qu#~6-E8u@dl&X1
zqCFl6Gv?}O1}ld2PU?!^Qle3E2<yLTFxzG>3Tk)lyl040+UPY12Addz%v%Qi^ISWr
zK^OXz>_c*Xq#QfQ5gi(x;59CHQuX4c$w~hWTz&CkE4=!<)RAYEyjw|`gnRe#qu}Z<
zsKZJoG8ZpS@;8y-T7Badb3J{LVhoD-xp>i<_J5j7DeApM_3(CK8HbY--4`e6g-+=@
z7rmVl704Z?*M)Bv4R^`^9I3nZ(xr=+E?N692`AkKc&*MNP5>{J36eX5Kk?)0?%PzJ
zSI!1rZYBOZm1xyibU^>jIy4Hv9G*?6^_<GCD|xSymy{Hg%qsaICFUaCqg+<Wk1P3v
zlE1GcsRVX5))5KbNQf`#zF$`ID?0nDN`$5?!n?28^k1)}HO)X9+vmr1&HbDjVD+1R
z1vytb<|HVswD)W-orsSP#Nxf#TsAk7FXXy&J>+_G{rNjH)a$<kRMwr(_os5{TyK7W
zaP7VM2PxmLvjl3{>wd{ueJeW~p6l!H<?69kjrZpETc2j=T<asOqmRLXD9n4?GK6sN
z*-EN{|IaG~@x~M=T}oScaYe?-`Eo64Noqx>sX6_IuQ4U0Yy2jiVsj=W&n6%g(GF#W
zXK}AF=+_3fjhzH}+*QxMAL|L4PEEq?7-*HMyxih!2tXP7ogLyXrh4=blWXLi5ra6N
zEsPM|d`xw1iM^3IAxP(zmdj$ATwqILB1Ht7Sq-<14v?m6V-t#=8Wt$p!cQxrr3@3I
zskgOY^RF9zi*~z~v}}XrCyvfwcMEAf<KvZ9@mmpy<zTv>y2i4#Lg{o-(vpBzc>u;s
zR$^%D5V8R1aEl;#(#!Gc6Lv=4eQ#25m>`beCb=tg9!I~R+C;$%YfH?7h?3d?xm0ad
zU|4}#IG;16y!<__eN@$Y(@UoM`*w|9g!##x?$FtSUCZ5C*Gi;`Xwrmvpn6lw|I2zB
za$$$Q?%Wbr<xs?Ds<YM>&N>oqy0#D*?su)VpThbe3sR!QhgV8Sf~d%@^mtm{$!317
z(Q`_OnrXTGXqlrNKd4DDXzI)&eXTxbd_(1*W=cBOACqY1MNV#6i?&uxD#FD&4;nYC
zmao!=gL*aP{!EFX{_j;K_quau`)NddYmpdV?|;9livegRW_I*4eUFymJv60dD9TnL
zOSBkesxFEa@jJirJylL$GYsKZW+fR+8@Dam&=HkziXedO7Rc^)n-N0>alZzb6byNX
z`X|@{)sb~jMV*a4Jz=IMlY$6H37&Y#SDYCeI+W-004rXW#v2h*m9+eoy+H$HUA&<}
z?}o^Rr9)o7H}JgP%#brsCcX1;Gt}fgy<(D#*3%MH>q)jHZ(mt{E|gj*vP7S=KV9VY
z*^7!>VU)Z8pEjy~#SGV^e^87RHW3TnUReM=?_-Q8Q$<uxCWkg4h+0aZ(10J<G}{5Q
z9*Gu1bEGryt#!!4p(a5xeu^ZD2-x55tRv0*R+V-%_MHLN4b6lDSaM-7fE5O<**Y{W
zu+<{<BIy6H9$w20?N}5u7d*fN{`o@<TNGNoaxNNv2bNFJ`j)V?k?*bM;?@yq(~Q%1
zv>#G5m<1UQV;O9F$Q_6LVhcD+@{!}KId^XnOaq48oJ(uO%%eF<A&JfKw%{YJGpqyY
zmF#@dG)tV9k{4eMQMyQ{kwB*`vQIY%Q|d<iFM@6iVd3-u$8q!L!>>@Dhd1NgK{0dC
zB?++NFUc#n%s9RP63Ft!hUpc&EU=*ZXVmH<qPqMM`!K-3Ngwl9nZEJ@j#*wpsWRil
zau96^C)AZZcV{W@PN73sTsgs`)TOsxcq^E5Gr%JJu%Tq5(%!C_wDqYmbVT+^<?@<G
zBgV7Q0(W}IA5>TRJES;2t?GYM)zwS)>Pjt>TM9~==`}Rd&ovVb-nZMam787I0-ENt
zCNYQ>-7SO1d{m1_&n30?I0kkKLLwQ1aZ8E*cnV&LclE((Dxl7`g(WSfxfOR1&aVo(
zi|XXV`&fEiE_NHg+LNZbJjpH#dOB2>cr%#lvR|r86qzg@TS8b82u8B_a<UL_t1eM>
zq2O@h{g(1FI(Ev-A=C>}UJmms<)ti`GyLx27xsMv6Fm3k9~7BmV!XtHq4~!6O#%0*
ziFy*VIhG<A`u%9R5fgabs5?KW+(y4#7cr9?p}hji%D+h|Z(uQH<tX9<^+u{m{r0sK
zo&#YCXS{wCJ_F$Db{m(hjX`VU9Xq$-4SmfP1`$(?m{M~g>REj_EO|d0jeEh84tT>n
zpHz46sCO~qyK!WVl$zrxHD#@iYSfgPdjhE;$689wU$Wy}-bkp_{NeiU`uGO5OsP4F
zQd2Ii9HYcUeTh<YKT6F#)Hx7H>7~-5RGPtFZ;YBKI5!T`Gt`tfBc&#G1=K%mO3>Z)
zBf)NOH=tubx&7X_<&FmXgWYcay*)70`#DCT$WzH8{U@>P--l{xKXf|^NG!r|6gH*i
zep6~5STA;}l(=`Oyrhi-w2`wmVjIVUB0WA7NMTr{yxTiys>WZZzb9z30G)Z**6fkB
zPw7Re&;G=R;&>Pwf;LO~qjeN%FUo+uODa!4Q$OjB@x2*Qjat5b8))<}pzX+uaVfmH
z_OM-hp>^%}!YNaj-le+k`mpvxtvas#MrhS`#5?+Y4pnpOZkkEB6?cz9XB#gkuKqga
zlzrYYMhpIN^JoE1k9fyTb$E3Bn9b2q#^X3@%mMENqcO3gI`kOrNhNvA+B<FbcxPz&
ztiGn$yUi4PC)OuAXJJ9g&{N)NX7gllk~^GX#1)3@tbuUBMZZg6lnj)n4u6fLlmg26
zjZvinb>#;-pkeJ#;-`(U5g5{STn}>w*b~MBx0?!3%s;PPN3{pg-%0B6pQ!vvfaNAT
zxozOEEfww1S%fD4PVw-s=mOa2BkKEiD)}%CJE_jU)~W4ELjTuQ?cXR7@%~q<3iN+N
zo!C(e=yZ5yD|L!WbM%D|Ci5-TTx*y$#3G}H_?qci<W>+aJT`=LE9ek9gpLKD<`Twd
zS&wTN@dyP@w*I5!{z&*r&GSCq*W3ap?K}y@1GYE%{w3%SO#$NnJ9Y8jlb8bVFRpp>
z{{wAJocjjUTFunLGOm)cwRED|A7nsy*W~Kb%Jq(rK#MxBmA*yg)mC&l>Ca!Cg3Dx0
zrsaCKx?#NF_b6wAsg`bQ;<gSl=&X9ZS()Ykmdb8d@&zSdRPrB{{HBuMQSv1vNQuIi
zDE_W;rj{uwXFK{8<xInrX9fAaN(xG(aPnVOBFe?Ng@0Lxzo*1j{C`sJKPx#-Qp&8(
zl<P|?{_m^g50w0fuKbE}|53UBqTC-U(fZxAppWak1aLp9wfub?5s=e#Ed|^gA`MZF
zI7;GPr6nrFotEenc@g^GjDCn#cqyo^KW1&>zRkUkB?d#SQA<Na4av%#{wI1x%hv#O
zVx2EyygH$Wb%N7A4QfBhq{OC{xrBYD=?z+9Q^#D3k?9bW7SxEw=4**0<6QH8jh(SA
z@!_1_{@6d_M1;Q1&S+4hkC<3}YHucK^WE^-1Be9EuecYTi|6jq8z67Qyd3_7;<v?Z
z6C_AHxYy_P8@HV|Zo3<tG}H{kT3wt+Ke1Y^wAk&&NAkt!^UzdnzE(ucKt%P~*@_Pl
zU7U3nOc5HlwmNg!!u_=L{Y6!+tkf@`H|2M`rAx~S%C5H#L=8xxBo2XQh*uQ0TCwPs
z3IRV!$R4}=<B7&;V>r*K4^ENmj*WCLPcJUdT&bNo$Dh+Egw>_M-Fivq73a>m>Nf4d
zpV15GkR{Yw%%y04Ap;X8tdSvLIVwJTZz3i*<P*5}=?#fpqNAYuv8FdMgXF4m&X1Kt
z!2$w^SQ+;jQru*qC9qTV-$Rb~bn-$6yr9jR)sFo-)CvIt2q2q?!-g_rysk|5_a7i>
zOb~d<g;<IPuBdm-6{m`Gxbg9J$;*|QD=Wa+I*VHpR|jfaQrKosbe#>jzl(L(>{94^
zV$Uztc-fs;FujgX_{enGIK6k4De9<%LJlroCKnkot1W+wq%>#^w(g{s{c)Y$qa>|v
z8PIDz*UkcwK_oiP2E1EBF{#O-zdYVj_+5ky-o>qJxg{UDIHO6_M|~FU^f2Gr`{@F2
zIzaGfY#O>%FNqm*%IR#e{5*#2KvrrP7)Pb+>9wB=cwzOk|71P0hD{u~LoyskmV_ZQ
z@An{BA7;jxn}}T?Akn$LK6dpX7y^X93rscEd=j_IuJj(gh;!`$rkZU^J9i>Mkn}-|
zL+7O#CjJ9DWnGcyH~TTC%hlo_qVRg|>bG0VT0)<zW36iP2JQ)yMRW-dJ3JU8uI0DO
z?C%uy2R8ERJ<bcMCbkfhX^YO2AYM5BRqbQ?L%a!kQ3zlvc|B)e(0xO>UUCJ?^?i(J
zWN4S|jMoo)5yu5Ojdduon4Biu)eX<Rmvh9Bh|by49-8!jkm+Fd{}w;~50Nym$y%()
zgowD=*Nfk}?$$Rek@&dsNIla~gc9OGL{*MA_8hktKxomakQfYrqxm@^{w@}I8PkcG
z%TrY%a@QsRMa2J-Nr*y|@#eS*Z{D*v&ahU!;?h)o=CWEab#4VkvjCF2+ZjdJ`Mc_1
z?~}adkw_Gt>sce?EV>#!yr2SSS}vcWtwukuf2ypl5I2IF0xd@VH+mlrE4_uHdy9>M
zr`xq$5#kk-!Tgwp#KgEiuG}Y-JfY-&QDR$Ao6Y|T9quOA%%f=~e(~ZwvBZsP`G-1X
z8gWrFh<=W$;8RK-Q1YOXf2722VU+7X=CIjStvnZokNsht{uw3zT*=QV5q01aBMS@&
zXs=3Nibf3-Efu!VwrO6H!*M6-M&o8@_hQ&FPVNX_{Y7t|+P;tHtGtb&{WeXZ>r1{c
z^_O%jlhg>G>Z3su)Rt1e_KqYF`<6!L>k3)DRZ^3V%*!#A9?|Ak)J1agA|QuB>~6g0
z34I0kt}~J8d35ZgH=~%ZI%&<2Ou@NN+w^=oG(FEDi}^{;$n+ek2KSS3b{Bl~-=LI0
zVt|1E3!`zhjE=;9F*xrqJrB=#JxF4DkwvsykR+zy^_e6_hUYu9;q`}Y()JE53?ebY
zu$t}s=BQt6WLGi`$m*BLHvJv2ryD@dq}Un_?u#r&`PwVg8bbDAV2E<b#Rfk>bW24^
z<;p>>lwrDGA4Y~Vj)8L|!`XujM-H5L<pUP4A#;(`Wfw9WbY-;sMLQn#hC&(6&(_E4
zyYcDYYciZ+WH=*?23H~1i7jAkzOfG(&R**55B4z9dhhQ6NpLVSAEoAgOxTOoTa3d)
z!*tv<sejO9KI8R6!ML{z$;|<B2QWBS?r?xa$gKj5%_-q{$c)M*!}$#)0DF)koG>W@
zG9aWRBXe<Y?<>TQ2u;-YnKY-key~%Hz?el+n;~yMV<b7(Z_vA=!9n_dI5^B*B<1M=
zpvVNCx!%x=>}5F4ePr$}iHy4!a*qM;ApN*_>waK1e#|>$auQ6vZ5)nsr-@*KJ00de
zN4CjHhM-7}w&fasL%7@|%w)n~q%8@@UtD%T?DmNEaPf`3|Hq^W9qQmepJd^$h*h6B
z{tfWh-vl4srpsnEmxgfuBE0pSS^@eZ^HTCFI=rkzY&B*mhJw0P@U?kn*?#)Jt>l~J
zo|kp)W)@n8+0eG%1nZoGbv8=~W1?xJl%-@A&#SGB^cLJGw(w2DjqhbNBGbzKgcpuj
z=7?R+@4zmP#SeW0CY7i=HW^il^(IOVxL0RVn2tF!$|y?rb}ZYLJ)veA6;UCU7~S$F
z%OYct_j7jASr)l0e8v_Q`M+!6|BjN!wE}EKiA3n&n7<ygJi=WH8nscb%cW8&^GJyg
z_w^re408<ILh8(Ly7hs%wvV}YZ(^AXri_-Nx~rw2?y)jdDV+*xD<fVdQBFHm)V-04
z`ccR5wt41XHda+kS%rx05RCq6Rh;n(7J0<ejV#<l6Uh@II%Q4C0<xkB@y&|$4(+%G
zoye({Q%23b0IhOn?hKS^?TVTWFBO4V=M50WYx7e@Tb8F~M&shze)Al6@#6`z+b6If
z_ak(&@$&v54YYQ&J{n}@jW0e$CQ_>supvp*@KlQjh7Z;&6tFEiA*FkyUQe~v>&^bo
z`-mYmvK4S^K*IRR!`!^hG49MK-qcFrOYer^5TiG}Y@MtYQ#g4G9dbWwmQEvN<7KpP
zaWaoF)?Qj+bjkV?U$|Lk>@0xc!yc&XiYz2e_u9)`k%5HyAY7E*tPS}lE%E{(niTs2
z4RQg(MEsXD#vcLIMjGcH;`NCSBt9SKGYiD#3kn?f@m3fjWi0e7@Dhl-7lR862c`G(
z5%I6z(6`>3!j*PyH%c?|M4p(0q{$QYqFdI#)JB&Lp@3%E%fFI951od9Q}Fjbr;VQF
z&4@ZOJ#_bE<F2}1ZEG?dEl~H`+gj^Qoi!KfObijLxxMIb;hFR`bk?u_*;l{%RpwY8
zN1Zl5NFt5KI6yhE%UV_$TUFV{6xp&dpyRNR0Fg0NdF4#0J)!phM{VcQ+eQ_|amRL2
zCr)g)N}q({meM3hb(_#iywoCrkSaA%g;+&M;Ml3-;>6Ll%_9op3v^MWu2{hiy5d8y
zV8I3!Ec_9=0tpETvEcVVcV;}+IB8RgEss49kFRI$z5nz0pK~f&Sodn{)1NH*d7~Ha
zv|DJKzOIYW7w_u;v4ak9efuCRv)hh99`^$M7>dxxPM(~#?0RkUQ#ZrMWRI3L@u7P3
z0>7G~I*Kv@`syZgrdRnS^>etnz4j!-XJ`pY^!dou9l1GV5Q#?c>#TH1RE0k&t<k!t
zB5fH;l`;V6vPkSI!%JC6a`Zs}wgRO{*=eLuQa(oIGc`pVa9=|nV41_bp>)4n6B0=V
zHX@nBlNU(<@}n-2Ogbsj!O038+I<rjA%<^hcTSB^gS`znVc|9SGUT#e^M-oO6Ipom
z@4i->rfE`cJev`@6dUhpr#&?_G<@t2Re6ITW$+k4m4`2U{|Ts)y>}R@X8K{!gIQy5
zZdDsG6Yr9Ts}N<BpM9&n-LCao)xdsR#<;<>uM~sc3JVTy5u2vqt|1UmH!={=I0`_3
z@cz9ti4(F)oPceJ4rOFh#YC#+9fnbZfy}BI9?c@cY#&)CJr6W*hGb9k>=*qC_8-b)
zf)D{XtWlR4YqKnrY-d&y7t}_6)N_)ZMmO~r-5_4rDNZ0E?Y<a9xR1D6qwqq>1l8r#
ze>-Sa&j2rk6gbvVa69iAg5(TAWFQcNJS6X{;YWlP_eP+FY=SV5Fen&q_(+kvaEd^X
z66|4*6u1kFiaEpnqnrhas6N?XIFk)2!!~ebcpF9!l7W&53A{1BKeZXD8asWGY>M?6
zxc=1bLT&XZ^ARXBO@49sQWz!TC{qMwpfP|l*TI$25XzJcWlF2N*A9a+p|7E_56BpL
zn3uYLWW8DI@p>07lmW{{?p=ST+G;ko={c&<+A&u)DD$^=9SJJIHdxW)t;Rp!|MBa}
z`t2LR7D5$*6rFKgI{qZMBs-j!h$l6#7!WOLLbACg0wVX4wrOPbvsH?pYDwLv0S+CF
z=~2TJtS7)BcTs0gvFJ|*V~w8(Kk_^K1JiFBK4#{NfJqT6;;Q<n6?ytlIC$kU=Xsxt
zgVcos23r!Sj_~(2Qf-#m;X^2DdXS>GGrtI?7%7<R7y;RCmg>cl)R;i&*Rba4*V((o
zIt_25UsGYDPPO?xOD0RN<efg9v#O|=;tbxV;?<$=(E{ryVaZ^3i7PnhB<$w#`;Gp%
zUDJ?9YO0RE#k%TtpDX%wnFR-#s2fdM-ut_({a3VD(m_FFgK3uY{Dwj3U%C@QniRsp
zEYCAs>nMTTsOc<*f@4kzP@gpb&U4P6)vPI6mCIz)g>m4o?6iF}{0)<eac+98EYQ>f
zN-^$D)kUut3>rg9an5hGZTb4N!QiC1Ij0aziv6=gJd1?RyZeE`TbW0sbO{dO9bO_S
zCKWybmhyr$Ed|20@g<<TaX_cNwNdWufRK0>jImR$_8KZH;!`~asA<t*tF;j5)t4j0
zl)fDFWio>Aje(dG9x*ZILDZG!=0hqhn{Ke$rSF=jg`SO;alJcX3}#ix0AuOFxE155
zx_kB}Msr%%+%;!}{Uu{odo2`7djz<u`PbR(-MQ1+AU7i34_{Z*NT0W9Wz5rOK5ubF
z!#-<hdyNflLJOmjM*TS1X9d3FT!tVW(Ft?q`B?)Po<I0*MNnMhr|c0*v2U~XO!>~v
z{=8)Gr{*}9^fxsIQcQ*c$NB1G{@`J_&bR(YxkI%&KO--S$sYH?ci7ALK?}Gp3aWF+
zC&isk;m`bzgxq}Uk{z=1z$6zPV;)Owx4ex?u_O)mC-#IIE)`h55nGx%EWz1|ob3@Z
z29c|Tw>)BIN6z?Gyy^ndPo9i6tf9x&kuZ*t-BA)oP7(%f!5fKOaX!>&$y-T$>lZ~l
znn@k^o&9y(R}Uk7^ej?=(RJxg_aFgeluERFVZB@{FYGLpHy3$ZroymKBc)BO))vc~
zY)kjkJ4?KlLs26^<c`Z8vxk#CB)vKtx=l$TUukt9kJtdP;cW7TA=ZCaem!X^k~|KO
zF-~guC}MP={1OL)AyZO@T#7y}*2^0z<e(?)Z&Q*{g%=(()RK}$IGLEFp`kx4X-JQW
ztHUt`7;*+z(8mIgy4%Q{HrbjKpv!(31@|>w2uVPKc(QRVKIV_b06y!kGE*u%$cRq)
zkv3XWa-W9F^a|%5Q{xJ>4WK6+vM^<QrgKjF7RTOSyx~ldR`2QO@R!#%E8E*XdUuvX
ztP~b09X_{R_^cb7T_v=q%eU4}`Qt4Dy#hNnjJ$R}5_@8!xlQeD?`6=v^W{x3h_;&R
zPLS(#)d&$MCrY>7Zgm`uO}QVmkoLBA)PCTTrl<gwUbo2+llsPdqCK`3INj*LDN+sY
zbqe*?EmW5JgIgPos+3Qs@Bo4z>RWcalU(MpQ&-5+b9nxu^hbA03$aYsiuwiwgI<gC
z!&`E*Z0*m-w?@Az*216-T64#BBALk7P6S_p<xr{iy6vFin$OO0N+&%xeA0kX75=ua
zfjh0!ozdbso%3|;oDl`9Vp%ZewBMJXsbIMVyxjBp-UCf1hMkvL?H6x_w{^w-qr0Mm
z6r5{;X2yCmJHA;+Oir}^6C)U0&^P9FvfP`#vDbAkEe&&)Fz!(EkUK@6#XaOnqDbYT
nC?OZKmQf~)TxMz_m!17GrRPS;zx(CNv6orT^D35>OV|DdrFKfZ

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_0_8.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_0_8.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..e2f715bc4d8a7f72460b5fd5b9e4b8788b2327ca
GIT binary patch
literal 54425
zcmb@v34CPNUFTbMYcG{{wOYMbFLt-=-n~1H_q5~0mJ_=j6U9zN(k-c_mP+bdrS7&U
z1(SA&191X`;qw7vQ0L9?Xdp8W$S}jc%$E6lV3^^}o8`_t-b{u^Ff)9B7hs@C-0%DQ
zpIfWcPJmIjj?O*zoO91T%YXm>?{aT%KE}^)PrX_`_$~3+f73<#Pl}6&`K|tHJQnj~
zlQBQ;dyDbOxc&8(5|fFzD@#tMtSmj5;csFwyOf*E*}ddqeyL}&XQ?n*u<O)f?^54n
z-%|f%zg?#n2bKmW2cv$5CWmOBSsY#(o*cHa?BdAMw#jXl&n=EFZJ*q3`TXM8(vHa;
zmhV~IxwLC?*V69E-Aj8W_blz5+`F`Ia^F&MvbeN=a{tnS$pcoeuy}Ck(BvV@_b%SD
zbnE1;ONS>9+jZaKk)`p;@uk})Z?o(E#oL#TP99x4HhFC6_~dab8(2KCbaL{f<p&o}
zEuEe`O@7E9_DB3}pG!>M;ot6W_s3q1Pu}V8@OQo%o4o6Zn7_;4{c6nLUCu1rT|Rv!
zIe8Cdd;GnW?JeJPB{6v~*Zce;*G1K#<UUIF`v)jFP`;Pz`?)^oAL9CuUB83tTl`zO
zzP0=g?mxiwVgCr%N9_8YT#x&=aebRzKj?eU#m0|Lt+0&wOY@6<acTZ?Z8aztmuK_m
zYPFT>{U=V$&DYMYP9L9HUOKV3Iy3JFQ}dPT3BO!fp0CbV=8pOEv$Lz!`Q^&7m6vPF
z!OXc6(~HZ~C+@lD<lWQ$?9AP>cb&dx_O5&VJMNgh_s;2irtdm^x_tMk^4;Y-PThUy
z^c}69O3j{1jh>oM*?jN#%F9~~ae1X&nVLT_zqAr8Uo88r(YaBT@4NHPd+$4a?;X?A
z_ug~&)Xb?n@43gn%l8@Webdv^{%QZd`%a#oDK|T4J@~}j)M~XlKUHZJ-DC`pEibKn
zV7_{yHnpS)ss&S(>g;l`R1T{7W5<g8ote+SsRK1J9uKomO)X9PQ%^lR?uEIaJQYxV
zt{kR=@=AHCMq%H~a-~+jTw9u|OwE;pu&1)RxEPhuKC?0v)aIua$tG(T{3ZV8mZr{^
zOLM`z&qW{ooiDAGgXPk4W$|T7@=wiIYmY5gX6NU|vti%j@?5EQxmKF0mgbk{!gTpE
zx2s{sFV9Y`F4n@-e06DRCCsqgOALVC%S^)T;#4gx%q*`~Y9&T6wNwqavoUH@6$Vo)
zmu7?Ir4pmmL*iljsrlNwmZtQI6=q^;v2=bh-2M3c)ZB7qYVnC^W*_Il!Tj`Ujon1m
zp(k1Y+I;QhGxI^U_FPR5?|o#owp^~vEc@l)y-$rN!XbOt>2htVRGm3j_E#6n;V`om
zOwH6v^Of3isaCF3mxC~`4O=cP&Yv$cpq|GcfAZ<)rj}L~%K?`IPwLg4TApKqpPs)w
zUkS4_8vP6}*+VW^u2!dNwMrO&nP;Y>JGoYQV#;4SJL73lqaWotSFy+VttLt8F%DaN
zZTgvu<=|2<Un>`vc#Fm340XOzyu>jp*3QjWi?geh8O=bkwp`T96{|XtR_jFh;?$xw
z%NP0am&?^+Ww}<IIj22SoT|LMv>cR=UyCt&$!d9VHcaXn!66=R$@LGNIJdl1K0%}M
z^z!og6aG|f>cm{Iyt;DYxz);3>{lkSazgWa;&D#)qfb72V$okZVMFEEHxBy=-(OiO
zU8vd&mQI%L<t&7K{(N<6da>-6Y_zxN%{2|lL~w*ZGn$3QkMhDJHa)w^)MNGdSgaOX
zh}S(oUQf)&KOG;TY`dn-n+Vf>`QrRcnKw%rm<f|?Y=1l%dci4vPlu`Mxv3RalfGIn
z9318{99*r;>jZKFO0JvRDYmigjt3{m&+yyvvp2e`uCa*a7`QDq$+&#aPx#5Xn4j{~
ze#X!GIY00BkSqATexIN8`}sG(-$8y;{Ojk}a<h9s5u1$*E4-SUOyuIZKL0(!4`794
zIpx3CAN9At8ov^sOqVko*{iW~?g~$t_53k^hn463o&GLxM~}bT-^1U6zt`W#-(J7y
z@8@ryf51P;-+r*c0PTbcl9Pk3H~%Ta1nPe%DtX#)fl7v>k_rE)f6SgW!g!ARCm7o{
z|D=D4UPt|BfR1-ivYmT(`gf5V^Y8ZW;oc7aUjIJ+?)2~X-@)Hq{saCy`3r8-oA?j<
z54{@WIeYwv{YNO@>p$v0#@~I>{rCB2{3q<WMgQIYlib<=iC8i=d4QFFf0%sy@%KIb
z=I<w1$MoWS1qc~>i>D{ftYtY2Cl)!GXMs^Vzi;Zmy-CH`n?A``jy<;fzkm7eoezEG
zTSxx>6aRJX`yTqr*RK7^Z$CZ#%0pjitaX(6<A=gTd8Im2Y?SkNKfhUi)n7HF=UVJa
zY$JY^ht#+G@r$u)?|WjtcP0L@czt_4R`P0zdTb%NkouvRubju<<QvH=acBW{cPU-(
zBhA!%Nwf7lX|CQwny(i~d+Iw#3$@;QygquxBi~o+r<XqMwAVeJnWJogl7V2dHmEyS
z;*<>45`LyOTpOWeXk*(|=J`sTCuiSC=*jglKenoRqud`2%Jnhw+w0UWjn#)qcht9$
z?yP53FIcT-x!P6VO}e|jH+ZGCr=GeJufFo07;l-fI&aX^-g>H@u4nw*>ri^sQ`tcE
zv#P~-@_KUn80-6jFHqiNBT~&Dv;H*VeKw!5g<^fEo~V2E{`x?D&@a56rr#u^*k9jC
zdZ2c2;ZS`SPq~FN)cd-(UL=1Y__wuN7Y>s<!YuW@u30-$N!Mc=<9>g=cvZ8uXXSTl
zw_VZv-%h>Tga2MTx^Qgc_*Iph@PRG*9Uqt&PXzB|p+oQT$uNF#?c~gA5P*FaUoHx3
z0)tmpmcg@?x#9sf$^l@lUp#R6WR$shV6FH5;<KxjifRDP*9MNQ1Z98TL2^OQv!Iy_
zpnrO~T3#D?Z@Ijps_%OGkt)<yF^nGvvh~c*fga1H^W~STVO|d|8Q{HMdbV6G*X;4-
zr72;C6(I$?S-e!9pF3Bp9xt91`r!X+wLH7JSiA%zF9zi*;8KqgMrPPWIyh$U1f-ui
zzp@OHJbwN0;_a=U;tP*FJMrYi6TH#8o_p8Xvt)}C&p%yau*K)!_xK}c-}PMa>@&s5
zcRl+|N3-#C*fTeOv0Nz$A6LWPPPCk>%`Y(;rY|hGgIt1M2x?(Ix++(E@a{5XWw}&k
zI`Xsgi;E>r`Dz(-Hoqps-5ZoItj_aR&^J>)5Mg=l;#4qx67atY^bfNou>O3lR0;=-
zEGgH@!Tj2k5NeoOVIpc_u2h;?oT^r-L+#4cQdzaZ>AuD@S9WtA_RcKMuYfL$Zt%k%
z(Me?%@a0l1?4K1$5>;ol!n^^Xk_O2ufQ~EW#cDXlN|aW3Aw3?V#|RhJq8bhexf<F0
zLNK+Wfv0RSgCT9gjM}kI!HBMMc1PD?VzFEalZ((vVP<-Db{3K{>{aCk?GuCp&1^(|
zg*_KF2F(Lx&<LAhZr%p7IK5nQdnfE)WauJ>8U>l9WoD83Jw{(Kp|kU~pjX4`v-x%O
z-T9STILKm@Yo*HSQfYMsdZJua(Hvte2hDMXIRWZY?d6p+iw_L1l-$Y%x2ex0`*ktw
zRUfTQ;#DeP?tP*ywTQtps@UJ`*}Z<4)ENs4t1}R>BBYqz(5tLKKXXEtm%^TB?Jv}0
zjg?Wgm7p>=zCRFk685ZC7`h`<gLhLBJfWnbdpQn=XxrMk;89&eC{OuDLzmPlm{-k`
zk`F4GRPrGu-=gFNB~L1OkCOK)IiqAw$*hudByQi?{$S)f_`$QfS61?(l4aG>=3|@N
z)(#d_d{M>l3z~ZEjBTw|;Igq2Hfh)5!C5siD)6v!?<YCMZ&heAzkeW}j_>5R{a+#8
zXV-=J-gw@#JC<fG9f)V+1My@$&)))nv*h@vdpFXFv^PZCk$Ax?P(r;MYrR6^#~WvJ
zCc|?ZKYe5*Z+MK~>Mues)}Z>X#Dh#dzMiS%YS4pVhNLi#AAdcyp7FhUW-PXzt$7Ox
z@D}tK_#jvFHj>=Q`A|(C_jbgtdh7XmW+C-rd_7T1*TIHB{EXq0EPV<4<`z<ZVh5OD
zA#>iVJ`6Tb3!B&SwVs6nzrB8jp25)@eSQ{Pow*V>oT6*t?p)h-o@?-4^L|gghY|JH
zb60tnW@*8dYMf)Sjyt_}XJCsveXTaTI@|QO-|+`pJ^G>(2FGIc%m@@pJwFz+zdUo7
z-gZ#s+hgm{9lXgPBQLa#ambB?(GS%N{_s_=W1J(J2Ya*pRp@g!-d#`~&3D{Z?{Uwx
zTG_2?Ee!jk=e_Ta2T(x^BmVZ`m=6WDunl^@r=Go{e)R^U8uzZ)dT-@Eu7>?FYjeQg
zp`L3!8{0Lz*8k4NmEMs0`>p;itN#%9pnAUXbFRi|nXK*bcN<OD>kqIudi}%$nzKHC
z4|ynVQ9SerRkE<_YJ6k&)fn&bqUaf;B#NRWy3YLr?Cbyn680}lPQhNkd4r#|+ou*+
z&P`Q|=cX>g-z^uX%SCu@ke^?KSt*vMX5a`Tlu65!Z_qR4aBH7yT$)*xsKXD3JFkIq
z<%{8P^+Et+Z;xr#++&5y;8?^HtJUt?yM^a&dUuC44$3p2#;$&MwQen)>bm)+-aZ^^
zq9KrafGB0A9PVwE>A{V1F!VfV(Vsi`==T<`e(qnqBOGZ}Gy%lSa(VXDmZhhSl6zB}
zXYe4&SNP@Uzci`-hl7W>UK?0iHSRg0?aDrw>EhyagaWH)nDacCxhhOnsg3V3^cU=)
zX>f@o%rxdcEeu}qgX1bqKw<<JRH?NK1nwN7OGna(%q-ZXcE{<UrZ!1U9msuo&M-rm
zSz?zgtu8rS@EvOX86}@k<@8x6(8#e!Rg-6g!xPK3CxyP37--plR}g@?k=s<JD#7=w
zO=4<salFrPp2I><s5?V=tx+0|(!;{ETUvWw@R~}p5~N7F=LVuqyO``?iL&qFw>nKC
zG?Di_&_UMg0~KVwyrGk^_<+YhkiuwUCrD=}=w&34j(az<poxLRUN1{+wHhNWB(g$G
zHwxY%&`g1R_C_{Qp!Lr9+NhzFCOyy~do~GSgj;4buAe3OcAe!|u#Ysp0l%=2*hpRl
zs=_I#1iY}2uq#gUYCPBn<`6d(e*-=*9$VY*$36i6_k!!?oz^DP(Ti*MP}I(I?HYWB
z{!$x?;ifGoJ6m3g>EC!_Vr|F$#pe(bNya5$AzC~l{MU9ts2iEDTM-e|xH+D6G|ESK
zd1z$TupE3?`FxW+M%Fj8yaHLf96YBx8Au7$8}}TM#9b3Wb}e|A6QNP1Y5AYfta^mR
zdqudv4|<?4?vd&$8y|`%zn=8cZz(?z_rC6VYkjtO%?>(5L)+*m%dZgx>4XRZghHlV
z039d&M9r%wHxgoXFUEr5Iwd<|>&aH>da9m!JoeJ)db;lMcW^yZPj4janXB>OWDRl5
zM!KFBQe)AR_4GSq7xrBcYncI%zq6JV!~JTkmfHY1ZS;sy+$gX`=t-<KWJmAjdwvE&
zB>Q@Tr~TfG5F?Ntxq9+qJSbiG0ME$n0D(YA<Vg!PNQo<PQjjiW#|2}W?aI%cPX-?c
z=|aSagaMI8IqH4fSY`+aA+_oQ_54QvRZyMleZ9~1=jYFRu-XIlUJzE#4o1v~3hVud
z9v~J}G8z+7tM@+;yYM5G?}}RG&Zl_yd%N20UGI-d@=;GC>jUjQ!K(N8+#8L}YP9rP
zNAN_?|MbEy)M6VD7P<>s9qj6-kA4PQty7H2&)<Wjg|@#>OA#Kf<xp44-p-ahb*Mh@
z04VrEsy+ni$G?wz8qN9;?=~bMNi@?!(|6W~HirEIBiiC_^%mY>XpiR9W|gsgpq^l^
zpI~o{u*(qez3x?X+_$x^S?%HV5q4F-0g3^hHf#czL2@IO8+rr65u(BF_&w*7jN}3>
zhU>#W6mtN=)s60bJNHKFBaM3v;4!{!B23JztX_+U1qAZ&Pe`Z5*5=__k+_Qyx?X&|
zyjV6l0jxyvTh|@e->PtYMwqzq!`VX476U4Ae#L9fvD;rt^G@+~uO9y%?*a;<*0D$Q
z?HxZECdD#`@lxxCld|R?5Mzp>Xyw7*QHMgnBeAu^cGNqW#^VoY^o!L8$u!!m9XK=Z
zL>$X#q9WKPP**8mYFu7JuI-nrGm-(<rWS+G@`&KmO7vR6t0ZTbpS1J{%tl`mESm7+
zdfJKEulJ(pLYIS3L}Hxl14S6R%PZv>aA#4nm$m+<qx<jW>i&FjZ8-mYt4dTjJ{mEv
zVhjUey)dcPwje>RDv`c}sA&nGr)rLKn^MIb^8cW=yo`__g}4BP!+>Vhce)huyD;{y
zB;xY}fka1Qq-GYu{b33gcMi#j=zuUK4No=9tyZcRR?B6?5dFrlwwB(&{`fWr)00iU
zHV~K%y{aGlsxtBOp*Md%%vK_;PVh4-O&PKXbF56s#O&yC%9q-+Z53QLJ|D&x(7Kcm
zPrx=W`prBop>W&;hxyjjckQR(JWu{@eyblN>4RtxubJgv-gr~*74PfWq^^v7xZx#S
z&V%ge`&u^b*&Vph_--!`S)tZL@li;PKCseI{8r@>Ly!`ET!}Clp~VQKN7nOR+3CUE
z-Z%vDqJLnp8*9T`Y~&dop~g>|jE<9LwE8FsoUBk6M`^*klBmPiz>7jnP?~@{i^@S=
zs+VkDryRdKn~>&XYA;*c)ICLho+#JQ62T~NY8tAt;)NrZOSNLXc)4_P{CKnK{k9Gi
zvIT<L{nqS56UEEL$M`*d{P;^nM~h1QXfm$FauxAmWn~pvVs|f5djvGo<v=Z(nxLXd
z^omD}nJAtbKVE!yIVc}3o<z}!;uG;H8Z3l<ZCwUsl+2a;i;om*s}QZUrRLFMgZ4jK
zZ07{Vj}{?b^~mT=4p)n|1zQ~kVz9<j!A|}JpHw1B+)=NhNXOF=Artiw#%n0XcxH$S
zA>u;{?a%DmiDCbRlJOjk+L$H8sydG&nGq<Oe0*)h_C{B=4yr~QqoMHns%WLno5M;Z
z1b5pdF%laCG+&EPTuZul`xMRGyS+|sJn3F5!RiNsxM9lHv3vAmktEt}1gDv`ihFpC
zuj)01I1t_q@5b65dySjbzv=q~|4jq?U6RH~TdVM$u2kZ-;P;ehq{e`Ck8|f1&s{44
zp-UGJHYR*18I5$`%|;rvxanApPVJg|X1Z(cIjT}T+B*e*pb`IfC5M{HVRB(}W|3*T
zd5xZaJL3)Bq)OL}cGv3|rMSe#4yj5C7>$x0(EuOOC>tf7iE$d>BPv}_SuFt(YX=o<
z!ZU#rdmoO2os{q45pFkquX4s=emA-CjP0x7|5j;p$-=%)2E0|jdsccbwB*gFv~ZH}
zHMZD5($333bY7Z$n!vPUI&!pXPRA@45aLJ%>mHIlG(!t81Q5H@dvRbmG)VtmPpqem
zK*QFEd%NmMSb;=vu%6mLO6aGYj0#AbQVaBIi5C;=>GjNdR@$t3)*x<X!sG|nJhxN7
zqFH#5L>Yem_*V?vTuXG#5Af&1+;%Mc?~}t!U=j6rDflBU#&hnh#xJ{GB_3`F_Lo<Y
z=3g!$R1q=fg#7|HHpAg)bBZJ}Zwg4dC%u2A6GYvwi67!A1;En~ze)ZT;-m245wPlN
zFr&F{{1nJE=U(U+!Z5xN`#fyeLi~H;&q7RpIKH0J+%@Mvfpk{$nWZ#`E|mJrY4S>9
zJ&!yPz4>`0v{wOX3$SN%aX%?ecL5f~_ax{0JiUH829pGP<frPnk=RD&D&_UOJBs6J
zN!I?9v9xBeMfxG*HZt)0xDU%~@PAx&|5C{=k}zc=7vpZtP+&@h3FyA@g4^$%eTM~S
zfm1S28OEoBFRCl+^na@9kXxyaMbRD!3zg-H+1o5lRnK>?Q+B1~j3&-#fh3#Q>5W2R
z97W&}Ny!^EI%o@lX`Kk8EjGw7@0@{k=Gl)s_7`b5b`zWePYm!}d_7%D)>1X3=G4tx
zL4sj==?wC6F{+1c?U02dM~7LRjk8*?)(d$*^}5bikIn}@X@v`;P%SW_kJ?qnNB++F
zdIGk(x1N%`<U+CT1rw6d&%!d)Q&2A12_wo`8O_=Cyjk0b=B@>j#BjcBE<a9wJnr-*
zT0+EB-@zYu^EPru=lT_+f5R+dO=#n?AACubOwNXyHx{g@W)ix1&UIx<m>-~!-EfPl
zYD<J8j`WaWP#J?PU#3b`_oy4XpO6JqZ|B{Auhw0FzqheQoXO>tXS9^NsRrrDQl{fY
zh{_~#tKFP7?-i*>lJT`6hq0pNw}x*H(5PdXb?BN0?*JECW)@jXtwj;2O9I&wvyh7%
zI2rJHn2`loRU#&8(?VfZDVj*{9_FW~5FD9y(kix9>w2-Y+ngWP8)<W<k>g2}lfJQr
z$rLIacfSKrC!;Zo6t3P+0{%&uu?t80)$A1yqE}lGAn+Q5@0G-B9BV7#9b8Er3~Yuz
ziHR9H==f2$!~A|#<3TtOe+`jBgv!`Nwng&N_MuY5xXYG`O=sB4-tI8%!t4V8W9;l6
zoaNQN>QR)BtvMp%LbN39OsT8O{Pj_@o3Q*&$|B;+J;q2HLlpKvEL3BTCC)bbE^9)K
z)B+eAH{mp5jr&_psGapAJj@U}XB&7aJ_59FC2WJJ1ad{BVvAg949{Sijm*Z_wc4pU
zw1*cLG~73N%v#=-*a|xa)UhkiQB%9p(C@IQ)iE=RVSgrqK_qh?a;aoJ=|r8r2cS5~
zokO*RjAM|!(ZWzTHC)+(m6;+h;ab)*K4vE$iv#pA_Bj>Y!84INz)K*MN9*aYnqkG-
zzSYY4%JQWO90R<I^drT?-`dnPuOH^{SJX5yHC+w<A9A)`f<R^YhN0cmVwhc$X&`+B
zpXXNN*d#@ehH2O*ro@^H!ceJe8nc=)-H;(|b&MYJpjtcjx%^)w;SiGPk*_6_*|#DH
z->Ba)$qk%q$4E@y=3^u<@S6C>gr5L50}zm5bBLJE)O-v=kK^ShnehGf1iI6`cAesS
z{IwX@vH?o?X|C_4Ws;uQOIKAhhsa5AKGCoo!rCXsvH<;M;r4oBA%o3Z&*$S<<{$#)
zUN(5v@2$lBzV#ef09`6A``_^DSz+yYv~mbnEl*uAI08{HN?l^Yti$YJ`bN6}+Vw<j
zR1YDi5w=ieQO|?wxzVe`G2=rw2Nl79w)F36y~V1z^|DiGD<^;ujBw)>-Nc5x0SP2j
zt5(a^&5)o$?zK+3tc8p&bjqZ`_caU^B5RZ4+K_3NUWQ|DYM2mC**>B>;W{V3x|<wA
ztAszlgj`GV`!Ipg%k_d8w?&2-*K_U)1=aPGWygCRa2nN1v`A0P;A6IngZ;{B)1$p<
zm`0?XiH6x8m}<!E5gHVb!OZ3>iYOKF^K$TK>cm-aNXi)OQ_n>bOn0D3IVQ97mt~}Y
ziHEZz3A1*=Xxk>=0pi9ae;e(M85`n&a9m77BoR)Y>mhHq*XNBPei=e6)8~yKlxdPj
zZjm|^J%UhaexKpDDsg~x21fV*ZP@gXNDw#(Ss;&+8!3ZO+TdW*Iq;H*&v}l+-^J(@
zjGAzyOMt3Z+>oJ|j8IHdw<l`C(G6M6#T}~XkgbuZsCfgdeA4GH>8~SLyMRUpnIL!l
zAJ#M^>Fxjuv9*Fh;OJHdWuPho4TcmRBqJ6E!ViH`J+Xjh!r*bg2tX?}AR*k#-9!z@
z3)vUnNb}~D@Q$g8KpSt|J3euJXu5c=c&Rv3JoaF5nm=5yA$ym~m=nKTa>7gIVT$^W
zwAMgE4<(UJ&7c%S^K+}qXp<vJ=U=JoNIsD=>Ld6qwKjazH4{03Lz6rGUCo8ypvWjm
z_P~_!NLSRXv}xTnGd3%ZUfR&iu=)T+6nu5aUgX&VmKc$UI5oHc;qV|2Gt-^M!o5=k
zHb8Cd;DnN{q4pqo0w*9zDX;t__p91-eeB+Be9f`6Z52A8dyn3#G0^`^h9SnF!tW3i
zhf_+T0udcA4Wzr$GNPzL4;c$~`0%N<I4UkStdudP*cmYj9<(anww^+wn3@w;>jihg
z&sxoT1~tb13lNa#HlZM{dQ3(Zn61s9!*1FA%f6m64Jgo_JK3l-k4j5+n6ZUiut*R2
ztKte!BKZZzCP<y}d+YhG+F3SN-$WQ+363yBVFcs^?f`y|P;~ZMGR$JMFjtz!mRA;h
za%;$7o2_XLvt-IE%QNRNCd88jvpE%?F=hf{KrMn2_#B{hkxm^WFs0lY30(KNW$eU*
zhxKrv(3~kof*)3eCn*bj)yglgU|NB-9&+}A2}-di#0tJT-s9K>ZH6%C=;nF<a@gB!
zP~y^UX1$EPWMDi(YuCI-H@db?ZvzI~>--cosyaN9Woxe|QG1WEk&tS?HIOtbs~dUL
z-?FEY)H>-!3`9pa9S}l$QCm!hv-)31STfFn&q;_*fV+lVuxqy32IkD4!Lg~c<|iB1
zQO!@rjhctFcs2VPoAyflld(_4F~kL?P!hM2_zN)~>31@A1*loco{Ci;xdaiQe;?$R
z%_0$vjONX7KH6{|M#Y=TUc(TpyWVT@&N4H$$whEln`?#{ZT^V97fLX`LBcM8d+ep$
zy41)p8c;r*wa=P*3qY4NTT87MVBx3v6DH8(Kxv{v0iy}!sYdCfTT?@0`3F>N%C3LS
zlZ~U;8CyGI@Ehzsk0m6^Ia};Da-!Mjro#=M)8;v=#CFgJ$Za0Vsn$?}4^rdbsQx6$
znX|#;YLT!55d0)>6FjW~Cj$=*$h}XszsRl0q1GE#mfcHkQRbdD?ge-QV&tgACZWIU
zMgJvtjQ*90htN2PoPE9fYa^)<+uxDssjFGXTxj>|*D$%3j-3+=WQ1otqzKD~M$}0&
z6D<P=r{YSu6MUD3dAE|TlQ_h(*eL@3h%Lr+=ih7KIzN%Q;&xlc&GkBPPm=ojHQaNQ
z447CJ0v^0Ar!L}q!N-;ZA4>>^aK_=^<iz#SN(0+E#9MKw_xjj$G0Lj^2v_4xq@1h-
z%a>T)DV`T-o&rgQ8mKno*z;=GSCKPUwS@gri3VXIvST){u<i*NE(y6(EfQi&CHtQ=
zBYC=Zrk!GE*2ZjBwy6B3z>X2KGn&N4Pd^#m5$>yArbBs97;s^N1Q0;c86<@nFc{|%
zR1L~*yG|JSk}<0s3^wqh@MIPN*2@4Qya!^n%oXfF;ZY!g>`o3vI@rpv<6U4V;i!Bh
zyE<c6IUmg%9MJh%PpyFTDnAFW;zJxD^+HmKS%ok#Puu8&RLWDwY&c~Fx-fuI2X>w&
z7c0Qs4AwCvK=vhWYN(E-BclS})|0^xQ!?yR@|i&ceIqQr_$UG1-u1pspz_llpt6^K
zdSrD-uYJ~QfvZ&T^R;aoqcz@rW4nxBki2<r=8W|Fsn?JI7@)>b;Z@{%0?W?<>%hnT
z4p4?^r`|K><9hm(Z9A~CWM9Dn7AWj(>h)~fWZ$>AY&!X~9R5BkIPuZ6s`fOt6dR}G
z1a|=6QMxwZn-TDB2fS^&P%w^R7>>5XY8;y2T8hpLwvNWu4%+q;MmJ;ob}p?6R>#MJ
zRlQ7;xC;{~CLPiG4vpVrj?=30B_$?6G+CievXc({ka9-W52=OmSCS+K65=?BijQ(+
z1lVK3!YZ}{r<?|TyuXcrh)<xC(@X|x5%xslM_CCt&}8JM31B*TuK>^DSfF_h7M0kn
zRFrerr!@;-rm%W1!%M>~3F2aG7f)j(*XQkqZg6l|V#q!(i9I&Z7!$fa=!v}?=^>9}
z-QP8@9gC@H(XRnH3HLY$vGuqU=s6=Sk=YAL46*bVxigl<UM;<lS;$J%BOzr3^1KNa
zS+vo3F6=cY0NK0;$Y9NS!ZEQw#xu<5-E7w#gOM`gD@;A&Pc6OB5N)ZWFm1G3HB3;W
zZ9~~4jX1l09A!#P;;ouxLm7)9xn>%IOJ?kAJ8TNJs(&p#5uDbLPASnm2lpw_R5<+O
zorSAE!};iCKxEG{KnWQQ9-f^TKNvhu-7sI0UkUE1IGYqor3<T5i_Z1LtntlyG7!n)
zHke~MLCNxRakFz?H@~dp_m%t)C4Z>ok4VBK?wpHG+u`Zq693w!bNkn%0{c}aRyDXs
ziJg&$m3vmni%MQnV*7kmxjISM8~Mx3l<~%CxBB^*o^D6=B)Jp(R$nE_#@K1djQ?i;
z`u?joiRs1-@7_W@-G|BGgV{nhoy}&`eYtdjvaHI{YNfNOd^+Es-Dh`G=`oxUdb7!N
zhBdt<(1>RE?fB9B>Zke_yjRPq6L(B6m&he^DgJu7G`Xbz0RyGDc;c>^6!3mEHkm3%
z&Y5<v$1fn=2IEiWa(%f${{<^`4xV{)?Ub{o+&bl~`7P%9AIDwO9RG3GG-rR@HTTKM
ze{#UN_D>Evmwo~se9*tuKa87Y<eq7xnzUye@yA~U$xV(#4wN7AZ}V?A2g+^!hny4V
z<Y;tHZk!+j+%}yf=a{*1%4ze($c=Mn<i@!xa^u|XzhsV_IBfDv&;PLhKzT0?rSJ3~
zd^Oc@+SF*cS2CAQ%>h9Q<m^f5Lz+eZBj)6(JNuh=Xu+K+|8f6aHrfNx9nboj_Mh<I
zZRH1}u|Mg*$Hsmrx_$}Q)2FP)EzOo%vy?yMzt4Za-MQ6&)_;!qJRG$->pyRO95I*D
zagd=O_CE#8snnW2SZnzu`Cyw@*hFwAI|MRXQluFPBN{)_ri#1~SyTNeI|yfmSV@`;
z<V)tViK_)A=KP7X0Hn(Kcy*iEZOXk;X2<DZkDownnuKiee2Hd{2%uJv_#$L_abbbs
zhb-!|JiCv4zvXk5$Fh?;`8m}Ytfj7qjW75pP3PnFUP-SYHxL8k=mHson<>JDQf{qK
z%OeHDsb!&8av}r;xR<^Vt7II1=F0)K?}MmXs}*xI?XMsP*ckA8YlDy=a<TxXh@?Ss
zuP4MdAZRRy(73b>=NKGJ&B=AU<;UuMW<LEB)ZBr^=%CvBa(^AG_Xm_N?5tijyVqTQ
zzu89OaJsO^Ty*x<2ZCR!?eYgyZZ&__W)P=S=PcTw{>KL=oDlWhtVgh4$pMmZ1TUj1
zE?Piyzu|aS3hKK53dxz^+m(G($w!s2kmziym?ILkfT#h>0W6emyo+SK7;Z<THB-jF
z4d~bG6U1fG*KnxAXU!@!I-7nO?#~YsE6Xdv`_$q4^`sv9z;M%?3`3!)hS#C7Byyj(
zh!2fYNCijhuQARr>-)4IrUEA0=SoWo&ydpqd~_hy;1g6ahk_Oz(-oAV<+T`&wnuAd
z4I><C<t5CMh|)BwVc+!hGL}Zm=gSqzO~PTBiDOxYIIQUmXf}I}AA3%3Jfh^oN<N}w
zkz{<x=_H5RT_~fMk|441;;tz8q~7V%N?up;DJ9>h#MaK<_&as^nvxdF*Ie$l)i!i>
zNaHY^rPUbmILaR&lkNqpz^rCHIljqK!u=biRR27LGiejkV9t)D+Icf@2aT^<$`wxd
zE{U&#bBCjsudj`D0h*Qv9K&t?Fo;-&U!$>}C1VEf@JA-H_H<w=k!ByzQo`J40~4Q(
zwDU4|$9W-h6~6`1V<NLTHc1`Fh0Mpj2aH`FPrAbb=lC=a2%b{%aU~iEYBCWVrbKrn
zvB+Qf;<6m`KGvnk#^jufqi%I&3hA1b1(N3$TA?)eH>t#2SPr+GQCN-%&9&g?Ui9N5
z@wEfpb7P*S9ZhCLc{YADLuxLjxBBZOydQp+q<}O4seCG!D{>!_YMj3>QIe>ZND=#E
zy|4!lKl}jk@I!EAU?+`7KYoVP5THeAiM|Pr5d(4Q)_UqmauSmjWaY5Y%j^PNKcai`
z%lFVv^<M>0Ud8K7AdY^egG)9{w;RuCfUNGn<JcN${tefxJq#s$4(m!q^uTDgIJ1fy
z#*#d^flUo#WgyD2Rtd@$V8B$RjOcyw<+WW04<0N&|D61_pF8{bF}$}Qd*mryQEt#-
zZBN6ET+u;}w*x&{T8|F~I!w4eJMVWVhjn7p@8^btFm|#Hpf$OypvhZ>LoPZC{=!WU
z>2OT)g|>pGBebG=)bMk2ZcsQy5o84q=LB!rutIAJtOzecuo4+HBh!!!R=iU<=pGle
zOwPjiTJSMVYV)Y$^xP@&F}U@d_M?=wN3bc(BCwLBl+He;z~o2%Z0%6@4&ADon{o#o
zPOjR#S$=Kvii>c;B-nbA3z>c1RVQlK))m$k6_E&VGnCO(OXh{OhhaUG7PbWPy0X(4
zZmXUnLIKYA9Y-j5Lnk$|=sTmzPbe{*`J{61QSx3Tos(u9iLW`56lgRdS6-w*H`exd
z&&u1oVCe|!tp(h^D$=SUC98XqKJ%t7aE4dntrtA$=1_~I)}g8DiBkFkULDxLdJ|`O
z35j#Vq=MjtgYDbNEU_%cOmJA+^v9?dOpsiQR8Wf81J-n?=(ly(Hb--Z1m8=uC0_Qc
zBKRcoavCUJ_i*>?w&TsVLEKz4_@i{mtHr7hiW>$l4c0J>iO3sx=2Akh$UWIC9^#%R
z!o?9uxhNh+T3IdG%t^$^<YjE~5+?CV;LXf-k%lIkJUpIDJs0kddXAq@R3A4oN>52t
z-($pAv#-V&fJfnwkxy+eRw}LC^1L}qSRuyT#7Mv<ySiM&c~zUu6&(-0gGpS=oo#ra
zt?g0;B2IMBVo}tiGpKJ2qZ>4lE-erqcAn~Dm~F9xpU{mXy6jxo;0wCZwTPdktfn1%
zfCj?W&S5}i{n<ooNec)^w8U$-zTMKgQ>1qd9;M$7RHT#KMA+xJn3Mk*ODM$iy2la>
zfHsm&VqH&d=q!Uuq?>)+=;jOA8uwYqbS(`65}HY}ka)5o^30ISfS{DiVQ`m5G)cam
zfp<wugz}GgA}WEQh%jBQNt%EL7OL#SnwFB1kC2S#&1x=?^eP<kF);O^aa(}71-2cH
zX;KHA1gB=sx5<_x1G;JTsz--=8V@EqhDG9PJl|9vCa?;<H9wFbDe#qaE>yVFWXNsu
zLq|bM?)*uupM-`oo59u_Nq~Gu_0MF`w{@@C&AaM2kpe+Ueg%fAkCSM*YOrnipyLS&
zX7?(?R)#*6Qdk{I@?h||J9F5#csh+{8IzF>7R!NO=B`Z$NWKcr6lrbS;ie6Uh1YQY
zxkO#dD|j9gjSF<3w{@Z#Bf>Ux`FRrJxna;`I2W=_q*n0vG?_oTc}9Z`%Kna~PVmc3
z77-Y{d}8fL_k`UH7k5k-T{WS0b&U%FFympzo#;;-Z*agu9wCG#T*CosVBjkb5%D0(
z=31H>MJ5+OC8AXJV3;e}eO<YJ)bACqdfEVzg!d96Qet4vT3PTfH_2F0i}nvQxXnW;
z1O_r7Q*jUx;}rb7X67F&u{ph!afcb+S9KfIYJ8@9HyU+f;z(E-9s5Q&j*Z4%%$sUf
z5G0w-W5_jxD{C^o_0(?F(Iz!*ijmn)6wH%NBghUWThl{?F>JCn02KfeS!ZG3VfWFN
zY*6A#0Vz>U0Uj*cAtKQIyI2_50<bqR@L(n)_qbP;yM=4$20P_x%$3Z0?Baa+QlOK8
zb0P5tBUhRDxiEeyD5{l_Sqb72cF$XuXg=o@W$+7{C&9!bl)Z_RH}O4<svR?-Ce4|6
zg$uo@aHM>rl+FPxd1p%8Ot67u<IGId6U2FNZw}Cu{ln{6w#j}%v%zd8$DsAw%F`<6
zm$|8@ZKjCrK@Fv9=fG=r?U%YepU&N1JZAxxx^G;%A8$gxcCJ<4@N04*4gtqRW&@<@
zkc+xDtwnKqIe>sY+H6(4M2N$4I3!kDmEVc?)Ko31YCeht1hCshZXLf%H(hTta`+FL
zowZq&X~m_rxs^o^v**j@73|BZZcErkNVlW}!LRAe{BtETVGMqqq+vJxPgL+vNkVkR
zm!wWqv?<O>P$4MKKhuqZ-m<h<o~;pgChWH>S4(mU{KS2NDw)UHAqM5P(P47ZO2WOA
z+?0H=0ltbEDKW;_OPsOY-QcUeNr$#kyGi_^ovRHp>=r}GaCX?vaJHO%q%@7jP2-x3
zRSi5GJ4m~Vs3QqZ9x-9K<~)sXA9GbWKXHAstlR-rX}xlnUNhFK*-iSXs8{xqEGp|6
z^}4Aakhx;7=&77gsI7hrGR5LI0xv}cQ<tLxThTCIn=hB91EN79cWE|JVHChADx_Jf
z5Q6zPw0L&jlWG+XRaX{?<z!mGhJ&ZAy#Rnw1DluKd+O!Td+M@j&cd8~9^7cy^HlV3
z$-=r8wpU1|S@qYovb%YLWGjmEa4Qa(lA4G@OTIOR6DKRZr+cM4`mlNESZ|4AwdS_o
zBMH_UZ;}_2&YuShCka?tYhVei#E#fb@Hd(%IpzaJqjLU6<?#Aus(cZH6qzhv1U*qt
zObAN*ByRge8v?RqQDzn-XkG8Mk`yr@L3$uZcz@}j5yZOth&yHM>@?eYqIAd9%v?=v
zWciC&)j8&0iErQ`17y!4YE9zc-XC3ewH=AgYA^PGcf8VPqfG{X6^-_<zKPKqampyz
z*U_lRA=Y>As8gXk;2oFaj^2B)`z@n-dgDx^Cht6GbJG6K?m1C!-^9DCK7J^iy3KR2
zsjtRdR6c1lAeZ`&xfN9IuQ3~`osHQb*D)Kq<7UI%x!G)}USl?_=9aUeTDI@%%6;RV
zpjtWA-C{O2_0^mWuhs*7it>Qq3jMPf*pKJdKHkI*4UgdaK^}9WQWdTa`;zL>;>^+F
zxueBP<E^?)WM-9#bh~1K3SAStb)!~5BtqSpYsrUBzH}7#`PTKRms<U`Xc0oT_O#VP
zraiyt0vQm>*vG_{7{qdJ_RC=pu5q{{JI9sJa@L*AszDq0M>E1Tb-xkL0TsgZ^OadS
zn<vkZ2LD0u&s6><p!zs99IpF3xkwW2Og{gkYW`=E&|94iJ)hs{Dv{)7f+hkC{;lf#
zU#ioiXG8!rOs__dRAKP%bicPj0=O3|v{gUd4J-SNAXk5`pWha=e4O$1rN$B#F9$xT
z=p90Je@M}DxEk}4sP@xu9ZwwfcE@jn(BG;wiOgng-&Ro7hU1<#BlDrgkK`*tp#alZ
zRV+nIN(H)u7d5TUAcmv3_Bn%_dJe=2as)6*XGE+Ma%jOYqd|;n_Qo=u!5F~W{vi2W
z^#?h$(udu_RUS@RCO19j6Tw-^71QMr&@l*gy-?Yv<4wuCVaE!UzUFu^eQ7?oP$8U}
zf|)H85d9{DRf4D%YJG4OWy|%$hn#0b12Wek#FxUJk#DrOGOu1B?uUE7Z%236ct3)*
zE(``gK#L)9E|~B15fz2Nsd^(uIKV+8L>T(6WW9erLGQz8@@Q}484SMA*;3I{{M6?#
z0s<BE&Y{D@Gv-&QLs&J&F;pMi*v7FR65)&XD;xX^S0W=NuG7K4;(FAX9n}Xp-g1gW
z7c_-6E6$??v4pxLx+7j~C8=w|1d+W7n$csb%@SNvc@a8aQ73RKEnRkhUWQEkW9{uf
zAsOH91dAryZHje+AJKhniQuXd)9IKB#uP86p!p!X#>sSy1+#-Wt+HQHix<g-+hj1g
zO0W`6Jd`T|B<5#{HNFiCgEqd);-6w^@H4biJaVe~Cc_vXZrRqD32wMEvZ!ejYLpz%
zzy{8VDQ29sS#dd~&Oo8obOUV0!qlDM%X%CRh6+wla^mF{!%bk=nc$c8BF&i!lQZfp
zA^+XvW!!hWC8(N0#V@L!<=y)*=9NxKGUg$)M)GrTg2uGYGcpc^su&?>8YPT*WSfHb
zU;<A<+8Fm~t)-okz6qLX$5yA?sJjk%m8d)Yid6dQM@VqiJP2nVyBc|M)o?81I>EKj
zzB%f`v%@YeV8{bYMkr%sP>Q94)$bK=BPX8TlpT3O3M#OL)kbB)PDNgunRrtVbo_|1
z8*vA&@6{AOz^!WqV@;ae=o~Cg@O8D3?kSK&Fw8dOTH%m@q-0HsTF!#jf|pG~73NvG
z$oIe`CDJIUUL^T7VF?!A812i7cSfqJpH}~e^dw8fiP%<qhnA|3E9|5IuZm;UM{cvC
zYO^5irn4QqrH9>65(lma86}<14tH<vucX*9X8ZF*b5N{$+n|niul(C}a&ys+#6>K#
z@!d_YO2B~$WF=0y=9GL6Np=q0V8Z(~K3*yYuGyr~+pG%e;Qu8;P4X-4gUM_GgyPuD
zc>_=@?|rCJ&;wqNj%PBHtKcC9*3fB4KSZp%ieJ9mW>RM-wWw+D8uP=HJ+Cnf&kfm0
zTHDn<mp60b$kRrHentcSV-iiB1Qo^z%FhUL7px}%z7&ROIk0QsRxhy=X5?*a!cgK;
zL3mRm!lV972`uj&;+AphBqPf%fPo25HwQ_{-|CsjLLzVMDQ<S)5@F>6HDU|Bbud+9
zOy>I0ht`R}s+hKVQeI*JE8|zS7!)H6BcgB!i8&%iM4nNp$TJFSI2qd;HTR}uH|8T{
zwxf_!7gS)h*uPZH=(83Iv2%T=)})0!Oa);0;m3622b7E|8B!uIBIjF`cYt5AD1*X;
zA|g1iTlJ>aAMd#mP(*)fQ_JRcg_BaZyw8l^Cbb15B)m0#n@4a#V@IXbS3IBt@wM&U
zTOu-r=!lZg*uRg_Pj!|A`7y4}kQ?dBsO7O}uJ$pJ^3#V%l4J_ra@6-Q0g`V$(JAJm
zdz$!J5AM>O(2!Vw?lJJ;>jw2?^%P0~qz{NtB{x8_w_g*nSIIslOqMMCyYOyvnwb#=
z+G!javn~A{nt;JTd|^Jqc><0vXx{%^eRdw4u!kU*jd<kqn!OP!Ve|kiuUNl?ZNy;@
zM&8?eoZ4D$0+iZAHh)LR@XnllS&cRVp473{wo;7eUFj}n!9julmlh3U(BC8P7Q9EW
zzH}L}Td>I@bWE;_h6tL0SX{+5y^tV+LNTE9+axr<S;G#<+983Oq&MRW;jHBnV1jJ`
z1X4}ja76cz5SeXk%Qp)*HAU4x<3M0Sq?eLwCXq8CPf{Sejhvsp8rKYrXQp(pn(W#W
zQCuh@oD(|sy-yL_DOj>FvWgaEaf@6Gay+FG;vJioJ;zanXwuh3e8N68yPBS_Rga-}
z;M+M%tr$xdTi!aInW`K{HpxeVA{Q$9LC-f|bCL3WzFSlA75B*^PDR?0R!0pc6x*n8
z7?#RetGBgnFr`yPbKMXR`=N=K^uFkXSNR&s5yYr^0is4J@D#5R)g&yQlrOEm?94b|
zme8JmhwUc9uoHRZRG7iQnpr#Lu1`0vPX~WT)$$Is8Kg67aRYqDXMUOc!8Rp>N-in+
zVUmVi&CI;}^gug+nG1j{#d`T7b%--1Kk`<P7v#2!>^zpV`!wcVacO$R|ETt^FjGFe
z8}KDgaBXKd_-a+3(JpQL>>{HRNEfl{?}5hIb%Id;j5C)pQ&Y%63*jIsr_A^9)&La|
z<po?VM8Ua4-kHPuT)q~>+VNAswzT<6Bz@v)2gmR|FH{tP8R;3<chxgo57>2%>qqG+
zMb97{dMZ?i)_p}^n}EfF6(SI%-}8E6Ba8oe;qy`+;j@92P7nIH-ZwnVA+t??bC}6G
zp+F%*L4DLj4pQsg=#vixL~0+>KYdZ7KlsjO4}J8|-_`>?6UdQ%76#f%g%+(A$Xs{Z
z#synj@CWdT*Zc4V3nXF#BQdq5q%lr;uMb4t>uNu&I`^orMvWnjqgf*_H`V5h{zhX)
zn`<&0t%<$@qBTkK#f<hfQAB_Q%%s9r$^>-CA7RC$-E?cZEo!gWNpZdclB$nbd%gt1
zn!<8hdqoQvv8z#QuL#|l`Y^sJ-7BKC&D2NfrTfkfyR&_ZHanZ|KFYg~was>8PPS9N
zbBppZ%6GJtTU%Fid(7Rjm2UL2WAogO?{vDcnZ`M2{M9*-5agQ;An<7#xBvp1^wJGC
z_crIhC<2L5OU|C*hycNZdf+>iJVbIPIH<xyO2$dndYt3`+|&{ugb?Lvdc5E^Rk)qx
zZ~YQK+jxmUB<<h2yI*W{x0ZDyzkG6Su#t0$eyFpx5m#i<m0G1ylmou6K%`3XVvZV=
zon~-UZ~Uyv%$wo!%I#N~VjKj=l(Xq+QJe;#ji+3u3ZX+WAnOW!V9pfdyBv9Gz<Ef|
z5^xUsl<Qa0tK=&x6YWI&RXEJWNYG%8;y9vM5Cj(*;gl-L_ZSp{tD*5Brzx9E+v}d5
zjABPc;!P`@>~HV1bopdd`HsoajvE${LG&#jHf$S(3scpc$jPyeo0m_v?tt2GYNh{S
z%XBWl<Un+{S>IuvVa@JgO~4)GKE)DOCD2i9=K+ySh$}<~Ho`Zgy`99fQp~qBe<7*}
z<seT`7YaOy|IrX4OZ=5Q`%8YOr(ob?(LV(Y-)ubFI8rWFoKRllry;m}i~`4)$(Il2
z45~U0<%n1jItnv+W19|g2aPfop`*q)Wks7wHzRB|OC*(g&a@dl_}q=xd*DQSOq^$J
zf_rE!ePIGmKt*Bctq9*>wFNGFsNTah;T8;oC4={XpY#>5w5&tx>EK(aDTAM0M2WOq
z?<4J#NRe{P|J5rR=>(+%>w`e>6oN~IN+uw(r1|wsP>rs8?b_~H)Vi#KKn-h~8YIz4
zc7Wwil#{?E`2R?-ps=%~V`STSeikcrl(}W+RrGmP{HhX@03_7mAE{j=d22FY?u5Lc
zZZHHlky%qbBa7j|hUljC1s&q{IMLe4u-_gjB{E_&Ebb+_jIt#cGXvun*gF?_L?Hs8
z=^R^Pgi;314=4X>@IN)w&nWo|C7nErxkLS?_Uch?CfR`sMo(xi_>??XX2gxaP2x1|
zQYTnLM7M?5yenFsk&83EbldpZPDYNdwgtt)ZW%I<CAY<fnPFSX^3wtjviS;Z(HulW
zI@r&rvM>*X1i~!4K7dMdkXA$MBb5|a33pX9r6UoyQv`$}Sl;?J0x=|rAO5bI$?Vup
zDv{J~8^Iwg?2n{&gx|{}dFErPH7O(U4Dtya(G0&g%JY}maHIHf3_-{gECk;+h2yCY
z^FfpppZtnjh_`I;Z8xbmV)wV~1I8sk!5-XV8wH27*KSdAtCGV?j*zSkJm)^XRrFCp
z@lA;1Hu=G00!d9_L~xzC557-{i5vSIiEcFcUs4(lJzfSxW8#9K`~|*0B(1f`Unj2!
zl-o^@eD{*40GOTd;cF_NQetStV>}4RZ&ATk@9>sxbiPQ~Q!6jQhfJ~e`F^DXw~c2s
zU(&hYYHvF|dnOWMyA2-GtKOywF<QY85&D2l`yC(?`yblxLVP=uM3b<82#~*)Oo=xa
zl$P*$h<JlZ<Q~cRZiLTULE0$Jo0yo!k7i3do(CW&Wi<7Hsx&tubpe3ugOGWWO6`sQ
z4!stu08pg>Unf7O@FnbPh%=9mv{`wQJD-F!ryU-aiU!81Kz3$942RL{^sNtLTRB4X
z2lf0?E&W;?r*AKK@)L?cAp@Sa2o#kOyN}Q%aK3l|-bQ?iN=kjaW^wgD8Al7F{$`P9
zL)v3$DLqWBmeZ$yQ27s9u*3{xH8K4w>Gsm~ZMA|v{8Q@%+<!+s2Qxp&>C4r(VfDGT
zKFT}v)rXMTiF{6R)z67Zu8Vjc2>v1I;KC45`V-FoiR<C|c32kPP!hv+k<!~<i`U0$
ziH*_MV0Xm)Fy7e0c1Umg(BQ%t=!`McM`1RkNaoX%ujfHh!|OXD4c32ih_3p7I=LV^
zrEGmW8oZr+Ob_!DJh2GiSRb|^Tf>ZWAaaraBl3fOzd6YNX?+CWtsSO!N(R?~RHJ6H
zoJPSeQySV`VZU8?jhvl$KBDJ%YO&KZ{5F5sDrTu6EdA9pBw*OsOZ#LLSc~yeY!FG!
zs!W3Njb*A76!FHmP%y6+Qg^j9E^-IfmmQDMBegf}F;1}E=&B(y5`MsdR!5gV#)ccp
zFv@0x4vI;JsPPtJYmas_(TZ7v??shQ+(lRoAks$PH@(m)a{&aeC7eudnjY~)pIQk%
zz(bI!+oO%rY&~E8?)&t7Q8wwA7cbt>M^M)8`x}p=V|w&f;`MEM<o95~{<B(z#xz9%
z|9_X8hDsUePb+WI^gHRmAy}KOf5?pkjoBKD{q;xL4E%q6^nI;~Ek04kuFM${|E!*!
zVphWRxtFK02r$m+EMiR29c{ScxQ2E@NeiGF$s_bIzSDh?&15vwYG`ahntq*WKuc<B
z#&4331R34^l5#(%++QiTqQr#QCaeCiE<dWoq}IlDv_#he#cr4~zHTrWw&Erhf0$Bd
zxjHRtYvH}HfLTUK#;htLLb<}n8o=M`{j`HZLVuqPGbTwZoprDIZjHzSI^;Ek9#zZg
zYXeizSr(`WnRqoA&;#@E1vt>_^U>7d;ZlgYHNMwoQ-j>L<+kr(=}qRHOvyh^U+uwb
z#!yQV>jY^wg?_R(jwEEi?Ri5t6h%}Yg;AvOLIQoJzVT!7s!`BV5@aMhu>9AuDd|bE
z0HmGMff8NomT&41%!p-d{D>mcVj4yEdFG+s8{sQy^%2M-AadBLAJZWVvlU^WKEpo)
z;2)VH4IsYflWA7t7FzNNI$*Bw$8iL;<YlQ3ss2C>%l%po{Lu)uYoapnM_VaGz0=*i
z1a(Fq)u#<c=M{F>8IbW=x@I00<t4j61a_H6gp(kMVn5hrxZX#;0JD%>ml&rv_%6~u
zuuT@+a>>Dc0+x(`ThQqcY9$VZC=pe>z75=Bp1k^etLbe=U@GGJDC`25f-og1-r9}w
z#}Nq)fn`R(9;09fC{4VL`H-AMSYYx2M$#K$kDn#ahmZ|>{DRpj3`AB7zb1TP*rO--
z5*UkChC51N-QN;JV-({6)M0NM4hfRlzy<_9%Q9NHIs*s6gW8-2xQnK;1qxaxe!Dvg
z-=S7U0k%%OL4mE#xt+?t%@L2t!qySbY50$7s1Iu(4=HD-(}>(hbZMviVdWlE@-58)
zcOKy2aozp*y8ABWba;a&lpIo-B?goWJn}J}%n8bqX=h<+pKA2vsme8HYOl_jeMhUv
zwLDxUGj?Z?Gy0aku+=zy#QuFn?9x{DnH@2ZE&9cb7@)?FFtD!mTh&t7hDVly#8R9&
zM^L}B$@P3h-rZI6An$O23HbiAk#`Z^S0tNH#v$)?U)~ufj4Eg&{<_FJ4k^?E(qPCI
z?7cap*jA!T$+5-Kom2y}*d$JZQ(V0rSrMGpr2j5W+@#;3%PzzyGUD4<gWxW1uHDfM
zw;W;pwp2rKx4IRj5EvM4t*=OgkE-wc)b}Qt<IVfNDJ2lxuV=O1=pDN3dLyCwZ`V^F
z_~xG4#Oz?|u0ycJ{zq^5Ax)?hX5AW0?Pm=AjNh_@WpvXw?$a@UcWVZ1@|xJp{A@1r
zF4f&UE-TN^HcTZyZ&}-jVOksAdgVGaz>ZRjEEvwwBLM9Ha(4Es*^Fz4i=pvmWfPu8
zZX^Z=gPK`1!M=y3s43PV-W8LPJ!*gu`4TKU0s^r&pG$02hh1z?<;+=JgSuu$Z;W14
zFW5G-3P6P7s&gt4i7M)BUMUJWw4!E1fnmNC@c9C~YTRT_YcfSmbio|Hc4XD0xOz<!
zdw}N(WaxW6Aazha2w!eonf%w~AaD7^W))Sdn+_upX=U|mSp4_MFfGEpC+mb2Vyd0O
zy&v8Z=d7p%5K`S?_rweC1Bc-t0|rc1j^LS;FyE!y>N4&}aX~rC`DP7qYFV>6e2g5p
zn1!HLQ&np{;l0!&7jj(5QNPsml-Zg--cFeX9~;CG1tV=)F={ClGOY15l@gTmSw+ZJ
zo$mW`LzjYg!g=TYp{6`Rf<$cDUta4wrH}WxfN^Akd2SJN%7|EjS{E#6RXMY(_yx6Z
ziNrsx%MWlj+$)o+4%f^MBRm&F_)F?%N(oCV69d~nMtkNhB8Hh12G`ZAaS4W`v;*6)
z(ga}R`?WQ0r92CzCtEgIyU7bcsYG$YOat=yNQ~Y9+Z^vcecxcTvl>B+2fcZkhs+F+
zG=7ee(Lpst^hH=8Vvv~JqJje;l%0Lno-e|AE;yYDcg=~2BdSB_y`FwCrLVlLZ(AQ-
z-_8!&wm|R&xNp`S)qd;R1+3eK{3A)gwo(4F(%L<d_3Ib#=S2R-XNL8$1ahb<N7#Sc
z>r#<r9*9-WQd;n{SiCN?oAeyM9cX{~9^R~^hTM-zxa0TWzEFUQYxN+TS|kGW@Pli+
zx9-F5d%GTp3p1ExXTA(uqkkWaRdSbNwf_2cy!<}j7%AW7)3xkd2azo(^j$yiT0lB5
zTpJ=R-&h?>*eERD4zrl$Yj{wx_nH&}N)xIPhv(|QLMDNtQJ*hHqVngp(b{%C*Sk{|
zuY8&~;Sce4$SsgX5H;p%d}9}?)t%JYE!o7nzT3E`zJoq?`NP!QjVh6_=NtR#yRXs%
z@`+!jZm~W_-LcyK`j|h$n(ZOC$KPhT1NA-iF}I%nDALkB^*y}HcHTu!1)oA5GIk~Y
z8IU?pItY#IA&iHl@OQkP0&MSH-`5e$FL#Phqf}3Kg!9vA_VRqb&_&NX={Zl&Pch0v
z^?j7@ukWXP7v)(!u_hkl*4p7K`p#qH>WI5yG<=-+Gut+eChqTHm)>kVYn#W@n9t_8
z_VWB*f1iDw@ZT}Eao)5Ks{a6^(f%#k$CQ7xegKKrh`*n`ci?lnOS~#Zbg+9wC^HWD
zhhEPxri1H;Y()<8tXt~0@UDDm(ZBVFl5cnq@F`Uv8^ENC8tZ27-*k*N_uXFb31rXi
z#Obe2g#L0O5(zJ{e(U<-^&_nEZH)ME{RmE5(Oy{}_Yc>{v0X*#S3m3@;jHBAhv8tz
zQF0_IInpSZO)v*heOisra>m!kdBfXy!<`*(sL=S@-xiRPHn-cG?p;5$>3zKFeI4&(
zyP+|{=IkHUlbM-gglpMXALsMtdhYSR)pJ+BiRYf+xjoGA&-1jCJZ;S0@^<3|kFxT|
z^gAml?kUt#-ui9pw{P)=Ki>JIg+uk*6=jUmbhLhyx4M(3WO&Mds2$%pQ9~BIaq=o8
z5?^HXqLT+C;QntlLs;JB-%SY1j0pcIgylUG#ElOAOkn=cmHdU0FDv0U7R-@`efA9{
zoJZW}A%ZmcAS}qnt>My(MC%XK=6@(@a+PoHpbG?oPpc;(+_2}|>e3V+6fuMOZ5?*v
zjjo*T;hZ`W4t84o4=Yzv@)0EqBw_YB5lDy%ucg{ZwI>-{@E#2?frD}@)Z^N&M$FCP
z`y;P@pRanF-oOU|w}`!F55b4jmx+)@92F(~Zo?^`Z{>(HzTS%uPgQQym19#rUmBPx
z@dpNapI-XI^l<&w#z5U?GR<dTes&&Hu-TwoF3znI)EHmZYnhu4BHZ0cP&AsLVkEp)
zxU=(a`@-OwXk=uyG&4Krs+X=Oj-Q@&${u0%<7)%Wsd=)pLd3Qj3`W;{*i>j*9d9iz
zyVW)JWNIbVk8Z3DHv4-9M<d<hlj+egq=1PkU*FcO*Cwf~vX@L~U_8k?^~j&na`(5(
z%ib<FYOtg&(xR1&;j1*ujG-69cN~%>SSGWXD3Y<*+XO>B<$_6yeC`U1>im3d_$@i6
zk&2=X?clhmTjJuw?0d^E+egI%(N#ge%F=uhJeZqy<ff5{CR#BOhG<16W&I)LTGXdW
zPlxo1qe>(iHPidxi@KEX-ohCNPiZ#)k;;s&{Ym8{vJNEE3q(%_W?}qp<yze1CEdNC
z%D+l({FtLwErO7sL{L>CM#vpJqN&05==9Z8W(Nq(vfBuQD}=snD6cIzGLldgbuDi-
z9CkJ5XJ=Qd3i+)-%E1&PyXk3gV5OHJPZo09zM<`m*ME|6hgsdw*j%UwqpMBZ_;n4_
zXl(;5nKgJenYsRs-uo+RN|70e5G<)V9=Yfnd<lEk+FvNl@jXnp+bv25o^LZ@Ph<ac
z44qM8Z?l$;rV}my6AeU8;dYJ!Nt)$k(NvFAqX7Skd}|xrwzci(RqFl&F#~_Vc>B@_
zr6eAX{tbL>ASJ%hxyr~{CLcd;sTV&AtAM87^AJ#}2H{%~YRRRz73WqCGkN;cmm{oa
z@@VyM?7fl4$=JEvjAEyfCbqurc=5tpdv6Re3b7)JpR2YihfDo>J~b5I`<Ln50fc0y
zjY&}*^)_T9P*9on61|E0T$Xz6(-S>8v0F%lO(vPGv)dt@$uKQCKgREVjSLtV?-~~^
zqb=ZWcbN>#lfh}Y5ll=LGMGn3AFZ!@);8<fO8Y9FnOdz9wav+LAPP(5>D%WEc%(rM
z5LBp<uMs~I?Iv7^al^^z8e<r{)|2~#H49%fa4}XbNeW$)ks`4iF>1yTMl#^q5FBuN
zVOSrkAj*H%&wWk?^7fTVzQy7`AP0{uQ3%!oYq#*BSOpu3%#8uw7IetaFy_530Py~`
z{mp2eF2YcIeWv;%0W2<R!|o$lD?V3tpPF7>DL!SYx~HBkR--_50c<tlb#i=W!m07X
zq(mcz@A0<FHS3`>QS=3cqsoCH`R;OM=H>B`;F6v*&C)qXhaXe!b(KA*+*u{hD=|<j
z%QOcFpHa>v)t6QIWhK7K%1WfR3jivBhsFTFe#VZ!M>W?-Ci8AKO4!c@->QO)+$41M
ziXf#OMN8}`)~Rs{<Wd#O?BXHoQ0W6@7;Z&}P$8uw|1+<U(7{N4ohXMKna%OhK<Gr-
z>I}HtVIi;aH-$M1TsS8*x4k8nsgBjW4a{Ghbv#T4-uGAeV2|9&h*FiV!HV#;LTz8{
z;~_9HU*QT5dYQJ2sK1`8O6__VACiES=DrD3<N>${Ek~TByLq--P`(!K0u`8Pw~<On
z98KY8_|O)_l!b;Z4VS&D$!wX7D7x6i)4t(UX?ZUP`7LJqJY~;lwuMri9DXP+6O7H0
zMu%F*`MkxQ)ID={V#3CC=1j28l0maQ${&TC{&e#-(r3<0@Ne9s(nq;#BHi)S8HzEu
z;Sc2^1C5-bsP8P*qkf|@E+@u&&P?z)-BLRj)0z?$@MXrPl;v#jKE2{Q=yBrg*)wNd
zxZH`e<F~o3{fdV0?IdBQM0C;_ygA>mr@x`{ymB_LAJgUku4GAN->ckbl^E{dP|k!{
z|6IA>Q1Ty?ys6|tB|oDir=k3aat|xHM~PXvjq1i;C9+x!t|*Z)P~a(N#DgTvjwCRe
z!$=FEC<}cZXw8CGb=f*PA5#Z*em>5PL26Y0E=AcGXd~4?WR1Ulf0n)9LqiH;8sM|!
zZ)J0cLAMdnx`$ZRnQUKvAbqFm4%|rJ3A*b^=aGaZh-}@T9n2>4ecmH>Z=dV0@K$<n
zqvyic)AzLHPd94}f`k;!bgXqZ_HBGnX75lu(1>RkY{w6K(tfHZ`O}Ds-7>qFOen-f
z898M+iP)Mivk(P3Wzm%7jFbyrqFIN?R&-pFU;T|f2uYmH_|+Kb#KrSb1Rvtw$%xH@
z{$g>i(`~0Z0qWQ#)S=jQ#4Qa+ss>u6r7ksj7{UQYUQh>kiqi-lBzHaUOa;W_1TV2T
z0Lk~MhHby^;{6Gf*wtT2OpyyNNbs5l8D|Exrr?jFZ-s-pX@Pq!60iaaepFrCj%a~Z
zJ!w)0UHe?_zM=|n%GlAwP%i|yM3sYPG}wujD_2K|lN=p1!Q~=jXM&JsU@O&&L<O*h
zp+3QF<P>Ii<hpT40Xyn(q%?``kvA`@2h}EMSx0SWunN^sy+c4(;WN0OHLbV246T3k
z!SW4kJwu-l{P0%KXz#;X^Sn*1<q-0!lekvrjXv07v|IN?t1Q~?)=1<$vJv(58P`|q
zEn4H>)GklEpW$u!kd=8;Hr`WmRO?-FwfXiWy6F8{UIVf+Dh#yMFthy$m6{IvJGH?)
zS5Q=`c3hx+Xi^F4suZr>_Q|Jo)8O8(tGDmc<>z(ZeH_|Z6dEnRS3NWt^WbZ$F1kPT
zX0~)^oYHoCoJJNUZ%_VrC0$SM%zs6gJBfL9awPwUDkC?L{}btV3v9_`PyWwERFli(
zubZAYnSP7!yfh9$W1~jikEljL1DP%m?~c~AhHRksrtngL;)5Eg_H>qLK)W*zcyQVA
z(}YOU2{s^-6lQ+nbIu?PX!L3x$RubKn@pBd@{RR}3~J?w)sKbVWX6>eAHFx&>ks<_
z{@|;MN1rv=B~|c97%BOlUoini^Sudp^z6jA$5$$rFGa!#x%&4z>&->JLA<Cy0!_kQ
zkWMcps3-2YD*8gfDmX+*tld(wCpXa=reE5s97qnS5!ix!ir*G1p@@hK)z~pDpe52V
z+}0j;)JA2YDZfX1=zB>b)l4|h8nL6sf@YPr@Yo+?cn+l*A``@VRSP4Cg)+wY7y@KQ
z!9%)drwO+@XuAf)e@uN}(~kXdl4~w%s|E7Z_p}<*>7eTo)jh1bMkG(@@(a2&XljWK
z@Uy((rx;-H54gD&`Mz%&)|OLrrURl%ev#!@3s(INl6tDzA1#!0JuDaco+Oe^K5bqZ
zYTiMGD{YY7PnbT)J{!ND!_7d%D3W5N!AXmW(gRwPb`ZCLdJi_BaAUyGxH*O4Dj?^z
zt8t#7h$GSlNt^j@^oeP+`7QvtGyL}Yna|-IfL+lvwQ_`@mP=bd#teN@6r1HIUj~@v
zb9*W?ZYc`^h7Ls&_+=&U*Y3Dix%)`M<k@E)e+HApPUymSL~XAbJ@rAR#O)Z(kdxdP
z#Ckxz7*+iJ>Qr0KX$U1_9Uo|uwEUJD{!dlc7~CiLwAwcnW`wEBD331@GL&zKww<sr
zhbf%tB&!ltXz$!4$i`>2cZ)nXi5E#9{QbZU1!-3V2PqZ@P&w+$H?S;ldkQ?VY3oS4
zXx^r6B<`DE(wm5A?l3F6odTT`QzyR2uCMpDsS`2C*MhDONS%ms(fNviWVN|$Yh`8f
z+)^i^mO_=_#NbVZV&mE=6o(PcOQATzUnvx2q?_jNcK*UZZ(w%jK2U;saZH?#cpwzf
z7=0-cIW<uUV!W@7_KT>TVblE{RFLPC+ZfRIv@s6bMsEdFkWW%NXbQwn!m4W|QWxrP
ze^UiH7?p6xA3&`#2<mM0aKw5TLIrtk>mK~!w{4$AUAPV9;!ickRKpLE>tE#X@Dw<~
zL4Sm|k{az`Eyq0Xz{jznY}{Fs={3qpl#TbHAe4%5EK)Z9j$Li{w@KMZ>vU~LZRZAc
zOx-w&x^X*i!Zqqe6pN@E_o8mxMV)<hl#i$vV+%zo8|%CMF>0cE+}O{kP$FJ+>c-!s
z?m<&c?x-EA@9=j32KJKM>+iJOE%m+i9d5P#UGT4axkjnS+eqE`B8KUF1|w28-ioCZ
zo};F2+-vH_ee1<d>X&%+(N6UXqlwgt^t_LrrB-~AQ68=rd4~gaDea1s_xSrw-S?N=
zJHnF+kbehl-yV7`#_YZL$vBo-2W+NZuc4lLO_tg@sUm|{Yqz;;97wjQM)f-kuN?$*
z9eOSP0BX<+i?lvyt^Z!L_0EOcO_lkEYP%j<^_#WaGkz_y0z2g2@@f{vaPx^;Tlj|7
zb8KX-Yi99QX7Qk5+_qW7^ze{>*i?15tlw(ub_+9d7`5S`e}wrMzo}C4PdnyVO39-<
zMT)^OdkShU_J-7D$MtI}v)fH&c4U2g)0%nJ-5veK7Nn>=>K|iA+*ZGh(Hv)e6yWKE
zfpmlBKOmS3NOj7t;8`w-J5*W>)<%^IIF=r6gP+yw@ncO2g=Vl{G3iktG60vtf*7xu
zg3SpF_yjeov%qZHy6~@T+I~(zR?V|btF}#Gv?bGQGjMCTy$VwV&nY>p<g;|;<S@_c
z@&ii#kmm$1DEBQ&jPm}*8n)nrsywM=S{;Z}a>9-#3EeRdR5VIWA~pL0K*JI~+K=q=
z$P&Ie4WQMZ)_7&!a?Nw5E{5n@G>WX`{!BEAS;^gX6GXH13=u!oJ|!BI^b#LYBHAkO
zmCPwQr=+Q53g)>SKk-edcf#btGJbUTsQI+s;0q*tlx}rt<>j_GEGFvOLiN*9_&8aJ
zp~MU7=YO@L$0RTRSSvTB?l+Wls&~e3Z=*7{(B}k;dP!5|RFqp*vZ6%FnBamENl;sw
zmO%U2eRl9i_2e!k%MaRJC7)9wol8(t&Um!1XoRb}yr{%>-X-NOEAgpJWWbqHZD}QV
zStV;qzNkjus+@EtL0!3ZCAN?LH|31rl?dvZ4&0B^pby_IX*#_4=nVw`l=dS_*gvsu
z?K+a9qIT?GOE<C?!`{ZXSiiAuL|TwEx{ZANR->J0%kz$qiYMsSi*#!{i2V|M9O7*N
z!!>wF1^;goCH{zBTXfb$n;<d2C71bGRW*Y2W12ssSxsDbQ>ODXw78;~9N_`k7*t?3
zO@QI_HYXk%c`Hzj8E$s`(87MIoh)ecQ*lWjS`2}hV@Jh?_UiL+S7Lq^w=}V!VnB)5
zVxL4S@PV^}KjioO1IB{pjRoxiz7K~{sG&3Si0M}=<t7ih_UU}F@$=|Zb-r4J3nvKl
z>})xJge}gx22(`utgg<SvluK*{dQ4R%PX~Wr%gTF>gnvV;z%o6dZPm=R79WC4MAk0
zUdtAPPf@EUh@)c7uX$_7jB7itF*sGHyEekPG`+YybG~}~B>$X1J*qAhXLpnK7AH?e
z@^|5;iRSmQV6Uqktea@=B)`UAze1*-*+9u<pG(73gglC_EBY{6PBaWcur!=^CgQvu
zmkp~{!T{7{jRfgxd?wAz@-W>*eTEh86%4rEgXY1^JMB7n5LzKGw(F|b_9+&=sZ?hi
z!J{N=<HW^r5pWt)Skz~+ic`fo{KfdV+`011`4wPsjZchA4(Y&glg$r%8|`fBgCA!$
z!kprE6FPgT%6GZM3hB#XLLsL2#cR5~Oi>bhHJ<Y%4O3rr6a0kw8nV8c4^gLqBDV*B
zU&&n>(cMa#ue6n1a~LUyq0B-LKK#fv8~=6!cMvY!XmD6(y0>#vXmDmmtJwI_fJKTt
z#&1=yn;7i~<cwD$0-T?~us7*cm+%wAu;(($!sND5swi5}svkmh=ka$?cGej4V%8g+
z)c2P#yG>t3>;sjC+VP#S3;T$sgz}Tsz6j$?j(#uN=l5XEuZr&733D#SUdEcb`yr_=
z>p4`as-Nd?<_)LHRIC|&Eh>BA5qj0m@!2{2MZ&|=;BtsKAoDZ06I${&XGUg<$^r*>
zVOKp-eT6#(-b=Alyo$!w+qz~X#2(=Jg*v<(ALQ$+_gqn~k6giW{cjK(3bm@;@dt1L
zzefE*MaX%>zB1!#+MR#Q9bySI?nvHbQug5qqYHw+&ng7}hmxNnSyM;~0t5&`a2$(=
z*LEGYX!P1iD#R%tKr=r_#L>kfpL;nnb8f0aIN<6yK!$KjGKWwkC_YU!?$7)7EetlG
zUtH3Mx7CB`E6XTR1!~;Wj-LsBP9yjQCBLZTmz4an681-It?%i1d@yPJUO>eC#kIlb
zE9J{8#1SBll`Fy?+Hl0LsLiyJUsdvJN<{nyzplj2ijGF`&vf~d<ib3<K?0*L&J&?n
z;vq*OO_4)Pa*GOnQ^})Bv}J>TrQ}zY7&-W_$?-|P@}($7<<IEOZ!7svO8&DFL-m#z
zs@G|~O~>kCGBQDMU#$@m))#8MS4I;07K)4{kkj7Cd)_Mpm^|P|_ln{G=WiS`(*={_
z-e`Q&X7llzuMqgX^`SnY-?Kpz^L7|aYwiSX-^*`RRyj8}m|j;XTt&lZmvE#SrqUPz
z6Ke<=0kP}I$rzrT=dJI+jXG5yg%8%9h8cC^+DSb|F|c$}j$xTb^Mr0Q>MSyh4^l2O
zYNP@@O-9GLNQOb=D9-L0CeZ2^VZZ&wc)HDuI=bWcBE#r|pKtXb8AieHr-vWhx(9zC
z>XDweXrJZOVK|JzbTlRmsCE4XFv)<-ks0Zrz15&8bp8>~Lx!;$tA2?Z!$=McZcr*2
z($L4fSqm*mt07v+%=$!a1j)xvjBy*1k6ksn_~E7v47(3mg`^eRk$j-&ss4dojrzlp
z<YTBdR@<?$v%cFTA0tRUw(<V7K^hWUz<7FN50a1F)Y)6##oX?pbf4rN7*3B;b1x>@
zMH?xm&1Pu*AE~?Fq$E3Q2kJZh?Z`X!k=us>wQ>jR$WYvhV?0d>*TZHGEy>3cJnt^J
z=^e&RBRN5iu#GSkyFW+JfXL)}kI6ua>-#tH)A8zao%}R2*pPJ$`+J!&$%>X3?=AKH
zyv@P-L7pwyNH2!XdOq{NaVt-f$um!vp|T_yZUo2*2L1hvZv5t>!{GT=|A5IBFh{nT
zIn0yB>*GA>AkR6pS+-CgelBJP<3HUodP>J{Y4QDk#!+=LTQZSaNan640dS^ze;uy6
zjcRZ)WBtDtpAGlhk{@ryb8o@IjvxLeIMu%aH`*d!%no2$)AP6CP`{{I1>SA4LKJg~
zX@envo0$`~v7hGPxkQ`bZ^(%LvIg>}8c5URPX@bCe}4lmGz%AMdUNAP=_AZgGRyaD
zyLr+s%p}I@Nnxf>F%oC0x0moavIr8J`F8B*h<5;%)tTqL9Sf?juOpntTrMvr*BDqa
zp-E#u-CZ$i`LE^ED5m}*nIHP|WWEt;cr%u3iC6G|yQUQ|uA<9!rHS@BZC482Eaz+L
zU{HxPWR8*$x^Ck@-=6vWG1~TO28*;Z3z3YM@H}j&{sL=|8w1<Zk!6P$ANhKq+A~z_
z_-Qht;uKoCP{)Y+1=Dy+8_HM8%T5d0Yh|be+O?oo28pRJmbYm@`x+Y1Cmrk9;y!=V
zm_9KzWnxi7parWH@t7-^cM&eMVH+J8+Z_>+C%ZhhmlchR1634jzHD?E!i-x@BmAoS
z`ce7#+;J$u>Unh=HJySutP%0OcyWHJXq)z!j5A!=&c8YDS_EpDDR@E8BR;Hg#lg>O
z!gbyoKbmZDC&pu*Ceu&=f(sH+DCA@tM+sN?|JQadzikw898Y4`aq7fz(vm(X#T8YP
zAnKBeB2*!I04gfA5mI}qdWr2kN#ba2gK8C(=1wKVL#RLxl@O{p!U2RpKvfQi^5_N0
z{{ctDg+tZw`Tl0tyN>NNQAHxJ*RzkA-TBRY=l53NbSzeONuC|*Z)2hUww{UBDdvtG
zTsD&!Z}yGIE9AabsdSxY_J#<EPuE@yiLfK_wX*AbtifPSZlLm%)QL6|-~v7ct_TB7
zcxF8VrP$r+Up!W`iW7w<9@TI+Sv{rVb6R{)@m(*%ohXAY>nf={7YKEr@CB$l(Ltf$
zRA7Z}?<Dvas7D7Bvy=N-G6oDm35j)ylne`L5<pzw>cNE!Yn}mu0+9+`3RUIkBB~nK
z5Q;!+Bci8)tfWvNZ36o^c^%@qN08q`^qh;@mHtABTHOAz(x-3puE6OHI<@HxxzY5?
z6sD=k+Ov(k+eV{VIO|yg$tnr&MdTr67kdf{mh>g0`*Z7*tAE{JTU&ENY?E4>y~r2z
z?f8hZ{WL~D)Ug`#VFZv@j~nQiZJ{wPp|;h+7vyw`5sIh1eRN_yWwxg)r9~8)E-`~y
zz?n(?kDRc9PS$zRn8dRU(0E_cY79+Y^R8+qo-fS3;eEpHdU=2z=5x`enw%!4pY*Qj
zm?m+tCxQpJpD_kSsdbh9DUcAf&~HNr+a#>laBK|87x+;jSP*K99uX}9o*oO;7b6oP
zLIa3J1Oo#>vNz(OVIpez2U*_su-J>SunN-RY3zr(al-Rj2Ysbdo6|e>2u?^gc*7xk
zU462ujp1iQetTQ*W|;C3o*xif!?7IK*2G#UCvABL5&Y`$zE)$bou9FHvqd6m@MOqj
zM6+3msGjkoplv=VVVaoeLn^vwq}$W+&M<VE2a3W_F=f!ASPUTqIaoqB-2y?9j5b5i
zKs_*xH+ki$CB|O}LlX{$0`yQ=>7~+gsZd><@+F*E($F)VE3TW9Q+C&rOL!`kQRsJ`
z+%=^9$vq}7UtID1i5<}r6C=VwdA&hR>@kT!Ik8C-$~ns}KmD@3ed&?vDHFFk>p*hg
z=~*PDBlBH2<0}ViSPtaLgwQ5|keXJxS%i>O!CiG0AEJ{>CQy6A_l<7zzB+-x5LZjg
zCaNxLc45MONDy$4V9e#X!9s5Y7S0kd(?FpC0WY_K0Obn81_se8(>Mv$QZiG(!(aed
zAWx1z{UcmA%Re0_+YW)C`)<e<Z2Q|WN`Q)J3Bm;;EE1yz05OB`L46ndcafsKIDj#@
zU}M;ajU!yYL$*3F|54Z&1U68+d)EMpOcQKm3^p>8wPRa>jX)AV>!XGt#JhAwy*EE!
zsPZ|9jp-Q*yhAD-Am96}5_v1fV%cH=vw%6IX}{Oc<Y?CFv-fWPefLcM)e}zJ8Dh@P
z2Vl;Ws>*9)m}4Mwn_=%3whfT+zEzoqN<BVBl0t2<ZVxeLD8dq%{HtO2vk8pxzSG?e
zWcvKULlwa;^>hNAKQ$hX)2a}3B#L&*FQ*biItKiRb4c9V->Z-l`<g*0$nD;447CTM
zUMkcew6)<1JHgs81R+S7$-^@Kv6@mU5G2)P@+LPczy#4UOG32q1)N=?B@iu9|B*4{
z_F_bfd+}}&EZz%1m!w~UuB0Cv=^%t~&ih$!?|s@~onPN(uQ0H7HVCsG{i)Hdzi96c
z8%}BmVZZD6uiE^k&8lwQjA;Mr#iuDB#+u5UM0l+@Fa*K5+P%yOqIG+>bYp>;_k%We
zciUKHhMY<kc_*kx=#-8p@Ci#~K_(sRj7-u<N0F&7))y-}X2@rdd~;FsbV^EGOdE0q
z=AubgYGtWnzoAs0>FKuf?g#7pn>6|@tY__lp^r3n4PHlRh1uF+f(DXB4=l)ziNTR4
zQwfnd91|y$OH0$)@(KWk5Z%l-vQyPrB`YvYb1C%9cWJU9{v`MlJ;wPMJxX=&(<8Bt
zSr~O+N<{e-=^p^Jn61XgD;PPN&cKq8mmk;k4egRM-Yim0GuF@z_^?;oe+mgQZJe)3
z_C93~W>&L~1q@bD0xTuiD$iACW)`Q3FNdh#W$^nzeQxr1Tm70Y_Z`XtdNd@wtCQZ-
zMu#@$j1s&9e}(0O&}r3a9(?y|qZ@Zu11L9P!|Exsc1OF^AsXTAQlEK1m0Pt>ed&M1
zX+J}qSh!cfMC?sbtEGvDnpS=<_QdjGMeUuf$p5deg27<<l;OP2ks?g$u|A7HrzkdJ
z^9qXO>RZ8*b^)7JlvzxI?rv%BRsokWI>Z9M0>N2o22@;WRtqV_;v{N-nVYB(Bh^qj
zDO9E<D7w6>0aHipJv;{A97I*Np3S@7naz7{D|lTsb8yJ~@JuBN*)pZA9n5D7*|C+0
z?A$nCM@gFLg7K(`pM~-49A!QAo|OqcM_si34@E)74(`}AG;6Kq*>Etkm}~kO!nF-1
z8zKhB#sNG$-{&k{8o=BUf_H1=wLq}cQE+$~fpQu|ptqI;cI9rMbavwYu*u~t;{*L9
zJHD+*%Y>S155t=B8q?lW{Ah(?t>B})V1)ZaWltaC|IOC8Kh|Bw_=P;*_v)%uon*)~
zsy!V&u1D4aU>J8`KyUm3qdfwAaW{SAZYS(R|8MS}Z>bGev<VqGn-jOh&Hce0d(<>%
z--3+}PMYbpWXh~u4uG0E!@Ep9<0o8Ax_>=MIMBjeZh6@`GJdL)<Ud~@TDfS4P=b=^
zTXC`><Hhs4-TCH)o5R{La6&bcDy3<hUnWJY6J)4=TxLE*d5iOTPq^jj-Qg%p42Vjp
zv{?3NbeDHa8*wJ*hpd9v=cP2xRV(wHA^)fPXiIzSC|ERI#;!!EV)LYwESg&(dBK*Z
zXQ#wkc`03;Tr4ix`Ch+jGT_k)#)23)>o)O5p7?a%1x#0<imonle{5Ggz1aF_w6syH
z^xI(@^vmV`8PO#p{w3(cb5qqy$xnskVi+izLH$7UO?|7Hpl-bbZ;#rwSDQz5#h%DL
zQO!FuoSel9*r($zGer)WD`4Y2rtAI@g^g+Nh|U;19dzp%dq(eR9g<_9`Qn&0lu4pZ
z4q!_zM?p3k!i~m7J^8$AtVBUBBSgF_ne1nr3yGyD(7sa?^mmUj7pP3mM=Z|GnK*GD
e!GB_Byz8?5Gt=(dZHFf_Pg0iRGnhHL^VokOipDSi

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_0_8.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_0_8.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..366862299d4dde634bc49396d087e91a270c65a8
GIT binary patch
literal 55832
zcmdSC34A2iec#!4HyT&t7z_q;_1uHuTpUs)MNvFQLs23@iX4g(B&r3V8W>=J2Bx|(
z#Iy^Ftf6E|5hbqWWc{%zFUB9QH$LJxjy9`RE<5(-@F#X0t>a`X*{mWvmZ>C;;wz3n
zUVFd4|EsQU93#oJo!!L@9$vkA_3G6-{`dd?9_{TdCiwaC)XUW)?@K0rRTr(FESZP-
zt^af~k?<0e2`}lT7L$`n`<q%yPo|TuEHjz4vfN~zzv;!oQrBdc-ODT%m%1mrmwG09
z>^i&HyVN(?x70t`Z`ZlS($c`>K-A9Q<RH)I7l)RHCWox7usFQ5YjT(6yB0^5c2DlM
zd~tDfY0u;y%Xcr1E$yA$yR>g|-_ric{YwWX4=f#=Jh)VzEH53JJhXIp^03wGSv<0I
zbn>X>dl!!_9iKeDbYk*^UH2`XTpFJoUph5;%C7qt?^rrLd3x#0<e8<jlV`20w0Lgm
z{N#Dd4=mofbYb!W`9W{U8}@d+o}RqRyTjY<jlP_myxZI3jlG<he9PkrZ?CuS<%GAd
znqRo5df{ef@?OgJdj}{xP`&qNdh$N54|-*;%c?`kTPZo@9j4@P^**le=lY0ul<T8*
z{Wh+TdB?dvUVR()AK?0gcarOqcKvp)$GuZrpR(%*z0`Aw@zYZ)Ok?q-`9-h1G{4qZ
z^{eIO+2WN(W2JuoxpQ;#jVr6uXJ?j|&MmIa%zOUSe64=YtJaq1>+`j_Gv55{>}q{}
zxprpdMq}BZxpHoLae4aOz4xBKXWE;cxo7q*7w(;X%YEKmcg@~+_w>EfZ@F-xde5EJ
zd#ZQcdC%R`cePup#4S}?EyZuye(&tcjh(u<yi%=A&7YfJTJe{!SG{)cT(7Ecz5DL_
z-g@D_yQZh_yZ4@{nLF>k_g?QUo=0!rIz2t@UGU!e*7Fx;s&NDD7oVG(TCLaTr)uq@
z+w|eF<)xMP&ezX1rj|574S%XupI!Ess(!tA=1iHtGxNnaw3#NxlVRb>sikRe>d9xv
zQ(>21o${$ZR}FK1b)`Dhps;Udxz?zzHI}AoQ*%{6?5?dYE=FZMpI@2s8}n0(WHXIx
z-V%SimZq*&D|7z5$3-9QU9B{${&Hoxws?b*;*<0B#$(I1+4;HgLfE&sJXdL~H7Zl}
z%KXw?n5(XFyB_Af>g?3&Vk6AX*O#VN!aUQxL<eZS${@@xPBp@wndQ}5qe2g+mg?bd
z7Di*LMrRt;%B;V<RH2u8NixhmIp284(v&{2#z;&pR<16F`!3E;%`Mla79WpB_98F#
z=ciX2tR|`sKEeDq<{LLI&HMGna}B+`_mS1ca<w+I>{b1DKRKQb2kl#@tBt8jedbEl
zTV1S%LyVR`HPfif*BZ-}MzvO7_QRqUY_+mDf3-@7x-VXQ;;HARmR1(4K9{8@^l48n
z&oRJH&9BYZ!orMtKf_0Mlk=DB^{Gaq7A9}-&Rle-t6iR+@|G^oq%^7g=r4OJ@fg4L
z3`vk+qa~ZuPhYS4FZuJ0YI%t-SUgKN=WFGc*ka|zmHB#kcC|L6uP-;2%lfc#UAxh0
zovU7-TC~TCWq!QnYQ0=rZj@)PXuXuDYB!da{p#5tNicGmdUbI&%;+8dQC@G!AAI24
zmF1=CIUcG`FE3v`=S?-H&dvGDt1IW8Tdh6GT4exh=QOtGF0!j1ed5`3i{8>X>ndBm
zwb9Rc-pW$tTHQvka=vmO`ylM|=Ic|_i&d{;y*;Kcx8X?pC&|xf6!fFN!XGv?`^W@|
zAUT?7Bo>lE%1Z|6`Q&Gk!<6mTkfkQVoL9X*KU3wavIbtl3=7&D&x9%ePJS<h+4_~K
z6=spP+8^tm;4&Oot<7uavGOafnL8+U{Yv`h$j|T_{~U;N_3KPtH32?LOwuPW<)yvM
zT*AwGIWO-Oye_Zkb(8Dydc8g`<Ms2e#NPpav;6Dl*K)H5zAG`C6ef7NYclP<OZWgR
zkf~<9cY7n=?w6A{lask>exq<JQSG|P8)j49sJF+;i{6;G7hKWp?eq5Yx5qo+9prDX
zSN0C^x6eE59pP_(wd5UrJ!z0XGdbW|^PV)wul5I{lBW#zt7IrDnea|~XY5_W-qYS$
z?;L&G<(>ELq}37c8KB`^l<fB2<K6AOh1{rjk9RNk_IUSsZ{_cpcfa>G{_gc2@ZQc}
zaGAcud(eC6<pl59?>+22LiqvjQSUMS9*pik>s|65xA&I4cY057=g@btFeVQ(<Ijbe
zix;1H>W$w>Gk>|o`5F*0Of6oRxYR7L0nRP56E6dkv|r!QCVPX5i8nlwubp{p-@pIr
zJ!21j?K_5l;_-jm{GNxt_J=?6pTGRn^hY0>Yt3<#`Nl(Gy1G)IDM#gp_*>vtZPmX{
ztBu6X#76QKF9>#f$*Z~gSR>`7ZYDpL40Z>JO1hB=5)0Xd-1jHEG`Tc?GoR1gOoI1W
z(UpAAM_LGaNxOm~X))*~?GAcKdxAZry^X#g8I0Uak?(JmXhrKL_iBoFc2PD!$$<Zb
z#-Q%pOj0t`$awk2aAOxGLmMNv7~`8s-dy;6T5k?Uy~I^LNe>rx`#&Fyk{=C*N%sUp
zq+<b3R`v!3)$_j|6u8<K>?7SD9C$71XBr2B?9JqB$$F-d;!CrtNBKqCI~ZhxT#)y=
zUWK-^+A1BO?s?UvXGQg)v$w%Pp1s1IZtI<^{UK{v{V&_-CKe6_gF!k-1^q!O81Q;t
z&Cx!iTR9wzksfIrT{sr(<z2_wNxiS8*30CN(DD}>Cl*eUA7|A1Uezd$*BIT6Q(k{i
zzNJyz|5|e87aMon)Qp^_=4qb#^~RZnvm57bsqDNb7*m=UPy26U;=|P1`7n9Cd46Wq
z_rW}iH_F18z~GgYWpHe5u6&q9a~N3bl@G6-k22Q}H+%0dKf7A1sRrP@Svs@gSG{=$
z$pt+xgJSZ4{^{j<wOM+1wYs9J?|ABwI+Rv9Or8b0b<fX%8mpD7)f@G&s25iZ@HQ*Y
zR_oP<y}r6MCCsoQq+mD8FIDH~t~Bar%a?^d_`h1O&aN(&UjmYs{c0VMsaFXb)9o@1
zoUv~L($8F7Sq4F#ZC)(j(f%pF@W`_hPfR?{7ro=TcU-<qwmk9tQx!T}e*T$@k6eDo
zbLGoVmnYxx?9-bb8_$K^bMx1$wTf_YJ?!0#lDiu7OZ0}}3wzwAu0SjJjj$M9RcjtN
zcNvnhTB$P}#o77A#R@xqwF)|$ZwhJm`qgWz^L!Pw&6EdpSf0B+<&WP93RndLgoO$i
zf4)(vgnf_H>s6*~x%Lj<2l<zbM5#8a{(N&v#bI`ZfoX(YmCDTGRK3nKJYAbws*)>#
z-#ztjuIlC>?44PhUjc;}4dI2|qM52p=#5Gv?4K2R5>;pR!lD7EiaO0Ffu3vC#d<i(
z%v4tRK)oIU$cPx`rXH4rY>k9|!Jk^uhh%MX{Xs3tyq;s;{9#>n*&SVn>BVX-%q&7P
zh56~#*;&ZTuve8^bP%5%XCqQ9?7pr(XdD=oVHU^Syme-Adb#4(QrN#p*F_k$3i3<K
zj4JiJjn-m-XXhJ!ue#G`<Lzj@t1FFgfXS&gDz(+6%IXUAMYXP?Ir>=j<9>x*0^*g%
zjg=~k0r+04xS8}%sm%=QcQNc$8|_7ts@1}-XGCXe68(3pVt?GS`}{DYJs0+@&Op$L
zuwrz>RBZ)XnjO2m6m~yrf1xBB%#5n7__ew5L%yh$uzR&e*B!~~U!=r;OvzQ<>teHr
z&TU-rAJ#QQ^OR@wbVX13v#R+)CGS=8f|B<sd0xq7CGSvjNy+0%9#tZ&<<BUolDKte
z>w}&P5cp5)-n5eUD_K-6Ek2gLEp=b`%&)5$-a$jJm9eWm3tYBl!Uk<T>Ay#h9942c
z$x|eE@>|zy5{HV(QZknu<G1r)PqNRhdy)rIMdea<H&t-85`9VcTu-u)?4e9)k(xaz
z)g0t{IMtKt<8QN9=>2SKr_SW{R(|^UmBc|~UjN%EVc-Dl=7S^G^R>YS)FD_RBP`@4
zU(K%Py;P7NO{^Chsf9Ec4GIn1(A7w7WVqAiK}&r+wI^{awO$PJ3)v4Q*VBz$0G<Ty
z=MBFUXiM0+Ya#2U_ka%;@>f&!k2Z>4PI$i2-RN29<+snv(=yn5qu(ol#q&3lhFx^s
z#dTN5wXi&RFuva%bkn0!&~=M%iA#H2srosZ*mS4Y?hNd3r?36Y-p$YSci!<z?G`=J
z6oaFQAU~X-#o}nf{_@Vf`r1L2?@p`>XZCvo^t`8|kAto!^nNJl@rG`tHuZB@<6vJ_
zyal!H`nwl;B>u)-LAQIS)hg^%Yhl<Mxte+{<wKz??DBRGB|K3sBT)a{LE)y_)fenm
zzxO8Aduu*d!``SpbJ5$QmK!}AqZ(anf2?(-FQopXR)4S6{|NV>y1w;$ZY6oL&=~Xf
z8MW8zl~^0SUitxzS--cRJao9IAX*e%vhP-MWB;uL-|@vp@ni4|&=h6S6q|Rvb}B#L
zc*uvmgc%GoQ!v_Z{4qbxJEj&_u1wX-SEjDR?Jbw5t7W)vP@z|beJNL`X5bGa1j>_H
z&tPZPVcI_TaAjsyLJuz-+I$U+t6mR>>eqZ&e0xn?bH5cXgK-g5tk$>Q-X|<~+q--0
zVZS;9V%*a1-uA7fJGb0?Lth^b#z+Vh?;}*1sfGvIWqNU|9Q-^FQuOAIJo?=|w_g8;
zZwrUp6-^W|vs|6MbH~yPM%TRoG)edmkVJFvm6#uR*nd04&C=4U@zN2cSM|V4*B7TF
z1X#bsm>0pzb>XpEWBh<2KYx#!ZIFcd*4XESziXa<R;6hO48NvI?UgX0OTj#c>vEBB
zB0md4Y20<ezorK>8b9#+@|+=uFu%l7Sz29k=-^{|`Y$TEsmi&_PQKB_Ow~=i5e`i(
zH=Ym%U!u2F&yhfgacWaF|1~|6o?2WS?=!UL5RqXz!+GtlKCV~jWns?Euf5NIStSL@
zRV3_l9Z{oOcyE7!GWD|lEJ;F`0(?;5Umv)j0KO0g8BLW^B`{7YJp#_rzXR!E?w3*r
zs5!>(fs}Ac8GLe--=5@Qp6%g@=7`~um;`81I-6uLLMTG7^?-z}$|@jDZon)oq&G6R
z0Hm-9DghNNr0t3g40H4V$U>}D^7Al#$wag0CEg1&_=2%_v~%5_%5Q4NJ$XHPGUc9p
zA5Q=f(i6=+_m`hT(jy_2_P7Y@h<b1Ch1xb+UAH32rFC<>;0TKs`2+}EM!o8PK>1=!
z(;}>!Szdt_UG|^Rojmwo^~O_<vf-{t5VuTG*d^+q&ZE_z<+1t*Ny5;z5OrTN1qye_
zyV+-h9aoQ~IlFj5e6!Y1{bX8L8##{Y0l=9VFWpE5nT@nK)vG!GctFXX#5&w%t8_iP
zc3+Uan0WF0dT#AhkRo?@Js;#YGC}@U(tospBxNHP<b-{UFU-9?aqZC;64&l~K@*n;
z>TWa&;+tPiG`cq6v^ToN+imnPxwIw@dA;yjYAx63-G1N8Lpc;)O<zpBmVEK6AB28@
zTIdQgSM&Zqx%Ne<gswe_g+6e9k+i1)_P?1V1y4dLe5271m0(w1*VPQX<vuVal!xdJ
zuwRq|`y}i8R2SaAo9Dh56gNt@z)!CI^}eW$qH6k$wbPA(pcks6dk?*5JjT}hktTrU
zRWgzgHVXP5pvTvqeSs8NVXF^WM&o<d-n-@bi|hSs4@70fsLfOBrOr0tyt_T_jU;B(
zciIn1vnhMW59=M*zWW7cb_1F~wZM=AYv;GLcz_lM+U;cNr&qiefr_>A`}Vvj3HSV9
z{FSX9jctCOUJM2$o=w?I4?@rJ@8c=;Z+&p>DBnD|&==2!u<xgW!Hr?Bhu#hH)D_l8
zZ!oxD<7^}SK;qgLDNDbWs(*o%vx~Ke^zPME4N1+yNc);qFtk3*y6ZOpP~uHPCS@5Q
zH*C4V&qD)(|Lu;~&D^K`0#6JDL*Jip+uqfU?j7aca4_7uX8>W>M3|mgSp~E7AYF$6
zLewn2v<TaY_*|S!vwX3-ST)8UUZDIP>yC|YS2#N(WU3!_o=H}p<8NJz;v9S6#U8#T
zxt<D=|6S@@=D7|CjSM?=_I#KTzZ)hiTj0VU6dw>*iN<H;!GEI$g+FtN<_QA=n>oI-
z52)vh^#{qco@pMwH1FgU%czth7$=ZdtG?8_gu~Bx)%uKtx{awt|FgWp|CADalK)W>
zVcnb*0*pjo3?5CA(abrSd9xRd6-pT-9Foa2OJ%sRwUz1&Sg$NWO0)l|=>EI8y1!U%
z4i%qoSBVP8cSoG3xH`YWyTgp0w#o1tb;;))WX(yMJk@Y)*N0TG3t_(BSYC$1&%*1Y
zIVfpVJ*N~AvkF)4N+KrR7r-<M!51~aV~e2fFbk79hpa<rJj_a8QxCgVYxQfZ)hdF7
ze&aaXb8k3ce3t|8nV2*6A5oL3y669tGRdo9YW`|is6{#&|Htf&hB?A6rl(@UbF?zm
zmpZc@6<jtsA0`*jp;Qns!1ykDah@k3P~0eo#rF7bSycWN-uz{L>mm#i@PZ|f3|yl9
z>jT}0HgH^|=!8^q5KIpWf=U=p_N5MR6;Th#1EL}Lcb-;;qpLCKh!M33pE<_;qx|Y$
zb7+TEJEQ%gpB%p#ehq&;N(QdeG!a2^A$2nyz=OayLIzNphLd71sT>TZdYSk->v+@I
zG^jo`b$~T&N}4i1k5?P$b|C249xdfo`P#{~N~0W<*DB}7&&E~nx3Q*>Y4#iU+hZS?
zD6f?t<M-^@voDq%xhSEbNun03b%cJkl~p8wTU&{qM~pIE_4P!ozbTs>u6)vn{_>sU
zXUp%b`qk6r^Qb0KRU-66?}VtXqe;JtTDW$9`H^yC6}pmVsd>8GqR~&6J1O?@(`D#O
zy)yce6ZNt!vvz|qGc=y{#~611<4Qy|J0kPD$c^VB>LqF=Og2zy@!k+cK*TfjbiQ}X
ziVOR%Rg6Dq)y9}0QP<{6<h6RBA%|^kY^fF#H(D53er<7_B(r^SSc!z!Zn>m~6QiQ`
zk`v<@_g$~>nES3zlB4t(f5J4g@Bfl=S)0eLJs*wa$8I4wU8+^wuTN0F5~;ytbH8=}
zc6I+uzoR`rpWx-ucliIDynTnUQCq)aY=@pTD*)=s^&|0jWcZGQxBHH$_P6~djQ^Gy
zVFQ~Yu{|Si=1b!FKz{9JBkI{5@po*WkF9m%uMqneuPbp!(Yj(CwSZ3a6|It#32ItE
zCn{afS}lPddu~1F#A618_}-jbfv;+gY-fE&Ib$3@O>R7IYtjFwDvc*H?Ay#-x9e}6
z-R^5`N%EZ<1~FS&mLtZ|HA~UnirYi<C_!jE3|(Br!yo}FR68=iAcf3M+BLX{t0`X^
z9<~=-47uK`>GiD9YuFNT@17vD&gc7=g6szJKQHT~Pr&7@o<NtD{$P4Nx1L`wNOu(!
zY-sWmCKZ4=(hl^0MM;Nn;2VFh4V=<jMh9pm*30qX|AHK*0E?)>EB?RYV!X@k?&O+l
zSzp@{*e|alrC+NcE)k9Ag!lqZHooCVJV=s+$D)s|L*Bo#8KB*-VISlz2}7iVAj=H@
zdQu}T(6;4)8I8AodiagU-k9U$0%O0BoK1c&T*!x#>sgIkJoag1t{Te%rCkm1BDImC
zf(eUA`OuGFMeceF;J1(svgqkDVuBaoU%Zs0cR!H&Oad+n{>jS*UBiiu{4Gj@VzfDP
zl9v6Sbg4aVZCa3JBf2n}l&?GlKWwYn^!;DeLwiZWtVvLe^)+(<BhJzggX2AJoo{YG
z>~Z!r6$7hba@zlKHD!%{MePl`IodQGS}I{rZMkOFG)q(Ut6OKNuu^em5trIKVg$OR
zFLkOV`EJ#SX%Jx~?FB<KKSYL6XZ5c$%3jiO%SgDfj9~qzSnt`nB)oF1k!fTb$ik_e
zzlo5-^w4>v;Bd1qPT1Tb;YK<RKRcUbw&2AVieC0r?X7O@4cgKS_l!W_z?Xi)uJRsI
zb@tYy@Zh~cRw5Dc+W~yGWbd<Bj0IUxTwx-TW7?QCi)KqBLL8GN|0nng|F4mdw;}yb
z<=y_b*+lsCG1ak~r^p%k>(z`l4hu*up_{9o|0PwbDfzmRCDqJ8F3&lgifLCqM<I*i
zm?~&xgu{+nk@Zg%t1D0TNOkwBujzhTW=_4E@BU*weKbx^s9tYts@RmvD^F`iM=6&q
zOX87S@-PyrVsan(3?<D$hsL56U=O`fqJBKn+Hvu&JIsY<nt9f2YgPp3GJx~+ER-cX
zPS$xTY{dNEq%K>etycQaD#$u;Q=FbcsATF!tJqO(%LnD$+We5d#ONTzJ2}zFSVN(c
zad%_*EHXpA6|G#qp9FN1HhUI!^~;5uDJWyDKKB1BP{%jZudt`Bgx<T70EBGlH$5>!
z1M!cV9OBn-27-m;D+m=F-IQ{ySWzlF^nzGx*|abShN$d|Y<3C{0C_Q-dvKP|`E4~F
z(Gl083Fu@^T}?);k9u1v&Q6NPUed-3dR^<)?yarXs~SYH4++#;&@fJ|)w;WF5N&gJ
z(2xS5Vlcq$F|LnUjx99g$JJ-8Y`?Ro2G0z6S$DTgb#@M2;PnM{QuAo!V$o*VZp8iD
zv?Lu8UNCUr9~D(QYDO?D&a^*`$S#H0DpPw<%hpQ)Bqk}p*hnvA7P1KGs4vugZQ7MA
zScxg-V3<&;<UNc_5Q8C#7|%^$)N`kQ7hOXP0mlKej|S6QHKU2<!PVN;+VV>^_zE~2
zX*<d%z9ZHeub*JQU)7`Osp-1^&&b)T@t0LrY#Gr_Erx|9nf%d)|2nr?J0>H-G|a(d
zF&Or^fC`ni3}QiJr5iGkHL*&9{occF7uTABLmf#KQqk16>c&Ic!m+l0%CxY3|7hpG
zBCay+rGdx507TR59)>YHpMaoaUwIjW#%Z*m5?-ghEZ5JzlHgi4Eom>u^&(GZXo&@N
zOEm|PIGJ#|WgWz@Czuc*zO2_?O)uoJV8ecGE{zNf2`P6A{ttWowUk#{?*jIt7v<T3
z&!>Wdu=P9|Hsr6Cr!JTqNhx?Fki4+qFgh5%_1qxObw|&r9+FQZVWGpKmWR}Gt5t_r
z#)toeP{RLR&3xw`wF%Tl^?#4#qYS$LU&ypL44ceqeFw-8QF`s2|28!uve0RW2A*lz
z9z+Ht<>sKNjBdc*$7&-)R&F(lVsSF9d&sf2)870NvM$N?!!%Yc%^tI4i!3dgUG55<
zQ!{JX@m>e2eqRl>Y0m$uOD$y|{bRx{-9k%a5ayiCg&7l=qQ?vjS_IH$&?ahBw9gG+
zBEHBZLsCbdc~1D4PqC9k{jgVN=htL)fX#+83<(Q%L2ow=z5|ewWnz3&{|mHG7sL~Q
zEJ2iN*W{w#Qfd&1PG4#i%4iUMPhV=dwU1iQ$BG<*C;{51`K@a>rDHI<2Y|!+hem?F
zK^g;nl-bA%L~#ZE3Ecy3`T=&=XM|kRn<y68AEP_+LNu|W&Y-LjCI68Z`F(st1T>63
zMQ7_{u#>y`lKwhsbqmbMBO&Cj_XGO;42>kxhm*}71HI9$m>$q=K?g$)50Vkr0>y{K
zDM&A%xiAjsfG9sJH8>%>%UvuGv0KQzn%v0o#gy=c*$Lm4Tk7mYb8xzRrTkKPrhMkX
z@-%<AU>Wu<Rk0GjQE_5S#z17U!!7Dam_lu2!!jV1(fr)%GHT&S=J~5sXbCW~MsxT-
zqo)lKZ5f3wK|>Qv{gB2%pfF)XlSF)AOQ%nYD{Y!<jf#zkqmYazWX6GssO<tMU8LHA
zl~{v_8kMjm<>UO}Y;RiAHKGFRnwGL(RI;U$-3X;X1PEEGD?iTt2%Yx<;Un&~qe2YU
zyXl^YC+p&`ky&f7+!7Go3%2kuUPSdF!d*H=cctk=iG^md;NR`RlA@4Fq6A}!N+Dxv
zo|8thgG<H823drS**USWDcD<E%mCUvN{mY=zX}lK33Qe)1-DWRM*)znCEtZnvinzf
zHEWtpz&&>gQE3rnm&_{@3tj$KXrXvZtN>~yuZJEAQ0Kkgptz-WfkEz@2$L&5&nChM
zunEKg?(nopT*ex^fbGFtWg25$ndQk(A&=3uhBhpasje)~T){#R*AQ&qRD8+!2dDr&
z;a5Q90IE8TI-=he=1W|{VqaOt0Nfwd%Yi&|rX2CVS5*&E=1-|sad`y`2~6&gxA&MF
z6(d1R+w0@qjy2GN2)i7uJnyZAz3~GTF5Obr$Jj^u#t*btj?nkk<<r}OyY@MMof>tm
zSn=XcoSlL38#QG%f^L-GeWpzvNkxpprpIHLAV}B6ueM|TpO7$3zy*)>AFX?L4UJ&e
zEV2#klt23}O>1Y(Pqwb3nx9M>?G8Qha^V%0>&@gR6W;~DkM2|@Nh?W8!Tp}Zvx)cd
zG?abDEZ)uPJz7vHtb%pDsqB@+4rR&BWv2D&axmC<z*u>e;wuH@jU$jWc%GNPIJhp-
zh<(Jz!$f?}R9Jwpj2S~}E3@r?ls{n_T@5$Tkbt3t!c?Pj-p!Yg5(+I9n?mcCG*5aX
z3ayg{mBGgIm@1-tvx)2=4&sO2Tvu@{iB?V2Q~wkN+k1LvyQltXYW%Zy7ca53{Jm;s
zpAw0GeQ6h92DKghALc{+BPwvha^Jw-aov(UV{))PCbeZZEIX9B4$w)7GyOA4443zB
z8Jd60oybZ+MjKMrlz`^@Oc^uWk~wX8C>}elX#E-<A0fjEOZZr~Zx|OT`qI`2@I2i%
zd2o8JwA;fab!|Y&7EnIO<k@}#0*M*M2IpU>vl(8Kfa+-OwyEA;E5<yksb2%@r^$eW
zWjm09qh<F++%A~da$sUfy^zN^%$u2Lj?`Ls)}h>*L%Pk;>2j1+`AM$EV|1LU`O7cy
zwFc}oHomrg3(4jkLtYR2YVypgS1>`U@IcrT86umD*VdsJstB!8Es|7AE&H!D7DbvC
z5*F_Qh8wjJ*`e|q!fr;q&IlOkr=MT-Ot`OJqcOQh7=U4L1SmkxdBlQwu$OkM1}W8X
zoi?H+Z&o&#a^ND7l92=$FAwNQJ&<VRZ(_;`a{}RGce>WrP@#jT3^y)-n}nT;k*Ml3
zcGcyfl7soF$y5_zRdEg`#e*<F$c11MGYes0-nY>QiBzOczf1*X%DI3^CT%z|SPyK?
zK!D)@qOO`@*1-USMS2Fl4P@AHmPI<`QTADO9ZwH4`C_94f_vBdI)G*`Pj$;w5Mzlx
zYomv&ENv__c5RF__}-1(GKWF5mU5#dB9Hs;dj-vrfoRMUUPg2$fc#uSrQ`h$Y=$|f
z>of4#%-v}V4IKP6riE3sg~0EP6@Ip83ePMqn^yiZ8(s1eCoH;?*$lf|3yNhE?^gjy
z?Fa+Q+OGzd``TdHf#ml3+YIF#f<CGj378mU%hF~G_#UzKCv<M+@|{ds40p#zeP7SU
z)LocHY3b<IVLI-FdQ#YTND56TsE#|Cz+Y0%sCnZSjiLIS%I;PJ20(vDmkkn}FeEdq
zV?c1H)1Z&{cTf+>3G{UuOkYog-I1J8rUDKs8NGS0+TA=CewWHL)c#wQXjGUtBVi2E
z8mHyP;j2`r-%SOXVo6J00=yQ|IY@{SpjJ9=0p31%rX03$`;bP8uqY=pQbF}^8P#|;
zH6Z#m_$Fx{%Q&%~bP_ygp(XlyA%m5c{vu_@HaOMDE#wyp675JH89_UfNs1&I?+Sa(
zfM53NK4Mlgo^br?5AqH(SU0<K$5mvF<_fcqcvDL+w3LWyD9jnjRu9wEXs;-ThK8L@
zK29%{m}aXNw3HlQp<!Zp7?kE7`~02iBXROCs7tbS_wQ5kRwWt=V<9fXykBAo{}u&g
z5xe5GBR%e0G-IEo^zy{`5&u~-VX-2=5L``h7U`)}uB}cjI@b}irZ>|^Ulfg7UdB-6
z!N2laaU*m^H-A})v`GHHRPui)ksc|`;GVhYe^u9CS0ZVzEjiy-uL%Lt>XrCYe^80-
z5F-gr>hi3Tb4qNLKcw6{NW$L8-({wXcTI=k53BJ#dW&t<^W@I)Yi5&$ObRoMH}|jf
zHUwt(rjk8<*bzQh=qcn11%A8wdYG>Qzm)gnirGT0*k3$oce91jQePoc$aCkIub#~C
z8~^l2x%%JaYuYZFI9I;cU?=XGxLRfeopCG7RwL(1yNAFFu{B72vdeqHN}OY6(cCiS
zXepOWIa<EY-1L*UTAGtSu9oJQKiTgb@+V8q4S#aLx!)6<;Qiil?*uNAE!WCPYyYG-
z{xTSBayW8){D61LyTcqGcX=Ol&X<!T(LK3e0`YOTbPkxK=6)$>%NHZ}%dyD)a&P2*
zxzGEMIbh1svQj<ZJ@9(cd%O4G%h}k`^26qUsV^pkfVpN)nJcDB4n-x?-bL>n)~CZ!
z$&B~7_f9K05|upRz03M`G`jY1UwzVQ9E(b-xQo8WN{)Nade1S6C!#x-z31)Oljh<%
z4mO(&`=10>)EcqhQ?s}vKWOvBn(zhuE?pK<<L4U>&Dbs{JIJpn60L#yE6hJ`0*Q)L
z4hVtFjS?pZN~BOpVWtm-aWz>#)yN?Hl*?nme=x}T_j_rCoEfN%l!;QG5g@ET<B9(0
z-4Jslr;$F(vzo~FTfWQkCCeA*RA-=(y9vjPQ6s9&`DD=h-={t!At_>8TtA^ba4ta<
zRVm&WZFD25!kt7Wiyl&SO8Z|()Y!WE=i!1;>3@HsmV)gn*1uxzpQV~e;{mU?F^CI>
zTpVCWDH%q}uD2vMb|LIl4nc2a1lJTCKh61c)be|RzQ}#^!_*za+;c!JcydhL6ZAuI
zP`<Di<J2cLLXCYMPNo<&?x*yCxyc+1O4NFpJ0;zTYBy~>aR_w|oJO(_Oc+VxyrvE@
zJ^o>ma2PM2Iu266W3T1*R`GA>{yNDe|0BvitmHeCFs0~M>(~<#KLG#zW&iU3U%Vd=
zg}V`x%~WxH1F*&)3|=%jZaBEfoy{t=zI6Hqe5dDZ0n;nXEB=!ji>EXe-86(<s5v2q
zLW3=@MB{bjZg0^cTBQ&oj+J<fkqir-#}mX~z&iX&Wl5nNy3`8~y-MAmpo)1Lw5g;m
z0UX+1jp0aVWQTS)!ohZ4QbNf+O)(qxO;0al<FtIWT9Zg79Fj#kHf>1TVrN0K2y~3z
zGy39TB_`^AQMq}N@j<8J9PD(lj6O;d%GQUwBL5kE)ALF$D|t@I2b9?AvM+v*E~U!#
z+q`7FAlvJ)g$_liIh?5Z7#7VT0wB=~c7o=RiK3)E*jz;aMkv+46p|m+K`{(jCYH;}
zA=JctDc?gK*NWO`j&1>oZ4Ws*Yd^d^k>}SaIQ7D8?E!Tr{Z0w2P$DY<0AK}Up$!a#
zHgaYcV0YQT`CB+Sh}v6sj9W$kV<G?X)B}$9&bW;XGw}p3@ZYKAJC&#tDBVQLm=@kC
z<waJ@YnO?b_pn+;;3hX-Tzcy(Q;6&|A&^wZR4q-7t(2V$-r*)P3f?iHF*d#v!_IQD
zd3fu%m=EfvM`lFV>PMraN5x6ke}aT>OZY!Z3ebg7RF09bT<b3*R#!RN18gAfM{kCY
zAmA|PnK!=}(~W{Lxm|$#3`$T?%tAM8q_lNCL57@UaAqrEj3(g!!@4JL{uKHyZ2dO|
zZsCz8@JqYW>SbEC;H~!<7{xM4<2=VNKi_z$`7ku~6|6TE=L5sr^2{oJ9ZT}-21d55
zo&g?b+9hac*y&TXDw6xf8_m5(jvOgJ|D3$IpSyhV3_jhDJ@TZkC^s0;+~0C8SIm&J
zoxoQn+w4GX862RUPrMW4I=SnM+;DJYg6^7uw1ZtSyXFt8$~bq!K^L?I*WuV%Ivkax
zqN5=8iB`~$22O9z9SY|tVz1!roaR$oW^1vai^wCyF_CpN+=nE&V#2}!_d36AqZTHc
z+jcb`(3|B(2Be<T5){Lm$egq-j`oUet8G3iCgbSVCAm{0#%x#=i*FK^Vn!=gKl&@N
zi>+MIKQ8_R>rRp;^LoAI1ov9EA}TEEgnyoOmQfr9TR*92_h#KREYOZ^!Ejgo3IPo8
z&EK?de2>{_jpw7PEYRYR_&b!lq~vjuZ3Aa4j;C=_uM_DJI8mf*hqjK)?cKAqXZ5vu
zWnJu8OWqcwB+upxB|u`oRx)eEu~<lLWN%5yQe21F3(}R`wc`@IKu)G6F7XkPQiK`B
z$O#8Jmz0@tSqK^*jYz`(LF)PMB56t4a*PQ`(qW#T(_I7lmYmS#`+00hpIYR~3D62!
z1}C<DY$uZJ=)%q-#vh_BJ}XgginRtLl^dAVM06GmcqM&pgF7;FyvQA<TvisS<g+f=
zhf#Xek)fT$lnh(mhAnM!vvj3v?O~otOUDxp8}84Q?rRIsoyYB+>G~%cJ<!H_OS=Bu
z%C}on4<eux4jLWT@ky0R^Vsv|c438BixVmVqwo518822XGFNoge~Q6tc3p1yv^Dpt
z0)Z$t5rt7Oq)RBh4RISeu&UX5P7`2^sSut`vft2+le*kIp+-TqgbLSH$%u)argW0K
z4NYm8W+WeWz6W@zBWaO@R1@68nwIZj^Y}NLXt$&E{r(uO#<)oPLc7}5_ad1&^wM}u
zAysy^T}(<D%#m@@?;yLO)ebh1w)a&d?5_>tHAM}kH`mB*U~eg`!}Q`shd?z?E)SMc
zt_vn0huD*RkY`U|QJFL=JhTmi5n;OAke~vsZnOGkgw_6eK5x9}yjLVZ3kN-{Sv^Sg
zHb`#+d`F0iUXat`)XY_1yWPpRBV#?2hY3xG`&(Tk#D<#mTRk<kjLFQxRMYb>Xjbo1
zvU!h&V<yM$kYYMAQ?l<*YMw<~CV<~E)sm-)^E!d-I<<9fZ{JY7kM$uLer;+mlF_6#
z;6QO?$F~&XDpfmfWoUCLWw%ixm$GPov5QmM#b%U~mkk(GhKK0hZAeQ_6wdmox$8u1
zZ(`9ooNq4y+=`0B#}MN~?F(x;Ns&h0|2`!$7a=qpHdKbJq3uME`HLFHrR`(s&r$Z5
zG-yt`=F|nb)Ns0aa_eB-4u8i(MKf(Ef1L|~JY&Ttoa9lAJ6@g!=@~3MKB<8%v6x(>
zlc>01vQz$(3@qSnJ$a270N6;`-P%#thT6QDs(-jq0Dwu|E%_xS1{eWvh9ih;XHnx=
z!D0!*JpT4j6~2KOXjnW}#83Gj)rfpViH+rP4VB27Fwe*8ffhZRoY}evjhr!YC~S`4
zLMxQVhDLLcMr1xVYQ#hlHzl(h+;L2IYGg*sPCue>C0w-S_g*qGpadr(AkfGVtqdEm
z4WI~cM%G|hlh}QfG#iw-QlL##Q@%`C%tK{hZs4M5V1B@uVNJp?L|%CxB?m~!P=bYp
zZR;EsQRcCBeZKmVuML4mBM}QDx1HpbF!_>Srr1$rX(AhL9kBv2fvzY{;a}AV87Yx0
zC(9CKZ{vqrRpW6`^>`dU%7wmDhz;=yRMKc-4k*eOQ{u*fl_NW9#vw=(AHsb%uv7*Y
zuVP0hlMRgpW0@Hxu2)BSPUZYEGC|J9iI^bNP^#w~)Gajs&Q>q1EBBYLScs;rH(t6Q
z7e=pfrCr|gv2p<tfzrfz1Ab{=ixM~OKzVxE2RxpRpDMpZfW#|!LDt%p-;O8OR3oZt
zj*-L<u-j#BoqdaLy4GeCAHa_rb3z=6x_ynO6h)fwuyD0nUBQ5@?%2R+8RXiM1OL0U
zVLq)yJ_){rb1jqXPpaTkBq7S^m!!c|#47fOUn7vuU(}5reP?B{I@=&tP1tW&u9jpK
zxT$*tVlw}^Enpu-D7;R^4ayf{n3XSszK7Xq@Q?ep!mQ3mVg;wxlJ=GDsSPr$6~ol<
ze7S}K1~^+xDL0MB#tRxV9=JQ!kJc2iO495)xJRzpOT*m9Bo_PmmKEn7AhET|U0Tgs
zt7fC=WusPENwVGy@@jQkJ79Q4Y|&fUo%xBd--1#x^{qfnQNh$&RABQN78~=`%Ct{3
zN#ro`BPxsnK1GE*)-Hsi{%%d2?evVE3J2>ei^P616=chu)aG4V-w^-yDNj9lBTPNH
z=2*Zk_dfX6u=~mA<r0u>nbclkmALBP*Q}2624`2(lRU1lLPoOnU_J<WwtwpkZ+hOw
zAf9IdVa>J8a~}{@UMe5VI3FR5s3a6-j)5R>8G90AAZoN=iU|PZjLP{Nm3PZ02?G5$
zu!*t1;3MjZKfyL2L&N}L6D?$B6G6{sh$>@O*;yh5>aB`dAXPVl;~oi)5ji8Raqo9;
zlY*L>`CC~@AQq&$wnzXslN)&50Lcr8g){ig_ea;(_x7{HiCL|`{?}5qG3%*R*}voZ
zs@%8et<kBhOxAMLD{{oAbl<0)pgZ6hm*bAUd0^{TM)mZ?`BqK7d0^W&yZ1z`eH-7d
z`Uu$BY1_wOTU+s1B=H~9Y-NZT@JDX`l>0WvB0JU^3v!#rLU-I)xI4ES3)O3lh1J|~
zEL6)@T%g>yUI(hxrMf$e#kRIuV^PaAx*<VPbr83qe^zGsh33a&Y|!#Xz8_>UCqh&K
z?J!`epDxdwE?+r~L20{gjJ&KeadB5Hw4r+hPvPh(U=YFZ%va?D=U+UHkAD05&KKM5
zwJ8rw3%fgNA=RE=bdeGWeC%PlOT=P1R(sX38z(zlp`9m;uvlac+v5xdd`cY3S|%6f
zp0CZyk3DmV)R(!gvnDW+)px3{0XI?4kr>*Ull~8?`GX{3YIQbDdHhb-2{Sk2$C%Fl
zKGpet)#=tdB4~+ycJxXW`mgJLZwuSIZ|do&eqk$M>^E{-?YVZoEYSEc{Yzx~(xZx{
z1KR6Ll~YGSen(UCaF&I2Cb=(pHg%fx^bTdG(wS8A;7%aay5in6BjX?aNXjBS3FJ)F
zf07Sq%c;O=aG_E;`G+w?MJ?qFd4evmCpZ&G1Z@swFF1w_72_OYhne-V405nCh}2c`
zav<tx{Q>r=ly(bT7vYp;%+q}}?f*FCiYD_3@Gh8hz319b+K$fo-vgt?Roo{`Y~pwK
z)OfREoh|hEUkWn*Z#H^^o<<*h!!KH{A1))LS0xz&t@mF0EthXSuQFQbA)1JNBk&V-
zVf>ZEc~vY_-{Yx4aVHoC^{w|ehA8LD=w*pLMieoWa+#oiK22*UQ906r^=XG^6`#e+
zzK($rSfF<fbtAD$@P*_JL|>yngTVkQz^-6WbS~<-0(W0`Wio`KBuB}QQ!?U=n}Pv$
zxO^>92u)!vizh0PE+Hrh0f~oPMLL@>O(<|hBQ;%Y1)iw72wAV_7Wk-E*4&>PP=^1r
z*0{9rM9(mdl}UYLfv-QS`{E`2D@sfjV|p0VzL=Iq%!`xw7{g^7)QovPt|#8l-EfzT
zHdhIV!ls8<B}T>kEWyclVaCwGU|G;q%o%R*oC26r)$=8z7$0hz@AyBX2FD_!oerTz
z#la2i?3@^4#>t@-yi=MDv}&<SVH_}1cl^Ji*Ws+F*a#ITcy6=RM3bHIU)G1jBNb+5
z)L2^1z?n6Cfw#+5HHbNOms2yPC99AWO2)$@<fT4x@~K>EG$oIUVdPZ9mg|uv5OoLB
z8TP#Ls;eiQd5m(+Jv;4mBk5wvmB=~$8jbW}GWdR8g8xq3!mJ%?=Xe80HLlSjK?`65
zDG4<CB#VcPIvn#pi^ZteNLUx&Ol@@OzUe)RL>^R34XeEYbz^RS`H465LL=2$OeRKm
z6z$~3Fh@B*-*^anaj_>cH#!HK<NrHuIdd-w5W_-C<P{DIdP=yY*yqe}!@pr7sj$dg
zMy?JfAd&Jx^;)7t$w)9M$5TM2#WPZ4t?9K#^(IS9Osuu{HqBh0ddH0PV5^0+i5Ml@
zl~~&)XaBDbzyGuPpij~7H#z+N&#8gUWAFcOD*FdYgf9GlsGKQb{}H*)(e-6mywjkD
z`?e2MMbUW7l<M(#v{~;RqkVeo8vkaE>@2F1=!*qy0)@6`s#cT<h$0z~BM7#v1Hrxb
zsAphj*1`6<->wLYMH&(qYbU&$Uvnk^*ozG{_ZLu#!W}?%g9;Ud#xa7}sCBx(L{S~2
z9wSP6&8x0;m?j!HLJPTV@yc@Sd39RIT@G7_M)z(V#+|%jid0vxpXRszbrOx5siMU+
z%RLK%9yBlg1<Z)hy<=gmUV02x%NXI?)r3vPCxuY2#%4sdOyMo(CqlykTFiNI0g?mK
z;};>D`CI?ALUHnrZezN8WGQ3jjlI+$q+0+=ZuQB0XZlgJb<vX&dXxdh1vDFC5iX!}
z?zsRG=~-fI;82z|hcYY*wPlP}eIrYb%_CjsHd9=%P)!w2DfwGU1lu}f<z}ARuX$;s
z$q7}ttTMxfb>%*+#B2b~mi1RvrH?X%n#Ytk`%_Iav7Lxl?5009bin_x0*KU~#?-QT
zui@~uRg7u&gp*ne5`ZS)w@*P#3JD2;C?4P#=JZ?FMr8dF2bj>BpbdvaX2~EA$6XvE
ztEYC_^4N&i_XTN!8JH?lVmFY>X)lF!mO>K{*keArry0PN4>vx9F_1M4kc$7IAXEPk
zN)J4!2uXlQ#I#iiEf4WF|A3N%N*E1!M{J=m;&Em)6w#=q1~o(7U*ZvPH7GtTM%Z0Y
zU)~%3|D`rJZ~3sBsGqHH?eiKrr*0^wQphS4zN$b#OHg@g`=07}auWv7a(awk^PuFL
zfdDeXZN-NPuza^t+z?#pUg;3bUR546{wqpsm3@+23)ud1mD<;J3!hci5HI@M>}^No
z^J&N~4q@XeHuY_SPYF=lc5T4z9*aTjCMnw_%@`+)Btc#lCbb7#Ah+dHhjkArnAse+
zUET2eQ)Cj<lo?IENoX_3(vX=c7T`vghq=EEMSkjSOjpdHJ|0EkBI-K<b>ICYk+l3J
z3$Cl!Toxh9#bw9i93kbIX<4$IRzNJBz~T{PD{g9fzEMAeGKI5+mf8`mEdIVVIx|%}
zf%KI_lp>Eb+Ck|#-*Bf1c%17~^Au&v;(J9ZpLRnnwi~0cZy5y4xdU`O+hW6~%H{_n
zuKfcOFY0?yjIVNH%t-{r`T$~dDa;q25!EEppfpPDwmMjF2iE2^@%n)!&nkA!onapH
zea4Oe{JOr-y1vk=`Y?dW7NJ2tv&65_fMc|OS-D+G29z|Ed@o5$iug}e@Y_mk#pbU8
zy%aI$$0;JVmpt>^p>mK>FS7oa+^w6l=!z@TE8a)6ev3SwC=`?XfMjU#$F@SUcJ&#p
zSN-heR~wSY6ZPK*ue0_5R&ZCZu?h2L>I=}am>81i%5e>67{y`{asm++xRx%a3$K3J
z52gKgdRYKn&YV#=)dSG?L45jqYFKoz?h9Nm26?W_cHPDGCuk{4%S3XarJfqQ;ASFH
zA$BRUrq}&ydZXY@rI{1IjY|yP_T8u(`#+CCtPFr7k9#=kfJm%R;PAuLL<ZF8-RP6c
z3KV)DGS7bQ$VhZ7ZlgpS{T*%4Hc`WAX`$3n3Lorhf#%gl9u76?8yr3c*F+mAQO=nk
zNCQj5SXY6<OYvKVc}pqkrJf&BoqN?*>sxlIA90PMe1Kb{N`E(6;~LM*NHjC*t7ay{
zDI=XTqj(ExxM{idpj!_Hz1_^)Q2hL8^t_^!CpkwY8w}g?oTb9d?h-M^^CR}Wt}r6!
zc||hM2Sd06ZJ!s}v*d#j+Sz_jG0OA7?j4_N&&3|bV6<b@TVpdu`PdHSdnw=3QEtz=
zn!6M3j?H<ioqgNK*wE{KrxToEBg8-26_N|t&XNvpYcQ8@6{g`cwBW)tY}3ZKf&p|m
zWkyk7wh>v*8YCwi{Q$kO7+)5<;Wj}TwnSj)b`conrj|HrL-aQ<FzQZ%MCVrT^j_Oz
zHT>C*#@}p>24uYep9LuUvwg+iZoQ&ea9_Q4zB$m!Im-zM#OAOovItr2(kN5~r(O_!
zv`-y0g7lX)@OdR-1$^^Sk?gb`sKPJku7<XQ@-@Wp?^L!(LQGg}`L@KbFc+Eey^g{)
zr0``mWMlJ-%KcL%zo6t-)R0Ixg6hJ-FGi9b^Zv)h$Rb&|_!d2?BxiEa7oMZW6ZKLy
znX}KmFd4=FibTp*IN9IXXl3nuRQaySkxe%&tcFNzj-Kr3g$p$n59H+NrkiW$+jl^#
zcrw%ekYzScz+@@98`n3C2k&8Hb3#KPp5W6=a{Yr`$>>0+t`C_+pAn=YO7S7Xe{3I8
zW6Thc9E}pUPtszdaUd@uLOW8V68H3+{Z+(P>gyk}X8T9Xj%{lC){dn=mwES$p<6O7
z$?;?4>)#*=<e~|)2ytJ8h()H0_6SD`W8)s>4qAIGa!=cNQ3WEP$=w6Z8ms^@<X{E5
zgRDOhbi)>Rn+(<Jo3t4H8uhxN9wlcLWX6CF4B8C3xfc5%yqWP|11DlN1<lC(2RUS=
zz+QB1y~uBQa=nkVPjXz!yU>zIua~ztc%}6LFh<t@GE&_EB31WDdlKaRuSVCsc5U|*
zbI@RCK{i?cJ2Ug<!w(jH+h(@ICKy9o=JT^S5TH)4I$tu|bpM*0zKJVL2=lMFY>7f+
z9?|U}O_I#>O@<kZ$7Ibt&=Q@s&pjeSlaqCx5Bu$vQsX1X#8X3}<S1Kl;Zd+Tf&+Eo
zWfT(vs?%{#M!@A^PKoMJ_rId9{zWDKR>@}8%ADYSK`Z$*HzifmTJ2-4%CKJRnzhV2
zmi$N7xg#bC;%Wz-#+F!!UsHqnndx<Z2{Pb3oNb97WunBM@Jo<{C@3OJ!WIKO7fV)#
zxfWKEh1tXM?ShQS`A;@rV$8OygkpJsCkEGtYWso$9BaE|xK`vU?e24!oZi&W;cIDy
zM_k`U436~GjQ>3ij+iqCOqJ+>n_a{Pv5-#^suFIo81xdtA#eUgc#nC+4T-`nj?kB1
zXK~6iXb>{1$3iadQW&OSXrY&7-}P$J!ud&YP`FLaVXLuYE77tpP7wEUhehT8nwI*%
zQ}XLdzNX|iNSdYR+<|Ik4>chtdz@{t3lgC0Q8h!nWK^JhRDs}+k?l#JV=E+Wu_Y1?
zUaSJ%F+W19@d77G$*w?TyOX&rEB8kvoNCHD0qXv*s_560d`O9*pr7NFXfuwfV5cwo
zb$dn~j6W;vZd8}xqNdmvoKEUMb?Y7RC!2?MJHjNPQA|@AE&kt7vtL)U4=86y><#6%
z?1isW_8+ttPLWAuMp*OCEmN_`v?~n01G5Y4l2h!7qrmxnJHqHF)SnnH{b;1L3uqjP
zgBYAr-GH-^>V*4;9JNz1qQ6I9Nz@`!h4sN#VAzRe!<vVLOmPHXN~RMT!t@g%Kt5;%
z-y*`XV~_z_5X&J{X?^QMn8Xghg1P{?-Ip4<SCTker2OyxE`=q5o$@-ulGIMxePl-d
zPZ+xe&~61SsTI}6D|YO{CzB}6)ZRZq`2Zt=aV%;x|2G<4jbfu4!;Aq7m6?YH;Mj*+
zPiN`+u11dzb87S&Yu6WaA&3}YH=>MO>kD?-ArYDN5u)4m1%t>Sg;=s&aSRlwO9a3C
z<#aO0z`~$cyo$Q6KG7I57Hv4#4ZF4Lm87Jg>tX;#UP%U{jr7LuS5Sh8b))Zzg;7`r
zJ7RKS4=9oT1S7CXG5{c0+p9&8-O&1;SOxhj4qft|zv<qCs;?02Mio28;fWXv;q6B_
z&tS-cuMN@Tl8d0g6TijPfY+}G3iSF9gJFCG_n6vHrV&M8+lX1CONj_9f$`y}%A^j1
z7$HW&P+*XO!xNo=&4~zRv}g4r*>|2_;|ng4vDg6*V$qr4^H?4d1=f$dd~5P*qY8>-
z(+e)w^s1&NEJ#M-Ux*{wdWHUKz1k~!$ONoWw+c)W)}o5EkpUaO%^$a+PAK<#BxOWd
zj_-JMs}MlJjPR+cbD&{_{QxK(Oh_D|fU>D2MWKXUFcG8lmqaO;T7i83QN77xzC;lh
zjBvGGZJjvHHyorBE+j&-`PMhxjd)Zv@Df$nsK{)}4uUAj-K#Y^?l1^sg>HnRZc1vr
zeVG0#EgX(}nD~?Tuze`R$TRIREk9nx_|Jef!}JrpWLvb@Fn8s~v`(<z8rR}7qFNCx
zZ5ZtTC;IMxLrDyABaFdmTf!f356|ln6F!-SSs>K^_eu<hw57vlzbL_v{~wfmNr^()
zIfSvK%m1jv<nbR+?mLu-8gt_LVO4I+=)bNT*5A*OYq2r55YFn>gCx%Ucv?pCLM&kq
z1~(O1B&&#+_X;#@s9}c>wS}#K@|VNBNenBO-M3y;>t9pypOlE?4huT0VhU=_;#(p7
zCx)F~ScLl^0E3PZrw%Wdp1ZX?Z0A^qwP?d#JMN8cEt2<fKatIVb>vZj{RwIV;UcGx
zmf2o0brSlm$mvdUn=Hap$b?UZ&?8WDZN0Vh*u$xPqV5vtuoDRCj(uC3W2R4i;77!q
zrtvX;>p#Uu2E7qVldT5QW@KK5s37R@f&@ujP}owCRG*=!5-3V`WhE380|syl{=0aR
zlLrCCLNOmi3Qb=A7jOb=4g4`0U7(m&G-i|kfnqvJ;TfEo>PwJwv{C;)Q&SZQ!WLNr
zavEWrhx_@G?yvn?Fo?IvzeXmNCX!}9sAnkXBi{=*l3ABrst+K|X}Jsi3ngTS>H2R)
z>aT%d7_@^5ix53Y$V$o6lEE&}j=7QRG+)!j?}Beh>ZITW5D0;7vV6Jqly)bPa}9xl
zhCwSMAPopq9NsxaP?9-Neu>`n#%Sd_SDcI;p_PwfeX=_!#nvdF5h5|P((V5=NR4NO
zRIdFrC`HySKOz3fNX`;Kymu3Fxm}FujyRhg(Wd?1X3n<(51>s9JGOv`QMYYA#(Hru
zp$#GoZ0rOWY}ea1zAa!dmEAXo2@%EFh6%r^KKvK@Z_vU29p!908`=4Lx-@X`yUP8(
z5`%6A$u{qJ|G(<)KURBxpqzHQ|A$JBs>~AYY6~&)nD+3KTqSba!|uG5QYwSfvQBT4
z-3j}azdgw^yH)lm3jgJHuI_g3*-S!FX?8Nt-4s>Z!mZ47>!th%jVrI;`X@+in-u&p
z#I_O@NC3<4Y-ZhPJ~rV(BL&@vJCy%uL^npHgQ6#WG6~(N`|`s1kP#LgROJ@Bk*y8E
zfmj^M2P1#BHs-ZxfxFCTY7>IZ(=$V`|6iqS{sJ@UD4V;sLQO~cf7_&t|644UZ<Cby
zh}Mq?5IjM*Npf%h-c3XMXH2_%XWXU4x6!4TSb@aYvIj5nS&rg}w;xJnAv_muy-h$E
zdGVjQ(Tclk``VD`w{Fw1wF9jU%D`{NzUJF?T^}XdE|C%LbUP2s%AQSuLmB;%oo7AP
z&KG$$$M3uOwR3;rksd_h!JaX91?@6%YpGeeBjgai2{B^YHH((sOJTB{mH9T@2$4sT
zZJ^M?$Q_x8xVG2Rop-g11kmM9BwB}vhG~EEkf5F3$DNMR{2eY^HVtuIgu@D=TvUp3
z6y3C0w%|Qt1subH`<5F(^DGp_vjxm?pqEvX>hKkfVu?DEDhZNt-eekyQfj+Ou>DYM
zsiV=;uaOsG;GNwGF;xhX#Yj3^hXe4%`1wU8K&ko#yC*)DDC$_CNKKcy_8A=Aa5$j!
zav--6#E_ID@@&<pou#aRKvZV02qTgEXpZu+g4|kqE&-2=6CG}FsJR+?%P+3ItI>mj
zE2S1m?=UV-eS&bre8ss=$3V2+r&8j}a^PQWM0K{_SMUSokst*^XDrRXp=J67<sJ8j
zz;D@GZuZ@&v#MMWL^9kIP9l(+?Gb@?7X!=8JwL8CzEg?eNJC-I>+(73g$HD_x5-s_
zli8<>gl#0c32~UwI1r}x4QG@o8jBgR`hQQYipz26@TmG>iQ&IAaO*E?MW{$8Jjk|E
zG^_%*V*EmitkH1bAm%WbKb+mV$G$~h-=bIlEXf-ec*BgXJThna)pj(L_fEiH&=2zs
zva4Ye0p%$h=&toaCes2g+DRbhqb9#L7G3Jq+y}Edv2cCY`pEij7SgT-VtybrVEVDE
zXkWV^jJpuwWC%$(!e6Faizv3dzGB{!I{!N5{{VN2=!b_{h`UW6nSUTr`zcC$pfGd(
zS6Ebf6DMTaU(RvFL`RE~P{OS{UN@dHJ*EwnMpG@USPQ*ZC02Ow#(g`t;`M#AHnc!9
ziALYmB5gpH>7NUpMBMxSVE4@=<<GSG+$cpQSJ+Mi2uBpyk@GWy9>g3EHU<eiiO>a0
zZWnlR&)UmowarP7(A?Kekiuz^(6_kaBR2U-1S>j)nsc^IaPsNKNMko=P>;#x8hv=$
z8{~Ttzu0NhW0<_{MZY~poqZCotn0++{lOku*^4eby}l1U?g4ssFyI_%(a8u{K8@e^
z?u~LVO8wEsp<vV-W-j-W+wbkN+~Ht<FzV*n8$m3*KiJP#?dGd=7Ns=Yqc@YEJ%Cmt
zv2X;sK820~qQ=|vY8Ie<VEtf6$V$EY0B_<rI@%bcjUprSee~&QaFFst!6C}`QeIHo
zjbpTVym8`Y>a}#eb#>BR@m@~N{%m>Mdz0RN*6EIKZjDpig9E&&*E?vZYJQAfjEg6O
z_c+XZwN}fFd)mJ*IE;91*gFIuJN&xt5{!*^Kd|-P=xq*rM_<kJ_9N>@ZKRLzu4BP5
zzUmlnI{y8c&!--+qkW^u$6c>w{4XLTcl$_xwaWFERh~w;wSIj4#QI6*h!bmr6TwM*
z{iL^Nw%5nK6T!GUt28*_on-wNgA;H}<UXDXPDUjsTP3q;zBa1QQ{%Jj-1Tw3<rLq7
z=%{1<bb5o<sfAAEnLF&84y+&DI+v}!#3OoI@8E0C5as7!Fh1(uaQ4se4d-}6_iI_H
z5xn!Xd)#_{hjEmr`Gzz4oke%&-HHAzwSH>-j?P{$91ZSJxI0$z>EJXyy_@#(w0E&_
zcH>+F5%b3RTM#Fl37m>ny>?^PzYqsaeT#PwK~wW0`=g+#_f9By@|Og5|CW-!qeLeA
zKEJ5+Ny9!nkPH7IckYNU%jB>}mVYhZZSdSr>Y1NX@>zS3<c&QfV74dprvnp)-B(tZ
zrZ~68{3dp7au;Z|XcDP!6}|OMM`-Zx)h8656P<(15@+q0fb#!H-|)w}pT?KH9gnoR
zw-p<<{GKR6gvZ&lrp)jlb9ZRL?*U;tY1ObJkEF0}g>ArLmx1Aj7)HG~gVp6~Upq79
zaR$arg+GkhL2mnfw9-7@>a07|)jUt;XXml7oAs;J^4u!nxp9JqD8F4t!hM^Ggw`YI
zMTsRBj%~i%IlXwb8bw&I%*@WY>Xl~t?1kCr1L&i0_St4B9-=2|D+DWSZ10W@iH6s4
z=VBDv&10&kr=*wCjpjhy-qYxUbSoTcbw%2tsB&{xT(3iZT4(%B>>w(qc{|CfivSY#
zcM9zOoCExxg@AD^U}8IgBzrv<<8~+ZbNo&hEmc(YmZ-nZsoXxeEpauXQB<OBS{L?B
z@`JGO?&=LY^4=GX<@c*B$9eVs+_WP+jbN<lo{3GM`XV*qUsA43hnk%CcU5NM+|MZ|
zFAu+|M7jqHXW}Qd71mW|6z-~WV$XcZAbnAm&glF9){QpH*w9@WsoGJ!<7YZNl|W*g
zJEKILklV6^^MmivzI{+POi+vx-7TGg6@ojr^z;_|8#%3tw#<D!9C9`1XJ=RI3VE+U
z;eJ1TyzL%xu&kH(V-`T(&Wm?G1wTyZ!h&w7Z!VaKQP`$D{rl>sQQZcT@=cgFSsVY7
zK1oW>kRr2{A%s`lA4P*O$Q1TAJ3lDw;uu)BRV@Mx-j_3BcWcM6d7WKoZ(K_o+R48^
zsg8U~iSh0dhs&QQ*6`G$!z&cnS>!)({d?6%V+<bRxxXgH;O8ky<YdJy;~A%Tm3fR9
zN9TCu4l_P$g0NKbG;Bl37zQ!u3f8P>f2Cv@!I=x;EWw%n!NPM_xlvN*vQ>b^5Yr){
z_()nc#D*vwv@0W~E+u0;cNhWO1u-YwKWyRv7s4pIvfe5}GKUaa-=jd$+w5N3yE7KZ
z<oROl4r6lvkPKKH-z|Kl5e;<2jNV!%gZ_-~!Fv#XeIbwiY0{mF9>CL3R&dWsT`Wd<
zQ;DAK@;^ne92X2L)rz+VML{qt?&AAM96bxB3iU<sDyW7`bgif+y4IWf{ok>$S*5GF
z`rmB~z!b~Ivt-efvO0>4R)%1Db5Qdfv%;Ac1+VLMIV9LYs&mAAFiA9Qu?aAW6U}3s
zfUF=_k!3f)+=5*hX2+7+#jrlqJQN4Eb&;()>od%loUz0plA%OMn&s!J?zr^TmGYCO
zO?>j%ay^QN=o<^o`Aj}Ug*1&}MzRm5)=Gz6aXl0_3g)2zXkFk<jt#HP+`vTdVZFzM
zQU*JJkZJ6Qvh^NaKC8q)>7sH5=p-6)#uJaIfk%~0scc%wgpwg8Kn`F)s}~@}V=A?e
zdyHo$i*8UV*bDj>RgjlwimvY9(Jk^Bi&1^2aEO4t0&~fAQG#Tthz7|!cS!9z-Gd=d
z%LsyKIy;2r1J4uKZFAW0*cU<ljs-XiTsRL$x4I>Ysh@AK#&JC|KLof8VmjR7d@1=X
z5*#hpfEU3kOsf}91lAnJH(U{$L54ZQYOnj2QoG*E2`SLs+&2-5TuT?B4vCF=5ASxf
z$#M5C>_v<AcsR<KtK+Y#-gjymPAT{Al^9Aigmq4Xxva83BH_%O<=XWNo*_<Yqt_r9
zY+?*D3&?fIK1O)<X$`v2r(_?J^CRWhL5}Fq=s3r?+)dR>mnSCtC+X6qOL2JhW%Y1H
z$u%WD3GW``kAkazN^e#&p1X8uf`8*Fp4D&CVy<T|QH((me=c3Jrv3lGqZIXBrh4?c
zsEo^r@$O3#^g_4voQvL0i3;Qn)91psi-x=GXEjLH-sQ`eE?>6xVG>Tc4RDD%{^KNJ
zzCw`P8T^T#Qg=O-7nQStf2S^MN)}Z1e&r^W7zHq=oC&o)t=xAj`2i(=O^LjF9SQOC
z%FQaVpzVXo{j!o@QSxt;$cWQ_LdjS4qOU745<<ddM;I98VFZTIl<gj0v+2J}m+>?M
zZET+#LVGv22B>5D735smgp;7eY2Ud*HkBM1NF@96g?wRGv8T{o=q1-zC>8I{QLl6t
zsI0qKEM*GWLVt09aP57?M<_p_y98?4=l+tj`c`%>y4PRo<LS}3#zzVVeD!IDuZn+!
zb@XF!pcUr5Z5cwi_k1-|#sBAZf_P&Jl%334cyUF>Dfn_NiY2w8)6|^)M#q>E(lvRD
zPO&*tl4lbTifD(j!n3&781!p{+s00VJnm`aKZf-LO=l*;;ktoVsmd!cXF~wW$nWeh
zZ!y)Qe}r7K=!_V|`D|f?=;o8EYfJ1|zMmkSTUxG&X>x%riHX!A*vx9UZFGP%T^l<b
zPz?(dZQ-XC(Ncy9t*N)QVDqmRPHMMnNy|1^e&Xm1_QXi*1s|`psvk!nmV@a5>Ke-y
zhtlbyBnF_`4~&<r#L(6uWC76O7D4c&*OIko?2f$q-l^a)K@!1DdUxbJj($P4iGml^
zmY4?-CA9-`nfk22umZJkzhFvv`FmRXsH%0+ORn~t_KXg~{Nhe;=xo8B<!$k^5@{lu
zG+`d7-qiBH)W||E?9kVpTjHq#iuhb@*4n~ZN5W0l79zs~uC?}8SRZ6TN|gBMNeM|1
z71@&>U(1<n=GWT#P6<(SvCEH^Im+=vnv_qGFv}Tcz}M;{|39i+`tQ#5$0S;LnVVbI
zqODayYCWv`pmDQmxks?%6a^O1%YTb54fVgSO76IGXDMk!eY{AFulIjOT@nKjre=2Z
zGM&;gypN`|3`N;0WQi7|Ow~ovB7Wzud{33ruNj8$S7s$SOdGc?+Q<==aEc&+{1(XW
zcAF7H264X$m=p|ozxpTG0o9RrP(__>{q#ne`eiaLASHO>rC)buZ0Jy4E&{ALER7Qp
zGS#g7mAyd&WL=z4p|c_KQR$FZ@&;bk$qWSpWzstjhmj`l`4y97#9vELt!LSmoW8RB
zQY5udWQl%ff4a!=*^7!>VU)Z8pEj!giW#m;|DYHtY$6uC9a#W9?_-Q8Q$<uxCWkg4
zh+0aZ(10J<G~WTUE^BtR7{Xnhfp4ut77q1oS}hNdv?2nQ+MRWznIBhaM`Pa^V7;uF
zZ~#j#3<j{Gpfy{ECI(wEsn>%3kLlyJ+|Z7-V&;Me_`pATs9}pjt0U)H!|%ZINm}0$
zmbT@4dtKvYB>2y%k^y8(3@X2!ar$)oBSnK*kl`qn!M2awamX*lz*&+HSwCsc-6?`u
zz>u4BX^oh9G)EaEu{ll)J{sR)9Z0Wa_tU0X;=Yu;_-csKMLLZHI%AQ2x<QyyH{yR0
zbYln$rw2Grnm-?Yh4MVSmE;YInS(A#fE9m9j@&Zi_#Tizo?mR3UdPJ<3#y-^Ru>V~
z<&W5p0R~R`n7_*Ol><0tIfPPW#)%am+EPxaD|znDQr?|Hhp@PY1dlS8pT77sm~%70
zBK)wSWTVpFt4Xr;sWEg?A%ldwG>^ZkvWHp=-02}DJ034}NO68fmHxJ>tC#NRN-dMy
z3QEH48k*^s2zASOyB%9O?7|k%G@o^eLA2;@8$9N7T15ITskO&2uu~8c$ry}VN|lls
zcqPv2gVR(%oox$CT1;~*?jW4s_(wGNHMMEyvGloI;tu|5PnzoT6uT_w?ND9f&0wm_
zl2n%{GFd#fgs>zKjAZc@WFg*GU83qj!QsUFvGTHY?Ua{8s28NX9Okc-m$G2a@pm_W
zVc$0}!E-17pvW8*<0Tdh%{Rtx3b;>A)RUOau@u43A4AKHn852s-T7tZHcE0`#7u4%
z?e(Cn{9Z~qfyI=SBZv>w8>uGsw?9^R4n!r~@k%Is2Ef(rHZEHmgVx6Tc5cHP`lc-m
zBBt17O3j5<&l<x~$;a4eybG3ez#HcKq`G@wql*#WgClE8sX2yHQ`YLJMop=?H;@`~
zG*)W<yj}11c123fA8zbvjBQZMl$s+bHRaOEHA+m>mnbz4pw!$;or8gtUMek1r5Wt=
zMyZK{bK?*_Lrr<BrPRc(fci&F3A(3oG}z<q0dyQ7cfcF7+_B(5u*c26w-<)`0M{rK
z`6^kYe*w$>{iv1>K)0iS#3BqwVN+@zFs0_f^>U|5iFZfJOWHU{8wG14v2h|O)8oT|
z6ozHWyS+oEYWx-Ydy+PLpfiuynmxMqB^{Ld;!k}tiHE^qXtT6G(m;{+iVWDhr1JE0
zjZ^L#-<w^kQ7<;`0F532v>kmVDTO!B9<gUH#?Ou|oHm8&J*w-Tk7_>}*KzGPBdfNf
z-m#YpsG8%qX(r)T+&g+Y+juQ??N>0T?Dvi{T239>F<OArquvQq9UfahZgX^u@i>7R
zbHF>vXpHZu4n0nLQb``S(LZDMcxRc1bNV&K-W{gcJGnmIISUI?hMxA$Fq@}>Q@r6U
zBd#!H=M01kF8YrMjFNFGIX}lG(XlFvese^rKwahG4ro~agX9?_Yy^h%91p;p0rrIP
z!0o026!Xuq>!|hs`a4NI<N@%%0xUP#$!!CNZK-I7&LT4TcZ!ERuLm3Lpd|ElMjdb8
zswDKgDEEs>K1C<~d{u#dw>q(-7SQSN%vS2HjN=@A;RloXSTz?9Gi?|XEs>v@p0(Tx
zqJ_tXaBc-{x`r|NQEHpD4f&Tf)jy+xHqrA}x%^$>D>ct~yl=S;PTF}Ah#71r`u-W{
z4^08$moy_1jhF)PPp^6Nbzac;`ENt56=oKeah3cnjgfID{}ThkS(B?vD>phq0-5wr
zD}9T~tF7pA(w{#+1((U1KB8`ZNZl}A@cWe8tlJvDqk|0EL7C<6)=)pFcJ?SYrev=Y
zX`=jnO7<%mP!*|}`~$i)wak*rwrH3h)!j=<E-I12$v>!EtZeejx;&)BR{UY*jwrc{
z%7o^esWg^W{G%#4rsT)<*5k@aVdS4sPOfpbem5;>Bj6=~`w^{W4F>@^P1n+cdqYb@
zR7jpO?p0c%o}|+fohC0r|IO%!XoW*T_55+#!yH=<+C*y_jI>6vhKPHTl{@`U>xh<b
z0_em#U&44bHdwptKd$e7C&`4wrk1&keWvLRE@`YKHg(Lk7?}=1snOCVoct!d_FH-8
zrpC_LmgI0jr$3f<IT4|+yE7Wp){mH2{nXz@(&oG2vj-3fW?y#)or~x0)d`R{6J7y-
zLh;+;wh0m>9^C8qO2%y$joa=9Ck=&BSgT9(=qFZd)tKFGekP9oQ+{-+K3^{*W+0;a
z>}=JCh%V2%2c`&(TVI{IV&Q&beScY1t1FEw7fkuxZt3!}g0kz>fmQ=jD2YR$8R8X1
ztyV3%r9!|@5VFS}|B+Plj4_;N)d#1@b=O9^SEd)2XRg-Ip68#_C`8qzz}@;t=M(48
zyXt21ApF@nfDTzgJ!URN^F!@BVZs_20+yrVvvU(M!6Bc*y-z13c8QLH?#G(m#0-+F
z$~iw#i3AG>AQBbaXGn3AftJ8d)&DR#&gtYp2fU!on$@oT2Gj}x0tg_B%Sa@Tb*BkF
zOedP-1YU9>mRbW>)>(7qsq!3de4H+Mr8;wU1vuMaaX+gCYoIok!iK%A=WM|J_pt85
zE``1)_WV+v!|ud_;bLrohfJ4^(|c!`qK--^<ly2JaxEifwdFsn1_!Od_)TirzpT5@
zD=Dd42K4MuI(^s90+AskI?e{XTS76Z$y)!4cuV1T5ixj|wyxz^K5}VBlc*o{S+vt*
z{MJ847dYtv!K1Nh=vKWnX3QC<v&HiBII;s-sbOH;DqYX6{eyr5t6%)P>$x>-;>aDA
z;W)A+44FCKgJ6A_8Ru>xc7cFI=lb@<wMStH5dJPO)mZaM+%CJ)`*aZJ+QUpW+mv?h
zMuH&egBFMGOEXOT2XxAMqR8L;=bbKBi-U;5>xFBdvX-@kzTCiC)#MGl6DEu35*~JV
zFh)Gf-!8MiQ`8^W$ZPjGFQmHILQJMDI!}st;rLgzujvo*Cg?>^08`2Foc%%fP38K?
z^;oX|^F$*<yKHy766{417w9b3p~PZxns8S)y7v+85ksPN&zAPkgfAk1+5ap2@qdJ*
ziA~mGRVGBl&Aw6oj&--bS&78Ql}GB?rXrLO5297&M04*6I{-q9PKCr^036NF5%G7i
z%wbF?XRb`uh{#<Z2NV(iOC})-O~%P_<KDbyC(f`|z4FpjW9EulFm-MfMY8~sd)rw>
z*!f9yu<u!pc_b2r=X%!6Ig74lF9%fMOv~d_wACze{8M#pg}4#a6lgK>zuEUxRO#Im
z-Cu4FJm0S6iV&}y(rEm3CANEiO1ZzG<QXMjP-0t9o6Y}UbordNT@g(y@rxJdi6w4S
z%iq;46UT{?LG*J>1%FS;!%7}e^3zJ}6+ffg-zVV=&FV{0_}Cv);m;}g1tq_z#Hd?K
zj4TikIHkQReJL6>P_$IoLffW!Ne?HTs2h!&-Q9;_#~8Vz{OVup^r`Lpc)rGI4DHi2
zg{~c@{+eE8k{aPt{b<kxwcpKe{e5X5_HB*K*A=pQtE48|GB3wedY3l8qArq?7Xdj8
zV)x)ZPv|SScim~3p0}=@^kxL}RVS?(k|{XvX`7x;N2ccmWHDdhj!e&yYVbZ8XLrFz
z|7A)EBnAlhJ&eZHGCC6bi@|w^>3MX=>qQdNhb*Gqf+R6LUcX6VWO%+q8(t}DleTwg
zVGxNChShB6ceVP(Ms_9BfUJ^Cw(0MHecb?ZCdJlh@?K;y%GY0~))2A}14EQcE;jga
zqFX9TDo+mbqzu#j#xOFRF$|nrGMv50aOA*=S3Y3zIx-hYU3MeGL03l0zh>7X-cTgN
z`RT@JV-G(4`%H#2j0|TNqrp?ibrK60n{Vt#hO>`42ZFtfw9frKC<zWm<|EWRfC+oq
zdW&&*WSEYdCiM@Q%xA1|I2iMGBe^+9?jQ!|${h)i2)R{&u{kAN51CQ9WH|p43BX>Y
z2q#U7fD8yJ$*#Gix9@f0M?@y-`%RitUO&_+M_|laQkx;~0AnOM*T1B9$AUxj{YY?x
zw@Av<3qX+xJafIFgY0EE&U<9;Es2bK7jlmQ?-2cfcIB&g><4D!$GyWQC&AR)#^D5S
z8V|;K(-Gctbeo)H2paNOTdrZ2$CAW4;l@Gb?qQRfgyT;yJLsT2Tzq5iznL_lLmm7l
zlPvr>vFhU|z6~DxP4Lley3EWz4dMK0c<bBRjf$<t48>5;h?dAV=b6n7cvJ}PO>)mB
zjM>aW%P<?-_D!(P1z2aZgfJ$WHo`n5vv^)@Wu&*@MzMuY2seI&t<f^AJV1Ei7RxMU
zufvtw+2zsX;cvsF5_QKWqe`*fM9Bg7>P`mJF=s~EiqgFu%eG}tsF|M8nEx!@Mz_4l
zvd9?Z_j7l`Sr)mh_zdm(Hj8|h8u(|L;iq)NR+LDD4vzU-G0RWWM6X6|gy#yGOvXG?
zlEeL_gN|WNBz@J{#4^KN+mE?-Z(^ByOc^ajbyuvQ?zJ*hDV+*xD<fVdQBFHm)O{@#
z^%IWaZS%~(V63W`vMLeVAsGGDnmFSXEb@q_+p=(vOe9Z==#(`j3&@Jb#WyR~JGA2(
zbRxH6r;NHg0Ihm<?ktpP{i>Rc9u<Mv-~@>B_4%o?Ez2`9qj7O<-#o`%{*jbf4-;{X
z_YpeTczJ(H0}VZ8KN@7^P04<mOiQgwz=kwY!!t1t3?Hl`6tFEiDW!W$y`E{S*TYih
zd&H0$*$TKdAYuIEv=&p$G49MK{-9P0zlfxU;SgIVy=<MVm?@mNjSg9_q0?BUaTqOJ
zoXlg4wU<^HU9$d(FWjs%b{4?!VGlR-L>`i+d+jxz$U(w<0xn7?YeT+Ci@ZRHCdIx$
zgG{J0bjFIx(dB;zRNK-x_Y$v9d?4}pxSv}fHeb-gbw8)V5GiA!q`*rc?mi4IEF6^1
z=Of}@$<VjXP2ov<wi~4xc_L3tK++V6deJTGUuvVvhEPB=?G;~7p@+^wz$y59ztcv~
zax$W>Ob^{X(Y&Xjqis!wqXk-q8Ji=_vJULNfe5GWBAtmLVl}rH{cSvxzKPEIjX(P1
zKmKFpSjvP>n;#^RMq`M)Aa+@7m9bToZA_6Z8v{C8r#mXO$93bZ5?fx&%KZ^Z*lp&;
zE6a;;o9<kfS;4W-slqGdzKQDFIboUHb_dMke}Hw2iO^O5K4mh~Ggn^n-$8c!ik39-
z$F!nP^40{WqbL%fJKbcT@l`HJ{XW#(j(d{evn2`Hw&r71cdX{fs8xP#2j5SoL7*!1
zNh$5tK!>!YIjWQmz><Yx*D<{RPi<$@+eQ_I;ly^5$(Lh?5YhsbTOyJm)sTov2%&@p
zLR7R?g}RCmZDKpNgA<3bT_6hL4<Ny!J2tQbegO+MERfi+=?2L^p)1(2V8QdAJ2M{3
zj?=W%mdEzY_<BC>z3=&W&pDKZBu5_vU@K6Hl$}5dCFNsOK2uY~koy|)0PnDyZ*%`?
zw<aW#L?hm1%MQ4mU6oUf{HBZKaRMWxgOe3HwEOm5gc!ah{JS~`9k8<j_gQ!iz6`lc
z4S!D^;E^o6I=gRmOfwvkH<n6@T#AkNg0C$_9ccJC9jc}cf|S9d3snxj@cs8emF&H}
zP&L~LgC5Kpd;Nf++J&oi2Uj7=C_no_<FHX|*LMQ@Z4ToG)4oy+IxFmG$2%Z4tqEbG
z2L#lOtP5!z1t36p|4D+x2`ME`z&1pOGP0>+BI!ypj2a9i3ubsUoq^dtvQ9c4=(HJ<
z9nF(MnK}<e)leQ2gb2Vvb(y>CnU%x^wUHn7oc@ueoBBXEh!=K>MI@x-mx2iQ0at4f
zUS#>9x}5rNQz8#D^AqBQkOId#3U2cQLlDmpL<RyO$WP>bHT;Os;;R8@A)6o!Bn%3M
z8!l30A5IYnl7T(!kOFssQ88!Po~0EeqS9oA;Y>E9B<sMH-gOu~NCrwKB=E-g{?ulq
zYV6E&WK%57!u6+a7iz0RnY*CO4Ee?3OJS6Vqs$yA1C0Tc*#uWIJt&hgl*ts1ubl^F
zLSIAW6p%6Wuu#^{#BQzF=JhUIC<B(4xp$q}oqDZuh+$h5TAMESbx;X*BB%sgVLv_I
zuKfGs-w$r?-r5S*5ULQQ=#1mivyXyHAJLwZh!-@k7!d7iLbACg0wTYpbsAZn)DFc@
zwWRJ-7l*PjJ?dAr`HPw}s)4EGLvhG2t91jS$zZJU7vV?#=X7BDQ^Uv1d=W4yVntk4
z7quc!9|{LAFL{Oc7vdnbJY=x-1JwlQrIBj0%n=_#QPY7Gy`A|*FvUp0T*nAVwNg|s
zmZZi6O239RPrpv>6YJEwj($yrjS|)7Pb`@%y^?qOv}aXOF~u3YO~tE2f33@`K$fs%
zu;1V^b~_1Mc>J~18MA8|@<<Kt>=u6<-RC2H`h*2LnW!60TF(31(so51`C<+9Ph@KI
zAKeKd%?QH50?&(F>nMTTpy@1zf}?&_4Vx46{{z4ZZT_Ldk{Q-ZrV{Bf;4kNE{b2aJ
z&5k{qn$yYxO)a1l<K9$V^itZOF{BiGZmq4GF3lJWPMVu@3c+NqbE$`C(Wmnse`xTQ
z;}I!cf<t(RS4oOVg--w{w<=9bfp86c38-%D&}<x(bIl_V67PaBc5*xIipq+(RF45_
zT69>U7y`Zea)g-DmxI1cM(~qS5OdriCdNF7y7JuIp~AB1289-V*E}usY_yEc*0?d4
zRUre6r3d3yjHBxDrRx~YIl9`hIV0?pj9KlpP$=yY;Hu_dv)bO<tCz`*i1)+Q72V>7
zIbVZG)CG>$H=nn-qG6vkwEa|jZfapP(x@Nz_gR7OIF}(vN3>|JJU43q!z;t@Rs_XQ
zpYQA#uadnVo8z3-X=)6NFc}6M=hx3T!SisPZ!r9ymfWFQou82x#bl3fz<1co_(9XS
zF4C%V$S1{}PT|k|frQ+AYLOkXdEX=#?PDHGt+%|5O0gsj_jk608ZH%Bz7t!TIxInJ
zbF}sl8H31G!do6Pvm<AGFJ5&T>1UdZHhe>mEg@mdlig7gh9?PwW5FAVTyZ7TXvtei
zeCro;cr^QU+;>jbao;$P^wG9R1xDA^z1A=RWROa9yu6z$=2nl^bJaE8HmES{(nx6)
zUyEzGD(e!R#L+shjZoC+BXY-OkA?He9;SWS&^4OVeRcM(JYu8YhO@~VCYe0B^6N-T
zk>oK%##qwu5zuy(U*cdeWJ=1AOVP!}I(b8d9Q1_!ZAvn#@ah>utzXjUQHeG5dnFC&
zF>!S`rU0h`2_eUj9CWvlIc>6yl-~0wxNqu0NCFbXlZ|Wbj1!9ie9FJaq^$5D!}#fe
zHd<41ABW4-rtO3dF3qt4^n^pEr;N{Z+1IwU(c5cVzA4h`Jsl5!d9j*5Jao~!H-#51
zZO9&dZkzB~SE?-~w5QA0x0l@h7J;6J9UDeon@ptQx3xoRZ@cUix_7R;DF)GcZPyp%
z`nqa_h?5hg)o9e4J_k#US}ml#ZD;Kd?$tyE$hTWH_L$T!R{D;|wgRW+CY&PG;NGQB
zZ^=SssXus7uIxzp^c5aJ@I&dq_V*>1ITyXIkfrBv{6*=H{=62?upqXEUqCQu*J&T#
zlE0v}EAh3_zlyccJqCSqg~end?<@vCf#s0jX}20d#Wk0mV@Z2l?rk(+RE7VoYv3#2
zkAG2%=k<#Vu{I%GI0{zYvS28cVNiZ1kL4Qh@?X^NewG$u*!imVn7a|))|+;Y{%hLl
zeeJZZPh!299p5Y@CMR0|i4hEzwcm;w^WJg?yQzC=X_!-lafh0R+$s7jX9`aeMJf+P
m2DzATNoBI|l2hYeYT?TfJ<A#Q?v{o5SNJ~7Yc8{qx%MA+;bCk5

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder24.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder24.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..bf33a7d0e1eecbea02bd08ae3f9de7fcb7c8eb22
GIT binary patch
literal 57902
zcmeFad30pQec#*rhJ6_fX7AYt<m}vOiWEh1D2hXfCWoTcBHaRT8yH}K2Ih8S7U&|P
zIU|XROi6Z}#g^n?<aPY)M1Hp7#7h!8o6fOc6x%P$yD!PfVq0P2_u@p}GL|^+^ZnKB
z4Zu*&cHTc}psVZFt*TqMmfv2#TI}g5#P~e*vu7*4ug7EmRu}D`G#BsYTl=+mEat^Z
zF)!{V=HsQf{U#QYrDWWtrAlc_%apSGCg*bt`BL8QrREC@U8Syt?ozj1r{{YXdP}_v
zeWgCT&dm2O43q|<dIn2_l+Vr&Eew^0EG;)bys)LT#p3z-k%g_LtrjoLk1lL0ZL@gS
z{Pu+%r5y`9OFI{Km3A%cF6~~}Q`)moEEN~_mi8{}EA6v<-Shhw4wMd9yl4L4!lBZk
zg~O%8cHKLFWMQl{ws2SJF1zlVzkA_m>FC0-(y@i(rQ?>?KYwE3Wa*^E2j)*LoGzWV
z_~86K3-^}pB|hX0dt1Db&nHXwc}KlbZ`<{F>3(m!x8r)O^uW_GZ>P8Gdd%BZ$<94k
zIejfvdWf{$-X7BSR35sPEIrJ1(c8=QUgaU>+eq2x?I&e_<zcQL;rf7gkn4kX{VuK#
zd55_^TzMDwALaUpH^%jtT|dV4UEbYX-)-0L_7X3~#*R%aF{Fjdv-4hYVfIRW*{>8A
zrwixn^`+XwCr-@F*3U0b9-mrVI5EFGHS75kv(?%OuTou{t<6?vj(M}w)62El#p<!8
ztMx^H>imhx`Nhc-4?J-4{z-3o>i+5bPCqbx--F&g_e?){@8koM_nkgnx&Kt<{>nY4
z?!R~Po>ooeW=-WrP0golymx%*>Sj$`T&h$jW>3s6EcuI<DqgE~u2q$X?!EWHhfY6u
z&*bET58OX7b?V*+9`Nq-JX-tE<m9Ax+I#4slc%RD%?er%J~1<~T&vAaR9i`RXu}hW
z3rioEt(~Y(EU1I({zSDlz34Ag{957Iu_C`yvxT>`xW>ogVeah2!lXBG_Jy%TnD;9a
zKG|m~VaBg4RVL~r_D(HU>y<0@g^B9KOvMkos>}29Q5xm5OA~&5c4D4rs(#U1;5WZ8
zaiLP4@n=0Qda3V1d8OhnmKUq@S4k<HovqcMSgcOZ&Wz>4-ucCua{WrZJW(spF3g0P
z$`x+c!mL-Bo>-o*hw0hc!o*UTWw;k;0JT@>gz5Q-de}X+xLmE5X~D!oE!@h)s83XB
zOubT`_7@k*v{Da=hncgp^`{ml^omt_Vq(60VLsgX<m|-EVs&Ev>8NL)<iY;z<Z_+a
zMApG)82|cg{py)nzgB;-u7~$LzFc3dRHqibivR4{v1B-C?>bqjPn2s@=PTawd?g&B
zxBQ8zdU>{5Uo6)v)!L#T7Bpci<@wnQ6&lp_<de@l_u|CD(tO3|vi}*q+S$b!I{3NS
zE3;M7`f5uPQ<ZXkcA-+MSC(pFZb~mS#d~)V^A~HiiF&;n#;?+hOmrvTN>5IB3+JX1
z8t3SvI7=<|1m9YUAc(R4;wzKyzf|!r`?K{*ae)_`KThjstHsN#ykh<QY^^xGT%A&<
z6zhvcy>GFmMQXWDR4z@-Td_iskGEK<6|0N&;?#LfqvAyM>cXO5Ies%nho)+k`ROpF
zXZQzsyal&rPn=&|sGOisWpZ)x!U=DpK5=5kUtC@~@#1pzEK`_%t)5VqpLmkR{+?%E
zI5F=noUo<>2^#DFgy$_SlrPq-Q_It{)k=N2TJb72R;TY}v4p+eY;9t4zT%aw?T6J5
z>XcN{ALD0AebV?SE~I0fwUbB?3*w`(dTcHpB)oW#oQ;1vK1|wHbz)*X%y^Ycvr`q`
zGi`7vOfmDlu~eAwPw_n+rfcUXmMZ>T)YW=@{|J}iz;bm~OOz#2cGVms+1j=<?w=$+
z#kb?LJG!d<C5E&T0};kbw989)NiQ`M^U_|%%X&F4?-jf*V%=Vk*XyOcKK}ao9pIbh
zuaB?ArgwiLHXZi{y`k&*QZgUU_j=C@g@86v(bb5zm8*0m<9)kBDA(epY$dmrzY(hx
zuJO$2gdvzX>0J)Nl)4>KDfKvHQtEXmq}1n7N2%YTjM9KZ6(HTaDnrzBw|De<sx<6s
z_g(;$9>1=7w?ru~dMCV-mNF8hoCAE`V<}r{-@V>_v~|?G-+KUrx6OOWd(eA`l<nMm
z*!wnOJG@7{cX4m0_o(+6zq`D5dyn(G+k21q1iyQtdY|;3x*p@nMemIFH0gW2_j=Fp
zyDz%`0q?B$oIQ8HH}1WkI|n`yOT|hDnHMjIsVAR&{<*jQU6N7G%rhw`{4g<pdi=~v
zj)ik#o>hDf5U1t;mKNt*WQ@J#5q#~~6FdLIzqx<=V_*Bo@LzfQU$1=MV_*Bm&HwZp
z&rQDe*w-2}A&UIoV_^~qF;#4&^Q(E4<E#2=0kgawyB1rE-{2v^Rxf@jR_l3R%u8I0
zza9^^2C;IYo(y7hsk!v$VxD3Nep6paU5i6XF!#%upqDTk^bqEP0%1PrA}j>mgk8Zl
z!tQ!c5D!MKC5ZRd`>3T?^YG0C&&-q7PfEXEst@SSwKyq*^`w`r57mcB8C=_PgMPji
z=gGM*B=zKA)Qc@E-w5|d{7NuNd}~1N@@OzbxGmU1xIM@zpT8XBxY`lyB-|P7_Ft>-
z3ewl&wb$Mk<1N#c=L^)dJ4gqaAnWDdglHt6()w$EPPu4LK~HXPW3b2nDbl;FMauc(
zR-anD$NDoiR}2P&WRM8@g8pE@>wYssul1}J>wD|_NG;C9uhG-1d+Yn>4%83wwtMxS
zTpjXy-&AiM_WD-$&BRvs^F2^MLVL!p#h;A5I{Hdt_2BBE;9#)#Qq2F)!Tz<oy#C-&
zuy03fG<HKhbZB*7aA57O;GnK9hOFzs>#<<}ll0{3e(JfqzJKjNu>VH9J{0V~7Wam1
zG*UeCXt15|SpE3iiC_mgPqO-l-%P9?CVuR*aesgP)ZA&}_b`TAtd;jvGeK<aUT-8g
zd_!Zp>$CBt<MsQlX@1;K&ig6#2ygYk+JiTg_K*i`)OT#__*l|^l&KgduAB_xmsU<r
zE&D#`b^dBmWC0juX=xEcqB>LD$5h@2=J1O9uAGb_m-emnJY0NXxmr~Ykkd;4u_eFa
z%{sg!TyPGYnFWzeF4ih5{m)h^OUnAxbC1^`CyQbHIFPw(b_V=kDPO2ut%U_WxNK<T
zR{4cWtx~tgR~9CO=az)x>}K(DWp?I#y>`5KPRNk|<yvKWdA@iVyj1kT&WpYtC6a??
z7pdTwy%TtA>cY|@RKW3DPZsZPeTpwX{=)b(<4^NOPrdlmxpPE|<1al|rm@ABo`3T3
zb5FfkJoo-$>8Tgq-%)HV6L!taUaC~fA~<ScPbb}xZ_x_f4%Cz(i2Qn3h^{JC4=Q32
zy17!Wy*0{k+>D25uYMJ@^e(R65_K@8pz$&6$i<6%YugEC7}r<xtMMQ{XV7FVaU=FU
ziHi{OhSMNLQbOQia%yQgOjnl|mafw8iQ^~3lp4Dp@f9AKhv<U1UwZUY)c$onZ+mQI
zuYl(9G6d9Yy<9&2$o%5e#C+{hB8}W*3IA#G`p*#XfWq|b{CpXRx?F+Gm|YQ#)#Fz#
zF3&PPkfIYF(06g>(u6<uFo<}W7Y=ic`on&sAS?BXKf5v^LMKcwF{Je{UoKD0Pt<DV
zvT|1^7AnO0p|w2q>`cYYp0H<Xes&4OZ6ux-c8R2|Fcq(s>tWxtsK+Qf^DZnHurI5T
z3<hLUwK87|N10RQC5A?ihdwnLlUY>@`$hB^hx4*Ov80BlZCd(+yn2{bIcA(ctgF1;
z(RG-duT;a7`YX&%E>BNG{f9lu+#pl^u)i6NDCMy0lG>ntfMOkHTF=Z{W9BCp%Wf5f
zee*P3bZa9ayRgV8k-y7GVFqk^w(j?+IlVT1jugMJR1XIjj7q)CtSB!pF-<BpCC$*r
zir;Kkm?V!c@+M4Fm%_9-lB;0`8oFNbK$X?{J*UII=2RDvO`qf|)LQYWGwhw1Upfyr
zF~#Kb7a4f<8Z&?EhDSBZ3JaID3c~Jr5OlTdriy>J8j)hD&xbu~bn~IKtQzK@7vG^-
z<9|pQ`<nH;{$?crp2F_sDHt44)C?zl6@YDVA?$j=e&H%;aZXv5{OZiu5&sZXmAar@
z>(QXt6!+gtvVUEV$pcfw^VHA#Pbe=8)P!feRavF{W#zn}prT+}!HfbC75<ch_bGT*
z!C3{ikZOwg3hD%IlG|jY)q<D)ck13n1?LrfjG&~M&YH4y>R(dwcN2(@)HrJzZfOlU
zmyKcfKdRC;9h-ex8f2xx25RscV^_*oE0-OxD|I(=mc<UnXG=XDS#1n%#{Cbg6~<kN
z0`gxaxQB1;HG;z4{&*(7op1YJcf8lGyW_j#g@oO)FlS+ZJQweer{V>EyZOx#<4^Z)
zr;`2Dma@M=?hIPK#B06r(Re12NhobN-ks=8SRJ=ldOG0RRF*b1KD|UF0@Ik*e$nW)
z02;y12JzKwHD8D8fWk<La`EDCrdP9GBFK)$R&({lToU>Rsvq2)uP4@0+{t@T;~z_G
zi`_`97J}?t`jz-<vYrW`s=&)x>Jb%}Cd>!<xwMzu23()ZUP#p54ON&CRah_7yXLz2
z_IO!phSpo_^>WaB*=upb{<;<om~Xo-a1H&`yx$dc(W1T}e}i{vrgpnjwNq3~+npY}
z)4$1`-d33%on`vk?|A*K8ljY8a|5HXAUh1D8x%%k_RBMO=xqm-zBRTA{mPpR(DLrK
zmJGU<(E7ok+Z(!(=xFD#`oZ3;a04#TwRZ<JSMwdW1YPc#mMgbeuDKy^<U#_P32J(7
z*xNc3^Pt4%w!l?(1-WagS8p()cJGL-_EaC@YRDV4GW)!3s=3~^wpG1r^>1%n=?%%h
z*YfYM{Eu-D>iJtg=SG~8sroi=r|||oUO#iA$4fq<KI`>%5l2iY%9$FWAm?`6h_CIu
z5#v2x5nXKbUUB?pay^(4h2e|1fapj~Amw@MHlJH}J8)e*KXD0}!eVi<QbhCviQ*Mu
ziHntqDWnjQ5X2v+c$hZo$Nw(HhdO~hP_%L>9I9RPBS_Lh1G_A75z+^-*mBJQn%3=|
zLfUt{yUhywl_~Ii=M%#nty>GH*4^}<R~rYLI28)TM}#+333s>B^x#H1<i;%6%KO^>
z_k3^njnDssZwrT88BOptwOE-xwQ1^UBe35R3FSXV@HM`C{#BD*e7FB@u5a}(ESrcX
zBDq1T-Sd~`CnJ1bJHzl4ATw%0%hmeWF2m*iHj4TmAqcY(V)b8F(g`IcLBjqgblE~J
z0&Wh2W+D+-b{bR#_V)j*%B0k9kSU8ZhIYd20&`+vdBGu?&#Ckm71%Im&N(Sso)M~<
z5G@=UU#vePLS%tfR=lTt9~uN<VRfSF|8IIga$<h|X0IVshvE!}iePMOr6E8)3^KvU
z278|Wd8Oney^~1UHAI!JqhI~2q^b1RCkccw3W)@WA(!X{@aGZ*LoB25{se!Zk&)zf
z5XN>8&2TakPu$MMQ@8t*yFm}bRtmInyE~Z^in`sMH~_-wCZ4;UOLkLwdwgZY5J;2I
zXb^D!3zPgWO!7Y-lhj9UAo2YF9h3NfK?|>mJ^a6@%kNk40}2c!bQwyB==tGJ2<=i!
z+t#HDQ#5WF2w!*K_=?hgyLBhL5S8EePPij#)SecaScfE{40~G{HsBGL#bJ}t7D8xZ
z4aWl=Xx-Tir$m+fFr5}|Z)I&^m8c<`-8&r>mI%lHF$Kaf{*NoTso>oTeu5wzY)2-p
z*ctyPxw8S2_<u!(npnjDt6Y|bIv6ci9@fAlr%s#^rS-Rj8d2k#7{~vTD*kB&LN)&X
zuHa`B7=!b(iWxieHx>I81%FGymlgaq1wW_YuPgZb3jT(IpI2a<MT@<-M%r&G{g((<
zc5V)xUGp}mVhR)yJ66*Zzf06r%-=&8Uqd1|mt0HTa1uVHK&j6qP2LB6wH){NK+8)`
z8vg=v#W(^JFZKZ>yDz(1-et99J8E&|9*fE)T)BpPPQT=i`5(Mv$x>&@%Q5|pCC68`
zJzRVdRgsimBHtv#iUejWI}mW0cucn<4!Ci1Ea#Zid0rlg40yKU&naGL3S&?LPAx7W
zL|OF9x|4;MR6b-s;*xdOG?VKF1z7-`W{skiF38QA_`t+#;>LR6$9m%lLdToIuYEI>
z$h<8+tv{alW+JiDYeU$qU@Ar7#%BwW9A9J6GD(TlQ3ja>0t`~hOV$%XYAq=tHp1GW
zfRt^q)l@5WH65f8>W-{tf&{+<tJxs4mI|^r(DBq!!>nb3jMxlBtEnLKuGqyr7bTd@
zLR!D8o|7p3daRybgI!qbl7Mxsn<+v~5+=h4_H4Z8Wnm3-Zzg%#?;uLf!wlww)TOvz
zzW4#2k>3VM4{KK-?5@MyA(YOE4T2H8XyS0Y^70o_{>Na0U@OH`!rDYJ@_o#N;jnZ<
zK($AL!dl-Am=9O`YOkx$D_lq*MDGuJU?IA;(PCQEz1oLT0rpKPBQdcVLEj^>i+{QL
zsi;)`LYjAfa9z3Gt9?;QA*yM3wZFY4gz#M+_eNsVYAv+}{nH6*{<{}{p&na<)ze)V
zoq=`r^it12t8|(+d4&hip-}eUQBtg+D>=BXWKU;Fo;n!xKLS&9F&zxTjPv)ggj%yY
z$h!?n{So!F7^i!K!L=c;n-*<yw|0;>7~G{kwO*wy9|)54_0!CaVP+YMy*Cq8E%z<0
zYesu$b(mSzXMm!grwy5EWq{bQ#Rk6sD-RoIcf76(DOz%o5<|hz=VA^(xV+K5@8aHY
zFx<G;03Kso#{Cyq7Vr(-h-MI;p`(@%wSbHiRlKBNw~9|z<|`&&N0wCl$f}c;w=x{>
z5IeKB(4a+PXGu0q?4049;;V@u-X?IiwU2J$9GSwDB(!0?ybd7vAK>O8fr12);n>Py
zTjZU>$<2h!EBnsOI;FrO#wih06OcnH)wsNgUfZkGrlj?)Ps}^yD7Md7U87UPQDtOW
zKtJ?00ivl3Ze^UB{Z<dQFPL|rMv&s;R(}y$-j$`w6r4>_+KQFF=c4=1a`kYbxH43D
zsg)&492<#bmXZnjRqGR#wxRIrHL3I+EX_#WJW+Q-y9H&;qyP8oi;D=O(}<;Kt6{&o
z({tt(5_2Lzb}5lO*%vT$3}AX{9vUf3Bi)=q0xvEtOv~t_7Uq|$wTsJ@3gYfQ6IHeb
z-2ndB76;8!O_8Smq}r3HdH%mpBz_@G%w7m{)yUAp|8=FN4Na8tj7-@S>zH;RD{PBe
zhR`xt)c-!KZLU<PmJx&_GoAOEamvDhx-Je2t<GORl`d1qc1<ObExj;xk_zVdE0`29
z@mk`Wxs<Mqy}O-Ax>y3Xu=g9eOv3IU6O8Xn6kz*QdN4i$E7%Ku8jK%OEIA0P*UOa{
z$6-ng!%*fDiPyF#kVD=+0CTB7u-EOCp-rapl$KB9lOdv|WJv2hMC!uyk|xII64#Oe
zQYIvWP!gmjk(EX1pe^M~HLuf7`Z%4GwZp`2y1)!RihQ1~)UgvmY=gw0VLwy6c;rgC
zUJQy?$|uK;H?ux$BS9jQz^^}S#XdA%yi$CE@A2cuUoARWfmC*;O_{IMP`Oo?meIeh
zuO%uE!8lp*RibGNQZ$R1;t`Yd6i<yEFTS_pSB@4>V*P|Q6e>q-STL$-tJ23@#IHVF
ze7sm+MyNnpavm)<NbIA<c9MGRXc6Ip9vQvK;absVUaLZ5SjN)6lxF@H6i8CyqziwU
z*jOeaVxk(tcpcLio*7~o;Dkh_?)GP{n~7oH#j=s^jocs9tkjAe&JIJ(6yhtxHaFH~
z>!4^fG17cBIleABY2)Ou6sfk|bV&}QczzLBKYlai-ffLy?%i$>Bh_fLB%|;D5eX;(
zW%#td^`nvYHX`XwM`Rf*yvEn{8iOo|#O=iGm0k84cglaq_woNk4f_VcSk%(Ltppo|
zUsmjG1)cOfVmYT&-q0*Z!(X6rKAP{?%Kmq3R7CNsdmEqju5LB0(TS}I@`DtJI?=yP
z+&a;*SL@okeUrAXltCZmOBZRP&4s~K)QNlU)ZVDX9lO!U=XKM9Nw98O07Ya6W!>$k
z)Jtgv>ss6#bTs_VLN9vyRb{q*?Hug&`8rxDcCR^kMSw)DWU@9OThz)%O2Xv*4agRy
zuBI)Q;G30$I5#beAf1~&88^{?M5AHT=SLO$qJkeHxS6#n@Bg6^nnPCV?PRfA`PcVK
z*Tt67^`mse*L$g`G5h;dwv_v$rQEFZjy>V$)R#8xqjvvYwaeQ5_Z0gT1%F2YR?!=V
zXrQezZWh0xTwMf_@Nq*1(<yeeCr|B-k*Q5{0jq+gT#!H;h307vmI-!1Ha~zBKnohH
zH<PPrBVjSw=H8AV1+$d&_Xp`UG+SQU=~ci-X_dgHDEUfqHM5#s&B>A{$QfSBj++Sl
zX2LBR*;BPk!hi2?8Bj(u(K%E=X>p!TBmhB-EEui-!cy5Ua50v55G{Vi)jCDihN69O
z8SVO&GI{}#yH5EoplpL3jx<M03gxDZfBpFPEp-C&@8O}fL7vhL5Fg|#1!H%71Zi>v
z($^K3QeQVd-9(yw|2h|t&e&>vQ5GBV?}@(vyYZdz)wKGq+5bs2it5iCsd?<c$WNc9
zt|eCs=&-OTynt5m27rGK26iUyr6dcOgNgGJ(oKDlT0b3wjf8>p(m{SWwwAp?dQfms
zVJsu-AMENQV5S~4SyrSCa`7K06`_Dl)Vq{pL<M#b3)7}JF#g|6D2-lCLLZNHyQSAz
zci8QWm&=Aw!uX{B8>-JL9i~!@c4g+;IVhSeVRv=0YNpBy6SWKL$0@f|cE+V&AZx8a
zmD>{|&|yatqwz?y+sM&DfeD#uMLfZ`wnjj|vj$e_XD{vq1ZZ%vC}kBSBqErJuV(70
zdb*CLm%Lfoj+sqf79Fox`NKAL=x@>e!p={}87+hla|JK`rq)-N)(16dgu6$e?O@a2
zV^>)ZoiXd{X#@{FLE3D_ib293mtJ`qE1n<?eU}?=jEws1R>2IF#hcW93A<Qd)|c-m
zJ{ET-1>$(ddVT3a+|9d*89(4vjZX-3$XnqdDxUw_%Jc~Zdo;u!SI!jt!;GuS(;dxR
zW`k%6n-bx$BhVa7f-RAicg;O&r-m*m<MAF7gf{Pr;9ozAJ&iG9_!pOcR71IwY!KKC
zWhP#P^iH9D*~w~4ye4&jD!wx4kY_afR`9KU3Uv&#7F~1U?c+kj%phy1H7EjgDUeKZ
z8lsv7r@apeEbx7CrmPRaNoS}rri8SEI)%vzL~CX(WEtDC9aI4eM7hCXn&tDC^+uYU
z7$ERQ(nhUdZH_sWyWat*8$?Nx{I!P(K!r&&17>Mo&s|Hv8fXdv1m1u>xR!i_Wo;?E
zgG&h@2FwIHIj(ix@lmxye7~;tAU=t|fzTzws!SrAA_W=5U~z+8yuA!290r|RFP3`*
zV9XYUM}LV|GYhdoY{k(t7o#C*XWm_2`cG(dJ^3x_3Ubh<l|4ozs&9zG9>|88m@~=F
zHR}FFb*K@f0Au61oE1mo{-zyjYhA-)v;}J$cqwiR3dB;F22l!>lL+b-<<w}N!8B``
zwQ=2Or)My*x%A6wuF*{e8)8fBXi%Cq=xYr+PeilQ(C?6F`cX6DV}1gkkzHcUM?B?(
z!CnGDagsX+>PZ;}Bb}s#p>T4zv^h&N<CCN-S<iZy+`b+M=ws-7%D<OqBA9{W8e?Y&
z)b+BNY_9BCu3o4vUarDd!QsjHs(AP#O+&2J!z}*mDw>>_toeUT%%+R~hSCa+4G9zT
zVQxXYDyYL`_Kjte5^Wr2U^3|vD=r8_rghzzQ;+G!KhP-4BC-IgwPl|#{Ba6CDwP@j
zMk<wiJA&|y{2iU#z`2_5HL=DUmyy808{)2$UJ}?0KtRgRBBDFfvoUB+mY0{J!}kVB
z?0t6Ib(-t3H)33m*mZ{M`ze{CCg#!&<;)|L6P!;rtfsMajxj7if7$-OnVieA8=&ip
zajcRN5_2!-zu@&$<6iG-9xQ<U4<-A)kO*?Z+Ot@aAbz$udBNa_jKL^@1gSZP*}?RU
za{ZL+ipnS-Vql}Op~|9~2UK&TR)=H82JZ|i{9~HZB42$rx<&BKR;7)cmXv>r8~;={
zPZLCtKw`&Qty0?v2^!>HYo$M`N*+`2Zi1D8=Noo>k@aVBWzY;Ht|HfH8cCc1QfnGg
zop7C#^F5%1c$oBN7m)u+!x1JiLS5-L!~V!nawYGsu(7z6w&+-o15WqoCFWGycPe1l
zUZi<?R53jb8-%9ipfi~?^E@+f@M~mfP(Yo)Ohl=OpR4{LHHCTgjmcsCM^y8>2ucOf
zq%4!^*(<W+BZIT+EamKi*1E;g1`->c{7)!veBmGqgymv(4ia<~xE@UG#NcfdA=Mzl
ztKP&gqN^r(<OZoj6(;DE;VWvQ_6UKrDn|GKZCDhLNDw%Pq9Bh_YiWZ}n&4p58Ss*b
z&smnk53?NyjGA<$i(jOPV_An{vO+OU3zZguc~E`vK4o;s)<{%LlYFdhGZ!xF*OC0|
zfJPSCDR;d;P&WvL?!&BOrQ0BIbgP3hP!@p(gTb17VqqZs5W5A*IdnG$kNZRbTB-pF
z;a=`0>p)(}zW7>(Hz$R6Opp6E@e;?!Zw*cs&lfKjr;5iOEl%>o1rxGop~AMSt7RwT
zr5{8wJJMPW2@|N%4Eqj9g*H30yof<mBoQ7}heT4Gv{4`a@2a%nqjf!z7dSNO+@Gi~
z1P4V%N&YEY$Vi*k%(P+L)ic&Bj$T?*&oKG`MO2^xWG@nb0ZR-YMVuO3fN=N#4>Rl5
z#=yO#1Z$wCw*QQRbxrL;J_k-fuv1w|CTOhYTraaX7hiGQbz6oG=-#EbYBcnp6QPMQ
zsPKCP#o?4vD2YVJOGoRjG>oY7(A38KZ617TJ&qEP2`hD#slJAdf(NaNw++(BUDGq-
zY7_o_@UxaP$fD@kdl3Q>?JpF>jRc*M17>UT=h@Na{&H_3X(0`0Pne5R3n=|EmY(<L
zsiAN~Tmfn}ubZ|BQfIxMps+4?j>*+K9>$maC+Wd30&)U(P-T)Yp1YX}a~M_6lqcE8
zB#V0O2+gu<P2C#ih*Xvqr_QrYflauWu`BtEF%u92ekn8Wm%--%t?#B%#|YTe{0x_H
z-RBqCLFa#m9u5?maUknwlyyQwx};o%#U;!Ev7SdC(QV>n?EBd*SR3ndbe$$cn0IvZ
ztal~sX%;AR=_a#Y#$M8t0tOJ>teMn}byKIO0fX&za+*AM(Gp4gHP=%ZZ;djMko~{i
zpE9fA+XalcGPg5G@<#xV5d+atO$UU~T+|fP;;j8E0)~vW;IR^-72vKR7wnp;widq;
z`waWTJ9B=raUJFSWZbBENQ>*aH<+~7;-8FtBF;u1U<xU5ONqZ6^Ux8bV%IS0tmaO|
zYL8!r2+-dL`7&7~qS1Cj+fo=!xC9%jHl)3Q@$mY5Z^S#(%s@0B!D&sd4%HvMYXqzD
z4HBLM+@l>Q9*hQ*4{Pnu*^ZN&>`b)14;nAP!a07zB>FBWO;k-0%2ST=NjIh<AHu>T
z(n#8#lfyjCf0@LUBL=;}-LvdVKzV6H-9}6_i`=oX{+y<cNr-L!%yYT1A*Wgm@fXQ)
zq|t&i=loYxB5Bj#|1}!rdrEM60^dN~3<)O8z#^vke$_?yikqalr;R22stPVCurBCZ
z_o7A3X;Hs2Qxp;hjWerveq%U2Y_mHOJgqC%(HEM#`WhbBz%g$S9BQU&28pa}Xp9Yx
z^QITb<<Bb+5k!LVWoOmQ=M-2!T41dnWLRts0e!?2v#H~_8m`rop~<Mjw%Ram%-4Z=
zgms#j=O`E8VOdZl;A~lQ5!VYgwiwu0A}~ZV4*RCYZ;ezNxYnWFszbcDMkk9=ROv^!
z8gq9*$*HQpc$rr;aHoFqB@=2O+N@*FYhiCyyWneO3~<X72)m<QMYf%OeTNK{g<8oL
zNz=94=d^mHK-F%ef|!|=QR|gWGQTCDW2EeqI<fKTBceOPe6=f#g!YaaXrY5_FcYg;
z<b+x97poCO4Z?1_P8#u&HLE2ITG&6Dkl6@0FAIQ3JQAyCud#Ch4g~_p?&L8vh|RH+
z0K9r1_(>S55J|4i*j3&`I|=`Dq25*RMs`)0fm87y4Ul;ur^M_;_?M@x^+G5W$YZ+>
zv~OXqAEOs`CYV&L8{TFhz|?}>lH#QX18nnY71%aN`9Da?kVndApe86e%+QOA64>oo
z?cD$<KivT+d#I;NcJkEPYqfTBmG*zWzGZEs&bzN|m2nVqw}k6CBm92)4fuZp(-?AG
zN2({F{34JJY~1I-WSDU(P-8uAWlq_&11Im)DtU|`>}l#gH6Kd3=jRvAa^W0{|8C`y
zl*zxWtX+*MMaGU5%*<(l-VrVg^k%i$uN~~R%|gLAnqe5axlRqPy+UOss~KtN%7o8t
z>svdI)&#3#qyDFOVvDp3lPGT;$@-$2E{tmy4Q=8>Pib1oPB!o*#f+#wp%TVjy{@zs
zT^d9^s@O+JV=Jikjn%NQIOXgN#`@aGhxj;pOZB9$5@A;)e3UJP15HM3N?YTkRW0W7
zy-HI*`>!dmUU^+H17XH=xn6ik3-v*ol7Uqc#KmAdp25(&H?b3{!NFk(A$t?4L^rI|
zPRzi2p(b{-qz4l$>pol7t$yto%nl$Y(H^rowi<WhJZGdPB6}`{p_+b?I<prhQP0d}
z=W-JCNJJTdyu>(zEXrssANJS|Ic*BlW;-_`g5LgD6nsp<hZNjR5T=dv3e%5!6ALdl
z1Y4>o%oycX3zOt%n^4wC!!RZI?O+3Lxk<X!%IdCeDwR|-1G%M5Su5MD3pUGtGc)er
zqlTPT@Q?!ag8yv_BwrXyoWn5Z4C@0bDn|4i4g82kO7^Yi#>Wo%6GXy7S-U*(>%rE~
z?sEC!^2EH`2x@!FYzKgUO_in<*etVm&w`h=@zizEW!=1`0KgOTlL{cvV}4daK|wbG
zyJWGzajFj!j!PD7bI?s^6A_qn;C;F=uE4k<(aip=f{O}7a=U5&NyR=(5cWiyS*I#&
zm2J0?`txd*dfq=ppl|Kd1i2U!4*~Lj+h6a$PMpWu<#ysh7k0-pz1Uekn(NMG_{;Il
z_vSO*q~~-GWmhJbE@TRQxjlA2of&1%T~98R$uignt!+~@tK*|S*QX|`yyXQ0{}qXR
zGM~z)`Ay_A#8O_#U@Sf{@Nbb4RKFgRFN;ccEqBAg-!88kp*N&JDNox5yboCk=Mke|
zzA)JG=*bI4tki8DFxbTCz2fcgcH&Qhy;0t7?(}(kyduB--d=AXzXRTW?*PAp-a+pW
zzeC=u=IbJkcME&AweCCZS#pmA7R1Yi-SNt}ZaU_?To6C*o!~iJqkAX4Q`f;1__&bY
z3;DQ^&x;ryE<S8tFXZ7uJ}>0qLLM*V;lg>m*zHZ2&x?3z56@0`lis^4MSN~N?!D)F
zx`ewY^@v)<y(v%rFVrXdn(0rdS3S@CVd&2O<{e6Kr(zy5bmu^HCt-C>d(V1jt*(PE
z-Fwa(w|6)cT~B!L_nx;LhnpoehG}ocJLkP*caC@;@LuL^#-cLc;eF8RxXZk5+zr+|
zANHLEI92P-t<x)o1-Z1b&D`Vueasl>?RW7r)uZVXeZ+H!4if!Sliu5WTS(ynonu}t
zkan_X()?oty%N`4h}X850W1cwaQl0LtiQ`kB3(_vsV6*1aOKSbahVq*rI+2A$ORB^
zVERg&*+?AYRpNPz4_Le~qdbH4^fht&Js!I(XX8Pygj~=i$a~qtiOjgMhjXR7UO-sJ
ze$BZaiIb2g;DO`iSQY4^KTke7dOwKaZmnv*JNl|f3)cF*p85c^3VRrxetm5S0l%IQ
zTZ8h6Vq1_sC=Cj&Ji-o8Ja9m_;eo^A+k!r`NdFRLwqs~JsG2<a3fUI)`=rk8sQsv!
z67Tf-%-nPrX}fJ7>YiZ0|CRbquV3kw^B1fyV{?1fiZ|45Xdq)32gaS$^%+LS-=|<d
zK{(77{u<k!K>%K3=X}}!tnPm=!5RP0Df)zh?@_>zqU5dNOaKp2cxYPm&-qX3#u<XK
zz2R15VE9~;hb!&Ac6Ea5Ok5idc5Y2jhDK$RS78{vFuAn2<iDZ{U)7Vk$bv1bc?k%G
z)f#RCjEj*cBJ-}*NQJs^?7^pLXDR1-lmHKwvhX?!xWZyLG@H;plqNO*b7XOw8-#3Z
z!^1iUSS_~$;Yhn@tX4C^(f0JnQ<BNi!@*WA$pt0tH05lmcXDzO)2_t}m8wKA;gD?d
zG37(D*4*ds8?w8gc9j(jD=->iUa>0#V?#!4l?L0rBSfz&nPj6&q9n+s8V3F94PV-K
z|BDK4C|FbQqY7+<tzUjnm!DT~R)O)7%}LQVCk*i&RGSR@Y0NH&%=iu>!blmo;9Nrc
z!)1Ws{zgdEUqXAlP5Q&WZv^1_b9XYRl2ulF#)UgmY~^$N#;NC<E2HZ`Q`5~w2l+-b
zVlr^j+^xMpV0#VVqfHi`&=My_CIbMYGA2!Hm@%zooa-J30OqncShynfM8UJXQd+ii
z*^eb2af>SD78OkC2YG<6xAw)o+cH7_W;5&KN-$-;C~LV7VQ^+W%zF{K$)5<@aLiMg
z4Gh%OAqt?4_crL-oU1f9xRI;0A5(7&;_gQAJ{(`!x4v)8r9?-ODG|1fk9tVOlYDEx
zLcseW&>;i}0(!Nbg}x&8A?e)teU+4CtxSkM0keo*Y<Fi<Jloxoh#7dw;3kPp_v8?0
zMjMJE3uYACd@1d^f)p{y-@0W!vDQQH0*F7XdvbP^K=IXg1JHVdjdKEp)GOs)s$n(X
zcuoVft^Xavh>4DN8{Ssl4Uv5w<5sy;!)CcSwagBf1v#AohBd6kfjq}rDQIzk6%*A8
zGX43hD?9e@-(P&`MY))H@!XTga5(eC<7ah6x<QeZT@BASa?f_W9XQI=eQdzja>B=n
zbJ*hqyiOkbV_IFd;%uE8;A_g}{U+NM4myu{I0k9DLJ3DDx@b#idOVRMwfY2`ZvQO3
z5VB_&FL$^yN%EE`GBmqdvM@Zl;cP}$<%mlp+!b3E4!FnpEwj2XzT$T-cD$Q(3X=?q
z{iNpOD33UT-DD0~m`t#=`mw|&KKutO2iDKf&8oR0<I%BCSCco#*Cy|SM4X2%rk+Gj
zW`lSG7J<p5smla2x+Cec=U&!WZP+mpE#$PrJy)sAR%5uOc3y5%s)o#(9#;IHq?a0f
z`Lr^>SApTy_bK+Qg0l)bJI$D#Us4|xxH*#S#>4~x)9sbL>w9J64rNL;Kcd?8)&|@p
z0d5H3f`r7{ys7g>mMBZZl9195xzLakg`*lwvVmzIOU?b#SLMvbL1aPhS8|i|xEx5D
z>%Xv1W_aS;a1{wCxpd|n?=Dg=Ov!m+IM6-|%`(iq0s2p9cK;jl`6YseWq~{`3Jn=P
z`*q#@j|vQnv{1<pQTFevqx!iiNt}>`HVfnusc)<tUf*5qc(|=K9a2kmm^#rqZIDQ-
z)95gSYz=eOgnKz8hxY!4&>OPVU(`Fvj%Iy>e}uH(RD%Ux)!>yQ@6^clx;t9gA$Ist
z>S6$6wMQjR0QwHpu`7y*J%kSBq*~2AnS37Lo`&1Gpi4Uka7JiaE}0xkM#~h*+7wBe
z$S{cm6Ck=2TuhPlY7%IhEO%YJKdL!?Az6FUh)O*rS$m(6qRqM*Z2;<qgGNlYb$_|M
za_}YFlVgeO7{KKdhX1w2;zX?|tmKl8`#(VMt>n)&wm7crPzGFqbx@{}bImhozl?G*
znE4ADi%I&)|FD7y1)7`w&+EpRVm1%IqL}94&B%JmM$#tXX7lfdDg941ockyuGR19E
z(PoldQnS$%SNIx`{*%Ez^iG5A@ULFpFHvs?1{MGlAhTI{o(K-?WMC~LT?9ysNk7;!
z<#Yf+dQGeo7}Oa%imkYqt8<@`&D1mCU18T0Bg@`NgveQ9S@5!AdDxx|!d~J*mL;B%
zQ0o8Wi6~hHBf5PtTIZ%6R0q}-EE9|s%nH%$3d2D+ecRDb3tKr7T(pn2rA|y;@U`-s
zv_3LOH1VDuUE0-nFy6o1YBM+!8IqU?u#{Nwd<oHgvvThQ6ybK0%C|{m9T_c2{_kr1
zB!-n45@!4eQ{+)Xe<szxYyFtrxvGvOs@KsnwIO?wh=!{Ue~TX)_8vpfB&tJ}hJKn<
zctTPN5H4}I2(cGRXnC5YQC2o>3<o=bcWl7B+9BgiyH>UwZraqC`!MIV0|&qb`EjPB
z@rvT>ZJo5sXwNU}^5+O}Ei4~(A~hP2>}S=ZUsllBqyAqd?T^)U5*E77%EX5gh2%zv
zb|(^~qr0f8g{)S%5JWR(f6NI}B^vPenfpw{`nYV`M-B|8#Y!V^MR{M(kfTWF0=Yzs
z%f1ftFT1ZRSC8twmZ+UJ2qkrh#ImFq+_O^VX4Z8wR_S7T!YrQ7Adq~6AyDUROA=q{
z{{!{R-&0_HdRU#9<$aa6L8Qm0*7cnu%1pouKc{8i@Bo%%(%4)qn5tV4B$X*(7&pk~
z>C`63LS4IAMH|%KX<BA8@dHF;BMN>;8lmpNmn}AK*#FB0UDj?G%-DUD+iRq_R6t6U
zQ-FsZ0?-54QaFED*j5-d;7Dqb8qRaX<Ma#|-Bx)9^Do=AcxkqB+0+SmFU0kAv_UI=
zK8#=X_o}E-mq{Gbt?#!SKAO+V;dQ_INieYpLt{ev4eVYcYe!G0NV6wi<3evLq^TWw
z3TKeU@Xn;To?rsWW}Kb~l6dHFZw}CuJ;|FOLD{3IH|WjOC@fA}dPeDd>6;*9J%xWS
zawt?e2VQe4zf>qjpJ><K`G<?=%@yYQ8<!tu7pqr4-%4-nrgPr50mpd%1*B<_i>WXT
zNO5w}hi*UGELFUWx9#(6gsrwRzl%M46ZI&o?Ty7vklikF>-c@T>1vyj8z8?~S(}yK
z_tT%!gT04A4|5kPl_l);YHmu{M98%1kN%<7%x@~VuHdT#4LkIIq=a85kf!~z6twaS
z$~y6@_yGJ3-RRa^GV!PDxVl3UtSgsGVw+OF;^8#X_DOa^E$Mk-+RCm=wrUm$d=;}&
z3ym_DSYtca!~6Ck9r|0%Cb7k~uGWY!TMQ+`&|-_k+H&@$(%qU6fz9w(DZ#ojgEXt~
zJeFkT;TMi;*3&TeG5>|(ZCI*qL-}B}a+g}OR;$^edg-WE=8`N_gRE-ZP><x7S*%NW
zDl0TQUg|R^nhbvAh^mn=aV1Ky5iJ$!vz79sj~8;kG}J7jM7KY?kw~#tA_V_G*5KKC
zPpMRCu(mXhuS(OaHZ~U9*b6lpHLzjW!^GLEe6N@;GR(W@!GxE(&PETHsBqoD_6QO-
zvmVsQ?&Jv)D#`oPA+`ZaSR@K9_4X)RhZ*Tz>qoky4(o@G@fJ6rF}Lv^<`Kw(W3Qy#
z{wSo*=<}qf2bP#Q<#zBlx<0vr0E$NG{6^`BIL!3>5(Yjp0lx%#BA@s!%pg)YiooL_
zkR^wDxLXqG)gDVp<82(I2XaK@A+0_VYj+=Mu#Cr@t$>~=Z9F-%H_~f4e(_)DTtQrm
zud(9@$ezQeT?)B)Uv#~$>~L&abFuHU@oKNNHs$|G)Y?D!Q?%BIQ(D2ij#@<w{fB!;
ztqR>Sb4<m!qxT+I|CUicy>ZsXrxU#Mfb~iHJG<xTy%qb`@2*_z>tWSx?1K$;HTt6Z
zN$UYQad_R0pkn_Oy^-GD=nY~Wy`ej<H{6{&^@j2_dc$&V+8fGc^Da>ATla=?<&}4n
z-q=uAvo{j;F6h%<TxKwfrrEGCmS6c;6FW5ap+5}rm=TpKcf45a)Q%RXjuy`!EnXgL
z<!vG}%ghN2OXgDA6=9EfR0@cMD_`5H^P!Wk9>r}#>-yBIt@>KD2!{i7wdD$P-t4?{
zox!OZ9wy89i`ABTuM&2#)sRi_ZX?i-vDTdput6Ib$`hnG0}3OY-{3OLyi}dW#X*=l
zL+FQse|IUH$>mJWYw%9=_1K^jtck$!|8E8Vk|0bhPlpMQ?_>>+w^M!-0s3Zs|If%%
z>e4eJ02*eNqem*y|G!j6PlE)o7b|tQWk0<hR`v<bSq4|neS((aP<qp&N%NuyA5=^n
zz!>6yeClyEnn+<Rk$L-g@@Qgb{4NOn%~Df{s#f-F236ahHYJ0s#;2hQqy@3s7kQDE
za1L|_FKSA!0Ss|b$2k+GAP-^%odTGoT*JE#F$|e7OhS5W_S-T>!o0%D{s8fO?FU%2
zQXJmHRRK;}Moe87@Ml80ya+uGItHPxc2~D(d6V*9*s<<vZ?ioZ6E&aPUB#uU9GcE`
zBk@l8%Q)okuJ^)KR4mp9A98^f^~(SXH=lBSO1x3t(h@;8o@U(peHW_7#{1#;e{R74
zeo73Ab3qu_iyt1G@av6eVLuBEH=(FSQ$gQslG=w*c~aimGvNPJXG!_%@zS5itcu0m
z74*y`R|nW@^tZ`{b5&Y07!0g!;ohJ~U&Np}|0^y<hF@G~{I7C7;!L@M0T#Hnnc@^`
z0{h^_Y6&;p(3kkHW}jeLTB|V0F*CSp=`t;F89u2p52Y`ECrAj(SKJTo%KiVyc>BLc
zFt*i6R86SelxO>YMfb&L`d?CDQW(=deO{NEEzbAy4aJNJ`xMj28G`(p(#~^P+9DIw
zWgO(N=Am72T{1h(!8cnl(`e(o%vU|;8oxq09gag*&+U$MDl{hy|6M8!w?|e=?IJ7(
zHn1ynf{H08W|!x2>2c6FHaAu_y?yJB|64q+#FkAt#V9+uy5D5EapFJa|EgZ3*;A#|
zlqySVM|A268=}{Xj@5;;TF<i5?ZrG=>NE-02T_-4=ZKS%X0)P_a>zx>Ffmi5VIU=Y
zANFG=*_%ljvGr1~ImMh}P)*x5Ti-_Cbx5p4-|71V->-j|fbFmQ;mu<=qJ437w(4<>
z-VA~OyH5$*;o4!C<}!0=hVbPiSIhF>gdO#@gpNO1#m1Q&1sn^@p^D{3SHw)kS4MXH
zEj`e2B#+Y%V{r$tpH~;k3gKqAF(*xabOt8J->$m{6-Y=J<|4^br8FooDQT0=%V9w4
z{#BDvg$0H#+Rb6I5~)&@uOTy(xTMtGa2qit9o;Y6L)1E>TaPHFsrG{!D(#@e_6z_k
z-)X(XHfrcLiqh`b+x{&*>`xSYQ$YgJ%6b&Ab7OsFc`Y(is-I{!iuusisH5vg{+%kh
zvv5Zo7lUY=cXOXA;6UAP0+pLiySdp`7=sBvtM*A=1h(0r%G{^}+f)MZEX8Zmasdd%
zv)OhS_U7XcLA;mK3t@4aa(ud(viUF5!UntY5^=f7_bOLR^BpE;>aNAMVr_#=Cy!zl
z=YL&o7t_d|Ev=><>-)5QdrX>qjZYU5F(0BRYX6x)T_}--v4pS+BKX07lI=sZ3lFkF
zU<)l@ayx9vJLb%>9_8#jXN*HspJ~ChFAAS|tUl=ga&t1liTg8<s{Gc@M5ZJKV^q6j
z#DY1aFyjm%SZAAXw6#fCdaf5nw?IqShb8ST#>Fx$R*3<z#N3qx8p-rbHb_Ui&N9)i
zGi(h_wnW#*Y;TtBKQqf*NUKjM)#$R{QOszy7BaCF{;Z~03y&D$Fii1ty7A)*?on_;
z!Keb+|21~JWVDcscGD2!@`?ij+ypqQtg?S#dm?`G!EyC+ePYqJwXq9qy#UY*3V%s6
zK~lp*&{OyzHvUw`4RT9$AilD7{mh6AL^|A1NGH&MvgQJ$4hvyI=1g_O;uwwB=yRMM
zsE@@^N<%c_B`}_n^8(y|%trUr|FBlnV5}7atRRDF6bEYssoE*D0&J{8rYb=Kd*k-0
zllLenD&Xa{iDDfZZuT`jB7aW}b*P!|{t-pOl<;6!h)|$l<$^kMNI~cN3A;F9qH&bW
zud4TC@q+OOygj04Fx#*l0l-F}4XdiH<OaB@RrUnm&0O$+WdaQ1=?VHy=DONmzN4W7
z^uI~5hP~}=;x^!=E`k2?736UK2BX_i?FE%NHVXb|JAupi&l`Y2zAn)9ZqN(fsbB}7
z6@+Mtq@>3(MiEw9yDm)94V*yC$w>2Nab|<tanaiR7J)!i5ctMKoGC#kK(G!ZkuSxy
zyqCEVR}YM3C%%nADiov7(~)~Jj&#AF{Ih3qG3GCrZ+ST_Gxu5U&=t0#Bh@=LDZ7@V
z@)V|H?mWDYHLIGOt=EpBh~Qkug_f%?I~K?)otmf~Mp(%KmeC$H>Om7YTX%;Xd7P<O
z_2hoc9G^w%l2%0xt`u9cZ|NP)Z3t*9+u%|sincpOOznrpU)B4f<X+}jz9Yz0^#V9A
zlm9bbBg%<qMrmVOb+tHD2ge?<*~GasxVAY}%3^U&ubpz&ryJL&%@HtqHiK^l?!Kg%
zXVA|W&s)0OqF_LQ%<ygNKm&9Ap%U)a`qBa{WiJA<<V@!`bk)a|Ox+vq&Purp^UOSk
zbp2{9xa9KWlJ~DPcMB9R<aPqS#2T(_Uk|=o**g%jpqFN?t-$L<R{BjS4`vX3sGZ4g
zBBB=-0Rts5wj7CLl!3*M2ng9@K`q{Z0^BV32jz7$sSPjuq7lCU9q2?XC^^>Mso{eG
zpRaI1%<bfa_>Sg;e9E0P*%)H;DLGL*b?nI^rh-dXi*|U9&Goiv0cv2I<wk#cWOg3;
z)|GEfjA(*3j$^1jT%2fg*ca8yf<eTuwdJ19MRXwMhKeSAOn0Vq?2CENQ+4xl+2A{G
zdYZ$}BJ5{&>V(%?<4DSio4CzhXsq=0zs-_GPJ-f?%BWeKj5LFuLAuGt;hJqNYpo+#
zElwRW>5$wG6+h&*ok{L<w0P>M^H<0rU_Lqx-qv1$V=QW}&29bbR?+(zI54(W1|BSQ
zKB<jSo90=FopT6krtGxk@zZ2(ZQd6Dx5Qp{AiJ~;X0a3Yt?R@vi7nXrH8EV?<V@24
zSF59+CnWMk5Y?fCSh2y|g~UcHr_G9Y$V-HDwOXh$R{Kuq5v>AtsA8tT*5{lVx1+i;
zpEA;I;z<S!&#^VoSdp|u7$1d!Xfd2faaDl6ORjdAU4X3dkb7VSfOleK2HnUS5dO)l
zuB0PTdM}6B6ikWLho@c?5VQi*X_p4=Dal9=zr99-W|}R7zO<J08#cwgMbK|88wdt4
z!lo|UnD%~T9`dQH4P|{$pmJc=!xhKkfQhyc8{jy?-e3rY8;8tj+t}&|kRXLGU>Q+2
z-c480yHNuN&2WFj>oONyTUSR{w*}j*-CeReY_u7gc^d~MN;<MN*oN9Je*?I&x;@yg
zV-r^gFV1e*oc4FxH>b%2d~O8(bdZ|lReCGO!Ekg-6yIj?owy>AdG>by-$X7#b|VSG
z?Fgb|$*#7}ZJXOMw{ul)%6CDP?KCQWcd!Q^#C>ndKW}k$G}yJaC)kaVfSiLAXPvaB
z80<3lQh%B{_O{g__aS>_3?9|9*VXes!i7i{2=82oY_OzgIJ#)t037)sGMe7+(dv7E
zscx8Zib%&ti@=O{jw#<axO!)2Cc}-E`1@2qtOOzgv&ZdIf<y$)4CqNNA!Y-9N}M+a
ziHdyKdDN--=aqMYfCK*oQ|BkH>a;jhbooz_zLGuRIdibRjvx68Dq((-jJxO+9}!oY
zfDb31p(wMXx>_K}Hx_0c_wt(B_i+`;>##KE%&A-TnDN0^6+5cna|(=|{u_$@bp>Y#
zOdNHoEa6ec>B^junqJ0#TEQPv``A7wr8aiQh=p!7Sj2*_{Xc$QK~@217)dEXDf0Ce
ziO<a=T&d6be~Z#MkP@XLNBcQ5fG+dnZ2P}!rM|Wc?Kid^m!*tVbGkIrQI9!XG4ATH
zp8kF6kCgdENYrb7TFn~P4UsecF14-Om3Mo8b=Fy=k1R>r->@h$zXa@;Gfyr4HdP}Y
zzLar8=^*_{ReVGhi!V$tpK2fAs(=}5KV&<`tAKXd9~P1aj1WpC`qQJxP;_qu4YMr6
z5`Vn&-}mm7TrM?)j4kC1-wU_#;9`fE$@%ao@r@?zjrC%C1g1oAH9ieF`V%D7evbh6
zMuPAXZ9vmV$_?zqb)fO?`?&*_z%F~lxc8hm1E~t(z7x|pN(%DE0d}#eb}Z-u-gKFC
z*~<72P+IEiBwKxjS9+>~Sy&aqS3VYWag8^XYdQ)x<-afJLA{NlQa0MilD~tTvOfbm
zAZ%f)FJZ4F+@$AFB1!p~H3;3m3Qpkp7m#-k0J*!RTw>c#ExPWpYrAJoDRc7i)}vuF
zze5H(9-|}tUX7V8Z7p&PdTja06Oc`e*=hEaU|UdeEAr=5q+dmB{oGcppmLEIxXJ#z
z^&ps}%^4=YZORH|w?EL3Beo8##4VST^Pc3?aXnHt4L}$6$LJuyC|Y(daY`x8Qs-Dm
z9zftI98;`$%`xqcxO7wfi)!>1HMNst_92yH-=kT2lpHA?l!uHL&wI?S0#~+`0_m^p
z+%!wI{X$<zlMPy&b<Qw-F)JOry0(%aA(c%*J$O@(m<+iMHl%%Ie{Ft6nS9+$(LCyn
zjK7ytfI+%UpqzhK(2s_FfKr32!__obC@^$YHyt=Tq0;z!!x!eNkbRN^6rQS^0Ga9J
z(0^ie3%*~l)xo=F-n?rTP=mxd4_UhL82h;J3)A$jIl}W7nQ+>H1q#>E4=~D{6}L4O
z3~@$Zn$wKqhGKL6NAHkt*zRvy2dqV5oH@A3H1ZE<@*Y%hNP#)ckl%`x{ukY;y+se5
zD5sVkx6b!JqyZNabA04hUH-T(A18*BNMjI<Bh+YwgHKifZdhDF;=j!4n=+!<q3p&(
zzM<Gd%9qts0E|xlwN15}GX;~AzpbjbwrMecRrd`wXeszRblLeLVOPDf02e&L+^<Ma
ziJR(j;~CADbbIU#W~<3KyD>#<%}sEDaqQ^`Y)Kp1*{!s7^Idy+zohxDI<^B@-KJW!
z8Q(~yvH4*3Gw+%GgSZn-B@Qs#ccKZ|4AMq!9K}L6KI$#)V@&d`35o>$H7OFMlmh^+
z4&avnk*L2r`W<{DR+Uo=iC>7%$U719TKm5_;n~tt-1%e-WuBCir{wI9RRLjcK1zpB
zGxn|yVd^`Ka~IY8YCZEt9Ip-u|Cvw7!Hn78wH(Y;hwVNhKmTQ81pvH_yVbPnc*6+!
zPsXF&gwrs=0|DwvQwrAWd7Vay1<HW=Ld|l<Hiw#2GwrDm;N3clvfcx@|9Fr`lr+HV
z%LiL9LS9)N;T?K|LCAJU^0_oueXN+&s!m+&_rFXyFgM6CCCS>2NN_w9Y=yb!4JEdQ
zB;|OGH{!u)J-IgW21Kn0cG?R;8w`=IdPYHKv>_P5(nEF)-Coa|1!EevMGDn_>kwV-
zH{CN(aOQ%oD3l@N?Uc%1o#`12nGd}oTG}7^!1#~E2fRM>fbsjm@ER<FDM3^IEyQ6X
z*C5)03`DvtyC`=Jq;~NQVz%PP`rDkMO)(RjAEt_NZNk!DKSKb9jlH@@Hmda)FU16r
z%(g}+DBc*RYPTF(wRk#rS3~0*^I#0z@fbZ)bJHH<v@DIP8lEq;Ua;pD_~GYne%ul=
zn&uRjzs2hLy4u?vTY1lV5sq9^!S&WS8#}^k0FgH0z3FCxH0K}mX3|N7tJH|AiHRkD
zUXPBPWf)y)OW2;j$gO{?=Zmt*#1ir1ZJpG&^3Z>CJ36{YZ<Xe2-7`wWPCLnI6dK(W
zop@rbgmsnkcI1?>_fSDwZ;9~xoknks#{TPD**s+X*2wd%jx9c2!OYR1w0QwGUM0+&
zzdDJ5im^r7S7X4og)5G!X>!uj!Wxd`F&<~T@6lE#A=E%Ph`(QhZ|I>Vv%jvioU$BH
z?6(#B`-+(^M68F?l$yF^PM4;SH^qrDE5_fPPziIZV9k7;d}BKt&ae?T#Pm3+&X9Ui
z#{YVkup29(vTUoB#4!!>4s8H`v-i{F4LSedT$nXEZuy*hP1|3!Q%x}b;E1y2bm+<i
zbe8i&wog4(ASe~CoRm>beO4|`^>nG{-5@`pOq!SBmQCmVR~dU#xunwCkgBsK9sQ4`
zHr6uIvkU*^ZXZc%KS3ffcw1hAbu5<7Tj)-ri^rw4JSxe3S%EU<5H()-MlLPx0E<-0
zIgLErz`;M&Dwq<R+4zVg(`Y`y*Q{%To(N&d$^?n&YHx&|`Y>fPH+m5YV$(fRm3|PA
zOyl~|9&wDX<v2)5jztFO3LzawPe5EdN1)Xosk4)$o(F+6oN2SG#1KeZDpbAGAifI0
zMjf^13`iH`R@)ggnj|TT!=t=v_Xj~Qv#Ue2ugj!DLqRX`Zn%Zis!53aPtmeo&`l2H
zayddS{lPHE1<j&7V?iQH4ipczfLv?~KBy^aRs_Rv6;9Fwsz9QY=B-^Te*(!8Tla%u
z(8mbq0a8<L0ze=AhCccM|2+}<_-n+oUbmr-Uofkg{>X;rmxVA4eRTO>0cBCjkVhF?
zoUe+bF_N(#^0223iTHapjenETYxzNSvWs_Vaz4gg^yV#CU|fzZ`K@j(d`?AejT@0?
zOWqU(n=N!(mj69#JfaI*Yy6O!|G1j^xEk`PVzxXDa=qK$K{p;#>^%w$P+5aJSFrzt
z?uys<n~T?fQfY=hEHE4*M=sek(|w%0sf@GuwPV3KC|njWT8q0`gDE@hwaB%$nzz;<
zYxHfM1KU`A`1XEH{L*Ij!W}-&&A-i*IH1NyC|K9}*3_{!;bT7jtC)!Tn%%*4YPAs2
zclXs3(06Rf^7lsc9V~Z)g8O6~`cC(?Tj+$51<v7MT%N4f(RVCT$b|w-y6tc^m3kgk
zUY;qI=_DKI#U^#)pXBPD=nDUoI=zV?efdH2*I`D{5l=7%{=M8>xo17xa+LKuk`4ZS
zs#c_eZ(z7JzM>J{qq-kd-5Vr|cdq-6q<}9EQEhMZurAlVk#PM}dg{0R>7Lp|?b3Ja
zEn16UllhONafSytpYttbY^YlE8N<^<x^D~1Xtr-%r{n&f)3dGfn%K-Fgb(vBwVfJw
zZNP5BRC2f=+b?lUD<hkaT!#Tdfl4xbpXFOSPXIvMCv~~D8b&i*98F?cdlWJVj>O?$
z-ZhOOf+sufv_wr%@AA2dJ-Pux1WfSkNC?=`kotUbvpmdVgDPw@POk46y)nugy<l6<
zN;Hbzw6iNYRSxx4SWxJMk!Hui4iif-U;n=A%**ObkrV4+4yV3ZHt9{zs$=)@T!9P;
z0zvAad=S326=n<Hl%EQVCpT(HTh(-ENY%ai8W#T@B6N#z@5uniVRW^#_vnMUWV2wD
z0tl(0-OzCr01Na8Xh&q3ld2wJryM#}Qtv|FsUbppD%+4|4zXBR0bv9C1TnNa3<SBF
zmI782yXm^*$pMcJ7)#kxrt95UzmjHJoB=jQv1ve7wHit}1#CP`sW{zzK_2jxXZ?M7
zg8)q7p}}a%-&L0v3D};v=q;}Fo{~z?xr8L*{~2!KRn8WUK&>+X{gh(ArlMa_&=RG8
zQJ0h44R_1HuVe3Uhe@bAo%Xj?j|o7a_oNiC`J+|Hp>`&Pmw~JQN2>K#6&RAz40O}P
zu+sYg8zoH*6=rWbD^B#HByDD&X|gzBrhe@~N{-wH+Z<oN`o2YLKca_!gW#>xJY=e$
zpz%3IL<`jr(a+M3pwBnKMU@SyP<FD7)|r?I=YG<pySAx~C8|Z3K<AW}PHuI}>d5L=
zX3&;7e4xO7JC~HLYv+=33j&Z7E*VGoWu!HGB5UlQLW79?rl*~nY;2k<pf?<5{%sAm
zk~jNEta^^rZZF5|TVyurIarX{FDHUdOKiyfD1|#-7uv~gsJK=QvhhYFKn*{zvUBq~
zyxw=J0ULvX6f^Vc_d83255}tb%Q5tmTiJE`#YRgxnN-)Zw;n*aAm^5SylXe&fuZ^!
zjyFdGEVBdWcoUr>t>QdTsMrTh41t&fs*xS}wSS0Ef_q-S(T#stAE`@cxLx+w9=6)v
zAa5uA<2G__M^m{2J>+)s?38fA^oqNJZPc;D8zSe<V04XR4bTnlykR=Qi+_{6#bA`Y
zqxHSPs5cC?zKhr{Z;QqD1-pV#H=f=IRR6AE7w@u_chM%7??xapdM*ALbSgZ5Kib#?
z(s;HsdfVPi1Gaas?&;8o=1+0-QSEfQPSjeno9A;*AvJHO<^naJCH+9KhxEO{Ueb4v
zp3@WS;xP`@4`0(sw~ebK?uypzq%{b;*R>|@?J~{i=IvS8*q%mzHrutE=l6Jf?0nXL
zLfgi8(_X0leLP9?w`ls=9}V^)@*4K`GWYg<UUzXkM~k4XJ6nV{W1n~6%`APifAxU%
z!G4}~FgVD&a@MSO=yRzrBp%@~X%7>=l=BMhdhgdeTATlD!v6%qXSd??s}-SNRzxzH
zNUk1QJ-m8^QND{79}bSNi7=Wgt7G2bU<^xY#D2kH?+9z95FCbsAx6rPDCI~aWjaY8
zMENN-Hm#GD$9TiLctZrhZEvWfFEqc+<t%0Hwm03qdSJu*Bx(<JypQJQ?nVony?<0s
zre}`fHEB;UHtL>x{J-nD%YTaJp5VD%^zhH~mM3`{k`CVTZnHW#O6-`v(-PyJLNAqA
zy=(REP2TV)I-fLmAh=uZ0$EK*gQL9Fy*wq$Q~p`~_}Ym&!r--&Hz1KXlR6QtJRkx0
zzuI(Cd!KhdPHM9v?IS0(4~*j%-v7G5{O>FHLj}L3fN#u{)nRX^vf7V>c*1U(ST?qp
zBYXSzD)S!{{Fn+pK=9T!0wjXwp-Koh>^i@^Fu{RAwtZtu$F7A&RZjV^s)~ezozA|d
zSY5%gf{zk}xfewsAuGI@ZX?ys&^G^RH89DB`Ia-+l^qSQy2WpoHyPX3{A9%x2ObEx
zMFux(@IRos#8sAt9Tmk!latOFjf$Hv>{btZ%4*uiUp+SAagxYXnI9PFqI%&wD17Ts
zqoK|O%672KPS0YAIPF&|#hGQ?mb16|X7-Ma2zPc86pbS281aTz+TMA$ePF`DM$4zB
zXI%F3t>p33(@xtX%zk{Ozu7g<RF`nISckz_*B{mu>elG&6?ICpYRnl`O1dAh-yF&k
zIWR81AAO5%m4+J4kOn5od}~WHUz@11#$0k%h#Kif^vJ)h;qGgfm%WqsWB<EUuth7)
z>+<zxnvqUo`2I5_i+_#CMxsczitpg=_>}V$EAn}r<72n)+?qe9yC%pqQgKq3S~zZt
zk+}FU_iW|rQ@)P?QFN8xr?gCcvFy)GI&$;3m29FF6Jdx}bYj*oDb}Js59;kqKzfg2
zuj}rwDE4<0Bz39%asGSt{J*O-qianWEt$11nVv6t(wT@nrC5u5ltGFAF%|n4#Kw*}
zYV{>0d|bgN6dYAx3lHaqusvFRZz#=#u&9=8*u#M(=0-z%ZSK#Jgeqy>@Ycd1mveS{
zdRYd9me>CfExY4taA2heUu))6A8j}qLgQ9{jdq7Q-B8<{>k*@?O=bFD)J&tb4X|Wa
zoDHwjKuRqbl4KSl_$+I-M=m3rQ|_LX_7@8CoC<8bTg|-2uoH7C>}t$^mZ5LT_MT?0
zveHa4^^eq$UsqsaEQy*qCZi?BuSExV$hWmkaGTaPqva~N#*D80dtS9SgHTH1;pngb
z8~tf>U<GqvtO2aTalXoN)KXH31Oh5;WXC}>LM@q(v^-gB`&NPabTW+9OdOT|?cKKv
zY<hPaz(%k_%y4#w?j+*fZ|}Z6NGrsO$V|m;`ciue-z=mD<GcS$Cch6M*=b`^lt;A<
zS_|als69n*;*P^pt-X3;LQWOTMaKq<;MUsh5YA-j7OfxScO_{81GCr5d2i8vF}Pfb
z6xOF{e?!VY4R?Z#$y^r8xcC>)m4VL_RyOC#O8qLHnOv<DmCeg^Ao81)z!I1pq9(nS
zRd5uq<M<WzCR~Vd!>Q;RYZ$xMll%M?^RLx^DOM{>3SA!{71yh{guoI;?WqsK0cYlh
zbP$bP(|Y;OYk!6vh|5_^?tn;mWQjsB7g#ySaoO^Q6<NCjyv-|?p<!%zor~|iD|?%6
zc%2KZ_WX49CCr5XtS0Ou1S`cCE9SVqxV%(6Yr49#FBEH$zepdp8h1~5j*)SCJb$mi
z%=3%|TR-#7Y^Y4+3`E}D@*qf#pRZ0`9UJyPrN@}~%0TN^DCgf$T3N9VE0|C)so-M@
zURUrT1$Qg>Ze@N`!ICa7Dv;XFrx`$pMgt^WQ|hF0ewv_EaGg=c#>@YV60+K)qpM4L
z16zp}*h>5?IsQN^Q5>^#c&;_2^9Ki+$8Phae8#)YkoYJ~YF(thiLX&s%trfN@HtNF
zIz6g)?9Wj9g^X-&t8?6rNH@DBm8p%^6Kj~ixV;(h7&wEt!FgQT<A@ivOdU>y!$>uI
z*_{D>iTMgwICPL{%dqO}x}ngncW_P&lr;BEo+A6dc}O`t%8KY~SU~Pp*W6Ts6bMKg
zP1|R<&_)B+rhPlrin70<>}D?_Pkon8dxlS?;=SUhHtFg~(uR0*cX$<GVmDnZ`-_cI
zMu%3%IJ)LO9&+Z~__(#}%o+cCXbALLi61$X*L-yEkU4W^oWHRInZ8%871$U{pCK7*
z8-7SP<;WwXlhk{T>`}c&UU9)^tn17;kJBxca~|tSQ33`vO$p06|5?4FRd(*&nRCv^
z_4v85qi$Y*pPKOJ3Bqg{N5WI=D}GgN{Yj-46tjN)3%dMk3e1({4=MH|3Vv9DA^*Rs
z*sm%0O$Gm0L8#y@1s_mg)I(D3vHjBDp|o)YX6M$Y8-of)707JS|BDJn6qu6MNC?T9
z9bsUUhY=V;ROZv(*Xa3wNtdk^WIiix86hT#ohL`_CwR4745X52ptHtr@9*axNuVbM
zJ@qHzy>I98ayehvf}8p-V%c18p+9r4a`)fP+>2bSD^oy5mc&(kZ*Cx$;{I^#qbke&
zJ+9X7w==sN^>=?W^FUktbTh{Q2&o6pFQcuyv5#_i)b7E!uXaz-<c<&Ez&^E;{4`wM
zw`?#<Nx5ICAhWEb5Mgr)6RzCT<`P}|;`071#X7{dqUnUr^cx-idd-aXK}^oql)Ovf
zq)`@a=C3;qf?=uK5>J9gb_k8g(c)NY9R9}OtTfi;CO>mV!qR;_!2EUlrX4)M<*kCV
z8xX(azswK;F<((On~r~(_s4hXy86mukeqKz94j?QI_I!cL~`UA4uAXHG@sk%#2dKh
zXlR=kEf}jO?bgEAyw_H8Q3&i0+SbHYFZ=AytN2X^&Eu_Ig&pp9L>D(<=R9qvgOF@s
zIMsu61sEqcUCz;w-O)bB19B!4#F5w}w?_NSQ9~#<4)r>6JL^>_h}u1Zz;f-v{hVpQ
zwUff?M>nk9s^(0|o}r^czjL!^wAW#!dESQ7SxR4b(%0Io(F&WicJrDjrQHy%mgK2x
zEvoA?uCCTwwA#O+T!C^w!`pIbogG`+cu&bxt#&!b=bTye(l)^i=sr&G+f>8m_E#Ae
zH{<?{CRoBH>{Y5}oUeIk5(@Gv6%O8bt=DwZ;NVwP+edZzbGq*if_Em8f26{D6ciPR
zN%#L)*~J5diK$I@)*jS!dy@R-!EsmNw<Kv#Z7+OHXdH{|cxt%t|0s=Ef8jq#*W1mM
zOm!9hP$W08Y~hyaj8mDnIn}AL2pSVL+K7vq6tuvpmuguP<h!M1#E(JYJ3>tXid|}{
z=5!8ERXei|cyQVAX+k9J=WIYEDbKv*=bd#J(CB&r$RubKE2S!F?bh}N4Qk~%fCF>A
zQr4w%$U{%Q#~bqcy@BgGI)XhzsaQ#B;o&e+_`S4b!iwg5<BaQt@v#JF(kxz%1QT*X
z(dSG#=Q(eDUM?=0M7$uKUP@3;oO11c0y$D)5v4GFGmW2#*%E%QrYZ}PMQX&hAfJ2-
z#>yxq8m@LTZ9q$OWVof>ldrYP+@@aF^!OM-q^1e`TP=2^ncvLP=Cghii1l6#j35?9
z(#F%^yfre3Uf(-SxYa@E>D|>2Pw@Tvy#%Pq8jf}25NMnGUY28BK$ZDf=Iw7L#?Rbz
zPL$12t&ZDkjDQ2kW7NFPbFKfkv_1>~TVN>q<8;dNywleu&5**k;hJ_Oa_X?*tv2o9
zGaUfApSlE;4YU0tB0;*=7mc&jMvOM9qZC5Z3@0cLHt(PfmSV~7Cr!1)ed&MMvLM11
zp)-Qzlx;cd0xL?<$QFd43wu%+J3w$;C#UGlF`UN3r11nf(~)9HiqZF?n#`E}csGDQ
z%eTkNex8j6m?%w=E6-*xIp*m@Ih0`|8CqwvBkc=%rdbFsHQH1lhlVC?{VnzCSxvC#
z75jEg!_>L=Kly%496JFNl%nBSqs}UHi<@-nB_{|npm$ESj4Jq4W7ax_qe-4Jn#BEW
z(wpB^=|50*HN+j;sVQ(r0ZEv?f<pRom{_?IcD1e4Fpqtm={ai>e`rD>Gv3UN%*$E>
zMV^~VY{x4J#M^D1z#tzb%pYhm<O;|ISQ&7Ta;>vr4oPEa8>rhz=eNG1HxV)3VXCG+
z(R%CRTYH7o6ZEvHD6#T46=lCvlqf9S1{9FBW|_^gob0GuDoRvcs3)9gy{V*ZTstM@
z5CVTGDTn!$l2Vqx8Gg6&3sb&^DVsaJ2$kii*eNkqD7P^r(;)-oM9s*K0L{2xM2!vW
z?{%TpJg?YV{~K&z#^7uVwRNM`e1_BkQ%Zgk_FpZL3R1s)O||Agl)@daA63l&D7aO_
z5vySkwdT#uYw(8Nv3w2{<Q5c`Kiz0keb}ylkyXS~U@r%}Vctrrx&8G#{k$z0Y$`Cf
z*GG6h3QrW65219_*2d9Df%z}(YOA+J3QS68>f7qu*T`ck%n?+WTX_?%QCYH~5f$cc
zR__k-><Lh4qOy$56{Wxoc6y`aM4h>|msX*eyx~-s{|9;Zo4Rvb{XnqI+Xg(?O>DQf
z-C_rW-N80D+TISB;@w=MDCBLV!h8j5_g$!Vh7IQ&!ul$K$(>Y~yG@0;XSKLNr4z5c
zr&FavYa&%8HSeKjsVZNgm4|~O@31eBBCtq$m$%nci2sFqM|e^<WaNIEw+G&c(R;6a
zGLE&^KI^GB>!`fmkok9B>P`Q8{VsRSu9PjxQTw>z$Nj*t18>A1LA82up3?iR^p7-4
zZ=bu{)T3WeZdXGqe>0bR#xF;fWe2>2*K;VFn@`l(!hf`$V=Zf4(~F1bMWg|YS6eS)
z&UnB(Z0f&*tA}jd4$>osQAG}TN9d2SJ1RPVzoU<(=se0(q-+E-@)T5Ogi^sB*Voi=
zcbgjS$m-aJF-z2TcGMS}lk$_3y_gYq1$WV!<BX47n4N%EPdMG#X9Rr#sZN3BKgVS$
z1zhx3MidG-mfzh5KWn$*$C{K5#Uyl&Q3!Mm<jN6%=}l?pgb*BC&)!jIC$|nMpQ7lh
zfUJt=nwD}Kz-UWO+GaG@*eGl0Vq7oreg#d9p8vcqzg@xa@*MvK#a>ik6#Tc=+WF^{
z`6UIHRl#ct1h-qHct<~w(I`I2>FjWihIxH7ADQKmd401R7=v@FP$n}sVI1@XLv$^|
zMV5DegFbSWclX@^(QMvBcm=lOhy4%gB}xi(UZ(%50&Sl6b=qM|apixQ%dr!G3YAfq
znp<RZ-u)_Wr2Y$f4b6Zy??QJ{Cbd*ZPOkFXD)v7`b!1W;(;t}u)0g$2PKD9f-EB0+
zCK@7NM6FW}RTP_6Frz@~9bZx(e^x<LcjM3L@@Mp@FKavp6tvVfpVi&(Q}DeCq{Z=d
z7*tDx<In4IL4nP?s$z=@u9CTwS)MA_7nb}brCd}X9gQ!^urv7AYVzv}Y!>|`#f%1)
zIBP-6O=7BEwu4D`l8BGomWxa2Q*t<BTCa6EYP52vwq-Zw`i=9OzqPhSdX)^?l0y8D
zQBIWQc}EDvSx5uZu4!4XKf#$%bn44HfQ6fIqB`%QO+fkQ_1apk<89)}!lt~cc!M^9
zq>-&A`nx0R`Z^_^R!<J|fE=#OIA#(T+L<j*$Ts}8uN+g{?D(Xkt6C=`+dM@w<w1!d
zR&z|PnAskk9C<C~<=E*aMpkSrE{N>Z#BR<y?Dht|KCj;x*@7{$T|oDtF!G&sW){i+
za<$UrWLG|2C^kOtnW)XyitzGyKAE1zyGXTOoOT5!@K0M?o;q(%g_;`kqOw+&>gP|J
z!nsw`xkY(O*LeVq3Z$YDbxt*SMT%;zn2$+$M;XU4kQKj~SUF~l+;O$RDNo(Ck<I1F
z`NgRVwc{uGbNlt9>{6PpFVbG(<jH7zzpMtvn+N2<q}Mt)I8oe9F8wtw@FI@d%uY_k
z_=GsBvuipdF)s=RQCbH6Ju8b{@U!F9;n>QUloCpTr4tiTgO%Cv0mjA{6onjiz1J`a
zdmqXSGZ?jN|4}G~f_SeRiMBJAI4te6w8IV)Tp7dBlXL0R=)z*;dS`|$Xq+>7zA|-T
z2}oSW;fti14j4BjoMBI+oOQkb1!iTKmuGsM!!Oi0eou^$jt&$8v0X%L#BNWMF9L1(
zw6pda#>1+{|3QM%pw-oUh${8hboYjWvucq6#^x)n=d&ao{ij+<((QM5;iePmXyb1?
z5Xp{PkNB%I>!)*5^l)ZMquBVU!6HN?d#KfE1TIq$>lvrSQEMJX!9mI?L*XWd;L>I0
zg%NKfRc_&+R6m3OFv0JDEVnTa#!%Qlse{!p<jvebn!y}Ek@>FJ#XYzOL!HWKUxK41
zM!o-EZRZ*r$5q8~d)KkoUVFWM#CbcL=HYBaT@q15OMn(iMbySZN|3-*LbfNd*YzWr
zT{mU5QPZG`7L`h<s1j)kE#ehHTd0IupnO6?2&qCv0@0;BzVM+%fGQs9hfwqTpF1-<
z`>>m&!j`XR?%bJsXXeiR-*e9Yo^uZ_X;GZ&YvmeSk<lgJt5@jgr)3PSbYrAd`5yjv
zJ>iU_awVfMaH-R`a#i&lkDkMSITWbiYz8$d`hf=^LF-KZcN=XDmw^@Jx6|7y$>KSV
z^m1Q4tR77)scn6=eOiJ$1M2UsAjuJ%GgC=_PFoqaa6!XX{}VVn!%%BSyezBAFHwG=
zl8(a5SxGzc0!MI95FU}%%5?QZ6TZy~Ug9nIzfvL<Nm))Y@HwCng3mEGR^B#d@fOsR
zl*u}O2AaYw9(5OTL^#?%GdHz_!|P%`?xAUUiI{VFui#m<7doJqTbwG*%;^ed0zH9U
zRzr=eZTxosw`#%fl)R#(s^s@d=pSw6%+UhNIyw0chKQSU<$+Hu;R6ZR1UP1MSum_k
zZ}F<mu{qBlwDm_NX4d_aw(MKcr{N-ceT}VZ59UT3|1K2p!Tp*F%_~_X$@jV0$Lq>a
zZ}o3e@>eBxb(;+Sjom6<JI(|?SYOuRe<=Bvl7A~PsJFzR{tS74*JpJL8zMExlq(RI
z0fIFS=Z2Gt_9IdS+Fo2SCC+6bQLr-aoIG9ktnM)~gz17;!>CQopI>2#&<5sWieac^
z@O#{kkhhM4eSn5&w7rABqIf=QQ?Sp9Dl7*Zas9@IkbDiTc%!~oIb&o?^m?`uiL+Z+
znZ2Phf)uPHAu&7LcRG&|D4R~lF(k4zH8c{lQ)n<gMZSpH=me-uH0JI=gVF2t@a>k-
zkm<UZ4D3I2>=-dSIO6r8!N?%Z*RCK9Mz7b;6)v>Cf|m`h$dy~1pW=IjiVOuiXcLrd
z?LW;ZlNG_4mJZlm4VYQzdFn%hvC>xj1to^i8`!u(t~5x44<t^S%T&%9<SY@lZzv6;
z_t*-(HzaOvD{0XnE58}TZbDNbUBxE!9+-=YuiM^;ha)PJ#|KNBOIwJ}yWR92!{|LW
za{ruzE~IT9I`)%0(0gpB%&576k_)Jv(t1F}9--tYB=4NH6tc9m;iq4r>@L%hY%N_^
z+3IaV+p&|aolvs1wY!1_#l3Op*yON3WCUyJJr<F8w;@Y!F<BbD30j1Wc)r+v5kCY0
ziF=1>KyqhyHA&O$#bfo-G*2+pbPRc;JTYmC7HRL^$}W1!?#gcJmS&_6s<i6o`9Haj
znna?ec2S?D$8aq`Q!wD|qIFoFc@8aG2Zef{cb#b#AUNBT8Kb6rB~MMesb^2KW}z~4
zjM(aEPM)i4J-PY4HL3rff@V|OYWQ;!ZAE#K{~h`7YGi7>pog;t=M&kxDk&S;x28gF
zC2_Y9VDn>FAg5k~G+LX!7_q=c`oHKIU898hGd)$tyCz<vgwy<iD;7WZ>cx0Uu2B#*
z?LxgYDhe>OziW^}Q%Ip^eK&cOYfN^LBd=VIGHC-cNwDgJ;Z}T@N8-qTqj=;I$w>IL
zp7=SOxDJuk5&qT_L76i0D*|4Rgj^F~C4_dE`00+C(0RG16Z6)a9X*5p>F5bPX*Z)>
zO}WCFx{G=PCRH>rVj|d^sAI(RHj12Qb%C@J&3)ZWLU0|EK-U*OALrbch3z@cGPX%q
zBAG})%X$-L$!c~}$PY=h5C7L!=@|;v{cH&Kic*EuEUHcr^?J>2Dw}F|WJ~R{JPZPH
zTdL(T#l)v!+>V+F?Wjkb=vb3JUoxRjLd^+$tusOMSC%AWo`NxkJI|199nkmoYmz6X
zAKlBc@{&O1rJdMa_cAf#P(<r5y7*ov#%ITw2^R0xWrI`kWQq6CdrP@{3R5}jwAYK`
z;e2skyX=}{)VhVITkTbSI6O9~*#C|mxW4!BM-MxaF%Pj5+5y3WWZ>YtBPLNIRmzns
zyq8VVX4kC1$(j|onvK^f;Vy_9Ya*V1i8jQfuNL;*lllbdL~UCkCBa4sY`cwbEE2FL
znk}D_uymhAxd)@oxWq$FK`MgsiOj5fAQvj1ekEhoTRO38vPK)}Ca1e3bfMWZ=OuT2
z1nESfv6)xN@;Q$zE$0YWJ+VMx=aj34cK>$pFSL^O;bsk4Yu>_*KrO|~7Op#Shd_vv
zyEwalKFge^4?jbl&XuR4(BY_!b8v2;xe*T8K-RuW8poGxGxylXc^lE*)9ERT^_BO9
zE!FF}n81VA2LFB}y?$pqJxTOG?HOS;JyCvdNg-NIA)-F&!nYQqz?4Pj&Wau^?MtoS
zA8)X)zV_DY>Z(&=>q9laR{l45xA_tODZxsEnbm|36V45_40MjEoCz*9gRAD^ZsCG;
zln1qY6XmLDqqCk`Ucj8`0*^eS)6c7bNGj`CWZk+P(|EQ38vmQL8oJ8e{{3uBc1+LA
zo$()Kvr!))<=e9QrrMZEN?6IRW|A7jC7uW#+;-Lk6on8_fV0p7+thy}3v8XLUY}$1
zNy=Sb$zo&D)#$V+coOFQFG;LTlm>J{rpbyEOE^PZtHC<82%AiA6HNA!Oswz*PrLCx
zraOwv<M^Al7xXRV+L*r98(~?~KAm4vn3(>!;k(r(9@4WJr0fXy4~VU@RGr1LovSrH
zY11=^;a9@{wi;vYyuj9_CXJ}UlOdCJn#ZWq(Ja>oCEep^!nOsV1kJIt8?ES;k#=9l
zXJ~faPXt4Ek{yR3FKy5w?+!Hxa<GJDx(R}$8Lfw)zG_IeH+|)i#i`kt67-;hp%6Xf
zCw*~wX?bP^r_fHrnLAO$Ri`;2va*-Jw?)=I6*1?pJ-Cl^8^Pc%8jBY92Ln5*B?d;+
zf?7vMZ(@r<jGmL6a3i+54?}pQxpVx(w5eMi$FSP(cF@uh+p#v{PZDJqovgHX<TKoe
zh-iaAtv5AvB)#6pLI4nfA=6&o9jvx)@x0aofdMWPfxuzzfC2X%f`Eer6E5=x3tcf-
zIED{S1BD0zKH35T6b%b1460RzaY6>BMN(kR0#p|W3eYG-s}WPo_=7r5JgHecDSusz
zvYsy{$pI?jIVcx!+KAU10K_=v{rd0qZfe_xm=3>OurV0G#toc5B+3rLe;hXYfenoA
z{sn*{TL&9igN^J&`Npe(jZhOm8=?kp!>pgxuIa+e3V*l5RJHYtV;rjXO)V5=m*^mh
zB3krs;e9cI`7r;gT|ul;<G{~ezVybgk4}I5L(bNDD+L_p48INLNTKDXvxYeaGC$C0
zV*^P9GX8g!cY~VLF+~_orEuoY>ukfAftW~S`mc!fa|IaVe@|Bz$W*(7kB2J4P3l<-
zbRO6Da5IglfsRzsB8oZK(H4t^H0a_RWVB08EVkE!(4KX^q1He|rN$IAmf$IN!nt7$
zgy_l)9v0J**_1-ups6N!Jpp0>COR#nB%QWn9_vFqhfYf*A9Tztq_IxRg5}NXuq=oG
zx}^OQbfp65Qhzw^|4>iwl3v9JNU9s`5r&*;LjZfD&#9Qb*5|eL0ttm$LD)}p{6|WD
ztfXNe(iLCSgU?XgAmR3G-3gDCWf7sfuT}$64AI*C@9V+>GyfSSHF?u2+pn8>mIGf3
z>k&F-IyzZp)|msDbZl*8`Xr~vk*Pb`o%EbA6bMKG=@dJi!ftyRL#~i?nsT|8m-c!M
zrMk^d*Tb{#ua4H)^P5D^@?C~LIesEpM`#6U?RxwW=1k(q?U#U~wNcBc%#oNVP+VRl
zP{1hw3?;geH*!-evx>f9nC5~++G?nYnGl7!(>C?TXksMlLC_y@pUlLlz0w(%Q)GVt
z(2_PA&zImjn$3V7$E<#*hHv_Ar`)cXorL*8WE%)zuRMAL4Ki&kl%;zoDg!rW+*>CC
zhBGJumIBBX=T=UhTsVQpI#m5GMx~JlP5*AQU$f<k;~MryL&Be`V|-qT(O=gE#$T$R
zxEYpnYp2iO)&yQIkFEc@BA{H83F|+oMm(hC3rg0as{V#X_zzQ}{GPX|@(O9H5l-L6
zC-?ut=^zH9M7WQEiO^0lt7UMpno;~Zwj}bw#rBVtw4$)P1YNLr#Bg3^WC)XatlNCr
z$uEw?ycs!%4al&WN+yg`5L_+8)k@$pltXy@B|02so@&R$m0`A!MlH@B!rZ|UZdVPJ
zQ$l50f}zWQBoylK=EoYuv;COL8pXV?t}W)h?`rUR#Yp9J^XpHR;*c$Ip}c20H<R0c
z>R@gz&%Z;2g>~+Il*R5$J~u~RqMA5$kiSDt5`08G$i%^+D~4v1K<slcTFkS2Sh2;~
z_A%Kfa$W~)gn#F9pJTd2z`P~|zp5{<34&#g0($iYY8@g9y;asB@=Yp$oYvmTnQ?|v
zCI<#lqXMo-(||hO8iwoS!FSb$y|^qGfN&@a#-u+?_PoPj6{cP%^Ddjn8(`WKe_F4q
z1^B(CTb3GT5F#<4mw%hDJqCOoZun*qpRf=6|CHMBWvaK8cm7#bX@Z=MiB~1fy~w#g
zSJNE71&$8uJ~OMSw6R=vfto7eKdO7Z)zzfyS5tTxotYb7T5`)HjSr?3`7EXXtR2D#
zN+^4|-liKbOm24NZC&p;T*Gi(tFf6BhP1Fm5M-x4*E@2NCsE}s6sCRQmak{0Xb?*{
zkzZaY`ZT)D|FaT}c#{KG!94}7${$}T6(}LXes!#=J$4i<B3=<(7J?y96EfAa#fk*F
zTRbs4CDF=H>+IBmw`k@4Ue#p4qZy33MR4pk$wt0V#00s?ew~3SdS!v@!(H+9V5?*C
z+^ANX8f%00TrjH<PMI7DPQe~NKDAO>4q|&buL$B$RCm>1G+@<)b?X`UThy+tO19~m
zTVmIYXgK5MePFweyYLa?=v)CCe}}SMWQ+0Bf%gF4AJx^~Q0L2Q>)j;P{yW2$ebnyU
zKd3`rQK5pi#E~-nv}p%yBIf#4Z3<^wn!;66hutlFkki|W(<B{jsgyHnYatV^XBi2c
z`JlIbKO=<lw6KZag!-=>TVfFCBnrxX?l=4XaLfH4Je)nyc3(ElUw`({@QwcidD2m~

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder30.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder30.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..cd724b970f38c6d4e63d926f0bc9dd709472ec62
GIT binary patch
literal 58014
zcmeFad3c@IecyTa4f{e6+%I=R;;vnhCCj2H%c3AfqOAB@$~A!Z0t5(LfbYHF0>2n5
zrfk!7lsHYgB(@XKS(!<jWKuV6mbOVZ&ok*{>(lAHZD+cS<El-cCUv$lwm6^f@4R;d
zKsjCJnSTlr2j@N8d)~AB_VYW7Jw1gOKVSI8bCvmz$7BCS5A8o`9^T7u?bqV5m=`O>
zyttQ`kC)>1H?fc`CF3qFRZ3f0rj+Gxaz3|^FXb&SHD6fhDs?S%m%8maJ>Rp?Tk2it
zEA`oPX1;%6pfnJ*Ggum=es+FnVW>1@X}S5~g)OBm7SGR*ENm@pwRmBEbYWX*o5j24
zw=e7{?O51Z+PSc+v}<8^Y4^gO(w>E4skpGWw0B`&X`hwrp5MQ4pmf0EJ@W?_4wViq
z94;NU=id1v3uC3Rg?mc(*mK|fy$eT6M;DHjjx8K79k;ap`4bB#OD8QpFn?;{bm_Fk
z2j}lwxW9Bi@gZ;6+v1IUK3RIeJL-*k+un?q9`v?*JKl_y9(pF`?euoN8S{2kvU3kt
zPG3)z9wBYFw}-Snl}D~8OONtg^!D<+S7k`~7E<<k`$^efd6ehJcs}4A<oTdIzlY~T
z-eH~(SKdSZ<2)bn#&{mH=O=i+$Gey3d+qtXUgG80*s+NvrnGQncHS#4%wDZ8`<3G2
zbm2n1zEpek#EF^N`i14m<5P<ZC+3%@W<7smwpu&kRjP}#wb|;-F>iKydbu{cSUt9M
zt-k0_T{tm0zc_i~p@&XBIO$DKJvjZq>4&Btc-Xt|zUhbWpL}TYfzzie51y(#Sh?@i
zgZEF~*J`QUY^mI6sri<TdB>NoZPvxbrAl>T_QdSMlD~Ml;<b9`dR2Mk{`(((<n+V$
zO-?@i(1R0Gr|y5~A@2dtqqmPtPELBKy+<B7d3vhSY@qex6EhRbwc6}NwUu<2K0LL!
zu=K&%+KKwaf(EGWPgHBui~d5zuN96REAn@0w(zz#*Z5dG%$=K9nDi#jy)>2x^L}N*
zr}#`I%=ne1%0!*S-l@fEy>hj_Fj1YDsrX@6b$Nb1N~3;uX~M71PRtWc)h~Gq{LL>+
zT&$F5{8^8OUfR1@Ua9zt<;Cj!HBt)aW^47Q7OT^<Gh?~1cYblET)$c`Pt?k@3o~J+
za+T~_nDr{t6U+1UFg;sam{<z4O!opEp!EuaFg-s}54)!pm#g(MJ(yUig<Dw|^@%E-
zsaML={^CNJUg{<BFmrCU{`A6xKC#M3Ow5-r&WAhC%udWKRww44iAMGeFZO3Am+P!1
ziVi-@{MTpe*UrxRwff6-y}aki<@#c!I<@Fk{O8V%CBs4c*2zkJqFkH0Q1O=ME8!5M
z<xfo2%d^$`V!2+a))xJ+paokg&(B`0(4np~XP$lj<%xx*`HIhD|FinEbBi+!@bj}*
zXRD<3)s`lvD&_j@LZw!(EY-r?ls;&R@9rYzFV<=k^?EgoU!xnDC@0@aPfmCX=cf{y
z=jcaqwp#2derqX$AjbZSuS~vhx#C~(XX};X0v|SioZin?i&xlr#rlQWT5)>0I;BA=
z))$NV-eOIg)JmPGT%MS>YK0;{-eRRztS;7zQx~+1iWAjq3yXf`_^lWNnyOXir^A%q
z;UDDn7Tlgaaba<xa)L^g$;HKsC%lRJ#EBVyae3*)%gfbsEMdmAdO|~f;tZSpeb2sh
zV%}RgVO<3hH1_`q&s$n3U#i)lmZxW{mHKkE;#F*}?mNk5346WS+Qj61#VcFi4{IDW
zD5<1B#-AyTN#jRxAsrj6okW6I5Fd@zV{`E!;l+dGZ2Y_8!=!E1ASTAcj90llJ5}L3
z(*}pa6f55wON9yl6u+m#bnU{#QpLZAwpyR>AK@_^Sgy`$i?T(^u9;&bTi<rZ{gcF}
z`0e=F9X-|lGE-WKfe2$I`sF3Oq?ekBd1)`>Wxbr2_X=JYv2L%&>-ADzAOHIKJHT(6
ze|`K~Y<l;nV$*SN&>MO)UrOfV`Cjiup%BnUDta36w(^v&WV~;62<3Xbl&$2}@;76Z
z!gby`oiGFwC%wxdm{PYxDy1HWOiH~Dg_Qan>L~R)lu;URr~;&WPi2U9?)8qonJNvt
z*1eYirN`e?yIZ1^m%S6-NlO`tQqBWD@3WMx^zVM}0s1=XJ?K3I!rSJ(;yvs=LdtgX
z9`(M3*beV8?>*%0^d9$~;O{Q)z21}j-R-^4dy2n%qIS=CPrn)C%|-95_YCQKz4v?1
z@^@d9|3UAZ_q@G#zc=o^K+b_r#Zs}-LDt2qVd~787oUIo-zJ&m%sh*7!VeSkr^nB(
z<k&bT=Gn#P0dd;?Z)<bDO~KgP9>G_SJ+<@S|LX_0Kk?O%4*%6>{?*F&J@M7A-}?RE
zdVcbaC%)QP2~p(to(Pjbh^b;DoxfUFIeyh%Ent<`W7lJA@teFP*y_bE$7($vh<S<Y
z@sG!YtwF4us3(KiTxu@;xtOO|g1@ORq^`#yC0P6AOwdc14SEQ3L4hzIbP*PUZo;l$
z8)0|7Cx{0l*Av8h>wUD+t9AHRf_LUg>nEk(FVzQ>b3IPVU_I$&>qGTnQU=$y++>`u
z$9Z$^3rW2>81-VyDmOy@h+hdtiEj-kT^<dF2)6}W2)740mGhT_98WuforF7s-ToW(
zT|xSKy!OTiVti%V%6x&Ab_eMo6J)*oTM&(uQ(AxRdsT}56!hlyJ_dXIze0MK^++Xu
z!rD`h_t<#G=8C~!kPH$*U(g>6c-?Ph7`2|&VtsFYAF0Ke_;p5lb#HzD+=2Q*zILy^
zlcz&o?^_zJ!(QL&zM0tSetr+skI<j7>+v(O*GFGVtR7rF6dVlpUXJ<yDcHYukJld@
z3ij=YjmB<jgbuCl3l6N^6CBjjrI39+_+~8Fe}<7<-A_CB*7vU+2=?EM*N1}r*W=!h
z%|?oM9u2k=9;+XpI}z-l<Vkk_@LP%1!^Dq$Htz4QpPD;O{66Mzi}mupY9@%S-S3SA
zhi__5cYQX#biDq+b*+yFDfu9^9^<PXT6_4W(jM`EjrtwiIzE>4A7?3siK{2W_~n(8
zQ_H>&dY!*k6j=a9Sz20zkf_cS_py}sfjPY5zN;sr$mM-2J&zV&TCP@A0_3#Pe{9LG
zc(V>K2^X9PXJ$cUlZ&;=O8;|}%94sc{rr<P$jM?DKMrK>nw<gvSIQSF*J@!wFD@Gz
zxm|v#QmfSM^_7JQ;khNDILj<vsm#t?sMn4c&kGsyf4NqfUY;*r0WTGOu=Ap?SBd1H
z+eI2UX5R$fn!32O2o-Sr_L<_nt)JqnPrfw%?D#W$(bF$Kef~Vr;`l4im+5Trl^4%E
zdH(5_i|1b`mY#m;g^p@tnXqeS_Hw0K7Qs;qdphZke2Z4-cA%yVLFCuNLiALrdQcII
z(9M-{?QLO@Tk$aM)vtk;-ow+|q7J4MG=2;_^6)ahwe18m%<Jp<)p!t}Gib7wxEcGN
z#3cxM!)cHrDIxGMIkmJLrmM>fOV=3p#PO41N}XMg_zI8BLv%shFFpPlTK}5fw>`G9
zS3vW483JmyUM?ShY<_WSV!rk`kw)pUg#Qd>{bvbyL1B7!e!dJuU9Lc8%&v&W>hUX=
zmS>qCNYM!o=({*`dBPui6hyqt2Zy;v`(eLPkd=DHpIw;{p%bQ;n9_QfFPEq0Cu+5F
zS*5EJ3l(Dh&|02GcBbN1PuMdxKf46tHWJSZyF}7fSc=!m^{{VR)MHefbr%*4*q7Bw
zCId35TA8nfqpYd&5>unsL!TOr$*QV_{UUmd!+F)8SW?H+wk-WYK0VB;9xKit)>Gbc
z^c*JVE7dTi@d~q(%hS_P|6z{`H^@{!>~BURN;&MhtUhQQpjd}l)-$u#nfb}ZvfBk=
z-#lFx-P%aVE-W%jl<zW9m<gMnt@}OdPOr_MBgHQ+)x!ZMqf#%kD$2`CER#x2Ni+1Z
z;y3#hCMn}<afXTNQkWJ;axKh2L)R-FsIpqW?{wJLT<RjS8Iyd4Rx3VjhP@N>OBdiK
zrdWLbA``DsW94t%@Tx{#Vd1iNLD)SHg07a`Qt|IqCsJ(n`LIWwZoZVBRm1#?;ybiz
z{12;OU$cET-s}XxQ`o&c1%o4sn(2hE0<bMEgk3M$U$_c-oKuk{zdAE^#6Lt+r7kGf
zdNe7v#Qpb^?7ykk<bf&TdFmJZr&JaOYQi(#s;pZ6vPxc5P*E_gU`Bz63V%w$2NXP~
z;G6>6NHxWL1$6?q$ZavwYr#wZ+m&}o!371MASh|2v!`sI`j?gbodn_|HP2dxTUt}j
zV`JL=kEymT$L5%p2H9z_fttL=+?Ddx$`uFfO5Kf;WwC?t*-}qOQJaHXasMOgg>e_6
zfc)19?&G)i20>wOe>@Z4&Tsp_?s%^~cgJ_f3kl1yFlS+ZJQweer{V?vcJnt!jDO0z
zlS=l}TFU+nk~3)K5^wazN8_19CZV+9cz2>VVQt)5>FI!LQ(5}d_~|7g5t!z*_De>u
z1<(k7Hi)lgtNA)y2NXt1l#3UCE4`Za5<zw}wwkLa=917qQ2pTUd_A$2A}8-bjejDs
zEp{`pS_rap>DS_`$$BP$ssb-(X-8CCnlK;a=h9ws8*qItdofXaFH~VhRAIeP@0#o8
zx5vxUGPK@Wua|@7%U+Kg_Sds$z<k?tfoJHaW`0-DMUVP|{7t^4ncD4A)lX3|Z8<%b
z)4xehZ>!FZ&N_YVIbMIOMJT1%+`woo$PPp428Gd>{pFoI^tA&@-x^zme&tIB=y`Wr
zPX=92=>1^O?G4>bbo6sr<6vJ_xCs~N`nv;~tND#vf-d(?E0x=<)ZCCaaxnqT1T{T3
z>}?&2c~IhWTi`0Yg4}hrt1lQ)zjwq|d#aD{G~|t1oqgUmwOsF7+p5vE_O~~l^o5k)
zYvp%X`6tMOdj96`xf!Qss=m$JX}m#?*U#GM@sf{e%zC|D#1RvUa;8Nn$hjRi<7+!_
z#`uocL>C*qR~)~UTn}bMVfZ30AUaYLNO|7A!_Vz|9k?!Dn7E8gVX-(_DI)rTMDdES
z#Kp?Q6jF#t2;z@ZJxm+*<9~<ZL!H1LC|bE34%IIC5hQ7$fnAok2<d}ZY`NwDO)Gn+
zkoH}3w^?PsG6kORd}Fwym9=ncU8et{`Z(CcsZc0BBD|?exVx367dO%&H)g?B-dFd(
z@6UJN{QN)smT<UL&;(yoi<Rk9o2H&N0{d-|Q2rAHU*(sdf7xUg-|N4Z=iB`Y%O;|U
zNN$j7_x$Dg$q1j<&N4j($c&oMa<x9T%W%2BjjH}f3BqiISpAPH>4cJ!AYuPgdTgN<
z0XK(1Gm!`^I}NG=d;5P*by6BP$dttyLpx!1fi<zPyx<Vc=T!TP3T&D)=baQS&kWT}
zh!zfwFV>$GA+kU(E8f$-4-JB_usTuo|7X1*IWa$ftJe^!Lve;fMKHGY(h#6t2AN=F
zgT2rHyi#(K-btkFI-*9`(XakB(p39vlLSH-g+v0xkW2Ie_;ZPZA(qj2e}aFYk&)zf
z5XN>8&2TakPu$7HQ+N84yFm}bRtvOor#qPwin`OCH~_-wCZ4;KOLkLxdwgZY5J;2I
zXb^D!1C#s@O!D6!lhj9UBJuqH9h3NfQ5&y`J^a6<$M0A00}2c!bQwyB==tGJ2<=i!
z+t#HDQ#7&+gs;mtzM{0>vhIc#qWb&Z4R=JH+S5W4>ySiLU~j9y20Y@5IBYW7LI_Q)
z;dr0}t(?tpO4P`YFlgcSR?!w#i8``b-sz~aL^%GBD-ee9e?q}61@BexlLX;lJ2G+A
z&iX$^&IU~4|5a6LViEtRc`OZeFj}rYtbs{Roj4;(>u(7)qRurjj{h@i_-7Re)%gF5
zf}c}h49;Iu%-ET~rP!}3_}dD;q~NbB_<03?L%}~#@HZ9wf&$|#TI|Jj(tca%zf7>Q
zb93nIy0<|UQ=p32v6|la9ipyc{vN{k8WO>|<XY;clkh18N_{SA@;>OR<+#5GT3&L}
z_!p2X#u1o!u@55Iebu${9&07r(Tc10L{u-~>NVtZ`b+7U|Dn6qEOpkr64Sr2<oL?A
zM~g3`Dw6U`<eOwzk-%(a2Ldh=k0~qSfE$@(Ime{V^YKVzz_S&9PVquh7=sdUYH<l6
z%A#LZP8MEL<&gb|OIEIFCf7|0vH&>E8dWP@kefB}fr&T7jrGEh^~Mu~jyHo}`$j5}
zc}IL&e?0MxL}I1arm)$-REo-tpDjdk{2GgvNlK)SGRQ0tV31N?vYrT1Ye@;Q5!Ma`
zq-=|=rdp}1=^%|zcVsmaB=|e9nhi2*sUUk39ZwxK%vvVMh|NH>nhG-SiCx-rNrKrd
zr1g91If>HWjMejNunTKl60ojyvqWe~!eki1o{f247S=HLR+6{<4x;2d%wRr9U5@+Z
zOCRJN`E8K&uyzH)?mEmJLg}2?AQ-_*CJwhJFMl!Re*!iLwo*(btW6Z7+$T&J4ofEl
zRC_Eato7Z5`EaeT_PX}G!o>ta^!}g+7NTn#J*G$9t9>XHVBeH75)+#d^gR~4^h4Fp
zM78o4(|r5G>+0=Z?Tb<hQA@+C{p~Fwgzxgm8;MP;x3n7cPbX;k?_K)EdTb3=Pq{EU
z1MAx9rJaFR?KFMz3J;+}q3*w>rdUB&b8ubFp3a)Qbuj3E45sK(Iv9i*=iet1>dopP
z-!>@qM>Nu6obC?>*M_`qdbCMy?I2$;xJzSdqe@>s7$h0%XILA<tTGgPZzZbQ?ps>V
z%=XagFsrK107XA<8#2|(0I^|<4SoSu9yZQ$ysnEWdUA;xL&4DJVh%vKvQgf5kT)C*
zH}V?5V{FT~{}S5*zM&h@48k*X)Dof=kddN_mlW)F@l0jDV)AulNyU$@I%#>U!0`^T
zGkXgSS|oOsWYfgX8NMmLnh4@;0%zO%=oZdXC`?I08^+7)0D}KPG7kw9B!~>hRu0=H
z?-WjMCS+dOcXrk(1r{+*iJ+Q*98#&q<1O^sUZplAt#5r|-XTY^ea7k<og$7ZBhv!L
zp|=STO<iz1<J9c8d$4`MyaP3Y6d$+yi^%e>E>))BY>LuWtn@t}<v+*MqlMziP~nwU
zktlI&B$8Q5Cg@jfOjO&Z!mrn)(s!^lBX#pc-3jd$R4|YJ->)w&B8*NWmY%JK{Tfcs
znO8{6i2&K9MDk=`z|b*(>8W{Wq%e(ia|Q{#xU?`WqmNpcU#`|JEmtatyZcO3*_w0%
z_+wifG*2}}n*JH}CsFhKf1ybHVwjk{80M;xp@;vQN=+M@DCL=%vMJUv?Lb!87PSJQ
z6|kuPeb(AssZcE=2uEf*?=|Dpg#&d%92Q!GzkVrQp^fcYN+Mf&Vd^9m%<->aQpCg?
ziErdmdNTIzP9o`I3E0Bkujeue%Rwd>-<c@D_Nn$@d<0go7yL9BKcrZ45LU03Cozu0
z)EI`L%q0?UY)>GEymJ8NQvbkScUFcrS;|w|K8>FY5p5+yTJIxL7p9jqF+P{Lo(zyO
zAsK{{AT^1sEJ_D$sa&f0oOaU3>7=Y3CU!FfX7Ewu=b1_!I}yY-NDLbGGsR0su9oY?
zpm?=>a_o4s=%Y3hB(ezn`lD9u!{f!P#i#f^e*F0BMJFqe%FeVY^OYJZx9ZX|`nUD1
zMD-yUCo8^cG;KkOW-(JdVv?TXsj=h5_gDPN(c($0pRk5P<%kUnMm23s`k0IO)klj@
z7VFCh6{t(eqs0b^eYDt4QjZ-iB3#fbqc1sJE85CyHE0aWSlXA;%>RM{NlKh_;fIKg
zWg;RbY9Wl*F^%D!A%+1?NL1=>f9JZD81`K%8|mIC{UNPN?a1NmFw{&TzA|iUV_mTh
zibe|~&94^6*F-06TpX4n)wWwM$zc@FF9YkxZ>8L~tx?T=+f8Dm8f}(j_WeI50VSXe
zpVs$&G}7KiB)#c~tYC%D_?kXrkPVTzlen|8%Rb|7<?s4E{-3I2Undxgdir;iV6*Tm
zioK(tlb%N`=akADy5(s2i*(LM^Br5+|E`^iD1HrZ<EMS7TU~1mVrzl?AQhrP^zRV2
zL3HfZy1s7Tq^~Pw&`0_5CAw&9VK5a9;-0(pH>z>hVKnl2-Lha2tXmd95!pf6aQi8Z
zQd+^f9ycc)4ZpL{i{5@sg>76rCwqOlj$Vr0Yc5_9AW<(_tPRK(^|FzYFnNCivPG$@
zX)7i8X7wP>P0J!k=ax^#E%YDNY}oSoF~z>9;D-rrWo^m(f24%wl$ClrS?pH%^`p{t
zsik!N7z6ROUK(nw{{ED0<^E_ZH`~1HNccI8r7ioY-+xd2vVQ-4#eP-6-&KHB^oA)K
zXzPqy#Wz%{iy#s{ZYW?n#g6{uX}vKswP_w;Rj`x`5@@5)Jk7x}!4Amg2e1NYL1Xn+
zay4xvEGFCJ?FdpZOG$r!kX}Qx<)xio1$>lN4Qz^%uO(MAtJ&3@EO~;Q;ic@jiNJ3q
z+@_H|Rl6ko_x`p4Wi%6=Qw5Y3=jlWO5X8uW(fThemHh$_V|fSB;#Xa(Qxt6|+83A6
zu3s&q7ZADYl<xw{Hre4wbGD>VZp!%A&wt-iCm{bGURoREE!_a|L4KuR?2eBhO^!hN
zx(ZVo>&8zvk>=QcoCio}Y&E_pi;ejA#9xBl`1bf}T4UE7|0Ehkjc1P3Ja%A|XG~Mq
zldA=ESXdNZM5}lcz&{5AI}`U(k_F7c#CZwnroKq4-xY(6gn{(ZL4G*4mc2=OP;gLT
zEF<e5?CK+6rV%t*R-_Gb@t+_Sp@1#adsJdX6?PB{)228u{@+X}jb2ScACGmrt=HLh
z*zJs$%Z5<G_@w_^YR?)SrcsP`W#-yBDOxOHcXhF9rpgNwwTtWLDYsO1#-(4NXstk#
z+Y=+uVMi0A@kq1VDA7TI37Kg}JjHKqjev1y53DlIUfc-?(BNWG$}UJqL@*Oy&D2x%
zbRA7EWwWv!Gn>3DI$p8zhi&fA-=h14ou7^~TL>TK3SRmx?XNEF4_eX;caK2Z!KS~@
zp0XY~WA@iG2p)QZwAqXmgM>dWz4A0xJV6@zE;rto8I9TPf*C4{H>vv)cCo)~EMFi#
z7I!8E;&{e-ed$76=Doy>AMmQiCxkiVt?&>P&;N=FeM-R|P4OpHG6nxI<C^l6qm|2Q
z5G`R#A{=%EnuAHOC9?9axmWGf)Fov+-eZE$X08bS^|RR1m?Nftap}i2l{+a0fz4EA
z;zdaB6xx@a?6$-kQun9gD}xSsM$>N<-|nYU$24oxH8<Wq9yHBNvZh*-B2bqC$t0&C
zs@ZTl`;fo_-xp`f{t%pWh8kl^NIR%gn4CbgX4XPhu&vlZRj@#m8yu!tK7UDHq{WE=
z0$(I;)C$(-m{Ym@4nW-?N|KbXJxTy7Oqv-mTl>x2^#rVemLNdj2JFH0<PEm9rSJ_d
zC4d+(6X@i)_I1aPnjPZzYw8c;llTpUE)iB`5!n(c$RGxb8|>okV=&<`=-m3S++zS^
zjwn3-%Y2$yh!tWhj-I&`O-VcR?#eQLLZj=+Z&6o}gSM>fH6l@cLlpKx4%Eb)Np`N$
z_AhBbjUWXW8_(seI2!qz4yf&QO^eYM>}}wsxGg9UOJNyADNs%#s9Tg%qk9I^tY_B8
zb+etG!NBJ7uc*65Hx+D(EwQ6RY5JgF>(B)vT9t-=heXqlnh_uC6Znkm5@SB%DJKl}
z5&(*m<Q%9cWf+Wfk{X7>DdE!QEX|BhlCEYw>tS;H@i;&qL+4Ze{k#*w3>?=OCqtmF
zm(65zWzTZ;Vs-IK6}}1%PsUfp!yj!LVyzx#^S`O8$%)CD|0l$3x%fAfR%jeZn3xZ9
z3p!On8z!@FY@3v5<1ho0$&greK^O|H8^)YQOd0=Bvph=U1yHSR`+VU~Qt(l!%<$J!
zsoXmegm0AZ7~}@d)q1aqHQu<51O{%1yH0vZU^4&#DL<Qt;Y`oQpgGxIUWx(V8ziy!
z*=^5hp2u#)cpkCm49^czGet|RrJE|5M<^#apKMr7W9b}YT7dqt{eLStm*q4-*B9eh
zB_kvzFXzAH^;F|t?`j?_fc+0O`@WC}a>ClPSdt)qwm4<M;E0UDD1ij2IfvQ7^o@G`
z)a#1ss2pNoqp_jNqLv5La-&s;W5x#W4l4X(TGAq4eGa-s@XdCm&78KBe~OI%Lz$-u
zB1j;yW35)HZG;33a<8@1A6Fw!D0nZy%D{^aJHE*Jv$!&71`^khYc!1{&H|~m3~5fd
z&dK>6QbIgTdb112|D@pvlNh0{bemy+WGK0kcTd<@+)i6`tj7VT`}7fWYVF$<aB45o
zJUy<M-i8fA({j+6OqzL~894Yg3N$F7PGBaYRK(9Ue~^~KyvD}lu>NCe`8@=sf@o5<
z$@J`1+3``p*>#q3_CRmlW@!V7jY0k=)Hl9xkPX6iF*^qdx(YlGCU#=*Hj0pH5aCsC
zVi?g?lRR>h)S(I!bjt86YNGZSfwU?{_yBF#6p%;|IEbPkk5X%CgHT%FVAC1!l8Db)
zw!@Eb90!b=bfk-4q>E!&hhnlqF-;4V7J+$CWAOnMbja37R7{h6tZp+GujsEM`PTuB
zEV5H_y+7122!-y$tYf9yAaIn`K^drsK!d?xEk3a@5Ppc=g5(^!8-vGvA^<JbfP`=_
zxyd?^7qTzDmf_1u;TzNAzAe1O@$uV(lf?_gE5)hevB!&({NaHG*|SjL*wwYN6Y??+
zqL>|Nt&W5V)M$o%2c$xqompPQAS#jwk7__7DNfp`5C3;n+wjr4k;n@insn|@H5P({
zBBLb#lr3bWO=}j~Fz*@}8x=<{t!ZSKeSjh=&;YU*iNAm)hL0jn4K6@9e2|x!b!%hd
zUQvQ|P)pl?R>8Wib|IewCm`6VEF}|kR%@=8)tig2IPSWwKnHa1(pNP)`p=2b#TZoh
zeS+d}N-2~?qT{8bbx)c`RC#D>WBxV|KD8c43CMz#y2?~v!$!e_R>j)}Y2>cy8F94<
z{{i?}D;Z=_^z6L^0g3h(3gTvh!N>u#wfOU#=yLyZZy{+R4QNl8i&6_H{W6xG_vdM$
za8q0XYBsN%z6nxiy`G@3u5^yY)jJ->m;5t~U>E^8fjg)&$rsPxN`*O$s%OfRoMVzj
zy-tK?Ikl!?4Rb^)ON&z%IHtfMT+G;&eAbu=hylNpnfJ@!bAZ-&(x_tuY-xTsk8s@=
z7CAxZf16$o6q<1$>*rK-LQ}e=Qia7O%mJ~UM<3B`;$-amIW1Tl>vD9R7DJeKbn~ot
zHSB3tDD&tRvp&W?(v$)Q5Z$VoRK~ic)6;;#_BlB%o_pwtB>r0KDU7#9SxCtK-|0`8
z)$pAHMqHUY86^25fX9e|=xC+`LTD{&iD`4z{v`oZ#$NE)3DFL4&yWlD%u-v6-;8}X
z=fgWoe!B4-mHc$vsCh_>H*+^wwAbUGj(sZ5K_6fWDRE1QzZ&z<5u{?*G3u=5PQ_|Z
zUV#YEzYp=tVv&eO#|iC7VYJ{99H`omb_3(#_2q8FJJZZSG#|lfEv^pLAERpotMLsI
zo&(&Y6DJ;w29ytb?R)LS$t_MM+Sv!q7hvH$f5IgCE+|b@O%cjdiSkJ|ry?K1!ei1%
z+L@EXyv=`=#FZlky}{kHoJ&A?X;a-sOf;+9wX^=5mX1k?ZT-yixUnOrS{?BhDRHFH
zgR|%T*Hj~E%isSsI^}yxaC!pYK->%oCd|Mhrulu<MfZuDq`9|^CH$%?E-A1f=v()p
zMXhO3zcN!45(ka5s&{^UI6Z8uI}$vttJX0VTD$r+Jg$jj-5@yBO4SMyS=rDS8yx3L
zFH*{%S0EyY1mnxjshiI$uyM4&dOgUr*d7A<h$-e!$8mLByC*}JQHO1{Y1~+@1M>*$
zG%?Rn9>Bx0ph&>kvgabM7i?@Xu(3p7h-Mu2O^x3ksWxz}L%UUncyEtR7Ne-rkMK0+
zazM$cs=s)JPc(3+aq=Y-Y9QLIW6o<~Z&jz@Yh?^@%Tx%vqf<q8oPPa)43&jiDHci7
zb=v2&Mx;Q~?x2F0nUzr+l}!r2Eudqh?34zv@zX~{Il_FktIUMXjv8oTfNU}ot6AiP
zS@0LT5kw8bZhKA|@sc&GB@9|PKbnx)2skebfJi(Rt7orsasdtn0?2am7#hUpI7t9r
zeE|F<3{{9ES7+@h@1dQ9|G8N2s&^y1D$KyCc#sCjypU63b|U=C+tzv^lnRuwQwKV?
zFxQXK3nvpyD%K5eGZ0{E!D&hHQiB1u`Sc2G8>IXnBxT4W<-4IKs5#8ki;EK2?OE;J
z04Tq!15ozRPM7TDX|>l{?dB=%|3ZDs+DM&mU)w6<AmnZdH*!Y!{p=0+e*@DPa=eLD
zPeA!)ARXAa&w<G><5Zx=dfd*OvSkNO-l<*k1VPx-)O~6_lyWc5FPi1Tc{cyODkUkC
ze?>*R8cT|T9Xptn(*nICJQ(QB>abrs*lk;df^l@iFm!XD8eDsg#!OZ-($JL&pWD&5
zb{?$>R>wyD&+x_;X%{9@-a3-?Wp!N`*DM;^#D|{Jw3D4|;Aa#wqW*+x7<ct?rLE}E
zAnH-YK1v!#L3M7dhK0o`XJ;_h*G4|X$I)AABz@HgyCUJEY$+UQGGbHO8Yiu4F_-UG
zn#S3GLxGLT#}zXWW=xkGg-5heAEqlASS3MR494Rb48403JE0mJ9F`EWH<3zo!%FSM
z47?X=VmDiQFu}I&vu)iP*N(~T0CE!Tv6^G6aVO4mMrtCm=TaD|=`T`e&cY<>nYrv-
zPJ$kZC?k-U7-y128I9$`9y=kYLt#2>=VnCE+yAnHPbm1Xf_n+Vw2@w6`blqM;njv<
zOAUn?qugp?k`iqT$_8l|rUbto9KbC%Nw->A!_`frl4@okw{$3LWt$DbX60{X#{K)$
zk<$tuQJ_)qzeR!M3uB4%80MU1e?UdWh@Pi|AJt6BzV-b0*dc#{NLVQAlqY^YIQrRL
zE?-)nn0E(3?QEHy0PwG?(X;|vW%lh^@Ujk`x*@uv%-ae8JTX7100KScXB89_bQ5q&
z77HAw`Y_?RWWlxu-EuY&fk_8Gpp0<^#tn&P_GcAbQXrDsE&ER^_F00kCpyeJRpF>?
zyN%ReP`5Pl{wV_e*1n4%7h~ZeK>n}$*ZZ#$7qE7@lX%#L-SJE>c9xIlx^o%+<@n9_
z<}=--=ah%CE0aqXGKId}9?MT>MmclWlS^f?O!h(R+Z5gE_|cf_rzWbr<pl%(6^VQ@
zpUS8Co5*K~rM!~CSbSpO-y$Wb{$@<REGpTx+)W36yS#3M-jD*NJbfGRK5R9dM~s5`
z!r;iGCodSWQnz`);1HwtnzzH-i9ZR>MtQr*>GSq@MgI1Cd%b=99q{&h2lzYa9rO<I
zcgTC)d|kxxZej1X-hI2hOYV`tf_S;G9IuS)ren^_1@YtF3Es0c$~)<udJ{~6j|=&|
zkdF)byoll9;v?quLLM&U^Fkgj<ncltE}X}U-QI-xyoi_f@a}{+>Akm7#OKD7-uvE6
zmvHx_5mB$moATuULSwS8nf{bU)$`0BhH~~dbErX1#XMvv=RlN`u(qbX=e%>))<Ku<
zJ@1X%HynzdC%hND7p=tMW=+ju+MDstd#_l|5$}WEt9;E^ROj2g4_O=cnAeSa!I~Gs
zzH<PlYQ1@MdZn-+mo|2od)$A36$8Ef9{x=AX!%4x;yFYIiT<fc?`^&<r0{^wF|QU#
zJ2^9H{xO1HiEA##Yg^0!7K2#0{XId}-{mEduBPDB6P_fv^5%fJ%!`rI%jrzy0th%T
zeI?FnB#!Ya@w~+cEMAyVnZbJcy14xwk5iVj@t{{iF6a{Ey`14hX52W#xzb%PAS~m&
z=3I}&NyroM!0~FV3UtvwPd++&KZN0Kt!lnI`l?6^*807k`T(>FXBeG+eQgK<zupjA
zgYt=DTaY~{4GOJ1!U<43a6q@=fy3h4f<CiI{~79R$Iy0AEqU@4vMuQMNuAqK`!O>m
z-s$z3x#=#_cH24BJ;8wgtM#2;ztXMbFWOkf=Ju==Z>rzWK*lZ(j6138v&@XYPr-hI
zaF`?fHI6-l0KCS@`Lh36<^Oqtv;Oxg`jmq2QNWa<<gMXM01r`kXj=5o`%f$5EWy~`
za4Rw}d@jkul}=x~Ho<i!t_=q}k0z)<qp`_rFpOT9Tv}Z6UsHpx>rGu0!4}rM1cbtB
z4YvWt#mEzpdDm*BLR~oa;JfH&Dd&0A01uY3@Hz{)!s0YEhtNHgCN=+a6mf?eglz1<
z!#W38Ew=;VNV{jORyV@Y_Vmb8lF8A-!B#2B1tsk?<!q^Ua&i&VuEmR$szfm1kZkfX
z<wLU8Jm>Elvb(5$l@$ytFdAWAv8x1QLq=?s2HU+OM4u~}WMfF8B*>;32K^chU)p#7
ziwbTkSX1z03T%dLTz*iGpI2~Bf$@^fMbWk<4DlUQpA7qH&Mu0~_%<TKNEx``Tter=
zWq{%SjgYE;37zpa=?~|=5rFHT%gLZhR$ZML7w$}Pl+T?Tr=4%CjIIMsO*a=E<Qv_H
z$-qf#xAqc&oi%`uHd%N=TbvY`3;>MEm^7_n#<Z4ku6rB+n9JT|<BHT11<&?MY1__a
zKaqIMZK{;pR4}C<;sw6G+86h3+XVfat*lQf!Ibr)tmQt0$(i*q??vb)e<B>iF;8VS
zF;G*7sDL)!+n{T6uF~A(My}F+T%#?ByBo#(aC~Lo`mr&W5*<~hMA$ZdG(xJL<hS;#
z1biO?9YTO0pjXFP7%O5Qmd>5OualCjl?l-&U>329<L(@a=eRo(F#}H-+$3@6o)QAh
zXhTtC!Hi;?FQr{qkRm4eTer+7)_NFS0P#nZCuc_q6kmNe0j)PVI44j@yHf6@8dmd-
z_cTD;`oCisF)`5Yz}w1uA+j%E+$xu9*en;PmN@~lAg43Hu!glbkmpz{1uYJ+Vxn3>
zraymeWyk*g`-`u<EEh8`pFeX9hciz-c}`EH8x&dD)$n{H_iV@8ful^_#|C_DCw!bZ
zhdoZf>*TRNuH9uj&i1(hzNT#6Z?bLSp!1lAW00mRlyFp{i?)QO#}heHt50y~_Rld2
zA!mm1a)&FEBwvXlL#wML3&W!u&Sqp)j<`g^U9n~1fP0<aGOG*YD}Lu@$Gb_VFv+0U
zPiZ}l@`@waP3Dk=$plNgA4_cF!+*GPVEqc+teLwq9v$a&wRm&<+Twkfi1X0J(v!%^
zY!Gk4BCvS0bXj0VcO-qz+{+rP4Lc^HjhuG4=NfI<ZVb27F34?4)sR`!!;1e?j8bDP
zpHboWD=^&p0mYtEa85zzpc%9C%NnBsnIp+=OiU0k-C5bYepEJ|P^MP%M^wAM+JKuR
zzzqRhkdRoLFLmC?5@l&v5>omh7aDS+a8!dy4lwOwtGU07RXKBM5LuA>E4fK}Tn;47
z^<UU0Gd%HaxQYanTt0iAZx^W-rsTXZ9B5yKW*KJQ0R5-5y8kuh{1QRKvOt~|g@z2D
z{ibsNqXNSsEmZQu)cpq<sD3gfi4&60VS!vC_4Sp*>xZiy54ZKELu%PEXkumWI%p=M
zn=@T0!<KChn}%n|$f?A5V+`yoC1td+a)uORG%kTsZOTW{Qk{TI+At3OFKKjSkz>a(
zHXVc?CHJ>A2x8kA1RZndm-TYw-3Fk&FC8P$A=3CU+GKIWYL81)0c0PjW3v<yk_a@)
zNqs4KGABJio|ckxZ<lt?<&226QZi?h{FmjIwdI#IabyzbC}0aI&ltz_Y7$7EEO%Xc
zFlsq|F<Cof1gqYXtbM@9+-6&iJ^<0fK_gV#rnX#OIrxg5_pwAyBj8SoLj|?P;zX?|
zjOUV$`#-=auH?@*j!&-aPyyVGbx^&L<Ib~a){H7Lc>If+ok_;Y|A>MK1zPw1FDOH^
z?AG{~71J8O6<J2v+}Z-)Y;FGtwg0)Mbstqk0=XkFI;fIMYNeY}3_k|atzxr!=-npU
z;dXtzU!vU(>@MIaAZcLY0uh|!$#`1^zzC<9lYTI9%IOz^^qSZ*u(LCy6w7icS0|sD
z&D1mCbYbokGt0S51k_n#S#Z8$d046pf@I=B7Vwdg!0R9LMie-M5l^rf!4T6tY5?mB
zmI=lRX5nZyi{YSKzU`>3g$W&bFS<%cUMHq5`T~Pa6A&3Yn&?okF70Z(7|&vEcNz?f
zj8sg#SW2vTz6A8XS<rU^p>VrN{M)3xjzpK1;P*9u67kB!i6a6CYUHg#|4d_W&-yvL
zds7`-RG*`5YEyQGh^DI!?~IQd&NV|UC8|S~hVGkG_)Ss@5Jhpf39&^=XnUF!Q&#qJ
zOb4focWl7vI(_5J##Xi*ZrbaZYcl5%1n0s9`G{tq@wnpaYn>#`=-Myo@#hI}TP*)~
zB4ZjFsK2HW{gQ&t5%qtXv~OtW1kv1JWq@gg<VJ{gH}a-qxM-?{2v>LzL^DCcm=n}W
z?BMS+*PN&fa0j)I5*TBPO-GoEVL&}Yi6Vmw<Px1P+dWLnEMHHq9kqKsQ9ErAN~#u#
zd`U65XSK|Qts7)4-o;RbS-hx0eE9}Lp!_-VB%am(vPR|~D6la-tU=83y(-&a0OC{Y
z#?BFZCOn2$)V6PU7t66|Y%LZ{jV=h1$`mlV8{}YiD!%#9Zr0ESmAX)-t;7!!k^L#O
z1CkC64}Nm7pW|$S?Av8YhcS-jqc~q9#iasLqLKnUoIZfoz)r*Y+QM$bC=Exhi;Qtz
zDV}6xpcZVGXD}_Z6Oor^D_2ZafoDYAYDb5^;upgB6@Ra)8s(bA+1>ha%i;g|f}Ck@
z(Krbv7Gai50kDCUY!vMn2~}y1#2Y;5ONBIb+E3vOk{-U96gLtqAlaic5<wF0AMVQm
znzB`S3nVC87mWs^nHq(CYD>>3onOW#$k<5XJB$(vRnLLf+{!N(iqTJWBJjea#S7+U
zbA85@M>!$u)i1Qt8z=RgCvU(po`nHv+T>zbOuJH?T=d}yjy7u*ui(l10*7p?t-|l&
zJm5q<Dr)C$aY1C+MY4`RpiI}=lw1t?&Bod+0KcE{l#cCvOnR8RSg9;w3s`eY!WKfN
zO*i%H+B5%H!J7(xgP>vK{+mkpEdpuruSoeTpQG#(zlwjs-&RJqzOp=DnXcnT4=J;r
zTq%i!O8JV1LrSHtUqMiJFob0{Bs;Z=1ip${shmbxOYE_o>*0NSl@8soR+Ct6+gEEu
zSS^N<VV1FTVsANHR_TRJK*2$KETCYESwUJ=cr{D1^YA&xGy7?nd`ydBro~9h3U(Wc
z4QrKLTFqLkX5;FmqgGi<vVIM+YIQ?9l8t7uT;;9o(Cm1r&m43z`H_>WM#99^D8Xj5
zRH)BZ%9B1G&i&F*vx*Yk+3!Xo)mn)#6S9YN`#q&vrNP?LJbo}uXWKZmY;!NvXw<-l
zX%7?UuJL=-^qgVdy$=Sz)O9X;xx|p2=qT(Fl4%ybM>D&VH%MS5GW!rmj3tN?g_e3}
zlw--v^se<Y-O+}PL&tm%(xB$t=6je|AZL!fmU3sWkZ7YTlnnr|#7s4}gTK)u%FP5&
zG)m`hl#cks%*rogJS6k-%b+L9iKoNFB83wQycq&na;ToWCGB4Av6M8P%t3k}N5m<z
z7(g2C@{u&lFx}Z9=#A3!Q!;xqy_VxIzVV!!iR<w-PBQ`7bNI_kA#?AGp4ZhKj!kPV
z_I);9?X}*f{6CF)`=@_~-WqXAFId-6uZW>rao?z2p&T>yRE!*b_rUtEjLPYYvo1cJ
z;F||*OxnNMy+_}z*f;-nmEz10yKdtcY-p=77S&JN2*^Ri$K4Dn_Fpg>>FtftAl5M&
z%5kINa_%-7D%ThdE4k@tsFba{K(TK=8Y-1n*-b`cLtD+!NYuNaPkV94!77^OFvM7X
z<r7Wp&^Y7%D9B?*RH|J4VhvO~TAVssyl}L5Wvo@UiOj4pw>K=AyJ}a3^XE}5AQEnZ
z?HJI9PriN>mk_PzQ?Iw$YtbTHJkZrvD$IGa^Uf^?H+FcKL*t89$MwBR*u^nM4%oXx
zMnBG8clN~wZD1%*km8IpjBvieW0-lRI*t2;Fm;yD|7U`K_o$f3{7i;y@J{sg*q{^I
ziNNvyxq|<jAWSSzhY640$r|2tr~D=Y^#7^K{GSA+F1;fHpkZb?dZiNmP;K-yNC5k=
zQg>VN)9YbnpAqD0&$aWYprtsJ-t=hFJoLc_6%z+Av^XGteLRgOQW#=n-Z`E;n%Eh?
z2SR_d)D&W{l|7q5RqKv>+f<%6HGUduMtTsdeUT4o3I0HL@S>(r8^FjHRi87D3i2RU
zkRyOeiaR{_5W`3cqbDTOW*aU;Crmf2?hg>p*M5LaE2ZRpJQd)SWgyjc5nm~!%R|wV
zpkol~YIk*uwl^v7haKy#_BQ*2p;Gg`-BsMR%311MH`4T!zl^j0?s_j=Ma5!$@F5rJ
zQNN6<a3LzUs>B=hEiDmr<Hg2(-*=$aY<wS=3d{}o-%pJ}aV|)#dhzjtgMWPyJ?v+r
z;er&EYbxlQP0~7YtX1k;e+K-&(pgi!fV}kQG1+2scLhB&$<+bQG5sA%;kcEa3<d*h
zTgV#}>5GUp=YQFy$f%6xjQ{IAk2o{2V1Ny-Bdj=_n!xsWv0B0fIP|4NV|b&KMl4Kn
zg$=G?x=h1dhEJ-@L+Q(x3ev~&Rrd#1=>G3%b^kuW*j6X4HGz3kR_^~*<%`ese@1~x
zVNA32c|B^iIKR#}6*DI6Gb|%#Wb%)cc7eyz7MZs$<9vrb5ABLul-X%6)!BjxNE_#6
ze(*5``McE9#XA)B+{sI)o^yis-=)EDdt?FCF2ZtP1Di!BsF-phe0epOjtFgL^U!9~
zlelvHukgAO$3f+AqwHk%ev{?qf`BRi*YqLHkt(I8)L2rdty5PyY`tDAts$&xKg$BR
z7t?F0>Lj=yMBJ!TDNX{L(T+ySA@?i8#7s?x@s(_eIP00@+$VL!)=PEglzxgqHEr8$
ze;a+*A+ZvDr{Aae{n|$eI9j_O-aK|QI&)X&7$DC{o`v}BST$Tb4AWd@4y_Y@qU6R|
zzM-(uzMjz4Dy!H}Q=)*=VmWKE(&(XBsrcc@j=!xJI*!B`j<^HZFKP&7;c%<ln3E<y
zIs=pAZ&$7qdcK5&VJ?y^RZ4>blae;+E*>Vd?q4$*Raju^qLU&fE0LN;<r*?WiAze|
z4VM&C(*OOcy+pk;y7h=+T53P6snY38?AidZ^2^psY@_yXqbTjJqwW8xUiNhb?<hzj
zT3L?*c5ZC1ERRTLruEa!PO%=^I(2mY%)eVBcNgx6<6;tx^KPDj1srJjO`vkiX-+pg
zW@9knZ>oQi7lCays7W{K*EW>^JV*7KG-v=q@oaX~hI9Y;S`hE$bWT|FrW~JcW^(?k
z^svFMyh2<q_`S+i)2fGwnY!z-?N~c})5)WlHTy5C?_wG`52oF;WBr)6AH7LauJO}F
zM9haMirRl7&=5*wVJso6f(U-_pJe;e#%APpK%KX8$?dQu?^-g)ev~8ioG}hjd!}L6
znJfJ9vHPS0$j!;TC+^Qcs`9sXHZoHw7^B)H0~t&ug&AiE!8+T7qpeTE(sR8qx&?Z|
z882yuF<h3>vT6*7B_>xAXe84!*&rRAXv;(=+OT6Z*%Ct^vvXf|w#}@5A+0{8RHMs&
zM=_(-TFAt9_;XrfEj(g~!!X6qE8`~=+^67#f>8ysEo_{o$!a4R?WQTlofel5xCL-d
zMP*yT(MNppgX8Mu`oy9gnd5}ndI6vrEB>lhf~1Crpr`O}Z2YMVC*<<#KzwEE`jrtG
zpLDpmkWQchWzD@v9Tvib%$e$l#W6^)G3Gc@&=`xMl!j=;OJJBK#|XF<nT_%^{;*cG
zV5}7atRREg6_<7esoE*D0*E7!sY;N*xx0ND<UI<C3ix;(&{&6tn`6z0$d^<@)%t!7
z*dJ3RObHK$g$M--RxT)SNI~cR3A?zxqH#^lZ)o&nO@m<xygj04Fx#+QFTf$A4ZEtX
z<_5T_UG@~eo4NM^%LEwPGZKuQOo4TDeFvl90Kzw@*0ANhL)<>8)FsegzKR^q-ynb*
z)j3m{dSg5g9aV4z|5XDp$dATmdcnI@>;SZa5KWPk^jO9y!fNY8he^7D6NovPX}&DZ
zYH+7HS_kVQ5Qqu_znh3NCFleQ)`2AQ)3}!RGB@KIfwAnww=hXf{_UB_^%>X2;H&<*
zbGR?_7tC+H9HN<PE_a~|$J~+X9h;QR%u#tE(-nFip3Ry~P0rS9$52FYXXHZ5jhJ0I
zWQ|TuR1YJp<Z{gD{2T3{37oCF3zR(WYpi;5ZDvl@B6Uftp$1oqo!U3`yygxYwAF2J
zsS`yzc_XIw!{e{(dr_b-bLHO=<f{4r9HYs%8lMrB#OtH9F|D>*9IAt3kJ)PCP8;0j
zoGN9pW@pq+x#!c3=hNn77(JW8Hv@M+qm^gS&lu0!dfcL5K!HsB?N~wsbN!JL?rm(q
zQuY!cOOASei}X?-Pcox#xKb<SF3z*^n9}vTvEY)+lS|%Lv~~+rFXVOtzQh`?Y+nz)
zTE#mMvY?k%t?j@!iLCUSP#(-6`VhPpe+v=4um~6^i6Q4m9D@(6i9|ri)(dL!CKTXi
zr9UK(o=F|1;V&BTi_n2i#DbDzT}}=E4)~LW3u0|2C&YI&7v$6K7R$yIo4?A5;;Cb2
zikKlTT`StfJhs-`mIbJRBb*z3^-<V)`&(DPwJ@Ru+PLbW_GodU%^6_SE*l0BzxI}U
zI}g!in;ROM^f3#X(v>pinNQ8l%N>O8JoRY}KZmfN)v4QKYmMtJD{kR7N1=F^pJvS>
zw@h&bX4EZiU7A79Al+o+;?B07we}IL7B?E1bVx3ZiXV1I+9dZmT0C{s`9kD^F&~`<
zUu&Ph6&W?x=eF^6yXXZb4vejxfp-kueQI;mrg;`(=Z=G#DLZX@{49lA2gAkxEpbjB
z$o|{fCb9+!`_>I&n8eO({hk;uZ*t%1*R(tOX(N#@f~XE9#EK2xDI_*pP;FMdLtY}J
ztKCAAvDy;!h;{*|U;jS)0ou(B=TTjmzZ&T_@lpeZ=Sm!CtVmiSjE}-Vv=|Pmcq%~O
zC0D!5E<l!k$UU$E!1FOOgKlIE2>;}<SJIIvy%*oV_!90y1id}V;p&`%skQpZLnT4K
zFw}NwP@|HS^zgUWsM1WcuTY%U*M7sVIMWFFt*-;Y04CqGWyj!NKnfxczB<0v2Tdw>
zX+1n~2@kkw3$Xz%KkN;LP{DDfjSi`;jsP1{cpjF?cH?XH6n!l%aIFpbBVLy|{Mx!Y
zy1FgcX8rDxjbo$FP|@4CmQnJOt-&@FdHI__lGW|Oc3slAI(UgwvmQiPs&!`g*V>Q5
z$>4l$1dbJVAgl6Wy_L&nxF{xyZ?pJL9G}P-e7pbeBZnlrkqzPG1mUv2cI(`>xgB#m
zSLGCa7j)WAqw9AEd+>PN_m;c?7gtAvU2A)S-Ixu?6-jZ{$#06mE^}t}XK7<^TN`p#
zvR9_(Q9FBGJO4c#mUIL2ox_q1wiOMx8g0joBacT$<@-I_j}Ni<4R=lv>G;vMFteUx
zD))7s-raS|wnk6<eX1blVl3{gfBTdmp@B0DI>RI6ZNQ%rx0^w<A~AOUdusj#m7O5q
zdO<<i`H5?~o6b~U{?nwdWKVczP&T&LaVURLHOwoM@f&^OEaFNNIO3i)RA+WESPLlm
z#?-v5GKS@ipLs+5`=qMmbv2uF9o4NeW{&V1iXBz(IR(aG|4qgIhJv#MCZM`pmN+Tn
zv}ewcO&{Ywqu@_yeQckTTN?vp<U@}-Ec)HgD^^gTb4Sj9l&Jes<XJ8fsGI+}QlIlS
z7o~CiC2B@45_E<EUFJ2~PMg<CeQggq-E8MOOBrkCbZMlc9dqwu{MKRh`1>>-Df7aR
zsMq|Ax;3l}5jg%X^{w00cPEW?vs$E~EJ^;~urV?(2%OF{zcBuGH6yORlyOt(VE;)q
zd_)b4Q_QlS%w)cR(QH42Jf^Y$ds#c;y)bdWh@(`ZKRt?EMY$tron<YS_>-Ogu6M6w
zb*UlbZt#MV4ixUlNKluiDLzVkqse?@$Jm*KDG_0fpN4GxDH3YGM}RveL4AokpmQYo
z272PE(D?cN<bXM_10OMtJ|}KL>O(m4#59hXg1m8sT^!;a3%Y<fT_$n1I{rh{mP$Lx
z(PH76o~mT<SQ^4>J`r^BjQ5u7x_UO{e<0{V<&E-EHsQ#Zzm1ZzNdrG1fZ=E{VXvg#
zq~}m9Ng<jw7~Q`LR^a^?k%12YySt@u;+Rn_dhW4j%QGjKIr)_9)x?>fBLf~!Fc5yP
zX3aLYwmC*VwgKgN$d<<JG-p__H>kKhDb<bNuPU~KVrsCUdXZ?jDH3oyLQqMEK}@FG
zlppGxz(7N`*t#GSH(*Zody-qu^-9?}0ADaP)y0ERwCr5)lu}x!&XJKkj=*I&Mp-MH
zYvCQi=@$DJ)#)wjbSD?>Lq5m8PwVt3B~rSW54kU%{#acFp6uub5@FfCX_e}@hkhYf
zHfVO%Ip6ffz;v7<+e(52SGEN8;!QnpGUYbeko%FnxA`$;@pZFA^Qb&B{$6e!2JNzd
za{fI*KU((zY7MRqSJOP9)X-Di^x~YnO5<M+-<zvK_em~nc)D&vWtNj801~TP@HK<|
z4&FTT=FPK!Y9!A6%hH#}I5&nLnx;3;5#GPZg42nw!C|{<0_K^!>9)p#A?^fBb9-{!
zkZjKX*j>sETmGhPz-|=AS%aG_Bmbb*+93tzpTk^ajPbbA|FXL!xagrL<(9PLHu(OB
zHQ_>Ij<?*Z$Dh#Slf-aJX-uQx56wn6c%}k)!wL(s|5a}LlwrjV6*s=}hGLJXTvl%Z
zG&-5sHnnQb8%%b7M@?^S(`^2l@(np?EBHI~*!dx0SG}?TXFS2$uSj=^TkLY<9nFt)
zXZj6ltIIgoG39K{EpP#H{~isEZD}~dn7`XzrI*^GH@K{`!#}6>t~R!(a86_Y&~SV`
zmBtQ))z7+T^$+5{GzE-K#do3|*$mP~ZX?B3H-0o)Iyaf*w<b^%^w*?Zl%fs*xH^Cz
z1jMBN?&$B}4b(Hv8e*0BjJzSSu65p=8>JJn-AF-HemaJNPYTOZaxTcOfLJ%5s6(h8
zdsl}r{~gAej9PxZp1Bdni$ubI_EU0bV>Wp$hc?w=%SR08ziR9OfVYubO{<L?M#6tO
z9-UsChA|!pP+^*)uwKvW_Dif%2F&+rmOH(<M5UT(PlW{U*4>x&9>D!4gFND;0d`+L
z*n(m5%IXN;&>IXwxI>iBrFrUO$D~$u!(+eyCBlKZL9Rtf)^0|^<e^|I3_o8eQ8q*=
z*LmEC2cz}m+Q<#aT9NGZ7m_v@B3<o_g3jneFoN}mY#qA2p0^6dHf)QOtpCO#y4r8M
zccApl1zS-vL&)2$n7z6~G#E0UeM9uLKk~WpABhimedclF4}#$}m;_Udru^H8$VRS1
zwgnl;bXj{*?>b2B(hXv^<H!2joUl!?5}Q9P731WDrN4HT01O*@eUI!`>oGoxg&;X@
zjX_YnF-_HOInZkHc`jE|;~W8DAl&g9y;5t_UgI<`jiwqxOl=ixyaj&vV}ha9?m0#3
zSJ^#ZQ-8Z-EALw`){zS=IN=(1aYtAUAks#@H{Dv0=KPV~N;)ZVl@@VrF|p*&>(!BS
z4x=nv?_VVAcl3TyHknu=Uc9550aqURukJ_3@aU`3{Mzv7w3FSglG7|ShAFyX#h3{j
zD(79vDQ54dfi_t4L0XU!(nf1E_Fvx1=B3`ZM_z0VZ1I^2rjG`t&FipnD`DoswMmRt
zj4{#~76Y~|Tp@oz{s{#wfNCU<aXH(4kG?vop(esX{QZo!V=r2Ae6!EYsmKB4enqi=
zsF>+T#C$mIsi{!r^k_PIQ>GZZVjRv1)i5^>*3FMoZfu9c88+jFn4Ton8COs0K!Cm_
z?8b7aEPHDuaV><rQ5(SD?EAEML+)ERA7)LqTR!hTQxY79O^ic0qGCB+@?r;i#4z~#
zHBwy!rNY&dGPJ4B%B8B_E|tBT<cC#ASS8%D>AL?d=HApWsk9Ej>Ml#i0bsoi4KLlh
z@K5f}k)+NTBoc#n<ds<0aOr-B?j-toTxiRKliZ~hsACRT<Aty1(&7%VQl*~L(!&iL
z{4?!>DKVOjACY95&8PS^i<_V)LRhjqL5jNC8=<E@%-hUuUxb1<<d0mX9|R<GxqdWC
zTpMg94pNhAm?aegAst6|KwPIupw%C%bGoFS2Z1!4cXQIj5J+1pRK3$Gz7D}g8@1;R
zNEhVB+Zj2UEGde^qr7hUgP@n$)gk)VWpbgRpqF?z+(K&AWJLaF=vgo5CI@o45+RrV
zU>M|rwo#tAAQ2@AiU(UjE_UP})RZ*Kf?>D{Cu;&#AX7^7)vlL6iEIg10l_fnV+8a7
zsVTPvppSk-AN_#;o(O&Xb>dmC+t9}^nk7wtWLNVmLKucVy8JJLvZ!Uqql|sduZyEG
zlCdB1u%``)_<OaJew*2A`CWC=i^sJ%pCA`qdJ7h`w)|GN7e1$|qJ`SFyeSVh+vv6}
z|6BHWL>IR9_#t)wed_9y>cbO?+4eNZ^`w1+GTy7$Qwj`FS%-J3ZgU6wXO!EydHtuA
zX6VBLqX6YNrjTa2kJE4}<E((~8gecsmsO1R;%@d}%5Ik}@~orntv$#deMk4sHg+Gr
z#NQCVw3!WYhfj6$r!yrEsPQ8dtY`h!G_bbdV?O?}Scv*HTZHM<Y9XTU9;hdv?>MIA
z?~Uj?SndV|_vtwFo$_^p=!B64SdBMw*U@)uQpkk@OuC(@H8p!4^<JJOmgyuL7{w-a
z;-BK_-RKHmj@F#C%?ABGJ+8xyq9e{S2mS+OuH3gCZaK>OUC9RjK{YE<!8b76nqScf
zPpfVDky_axQ@nfIcO?b<NA<2I%J;uTkL$ijxc>cm>tlbmw>D9`bm01m)+X3w{Ud9f
z;RVjeeG3^Is@8hO81<0yZDSeD_RZUL+~4zhw+&tso0*L8J$y@Tr{-M;zS}UB96ZP#
zOdQk7$mTQGVU*xVwOH&q8mnC(0HE!Y3SGw!qZKZWCNZsZ3mF7Q;&3qen#MT6lWlie
zq9&+!`FO=f-2fp1CU|xv1Rg8=^U2N1u!;?;aI87Geq{8;D0K9JZ6kYBLs?iil-y#6
zN-HcVbmB;}?cmgjCD`2pz8lOd8cdNB>tGJI%vmw%QeV)(?&G}z84?A8)Is?md>u8+
z7QP|B7#2@%)S9-M>Cl>HXimR|#eatg!y?>!GQd$8L+xxp`d}_OKp3R}LTczYbUg;Z
z0!;$?5n1b`sz*3Ihn|(xd(eSuiqNRaUZhz<EEbkQ*u_3Y480B$L8+$ofVIR4yDoW-
zz=H$^RQ8tXdN&raq?sOPfWuN8B#@=8rczD_8*fu8PJUmK$9<JqpD*tdfGIpe7)|*f
zG~`7Bjw~*Eiz~gSq$YGOD2e$06Ipn%vyCHA>x@D_qu4)E)&EsNOPKyeJx-Dv?v~MC
z$9duoGf{UN?pM@~2|%Ftq!zHPs9ne<dM1aL(X0O_YSn}xhNQFt-7+z(^c-O0buA55
z<}5rrPV}NAZDzA+vN&O;ew|B7j@$v;9ACftzDaLCsvdlU;O*1AWU8N_@pFubHmV__
zzeYcTKHmfvRSu~_*~wm7cW@@0yGxVq+CewAs5W5&ty5ZBxz#PJBdc3kL0jhVxdQv`
zTwu1IoeRt@2tZP}fE?j3Gp*GVS!n+iw2IhtdOE$yVW+tQn!{n%-_~F&WwVdPs^>}V
z_HwMgMOKsEgH@UR<p$Dei4Dn*QpoYT&{%du#kE?H-8UiuTKIvLotwAe^}bsR*c}X{
zSee(p-&r4gC|1p1iJ_s~%1P8OHhRiUsCt$S_5i{KIn(UpTe}et4Alp5^f?+}tsOW=
zpJ)~7)v)c1hfNHDm;<VjQ~I_49iaqyUcb?ee^ei-OKZ4YHrXEb+uk5wCk^B_N^M71
zxdTn)cFOFOaKbc;yMk@BvBMjp<j!DpjcXLp4(_~ZTER<yi?YRFl(M7sy}_tA47I+C
z*e-92#r6fef>Af0-Uw9xu3#76vXyVq!I$quAToMA{@rL*c>jL%u?eK{9CP%xy_E)R
z?_S;0p%cxY;;N<E>2|HC^=3Ek=Uzlw-cHK}T0TemfnX2mdxO2C?;t&=H`c{t9I79_
zuA6lmPe<Gnz1c}`5O%NYP2AgMy3x)1v$C;2jqz;uYd7!j@%Gq#u>XX<jq#<uQ2qOO
zlh$w1G_*e!>_g-=?CoXk?fblPadbzIV5d8Kgf?TJci=7gq1eBAz{X%d?>ZP9<XgGh
z);sjM)E5$uad~x{M%j(tZ+7%H|Jj89DTL2%$LX(jg#NN4lF3AJ_0a0!)g#REJ@oi+
zaD;<}(OOv@^9~1NSXU$V3l4in*eiwLFdPgqQjSC^M;a;9NyZ>5Ppz?O-Q+yR7v94c
zA^>jtLS40?^=&S0sdKM=>F(778@?w|d!*xgv^IA)de|KOqk1zVa||y{dxEi1_uk|G
zP48X)Grac%@9koQe}S(&$=i^0@Rj$P-N8{}$MicbG43h!Qi;`jR`1>93xBfnO>+l=
zd*z;x-E=fK%2(abTe7_6pVp7Bov0%WUORac5{Wyo6Vc8C5|IDvO{cgIcn{(fH+zOs
zkyG4<#_>b%|AD~#A1e4`1z%CXZ_L!yVQ;6p+E0LZ!fu&eHjbSmd;9mQ^ScUuTooT8
zczYWG5<&BbC4?JxU07b2;DRDM(y^uEBtxSqr+ipdL&CvMYhP2Wu3%Zg#|XmQ%Oa4F
z72Zm>k!sJ-H~;<WV3I@hE$6W-I~rbmi{C2mHFnhbOvMxj9tgNa1~*&qUsYS;DoewT
zisJCeN$31V)lL6*yN7dTH687*9-HvExn!!$9~kJOM&UzLzI~|CQD+uqr&?yGXR%J4
z_A8a*%rb7$Id6R{d)H2cJ39%AMiq37c;72+@62tVm~gPs^Qq|>SG;^XdHnRW)Ak6n
zA7ANj4$ZUGC0sk!VKCN>hYf{>)rro;WpgsuDAbzH=Bk!-Kg!?|n8;yr@dflP$|?;t
z%1Q$h6~4WtS*}e~Sz|3Z%S6red-Td*(RBB<%gf%)+p_;1s@S5HjN$uuGtEdRF?@dr
z$>LupvXLl~{o=d0S3c!D(~5jv=V;lTMYrzHDc1y<Mk-F~Q5(k{I}#Tk=ANrud)oKW
z(Tc9}`;?ZcFP8n8Nk?veMad>wF%gDnMJIFl8O2)Er<rn?fb>3n=f{=%cNP143X*!v
zD|lA#|NBZay4IA@l3Dwb>G`53osq~{#ai5>OiKJusM^0EHg?QWtFI{GlL|hiK#Y*v
zcsN^x?a}VLp)?c1qFT0T4+oZ58x8HXxm!mPs-$((TMLI=$=T`YWf>J(SzoTO@46ct
zSn0vHn>q1E8;*w1xYb{y-(gM}>YH;%Vsy2sOmC~3Mr#{j$*wrNUZ;VSS}-KZY((&p
z*6femNH{0pJuB@W6y~{w*zFASB*W><sj#cD{@I4UDcgIRrOHY($<%MEBfq7<#8?tk
zaV<wnj9-f`43RHwTi~{=ZAE{YvX@!WwSUK__GS=DNjx0=>;HOx+8k!V92jc=t8koO
zl{jiCsYC(+l@7P#5E`MD%tu<Dvvs7aKzq80##$zhO8?I8I|UBHyF+0kSSDt;t3x@7
zc=tQI?+nrlXNV&I+b$h0^&5rsV0`y~&gAzYBs*<Pir%Ky2CWBjeAJ$zFLBrJsnuS+
zF(F3_=C)&lm2hkCb_i#(42$-U@w<}CfPp#x<vhCR3>jRmL<;NEbmk%DpN2cZ#$+yw
zZCv~d=*qz739FlPb)|k4&rGRSit6TNIuQBNN?;w#sZo>O%2GIr*Kst9dJ`_hxZzav
zj5Uls>&<=siuv5?zZ|QTC55gJklHVC2(5ER9M+qIaKM?lAzgeUSGZpO^E$I(7w~eo
zle>@-9$BIg%mr2say_>^WJMP60B`e@WoQ_?UguJL@5<h$+h6AvtGzr!eHqi?Kc@xz
z7{N;M<%&7VFD@??&zY|7+)Kq;<e$=qt;PjZo@;BI9?#z=F!LgF!Ew-hvluE9IS7$A
zxjYDx>+Y*l*T#nZ&*(KKzB17IW$O7il~z{lBMK%IOe*+<f{!csu!4IPe5VS(rC>>q
zmlQ~C=hF?KL!$$dt|@g=CCz(m!3{<kD=+`MNiAh{a7Ry9^aZvPEwG*VSxWqgcA_|D
z=U`oXO7|TOvX0$>O8KaF2PE-{n$$i`eFNX7?3j%<z2I}4<aK&*?>Mue{tFq|+ScH>
zQ<83VODa<vttZwne{ttE;4yFzaTA|*I`@cYwM-pOgv(F0dO5KHeTn%BPdIsyY0I$M
z>$<7Xo_BCB4U{zbCQp$C;5?)p9&APQHEbaFXlQOJK?($<ji&80TxhcaYtw$j>Z2-t
zL&eQrM4kaJpY{x&O2vECPj52R8if(Q+&RSqOzdWeWq+|z%IMJQ7+2*yz)Q}aA0M}V
zojvP+4;_JCEAdAT?6n@<H)PJ99p~Rzf<k{@y%pFPOP?hfYa9NMZpx9zNGGZHJjJ7S
zjlAN4pRum9<GfB;s^>i9lcEF+YMK(3^Zq$~qBVB@{Mqx)r#8+7kGggJed@wrAPBQ%
z92ZY<*7$Yx^{13xP|U{F3`76A(#&<`4=eVg3VuX^A^)FN%%oWVSh0Vi;GZe@cM3kJ
zz^I3m`eSFPzfEc53e3)}Ul~IRwkeR|qW_l^j3_WAt&tF8dNjhoC=Vkrgs9AizOUKy
z|FRxiJIH)w+BQN=61zZ&+E4Onxfn<#(?Dm9zrBBudn|#T6!g@ei1)se%j4ocU)X|M
z{4QeITyLR2bH7UW-^tvMT&ycoKt`6tHGXeyAeSP4IQB7&BbC3$wc7noW_P3g?r&rs
zYKxz4mKXpb_23O=w3QqC7?)M;9*q0y_Y__3_yHW)Pwgas8m{$QHW;O(T(VS<Syobr
zu(|aK*YRm{!LGA%d4HB_9pYQjbV6tP8(lVwtKIlbs1>$pLQ-%%PT{yw7H#GqJPm?j
ztJ@Y&f<|@-jhF+=)HwW&!C7gn%T0dftc0cec!Bu`_f0!^fXCYfXEz{z$Nx5_2#EQb
zirI4fA-*49sq5M+i$QXZLUFd#AnBZgP!Y+Iw>bR!lW9J_&1pDr&(YAfE?O{FZ`!Mk
zuXV5E<)RQcGqkOVtzPvxu~+e%4xz_eCk;DX@`x^O!p?d6&H%~a3dIQ06=0lPcRA-r
zmZP(e2jpNTh$FE{ZjH{KqlQpvoa}X!cGjy<5ViXRf#vRn{G4gPb-KdZM>nh!tma_J
z-l3~WzkRcJw6|fkdEbWGS!%!Oq_4GEqZKyk?dB~}OXne4J;~G7TGZBeyS7?i(dz$(
zdIjqJ9AC>Nd3H@|<2xl&wbtc~pL=c5OWOi7p!+1FZ%YlE+y9MeaVzd)T3`v6uve*C
zalY1}Nhm0*R5*C!wZ21{1_ysbt-YbgpI5%Sc;1;v{;?|WQBYJMCf)zfDlQ%%OiXQh
z+V+T++ZoE6SI1q2|3{Mc)b_$xg~qYSj;DqT|3PWQ`U}4=U2iu_GSyZ1Baz(1vW45G
zGfrjR;TEaJCTJ|w=x{D-QqTgoY^rBXkngsZ5kCfn?+P^qD2nQ-)^rZ<RXei|c<|Wq
z(}YMm<Jo{n0vlfP^UgX9X!K?Q$RubKE2S!Fo#6Hc4Ql1N@B?$bQr4w%sYFk{#~bqc
zy@5A%RR!mWQn8ZM!oy*t@Ox#+gcZ&2#zEIh<6{Z#z*)Q!2`1#2qR*Lf&U3%{yxd+)
zF(vjRNT-hy)Dx#%dyqiRRoFx+OyAPd_?eh3;rD8(vLV@|Mr;f6$=_hCj8Y=9T4Tku
zfR^aUa7(-AVC$8+QvHOM$0rFQHBH#x>aiou{AQ6hANiX=tmiZ_f><);7sSG$Yh)6=
zesr2}tAo(fw`&}p;`eL!6QCw*INJsG^L?B7dX}?YK$ZD*=Id`I#?RhzPMghnt*-ZL
z%zy*PW3;@^o2~zM7@;;iZ7BLv49d%V)7K=;kixg&u68AI^swQpHXY%!9RRtXwgi+7
zv;78<AYJQ==2>bZW*gN}3ZZF+8=D84IcS5WShD=2sg}r>{+A;QB5V;lBUnz^F|;nQ
zq7;oBK?u6ACxx*C1jlt#i|#nXNj*#&Z;*o>DVC%deLt$njM<NO1NgK2_ITOPbI<@2
zr723~IqoHAJ$)#LG8`pC>uh$UeL>GO8^NVUhYIB6(4?(j(WpMJ1@@w1->PMpI{(6%
z7cg<`1W-_lwp@)mt1v8X(P@;NAjp8;d9^aC;5!?u)+roK@|4vg?r)Rc{H|*Mk&3G$
z?wU|7fx8Mw!t_-X(pSR7%GI!|ZKsBL?CVU=S(Erf3ksR>X0ByQ+5<)2n@ViQQwhY|
z9o_gKza^|6=s4sG$OYILaFBAhvtbQMV`+z~+eqiPzpO72G2UUSrZLfe>*BZe8oMXx
zX;V>R<!>s=eyJ!?Sh@o!AZx8MTVpxdQMXi-sJc*3IMI4jN!fUIO3EPw{!&s7^H)kr
zS^j4DyOqB%<!hL-x!aOZS&oXG5@UsO8$&W(YCuWUjGPG2iu)zh*s%Uy7i!H5immnE
z;7~ILXIp5k8@1-Mqz;%;^3$;Y>WNg4`rFr3YYs#y<aqt4Y6d{Ttrm`03xlXNZ*AU!
zH}tOcbEqJ<ps@VeMxW}#_WVoiBHjXfIp7WRRZ`9Eujd)(ZNXqufw{dt!uwHpVutq!
zN=F@T9E}v1|Cc>&^|nZXN$pI1TYdW)WlV)Rf(mmhU&1pgOO7<6!raa7-9ec>0SZl2
zma(~_6qvzIZ<La#GuQUgD-@GAoeJ|GQFgznJGa#j1lzoAz=Peyc6-|`b}-l-Y;&{i
z?SLuX%`=KZzD6p{*RXcqg=%NmaLysDuM(KtNrk!FRG52KiyKrr@!I=3RXX%0QdQFO
z9$J>F@-=#SI4JTB`vNHfi==mXdrgJ-pUFGIo4O$*_uINXa3jX(z4qxi)?WK;q~5Bd
z^131O@4VET{+soC+%u<Awx~qylZGGn1H%s7h(CsE_0l}G_gn2BZPwmCcdw~Ozo62t
zg;x1yDff<Fi7d+wcn9Cip=@rxQF9Cb(Rz>dto6(&9%2;tuOCIs84q}eP5pOp^^ncm
zK}O^-s>lKF2;(t!S4HO^bd0eSokw|#l#M_}Rt2gvLaE@6>(|t9_nI2+$m-aJIZM=b
zcC;6plk$_B#8?sc1ozOJ<IIoTnw@}GPdMG#X9ax$sZN3BkMmec0T=z15rqPd<@dJ1
z&)V(yu_mQMH3^+{N(^)j<jN6%=}l?pgb;CE;o<vcCwGushnWp2pQ7p;0<x-}Yg)=}
z0HZBAX`9hp<EX5mix>0}FDhti^!#tt<4X$u1Ml%)R_we2qu{@}*3N%Lg+HjkET7&`
zD}viCQoLgvC}<R)<aBnKNW;87T92&q$h^Ke49vlKH7JvrTM1`AV~Ea5*ci$3?w1%N
zXL<L)T@cOYBZN<2yYAR8=_5X@;B^Jxu0R(n`X5oyR9yKJJdU0CGpLNh)Z8M6^Bz)l
zBlTa>XJ`eqc^onU+N9Dtl}RlXl9Q`^MMeHUQ5~5S$Mi>L!1N`(s8eAyc5fSvv5AJr
z_w<RT9GX^aMuB8Krrz;q6`ND=yjF*FH~vLE{+wc8QrdummfGffl=~MHe6Ip&ar}A3
zjMLL*@)z`2RbcCGQ8BqcFcnQHvpiL<FD&_&lp=Ys>1cfIA!jaJS8Q2<t)lN&%$AbG
zSxee(5>xeZ989{CM116q++IqblEXREdacV*qm?_gExR$-Z`>FC&9yDkt7OoY6yk@B
za-uHpJ3=VVLK>KMP0M=y2@Z~;TV~z`EZl+<)g2gZ0?NOr&(>}oZxc@zHsw{t8?*@|
zjchg1-(6YP>(uywMsk=J<Zy4swU)Tm&TMf)w&8btm6#&4<0l<G)jAp3=B<?}4@wNN
znqz9k%=YM}%Ih&N$4NIavSMR#OJuh+c5}C4w>RkZdHu%77L1YY0=f@{kw2}ovq=7z
ztCc1vyYgLyV&mt16Sdh|5ndi|DAUt;9I4ie)2_k<K5J{sQy0w9P*Y=GRME;({laNe
zIJa6lzbNnNx+9>`fK)W1&S?fuNl~j6^E)YzDdRW`vg)@IE60qHJFY%B<*9o%vbj7t
zzc_WVcKjs&+=>0DxRj>rtF+fRc``cSFROv^=7oGP>9r0HP84^OOMin0Jd2|?vzr<*
zJ|T|k?7Hq`%!`6Sl$L>i&qn;a<JIBV%9xZAN`a*t8c~ClIq(6-#uyZZ9Cp3eF$wzs
z$_z6YwP*ivD20M}ubYXsJDs?k?X&d5E-zde!}*hQ^VAr^V&slzh9hX)Yk8qEb#Vzu
zT*nEFq?ryFH-!aZPotg<z5j!($}lf)^*D@QsBzt&7$IF%C<J1sh&YJdo+fVuI`V0E
z`!$S*)r|i`1f@Z1tN9W&>fcoE7ZsdSj|?z2KWROmCFy9Woh03UcNeZZku@~_Z3iOR
z_3jaWb$0!7Zi*hxPH7e!KkBdu5y>8Ebvl956vTSQDRH%%hf#2la>`J+i6OXjnR#Kv
z+enqGxc|4dbB&GTs^a*1*Rj`Ld%b?ddDhuH8hb11HW5X%P})F=h}v36DH6E-VtXBX
zT|bi9by`-NHUw2F(832IBuY}XNIZgQi;4v0jRYS+qCy~q+T~G%#D^9kR8*>d2$<jh
z+?k!(hn+`7tlXWs^SF0r?!Ete&iUVSn5o{20x-^RznI(5gCPogH!6@DMBa|4(Pl75
z!;pDv?8GiyhG9<SYoA1-W{dlt#7!-VQ}tHvu^ky*^1WJxj@~O{Xr&7yt<LY}xAVL+
zj>?&g!p<d69OABOIeA(Rzj7#0!QBjMRP+OnK!Vnl{B{{_4L5=n<hK($Dv9E0j`Z+c
z-K-vsFRO07;XXCNodM<dRFLF|1)8p;KB=uVTR5R%tM5Dx&@j~6k$i?#<rg@=zmlqj
zv$Hep$O|07Jwb3pS}W7l4^DVCD|nHw;Qd;OR3v4&#lYu)MhLoNcC@@>)M7BGB`K42
z{tPsQIlSyHW{G+<HZwo9j1%l)F6y;uWtkXvxqKmS5nyP5`RvkEX=Yw`Fcau3c3BNI
zuC(#P-tSd|mz2D$q^jf(N@yRka{BH9%R0IEPKJnEvgQ7dEaMjm*916gb6GH~O>gl>
zU1M{eKWXdFO3bYLingq;sB3tC)$UVlRl6}a;tY7PfM4!cmAjy1i6qzSW*@I9L#@@j
zL&@Kj*xhY1_z!lg`0aSW_sjYl9sZ}1H<bKai9x+32K5h+_qw{(?QF>RhD^BvaTy?3
z<8XQ?p$I@CRiN#~JyZO21`-7;^G?hAb@$pnqg9wLcrA$J)ZG0F%Y-{Hw^Iy5C4=94
zyc+V>QLwM0A{uS)<bP2-pY<u&r$iN&gN>+uV^cuBhE}{;{Z$Sc*%G~;t$6&@HdbbD
zsth9q>qtP%4)&eSV;IV&({T)nY)uJ`#Ox#*jE|8oVm3MfN)wH_%hzD^<h$v+Wi({E
zFD3)~g^nE|X8T9-y=XAfNb}(xq`~OP_i=|OTHhg`@$bl;TU?){KSD)@g6-D{O1Ac&
zVU)>;;7m>X?Wy|BEc9K<LxZszEB=Br2GJYXxIwNoNCS_=kC_`)t{UJf5w~wB4Wakg
z4!t)ZZto~*(I6|o8N+TyQz2c&7W5vNi;Az=-f$jQR3?uPl(v?(5e;~!={<(fdu-<U
zxdvTGYymp<Q`eyP*vUB~<_Jnopms^?0Tp|gGe;nKXRW4?rKJr&@iOP`HXX_K(w@rp
z{1&twyV%+VC0kp2D`-&M7l)2b4*P>fu$JCq35j<Hvh+5SrO}(9Mc9nzi=F52L*SFR
zuQ3fs_SEhsX}Ya=v|gI#4F;Ny!TbnsOq!x4>U(WvH!WpvWiMq*GtvuHTIKWppW07J
zB2iPisL#@4xEi1-=+Ez_cDeOy2ZefnevfGuAUNBb8KtCLB}YknDQ91^W}z~8lvwU)
zPM)u;J-NASL#h7_L9?lCHT*e=wxT@A|4v@J5}6v$=)tVP`AN=Sk(7<>8>*07N!%?2
z*xcx4$f;K$jfT^gpXq(rNdGszqpOrqf3CYQ-Zk+WC7kB2yKM1uzdp>YlB*O%f2E4l
zOQWIyGyA&=DKv=`YSwp?N4dvjCpq%V)hLrTAd>{E3=Fs8W4sba{u{v~k4Q$sr;WtV
zq4*v|R!8{TNCaid$e##!-4b$5fRzy1ZsMmqYC`A5?hedbZ?$(1yxHCzc-L-5xlp;n
zn!5A)0wz^7Fk))3H&Mrk>1h->AJPqalxXhj+Lz!uB7tr!e4gXl-wE5ZTxD#N&Uhjo
zhnDpg%#yXtmVh4;ZyWlLr?Y1`vF_g{f3GN2SZGmof@r?S?547*cGYaDy_SbTAZkms
zJf@g<HH_L((}5lJh!Y({>GMSs`XtoM;(wh9nzy<v8S^-dF<g8GY-^vsH>OFRn0_=b
z%gRXtmFITic|FL)j6*eAf6>MKnjN1TXC_#@U$^xy#k(aQMIS6>PZp-K)@ZL6#l!jM
zyn5L+%cymTw_EK|c{o8fso1-q7q0#u{L{;>$(RS(3G9GiLBe<X-5!xBkt*fh72eAh
zX|qEsa3Zt<S2NKvCEWROYE8uRzCj%#(pL-nZb4m;Miko)DG4@;Z`*DBVl@G4qS^8(
z2}{pelzTAh9P>JnQ;>?Fd?GXJ8OVjor$5P9^_EWTnyk@Ay2<J(31?{b%mv9^A3!=$
zC~oFevV1NeOUpR|R!=NY*g55@q0KuC{smUjUfir9Yb{s^5~!tk*}|PC?hpujavxXs
zEo7MU^x|iz!@2iVWI7zRaShH5G&jN_8_3#QN#XjEZRQ^Pxo$Jsdm26GVtwU%VN1<-
zpNr$cYm@g0B)vXoJ3U53K<yb}H8oLwS4m-8O(CK->O#5}Yr!gu&Ycx4SlXAc-QUw-
zUw!qzYinyxg{_WieyzN(@NIL$-qRY*8_cXGe3)=<sAZsI%zG1DLW8U3;%?!BHIzrS
zdo%Z|ri{*dd}R@HrVB>$tcp3K6GT$kz#{9`<(S5^1<-h3rPk0@?)T2HG1)#nGyi~h
zmd!?efGOY9HdV)TLc&UBEuGLHF7ZV0;I=a+pvZBcf}RBy*rxVdv%uD=>eU@<T~hAy
zO86hbXc1SVMZptScVs28c+K)h#57s4V+nDHYc*J>7Il;9ZJfzof{7I#;(0q5CiO%$
z^Ef`J?E`&Hxi+S+_eEINv|rbU3KP>GH}qTG;xWCO0nQ!f`2n#tma4N@wqq^SlQzAB
z2!6%Aud6av&o8lcp-CfZ@MOqjjphsb{+FQ<)iX{9b@M?9nqx;7TG4I8ZJv(LQ0=;Z
zH5j^!>?jQ7QwA;a?hrzdgC#7}O%No_Xgvh=Rs*uV=_`*cP0dA=pa&cb1?VB)=}Rli
zD>JJ&igp^#>}(Ndo#uqd=_$dtCDuM^0nz${`#6u`5ALF|XmP(ku%lXHU_>pbb#$~Q
zwiv`{If)53VypQugx558j(?akb*tkTR{PuzS~}uAh9mwYk%!UAN_$5>!;OfDHVA~R
zsi7h1^L{Y^01+6{ZRNfGYU>v7YdsL?=QcGE5UHL4_iciJg9H;Ua|R2Y5m-2i4^9Jx
z8U%c>1q3KU7E~Bis|@3W5lo4sz?uc9E)W!;QHWL}rkL>ub(|PebNEvJvKVFEk0;0h
zDxx_k7jW8$*Bb!DIOcu&&G&4H?LbV2UoO}f@L}Tyt{)U-2jM>o8-2hAMtAQCfFe@|
z8ySO*%tZO7D}jwb6F(QA24BOhpViLk!ptiF55rWo^^Bt&s`gGT7Uq^|Ac`zn^ls;S
zF@bqMKh;h@eyMTb=Pq7&{kL~dfA}_M>pa8>4s(VsfjLrWx#?`U+Ut3ZHa3vdK*sx~
z@@`O-+NTK3sT9uq8C`7{(;pFuO#k(!s$ldZlUg^#81EU~T_98K@;eVj1e=tz9_T!&
z{%{MG2th}xXc5I6>}ZR{LK<}O4KmUxCl)*FL1^CweW*1MQK&HmjU{-TonUSlf)Gua
z!NcNQGMiG^8#L7fv?pi`z(k{El%&zNFJOI$=Fn(~?t_k*g*4V^S+KlS4VDEFK$o;%
zg07?wU1|^ac;C_6yP!{TD@k>ey~2PqZ4hQ(^xHbkKI;qG`W^`<wSur8>iGAS{6I;=
zK%_hVL@z!~X#<4dvvnuDR+dEs?!H<LND)MX&!Za)%)IB6gz}~pHl~O976;A-<p`b9
z?H#N#>&SvkI<`JCeU!_i$kdhSO5~j|<O@hX=@dJiLUHHQhFk&ZH0g3JFO}~xl<G1&
zT{rK(uR2m^&u<bv%l8@jWcgQ<b%a)s)~?47Vb&y$?3e@`t&LhiWsbze3B{Eqf(IN2
zz)+$ac_TZuI;V&nhG{Nvq^*XUm<dprCyl8+)+R=x9{BAM_sL9*+A1AUIYssd04-s&
z@mvY6quC5-am?y>Y51n?cF66D*-5aB)ocSk?3G83ph2dNg|c+-L~7u{jC<=uz+eU?
zz*11T;{58dV~ewRtV7lB;#L}Y(Dd&%`!!pxIIcl^G$j178pfkajQ+YIFutgE;$~RR
zt)05RtqHtX9^LqJ)qrwcCam`<RpQf1KBMHP8rs4ZYmHTp@Di2q9%rNc&Px<|nKacJ
zPTxU^|1X^O<1$Kw`w*B2?G&?G8W*c+#k6BfA|G6A?`TOY3cE_s1&c=v=XH(@VKR?(
znNK_U#gUjdBj>O_88%%>2ayW=yQR5X30#J92#>!+gX5g1MTku^TS%c6XAfcSKnb_2
zhRR8yGBv@_<$X2~>hR{r8pKn5n93T(yeHQe^WJ?Wc)e<*a+>+|$4XJimbg&fH=Uiy
zjvYUcozL-ekWjJCy^pfkoyleA$%|Lx#}DxTppyhYrxs-5;NWFLvq>P<9gG(9933mN
zINLgtb&)eVU?cq7m-`&kr3TFFLhvi<c});3a}?049;kJQDD>7?he)4PP&uu=l`~^G
zjW;=P>=NCjH4N9ugYO83y{IhcgK!`VMx;MX_IyKs6{bEX^Ddjnhce%@`cy5zuZ@0b
zYM23t#DHGz64j3YU%MN=S;Qyo!~Q>|I-I3=TY2X_q#{j_voZ0Cq`4n)?JxB;9KQvQ
z4(mSCYsr+cTy}$+I>Y<Cp7oF`N%yZN@#H=;Kfb)|mPZ;NOe^wPO8;>?gb|eR_Hw;V
zGhUqB>dv=yz2k5VgLSQ@Lsn?g!ZLx8o%USc$VH<>mA6=!_JmuWR9K1!v5XV>mBpe*
zrDNV-lxW18?6(3=7PKmVe6>{I3=#ILqfPa(qhJv+jNq~mAbFZFs(D+iNbtL**|{l+
zR$fY1Cl~WecD~o6k_>n>gE6-Vj@>5N$P<c~AUD~kD=<Z`E^>dkE1q6#bu^k=E0wB7
z>YzOr;A)slCWrk?u!oONt(I2&cwf#df;be_-Ss#1Sv5h~dI#P%)oZ(w9lGbX$USQ`
zoN@C$uv5of0EuyQu7Hhqjj~*9i}4e__W;iu(cS)~#+TF9jU?5+yMm8>w>`OcK!>#G
z(AN1FN6PfmrX8?}m}^%dAljBtv6_6ZdxZCKd8{~1(jH4Dol#p0nQ%SJNZ`x|J#Ax*
u5X#fSCjR1TzjADeL7-!(J@&Bs%e*Jpa=!z&Wv+`ooJsM&FLQ9{rvCvNgKcU6

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..89c92d6c84b824c3d21fe3068033d1ebd6c75255
GIT binary patch
literal 58014
zcmeFad3c@IecyTa4f{e6+%I=R;;vnhCCj2H%aR~PqOAB@$~A!Z0t5(LfbYHF0>2n5
zrfk!7lsL`OCbkpMS(!<jWD+NBmbOWkc%DfnTc1wnZ9CIt99M1nG^w+=j4jUR`#bO5
z08mbsdFG#j#KC#b_MZ1Fzy18qVoy&Y#?RnaXDjm`jmQ3t9@>A>JiMFV+Hb~VF)vn%
zd2uf>A1}r2Z(<=?O2%DUs+6|0OexFX<a}-+U&>ovYQC`0Rq9&kE_K^;dcJ3&x754P
zSL(Cp%zXdCKxrUqXRtI#{p|eE!cb|*(sJ{|3tLKCES{eqS=d_IYVpGS=)$(rHj8)7
zZ(rC^+Oe>+v~yusY1hK;((Z*lr9BJ9QgLB#Y45_m(mpHKJ->h9K<R+Rd*%-=94Z}J
zI9xhx&%N_U7RE|r3wM?7vgf|}yBCg@jxHQ49a}hFI&NwG^CuQgmQGrHVE)v?>C$P7
z56<7SaBt~e;zQoBx5XRzY_fEpchnp8w!INA-S2JpcDxZQJ@8b_+v)9kBj)X@Wal2N
zoW7PSJw)1WZx3mEDi2*tmLBH0=<VfsugZ|}O{DDe_LH)|@-WYj@O;2K$n!yaeizS&
zyu&;nuDpx<M|nQtjqyBY&yVqZmv=YMciZ#3y~Ingv11cUOljfr?7UZ8n7vY8_AAB3
z>B9MXeW~{Fi4!xk_4CV<$EOw-PRuV)&3gXCY_)d6t5g?fYqQmvW8Uoa^m1)>v3hLj
zYJJh4I)7qvesS`|0}q_Mf6|+tx_|n<(+^DF_n>#rJ<|`~JNdxmeWy=X?mtzzzjDv1
z`|q8+r`1xq*;2XDQu8ev^NufF-K>j?OO@)x?1|ZhC4cc!#cTD>^{VpFz4t!&(CG*7
znVfv^f%_+>PTl*!1Kxd}M{ggRoSgJddk;Ny^7K@t*+A>XCuSy=Yqi;lYAfjueRyJV
zVd?#|wG;J;1r1Q$pQzTR7yX5bUn?9tR^;#0Y~d|!uJN&Wn0t0&VbYs;_QkP8nD;9a
zKE-D$VaBg4RVL~r_D(HU>y<0@g^B9KOvMkos>}29Q5yBLOA~&5c4D4rs(#U1;BS6m
z;zFf7<Ij3L^wQpi@=C>DEH75)uaZ)DcD7c3VzD|sJ2RFGd*>Ht%JnPt@<gpXyD$@G
zDp$y^g;}pMJ+VAr57V=?g^8sw%XBZ$0a~vx2-EWu^{{(tak*MA(}RhHTDX;kQJ<*N
znR=x>?Jq8r>7`y04>Qls)}LIM&?i<IiHZ60h52yjnc0b%#p=ZTQ_;wt;l=*!<Z_+W
zMA5;gng9B1{p#6SzgB;#u9x>bzFc3dRHqibivP^BW65yPzIC!vpD5R+&R4wU`ARs%
zX!#RU_3~`BzF4kTs<lNwENH=2%JZ`qDs-sp%$cX3dud`}X};p~*#ERX?b*c{2Kc$z
zE3;M7`f5uPQ<ZXkcA-+MSC(pFZb~0C#dmiR^A~HiiF&;n#;?+iOq7#vr6(u6g>zF0
z&2#jlI9o0D1i!TuK@emA#aAYuzf|!r`?K{*ae)t;KThvwtHsOgykh<QY^^xGT%FRO
z6zhvceQ&X*O=_i1R4z@-TeU)wA8)ZzD^?fl#i{dJM#YKh)rCdBa{OkD0Zr8^^V4BU
z@9+=udJAsNo;bg_P&q-R%H-nWg%jRHed5H7zqq_~;-%&4vn*l8wR%EBe&P(9{XI{=
zcw*jLIAL7{5;XSz3C~+vC||7Epq8g+tCjk4wc=H5uI{~$%@X!{v$cuI`HEMzz8}^&
zXi!o~e~dp<8k5G4;zBw$SUZUXu^>JgtH<WzLBfj%$=UdK#D_`SszFSQhZ(POX?Ci@
zccu*vg(+6PH<k(${waP>hw0k+iKU8v7j3mZ-#@}*IIvuu)fQ!olwC8&NVdN1jQb~v
zPx0IFvpagK{dK0a5(5#&O7zQ1cu6lc6Z6tu#>;v+FYgt+E@ItYkJsy^ygvT*^LK#X
zH2?bewb=CTPsFC<-k>-1M!uBH$Me133qm2Fja2kB;%((AUCDUg>=4Secqv=St>tgT
zDurvjb2?!NCQf>nLolUohg3>E4w;mC9SSM+In+_=cPOJY;7|of_pZti?cD7heIr#G
zcCC9a0!oj+p?0@KDKB{^ypxtP5~Z92eBNUzTj}4u-hK3S)Vtq%0ED;Ad)a%?dx(_n
z<UQ<t6R{oMBi_5n+vz>(J;vW%-n+fW`McYDkM{(B_eAZU@t%Am#+!@YS??**_j>R3
zp62hqDF6N5v)*&|-u>RV_dGcVJ`qdBN(WgNuY{>HXI^;jt$&+jmNWA#$_YPA%%2`V
zyOLw$oS0`9p992c`@f~l`4$CZZ+Qe?KK8`UfB&!V-~QN_KRo<5pZZrT-}Bg)zjpJF
zfA6`;*B|?GV<kk9KYA=o0wJb~jdcEMUFG;yd$oX7UXNXit;KKfl3=SBzZ9$Wyf5Y@
zuEjqZ54Hxea-yCLVsojv^k-t8VhR4HK9{-{hm>ILmoq^xVK(R?%moF)e9%Q$2)YTo
zf^CG|^`0Oej9g0)@2&UIO0U-8n+e{TC#|29e!o;7P|meDDTDQ-m#q)ghe;V++j4_(
zz82@rxz8o_=3vx|EvwuJ`6GTM7$v?npmcdO7$V#jY$4np<W$aI4sty02zC<g40ijk
z*LMZ!Yw_Ca?~C!3X)E(NTG}0?gG`Y1@^3;kQch|8weMCb`cu%G+xr;o@&5+tUDhL&
z{0VDMJ>FyE8JjBxgF!M#1bsn&FyM8+nPJp=R*Ut$^?jrkXX4iw>D9gU{c{KE2l?8)
z`c9q>dA)CHv<`cHtNUhRtNZyqP(MO{#;(QB#9kYHHL-ed^-ypy*n27F|EFO8+Ff3M
za46WfBQ_elp%FT?x-U4ec2{svPZvY>_23(^VE-9La&<rL++E+lb|Bb)BVHd0_Fs#8
zLpB>J-gz|GPI#<-eC|ZBgOVrN{ljl2Ru2<D_UX94zkX`&H1T_w!!6d!d#afrwsx;K
z5*)svIo<W?_|ozEeb=-;?x*Da)Ov)kdSLCr8%le~12*b+Z0q<~(tnht7$&Zq4C9wp
zPEIZRKInD+YEfhX7-eZ`5kjImQ{2Z=-UsIJiu<mdj3Srzt@J!xd~vy2RSA&OO8>DX
zzv9h0yd+$34xE_<kxeevDl7fZR4Pj<`s8zu*B~d0Vf;9dxodU?{9h?ws9ddu1--az
zXyjJ;#Y(MGx7Sw|CWPmfgyJl-c)2n=bG}|XUOXpc$p7V9WqNtOcp1D@^uf-HzFsAg
zgKigT;Fx_Acx&py(jrvA@mpt#cej3uuRQ+Z_|xN0@kLL*^yIm7M2q7uKUb!+#g|_=
z^Z2<ZUn-t^zF2zl#pgS!jb*~Fnb}K~YFPwFE$r!}JMt}Bq1%C)G6a!d4-3&#rRqUN
zEJ8O|%C)zIVs6I6v{%0hT6z~xZ;3jXQqcG@?8w7Q{MNP;%rLL7<yYfDe9oZBTH;3R
zyAl^6<PE1mill_V!{pS`a+t0zFDzYU+!M!7hADM+J>n}oG7r%OaliEFr)d2vdf)cg
z%3cA@<7Eh_*?PHr{E_*^sfqd8qeL2|#}fWil=Yt`;01;0+4=c05Oui%nK8Q}8mq^z
zTwI=Iejr6BJfQF5%%ur`>|qe`G9Mh~8tsSuMnP8U6@PYRLWEA3USdk?VZL0RnxClE
z%4L<VPApW2^+RiU8rhkOTRmaV)couch}%d!FYFRYTVW|)E!V@oX;F_+an@Z}FkoL+
zCz%Y$q-tfp7LKx}%1ca*UJre0G$yO67WRwiF%IVye_}};PusHe2l@0ct9q<Be^^g>
z%h7X~oUc^Fl*TK}PA*SRL;Z(6D%>Db{jk3ojVR@?>yrAQae!hSW?9e7T4&}b7t3xJ
zgnjdLU36<BA-k~1EK$D8NMR;ydbaNOs5`wje~uKtuv8BRn2buj%&I6aFR@H2H6_i^
z$BN(VSD2)XFY+c#RF}fEIFhSj1{%6v@j#W;`aP$^zUERFk<FOoE3{hiX*2Adm|r>%
zH!;QH^B0+TjT$R|>xNe~>Iw^&wF|=Tc@T89?3Rjuw>ptxtIvl$>U8s^^sE}@Ul8A+
zRpWn91^b%qyYXfx0G`6`<tZ2(QPfN)d=-FgaUtw_(f-0!(BqtnEcw-$u_OK=nkscc
zxz?jeu_f-mmt_A9y(SM#5zkXU??0ikFi;bo@m6Kk@|RWef`W>IX$3P1L{#`w3f`yS
z83oTOu#Hqx%vVq+aEsg)BfS>9^uJYk7ZsdW@G*jtRyupi_NjkK$=^;OK2r0nWw@m^
z<vcc~-T#Pc+j4A<X=#w11{<i!Ys_6KU#(nrz^>HYC|MRe7@sZmbQHBYxEc39q+S?z
zAqvQUjo==BYp)X&_V&j!@$LM!|LczT+H-e&cf63W91C+6_Q!Ma{&*^0;BPm7bHw<k
zyxXZ{Kdq(g-yk`IRxa^+Z+tYKNn{dA8;*A;dK1>h?UkMmxHgrgPmP~mA`*dVPHVqn
z^jZLo;AeyQYPOoM!*xJmq(r%R@i)_}SuYV}M`NqGdSWgK{R7nx?#|Z}YbkQ_9@O~9
z65C=o603zEJC}YnzM8CO0;nqRa+Y>P#ia@JL4GdnCAR_B=du?PwRb}mW<(X%3-zwK
zZhm{bEG<Lpt@V02Xuj;VxM6=iiw4ZMJr{U}ero1-1zq&0FUa5ETbileE>-;$71Nf}
zV>$hs<n*@c?C7l1*Pi3`w_1c!ip>p-#)9lHlx|QMjoDw`xkFz&p!BV=Rp?i~WPqM`
zxAkPu^@QFJ2HoD!jYLO3hcyoNWrZ7Xfv&$hpt+jgxFzUv@3d06%}UJ;c_SAR&`eO%
zbHm=&p_m6HKDPy~vMb14Q@i?t5%qgVY_+HQ5KlwisMXo$ZBxtjuC=WiU2A`P<4Ipg
z`Mp+thn0VfJgDby{GJ<eYNqPjyq(4y^mzTOjUF%gh{mkf+eI8Pp(tlsgo2#gaU;IA
z^G1yCcvW<<(R;=5o5}THMihoG;sT;0HG!1pt=s(Ey4!*4;`xb7$P^Zfla(T(A4n9h
z2uob7OiUq#h=d^iIMu_nQ9u5-DL&K*?17?{OW{!Mq8~w$78=-PiHndvh{cv`4$!o+
zcM56WF?X9)_A681`OY_nJ6c%_r`BcqFQ|`$O`Hma;v>SFs)V~+X?k%Z9dct9Y~_7<
z|9k#w_l?i~#y5q-t%4@_np&((pV~C_v=P{EiG=bWBlt4E{QS!%yZCPZ-8|pwUsyH~
zO+<2oRJ-Rd%}+-7ympr9DL`h_gqExIv0aAC{cTkBKTHs2BgE=|R7oe4lmrR;pU`6q
zwFtO544R2VVA*L<71-PVE2@*yxIv~Y&KTMWvkR<=h2;f@Xg;IbpI2bhoH^&DXnAI+
zW<s=ZXne8$v<Q&}dRg(F^nGX$goV|Ks{cRh1<8r|`J26lP#uah94dmbt(S%X^)ko=
zBOB~}{%4hvlk`p^W!Dikx{iMJuac(PUzsEj!YCvXAckC`7r>uO6b!M9#`_ce1C5L%
zw}UXYgJ_16nRw!ME}pvGpWF?47`9rVjoaPHoKV#5?!*BQPB-z~?Od{(+S}tRBZfel
zghqpa`yZI(e_)dT{+Ogbas!Fy|L>T@|2}QJCid|EnjXJj!4D`fl+a};A)@DpJ0Y}7
zEp1zuDooMHG7!El-}s8se#^QOUWn@Ndneoxb!txwO{_x_QGva!0vqs%E8?)pXbT}U
zv4-P;4zzMM!zocCKf<7e+gn9jSS9MnW_hQh$`axDKdwL+#{UTgHx;~F!A}x|gYC$~
zRXgkd6ge9(iT^iMsfk7WpXRYN)WK-E`mhEjId$TUD6PLG)QCFQ#5n%XsNtVgAXMZ3
zFA9E6fiXBgub8nje@C%jSMYZgd_lo4DELJMe@nqXQt-DG{E`CWEL!ZvHPU`x>Ay;_
zvU79j?7Fu>6;q&!*s+@4_-&%DV*Vb&_!<(yx#U{vhLi9q1xkG`Y4SektL3=A2U=co
z()j0)E5;F+c(L~**?q;e@-Ay7+tG@v_gGXf;p#QybNWl^nE!!0)+}|_yd2ZNvE=y5
zwug%^p(>K{OXQnmSdqYNWd{N-6OSn?;(!~OV>!p9&hzm|WWci(e@^j2Qy7C1aB6W0
zA<Cj(R!$aPQst2Sh)Y(kX(rcA3bFt=%^FoJU67kK@qvlg#f|mCkM+hAgpN0ZU;BD0
zk$GEuT7Nw8^+aN&*QT)9z*LILjh`(<a{L;LmPtyajxxwB5MYo}Ub3DDQfo;Gu@Tk|
z1*B|?t)^P3tLY$(P<LcC6D0UMu$m1rYpEc6107EtHOyKj$cW8Aw3-Ss?}}aAb5VlX
zETr|j>N$zh--y-oYp@GzT@tXab+bfhNy20p!JdtIUKZ9c_hyo}{Q;unJj`G|NL`Bi
z<%{p<9r<mL^ssgX!tOfE9YX1x*dQ3eizW`YCog{?<$nw|2)0s8C9F*pquj?#7!FG(
z1XOz@D6I9}fcbE(ulBn3yuyV9LiGNi2Nt4h8$G5+-K%{l6=2_#G7=M;5%fJ0yZA%Z
zPerxz7t(zDgX`+;UhRug3Q<eLtNraQA%ySp$Qy}GtGBco^iL;f`5#>T<$7!lR!_Mw
zIs@z4>7|{4R_!!>@(K^2L!s`!rKVUxS95S(&7RJhymc_>e*~uJVmcUv8Ry@}66($B
zAm279^+z<)Vw~;`2G@qXZhEvyZtWmnFt|%&YNJYD-XA0x>!(;7!>lqCdv7MH+U{Fg
z&&>AF>M*OS&j3X~ZyPez$^fxpiw%AbRvtFaa=fk!DSC2|8biU*XJQUOxUx~+w~;p-
z3^(!`z+-I7xc?&C0=}Ue(G0>fbkq`}7LbvmikB4ZR`E<_zGCuqWJ$#juR3XYtHALN
zu`_!M4O%32mSoe!&KbTbzM2T)Z31W8`{)+VQ7BAFLL0`*>i~lPeliaU6eNfY$5sy8
zChrtZZYE@2*>`r<DFqfWPKltJfE-e(#^X)&+FqqLC9Q9LV%{M~v3<ts8l57JDkIYZ
z#-X<f5KUchE92Dcw|cOB!Mp=Cf)pRO`isc&t}Ioi;B1Q0R;=_r7v(?0)5C@0%246u
zR*@)iY$TFdN+#%6ZA?_#royk+q|$e=G$VEMMBNGP7E~~g{@<@JE+ULhBbJ`6hW#2&
z&zV<9%!vTmr9|>%U%=2Yfa$4uXrwTWbaMs?ytuS5Eu)WGm|w2eE-qIph`aktRN0zz
z1NdWG95hcgMVkH@^(RsD{C}ZH{6d(Ry%6TAk)enGyGl(PnkeO&nX)O?G3`KB*cP<{
zp%t*G|2@{)T&YkkBM3)kI`1{()P)0eLmU=bgTH<$U8argT1p~YdSU7$70mIkU{b`y
z>xr-DQhGA>?sg*SVhPy7-mm2{3Clqy7~h#F!1k&3V0;8tuowI^7(b*~au8OpmnSif
z!_*jtq0A)`uWwHvhrE3N=2HK_Ubk0<Hd)G3+CGh+3=wT5Lt5`4QWvI|G%-GxxRwl%
zG9ej+k{~sStSm|gZK+(U`J8sr$LXZ39VT`&1ZMD2<mahM9Xk=kHb@K__A|wcN3N9X
z#h`eld~)n~v*^P%6C|<-{QARI?StdRE5#@HJ%0T7Yegq3kjl=qDf5*YD!1y=GWxgm
ztwi-97$+;fYBX&@ie@oWJYtfb;;FIY#rIbH%F*IUte>!kLgk1J3r00<P5PLN_|=Dt
zj~DC92o<PH$)m*viG8%#PEwB@Eh1ddE2A$tTr1kjYc*&L%UIf%(#-#y0!d1obm51H
zjb$PtCTby!*D;OZogsz+PDoVhZhz;xl^FJ2EF0<GDE$GgO6|zu>@d_!A-*zfYhzup
z4vIz#Bh9ZC$5%urZCo6dBGtBAF3DjO&o2S%$8V<Gx2;jlecKIUq#A9OWcK|(Cjlj(
z44>Bbel*hFMkKxIh^%0R&-jWyV~`DzxShDYvdcc>PUY|TJ^o*)V_zc}i+cL^lwh;)
zYl^+Cpp%|QEa#NU8@lCa_zQH-NAn$9+5e86iYR^!Z{w$Zs9RlY3}S16{2&#gLG*7E
zw?TC5)w;fJ-=wcAWza|Y(nY#xYhf@I4dR|V^*5?<$6++`dEK&L5v*GlKoQwN*>L+Q
zjZ#{{x*j(t9Sy&;(2L%FRfTO_J12X6xsG0n-D@sh5g<`7S*#7n7WJ}`k}!FH1F}V_
zt7$7G_-6GW&P~fANavPM#x3+8)oj@E`7y;lui%FXZf0%C`+uf{=9HCsJ6Y^j`SqjH
zb+M&%{Rji`m0lWZtp5I#ZRP%GD>vJ`<4E{9jioL7sNerU{jz@lL&bhw!QWSaRrH1_
z8ffc`TgBH^s*4~JK5i&rI>nCu<Y~PzGqq_RU{$b`3leCf&^*n-GQkeW<_E9>XhCE3
zW^y%cBrGP|<n0JjFiS~)e~?~7v*o3oUIl!VRt;>5lCLILGppIvoGf{QoZ+SHxQW1T
zCfufxJyp9T{Ez;w0cA83ol^yr7U$_i0uaQ=g3<afES3EN4`X=;(c)KJt5XzhDB2g7
z(XL-9qZbgl>y+;T$~M{INOQKNP;Scj*Ux|7QYRq)E?!z2<SpF*@j-s2VC;^MAWe=y
z`nn2J8tcYSH<9Mpf0PGEXKXb-FN=-%cg0_X-T2n{YFcC09RDO5MU7{U)I4@zlxIv+
z*OIFRbXZsvUO=mO1HeBA13MG<Qj!JC!Nhq9>83tUtKSiWjf8>p(m{SWwwAp?dQfms
zVJsu-AMENQV5Si?SyrSCa`B%a6`_DF)Vow-L=|=r3)7}JF#g|6D2-lCLLZNHyRFyR
zcG&HVm&=Aw!uX{Bduq=b9i~x?c4g+;IVoB!VRv=0YNpBy6SWKL=P9>TcE+WjqiC%_
zliL#`&|yatqwz?y+bGdNfeD#uM?ArAZH<6&XAi70&R*OJ2+-hSQOYhzNJKCbU(M80
z^>iIgFJ-f`9W$G}EIMAX@`r8i(BGo_g`J;{Gg}BB<_cc=P3^BP?GIYg40n$}+rg&4
z$DXntI%D?NQwSb<g0$I;6@!F7F1_+JRy;u(`Yt!#m>G@Pt%4aUi#MtJ5_Yk_Y%HHA
zJ{ET-1>$(ddVT3aT;|=xj34l-#wUb1<gM@!70>^Y3VlMs9!>GbRWb$tFyor?l%ti)
zY7i}9OClV01e$|MuqCqcuDMt3)YK(qJl<o1&}Oa({`Iri)0iWse{t!@G?hCk27%2~
zX5vLi?-bgXo$R*6>r(fp;wys=c}CN372oQoQpYrF(=|8VJ{~m9OtPk0lOj-;0?8z&
zA*$JMI{T2o0^b*B%Ki|XbcPyZN=Q4XQ<$7Uv}V>qR<NzuK~=Colp7qTSw4S3U!=u}
z0RmqnZPW_Z=9p8t{0>0fAWD*yuRTlvDomOgFkAbL+_eO(ftDaZ;5zKVwd8fSwWaV4
zE+v2%FcawHxb}6&kD49g_bciT;*<Dwgf0<QWf9pDDaarOiyQ3X?PD<EFzDR+u-qd6
zV~!|1`m21JS%?*4D~_JI7)?n#^X|$renO+`$!}3tkb}0Y>@^}$eM1!XLJri#oJn@B
z(e|%sK#d>;7#q*!tT-C^n+~Y$bxn)W7VK@{rMN995KCbhL@7{CBB)!GQ=@wZ)2wIK
z$91!vp25K8(yytzMmH5~iY>9DLuvY;U+d6$B3hM(euqTUkD3u5>l65l>=I)>;wdK#
z_7VV!ljIzzCuJCnbdnl|!YSd><}A&OPm-=?J?mj|`_VW+A4BI;{=K{t!3-SN7$-xZ
zu9wYZb7jwR^+I*=auvP`4o}8c#ls(N8e**;X7j(Hs>zATn*UW|wp{${N-H!DBuvbQ
zxdol7pbeARH?~bmv~ie$$z({Zx*!aN)(vA$Bc_bs&@7MAcmY&v+dg0TY6?Col^On8
zDwTUXg7A&<9fRD!xmxcvvBn#>k-)%pao0&N32X)+AmwKhF`Vhy7&Ir_%S$ofdxIqQ
zKD+HX&GXpx7|$d2oZ<O?YNlw3wRA%z^9bbx=aUVqX)K*%ObgIow*PM?=dzpz==yvd
zt7L@4<mLPqy`E~^>s`%*1+f32X5Z%$K~7kE7E2Pu&laaF7#xu?7$uM(HRmupn7&c3
zpL$(U9hE~2Y&14hS=91?T5h!JaLm}?ok4|vOiNnitIt8V2)@~_w3*YE@=uZRe<<@b
zK?DgTcC6JZwT+OVLGHD7`lD*(F$M1?SQ&VsVaFF)e->8;%|PNRa*d{u#91J<mLbgv
z*Eu=g14@X8NpE%m`JXf#VG<+Mm2NZaj|?SO^6m*6i(6@nj`cX;bdNq_POW{b0#5Bk
znx{t<)7!8?Xj%?BlSwnrGXn>|Mu7$e)CtT)l#2Md>JQRVnAg~t9M*qCEx(JPR1i(d
zHkqEiB0D|`IJ?eL&K~Hk+bnG$u`$U1g!;x84zfYmE@tN-L05t2!Ng7s-bN8p4I;ei
zO$;NtYLZ88k~&mjf=(HJMNQNmA&^$Z2p^ygn*tIE0tZnP<WXuZZ4gQe9BeuRUJ~&+
z%Xat?j^lt)la6%pi*#`;>rhNqD5hzl(jqVqYAoKTf)3dliHd2GkJWAF!e#w+B>y^~
zkwtb&uJ@-J2BFY>n02gl8w8HBIw%7b5oj<Nti>l52Eq@qTacVXcVqClPXwT)8juj~
zB{x|I@<R5-*D`!LDSTsk+_#07I6i)BaI$#5c)2)LJoadDl0Q7KAbS=n9J{(&c0yjp
zK@_tit<{k*ff~)Q?|@Wjvop(!7(_)9;ZY4pB*jS^_2K`aY8yUUHxhY)LzB+^g~mc~
zP-K+kpR$FFv}w&k8|GajW255er8SKVvky>21sXv1BJmfn#PCtXslf#ZhxhX`vu<rn
z+{;R^4r*!pPb*m0)h^_7-~<Fam8E2Y&T7r|vU+pz6~|q-73hHOUHYm<NB=nyx)_5B
zzei9UPAP?wNOZh(wC+jMh$;_FZOq^1!Kc>aC;?fpQdgPkYuG4w(5iUbAdTELJtM9*
z;ok>8YbAp$ik`g}At2HILP6X}Fc>*twibV$6J73K?oA{uqyg;-b5UvmrC-L<^Zq<7
z6mEzsK+WcL(>Fottk)A1)|Jk&xO&IK_>zBy5ey?BCvXQ<Ci&vIo2f8|QT0rDl5<S5
zsMm?mET`5qtYMBwWodEhJjWC`go_!wlFu450Wsj0GV^{Jd=Ak1b{chzfGy4M<Pom>
z{30jl{BP0AfkHD5Wc{3qPH0M(RI0GJggGGA^XMbGO`ME<Kc@w2V_lA}(_#qoj&7dy
zu7o|!3S}PMV%EplN19T=0HRwplge1Pbb1;v*ghwx#d8-uk;GqXJ%#bsC<_VM|J(g3
zvl_l#z=$hzJA)*D1n?L!5FO2QKnSfxEirA*+P@@V%Ge7YJ0aQu?iq5yo>^*Z@f)%4
z<a~H%$xk+(qmrMD8#NDU@kZ`Ci}qUlld(_4Ip_mSAti1p@mFFVI)YT}8b+Pf+^JaY
z@yiea`u72TSu7IK=s2MrDU23ef&*0>(yn7XyuRG^cxRd!h~^_Wt;N-$`eSsBU^Tu$
z!gGLobmGK=(SY({uYI?jIJwEmL_7PS`2sAQ<4>4G-vy<KswqNwDp5Y^=2YZESa?Jl
zNjr0Ln78?_khpThpf|XCmU9UxFKw#Zh>2#EJ9gHe)6y{sv8|tZ9yfO6RI4NYA|;MA
zdT{of|Eg*vZTb7ZNvC{I2~JPo8;F}B!GsxD#5BLJy68S}lQj3Xv4me$#U%wc1byp1
zw5T;L>Q`opLgJutR`t%W4X1}~bw`4yb=5k?LTgvQhQ~E=tQ!P}TB%w=A}bpjV}s*-
z=><yp^9n=+kzjn;v+Cw^3TzxLuwD-`Ew+b%K4OYF)Nx!L*Y3&CWz=C?Z5lV0>%ct1
zI!(-Tln3yzEGQCiw(Plx>jfKI3~VeB7@`@6eN*GNMyd^5>(Fl1A>Lb~lf@{i^dmft
zxg1b(s_HLZ<`WIvX`Fn?gc^u8>zMOe*jv>p_*xkQ+%gry?&wsJ9j9MEAVX!LR*FT^
zbe;A&tr02Ev^%IEW@cs7MrD)2ZwcraDLbV>Z2a^QQI0TQ?FuuYv!e!D7$BR>#A+5f
zVHW(wZUj+-u-l%KM!aOrY6*iD&W|Q!HUiGe0w5BP#Om2=oLqoIfdI0cJcb6bIZhIQ
zSMLKq2}2bk$<<kV%6n)h;eRgFyXxJ@t_m}7DjuW(GB4zmn4Jj!^0u{J2&DpL?9_qI
zEzI>}^uoymlZti2+YAJlT5wuYywqTTZ9csM+XgBB2T2+7Ncm2v32F{A_2Qxgc6(NP
zHvr1-=m3;Gw9_R!d0OqYR=ash`@dA*vNlrZ+t;?rI0(60!i}5}em{F1{@=hfh8%Ao
z)e}&D2}lPv?sH%=%s3UOu^zWFr)=4QlXq&DJVp@qG<Ba^52f4-^NVJ=aE{G?w@OLM
z<X=|NuEvt0V8;$-<+MQW2oDB&vpVe84tCpCp<o={Fbv(?rv}$vr7@G$j5Ks*!smAM
zt(`|}g4MB6|5LoNMcRc)l(&v#eMwyx#x;wEHu0gSH0@+38~7Q;jHo}M8pd6HRB0=E
zG>Cdsu@95RQBa*5t6^br%Gnu=^|g@?@p1H)8cAO@!mdd8C|e2#nvB?#w#G@TTFm8p
zm8Nm_UsqtG@=?VMgc;N2M&TiC)CcKG23AQB7lZM521D=O#7?LN2Ztqu>`kN+-LO(S
zF$3>~n%K>j9!#*U`)pgc#<gQIJAj--d#vWzYTSwQoROM{?70+%YWj=RnX@p7dS)&=
zmy@7JBFYHlCB~U#QAT6=u*Xiw=}?#s+qoGL^!C4~;A09tsNimbFm0q)n10-wSa_u&
z*iu7b#wfR1n50D8g0evxhAF{s2M2J=P13Da)^K&xsHB=1$SobpTG?hpuvz(=nQ{Lf
zb>y^yhZJZO{BKeq`NCM@9ELe(*&k3*F{0<_;D<F+vTr>%K6c2TAQBeJI^~I94~~9z
zm&+HIC+6KjP&->@Cjk6wYBa6DR+)W!7QC#3r*4QYEAy5D08h+MDu6(b`B?=81>FRk
zlEnhYsXk0NE?KaxLARVuL}1c^_bFpsfpJ5knf+M>7Zr%)cFX>gihY_O?1>JuPE|N6
z+ioNEz3P@m-akd4-`aN&<YFv51jzq&|9byb;yl(aw-XP#usfdV#m@54Tz4+RzZ}2$
z-h8H;^qlfgc4czuLZ;A{+hh6Z%qVB>dUB~umdQS7eVd|N9X}d#{nSL2x4dBBzao)O
z=2Q7Je-rr(v6NRb7>iE~{9B|1)!&H8mqjJJmb>BLZ<p7N&>K>ql&5b4-UqFQ^N3L}
zUl<&D^yCF2R_Znn7#w2sUiEf(JMkyM*(h%}Iep$9ugKqiZ?CtHzXRTW?*M-Xy@TE%
z{tkJsnXijD-Yx9i*1K=DcgZ~xSP(B4mgALi-E_=(xgdVrJHdOlMtLW_Q*VGN@Npr(
z7xHl-pBFJaTzts9UdY3Rd|t@Ig*;x!!-eyBvD=$4pBM4c9^ResCcSr8iul}k+<VU(
z=@RaqG$QI1c~hSJUuaDBHPfHasCu6H!%)utW)3yTshEch<s67|64utV_l)<fwRO;?
zd(U~}_6>)k=Lzq5?*%JyxLH$knD%D8bKc9AbHsbU_X=M#7S;I{?*rDxUFLP;Zm{P0
zu<u!bQ?=eaI=xa@kV_jo%suYk$BKd8eiwhHdbE6^AMqTbgGB$-r1v)87E*XX=a^Ru
zq@A3ZH2)Yuuf#PM;<YVi0E<B^-2R>*>+kZCNLN#E>IqL0TzPXqT;|0{>E(1LasdP!
zn7$HcH4?{om3ZFb0~RmLsLWtJeNEhckH;y?*?7<^As2KB@?OqxA~SBB;ausi7Z8?l
zUURNT;w0n=c;I*?Rt37~pC=z3y&u4Ew^lXZ9eq`#1#A6YPkjJdg)@v!zrHqvfM0Kj
ztwH%ju`S3Rlm>-X9^nKi9yp-e@W5g5Z9$(|r2h<cwqs~JsFpnW3fUI)`=rk8sQs9k
z67Tf-%-nPrX}j$l>YiZ0|MmJ#uV3j_@|SHaV{?1fiZ|46Xdq)32gaS$^=W3t-=|<d
zK{(72{u;-gK>%Lk<b2uxwDSKd!CC*i6@5a%cPU^>QS#PsCV+=1JTxu(=lmy?ah70g
zZ@3j17(SQe;Yz2kU7O%K6W4}=oktT?pwZamRTxGuOfD@h`LC+M*Yu_?ieL+CUIId4
zwT9aO<6`8A$h>PcQlTy!d+;6fvy}5ZYJdkzS$LfVTw!q<nnUOwN|T!Z8H%{W4MH|{
z;9;Eutd`q>aHQQcR;wG~XnT6(DaqvM;b5zj<bslRnsT<(J2|<CY1iU~N>w75a7Z@!
znDQZ6Yo7D>4cT2#zsd@R6&Q^$uh<oWu^}V2N`vj*5u(qPOtLW~Q4(ZR4TFA-hA-{A
z|9J&B6s#%uF$FfmHZDJ?$ImKwR)O)7%|+3+CJgZ%RG$p{Y0fT)%=i`}!blmo;9NrI
z!)1Ws{*92Te+ixOHt7%Nz7c@ypUcUhN>*K+85izMag@)U8>gMGuZ*q(O-(l!9poF`
zh{?c7Yq$0yft@vgk2YC&LR*{^nG67o%9u2*VaBwUajttD0GP|(VB?C^69v!qN@?59
zWj~g9#BHjS+f*>6AK(SPzS<Y}ZrcR?o2{&mE5VfYqO9dUgvpupFz-d^CVwIv!!b`~
zHZf3Bhp2!y-rJySbFR|d<VLR2eq5t1h`Srb`*3_^-}<pJml7RSrbO5_el$X=p5(Xo
z>jZos0v$qtAfQ*rSr{u~AC%6Wzps&!td$ATCtwz_i{tJbis!gH5-|f$8Qdgs=$;Y+
z&1geWWWkJLn=hqZSCAqm`CGTlC)RoxT>$ZilqY9L2^3#_Hvp|SI5;OzNV`(*r5aZA
zjrTM_+xov_7%?%>?!ep1yCJgAW85m2YS=6nr<OSZvmmE4z_5n3IFRR9D+MhMuwtTG
zL8d={b!Er?{rihAza$qkFP%Gc42LsMJpQbnNH-|5va8|wM(){;w*yC+x{nR`+D`a5
zaSnT&fY-@me_XrEcAV{V1AI-{yx(No!a?UT562))S194AL>FxdO^+vXq*kBc(CuGg
z6hh7n<K+%lCP}^$MTS;aOBRMlH=NDLsvL2Ngu7zP!U6X>zhzbz##j8#&5n1IPGOQk
zv7gd<9OV^9u$#;w3zG?!c0ZQb#E1WM<-qzCx>+-KWIQ^~>1y%j__f9RAQ9)Gi=`)#
zli48NfJI>OXz8-RjP6MKoVk}ZRvUIqL>oEnaL-lRvfUVNshyYGl&T@KriT^(rx>Ni
zSU#n~?^R&9^?iyxqu^Nuor7k~&aY~W3S^EXyD>3Az;t_M@A^^MctV+4%^y+i`f3Ai
zk^naZa6v+1ZNAibBTJN}VM$2ohg@jLiNaA0CON>gkFDnZGFIix#X)32?yuw~>2W!b
zG}nJ&pUm*Yx8W)hP;%+)Ilf(_UYL^e!f>E{6`Ez3c?0yH(CYrzl=Djj4a)*~S`->G
zeD=G_{f`O^i?mS54^#IaX`uSalq60_LWc!%iPYCt4zC}sc0AnHn+~aE$DoOoz2l&n
zh;Ghwr3_oPIcyr9AtR>}<Bc(}uauP0#>yE|jM2CRO0_8;MN4%8GHJs&_`jmjkwuOj
z$Jlfbew5tb*C2>(XApGEonO|=k#`z^_P%tCK!-@<$7qwq5vx5aQ3a5FppMN_L`Wjg
zC@1x$<jI`$0C`$U&b?jQIhQjc)=J5oQSx7wU)Gji(!`NToTGp(q&#CB)2m4!d9vJf
z@&2gg_=RNcj1jDQOS1MpBXgT=HTnQV4+o7<ZJXM1dF9~CcHYMlIgNljDGn9X7K;<L
zqA;FII`01fqqvek*El}8vO@)MGuA=%MvgnrqFFPl$l&oWYj!3XC;vkVCKPDh`@f_N
z&9YnLUsOzM{AOesWpir_e6zLvBh>!qn$~?(5eeju!04b#E~%AnN-_KxNVkg3>Y;a<
zY=_(R@qUSRJFvTeqkyD=iStBojwj=782}@kVov(O#3`p=2-0g}%fQagkWwtm#ax|y
zW;RpLfYXJ!Q_L*qG7(T`iDkk0isfOcG6<512U);JMgp&Y%o|bQ3`RV`Vgy4>^QZx=
zD_ABNE0~3&*(`>GZuz#OwiYII<h|%B9eJIYy5I{8I!!=i>}aAxy}Gok@nSrSx!q|n
zEHY9t@nR{l;`tKL`({Dk353G!Ch>2R_Bs+>T7o~;{7J+s6DN)cAgGbI3jH&U!CmX;
z?9NSfY*BrVwy90o86ujlI=nMJZaCKrv6QF|SsJ=;QsFmADL@p(-6q5qDWUCYR!mvh
z%P}3CGTyNPr|a~MGaFmka=2-)W3I`ZM-ZF~7vv+FfyU#Cudj8IG^1<3pvPY%z-_Vo
z--(QAXrO*xBl-meog?c1G-+Sg&<Ucs!O8&B3dxNS?M~!P$8gb93lXmHAc$sygfS<m
zmDs`GXRbL>8Q>0TA0;rx7MqSR7sG&hh7v^v7sw?#U$%RgmRY`@Tsvy_TB3H^Ae2-s
z68Vy1aL;O)3tKnHSiFm&3bS}ogZT0dhCumq<VifM|3!_=KT=?0dRT*)<$G1O!2rak
z){UJb_)K^Vuc&R`@Gh2P(b!rnm>OLWB$X*(bT`Ps>{NX7q1~*Z4Jvh^Ok0WXCnEb(
zXa^)68Xo-QVn4^(0@=6Ak`7}W%SUm(Mv6-Xq(mhJcsP9kt%03}^R<QDhEW=hTo)PR
zyiz>Q$UrUFF3(_EW+x&q%~md(sshi5xYdphf5p#-@yq^RRW-^riL<-)<Ceq!^LaVb
z-lB05Of14InF3$~E7>U8F%qiM9EsO?(3c8n>a?H286-V?GbwH)SU|EzXC#6o-ap)z
z12ko;@+L@7wk{eCMl&@E`_z`6Q98ejO^~sX!gm-Y6sn#Buep_9Efk}l=tSW8hl}UU
z&F1=y%MWux)~layr8iFMIZxhzV>}B3(zMCNu$XqGIJxM<6C7>UDqhBu_jwN4R$GPN
z#d*MqdQ{ZT-Qt4CvWsLLzfYO2wJEt6@|%sdSpa@N<0&26dzkbvccD^Q!WOXRmV_;Y
zOq*`%x3y>fv4S@g{0>3G#{GAd@OuQ(;$N2XS3XDCCw>+Gg1@hfZhd8WzA|0MjUG~F
zJ-Jd636=5{4~LXWUB7~$?qCSZZb)`&6$yM5vr;*YvX<CmJJ-Ye_9`8^VXY>y+_taQ
zh_G4=CBrOZ=fvJ}wye?%n}C9Y_E<o{7PEr1s_<%-War^?j%W7MF!`7k!%T~jmKE$a
z6dTqmxwM+KR?WuMOGmA;mSp`JWYy}1b|f3kV!6s&*`e9-QlB~KWbz{?SB->;D^Y^Y
zXsJ-2t&}HyJe>Qbp=K2&y0hPnM5?tCVJ2h`>Gpd{wMv7vrFr~dn$EUyXxZjosL`l_
z4bvVbp1sQN71MKudG|gT{8HDm(aR-<>_kUlkC05W=v|uGoxDK;E0NiUIASb8lqj^+
z+oK#yW~O(opXrV^Y#ch~dyob-=QiKNyaG9M?A4Szdxb<BU7>6MfF)+CxgGqC9#L*4
zfTB@4f1`B7CuUZD3F9G|mtO)sQBFJ^CKf52P~goF$dW_#+%0MMYLBI)@njCt134m2
zk;MSgaF>syS%&G(4nc2}rk|488|k$ifANjy+)P}HuW_0Q$ezPrUJ99eU-Z1L?r>~c
zYq9Us@oKO2Hs$|C)Z4%KOZ3)=Q+mO=j(SB5-HQ80?F!|Xsi$J(=(`8je`QooUz~OE
z=>*?AU}Mt$&F($=ZpFUwx2qIqhS+r*$6!NSjj^bH(ndfIB0lP7P_h4l(MWG^j0Ule
z(NK;X4VQDL(NMX@XjsWjM?<A--35w$<Izy5yvlAe8XMYbjz*&11%29!I}TRSG>0L^
z@+%)}Vu!{V_lH3qGon)E>KALE+R@_F(c<}|#mi%@vQ1=Wg}J?9$=p@DDx5!$Y5|dO
z6KuzTK6vuAqqu}<J)e55)n1Dh;o^a=wo+lvo1J%VIk>UI!yFo4v^uWuRl+WgF>=7(
z9WwfH_PVn#HfRGwd4d#YoMD9Xbsoda%hhSzCxof9g#JGh{JTrVOy*}YWP^92ug3<R
z&`t!7|IZct-vnV|c{)sZ{7%;JraR>~5upE1Rp$RBD0S%_5daM{%h4;9=!a^fr$GYP
zhn2e9il1H&EBlNfS9`9VhXpOgq4cIlljflhKB$;DfT6_!`Rn6pG?Bs(BlGt0<k7^=
z_+1eCo28}@gRSh@460gp+}oz|ys7ciP&3klSnczCNK5bsx`P)rh1vi{zNq?~c~p=G
zv4R`{Oj6w8xrZ1=QW!lUnKs*S89HIQVRe6ic)s=nY+5NL@8PKcrz``ht_%1|AzdDd
z9tRzRP*=OFTeQ7Nc`xi(ceS_K9}Jb6@9nPQu2s%b=em)mr~GA{{dd=U;VLQ?>w^!u
zK#%%mT!jl!xm6|JsBdYBpc^kX?)$zCwPxe{xKv<n!2f<~42p9>V%3X}9~}Jai|AoL
z8x0qvs9aM)-)xfBkz=h=-}*D)|BcR?@&)9jKa0s0o4YIMnMtk=aE|HkQ3}Vc^kgs?
zSldG0ph#attU3RSE=5LVJZJph=6S@Ki3I~}a2;XA;nW1S$BWexF2JELB^tvUr8Ht;
zk}GU*1=D33<}!RzWgbdjzEqGtman)!xI*{;NUQsg3C6ZMX{`y&o3e8MZz^AWrvEbv
zObTO~t<UOFtHt?szM+^gVV`0dIU|$bQrdYQOIu{#x{UK3_B^yJZc%2Zxm0HhCLnE`
zm-)fR6y)zyPZ#e{)N>~<oqEm*+W(LS!|jm;RJ#bvfemaHouFdMiSXssTsk7Onax9+
zO;6&=@xR3DN*o83!;P|&+51hFn+pP_{NL1vG)JnGno?s)owiP0;js03v9yM;s{Jer
z;9g9xrK*$Qeh_h^PNg^rY(_g8DTmyz3==ao9mZF(CE~1Sl5?Nb5nC_Som2WL2Gz7}
zv;A%KU5CU<^qqd6;P)#ZBH(E4et7fPjp)o>onwGJCwUg)w`0|C?J!JpnK`sh_=%Dm
zXZeQ0M*CVqSF5aIKTU}OPK)KN#Y&@xVx{7TBRl?<Ug$UyV>sdtV85Utl!e31ZevcG
z{OAl!j=x>GQt0^-5{9`*vQ#Mz3QS7cq`P>S(7J!sWK>~+sf$jEn5;x<8kK9v3?(ip
zbvIm6OiBOu>-G}$&gj-7ifO6+u%=3<GqGy}z{)RMFR_i<zm1}_JC3&h7kb&(6uhk<
ziD+d#3fQ@^y|O$anVHs4H#^08XzSF`^)vrYjoewdBaVwnG|szu1{QFj;WvTGO{Y2C
z?3j(gguku+NnQlD*`Oxfs9)Pu0`LsgYto<r2*tD6Q5(+v<7+{@m(w|6&6{$3x|zxO
zuh7E=yYez|x#0IIS4^uOCT8lc$F^hb@J%O=V%F@xs=kY9<UE*m(~k9H+J5vVO}WNT
z7ZEWZq9|(ri9kark%h5@unHph!GDtNOB<V!+W~dn$|bkMmb_!h9Q#p@)N{r-MD3Y|
zU1zTF$H(rI4j?xt^Padr1F6d2+S$lVrC^L|mkeYul@w;2Aq4Ae6OOh%2}{rQ!sr(0
z31_^d8OCr~M$4)(AeNY1NuZHT&t!vibfPU2ooK_3(PT>ueay~%+1WO;`h~Rmgi?(z
z`vb*{R%;;>+u_e>iM8;EAr8Y7zo?9#P;ife6ADHZ$hNR?nkK7_WVD;67<XD+KHwI>
zvnndv3XVSFlOG&cFV`m)?Z_M_%+?D4%~<i*wGt#XJOn+3e`DiMWjG<1R|n!NTi36Q
z$oQng#f5YN4Jd2wMe48+CS=Z3M=Xv(dW|v1k%Gop45c(gBVGc-EICHNwa9Fgr}2li
zq6K5E5MTuv%&xe!D@fH&p%p+JflO6`1kT;<(;)9rP*lLj>wv~OG~66(Mnt}(8miX!
zYQX-SDq%`^Ff2qUP_S}Ac|!_1_fOcx?G=q{YJNwfCu<rEOW^GhJ%ibX?Ro(Y8Ex2A
zZ8bN*P3^KL_}$FC4_GF^*q)JK>|_e8qw6~u1qTqmPPK+D?``7tNu@4<{_+*%aQ+4X
z)Tqvx%G4X<f#|4$EBLP(fI)sVHq#5<sbUA96@+Mtq@>3(MiEw9Cpt{h4V*yC$xQQQ
zaaMyn&Cxno7lA-j5cu6hoGC#kK(G!Zk)Ou3yqCEV*9eSdC%%bEYVvPSMXt}dE(Txq
z&peC!GJnDR*2^K9x#n^gx^T=Lsot?k*~}c37cyO;=i%9`+0^80y><*m1b0R*wA_f<
zl|$C()I{|#!b&d3jLyH&4w}H(y1PKh<G#kKC)Z}?R4r1Mv>Iw~rP!%`L(gmOut8hh
z2A4Wfw39btYCkysn!XnW`Z8Dk9YL<D55O^+e5>&pQAxZ$N*mK^tHq%@IQEFGChoMs
zZO*Au7Hf7!?UZ{y-FQB2PKMF58GJKv_cK~~2K|ijyrsu23I-I&#NUo3G%(knDdFzM
z1}tSS0<z?&=l4i2_3<P#`i3jDQtrY$D~~B%zZ(lKxjeb#eMxJ#K=ndyC*Vu0;mY>)
z;Hy=<10f4~Y1P^ee1phJzX|2R45AOgYw<S`(F==!fsz<<j>Iwez?w(|glxT_7H>cS
zZdUpO^5~h=VH*CT5x)Q(=tL|iIo9RW@b7>>S-2q9c5*^|M{_|w>29%XOtJZ^oG6|;
zcBY6K;?mWkUCd)^y=_^58aTqa(N`aZowvVr^;-)gTA+=q9%>I4C)%6=M(wg;5b<ko
zxwrEWUADQQp-CUJpebD`W1jic+`QaD_|8+G*6=e3`&pg3J+{`k?y}+*ZgUihclc@6
zEON^fS71in;?|`Z^bFEXHZJaL>sf0b!D?}%kx7T-(x~`Bcce{npQFW7N1ZQ3E*SIC
zY4El730#p;bA4_bU$={%XX3!v+8KDq(A}ptM{SyCA$IOKsF||Uw#Uy>xOFgG{NEDi
z<bmwJuWcf0ps;V<Acjfo%+~LT;qoT;oqk2Tqn|bs`67twP(rNO;O#<UqXpGw)jQ-R
zLb}>5G#RTcL62w`aQgKhvLB$`%y1snmHDfYZWAvxV0f;?fyRoYCBpb93`C3Jpo*sg
z^j&hb%j^PV>4)3{D*!wnBQxkm)`0L&9(yGniPC%V{fjT*E=17Vqa3czDVSQTk33Wo
z^b136mj*Q|SxFCndyOj1H2Vt0X?^WC{E9P;px^pB5DZ}QO<Q&h?s=pj^5CoEYkkn9
za+lV_6PNISo3;=e;PS)XU<ef)SK8>1+Uf|fA%*8*nQS+{R!`B_(gN4okU!#enZvKG
ztD~#if^F9CF4;IX`V1AljcXYtFWDMwLy?!i0VG-79&FbojjMwfIW_A+gr!<%hJUmD
zD4Yz==SJXIaR;(0AJ$vBe1?l+qWCt8@5J$mjKR12|2}d^vK!eDPEHUm>ua~pZJXOM
zw{ula(RV?o?KHZ6cd!SK$9-?g8*p)TG}yJaC)kbIfLxIjXPx|}80<1<R)3i`_O`Vl
zXC-@OdLFg2*R}KC!(mA`FyA>W*<f4IaI4XF+&J=hWK_Q2qy6{*i{EhP6p@Y}Z3{E&
zIi_-7<LRAUr)+EV#NVe1VlKww&ic1c2@)DO!=N)fLf!`aDRH|QL@N?w=f9`spI6xl
z0<ISnl%1crs=Mh-_2oZF`bzeMX9i_sdmV@JeX3zznT+4)6K4@un!pkFtf4xyi@{nz
z$v39vC6zHOZ~V;b>fgszC9kX5oa?A=l`(UK-%;$Sg3l;02K#R-_O}$AB`^WirLx3H
z8K*sShHUy6|0xAurS-9WPHt@skdY5P>agf{Kd)FpfzBN{|52juOOa=}NT6>1=SqFf
z+gy~!^_Qp_xk%6%26UO%Xgh6QEA_QK=ybE4?<{4knbW0_j&{twi}72B)#LBcc%;k=
zL!w^uGwRl`GDP6`yVSRCSKpm9*3D{>hO#92f5XPeydZEo&-}vp+trM?`clSCrGxz^
z)$kEDEKV`YdNPyw0!Fj_5b~JH0_<h&i1)(80V9r5iT?B`auwx{pmmnDSmLWY|6T81
z$?8%=$lc%tB^@Z-mXV+?O;dc7_(qfY#*VQw2~#4%8b1x$`V%D7{)hl~N`m?lcR=Sz
z@(uLFRiW|o`^f=wU<W>89DPpQg4Bm_<cVn<GX;6$3cEPOI~H^Sak@<6Y<2txs4bOt
zlB31KH9b|y;ITA>*L*DK;u-HP*L3x4%70(bgUTD_rEJ2HFMkUqWs?SeKmfzhV!~cY
zy-ClZSdv0CYcRTh6|BJfFCYUS0Csmv;lwecTJ+pw&z5ISFmv)L*Q<#$KSu^U9%CT<
zUd@_qZf$dnd~5^C^N=l#*=f$OU~f=yds3<!zh6~s2gTH2LG>cha8o4Uc7&jk4uhCX
zw<$l=Ie~$OY_WAgCT_r-?Dr(Moa>dca{#_zXsU|`qiET=;3=iFPMsqoc^rYuaE!86
zHrK*Cg3~Sb&#Tj0)agzx+J}6OeUH}ZQA(tAF&}bYJpHk{3Ow1-4J5*{ebXw{aS#1M
zu58fktaHBUi-GAlMYfd$39f7j>cyLS;AF~evLW{)dvEh&%Hr#0iRMvxWc<C{It<!n
z0p<L=f_}8_1JoK^9j>N%LaCvry6MF^d6mY$9KJVKh3=DF*zjcCgvu-@M*t*Nx8Q3A
z`yIS_=FOXD0o6#H`<JCJk8y4cKQv8mo+G?}kp-s{UxUMT)db8lchhZ+1w-5knCAB6
zxFOk`|B*YC8@BvS+ko9DjI#zeSw{Xrt+hi6%s+>@$Qa{srT-;&OK{OcPs%N6$8GTa
z4{E}N#vE_CRgXWR$H$4`l+u_+!ylTBaPUk8@P-u@WdAGN_9?@P9V%{o<#oj#Qn{?&
z0%&wHv2AMAoHv;4{I;6j+NRn3P30SM&{ptw=&|!d!mfH{0nT`WwO^6$61Ujp#ygrH
z>CW^U)K-^qu4BsCnp@xk<o;b68r#xvgfV}&y-F{&MQ?CfXNP}I>s@VZPvM-#{-NRc
zS}Khl2&<oU&*~q<eQ639or>>7JF*$1joe0xt#15iv~+GV$!|@dCg`t8xhO>)0C05x
zKM06P{oT>u!Rx4JoHfKM@fmqTVqNRJIX6ltWV?}qsQhFM1)mg_r{r9aT>-IfK2e8I
zJ@&2+Vg5UeGa0q~T0L_;ju(l9|MVy1(8g@?S`KZh!<LU2(0|3)0|0L$x0+TP*Nue#
zWIQ^(I1OVw5TL>|MPa?3*X@^Br3{$w)hu^<bBRhd)1C?m-mSYY>pg(`j|X|gO9Sk_
ze6R(><dxMCzM(f5gm8x_pG))9$Bs#@>W0UD{|kfzbAw!qlC0f`gvmp}Rv3Q1P@-&z
zQm*s39uG$A$+eN|khLP&=`SR0FhsiA83mouhhPNj57|0&dp&O!jBVH!DOvxGLv*#@
zcke*ynG3d}WQLHpTQPfehiEWlKKq8~X@BH%<3AD~@cPW-#-9YkYcL6>7)|-N5Rr{s
zgKP^jkm<7aqTV%-+QsX{Y{!rFw>e>(VkI_zSSrTJ2}^(FECCoc_SzoVt=40F6bnIe
z+!}+Rcw?HX-EyGS;`3atrp7q}!a%s=HF~AiroF~#UK&j`gqYeY*mw*4@W%v0t=)5q
z)GxDpzM}qi$5!66UaTV*Sa8BM?&6NH8bG9td~dq7AkFzBy_s}U;wmlT+G1kKpVzA+
z=Nv{^w%)%;)*tBoqHHp;M7(%gHv_Ib^k3bNj^WW)rTMkt(P<~UT_vYkXbe+y!-_Ez
zHdM~Ll2gpyO9O4N=KZuFC8UkkXzahdm(5GPZ;iar8rb4f6-*xuN}JbV<5t4V`Kyx{
ztr%mZGb{#dTew30fcz5*S^(8Z9^-Ph`yPFDQbSFIgZTRyZO2};<oITvnNyJi%KegJ
zzoD4vN5p(M?Ww6y=JaSfdQ+wtyJ8&93Dqz+4%W?&Qf_RA!x=W?hL|2F)frb$>Og?L
zCG5s>s4RPHC2=i;yips#-|YLecth@6I2UG3wp%{uK2s7LhfRz_IHF=XUGicFdc-jJ
z`!!Ns1f{~2lQOia&&s8$-Y%8Bo8$*oNLVG@vgx}2J?7rjFsZZ-!0Ikb#{poy4Gk~d
zyYNr$_K~E{7$g#dx8;>s*Kp~6hwdc$d0c4AgOl8)6{uqlSmTAS<<jB~uu`R-)6&BY
z9Q;e|f+;bYjUSO@n$0KpHH({|Cqh`VJVA=O+8d##KFr(9ZC`|fIOLC9r5^+&bGd#r
zN?aRkB@R-PYnUY!0wEnocR*aHNubpqsdKudo(F+6oOg55#1KeZDpbAGD!vB6MjN%~
z3`iH`#@iV=nk*@b!=t=r`GcUB+0`NX*JW~{p`e#|H{3#M)nr8er|4NP=q3kpxf~&v
z{$Lp7g0@kfxF8WF35o|>KrVLVAJmjI%YtFJ3MXp<RUlJJ^VP1GKZ$G!R{_B==wk%*
z0I4ar1fY+8Lm&Ns|DFha`~vZ;*KO$Im(7x<KeDU&H6aW`A6@<zL0Qx?<Wa^x=eNbt
z7|GZVdDzp2MEt$lNx#qRwfwF+>BXa3oR5)<F1-Z{T3ddr+Y6sjRnbCiTi%ohn{9O4
zmj5k#JfaI*d;E~P{~mSqarNOb#cX>T<a*q`K^gB>><I-1sI0>~Rkyi={WHq#+`Rsi
zN;C9ffl+{R98*ZM+{bA+m2p<Ub`3cflglbbdvP~=FlD#P7J1gu_SPO`kG`#YXB)c@
zU*fNeU)s!uxWlKq`O}#a2h{ix3f8lJYZ_Qv@G&2MSu8~Tnk~X~YPAs2clXs3(03fu
z^7lsc9V~Z)g8O6~`cC;eL3F~%0<6Xxx$EdVHYwyn0Vdr})S8+-k9sf963cXw4UA%w
zI`L2O^iFhzFGp)m+Gc}(j~>@yM$r*xnFIemGFR?d54Rj;{f=aVf4`a)so)zJZq2V~
zgeTRu{79{AkSX4|?K_eJ{=<4#6XpBgq{nq%BwYVqz4ej5+*_NdT{>`mMQammvi^}Z
z&hP@~<GzKA4OMGBV~l!8`L?l)X8Xo%I_~c|z1s$_iOoz#_%6Pswo~)21K(|!N)8@m
z4<?RjWn}Z2>o7`iq*^TY42{*!69CZmNrkTChtUcbN0XS=xrGdZBXKyGd`)AV;K{Z-
zEm0HHyL`N2qi%o@0TVns5(1AE{@LVaWmv@qRXEn1Tt71UViY?1z_yV+s-Y~b8%l1m
zL!}iK6gqLF*>-U1#1iao0pAVgWeujtiFGiCTjs2obg9p4VE6G}feeWPLF%A<5WbEY
zW(!}JUkr;UH)>5=&2(r@Gc>1P!{UEHgkcfxJsIGrjG=ZmAbl{G93YHR03kK>8@e6?
zV1Xt9{fMk}Qq?1zo<q+{>RsqSHAQGtWiQe!Ar=eEAnanFAckItiJ(-|dca!Zgk6_B
zN8mvM11fvVbiEskSkg?7Gr(af4id=HR#Pb_gpIc;6(_$h%HzJutk0MC3BVK{A&jQ{
z4;u0!0Y?@Wy~UN@Q&JN;7nDT&|A{QT*xAMrsC7o6pHl3%RP}#V&=RJ9UXPRHhP!3-
z*KwY>!%WoOhWjP8V*(K9J*fq3D{2>ViJr;fW%TNQRjryZ#E_I$pj#$}m7W1?yr!k0
z%AAE~$BAB)q|IzLO%^B2)UR_%$&uS&o8#+u-#6&(ht-3x6TEermrV5&G=7c|(MB~y
z^z-y1=<`i*QRR>-l%4FQbq8m{xw|y!t{rq^i)s@l&^o21m0R7iI<mTz6|`jzpDVE6
z&IM-c*}1^nf&e6i3&;`vGSgZ;k%ji(K&yyNr>E1K9Cn&3pgA07{cR1lQa1ZYta^^r
zZZF5`TVyrqJy@05Uv40sme`Q|D1{ua3yozrR9veC*?l7tpoJe;*|~WeUhg}#fZf4B
zij{fw`<?Z{2V&Ly<ro^ut(-*te50q_gsNxRU=JW%kTcCbzO@_iz)*b<N1vks*4lw{
z^odrHUJcvMc+kWUh&iAdIi+9w-w{fX=k*)i_-FNzy0nJdWs~h;zwHh3b<#j?qtteE
zl{?TxZl}yn2`5amxGUI38#}xqO709s*SJOj?cmNErWL&ScPLv7Mkza5-y4j2!%*wH
zi0$&WSZrUgD;Rb2>5V}3?+SMDEnE2(9enwA1R|r?;@^o@h4=4AADciL&oM`D+nZ^?
z_U_d^9XiqcDXv<ooo?5PT5opqe(pu2<?Xawpyg*tKM?F8eQ&Ur^c|$<^v1e)j6?Oq
z*L1UP<LQWdqBlF~4Z`kqy@`9fOgFlDe^xg3r!k(*e(mP{J>DL>5B8tXw=ur77pi|B
zZ_@fLnuhkrf_;d*hP}P4y?vimE{^W#5$tqlkI-i9^A5ZzKNR~{57-#&=UoSbgM2G@
z+j@sSllolZ5iYN8(<r;q``wP-=0BbAKY{Ss?Ku6_j?iCrL^7F3t{z%Fyn2LLzKb3o
z4vuiJFj_0CW8UFl4C`vde!*ey2z#Xv9EO7-M#_;W<wzrCI>{JB<*79`t(%<3_`<vR
zLIl8VU#P1#w7$*7Ep_g;FWtR*V8iz$Y7cdMkJjeyMh}~#e^hT~WRBsbX-_aV>fU?&
zzv;cpe~I^=;Jsap@GtR|CwUu^4!-hkvpYCS?3jM1CB{94UMjJA*XrGyeBn=azG?12
zaJSqOvYU<uNBOFIc}teJ{L}jJwG(xO!D}aPKq7Gmb|Tt&KmziAyXh47KJR{<;%3iK
zDsqbZz&L*B{XY?y|5F8juHZ`w_>GymI_&LKSNjPNPuMN9%f_*DWN-gob^cJnkE`MX
z1aEC4Kq6=!v4n8LuJg+a6I@VaM>@82oMdP;<&+PrYDhTPY3*x@)fFr&_y|Fmdr1Tm
zvcj9`Hd5^=`sTk^9ZYhFzU4f2Wk<t{Z}FSuy~d6@pQ)JQzykrd$lzuR{wr!rTxDt4
zQBfQ|Iq96=sJiL@ZuM}^tfr&=)ngMLH<wJ6`2z!8)F^y_%C`<RI_k`#>{QF_^eoni
z(|)B=oLR<AI_Ir#X7AXEaAzk$(Wrus5$}7Y?VY*p6B7<LdOkHh<BFGWC6Aw;cG@0c
z_Twx4&7pa^x`b=TIt<3T@vx!LusYG1xNJ`58iiWZ*<96<?nfD10uwn*E<TUGMOmex
zMp<cKqQbYfG|RP#Dr>AIXPKy(ewSYPYntx9c6r%5d0Y0sO%+?Tk}-TAZKfINB!=%#
zAzA!uL^cvdvR`}$_sXZ7XIhcZ>l`h+v*_0SIpvxl(@4chJ!<2)V@KlR!`w5Kt55nq
zI$F_HexK4Z^~JJ3GwI09FDcnXD<;Aat>|PfKciTS`ZQAx6Oi7c@BFB8|Gr}XKtWQE
zc?D1F{r^yDM%S7$S~6>2GCg1Pq%#sZt5}PBlu3#IF;)8)#Kw*}YV~C$d|bgN6o?UW
z8xLoTuszy+*Og{MSX9e4?cu-@YonpPHh1etLY1^`dTZg3D>*woy)2_bE9=V@_8oVF
z11mlFb~7jbXv5JE8n^nJ^gGNcLw$4ZNQ|yFmFX>Y(`ao2EZG%j*XuNpQVWJ8nT-fO
z(whB|8wuwGyl18TgTg$w5WAgWo@6+kITdy_)<4_OH)VTIvs77WCYkzOb>#OHm>5ff
zDz4>diScXEg(32#Z42C%wXNt+QuY!ny7uq*)ZPq2DT#-pfBj$UPn*Lmm;+-CU=@z@
zs}e^oC6!1Zpwi)X96}@1lKDu>bGD9j6=+X4(OApGQR&~_eY?Oxcy}ml1k1z>cXcQy
z5$}F`_w7M?;S5paf7_+QrGCAT9*pn)&zbx_gk-0UNzvQX+MxA7j*r??^d;{4J+<1a
zHzwq0!Q6Iiuo7<V-45YQmSNHUF@9Ha889&Cznn)GogstEl}KTIn$A3={L^qJ*qF>^
zv5kv=4qX}eJYjWnuCCOt;+ZMcN>SasOa~%gS_!O!IW=n1TUiQ6@j8xXQE$S97&n}X
zp0S3pXT7=4UooFs{g-03vZT=U0aE)V4xx4Kh{Jkw5DqvqH>8Vi<O<ize^zHU>;hiy
zc5)X|!Xrx*g1NxTL9XYPhpfl~9^h@BvJ4Gl*Xvx0?_Js3bo=YvVzrlNs4rnU{8zMK
zA0b#NzEm+s`Nid>;<Ki!d-la*E%Hz4!&c)0D$lhwPLJpB6PS5{x!^cxzF7>Fi5!H;
zn_M0Q$#wVDsjFkd{-^XB6JHr<{UY`J8%irH_8|om3MLhNOu<JLd{DvN3cg*1-&C-q
z$BPQ2w)5!*(4o-*N!OG*sgmYBw%`V%jFp%FournsI=G{!%lZP_i5A#S{4^!Ls+}l~
z**RF(p3;4XgREnBpi(~S-2q8_q9(OZQ(wooDLZDPO)vNyCwZM7+&j)}sQ*Gnwzf4m
z?v$ik-IB`GM(c?+%wODj4R{P3MBKoqoz6YtSuInC6XEhxtzJ%SKwo0M!V^v&WZE*U
z_PTB;wC5e%O9LfMzR6Q$0XPpShX-2`eGMDPJsO&uN{|8pX`^ZT3>VsLz}mDQvHGxz
zUsrLn7m;VcOQ$`<r&94=@za|OwMJotFLzGy028|zV%cA8lrlQBI>uEw_wkam=f=ma
zUuVzy-$h5D*Gl}61ADDU_YIk|XUF+BmY~pIRc{40#?og=#@dEIq?>Z&5z<NOJxB4V
zT_dk};AgDs>^QGemg+eV`J^ZTgPNv<<(&VlKG7OGckb*t=TjT!f=AuD{vLJVdkMm9
z8OOy_oHc$;ef=q=7ZkH`HN((fP@1`}{9(m@RKbrZFy#N!ikTGaA1n4x6#O#<|4zaC
z6&Ur9Qh)3W^|vT(T!GoS^($jY!8QdlT=f5%f)NF#q%{&^Opitw80BFEh7gtc(DyZa
z{$JN)YX_N+Oxs3?Nn+<IQTs_gEf)i+WE$wK@wfL+a*rg?lY*Z56Y<`+b9r3c=L=hK
zi{C{oo9iv~XYN(${@a;*k&AU@3dqQkxW@0z4dhbf563=&aisG1xK_L0&g^cq-~IK>
z18wos%@PA3q#nG%jJ9%PAK|j9-Ggyo{hp%B9Y25r`>CDePs6o-%Lb#AluMQhGRsN|
z5jM9z;W|ETF4%QeF7MA$twVe(noj6Uf1}H0akU%20ky(5O-Krk$0-~)%A(EugQr0-
zY<1h>Nzlj+p%HUnnHq<`F*qxYb-BsUoRzS2A1^Tf;J#@G5Ab-a;Oqv(@A%)s6ag_`
zRWVzRKg9RrD|KCaWid$3Q7F!q8YG=_5Go=$@)n1Ge=^O-w>b?5?l~IT)<p}(>P>sK
z@wM)Cyj&ClXNI;lvDGU+C-y3S(;@VD>!e|aOCHh1P1re4-x(koT%j02x&n-o>n`W~
z$Z~Y{@qir61aTxb$*s}ZbJP$jjg!5O($0Dn3Ziz8Ah6uMke@RRxK3AC`{;&sg4G->
z**kPK>9=n7j`lXJHt*X|J4@|1ob<IeYqY{9z1_SeYUw;gt0#HdT8rBHPS;lJD_Z^E
zP_IC}pW|z}B+sr%ZG5L>s@A%k@pG>&dTCo=26P{1^lhnObNjzBEpEkqL<=n8680)p
zE6&$CGzkS|l?n%Myw<lV)8OFmsI}Mi_>0PS7tcEr$v;=+Jqn5n#H9QGS;fTzgo&w5
zPum{RayvtL^Xj;(@c&5Cp4wjcvd}md+40nH;Xf#iSbyP<rR(ixNv66Ae<qTfShjG>
zbjGR7+uS16*aVG*8Xe9>O$u7zmQD4n3G&_6GUCUe@ExJ107X$f)tb)Xy=rIH0S_KK
zewq+TXFMAaNnpcEe%4us0gc`$0GR}hVx?3itrOhdph2xX7k*%_SIW9nE|uuX_jp5I
zzc=uPuBzZ1Q7TrFT6j2&6n-x+nXsby-8kraaeOSn9XN}ZBf*3mQ}j7g&Ux-PpO@Q<
zDW=4J1nKlqf_mbVYxfh#xeA*oh3Q*b8b1@WCH!73RW>A>)QD|CKKUDrl~GDWR%@)7
z7SIwM8E$F!9BjQZSE?V=^7uGGq^1e`TRnE9ncpnZ<|BU-i1mynMi5J;{DN3Gbd5}+
z*N;vUZgmiP`gV=O6a0SVUINr)4QIQ+e!g!rU(a&33#c-`&V2pN#Q52p&S|qbuhsQ_
zjTvwNd5o6Vd9(HZ9wXF-rwv7af<bwSZ~BU)8B+K*+|{l`jvh9A)utnSwgVvd)0Tj;
zVYc5P5~OQ=(L76S#B8HFN+C4OaAWgeGY4(36ib$$G}RLM(*JT~L4++rX9UYBJBHQ;
zR+OTVBM3nk_M|X&fZ(`pYSA5MIH`w8;|+4KBgK*wqwhsEnKAqEZUBFl-ySdfSq>Ut
zqBKRRJjcD{tfvp<P==#qXr0ZDv@htHW+S-N=um;29GbNCOB&VZw7_0a?3=X=Q|F#P
z^E@Vwod60-(Uz-GXBCFUEjo>o69gI1JEvAg6?}VR)jEZvNuIJ=#Qkm3n?F?TKT~ma
z#9b4rC2&UpNtnKZLi%!;Sh*5*we8d}kA0o#IcpMsXh9(}-psX3NqeBkdsB(+cq)N-
zyR91^<hO+N109E40l5G>0}fK|b~dabX)NtfbsOpY)))09BE~yR)ifsBZ(aP>US;<L
zJ#8vVto%(y*)J6(3QKnY1!S#NW@{`bJL;B-5>*%K2`5@_Dk&S!PDwe0z+XzrVg5=<
zDa+psf4A}%rhE-kHg{VRD$7x^Q(~-8ZevKMOARQAnvoL$T5-RE8XMN%>q4!0Ua__Q
z>l|vv;A{)6b)(jNn$!VPN`4adUp<iuQh)oJYR!Qtg&eORRm}h>xYfcDYhe(z=FQDp
z@P^*8ehwAn78I60+vrn$*q(odUBp{pF9*C~zDlaO{q;QKye$}PDloU#M|eL9Pt5Qh
zLg}c(jiZqQ^Z&A^t=<+XFsYrXZ>w)#ql~FAM^IsI<x6-*Wyz67RG7Qjy*ntgCqSWz
z$}%=rlmau@>5Wnnb>`Y$dWB;0hErkw7G?LFx^r9oK(Ni*20YkJY`3@FVh4lW!8SMB
z-VT`J-8`cx<ZGnDd=+c=U8r`34d)!f`YM6Rom7~+O@+B<wYWj06R*9eQ>8<1B2^_V
z@1bR>Dqp3Shl3*DurH7zut<8Bx7SpN|Cziayr~;9a=)$H1J`4W-m9OCW9_xiM(WKv
zDzED@|ISOj>Az9G%RO@{Ws6GGK5qDNKQQdT_4p&GRxi#|d%xBG(Pr)Kb9bA1^m8ih
zT4<GTmU8d-waBvUfOqhX9LnbA8#TA^AFcOT&sxun;vq(H|N2qHobiBn*wlXqR}a~|
z9b`lfqlz5xjxZi$cT{x#NyivV(Rq}&NZANvWL2O#Ba{m6xPDCyceknGj;xMtn6pG}
zXGeRnIVnH6NsJY7S8x}-InMmZt=S29^@P)%eOk~Lkm?j@{y2}N6mZdB8Br+USbldK
z{H)!IA8S%NRFlwIr^G<lK&~7CnBJ6jP6!d#6&}8Cc5(;Fb(q<Z@+qpmE+DJwxu&Ju
z1~A%^leQVnHIB*}x_Dk6@q&V;M$i9dJ-(>mKky#^CB@DuFbe(~Ywi4(Rrvi1%<}1V
zwIaCPBE>t#fr3WyNls^%i8RdXqxHxtkId_v!@wMzQ-d;@xtVb0GluA_gpH9b?|y|b
za+Y`Z-2u^TK0^2ew(E}ll0M>t3SLw2tqOFpqW>WUO~sWz!Q<G8zl6#tOwBEFIPU>f
zH&XvaeTG&*o5vv&piL^RQ<>CKAvw9qmsI5c6V;JPaZG<?225Yji#ioXV|TaF7@KH_
zd{3Wf%Ask+W)w))W9l7$R<St+&uMi?cjI5s<IgGf1*HurXsK<!OS#{x;JXz_i{sBL
zW}KcjlfR(HssdYgi;BtpfvIRpndPZ+ePPMJs1(VAO-JKv4>@z;x?;-;Y!!XKVz!hd
z&RWuTlbEWP<6zRAB;q5t<@QqglpM~P)@xmk8m-)^ZP|^ve&fFAZ>(*RUL}LJq!2%3
zloNG%-w{G_7Sh19Yg*RpPjGM)-7@nIVBsd5sP4dM6HtDiK3ltWyiGh=*pyclZ_p-?
zG_uu1e|KbEuTkTD8p&Z^ki)$h*IMFMJF~?J*@oZtRbq<Fj-Pb&RO@79o3~b`JSZ{5
zYL2NDGuxw^DzC-794Fnx$cl}{Es@>Q*v;LF-QJ+r=k*&STQEkp3+O%+M*g(U&La6=
zu2!0y?8<i(ijAN5Ow?v;MR<9<p-fNXaim%=PP+;d_^hohPn|bMLrsl&QAI0D_4B7q
z;oNHJ+@idv>yCg%15(k5I;R;tB}J`P%<rT;ri|k($g1B=tQ<2&?zsBkl&9|5$ma6o
z{NmJw+VPY8b0_wr;!>KfuhL%Q<jLrOzpMtvn-}uIq}Mt)I8oe9F8y^L@GOqn%x-GL
z_=GsBvunDOF)s=RQCbH6Jsa`wj#r0cD`QehC<T^oXhaQG=D-IS8)Hxua@h4=!zAo|
zC^O7p)SmrEp%e<@y>2Ah?sVdEwolU!yS#8^4ChbI%~N9ti;+8?8IGWFujTp5)P*G=
zaUCZxl4d$!+!PjsJ&k%c^!^XBD#N_I)#EUJp~iK8VuW;6p%93jBH|!+dz!ov=*Xwt
z?bk3KRx|z&5tIh4t>#PAsDDGbpI7j#dSrmH`AO^fEJ;U0?Ih{;ySs4RiL9aVZ#xjl
zu6K|4tF!Bub5rzic1p9@_)&*Nh)DKOtJ4XbrXbccPKm4CJdA>alv9SnO$@=M%ghTS
z-bSii#r?mvooj3yR~5(CyN<o~+UxZr&a=+u(b!v2w}~jCh0+E}MAX(oN|C_r7u)OD
z>-v$*uG6yGv>~WcffhaxAyJa5MdA@eTT~<{ZzT8t5)}d=)Gm)ABtEnVp`udtL%{t0
z=g#cRKI}XyV&(44oyWa1bMO7%bI$*s!%X#F6o7Gl`^DUb9t=^~yHSDMAo6xRjW&Zh
z8iveUV<&duG7NJnU;88yHCx>GByMU^oT|5SkL}3llJC_jbo5>sLn~bvX?1=#zn$lu
zaa7J^6m~9o;t+RL%gNJn_?1I}3hri5qoN;p1QN8a<hRRcYq$}tAitg1QArd}bEJpo
z>Spz5d|7qt4fm-D?hGiur-CF$EYNf%^+|1|*}@48TYcwofQF&gj^s0}D!;(_{gqTL
zoSmI%M_%9v?g@e;(ps6WesIFGS;32Z1@G5Nq#`NHEe1XZG(ylFv!mr5qZWffElHWI
z^Jkzb%;9BsF-z2=v6=a)Wt?Cab5XBNE6c>V%jFAsivU9d%x9OTN;C7igPA~QvCC?x
zaixtP_I|G#yrkr1B~>MVP(u5NmD6_@Sk}qKcQQoWk}datWEsCmxF*0^o6CY>ZF-A8
z>KdE#{7GAXR$^w|SF~k)MP0-Dt9GAatJ;mZ5of@Q1^jZqs@w%7OC-5oH~V-^8EUQG
z9ZLSL#O`jB!GEw@#c#(0zF*eg=<q+4yrJaZN(|~PF{polyw}yOZf8TjH)P5ch|2)M
z8i&(E2}J-BsRC^;?wR7JGmt1)nRi;=ue;aw8Lh%}!D~Szr{?ZgSSH+oxt(GdDjEFV
z<JFM2j)Hw1713yWC;yA$`K(XDJ|(KK9Bf4O8=C_1HMHW*>aTLp$d>5!Y{lcJwy`pM
zQ)L(_SVsb4cChbs9>Y*JosMHrWNS)jBxWbkV0?^x5wp<=P?~7WUA_jRC*MuqEu$gR
zeK8r>FLdk(G21_q??r==Mw$=rAPq)OzK=UR(fSVgjDJV&+~WEq{ShiM6l}jvP_niE
z45Lg&1ZQg6Z%@^4W})v=9vY0*Sn(H}F^Jy4#tm|%K^k}@e$3pca@7D=iMV}3X$ZZ?
zcIdqUaeGHeiw0Tw%@}qwnhNPEwxIXGTvU9`_J;GgqB41WptQBLjcC9-P46*;-eWV*
z&o$^mVhhl*pSlLU$4<@}F-K5x0<}w852)C~oH+u?J8LzCEG=#LiI+Kdx9Lc>m-bY)
z=eMBk*u~Z^DB0TDTS0^3zBqJja@Zd<g0=J>OGvyskfpboEREg-Ey89zU+g@G9|E7m
zeT`{AvZr=8Nz-k`qxI4>Z!pkw4CY68W6~5YQQvDTyJ;zVD|;zhnvq_p(kh?#|I~g;
z5{a79MSYeY!_@#yL4STXwacwvJ1Erq^LtFQ0KwVb%qS)0DmhBpOF8?RH4Bx&qr`GY
zbMky$?a9qu8%q6e2%1f8tKrW{v=!w^{&(`)mB`e1Mh|8U&QEgwill60-%y3zO5$!I
zz~)9TLr%R4X*8U^{7mn|M*6?$9bKh_`g7fd@ve#2DB(1B-DQiP`}JXFm0YDD`YTnW
zUK$kznAzV|NTErjP_w?9Jjy*LJIRq>u11-(0huINWnj1!ALErc^4|y^c|<Z2K5Zm^
z4#oE%vO2=wMj|LvM*c*=>z0sf0<474b`w9{Q4=~Zc6VUjdaJ#A;LY~#z`J%c%7w}m
z*3_NX7ci-!fe}-Ky@@(TOi!c8`H*hVqeOFG*S-YT5eal-;qx5V{!ZAQ<tk&FbjB0$
zIJB&{V3w?9wgmi;c-zo_Je@tmiFN-r`FlmF!a|Fx6GZbpW;c~hwX0@J?X^4%0#RG4
z<uS#?t6|iRnhxx!N1W&wN}n&9&?lj07XRx^(7e@U$(YAsjN#%lU|ak2y)jMl#Pp+i
zSyoOGs64k5&+9=ZW*n-~`im~+*X;P*I5WZG{kpAxDc&vdDEeS2d$KT<wMKirC?3u~
z=he%uSw^ixyxnS#%EJk=NyXj;y>RvS;GbT0O~yRPPGAQF3lhH5@Aim9iBu`~uJB&A
zNShs6ffJz>xSENUDdEnKQ)?og_YLY0k-l2kcMIx*G@{scNJ+3!eA{m07pn<a6U~-S
zNmzQ$qTGW~=a|=toPtyY<rA4%&p<9zKK)6?s<(7v*JO=0(oI%RNjO8ZXD&$Y`T)|2
zLUA*%lI3#&Sz68!uzF&F!p<pI4Q<|G@Gr2E_Tpv@S!=;UkU%ZP%NFiDafd+Ill!>3
zZz02+rx!m%9nQU{BGcihjcagjpt%ta*+ACbN($GPY%}-R&vl#8-qYwg7waqE3tMWw
z`&=9kUYop6AnElv+vzbP0&33)tEq|dyGjb%Y6=mxQ5Vv+SPNEJbndKZ!P35j?f#wy
z`|7LzU0YjoDr|LB^K0dOg>Rc1_MX;g-e6`m;lqS;LoEXxW8RzK5*l1J7k3L6tf4%r
z-J7{zHDz?x<134pGhHx}XI0D@ogk9R1{PVjF2^*UEr7=RDz%2La=&+mjmh@unfV92
zvurl%15Ej*wy8R%6B1T3Yw3grafv5_2e+Lu0Y#1j74$5yz&5qtngzB_Rj=+?>ymPp
zSHk}gMvJ%_Eef8vx+5!z#cP&7BBsfT9ZQHqT&ux4wWym+Z{tk%5=^Y{5YOAeFsUc1
znaA-#Z6D}s%C#|ly)VMDrv17;RG66lxS`+b7LV!O3~=r+&ku;Lu~ePKvK?!op0w#5
zMDQ!_eO;BYdVYzm3r!kPgC|2KYcyZb_rDB{sGe~;sGAQ;&>TCu(28ywZu4|}hHBUS
ztHID+WJh5rpE78XcZU#y94uj(Zh|0bM(ZJ{w;GV`O<#FrX=*N_1U=wjC_oSSPG4GC
zUYS|NQMA)=W@n2y>og}sPEQHGEwT1V3y9Vq+{bwge{dI#MT`6WfgRNn10!ldt)rtg
zvBe-p%SlYQ5nIiNA-tx!bNs`UsaqY#u-fN#(9#j_F&yzHi9C!>R@yuA8E!;Gv_T+j
zO$`l6pZALa0EoblZY%HgS6jDuU+aNDKewrYfJpTWxNj2#93+@<nKM}EjKIQCd~g~l
z)F9x4Eg(P<vY^7CT4fj~j9^M61=cJ;b%CG&jY6~<F~y8OsN=+#n!}g!m&GXSemp@A
zP!Y{Rxq#C~yxssH#xd{HZ@y<sYzJaG{Bps@fDaotaQ&btI|%<#*ysZ`FuHqB02G-z
z*vJ@cWG2ctT?uRin)tZ@HTW84{j7FQ7iL!Ze;B5!t!EtNP_=hzu`stx15sqrqIWyr
ziwVs8`KfmL@k@;ZKX>uM>%YBw`op(5TjwE8aF{c63Cxi~%S~s))n3nQw6TGt1~T3^
zm3M=x)ILRMPNi_>&**BynEr@JWcsf+RRyCTnbf)=#(2-@?gE)=m*06PBG{yy^+4xI
z^@m%iL<l-kMT;orU`Jal7Sf=LZ;+8rIkDJT4?_Dk=tHf6h(e7iXe`0w>;!Ye5QJ#T
z3?3HelG&8P-k_-_pgloj045qOqa=;CeF5u3G>1k@bRTrgETpkU%Yx;tYOpMb0J^08
z5_Bbf=u&&Q$NP@n-UWS%TS=;$>=g!_X@fBPqTkkO_E}%h*7ry_sTG9%P{+To<OfO`
z1|r?@CwlQ|N*f>qpRGILwX!TCaQD?}K#Cw5d>-9cVCFriB$PL;urWQ%w>WS%C`ag&
zZtq}~Sw|LR(y{fC>7!g8MW(JqS0e9(Azwi9NvGK96pA~aHslIOr%9J<d8vGlp;VXI
z>AHFMebtdVdw!GXS-#KEC(FN@tRu97w01py2(u<}WXB}nXl>LADsv<zPAIM{5j@~H
z0EQCX$Q#+I)j37vFidlSBW*R*#7uy~JZVhru{JRh^}uhBxKCzc)K=+;$|<rx0B8xD
zjps^m9nEGyi(^*5OT#y9w?l4M%ua%3tY#bVVXr)L1PwBEER>~tCsG3sX53pR0tPcE
z0hWTw73Wuv9b25mV;!n~7q`;LgQkDC*{|7h#c>VVqaops)i54aV)WMyf$>GP6F0+h
zZtc|lZB5|C^618&s|J+oGGV<>sS=-7@);#R)zB8USZl0$gqNs<_c$BncV42%%cQB+
zaQY5P{D0xJAD2-g+=swKXs4Lf(zsYnE2bS=68Yd_dq+!JQP@?2E?7KbIInYL2$OlN
z%Y53&FOI~#899gb$*}24I*3%@-!0AEO5ie-LwNip8XV_5EkbOX*+L4nIC}_l2THhI
zHB?Rtm8l7aF7LB}P=_}^)*znh!&KHN<~_N-nD_20!Ru8cmD9|xKURuDw#0?<zUk~t
zcI^0p?0k-&gM^B8?tPTS?o2K_PhPwlKYoD!2c0DNIkg}Y2L~@3noR<+?qIZ-=jd3G
z#o5-Gtc#q{0UP1pzTD@SE;V3Y7lL0=&ufBUnWKPS^+2sdM4`9FIz;-Ug34*_t(+Ol
zX}rmSW0&YItzo!M9(+eQ>_ufkAA|#0Fe3e7vgaH6t1$IBnRnSlK9u>M)u(C!er@zi
zQ^O2EBnI?ym#BUO_}bm@%_2TwANKz#)!{6~+sZrdAr)zYoQ;WBB+dPZYk#S);rK0Z
zbXfP9UQ4Eo<+2;p)EVCA^{j_nNxFYEi6{4&`SImtw>;ALU|NySQu>eEA&j7ex0mZ}
zn(^Y~R(HOw>m7${7_4hG9kN1`7M2N&?6l|lMlKp9s=UR*v?tv1q{31>h-I9}uPhcl
zDjoCwqC_L!WWN<~vY=J@<Ey0tXNa(09c`+Q9R-VsVFZ_j0LjyYQO(<8MS|Zg&CX3p
zwDMBAI=Pr%vh%$jm1Mx98H~9_aO^h8MxIc_1i8sRU4bcjb&>nSUGel{tE17}TB%eu
zQU~q109V6YGCAyDf<1hEYPGcD$NO?#5yYXW?ykS7&#DQ^);sXFsb1Ta?9e^8MebRn
z;f$O2ft@<;0!WOba|LX?Yn0_;Ta2IZy$5*Si0<|;HNKp-ZX~Jp-4%T7yY0!n13IKd
zhqlhgI8vscHtm2-#9X@y0nxUEiq+(M-6On@%VWi9lJ;0K>5ST1$b{=zMgnI(=xG~c
ugixLqHt`o%`;}u$3<4cP?Xid5U*<i*miryJEpuJ$;Y^DEeVKzpH~kM_wQWuS

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c5c9585222aaef5bf362e3f1ab39d43e1986e008
GIT binary patch
literal 59692
zcmeFa33O!FdEZ%k6$*tNjYgxl>b=qIjm@2=xHel0$tFd%M7boY2;ddaKm!Hzt3neE
zKiD#xl4x6`XxScbqz3Ic_Q-MUjN@b!#fzNrIEhE&lVi_#&b*UFUSgS!<H%d$acpOP
z|L?wfRai*LvVD>{Cr#GHefQmW-+g!a_V2q7_w*DK{M-HBmuvGcCKJD*i`Ks^nFskb
zzK~2LyhJ78CB4*qvXZpFsfBbUopfcHO4iD9l{|mb^M!@3N|)Ws%oi8BE8PpFO3ALX
z^F0f_mEMKEN}pZl=KB{0Dg#kFgOx#^&(9An3{{4#tS~>ku%oiW@?G;I3p*=2Enl1;
zUD#FGW%=&;v4!20-3xmvdlvRq_AczJ>|5Ah*}qV(lot+E4lEq39JG3+`9lkbD~B!L
zGk;{^Xyxd_vC1*K?wvorFkTs7I8ixa*M0N1ES#*ITsT!ZwQ#y}+RFOp&n%p+oVEPG
z{H+V;D(5UeIDgy1?Umcf4|&7h4sYc3bmb23q&MpAdMR1C(;M@4zm%xl_0EL1$J_f-
z!rNQR&)r=+cOz4|hq8U%e#-XO?zxe!+{<;@JHYh;)uH6;C^_gIqU2ERUas%s`mlF|
z>mzpk4z7=S$GARLdk6RL=lZxe&h@xmKfv_~?-s6avFiuD)H8|kQxi)}Y4L^Gd9S=M
zyV6|tYvsl1;^k&@sd4X_Gc&Wz%gd9erxq8^%r8&Pdj7<0y>Z5?)fZ<Qv-O!%-t6@B
za$|O}eroA@bJ3r=d}eZfaq`Svcb&a+(wm;TbNY^RcTL}Mw|CoZ(|6xKdDr9}=g!sc
zytQ^`?Y3L*ynXVvc1zW`rE06C_$^!SonE@WT^ARZYW0cPGqVdz{^GTo*Y2I`RqdYJ
zZ@>GVb9di1IeGV8cTP;*diz~>d3Sgoy}f60a?(5J-E+^`b5pgrf%c2f%uFmd8nYAi
zcF|k(;gQ9KrT5P^&NL?$G(b&%qTZNZ^cQM=qj>66nZHxB#W%IFE{-R|!s8PQlitMR
zPmia<F26S6Q+=iu=KR`HZK6qG@6=+wSzBo?Ow=c4YJS*VU!I?j%6LA%G~qXAC+5j!
znpeFA{&p=)T&Y!O{8^8SUfR1-U9I_x)y4Y!bxMkl&o-KmEY_!IXT}R*@BHFSwYk!)
zPBf~s3o~J^w!-a3nD=Va6U+0>Fgx2=m{<z)O!opEp!FJqFg-uf3`<js%k^fJ9!xAW
z!ksLP=0u&&G;7sqe{rEoFZGgSn0tJ-`RKxgKC#Y7Ow3oW%!hl<&rZxN)+grQ8IA0D
zUhL0KE;m_CR2_Ub^WU6pUcWHwH=55h_41yFmYa*U`qZLV^B;SBJRJ_&w@%iY6V=Al
z<(juVUkisAEq`LFS)Hvn7pu)$y|L(rMJ?D`b$<3rjSh96KmYD0o|#x!ny>j>_P<-7
z_W0rq1N_A7%50sozQ)qTRIS>aU8pshwWUT_n9>JL@!j3z{KZCNqS>s6$?J3@7v1S<
zm!~JZg-cT@%`^Y>mu;4KgkK{=5+vAr$<@gxuhsk){Mlx$yue4zpQh)t_3{hsx^naK
zY@<BAT%Xc_l$(oXeQUX)4QjQ{)UHj;+hfHt|GdRoqg-EXmZvUj36&@6*B2K3+Ud_E
z7|cwgHa{I^^bY?Buean+XU|+-T&SJlq1xo);*~SrM04WIjK8?Nbmp1m`r|BL#<YG$
z!+qvF8~x#TKYeE2TR3A~1q!tG{u$3(TBu%a*nn22XX~}*a=qr&Y@TkvgN+jQdb5p*
z$@!XBwZ0$IIN0E%{c-YB8Wa7~Um+VCs6AwYM35X!G!t{lAmt^4^lb9g<S=DBHGrv$
zVa}^vo1Lohjah>~VTKj&jc3A?e=EOpVYYF3VyWhzpsn_&`^UKq2bSxz+L~;Qs%z#H
z#jamT|19|_e(}G3QLgbCQ&~%Z{1O%V<fXi{mzha;Suf}1y@J=}6}@h9C9lWp^)g-`
z|Ml~CfM1sX`uJIHdf&$r(@Ag88+xg$lJ=ew$^cDdy!Uz|-cEAaTF!ggA&(o$O1@TD
z@4A_&6>spa>69UpB<0->nN&&+fmC`N;;8gGbW!PZXrj{Z&_iXwp@qs|ZOA+EdeXba
zJNZ(kGVEISo&jv0eo5`_h)OPbXS}mkG7^=%4^VlVmF)E1@7?a*L0?C`JH5L=Zo9l^
zy}P}8C>irU;N9ze9l71!ecn5`x5vBRdw{=ty$8LA_`A=0*n5P(`=fTxdyl@9;LT<4
zg7;3!4|wnL-p${G(fzOY9`~NG_a5>tdQWoa@W)v)l_RW$N|-r+{;4P4{OvR|o1145
zPWWMJ{@le2s|7a7nR#~XCBT`s`kUH}Z&ESwrbqInQ;+QVcfWq;*aKhshT$K3=dZ1P
z=L28*v(NngFF!H)kq2g4vma%?{6Lrn8cda=a;>8RKeg3(l@;Ag+(@h^Z}Ng*r<c5v
zYm7BhUg}2j#bmHENL16!Oput%&gH&8;ibu?`I~tob0Z1Kz*?{7gI>}?&_mi46iJIg
zH)(fJA}s~GNPC*SK{6P*ks{yM?57p2sN5?l-q}Uj03`$dPc{d2=SGr}p=QR*H;0=$
zC>dHGxycycNb=^w8)>~c81)iY^dvo;+v)#eFiL(j7$)5n43UlnJXzfx6jaauouI(g
zo?s8@-eBKrNk7xv7i4cFUrRPJ%@kjnRXxhj)876d8{~q#*Yye{qSaRE0Cmr*E<G!%
z51qXY_Vesz?sQx4T<s5A%j$pGMmI5cAQ%kNK`Q7A`hx+l^h%B~>{%-}4>k``TAoSX
zVBFUZG!M@mX&&X<59q78I_CAhqOm*f^{pM8Nvs{>cX;LD<~aR2aU*#?@!ahnO06AP
zI~p7b4qVC6m!A&~t>5DH2S<a0yAz{{n;NsDYX^hF>$e0)boE;D>d!x$V22OBln4%;
zXMEQV(b~!8;q@cIp_|F(aB%2G(i^flVg*!B1!JVAn`h?E2D_<wD{x@=mDJiX@~2-*
z`5$VYo4bwt?ab^B>+S7zR>1ll-birlre=EYYpEr_dFKtSn!BiZ7f*dFUw8NVJvUW$
zuO|q(^Wu2gzmLTkrdG~|$!n`;r<Q#m#5#YyEP?=hvb3}a2~nRZA7o)41aEldgDYpF
z%(a87J@=NMUar?w1JtzIe`?9Ed9w~H2@_lbTjoJylZ%boYX4)k+LEe1`ou#G2+4Ao
zJPmB_o}B^n*Q!@)*BfC`FRmIAS*<=@Yt)+d`r5*Tu-uXmoZT$HP@A2(+-#gKUlJ<h
z|K&z)dU?M50$8c+gO?Y5y-LIe-7eF>Df=d{*3^}yMezUW)$`?B+W*SWKJ@g(cVB!b
zU-alRk6yY&wtVq@PgLn_`F&5Ff9TSq&y+7cS*|?#^phJN8_$K^Gqcxf^{PmXM%c5F
z;^=Bq2_*-3s*pl{Gb~0|wYmrWun5Imt2W*gl3~oUUi11=Eru&^iawZ9qJP4T!iCT9
zYmAZ1FsIKAt|f!yoWYm%)Xl`Vr><t6aaau^BqQ`4rl*#c!)$$dVd*;Ko;rOt%&4=Q
zP~V>x@0*9*f~;S<|J(Emy)TzoJs=Qzx(ewu+pJbk-#5QFH8J0~pG>Rvc*=i-y8fdi
zyr4KeJ3n6qzAo1wGG<prUG@02tIM;@4@BsM2P|Hkxi;aCe;ufJ8B81&TJ49u4>cM!
z7T;q1QQwD3>o;1m)~xxns}m{?vr9~IGwiBXr{*UbjcQd-*C!Tg<ocnyJdN~B&8?=e
zXKH?S2~=*xo)>nDxUI1?uUDI4-?Zq+s5<K~EE+1Ps?$skL{q&s-v~!pW7Q?5O0S1f
zHENSp)(HDW0vWIKtUs}&56Rjx^#>V>Ft6uWf&Q?ry6ldw!}NTu9%eLtVSaLXdKx|;
z>`~<w!OCaH*@$KiyRWGa8V5#Ym}NdQYn_>&T&%iX5%$f~by2Xbg8afFvqt@HBZ`^8
z>Di{=qwe(Dd^)21%2G2NU~+2BDyyWryu`ApHB>Z1A8UTxuP{v=UqnxssxO6E@g~>9
z9Mp8P=7C7-&D+j}eeu#4xy_h#)o8Wm(`MK^F~4*fzG8~S=r6JmG-~XGom*bjdR92P
zs$CJ5=0V@}s#`k#E$T#utv?_3sMGOF=~+GOdP*FKR*nBYRqTt~cjL{@09=Ko<tbPl
z(br5T92S6XaUtw}+Wx{{(Bp!tEcx}B@#FqcnyPd|$2Ox$u_f=Hr`Z1{y`~EoBkrep
z*?&lNVWlQK<FcxH%3o5=X(i7qnNTvRq^jh@N-ijQr;>Lmv5i$%Zb8W+iCg5h80od(
zr~mc3H?L$$$wx>kTIuXD+qeEz6@Q#W9Hr)2%Wy|~%DHS!yT7WZZ8^qcS{Y=g!4PWl
zT60(Fs@Gm{z^_tj)vSs!OwLw%HdM7a7*F~iR4<Iz5RK$NOL7~(MnIA{P)znGbICD&
zo&S}Ry>?wn?n@PwOWEC2!PQFiCf#$TWFc9iOlgstrBsIY#*!KPZ<tzx<oaoCG^I8M
zsX3e~rF!|h+Oq-1P36@e{`K;c08Eq6_$j000_X)FT5By|A8f*ZKy741&3MUIvTJ!S
z734<~YlUWNE)8V_4*>q|YNpmR-0AY5(Lb8nmAIK&D+c+w?1z$T>1Hl~-U2u0X-8CE
zmb5GAn#+3WUBLIb{FPMWBh8|h6TR5%ZkFbH`1N{uT80W-@AC>!gZUdtqX2Z>#dTN5
zwWvktsrY_(&`po}gRYx=OI%uVrRt}snT|U>c4uIlJH72^c5i&9uk()A-)<3#DKR%V
znh5g430f?UChRZo+^w%2RQb-tn&_H7Z-AbcI(jnbdP47qf|56MGqs_g!x{(svf@qn
zLD%2i&|UF2?g+ZwJFQk>yIOO@-pG{{bQ6^B+zxN&P{I=hJOY2&9TaY;U46k$^?P?>
zt*7pDHSCSrGv~csYPnfjAJynu`(v#ueIfN9w)(rR{x@<D>isLf=Vp>83(YZak8uq>
zUO#K2$4lR*G3)d8l1ES|YMK^BJMOueT;F>$!FPPNS$q+$6uPf0x^E+Ri75?VL<V>|
zGl8t<&Huu`)mt2>E?=IwhIC=EJXtFv0D@fc$}q^~+Qby{h)5K|-hK}agju777!vYB
z8$mrVw012VYFzarIMPM}d#!L0LJ0BLa>Id{_U%1F*>8Dwmp$y)roi$W-x%(0-&(kJ
z(@p;|^>Hx9sL(4uqP(eExUXHN7q`kGHfF(9-prwgKUKQ<`ak`;aJXI31YlE(wdq^8
zEj?%C_M0N8{CAK<!~VsXm3+{@pW@a2g=G`hL<Be}wKRWielo)5jSEar5h9}@q+D-~
z?=wv9?^3hZNWy%CR{b><ol#L56zspK%l6EFK$n7X4vFR>d02iLYy|@MKdJ{a8aybK
z#TmmqVSa(7v9P@0aLuRm^!F;Uxz1g3BDF52sbLbeaOmP<^WCCF7U*itbHp5y!}>(s
z|Fm9^o|vB>?=_t2u$-Y$(U2WoH5{myK`9vJVDIz4LnQ@?@+4n&9Z{p3NLs&1+4u8n
zJWY}i&HytM_^%h(UjQcvmyD+RQ~h9u{`3g=LjUbc4|BgiwU3%({Pv}UUCLmV!~9Cg
zgFIW}iPaIq8!>IsqS^jC0@(nV|2qQtD<hER&YOrl|Jz3({%h<r|JzBvCj8ND_#-0y
zhc|+AS6bP%sZ?mCb<5!Wru)W%R1Vp#w?ZP(^9SD!(L|lv-^MAMFiKS6K)b>gq~oeM
z)HB+~9x;k>?9$=(o$U~iYov{a#@f}}XejE;cK6Ojy^sXRe?!S%4kZnCq9oTAh{^vS
zX=w{a^4HaVjE(%8TvmoQuxqXb7z~rnIw46^*54MDM19_fx%?kc>wim$P?rCLO8&MI
z<CT6$Ipd6eOu3&>^5aT=SjqpS<TFZsM9Du?@^?tSdX(h<sGbr2GkoO#UF8fXZikA#
zk-NXb?{5kf?b#j`xq;iDbP`7+HqqEFXvKtoA8B$O`R81EJ#*8^GF1XEGM6^VCKM!6
z-uvKIBqL6~ffO~FSS@;q_aoIt$|*V&BE;IsZ)nFoc`bS}<(~WiPe75TFRt#oxBLvM
zEh*|mWJ>B5iTGA`Ba$<Lo^C}vdh6zR!LiYzq>v#&2GsnT^2JzO)TrdA7MBq7EczAQ
z$;0ic9@46#c8#b=TGUN61JVk3;RcV^x*>`iBh*b8zbH<zH<^OFbll@=uMK%zeM-VP
z{oBE>z|Yvm<7p`XP*s^-1Wa4TOE*(NW<4#D`<0x3G@xWxVhu5Qt8^{9a(9qL=zMl9
zw{jv#kvq7S4|407Ab&IIKiouFw4Muc!m@~cGePbhiK`Dko49)Sv&J$)_OCY!QUJV^
zXm+imGFb1H0C>H`<kFgi-fM-|QY*P;&(`~19@e$+O8Pt^?B{+Rv3D2jYFCiClJ~!G
z^|QRKYgb~f7v{4_T57_6B0?{SQG-?ea<dOM)vmm*D;WgkdtlXIeZ}~~enmOhpJZc?
z>Y~Kx=DE)X#r6K1Fi5Wbwce<WqH6lhm6OeZpa-_Mdl$WDJjT}gP>aB3s$?V~b|~n(
zPeS*npCv^(+Ui4=(fH1l_iuXs{9512eNkC4YV*Wee`lMh5V}3?jU=YkciIp7r&IQh
zAJRLne(G6fb{(cvwP4W(R?co}aUU%XwA;zjPp^0vnjvfDPwaWIvhMl8_$ymI8r%3h
zy%-Goc{XJ;J%~Vo|2~>h|JDXq4)e`}bG`9wi2eK4U~qldE77|_p1REX=m`e*YMgDP
z?@L_$EM@7}QjMQv+wNd3qM&>wRY&DCH`2al6%4Hnv+nu~0QB>wAyXF(kQ=t#;2Q`b
zU_kAT*Uj9g{Q^%61w-GTaNFM1jqV-h-f%G7x@Q1k$3_1MHV%AF2|*H~K6L96`4*Ar
zVnQI9-D>%KZN6sO0VH4LZ&-7({&t1avA~?2*A|#d25SQI65o<sO9jaeQMthYxZ@|O
z6lNq#4wKbQfWdz+H;)P=h>t@se$2N2MltGk;_=mk7iOJ0V-bUz2)YT_AwO$fA~w%>
zwZ@b**3F4|hnK~o8e?t*mAJp0%pn+q-WV{NqGL7Z%mY??u-CyP1qFyyHmm(*q>d{~
zwJ8{|vh*ygeNRO9ALHuYVtI9__`Y_PsBnB|BwdyC(x0_4(bF~!ezPG{1P4=dQk73M
zop|xGDt2MF;5Qc+5u0ZbpwHIBehsJROhP1(Mbzy|BFVQeP-)DnFZ9l1^U!Hw7GdlR
z3IuVKVOG{Ejj(IE-nhD4t0A55Gs$Xu)(!fP?{J_!6H8$Iht*`N;rTzWO!7*Yn!OSh
z>XC(t|1<VR!x@z>rl)F_0obG<+U&?$h0rQk*8ilnHdiUutH|vU@y>g3o@e2f-8hHE
z_6%%ZPd;soX+6!6Ng#;q2f-k`wEucRHDXkqa8pccDme(-3HpI$9ZvS9_Hh-lvdMj7
zX!-9ftqez3V=%QNY7-&n821nJ(|@Z&+pOCu?G^pY@e>|1+@-;53ZqESnVd`ANCybO
z5L!Y42+1SHi^{=Xs+WncvraHNod(UPCid}}W|30n-#cqf3|WvSA;xGK_mr<5U#T|B
zL3yQmcKmc)^<Eoo3YloXd9OY8!HeaU@+16CpFaIu*~u=XRWt?9e64|2u)eg6ig0r)
z(eqHBlQmyY#KtORGv_HEH}O;X*74KjcR@~^ET6?R3sWt$q8Qm=@zc?yk3EZDzqkBQ
zxw(vxgJ-FEvfLu&PnJ7L_wkcugdKWi^d-j{Wm{eC1|yNsmme6<`qJ(B->Kw`5~DEx
z2Xf;O3;YT5Q7d7xiTw`m4Y6i$(yL0T^Szr^T-bNDYAiyl_Sb9mY5OJeT0ICfhHY(Z
zsumMAS{PY=wm6QH*}6EaL>hRvT++kn##sW%i{lyhUEj@P?z_H+9HsV2X=dO5*A$?q
zlvUQ|o{vTf=7>KqwUt$@=@X3bN)0Ag_geSgs_tLwcl^EhJO1bPclf`qQhkU28%mDE
ziAD?2qEnM^`3gs+U)k_2hu-#EBD&tboTmReW~RO3zKa&4nd$hV`{R$X&$2HX+vbZ_
zt0151wX5;hWSE)#Z}l}%?XP-nSW%nS7$C_kzuIf;Yx$CRg<<jFY!#!PU5~$G>k8Xk
zH~tDqSL2-`;bXK@*cmM_8hu5pBxQP!78s36*Robiz|@{w%RwHaZ8EUR_vYN5l8Dpo
zsn?Y=9`yUjjpuE5`Tt6#@zM-?H%biJ^*68A?yGH;^EYW2#5yJzh7l8(uJ%XJB5v=$
z()a(QzRzIgKTys%{l8Cc`|tOETb1HL+WgT29Utv>*~{B}`pD)_cU2LDe3;hbap>YA
zUI)`$z=UC`7^KhvqVJo7DZG;MkCOvp0yWVSzLH+c8u^W>JooMlGHcAge<8@Oqfhj*
zPMZa)&FTqkywV>^ujSVAYX#X<1qE9h`HQA<fY)sUqYg38mw&d6Z_=BV2MDO90W!(S
zZ<525bsjU$s{bYz<6RD5Cs$m{`r4Lh!{Rdb3@cT1F(TQWxr6YHjdeH@uM(;7W4+4e
z;qP18h@-w;gFnby5=I*hLjOpmTS|?zu+^pqrZnFASK=3sy^yHvWcVegliveV`QhYR
zR^t|reHtCE#<D<ZR})f)+Gu5Oq}PgQ4YBCFf*$uK_-rm4WU<Z4NTx9dJL;vR4g7TK
z)dWm2?68*&x`q?$`J0pm#Rxj(q<#BcdeI&?69h(@H_Jzp^2G-bZ`g7^q58kCWH(8e
zH4TkPGR%PWvI^5s_~RwF(KogqmYm6W)i7n4ob-QIO<AM=L5~i)IodEC!dqdfzF0Q{
z@r8-TmCZ9$SgJa+-Z$F&VgypNH+7<=NpIDN$;tw)YA+b&@PlL+byoixqwFP}@B@uG
zw!EzW6ze@ZlSK4UZ)Td=Cc1TM=ViHP_K|tC>If&EyJ&NV#vE-s;)&@bvxR77uIOc7
z(cbFT-k>ebaA^dR6*0rd>?-e}rDtzFj7X&?$V$Q`@kxN##Gorq5J48IrEsw|I~udq
zqM3(_yTS_J|2#$h&ykR~A$^bXZvWeCB1!y|>ex*)R}oL+)s5i^3&^cuhH9Sw1yy=c
z$^Wb5n@KPZTCC4FgAg+fkqDJVAzH~+MmX#UPzQD3uViB1aIgB0Y9KAk{T>r2$9F%e
zr;o&mA@H8oRIw=+mwreyI!X=c$|UW|B@dzjD<=1l&rq^D=<s;70_>qT`>7w#w02y)
z>ke|EnP#3f+nN=@xeWL$Jq@$Kj+4Vl3bzOU&#B86nc$DJiIG7rz{u#s;^YJ}LNm;=
ziXGK9eNfJ=&3~XT(W;Z|E+@7J1Apw;+}#*Hx9Fxc6&v@GfN#@gfz7Udsc<6&v!m6=
z{(l*U=SKQv_OzAIdsh-bkr@4vzBr}#wf?EeA%2Y_338(3%gBfvDV%crb<viCsY`8I
z7zD%A_C+>3h5LZK_=4R3(|pb!s^y4~x*APDr@X<{WW;2)vYCnKB<buWZM>k@wO;Ms
z+G_o$G>AF97IPVhIA<o)y1Qi%ZF8#yBU;$ZfJgC+kTzDrq(>!?Y@67E?MXI+XTL#<
z4rp$g>g)`LOxJ!+ozy&vFj+0yEZdE^e_8J7XZ>?Xc+S9qe^`vls2NYPIMe<l5?G8V
z*E03{wQRi<K%zqV`DS`9GnYl?N`0a3tCOx|&PvSCC(ToZX5Pc-8L1{lnB%#N7@Obf
z-_ENMD#1h|!NL5>TQ)=4)&0x$EA_<}>adD1tuk>dANz*b+->a``~92rXnJC@;r|<Q
zwrczvDl4{}KqltH!h*b5Xv5SQtsRpQWgh0>n;8syTtJ0Nn+CC<vC@rys&O2q@dW$5
z#BP@$B7@LjBvnX7Q{So^4{eF+8rn3!Mwu42?jP;^mnA4jduiY?FaS*hyN6-S&L*H}
z*;ii1pm7?jD(NE9UY6^pUrum+(ynt{&+}x4mRL|XRdWylqUjo1R_hFVf(ZfQ%irRa
z^jsdF3p_4n(ugz>cyqVl|B%;LPkH@oUBG_qtax_djZ{z&ww}de2XVUPsSD;t@C_ab
zB>pWpj1GoxJvYd6-O)3uhalT%c<8XG<sr4)Y89$oux|J-2_^gtqq;qK@Uz_dzjYfk
z_ZDQ3)D#Qa#x{^4qV(E3|2}5Tw13Y0WZ<cm(PU(5TV5SBdz9-)Y+{=fDf?_Snt1XJ
zHE|aSYdh`DE+D{^swPZh|GQc;4~xhqb+yY~VbHRgwd{D0166r_LQM@B%~{apXLLEJ
zoZf}4M{K0&tW(Vr(hNy_Sx2@AppBqS)Tn5m>;Au2BV8I96C(TlYI%U9QWW*WUYVX<
zk;x<#oC$5EU>Edu!{9pr*&5_urG<uy1R(p(=1}4}_v=p$BBbk0jY1g>Vo%bW8gA{Q
z*7LDmPJ2jz_DOz?J4vK6GP(zV!+wB9g1$k(27Q!S&k96w1^o%#18({>yX#dUm-GgT
z1^&nAj=T^}YzWk*EdPub`4T}!K*Q)$Y`}a>n{!uQ(BEwVZyvEPcfCK>=NsrvA55;6
z4D?2~VtPQg1sx3d-cLqc3ltwh#2`I~BFZ?ReWLuV)Zm2hE_d;a!45L>N^(8N7gNF)
zW-q$smO6cLb#StLx%@(Ts(kAH@+5z_U>Wu-)bMh<UUjl*#^AC()4p0A2~%hRZCD1R
z0i2y#Uc}}s5@i1Y6<RW>tkE34l(vp=-vl|k1Px8r_@6Zn0)+`9nk2Oa6*=`_Txr8x
zYgBAR9EG&5QDMe`iD+R1C|#5kf|Xd88q|Q-Px*&<jsHAPwx;W#3JgbTDf{?4Z=vtH
zku!n_kT}(r(kZ&>(0MNqK4RZHD#T#Do0k+?UHnZlD^0qWfaspHg@?`&O_m6EsfgW`
zrVpJYisFKQy9Y~(el3YklO-zsnrZEh8_5nX6(bvDksoJg#KNXvZ*4IHX!Gc#E}+XS
zK#(U;K|@E}Ofeh<K(>~A7jAU!zrrg3K*|93q=l%oi0%@2NJ*FfMOrA{6f1z1&@0hH
z0qVTh6BIYqF0izEFNVn_e+Q!&MwSr*ae%uc6kWo9uz=0_Omz|`GC45F^)`>2poTUq
zkf|*#PF=<;1aD`&1XO&%_y?!}zml8xs~~a!)yHYn5&gC>zm-c^?8}R|Ec&>g1MCEn
zr0a6V47X@nuc%gWaS4k{oD)!@luR^^I|Z(pjqz^B8fZa;U5-|s^;W{3_<<^yZYk?y
z>?2KyVL;HWoRhk-X^r=^;I4g6K?_JLR=l{=P-I~IM$K3b-37*Vy`~%=NkxpphR0)=
zpfxCNP}{Nb1rnx-9pJJ4qjm4Dp%LtwMYfLr+$Z0oY3;1}Ev@UQ=C>q`c88vLsqiw(
z^+xhr5+8@(M-{G;q?IIPukhZ)(}@r8G?abDd^F8ZGg?q7yw!BQq3q?vHf70;Wv0aJ
zaxmC<z*rZ_$Qhwz9DxM<v%Fl+I}kMNBR(D`;ycWF2N0Gqw-L<++g>;!OrwC~25PSe
zg{emMteY=kAl!GfWXyf%nC3}uL`!+xpfcEa7N-t$sWy=v#6kSf*Va{^zy=X{;(F@e
zLBZCZ-rDY|e-|~5x4U?Ot>qt6Gsl&TD<RB+m_cm^{~z-qJ~0+r`VZg0-mSXzv)saT
zC?1pgq8pZN%3KE+&O7zs-AXiUVc({qDXU}7=tD-K(!ccs&G(poZMY?K+VoI7c3RQ;
z86H<FtT5;ptzWG*(U-PHfamG91&dQ&rrjRiqps~$qWNzRTR6z%*?s~7i5bRk=CnF%
z2pFnkb*D}B)><*<QOy|hoa6!=EK`sa94)&q;&#ErmID(@>V-VUVcyKe)scD&&pMP_
zcSv`2bg~>}ReqeSad$^+x$ZB%z}FhE)7bdh`Yj}zcMN$W?5!)<K%<I%bCn0eQslB_
zQ42N?#ZXmfm1>cwUanN<G#28ug@nbsfZ;}MM7F8?W<ebh@j7J)kxU;y^-Q?0QD^++
zwrc=}!Li9pt>qC5=D}Xtu^OaQ$939>mb@8FVVQ$TR7y54V7xq_BXwV*nZJQ+2+RqD
zkKO57S-~O#JY~3X4%{T{RE$Jbui90Yhn5xQr!LnIgjK~Em=q7f03jEGNvvjsfqCD0
zFC<ctI(>3IkZZ^sE*`Yu#9$@Znt=d25JX*d!>oe=4iNMVd>hD4=QN9S$fN9&qlP_)
znS8NP0>M3Ny&XWaho`#bdVuXqueDL)DoY!4%^m9_O}=-1r>vV0t)<*(iOA#r2VO=Q
zXdoIZla~<P2_Qd{Q0aJ|1Dj#asc#K@uI6sFg$53mREt%#iNNoPb+}p$mBLf=i)LzZ
ziH)Ap9{O`0^Y0@UcDEK3732NNnz0>WU|IXsz;bUJEIW|gUVjlhj-Z@F(6&znOboJp
zNb6AT0(_kgPw3nn3p<&z819ab`XAG?ht%EY$%Se3(2iaeN9}}qv#KIg=_G|!<+PQZ
zOyCEUGiu(rMPsPeRQ7-xFaVlX?iyt{-pVVlfmO(@&MacQuY-C>UPJ|}!SwY+*d56k
zW&h!zlF^$FtF?`D;j?VqQ2UQ4XV@$0M#vzoaawL1?$H=YE+Xli3?KRduZ46D5~3ea
zE3*Xw-d=d79Jcp+kVcBIC?_+L@$}iQZjb5|ud{y};G3j*Y=XpE(n;`~!JX*qIh4-G
zBaq|I5Ho^JoSVzf6(rh`JTiiICX*CNG~N~VnCGFK?c~I07BP-r{VebBKceJ)N^T(u
zvqp1;*@wJ|g=bq@R5cXljAUzsX==1rltV+q@FGM;!O^)I(`=2ZhN(oO2@_#5JPgX}
zF8loL>LYRTZ&OFkDY-|9M#29&lDEAl_TmzZ{RNgc5F$Yqkt|-^QrkaBfq}|P7srqK
zAEK?WSe17zkyUWHEmf;mmnY_(1Fre>ncIc`je2udiLE~yn>+|w4!~{*eck*uCBLcU
z-zxbnC0|nVrjp+w!CM)7BBus133);tvh{2W-Q)!(DiAB{%htqgCZis1)8#!%?p0zN
z;@!$UNfP!%PQ+6+T$Vd5z5qBM)cb8$-%3uu#!DoHObR2MukFA7cOzT7OTSdI)Qc7A
z{e@B?$A1NWU4^dR5>s5@$GuXnm@VXreZ~EDKU)~>?=55sdF~wX)tf1P@joFi{cHRJ
zU)7F7LIjbDfm>pb5Kkl{fKG%AVu;l8>xG*R6eA-;(hVI@>GD2kwO#a)qQxG;k<=oO
zBq}9~I)ZDc_aSe$w}%)Wz207LAMN#d`@J%M`@I9+LH-VShrGl59rTWPNBKMCJ!kPn
zl0+1-_j(_;_bBuf$d9NZuCz*6tWz$ki1)mA+B?HLc1HKkdbhp=vLL32;*2P!h+>Q+
zh$u2)QAQL|L@`DbQA80&6j8)Q7}@7dT8xooWxqFt+0lcqC%uQfhhNH82&1J@@H|gZ
zMl=Ekqmr6M8&S!jsASsWj;Q2tR5If|<~?pt9f?Yw@Ge>(k4D#*y(hh=tj4jZWY)Xn
zz0XRH6D8$YdO055`GEKJ_Us9ZUUCZvaW3q89B@-_#!lv|#RY`|GPn1O{vFKsg-hBi
z{QL3)Q#~$cN(T`&S`&>=u^uc=h_n_EE*2pKxg}nv(#fQPUddgqBpWB18RVy!=fb|<
zALRXey)<&t3`}>*q_Iy4mNuU95|s3=i$9VdD@M5FSzY8Yo+aO9`2oupXH;jfnY#hg
zjgKOJsk6zT_q$WCO3sSx7JpadzxcaC#8ivdN1NS9yYO(85vzw_p3=T&6LmJX{`2q(
zt@eF=qMm|t>u!9}VxjccMO_bgJ<UP9Lh<}``tkK)1oe7LVtohlVwW31l%SkqmF&cC
zmWU>@T(9o3{8-Qz#VDEMsoj|S4yr9r(O1TTevvnf!5$2k-><Q1?)8YIg2C%PO81)&
zZ#f`_iT^5h`gJF&{gRDrVy?V?Kv7DJ{vWt#^o{c$KgdM+he*O<!XPwo6$iL`t!M*P
z|J!u`6_N}7tI9s4<P%DmT$Hs90x}R^i}+TH{-yuFgwHq_?nLHBI4uP)lRLfpV34c{
zg~P#(ULvZ{`qIhkFtMJq7ELcLF8S}*Sjd~4I9wKw%VIx-!l|ub5XLwu+LT34Yn4LP
zIKJcwMzT`yJe~jx<2i7-x<IHieB22w;Gv{x_#dE(^D_}DGvAX<0nXaNB*Kx-$Y<?t
zgrlA1Q8X+Q*oTAdS`tP|ylVR2O7G<4BGz$>S88=he!?O7F<=deP%n<&;2SEPR=+Ar
zq{eX+icAXp2Fdu45qXut&X^a`=SmRT8j`38Vy%G_ghs=!=*z20KCI+9C8D|98nkiw
zpe|)j=(pL|c-gjBZ40foDQ3;u6_GiY$RvdKAa!KuDghH_9?^dzl<L0}!Y0*07Y!xo
zOl9p3)S);&<xA9Z?WmR2(M>=!4rLOPjl4dQ=l2LdbI&n<4S=JG+*1NXl*lvzkXgk<
zYaO$!^_&?a*j+Yy{wAPBB&Nc#aGeognah7Pb)Vbt8Moo#V4mSE{?kg{ql7V&X5ZG-
zH>==9C8FgO9)@X|^{{_N%BOfSI7C?tI-3$`wWF$*rp9J+(gneD(;5ZA`(cf;@InFu
z;c{~I;O23$5Fr~LnG!**e;OS<D)zl0Zz{eG$0<_4G~B6jjD+P{e;Kic%28QhcyTYv
zIszR4HgleZwh}kIUa+z10!d`hycVh54eu%yVJXOvli<$`G>p3iRXneI3T>4_$<}`p
zz<(33NP%hEl`1jQ@+WA$$G|P-P#R}#2L65dfz<~g<S%0dtl)SUaF?f+@eNu~03l#$
z%YGgpbE;i}S_p_RQLmvcn7_Wd`_Q37<@Y_Ka6!*pI)94bL61E2xUMKSSh2dd6~#z_
z_D**KX&L{I5Bl0P#HVq=mz?z2H({^G4F_*z5%0FN!MH=Z@x02)%AFt=4!YonM0SZ|
z*@UAKk8~8oA!`&HuQ`Fs^hfn59^V8kBo<AYPi+~?9aPJbt`VS*=s~ixM<gThu{ghQ
zz`f3ITLy;7)h+v)05uzhRR*$tR7-G#RuY-xw#iYMv2C@{N5$D3-n=BYYs5Gl3+eVY
zG1_7eY^_|$dTiy2<8;B;Sa%X{nK9)}r!COBWpx=@l=jcE&N4B&PDQ&n>rhRTc5DlV
zI~tc2`=)LPEDn?7&oX?i@f1{XAz~g??t+qcDtQ;lmVq<g>8CYL>UAPL0{e?-@4)7f
z+3Iu7Gxjg5$N1^J231=^z91zbI$tQM75lZCS-D+**RwYfl|OgXiQZ8<rt$Q;zz%bN
z8Kr9O>QQ7)_E&NsQ^P49=#+v-hJCW^6B9_}PvFNj1iE=&5|uEcc$DEl=Snn_GYjzO
z4{N3UbL#m5{w>3W98C!8)O6!#l>DlaBXMFp+4DU63mPUFyEutoAte)bY!!}e9;{BJ
z+tHU;F1lgJq@#SxAxo}^+C!)n4O(wJXd0a3(cq*mwno6dQL;=MDQBfI&I>{WSB)tj
zWlntps%^_C_-h)QZ&dP@LomtRU(pamLQjP@kX7XIxBL9gp2Xk3L2mMcbdPULG*%@K
z0i>6k*g{40EK-wd`szA&WSe@PJJLg=rNxdS>mmmlMNJ)9M@qKLvdi1DOPjDWUF}+V
zkZ00T%(3i>YiR&?y4rnp4od&1y))hTShEB-qPL_QpHjZvntBic^>EPW<&G(>R#%U_
z&m5qv5U&j)f#H17SS(L8%0h0g=(PVVqqo|1spVI?x?2^97`K7=jv}yLz*50TA0w^;
z&B}9(k`oa~AY!ZC{|4RA?7Ee0!Xd3}qf}dxAe&fQ+uJSc37Ytqn$t4RN$TT#fCyED
zcuZpL0!zFEOi)0I)uV4W(+-pC`~9P|8sl=o6+uOV5A$RQb(q0(5mkUQ$YWB<VBU;V
zSp?a2K?|_4GtLwPb#)NWMQSj;xn^!1BUxc>rWa3Eq~LjSd9b{4U9fpMWYy$@Ja7Xu
z*rfk?-il^mFk%1}Bj5pf`?Y8o{IBEl#*5C6Oj6Hq(5>=L4A@4Pj^G!)CCA~3sVlyq
zrBg~o7MAyEB0o=QWpAsCM38e^*uY$5A7g^fN@~^f-=Qb&SF#aKg=40C=ny?SqFoA$
z-`6}3^OgkoPxf{4_;MsIWb28|bNki}#rs$vqP=WWd!CFYwF%EpG$Z`GAx=~EqgI9r
zo>F!jC35SG_7}T2rCn^6RCyWhF=cpZ@7{v+<;>(v)>d~Mi;aXV_?!#)LLlj);(;^7
zMAh+y^_|SlDBT~{<!49;QH|l75j)VvqG$c98b;A;o5s?gr|i!)Xig;RbUeA#aC$4q
zdn@`U9x9q?W65h=2;`aY<DwH6OH_iVwP8vI3$JTxU;r+$0wQJX5)f;o{Kpwsz}s5#
zDt=lRw%OhKVb_M*ypd`=-z)&YWTPOVGbIKX0dEYh!WFy0Wd(~P3G>9qgDUe4#J~gK
zIVQf<|F}lvqe^Tnk7=kx9)@{7Ru8oJf#lStfpg@b$*^G>1s7Vu;|i>p$hf`;31-fR
zxC@%G=N2@gG2O0_DJ?tw6NeyFYRm8aWMuUUJ%LC@BSTPR4ENUIqkuCq;ll*T?xVb4
zr^J;4ZK9g;Da6eI`U4vg7fKDA5u-hfns*YEZ|6OL6wU(>1!3DdhZ&psj$WIsy<l1r
zLSPa(J#wE*UJjEl_y>5-QM75I7;YZ10wKOHD_;5zjgXOI$#SwR?ZXy+vQ;%62UU;9
z;Uiq=JB8Q~XhCI;CT0*w@x_$5abV@h_?>YG(!_Xi-wiC4mCP%kN7>nEEEvnoC_#cd
z%5y5`$H)XZ8z%z8QA4Sob5OUi`qRa7^e^(MynJu@vc*!}eB*_CaVGYfm)qqnUq=_H
z9VksOY2cUkwKy45>69lIeQ3dx@l)j&2-JQVf8=_*@;mUToM=W>%|)4@Qg*w{t<!ht
zrfY3VL8JV*F*nt3FqGyBE$yBrJS<$P)t0b2Y&bSLS_ZierP!<5FyEo%ypm6nwCv~=
zvfKZ55?KqpAhokX%d$WGI$<L}r5h!ECyRc%NpwWS+`4kLBonH15e`gV>G-j30{c*W
z@QPJ8D0{Sq3})rapzmRJ8vOU3%>cghkytCOwWNJzdup8wYsD}%Jaew8pdHSDRvKxO
zQ4om$+bg(Z){oW{0qfH2I=FAH*-OLR$Br4NhnDH>u7v-(waQ&u&0DKxSnFk@R#{0h
zxefAabxS*7c!llLTiKoYi<Lf$xysZ>k;_^I6Dv`H&1a?9oUK(SeL~p#m7(|%6*?#R
zRw0kI3t=t3OA}{1J)@^8gN>zmLYtc2xaHVw^DbOyh=0qJhpES}^ILH&beDS{e1E0;
z@#y6eL~feY9znpk>Iu#2C~t7aP^IKi2=FKZB{rXsXL~o#@P_AY4B~knq#ez*&2ukJ
zBM47?DB~Qv5V@nCl+^(c#Ee15K-8!%6_*3Z8I|)lD({98lkNI%W5{HG!AH~+Z-+rg
zhNv9`sGQ48qZ)XX(1LcAohIzK-m17EvQI!%Uy`VP4FR|;-<{Qi-X^O8YUXcdCBd1K
zy@5sJxRG4PRSHO6zy=|MLZC0Y-t_EnVp{93@3mBY%z7&OgFkY8RqiYF*637v!CH=b
zMULS2?)$V8bjJ)zmE(@Sd0_KbM)mZ?d6%C~@y!EUzS+G;->uwN{&v;EA&H%~bquz&
z6^}&{r$5bBhNvBX;pR`duW~H1W390uw_z-F$Bl)%^HyV_dabdrn%j<rYT1ell>5qK
zp<1YS#<m-aEp4^NqMm7XLxT1aql7gyEnkYl>PKU2(25FhFUVp>gs9>*V|COxS)Mvs
zzI+lZ;C9^@d0Ax+<5;p7<n9qX_M@kOLBx7C2cr+3eeNXDGTPU-KG$xqO?hAs+TBqr
zEO@i?E?Nl(uy~kT6Hr(I4!l~}jb|hw37q?quvlcnTyj<(AEU&ftQ+LQ-23X&#FGd!
z7fAi@6R?w@&v7s&GdAER>KP%Tk<60J`QNXapC$=Y%hO@X<2TtL8uFAMV><u4Rp)z1
zD&2ZV1TDkda`Z|S`roTIdRo}tzNu2`sD5rUVC)lMvkI=Adj%RFq)&-#Z+cV#7D0Qx
zsdDNt$nUU;X=E0kN$yFWPMst@xlP%LbSAaBe>)J`@usORy@h`gHwjPj=0xKs_<**Y
z51a-UD%GBUPzuffsXUYs&L3b;a3+ul+8oMWa10qV7FDPb&E#BGRG5ZDYC*ZCaD+kQ
zKK7~9q;p&s;gn^8)qN%H|0v~(==3r0E|_zzboKAqj?Vet38TeT+$Zd!;&+$oyjcNn
z=Su$Pf{g!r&7Ppt?1gXmCCl}}WrXyqU)Eb|Jy-w0<y+6Ij222n^s#RQesWYq{FNLU
zp@S8CkEaI3oglF4UF&NOQO=jq%YOD4p=416XM(=jG_4&+$x92?r){29j3h7nI#yxq
z?Ct>fnzaEU2qpbb@D76T(xbs(0JUpZFepkFMQDM$pLS)kU!x>P$&XPo;%w4_0rt3D
zWQk!qfsylKy+SB=NJ`1a2q`KZSeWKeA{`uH`tK@SQEeW=UXf-HKvq}WA41&wKc-dw
z<0M3~5g+2rL1K})e?|AjP5KK;YO3-HC8ohLP1p0fG?|aFT((7Lt)8BkB3IcVTia#A
z60z+eSBV!hJIygkJ1{BfV6p6c158U^<T)LsL{-ls;zbHMXU+O6G#HLWwpN|;EC)F-
zk92~FDJRKSbb9HA&|k)$pm9Wa-SK~r*H!Q_Rg{vd6WzBNYYu*x^1n?V5|30RGo{AT
z^5M>`;KIFCGA-ppT{~OKt3+DT3dx~l^F2ad3OFaB%B4nAavvK;QZ;P3k_@=0JDAR}
z>y;Ptt|#PTDx#@~UDI}Z-H5tab|r!?7H~Yz1unZ6;J_0YaY$5ow28+d*XZG(1~6Tf
z3?Gf09j}530#=z8l3NkVFbcns(&0I47;#ghNC0X@x3Svm&^P9-nZNj^QIU`FvKE_}
zG_bv$_sI;k_(v3%k})VTKRN@W<Np!2oRywLh+!cTCDp<~!B2^obmS5<-1M)TP%11k
zmyw5xiAbcrQN5NhQP>DifGDcSloWyAtZCgyxZ2wLTbel&mNO$IOze<15iMw|8gQ%Z
z?7!;p`~OHE^j8>u|8r_!<JkNESY<z@L<qzGY2{3j{WIh`N7w&Z-P&eQ!#!IEs;bCX
zX5jYY@o2N&J4XBD<~9CyjchNh5+liSHIYKwp;#-*L`0Dk$hie=S_guA|42QP^ed6t
zI`>;uiCcsknu&&p4uHKFQS(|NY64Lz#85hw8phF?Nc14n!nSQ|ixE3WPjp^{S6eZw
z(l9mCbX_*EG@!59)cM1d{j@qSG@k&tuy}WG9?ni@rmX7mubZD(E3?TLrKHg_rM}o_
z7$LDVAbe>uU`vGf0Ru3*RxdpU(`792Th%N8bQJl&V60J8%S;F4yGEc!K#Vyk&B?l|
z%l{?FasD<wp~&pKqubc;l5CT(Jx$E@P=f%A3bWYilgN^}KG^pnJ?Rh9QhQ?tE)#G)
z0b)0GriASX;ivOK7LUN3c?4n-sws1Ht8>&ea&n6;HTdX7RlG&X|DoiJk~Y?qO@hB$
zGgDK}5T*o>4o!YSH&*Pf60^+ADE9}dGDfcDr1Vczs(COJL{v47%y27VkM8Ux!HblW
zB*5s+YI9=Id=PQl+bq{KQ_p{*B_WAuB6&1<0_svYN%%xTI>&H|*t|p{<C-{3h;$*e
z;W+PUGEMj;6PlOmCoPYqe`8OOCM1Oh8@}0a%}FnXft&(N5T0c=x~GZ2Arogzq%%-C
zO>m6=;ULraFj^4&;0P>%j3m;a`UiQNzh6mN2_G&WmrVpmJlKqeLXEZb%A*>rf6XId
zMi@UVM#x@BKt4GB@2Sm=n<(t&yp`5LOaDS6=kyQ7R0`pR;>r~Wdk$t#ZQWWOPi{a6
z+GdaNGe20q8Gs;D>Q<Pe0RIIm#hJvF?v_%)EQ`m~!|zsNOYM8ewP5eBt5jcC=@xpc
zt{`vpw}{zB<+&?MdaNfRR}fe6*Ys^d+zBw<R`tMJJr+aU4PrNG{fsq6v>;y@6XOFl
z5ajZy!@7r9%q+*-4vP^)L~luWRs^;Z<pL9%z!e4PT<`L5+OVO>Pe_4PEJy#|8HLB=
zlp~H=c<gaP-1!T3<bs0cS?D%*4iRw!kXoLalxfgOMXJ*|kRH+R;-)5Nn~hUwS2*-^
zp&e?@PR_DMrzYyh5XW-5Yvfl*J7_^?o9-Mij{`^Rp2FW*;Jru()NZK7hhvTR6{GVx
zr;d(iTYULM*}R#=#((hQbNXIP29`Ou@;H)XeE`Ap6n~G;h-wo4Q!1u*TMo7f%0Z!t
zRgTCottJj+A_CE^l|1eOjNPs7`dsV!T&wDXs9$Um8u<GG8gbCjIPYK6<qjnSO0FsS
z+axV)_ODcsYVF5L{wmN*agWSBt&c1DDzt*`RSH+;S$|CK=FM4j#ns6r@84?u7HKw7
zD8e7hxqNkOGbC$QkFm6td@M!`xVT<6<pbD4gt0uyUqKZi%mP+Qqj5hg%UaB*L{G@D
zEr(bS(s#Ss&hmC;AH;K7PD}j7F5pVSUvdJbwe;0XxE7(=m&GlcJUt;MEMAU}x<go7
z6KtVNC(5^;I$y@ha_M^6PLr~g-mxga3^>Majpj&Y7Y}mN^X+93Ezi~&C5?N_6CL3q
zqjuRYs4%p%+}pW`&WhdA(4>zk)RfLave=DkZeGzfd>6}6EBG;#8LUkme%x>&S=F*F
z@OTtj`+W1?_{lOyrE!XD)Gdw*n?a8vb!+R~*^ZvIcM*Ig$8ec6OtGQLA9PNFl5?Fb
z-+IzT%jA?i9TiKT>=QWKrQ!PAF}~3r8ppvNjIHfJv=N=%Ym?ML4+ybyEJwo>xV9($
z7A0+OdvUu<1RemSe@c7fQPuC;G>Tyw3%ku5VYnL2fx;WL&g=a{Z3$ywx>Bh!CwZJe
zXoRACi_zG2565yMp<Ky~Xe`mVoN~H34$4V>@+CF_M4UM$pqI1YKl08a<PR90!<-<t
zA~}&TKKc+5WrX_Ts;H1uYu#oGfhRhITnT+I5wVexl#s3<=~SF%$%LZvUZOq|i@6(V
z`&e{`L!UV2i8_7U<1bqhgLzj5St<!p4}W`&Jk7;@h4i$(_N%YRV{$>i^>rW^ASU#W
zG-=KJ3qOtIM=_k`7uW}RssoOCxa!4UKC`xi+yF;<_69>}7&(qfo_}j2fP@S&r+0LG
zudbr+r3Ip2bAQC^w$N!i*GAWN1-q>0-E)$nMZJcg-qjqUe3&CoccG`m=9AGG3&z%W
z0Eh;!{`RvLM&7w)%g}#r1b&==Dr-8<XeS3M?PgS6ewXF<5Iklt_r{PU{i4?cE!MlX
z4;dJtVvw8f(NlA~=628RS=&#3FO=CHBkA`A`-ySg_ezfOD6fqMd)M~|`!IayCs<24
z3f58%_A2bcSJlRWjy4oZ=70_(joLZj+A(@>@UJ=WO9@!-0>5lA6={VR>2ULo;z(_!
z`a2^8iaju7Iz=Y_r>!FKs^KHyiOqqmzEX&qU9FzN2^v0q53RLWdI^(Sy^4vxSdQW>
zsdfN9AdDa4Plc2GAa@b0yC_Nx|D0gUEhHSRDd0RmaUGk)tmy=!Kr^fPGoIPrjURCA
z&C_~Du*C5xzo^UK)TO`}rwXD~v}5qvz|3((o0y?*&?DchN4j(np^N18J1P`IV#;P`
z^J1a}W6@Xjs_RNdNGcNJU8_pymUD_lXY;4e@pmfuH#9VU&`J4?aWY!tpXrvUWB=!s
z``?uOuS$MHW#SAfQEb0R^I#DGD}65BUsT57uV`R7Hq%*4bX(+c3lGq!^mSaw(cIi`
zD>-ZCTxDcKI~HNb__V*UuO{L2C<VnxH5>lVsayX+H$)Bj|B0kha?d;0Z=E9;DQ_!M
zl(a0ZEGPv&lNOrC|G(6X!~&I^n@YnO;3^xB<7!$wW}c<ic#5k;25F<T4#|EvLRrlf
zusALz534kl$qu4~CuAYz7QZ4nf;<z>afl-VO^vJQ7UG6<Kp;m3Ze#Y*+GO<M!@O_G
zs5LSzO~fPQ8($^~9Hx-41j0x<0^luyB25_ZFfayegN6J_JIuHM4A)U}Xlmgk!7gx;
z-9gsB7<2=#x=ki;^-cSLzLt932Jd_7Kx^7kD6Q3?8#%tX&jccdjQ?8DgTDfLU0KQp
zIW!eW|IyuoPY~VU-A&pny#VE1Xx*eI&KpGUUjvKq{$EB*FaV@4NgIcI*sn*|J$7yP
z6nNAiT|qW}|6CJxAD_>rx0Py3^BoklHuu}bAT9c~CoLY7WDK@sW~Xud!91hp_No+M
z{=ZXVJIh3RZ*tiZtj8h-_kK;1oM%iT9t&ILT`|xS#I}!#B}|)>oSzN*?3FSL0rFt0
ztK&DLY}Lg`s${f&U5qIOhJk5u%(qs2n2#t;w+X&i{ne`2MDv(?>vwAPp5$f%8LgaP
zdWp0vR~N+-f<MO@TGB*W%a52`k*jTVWSat_en}1LU+f{clZ$SCwnPszQQ~g-CF97b
zED2i-@C+tdf`>KP66Z0pe7jH@<^0E+@W5s(*pH#Y08b3A4b}Gq1q2@LlHq!hD^x_f
z&q?QcQy(XTrxhc0O{7HnO2+@rCg-1<7q%lVQfoU15QOOyF;@}M6LYl~^bp4-ZytgP
zyaNdy6vbQ}p)bG1;vB(?Z*bTSIe{nNp;&Ce&>X}U=Zq%}ol6Z>c#E3DR%6>%qGesY
zNbuWj7M1^RwA8<-<kyt^x{}`@S?zzuoupm%P&;#W;OQ2(DMs~(sy)Um7{;YQ`KSUB
zU?W?mUdNRh+WdVggoEd6fJApL#<QG<D?63lYULIsH<a|yGS>OL6L9L}Z-1!r%SucT
zE|EC8F^O@v{h~kCjg6lbb~kGaFyRyIi<-2<L}{<K-VuMY;fZ#bS>hv^=DXn*zkt90
zoBD3S0f(&qM7fPpDvN{q_q7-F*@?^uYu-6rD<GkEMF!nP5o0{T^b_of!^p1oq_&#o
zwSt>YiEr0Gjg<Uk)BGBWK@;>hM8-=E3h-PTAhd;@bQJwP{BoimnZB$IzKpyFvYR!p
zr~#b4o{~vS2J)3KA-z0T@GUGPI|jzUBH0e1n(bX1!o+&`Wt2}ye?QmEy`04PDdm6a
z<BClrW4(^pMD^o#ACaT~W5$XAv|Evj>P5Bjvat)_l62mx{ukiF2LhDDQo{Pb-Rx=>
zo81`t3|PF}d@!J+eCnmnQi%PMPKa&xSWpfm&B(C^*o`Q+S9*gTcJNPTZG;e)y}=-?
zf)GoVE6$b$b?H1*dBa}g@4(!kSG<Dqw{fvKWYTO1e54CIUQSAu4rvS_|8g=IZKl_E
zzKmQ!i~@a6%#8*^l&h^>phWr;j9_3RgAyV(zEU)1Wml}y{#A!AdCxy|??G`~2zH{{
zhlQ}SWP5e;ZZKpK7>DR_zYC+n6Mw+ffY+xmD)jo#g5mYCV3#SOWolDIJ~whB8L6a!
zB>_+008Mhh&j`|tp}-)+c>jjTdQ&W7`^OS8YkDJ)FK}T2MDE8zIKk(!JS0hO9Ci8D
z<kd$N8LiD}y1SaNAQ{DgN?(j$p;u|W+AB(A0@kQo1tv&sQ9=SufDO7HZ78FPvRr<N
zz2sPhhd0ZR6wwS0yunHK5%vS1bnp#vm=emO0JCW)p_VqBP$d&fAowTAyO1a;_kJT#
zZC6_-6!Gb|w4pUzxHH}JHFqN(6}6P*XQMJo#t!%Cm-}UFbld?yh??C9Mcw>1`m%MH
z-b)M8^tv7<{_;JPMZX2#S{-?+J*MS%)-WbD6B>r;^Ss3WIg(CBD9l~HK8cx7iwTN}
zCHWAEXz4&LZ_;=Fca+2sw?$1o;25gi?kMpW^_*tNp@lzH&Tz;__3+1Zd0x4Ka{pP$
z7nFQVWu~zCW?h=b$LVx*`3WVW#{8P<88g(@^H^vO>+c#>#&<jHVhh0ps}E4>%-JX9
zbRm2hmN4L|%EDbmoMobD<Axfx`A}O}AqRzA3iBp#u3mEAD!G!wJAa|vf3f%I7$Wnm
z`lQPKmHPV&B$eXISs6VwXB9YCZ<pCbYkk<(u^wyHhC8<19KWnpqESd>GoT&0=-_u|
zOvWInNC--HpvBa22)H6=$jXzs2pb`rP+8WGK+d(zAjEbIr}l`vOJGEiK+t>StJ)n?
zCJUn*qUJP@M$P>M9~tySI8DYsNOCdYG)x6Whg&jO>YRcXgQbZ5fRLvj93?}=eiU2-
z25}4idw7!b%q64(XMG4Q26;L3zzVE1aqww&fn!=R*>UzU9Me$>)8Mq7pM%1aw)6W;
z+gT(OV`N^6Xe-Xc{(MgNSAH`X#Qo@Bt_{({l8LE?f?o1Hup^l@iL82o;bZ*mo9p)q
zc!&K#q(vPFhQT{%LKS5Vx-v001ns&5ykn6dz+7oF?bre1;)Gvd5QJP=zTA3ByA{M;
z#E=Mv!7C$R4JcH_tN^d{8(!%L8}!6@<r-IcuN2{xk6_HSGw6?vn_d+rF}%|4e;%yH
zv%)G@KQ9S6@A@$@P)2k11I2qbV3)UwH1CcDyNkK*1R(yuXU?}k4{F4~V-t)Rb^GR1
zwBvw+k&Xr<3~p=(8Eo6zKK^DswjE^nDu6;na<+lOpR2Lor~kgdga3QV*?u;1`pddB
zc<>*T`vWD_xYz$txp>R_f2g|~cf0>bDl;@>iQ%FIWyLIe_;H?3<g|y~L0X6fh$5gf
zqP2tBxBM;P^HGG^k^rH#bKU-K@7YX3v8lGR_imA&$exH&0gaKE>efs7SK_C<ehtB6
z+a?8n3~4juC_gi;n9i&j(Z?p7Yo;I^@vrkAjtEDSQoo#347G1bLOANa+`3$Foer{c
z6XD3#hT<s78vxn0W9+(6Pj;~ob8Wgq7&g!EziHCu0yF7Io7*-+O($OZs%aVjcbOk{
z&69lcSE6Me)%wx+K+9~AzP|N)|0ao-cg0;wd=*`asTF99O?&WBegDSohwODyr4vfp
z6ok<i+D@(L+qSO_iGJla9bYSBkL|Iq2~OXkj}md0$Ow1Fn3X-70*5mCBiqk<Y~3NL
zT#nyk{2KZQn56q<k09UGXqSmyOHIqaC5QA)h!IP{X{<cF6gKc#`2)a?5Pulq209kJ
zCJ>5<ZF@c4c~@Q!0_X}+Eog=teA?eQB&fpnaHnH5(;B{F2d$*x<19$Dfniap6UdrX
z4DLu)z+pW21G)jcPXj2DXA_v?tVXLQ)$@}YMIF*1v62&oj-?SPrM9Y++YiMmWsRPG
zMqd0j8NOXa!Sey3@)${HE3*&A7{}451Sr)wXZOU&a+U&GB$+p5uD*)rBuZ#XFQF*b
z44`h7T}`V-{WN8mQkfk!(nth9f`s|lz`|O3CIOR+S0;Xyn4C5BmS0+VPqT!@G^M8R
zI*89)uOJ+epYhhzVKA-tsgx+woZ?y^QJu~A6^(>>Bx*=m)06UlRHOWNm3IaPc&#pa
zi>tl2O0DWbp_1XIa1!AZZI1}FI~rW@-8TcNm(`FOb%@tYji2LgxKAda8+^|<SP{E;
z_+L;vmz6++N+WVqd94@@pEL!8Y@+>tsP@I?ICOYKl`Jv*_ov)=PAlRt<vQm@rn;hG
z6}}k}I4QzL!-0dCQ(-=Hdh;Ip3VnT0ua?>7o9B4Ll&w56r}$|*8p;#63i{x_L3VX)
zYM?x2RNsY+Fy+Fo+Ic|c^e9J37F`OJT~;d0wH<3CYdcv;JLU+&gVccO$1bOR?Lx8c
zK!%ecFzE<?nQkqj*p&aWx#Oa4W#N39J4IBj!z{#|!A@!u61Dyll$M|{bN&}uRC*Kk
zc=ngWc`@<Qq9mYr>yForinj!f*=}LUTIjhVQ}_F?-@Sb+Uhmtrp#_>rG<&ZUY2#jJ
zso|_JIQ^n8*m)yK`BSYvH~XWK%WS6sq$3LN%Av?X2|368%|XJ?B6UIS?849PT6xLb
z064xEn){kbQaAz_`j*%Z$W8tpl9j~Vu-6|^xSwc_G^L0glL^0vw8<Ofd!@qNMZGbU
zy}ME6j!|ci<SVAE-5czpmEGPDHTMLg>qO5$-MZ%{EtrJm69f|2xn2%NsXy905R7`m
z%;jEkd%Yc&I~eQ@M%_GnBXBEwgS~v!PQFS93}=w+jNV9oav$=W#M~j2+9{M25H;Se
zSF!-@eQWzW0@v!@`*;)QjMBy!Z4?=q@1swLgZ-2r2o6xboAQF%ZXTh{qs?PCQm>^O
zt*hhiiudl}y=>)8?@fApO|`tun_J@)_h27y>hbp5QOPgTi*fN}@E!+gQ){)%Jf!`*
zgM-NDhP?v-vV*VdF7Xw4H=uvxyHVU6^bWt0=erNB9k!7^#Ji3JNBF8Eyy@uoXWmHN
z*P&<+x?aoppG8XU_L2T-mFq97Je^1-){d?nTRYAiaSVBIEI5wyXf)ev<KD4g+#Nt2
z9P^H|{)@peI3{u*O$EoJlH;wC=`>#()#s`4X&rh#&bOT4TTX2F7M<aub!q{WdFB@T
zrhRLNH_v6OFY$<;)I0dvQ-s~xAB>N>H=O=!e8U;u(EVCgdIaw*?GjsqS-8c_DNd3*
zrQft7XWxnfEwy%H?Uv47&m9hKQ3Oj?^2y*NJ-waw^0arpd3ya!6B+aR*_#k29Q2-w
zR=sv();|}APQJstlhDa|k^NEV<hzJBlkvs+`+r}_KTz_6O85~kkTmSws5kdz=pL42
zIob*a0G|61J@a>zyrBohX}-CO1kCn~{&Zl%u>11z!bClC!`ZRH`=r&PiKM<#4Ck*o
z<cI&TKB4f8=p2NWc>lMFt^cTR_)of@COCoxfVN;AtGinfqRa19ydraEKVLI#h6h63
zrUgF*p>G6nW!RBN_!^yc;g0lp%v`Vb;9b~I?1uWO36GO-rmFm5%*qU%pQe@7qpi+5
z<1lk%nw_4-LUY=$)ygx=1nI}C9-{oMIuh>LNF=l#K`}}Y$Z%}q-OlL^2V1?Lnx1jh
ztE=hL=cc0%ppU}Yr&s&qA$oUxiFl*Ut=+L9(eOI%TzooS*hX+R<cF=Nq?po;)q%LZ
zCs79JR%NJFSL&gt^6HMbUWWj+!T6cjK~&CLidK^*S|t!wTk@(9)*JXOkxa(Q$~NP#
z9_IUxMM5#B(8gN{B-wkqNY~r3pX0Z>kjSE{H^u&TQd#@pn(lr=i4l#W5^dAENQ{yn
zgoVdy*B|wL)X}1`{63ZCFqrda5Gbp}2*wwcGqH(=zzKwZK)E&@YT}%X=F{Z4Yr6ZQ
zE^jE2^1<Tc_<y3ymy{TVYkGIFXTD^Rz9>s)C-cY3O{#1{$u%X9s_f^<jh}LK>p>M<
zQ1VVC;)L9mB@QQikM`}mRQ5#@Y0+(=gab>2cyB4{?Fazmv?|&(_l<DK)tsH4UY0?o
z)%7z>*<0=*=Tg{1KtMZ20>wWHPh(O4G5rn;x}m<g=rl%Q|E@mmv+AZ%-3F5KtIi_2
zZTApTWM(qN)QkJ0@FFfWea~v=2Zddn@9p-Ig%rY_cPi{|?HD$%=~u&^xK>qVCJlQ{
z9r<=8COVUFOzvT^il-4BVWaTwM>G}5c6@gK-8^?qtNItYlgP;nKz<ZX^D0{!+0ZyY
zJ_{6+Jk4EcW>1p#8^a(5T_M}$)7PIYBRF$`+$HkXf3Wb}Rc@3t@)?B15Yr){_()nc
z)T%6VTp6LVDH-FrgNV-0i8<l^K@$hKm{`%3^;Y5JW!Iy=N8#eP*iE;0XAB@D=hE&l
zCRe)wERHK6&i9B0<W2}{D>1?5GC71~d=K7(Lm1}rm=Px5K#d1Y08c|%!96Q|u^8nI
zC3?2Y|5l<eMNhN>VDhwVs4W;KMyVDL8S0B@XiyEA=vrZNbgeh{`F~__*!r*J8b8w<
zfGH*>@?5_}9JJ1XgtRgQ)0=~u=ddNkYxcTcmqUx4Cq09&kV&Fpi%o!0ytsOVqwN(n
zE%FQim|OHT!|Yf*yGZs2Ru9Cnsa>eF&iV{9Ht#I4lUk7dB&+3TYIZzBd3mY)xTzB#
zf4ba=0+jm3LUT%+4^gqesh9r0)Ps`T;1}E#S3_^3;4q3T*ah0Wx{PDO_3<74dsN%N
zo`KHK@?1yMwh!v^LrM&iendHgb8+N;|1mZ2xRNWnJFmp#-a;EbfCCWF>IFE_m~Z>G
zCwQh(bYoJ*n$5qcg1j7?bhVd9H;HE~`uv}0vz?~dR(OBOk)<ITG?}y<nY8J&q~XWY
z+G<7gOj|9ULGwgR+<10C?8>5k3)9%ro+IP@W!>VIB&Kn;$r8tb&D>aEGKdj!lK_YE
zfF;;ot_drGRhSkpjw~!WjBmIi)`tvphSgs8O{I3dn`4q7ySZ-?6geu-LmU#hSk$DO
z&7}YTFhDK3<H4AMLXK$4M?-kGsJba73>_M>I-rrfp&Mr7qEM07&UuD3rHWpGSg=_!
zs4M{2A@}Iw*%1c9MdC4VwU3cLyp0Ueq0n&-a&j@&FJAQDPlqpDh!0U{a_KKCxu)a=
z61|&0IzmB9)_rO2!i9_cH=g2I{gQUFLiPg1SQGK*!Ub#E|1BP+sP_`pqt``cTw;iS
z;o_x>c1zE>Ga4vSUAa>A`VXobqTeq0|5;gU@6x3Umo8cRFbF5z>VHai-%Aqat3<Y*
z!ZrO_b$3?f63=v4$Tn2gRN3>&O)8mCVkE%4a-UFQ@s++yx$jf*86}657!C1x<*q0>
ztOQm);jH`qcjbOb$^WTjOvy7!Wa#7$LpDMpp)w;6jQ%h(L&(ZDk*|66FX=L#aA1yY
zcawL$k6SYIZ+wDQk#^;1Zo_}QXA9X>a%3Qp?9CT?IEP_Ju~g_LmoM}d`ir;csNH`X
zD6P9#?9UXkh2G)-Co%LEhKt{n=;hu%)gTJc9`~1=J!55OqI-S)y|geI*Z9W5K3_eX
zqEqodVIci8pwK$Sp=~C@X%6{XriL%s>%`>8E+|{c+Gzm_yHxN!o{XioX)e^9{zeBI
z6K^?rlTNWyQ<7~HZH!QwGRd>x^;u91yWMtq8boquGyfu{6f~W=2<KyPR%-KV%;j8=
zg!CXUu;BHkhV&0}xmt7<4dQ|}F-3Iq{d&Nb+=uvnBIs^vxhe+Ag^wlRR*QIZVPiFg
z29lAiYol}01T`$$yTy<P3L44WS^%~pY!&E*2ekXO#C5{0LJ*)o*cBtNXMH>zYknLa
zTz<>@scY<89H*&^5}I+)w9W@GUNRX&bBDkMsHZ4yYl!^nPuU$sC3u%24+TktIO&~H
zTmp0rs!fQ+sJ28th%2cbkjpft1(Ov8i2DUo(90La+DC0I$KPE2_v{%Rjr-Z{-q6{C
zJ<HqTXC>Z5&}k-7pnX%(|BGf8(qWsvZr>75$#W!EpSHHTtgZW9TZj+$xz^fWVSSJ}
zDd#>#PfB2d*vOvr_*zbBv}1Z(-zkA=4m;+Y*)eOGqbRo(XH<c+k+jvvjDJp(C0%&u
z`ePEU`~WvMtwmd_CL!Tr-3O=J0y`>*V(90SwB{etrO^SOQKha96!Wk3%6O3&kMI94
zdPdAZn3~$wWBIt2;oUr`Whg3FVT-jGdr{Abx|LhgAU_O4qMsRwke9t<4m-y!i#Bp<
zCRQg1A-@UIM*v=ZHQ@U;z)~=zto{jhp#IJ~sKQ13Pv(>QXR)PQKuYk$OTX@H+R&%G
zR0LRYd?E*&WNKNtz<YxR$htUy2GhezJ}MpZ`n`ddbgT-VSD8dbI_Tjr((1i$$)p<b
z*AlVqX|^TjF)h9jNiY<&q|X_n&eyIH!HEK+>IL|;QT11haYMQY1>j*5vEc0l5@>rL
z%S72NqJA=&v;jd>R04$t{J^gH4w&^}&8`+hxT7<wv~|d0?Vhm3LlUWH!v1z=9i8UK
zRoaIoY!9#=(@Z#kg}tk-kyb>c%|jD|t(fj>LH|ee@mg+Z%31-L!2^6?I}W(aG+S}N
z6_6Q7rI_CmTrwXi2pAX8+d>@c5X#o@JFq-X>ziU8xBR%*G+s5`|5YUhkS#H&e3Ef`
zru~s(AElpMjlyJX`N(aD{6Y+zB?S?HHl+72$OKuykehR9keGS2NExKEInED19N*#N
zrDL-DY11-sUkYDb{6zL5sYV)|F$cbG5T;a*IAa9eSjEEi0gjV8M@Wb6aq>3-ka&Xv
zGND&8Rq1OeD|2QkUjhl_`C-rWIxZJjQ~d(9x^R<IaHl?WL^)jWP$ZirY#%o7YzS8>
z_lofXgj~uAd4G_R-CpXuU1S;-*nYvJ%%vyKKMD5S2(%y|4J{kd_B;dU)~LqONr?;^
z9?)DGfv$5W{acg>Bs#IwKT&`B9dJCTn!lo3KdGF$>Q1%QdU;FfNtj(hUHw9sT3rdd
zJN9(gg@vH0LmQHfX!Rk?-nwPf`-)c6KHe-{_ZU`p3UnjEgYi%VBFn%*q0@x@RM4g^
zD@>|Q^HuMlsNei&H1{R7Y3D4p?Cvxs+MM0|8YkFpK~IN16CVfDXZA~<iFQ-YnOaMj
z7YJCg3JbCuZ|gJBf1&koGXGek*}8Tb%^`FS(r6CzR~k*(HRt%dlfUrv>)7SF6TZ-H
zj*1-<V}?2$i#Q#NK}~d)*wUeAMQ`{bs&8ZmUN<_>Usi6tU!Hx~&h4PR5<1ZDp_G%A
zOoKUsOhLVozEXetVlC)ERKgvvA1%-TNW0y}C2M2Q+W5fsZFoa(+rl8SiyfxvoNM*0
zIUJR|$d2P(FscLIFyAMA-UpgpjQFl#FxGUAq3Il<ALvU>)44m4Bk5?Y>HI~z-s$a#
zG@U=x+|?Z81gPDn>EzTXZztcxHJVLaNzrufW4G_7&i(+sCz?%8caokn*yD{-6Rqd^
z0eXh+@@7lZiLn9o51D3kSMzYN%i9GE*+*`lH)gpb!M<RZn}2UNZ1z5`(K7N?vQz&_
z?EUwmZvun!wxgJdrLZ8BrgNWZI`^-YJM~PwJJM*<#(vr;SR0A;V?mi79}J{LEK}a?
z9WedluhQS+v{{1SJY;M3@XF_OhVpYi@hwUG4h}-3rTvj6+O?Nu*4`yOsGn<|aMyU`
z>`;wHv3U#l^bj!a@XJZo&=SucvS-i7&yLNVG%f0#s_UMQYCjy;aqX{0rf!G5BQF)u
zKgVy=Ov1Lfca%EYcrA7HSFy3|^NuoFPDk1?T3BKp_Kul8@yOayo1-I)$1!x91Kx2)
zV|-g3=~3E~p7N-T{wXukJIy?t(a*Gfx0tr?_}X~qEX+wm%9(b|=8510Z#d0}E7aW?
zgX4mW{#k)hGEPtD|CxYT1}NvRjwltVt3KEP4I6)&JY^IPkHF0V==eRu&%#uL<V`Oq
zsX$Wax<~-Goj4@RqaOiSZZMwP0uI}v(hfyN<Otxv{-4thC}|tbq5l`UwN-QI{};-A
zb=zblf3x~QKgoNxR}DJF9xEhas=-(aWelIu^6@Zhh(%8gX_!YzTFwU1!ec`?XM;9@
z!<altMKi%6Z^{D0h;8aeJ_kP~e5K}5Bi->9IBEMyi2V)z>Pkh^iuhU0NRGs`f`4^&
zoS&zy@v~ort}D#UE#g4=yPAqw4cZe*E-Eq7(YUg=(t@>hUrrSC1wHnnlD}@9mkFLs
z@pW0viwp37ScyrgCY75Y8Na222)dwe+Nj_1i;Rm?ZIzVkQPQhK$|=83Nxza<^W+cc
z@^9+}f)s8)x0Oz}>h2v%Zd3BGsthV;?CEdlu_0X!E3x&yL%9(p_fxr&Tb`;m7nb~;
zDj8MsvwCQka?&CBW6JGTqSd=$IU5}>Y21=l^7nC@ps%KWDdF+ZQW_Nq&1DYO&|;-X
zadeWr=>4xnWkf5S7Odxw(jFGtP<s7D;FgQG6iKnth^w|rsCA6e+kiQ7&=-(HKd3=;
z^3*%^-FJ~(lr+^cmoV8hT|z8Pbv(BCnGX4>k<%urd>fWqUWo6}*ok|=4zHjyUi)`A
zNujU1QySFPKk>5q*LV*}oBf8}9zZgfeLZPxwpi~To%4Dl;T7;P6x%JPo48YA#l1eS
z-<a;AG2Pu@q@gg1ly+ejoyBs!7IWRJuNKR#e-BSIW*cRs4FtNGo+g+~y;+`i4@?k#
zxv@NT*&;~ACIDqstt~Y#pEG@UyQNEuG||w>7Oe)PV-kNrGlam3TCG|9R0YGiNE9f0
z{4=T5Q^t3mRv(;h*IgUoUY(p@oVwCDeU|@ROogbrbh?`#>3rhs*(d~qrd`;xb(ka~
zg+|O{isBb4Fqy&%8G_=Y^|RAZvC1Kj=I@5iJ?#<|1=Wv%y-6D+UX{asq8iB;kV7P@
zc+`;MIRhnu>8k%cIZor{&=5SM&A`>J{U)>uK?4X<oJu;joFfIljraQxk*tmr%gaUW
zY7JaDimx<-YaeIEUan1DSpv#7S==Ae;x<4V3uD8c)^j%C{ySKAVV90q;CzFH2FFv1
z0mJQB2E|UjjU#+#nW98UHss)<Pp)OxthW67Nh*WZVEiVv>_4cx4=LHBZW+*vziTU#
zCqZtg?IJXs=$58fF=Si+6>*@#6(eHrE^J=Qv6$q-lqOOC)aL@fNBA{fqzgpCLH=k=
z8tPUrjb(Gjsc|s|J&FiHhHKayw@TNtE58tAA(Ve+Ew_T19JzzCB}b%$Z8K-c5C#&<
z<J?W8FOZL@Vc(Is`Y`MOGT=F;8Uw$aa<qOoQI`uV4>HwkQ`)&5L4t%37Cw^uQWg^v
z0yVRqDDpS|hEwlqaS(53t#I{Qtz|8tUv6T+YGMc835!L<2}L1-E#g`Jc9}^Y0V4yL
z$g6id$E1dsLaeC0BCfhAV$$Ps)xM@b#Gs%Tr2v+aLwEav?i<SWx;R7R`raVG9SUc=
z<MqQ_Byo$*VldhrpUr47iK2Vo$UWjpwC<V2Y3QPn0L=bRF)RL0ldP7{6ZlP5MZ{*m
zUjBwPx4!WNBm%R%R8Oxeya_QPT2+p%?mlLRjA+rRkhlzhquCi^O8<YoolR^UXBo%s
z-L<{;u6NfTaT3RI)=3(BBVtIDR;aC#f=X#fVyZNSVi3aiIO}!%#q2r_D+dl0oGR2q
zg@hEL^hP+LiZdX-IrKo)L$n~lfeSrR4xA9-_kZ4**<J72agvs`c0BXW%saC)@B99r
zkN^8TD|r%rO)V_Vtr7LTTwo}Q!gj5%k=?Q2l|0L8L|gUpt8<lwCDmXq-bI|x8j#%8
zrt#?cztRn6u8^3O=oFs2vud{^c~$#Kg2Dt_PA3>vzM3P&7=f{fCBcyzT1@z_W-bQ@
zU8Lyge0AWJ`cbY3<BCiU{;##L-uoM^{Z@-vEq<qkbwPDDms9RMYqdVSuSCdRDG{&U
zq!!u6oM<IUhU{i1RPZM)G@AII)#5!Z>=GuozRzZjSS_yw0g8XF-DcGLv)2Big`@+w
zFtI>`z=!IqXIa6$21IL0l_1(2F!5oaTILztZ?<;~uE!{A$NB0n%y8D)k4tN0{HW)@
z6<lBSh17LBjqp^TlYBL5U*ubUHO`2=Bk%my5Y2tNou(^foWq{pr_QfHo~&tN0c&t%
z4>F^^Yik7EU3)?{dbsUuHzP1wovmg_{Bo{SXQR&tZ1g0Sn4hsnY;<fITu*%MUbN^R
zP@2Y3((V5nYoKbzRbv0)pf|A5gFRk9c9;y7h<XjO!=$~e*<r**Z&Qbt3+klqHZ=@l
zFM?}Lcm6_nUvzR;GS85e6N{Vs9k5?FfR#z{HmY0~ON`dbw>fGE%ZCj^l*=kM_+0{~
zD*i5K4sxcr>V9Py3(hE<=a8#@2n&wpbTU()VetkQ7uj6)W5K~?M$Lb>?GcaQW!m<C
zSvgQS$UOfsv)~M4!P&>t;4G{<<nYFVa~KQG7{`oTFfj!ZJ0klHyz>!`9Ea7Ox4VTe
z9&pu}xykVpX7L%V9NilA_G7m>!rBoy=vq6rg+0i%0{G{Yusvivb6Ifiq5mJkhA?F|
z1S~+<NcI&wy|LTGln7Yshs~CgzctZlMc|o*Hk%=DoM$8}*FEm;iLD7*$+4|tTtzmX
zeg+h=;(6D1PEe<~<XlG#aM@$rwXk{&coW<QstLck?R~%`Kj9rUD+%mxdk&LasjyYx
zO2@d)@fIt|5HjQwb*siZ!g4b*v*LWH#zQvy#^YNK$PN(79xj5m_kXh`G<by{&aUtg
zuz%Yr>V?TYXxNXSN7v;t2}KTY9!Ix+U9~X!wopl>Ru@k0S!e#6>PZRU9#MNXQ;eC!
zA{AO(ZDiX=P@R*g&PIqZC7L>Bdni#fxb2GcHf)qynCk(NzRe>E*_Pu(tPWM?P+jh8
zqb?ulJh}&#CC-i}o~2Z8BIkg&+EZ7O!#MD*C|kEY;c0qS&)jsQZG<CJke_Goj3XSm
zRp}Xe+jSNBxGMO)UY}zeX|F_t295ba%JMT*@xCOA5zb2{5(%56(m9;X9dQbCN2jl2
z7T6tqx`H*E@!g~{r%hRwzq&W_Q}<gLj+90}wUr?e6KS*2Q=JJt)fb$?Ti2Q2HC0tg
z*&;as5E%W<HEG5hu=I#I8<O4w*5s6gP7y3=Kvq<c-mGi_$c{IW6WJBb(<r+ndyCV>
zX(ZF~byXXjDgm=X?9}{?(p=t}<tedgT)@~z*SPa{V@4>hrFA``ahophzo-YT9<7fa
zver#6{sb$bca``-aiY;DA{`h#SkZ-{7){Cd9(vgmbuW7@*LWQ%q$aj9+?t$f`s4*Q
zrbuIKuP2t&zi2Jo-7p$r=sVy3vLdB$rrDJ&tx!LyoTJTh(^%<9c#IY&mW-+P=Fo%r
zz@PNOX4kPJ1*3;OSJ4?gh%~92Z&FV;BFy*DqR2tAh4?0C@-ktY6b=JNavH%z`j_U|
zejlg}9nby5@RJ@$%s%#aFB7ki*s{3o$UH>wjOCnyF#+5eco!BTN;x2j0+=)SR^(OA
zv~yFq%vdKF#S9`%-|`4<5869N<I+Y@z(MWx-Hs71qzh3hiFTTGF6nDz-_(|ogi|xs
zrz%RHY&IM<P?uV{9A$$oyXa?9h-l6AM!&;krCr?C|M>6y`}cXrFY`7Wu`&pc)(FP?
z9oyFwoNeB0Q;MwFnBJt9+SacvRl!X+vF7!<HjQbTGRpDB`U={n%N%rF<CqbMMH1S@
z@$KeinccPx^7x;l9m5p5?oYBZo0wl%dd)w}YHN!&Yoc1TbSvfGQFK6;KWEJJDwp?Q
z3@5kKo@Dq8Eg>y!K5}$NP7WDFq8|J-8x?|A;Z2IEw{9sBV4QTgbO4$x61!48lX{f6
zeUO4J2`*B03@KDR1*X9BRIqaMOLlp$Asz4q4zu|OPC&@(*_y69p|yWeQ^+TYdb|t3
z9uPeZTTSH@fV`oHq|Vzxh&WlHL%VOwLukOanE!d*gl^c`fLkoQ2499;c57F3YkRWr
z>etQdp3>Y>vMUi6(-a%;L0@|obfe*CXQ)aU1SzOT9jY9B;rVZYD%pFRp{l<Y&Vw*(
z?1k02VuOkIw1cY<Wfby%b$xApVUq~w&VHN6xWTZm)B|?xU2t%f>1vwq)dT`+Mg{_6
zwg(^}Mp2Ad3JJwhusGX{qL@83O_HX>8si<!CO%_^M{7Hb_K|hc(LqyYNOo+`TGnk~
z{~-k{2oZq8EV;k2Hp@auIir#&q1N-Gmebm^^iqEqVFk7bvQZriCcU?LTJ7*+fGQ%q
zkkjY!@j^&}TO9?re8mtXX$T?%fe_>;#DX>ah|uDVHfSN6AVee#3WghQ@(8J9gdjOc
z#11KN7Z@Y6pZx>Wf<)x?7U|9eNQzSit~8fn^dLYe!I8iln*%`pPhy6r_Y+~!>xci}
zzm-C5btv-<P$s=hW;p7|G@(oul!3$m%A5mNa!n|cGnC2ARG)kplnH$e#T`J#(8Ff8
z)_Uek3!8jjfDUEAGRZrv_0O%8ifgnSWdSbRqy_w4L%h*iPmpc6$Z`5SSNzZ258t~y
z|I+!OguGvEaB1IOaA}G9oQ`FKp8?SqH6Yno69JKbQDvXx;96pi1iM<2S*(sj0}XnV
zne8G1)*f)kzogm)M77?auW^ZRmFw&bOg~chF*9ETOq#_a%4(P<E-9}Z-(>tK|Hm3Z
zDvaUZuCuiQ6_+w<r;RcjY=okw11VZNfMwZgjk-q<NK_I;>WBQNs4cW>$n&)8#4^*L
zn#*X{q_OgPu!&oMSz0CkY12u|yo)K$;H@u!H<=XcQ>DYS)c(sn!FnU%7(V|zU+c1G
z8q!Hk)v;InlAgM~=c`(Kg#`zhs2dGh&inhS_J2u>FYBPT==Te~2_a1f!oe_qqde;<
zrd+!x+<=0e{){R%C+J59fS0NBE!~#Pu*rBLmgxfi^1jO3!{1d7Y|_-6Ru*V#0katQ
z=A9ngnHht|kXf8`rB;^m(guUQ=H~1JVY0RHCYD8u&RhMO!CRhBq;v@mVI7_#P-fm#
zpZt_GEzONv$CrTW#sTH^Rnlr~fRK0>jIWcQ+bk+mjSK%6pr*}+n~CBRsVzr{d4-Lj
zEfXbts}scRc8G~F55BHEH(+Mmbc2}+ZPz?4v~0ADbCqr*GWRIbj`^8zD>kBPb^HQG
zbMCHGHD`qVBqLZmEfi*Z1h~p7Szg+_c5P*maEW+7T<%)qeVVVK?&&k1w<yhFn>Dok
zDh2*2EsX6Pwd2+{EAV|+jl=BqOa^r^6`iA?I=TDXilF%8r#<`j;|tvFBU7F4g?*rd
zK{22@pMAg&+ymt~k~=PajB<za!QLcPipd^tg71)+@q=b?U1XF=lQm66?Zcn>V}f+E
zRv}<y`7M)Nbc}f{Rc@gi6?I7(?jP6_YPgxL%pH-YGmO}-wk)-MfQ-RpR%W?;z{rlA
z@wVvI8Kj>SL2lSWkM)o+1_}5m2_q>9gL@(C33A2JP@^T3B~xB+W$|dX>bTGEtmA(A
zVWf{uo3KFdI(4nGn*h>IC8{2u&oAVsHcsc43j8}m7Hbz-N=0lg6!J@y#cHvQ(|pf_
zqDBjsyHO_|emL1f>Ys02*Jw^}YqKtoSO>7~Z1RR7*5fO`j<gg>9=pgGe^vK+ihHUn
zzf6b0kSQrcE=4z2*2x>n52-)6-=-v^3Qs+ttF=lRO)Bve7jBj`-~cqK!wm{B<P17v
zhmH2_-ZnC)4Yno)=&?VHg4;=tcXlm)8f5mu1AbToz$g4!24&3>GK^2Pwb7cAYva1T
z<GpLTU!m?l>Me;Q5709!GSg>#rU_sB3Y}LApYctRR?W2}{_=&T>9sXCsn-~PH?(75
z<6E{1pLKDmq9FG)`IXrN?s%J&J`FcEoQ>_f+1cHTr8V-qJ0pVLolBsLL3E`w?@PD#
zWerremNim2tgn=PZWf!g78*ro2kZhjN(vY;y;&)7M6b@6Y`GuX3!IXg3?37C-sedL
z?b!@kQd+Jq7U!gV`YD~AT=7=zcwcgvbJ6PwS;P*<UzGmn4{9;QVm7NYAQ)_}P(QpS
zf0)wRWMgS`Rc^K69<=3h)s3j4k>C_q4%2g+mGvNrp9|O_3QcXIxzd196<%A<z*h($
ze^iS@s`Fq&oe>3V+5%&YX}?SFGL7XL@bV9<$Yq|H{$>!^!uQ8j+fV5tz9H(T*sNtQ
zg)jTE{c``b4$>@Hn}udvIJ54X-Nd9vYgG-jgUAq6OFrv<!a2R+^4D1;nDZU#BXY86
yU6btii%qdqnI_z&W9hhpUL@mv-O0r8cRJkPGr4fh{Xai=l9CkPZ0=0%$^QX-_7o8S

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_22_15.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_22_15.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..250dc07ebd628ed2b45f4a294c85ae4f20205f00
GIT binary patch
literal 58025
zcmeFad3dD9ec##lh5Hx`=I*%%<lMYziWJ4+P!vg!CWoTcBHaS;HZZ^d4b0n(IiMFS
zT0==xWZH7<9JVAIkk9ctvAvGs#77c4m-bm_edWn!-z2*^R#uoePwdE7#u8^g-(S6b
z0T{|T_St{ZKv&mWZ&khZ*73XRSBpJ8g&3b(AN@pS{^Rl3ztKheC(Xrs`PP0d9*cRg
zQp}5ciTQXbZoi3zWGNYUX{l1$(lVtizsdRBLcWx@d#U-tLRYD4p}W*=*XjA5h2B!{
zLSL!Rt~2xf3j?KrsGh;nAmy|3LkmNtAxq264=-#hZL@fOeq>>LX}iS>^P>wpN;@px
zHNSIVS83P6?$YjsJ*7PhdrNy4_LcT66ida0{iXd22TBJlU-$gMg+rx77Vnuqyl|v+
zWZ`J(s9pEYA6pnJjV;_&y34Nn=I>rOUOK*TqI6>6Wa*@(_0OMLI9)nz@qzg>3ujAb
zEj~DZ&%(W>dx;Nu!`?P;<nzhWeco|z)Z6h!ymY^})7$k%tn|QBF>klG=Z%=Rr;?p}
zuyXcBs`L<Pd%b<6?W;U=BUyTw>!P=x>;1|@%D0emz&l9F!OFv2Kf?7P?=aVg?fN}j
zAMuWIeYEl(?mx=)F>j3PF}r?@>$|+WxxU-3-|Hn_ijAF^SYk*ES7+zF;==5;`m$ds
zE>0IN*6T~PhfkfFnXO-3o;*3VxNvHId1}`4CuXa)Q(mRII9r>o&YbXOr>B=|vy0Ue
zOV{g*{?x@&lk<y{ryh9V^!=0G^wj;+_nm!U`o0Ifd+wQj@ZQM>Cht3YwsQZO%Kepl
z&fI_R<UOsL%FUX}jhdQI*?jNh()F#HxVTiQPRyQ~U0CuLuT;EN>s+fU58Zq3gAbj3
z@Se%Z2Oqe9V(QGj4?N)A=XtdDp~=Zf@2vOGL#NM9RhkvF9(-zMV!2kEov5~w?$CxO
z78jO2I9oeapIA@_)%}TTZF<pPsQ9(Qi4#SBr)CRpYjKT_#lzgQ6AP2x#IrAsCBnR4
znefRzQwcMEWvMbzC$V>Gv0ATOt1nDcCuS;s*i~JgpO4ZgpIw^p>$4N{L{s(4-U7e*
zg^5d*@{B+0anVbCm&z*@f3dt+oxe^>;n~?*{fWiu^z6)7F6^COoGI6@)yosL^6bJ)
zn5kUjb}h_$mFbD)`Ffb1tu0I}g;|Dsfd)`}g-)2BpQwl3Q;W;hdYKkXEY!m7OpN+O
zmB!R7<!OI$p-e0Fka(DRcDDZH!h~M2N>5D8moLqSyU)!|%q&(X=AVjs_8brPXD65I
z%qFr9KF#>oXY1F`&-%6cOLaZG=kewGVx>B@=vDk@o*heugZ8eImHI@vHg&P$EzeiN
zA$rT7n5vg&tM$cly;7|$`e8v6wo;y-y;Pw=UFXg{{oG3v3rq7ApUeKI^=i*9&d|Zn
z&0d?WlGaySnwYAT>$3}$TD`JV3v*L?p()<Gi<rMyt4-AF)i8dYW@Mr}`Br*z!dtj7
zmC!gxAH`W}u_yS}QUpPa^%q~6eEv$szv|D{E5!v~Z2lyzpRE?Jvhs@ci?g-j^m27d
zol>kX7WKZxnii?$I#szcF>l2RMLynQrB<vi){9dYHI0fB)$0q3e&yuN7#*6bRpzI|
zl%C-q=J6KXT0C`eaiMaGLY2wI#Y?BWiTcE;8Gmtk>C{Wh)n}Q)^lSB$y8P5R7W@02
ze(}`2w{Xgu3M6Q(|5Ki~v{1fWvra8f&sHn-<!Z&N*jU|rU-|6W@|pWsGGVVbTbr1i
zuXts9gQMyVbxkVikMT35o@sm(7wWMt+f5{h1@X~XJvJ8)5?(w=&c?qhK1|wnb!B2a
z%y^Y6vr`q`Hf_)-Ofmnxu~eAw&+t7PrfU}`mMZ>T)Ya+){}`9yz;bm~3zdaZcGa99
z+1j={?w=+;#kb?LH@d3*Wd^nq10BXnw989)NiQ`M^U_|%%X&F4?-jf*V%=Vk*XyOc
zKK}ao9pIbhuaB?AruTj-HXZi{y`eYqrDQ&y@AX~~A^~}%qN@>aJ6GvS#`{)>R&K;g
z*-CCLzaFa;Zt%?MgrS)@>0J)Zl)4>iDfKv%QtEYxq}1mSNU7f;j?#cb7@*#}DnrzB
zw|D%FRB71N?!5>=J^6;}-4><1<el<PTgph3aseQEkELv<efN6z(biG#e(wR$-VX0&
z??LY&Qg(9hVeear?eZS+-ow4!-lN`Q{O<AI>pjlzUhjS06a4Or>OJQ@`9_Q<7rpb|
zQ>5?r-tRrl?}6z42fb&#=j^!$y>ah(?i~75EEOvqW?sA!rp}#v;kmc}ZIV&W%riA7
z{4g<pcKrNGj^%S|p0#`dKqnyZwwCDIWQ@J-5q$N;6TAQYzrKIxV_*H~@LzrEU#)!K
zV_*IH&ENm6=O$l&?5mBL5Ji6Pu`mgom?}2X`PID2@l}1bfLUIT-H5Hl*Lg^=-HTs|
z)p|Y<^Ab1WACCvygIGCHPX@8M)Li;=F;B4szo{>zZp0xdnET~S&`X#NdI)ntfiNF*
z5f*}O!meNkVRyYJhzBD#62yD!ebmycdH7a>XXZ)kC#Byn)dzIvMx2zvdeY0*hw8(m
z46bckr=M@cd2;RxNj*6j^<vA)H^Ti9zY>fR-yV>=JQ@rU?g+LK?hJCu=Pw62u66~x
z33ms3{nzVzg7l4e?e!1Dc+0fq`2scV4bnj-$a?v=ARWo4wEo)nDi`f3=*jJE4EFhd
zh4e0Kk#hco)u$Hkv;K_D6@$Sb86<+fpg$P!y5GvsYdx#Q`u_R>Qj0V38}#(*{`$eW
zL-oVF?S8!{S4X_wx71rly}s20GqKf!d=J%+(Vnpz@pG})Mqf>=9$q~X91iwhiTVF2
zIJkC~*B=}S4(y7J#@5wCM^+C6ht}>24(sZ2$hsbUBNiMyM^COEq@KI$2iFb-2iN2E
zq2S<+xHn{@k>Z)hgPnvY>L=$;1-r<3n$<u2R$}!i@e`ko`v>c1=FSqohcVn{t-Pn2
z31VyadLzNnb&cts&&HQd*6+Ka`Efrv@2AuwywwA153VciArIK7@7VV7v84YfQ!z|j
zI~~Tatel=&_I;4-{Pm)U1Mtey(jw$Ub*6ZLseAzZ;S~>DI~_%?99ZdjxcK67wW=JT
zsFnT`OMb<hb=XN5;R2X53p$%ztW{R}pQ%)ql=aEy9<M=E7Q^^SAamF33{*g+e5rE1
z78dm2vLTaO<rgcpO5Gk`S(p&ETN0wPo5ib@*_n&=+R5Stp+x?dYnAEc`QlZuQ_%-U
zFZz0v2oIWFq=FOnPGGO8OG}H;0w-^sE8gAu6kmD##qp=dpW=<4eCf#x7l;<eUw*Dk
zV~a1paPIL7Prg*V@O-iK<crUD6dTKgT{E*+D%G;cj#}8$Nqyv76hpTIHD$;mzaAE%
zt4h^_mRN*(u9RzU3&q@whiR{V9klcwuHF`nFr}dJG3?02OMGiP31%4A*Yd0JAU<c%
zWG%5C`<}#Q$a%wQ5F{y~@h~~Hv>c|Z%L_}_>G#CR(_u=D-H7-KkIX}QLE<kx`Wb5f
znx3~awz6M9^JEz^YPMc3pL}F~acW||_9&4??y-db6nXup33xzZdUk%k3`AY7Ky1ve
zh|=ouE0>pN86ODJ2@mMIICEvfAA1-?yvz%SxkmkAztNGEdc~hznGm@Xrk5DfdYCVl
zr{*VWwQ^axs}l<qV*OBDo_cns;$~0SGc`ZE1mZSA&kMUm*jAW|*UR;=Z(1~Dl%07O
z77W;z)kp>dVyRl0uZ5$`sqzv-qsK#~8l}mss)hX`eT>g}#h+MG!_zh`{Xt$m%&HtS
z&L7rQ-tOo+OwL!TVM_fKW+#`Yr{Mv@9%XJ2tA5zuj7Icw*mXs1P(MJo4l}K1X00*v
zlZ$1y3c|j5nl9?Kk&s<jWR%F?Wdtz;Ha%PSd(@m>8$U;oUs|e%0}Mu`US?L5mzS6(
zm70=fXk*21wku4M#}|1MCaOzeT71d%FasrBuXv!!YW<$GVPA8qi`1r1@)c^W_|zHp
zPRuV|gr}Hd^7)Glyn2n9zkSoA8fArr%UT6t_dE!?T6R;#zgvw+vDD|o9yPl8P+C?E
z^Dl_|(5&%4tc-ol`dxps5&%zO_wp2Mj_7KJ6YdJYwzv>>y=cGi7PL60EK7cMX6%@M
zgsMti(69AqP;83(?<d)RLyyS=Q^fhyFZxd?FKpC=XB<{prTk^(yriI_U|PY90udGd
zl!6Z^ct*jq3Tz?O6!R6-3EU*N$w;dOFa2-Vy~_$ND)<CJNi&@_W$V<xqU7%+5I3oD
z)->GK8gech!|s1frENMk`?NI3N`o2H;5Ejsl&@B<I$&4oZsaVBDU8pSdOEV&7~G8e
zA5kle#}EbNzeaEm-`eX0h5h~UOnfKb_P_3UuU&V?_r?neyJKO_!v1(J-XBlJ3;cHT
zn<K`b?%hr$`>8Eue}mi^w0w!zd*h?=Od^v|+Hkx((VMV3Zm;xoz_qC?ZEAdaiAWHp
zF|GZQ(Q5%Tf}aiItJ!M44%Y#NkrL(N#otP=X1zp^9gVH#>WR4|^bb@&xI14@tfjb<
z_n^i<k=PMiPplS#>|FZQ_-eAA381RL%US9X6_+N=2l=_Qm)rqdpUYlK)ZPnKm=RT2
zFVwr{y7~5aS!#yXTkG|5(0th&al`()77dthyDo4I{nWhQ6?D;}z97HOyEId~U8>qC
zDyHpDkKO6t;!ba?%&yKdeeHL={#K1pO0l_t(O8fjhSCiRqcQvCnY;A114`c>TZMk*
zO$KOrcUwyaT}x>FV9@OittUF#IjnxLH!G~e1-ka`g63+z<F=s7J=1dKw#qd(<c(ZP
zKr=y2&kcLqhhiR-_}n(Q%B~=HL-pznM%3<IvDKdHLtG7cqgLjCw?j48yVka=cdh=N
zjVrw&`S)A?U6%hb?m<0&^XII`DVeJ8@OB$-(Bt(pH+sC}BkHqWZx3;Vg`%9P5ejl{
z*Lr+y_j-)?cvW<<(R;=5o5_t}MihoG;sT;0HG!<>?c02A-R;12@#4f4Bnykh$x0Cc
z5G0CMge5LkCZ>=@M4}LXoZ?~Hs2~436d&pY_CV3fm2jwb*^eMe3k~eC#6?ISgk#G!
z2WVQicMECX@$L>Q>{q71^PNu&ceQRUoY`>Ge?e^=Y~oZX6dys}R3+ToO4EZI>5v<<
zU@Pyd2jBPSyVpPe55FZGZe=v_*VJNV`plN8XN|yqTO^eK7{OQh^7)rdcJaObd%3>V
zzp!kAnuz2Esdmp_nV*dCdF?#IQ-I8<2`yLaV|xsj`#UJ=f0Q81Mu^q_xROpODG3tx
zKc&kSY7uaA7&H?J!m`t#DzLZz=Ts)8euGR|oH4W$W*3+f3(E@*(R@y&zo@{5Idj3u
z((;T@%|vP8(D-8gX%Qj|w6fwo>HE+ihzzR}RsVn11CkT-^EZ19p*j?2I8+2<TPqC#
z>S2%xMmE^<{Ld>TCmEgu%dR1+bOZhBUnfnazcxuAgi%N&Kn%G=FMvOnC>UZHjrS+`
z1C5L%cY-i>f@p@5nRw!ME}pvGpWF+27`9TNjoaPHoKV#5?!+MwPB-z~?Od{((mUfT
zBZfelghqpa`yZI(e_)dT{+OgbvW~>_|94E{|3xjlCid|Ek}khr!4D`fl+a};A)@Dp
zJ0Y}7E$!HlDooM1WgvXRed8-i2kq9K@IqAnz`Nm&s8Rb`Xkr7Bh%)SNW!QvATo#8-
zMq3D>i8UM#bf|S_E1VKl@*{LwxU-eDg;k=4Y<2H!R9GS$|Hl;w!}vd;;HHB2D)>o)
zaIhVjxMJu1pW@CYOyd7l6>4G;|EIYu4RtVDt~{)PNlu+OBTDOU2{od|H8GC=GphJ!
z6$sV%|BHg3Q(z3vUsKH3nZKpjuPXT43cjS^uPgX@1%E@qKTz;D75st%<1AY2#SPMa
zTj{?{u(Eq==<J%eNg-39h}f~3p7<T2u44W^!uT2z!MWsGYTZfrlmewbmo#}F^wo0Q
z-v=!(IcfY0$Q9!VOuX0!k?g+WYI%><lI^I)m3u5ImvH48@;UvIJLZ4rjwMT-C9lTx
zH<lb<+3|4kB@{+dfr)&R3@Z|tt?WX;W#Td2ia6lL&9R(gQs;SjBr@RHia)1#p(%_(
z6*#rHgb-!XFY8VgUQ+px{fJA}UDHo)7!+gyaGEuWR=OZJYvKbFuZtV&g&*sUCkP#H
z2EX==R3h_^__Y3b;v0#?O0NxJvx2D<g&Uu3L~?wMMav{5Qb!qNb_g&?DKA-11gW*8
zgxCmchXPV|#8y+S)YWv5MyNZonh6s84y<N_%vvhQuA}3rqlQ__1R1dzh*ncU<~^~?
z`z}i`n}xJ~Pdz75`Wvx&ehqeEtxE#dwQi;eHA$EZBiOU~o|lC+%)OQ5X}^OgIS(_K
z4^mg+e);kTc}9K*Bt5KMfv~#{bB9nmCpHL1@Un@+?aIqvO8K9F4T7x{QweJm#mM&w
z6Nbam2?5m}2?}d{>o6a#_SIfjpI5k)K#1NS^uR)N?V!c9sC%^!r2_1mQbuB8GlITH
zVwZoY`kAOy{!*HEe{e&&y{mmuN+GIgc(uR1CWP=^9`{CK(`qfX2L00sYW{ndf3Y50
zgVobr7@dI)_4HEDK&y0`HhF~y(4kQF-%?Vnpes4Jp=3{INuD|w^gjYqbU7Ui!i@9x
ziG*6SI>@^XO8pV_v>2y*gTb{SubUQaakqAuHyGTbKDAz@EguY$^z~EBjbUaPioLfI
zRW0{zt!qYmXmyxb)n|aBpQjC(YGr`fu*C+y04on0XLr1=ODS4%nG!?6(C1<fK)Afo
zz3<@Oa4_7s*8m=4+s6GDSr+gO-H2uoo}sCh5Ve4e6otH`V7H3rD)SYSuOmw;estAI
z%Uc;vc8Hx>TWHWCv9lzbCU(y7PVv=55N{JW+uBFBaDhx=N)p;IUfuu@{10;Th(JMt
z$Z%}ss4enN;pA3A=9L5IXPr`D5fhaNstL#;m1<nxM7!-(YE#ns)+gp2aunNVtgg{1
z;;1q*FrXiLn*h<&1-CLz*?y}B3mA+%P$Nk3ajU<GEbrP<WeU!wC~d_`-*eIZXSjN}
zP+S=*yxhtXC60|mGE2z>{i^keO50HQ^_mp>4wh!5Zl0(+q1}Qq=CK0s>x+vBqtl3`
zXRBeqy3=z;77}wJKz1pSJlPj8bPQm6Y91OXOe5W#K>{x>ElkVwqZa0utF_C^l?vkS
zJ`+{82HgPu*fs~vQ%#Yke@^X5)I9%RC=$OECT1^%xoTwY;s2&m(}pHWc}AveiggS;
zkQKH?EkkG-Eb4!s)izfuRLcm$k(thW%{XP@KwTGyg;wWpoJv=zW2dH)$d+E1I!Og{
z{1r@!n0P($ja*7s#@^jdBwZ{4TiE;cTqa?6kO{_jCkn8ADm@q<ffeipKMlr@D3%<A
z)$8R-jN>pRhG8gkiNxzW6UZTNAA-5mAK2^m%Fq^5c}mNt@yQU;QZl6VJ|cBtdPx)G
zbBP<t04WoaK`04QlgP@VbkLUarJC1iCw-hw%GzOKFI`|}A4NV-Rq9xaAhtnb(6FK@
zUOslMTrUR2Yvt2pC!1Lxwvix_N#NHXwqhS1FJ3D?!T03Jldlz>tUxL|)27T<YN*_*
zOUvlrHr5iAhhUtn_$tw~2q~J~O!1gWdWvVpP8Q!^@hiuRr?G*;CJL1!mMoaov{mV2
zG~!nuE<RqYFC$c-EIE%C8zlDeVmnDacD#siL63~y<Y=vEGp|*lF)U+gUrICo3koDD
zangk!A~u$Zh?uB`FkZ(%hG&Kt1~?&6sk{A|8)jnIce!k&dn5OUG%K|thqJ>_Gllrd
zu+5DP**Yj1O^h^OO^&aLPTD*<EJdnqH(ip$D4t&e){oyzxp!Nmn0vQ%Vx$^vmSpt(
zKPCYspv<2(wth6y-bN(7>4+?2h1d9+USp62k+_|>y|Tw%<4*bS_&)xhs$pLz7>io^
zca&hG@GFYFqo9+XM=a-*%A1<yX!uJs&PVedTRHfyjfyCKb#LR--qo$9H9E01L4J?|
zQ78JhiCZT+_G&|0cW%+vl``m~eC0Atw7D>tiaK%Mo!T3fxMMdO`MhCTFbOtH3!sSX
zpsc(7lzJ(xU_*<WgN}yZ+37`3zpl*IubqRvF<(b3#qKpHuLzK+l}y$KWQ$taNJ*Hy
zzX92z)YY`*5`43A5a*_45u|g|C*vmik7_h*`uvz;UsUkJ1UIub<^4ZWLUYJUy`3y}
zEC0q`>AKufx_*p~_*yR&HD-T*%9e6}w3M55-mxeAochwHebny1r*>Jp|Gr|ss^IS`
zz&3i*5Dm07#?9h4m8**&5<YIqU^>N)_T;I(F*3DjE?`x#lnWAQqtHCf!7{-P$mR#I
z0%$>F^;U8<Z6qv)+uYj~q+phk{=p!<hGxr4JG~0{D6JCM6eV9xu4Y!Vt2tTn1UbV?
z*>Mwr-%PkgBYUcLN%-&mZ3D_^COU@-C@s#@i3A{skp-jmUsx*p1un+&4x+`cxmu^l
z+EBDFE~8z)Rz@!%a@Q%}1(a>D!;$7_Nuk`7@oyaezNJn;{yjXjHpo-D0pf#vrC{ui
zk04EsK>CIPQ|jx+r<+K#??283q%*b}UzEj0{Cnar!ft$fd^N4UYxaK<jiUN9M`|8B
zF!Ix<sT;}F0y-=#3NN8mTnF&a!NAVMy_94Db1-pULb|ChQtNlcU?X85y>yTtj;&?a
zNe>DRDvV`h{eyLV1kBWfCd-PnK`#Ciq#_isiF%iEjHtjaVqw}82gd)K38m4iN$BIT
zZnyM0>khk}33J&HN*JH?e@pdQrNdN;>8^}jI|oIRCG4&)R?T2}VWM_v<2dD(%Fe{}
z3uLVosB&jw1Ul?^Vl*CUb{jc5C@>*2t%xW1*47B<ch<lv{p`h^fB+3H7Nx9$ghT{0
z@zqQ{RZrK^^pZC#+cC4r%cA2ID}U6+4*e~<U)cHSIHQH|VXok%-_rW(()yq#jd1q}
zv>j~v`|K*~p)+QEJ%!+*CrF#kSTRWW<I*cnW5pAsq3?3zjge8G-71*5vUroaFJTw!
z%lh(p;$v}VQXr0Jtk;(=#NE7`nDGN%)%b)khrAUYqT=~qQKnBR*ry@>q;jU<A7)%t
zp6+PoG8;ro*pvu|9f9Ux5^RZVy=(4KyESx4nUMFGAhdZ`1pmfS>}iY<!@s!nV;aic
zWP`wFC^PXQq<0GK%WhU%;&rL}Q}LBShdiU<w}Nl?Q>bH@wdk4)?*JDXW(HY9tw9l}
zOMzsP(-74xIPHK)V2AIEGi7}UPC7%4F(srO)G16(AX+nPA<Nj7?XU`1Aj%C6)9jzW
zq&L#!!~lUek~V4un{&*m-2Dzf-5^Sm<gYzU04hwH88A!xjogg{tbwK=K;TW-gB!^=
zS=N@qJGhhpV!%wGljB;~9UoOY#P@4z58{*fn+RPZtjZ*^DN>L@3>G(7$J@(b!eP+4
z^<ue40LE-oc=VTfHM0;a#8w<Vb2%E4cIMsXrT>IRH<I6?t{?|(TG?YnqWXp??15~n
zi8+((T%+z^QimEr3NSXF%UN+W?r+(lw$?Q)Mq9A9ftTX8pg=5zX%MAAIf<ZdQBIBK
z8BDX5SsOQuc6tT_n=8Mf<{I5pupzd@js~S^gTB_Fi$pXl4gC&@rXMvUKISLz8QCSq
ze8f{u80;kg6sNg!sGgKzFw#j%7z!tcOPjMaGd@YWlJ%^I$?eDE0DTOd&-nN9OawD<
zTx0AHfx2Eclg*WV%hgNO#j92LDmXkDUlosjv}uU7dX&ZghKeR9CTsqm5VPsxzp1oB
zV_U++e3)C%-U{k4nSEo~q(mEs8JJAE#EJ{TkZD6V=G0@l@eei1qf}l1)!MSp7ycv#
zAC<}se?67Ty%RzBM*faYZs1(a_nKJa&C5t&;7xJYNiPX(1|T5iXA#kz>Dd@GC(FxA
z(c$}pB=$ag?K;i%*qbq~N9;Po_5GAgQ4@1%T{-gz<pk%G4XbG^ons6O&|kLyZzbol
z>;~xiVjQbvgv8v-`7e4s)wtKYng<JD|3k^XFC>DTu=XsLB#56aPF^rLB4aR0AVF%*
zVRkTmqg+4bx}q}5hZxvsY^buR<^k2*sMX<^vB5io3jc(rw8&SV?QRi#vsGy$rzPc|
z;l}@=n`a3kNFcFetyZaRh6D|AueH)2RV9xpcrU@qzzYpKzR3EsxH4!464#MyG>s(A
z1F1C)sZO}g$@v~oLOe`*vkS=oq~QpY7@@9on_+)sD7lh%SJ+tGN?UZS#{s8%^b&Kb
z?b{WwcQ4XBJ*t?Vh7Cf~a?qJfnt7fXIQTU(G$^1>U?!qe#Lsnqkeb50`o`q2{v)dS
zJp`qKXi}ER^z1d+@sYvVb(V5=L2KP&X#<IkPW~s9H@<L?1;TPMI|m863S18+c4P21
zijZm$;Z<*97|~UeJaU87p$Zdp%J3C6QG0|yS`{OFfHo`&NF)dxL{X4OskO90C{1v%
z=?r*D#OEx_;YZkx14d0c(#0>*#IdYHF<GIQriDt2z&xzJ_<%AxWNRcUrb#|lx0y><
z_3KFf4L~D{?3BCSAF3OKLJwfpvC?f2IJ(tA87PZDgTY`;KCv(meu&+I<Q%#igU5X$
z04>#kgm5o+lXW03WM6zO!<&=BJEq5dn|O(n<F^JUix-Pmi&MoDj}|BS;erX-vru94
z)%CIy^3o5Ym>p@YhJ*>!Xoh_Uq(YmWSzg2-Dv}6~szV|vPTHss|94f|@X>~z$O{~r
zbnZ{p7lMN#qa^>7Eo7ukYi8Os?&=xq6-O_vsb?5{fFdf;0J0Z}zknr%k0MSDE<iYZ
zkcXLdYh&PER)RH9Q`>)9!G@-GA)f;$AlRuaB@;AObFP=!n~SeF?z$~Q2Xyb!TQwT`
z&xz2)7*zN@g5q#WDU?K_<E5i@R~klCd1z{5{tgd5wH`+a$b^-;%2Z#&M!|zt#oGpH
z<gV!%akUBmKKNP78DvrP?7s{FiS`!?Vm(1;<bc_l{CReCxxd_7NLok(+7srY)B;Ms
zjHT!Od1@%Eiz`6Q=5^CHLF%m66BIV&&M~=q$HVxNe~umuBOoVm2URBd;)R>3Fo#j~
zOnH)hOtPrgzR)bY*3_+Gj!0!`aq1%56xfQ38M~6t8#4hh;FmJ<ei?iY(E3g)b&P;b
z&F|(CuKVI5JLvpx)5C#6GY(|^oU%@6NSBnWu(*UdAlCEfBf3qTjD0_Q25Vzoj;_;W
z2=k6^p7pMUJ<S4TF5P6-%h*erQosPBn>CZVv0>`;G+?m3PEM2OE?Odqzvg-h<E>F9
z60-ky`cq~#e7k@VSLSvGN&X1nF=8M(s_B3bnv0rZTAa0iNx+b?7Ccr$v;y2U<bqu@
z)z;$cvF~P=cxTQ}H?E_cpN<<f4{7m6?oB4`jrgZypNg~52be-i+*0DN#5{Basn`vS
zI;*)evD)KTAp-RGA-+r&iD<N)(6$st6E4Aqs!eHcVm!Pt-<$EyG&2y*M{rt`t3&li
z?;62se1nAN0QYFei3g(r<-=P0UfXeUlbwmS_d(+YSh&DXm_*+NrHQI3LV3zjKJCU-
z<U?3^L>fukb8?iY`LB?;a?GGNxO<j;2`DdZsN0B%W|2EK)}PbVF$uBFpLs4fH{?vK
zA^svcjx}0v{(}FiN+fOi`@cq`d`}5ZPv9Gfn<2r38Cb+L-><spUU7>w_q4HuUsb^+
z1=a<98(y@iIW6i}W{N`MpmApP?ynE0hi!I8f~O6|I{HF$S6{>98aU<+f<w(z%^;DL
z4UMtEao+S2x%_zrB7#UTzU)~w^En09j}};~2N@PyLqH!f#cb+0sfKIyWN0$#u<bUC
zoAY&G9$}p(<~hy<cvuz`2{>EUT*UQ)jV%T?mIw^djKjXE@mnL+2Cj8zx9Sk@t<lM1
z6jk~$uEyLQP;#p3FJ9#p4cw`pe943wh&Jn(^IF(j)n52o83Wuh1;XxVSCMV1-`F8T
zWuaEGMbdQb_BpE_DNwcBs32x$Wz>3Qi_C8e=ol$GrA};o`iSU`FkkH&BcZ*c23qJK
z8_dLN7CB)S{KaYnQG>ACu9HT*WX);`gBJFWCS*1O&dUNI5|6~{*&FO!fJ1=*vO9SU
z4PtZbBml482YwQUDnyd2^LCZ@&`!etT&j1~yOCWLX5dsjNCRYE$SE;95&q?AYrPOk
z1@hRg1MORw>&NJYoe3rt>xQ=(2r#u^x1@Nf!2sKQS_QTZQvMH;GUSo+-B1&h9A@an
zMG5TotoCjKl;70>D0`@<OLp?q+H19TbCvdgp}uWxq|UpqZI^Kna<_!*IV1dj_D%SI
z1Jf9Cyn$3tK=~yg9oV?ffyprARG`Ls+{&D>X$MZ;tyS_ELD<vOeQG|Gaxcs;n&rX;
z7XRJKB`K4CRav_lQ;LioE0~$n0=*+#80gJvvtK*dZJULHaWum)baR~=Tzi$uOja|}
z(3J_F+t#;s9<2#h$433n@Wd8r7ba2OI+FDzHC-6jEE?Lxhn~{3lAUbeXB0D{{*+1>
zclB|lt?1Gq>T$(BN*Y^1wQsD3g~b_XXE4^+Mn1&H(Oaq~eU%8iBH^QKDI91rVpG~0
zC#`BRm+x1a`q_V7f%VGA6*CZKOqc70hqO>1q$wF#B|%&a#^V_by?Yb8p&A?<mJqTx
zkxF#KO6|rByccR>FH3qb!LshNW!>u6j=}5zauV$^n`5hSC(d(5Y9g}dQW&b~7pXIQ
zVG{MsTy`!eL61a~5y(r7GsvQh#`0m0?U2)^Fm1MTBO>VSe_6pN6nt30-2`FUNUt#c
zxHqxzN<*-vio%RhZnZE;j<yM9oiq$ng7*$K;Fg=DTdl0_>ZVdjH8YS~+LX1j!@6Lr
z{5Lb>{yl2QSp^R%P%rr3qCoP6vBU)obI!9qprT?#FVMh`YNTY}dSQI*h(AFjER?m&
z6Ym~u{p>E6FE3BbyN#f>x6F0`_%~E(T7k_nd-p7OSsPDX7hToOTM7U?F+ZsQ0zKwu
z6%-V76R=Ab3mm8VFyXjl!8QlobT$!zNe4cl8{-O$8xqaz&nmdAKqR-D_McYlvjkyJ
zw3&6P!dBUK8>zpbW~t}>GX(n9zKb9iW8xt|{;&J%{a1;LSi9U#Jm|vic%~OS%SUtF
zxeR|fzWLsKrknJf?xF0;<kE#qp)a@3?x!=O?78d7r7~Ft`>?fbie`0u)aUxtM3uL^
zVBo(Zkx%AR`82<Ye1=%cD;bQ%CkFm4QiAGl#N^AOl3mNKJNVn>btCkK6e#6s+kp3B
zE8#q16wDU}TOK`m!HAW*%>xFT7`<1$UEXf|Nw7D{+smCkZ=YA>x8K|E9pHDsJLnza
zchEcR9pQJ#d(C`Z#PM!n&$iZmyFE)zlE8v^xv)E483#@$oR<sYC%scVXM1$-w0GtW
zFa<s?<o7~8F68qfhKGxfnAZz=xRB2adAN|r3wgM39xwKK6Xx?GUfRdA6W*lv-bxXl
z8;^VMdm~-K>63ayt>WI4C;u1flLO84C)BH+XZ|pB=V0>=CAd>D4;i|1D7urdx~9En
zyl1Vh!!F%>&KtLPI1*h?c+Yz;SdOF3k{ZLbH{)IKUbZ{OybpS>@HS&nnQ!wxWOdwS
zUN`OrYhDcdo&`8n>&>mxD}@C)wz19J<Nkfj80hWy@H5q;=@Wg#bBGQS{Zo_P+k9I{
z;Q^gvUM-MzvS-r#V+6et*IbI%wwVDe2C;Da`+}^$$4eqzO~I)rJV|in%>i+l7bB&Y
z-I>S*5O84nN}Sn99OG5ud5aHNyfC9YgZ1<car-?UyDVqpL9c{d&?U%w*~5v<xUq+G
zrMq50SjK+MxgLp=kSE}Q<CRzy=%PPQK010ogyC+jYQ8)Asz?jg`n{g|0JI8w7@dB7
zZ3qFso)BAu@`+;GkUc033avcC4p2OBK)2z6!{R%FKC?*w8OrR$&~{KYdGZyqBk1=@
zo!eFWF*7CJ?e&?t=^oPd+CJ2M!GQm(_1#{-(k<sNT3^QI_N^7y)oy4YV;2X;oz(Sd
zM#ev&;2=Rb%ohF{+nzxHUSsEc+5fEW|9OJ*{`V^Sgo5u;z>uQkt>H`nKT-H;TJ$gY
zPwK{bg0cPKc4T09UXq_H?Y?$(g6m9N8xD4EO;CnLWs}!o7`-sLw7BHIstRAzle)-)
zEv$J72!+)eZUc;qktZVavDHY0x^V2lchSyL&hsb%9xP?ybrx`l#cpUep?fGzYX0ZQ
z;x;!3+1Q4M4Q{YnZU@4VcK=wdW`v{d>5;!AlcR@&tz41|O4@14*;4Q1<RYeBi<c@@
ziD1GZ+2mu&hh(j}&)+v>cS-FkD;QQ_G{U@M*9gXjjMyp-wtGj2URN^7MwdiMkWDoV
z`qdl0wD0~G6|5^*Q}AO7Y=o^}eo&X6SMaO?<0YGuqHRtX;ybK18TQkdT@sn`ZA65T
zGH}7Ug!YHa0K@%_kgC6g_IR80hkf4&!1d?uWKbomtoDoxcc$3N=k|?L&o@>^H-M(5
zn~M(ejb_AT;H0@*dy&BQ8o)=JEIgqlPKrzh07hj@n$|F5TFW@sJq`fOW!G7_BK1VU
zv%FGTwsYA}Bpz{#D&-axOzDSsfUmdq#l71yLH}km>yt__WxXhCxesA*W<AV%5xU8r
z2-|SXUzrUI)YK6QppEx7>DrvbG&i`B!?Yh)Zwun?Me#lyUpcU`Z_K4cN0BKJwvCT^
zNX3(UYrjgs`ytRF1PB6pwVj2&BKBeF-1&Wtlw_?;h&}<ch&^m~XHz`e-I0hHc*@`=
ziB0$95NJjliXsbU6x)0$?Ye>#G0ES$Wj?XiL+=8JKcahbc9cNz)wd34U1#H*Kq2)?
zxtD5K%{QLY0Bsw8$1q}|quqwLmG?qqU&Odoj@7VPE>11817<-^XMkZ1YjGgYiB<|)
z9AL#nwSr84{`$(Ug9i^5Uw%n0W?s5*?gS2Jo_PFOU6F23WMxmo^NpOeoook=GIbvt
z@U@)qapD~II03Jd$Nsohm#sKk=LYzivU$JBwuOVvZyt_8nyyg7QHd_v5}F=Q<VdYP
z!KT|kM=ylz8OGNgu1u1=C5jBqu9hqek8U`dkySb35(#(3mW2cEaem9JE{w1Eor@js
zCY{10gJM6W`8diWj$t>MLl!0zEUkVlv561=;mV<nGjywJ?#Os_?9<ic&GEI#`ydhL
zp^K>}k(1dVuEQcQc{FvIU`BT&efHeT8mkRECZdI$cDUy{b=hhRx79AnZA#UUS<}Oc
z|5Nl*qc5LQ=JzWw-1-5<o>B0ug3eAeX6KjHM+I(<B)c&&LBMo-W&g%r*}OxUQq7O3
zcD=O$H%WjS0=OU{u{Lk&ypbi!(y%0?^g}K*<V4}929s=HI>1tMzw}i(b9oS1ko%R~
zBt0%SlI8#~?2{Ru_%<9y0!pr&zreeT)C*H`VHggy&qA{dGjD+Y6Pn%sntXnVpkY}c
ze~UsxhR=RecmJaT!y+wI^23z<2kNMPZb}j-B%#d$xkT#gD@QkWS34eVYfXpLvZK?)
z%HFZlOhh+lx>C9=+w3-V&ybN*iSb4s*jq}<Xno}jDaNQ=0;SrNkD{eI0hzR^AN*fZ
z@5my@wqtDB2|vo+-&QAxZKo5o&7Cjv<=DG*Kzmy{dZ0t3@nh7<<cQTCm8b&9K2XPI
zDIz2hXq1zBQ|`%}^bq$nm7IIKv~w<JM6BhKIiuviOuwv6zodyHlQ>5KTS$3EKc-ic
zK=Nd{>+=0k&GAdg+BqXw^^|1o14iaH>uR(Ch#n3aq1raI<?_nmmu<g~C9)d<cT#LB
zs4W&JYDHl@mvqwq0eW#If1$B`a%Gn?;AX6Y>W$oYo=3B0RFT2sU)1PK(og<J6ig`4
zy!U@WH#Ev_j(=G(&GDO&Wt5GrP4KPe_K#5dpKDkTP(&n<+XADFD!HU)x+%r*V<6os
zHmgV8ZLl3~*US4Q>g~Yp0*(Tb1|}{N!8x9cw`Bl~aEdYM2NS29ej!M&i7f*=J3~sb
zESGb2?lZEPdIp>>%$;In*_VlcI!i1I&Q~lCOO-*8OgzW}J~9$`{bQbp0%tJd2^J$5
zVwy*FU|qp7!C1j89L;7i9CXvS9ksPEp(F1_S82=Z#MC8UV9;p-B4bAr9qQ4gJ&gzB
zS<J0YgJF@8iisCXi51V6fZjI?`c5De?lg&io3z)F=+YAWzQ#`?UYR(tMF2sKyjAGW
zGzNEV9J4!D)v-kNI$EYSWao%zxa#oE__$$TGsIG&I%H|+zDb4OB&7gR6nBddTcm`R
zr&%#&WiQ8Yu*-PYCY-L_H_mKqW!urFy^gsia~?r(F<g+RXgV5?E56>=Nz#n2{gN(!
zo&e{?@_#2XrlEoQYwFQ2Dd_A`|EEd&hPqA=&2?4=m{v$`hG=&pZ#ue*s#=I}g$qG6
z6C{i|L9N6N{sD8%iOK+XPzT6?F}B!rgt-_7)HCEL(z!q`(fP97!?euq>&n%mdT%6Z
zXAMG0)gqBEDF*kfl)12Vos7l17^*Of7d414-(U!oKU<!}v-)3F&-?=g)~83+iCNxP
zc^eEsd}>4AIfBoG$MA|;_6_f1IVO$G#e%8P1wm4o0!DX(Y|Ku@x9-}lD%zw{7s|Am
z_(3AFKZSNc(xL9bPcHUz>@AReyDaH2#<BY-&euqBseqIyrvMMT51=)$({R4Fu-h<7
z!;$MEW1Lrt$LSfU1zY7AOv`LX<dxaVRZ~^q84>5&(dMuC#V~%=->;%Zxh8RTx3S-H
zc!9nsm)hIZPlAa>m?cvHY+@xFSvz_{MVdYFIv09VAx-V}Q#ga9hj%8$^#l`0_UQCP
zki`3kdvk!MY*pR@3Ch+*y+LoLMq!`Y(lbiuOWy<;>nVJPkwc-%Iq;fW`Q<_}`b0Yd
zFFstnXl^z)-njZOJ7m53#a4P_r=IiY4LHWLFd$8fTnvk8SBjI1K0LwkW~t&;{CQtw
zlWnz?`916hoTx`xZQm^}i0pQeTPN?+O;_8LTnzcm%GxXdzn}h;j_rL6dYHRZsVrd&
zSaVatCPJo7H}&gUGyhn@8w!4dpkd?wn@adC0%`HDO8G0FqpTCZihsf1){SnxWqH0b
zUB`_cQf6JbToMVD@)Zw<luF$=gP`tU2+OWZc54<1d=;}&IgK)xSYx|4!u$3j9lBx7
zCb8VMuGWY!TMQ+`EMw=y+H$t6(hHk_f{pfAK*1I>gEXt~YL;Z>;d72_*3&TeF)fCf
z79%Yy*c~V~tXA$)Yu0Kt8&@wK)yiCw^=pt-t()qRY&47IDo<sFX2(l?=Ae_okDOdJ
z5+<%i2{xjoLVdPUp7il^?w5v|MU?3Fem4>+)=Gq#kUgYZ?<tik4c3<C@q=kP+s3A4
z8+)NfqXsq&dzg6kI^S!i=M3}ic`*2;u4ki%OAOhGj=~-xnP%3zG_t#Sf&^9~vyZUF
zSb``~XsLHb*_O;m@7Xxg9d%egbd2{P6>7|FyoY%Na^~2pDYy3ui8i`I*#H1b%v5tH
z_!~W<+)MyPqjY|wbi^lSR(=KJA(@w70X>mVJRK$$DV$K?%@D|vL-pJ(Y4>W6rKIs>
z4$=cTB2JOT0Mc-GA4#(e)14iHo+wQ}IkW5OwH&|r#&d2aZp7Ev%>-o6;V&<R%)KwV
z-cWWpHm$kX_t|*0*IJwM|1@gtpZ*zIYs4w7U|vV9B8G0oy`xrz?wF~kV%*Vt4{UtP
zD4*Uq>*CW1-g&_Kr2U=UbM)Sdee-u$F7^zu>NfYmrn(w^QT?>_fE+}8+>M}O{{_8~
z-r494VjaDqJFYj}ojdi0@-=$Ha&Flh%4PE|Q0$xchH~YVcZ=TGR9CY%67??V(_Y+h
zFpH+y3^A5p`9u>tH1@bZ4Dy%}l`2=iSOe9L7pIOFFCH&m9c$%nA~Va(84gS4uG$r0
z|9Mmjh=iM9+XnRE)2|)JB}D7`%xkUsTC@lU4|KKV3Ul7<ymQOJxg8$n(D<U&c73lB
zcCn3-4fbx6(T}s%oqe%E8yLz{q&VXYBb?vlGR(YOoyL7am^x4B|1-hAyOhmjekMaU
zcqjULY|shqMBw=UT*3cM5GIzV!-U6ovW7R^DZhyT{eP-F|0hAIOV5Y^XqZ`!9;rk>
zR2@AH62M-p)ZLc->_%AGX9T(GbM-tdXekb*H$9p(4}I`K#l#^DEe^?FA6KJ^6owd?
zcTOgcCw9m0g3#Y8HH8>#W#3j%)tcj;HkIc|jZZ_(NDE@MFY+QS!5`=jUepw70~q<D
z>T~8%K_0{kas)6*afjy~Vi-wb^n_&EY{O;fgz1Kr{Q=_n+7Ga3rIfsfs{)*|45Yd)
z;VXr7c_?}ubPPgW?XGUq@+Rf|uw&iT-e!9+RBArAyNbJ3IZK`EMw*`TmvQ#rUGIgf
zs93BIKI9TD>X&g9E=1*4m3X7Pr6q!Hyx6$+`wrBajrZeFfw=+y`zbLf&IO58FFt;7
z@UJ(bh5alvT#%x2O$B|kNoq%qwMuzw&w&3|I!nqIkeB{ECR;4-uApZoxjMi;roTfj
z9JkVv!C+u*8}|l9`XXY@`CoP^GAiRb<NrF>BhE}L7+`^G3o8z%Ca^tTtd?*A4t*)n
z7~UwQ5et(XWP>Z1F4HiV;gc%!Q2O$vg7mR`&HdmC-Tyt!?%yXE+wP>bCNOWx%Kg8p
z`{FbGpHW~^7}IQhUYD9J&ad;jV#b7hhH2!CO#YG5E^=AgCiB*1obRyap<Qu{GCR%D
zI@>S-Y2&=i4?d<Kf0uGPeuu1{+j;5Kb579yyHptNj4YtqMOY4OV6*516;n=xFR$j(
z5uwd&ZrW^m64xF7D?F~mc2GIoC_9<G-(<NtAYjV>HN8l)r%I_QRhHCl>(n(iTW=Ig
zs|%}I&$0mS#q?ULItlIv5jSdAij%-*w4#x6$o<MNF;mlFd?i~V_If7S_emMC^-|qA
zrJrI@O*^()-$vhcNUTKP>H7rVuYH7ot+fZ?&137)p1V5R0J%<bEyQozs^QvUnC3Eb
zXr1sAB{$CU4TX*Njf9R?S;c;u90i;f%UO%%Mi0eI#Scez{B1qZaU{lY#2vtXL0u>d
zhnwBToHY5-8JHY@r|wFj=SxT!<|4^br8FooDQS~V<Y7ST{&ka4g$0H#+9_hP5~*pF
zuOTy(xTMtGa7i&G{ok+JL)1E>TaPKGsrJJfD(%k1t_=Vyzihq4Hf#Sji_-4c+y0;G
zVP9A9j)Ekjm5nH1_vZS_@`z++T0h-v6!W32QO7rq{JT|hXW@=GE(Xy!@8%v@z=68o
z1S&V3=5(uVHU<;^rrIZY5!hyvnsl>%ZA%HjGZe2$g9acJ&t_X~*!PdG1@T@^=Y%zH
z%JJ!DCg;CG3mfdp%f#h^->Y0Rt$LW4sk;{2inYx*oji(Jv;VT%E~b(FU|LPPHuh=z
z)|)i>8lNsAVm?Gs)czBJx=<nuV+mmuMDT<EB-@uZHY2wKYP{u3?u0FQ*PJ=lqa3N{
zjB$wSGYz};T;Y$8)h8W5ZcgSsaeoF<mEYR=$V{bRjB1w*WH6N!W}G1e>+BGYwl)b%
z&-KFS7HA23yrdb%a9KvnDls6In7fidBblDb2I*)=TPE7kh8?5HmgxGJ?fbI5ZD#cg
zY4s_k8eR4~iW#leLMFDtpV1U+;Soa|hADntH-19FJqk`K7*!zK!p3fztQL~dZW?0T
zX>s^~n*h%$t86RS`iM_{a9q7ypIEdlbL=qNC;&8L#b4D-kks%9^c4P$jX#y)gj`-7
zimz<nI5Q&SlMWXb(g`%6thpDd!$O#lIa3|6I0or8`W#0J>SHmK(h!Y!2@JF37y;KJ
zv(Y{EKdcos7;A+9E68AW#i3n6s&)pg0OAN_suCoy@9uy)d7px!0$yGlG&Z2&W?$1I
z@+H+!wZ2~+_Qw<nQ^JE`Awq$Il?%Ezq@Z*Cgk79o(Kx2&H`IHwropfT-X762m~GgO
z7hsdord8EeaueLtDtm(OR_=YkG6BZ+^aOn;Q($df-$gGtfbb29HEelr6Sr3?bqVyB
zuOWx?HwmCdwa-+h-WU%=TNPZ!f7JjC@}sepUhqx@I{>X9L{lUsJ(e+wu-e+uVUlj(
z1Y%A`nm3Cx8{BS=*2cOB1fqh#?<V3*2|59Sbs&lSG_K{n%z9isFqWP876z%wzdaSX
zKI6C;eAPemEbhzv1@l`khiK-S%N^*#Hg}|YCnjYxb6j4?bda8hXR~Hile6{O2^0~W
z9J$bPBW6bqS*24G)uRY2IU+OK|3*D%0%z;)040x88>^mNo0(I!NL|vZsKJ$Dr}j-f
zuer?zZDku=>O|3Y-iWFF@c3(bUli!e9Q=0-xvE|O$7u4c#%n}5@%kujOslRIhw9+i
zBQ~2j(+0OWXG&SD+3B@2?)q%w`m8w_M$cyO&A{EyXyzI8Gsg3lF1INdP#_b3+m_J4
zTz{m5yBiCzl)VhdlB1sABE8hdmCWcHuGC7oOY_V;hIHd<EV$(I<dXLl&D{dU3%T8Z
zFR_LzJ2!%_R`w2rEa;_KYb)>#A}jqSlm|12J_N7D-$Fz$ECL2fV#qlX$KV5NA`uX>
z^@3VlhXULx_lM-sGpWrq{Gt)R1Rdx^EGRiP+^ON;0e`Y^LCo#sg!qo;gnZJSW!V^F
z^H(`hJagh)5i`W4>qR@5$L4z5v;Z}*g>$p7J~BIRe;dlTCPp+t8%I6V9xhI_IRlL9
zWx*ig*V=MV=OQ|Cb5li=K4w8vI#R|w^QpRdxr6YXr#{W$XAt%?J9UO^t#RmO#ZBC1
zFBI?aX_hQ<$`l7-M$O{Xr5W@L(oHsw?`&&XYaPLAaiWn)hvd?z_+huDO>&>(#WTm9
zFGLO)^U-PWw)P4fkx_GPZtGvSik@fSz}Q+Dc*oEQs5VAznr9()&N!%<veTBw&yu;d
zF<kuL68q$V?7yvLB5R<qZ$l@BN$kuvu8HCD7N?$mO{=4yIuiLJi0Vi}tk~e~LSnN8
z)mFtj<RwD7S}jx=t1Ur~Xce&g_3yJDpxw-H9@UlktC4OKFEwCzj>Lh+ilimN_$UlS
zi{YS(s{-_0a<$9s0%Ylj+yg5BJRc)7=tkCn@J}9lB^`;<d-45?FX1kBxbBSZaC8n^
zsL9jEJ$`8yhT1L-YE-h49)5d`D$O+83bko%?Kk|2GmW6%+By&nVDe2}whiuiq#*L(
ztL<xj(4=yg*25Kt?tq)N5gXv}!`@&B6&y#~cwO*~Bfy3fo`+?!-FRDFMQ=+D9Bae<
z5uAY}A^f+mj;`(qc38W+WaHRqGgR~rj%Ad*WP7j!MP6YYNV2*!*r`JrR|hY%Yu1AZ
zOSSe4|62Q2I5wAaa19(Q?m$-M!+JZ1&u~yo6yIU--8epxG5AjZ-+R4KW__!BkqvPO
z5yEA??e@7HbGzntugWR<9_X~)M%V8R_HmGD|6B3~TwEOu_N?s-_F^_5M<c~qC%-8M
zd(4^DpQVodZFLM%$9|cfNA>J?^^6Vuw>vE92If15C7Wz38g4b(wi`ztkBrLqd$b-U
zmv!VnSFtI+9UrX?v+X&d%wOl~-TkNRXf($^paSAB#^TNdxK9a^960Nsb6i5>2K<yb
z;ta|av9WXFQ}Zt>?*sv73<}iFPh8glb*BIFpCo-Hd&)Dbva$V+P5Fx|VXm2s>F5)C
z5m%aE5{Ip!KeN-pT9C;%&gLcMF;s6%&FgC4Csib`^VyvLsBYykn}pv`?6`u@DKIYk
zZz}dT6r3k8Db<y-WJ(#QLUYz^dKv#I1%E>AV+Wk@+PEO2A$rtc5%GRrv4R3^K5|Z^
zMB<ks_i~X`-JH;s`kc$TD2+2P(KT|KptBC>GS|_z<-AtvYrD`EXWQgi%2+jLOCufi
zn2#4@x{j*F->3danJb1wz2;}stYO^{mE-SG+qzwOx7}FBt3_(cl9T`qLnCvAz!p7o
z4&(1sHDc{c88?&;0i0IF$5gS{#Vqs5?B)wt&h|siV=oKDm&qe83=@ZpLP{n2)1wGi
zbaw>RvrNVkf3o}E_3oFzE;WSk4Tex6g2HWC3hLxE#Yc&6wwrG(8{3#LB}%OEX$aV#
zAffhq1o%`E=$Fg`Vn^a{Kq&qSjj2D#9nc5H;3LM?=fpBdhX`Aqn8rI(kT=$_i=Dh<
zK^I`B%f!xB#(#j)(rYK#Uo3>vQ<Wedb3+);$AT`dap7`9=g+464+K5vz0qLGFdQNC
zw~<qZX&?!JG5d=NdnNiNJ%@HlTG6Zl>i$(w1JA#NAbbD--Yu;YdyQ(*b&p-!J@bT_
zlViDF4V*bUG6?b*9pU$C)NFBUnPcf=3sCNdY--F-vyla(gNj>|(%<;~Dq<@rrV0xx
z7s-d4asjs@1eLTS#00xdA)+=53^W9cty41b1?B|6r#bLkkCd?k@C9pAojw>v%gzr^
zDW!Spycx;u2&{(VmNm0EA>L7)ZnA$-jozk4ce2txG<59yG*6F{Bc;>%5dPxskJ(k=
z%Jy#{7M1~=W~ugk=nKuVNxied1E()8reiDFb`m7LvMHzsZ|R1UA-BPX=8ufN&6z2a
zubU~FNAHpG_jB|xxR(i(^Y04!QN0gPYH)S9n&t}4hOX+S9B2Dg8Ygo&;9M2XPjYI*
zlXa6UGo2h2kXYS@!x@ZtaOs&hm!1XmBXJI4meM@N<}sYnG+la*@ccz4oVI-p4%>MX
zaL^o3w>=gNaWG&SS|o0mHs^or4*7=d{+4yXa}>szgIi1^|FGuT5e4R?!~A57ak<j}
zk~=E6=%Fm-sI-&T`TmDB;6h`L!Q8IPpU~yw#PCdMoTK3njYc?lt^#<&6bst_6^{Ir
zb;T}aH|Fw9#U4_=teygBbb_%Rs?|I=m;n8@s@~qF-uyM)H{_tD;P299=Zl10^~wV5
z@dR_fA_XSCvCEBTG+)wf>^G>bCgWkpG_*B0!3D_uyVNzdq+ttV{$6{O9%_@`;IcLl
z|D5K#>e!jWLyi5R=J<LljWGzbpLx&hAH;`g3K*S=??y$k6{L-PM~b&@eAHXoJelNM
z6Q~LLYtk@ETL%DK9l#j^vQmF<^gH+_IvQsZu}XYKE|Hkm+H}sb(g_)Gq@XN69Yf0}
zt>qbc7-UsIv72MmA@q;Et3%iT593irHNRHRycx%pM8bdiQ}SwKhIuWoHq~Lfk1WuC
z#drh&Z{uz?tvcQ`I{wq~Xba;s-0?tw9@Dgi^?F`MU}BmwU=CQb9PG_WD%DJTDs*_a
z4#2GU0Pa5-<dHEAu=?`BHms9ZR!4Y;-e3^w9m;$z%~hXK?>gqO-~STfz}z5bqa<tV
zkwkeY*bdjv8%mxHWy+ZzZ^nbsdU9>#O=w%F{3Pv#t__ArS3RSkGujZ0U;-k8hi<Ru
zt%C6mJ0eZ%zj27J_S^0mXg+hnb~MdU^L8|5uMQFohRku_5H0PG9B=$b;saivx!w4K
zV0aBq!L*|({}ytxksFY0K?Yh~CSR1h0aCmCCNW#_WBqNO*ru3?%@0$>*g0Y8ubn3V
z!^U3QC*#$6jF(~}NSIrr6BKU@Q?*-Ov|7xbyQ`sb-hi+Y?s$wIskv#7aq5>wRSij|
zmI{X60zdqibf~p@&XD?5R?pYe-tO4S`!>pU<Od6OxW<9p5mp0;w9)WQ-xj1fr=&NN
zPE=f_M*Le$Ecx?#bmXDKD9hIK7rFI2dcG)|Oe_&E-q!JeD-Zowx1*zb^j2xU);-$t
zWJj#zGzyJwijG+^X2QD4xma@A*?Xy=4c2^+8l;J|-WrYlm$$NcviGf#7g`-#e5!)|
zqd{qNJ#5@cn7Md;63Z3ij<k`*fNcv`$SIJ2N<j;t8p&f^&Q9N>txj~Pfp8H2AieF_
zi<U6oj5Kq~a!7Z-qS!xF%#<WzKAZ~G^eA(>G)28>RE%9Q4(F6gm@fxw=Euo5w#(rR
z8*xKSkCW;wtS7ZAK<^TEV?I=t(Y2B|8$vFr4d8F}eww@?hb~+QvnJpzUvRG}F^<C~
z#vvS2wwz9Sv0Xi482p3ksV;(2;o50g+tg>}S5;4!-rf!JgUTeV5^mda-v1V3Z@QRN
zTDxF%przvgFyV%VmjYh+CwKc;QX35tiNV`)O{}xHbjU+@5+yx;wB^Q0KGO=6F)ys~
z!q;<YaR->HQqHOB;RX)>nO4D+7|q5<B$-C@3BG1_6ZAv~OTr37sjIyadg{Z*&3yMo
zD2Sc@2v_<+K(d+ZN3F!!!ItANB{_>(q9G8{NfZdgwWS1F{gFCbOzL?MNW+6S+f59C
zw53ATJJsTA5Ny;@d(MD#LB715rK1UyqBuOtYj%GS^fJ3TMEkl-I5ZUW5=RIHUt@wI
z|1-3#7j%;Yxm=BqOMfs7azW)NcU+K&5(mYDZ6Fuh_YZ1Hnt8!6T!j-jfhzERl;*8n
zD}NFJ6aE5%VbI42=mAnwz6n4d{f0jJ0slP_`uOX_vtGBMk6$!%n*PYR=2wI;41IL@
zUj}7S%8*AHBb{FtM`I*oKjdLg8xryNYbE_Qqt_-3G3mvlnw*bu7lnEY7POZ9cDEKj
zr=p^T+LpX&5Vl(Awk-c!)_6o0w$}I&HUE8T>f>s|V~W}GG|2V1y@PJNSFtA)7@)ES
z?^fC73ii+GZs+3lpH!Nm4-1R}l=qlIn(01H#i@)l1-7%uIh|alF<OgzS%WD%Vz$V&
z_P4jzAZzp;9X{JweK-_<UHsBkhQu9?)y=8SlsKTqM<`g=`qtF3HsNDFPP3SZ`kFz)
zbZWH_(RcUN6VP|;)$;d8^c^gBgM#~X9Qsc8wT<YMkp-BJH|}nr?^vXe3k8^T+g5A3
z_B=YiJX0*wNjA`nP3puy!_~Xd6~4T!Icb}9`aQbbfEh(coM#OD`?$Gs&qlcADC>76
z8~poKtw;smz;J7PMI$_^y5&r2Ws_j>&UN3B6!0I`vzjR1{}x?tcq8HZ_v@*T{Mnw`
zMD0?9>n&P~V2k;WtZ{}1I7jy_WNfHf^BIfOBf4)3%V@T5UZ>;!p3}3f^P1SqWQ6bG
zU23~E?%MU;hN<M`K}KQXm{vx%9=Q(71V^gHV$V=n?IHmH?SS;?+JhL)aB(z=X>DG}
zAUG0-gWcCO77CsWywegjLA}e-D~9R@2oW&Bvm+t!nBkvKZk2~wY*2;0&B={Dqc=vY
zqZe%J+2iWU!iKKoC_D68VL_o2N1A~LTTd*(4jAxVXI@ojik#R0b2w_wvPq%(ygK#(
z&lSj!C=jF$$_L?VzhSoU4LQfKcyhDqv{g-q>NG=j`WhDh9U^p#aPR2=Z)J3~GX&{_
zxnvh%lmZB;q2$mR82}5^2`EWqvXiPFV+$S1R#NXm5vn0Vttz9DW)87fC~Po}eS#Ru
z9R`A2O%npEiEVaWavy;k39P8>Dbw|C%wkD1WzGOQrr1p&b6X9iJP|gYrc^xrz9_f*
z%CqslTu=a}a0_8H<$q9@7YW$6xacje^q!Hf(D|Vx;{Q+F!j+vZ9D!PA8TuK;{*j9Q
zuL@ei^e^gilDpwvS^jlwD(<iqbw}cUMfI2f1bR<u0b7b%g`A{ka(G$3`hTKYO&DTG
zN;A++6T?c+05)FJ)KFnI#Ixc=FG|v8hMOjf6K3ky=A`7vZLrPBjjQjQwDzNF!8Zur
zKFdR<`Ux7J6GXI74H5k{+7a~mCb+1wQx(ciM$<ZoGvRz*nsnE8yRk&I2otED(o)T>
zZd)B$-Odc!HizRC*l*_tvvuwKU~WSIlEM$<2)~T9W=~|M{Z~*eV%X_v3nx3C<_f3}
zhnaudgYD$aJ`$^5Ahp}eG5Z#oO?nQdW%kQ4q|*``az9Gpj@O0S64hQ?4Kn^lBtQ*6
zu(EsWI=tR@s{!MKffO_I`u95%gb&54`KvM1l-t>k`o%^|ISy6VGQ=K0xFC<3eY|To
z;(?+1Al^Pl15CC9=j{{KBCQ&>mGPj7ArNywHL_K|_P-;P;GWlSbmJe@N9s}??v!D+
zhY`0o$lFN`xr1CgQCRLm9l4V{yCs}3_2Qmj2X*Z7hRC@)7+vEm1yqE)*G)Be`EQZ8
z7>ts4w7x$W^@gF=_Ym9TZL`>cU{5gW#?u>t>faOW;a#@#F53O_od`rmZ^XYFtqRXS
zh*CCzG@iYV-j27@fbG4j`#Kb(`7@liR6E<Q8nxE!<@p?rNX<K`xj@a&l71-INBaI?
zKk2(j&*_PE@fb(yM{nqO-Nw~1cSUP<(;9@`8(I_h_Lzcn>-MZ{Zcn2>o9)`m^LxB~
zb|~yWp>1QlX)jd&0iLA!TQoK8j|B%1c@2B}nR^F5ue*4=qeZaOoh?F}alkwDmYh)>
zTs>rcaFAym4i59K9B}I$`CRG?iAOlWx=pR@dha(oTATlD!v7S)XSd??s}-SNRzxzH
zNUk1PJ-T{~QND{79}SMNyD*w7t7G2LU<?y$#D2k1?-*;P5FCYrAx6rvDCJlqWjaY8
zMENN-Hm&2F$9TiLctZrhZEvXaHZ;G@&n;!{wm03odT7)8Bx(<JypQJQ-bM?Xy?<O!
zre{vzs%c*^HtL>x^1tc1%YTOFp5nP(^zbk6mZy0dk`CVTZZkeOPV9uf(-PyJK`)h9
zy=(REE#B}aJD)UnD7afb3Rz9ZgX6r_y*wq$Q~qiF<l3n^!r-;j>yStsgq?_19*}_h
zzuxqWd!Khdo^iA1$Q60UePA4C^!^_R%>SW+KUVM+1$<+st`2)U)zy9i#1nSQ_Oh|}
z9NF8ySDD{c&{VO%y@M){2%1|gA>6R*;_|`-rxe+~j%^*=85&hN<-@8f5)O8%`<h~P
z1<MLPMiAy+5`l!Q@MgM=RC|iH`R`W)lkB8#d5~S%)o|ro{8qWJvHi~HDyBH_K)@|B
zxLJe$is}+qSsHd!6gyB(I}bQ2Zu-AlJ#03sX@7t9#DvH3B~xX7V4#cYz7J9O){#a-
zoo$qDZJC{(#YA!1uT+XN%lJxX)Ah~l9UBqu?j$H0MbI(gg0Hl*^KSdVgoBNiPfgFb
z?B!d@lV_)$wnv!#<Vt_DYo4wy;oq?igR!AMtSi*5PIM+No9($qq1LoPSEZ!;(G5<5
ziM%EkpGV)KTcx2!UTI*W%(u2R^R<a8Ys@8Qo~V(2j~@9e8t%S!dD*+UT=u_11zWU|
zF?=6yrWxrZhVKs{S^OJBHWNiMVtf}L%V(T>T9MBiye+#8>DK%?-8Dg`k&2VL)WUIl
zkHp1?xo0ZZpY(lnw4$s0KBZ;qi)DXi(vh2AQL>3vOoSm?(aBtXMzI$4X{H<|AiYQL
z`ElL-yNdli1xa1z6+Ery|9zzyU2Do{$*g_J^nB5i&Pe3EVlD1bCMEtSRP0|68$02s
z)t8m<Nd=!$AV$b7JUlMK_GtCJsWcP9qFT0L4+oZ*8x8HX`CLa5s-z9WTMLI=&e`ed
zWmy(lUSIyO@3<NqSn0vRn|bm_8;*w1xYb{y-C<5Q)Hdgn#OP{Mnch+}jn+26l3j7e
zy-ouuwO~k+*@)mMt=S&=l5n2Ddsf<CD9m#dv0EADPKGU-Q(;$Q{<92yQ?~asbCs25
zlBwTRLw-wviLoT8;%ttV7{3;s8X||<Ho<LL+l>A+d9N^|YyXZ{?ad&Rl6W}!>;HOx
z+Pr4L92jc=t8kL9avZmmR3d?ZN;}-~3XM=p<|8fl+1l4tpgtW(V>J^;rGI<x?E<^u
z-OjKP%o8&l*r7X#c=tPdZx7N6XNV#v+%6q1^&5rsV0`a?&g2guBs*(Nik_z02CW71
ze$<|#H*sh1sn%XSF(GdY=DTB)sc>uUb_i#(bc@!H@w<}CfPvWr=G?kyBN<$-L<$?z
zwDBS3pM^WY#$+ywfn59x=*qz72`ig(Wu<-<&rGgXipu6?IuJS3N?;<))=`t*%3L^#
z*YP%sdJ`_hxZzZEjWvv2>&booiaFluzY?pJC55gJklHVC2(@!i944HDaKM?lA)S6B
zf4E-$^V+y!r|@#XlRK3X9$BIg%mr2sb4It^WJPB10B>`bWoQ`VUguYQ|H}TR?_cK|
zt35wmeFgjBKc@-%2*FD6rHXmWFD@??pEX_Gvo98Fk&{Xvwi-WBdCsnJdOZJtz|0Gb
z1$#pC&1|Sl<RwHd<?<j%&b+TqT^}3vKcmN(_{u=*mnrA3E3K^9M-)sbm{jly1s_-N
zVFh<9_)cYhOTm&ZFDsDR&ZikbheiV=T~q3$a+(X-g6oVjR$l&hlUmAZ_l~Zv>J4lq
zT3{>jv*h>_tweFm&da*ilny-{WFEU+m2y<?c1hwGHK}!)`UVb8SuvXpd%@>;%IkFF
z-m!5*?H4k#xvkD|TP5A>mQ<!TT2HKD{^B-mz+>PgVjahJ+Wd%nwM-pOgcDFTd)c-D
zeTn%BS9p4mY0I$c>snW6*Sj#?g_7pJ$x~zjI1ed@8`}qXcEbj8pStFz5~M&t+GyH7
z!-X~)ur}@ctUjvjZz{Xli^x6Tm9w7VQ>l2b`ROgXTD>sBn>){VfQh|yvFtB4OBo$n
z9pk*5`*_It3*+O~uJh;p@1Y^kYbAc<#a{E#y+h{w`EmZn5@h=GYOTP=So%E4SljSJ
zx+zB<A)Tb&3uKS#HS&rJK4V?y$9bG?sho3@Pl^&SsA)=AF8I&t6|J%h7tUXBj<xYH
zc-+nF?^6^00zsH9<Gpx_4acvktv{vof@0RMW*GX{m1h1ce^{{}Rq!JU4Eg`GVkX7<
z$BO+E1^-OJzf<r*1x7uj)E?VN{cTDcS73H-{kk!vV21)3F8Y5-!H5D=(i#adrb{CX
zjPfu7Lx{>8>H8Wz|1ayZwSvr1rY$4HB(aO+sQo0bmWzQ@G7WUr`0f3J+#?C}q@bt%
zM7;N%TpmC7`NB4Q<989u=6VbLnR}JH|90kH<YHZ!0y45B{_%Tr1GyCUhhrbZI8yib
zxmvs5$?R>^-~Elu18woM%^U+Dq#j(tjJEE^KE{bvdk5pb+C4>+J3fE|`_xYJ)9|m~
zvcV`N<(H*`%(9X~gw4@U_>WJUA9ih+%loqw>k!|HrV~2TZ*<};{&wT*P%A9cgrwlO
zox*#gEZWRTcp3!5Qnw|Z1dZ$x8Zj@Hsd4xlgR|0Dmz(^|c?nAo@BniX?wfY-5SO<K
z&Tc^bj{j{85fJlLWwYt{L%cr@Q#aIC7K7xxh2n9kLDD%dp(2tamvK1x=cYOOHqYU}
zJx4>^ylBB#J!!udzUIC5n2SPS<Is*KwtB^9+g`<QdW9ZuZ8z-j%OkqD2|MR$I~^p0
zD-<J0SAcQy-{m|Y*&S_wJR~nOK^%!qa(lD^9W{h<<7uxWx3gY_f~egi2rQp3+|QW?
zTw5%xessgy#%f-c>=`<r^xL<3MtdDrn&)jQjp4ulhLgV5W{p<ZqP1JsM5!E_?^a9l
z)U_7X_1&(n)?2jNzo}e-azDr0a#Eh1mD+eu$yBX&dF1DCTlCU4!3^j=PVd`P!{+vX
zV_4jb`<NzJ!X@lis%D(8d1w*}@+uV$-gvF=&`pDbzoFV**X7UazB_&1nMnSz3hz@;
zR3IkZ|If-U9w1CiZMo(4n5Nq~@|$bNJ%#^AlJ?Zj!dHdHvB-|6h712eX~g;qzb{>H
zH&ZgzRrn*3+{ChlTc$HkW!~W^sm3B`Ow?$HE^1QH0!MACWlfOpmX;Ae28HhkH3cY&
zYN_UQ4i{FtvkrK0+3{&YByIF;KqP?;FZp?A9R@UdqX1+QG>Vl{m9(~TdxHkG@|^mC
zxn3#jQaM?oC*R`@dHvqN8#=Fo%|xkKNowKYFjDxvykx?P=6mC%>&5Z01P9?PUX273
z@=np`OgZN{<a}PfFQyn0`v}tMr3CfFDc9~NkcSl(Q3}(yv^0JuW=r_JnyM^F7O4^2
zf_!p17%QWch^*F_F)g4aIx^hW?mpOBW&Tt@q3Q8Sf=Eph_P1K>NHf2grOi?P77*(h
z4U8a`O!)<|@ah_wL~rb!Cfw>E^z`oPhbQ=c?Op=ZWDSqIz<%CuD~HeWxC^K<=gz$S
z&BXZmo6d8yd9c+Pe~l4v0C|j>H@LL*{|-IWhNlfhe~M0diFf*%q#089Hhk8uMBW}Y
zz15aIe7*x9_fwaEvSGI0AQGf&ebG2eZNzA!I!Yll&2Vh<VDk>zU@4aDe$rG++?W2B
zEej%S5jrDSPT5|xF0i5$jch>(y09mOu>%ChbzF-MI>S>vOd3y+mmMjVq!@ias>zJm
zk9Pz3vwVBJ?C05NfQix+x$^A!lE<DtltUTzlc9AsJJP<OXPSlJQlm`;@^om@)~~2n
zpVI_;L9uVuG)!H1{@nAJICcUkC`FsDMx9mY7B}hCOHL4EK<|QT8CCF|jalmyjwX4^
zY7+OiNpF5vrT<9T)ev`9sHVUj1tekm8Vc#FVPfT4*wwaD!#wtNrsu3l{Gkbj%y=vR
zG9|5nBF{}FcH*uC;_bGMeUS4K<_~lnas}i9tPD6v`P|twhorHzoz-om^V?t6n}`_i
zFjZ5ZXuWmut-Z?X33}R8lvw$jin3oSN)(oE0}9Amv&`mLPIlBS6(y=J)Dup$-c(XH
zuAP!{2!X$pl*9Z=Nh!<U48Pm?g(+Xdl+7KPgvxSM?35TQl-n4R>0|?PqGn`AfM(n;
zp~i;w_qtGPUQ}$Y|4nu_V{o>O+PYC|K27R?DJ4G*`>&Qr1*zY@rdo3#O5u*zkE&(>
z6x^!enAI?dTJz@CHF!hsT0VygavKWEpKY|MK5W;&#46$`u$KegFmEN*+`)RDe%=ub
zHWipV>mxiLg(qfs5219_4#&|*f%$*g)pl>26quCG)OXZ(u93%7m?Nk#xAP`kqq1aQ
zBPz_jtlnMZ*%zSDL}eM9D@uVG?Dj^<i8^y_KdnMBx$ab$|A@Q?P2IVpekj=C?EoI^
zCAQbwX|cn>-e89tZEqJ$@m{V`6!JDwVZMsB`yNy~!-jK?V11Rq<W4Hgy{5w4w_4n!
z(uvpJ*QwH>HIb^4n)gw&RF$vN%A-M%cQ_D85m+R>%iC`%#Q)5_V?3!FGV-9!+e2^0
z=)G4z9mm@1fc4Z{byQw&%KSSo^``$u{VsRSu9R)cQTwFf$AiGILvO|(LA82$p3(=c
z^p7@6@0`2a)T3WeZdXGqe>0bR#;-({Wrw`OZ{$!mH=n4nh5u+h$6D69rWcRUiw8IM
zBIb;TyrZW6JG^?t#_cdYauikMfOm}k7`vmQ^A9@uSc=Z$JVnYzAS1H^)fu5wa3}RO
zHQe2%hC8-8wrR`~wcQ={#pa~^<Tx>A#9hH%wB{t^Bj0AH;MEgOclKF9UqGr;p!wrm
zmQuh)e`Q3WfMfZ+ZSb>pD}JI$=}=5UXYCRLT?4st1Ymko+BqRaT<3WBzS+qg=B~rc
zhLq1x^mPGQ70)#-<u-xQmYlTBXs)qe*3iZCdWjblG&Oqux9ajm1^<EP_%A7TL4i^5
z-&||wzpTt3RA81*ud5cp?G`EC(GO%aicfMnJ5i)zULVazW_e^@-|Pm);DRcY$;{1!
zGoLX;XF05oWO?^X^pUf?yYCK&X6qipF|eI^?3eTsA6D?1f^S!#(-r-XC}=9K{0T0{
zPW>5FMqz4hk<EDzsJM~(FX}Zk1KQjUnE-84X`RZXmI}$qRlcGu|DULiOp0UrBQs$7
zk{;BlFdDnNjmFqQL*#pUMN<w<D>kD*vK~|K__K=5DR@q^L%JLPk}iKvu`elYKtW4w
z^F6xz7ZiN20%>vldBu#=(_-=$bXiqk^KMZw`9LrgO)0ZHRjw~A`InU<d9dkde61m8
zE?ie^S%J->?^n#GlEhg{T5b|k^|BpIx|2kF<hFcYN}rO$L(@jB%W<QXJGCu)G1qS#
z8vV_+Ez+xG(3TY9M~rf!EYCYeD9%C}n08IeM*Rt1j-sPx-UTe&gcH?47;OT|zo^&N
zY8`JAPZqZ1RmB^$2_%hdHPPQ4S=Sqs_<(wHm<QzWamLw}_}0#Bb3(S^cYNiT;%3Ju
z9bMHr8QJF1l_?KO46&MHYQ@a<=(x%oF)znXH!-qeWARO7M>TeHz+$&I==FL1#>f_o
zk?jJy4~3Cat@E=;{+FwjCMUb{U4>%f^S+7NY^?|{k4u#4Y21!f>&0nTU;@XrwdJXc
z=545{F)u1>WvPDgtSOvZHC<Sg3w0e7(5OHv8d2v|gS(`t)`~fwl-rbXJO)|un~9YZ
z#>kyi8=UggT^re4o}6Etx>P%Pnm@N=Kguqp>Bb`MB~G7?Hu%eGV7z%MA540!gM$;r
z-Q?0==K}ZQsLkxSMvPC0qdL2xgBkOpU=XEc;NP<m|L%BoIJPn-rG!#o>6k{;U}ZLZ
zfUz+KMInb>?+r}CK7caA3`XtRe-uigAl_>|(RQ#CC$xQ*cGwAqD`R+oa=xA#U095K
z^31RWjl(T3R;Dg30g3B)f{`@S0pq5yAna+Bv#$4lkXae#<+2{H@e4K1{1YRj^9qGP
zY!?w5vD?$+jX+yI?SQ|A@vy4#e~6$oXmvFoqDuXB-Tk71XVoGDjLlct$Y)798fqm;
zx8L1`|4w8LjlZ2hBs=3h;;+tcoX$<r!}%$VV&kI*ix83Qp;o66cuqmAXPgq}yLlJ|
z2mfzv=NcPXS;q0{Ot+nO+UfMNUG_qEmR+_TB8A<E5fot+i5s9XSxj`{OWV_Jr}VOW
zX1d^H7S=T+sGuL*n212Y#LF51gNfqZL_cUEF&blFykJcH0LB;$5<Z~y_kZ4V=A5~-
zdm(PKXU_e+=bZPv|M&kq|Mz)1oih|RF$_zW%nK23xK{Dv=&9a?0Wit$pp@IlgAoe*
z*D9MELf)>Yux8Ll!;yJo;>2#^hT%@-ZJ&fuGsk&P5~~)&slHauu>+1Sey>`gt#`>8
zTJ6C}tNnZV?S9-jM-@><x#!X+4sll19FLmAuL25G34@Or75l)0h@f>Ozda^fBbHzl
z{&r$#HB~yzmOk#QS21E%RJQ(TnUWCBfa~{FVR9q{%~ms?)LfQ1BGE86@HkP>aMaos
zFUP3zbL>A@&D3++*`BuLdA1OqAlxFWmD%ctr+n)byue%Vf31m3Bo&3lAm)Ht2tLRB
zcxC6fC1Oxb(n`kp)6f*>3AMYFC-u?9?85X4k+4gJc<83p6%ySQyrO4mV5opzetEh)
zyPz|e3v>>@teP5kwaHuk->U>KYVwjMHBJ7Y3H2jU$=*?9SSJVHMiX&EzB2fc6@nuX
znt+IHZWRh^vs?U8$5@}|Pn!F)Cg#?CS#$QS=+p53s>LUntM%e;BqH!qk>K2~XzijV
z%S;OWuJ`e(R;af6w`lTrP3-K}8T<!}HR5*M8w6+ljW+*Nlh-x*w<ZSlHZiEbmvyh{
zv$~lX#omxBS0FA21Y;adkEE0aNTLd?y~Jorp3WhnU}WBDg~0Az-)FK4vjwk*>6}_W
zzv2ox4=nZ+%~09k_fEf#ybTiUt0;(C+uQhGlFnyi3HB*Tg%w~UZr|7vim#CsZ`Jpz
zD2>cXUe8=Id1^Z&v$s@7VZquG3bVsyXY&|EvgvFb!xCF_g=S%P8Vkn9STA8VHUX|C
z8FNoy!RYgP`F1N<$aG#@2KI{_J0{Eyws`$mFtV`u=nS%8^mzlE;nDVI@N&T!Idhxi
z(|nIek&$2rWkQmz<)>+7auPUG(m}hcL30azm+N7{SWA?C!5+id4XoW@tt?1G4<?UU
zELDyg;wTBXuPKjU_t=5FHxzE~ENjppBfn|GuE$a#Tg5i)9=MB2uiDb6M?5Oy<3r`k
z%G*f^yvyt!BiKE*a{nBIEhMps9Q(=3v3u-dpD~LDrASb_W%WRcJ<6VAh`jSwQpD1-
zhM#zeefOA+WJmdm>JD!k){foG?M9NVxxG~^DDI6T$7T)7!zQqn-D4TXyAv+G-MBP%
z6RZeZiF~o^G2##eBJRu00+K(qr-e;-l#VpAX`Wzc=@|CLcw({?EmPhrt9z&^d#ii7
zwk#w4NTqdsp8u2kxspWGTwT&<*)d!RuoMh>dnjFD<I*9a-tS#umIVaP_GHGnQlVPl
zN_)A^zE;aZb@&Jg-LafJ*HC)a7Osp~|LcNgGux{9bBwmCb>jbyuU!nMCOmrBYjDAm
zy^CVmaNo#;+>UX#A+Uw<3*f1jz(%9)%g^*Ytfl{(p3x;T)Sv4tw0AAKMv>FPRToS@
z_v^*XX>y4~^j9iKBO8?jnBLzdu+TIt)V%M;M>)sQZq_I+S2HJVLM9Qb916G6!#olv
z{u?7Ak3>eor_Jc+Nb(9Os}uZfMnUN^@+SdauSl*DSdq|9qn~c8k<PbzyKry4+1Wev
zMrUsrV!IXPBI62U>K@k{Fjmo|h^fQgRD&X>ubFb*uM_lXqQ0-IUxMox1Kpf_KE|=X
z6Sn6$%G4&^$y73lEbC2_CF{9up*|$pG4dZ@d(X0C!@n)TQc0$;$fN3rsMlwHQ~6YT
z>b}%|TZcm+?n||GbTJ8S826)QLqF<aM;#;f`GS!?k(xPzu+u^F*H*+akE0kP&S$8$
z4&-|i>f}l3NA<E*1u;;CZzrkOeRRy&RG0OaT;i{}$@xiof~9+O+Tc(^ToP9FzH<I#
zaXN36_G(EyT#(L7=dO8LtwTKBTA!{*BxGa7{y9BxeedBvJ?uKiJjg=m2LuaJLFDhw
z7^8$$D#k0imu<3UM_%Am<OQzf;@1?p3zF3u#q+;G8Di|KjeobOPmoHK*nubsHA>*y
zZGK~Qf;H9Z`4q#_eU=m+j4~(uZg>i;2+1d$S@*zNq<s1l$7=L+;@4!1HtZ&^yCkQf
z`7;;AyFLIrQ9f?^Rq}i;!le})0k0<>DEyoX)zIPJ3jT#&(tg6M!L=4G4++vzqHGb<
zlW+*+KDnEt2NrYmdHRVn)a7D)Dm@*6+Bk;b2I?CTkPT$*uV#pU$vl0J{T#Oy>phj8
zeeu5XzVM}by^ke{;I+m71Wa$h`A(0K7EntjSj|jT-c?raRx^mGj=H?ArFzE7k_%@=
z4VLvKs`qy``Bz{0@A~?>GhypPwYXOPS9rIDQU58m=1p!^BOgZ24Ydq(OnGm_C33i`
zFYX2|SVehAi`R3$TE^t8$5)qdXS$3e&*(B|w1Y${n|Ng1xE!;1wgDRdtCSkK%02!W
zW{!5w%r4yPpJle$9$>~d)lHQ#n-W>ct!Gne#6?d84{knZ1Vw=ZmH906z_!%ix(Bww
zRIksm{z)oaUYYnqI4zQ@v>14j`WzWaEM2qW5z$Rn0$K7L5?T$_sioeedz+-Qm!e}u
zn0THJj!E57-91hqYI{Ln(^_lOH+mxsYuc~lBZG<AkDI<*o#J6Vn<4fc<^BP&b%v@l
zShj0DvXizvgBX4#{jaMqR?aUmcdo@EYVc&pWR>O%djA(-5!F4;gk=jr37KP84_49b
zqaD7s&r<A$e{~qToz-y|@-hZ3itZ3Wkb@;W(=8Ar%V;A6_18kNz1b@dFHg_MjGzY`
z429^S0P4%DE330>M2mJ7&iq`7V4W6($k{2uw`ImYX#mm2jr%0G5j5_Sv1o9A(6D1#
zqG7}=sBv`ECgy0ws5z-A*J7*s(1h15cTRtpF>|Za7}f^d0#-VbK1MD6QPK}%la=+3
z^)xpUB3dI5)uyJ3q}Th!5C9}#$aYlr2BWRpJg<#FV35<)K|rE<8r-)D0uB<4Tow!#
zx?`|#gczI#3UvthU>gWfnk=L+m{w`V$t9SPNP#g6NL>&pAfu41MoKZ$4{AG!rsfk~
z&zGer>wP4}8lWP+2IB%j8;N=YfSAU7K)+t!w!}_oI{I?K#!vto*Kqu>Bs-A*IBX06
z8#vwlM*)gl18n3BHgZ#y>n;X1LQDL7h#I^N^M2O4XNt3H{J#~Ys*PtHXH%_zdZ{?S
zLIqKJ(UN~N?~4x1`}wJL2T4p#13&-Pxz~Ps$IORsa=y+(?BFnG<UE)ogO=;gMx(u+
zQ)^=lNgZVTZ))8&DpKb(`8hR3Gk-=$8^#RAD3RH}-cS)teq^k5Q;hMS*4YIzwVvSP
z;S%92*Vzbkp3wJj1BHk{N2X{A#T@Ktj=@4|bcqcz)~z5GyBa}g-zMKsdm!SXrW7=l
z;BgkhzF`DHRAm|uOLj?bO1W>aRFl=7%rO8Hm6leLO53@J_aVN9N=u3#Y|IR#u}aH;
z<;zrI84v+<$@(SeN(a!T`f#WJ9X-8sdKEV^sco@G7;2^s!t9NHTf5n7eO`0lW5Q1D
zAnb?Q{(Vh;ph;6hq%;0R4?fG)hRDNb<4$<242uZEeYI<lVu*&GM<*7T`Oj(+@lCtf
zgl^_rY&aWUN9dI8>|&HzR~}^2wvCbLqZ}Sbrk+$!%5!8WAV`7e6hEEvaeG-qu26KE
zc57{2#_Kbb>M=iEFVB9UHrC+JZ=s%*yA6Hv{Ht>vp%tRFtBFIHH^z~l5W&&ds8vkn
zFeY{=tuB*6;5Yz=5#7Wa`RTQJrR6Y8bJ-(pG}P2=h{D`yLiMrUF_QEksE>qCreoAm
z?TYgh`5yqZl=a37WwegwGoZ%NtKY5Wo4VVjuq);#!82C(4Fs@P89R&xnKBkDvb~e8
zfg3aJt)qZp4@!Wg%yOlLwPVMY=7?B_soy29H1VL>->vs+zFcWs!}@4Q_+wR!hcq$y
z>!!f?g6fIuVY#q&`uuH7;9HgP&0kj?C^uBX`kzuEKCQ`TH2JBTwy4HhWz{XbNFn@3
zn5n$;JeRycnQ9%UZ{dpnFPsjNGm64}2uwtFid!v9h}EnT+c77~hZfsEQr3vVo-%U5
z(qY4S?ITB+++#f!(@t@5MDu1995xWcW~<pST|scREN3f&%SaB<@t3J^?DLcau~~Wx
z8O-7=LFNvW3A<{joE9om5*%IrXG2m)G(W~5o*KYa)=cxB*qG+M`(p5V%|zu?^Q(`Q
z<B%=7P}w(=pUqDkKagK2@N<xSu`axiR<Sr+$S<%iSxX*2!2g4e34TsB$mrnk1w*qj
z5c?cV7V|70R%~##eN6U=oY4kr;om;*b4r&wFmDLKFYC)|fnd3#fL?up8i$BMZ=G?7
ze3Qy7r?IyRW~`v`#seqL^SQK#;RZhVj;Ps-bHM<FLoOI&f4J;<hruXJy-xaF){&1m
z-!pnuZNRUC@6yaLLkNihy~25l9|OKl*L*XGPuPe5e_CZY%jIq4oqxYBX++N2#EW8c
zKjPS5>TNiE3mP59eP-9w8B@9J0X4OU|9Rc(A$KL6zm_JF`|QHx%8DBvX?ie?$Y&`1
z$88f%P;%TW^fuLasdky$-^TS$qBRW1wVDlC`ALf_WJh+^bG;*%k`hziQgOx?Zuv4{
zDIdfNLF8ALN<M{7_<zwvEnaQVE^xA_QTdZ=<sy4XuwNT*DUWRhi%4XImW3?IGvreB
zY_K93@RsN1r$w#&jE+t(dCRuH-=`}X@Tdo4VG*3VP29*AiWreQI-nzPMXxP!ezYsT
z9&BwqzP5f<iW)0}mRy#rQ4Tpe8XST@d~$lNyc#6^a#0bap{VU?JZYe+39qeZ;BQyC
zc4)Fw=iDAUXI+LfY0(FEX}e1xF^SC;u<<X~Dwo@0@<b3l!1u>=wtuPe6*PA(liI-T
z;mf|m?%Y41O&WA)<9tjbW%g;a4p>Lb)vE{)ZBD3IOTX9M!uvQpQJP`WnMkLdQ(FU>
z2t7+n;M@m&9TT(=TBm_c{3TU?71$DoK-WlT;sN)Udrvs$eur+#U6ptsm*M|F?%>FE
F{{xtAaghK3

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_22_15.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_22_15.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..7b952c4222e813394bbb9b9b3a4d24c424684587
GIT binary patch
literal 59703
zcmeFa33O!FdEZ%k6$*tNjYgxl>b=qIovUbyYqPbGY|>;)luM$D0A2wNG*CdlDm2m1
zgDtZuiMB<GR_yUcYS4~jj}phpI8H`UyvP}klXx^fIrfa_%sW}+C6?(pj=Uuv$9CrT
z|L&_-g@u$XJ13cQ(qvuSci(;Y-FKI7|GxWhPfszyzd!lN$7=I0CKJD|i`Ks^nFskb
zzL-oTyhJ78CB4*qvXZpFsfBbUopfcHO4iD9l{|mb^M!@3N|)Ws%oi8BE8PpFO3ALX
z^F0f_mEMKEN}pZl=KB{0Dg#kFgOx#^&(9An3{{4#tS~>ku%oiW@?G;I3p*=2Enl1;
zUD#FGW%=&;v4!20-3xmvdlvRq_AczJ>|5Ah*}qV(lot+E4lEq39JG3+`9lkbD~B!L
zGk;{^Xyxd_vC1*K?wvorFkTs7I8ixa*M0N1Eu5^JTsT!ZwQ#y}+RFOp&n%p+oVEPG
z{JDkmmGhP#oWFhHj>;Y6hrD5Lhd1(ix^ky?(i`=5y_Brn<&Al}UrJQ&epkZV<L!MZ
z;q9&E=kBSUznQ7LgR*_ze#-XO-f=TsxtHs*cYx~yszb>)P;$^aM9HDry<FeN^<nP_
z*GKI7om?OFj&Xgg_D=5K&-HO{oa=GBet_!}-fdjpX4emTsb>=7rzVz|(&7uV^Imyj
zcBQ%O*UF32#VgI`Qsdq;XJ%%bSC%JFPc1H-nO~lo_56w1dgF{&t1r$rX6rMjyxHmL
z<;Lt{{nXNp=Au7!<;>*#;^djT?>>9iq&Gcv*Yus|@1DN%9`E+sr|-FA^6tqy&!4Z|
zb*^?-?e=qb-7$H4yQON}Qnl4m{FbfvPA}cqu8WIHwfe;Dnc0OUfAM<FYxmCes`idM
z?zrb2=kK|Fa`K+L@0yr8cgNj#dv|&sy?w{z<fM1rd&fJ@o}a444YXf;W@ci!(U_g6
zw~OAQ505M^EWLlWai%%3paE+76ZOXQqQ6k{8^u$n%KV+0ExxIZb!j{q79O8inDizd
ze|kI>cKNjlpXxKUFz45nY7<Qgd#4ub&Du(HVWK`UQ}e^_`ttmIRL1l9r3t?|J26i-
z)4b*_@V9GW;%cor<Ij3r^wQqd>T1njtS;8)Z%|Tve74biWU)RyJ2PGgd*>Hts?C*V
zb)r$7U6=`TwH0nR!n{|To>-o5hS}N1!o*UTXSx^Y0Ik;;gz5Q-W>}h9T&_2(^k8D4
z5$<GRG$-nGrdg{_`-=-zda0Kr!`$Pu%|{m|^oeywVq(5}bw1p4VRm9>u|6^Xu4rU0
z@M3><a=FQBqUzv#nE&Q%^Tx$lztMcAsh9UWwA@^*)u$G{n*Z42<LPkFzIC$JoTxUY
zuGGBc`C2%{X!#RU&FXBuxma!1>WxJ|ENa2ls`Il~Yjmjl!iD!d@yx`+(tOS5vj08$
zw8s}`7~m&nS7z&!^);3zrfSvZ>_V;4tSvRd!jwK}itp|w=Px!I6U}BlOx~ayx#&(;
zyF5MNEnJ>TX`cC~zihL_Bm5c}k|4p}ORi2ndA;Vp;LkQ|<pn-!{xm(Gt(RY5*Oi-B
zW*g<{<@%Hcq}*IA>s!kWZBVOqrgnW|-X1HK`R6Ux8s++8vpjW0OQ<|izp=3B*G_*n
z!C+<@wfX5Vqj&g6c)cZmvUujo;zI2V57j0Y7q6c2CYlpxX8gtFr8Cbg*B@v3GN$!2
z8tyX}*ys<x=jk)^-ohE{Do~)c_s@9V(n9rG!v?fEJzK9em+LjJX7hB%oz?T_tLN@y
zyM(>oY-3__zUEc!3yx_tY>3kSIQc1!jQ;7bP>v1P9x_29NRB3&iMeEu@{&P%Hu-9D
zn6jN3#?+-S=hd#yPSyC*tU;kL!@BpzGhxa<$M1ZYZCsgHs`)2qt33ezaW2Dw<@&5P
zC>x~enmI+W>sQi0OMZ%9{BK{BYrMvU))Jt=M1?+iDKG71W)fc3%XxXP;B|RLubW)S
z>+yQMjMvA1{rnx^m*u}cewLfw_ld-G(i`-KUh1l(y{CjYKo%MA8@&;4C%J4b=RNIE
z$jxLWUn{J4-AdGoH+k1|%1}y@@@|JxDkX<TDm@N;RC*oKsPs8xQR#O`qB7u+LuIfw
z<ehju>D}g?d?`~IcCCBQ06b5>q;_{iC6~Q3-dQUdiAvrFz`Wf`c6#sk?(puUucO{w
z-rb<MUEZ_aJ>EMg8S_5i-Rpe=x!vA<-aEOs$GhKqfWLda2fc^*yU%;rdxXFHqjoQN
zkG_=P&1LVR_b$p0c<=V!!{39^{crLf_nxr#9`Y`EPjct*Cs;C-Bdmo=n7MG_sVCn2
ztu!;6n`bFb_+e`P{H2Sl1-8qXdG_sPfSNY^o7$3ZQZezSNAl%UkL>xkzjoKy17H5g
z@Q=OgS69F5fiM5*XMgXPo|ydT12e7Jk1}6*AWQ=jrpi&d)=`0<+G@PYif$%uCf1X;
zctNn!OJ2=2#+oTFbu;;5GT0d;s_AAXNX%vDazBvp(&W<o&AgGhnS^j)tyl9wFKHp@
zA?*r^q{X0{v^yw~mV#ZRJ<Z-A8I0Uak?(8v(~4G9?v)hp?4oRdk^%oGnuEG?GfBx%
zGvno(!_6I(46TpcVvKJld2`{7wB8(ydWoxgk{-_O^nW23B|jPrlkN(JNXG)6tnLm9
zs^|Z9P~d7$u!nSSu<x~`pK0z3vNw~jB^#M$iZ9Ko9_1HkZ-0;tazWnfdIdt!YO8dB
zx@T3Fo)y)H&fW(5dG-o-x~+Gv_J^%y^}lSRo0vNg3<l{S74!xD!GKqKCC3={td*Mw
zn};YZ&m?a$?rR5{hv$wokMivY^wnG)^Lk&=*d6!!)(*}j)(-JIyz+2!oPM3SnY@sA
z?v4+o){d+l4UPl{uIA{=&jp9pZ}a+tqrt)5iP6L@joHz)gTdkT+kzvydM$bF=blZl
z!v|kV1cxp#zH5hQ?PT-t`jOz!tz>gJICL}V4cQ#A0;;EiG1Ak`GjnHy-PAk>92kBj
zwRVjB>DN;JhnnZ-ZYO^SGrPljdq<rWuzsgE5*)jwncn+aYRPZjbyKV6Zff4mQ{T?l
z-Lw9VTPnNP6NKD(X*})U$KniAD`&&x_0_Xe%f1hKoxf2QSpZg9T3UpVsLzxSvak<=
zJ-qV4m9tUi`oYznd&^HR*XybQf?DlAwdB{lS%;g14=#f<^B}Xy#YSzl|FK$aNmU<x
z;-LoQWI0Tp1~zxk&Vc`G)vL7|jj*T}R}Gb{R-djlYE656ZDB&VZb@j)ZkAuD&CXnD
zHcpo>3nB9Va-%lAJYRkR+*J0#&WpZYC6a?~mucXXeG|BA>gv)WRKV%g3+3C||H{uk
z^z^0oTzVH@^yo8>UcOAWeCd5pROxK_eNSC@=<=h_lrKM7t~~nmlN%lz&xPGHv)60&
zstAro*t3!D=xWmnB?ow_5JY}6EJjzgx(5}p2;E$(Hr^DHVa&2#^TtvwhAVH1N|;ij
zf5MHzh0pM7jFHSRr_T+pC4=Og!I$;at;Bbxu4SHaSPfDnBP1TCr<Rt(Y<+oQ=?3GT
zI(;_GsI!|;-=7rkn}_IvxL>;eJM;>@FPB(7AP{=G3IR3StX5CoH@`SFG2ghKOsn>I
z%728q{-Y$kpg27{KVJpDF4rJ6W>-aH_4u`G%d^Z6r09ePEMA<sKH-ml1E_czOdJ+k
z?T5V&H5xS*-(vky--lM~H;S^>togI66Dkg~OH6Sy?5b9$<|i7BYE@6yCl+et`k}o%
zjr2^-t){SNYJPSJRBj}n7j}!pt+6z3RGVSnw5Z9bI_oej8Y-x&(@YMeQ@u9d2uE3C
z)g`7%uZLbWnv+%52>V3{8OQUiKe40_$=WjY2N{Ymujg2S{;;mP?2fL(^n9%zW;A|b
zesXzw8g3x$QRNod%4f&fh;k0Qud5Fl2S#O>Wj-@&otd9pth!wh_RZ6E(Xp+9{K6u$
zM*VIhjhVpd*{0v4?)2JxI@0{=QZpQ2a%#;gtE9TT#ImV1R5U{$Yku6XFijm_L{FHi
zFNIlgC^y0!G<CD)fk^Ak+s}u6@zNL3&6sr6Xtn0kX4pG1zjOueVv5D+FR~CcYV3ra
zTVB<ARyew<T@jY%LErVNTRQ%2>O_XEKOgp})A38`Sv~A}N_>b`jsHGX?2FrX<IT<h
zT!p3ODHtA6*-R&V7JzPXA?$wI{=#L@<ASOz`SqFc<Ni^as&qrmHls<gCGTIL*#8#2
zrVAJ&9;kW6e@Jy<s3tt)wW@l`UsBC!CC@9FP%^2cs^r5;E-HDKl6Nbyja64}LCGSC
zTjaJF>9yde|4q6#uVhNeM@cGL>FhDvxBfL1e}Y7OrRG`7a7TN}xol0lzpAHgImTmJ
z8Dyuy7HaZZb64rA*IsbIuTpB&tcpEM&Q^LhRJA!6Px>EJFO1_5jpRQ|ay!39K$196
zO!g;p$uWMN|CN%xc3n#DOBIz%+1*sZ)k^dx-E*a6Az7kKX_1<xREGA(k{SDNm|BD6
z`e|)6r8WksIh-n`dilHBvjN6U<<%el_41P#Oq0;~Nu%Tf=mj5IYb{?NY{Gv)ZDd5v
zc*$3?Yk4mf<VO>0g=T6l4P^um0RHZ3rq(mu>GGh_KbG2+xRqKf2Kl+{hmvdQW-fr<
z0ypPrM^s*xv@7VE%X;Zu!1uZQ)l}o7&7zkRz1Zw-mgai+^?G?)h6-Ho^9oRd`I||j
z0Ce5Obyvr=s72_h_<ncLO^^D6u3LOdTv~Fa>ZhogjypYeXJDH<z3pdqZ+xb&^N!cw
zZV`$pF*i7x2=c=TS}cwx>@V-!t*;$a`Od_e=$bxnfS#8+dNSyGLhpxyk~efKwV|KG
z8VCEb;w|_=*WcaHUGX>W2)f-ntyW>XT64qR$kh~d6O``U4sYjB!V?8N0)N>Z6mF_r
zeZfxkdv{{3r|xq#?2Xzp7rb3+xmj8t)#zILW34NFA@v`&`n#?EH**i_{j0y{R+1+R
z%`tC}aSc6QKWn4MOW&t4>+|-KM_edsnifSn?zxp*-+L>;cYLl{d=ah`y00v{ZzFh#
zDGgsl26#F%fyC#{|IEMD+Z?DaUzxa$jA5}nSt}z3f?V;+Fv#WF#1vAANFc)AelHD#
zS)+v*67oYEK|L_Ec0C+wT=OG1(nbP%t#A=S2qD>W!-1Lh?L9)-Z+Ul@J?z(}!15d4
z818Q0S~$1qrvI4wI2dD8=oKG<-c&8z*DlkGTjdZNv*0Rk=Fr2RF5P<lpL|0&+^%S1
zu&Kq`^to+I&l|b@rU)wkog~q)e<@}qAN231c(s3F*@QL`0S-zn&0n9NjIep*A`?`E
z$Y=;D*PG+}43qo2)a-SVFdv~+e@#VaRFnn<`!DLUJ@X&XrC^*xqPa*amY)V&fx!Kb
z>A{Q!4@zZm#xPHqUtno0EH5}*^JzW(eM)SubC;c9t&3@Dm|QI!y0qAQk7$tvx?1xb
zF^BxHK2i5SqZgzn=I6(I4W~LRXJ}M3WJgyG2kK=|3Pw5D`~1JAl7a+!k}|uFsL@R%
zt>2{V2lzFfCP@fqfEfz>*9+_~fD?pEMpON%elSCSdIWr-|MsPax!<4KN6j&Q`%=O#
zWiZQOex>9=o-Ogj>WJZun6_xqZ2ujBYyiyv9fAC%5lD0AEkvIG?IRHXHTIeRog`lu
z{^&OR5t07G8^O6Nt?b%VDzwtNW$=E}ePcl?hwRo{A(80$gKvjuqE78^<CIMpC8}_s
zU11B-aaA1Z8Es>a7{xet>2Ukbb_mEd(ndpL?dokb6m@31d*`EGNCM=)q2w=yk_J0b
zl4}dZ<o}Pfv;`yi>uNv7M*b}>D?=ODHP-?RhDm3gkR&SWZ;MKzK5xWa{tv44zpg|m
z%l{!Ie?y7!N<XZeaYjF;+>a~yJ4${;$^WF}vr7J^l7FD&Z;^cMD9Qg(JtO>Q_{jg;
z${9}F4i$YfcYm4Re<@V7XM0rS25y7uNgR#XL}R<46%+n_q{(&UpL6N;%q=I&R0+Ju
zT-qd?P>@J@?}J;Bj5zrQQq*K(wdf_@k5n5er|3|K5NjvDp&j?+_2|izd-4N30Y#d=
zw7Tow@-ry7q^=W@DXCi|;#=K~NX`U$x)t&0t()To$3}~iLWTqxQ1ffb7h`czqmrLm
zTtdvV=vQ<n54Wp&NUM(8HKHQvQa8~INGsrl8$4R;hA3`~P&Z-xqBzCgWD4%magVFL
zHso>jDGBHFZwJ2uKVut@r=<WuRb|!@Fl`wx-Ao0U^|VCpS9AW+fRbH_HN@nt(zWc$
zJwX<s^VzlB%84LF?%-NJ$gO9B{H>(_a1&+GdM?Nb%Odv81i5!6u08y0;@UmW8p{aT
zzuqiJ0q|0y*|m<!V7*%c;Pn!dOKTE(uN7WPt>l_LTkm^$Sl7ZU=?jRkpZhh$-d(V(
zT|wq*-v8pY&+)ddU5U9~n9m|<sR{du2)!Ui4OaCl%|6&vyYjlOW)PI`fmMU`72^y0
z73E-ml8rs8ixQ)o=ROw{*ZXh5Ai4I}dZRXqs_8dZPBsUE9@yUQUG$#u7+dQ@EdraV
zl97bip`h<R3EiK5mJ}6fs}EU5<GWVgzv=l4Yke#CMP<dP%@b?=oo%8*==Qibl9*QC
zX+P+nPT4ztSns&@>1UbQb(m7sf<+rxIlHOFeY7~xZYN7Wz2e<yhOCu8w&%sly5|Su
zuWa>bY~%CvVle3E*_6%nAOZ#c`&df-TN_+C%r_6t^~SRy_V3$+!S!LUMDGT9>I&<l
zCm7tTaki1ZFLCX2l%-!wHGYC^yMwieg7TGA9hKAENc);qFtj$zy6ZCl(9fHOOkFfU
zZrE~zZy<z#0ku0`H*=r%3p_Ct4E;dDZF^TYx_6j+!@+Rto&kg%m;5K#IPf_o1WAbc
z(6CG7TSTIZA%SFetK|!|`I>15kbIRtvgTy{?Fy%3fjK*`Eiji1)&%Ayz9qSq3X&b7
za)SYI$4^oz%t)3TCaaqOga3`(JSvbNJ`Tb7G28wd#i-kf$5#(toOSAqMa*g<=q6x?
z{H%3}*gWIa8dK6(Hz(#DUKWdLjJXk1;{I|nieL<SW58&Nj@6tq5?JlQY6rs<6d+RB
ztoE0YI<73$reMIz(zC4gJrUi1jH`Q#<<+6$``T5a!ttGvbXC$zf7ZrCPun#3&4vsT
z98AqgRX)*l;>9bf*oF0i-&|ZoY@S7cK3fm_HJqL^43R(<QMW6JB;US3r7^3%&^wRK
zL#KsVgt0Ry5X4c2S=p>K!mj0d<JxkqhIGEqB&+RNH|RgU!-4ipEP?eOR+FiQ=l`5C
z$*W;%_G(zDM|LXyPum*}XH>eFo~l^}V3~qwvm<L2LaSg||5Mi5T%}mABDY7xJMYDL
zo`qX>;~W;-Gq8C*`Lr>n^)yE&fgrLU1cUI>{_6$Rh*5RIO);&h<REM(=m(Z{IN6)p
z$5q72CijV<<-fDEG8|ov!PJhZO@y3d+&|1u|E&&fvu>xfSM)E(Pk795mj<sXj3Pm2
zaxQf<9UuThXbA}*B##&`DhGS1UM9ZII>G348Z@7p*vDs@T}qjM@2WL1XF-~T7^7w4
zQ@(b5rP?e9<(2B$@zZhDdu_BSWTO4%z4q7#FO^rykMKKv`t);UC%ceV(G)!MwFX+j
z`qDBg!p*Hj&qIAq)_gq?o2-=0sHc3~#82gO<EP8-hMYKAK8t}C23u%FF}1<&r=v+9
zs}{e0Z~38ea~UBA&r<VbxkbvKEO(Ob<0s1qJM_xvOO7?lwz}F4Mk1lFJTRX1rQ7qr
zOUW4}Mq&OB<i;Tu_!H)%R>EWxD<0k(V$I;BSCvxddpE7Pu<u&cScF#XZ_?`1_Dkfo
zdJt#~+uGPvEhcQVFtYq?aU3VJb#YjUH1KY@q=(Usvjmct#xw4_zK6%$cYQB8O6`-<
z%)bAxC_qms+pNt!AB_~u5r1H6E2~)3Cm7+C8ceS4weG)F-M`ZB_`C6U{Lk(0@PAFE
z`VRlsl^lr^jTWLsrzYR>6^=^3y5U<6z3sO|biI8!P5)KQOnb$BH!Vgp)A2?3#~)*#
zWnVP5%@?g!K|a;%*W$0qFf;q#>T9CfU-R6sqBgBDK$2O0wb$6!^Cj^L!{WhNEJi)M
z5r4<l6}Gu<{1uX}#ydsA$7rXpGg@FY`ifRb%Jd#BFdCJvWv!NgsXe!rgFHstWMGx=
z&AB}#5vSWzuPbLf==YNw&)e?u|D{Ufr5W~alo+(@Z(gt6*V-!QZ_zM_bxbe}BPK9i
z?T?^E+}?ks@BayXpTWw%ubgrEe~;Yu-|zpHD#e4e`J)FqKHBZFm$&)!k<Fj(sv-#a
zFs;Yq(8Wc(4yL()3Byt`NTCBn-!}(Scs1o8CkMm?YN986CB2q4@*9JB?%fq+)|h|)
zVvt=&pXg<sHVagn)f3owr9YHj%dO?t3bLsR3br=#mrUgVuiFMj9b%rZ{7f6)q&F=O
z5Kv76WRjELAcsNgJjR?=|4lB&yBxqyuDF)<wJp<z#bxXnR;uV?M6x?`2jLqV>u@Aq
zB~szXdX>$?-?y|8M}4OTe~`B%j5ZpC{*g+zlp1Mat4$9~X}tBX#4jFuAyL`M@Jmi7
zzZa(R!^yR*#w{NEG&)?3Wr5PJCZrCv(aPRTuNBc6V$pdOJ?<^=*<3crVw;tbOk)mq
z)JsVl_?gtJ37BHoVJ{nW4JX#~w<ryY5p>E)`}RBZqCIXV2#hpumX9XoOAjF4u;qM0
z^?y&vZjvx-8XA*im;vh*6{ex^$4hRbZ)`s-Im7X)VahN$>HmzHvPS>C9vyUZv|&1g
zx583=v2JGK3loj2n`fx7RCUI^Z?yNt2&86j>O@PE-l`Fkl?7VWUNFkxhsZGMto}7d
z*-JX%2O4v1d0GD{)_Zm)iRh!=%rvu2bnDd4%W}``BlBq05l%dJ$>t7?Iofu_6Vpj%
z3(?A4(aXM~z16L~L0g*P(g-9gVup{~Ro+8O&)#|%kxEaHm4r*;lK`=aL04QMf-F=^
z;ZkdMG-j(sGZGhfg)P4S1&aKiB_VG^`d;PT{<qmglK5%Wv72VDBA&*p8^aS8kXyqH
z)ja=;s`R3g|5wSkl3*OPSf6nQA!ZsP5h{y9w34liaM%%`4(h;P$<V&xUiBZ;Kw9?u
zJtk0&?|w>8ABht~;61IWVpA?I{jg?qlp4~NN!pc59z+9HOzt6{p=5Q?;qhn%*h6pj
zQ$L<*?YMZ?9ppkY%{*(iH7kO18Sq(p8fJkVC%2On&JX^dQI{<;!5?Q6BZFLkk<o|6
z$q8hHW|(CaJF0E^pqyKqe_vmsRVUe9PHYcm{@Ag(yD@xj(M@S8Htr<>-=@t1n_c}<
z;bsbEN2`zh|1u2E&GgIcX)B@kt|WjWG5RBYX-e;F{Zo@e{2E0P<V4Arkr6pkIOX{3
zqAdqgm)f*22!^Tci)?lZ_W^nF2)X~K_?$mb%Ml@UEt-H%d4sFTh{<eaGZWEC(%DPe
zctNjgz1qFC)%s6r5OaDh<}wg*&P=9tcgrB!=2i<vw6K{0kK!32ZLEYzk4hlfHn9cU
zlWYdhew`K_(A+fD*%=I(uK%n$sd*G(vRbrRwi|K(vfR_p`sa}FoPh)Xuo#t5GoEB|
zru|7IuozLUW$O29*?K8}M1}GT&GcMmE{n{S`a<2;CSA##m6)MVnx_iQyob><Qca97
z$8(o3Hb3Xz!K)D}!9*g#!TicwHbdFf{mb>M^~D$Ju!=CPGI1*(`$%l=wswsD{w;bm
zJu%tv|1~*VHU3SN6<e+#6Z2tVLB1@sVd{+5j>(8J4|DL%42C@}phBfhgILg5>Bc|N
zI1bZzg8g1%w@VO_LFh1&Dx{*RZ`F;5wnTLeZJJ-BObc80k9Pjc5)`DpH1HT0fTn@n
z!!TxN6Hv75D=%ZvIE_`6bP;JU%k|SQC%8Uo*Ez1|c``#wET~(mIfww!bPX-5b%s5`
zgaGm7Z}CccE|1Rz9v3rdM4AY^xm)mm*z2pOy#BQ=U_W+NJUj44DkunB&tkEIINkEp
z1#=_#29E?1{}voZ2gA3X8|1m}=o!^RkZm+PbXe5#kXmlF3e_%HH~bfb5`KnJ-5xyn
zS#JH`x{aB83o=M*iUn<B8^{n*dhMNmAG2oKKWBb2@KnoaGBUL-uMV0$$_*qou}zAU
zeYP4+Jo$#2xSNEvo%UuI5MWAG6Q;5MT`if1MP!q@+U2e=Xj#o#cD%=ds=Pj-riP5>
zEa>vHx*Swa@50t2Hqvy~sb&djh9thMBU=Q}M$jf|RJ6|x|KF*RE{%){k^O$PJU~(@
ziuz%%OwX>!WReQbgtk(!3wpa@@Ew3`4f3zhLPJFYko{(JDDj;8^`{0A()FfBp^OHx
zC+STMxAsx%`B*QfJtRQ;B)`U8BvKg}-2=d3KR_cv-ymRvKFX|T1){iu{)FxUH+_cP
z^{S9ddIQA*|6_DVUWg_(1Zq>3f69w|i6A4OVe~0BU_PeJxvMYe@3w$9kJy*H-XH1n
z4fLiDCRa-adZSx0J)qly4hDShCnK%}iVq=Tke)*kWgO5xQGQlxa6)*OyLiT62bp;#
zxt`;TDd7vVm)vqooxZd>I9a|@exW>7K6QV2l0RIq40{%8_`2PwI@vU1a7CYKU#*UW
zDYSq#ECbR2&dw|^VsjP=vVWfnEtyo-XbxXWTSvHWf}CA~h9+zLPZ|e-!h{h`lG=ib
zoO&>>v|+9_DmEgHLR#0TFyp{Pw6Fn`E=mf)N~}u_YQXEK{6oCPf1W2>({)e<h9k9<
zeY~Ex(0AR)89@X{oN7zy6y0>_ycY-`vF{xfVzA!LOA4(n{sx(qCf!RwbkEtsL+6Ml
zON6^r#O_MdhfWejalyaCgC#}3mPDt?5|w_<w06ghWCxdukqxrQkFzskVN<ZTwwM95
zd2~`2(Pb7O$P=ibp(Adk7>)uUTT8wRH#+xU;S~TNWq^CqLR4BrcL_YCq|5&jEfjBw
z6+lbqmFS@Wb>8a<ikoT|SX#Z8!sL>_gHa44%Lsuuz}*pwF5^E~z-E1>I*AjR92n$)
zo5xL1LmL*z)Rq>fuHY4d&of>ED!yp^15|)t$<6y!5IKPA6Ey0Gep{H|&Lu4Nl|@_@
zeVot%b^=M#bva{(+cd3LRjatTgvBM!2`EuYCK|_`0td~;c(-E>v>?JRM=Q^ID`8Lk
zK$T0kl=U(8k*35jAm~=kN!{4A#(P?D*FLA91*8=#UfgLYGBAFlW-N#90^_<~Q;v_M
zB1U1u<1tLo8WcCE?b!Gt3Dd+5@L2!Rx_8&m2zJdPTgQLyQ}5NZcGmp1)^$|#+mc4R
zLr=U^c$wvTGx=?aPr&b^3Rg+eN|Lfy_{PN3i4X8Jlzqm0G|f*lT2Lvx)pWg~?B&EZ
zWyy_Yro`-WFxYs&SQp928KGnxfdu@syj;#Z5H#!~J{~6GubJ}>AS`2UBbp1gy>LR9
zMghkS)Lsz^Q;q6bH($a)xbJAmnETE#&6D1Umh!kkWw7xqP95k{Z6Z5}gZQDZud6=c
z4I=Wy_0+$Uf~`G0*Y2r*H#Ls8yLgeU<sVZs$CZpLA=rYLL2U>BAMqhRkrrC|58uGv
zIo<jhZecnUk4b&e4a+uVt^*9`U3%~yB^tJ{Z`06})iG!EA)`>~-}-^(drZGJ+>$wM
zdMF+{t!VuWkE<0{7<7!*uhyFAOIstr^K{#S#VIe-ZV&HK*Y+yW{I`cK9AxrrKLLTn
z4C6O*TAein4Arr^)24cBtr+vDW{i1GasdvODM$*AmfaU|yI^9=fr%ycLLTEVZ|2hK
zNWFz;9m=gcq`NvgS&p(QKhD*-yQ8&S_ZMH_YYo_GY<z9~7Lv_7hP)B>))jJ~QN_Nw
z$^&63a@n%j1)GOrs4BEdwMbMiSE}<G3-Q`Q!s1=PaHBRN+f;tDppJ-moic<-rjMU`
zCfwJkGyZbhH2}ll*yN?w@`wfVU@z@h4N|J(I&DNt-i)TO%)uloC7Ty8ULMeqx-Zep
z-^4Wp<^;mW?sToJU=abHGTb-^ZW4AXMxv@$?W)T|%L?;Tm+J?@s^SbxiU(nUkPE>i
zRx`rDyl=f15~)a?KDi#qHDnGK587~Iuo7&|K!6<xqOQ7O*1-S=2zmy-4P>WtnngO~
zQTC}(!=A%TzSt;%;GVVK4xrh?Q{8es!1krr+9+|ArH#4fj`fiy-@Cq3)=h}kQf{<F
z<Z=H4FQW`J5RH||ONj0Ske^AYbiB`j%`oTGw+22}bLVWKfrBO0Vij#7@Oxq%u2w^(
z@YMXGnOa<Cqo=fo{*1@``^bgetp!EJc)zk{Y)2Sa)_ygx+}j4r4kWkNUj&aMDCZEg
z?Nb30gKQttI#jy=U#G(pIyc9{PNpn|yW^w&$Mx(Xb@zF4VH!QOqgTaIJE7jJst8p&
zNnuqvZDl7D_(A22nm2CI7^*duJ)i~*fToqZP8p83^2%#q6>`p*MU3}#P!GvVs9-gi
zzMcrXBRQk&KO9stdh=nmws9_emW>-~{}JU3dnMfn8KgB%%Z<Z3G)9t(NIECOhkn3o
zA)SMS=m*rwY(apx7oI7H?fo94ks>V0$&6$?eYUIHqdLXw?B53XCTSj<AhDKo5<F*c
zC;EC0r8Dvf<oGkhj9?Sz=JIm|iFPE9jG&##Bt;U9cZEIXc_?Q)IWd|=jN?~7$2<Iw
zDtVuh+epH!(OhBnA#Y;g*_IYn4TU))*&1P*8toP3(9kfv2=P&HbgsrUTcfIBD$!`d
zM3@W@gR;8IK7YIVNSyrJ)sgc`-l0UJ;C}<j+ujp<aT&(`BFh^Hksymm7B6n8?H{DT
zK;`92<464u(N<Wj%Da~MD!AO1s?}@D6Z6gi*L?cS?ZW?Ny*aDI)}M_{9t151U^j%m
zZvLv0-%#>zl>DZWFDrRd$#0Y3t&Ba9Q-hd<JfRNRdbWjb@&XeTh?VtaYvMMOQ4hE4
z@*PU<Rbm_BJ<2^v681z+#8Wj~mOCuI05~7i`)yaBBd1^EC6YoWg%Qrz_h0{ekgeUV
zUn*JZ#ftR)LaC7BzXHFmLRW8zDK7BiUMW}17IMYD;(oiIEsXZ}7BYoAcaHe#%@n`*
zpOBaSHGZD2YR4fVf=I={Eip)lCz26BCqf1>L~8l<!Yv1ik&z+kh7PE7c^|aeF8WB(
zVvpcRYLQ10m6Amr!8O$Tkhk00Lky2zZ?CtH_WHd2UYWoB-U06*e+Rrn-eLX@dPlsY
z{2lV1v-l!OB8u31y${=a6o3lkM^q74S|vEvDHm15d)_<ko#7ojqkCt)b1#7`h$*5t
zBZ?`a7$XTHicDCP5k(YHj1ff?QG^ji6mbzo_IZ;QV<cJG?@eKL^x*4B?;-Eum$DUt
zX=xNZ&r_5UjljXEq-N1ZRB|XPnYOqiDmfgL%y^G^kK0p6qLL@POV-Du(e)MYN$)AE
zaV#pC^)7qwvy$UPNqLrDjz@Ps;C+)ld%~iZ+y+9N3;P}i+|-+~llf|KK>>lx?fsH}
zC-Z&rvi1u9zVg6SkIR|TK}3z#MB~$}2a6LTtp$XOMF>G|iI=H#GO3_fa+j;g#))PI
z`6=ePu<!Q=dH-H7jhr+C)15MD>{EiJji<Z>CB5t7kL1US5iWUF7kP|l$#+?P!1Bcz
z)fsH&Zo+iqqljPXY%=Kmp46+7vm(32-xc{U{;m)))#8oOW;fC<JX~eO>LHk?wC~wO
zoz1QPJiJ1yeczO*r{LVW8(*?mDE)O&*8^Tpa}cjkJU^X&e0>-}z21^o-+{c?<wg)C
zD5qE@JMo((qKPcmtGg^e7W73iO6GWKH|D;BYRgmfm9d~-<PBr62gBv}Ypj}kJtC=K
z@VbxE{pQ144v1mmzsj9{-HB?yWFwoHE3Y3=loF%=2QC?X<NU`DGEx2^l5m(X2n}4t
z0q$Na+CbI+4&8r+<f8wovJWZwq!K0<Wo?6i3`E%?%GIKO`TsBBGY*D3k+~66OCikU
zPVYV#Bx^$9aB!oSh$^(cbn*sFtmmvn(@Tp>{`)l+@+K!5mqq2W*bkv_YAYCoF;0p$
zWiixRrBF4FFL{EItQ0(tC&0pZ4qT}&5bz8icR~wzC}|r02dLuwOoYnJ_heI;vvx3v
zaHKQ-S-TtIXlHp86Uzkl;b6O#gpm@jn*O)aJ2|<Cb=>0BT3wQ#a7ca(SVJPzi=#LA
zhDxW^uZj|>aU6vrlLEg%GCpKPUS+T|=0)_m5`?yfBr1YfYv2T-(eNw!@~V;#D|t?d
zXfC$~ZCpO6OPLe;ZT2-@w(V8hLaS|xS+jOkWX@$W3E@3R9T~bxz=WAc^xp`j`Y(mB
zNp;XgLkT)lS-S&uD2`A0617}AYGrkF6A+C<nZ#rxuTSLpJ;Kl2bIe}@;AkTElmHPW
zGED$vRx#09$1H0-XT}J2myMpk1!xh8sc<Y@X9QU0@*hjx=Qe!CZFo4CXLyVMw37EK
zVa%l2w>9;xDtJ+eXnBQ)VOnNA?4ObHDP9Z?Q5K2LrUY8;sH&x@v6-B7q43<aMxpS2
zM58RckibB=oLoJ)d0Z?+$c9IzL{RIWMn{i|eQ(H{if_YliWD#ncd8sCVY${{My#Q7
zR2CRs+>5e~KnH-$oM)k}#0{?(Y^=IK5*akFMQV4$yGlh^3Nqv*_%j0y<8DC}&+DE-
zTcuF4_1^;U-@+?WV48NNO3bwU30m(laEm#V##x(ze_wfE^+5>vD_8+5L>>m*<*8+S
zgBBD(2w2*(p9jdCYL}oE0wPS*Yp4t6Z>;V<bm&m|ea|Rd&@-1WoFaJ8BM&{UE6NR4
ztnO_^F;cj_)15$C#{c7kzBUc<X<YCnCq4E}*z0k_!5dk`yDe=n?vQRgud=doC&-0^
zE(9WxUE)|a;i$wT9R+d78pXzIPT(^AF+GaMH(?8jMU&=JTgGw+)v}~(1n47rknHRc
z$w+)G&MzEruk+iMfnjoW%f2Q+%|>CBfvg|Z5*(qGMCQ0{a#Ut)TW$1FaW;oHFUjp1
zF%HK<y1h+|w%7w(D_61}Te;#mT`)G*oy1#aOnJ*`3$$)oT}BqA{j;pIOpLBm(eBMU
zRMVs#+k)Ya#udfBsT%@|!{qq03}0(J1yx*#m`9bnsN`Kr-c7P);EZ?rDUFkQok)+s
z{vz5tuz6&*`keEO{mbexetNG#)s~PiNJ)s!7fNcyeywI!?$F=$>@7s)&mDE5ca)B4
zJiRWm!`xp+shYcX6q%F#l^n>_aEb~#r4W)~pDg>t1QP!f_;DS9Zr+zfCCn%)WjN5e
z63yhy0zCS|T511`dcJ^v%Wxq_6T&(*-S}xGzoO(woET5`JkS2ThDpXQPU2Tc$%GwS
zg=3oss}t#V^d**yZWuD@DBp6(k}IP25Nbt()*BC+2IqJ*IH^mm5wLHREYn8HS!s;(
zf)K$~W6DRFQ=fop+cFCNn#SgvmAvH;Omg>^H3X5+Q=tuH6?y#aKEJak@%L|#oBR;n
z;~Nu=RmnpD>E$N2P!T<g)TElew$2^dre5HV^w4N&v7^Yk$bm*tQ%Ba3k}b3B^0w^K
zChSaCyH+0LnY0vhEW6@b8o-^dc3+!=(m!hNOgBE>EWwTFE$POmm2bDE9z;Mr95i~l
zV@j*l)g$jS2PiAVYlBE&IA1gt%M*>Vkee$y?LW)tt#(~*`IWBjRs|x)Z6Lm*III`3
zR4~%Vh^s)e@*Jb&L<ACu*lPDbq8plBx3Wz*q?K)yYD*Gi6KiXGyJbB=6aPYUTIM-P
zeVh*vp^6ZXNvvI9iI;#03P`bf^zCNaVRC)Hf0R~ZTrRjGsA%wEo(!Q5Gk7kd3UCH_
zOiCHdn{g_OAiFMT0XBBVnPQ->4dS^-4W>8O%&lW2E3D1*;>n5>JWnnUmRGI|HZO;)
zntYH4ZeRwR^uNGc(F_bm48URpJRons9u0&44Se2s(fN@{>KP8YRo;mK+X&MU{Gzwy
zI6N_R)fcpMN{PtA@;*)E7bvalZFP|da&8M7n2YRVOwd_Lt$O}D^~C*3Ho~cJ%#;ru
zqGw06OJVVQn&)BOk^uk7zD^!rj--WbJ+XOi-@2iAAL~Q3mu+e<kkO<z;rWSXgnu{0
zX{vtI%22^m%5I}XZk^HoVi%{hi_MZMFT*{i3@`27Tadn-nViYm>W*Wvk&p$Sa{*rn
zCtXxjaE6$uI=-;Jli3-i`y;yiED3?CF?=&(2ijQltba|zD0*$vSo-sn{h0>MiA0@_
zCzl#dZv}a8MgPP@MKf(Id7TS^JQIFga^hl%O7OHcOvzy3bxjQnz$I2dq>NnxVvUsl
zI0FlKTT5QUPYc5~yIVi(+EAM}Q;p}F1pt_A6eM(}!~i4Ujlos8VmG*~U~wd2p7?lB
zWxjzJcmO=d#JBpN(1?6YiH+qk4VB2lFwe*8ffhfIoZ2*SjvO=@HcX@7LMwP&ffW-O
z*B2qd%=r*^K{NK;f<`o^+ch$!Wv74Q5QIu?`MsZvtX`of5Xop{2#Sp1-a32~a7HG4
zm;l**l=thDxKf}^R8u~MxH&+7U?bu}sbMo>w1-jiPJ;5Cya$lNsQ{uNY+L6rV>92;
z>$9~NOiMxtOyZ|U?sLg2Ve$q40M9v!Hcb@6%_CMI(DxNZP2ZsrGEyvAPL`#8*uqb?
zs>b7>>hU;ylnZ^Q5E}w5sLavC3<4>>m=ZS*tQ;A?GY&zT7%%R-fu*vNc?I++I~$D!
zW0@HxNN`7aPUZX<nILE5L|`~-DAjWg>K0ais#uQxMLv~R?k!)jSgM<Eyl^kh#9s4C
zyS(M==)$!Fr3oet{L;P_Cu1s|^5miqEqF41s{8`s+OOb`TyIx?Cmxj(&8VumC=*o5
zZkM@r`cB<+txYLtlpi<druq$r(p;gX-P44Jg{!sN5>|%|$3{oXAlIQ3dsQ3euPM2p
z<WnRqJ9>re_P>)v)&eg`?X1wU><_<A*vL=oMoHhvqMvRO9T73Nu3Rn2geqNx1Cv)e
zeyp3oJ`^9kV$}`G9<3pRS@|;PdzhUD|Gj53fbV=H)=FzFX<yl%S|`I=F-#55oNFp*
zhclp+M%rW)L?XcU3htQoqcugqx-`2E?wf1&(lGb2W5(&BWqP|S;lFOJa+g-~)~XrS
zdfBK|R+3C^gS=YZ(heA2Vf*w}c4z)lrO#rnGWAj9vR1*wN>pI;St&MWYt>1gK=yuR
zD1JnR&Pl#i$Ybq7Sc~u0#Mw^I=&8zJV`-kyrlvP;Id<E;3l|#V-!kQ4>hT-=RvZi6
z<=zM1U+I24dbtFVn<lkK5HPNKLbE!`8=NszDR~qEJc>Yx%_rp9-pw<-;dvW_c%BDo
zM{{lS+)L94!V@3LIL9tT?x-habpQl0W6&`WHL6R+<p6R<<@}AxyP?EnyZ&1kGTC47
z5%t8|VbGBwY6k%-=Q7i%23{q!pj~CB2|KR0DsG7E6A;yxB&uIS04~dSXZ4`B$*O>w
z`CC~@aOPxhV9_{kCf9M50+JW7LCByG=!>p5Jv*G3*81yvEma@0p345<4_#lC`zpOP
zI+b3qmZM&gBlx}hKJ5hEF@sX&xT9|#*!-1IJ$-TB<)>48^T3vGcJI-5EBDpEUA1sX
zVyA5#gDq{vW0Az^PqURFYR8|u`BUy|9E<E&Yb?lZ7z^ETW8v<+)mW%rYb>nhwqv1M
zw&DWizWP|G7V4d`?Z#qDTdlFEXPVuRpuNN>VGT{om!h!xu^1b)q5|9tvX~Jes(8&<
z9W_ptr%slyoWu&aT{lKvR+%F?mMjLjdjyaD=qX?jv7XJr=!0jUJ4v*R_Vu~v+U>O|
z59~p^J8FdmZ+6~AE8!Rx4|8h*3M;^YR|~uGj3gw1b6*k`i)@%n&dTHClsJ@ilU$g4
zUwxW*5@F^dssH@~b`tbC4#s502HZqFBSbWkS&}*b2UPPjBw=cKI!t-|CL2UUp7LW%
z=YNmtd@o6*TknXVWtdxzUa3O=`_x8H3)|Z_RZ1Pz&u<2deFAJ&!L@U*K;wh-DUt0>
zk1D_-Xs<U_P8|mM9X2tI%;Gc2J;~FllcXoNDLawQq*nKD2SPjEG}WcI@K53<;Yr?{
zX#6-I(3bOo)8Im-+Vc-e!5JWxhcd$X1MCUT1QJ1;L)i<CA%n)E3N@mcoXd&|(~w9l
zDAyE@FlgMzK9!nuj_V?vvMjK=ucrMUrCbr6J`UambFP)H{axG9Isdz0w781<gk4np
z?oypME8y*1$^U$i@qf436O@|0@D0CcxjwjzkY4r6dTXuc+V8u3>v@&YLWzhz_Km<#
zj*5uCl0zeOu!8UL)S$Q%1XjIkea#`t`7(Oh&mJR`EQ;Vv&^Mc=wc{vxX~Fum&9jP;
z<YiySDvX`o9pGNGHb4ZSr2k3YK@eVgG#Ct^cI^rVMd_jlEpYcUu1xl8l;kM+F-k_9
zO<FL(9+!(OF-#{ga$c-g2;~k*Dft*7MWq7^(;Q8tg9A+eU4<*E%|qBL(hLH~>WceA
zh<pFXw95Yu36X5Xhd6VPSY+;B)qQc3{(_R4s(ez3X|PPw^}H@k=3^|EZP8h)rzfV!
zRd&eMcA2n5Y<tL6;>FBPb6C<2OiDUfEIZ!-(~=i?P6sMc)w76rkwVT{v;Hy-hGUVf
zRi`}5K@Q9#ogiY$N%9q)Ub-Rlm$4^k91&i3{2$_V6?{w;rKIXa_ie_SgCC~+@6d<D
zBUQ;vsj;+txHBubaBr1NOZiaO&X)2jk(RVVawyq+kC2xF&Pk|psnL|&$A*zq4O^}x
z11{<grZen%<;A?~3AvbxXewgYwB24eqAr$QiJ*%G9M5xs%kD)u@B~I25>*~;;&I3|
zdN`;7OjjktM<Zv)t6+kFRi=gHR)jK)!f&Q@c+MI|+|(!%fLhUQtoAzejd^S4FTH70
z<YT<7#bzcAY;Wg%GD9u?5yhos3`)$8&cNvSf5<Ior6&<$ScpVPwQx}IQ{p8Zyu=JQ
z{Tn8f3X9BT<e_3B5~*)guO&<rHo_AiiYhWCMc}t;S~n7|w)XzIX3m7=%t#3nJLFA7
z3)-p%+$uZ!uR8qxKhy{PC5GSsyc*az_WnOo*-t7F!tj4eIa6f+G`Y^v^?ydUwi(oL
z&(?veDl(QCxc!}Yv{~;RqkVGo8h^V+wii~3kz~1=NTKaetQBP<qDTtl+=4c(1Hrw2
zsGdpsl}K%!`>m?PEkX^=L_<Uez+Q}~c`Xq&fhZMXD4j|T<LFEzdXQ;h+qSjEh#jOS
zIxoVjt(aA5n3`$2E*n@H(ARA0{9(#|N}U&)PXJt4yt_9KXQwk$R`vMT%}=bA+2o5-
z(&(8|U+goCkXRZJzBC!IB|`jw0hnE@mmY)ZGM4zQY8C)Giu_+N)+nlFrUUX_BTyqC
z#vGL9WL?$e|03i#e;c1vWOm-sZR~eRwn^BYCgysmL4ZYtS#0%5WXW6~?0b=(^ap9F
zy)grq3AmmBv70(m!ghr4)A=BaM_|r80x=2IlsUT9Icgd?xy6<meDtC!-lpXLP;y2|
z8*9oY!QZW!sVQd&QvyhbCO@egD|T0jS!QOG`+ZdzBiC|L`o}8OJQxZhsv1XTxRtO+
zclMItMaoGMVDx6SIk9Lyh`8--mTQ`+=RekxkVG_*JeoWKbt#-Ae4-$oV>m@@ULuik
zO&lgfx)9oMocA=DCj62K%}e!@mdDb+u_s6ql0t(G-)y+%q?f`#P5~we&oUd`(?sBq
zi8ChB8K|5lIL7~QkZF7vEeL*a1eQQX5@}HVgS^e(ucWMm50{V2CW0d#Y(_(&##(yi
zQ4Q9=;*l^Tj2{*wWG^Hj9~}R8)#k=c6n1moO6#Ddf3A^p`iEjFh44ag<qCv72eYTP
zZmo_dHy{LUvq$)uA1vPtK#(bQE6h=V|ALj`OyWv+OQ~R%#bfH>_b9QY_Pyj<u=m$g
zs;{eb3%ylWkT?2U#B8JT+?6Fg))SE{h^zRk`Zgi%1ek8Cdf=@diy`g?v75Ah#u_79
zkgtr1@qro$a{1I@-9s#9mg8-Q#Rwvzw<J6(0$Yi4fr(AviUM@5cX>E%*ihsrq`)ec
zqkr#;!sBtu5yvb%_BbK#`~^F5K|%8@belVeh`0erEl*9#H0Y!v)#)@ykLY)CQ<JmJ
z#woNb9DTab4mD>dXIZ0D6ZK<=V>#nB@++hrw4k$1cMh1xv7>cQ;qNT)UZewRH`L<8
zvBvwV(fOQHN5``*zI>u=-b`ZSKX~ameJ>^h%bZ?$9LcdhfZ%zGzsF}pH3|PI6;rz{
z2ipYYpwPrBM`V~*69+O8f#_T%kGlY4cg|g(Z(W~nReccki!DL}e?Leg4jLNg{j0j%
zp=3bGbtQj;q=n7?r3zB5{aDFg19~a$k-4Y!aV1}cR?xjl;p#l=kICJ<Ig75iI=STi
z8?E0W%_a&(_+vSjua0eoWbNuPme!Jw#fSkHH_E1b09%MKmM8fus3L?}z)EQ}?q_9L
zi}{r32^qHK5bHtuZdcn`-mdI}cuvb{iNDwdTut~(PQbL5zIGYcA~gH5xMh>4C&Yxs
z%kfcn6l-gOEp+Kb`P``sWvnchZj|jbDO>3sivrAmW9-&wj#PH>AU8eVUKY{vY@Jck
zxVJpf5iT-nm+gWILp#g8or~zi*ewlB`j|pZ=^P}B-Kgf~6<x!3u^hF6A48eJ+SC!q
z4HuGCE!zT*N1?UPH~)>FEOS&EXSqh*;;670^e9rdwoaez=vjLg!B=t&mr27E8>;+4
z=OidO*U9p^lP+2&r{w9tSo&n2z}YSh*XNG$jrPzu4)$PdZ3m)_=mcMzqz-yOh@E3O
z8m7RtJ@MBmX?xp?+g&2?03iL-+8d9me&4214AWTHZQcmO)o2bB-l%nI?;mJO7z5Ll
zN|iau;{-w@6y;lt#<qJnmJ<o(N@hf3iN=+b)6H>EPV!SPu?ZmJ%rODIoCW`pcOD^s
z!0;UA1hEy#iG=adhlnU6)E8Gpg``^RHd_ch(IMnY=zEEXjf|v(bOlML;xtPp6qWZ9
z^_f`A-8dDGMRz##39oeO^l^{BY)K5}T^VGlBt$*@?KSc=7xxvi)B4)4z9Nsw1^w37
zfnb1`(0{W@Yvy1086-c7;Vi$vKFCuYaMZ(9FUN3Z)^?B^;7HHjU<eH($1-`{AhZ!c
zLWY>rJ378sSJC&<0@1IzKf+O)X_$bWYolwsf?d}0?m0=(qFzH#?`jTFKFpD)yU<e>
zZvi^i#)7f+9RQ-iYrplZg^_n|+4$!ZeHwn8fGTS`&S)nGD(z-eU4ED4_Ygd0FZaff
zB>jTd3oX{SwhtK?2WBBR-=nAIcFpac+q1Tx{9Y)tJx0>+3-)saZ~rSf#-qG88th%)
zAMA_5z?7q4E#+XZ!Y+JGZ5-%mLm^=f=rGc#odd2NqxXjXiUYrtfb}l$%NA3SR(O#P
zH}5Ep)K;p$GeV##Sj8h4Mx7!P|I>bvjMb2mkj3U`R$nbn&8}8=VFwMVzL(Zoe7z(}
zt**tCUo1><=2Sa?AyCH;@u$LZe$c!~*Ig*3hJRiF<~9<}*A#@FpSXcxV%8LcQM8%W
z{29*-@Wu}~9_ML2BY@(Vm0!^1Z|G96jH3lnE!tUlZHVUBqD>spNA$?I>X9y;M(DzM
z{k96lm6*oa8NQf&!T9u5z3PUN5t52zdDp9wyycvB(HZ{fbNro3{xuDaA9SLA<DQJ*
z_@}xh650Pb<^DG%|ErQ;SDDy@N)+HPQa@N2z)GJB`WKaP0xW7+&dziu6Wtbm++qYY
zDt#Rn@-{ag+)B=xIbRvs(2j-KF=p-0?W;*RLQ0V_Qq6|{v+CBr*A0<G{(mH?l-%>q
z|67L%MjG6Tv?VQbD~n8l<D|vs@&7M1BiTSD=cdvS2MEi?<G7j@qnT%^HJ;)skwN4r
z)kCr$wos;X1<a0%$-^p5WwL{4;)z>GxdpLEjv&;8eH`MPK-1+a#D!QQof62of!o-A
zw00VO_%QFAGLnr<OC9kD`NmgB0*5XnF@ZXgq5z;vv`CW(JPe=#<zOa%(oQrk0LXRT
z9O_!wNl*;zWOtDDF9qEIt!@*{TYXbOptPl4w*md0Iv|_26i;h4=th_?7BqpJA>+Rm
z^x(9Ba#!Z^K@N2VDgYF?AQeD2F5aZQQVvkwh3ZYp;=F<N{xwhu@BbxK1p|Qll2mf|
zi2Yi0-DB5wPtivW+!bW#_s=w8_wo5`dRwWsG~Y=<YjeL{Akw05d(whJNzh<RW_B9y
zA8a&gZm&uU=KotIwzEvG_a>Ju>3S?@aPQY7$%Dqk;<3b4E*1kVX>9wfSYov~@%h=X
z&t56p5Fif*yE=n2%2r*Vq)JBX*9DqVbQm}n$AN3bhxv%yberJ&)L*TNO@xp6xPF&b
z?@4YZ5Y);Ork8NL@^?{CA((VLq9sz4$^3{37rEL-QMM^6>X+4^{>5&BJGtoQXG`=j
z6D1axUosAl%960f0OMfdB^X%~FmWg&%eM>7QO<w72_tL<gZ)@54DiI@+E9H@P(bR@
zE*Y*Dxk68*`y6+!H}!Emcv^u{*F;aGuV(!3YH|p=xnesCBek}JC_&gi5qK3jJ%Lw?
zK@Y)P^5!O(z(tS<LQ&w=5&H6*EY1;J`38sWq!SqP9SX=649!7-LFFV3r%M-Ac#E3D
zR%6>%qGesYL<HPz7M1_6wbZ|)<X4sanv!29S?zzu9j9IP&^>d2;OQ32DQ@+Msy)Um
z7{;YQ`KSV6U?YR3UdNgm+Wb8#go77qfJApV#<LuXD`S=2YUMU1H<k3zGA8=G6L9K;
zaDSlkD@sfnF1a`gGRbnc{h~k8jg6lbb~kGaaN-l}i<(r#glez0-VuMY;fZ#vS%M^)
z`n%y4zkt908~Sd+0f(&qSh<a&Dhr1C_p}%E*@?^uYu<TXD=MLOMFs^%5sN&*^b_of
z!w9hUq_*1UwIZBOiFwyQjg*{a)BG9=L=*HkM9WJT3h-PTAijkicNG0S{Boim*}tp}
zzKqZZ+M6}6&;cC4o|0Wm1{#)FA-z0T@GUkZI|lB+!r2a?pY2^6!p?g5Wi(KTfIr{N
zy`01YD&>Fr6ADlyi@lBjMfKx$AGxFdW5$mFv|Hhe>P5BjvhfVxmUJ$w{ug1z2Ld$3
z(!~0|)$D2(o84IZ3|P?Id@!K%eCnmnQYikCj)`sdSY!^w%?Psw*o|noS9*gTcKT0d
zZG<?Ny}=;-f)GoVD-M_ib?H!4xx`-Q@4(!kSG<Y_xN)gDWFl>-d_)X8UQSA&4t)$Y
z|8g=IZKl_EzKn1I-Xcxk6LX`%5anuX7bubb1S444$f|^}jjt4qW7!q!wSUE-OWyMj
z+<VX-7lNJW_u(V#K-pd$zZ(o$7{(!b-0xzk@Wk(PHQ@CrmI}T8(_nagEZAk5XxZBo
z5z>v^Oh$TXU`fE!H$jt}@-u=+V<<4lFy6l*yxtUx*#5DE%%t83<cnNb6p{Ne6Hf4X
zEDwp38%JHfHF@<>g-2_%o9?bAEJ#LypwgG(SLjt*ul9-(nSeFwR)I-VTa=JM6JUd`
zM;pqhqAZtRW-mE@;o;4KB!xAD4R3IqeT4l0C>_j09IJ$~D9UWwiK(R$Ctk_K5(xfD
z@-8k)%DvwRRNK|o359+7O>Jlm7d}nz_`16hkBVB#^0QGHC1WT2^vegcH9GE;ALPw$
zgraVK2YuN(Oy5WgQun$ZCjR0*lv%$;;94Ddsy(LVch#^cH9H!H=?lEX|5=hwMkvf(
zxiN{YP>TtQi6uD_iD>CSEpO6y|F@OI5Vu85J>VFs-R?Z`7xkQG$f1QlQO<D4$Mo>W
zb$LO#f^z>!$rqJ;TxF)U`Bq(;+Q%t&boogoqQ?B1>KQZCR`ytY4(sn4RmOKa>|zVS
zq^l25>TKF4<#{1|8J4i(s>;k=MI30NkmH6Lw)s$7SRto{Tn_Ulb*^4^-zwpf!#jVj
z+<&(B=qw_0ulkhA{-yf+^CXqx%2`=HHD?t)S8tc?L~DK6*0COI)rLE^-5kH9RiaTy
zWHX>0`RU+%W=zH)s7MG(#-PR2aR|5~2g%Bvxd<B}!%&&mk3i0~Z6U;V45#*pyh~t7
zkwDse<ZId;Qzj6j8=~ekk4DY?I3F4GL^w^d6vVk$aT=zAqQf^CEOlNHjKNX_fI!mI
z4~~+xVn1500fV>&|GhlPq2`iQfwMk@DucYddSC@snt1s%yTCE6!0dSZ7>?;Eg=ugq
z&(A~QN#*(drt&Njk1?__MZOj1VShfa`zyZ@4B~_IFV=?WVaa4wLqRY39@vr0n&ei!
z!0<8t_RaNs1zf~_KT@R*1jFDR)S(Ks23?t$8-jM-0p7825MZvf*>~)Kad8qbFbFZO
zvV6Jqly)n~ya*%_41-rjz#34f3S0qR={LO64>stD@yd0s@?I&zD<8$8X=l(MTR6Qc
zOk#MY+y4Ssjc0{bu6;pba^Cf0VxWxX><5bXY``vW6>i?04R$wk-3dVaf5)6}fgaR|
zfyX8oG3xfsr)kFl1tT2|Mi|`K4l>xbw|)GrdTcw$@HGI1h~#Vog+EhczeoRlfd~J0
zm9zb9<n&i`Y4G6REBE_Ks&TLXgL3hf_y0h5H|}=-4^?Jp$P&Xv3CfCD_VD98pU7zs
zyVJA~3lK^`2SsZKvv2uZ!ttXBvn2sSYv;QC-QKgAgaTA;XY}17Kao8Vssb7#G1aY?
z@~^~CdHouK$F@xh{um->$WeY~Uoo9oGop`8IM+-;IO1gIKO7N`CaQipsX%JqmV|KB
zeff5|2s<5Q<tD<BtqsLdluG~tY{%Gjp`+|#A?Dh2hcIlO-+$Aj%|&L?kv6w)hMG>i
z^fl8m{_ik9?3yR}<gZ4{JgW7h@qw1vB7%ME_x_h8UfvydDe*OQDW+DSF*fbNNA>+1
zw;!_CNtI40X;TnJUuZkELU7x@HYEDh+jM-bEIzi!z9u+*r#?!=T_Pjg8Dm!VYziF8
z=#Oka>#>Q4q;ffakMV2hBVdy5mr;ToSfgDgb}cn6Czl-3Hz7vM1*b9f@KPAUXXO+C
zJ3<g*gc~SYaGgLXBDU@Ibmv{UJ_w*IPPL#JzVK;(<B*^a+ryoX(M)UjiXF6)hK~aw
zQ3-}crA{DgrZM;=Spg^V;1uWv@IDQoNS;k#jsqL5n)J_4Y7})+hr~)w6grkhq?FpK
zS8hKP>y<Tn`WbogTV(il5e3f&#LHtOox#jL7-PIgqY|K0<GkGyBg=sbsFGykl)3gQ
z?vrSuDZPxgSTlgmS;jT38uinZq1ZBGYNU|}egp~gv6Y3j^h^RK7uQUjE3rFk>Mg&x
z^4?|%vuR3A>2(mtxn4my!aw7(sS{yZ?^7wEra8*BKB79C?<*t;^GN8BvZg2H|ENa!
zZ!7Pt3~*gt^cGip&q=rH;-Qk^rf?E56>X0Qv^yGH@ZC2nsh8D|8g+=*?2Vt}Zn#f&
zpc@>}H<%K;p!i=@J6DuIgGwWERC%o!PM|aegbbtof1vio<~VeCM3pQt{P!o^cup(g
zFy%VzMfSR)VHLv}5jZKrM#F)F*i>Odb9(b0`zn2XP_LHl=9}kv!<4N&GN<@yI~vLp
zxC;8<zCm_%3~Qh~Wm(^aj4<V5uiBwN=J_aZNfuoSy<Jv%%(WeBBWpWZNIT|;#Dmm;
z>Bl&yeeL40?m&i<Av);@f0=GAqS&7QiuvRgJ^Ar|hC4;{t-~zDoxx6O6Bo7q<CK=5
zFmwKwSX6owMtSy^6MC`p(xN1)c<YYWjh+`xXh#c6)<Vx!*}LC=<DTtX@p|8`4K2`2
zqS<@3NE`P$a}8&T!4VjJ!OojW%AacWx!E6;Twyy6ARSSRS58I_O2|3xZw?ZJ7O4xm
zXBUHZ*UC%g3&0t^(A?KelEOK_(6<D5KyLDPk*p-<hQ0oX!u@1(q$zFenC$pHq)px+
z-zz=tF6xb;@!gF+cZ@oFBwsO&?cQJ)t?c%OsJSN?T_=PFI@dk7Xu%{bpCp>V&h>II
zO8wF1fnd}dW-j-V+w1MH+`(XPFzV*n8-ZKd8|>w)cJfs^WjKRmXY^+BQ~QwLB<2pG
z*-oLPfT;0yy^;lJ?_1m75xrLL-p89bY?L;}Xrsu;d_R3U9PFq3KyZNa-IN#9cJl~r
z9&H}GnR+eVXk8t5SG;!*?`11*dT-L(Yx?DF-rO3exCi@qQ;)ab&P#rgUW|(;gZDT{
zn_8=7<{|Ch6C6Z7H|!k%kR5zocL}n{y8-<h-;Lttpm+F{Jl}n2?XZpXA>MT)IKo#Q
z;Y~+>AoE7*z7B1B(Dhoz{~S_sw~zE!t6YCs<>^E!v37Lr*xGUCh_lFpW5IDeNTb<a
z8~2U{<L(sd;Fx!u^<NB*!7-8hSSmOkl^k!COsDzUs6J1PPwV9KalYjQ-*RHZx9A`j
zty7Dl%rm#yH|<+Hym>BLeThf(q~5{To+9?v{$PC6z2Wp<;Tz8IhVIw0(j$0hX_o*S
z%))JEQ*n~qDgCAuKKmRBwA9*(wc9#-J$E>`O<^oq$tQ!8^z;te%hTS4=IQk_O=QgL
zXKz8AaN2t+TJ_qES^s<-Klx7YF5)NWMfOMWlkX<zOvV@M@Bcj|e_zQDDd9)ZK+>>x
zqu$(?wR>2S`DiO50C?_i>Y2Z#<PAM2PV>!OBw)5@^rr(8hTT_|7bfbFFV2n)E-0-Q
zO(gZz0y%%(aX<Wr^$CS%MCTy1Bm_X4*!mCphX1JhX(A+86ljasvAVkzCc6BM3R+|y
z?H6jM&G0~|+qB@PAoPtOt_(Z!2#2GyF6NOQkD2S$9$X9?3gA#bHQ{j_&Qz5@j9HnX
z^V77ldbHJ9XCY?ZOtaIom}yS?wOV;*nMnP()<cxPRY$@-8;OM0BPd3R1R0KPyxTdw
z;b5!xQ`0l9dUZ8@`uueC0rXKg`}AsmJVfuQFA;RKxwShsBpP1For_P$7262ThMckW
zloV6Cu{sd9_aw?7-Kq?=>PkHnRbJf@*Xt0VHW)t>JBZ48OVMi5M5{!jYD-=f$9e<5
zC6dWlS=na%<zs#Skw_@!6xw(zfh40(7w&pH_H+E4i;FC(dQ$*!CzZ7iuIcV4l^D?|
zD$zEr3&$w=L0EXKcH>dsM;$F1%kNWJ4y!qT27$6lj9`3GITM>`2%JFp2bF8np(f7B
zXg*DzyQaG@>hh)%DIY9Ij{nEHd`XE>xTbd(d*(|9>5H;-b~1mY+@#7Tlw4QxsLFnh
z-1sR+w;oi%MJ4Z2B2LI{S%PuG_h{d~TV-D&krv$+N;t4YocETZ-p&C)POG9#bKeMu
zT+P|(>1A1ET3tWGl)dF1a(;z9L<O|7BvAaL@H7_nAJOlypd0F&3sGYf_V4J^KBsOP
z)omatzv|4Q+jb8jMP?>LV7<6MiZSBi)Ay`)eo)xOA>VE<SzIA}dZ)tf){bHGntnCx
ziEC9=X40_N)RFI0Vxlt%$K)dxt9TmGIW~&%eneA|YzJxg-@|j)wW@!CJBgfJ0pv{K
zG_SIykqwP==Cf!q$<y4GX7(g$zcCDA&=t2`j(z>fGJ-P~&0QjI{Ra!rUFAkeBga8l
z3^5%NijSmKL#@g($CVK;o02h}JBaA~yqFX2A2e}*3yc+AS#K3nUUog|dlWN%i}7@O
zcg6rhaxU!-V{)|{z~cA=;(?E7Kt6@Awh|L;E|XVC#`oYoIEi5{k1b*H4b*th1n@MJ
z72LDZ7mHEeP@-qM{BI`&Q}jeD3MNm>klG?*Vwq||k)gf_i3Zh>iLMnZN7s6DpZ|vz
zjIIA_uJO~&0hnR}BhU3q#6jzvN=Pe1Fuggbc}`qX&}Ogeb$PYeq0%!r44EVvw%7z1
z#Y?M4INx5O(<1i(fVqWGGt7?JvkPZ`VD&&8pxVVt>#WZ(WBASzJE;ZPPqJEmre<d}
zl$V#vkDEI2@u$m;C`zesEHp>8`4ANg9DV8k3q2^w4bH(`aW(WdiV&mlf?c4^Ys+{i
z+!){CzgM*l>>23%9M5%xZu_7vKcvJU=|`0_I2VWS_a9ROk1M&VyYos+?k%+812_Ny
ztzLi=jrq23dxB>wMK>l@tl9iaD#**bNmqM$bdz|-LeT%QHrr{MZN>PP99bHoL6b?#
zn@O8aOB#MWt*ur>&$QL-88lDW#El0C#QrSmw=j(@?Kv{eY1S=nNn#pjn=Em>*vyv&
zCWAmBw}^5mH&`O><(jY}ScPfv;?2U6!}x|P0)EIaXISlZ-%@JVyD@8q?B>2nP~@#V
z4{=ENVo{TBHk1DU!vMADjt6533OS-F9}VH%rs}4YFm!0h>VQV_rf!&xi{eFIKkpgR
zlqz}!V!>v`pt1m5huou!XGa(a7mmlk)jmf0@HR3;heF3W&B+B?zjVoeKOMe!5%IV?
zWueKXzpUiCk{3wyZvN;T1ua?krMZh2FY(`aif8pp+VKk6ixgu`#Gi{7t!e)^d6c5w
z%T$kE7nN~|HU7m*moM2ZJ?9Q;phR`$Pu1%`sBVaUyX^lbWv#u-moHwvZ0*A!oOG-I
zDc$`>k}zK-y!90R>CdXWvnrQ(ro%$Ep|YmRo>y*C$%GOk0p^waq!J6V^xevRzmm@?
zIi$pBh%YF2Rmou`u<8kC-S@vM_lrvYPbFhYo>3x0CwC&U5fTZN8F^szhmjdVR<?<J
z&8vS|m+^!Hb8Netyz713lC^*1leCJoD@Su1{_8zk$flAb1BqmBzR<&A3_FUYLN~d5
zp|{Xqydy{L{@X!m-Nj;mrjRZ476&+vp|>zx{FX#7_x7m<p@8<dzvS#0D?1b2>+A2O
zh0(aiHy8H#>e&>XivI}%>7N0G)=>^^GZBt-$k#G89LioNFh6!d*-F-q3{dQ)g75KU
zEVWH@q2}~AI^CF{%gI}Gik+I0Y@3i{#M6{Xo<*?Ff@0Y1w#(BXlDnGu7cr%v>C7cK
zAA_?}n^$8l=b|K}2YG=-us1cNf0)bFqO)ia7qp2fqMPs61GeNo#P<_McT>w%F;Fgs
zEK#>w#G8v9t0^>)jAUIKo#Q5`VIkiwkUUV(Naof8uoYpeKrcL?-LECCV{R3P0R6$P
z7=b<O<K|fN;~3#`Ufxe#W8dOnO<k1GjDx0iK7jF($rzeD1TH{5g>qX%<X3;n?kF_D
zyA^&YNFu~Z?~H;IpleWV;w(nBCGtUBN$r4KrZFv;tWZGQFPMT}4k^|?YHNA_=IXy|
z&**&I&u#aH&KB%h-WESA@g{;!Gl>H2n~MIQH?xoq+w^t&mUyawQa)Fowzj&gt@~YD
zh!6L<*4kfTeULdRhd)J6N??N6$e#50T8?V8vwB<KDS>JZJLa6(F>9HlD7O`7RDrXR
zwAIIqe_oR%U3lmEV-l_W05><SMO&*TA>m=&2dCQtJ1U4`=;u?k<{#0e(E*=TrLGPX
z^RM*Ec##;7@Bc4)M$AB%n%dTF`M8$hJv^ypC@NR6i?tYgQO}6Fm2cA^KMX^npBage
zm%U>SJI5`HHu7vHU?&J6zX{Su0A77H;QKbfQZS^f{t0%V{?0q7!bSW~=9Bto0j66(
zO7O%>zwT_>(5Jjq1Xyu~BBz{WYFYWgdxHkZx;TXf)5A(WDjo9ry@8i>whHc7nM6f8
z=;1KZ>b-Btq#E(p61MGWwk3x$Exr&*Fci6@&l#i6*RB)Bi2|eQ1^Bd4^;e8>L%Ik>
z;b9Z8;O!U^XnP;aMA<E(elnS~0YOw$0)+<rz^?fYnDt@Jt`<YMqcgO$b;ttlp0LG3
z5~*jx{&r^_o#w|?+9xJ#53nB7OgMmry{oN}R#>FXLlc9onC@#q|3~!kT5f2{T2Yz7
z1AJgR7`V(dTfx8;l^IB-z~2&FG9M`j7#G#s;vDNF%GU5Zuslxdn*txVoVnLDUNznS
z6(t6cEitHkl5u*b{gGlHrJr1lVr6Xk$Zd!GVho%m1rdNYr1#Is1X;k4n{#Q9n0d5F
z8Kkl~4iP^b-{Iq>W3u~c(=u^i3Sa#EMD`-7MjD+lFTQRNrc{qOV+7q;#lrOgj*~h}
zNGI=c{5MgMc!Q!cp;t0h>1!w}b7m=D0tw{#VbAnBE*Dr+{XDh0@Rd`9r#^H<IsEWY
zB%38{A2#o72v;iiitz%3T*?V~e~6LYUh2DDcp4Vie!-;7<tHyZ3HICwv>+f2EgRAH
zJOk&}sK(Jri3}PZ&|DgUuEQt&+mr|-I<eG0R)6{(a6G7*zpPt7p`5zvj<(f$c}wX@
zm|a0#{X&>pT?xB8_H@{Vg`lZJ8<LG^^&!mOx@FY+idNG;-Yi}B7*=<Rb|b-q@lZq~
z%fLaQ(}ew0<fbhvOsY+DSnr^y-~6XE_hq$dhb^`2?ldOaoZb8yC)jR5PlrAe9|zNC
z_Di3Mc2gdjT1%K02w1WT3$hz;>od`Rq4jVw|5&5hx^^1PA#@JXXb$sN8co?X=lHvm
zzwq?y*yXunzR+%tiX9VUhB_RJIGv3_O>~yn(xGQXZ}=jrZ)68vH#*Q?Qf|Fpo_*NP
z?V!C9I?(T>l;f35gE@jsLA{Z_Qh)nmE$Bd0!X2+4Ezkf+yWPfRYh%#b_`vpUctda7
z!XUDX9j57=YxS%-9F@Gtj^kZ0ssr9I-zR<E2bx`s_^x0u)^v`c=^UXS=u1u0xjT?I
z>1eF!`~|z->FtO#oj=^%)g0p(sNJUN<mf1GC*Q<1noayk(RA)(x9_IT{s6rvnoW*)
zlAbfz<Bd`it>^jydWP=uR!h@~u>tiDnPzlX^Kh`s+XW2SM{b`tX1OE5zF?P|e{VNz
z_CBuBGV)cjQ~wF<{r9480)z9mqnL@MuppGCbDwEC_pg;Z^-R1w(rD7ge%dHl8;SK}
zL75&O45URYQ{L?zF#Y4N(BI><S%Tm^WNY^D%I9^E@^e4_ZAttN4nm}*{gEcxwU=eq
z-X%S#pKG3Q*LdXYP>n{hc^mlj5HRlW%SqPI63-s8XD`Ihj?JAkE$UsW>z<ElKOEO_
z?XO0rZil@iFBQ-~$8Xb2!nU|~lsemZEp_czu(9m(jxt(KN7^x3SYjUbj+s93$l6hx
zqa%#RF?5>)-f>1_d|MsqQQDK9@~DmeDKpYL%{-jZ&$NBFnYQov+IZ(I%t=Ga!FJ5%
ziQoipIL(MF-rX64<ARI+S%Fb9PEY6mseo7pDCe(^C>5xyKG*>b8-J2KWfTsNz|8^Z
z_&vkVVpW6WO)n^^KvIXhNC3B;I3&xX9|c%$FrM224%?#A4n;=f2;ji}pV1B|X&cR<
z|L3~3RdeY7XUctb+hipFW%Yx8lJ{({8gz<1R!G8BgRvCK7(S!r<6+hii=G<NFprY7
zoDHId$A)mu25ka|F?o`TW`aZBlm&(n+tiPI4t`SjO3kB2y7Mh?()N=O`y2e_m5Qbn
z@w1wd9EoWK|MKcMKTliZXTJtrSD2Yw#DVg6G!?TNv?r8YQevc|ab<6%1#9cRoG9pv
zdhA6df7LoK6FiyX>x!Be7vTSh5|dI*DmOtgep?3-bWz{5QNQIE85gJ8Dk;~aq*sZQ
zQ+}V4ekHNy$sf?=-_Q#LDcpW;E1k~i?wv|*SMso`3@T^r>96atAzcnDvGu+~xe+Dz
zQ@N5`o~kw%mi(P68CCK#dT5t&(joa{%I#L7)w^Li8yzoc+>%!E_j8-bucm$};qlN?
z8Wo7oWe(QRVx>rNbdtR2{jW!5L@OK_tmlu?9v0eAdi{jqmW#I(NwLz1tF}t0b(YfG
zfH`r{7m-6hs6lk{)VuWEcavO_G}SVfG1)X-LM%;nJhu3m4*98((<Z5W8<tyMi0{?d
ziF?5gub_io`*%1=p|87B8r0T5@v{2YcrQtt{f6BhKr)zpJ!x#VSnnPk_Ifkn74R_>
z+byP>;8SA7y*{tsnC_x6-Q8fMp)d-Uc5xP+#d5tCbKR@27R#-F4^K2^8)c*oM7x=u
zCZbHeS)O(eOb~;)u{?Ff!brs?0A*FJEj6#4H+^`!rOS&n(a`Z0tp=oH5`RE5#KDSM
ztyvINMZ~#8C@6dUv#Hfn#&@1pADnL2T^r$Eot$5sy4pB>mj7H#g{Zo8x|<*AeB$ic
zC<KG1UD&gAq9h`PM$BW1;uk6~nZgPgBIBd=vm;Tl${~;D@1_nv?GhCQ)sKO_NgE_y
zmBW6b8p#)sLnNwr)R5vi10{j!s{cGWj^yR!5Im#Jz}2q(CbSAc1Bg_dN;<Zja|OSH
z_xlf#td0}V%Z2W04O}@2vNVHh9|y-?sZCv70?Ia7+#k~7Hb5H-W5b@-b2i}qJ6U&O
zm(Ez=kb{K=XH<#-!|hlG#ZJAABYbC>qC`kG<lv%Du4UJ(w*31^DudQw{3f;RKd8G8
zDcPfL8PJQrYb%o{L2jt+A~c-nmZn&NWLy6gaiGE#BVz9^ZeGi=nB?M=CQ<*?=K{Y+
z_%&Xn3xvZ#{%A}Z>Q*m}Wpl=<aWMuxiU>i5YuFsOO4qV0KObZvlz)0Hw}P1*xr4GL
zN2G*pGY82K3lhuY+%2RpkdLTg-<i1fFzf&_;5num1HYVdw0;kvmkTQoGSzHT+PMQk
zf`kwjLz4Sa784T!HM5>5@;Cp6Q}1eV5Oim)aP8ZzWi6p!YGS}@Vh7#{i$%l<g(8A2
z;#vN7nMoc|BLkSoYxg+Eq=uM6tf;-B!T=GN9+#{3HT@w51-&Q*u#}v<+ZS}-RIb+r
z8zR^D22t)%INKesALb&7TXYtK(eC(wMhi?7-TP+l5m%yh&m>Mmmy85p_J5LD@qdbB
zwS0l-Z?Y;PK>LmIN7mf>#uJb*%<@t_y{hmg#E58BIkvj{|Lg5+V&gc=IBxH*?X`Ek
zyZ(riIF7SU(%2gjL!z`oZIu*MN<$J;r70AH5Vprzuj4Of*J)TeaH!x^p&lwEqzI)q
z!U<KJ0rAbD2dW;T1qlvZ=!tURgb2U?^Ulofde@GVw5+w`nRjO1nVosx_y2tS-{+aM
zOe1P^DkLrg!%?Y7VCj`SS-++hmgd$7{a!9G6h*PS*4Ie!Snx`or8c6idimA4%EFRr
zFc<G4&Swos?rPI`^!#7x1~XU4&PsF&&)r$I+mXDg{bWL6f-R>L5i4KKk!OtPSj3Xx
zNDVC}{8uxVgM%(o^mM*D@JjtCSA=mzCI|o5T3GM>jn;mv#jF;;)55x-I-5%@cb>Ic
zAKq8OWUrJ6T5nQ|Y-3Kek|aZlvlA-#lNK6H{LgCfo)&fqlUv_svqr3z*Mca;Ki6(E
z>it=3f6+qHfm@hZph4h6_0_Yi;9diwHKj@rZ4Q|DFi<V?4DL7EI|kQdl(pl0^%tf&
zYwgFSHPU|6li&)julhpjx}8RNs?SNj8nrL-Ex#IP#NLs2ert&4zTHmK6*A6YPw!Lb
zS2R!7G_im+II;(sQQx&Sg6^(8Asap1cD9=l7_H7$GbDaF*QvA7=L0r+5=+d_*dsPN
zHVv*PzIHEK^baUa<0$F&|BW?JHRCF=|8US7*yzC?uOB;121`V}2H9cKUe@d|;-a^y
z!^;JAQg@pg2C)~xwWd3NA-pd-xht7x$jXVu&HWD8uN%P1q<9-uu8So`>*d=VHH78E
zh9Syj6&w66QBxI!moo=BQ(SewGK>Xh6wY(VRX>CUM{_!vsn4)@1B;7nF8i_I;4-7;
zzuWeRM+7r%`@gIls2pUT|Cm{DhOyx6<7sdfRvi*~W5GF$1!s(7#x0_lqKO@m{RZCo
z2uF^?YR}u<!WR#?>df5a_zAQ4j8=|rje7gB+Z<u-2pn{+9oxbl<XQpzb4u7AGM>3C
zIQP*14`D-?G8+OGAZ#T2ik;rrZDL9Uto6fY%gNuGXtW~m%tD*ZkT=dVl9lToclX5B
z1g+%Q)-kRk8&5w2idgZy>pLf?Q(SVcBL=waG45JeJqEl9?gQ0?U)}aTV3MEkj+&JO
z_P0HUNv>4bDsZJ^T<3U;m1GDR@`<`t;~inS8JSseK2+l&n|<T)EeE6s2xSi!M%(+p
z*%BJO!VhOx_z2j)?G*LG<Q_EaN6@3|a+!o82RM(TTfeSa7=2r)q*AL3C-<x~|4sFz
z1aOb2J)0@UOk$A=Ev`1Q?IWnpNmOSeM3@py9kV@@C>q>$MS2@HN-fOwfJoowk%Vl^
zal%%IDs!kV_q9=%4|E>g1IrR;M-$IdsyC5yz+3I9E6HIn_*RsyTb}SVJ*#JKy3sbm
zktxW}vv<Z3j@+vB4885TihNuZ{9domF^;rXB0_`4{2*od8LD_+62%DTB@>B+O;YI`
z&gPCdg}I~C*D(w1jy_$%n$7rbQkm1HEX!Zr8~LgGtqey>qo3N!kcbJl+32axgr4dP
zPT{TV%<r12Dy3|Z!~h74{^puA;|*AP1fC5^?*VIaN<yaymNXzMDoAfuN&#fYo5+dm
zisosQU8cRo>Ebk!Y5BUU4NjGSSs{RG{zhpoZ_V<Q*fcI`?4xVk`MWVA6xY(a9wE6+
zm-k=PgI15$M-N%+rWb#LmC(CN{Gd1?=@XF-j2^5I!%&Q-<a-aj?1{RUy_RddjucW8
zTN!RmPBne<f*Mn#F}BweOX^>=7Vd5s4KehcZ+}^lQaID>N|sirpH$A#X1QssbYwk7
zixW%6RC{yi!F=FPdSSEc*pY(K!=9_?j2=Xq)Xg`kryCLGduUN4BH2QGlQVgluuTew
zfg?GMU?TlXb8Np4)P|1deq#7Z4<u$E`@5Hk*GGU^+;*fNB6!AfPQjP}?hL#O3lXJ6
zkVFB@8GI}JDrefcDO_f(6O3X8k*05X1h)t6ouhGSBPigY_WEwe2p7_YD3wGz%{rI#
zHBxYD%Sghhnd(y&<xn;ojvA=TuUw8&!j@h1Gbu#0=6a*wVY1RLZtH*i_x}C+yyKU7
z8;)2Rghy)x<Nc29YYNUbZ?-8#)@)2~(o1dY*Osc_rkhyvdR?2wG))=hcw>D9ZPTR>
zx~_4|2*e@@?c(@$^Rmot+Xi|3&(V%y3SIXnS(#1DFD$+0pJlbRMVmEIEn2#jlJF=x
zpi81NW_p!NelUiU+i6cSe1?{gmNp+bx+5ot3?fkvewvL6!K?5l#nfB3lnpRWPFy+w
zO%{n=`JTx?O58q3!IlXZDLaM~DxLyU;CU)ox%nl#yw{Kp_yUL7`~xQ-<n?S#*PYPX
zzo;qXlSDn<MPUz!o`$Waatc7+&_hz^?I1*)tk9v|x8)%;;9Jc9ylz4_>}<d-7G8rd
zLoU0uE4sBkS$Os9=5<eLZYkN7h>K~8jrX9hJqx<g@Ut^ir3``;)T0hn4!-dGw?LKb
zz0FY7Ukm3!m^Jpo>RhqG#CzJoRfsZ*`@g!rw!W}Qm~&^p&12kP*jN4mJN7O(xXN@j
z&G%{o0W~870WsSH5D=p%Ml6MdVkua#ZAMYdo|+~}Q(}$rj%E{|F~g&^9Y*`eI_c=3
zDKjKHwr4HtHn9JYhZTefz+sjI;8>ewp`@HqNt96Q`BBSh?OA%MKa8*fTLdYojs=t6
z+dQpycric~5njmY^Z0lnq`<9?f?K{~2$D1ek%2%6@)Kgg8h%7*@kSf8kWCOG5(Wjs
z4L5m&d@@3i93*0g6u1kFk=f7w0ct@a@_LJOX96U}DFauU%P@KnAe7)p;El}zAPFde
z!_)hTu;}%}|L@;Qp|(1d`35MHUM4*pb!3`QCJV|yVgO~%fh)Ntl*t*&<YuZ*J`Bo)
zzJ}rsAY<rZvs-ID^QDDNzAr$BGGLkH9oG8iR!YS+T8>fymu=Dl{;nb3XsstmIb7s8
zeV!}+=kAB^U7mmGd{9F2uQs@}Z!fsCM14-jvcb=Q=!+VVY^;fZ$iJwvPjYZAF-MkN
zElDp{$Dx4+J<7~>5dmutIOJba?E<1&Z_wAcM7YXzb_S*&sr#6jF9IgbVi9FE%oUfE
zSB`Hoew6=XjUW{U@^9DKT7im78MV_!nGH5VQPY7GtsTI!l(k0PqX#4^3BvV5{!`Qz
z+BM{P+I3=?=}*mNv}<x%c|F*~Ex;_TlK-^nq@~}*6ld_(7r>iLiuI||VQy;wWu9QY
zk#G#3f1a;(*)t94q^9cFD}G5&-QM$6t-Zp6gG|(o1}*3PeO3Fvq{WwYP+Ro-h2Dga
zrUT(%n7>h;bre&s-4kv=!A^fh6`K?EqXWRp)cKZfOJ>+)JQ2%u0e^X4<?Z3`DhD=c
zYECN)G_`<PjC=D=5AMv2L1V})PP$SnOL=L7!CrH7_JJ_j+ISPoqDAMee$C)5&nHs4
z1c$H=PZ20HZ>mpzN}87D#;xN^Ky~AQ^7<;dH8wy<ybH$H$<J*Tm9EBxe+*F5X2Z=y
z@rl%yBgDMIM$ndt628?5Vs<;k#Fz(PSDqU%Gj6)UOog^<o)%g*TE@9bw-K3p6lur&
zOt=*rQMEdL0i!v0SE`yb!hVtwteqALvpoV_rIsu&ZC<;!vPif@ydN%st?@q1*HHKL
zna^94=CI8g+J2P+|CAQSc8=O{Ynv7LzN^Mz_If6Rx|oX2QBa-S{cS~1{PEMCef#kR
z?)H(X&iBGT(7~V>P@T^{;0Nx3@*K$>mp(?hLrGz85-P=Hk2k@0NX+;_Gq^4?N~g)1
zrlR)Y&-^h#x>>6bFtYrXNiI6ZJeDfAP>zbaBn|fu><Km8%vR=(NYfccY*$;B+CD(W
zU@|MSTs~lAN6vU#^y&=KPl_NnY@x?`NEm|ze3XQdl!U>(kop9<;%KPR63UV(ueY*z
zG+TAt=Xcg|Km9P$$EHnKpm&|RR@qGeX{QoZkI&~9@>3h9^GgN(ogtOAi!7xgHWv!{
zCCXy8*v4tTXF^e<h0EQj6AwR}>>>5fx2|h6r?<6Pmq)AvSa&vg!w~E7m0w3%iX@L+
zWQ@P6`#i-x)s<hS!(hmilp&X*n=9+&4JC=xAKY(Kl2L`H9?;cVC5<MPc!~=*OB!$h
zn$+P21sHM$ow374`*v>|nbQValLGYEA4b9Lq{lnE7C#L#d*J~;tO4K?{w#yCW(gU_
zr`p<RP06)!-QMxuHQldJ_aF6^#E}Q+nH8DoGd|OVuYHBitA)?_rbw&iS`vTx!qW8G
znw!*XjK3S&F|hG1+l9}%xKvS)dz$>p>;ZSYO-i4J8yn8X_TB94?#0p?N#30iLGR9G
z(ZwLTQkwUrTl=yGDqG7MsT|f<%04%XO<D_$qO${bfg2?SjF{f6lsKYSXH2%-kL?9c
z$xjB4iDd8d<b(EX1}%9lR~L(OQa=5Z&Q7j)t9HCExy-rf^@J>9hvP3wfAj~n7-BJ-
z)fo^BHdm-0-jY8|X>GExG`cFc+HeoratZ53RMALq3M_}|xy{OYkjc*l>=1>fHql&Z
zz^DqZt!LmX1du<f#Ua&su%XU~f;DY{F~+ptrFWUeat(O-hgIY<&rE+a2yEf|<Erha
zbP?YW^;2xtvX{b_ec67we_97=maNS}GcKH2_swo%(xbJi2HHVn2&yHYbwA;p-f;Qr
zEE3H54)qZ^S+uT6cKpSrSgK4D?$WVzTtP3A@xJb4V)#29?(dmgxaR(!A3RA(if=Y|
ICimq30AtJ-$p8QV

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_22.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_22.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..f11ec493ab357035a072e4726745849d04c7a1f9
GIT binary patch
literal 58025
zcmeFad3dD9ec##lh5Hx`=I*%%<lMYziWJ4+P!vg!CWoTcBHaS;HZZ^d4b0n(IiMFS
zT0==xWZH7<9JVAIkk9ctvAvGs#77c4m-bm_edWn!-z2*^R#uoePwdE7#u8^g-(S6b
z0T{|T_St{ZKv&mWZ&khZ*73XRSBpJ8g&3dZCqGe{|9Cw1Z*<ZANptaDzO`SA$6{Wr
z6!YR<Vm@Aq+izkaSxUxTTB?+`v`i_>Z*o4jkT2!!UTVIu&{gVM=q`2Jb$Y&Mp|{k#
z&{yiS>&$%r!a!*ts%Nk?Ncrsi(85q@$kKB2!wcI=+bo`+A6eL5+HUc}{OH1t(hiGv
z&F@^;Rob<%yR>^@PifD>-qPNMeWiU1#Zqx$e`){1fzkoX*FAr5;ZW(2#e3!tFB~Zy
zSvXocYS+E<#}>v)V+(hc?y~E?`MVd6myR!-D4ke1SvqNH{qv_5PM1zwd|>{}!r9VU
zix1A<vv6<eUgAUEu(!<{`Fyf;pLg6F^>(}wFWv9$^me@wD?RX3%-iklc_ZfSsbuFK
ztem}(Dm_HnUT+_1`zjCJNR}Svy6EladcX3J@-3ts@D7r4u<|h1k8pj+JIwWAyM7PX
zN4%q4AFaHH`;T&c%p2o+%&s5f`Y!KouJ5+%_j-w!Vq+&JmKf5))!BKkxG;OIzU)_u
zi_?XR_4-on;Zvt(X6qN1Cr?f-E}WWQo|^UiiP>uHlvk-P&emqDGbg;+>FMR#>|*uA
z()IeHKXvic<ox2~sRtf7egC94J$3)|eP<t-zVAWro_nSrym#_}$@|Wpt=xa6a)0HX
zGxy&+c~7gRa<is#qo(FlHs3qBbbYHPE-qE76SJpg7nc0RD;2NRI@hYoL-*eM;6rC0
zyk~Oq!3XZ2m^yRs0}pujc^<8OXmWDWJL^64(CM>Nm1YI42cMdmSgzG(C#tQaJG9}6
z#f7C0&el%VCl=H}b$_B-n_l!6Dt@hS;zW_(soBEYT3q8}@i6!7#KNRE@$8FZi7@Y1
zCVaBbRKkp3S*lFbN$j0ktkx^n>I)OqiJ6KYc2$?>=c6>rXO|}Y`s~C!(Nz7ix4>_H
zVd7GyJmb%LT=Y`krSeL}Uo0<H=dY7ecy_i{e`2vZJv%d&3w!4mXUg?!_3}imJi9Ox
zW-8aXT??~bWqM+Hz8<D$YYP)gVV2=upaIlgp%bR(C+cDM)Z%irUZw>T3$<`N6Qe#+
zr7`tNdD>rGDAP(kBpzm-ovlB)Frinh(i0Q&<xBJ7?sKyfGmF)U`KO|uJ;#In*~#TP
zvx%&OPc#1Y+4}YKvwp4qQe6-4d3?FPSgB4edKLefXUCG^puOv4r9M%vO<k;b%k!0R
zh~Dxirt0O{YJIU>uT*P`ept|it(50yFI8wz*ST{~Kljqa!qR-j=d%B4z1p*jGj#BC
zv)5*;r1jO7CZ;Oo`s_lbR<A78!rYWzXo~mlBIYmFY7_N(HH=@U8JXx#zLlPw@D?si
zB{a^_M{$-~><PZL6hRPU{l!-%pTAP^ullp~N^yY~n?FhGXRF1lth{3V;%u!ry<DAA
zrxfdpMZIsarbTMGPF1c<%v-TSk&m}nsTHe>_2Se;O{3yO_4>l1UpaX*Mu(<qmHFu~
zrDynudAtR;7EfJVT&SF)P-Sv)@zN=8qCRnI#$Q}sI`z_W^;xDc{aQVxE<bgS#s0ph
zUpzJMEu6BZ0tp)H|CHw~EtD_UtW(R=v(-v{xmximHdgoESH9<T`RrMiOxWwq)+Q$B
zD_+^&;HY{-U6V@sWBg31XBr>Hg?g;Zb`uF=L3}h;kIlt{gclEzv+?hW50kcCU6~jU
zGhXG&>{NxfO&c@{Q_O#FEEOjFGknj6>DtAKrHX$Sb+tOdKgMM^uw0$hLS><pT{S02
zwzln#`=^Od@$LBRjjn2cnSrguK!>ps?eY>{(o4<6ytJ3`vR=;1dj+qHShv^X^?E6<
zkH3C?2l%G>>*H&&>AjzdO~<`KZ|IGDDVdMwd%YKgNI)K`=xW5<&Q-dS@xIlel^gL=
zwvt=Rug5Bd8$5G5VQ3~!dY3~prEZ5>N<9vxlzJT^DfKx7QtEezqcq?U2B`P0$`JM3
z?Hzw3RT_4+doKb|Prjjgw?!#0d8fS7mNF8hTmXpPV=3Ed-@V>_v~|?G-+KVGx5Inc
zd(eA`l%3pr*!vb@ySzud_i%5w_o(+6zk9s*dXMwF*L$D$1i$;Dde3=Jz7gZeMen@#
z6zTiD_j^zCdmy_1LGM}bIeYFwZ`^yHJBL0MOT|iunHR5wsdMLEc<$|gn`D$T^GwYN
zKTOP@9Y4R4WBHt#XDwd<&<O~<ttI+48DnpI1YbSz#O{CpukYXa*jGO~{8yj)S1aH5
z*jK-P^Y?%2xyjcb`)Xq*M3LWnEKC9?rizVpel@Rhd{tj9V3yZoH)3n?bsiFI_u^M#
zwVn^eyu^+8$K%2FAXZM)lR<1QHJAQe%u_7EZ|V!F8*vB<=6*R7^b%%+9>QEuAj}6{
zgoU7+uq)U>*j?`l;=#y`1o7T_AGP#q9=?^}nR(LsN$K}X^#R?v5hrD^p7gTyq53c>
zgKOK?>E|19o}BwaQcn&>z1Xtyjc|X&uLPsSw+G}dj|M}8JA!S5JA<6^`O86$t6jlv
z!rj4M|MmKwAblfVd;J43-ZE`@zCcZTgLIGyvR?izNJsK1t-toY%0+t$dUAUkgMI#A
zA-&65q?|ut^{K`CtUqIO#b7W<28p0A=nn?G?zb}ZTF+{+zQ2Be)Z$G120gvHzkYD;
zQ2j7(yI=3g)e*1vE%nw>uW$9hOl<Wa-$V6dv}f!_{9Np{(N`0zhgXjThlBlBV*Y;$
z4zAth^#@0S1G{3Qv32#(k<|mip|!h$!@9a0vaSc;hy@4F(UYqOspszc!L>ud!S#53
zC^&c{?hV;!q<H4>U?<^;`pLOd!7g&1X7vxhl~_GW{KRMD{=xd0xwFLYVGOrfEAOdh
zg4o)<-biqCU1PfEv+<>q_4{sUe%w#a`ziGZZ}q_1gX>Cr$OAU&JGOm%Ea^YWR16c>
zPKWU;E2pQHeIFz{f4wN;0KBrav<Nv-ohcq*DjxuUc*O(PPDha|2UdC>F21;2tttm7
zYNh|gl3(#=9d;5%xB%wNg3cxvYn7G$XDXE?WqtCw$7>Lk#V~#n$lNtM0~JsyU#eWM
zg#|shY{=wR`Nc}DQn$xf7AA!4mW1f+X7OrecIINecCvUuD3Sl=T4j29zIYYvRP@2o
zi@qKu!h>cPso;dY6WD9&($XTdz{y+Zig&j@#aAAGas27=r+A|$UwZPw1)|0Am!B)s
z*y77CoO}GjlP?u7JYOt5`Qq~(#l|vW*UapdO0_JqqZambQXlyi#nA0QO&PMtuZM-`
zs#5i!B^IHcE9KhTLNPbvVcM%-2Q9satG7iXOetu53_Ei165rZRf*Hp3wft&4h|d``
zSxc<Pz9(@Ra^7$n1W8J0JWNh4Er;pq^1{+}`aN;-beK|OHzK~mBlD16koZfFeumn=
zrswU9t?U=jJXwZ}nyr`1Cm)$#oSK-gJxZjJdo1BUMPC1D0v=G9o}HgB15uYN5F4{A
zqO^Ma%H`!*#s`9Q!UOs)&Rm)B#~ua|FZ04-u2Fy3Z**j(Uh!vFCPeOp=_Q7=9_GvC
zsriXotz1^_>cm2YSU(h(r=FduxY-l-OwG?Ofw+y(^TI9>wiTx0^>RJzn-&cjWoO=n
z1q1eFHIl)ASgKa$YvCw!s=UO|=<!ggMrksuYGJ=fALDag@h6tl@U%@!e~?!Xvnt1o
z^M`enw>!EHlk=5om{Na**~#VUX?TFJN0}SMsvq_@qY=Fvc3n{$)DO_D!%XX$S!>Mv
z<YL*ag0OF%ri*%QBxDyB871;}89~f|P0!Z-9yO=e#?KMtmzL_`0E1Dfmzfpi<t3&`
zrKY4A+F0?M?Fy6R@kQQ*iRx0A7GH8b%s@%kD;}t_TEFLP*w>uuBDLw0e1%#oK6Qq@
z6Z1<K;VGt=eEuQ>uU=#3Z{PH&Mp<FuvQ|OZJr9DemfckG?^YvHEcN-YM~!Yil$KS)
z{0rhfG;911D`Q`?e%Ifu1i(|+y*ve*Bf6U5gu4Q;EiQyzFWN7>1uf1g%aUK689U}5
zp{i0B^lLpD6r1Ay`$_iS&|~tz6mdTFi~bYJ3mY}z8HZI?DSufxFDa-fm{u^OKtzQ<
zrQib!o>B0u0$WHm#e4;I0yoKRGSX_nOaI$-@3Ml63O+$l(oAPf**f*FDET`H#7%0P
zH4V45hMddBu=^iVX`7DCJ}nKh(qINPc#W|u<*Sve4%n5t8#&8j3gfe-o{p?G1~=pW
zN7M@AF+>6RuMynCxAr<gVSj%-6W__V{jWRTYuDZJz41cA?pT<!us@!Q_s3K50>9n-
z=7{m9d$&``erikE-ynAeEnnjG-uP%dlgK2LHXQFx^d_v1+bcaCaBV6}n;M^9A`*mY
zOl!Yn^jZLo;AeyQYPOoM!*xJmq(r%R@wd{eSuYV}M`NqGdSWgK{R7nx?#|Z}YboyJ
zJ*e?dBzDBs6RU+FJC}YnzM8CO0;nqRa+Z2T#ia@JL4GdnC3gVV=dza)wf905W<(X%
z3-zwKZoWNUmYSjU)_T1hG+*{c+_1l{MFZyBt_xg4KQ-@n1zohLFUYU+F3r?#m#TJ(
zifOykV|V(uxYOGzv#YaAU;7=ezf~iYQfzKuG!|rsp>%`7Xv}_j<}SVMfYP_eR-s>c
zlL1=Z-PV#p*AiMk7<79>>xqtb4yzyR%?j&qfv&x~pt+jwxGm^%&$L{*t#Zu`c_Wt+
z&`eO%bHm>Dp_m6HKDQ07vMb2lP`!GC5w&|)Y_+HQ5LZLqsFgY3?NH72uC?vzU8{d*
z<4SKx{{5DJm*szqdr;5c{5k7!N~Y>Nyxqnd^mzTujUF%gi2AJ8+d~{-p(tl+go2#g
zwH{yFy&mH|UKL$z^j>lNW^yB#5ryH4xPa(LO(5%e`!=6jcRO%hyf|?M$--iBvQk6<
z1c~AmVTp^Ci78|ektoC;r+An)>c{^M#fLh9Jy5iAB^;_<_9IBrLIZm&aS_r7;n;G`
z0h-qB-9p-Tyt~5+`;{s1eCHFxU9DRSXExmQUr-wdn>ZB;#Yd1gRSEaD()8d)I^@PI
z*vk9r!T0_7?)A_A!*2<PTNzFKHMLloKC@-&StGFD7767)M(|a>eEwyVU3{<qUaoKT
zFD#p&CL*~(s@?Nf<|iY3UOUh56d*HdLd(_q*dD{>{tk-zA0-I05n}Z}uB1~+N`i#_
zPwBFSS_Ir22F*l*u<SIb3heFwIh9GN-yl;KXAJFx*#+jr!t#PcG@n!HFDkHM&RlS^
zv^*nJGf`SNG`?7WT7<{~t*m%Y`aU!WBE#xL)&HOMfaJvd{LNlNs1C&$4i&-J)=EQw
zdKhGakq!1d|MN=8NrorEvTKMc-9W$k*GW_9uT2sNVH6Sx5JN7}3*gTs3Wiul<NXQ#
zKqDi`ogj>zAe!N1CZ4#Ri>Ge)C-;ILhOHE6<92s4ClqzNJ8=kv(@i{gJD2RH^v?Lo
zh#`<Bq0u1V{s$)cADHC7KPIV<tRwOK{~eR~e^Cpsi9P(kq|5JD@B<1AC3G1|i0Jv@
zP6+K%OFK5C3R5(083^BS-}s8sLA!M)ybzT?@NT#xYSg|Kn%ICOq73_688+b&m&IX|
z(H25zVhzUw9cta#3a3Ps{0N;E?rddkVU?&MTirVw6_yCc|8WJvF#b;{xT)a13VxCx
z9BfA>uGo41r?|5TllXsCg_>Bz|7k8uLmiBkD-Ua6l2a$nh|>C7LXD_#O^oCJj4J+F
z1wu9c|DxdM6c~f^*Az2$=5HzXs|x<Mf-fof>k58e!QW8u4;1`O1;3!cIExm0af7tq
zR{Ad!tnA(zI=kj=Qpgl2B6h5%Cw_;htC+u!FusOFa4xx)T6YpYr9i3AB~9K3eYG6-
z_d&}`P8$CLa>Y0T6EF5bB)hM;THa%|WIJkc<sOU5C0x0Nd``dQj`<(DW64ry$*VE_
zjU~rdc062s35AhVU?Sfn!-@oEE4vVInRraMA`ZB5b1dhW)OlVWi41tQ;?F5wXbNLc
z1x_t4Aw*g9%es?=msCDvKjM;g*YuMc1_fCFoMw%pl`hE5n)txP>*B_G;m3O82|~x4
z!LNNImB_p!KCM5V_(meJ(rZK5tY9id;l^hhksM!R(K1Pi)KLbR9Rdte%1hQ0L24~2
zAvVI=p@5VfvDH*7bu}HN5$cYtW`YF21FP8}vz7|7>*#pusA1MJK}Ku_qSaK8c~9)}
zzRMEKW+AQLQ_o41{zj~xUxQs(>ym(Vt(z%AO%f)<2=;8g=Vf6Hb8jVi+V3Dr&ch7m
zgVdF{U%vc7o{`@HNe^pRAndNg+#!_Ci4B4gylmodyYljvQvN4kgJ3JgRKnUsG4g%F
zgyFDsLO``gg2GzgI?RWweYMxs=M^p`5Tf@7J+Kg6J7_U2>R#<bsQ~+?l#!U&jG*t4
z*ySIpekLlFzm(?PAKXxG?`mI^Qiy6AUhQwM2_bx!$Gwr*v|3B8LH~4un*ZMAU#!R0
zVD)qtMrU9{J-yU3&?=p#O<v&vbSRYlx0Dns=t>T5DB06llBW&^{g1#DT}}ssFys7v
zBB9o-4)Sh;Qh!7}Eyn5IU~p~7>!w9p+^rqv4F>n9PpwyJ%Ljubef<=3W0+ZnV(+a)
zRm*)_>zdIXS{-Iq^%<b(=V?QxS{WcVY_Y*Fz{<nM*&VOzQi_&bro>P%^tqS=5H4?Y
z?>o3R91J(^HGs$1wsHSOmIZu6H=-GYXK1P=L@gjAMIkRK*sbEZ%6!G->&TLdA6<3Q
z@>YhE9b#wJ78<lj>@3NqiJddNQ+zcM#M=bUw)W92Tp&}Jl7u#lmp1?e|AX8-B2bVZ
zG8|huYKy#6IJuROdF8<QS*H|O#6%^6Y65adr5cwv(QbQ{+LW}u^@(|h9L4q-t7~+M
zII4^c4Csg6CO|ZG!L5u_w%_W(0tO=w)Cf|1-0CkP%e%HznS!$^N?Wng_gr-U8Ll2K
z6jz1{FSoKpiDM&?%u+HzziNG=(l!)+y(Y!JgQXd%n<wf{Xt$t@d8`2Z`r;zO=rm&K
z*=pFY?)03Ig~Xf)kX=e7Pxb{29Rrx2nukUT(?~aGkid&e3)3?FsD=6EYVGoJrGmJ-
z&qS52K{tRuw#`BFR8yqspHq7hHP8PSio`F4iP=kGt{Rzp_`j*tw4sSoo{=e=VjTky
zWQA=}%Me-yi~8SZwat|Z)iQ!`WTx|8Gfr7JP}jv_q1E{tr_xpG*r}-`vZWWMPEx@f
ze+82wCSFf`BbU;Zv3Iu<Nf%4N7WRHUmr2+iWP<VCi2`h&N)N_IU<G@@PlNFziX{hO
z^?JDy<2X!-VHnC>BJujp1aipRhhQ%C2ll$XGPK21p3?GZd@@9|lniOTk4RmZUed()
zT;fJDK+1$<5K4m7B(kz79kiu<spfUsNgt<^vUZr*OBa~gN0HA{l{(fUh;5J<G^}Wf
zmyca5*NZ{%TKV+Y$!6AvZ6ru!68QCpt=Nahi`R-z@I86*<ZDGIE0D_0v?=qI8Y;Ky
z(lYwDjkQGOAs8nszDhJLLW*WLQ#@vpp5mFYlg0N}{L1m-X>6ddi9+RwB@3oCZB_ah
zjri4vi;oxU%Lo-HOU~oP28n&V*iKT99WNqW&?BQaIa({)%xhI>49i&Bm(tAtf&xiO
zoOI!bh>c|;A||RKjMp)c;h7<Z0ZvF%>TZALhM5@lT`n8x-pKtS%}TAv;p{NfOd-B9
zY;$8nwhoF$6C=%6ljCcmlQvHdOOa~ZO_$^_iszSr_2V~F?%mcX=H6|c7^y~^B^iDH
zk4ZoYDD$U{tsjlFw-HHiIwH$h;WfUd*BE3$ByJ~euk5kcxKsW+zK{Q>YS`Ba#-f(~
z9VOT({EA}lDCnf;5z9HH@}_1v8vYWE^U-|ARt~;vqauo5-P`!IcXg|2jZSP$kRPN#
z)QSFW;?{|dz1q;$om;eZr40HgU%5;ZZ7vL^qE6g*r}joA?%0h+K5v*7Oo9#50w^Lo
zDC=%NrCv%a*wEtUprhe;c6!m%uPd|lYv*8Z%-7LMv3t$QD*_~HC6l!Q*`ii9QW7Tb
zZ$P#vbv13d1mCP2#JOo%1nJ!L$+(IBqZ$pHK0l_|7Zv<4!Og5qdH;`;&>XT-Zzqf0
z%D=H!x-Pert{<Z#zSc`cjoIIyvZdS~E#+pNckBs2r@pjlAGQ1Msa@9YzpvP@D)_q!
zu#Mg{L<4P&akKbM<?14cgpZptm`<^yJ$Y(xj7)8s3s@B_<$?s-C^S!VuuQN6viSk5
z09w#ky_H-|8wrcyHurV~DVU|Ce=ta|q1p1%POkz!N~;7mMafr_tC`j8YEG6sLC)|}
zcHBhZHxq8r$eyZQ68?LC+ki5fiO!({N{jP!A^`|uWWi|t7naI?fs3)cgJ|(<uGT5C
zHWclP%V^iHmC*}`+;z%#0c9KPaHKg}QYbfN{2Rx=Z>bZIe-96>4f2$3fcPL^DHyxs
zBS@1YkiMb7l=`~y=_b<b`;T)0>5Q$$7iF;#|DO1Zup8eVUrnp;n*E<dqp1GOk($R2
zjQsRz>PB+4fDQ|b!b@ls*8%)<Ft9UmFC|&P988>-kZ$UW)cRd9*hm;iFCFBEV{6%U
z(u0D73S${r|6pAo0W<ZW$+9ACkc<BWsR#vZqTZz(BPy_qSeQ1&f${%lLTU7B68d<o
z+bzA$y2EZ~!dy0l62>R}-%@>6=`fXIx+`PX&Oy;+3A?L{RWn##n5bRaI8M2xvNJLL
z0$FPXs@$0vfet&K7>!4o-A0ZM3QWjME8+>hwKW3zoi(sZKYMW}AV7nQMJcNwArZk$
zd^J-~)zft}z2wcxcFb(@vgmll${)3{Lw}3z7j}L+&S)Wgm@9bcx3s>xv_7awBiua#
zZ3mnFKD)|#=!{ukPa$~d3DRaWRtysUxb(`?Sn&jD=)2r_V`S84w+d#iEZ(H<OW4Kw
zvc7zt_*mST6o}&)>-D7zaX0TKX8eFxH9jHCA#a6;sCfQYl<89n_GySeshlbJhZ$Fu
zr#qUt%m&dCHYLJgN1!>F1Y06o@0xqmZVg>hCgeRP2yNaK!M|}7dm3ZJ@Gma?n1*sU
z*&whP%1pcn>77FRvYXYGcwOrLRD5O7A<t;|t>D}J6zUjeExP8yJHUm8nL*Z2YfuF0
zQXrY+G(<HEPCFnH*x~!)Oj#d-lg?0MObKZRbqbRch}O(n$TGHNJFEg0h;oC&H2dc-
z>5ViwF+kvrq>Wm^<{Wb>cfSKrH;9rX`D+gofC`gl2F%iaBX=VKYoI9z5O@>z;70OI
zmbInu4lX5t7%&s)<ha&#$4Au;@%@_GgZL!=CPJ48t1^jfiWFoJgT)Qj@%A#9a2RxM
zy;$xMfH4~t9{pur%`C(Uu@y(pT#kmMoq2b8=|7>-jpVneE672cR`wW?sJ<Zzdm!6t
zV$LKx*QooK)S*U@0*sC4a#kFT`&)LXt#u8H(H5+2;H9`NC=g3w8bm2jP9msVlvAU5
z2Ggu%*2WE^ou0wK=E|?Axkfh?Y=|wfqd{rfpszLPA`#6>L%&0!=||0okNF9FMs|rY
zAMunE273tr#cA#wswZU_jC7I`hQi6=(&jA9j8BrTWIgL)a{KW(Kp#WrGyc6i6Tu7|
z*BJXlpsttAWOHTTa`jSm@oE*m3Jy=kSH+_rZ5m>&9%b>rp`yu&$(sKs#B93wZz`?O
z*p@IcALbUcw}LuMX5UygDbdDZ1}2j(vEqU-WZKY;IrW%s{6mfMD3upLwYKc@g+EEb
zN2M~uUr(iS??e#3k-wvp8#q_<y(ZRp^D+_`cvIYU(n|uH0SHL>SwwVadNu~l$@212
zbol-tiM`KWyH0aG_GXOh5xdTCeLp2r)Wlp`SI#^_Il=j4!)h8!=NQ8R^q1}bTgkaB
zy8*hs7{@9ZAu;!I{)=8uHSYDU=D`Bk|4_2;3yB~ntUZe*3F2pqlNSt*$QX<gNRXOy
zm>o>tDA!N9uBeRiAqF-Y8>%d-c|bKcYIQhfZ1B#Y!at!YE%Mc8yITa`Y*pIGX-WBK
zxbc7J=2?OW5=iV=t5s^7Awh%OYpwK0Rmo!t-b=7D@Iu3mFS7nDt_+%i#C7BvO(Ti(
zKx$1xsuQkra=r(Y5D$~y>;m#XX*j|pMyM;@X4oGYO0MMH6*d;P(iR=-alq*wy~Lbq
z`*sEF-HS9&k1D38VS~`L9CRj=W}asT4t|Xc4GO3en29JA@pIiDq^2;hzA-th|A=aS
z4?(FQnv`WSJ$p@dd}MHTou!;z&|0@x+CXBXlm7|jjV~Nzfv{Z6&Ow5%0@s6y-59)$
zBBUBbc-5O2Ms(FAkK7=2sKNxDGJHi%)E*&_R>cS(pbd)x5(xqaQ557+YAtOLN)sGx
zIs;x3@j1(K_z||_fKii<bn%NcaV+alOjanSX`#|0Fb}IQKA?;a*&2z8X_AlCZRXNd
z{W_9=1JKAKJLRtThw28Q&;yuttaKX$j&5~O2FfDPU@%ydPb>_CA7Zy4Ifw4X;BlV_
zKua|sA>7N|WF5#0*%x2S@aClOj_GmVCSKy?_^rXo;>F_C;#Bd(qs2*nxL`u|EL7Nh
zb-nC_y!3-8W=C49Az=bFnql7osnBL;mKQOIiX_6L>X1l^lQ!zZ|6P?fe6*n_@&bn@
zo%>Vuh2WsbD9Jx%3mIwCnwd6@yL!fY#nDS^>KR5Kpoj`Ifb2!$FJOt`qli<33lI(;
z<Y8vr+8DT(m0%6j)b^iNu%W44$mhTb2zDw<$pnqnoa<%w=He@kyKc+S0o}XwR*i=K
zb0Rb`1{Hpfpg5dT3MG-~c<E@}m4*>j9-7*izr%x1t;bOUGGV2zGS%0xQShKu@wP!4
zxodhxTy4U?4}R8i23Zt6`!7R4qWy(}SWnOyIbgOXf1Vv(?l1Qik`~f{_Jp}8wSdwu
zW9fN+o*D}4;tEi+dEK;4kUH!21ceQ`b4;$@@i4yRpQ8uE2*?TCL6u3qc;RL$%wbeL
zQ=ViWlPv1BFEq=pHFay4BT`vfoVv(11-9a1#;)Y^#!NsA_@&IeUk0B8w7!!{9V1{<
z^Sil(>%O?i4m$ta^l+fij00Ier>qki(k108EG}UVi1j@Bh;9=nW8cr7!P;1tqw6#o
z!n~uKXT57-PqRRoOE;PIGWL?D6fl73X3eB-Y?wMd4H#^%lhfq6i<U^@ueqMWcx#l2
zgzW#F{*+k_-!5RpmARckl0O1?j2MWHYC0f<=Ax#U7H92W5-?<}1&@^wtpIloxnS2!
zwYB(q?7P_|-kI~$jq51qr{hM=Lt4C%dy`3fBmU{wr{Zk%0j7`=x0LuRF%KO<Ds}^-
z&T8&VtoHa-hyeY4h%b{xA{uQcv@M0vgiElYYE#;q7!PmE_h!5^%?w2I5uDcK>QMdB
zyGF1Y-yq>Rz&+Y=;=yP@`LNc$*LIxTWM`u7eb9IT7B27;Cee36X`*V1P@Zy>PrES{
z`4AQ!kw((?oE+t8{wpM|95d(*?w(~|0?JDp>NaAcS>%q5_2)EoOhRn)XP(Q=4LQ?l
zh`&gVV~rM^zu>>B5=oo>{;$y}-&2Cq6Zi(=W=Jq$1{N{R_p2_tSKK1aJ#8%ES5<IH
zfptOOh8HbrPK)}LnWB(5Xq;KS`|HE$VVm8N;Aum#j=s>`)z|R2299}y;7~JFGe~4*
zLt|`koHxBhE`MHuh#(S-FMC$ad`^M&qXpLLL59WF5YR_VF`GJ0s^MBa8JdhbY`YEP
z=6oHPM_8wcd5&`d9+m|~0?w8-7jeB{V~c@}B?3b<<FIdP{MJadfomPwtvbYeYjm<0
zMU{Sxt1)*6l$@&ii&uF?19$2tUoxQvqRl$yycYIWwHLlt#sIfWfv`K;Rb-p$H+INS
zS*Vq4ku+Vqea@;!3RLYjDu|g`8MR*7BJ<k<I!4M)sS_KYJ|em!%vZa{NNDe<ffhQ*
z1~ajmMNXInf3X@t)FAA(>!cAcS+iQgpoRUT37L(6^RfVl#3QkK_69o_;7}ld>`op-
zgV-E93BarOfuDq-3X$aMyj|rzw3F~Zm+D>hZe&-588{UW(g2wka!SljgnxP3S}%lB
zfjqYBK>HTv`Z0Q8XM#z^y5VgG0!%H~Eh%1VFu*pSR)KAUl>dXI40)t{H`D|rhZ%Zt
zQ3AU?tG$~5<#%-e${y<JlAS!Y_FApoT&4YAsBc>vsq^k@+hrVt+%4gH&IrGseG~rQ
zz%+&&Zy?nZP<{zW2R80=U^2`&6{xWuw=!pJ+JTdIYn41k5cV{6pPCP)+za!IX1Q>I
z#ecVQNy_A3Ro1S?lp<rt3TEcCK<@|_270sF?AH!<+h(C)9L+Ec-CU;z*IuPElhuqg
zbY;Tlw)L%@M{9!Bu~GjsJh4UEg-MjRj%0mFO&7*Bi-tDwp{F#hWG5T=8O4mKKcy1J
zU42|>E4nm@dR(!OlEzk0?Hj9MVR6RU8I1L{kq_~4^p@&LUnRn>NcbpQ3J02u*p#-$
zNvm4S<@=SUe)eBiV7>Bj#SDZQ)8%^MAuZGgX-WoGNe~x<@puM9@7~02s0IgzC4}rv
zq!Qh*QoAt&?}eJ!%aR^Uu&n!RS-1MNV=y~_oJ4!f=Gbc7iSwM1nuzSV6ozX0Me596
zm_$7@mz~Q=&?6CL1o9H&46-Prv3%HLJLI$}Oq=c8hzNT7Usmu51s_&$H$j*-(ko0q
z?oBMb(hzK^qA+8WTP;kIqisT2Ck?}t;Jt$lxaB74Rx7K!x~WuB%?#w0Hf62surAmt
z|IN&}e~%h+R>4CG)C>N%D3E+%EO7zDob#*?sHhmx3pDVf8Y$VgUKk%c;!h9>3uW!{
z#JdMuKfBB2%gYn<ZX>AeEwdc}{tZ={R$#Nt-aQLm*2YuUMOSt6mI44z%ugzSK#%!Z
z1qB7&1niQ<0>`O7OgJuCu+2d?olQhw(t!`?#<&9GhD0;_vkERN5XtSP{ihZCEJ4^4
zZDyUSuvNC*M(Qu9S?YQJ41vD2?;^;>n0N?~|LgvG|5f55)-Jab54x~Bp6SKT^3hy(
zF2i4rZ@xF5=_Wm=dnmgyxpW~@=*#W1`{~Rmd+vI2sZ5r^K5T89qFEgu^|?MZQROW!
z82GP9<dgYSKFx0;pCOj=N(N)`iGhEMl%V<>G5NBnWY==*4*qs|-3YxQ1xk6^HsF2O
zN;r=g1@nc$mPb!sFk+={^MJu7M(<T`m$w^#66}rg_Hw7s+vgSe?f3S32lyTE4tj_9
z9rO-+NBAA`UNc`8alBjDv#oXCZqJgFB(NY}F6@q1#(~ob=jDR<N$(WT*&f|H?VWi8
zOo5LJ`Mr>j3;Dc=;o;&V=Ji4zF68q<9xmkZLLM%h$BVt*g!#OPm-g}Ogg5EEw^GFC
z#^c`m-bj~l`lKFFtGGAi$^V7=<Ulk13H7SynLiBOIoP~I3GP(PLx%1gitZ$=u4(TX
z?^&zsuuJ!z^TzERjzrfJ-t*oImg8u%q{cAq&3G5Qm+j6m?}Oeeyv<lt=G(jvSsizo
z*NwZunis>qX8}&tdUNaaN?}2cZEQ35xPKor273EF{7m&|`a~b`9HN6n|J0=SHs2Of
zctGcvR|}+_?3pzG7(uVZHJ9SGZDs(AK`h+<z98%G@sdbaQ*i1DPZC^tb3k0?#YpL8
zcP4TH1RR*Y5@$9N$9R=^-r@rmFU%;<U_E_9+<uS8F3Z_?&?_MqbP4ia_HZIIZtUS)
z>8=+Lma$)Ru1De|<Oz7-cqLW^y6Df7kB;6CVYpkXn(vOjD$;_rey^uK0Ik9vMyFq2
z8$!UZC&bpEe4^MkWDiP%LMxB30~8M&&~13&u=tLk&n(h^hB7-bv>jAUo_vMu2>N|e
z=XTY8%uI=QdwphZx`(vAwhwh*FyQ}ceYe-Ibj$gR)|auleQU*awHq49*u{ZyCv|<A
zk?{{GI7ko<vxUFLwr3E4*Vs8<_CKrpf1cpH|GkPnq2PNIFr+AXYd90YPZWNd7X1tU
zle%%9U~GT59T^y&m*nS4yRTiH;5rl6hJ&436O^G*+2nN?MlVb*EiU=5s>0Xwq%N{x
z3u|5iLSeOr+W_NY<cY|9Y&BA$E*yLCU9_{5^E^s`2TNIaodq0Xu^XCA=pIUwn*TYn
zxXlehHn!nmgBz@t+ktST-9J{V8R2MqdgL$3<mlmGE0^Sgl6IPMw$wX0xrk}k;-yMe
zBA9SUHu;$HAz5qg^Y;zeT~fQs3WgOJjWDm+HG;7rBeqI|?cNch*Og4N(Irt5WK#`;
ze)Wbg?YsX)1?vjd6#SS18)560AJpaN6+ElJc**9ZXqyv;_ztU0hW#{VmqccK8xdio
z3|w$7q5a`9z;J&fr0OrBJ>Dk$Vc$0baQ(SE8C1zCt3Bhwohi2Rxqai*^Np3!4WOy%
z=AwgqqZu(7IBD+IUL>%+2Jq1)3r}c?lOmG=fKeHfrZvo%)-uj@j{^X6*>x7KNIg;T
zEU%Q7?OgT~iAUU`O1VV^Q~Dtu;Onh@aqqTF(7)Nt`lJ#}Sue_3?n4-ySr7AGgl_UD
z!ZsZ9S7rkPHFbmnXyd(2x;E!9%?)nkFzv_H+k&`zQM?bwR}O6K8*?eqQDjPlZR4XJ
zQt>3;+OHDueh7330fK;DZD*mch<#W(cYa?ZC0Q#IqEEmqVh`Kh*%Z%qcO+s4o-(*e
zV$(f21e(!?qR4_7#Wr6`yRINbO!BvGnNO_s(7OQQkLaG99VJkF^{oS1*V#BHP)NN}
z?xh-5^Nr^;K-<ROF^rh#Xt&{Q<-HKu7cp*?V>N7+i&M+&fLV~!8DLn$S{%r8qLqRc
z2Usyttsv8%zrM2T;K75%mtT^LnU^k{JAuQQCmw%RSEL&hS=rO@d?P1qC)<IeOx?!@
zd@Uz@oH&O)PQdHru|KZWWh>6sxdFbWY~F9OZQ-Evn}=hNrYn?iRHBQvgr>(6IZ~@n
zu<7>C(F-AahVgZWE0ZK|i6TR@t0fD=qZ`g<WL1v1M8aLMW#NE(oZm953*#$(=VHgZ
zNvAN$px94oK92H;W7tjRkcG(vORFDCY~sUzxN>OY4Be`lJ2D;}`*byVb9`;`K1jrQ
z=wj+g<YYF8>#zt+9!*^)n9&_cpFQ`o#%jZkiD)6G9qzeKUA7v-ZMBPXn^HAo*7UIA
z{}jE{=*y>+`TYtEw|+peXB0fEptIAA+4*JlQGuHy$!<(c5HQ_d*}t(@Ht$fTRP!UM
zU2ko`O%mXS04_*Ktj(J`Z)AzGG%N`z{g4X{IZ-&O!6X})4zSeRFMU<cTpmOg<bEYL
zNsr5oq&WZ#`(%bEz72<wfRZccFYxXn^}>`~7={Dwv(PNV%p0Kpgl6}@CZAs-Xjm4=
z-=ffv;j`b=-T$b-ut*D){4izzfjX+6o07x{Nocb`E|L2B%F&J8)sBbTTGJu5?C3PH
zvUlt>6Vc6?u9R-eHoHyTGi2mcV!Y7@_Lh<|T3<OsiZLpeK&dw6qiCs4KqhVK2mhDU
zJF>{J?HF5j!jE$Ix77(^+vx;tbLY!^IreTH(B77g9_SEh{1|mIIbyX(C8_|j57e<)
ziU>&r8s((klzTEKJ;XgtCFkBQ?VQUQ5o@_*&M5gW(=Th&FKOb)B+gO57E+$kkLlGU
zkUUxLx_p0BbNo`WcFqV^JtbNDfRVY)x*BZ&qKAV<sJ0Dlxx8}tW!vv#iR?zeofMl2
zYKz5*T2UC!C7twtfL>h5UubNfT-l`zxEbr9dL#Fp=h3VgRb=q^7d1MQ^ppP)1rrK1
z@BLrU4UMv!<6l-xbNpsx8D(Q@6MU<={UenA=Ni@n6cGvJw!mnkN-n9HZb~uy7)ZB@
z&FYbN8*GQ$_40m+dONVYfTMt<fr*PmaE>SAZ5aR~oMKG+!Ne)2UkK7`V#~nJ&X7_p
z%jI01`;2U+o&l!|bEg<t_GKcV&JxRl^A*d(Qe_Y%6A!Y0kBkIf|ClGDz!{8qg2f1i
znC4L(SXZ!2Fjg=NN3&TB2i^2-M{O-k=*WA~Roe18F?Goo7<8I|$k@?DhkA5rPvgOO
z7IUl9U|3|NV&cV8V#V_%p!dy!z7q(AJ5A!>Chc`3y0iqpukn+JS0+wu5kOEQZx#A8
zjlo?T$L!8kbu3Z6j+Utn**PK_t~$IkK5p3846&4`4p|zyZ&KkmNhv@S#oZ#r7Ac|S
zX;w^G*~>8;>@wc938!oKjWZit*><#PuVb#soJSB`3>V}nnvTZfim$hIk~E`hzog5b
zC%}2J{NIU;X=tGSntJq03Oak#|7p^`p{^4|bDfm|rWKN#A=;hDn~v_Hsum(#;X)A2
z1PNnKP%E*6f52RGqB6i8)B$o}j4d`DVJ?ON^$a<RbS{ufbiQo&FfFtDx^nfX-W!S9
zS%XkgwMgVkiorcAWiD)8Cu8w0hAPbBMGfN1Hy8rt&z2|gtp1nPGygz=_32S{VwU$+
z-Ub5@pW4uOj^H!lF}$LdeZ#w0j!9#4v0!R+L6B6YfYIF`8?#gKt-E%siZ-d#g)(g>
zevpXlPoW)<bf|mqlZ*WvdkbXWE=xL$aqK>d^EFajDj+4wDZs<-185EGG@P$3>^6+j
zaOAqk80VGZae4-7!B%+&(=yu;d1bb8)l?ODM#Q;xwD~K3F^pgJ_p7K;u1TETZS1!k
zUZ5|^rS>-UlVD;IX2}!)n^?(4){dS~k!DZ4&V}AoNK?E06wV;&;hjlwJ;4N$Jvu!R
zB=P>?-W;GQTa~v!g0gi{Z_t~mQP`)p^o-K^(l<fIdJ5lR<WQ(`4!q`8ez{PLKGBZA
ziw_qsnw!myH?BU+4q2~$v6bG~sptH81CH@53`o-=7sF!OmEz>04^MEsS*myyf8H0_
zWLs@zeh>QrC+bmF+jol#BD-DW*2(*H)73U57ejutvNj9A@25YdV|yQi9_B7pDofY`
z*4&h^iI8d2P5rvo%s*D}hJxQ9XxO;_rV@UOKwA8(QvS;4DC@+p;$QH$b)#EvS)Q*<
z*KwnVlv!6UmqbFPe8s~drBXM}AgDVS!m{g<-I_%LU&X9cPNU2v*4XZi@V>oBhi+K2
zNi4Umt2H9b7DLG}%h)-wwwx`i^ui{fV52=2P_V_!Ak8Ygnk89z_?+XK^)$?VOp9Tr
z#YoEvb_a?LtChRdnzdTZ#??zlwKA7v{TgId>!x}n8_i<5%2QdP+3`}JIp}2YBPUml
zgo$fWf{kdYP@k=oCw=^!`=y~~5hc34-;G3ywGv?_WDn`qdrGBBgSDl3{9u~Swy|m1
z#$Kq=sDVwx9wwf>&i9(>Im5hr9t?h|>)Ghx5<_;Pqp(LvrkV9FjqGloAc2+0>?3S3
zmLN(LTI!uqwk0#tdp3@AM;+D=9pgPng&K1k?_nN+oH_Pt%I&>EqK&RlHUPj9Gu7M)
z{zi`|Hxoe7D4pLZ9r1~om0!VlNap2NKu_coPlt&`3MUkJGX%2aP(61`+P&IiDQP^J
zgY-a-h*M-SfHd6QN75|AbZ3X4CrZ;#&g^=6Eypju@tm898}T)EGXdFi_{&QnbMK3;
zH<TTYO=~XpeKub0wbrKmKaE=Zr+<dl8gWW1nAcIOh@o3?@2FLwJ7(&s7<csE0~_Bm
z%BMHZy7+X0cOI}lX@6(;9KE+<-~8Q`i#<cEy3KvCsjfy}R6lJ!AO{g2cO$6Se?f1g
zcQ$&1SVwQ@j_VC~=T5z$e2w0)oLlyWa@o8K6#M4Ap<H?8-J&-()z$2cM7<08v=?_A
z%%W*FLyYBDKGDPujXmxUgFI$LrOMSW)<Cu6#i`@Pi^q#s$69%t$jmZxhQpG%t9C`$
ze;$<rBH<?3wgG+k^lQg)3DLSf^IEIE7A?ZT16^&o!kjld@7!{5Zik0CG`?uHUEiyO
zU2J1ygT32i^y92`XJ2g428Qw!Db6^<2<JDs3^OlRr*WSUrp^=k|4i`jE@d;BpUIF7
z-if{*8+1ZD5jg%oSMYxmgo)+pFyZl?tl>>}%5Ne-|DP()|4C5l(la6e8fKQGM=H?|
zRYy;Q1h5w?b+=_dyAf9Q89}c4Ts;p9T8cyIO^+tcLmzxlF>we(i$n6)$JJ;eg&{`f
zos-GqiQVzLAoRCNO(6zb*|!x`wdS~|P33t~<I_+x(t=p+i@Zon@CUks7d3_207ky3
z`kZ-GkO#4X905#H+~K*07)DYUJt3Jk+i)2=VY*>ue}H(t_5&<hDJAdWssN`f1F5b{
z_(~yN9*Q0Z9fMF;yQ|x@yh(XK>{xfTx7i*Hm734(uHvp$&Qj;Pk*25oWt{zY*L&eA
zDi-U554l8(`ej^&3sJdMCEh4+X^Eg4FE;M|z5}&p<NY{PU~a(weo73Ab3tO&i;o{1
z{OgTqVLuBE7o@0MQ$gQslG>4Dty13FGvNP~&XV#4<fT84$rg*dE9jX?t`4w|>F<yW
z$E~zvFc?_d#=Sw2zKB?J{+C^fjLNvq_`lBeh%*xl23X+Q!ivMG32cuSt0i23LtjcX
zhBr!S#KI&8+29JM%QVbo_@v4_l)ikaAbl)fb3eF3_kT~b`}YaPwmWI93Cx?aa{sUD
zzW7Z4XB3ze#xz@>*QI8Q^Xt5>m@#3WVH!CjlYgYNi(Hnr$-H$L=R2%<Xjj~#%uaK(
z&NfUy+Bh%sgO4f5-=&<6-yy5#c3wL5oD;PFE)|A5BMYc@5taiR*ep6h#gr4_%d5F`
zL})Xcn>L%C#C6C23XdzX9aIiC%1&nQH(72D2$=GJO)t{ysZwf6l_j;?I(3cB)*Hpr
z>cXnlvn+snF};?mPJ;VE#EsgO;v}#ct!Shia=$W6%+z!kU&)q;y`D+-eNskjy;OHj
z>8BV}(~hmyx6yYU5-ZVn`aZ$;YabzCYwbaJ^VoW{=dR8+K(3Qq3-Q~wYPfb7rn$@<
zS||KO$&IsoLt&$RBcY>JR<WNZM**kBa@JzG(L*s)@xzfFe_Ic99EmX;aR;zpP#4O=
z;bylnCry5I1}4Yfsk>6>`4SR_xk$29DGdrtO4_6oc^J^Tf8AtMVS%BGc8Zv+L~0u4
zYsd^GE-7_4TvAL)|M#o*5Vg+e)?<oks{OErO1m?$YXiW_FIz9M&Dy`sqO?2qw*RMk
z*w+=jqacZBWg`mMy}7=!JR+Hy)=xJZ#e8UM)bWiY|8AAsS-2yPi$OHbySWDzaG>rt
zfyzy%Io)cTjlqP!srE@;1h(0vCf%%G+foAX48?2GpaBTQv)NV~_Wk2)LA;mKIbqG4
za(ud($@#C)!UntYGI6=!_bS&+s~#q1>aNAMVr}zHCy!#*?7ytGi)my(m{!xSjeXj_
z^(IZe#;1#jm=94Dwf{t*E|kc^SVC9@5&Ymk$@Zm<&B*P58gKcMJ7G)SHD`|XC`al!
zV;rLTOvA1{SNP*&^+^Yio0EA@+@FC|<+pY|GE*rSquM0{8B8UG8D|K=Iy;1;txdwx
zbG<OS1zN%$FKLD`T$a(YN(_i4=B^~rNTz49K|0#emWg(>VaI5)CAvOl`@U>%n_2xr
zT762XMwk7LVn(aAkcqAEXEeoHc*GEgVTzyEjh|3(kAhPQMit1mu(6vatA%8=n}!&7
zS{y#$Ccv}GD%%RSKH`%f99J*bCl+nX96QW53INSm@mDnyBsDw&J%xW`<4<KcA(vN&
z;w#%X&Wy<Tq{GF9bOH@1Ywktrun;C>&QwP%jzM~jKF5)Q`dAF5G(;m_0>dmhM!>bm
zY;;fk4{JpY##$l33Nn~oacEbNs+~bAfH(q~ssst_yE~vx-lw3bfS1<>jSXnH+1K=l
zd`UG_t?yTd{V_$tl<;6!h)|$l<$~@FDd=23VHc-YG>)nH4fUR^X)r8-w@360W*fHS
z1=wV?X;rnA+ypnZ%AVl6m3tqsOn|XHJwe~e6j)o=chL(DAbf*j4O`yZ#O;+zT>|~(
zYslgJO#-M<?K73BH^u|eRt1;wUo`-O{Ag^Y7rax!4nQji(G*EZk7bM^thRP^n4}vx
zftZt#=FQ^F2Dh7|wXrS&fv6zxyNNhcf=+;79Y`WSjca)?vmRFujAbXjg+XfaZ%;+8
z&p0jyU-i#Ci~BNv!Ti?CA)2}7atFGw%^j)UiAmYa9G4d|9i->s*{oUB<ZQim0!0KT
zM=rG7h}n@tR_WA4^(ewhj>wGmzfljGz}dPxK*{6O#;PaRX695aQkS$UYH+34seMz=
zYi_eaTiFJeI#INpH)3i(JpP*A7X|t<2mc*IuBsQnF`9g<@fuN1ygo`B)2ge*p*lGB
zh|MO>w83r8nNk*Oc6#lMyFS~vK5I^f(X$zRGjR7ant2BOjPbms%WVn<6v)Kiwk0$$
z*B>e2?#2QvWiJD=<f!MjNH6tqB{TYlE45PY(mXSdA>Ftd3of}lx#WFCbGJb8LT)$U
zORV9_&W+%!mAwNY3wmkR+6sJw$V$Hn<-rW155a5kw-C__i-3WW7;=upG5ElmNCbpz
zy`UD?p#Zna{ULeuOlmU?zi7lSK?gb!3rda+cWU@|z@IE!5OX^@A-<zIA)j<-SvH2)
z{8dgA&zv|{#0+uideIK%vANzhEkF%y;oR)2kIc^7--hz7i4jfE#!(Nohl>+!&H$r&
zSulwBwYJ>TxrmP3+*Hw|k6F-^j+8Oae5!6<?jU^UsZVqG8HD}JPMslJYaDu6aTB-M
z3&lHpnk9>zGQ~ldQL{L8X$C!mbd!zaJKI{;T1T*2oM>dyA-Oave%NhklicTc@yv1O
z3y}lHd~_PTt-S(AWYk=n+xpk7qURYnFt%0(-Z69ns*O>b=2?iHGY)E|?6l?avt({<
z3>W{m#6Ec-`)_NR$Qmf@+t7(&5<9bvYht*(#i^%X)9UD_jzqo)qB@cgD>it$kl1WN
zwN>#Bd5Ms&Rtr_eYD>@~S_SNW{rjv3Xg4#QM|EZXYNXr5OAQ#FBXOXyB58>*J_-ZT
zVmPScssMeLT<tQu09pDW_rMAO&&S9Nx{);?{FBFCNk^jeUVQ)JOSp?2t~;YU9G$}!
zYV!1Pk6+q_p|(qd8kMZ1hu>bKN;A#2LTy@G`whS1Oe5&GwhjaXn0!;0ZG(FrDTqAy
zYWrFrG^yOB^>D?ZJK&~m#0EJ0us0Y&1;^1gUKf1h2(TfA=V6&_H{MoP(c4l3$J%gz
z1ZN;g2><P?qpLfD9oFtH**G@Z3>CeDV;Lnc*&ggbkyls;lC16wcIuGE)xpc`n)M*U
zQmsA1zt+ALj?Lv9Tm#37JCIfRu-?w$GaM8X#dlbIH;zwa48GI<_g*iQS>NhjWJ4T6
zgm77JyM1oQ+^)IZt8$9I2Rd!H(e-<SeH>)k|CYP~7gtAvJ!|`dy_gNi(MWOD$#06m
z9&={(XQ^X<TOEVcv0tX=Q9b)zJ!3=v?G8)2f%(p1$tK&1hFguc?Z%PEBct;D9<9d*
znEcM^2v@Nwz8xQ}4YTbzq0C?B>fQaP>}WK{KcE8QFvjA}1h`KLk{meepmSV8;|Bbc
zIN}V-6|u2%;#2c4D(?gVXABC|&QDy|0d=PT@}DGqC40&<tFp2Ej!pTCDq*gfjOpkT
zdl6TfU=oL|p+B?J!CH{XH_qlI<uO!mOwH?R-zQZhuk+cQ|EO-|F`I<nQ0%yZ&nYl2
z`)?}tHx!&FFe%lQvSdmbr$TeqY<e00DFuH*?PCX=@Y=W_qak|KU=i_tUa^7#Z9Z~N
zq(tJEBKLBURNb7=mHM2^xhRb@Fwr$~nxL}|=rY&Qw&lE5>TA2u7H8Y!S;|;7XG<d;
z^_Y(rW4ey2#owp?NSP~!M7`!`)U09M5S8QaQQNv*dAHqI$E!tZ%94}-4MQVyg}@d)
za}MM0R5fDlOBpwm4gs82#m7{!*u^aK$?WC}SkCrC&0{YM#Fxn<E({Zgj6zB!`qQHb
zS9Es-)w4{-5`VJ$-}Ua7z%Dg}@C}AgB7(wgSqkdpG{r}WZ?>CnEF0UHFeOT?@o5Oy
zpCF<3dj$AY66lxA17b(wZ$K#i3XQ2h$Q{rJ#^58y*5|}BNQVepo|wivQ;;{-u#26%
zV?h^Sr_037R>psT($Z@u*<UP#(^Hio9&<w&&c}i-u5saVL+8(?{0{^@=)KWk$}k)u
z^S6;xhG`%PfHC`v340~_COwCCNm|jY0qXu$Py^4ugdltX0NyRF6MKzn(RGhq+dcDy
znUiC=UJaZ%J2D9J7#-pFYSe6TYnfx|V+&C3hiq!hPP35(ql1cDlhWV#{VHNBD5eSv
zDi_Izn{olSA_SGRBg6!|O(CK-3k);_jIC2L@df4tz^6IzT#uBo1MmfFQ=L8-Ma#|)
zPbsB&>bx1r?Fg)f<CZnEIU(Lroo=#!QH|cFMt8E(J~VXf`!r9FlOv_m`4Il%?vL43
z;L7%IAQqMZoMx%^eCP|!vPr$O!2_o+E~aBE*>(~ny|O8&2XE<ylOeakhUSlqzRj5_
zldqd8nn&-E@%MA|Fu0cql=JTj`cb_PP-<{>xSHk)&4#Y(rW|McRT?LAIN)3r&QEe`
z!;^KBEHj-P6_8lnhQk?*cyQ^NH<z9T^doT&VV2T7#^y1c(KKCpj_~|NCY-i?4G!CR
z6L8QRP`5o63~?}E8d@Z7m^SBs><;;c?f#Z^z;hJFnS)zQBmc1G+7Sijq{IAVjB&Zr
z|B^c@xagrQ<*2li*7^R2HQ+*Hj=|in%b(EY<HYbxX`G|s4~<4Rc&-9?!xRhJ{}qn>
zly$`}WjE&XO~oEkzO0@CXmo<H9jetlIG6zawyNIVrr!KD-8bZ*rQq+<W#@~8UG>TW
z?C}J1zaj-DzOl=VXEa~ZZR|IwttR7P$27DxH^BwS{kzmPwxnSTWBy)ylpbo6-r%w}
z5dWO!yXx4P!b6Syq2~B{DvdD+v!8j->>tF3X$lyfitk26vK6F_d`F76ZhX{R+B})$
zTN9`W`fJiKN?Qj2Tphp}0<uznZ}dC(COR5t60u5rMlO+<*V=T>vC;_{aHOCtKOIBM
zC#~ffc^G6>K(U)+)FJeby{kjm01x9)Mm4`y&%7DOl|;gS`cv|1V}^MxuQt_TyN@i;
zf5mtN0B_@NHLW_{G&=s%@n{R<G~Dq(fF9Gdh4p$~M_^)_GGGo^vmETrNh;M$dn$Bz
zw+_Ip_W<rc8RU^M4Y2z1!8WXuS5`-Ohu&Zi>K)2_F3nY+QSUnDvETm^;lSJ=XQL!*
z>yboxDA*3y&l^gf4Q0xi9&g5j(Ry-i<V|Q>sQe`Dg{}>TNLM|hpflPKj9>yHgNJUf
z=dFVA4m%=E>%Vb`uJ+sR8E8Ip!FDvwQ1f;)X0Hws4Tj8d-w-YBj~s9ON8$rspSj)m
zgJ5_KPQkRJDgPF7vXL8*Z9xWFT_#_Yy8%+W{3bD5@nijMp4g_CiOmmF#n?Gv>93t9
z0K>*!+b84IdW@H1B1o89qZ1Tw3{$mRUbI@wp1Z4|ao&Ki67G169;vx$k8$dkMpX?-
zrj`nZ-U2`Tm~^PMdd`sgRaVc})ZXsc%KJ9TcH{>OcDTlY+!0m-h_uo0P2U!zIj5vI
zlTK7zrAGW)Of32HdUWKW!zjzv^B1}GJ9@q-n@lVbFW%PifGZFESGS|1d-PUmzSce3
z@?=M><TMJ6Zi<dsF=oQL%DGr_+Sz-lpbgf1kQ$_kwB8zx{g=10d9wGdkr!GWTYRd5
z{i8u?b3JU_N|?ELeG<zR<BqhE#ei)KSI8-le@a0Mpc=_zT+UA4qpePKsDW@0{~*2X
z*o&4h-;6YK%5q3|zoOVbRLqnlVm_P-)$}NHx-><-X;h3|F%IXHN|-MPYv#wvH@3^+
z3>$GnOplZ5EUYKBD?slOc4IzNmeIA6I2%GPsSV(7^?sVXA%`wp2(u>OEnjf2DKU=2
zCdMHgQ?{H=da+$SVi^2`>ZvY*QsLTZS=-cS<yTctm)_nD@`K7GtP*b9a^C+IV{f{c
zR9d@Wb)co=05IW(hL-|f_$PP!SW+7e5{bdva!stWxOB)vcM>H%ezfJrNj}pGlrb-?
z@xs@0X>kXbs#4CW>fr_s|Cv_7lo-v%M<kg>^9jCYb`$hO2us2WM5(L25qj#w#?5^9
zMJR}!{s>q4K|r#Z>qo7`*}<0MFeN#QS)w5j(n%Bu#I>aaTK$naTTJSC5J<y=H``4N
zfwZMU)jQSVYY=SIQG3pSbV0tnou#7*l%hC1%4>Fi5cD#;Iz;=rOgJ<Y^b$u11z%%=
zBL6eAtQT~X1G!v{kV}6s401u`D0f_th!O|IgKZ!e+xHJ@N}74WFkFQbIDsngew5~|
zT`PYQ0TccLf??3d2<QP)Q@#m6AN__t`T_qv5&HP+#Is(vp^slQbDI9hxaL=cFbsWk
z`CkTQQOb}<86%xv7e`|xV?X3!Pa6{P_iH8nHlx>azUrhGk7{y0#$6QZEm+W6^4r~7
z_?(K07HV7ara{<hq1&?jZ&~9JUD#UVN7VfHsi}{v4UZ{i%hMp&<Ms}^@m|HAP+)+{
z8oXO&n=9Bqr@Nht*MCxJhCVDX3Q*o-3TdYMI2ETd&J@_rBIk5+nZ{@>?qv<8?1<SS
z*V^CST7#_7cXaq{WA))s{B`k5TNx5}I94~OI#c3+8XuuxUF%y@$J&IC`8dsDBI;`f
z3Dc?7LPX!)S5H9Su~*CAAJKQP+zkru({boK-PbmvQ$`kGI^MXufxcsrLM{|w(rsI<
z>Du$?`0`A#OefhuFE*(Y{|r~}MpyXqw&tX5*6H`?asy@*9dVv9@bBa1$~_z5mZPlS
zm2B|uSG6J)d;`O+@fD5mr0SM4sg+HF#XHx1M^eClSkG#reE(Z?x#5k3>))@ZKJsUK
zY7@0f5w5prErKoPKeEOd9^f3^w~(=+YRzXXQjh4qEi9wizImOF`+H8$w$5u}Gm{a%
zhj*#%*0^iecN?aXmj@YziDOzB*?QzUEE6257K=SYWwnb00JH<rqiYXhG{eQwB&M}_
zA%oyZ91eD0(^x2YGVo4I)CBb|N3R&F8z4l$1ka9yz+;AgKDkvMX0bsP_BJOs_Ke;b
zt&U!>t!Iy`D+?RClB4X<YlQ`cP8?|l9&A0a1Uq2Bcb$1vohfo+1I*#5Im;%6>htQ@
z13XtCL!v;CIw&86ul<JE!Z+j`!{W)!s?%0A9jem|)#+<k{C9}ZEyBI01H6^d)y@#4
z59X3xgi#70q=u41XJh~@P$!@yk;zV~dW<b}C|gOr3q`1g2(_w=Mw&UqVxh3XIQ9u*
zD0dhLay3l|tR}YEb;*4MZX~dxvZqYfyD^I;&6GI<?3iLVfy`|+l=4K_c$!l2^!uXR
z?kms6`*J}6n8GcD(Uku|U0x($-{PXTxYBz@x<coNl8FC5aSK;=wr~V$on`1}6#GXi
z`oAh@3Ddu*%SrBrdu938v8lMjR@5Dd`xVt=0ubmusRe8)Y87&lp2^{5`Rf0PYBgbq
zAt}v3H%$yHJp<TyO;baK*$~f)6TK)&n;CAJEKZoIUz?MXBe%geCpWIXZ_?V2ss-O5
zc>635nd&EKd`=M2LN!G6*Jwx3=bPZ7%1%`%I~h&uAkKvId1=yJ+wH~@)gnxwdP++*
zx4Lb0WOX|;Xxki)S75)LAI#RZ^Mkn!0Z0lzkR$vu(waSynf6~nwTNM-r!Ac9c$zDq
zJ{)HLZ4b7SH~UDedV$n#FURa#WH#wJn3mZu$B<4-Y{>m6g*#psYD-jmZ8gaF8<7At
z{J_fYt?Tf5->n9W4+c`q%<JFpOb|X4tLCr9P*ZMaJL(r3E#){=UCR)A0O5i>YWDH2
z-G~Q<>VtUu91Sqp4xG17RExA~*jC1aCWb)F0oBM>{o4PIP=b42ztN3<R3E8Jb+}W8
z*&asR-XL!$HRKL*?L=X@3w7j9^6Zvy!qkg<f*sVc%Nru+?qGC{vlLJf?p`<5;N`zX
z-eNFH-qHI0VALCiTHix#kGIWY2ZBAps2fjj1gd{eu!ncq&bw&$%XcCW8NCtzZnP>q
z{~$`)1k!l+I(j?aN&~j{uI}qljONd9-cs#syK2;0vzO;{I3hLgq~-!OKTG<dU?1uG
zgZ-rMB0Z-k*2QBSsUN+e<8>QX$J`aI*-dK@c5i4++}mRc(yiOGvbjBt{%p2uFVFAs
z_SvDZ|Ae-U@us~{{ReoG=5Nu|v_BRcK;$*-?Pu;C_`L4o?T!|~PItBlZN>rb&|7jw
zad7pJ^}#`&bvQW8yK=y-cjR-aFC-q}1nV}nvg^Iy>}YNNvkCuG2%p`G)2~*9epwO8
zWFonGWcBFkF-G|=T6{D(#_qysuB?uEM}sj;tP%SKN4;aLl|pb74u%*h$D)*Djg;vm
zeGuiR)Y!C+a~|Uj@8S&+0JpuN&fC!ZHb1wNx!c}!@9LpV?~|xK)bT!=n|m8AZ1(<f
zJ(-?4fvcu{!PuyK?#chA=Pv&lo_mVtcG1JXz+0Z?X-GPF%e&3^;5e}p`c6xXdj`E!
zV)d@oySI45pX_|n+@auZ`6y&H9S@H4R`>FhEKm8T^^<F->Ij3^POn2EaS(PQT6sVM
z?*DqzGwyxf{dmUBo+DS}8TWy4oYDJ#ATa-j3jSEZR}}D#nYudc?NnF$2@p@%E!)e+
z-g9Jc|6XN&SHX{~-~$A2?;t=TXl}8DaKo;P%L@~nQe^u&wsmZ0XjJ8t538z3IM}J~
zYl_tsEGzgJL700<1QN2so9Q-E?J3&kzh4bZvXj2$L3U+V!<BFGTjj#W_B)@enBu?#
z0k_EDW)1!;s!LpDY1mOw>_9o~Jm9Ff>Hlu^u-UAp{r%Mw6CTHxOqKb8fi9{SK1AVL
zM;Z-vwo$gVWp;WN6UAx2QYp?X<13v_*Eh3wY(%)blb~o6LC1&-zS7RlyX^xL4mMgo
zH9h09mv1Fco}G5u9%1&AEB(!`dAhoUf5$ou#)kf|u28o+(V4hxw&xm!TGIwym6Gm9
zH#i9<@|s+H9({{$m4+I5rGbeu-`dv9*Cwj0F_)ZqqDJ~XdgQNYxcl1WW$)&4+5Zj|
zY|%=_@O`|QW~7rCzCVOy@ox~>Occq8@m+i@pK<PKMLuuvw(K^fTl431*94hHDo*NB
z3&-s}5*Hulo~c}a()ZEPimvkel$NP4mi?JYM{a&a$tGGc5r$|*Cv*83#ah&-nR1wb
z^d7zE$94DbD)#pjBz2ir@U)))_myUJttq1=v-Ty^^F>cOBa!opwYW!_l=z=ev425q
z?1ZCMUsl2=6?{s87$LXt@VE%uqt*AO(o6`8YT1T899Uv*G_=>|a~(;jk~R!)EgW(=
zXQ!u^Wm#x>efh(_<7#kVr3VLZ=E)yzI2uCZR)39lhdJF)+ni4lqpMA2dP~hTTH63i
zcEuU@It`@Mf+0y}BZ8x}W_#pI!g&JkS!sWvFwarMZe^G|8MbIng<XyL&ocB)+1}I4
zRaTlwrhZcm`7H$|#*(0lvpHH~{91Hsh#YF$1h;8zGy2oyy~2#H{X1T@H-k`0;^FA8
z|Lgr}^O^;7V5|YG!b!f$aokc;i39>F?Qq8{G(s(zkF?xpYhPD^`g9zP)l3|f{_VZD
z3+#q>JHtjWPt0&&hwdce-S6zZJxD8@A&Q)EyL7nJZxqsl@xA{!lRtov?5r^<dYWn*
zv=+$wQG1Hs#GS#XT6^`xguE@7?~YBT!mYL2A)Lw5Em}Xu?@BHM24)kObL*mwWN^6>
zDQryB#)p)D7VZQalesJga`7*qD+8Y=tZdGemHJgYGr3wRDw~(-K;%#>fr&6%M@@Pw
zbKxjn$J;FGO}G%_hEvfs)-ZOhC-?a)=6I|BN~~6v6uLe@YQMxG)XqV1m~all0cYlh
zbo!0_;d=ScYvYEU!pi|q?o>*6WQjsB7g#yW8QpS|6`8>Uyv<#fp<#@BonP_&EBl+i
zf1Piv_WX4973_!qoF?oe1S`dtD&{S}xV%(+)^v5xzF4e9PAYxaYWzUuIlIQ`@%#e<
zGcPa}><P^`v!ODPmk_y>%Yz^}^S(NDeQenOj2>g+D+8@xrkua7w6bC!Q81xkQo$z_
zd|bhY72K`hJC*q@1xvcTtUzizpJo6Z8V!(iO{tU0X)a_7t~1J5dHLT>YALJTJG#26
zH?Wmxfvv>PlH*Ub62&n)FY8)UI`nXmdF*yo%2B=BC5dCyq}FNb8#p*+#cVe01)t+7
zuhWft$Hon{U&zSjwmQdcm2|UPQkmLlJ+X%Qi`%pTkAatnbsXDi^CRxnGIcl+PC(V{
zW!nbyCFUz!;psu9EyJp>Yh9sT@4|E!N}Bs7Pmu-SJfs|MY(?}nY#{fkYi=q*3IwE$
zrtLFaXrlpZ)4tE@qssoKvYWk#+yh=Y>lr?kiuanI-lD733nRR_^Na_W*h?47{$jI~
z(V^8b&da%vhn&AKK5p$gf8PHd8Unpm;zwTWH6PtOWX_)-=Wi@Qra!ON3T%v}&y$R`
z4L_uta^w-xN$R~o_NZPXuejhd)^&cI$LW^JIXC&FC;@|-riA5!|EyloD!Xvu`~~M&
z8xMoW-Ms!jHQ_H1gxNCQi>KId{F>VOQ%Wx=X8mf0p?_Uz=D+fX75h;IKcc{p|4%Ds
zQmlWh*gsM5&lLPS1s_yk)I&<`v5nN<rnGSdX6M$g8$$|qD3IZz|CbbuC@>|hkq~3L
zG{V3r4<j&ysLYYRuhH}WvMyUI$Q)(bGD1ueyGV}OPx5NH7)T}4Kxd8L-ap7al0Z)i
zdg@Qad*8|B@pGRqY{NHx7qM)vx6q%tSGoIdXYNHV)|DwBBTM2Rzc)9KOL2cV_A!hj
zb$_3$wfmjS-bVf1-^e`B7C+m}F#tm9!6nRS>u&60oLIGYFz&0}Q#85b130ix?KD3P
z|N1Q(j8am5St`gZD=9?S9Q}m<__X<9*M_;gKTEL=@vUe&p)>tPC(hz;H@*(F!ZJ-r
z3Xa<;yf@0C&76d%K`<<JTjEL3$S$D~^J1AAhrcm6D~)xz$<LgZu=D^AFel-@X$KE+
zd8^>;2E_09-^LIDF<(_Sn~p!k`{OWmLw#j2NX}a*9+w&<o%0eZA~|vyhm(J9nxk*?
z91h%bG_=i&7L3)C_G{s5-fNG!C<HbR?Py}FSA4eZRs5z`=<(Ke!w$bZqKlibbDp--
zK{B{PF@kgj7$^T-&I6L&(FVvv@-h>|k=P`+M;p*lLnt?%_BwJq>s2U-+C756^7+F3
zoN2(d#lq@GH>_=}=4HvAq4P<<eXD1**I}i3-lozR{`+q@>1%D)XoW3WyLC;J%Axsg
zwIoknYf)X_?dod1MXUXr$`vU0bG$7l<=I)OjrWvH)oPbVeh#-qFKrXdfbQe;zD+f3
zZvQui#m%^nX@VtO!hWS{#`&6uCZQm&QsLl@*ZL0KG&uMhs_k`M{=DwH)90Ou<R7c>
zJ_SVuV$%KptnA_e!o<{;TW*hOx}77xxpv%B_<tm6PwgyxRcIWG?09Op@E?>$tiSO4
z()D&TB~x96KN87JEL*r`I^$I49gdP}EP}>FjdtjwCIu~U)TUb21o>`h8S!IK_>NFh
zfTF0DYEI{HVYNH!fCrZypC&}oM$ZOB64>yPpLf<_K%+MbKqf(>SSeLWYa6#WXizK9
zsUMi@m9j3ClO=lcJ>HPl?+v`6^D5X(l!}$479I{Gh2P6dCah?_H(t7493M+?5YFP&
zNH8Jq6n)N=bDl%a=jHoiiXpL&Ae~-HP*0q4?S2AzSYZ*RFnvo)<7Z;Fgx{;F%7SE(
z8nG?NC#Qq4GD?ZaYK<Aw0$QRY!)@*EgRNEOPxTX;9-kzL)HGp#tHq8q^P5@P9OZ8T
zv7XVu2x7^UUl0qgu8~Rf#@=bdtqwv@@2-A$g74SvB|uHq@VE=?=l!;F_$-gRfGTtD
z%-i2gjGw>hJU5#MTb=RO7y$>6$EbOOOI!c%&_iu_+EDbT=#-aur>{wxA%$<lXYESl
z?P1efZP~-;I{<P&bqOdNX8R2yLAur#jkDB7j5eyH6hhMs$2Jc(@1PBqV#)3&O|`^*
z>3`X>Ai@@*GlJ!m?M3SXD@xJG7KES+dr}xXKyX~gwdkNTJk`Ub@dSC<kzz@T(f6a8
z%$WUnH-JCOx5vwVo{a{WC{2+o&z>)N?CC=}lwm&^T4%E(?F)LQSqLsQ+EgG<hbC?P
zihA`qO|Taf`&LcE)P?8IJ&%cFCxC)dwCQTpS%q$KlTN+l1VIM$E~u7K1>f12wNBw^
zlBcXDaetfi=66;4kCa^vac6~U3fxgZ5~i=AkiHrwR<4CzZ7Vg*V_#=_&YHv@no!7$
zxAHGj(i$l8+*D#G?n)ruZtK_wIWJ-UK*u3hKrX<_fP<9JolSE{8cW+*-9|dU{bjw0
zi17|nHT8+sTNmHjtE`@&r%gqPmA|Pd`=z2pVd*xYfUGskY>wq*N8M6UqUu6D;Y90A
zC1vB<DJh2#_)AGS%&(M`vi!~PyPaQ{@-<A^+>uGBEJwvoiLpYtjUky%HXtWzMs@^f
z#{Cj%Y*>G<3$^A&#n$@YWM?x5XWOW)8@1-sqz;%;^3$;YYKc^k`t56~H3y;;?s)yE
zY6d{Tts0J54TGpPZ*E<KH}tOMbEqJ<p|JefMw{xxcKu7NBAx<!Ip7WRR#MF!tmo<H
z9l>Bzfw{9j!t+shVutq+N=NN*9E}v1|Ce2D_qIuaN$E^|M}6lSc}#^lf(mmxZ^AVy
zOZGLQ!raU1-9?^#0SZl2ma(~_6qvzoZ<L&<GuQUhDio9JPKEi8$a~P#ojdA>f*sxt
z;K5#Ed%c|&I~?o{cDT{@cEJ?y<r+mHZzC1vt6015LA5h%IOhn~R|!n+q{7^5D$ISW
z#Z4-mc<p_iDjixAsVb>?A2mx=`6{hE8Weel1A!EQMbf*x{iZ_v&)hr4le!@z58Avv
z^k$6Sd-c<Cti29cPrX$~<@Kh_zw=UW`ft?la@Xui*`^$|Pa1wa2n;*)X8aLUtC!~~
zeb7q(XtVUrxw}n0`UT~7HMH_KbGc{yN@Q7f$UFQ-4rO!mi5gq@kJfXnWvy#^@d&+m
zaAPlG&UnZ>YU;nkt4D0y4$~t?QAG}T$LNo-J1RQ=prena=seC-q-+E-GAmG>5lRJj
zQeRWU-EC^PW2<AE#w=0W-BDj`PRdV?6Jti)72HK@PBK37ZFUM?J>hg`pB3~4q&fwf
zKh9+-1zhx3MidG-mfzb3KWn$*Cz_NF#UymrE-}zGkSj+3rZ=UX6GFswj)(7?o!nvW
zI?QZH`3yy07m!u)T+>o+6BupDN!yI(8vA7pT|BRsctJr^qvwCCE?-peA9#-cl42JW
z7zO{$wRZl?%KSkEX8H8GY7yLSk>VZwKt`kZB&V|zMH=Sy(R^f<N9Og-ZeR>9s6v^{
z+)Oz08AEiI!}>^;cfUj*Im^5I?to~v?jall+nL9HNiXqX1+OXib_F_J(f^2orsB$<
z;BxHLpFw35rsfveocDl=8>#=IUPCjW&Fzp0&?c4EsZ46AkepoQE6Vc!iR#FtIHo@`
z1Ew$OL7fVtvAf%7j4d=ozNc3-<<PWZGYTZ@G4+l=tJs`^=QKN{yYVmS^5+!$lF|kg
zwA41=qq~1W!S^bV7RR4g%s4$QCVxSfRRuQh78R2Z1XIzJGRsrt`ofZbSt*hSn~uiU
z8gk~sb;Xty*ev>f#cV1`oVBFoCNWhn+rgweNyJBP%lD=9DLFheZPdCPH(I$<+p-sP
z{l=lu-(1@wy-EgcNg;m3C@0GDykmsoETn;H*R*WZpWx*vI%?)!z`{*9Q5}TQCZPO_
zdTp)N@iy^fVM|_Byg{2l(#Td5{oRpuy+Mf&s3(VcKn@>goNbA3?aVeOWE+0RSB@!e
zc6`#&RjreeZ5~~j@}R^Jt2w4t%xsU2tGp5Oa_n>yBP%u*-$Zs)V>bsZc6)<fpVx1U
zY{3}WE};8R7&+BCKa1plxmsy*vMb+JC^kOto2bp!itzHdM46t(?MSs=oOT5!a9mqk
zp1Nq>hMF4lqOw+&>KD(N!nsw`g+;kg*Fgb|3Z$YDbxt+7ONwf(nDa@wO&P~ykQKj~
zSUF*g+)1^;DNo(Ck<I1F`NgSAwUej$b368<>{6O;EYe=$^yz4WzpMtvo2T-@q}Mt)
zI8oe9F8y^ba4(M9%#Lfs_=GsBvl}{?F)s=RQCbH6Jsa`wj#r0cD`QehC<T^|X+#ZH
zX2S;<8)Hxua@h6Wz$EMgC^O7p)UN$Up%e<@z19<L2Rm^>+h=KqonW{!h6gC;>#5O&
z#mFbm3|r7R-11^&>e3RBxQ-_nNi!WVZVC&+o<=$AdjAKRm0?~k>+u@DP~*%$F+w`8
zPzc0!5wQ`wJx$&SwB^$d_-hyss~Z1@2ug!iSMwpN)L+-#FDiIeEi%B^e5H+imZYPh
zR+4o4-Cg+aMAp#w+X+OnGu|Wq>iov(+!Q^WpVBBcK5DQC5y>8EbsB-^6vTSQDRI8r
z|6AL+#zt0_aeO+{ZKs`fI=yU{z0jRymu-hgVK-s~MOa1R253wc6J7Yy_H^4Rz3iTu
zE;yNmbqxtB=m$3@A`md~vPQsQqIfsa51L4f#uykc7!yB$F$RN#4`}`UpZA<OXD;nt
zh}-O$b3gAn=RNQL{XftDeI5@{aJADpLtzucuyo115b=g<6)%pS>RlKBll%@!xs5y+
zp|F3gvbiDT?RpAp27NRfnKvd*>?Up)?o{6PNf<SAocAQLYB8MZYvmj};OOG_sukLL
zmz<&19-Oq=zn9<c$DMOj5oMHnE`8z<XI0Jds5$&9pg@%{_?S_#4?KtnT1WERW3n}3
z30C26Cw5j-rPFNb<Gy+oBW6Wq>yMTx3E>R5eqR+PM?%nSHS<Z$Wtk%q4RZsJ69o-N
zt!?pgj4D6J{)5#_J*S=RX<MFW3*iaEEwWmft$ujQw_d>uyaoT)n#e>_QCJLO4yc9T
zbIgxdc8*&j2Gu03WSl<@O>v%3yGwaeA5F|IOs^0LyHtpWZdzR-(OtnSdX@%;3h3pR
zr^~YoI)k}D=kUv_sc~1Eyw(4`O7Nm4FKJTK<PVxqKN6Mf9Yuz9a`0_55jW&3gCAKT
zI1-@=h}h;<p|Cc)#UFKy^?ClJxj$=SZrztPXWxoG4garNe3H3ZFYZPn0xuN_&i#tk
zE^4yOq|onrAFpbKYO8;XCV$t&&TgH-f3R31ZpXbraMs^w^FK9tU6X%nVo+}rgZg_}
z_nJPdo0(DU4Y_g!;&MPR#^LlxN@;*3s=(SyjHcx293l!v=ABju?B4Z#CaW-8@Oqfe
zsrB<Ku8{M<Vo%Wwl?{IH^y|pmAi=(hf~d8<jsGR-d^VO~pOREq0XE|HjV+=08d>pH
zeXokr$eiT$%q5ejwlgw&OLY_$tSzB1J6v`)k5MF>&c-n;u{Bp{7G|fhV0?`A5@ur)
z;A)aF_XHM<KChQ=w}ORC=f!1UzsRv;!t7v+*N+7w3!9J5APYvHH^3PlZGQ$Y7o3qZ
zw>dt|_lOi333gB>B-vVinpP$!fioo?w7VKKx6pUF9u|zXMCljoF^t{7+6~sqf;9AC
z@|eX^<)|T!l5qQ)@(6a19msn_;r7n51`RUun>OrvEETd<Y{Tw>yQuW4Esc7_qcT1|
zRKBddos_`4%<eIQ-D4~F&oS6S5{t;OpS&Eq$1e66vuIF?1hrdM52V<m>^X+WJ8vaL
zEG=vJiI>=SkJ(6el&`4n@U~&?*v;H-B-xtVTg8In-Z*k>*04Nm0&CelmSMa*;nLfU
zOJg^|im;W)7rP!K4nZK|zT7Mz`BQsZ*mOtfNF$r(35J%AVQ-8lCQH#W<-M}Hhnljt
zx|eIqGSZJ!TG!|KKe?YPNkq-nC4H71!<7I_!JxN?(iJu?9TMvO-W6t9K;UdoW}GV(
zss*mJm+S0nwJcPJkC4zE%gJ*MrDtv7%82#9E@(Eht(rf_XscQ${_ptO#c*oEqldi)
z7cAMkD3%TPjZDbx7<U^2TNu9po_YyvH0r+mOwYqw`oHNJT_QvMxz0j+*P?3_IW1gu
z!Sr*#Ud)^(mq<i^rGhlFQAvR5{apeJO~XRX`)+)ca~$nvjpA}ObJ8Yc60ypma4S8`
zBXQ!tF(UFvWF&mrjDC(JuYj^T!QW;SlrAHG65#cU<Qjn$3GFoc>9!i_e5<z$_tu-8
zy+dzw_J$$0TTw1Dt}v$ValHX!6-|nmI_ym~C}R4WDd+t<L7yh-`?~rixQ;Q<&B^Cu
z9Q!+Ad!D0AZPJ}gC6maq-b7ilp4%4cLy{dM|M9i=EIT&*+Y&65WD1Kss*Z?yedafn
zPqnA+OYOIHI0WLpR9i<ElhB57KWaAgqaJqDF=C%D80izKnIi~09W;M!MI7@uiZSAR
zhHC3TzBi#xo|Jx6FI!a*16BBTl6u`o$Ba#NS%1kT{+gSdpQI;Px<{uC4kg4TVMXsN
z=T8==^HyoEmc+va>AZCAny1w|#M7<y>3T#$HdgGP(*xJ{9{$tAu5-+TEQEeQupkvg
z{_c!1N?4_0yrO&ACTn)&1x`g?;94$zO_94GS*=k#{~MGc#=hG4cZ>Q2sYHn#h>}pF
z1iszoH&!QDQ>~s)F)ZC@N#VgLbHeY2r@)Gke8QP^53EJXr$2G5Mo%YxO~z=$Zt}WI
zavGXHb5XqO1F#e2<ECFF&*vgsTEP+Udg6h?&#6!i9saH0U+5+6C(IgLYtizMAT1@z
z7BM{uhd}O=yE%GbF-M=LpEyHZF2<+Q(-EkRV+d}bz7YZ0K-T_hhWMAv)A!iVaa*z8
zQ|Z|k?<?;MU#i#pSds`{Tl`PJ^ah;o^cZOYwPb?T%v9xFW#w))gNW*=%j;UIXRIu_
za8}e{Szn@he`k|_^_Bmwudh23wmwvgYvq51cUu_spHgey<YqPUVdUIU%RtAJ_eNYI
zhpYPHZs3Adl!vr<J?E=sOwM|IbqRN-%SiH!E^|gZNTjlfN7jwYF^gv#pz*&-sgbMP
z<DX&XXy?rA!oB`kW}EE+W_(lKR2j1=k(JzfHl;>f^hEIB=5t0+6gW_s&q5DuOZ}~T
zU>i*J`W)+@q{8Kui9dwXBB@G?fhVcYk&(pGH7gzw-DD+@CC?$D)li*U>P@=0NjiHe
zI#z^<=jq^>)E(8`;{>9%7xXo)wKjdDH^Q){{W?A}n3(;z>ATe_9@euNV&75j9}rt-
zs5*mXyVfH+Y0EQ+;aAfCx(Z|E{1S8LS}dXlPlilZX}+NMe*qRz-Q!GHwg8lnId=76
z72Q7C;cNRW#cudlhoRe99fu(=W6+}L4iN-7Si&>i0ztBjHbPK;Efm|Ez4Gw#^nA<+
zdceU@h#m@{zP!4!I=e=+XlLQf&y@()X+embof3RoX6%y&5N+JJPjVYU<1QJC2KNUI
zJEkQXM$Cd5M@Ma9jz)}{lbUiZwyF<Jc-?a6^oJQUw>phsZNM#Hr6cKM)Z!l{{V+CJ
zS?^d+b0Z<5H3Cs>YN|+jy<ZFgKmvwrM`dp?+Pcm2+6V*&IZYh|B&w&ueVZWQAi>CG
z!C;{~1`9`s!D*mShky^ZfdHk+LJEUvm1dk=f*FYv7_)%X1%U!G3dw4u6f^yxwv%XT
zKH>FzS&FjWM^dZ-D&lJ}E)cYls5bzJY0L-o>-B9*?1ZMHFBfbK1+Z}q#}7-g1No1`
z#sIK^)7^g*pvX1AM$TX(H&warVqhb*#LtJQ!P_wJXRUjtIJ?IGTT!aoc*b!y)%vHG
zit{T}5TzF_`8V^v=)k<6pIUd2#MCtK^KYGd?YDQ#eE25k>pa8`4s%A%gE=y2x$bN<
z+Uq&BHr9~TLB{{4)?K3_bxxC?Q&TkaXLPh-%wUWXnf>bx6~W|3##%SU82@RVT_98I
z2|gY!5zca*jX>uKeGfNKhzN9Kik48!!H(t_ETl%4*dSxw3SzOV5rp<_@(r~IA}(r5
zK~o7HXCdqxMj%90rtz?3m-MET`vyxjS?$Rj129o(X(g$&or`!M;%lh1r1-(c%s?8e
zv<z6jOcj;^5kQx$UxKc509~pNclzJa(>teEaU+x37JG!DX4)Xk-srcro4wZOHTOLx
z?9>jzeyHu=*W?G9G&Mvz<4^S9vs`V6JbX6pgvZLTh%nq&y9OzSXy|!#Vu6|etR@lP
zw2MvXX1>LSv*C4wPT9^bMwxZxK_+e67@0oG;c;Z@N%f>WM}`7|6o^jo(<vXfmo?-H
zMW<=E*4AabK0~P<^V9Y6><4ON4gUNV>RGwl&?nEoI@b|eAzHhdID~m)9Qg?m9F2`y
z#bgd+Vu#Y|G8qJp17H}@O}vqxUYl214#PB;J<>))P0fZV%$+7wAL|_>Ne_bhNcdzr
zMjh3zI8TxP0YFPxZ@f@O>u5d$Y8<`#-D<w6yIl&qVtx`lV|Cv^0DG0O!&s0hW3eLJ
zJLwv@G1J~U3K;gF1X#)}S6Wy*c5G>mh;^9yUGhp351RemdcWq&mBux!kA{RlR>gQo
z6O+Gg3XCtPp12;C3u~v(-^K*KRT<y>b=84#LnW;LDHY<=ntVo+pQ>q#YOGaO-NK6$
z!heLB$~(_<$qSUJ)^Yk4uK54L=^#0yDBOp@L}aJ9)v|<G%_^}SbE14`vHc@ujVSCX
zBNr?kHk{W!a)ik})?+d46c<M{Z&tx!12Js2nhnzx1ZT@~wlcVk<PaTynF_}~Pe~A)
zrMHm5EY1>S?m(HatA@&Hp)w`G(dB<OBy~jdV+`V{0bFIxH1COxY2Ldp2CvslR8BR&
z`dB#**`f=TeKYyl{KWAC`Go>M2gw)f!ux0yi?fCN0_&2s<naUiKj@g?=Tw7?4h~;1
zG#dl4&%tCd&+=i#24~yHWS__xZLk*p?ejjTbg2XLh7kO+zPuI)mOBdQ)fcF7h#2(N
z8HdO>smyX3d#hl^3L0-baN<0lOM4h@;Dhgon!Pv|3_v*Kf-&}o%bs@_jKb9Gq~B#7
z`H1s9qgT}i{5tq9%?vYykQmS_oTvCP;OlhFH-q?uefa;URfe-%-bUW}_v?~I<g87+
zC^q*aj{T+HhSRs8(P7+Yc0HXjmCGJbQ+xQI*S#KcSJL@wX(GALE=;bhxZ#ne2h)gr
zhSGoBHsJ&%$Gt*tQ;nBum%06IT<;`W!*E=y*^rf=w75ccWM@6sJ8~&0G36~4XMEw7
zFB6vXL97r&es!tjQ|N^M7fsaS)duYXCyN@DKe<*evWEowwegnn*jBKJL`G;?$dWum
zE>+J4E0O_kd2W7M)XLB3==74eZ2S9tx{?8pdN39i!KvHCjeMbq5xJuSIs#Yp+7jnS
zyW;D?*2d#&>sO_yu`+1MWw{#VkfWo)A^5{5r`O7>LDDZ56+s$`+MdRf2CACy+Ij~5
zc9m;~COdV`?Xh##WjK=-ePEZiy95%G*jxb{|8lKzxh*D71knS0e@tilmnvUDbJsGd
z4cs2S>^tnv{R7&hL5DWZ$23xApEm1&b;Mk~iU85(go?HFd)+O(kHZtC877^HblN$!
zHIRwWv$O=xebCo2K?|XE8rZ~NQuSAXEpZ5RjdUg+aDTb?gmdn9=%(CNi3f5S{tx61
Hj$HRYY&vk$

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_22.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_22.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c33b296c35c9bb6c4a343def7bec5c21bfd92328
GIT binary patch
literal 59703
zcmeFa33O!FdEZ%k6$<;(Xf%4O-W$!{*xV_KYqPbGY|>;?luM#Y0A2wNG*CdldT65I
zSF+5eB+51=TG1qqO*&{Nw#SaGcpN9AI9}u|_KahDd>oH+&YPS>@g~vEIP#V_ikJEQ
zzx!&zLQ0nHlgv43vM%ns@4ox)yUVwK-+j2Zx0vAH%A+5tF1(mb{JJih|FUEr<X8Xs
zWFp}u$_X#&r52Lqr2S1TrpxK1E6bF#R+cO0`I}xSEOwQ<>|SP}xY%9pUhFCN*mZWH
zcd@VBx7c6qx9i-(z~W$eFm7k4JjC<)h2h2F^01W^7Dg6#ly_LZYhiS8XL+aPiwk3m
zyUM#P-@P!txVyZ2aZh>A;@<M!#eL;{i~Gy_7fa>R;(_vk#e?O8R<CE_(Bk3pVaxX}
z99cYCKDv0Ue9W%<7LG4YlqVKXluy`o|H3VcC(9=nPnAzCo-UuZvVny&i)YJcEkC$$
z>*BfcIm-_%+_rdo`F8Td-iWuu8+|QZzQa4|jd{CXN|x{R#=YGyCCYcbBjN4w_P&(x
z_Ez)rcURBd$dvD)Y@fHEvi;S2Zlufia$WKcaD6~^D0v$t2fag-9ID>S^?h6)_Kt9U
z#IE1Y^-=E_*T<@F=l=a%ANMA>p0Mi&xIW?C!u2h7{h*h6CNXhpa+xVDzA(4ol@{k#
zgB8D8TAC?d4ua+Sy=TtM&IOlOrcO^UEuL9enV$3f$+=qnj90BK&DG~>v!}ednVFUP
z+*0k-^7UZJpT2x%YGG;W%w2b#y>rT&nZ9%8j&pa-+;O*e+if#<-#&HM)E(!}RqwpD
zdS~^vTkpJm>b7=El~zlYW=pNNY`u4S`TBNUTw1QyCg;w~EiU^@*Q#E-cdl2}dv3q|
z?t9MNecRO3-FMwNIeqKxcirXP;d%7-o~fxR@0@qfJ!j8NS6dCVUwmeEa;09Mo2<2q
z-lPwYEG;g-XRdxGm|WBV1^#5MKC|R6R{eVM)Tt7Gr{{`qXk%TRNJfRnCl{x@$;Y3b
zNJU+Kb<(H$Y&FXH)#d7BKw;nXQZ1;i28)xm$=RwOb=OuF7UD9V&o58<!QA8m*-UWN
zTjX!o;^dWTW!9hbxagz3E0wjXzf@VOEnKIh`1o8scx0(IGdDX?i24?mW-GyJP?@Y(
z<`!q8Ty>S(^(gOEXC_w`f+#yzUz}Wy@=W(49ia6pgD|r&8ALtPODnaYLJuYv>(NdY
zMle~UGeNa7<1Z~%=%rqgjB=091&=OH>Jw{>#N<Nd%0jf~{M_X1Qf+eK9r4JX=f(cq
z)Jni=qUzAQnEzldxPD>IuLsWrdU@|dE5TB=HofFk{l^}kNJm5Vty9%tvQnSET=iBK
zs?ji`<xfrrmAP85R0*oJ`jQ_NwP352g}EzLI@Eps{JWlbW^!?Pq3Uxv@GgDY<4dy)
z@Dp>Zb2ZBP>&uhV)k-k8Sgi-u<$6?@)(1`V-QDE;rFwlb2x?LCI^D>{ce>i;=}B+#
z(sWAm%s>5Qn<XCMSI>}y3HDxcZR*KuRsRKlE~u6k`KX1{^n9*XdVyV63NFvpOEW9A
zX$?pzSSsmTOLc8ft97P&ZF0dLE0*}@EmiBK+EP%OzN{rwnyg)4T=J`@zdylXX6n_2
znJA-o_(yoXC0|}Tb9rg8dWMIpQ%g%%&Ull-<e6E2X=VA$Gb^>nS-y;E?Tm)|%y~BY
z!|!_f%!0Rg#<~g=Xzu+pp0~VMxmvdYt<22Ts=-RF>Q!x?Zoi{)+u6#wb8MHW&zq}H
zPAycuihaQ`jfM?T+Mgglt&!0`{T0fw;o3tcOoYj?M39(IhAA%@rstBMNRCjpQ^T0L
z80Ea`wYljkUz#;26lGZV-b5x!`M2^r7iH_0Czq@K3EFB8fPb9JXmF)Arwz&mskmlN
zQSADa^v{x?=GXeSFV59pWkRb7P++1=pS+Zp_A;{xFYD#JyjSqLyrS1luE*>3`n-(S
z&wm5_9psngzkYs}o7wlV#7xo~@`hjPDyO}tgg8JJ8Sm@7QEw-?Y&GXS?NG>#WI10g
ztasf^REsxw*G$S#N|N$!hf>Nt4vmz19r`HuIiykUcgUhV;E+Ul&>@HNP<7Zl@mkWm
z#XI>@raa<W_nrZGo_<N~?ubh+d1t({Rx%owJPW|Q%}RE9@9}Q;?x3$@-ksiEptxP$
zbKc$FJ(P@l@AdBW-bQY>cc1rm?(OmJ_a5NyUhhHgA^z_39`+vL@BX;m^WLK`C3thm
zyWqWp@&n#Gy?62VV0{1Uy~n*L?7fG)i{6vmIs7q}O!)|Fp&Vt-pMUC!H@=u=W^)TH
z#YsO(Eu6b}VXeS+IkUjNy#!FxhJQm_@(n5`-tb7iaO#mgzw>K%jz92)Zy5Q^JO0Dk
zryuyjpM3xC{qhr2AAVrAIs0+uOAkb8V8V1MF4sCL@Kam$Pq3nc#Eryy@+L0`cY4Vy
zx%zmJ@=`aFFDApCVWN@_GGSsqJD>Z`gqJ3l=5OZp%#9?318cpK5Bo?9VJ~S{SR^fm
z-K5=N4{1-hi?lcB3zOmKjTHI*V1QP%qH?dKcxM-7gOm*VKN<|_&W$7`!$HQ&2P44_
zN`}`*Z!*R=lDxU_dRlJ|$GpT9JxLGecltjUj*%Y=M@V;t!=&ROPgZt^1=aI^J1lUu
zC)`82H{ADX($56@!t9OYtI2vMNb#jv)ua48?d=b<VJ^&jU9UhWT5XjMQunOt(zBxa
z(AnE?KhIv~PPg^W)&79Btp1m5bQALj!l5u7ro#SkARP31Udb_ry^T_EFgQeMX*PL-
zac>+54$mJ6j`HmX^wnG)^ZH)V*d6!!8wY0-jYIqnuRa`1(619WlIIi8-~Rqo<4EIZ
zcqBY<B}ZTWMR;ia7H=Rt8Xnx87)#vLm>q2#3=gm05+2djtI4bX;<*GneCVY_c<4Oi
z+c-pPCxgT5N5Vrllfg)M=tj~TwmD)2R8EECq^E;3^Jl}|)VviqF!D;Oag6-wS5yA`
zgLCt@k-wdp-C@1Gy~YYyzr!01kKNQv?|n73><4$=(5ktMns@QkH}iFOuitZ1W%qi5
zkUK9<r2YF?oKb4^Y?QpVc6NHj_d%}<*GnP`z$(kjOAr#Z+0sE4_Cc_RS30<QHqKl-
zxYm1b>FJePO*KGJYXhg2{i-+TaFg)CC2(dQWHz-_udWR|R;@0p>Z4CQREL}_Mak2^
z=I*&!@PD;(rFy*{74_nZp^~-A)75%4u-8`?Cxz>lh34#L>4oat?B$?-x^zhhk^fif
z)tQxr(hK0Gk`H!X^7Sf_9CW)x1E=hpz+KZ<mY1LcPOqIW-O~P7dhVg8FTU&IJNTkU
zpLz7sC9<W9&puJ1v!!RBI{(n6N1rKOda_i0^yw!zJT{Swx@YIERcjRy9QCMoBi+%}
zrWJY|;Hf|m`9V~Suc|c<Dq;z`xmu~eAtb|?Wxe3~a<v6l-Vl{AtwjHX8-)v>;a49g
znPpC&A8I7S<h;R`_0-M8x23LTo^e<WQY0fJ9;K(3SE6ifWpVjB<DNQwHp-~8n^51E
zi}x)+bV1xN-~TOoh2ED-tQ`;tJzar-nhPqG)AubbO;0Y=?<do&J(2Psp|1ZZ2`?zl
z%q=WbfUhf6NR7EQ(OA8H_3FwT^8+b5=>dzEX0J{96K?|*uYif8LbLs-@1c6V%Hmt9
zJ?i_=Y6C`5R)eZPw>GKbD7(xQ2T@n0GQBWauU9I1x;D93B{u-=<!Pj6t8O(#z0(VG
z%b;>2`MjuGByN?ZdA$-u{WGE_<La!#sA#C5qE0h8kWRJgLOmK|ja8PJD!m?h)o4yu
zSv?vMA!HoSbN=MAJ|t_))E{CfqP(7C1^Oep>ash!j?xR&T9ncFMfs_fnHji&s8^Mn
zWGkQD%Epv))O}5T&^RzEBP{dTIqS^A)KbOmil~2qu8WRs7UUO~m^JEm8)?i0&dddV
zue#G`^XW+QE6YJN$mCRm3ag~Dvdprn)>Sl1AFF<=Us0MmzKEVERa=g-;!v(fIcVyj
z>VZgW!ENWF{?^hL(ao52RcW>A(`M8+xv+d0?qZt7=r6GlG-~XGom*bjd{#KRqFoX7
zEP%di6}NQ!ThxgRTYn+yRi|4orDwIM>nZUeS~dQ&s@UIZ-;Fmr18^1ftW3l3h{|R<
z;j;jAON&wW)AkoGgB}-DW!bOIP8|1-(p0${YBq=`#g@E(o?`zS^_nhVjCi2nvj33k
z!ca|m#%opdl)tQ+GfF<FWKzkLl8TZKD7m2I9ZKG*#5Pt<xkV*QByN%0Vx-rCpZ?eD
z-hz^8B_Ad!Yo)WtY~T7<Rs1m$@s*lqEyEq{Dd(~|?f#mcw&mCw)AA5I4Yp8|*POd@
zSFQSj1AgV6X3dJ&!{l7KcSBX1gNda7KJ~&l4$(;db0oL%tA`|s1I6S(GM60Z*ZE&h
zvd^x2lKWCc<x+MxRdBTueM$FRPqL8gp-gF!nmwru?Tsfh_TLD#hR6-j+E_|$3{i6=
z)syPu?^^E$7&o0)fB4tOPhv1lLjA{$k_(|1d}yskzBUxVe?V<yM9p}~SF(+~mkRS^
ziAEtv&8MM^-~qtjT|sI+!<{Y<8vP@wU5T5iMlsCKXWyS}q=Q@ty#;R0(~hXTENNHR
zHJ|m;yMXWW`75dVhl8S*6TKL82R-w>{QA5+Ekgya_j?7X!TgP+Q2@H`;<~HjTGS%+
zRO^0s*iDZH!mgWqORKcUm8zejW;*Wl+MU5|?)0^v*}d_Z{?0qzK)XdKro{ZvSR%}i
zBxtcXmaxCPbGN>BNaZ^d4be6I-XJ~i>FCLj>j}Ld4tu=eo2d=`9ML$~mlbcq54!&D
zhVE*8<BqV~z0+zHwyQNi;*DNOK{r9^&hPMc4ktWOz@zY&-C^N|+SM2ARKIs88of22
zs}XO^o;mOBQp-Wl`j|%7+8=LT=?kg<u+`sf^*_WtsQ0h@o|{RYECl1;9^)E%y#dxn
zua~}0W7hBOC6BmJ)HE%McHDC_xxV*ig75fjP<#=t6uPe@x^E+RX;B)!hz#&_W)g|d
z8-LEfwObshE?u6yhKymUG*vAj27+AiN-)T!>f|(1iC7@Q-hMX?L|LPS7!vZs8$mrV
zw0bQXu3z<IIMPM}d#!K@LI@$*O5K5(_U%1F*>8Gxmp$xPr@`_Y-x%#~-&(wN(@p;|
z^>L_$QK46S1bWleXkWWbFK(7YY|MeHyxBt!f2!x^Yd`z8Xrx`y#9-4))tOtjEj?%C
z_8TIo{I`?D!~XdeEBT;*KgDYUiz_Cyi3xB}YR|&8g{c^u*Do+ZMTm^Lka8`U*k_pB
z-=$`+kwp0zt@;fWol#L56zspK%l6E_SC@iu4vFSssaSpnYy|@MKcWXS8aybKrCGx~
zQGSu7vADA6aLuRm^xsxubDg{71Z!PPQ{Cih(eTBk;9a6c7U^o$bHp6-!`fuc|1P~C
zJ-M(j(Puc-VL3yiq9HrFYB*3YgHkZc!QSWp4V4rm(36zebwrJBB5D1AvhU<qf0`sA
zoB?Ji@LwOWzW`1UE*VP=qz1qY1L;xlh5p-@9^w8#Y9BSn`Rz*yyOh8zhxzp+5Atje
zPppj^-e}PlO`7e0B9IM$`F|pizcc~~cHTtf`9D4a@n2=1`QJwJHQ|qL!yhr}Ke7>=
zyVCNmO{GFB&07ZVH{CZDq<qM3y%`dTpFj9kh$imT{x(k8gi+!O2ig_3ARSl5p`Nie
z_GqCP$1WXi-`NfUxklP(XuMs$jfUdRY<KTm+zUy7{MVKI#Zb~vCrWZ{ftdWiLrYsQ
zlE1F@TiD3I$z^$X1H0y0fWa{7tP_&NWdm(dN!;g+n9Kiewf=XN2xa+yPs#TvF<$9=
zl{3!hGs^vll7FD&`;`2zO1@vo4=DM^O8!2{SC5kXAJj9#e}<3zA5zY6;&!O$L)`t>
z{QkO7(Vp#5ksG)TswZ(YViS$+f>uoU_mL*ok$=vo*E2VrEK?=$BJ*jJY(haI<-HGX
zMKa>#>qt?PiM67acn?x-q@1EdAwsO3{DyYilh@)WQ|`(4@&puV`r_KIdrQxt;F7vd
zM5d%}v50SNHzGL`=;>C>qc?9(6dW5ZN(vbgWI)xgDqn1gi%?rmFD)bHS@O%elZV??
zJ)~7f?HW;$bg7$Y2Ba16!gU_4c0&}`N2!}Ieo>raUor)E>A1(WJ{$5@^=S#`^lt~h
z0zYFLkEf*oKviYd5io5TFCC=9%z9cP_bWO7Xh_MfL<2E-v$T<2y*tbzbUxe2t)2)|
z<PJ9SVQxJW=5HqbhXa&F>$xx|EQ{DT6XxEYxccyOiK}-%XDlOR|9Vi60^p@Y(6x@r
zV7*%c;PoCRm)0coZWLZkt>%K>t@ph=tZU(w^m#<s&;J@??=INYt}t^Y@BjSO&+@je
zU5WWVn9m|<PXPOg2)!Ui4OaC_K|gG&U3pzsG6>4|z^cLeit&a0igU0($@(7EMTybP
zbDs^1>jO7okX-wXzPOE|YWl(I$zU+-h3)O$MeiAp@kT#t5!g(Xj3&emh5h$Q=>GI`
zq^L-neaJEzpI&{>rsvN$`d9CZ%ZhQECmI8tZK6Wx_P95im{H$pKOC4z**m^h@3{J@
z=a|`bm{Qe(MH^f_yQ#%}v^dyqCrdxQ;$3Kltd&2u=f%pp=Z9Kf+3eBy#^>q9P&mM|
zDVymb1Pc83k(Bz^7+O8dHxJGCwPr)?-#3Rt>myzdy&K}G%dC&yaA>c_*+%-l#MRGI
zmVPx=|53K>4%Q+H%2!e~R8I4w?Q2%SaASmZ*KYt|fHw`Bx@eHxh~<V}M+gA}YInSD
z=05Egcw#sl{?3Hk_O5Px?=bg9!ja}Z0|+}V`cJTN;B$HqBq8cU!!D6;5s5B_1d`dU
zmCjces-_)4@>TkVhLiQTE1Yf#%-MNufw^R`CNS^eTat}bnCuXh8w`Luev(R2MzZ86
zS=j^_{IBEYQGo>UaR|nb+4kQkM%_+4zIO1!oKt5kVOA4EHvv23XU$8*<{7VApO(ft
zm|SppSuCnC=0;G7`^(8Ff-&f80Y+1FtmT}Mz*;X>I~b;*0FlaOZJ>nIado*m4Fgt^
zo@K57iTM6wT-{qNtqm8SZC8m4Cw9itRY@=XIU5r_ZPVZfbr~c$n3|KSd@^w2#mlPL
zh4q3TEG;26&musdt3?AEPR|*JNFa--+m*zUZ(pF&m{nisoyQiS)1oZG*jW?^;wYo6
zY*y+~*GjE^b){NGI^S=S)%L6#^q<(_KzpVof%P9&lc~Dr|4U_(SEAJ1m8ejQ?Nt1q
zvNsyeD0eYE6|)S$G6m6QN7gDtR>89VC#|*la<NuHZjXp}!E5Dt7H-*%b5v~4z~=Sj
z)5f^g(>$33g2(|748lwMuMbosM%4*7#k8i9L$IBoA6V9rWM66@S1~J_+$V;X|IX6N
zNPIO8Q#-0Q5ps@m|1dxOw>G@Zx}DZu(Z3u&;W5Kq8oWRlMS{-ceCkFzL;!}+5)wd2
z9x+~A4)#*LOzS%91fw%)(0pogAD?M<DJA~BqZ(k&f;0&+M$^QnboKaZB`AfZ)ymn4
z)2*ua+GtbAMEk+L_SpL_mR3uT@H>6_^z$VryO37V6g&&nI$FWn@(L=#&8@`GLw!zF
zeLc}KSt*%OPwBXcpGvn*oG!f+a^hs^ECyN_Y@rp!)CRksjwXGqTKw9*rH4wv3PKK^
zrRK>}laxPM>LlGKPL>dM=#}x89IKaXb+sFeMM7VCU?S^Fx95La$r&X^Vg5~W6A%mh
ziSltPQ8K`ahxbNUGdSs0xu^5Jn^s)Zf3;#PLbLYQYxQaSCGuK52sB1)ZEUL6B5brU
zvixju94E7Naaf5o@NT)JN6?M41d<mgGVZ(n7LU2_`gU@Z+9#!%egEH6fSyveS(|%4
z7Au%z{=n2$R<Wv2Fv2S}lw8|u-G8&Xf2H5?!>#Z5-`d~d|GG-`9sX}9InqirT8I{%
zntaPwI4b?hhHp9amfsT7_4eg7{Z}zF?G^XG&|*9@9ba^R>tpP*?2E>?`J%N7$ft7c
zYU^t<%*_5b`<l4+S3NhZs7-4OkYtu$?KSqbd`WACVe#NB7UP~>Z+*ws6}Gu<>nkK(
zZS52ZALE_E&S-+s_$!(vDbss2!Dw9C$XYD{Q+uwFgFHstWMGx=&AB}#5vSWzuPJ9d
z=y#Bt$lLDn|D8%(OEc=*C^2Z)-@IPCueMdr->6{_>zH5|MonP4HV{LLR(t=QzW+z{
zeFiK4k#ffA|3h-yf4?t#+c*E{!H$o1yX@s{K7Dlar@N{MLOww2t#Rn$qO}fYxPS@6
zQZY=S14Q394^wz0<sT;p!~|-hCwwK{$Qt>L!94fw3^NVp-@g!M*U=|>S*Ohc)n@es
zHeTuXryIFOzEO}(RamgKk-umv2YB5!FzOKVeCem!_$Iw+d4Pbb8X%LL{3baJS{E?p
ztoXmp#YC3_*vVDbvc9%y+OV{OJ;Q1RU5rR}XYL?;V`CkSwpNK$_$|H4<~7&9yb(uz
zn+AV~w<L@<8iM|jO1CF9+Qe3y9+=j6>t7E)<qca2iONofUveh-?J$)eNH($>x7OIF
z(cx+=3zT*RkUG>xD|;i|D55pQqVo!R+?(LD`D~cQHY+2U#ysq(my$N{yHcM>z!bv{
zd)csSB(a{qNoiP&p;J!UxBr)3w8zZ^fwAVz^6{j6{sF`rwwzC>{_iW<O%i2ILt~N*
zGhn@}!ZZ~AM339(8{3b1oZ)!IFlCgS@;|4htkEy&(IGcS8>T~eE9$8&)yzzMak748
z^9&W1E6%w0_4d9Ph1BdzooH&(n>AWwWr0?;7i=#69vMcR)!$&0y`&R<pfSgmm-U}w
zy=P~Wh(2mTCddZp)~TJB<(}C`=FzGnoOu4C%^ezZwC#u|W|GVnqLulgmwiQht6O`6
zwlu>%qmZnK89r)Pc@Hf;d+T9DD!pM=5-y2PLc}HpU2%m7vrsLCi_O{5n5`AfNL<_%
zw)p-}Q{?|M67n{r?^fRJf16DtiJwv(yJ_Ys;%U5^F+5QLxi!pC)$@Nrm0ncxWhLK4
zf^pDNZPpotR8%MtDvLt2lC6wr#1Wtl>cC&g(7x_o^<UILTK4<BCQxqO71OeLGJBg-
z#im?Z{$9=K7}caJle8<BJctIYnA}4?L&@5Z!{hM^u!r6lpnhwnwc}d5?jRSMY35n8
ztyvMA%Ye_)GcXJ6IJupqaDMRrjJj-*N&YyS7-l5_Ff#hEI5mlk&<wMzVn?+jdcYE+
zNOEp%{v&;fR-I&bIk7#M`D4fC?zZ4_lWs~=v3@TJ_%>}8*zD?;3O7<PJ6e70|CeES
zZlqskPg@DScO@YdiP0bFi_>~v^Pid==2tJ0ASX(`jEu;U!YRjJ7i~G1y3{^^K`>0+
z{+PmjKwdmT?*9os=MU6!Oh{dgC!kZ_;A%2rGF#cqM0ApL_L4SU(CeD7wjnGuTmNwl
zVotBcTm~Y}naMQoZW%<|+-kvy7B(~BQ9L80jg>IzaS0^bCbnRElFi_mAd&-`o2EKD
ziy_mspH?R|k0MOgiZ;u3qt(AG_w=*=IV3!9;J`mDMrF*5Cs~|ne+mgKMwE?A?S3s=
zF9nb&Q+_^3&u8Yd$Xux})O~fzmCRd-8TzDosu1KojGmEdVuU%7yNI#*t^V!28le(Q
zBoZ9Vue=pAlwI4uQoB-HdZ7lZ2-7MPx6-k1Xqme;j<MgrQIDo4r|SOa$=Ry$Z>X%;
zbOo7Qhzg7HWuXmIXEb+AMwEG!gKuUq>~R4VDs39Xg2qZWepcf+LgNYcdk?!^f`|-4
zhtX6a6;FM$Zfj_pI97|jUZRDq`$s$fWeEz>UK)4|3_#Ps?qL|Sa|tL~_LY}0Xq?8X
zO1g-&m*x8DmlIr{wCfz#3p|;jB^K08)f_^AXu5``)jGqTU_yZS^0#;;J)g(t0*{N?
zG$Ks|-rOzt-|O|)Qr<wL3)qjH70(X7o(c=X)^k|wAWpYDb-~;SzQH4*#J>fH(ZTS|
z=Z1K$JAOv>5M&z-4;>b_Jgk<RtwOa6){XoHp@g4hRJR8YKGyAtf1}%Zk|v8~lUc2A
z0~umUuf6l{W7bUj=gdzApK2OS#-_HVwIQ=dxsJr9Ws@Rh-{<w9i6>uI?k*D6cG{a;
zM1Uz(O_av|cdf@fEMl9~wJvvsLCad!vJ<@yR2B3IRW)QZCk~>qW)4@+gj>3WrUEF+
zIqOujgfv4EU)GUL0%#*>6E!N@=eqy9)QP%mb4iHo531!Ml5$bh4|`>1ZdE3e7@Im1
z+H%1z=<SBVcK}kNY0Q-Cze)>r6$wBNn9ZTYbM7~g8bV0dml}gI8p58WFE!HKN6qJ3
zdO3k70oo_|)$b&c%E;&*01o>B8VULa0UPvDW<4tq#TE1?bPu@cyVzZy5OPUxpjhC4
zjPA$_(X<SK+LYy=@*;mX-w*>0qffB`^D%ACU3o!&w*|a;#J=41{z#v1pf`Omxz=N#
zH@?-P2XtG|!GQ1mWW=>V@gYPE)AJ~zj04&y%Fjv-P6+RE7ta{%ATzHd*K>R^C46D_
zqFZjM(-+r<rb?GfFO;TBr|vIJ@rMhRVeeuUU$^TOC!1yrF6%SxtJRSxg%;3;Wl$Qx
zx!IK^Y|dgqws34yCY3ds!~YdMZHQ<S<m?hOG+E<+(>MqeCX8s3)D~3a)Pt=`8|GT0
zVk6=xq;-u7GY(8d3mZb|qNEV4#Jbd=2E2aC$3xi}4mYRkkP578TFSnJ$(zA@H*!W0
z0TQR`aymsf9Xjs=!pH1;M}-zx@1}c&W*2{h%xXaQ5)j?<w(!t7qRA5BE)}u6()6K|
zL{VJuZ}(tH(XS=ZX|hD6Uo)-UaU<EmrD9~mEb`;*tXS9-?5!<k0Bs(f)CF{z1qkv4
zDro44n<<8)0La#o@4}7F{a1Je*hm@Rp0p5`7SUY-7%A!UKTiwAn_>mf5_&!KP=Gq`
z^@hbwwF@k*zKc<E*&k&Tqu4S+AP#URt<NR=2MgG&&sL^zB9jAy9B}iv32JDg0-5Ua
z()4A#LhyOUOF+dJjDLU%@XNUczXBo$P<@O>9no(K^P9PZ#lF0R%c6foF9-6>ny$+k
zGn~}4UQw;$(lQp8I47V)=`qnb?i4s^)+f3hYoG-Ybvasj&RdOoTMtyYbW2$uV;^Zs
z3<H90<($%uO>4Zj33u&t3R*x~vEs#@h9U#wH)h6i=q@m>>oeu}XewqDHay;f30i~V
z2DKgQKTpCmu>(BTf4uJ9H8g@<v&h!*pZnyyHLaaBKi<5KYkoXwv^(^~ONEzNt~Zh&
zPkao1A62+Yl2($Gy~5Weo=&`%r=jdK=A&tTn(=~4;jO0Y4P`GUwkb<)EHfo$mxIBE
zA6vReM$TwZ;|L_+pX23n-hrTDAMx=p5r4y+cK~4-a~siIu<eBtqBIIPZlLx8LkWec
zM&+!VFJU0ucT{ZdJI6FndLvrO;|7(%#&bAzpi8xh>>v(W4}EQ2_3vn|nz*O_ofK^C
z>8<Ua`gc>~c)N=i*joN^H8Y`vI15m^Brt>84*rkxA^sT^IQ@rjVDFr6eU4k04z<Rl
zw&aFon=;n{I(e5Kyhn+KE$ZJiG$nP+8GXnoRQk68p!pv3Bl<7akZgLWHFjFj`WYTq
zE37aG4Xt0THPM&0Mu6w(wgrn*UZ&k1-m9)jVYYcM4l#MQpMXGOhVh#@t<D+(hU!?`
zX;ZzmRtxi}W()J2<N_QlQ;-xKExRw~cEQA!0~1T?g*?V#-ps|d(OMJFI+R;;NOx^)
zsuX8cew?cbcSmcv<}ba#*BY?X*!bG|O(dIl40%23t109_y@Gvng$JUZ*k#LN7i=Di
z;fl~I)nZY-T&d1!EW~RI35$0D!;RU9Y*YD-f;uANb=nXjnSOrinQ&jd#`w!^*8mKI
zW0RL^<Pi(z!Cu<28l+Umb=ruQyctblnS)7GN;WTGygZ;IbzdUL-@r8l<^;mW?sToL
zVi5tJGTb;1ZW4AX#-gfE*j1N@mKElwCf5&yRmEAD6c54xAs2#4tY(CPdEa^;BvO$&
z{c=5!Ysfq<9<<@aU_G!kgCTYxh`MTqS%*R#Am|zRHk6&tX%^|QN7*OG4110+`C_94
zf_oc%9YC{}r@G~OfbC14wb8>>mNw>t9qXe3-@Cq3)=h}kQf{<F<Z=JKFQW`J5RH||
zONj0Ske^AYbfVvZ%_!&8w+24fa<|$-0|!g0#VXoF;P<w4xLOV6!cz-NW@>SXjh@yX
z`ZFH$-%c*-ZZ0S)w)QJ)#&(2(W$jl3%YALI>_BpR{cVJD4nYe}8!$1*_I|BHwF~fd
zIy|9sb1dv+%35%DV$A=j8fZ~>Q5rq8qgM~;6He&voRYSr(1e28%1$Qm-O3p?Z``6W
zR1K9qs0IvxW|X@|8IHH|%By1)a;r0onCS1I9+DSP!D=vlJrQ-saz@#IIH+Xw<|Art
z<6QUz-El+hpI6SXSJI7;L0aRq+&J8$F_K(F(m5GE3;<dS=^P})0H9W83j(}-@Ju;u
z@An{$6k$<HW+db3w_V*H)z)lofNzrKu?Z55q?6z|gFDgJ^C+E>M<B<aC1wPhI5(f4
zFG#c_d1MUjOeQIkXre3XHP1si+sTR1%|(^w{m&}-u##t$+(Hs%jpmB74|$V|&o#BE
zYADJX$ySfj)M&3LhlWPrMTn1rqjRN2v(+maCdXH3m>3=gWo?&z{&w||IQh4!Bj=Rd
zqeP?Nzm4QA?}@#*1Y>`J<&7hGf-E9gytt*de@L?_P<iR%#8LnKWTIk4-nGP6!R5B6
zQn|V^x!@dd&8N@YF8mMa%~>V3{%maWAZR%NyCL*-^VgL8rjp-M@&zSdRPx(Oeuo5a
zW$cNZ8sy6=1coJSb^5l@O<rK40<p6GekHb<jCvsa5~HBxZAxrIKu0G0r%0mS*ok<$
zipz3`#TNkQeR`D^u74|ue)X3~3Yio}IA7a;1Mfn%c9(vsWKSPfr1uwk3OW8O@armc
z_4P2t1%BM?$rZDOT(Q5n-|lA%V*`DKOd-#mBi5s7I@9_m<fVW0pW&<8aY%?DQZ{f)
z3=-mrWCYNOkU<QQYJR<N(}7}SWJtQ91Ik_A`>eK$K2o&UBRG;;<dH<V$D)ql8tT2@
z+wJWkhDV>b*V{*X{oa1B#NPq$fOn9;gWe(UFn@=<Bi>Q|4tvjAe32v(MeM!a2kbox
zKn3z6s)#GC5FG22iz?!M&^zs&;T=2UduP2{UjkVWQ$%q_6jMYoMiN96nY1V)iYTHO
zBZ?@Z2qTIp;v$Uf^QJ7uNV2@&o5t+u!Pk=BL*BzLWy=K9(kOVIrzj&DfrD{L)uN54
z<WO8PV{u1RayTxT^&aycx2KN8B~N%4t&d0J>&xDg-cwfNSX?saUGko_lH){4d5&IA
z#CP87eZ4(<!lIYl0z#aR`X2|})Pk0i`C4&N0fEfz{i1&d^L^oxHVps1^uTnl%bC(a
zM2*%&{Zp(5ixVQP1%!)52tjU%m#K6zsjyFSmn+Hoi6Ddg6!Toz_xr=Vf3KHDPMU$~
zPMI|JDZ$eEQ(l6SzIE|O@?*samprSBJjS!+yDUFw`QohV3<bFxFx~hl;+Hy?4Ez38
z>Jw~XY2xsAMgEJwD@07CczrDBM%smktBhDZ1oM>kKbNQtB@_D3!z;AX|MiJl3eK&&
z{&|aqGEftBJ?QlYLwJSa`RVlI>mvy2^_Il?4&=oyH;O1hImIg3iQg;{O=P)V*=70h
zus@DbGS5@HG4~x(Tb`n?jE4gvZy19;7%sm@V-@W6h@^tS>pn{Nn-6a(B!-Fq3GNK&
zPF(vX8`;EsY5jnrlo<U#c+rVZA7zgGgGvsOL?eVjsN*UQaQB+g1}gry=>97t7yM5s
z`;d}vRl?+=tgRD}fhb!<xmxls{r@F=#=&SOGB<*1DTJBa>D>o|WKAd>4Q=!iQHAE0
zPF;tI^_;b6dU<Kte~-rEIgLd(^{^SX*bk9#YBLywF;0p$Wiiy6rBF4FFS*D_mJ6Q8
z6JX(T9`<UHfM@u)6I#GSNmKV_Ea&`8gv!kKWK)>4b})%(v@`x$yBpD1XL%eG%Vf#X
zP`j3dkrJ<({<qvWHMN9w+|re5O_HBzSbhvxLn72`MQ`v8mCmSNWhGMMIElJU3j8|B
z#IO;0<)O})7xCvx5ZWA)xCmmcjuV7N!+)Q?{DVp=N<N^(Q(|k-#^wFGEGucVudQX<
zUbRiMdPKc4+^Si-A~NR^nS}5jWR477C1AqLBmQrcQvH`g*rYn>qM-zxsjS_BIuysJ
zd=Ir;J8ETZY!eV|g)(W8jl4dQ=l2LdbI&n<4S=JG+*1NXl*lvzkXgY*YaO$!^_&?a
z*j+Yy{wAPBB&Nc#aGeogna_VDb)Vbt8Moo#V4mSE{?kg{t%Na?X5ZG-H>u!7C8FgO
z9)@X|^RRzL%BOfSI7C?_I-3$`wWF$*rp9J+(uKlv(;A1u`#z1b@InFu;Zkz#;O23$
z5Fr~LnHE8<e;OS<D)znp84|t?$0<_4G~B6jjD+Qye;Kj5%28QhcyTYvIszR4HgleZ
zwh}kIUa+z10!d`hycVh54eu%yVNaMLC&8Z?Xc%`3s`#MpDYR7zB^&-C>w`D(iWHcp
zU8xc?O@D&sdkoxS4yAF{X5inK9$0%2LjE#VzzUIv0e5M71>c}W1rP$3HtpvDGN;-l
zsD*$CleH@9f`#jAyAK^YRC@Lqg$sJ-()m*a4|?RG$8|-y!HTuL%_v3+w|BY|NXz(t
zV#wE~AwG=@zT~9Gz6pD)+;H$l7V&ON8;mou_Jb-rpeo`Fqahapk;pEsST@m^#3LOA
zt&lZ}jTcPfGW`)fipMu$3yDRO=2M%-a);FNId!d1(Su}Xk4Q%1V{v}bpnIL)whWAt
zYg_g;0cti1s|;lQpqAh$tt2wXZIh!iW7}$@kBPH6ym?7(*NAaA7Sio4Vzjk9u(fgv
z{A}fl<8;B;Sa%X{nK9+foS|y1TUM8mMQQ&m>nzhk*Qt2-W*w>tXvel-w4;7mv2SXI
zz*=E){CS42Ii7+lE<}tW`wO~!hmv=aY#BJ?oqj^&q+Tb|qp-h-_6}?wnXNwOJY)Z|
zdW@gmYf!Z*<O@?0qVt84TCra%nbq6%cRhO(QTg*no#-8<V;WDd3+yoWmr<(Zt{z3^
zWPc?GGBuo{f=(-hWYjOqJ~4sB{{((qL!g`YB~ghoib@#`cCJJ-IkNzd{tm6QpQD~H
z;NLV{$kBwbPE|L4O3A-ba-@|Q-k9XspV2V&QSKyug_KO#u~j&>d9XT>Zbx5Qa?uS#
zc0cXC>5wH?MeQNfiU+N)HE0@~<MH66E;dKNzEQGF8!5x|_U#dZtNOH$GN(2P)wX36
z{10htWOMuGLm>S1uQdckHv|f8AgjpZZ}s_|J!yUa2D!=Kqcy%UQD2ih1dv_|u!V~0
zS)?YF^wo9l$Tsyncch0#ON$*v)<q69ikdpIj+AVfWruBro-<)*y3)1!AkU<wm}A)$
z8)*P{y3&1h9!meHy)#|^XwU;UqPL{$pHjZvnp%p1dNgG8a>tZbDr-lcH3ujw#A|~{
zU^rjYmr9fMl8~D#I_<xg(Oc`f)buM|+pP*jjN3qb$8lIMV5wlFj}cc5&C2tPk`oa~
zAY!ZC4|QWgm$tG^IHZ+rlxkBFWD{#^d%I;lMHBy4b6VoL1UujP01>JP@tDNg1(tXZ
zFhKz+){ef_Ogl`j@Ar?<Y73VOt_Ug`d{`husKX4Ni>LyeK^~J*0`q2^$|B6J3tE7U
zopGiZsH;PGE>eT(%>}u2jAVtinO;0uk%H&R<-zjGb;0K4kX4fp^S}+vV3TG#iDqCZ
zW&oCA-~oC2wP+ap`}n+xqVpq@)H51#tGp8fwh^Wy_(gBYad>k2iZ5vCloGLprR2rd
z$@1Q27l|O}wy=S@*gnPto#oV;=ijd<9#FCoPDSIUeCQB8JEC0*i{H~ckMQmU_)qqA
z^7wKjEoAG7&2#(a4Yl^MK16%jruIA;O=<woPc$R^yCF_fwWC&s3Z7DS8zpk<jQ1D2
zIHg@|mQ;Be?lEO}Y46^G^ySRtOxD(R9BUZ~S@1a*@P%;FMMVW?h>5D>3+p?Xol&~q
zr_1jrAy74jZ$|7u8;hRxMXLGrtz+pgQue1BG$#^uI-Xo=B)t{ny&3(}8Y-G;W65h=
z2;`aY<DwH6OH_iVwP8vI3$JTxU;r+$0wQJX5)f;o{Kpwsz*{4E6+bNu+w5-buxmqY
z-bmFy7!&|tvQd!GnGyqxfHwwL;fmehvVz4_MS0@mL6!LiV&DPr924K_zpN4Ys1h5?
zV;U-vhf$u7)dNj_AUVBx5gIvYGHjSe!G&h<xB@FCHm)y1f|>In?t*6Qxdn}AOt))f
zgIZsx)Ry0S$jIsydIFJ*Muwos81AjZM*(MK!iNcv-A8%9PKhf8+Qc>GQ;3@b^anN~
zE|eNJBSw1|HSZ)S-^P0YDI*LF6ohT-9A<3hJ9=%diW);>ltw7fbe~IJj*>6<2Pk$F
zZJH>Cn@6lbpzq6yn!ZCLWTaTKlq^a6u!Wy&R&9-gs<+1B!(8Y)h1d{iL1m65W)Vp7
z#gw>lVCBg8opA`$#CUPv4J?(F%qyTr+1Y3;7|YBUL4rHVb1LV@$b>l?Cj!G!L#du~
zP`9x56PvvuFW+0bY_U`~-+19(oQb{Qa=X0g>*&I@1EmQj4gAu+7AIpWozm2j4=s4I
z^;GEv!nI$<AGy}9{B}GlCxf`ExhNA<%5Imqb@~q7bgfM*Xq4Y-%uTg)5atRk?Vct)
zDqN{nm$5pm+k$8=gItGF?33CspH}jy5{2n*+R@9i#QzkDtOZ_>+F7Ax*&lw5u#ta5
zH+uA)Ec%&%=!l59b>(VFCRFYs9GJY)@nhWt_M!OT6)SE~_Gk?m%*vNQ-y`fa`0qWN
z0et5pEv>ZHk~nYMQ|n|{D~74znR9`Hb~poCX{1d?K_mifui%bZKUz}+tV^@&;J&$L
zFO6^?J7%07nx?nA68>w}DtBo$Z>^ePt(T2kWhKevHq5KlE$x8e6}C@rWq0N;misN{
zDpMavE^8J{uEqs6pXFjOSFKF>1hV(b!>vbD=$z!6g*?_SgthpanmF6(89h}VsxL1P
z+SK&MO~-DVcWr$`{9C3xN<Ds^->PGwyWIQW`^(*r$1j&4a?_;t3IevOp3tn0@djrM
z)ss950Uk%7#O4$7Z13h7-tfGQL2I6eXh(Bx^V~<%2*MNZ&p5{}MDD03Wpw}qF=Nni
z5H+ex#pM8U#^wBt%e$e(WV`-F44LdN_=tMq?J(%b5VeB<mGhYyR0E$Nw4hyOX9zp4
zw<>Ok>=O{x_efOVKmabwcW3pWx5=u2n)#bqNpR+6Z(z|lZY0-nl>(9%utCV65a^Gu
zH$6L&n9=&{e>GJbx1P%W;16A2mHP_4H9D1Eu$JRqkt6uM`#$Xi-7$kw<+!779^Cwu
zaXo!;-sNXfeDmOzZ+7p|cPsamzg@L(NMfgL9fK`xwZ<Ze)1PK5L)4BxbMvR%S2-5h
z@#a{N+b|Zo<Ho|>d9$%lz2;b0&27g*wQR+O%6;XrP%YFu<J*nJmbRK>QOg9~kf43U
zC}9oF$d{t9_K_AgXhsFN7i2LjLR9gZu{x@sEKQ#*T|S8waJy~`d0AzS<XE;C<n9qX
z_T#63LBx7C2c!3$ef}iTGTPU-KHqMyO?hAs+TBqrDtL1XE?No4uy~kT6Hr(I4!mmA
zjb|hw37q?quvl!v+~ce~K1zv0SvSZ<xo2xL#FL0J7fAi@6tI(^&v7vCR?dK%sAq(T
zMlzdC#%rqicSxet%1o5<_)XP`hCJ=JFrEMHs`Ix=%H4WL3@tILk6)=m|2x!1Zxh?w
zH<f!js-N2o82bg-tb%LjUV+93>0ct-mmX7qMbKVfs+2km@;hu|8kxmsl6#V;QzuDJ
zZc}z5ok^|j-wuR!ylJ{iZ{eTBO~RABIZ^)+KA<h<1E;}-O10-7l!7xvDi39Z^9R@y
zoCzd?Hixno976_;MHOm9GdY(P6{aDvT2QVj9AQwuk9{gN={(m(IAvL2bze#QKS;SE
zI(-zp3+8O}T>W9&(K-LqFj`!-`h;Cn>)kyy-mHMP^F98Lg&F^MgWj+w=!0+gMa%WW
zWkmF9K-OE0-mAaw^3CT}MhiVe^s#RQesWYq>nk}lLI*4O9#0L4J3(O8*XR$1Dd)@R
z<p6t(P_ih3GhzQ+n%0h^<fR4c(>Bj4Mv|9(4XZGAc6W$-O=FM<LP`Hyc?Usw>CsR)
zh}yL)91^9AbhW_U?{a0bU!x>P$!91Tbv9|?AbVUcvIOm##K?K6Rwk4?B&FnIgcOwy
zEJ|}Ukq!<p{dWbfsJZ}QuShcpAS<ix4<YXTf1p+V!z4tq5g+2rK@O<YcG7}!;wF9B
z`8i3W>9S0NWty%J>M@h~7|Uf_bWUZj=!t1^<sGuMT_G$H+a7Y2crkM`9G0{LladY=
z%g#5zwB!cQ=|Cl_dd{8GDde0r>#xvYG#=Ypb;`3G<iI@A2_mMQBwx|#r5i$j+42N!
zMTFNK|9f~{86Q(cDXBQoeVei7;D>4dTlFEWkt%1V)mU0S+?iEexVK8CrF^JqXG?jN
zNK0BFIh1U^N6AY8=Ok3Q)L2UHV<SkaMl9DO11{<gr8De$<;A?~31`)$+}f_~_PP;u
zE!mX_I{hBuSN|XhF1r`tz!Mm8NK|<=z~hi>^l(rEn665Ok4DapSHT1Ut4s^Ytq5fp
zh2Kc&@SFxl+|(!%fLhUQtoAzejd^S4FTP<^<YT<7$zLQ5Y;Wg%GD|J~5yhp)7?c)2
zIt!!Y{}H#Gm7YY1Q6Uy3Rih!nPl=aw@Dei|_}5J+6&0Dw*h9rcBvRj~UQ?JTDG8nc
zaa56MDFR>Av~DC^ZSDPC&729#nUNk$?2tDREoiG6aI5U>zw7Y(Kc)}*OANpNPu0N2
zvG;#mWj~>WVM+KushlaYe~Mh^==z`2t!)N1+Ou__DvFF{25$eLHQKEAj?q53d5ynS
zBijqB#7MGSO{CCvC^n(AL`1O^XgiN3UvTfAsArOXB~n}Geyb{Ri%>%|Q5Vqxuook0
zUQ0wxAWDT8N~co8I64!H9%NeBwry=PVp1~dya=zlYF4FDYBq3PHn23LLEp59k5Kj#
z>b%f=0^q{p-Mx7@JDr)bs>i=>eqyc6CjUQ@(dd~{U+goCkXRZJzBC!IB|`jw0hnE@
zmmY`dGM4zwY8C)Giu_+N)+nxJrUUX_BTyqC#vGL9WnI<f|03i#f9v0>$n3nM+t}|O
z*(PConwamU1_2foX0h2PktOr}u<u2BG7zSv_QnicCg6Gk#BS<L3EL6E&*Z}_9)UUY
z2*e~*Q|9Q_yT}2>gao(PQez<UqAK2^<QJ8kQPRemf2zwpnwhF{hA<_7bZFA>;&t7f
zQDT;vS>^sfRd$nWIw}2gl^VsWxhAR_M`pN{uvd5XsW6KP&m<T<s05Qs=7Wga-e$R`
znR@;+EeT0P6Un2=6Hu4JNx~-z(m9S(#O5Uu8`reLgh&@c8xE(;kO|<IOlaOyJ85|=
z{p)+eG$AQ8*znDUYfgG84CEAGg77SJ@jXoh4w*P(BAtQC3BWP_2f|GK1870;gCnp6
zGLl$>>I}d3^Gd&@gb$aG%O--OHQ0=XLX9=`%DXgJ^5u;(!uU}!M)pDi^1<={liJ+4
ziK1@KTWKD&^h+8!r++A>QV1^;SFS+Vb1-{q>(=UcasxupHhYAh`N8td00fy*H^Ur-
z_%B!~&LpmMx0DKISv;;D{w*c8)V`fu6ZZbPO7(T+ZlSlzD)L5ui<oUpp1aN)su{J(
zRs0Y7HX-f=m~N|j;LRRuLEH^uH);KhHAb``Ul|kQ12quj@~I=bhgi%k$J-8zF+@af
zNqANSwi4w669eFi0(7o-c{pv@P~<1y#&oqf)_26=@i^s(V-_BJoDg^Zq8+)Qpm`R$
z&7DI;+yJDOr>0~YbW)M(bQ+{b^t)D5Q*%N66xtPzK3!~wnzNI$tkLPo+A+kjobej_
z719n`(7C{!1Lkq;Xw6gjI}5xQ>www~HTiI?@xEeoKIhcY@obYXpDdX-li2w8U3^~O
zi^;$Wr&k_Fa;y&^c%I_#@fmSV!hcG|)NaecHbJ=<w#8(aRucy@5rOE|avpa9#_m>k
zeXe<Zu37a#)GxLO4g7sKZ8>OY{KBv4a)**ZCD)XE4@nc7nIlBHxgX2<t3WTsJ^CDV
z%Kco)SD_hnuUxpY!1`lyH*e0OE3Ql}d%vmmTjcRXp$LC0=km4j&5*2Jy@jQ<<Xd9I
zkc;ajQ$Bz#L>SAH{1sFY!Yp8=G#dA_vaH2?O7w&b+j5BYAbq#1?JRFs_dz_T<+Q|K
z>;kSN{ADL#YNW4T!nFv^zASFp<mpNA7p>*^s5^?aIl&gXbh32osq-bQESImB>@+D`
z=^cv#%z$I;)@Y7YcJUxLJ>Omy@$zh*QBuFRG}#d@GH#dcf(k=B%e|e8_{7*P4Ndu&
zLQU%&B#Ygs<`xuP!*{V9wSpf*nZerB5y<uCxvf>VY+IvHdXs;xCrcca##yd$w>T<n
z7Cnm8t<BSCJ9^gMMevmz!)4Mi#fB=q&p8Q7&ULbM>q!?alT-3^U@Tv2pTOBJb=T*P
z@omxvIM{=+wH=5yq7!^=k~-)CA$E@CsG9=U_Qc<%r0s1lZg-i$1AufXQ|zp)sDINa
zMrkbUHgANHN<0S&Z`3@s_YbrsjDhJ&rAnOSaRQ+cit;T+W7|F4k`o=Ii;@}9SfYM8
z<#cl#l#~4AOKbv&ICD%uFK5Aj<ef*zA22+JIYDg2aw1`T^dTb32=&EPQ6Z@s-DV4c
zCpv^&5BgprVk0BzLArvZQ*oLl6N=0Gi26({=5Cyd$KyL3`h-_Hb^5u-U$!I$^R5iC
zR1%_I{`MJpnrrnHveWuHpuQrH$%O;f*THa*n9x5E(3<%deixD-#c-BiU_a#Ph}z?-
zk7GD9jUD6$InuK)97e;)u}oe!2yGOQkRj&uj*jouRs6lQK=f<wk8;#z8YW<8W2~_&
z++{uQo|hCY?llDUu3(t*5sp0Fg`To_6VTBZ569Pc0EmXJe(^aABk$a@@y{pvH2gRL
zRT?_ZXeS3M?PgS6ewXF<5Iklt_r{SV{hZeaE!N-IhYXAZvyhwb(Npug=6BEUY3wJz
z7s_mpk@Wk*{T#tN@Jf#HC^g2yz3cnKeQ_9=QXH(M6z)~ng|DiO108KBB+LOFMjE$s
zz_nxa-tb>>;Flg?y$k%Z#Z;sjUZlg#JB}l@mFn+|5GeM*g`Q@=CZ_qd{%OBR#%f4O
z$YOIetFIKNW>>Sju!DwF-%V>xzFrcgX4igQ<)Zvr!W3strML`G#}D(T%yE9uyhztw
zD5bi8n*hv763*8YgkG4uj$vZf6oYZJnYH{G&kXP;4mcj?89gI_;+U0xrOV&crC=FH
z3!+-Iv+&vw&9OzBIHGURBj2P)x^x<$3+MGaDil{@8fRztV)6y!)4xHlT2r!<#AJEb
zDw4eAoOaO}{^@gkd~sa}oQZ=@)NkCA5gh-gZfT(Wf2rKhDESvkenVwq56W?XzgYcX
zVF1hhF6due#tE>fVL3a~nM`zB_;HI7P%rm)T*%wpd~nM-Yvx>ebVEB9X2+PdKiAto
zNx~6Qij0v8>i$1dw-j>FNFx7_$(4KD^UnWUhY7|S+_JPKO>--YOo8L1#pm&VRn15?
zP|mrjG{gbIvhg^s_Qh!CS!(sCxJqOYIZE}A9Dps9>0AM`<6`o#N>iEa5Sn=67E*3O
zERv%LHDMozIVaF`xe9S1mPn@ra&F)@wja%%Mjt-F`=*U#BcrV%ImILV>iY1|p$kb&
zppK*{0O%4e(j)>8186`wn8}~C6O9W1a-BDax)yd46azci9cKNDVK+dl+XVAg-&7DN
zZK>C7K)<&J$fhmD(^?C=5$1~pO(18;_^*b&I4z*umAQPFLtTLi00k~c1<;L)H))@g
z1C)26dXus^Z(x0(0V?7Bzl^G25OCikl^i}|zZPHj+O^$N^iczM1sVGNoF?o(KA%l*
zE7g|f+bL*n?zaj=TJ&vCT2Lqn8f?kT&EWlmjYie&RcXO|33YtiSti%}Z7ySXm$sb2
zy<d|g4;mATw<NA|u^4PhW7}uN606OL&(B8v_Db1?0C_Oj)ft>|w&DUMl`~quF3^;s
z!@#*X4*YX^Wj-c1-6r_k6#2iWWFvcSKCYkE>OINL1cF+5!t@btSN<*vDg={`N3=wW
zGMOJW;UZVtD9ScPMg4*r)W4RS;7%^O`PmYcY+o!czho;sDoer^1B`=-mtbT~z{H`9
zEZ;6PM>+rT07lph1_!WK803ke#&B&<SU~F0E*Ys6xk68*`y6+!H}!Kocv^u{8=@!D
zS2F&m0}eqqS8PXNq#8Sj5`_H|fmgdM@M<yaC74Uz+yoQ22ogak4!k-_Uw(_lIf^Ua
z(1@LM0%N{I0olUgc_=WboTTA&>7oj6Qgg&=Y}-n_tcw?kfZGyYwHKBDAGFkeP06n-
z`3)t%NwPNZj5|)d<e_`!0KwBumQ&p7QB`}4S#ZLiq~+rZgn^Ci+WH)8YH0KKsT>WR
zuL2UWX@k*tjze)}tfC094zb-(u8)>6(dV6jQzwM`LzQ1vV$$$`N(~fblI3puMgLhh
zHhxyr9aI<L#3$JoRjG&x)m~}7qxH##Cmf_#mkE+&>hHQ+`~v>|Z|S=Q2OP5cugY!O
z3ty(}A8IeCjYMXYHSfHx6_rrCB7*{>h((@Y`U&>LVFXxvQd{lwnh{Q?#k}jEMyi)z
znqOUkXu^TIXnE;E0iKOP;#=5pNAcgoFDGiT{Yzu$WrRM^-mH0r4&VUxl<Z<M(6Gb`
z>EpSAZ?PfSF>nVK&UP66Y+qv-JL{2`(Lf;r{;?qUauN@yl>e!ZDL|1d_BsL-)sEYJ
z<c|Jlj2{7LH^UdzifZF!;~74lbS|s@&%=rjhG>eViS@r2bOptr8*86I3!0k`2X&rL
zt*5gTioZw4#0I?<nFDb%!mL4dBO312zHo<~{*!5p66dln9D-jEV##vF0kfbk9f~TK
zSRMH^I6veSub=_0Ukrv#qz#pih+)UeNeR@UkD=yYPKIMadVS~12p8Zj()2wsKNb#C
zuC{i866sGkij|G5N(kHdO3^r$T`j%#uR3(ed;YO|58C5GxD)+8e1sh++o$7q!(j`<
zI82WRTr3ry_<gPhy?(_~q1S&Bj;xP|yG#=;dz&Icy3rfSST7AM33>VkXp&QYMiFTY
zhXxrY1~!D(n`RN)KbDZ0)Ej|(feVWwazAFm2|kbIA#rm3sLMAeuQsOeXl-`W-PMGJ
z$v6;H`eN%9dX?6zy`qOq$QpI4z@({7N=Tpyu#wjOvJGWaQI5+mu$LUa@bG3qlERw7
zhSxdHKE{3kln&;h6|01@ILd6=iK(R$Ctk_qG6?=j@-F&Q9DBtG)OIKCrW3!V4XxqA
zr|F)rxf`ufQA=5VHY#Ie?1Y~I`Cv9j$DQ(nyxEOV+)Zg7whq(R(Sp>yu7`=gcn@XP
zZxOiGMxSbrY3UtREJ_VnGfY3tOSZ&~j&hf;Phl(6WP)16k{pRdv~-}B-=XjRe^k<f
zxJ_#60mo47cISz|sOQXRK~fz5zbP>s@=-nfQC&W&oCJpcpD6hSB_C7SA1L=tN=)rz
zik+%1+iIT~)iY+Ot?UudcKvOTo7nBJi!B6`u0BAivuU4_=Y{ZP)PogQMP}|Q;y@FH
z95>Xk&4=2;jyNskQj|BTbLEoz)<^Z?|DweB4<wqrM`sb4dzIJ`|KF>>KTA?BuAY_U
zQ!uCKxq7>7Cz|WSwvP2!TPNDF?dJGptrF3siEIY6BR?IS&y2|!0u>2C$r!YlIt~F>
z<RDqOGZ$eaWEd**`ccTawk?F%j*-+Jk#`9!DH2F~k9<|TW7-5_bVJmf=FzCRAK@dz
z-WaDzmV!7JD^9~yP;~evgQd<Xf-zW%01!xe2Eb9WRvbXfHE0mG;J=$EIn-Q|Dsa~O
zQDu;qR}ZYfYJitd&;^cZ24=_O$8bzXDNKV?dHxs_o>ZRSVJgoe@fc$pQ{-E%JnYYp
z>Hg|(ghTis{oBScJ?t@A)o|EHz87{R(~#V%4;Vhq-~RaluYil#@5idt!Egk;gE~~9
z)}SjB^TW`tJHR^@4g$=THv5hpFfL941_mL<RhBQep3-g^nHPa1!V&PwC|CmuRe>wO
zD+7jC2EYcrExdA#tGw3}<CPC%(X=xhXjwRYLYTzxO1J+xuo}+_t6cq@#N@o|Gh(2O
z<{SWu_in&0Zx(Le31}p$*&s>zPt5ri=s}GZ@Yn<+#@xR76zw>m(1sBPH@1TeTJVk(
z{U+tMgA88<P>4y+Hc<F8HTDPe-xqlBzoeY)XCtS7UzY|CeowhSR8mnr17#bxy#GhK
zyK%Ss|5;^6^n@jb5JB~cVwOGpIOU0)_OLrm3$Xy91awfeb};*vzdbm96k)a`Kxpn<
zx4+waHj_|*s_l%vo8%|5Cqh*)`-y3>UHT_HuDpJA!DHJd1%DioGvp{gv#*%RG>qtD
z6V3-I2uGak{D))0(L~iRClyHT<4Fid-Is5di?Gu{R&F93+1gMXMY#kZz;=vX7dpx=
z7GkbVcL>Ag`TY+~+FW2J9cgphW~k}JOJ6lD<Nq%6^HtI^k81sBe4u5vh+yCRy?>p=
z%R5_LN_-VvYEdiD7@PLsqkNVlIa=Ef+3S=_CzZ4*2%|5woth!IZC@J_{mN}RU?Pi;
z?Xj;3PT#JN5^<Nv2zRzHD|<Et4rTO5x1aTviARxVbNn9TSJy|tB;7Bg1UazAyG-m_
zYDP{jIiznwjF=0~VCvzeFoe&_DFAkaAjAkaP_*DWflx$j+iU60yK;RHKv$e<K{I^e
z)BeUGK_9k<I~}8$(eM>JXeA9F2STC}jEYK~K-NrS@JX@)PUOKU&<)_*HGm>{Hi0<~
zY_w|9KR>Ba9N@hYD>+f<cp8yXYO7wk{ZLD<tkKiY$crzM;oC(Nd@v+l9wX@tX7<4t
z<2@Re0Hx~Z?4B4|4pcyuBpav9)lc9)i6)xTOK6KV1L&M(T+^&kJ53piEi<M@8aZEN
zw5ipyvapt(Nx<adnu&8Ic4vX!@{6nQ4tg+~rqq;P2XUP16NDrDGcKDt5vKV*l@e;2
zqg-pFs<ZjNLXt3#gbopn3rzSS4e)1_cNSE*t}b~?Ykjv$x9Z}dlHsOs5-}BRj|j9o
z8eH(*Sr30)4H@BVDC~o}e4e|}KG}h8a6sQ+N^B0Z9g(A<L8TEns)n^<IDyg>5HgJR
z|97=2HpijEBdTPH;lDrO#`9Vchbh-#FS6Ga4XYT=h`>n^HX05b!lnuvn$w&2*jMQ5
zLwdDrH{UqN8>VgLkvYXr+tE;-z*X1}_YJbEr2<i&vaIhyMwoK3SM5+B^L&)IB#SPE
z-YzRWW@ATVw6T+gv}2w~JV*_gevEV4*DfCG4rDkPqLYsDm+96bYT5H&HlN(0CqLfr
z;!Y8L>j(>RXSkEv#6_+B2&Fwxm^uIREGoSTqdfb|3BA~PX;BhYym`m#M$d~Tw4;S(
zYoYgw?A`Cbe)smRcztixh8Ach5%gUt(#E~cT*H}Sa0EtwxbsGm@~4`84hG_q%WS7X
zq$7&)%E`!K4|0zCgCSzjB6UIc>|)UFT7AiU0XV}Kn){kbQaA?~`j!9>$W4A2$x32=
z#2bhy+;0s=18HN&WykL!ZSscrUg>dnQEwcL?{4(D<J8$B`HE?5_lCP@Ww$p>%{}4R
zIw3UBx$e113npRtR-y^)TrY)V)E^5Dgk#<abGet?UT=rx4u*TfF*ncNDBQ~4a4%oA
zldsY#!x<zyV>gnY1dq^%Luj^BC@COnyj`zk0owZ-`#Yl7>fQTz6Nin`#yD*h8JX{(
zPlv<(lphEWP`;b;g4zy_(B{$L*p1Yy>3Z|(xVz%Ldw6eY%X^dFUehmc^XBF_wR*6R
zH}!h^?Y!g{>BWS2GI)=JyjN?r#5|<^yTgOX=SI8(0J4Lx=`KMQdG~#r-;Lttpm+F{
zJl}n&ao9%s5brt?9^tEw@TQ~RnRz{RUx&6m?0Plhe-<gZ+eiAVRj$9R@^m7VXdG=E
zYaC~eIEy?y79PigG@k9ogm)~QaHmj*$GqdL|6+Iyj)~kyQsMEq<ao1WCe7Ez^?7Py
zMkk+7@GU3!mJ=JkMF+WPomv!Sp1H-oX<y^;=DBS4r8S}_^$x!F6tTDVhZAG&4X6JK
z-*AREbibOF9>F_HyGO0(x0p@EN#^a8elrT6eJcvIRO3YBmd;+!9}aI(7)w_2$?zmS
zy`A>*w0Ax@y?!P@#=L&^Cd3J+y{F<;uicpS&$Z$w-{IX!{N%jI{y2W}T?C!U`2SE~
z_eYidV<q3CgkQq+=2729y}2)I_ozqaqs@o_;JLrAXMRx0>v~X}<{P_6z--UxPX{K9
zx-YLRPS#>yoE;lnP?{~8Na`yEa{ij*e)x~*6AI6W&OvBN2!J-R^~d^#|El|GA|zN8
zXp7jfw!0Z7y7YAlT4WyW=c}g8@Ia{BwBTnT^o<}ck2vxOhoiGD=8+z6G1qIoxER(I
zz@c_((&IRs=?Z@svl2t+XK7{aXtT4<Ld?9G=4R$F)12|E)za(=k@|71hbVuujzoJl
z5(&*mP>d1@G8*4_w{v=<p=R%=XJ%dX%3AvLxtaI_=%aA<>9v8@5WTCmOwiF_Yj<o&
zG`x;G7o&hHwh^3lIb-W7DW-H|ZLrnelPH69t32GSEA>!Zd2L6lUWWj+&iI-5P*l#F
zidK^*S|u7)Tk@(n*3&#={}_979x_%|wi$o<Sl@pn5=!k7>w61<B%@Ck?s_}+bK+JP
z7g<#GrU2khDr+BH)m;<08qp{!(Kf9M$0+$hRCug<{ZZdX9W5Hm?^jt4t2uubfwD?W
zsQaRFCN>e(==6l&tz4T9{Zn0L87*fI)6nG&UB0A5$_ERQ<9}I~A5mfyuIb&yp7}eK
zh_ZBcGXF)nDV0qskyVX<L1q7v+{7tIw;oc#JCwXri8vv*WeLU!-=lr|E|r<27%jRj
zlxT37IPXnGy`2MqoK{7f=Dr>cyP9(|Gb^&nw7PzdDSOjB<opVIi3(_ENuc;g;b|=D
zkJIm{pd0F&3sGYfwv;aZKT$W0>VAY=lwWgZ(QUg26`7d~f%RJbQH&86pT2jk^Mj%;
z4*7O_$>Iv((>on?H+Kx1*YvAVZ>v^CWhM=4d;4nVnLJ!RVl5R<JwC@qG2V}8Dw6FW
z?f$!Y<eFCX&v7S_lPiFnDV*k2wluP#an5`eEhc%IyVA^_Bpon@K@7U$wkrnbK(d73
z%tdpT$Xoxx!gE)-G1Ay^5EesBhlJvzY1L4xvdnR1#LK2+oaYWAIzK1og!>0g9N+?D
z#aGr_1xuG*kNO_RjNf8B-QJxo03kV-c84*!(uaV>@dv~MAJKq(3Sn&}CfHmiuaJ!I
z!FzBL!+aiF!sP3y@t_IdX(%hWXQeL|qr9O+&vyCWObDj<iDnc`o|YlCMa0B1)q)~J
zeGw83sv#3!D^`xK_2z#64=otmz?EG6CxbzlVge)26Nng3l&nrAq?KWq-W=3CCoU;y
zv)A>Syjtu~=~*0xOcD)SYyynp#kC`xZ?Dj4v3mf(+(M`sX2<N=g|k1fcAyoY+Qm!j
ztj{oG_|6hLrL{OnvQ~PgYG*W*R+dYTn>z9Fr%UxXN~v!wG)J}h5EY9Yed+%@Jt)Zy
z&cR)+YUphgAx7Z^yFi;)SMW}_KC#1pOtlT{8R+~h&vk@ud!H`fuf!nfhm|up_pEZa
zD0y5BJfY-@E*F%T+*@eF2XFuan!Nxg8uM-6c9CbwMK>lDtl9i0Rgjl=ldksj=qB-u
zg`odqZMM@it7vcXhLs#y8loZc&cjxlPWRx))7ol9^h{gLo<Z}3P26~ZK+B&+{T8ON
zr9Dr^InBDoElEuMY`_x7i_LslU@{04a+4^Ba)Tw}UM_$Y!75CP7jG7p9L6_X5%5EX
zIU{PX`=(O6-i=u^WH<Lsf+7pY1&Bk!7mJ#7vzhe&9|ovNcRbjlppYZ}fhrlovnesf
zrJ+MZR);i_H&msM1~@v4kdx;;Lz+@WuR<)?tQb@lfa{QZbo1;e1L4B)7`WQUNFUxt
zhUie}1gAOOK^qq?UA*W&$NMf^Ks@eFSy<uHzoz5`C95R7`xt+8j)In~`_kNn3m5rs
zBE_@%CGB{H>;;OkCgRV93)Zy%MINQ7?-JGH*TrRAVvT>{BE8TpJ?9Q;phN}or|NSj
zHHdz@<o_wDd-l?$3zsfg`!EP6-Rl23-F=!Q%2x<)J&k|*bL#HA$|auZu#gQ@7N~4e
zIYCQ*N{NvGi^{#C<kL$2rjqYe@&if^E4iWMr<E)yIih60l7FS-=au|lO8##pvfuPC
zDUqR*I}zCkiKNPmJTUsh$P6JX+eE(R)qhr(tqBL_*mgI0*L%1nYyY}vDWqLFq@^4F
z>pNS>rjnzBiDX~C(92;AJBmGpZgTlTUtyqldyd)zw}H~Si^YLVAzSDx4sskrUty&9
zjfp<)?Nbdx0qt>r$=Nejb|${pKhQ@DW33t=D(v&svuQfj`X>yee+CqqM>(|3L^#qR
zU(Hl;D0_{-{MZF$%UL@zK(Ut!zQ>a-sco7IHK)Jv>Ba<IPTr(b?9`NG+k_lbWKK--
zEP{O&6vJ+}U7iM!+!^Fw#FT=jGZ*1}49-ezUTJYT7bGD)$O|ljz5kyy2Zy;_D>{n?
zaY37yBD(n=Jzz`jc^)8)?xvP2VxU|MS)y(=i8mKJ*2rOjWF+g_=o~je4GZ~hf#iXL
zMlv@SfUO8y1$yBD?S3tB9doNV1n3WTwGh~IK5mXxzZD}~&dd9$YwTMqSW_1zG~=LY
zjSpbFWHN^44uK0$Podlzi2Q0#*&T%@c&EY-g-L`s>78+K0(1?kO`OHJwnRRNE2$ll
z%hYEClNAbx`vp_b%OS<uM{O<d-(2l??HQeq``PW@(Ak1L%iCJdO1z1n(@dg3`=+A*
zr63FGuuWgLZ;7W0DCKjt8EdP{+PdGhh4^rvYpwki)(4rBa`;pHqy#32jqFK}ujQyl
zJFB<(of4?#uw%}d9kZ4>igH_V0<p~Uh}PU8m;a~yu|?Rq{+L87KfujRYth!KPXlU5
z_rdA5z+@Nbpx~=&UPtOVI^g?NsjCCU{Cf)8i^O<*|Fe2V%s`Zy-qvmTxR&AFJgH?U
zDp#?KwHW(Q&xpE}Z_^Mz3`63d8HtdWy<-kL$1RIC_G~6#CkP?G3DUdWW|WZu-?spk
zf*~dKPp|_LB=4Y#I@|o$8)q!QbQ4Gko_OikoJ|}0l$VMCE6z~ll#@&~D?fN|$N*Uv
zr_f+}SkA|#!`^^5_>#_6!Tl<eC`$)D62)4*XO~T?(fV4#wmr?Z<WQ!i7h(y9BA4_#
zW7LJ}HNrSiU{t*TpEj!giZQNB7ojLTY$6uC9YX?b?_-%LyG7JbCX+TGh>A*}(10J<
zHQxcVDw<tEplC;DXld(^1>8NU)pCj?R?kEO?an$n&2Lp{pO~;cz<NwG;Q$u)uC_*+
zVUacuO$%(b=)NZOe?%Xz<%XuL8I>74zz4R2fy+#@84O%enSoRa{4K#H^O1spaZ$Z3
z&aqCSY!1Hz%M-M|De!UAnfpT;ud43<8zlyiEitI9Lkyp3f27z)=_l9XSQ%SBa@!%l
z&;rhqf(Srs&g;KK5@rEIZqB7aV&>5zWsu6|I7Iw#>kc0;9h2Qpo0f_DQuyNMC$bkw
zHPYyedGU3FFr|9L86)h*Di*E}aGca(LOOYm<G+c5#2XZq3B8i3N?%1;nKMiI9*{tu
zANEYI;c|gB)z46?3tu@!c<M(-l*11XMY37K_G9zThH#~Fub3!6$fcZ+_j?%G?WMlk
zg{NVGl?0PAm!3TTB-pcs*$ikwKpI*$qV0nWoLi$BM<*pRV*H@y(g^gk%|-6?l9DC+
z|6KhUaKQ193VuaT|0CtpRd=+l*2|kpPonH9>gpGw)Y@v)-La>mE-VC19a@)cM5_;B
z_SP+<-j}tS_VH%vy2r7)Q?wfi9*l<~8d(Mo3Y{kGry@6PSz%Ibn!|brMg7MAL33YF
zn|9bz)9y}VqRrXOuYQ8<7WQ`NGx2dSedd7lnP@lVp{ccmd4Yf>tFR!u@wPq_{TEsf
zC-ZM<G@I8>qdAPuK^n~w{z{`MyXG8!ck&mWejU3!cgz>s%`vfKV$4v7V-csbF{p{o
z5?ea-tmqA2MD>mA!0Sc_`pe3#56H6*+qoUI*Mko9+bQLECDUMzB2!Rrq_5Q9{+1SW
zFfQSaH-Hvs5TxC1<C3*8WNo~6`!>Acw`^eu*~Jdibj~+>7L3FtFS6r!7mVtlH^TQx
zpZDINixJ-y4z)C$<7hfZ=?D5!({%0*<xM))(scfuUGMaE#G20U4R!_N90Rr6G@TqB
z<?ZB~xJI*yKPj5deeCw#)Y%`R_e8VF5l_-{hI_m*YNGXAKS0mWUEXYJIx#k&{vp$h
z?g|cvyS!b%kbUI#dE=Hl67CCkx%u~Y!)EW}8Z9GVB|G&W#om7}`X;C8Jc^lE3JXGM
zI`^5TbAO}Msb}Kdu||_N_R~he+DNP)3rqC)U??qOiSlmmfaxE9mHr;5%^nEOL$+oQ
zul|@0QhxqNKAyzy;2=a=+8+(juDvX?_Aco`{akRuUE`6nLpAEf;1=-dAz<9$my@ia
zWu84`&z^5RJ3fEXw5WHgu6sVN{cx*}Ykw^^bvx`Gd8vT@x%D>9By5X&M^9%Pucog4
zDmIq=-cd%&=}0?93rozy-Z9fB9%&r4IXc3497DG`=pAP?Cbrd)9;H3$DUZTFq2Dtj
zz0>5*=x5r#TTI(`yfM)^3-i*Da<Cn<c_KW)8%{Igig$O$;JDzTufuDN3N<~Q|7QYX
z8K9iMHmX#huJT|9G^~F)dCDjpzz*CTfR5iY{47>ANZ#~<k_sesxQhgE+lfQqL=ztd
zSZ*+$+X4>TqS6jUM(haSv=M(!JD{X(G>5)=#@p35YWxrt&4$KL>WfoKM8*I0>IeO_
zI<c*4&?)xJfa*MbYe}Ju;WM<@8fMxsZhBb5JVw%VHi#D<8^Spov<V!(_(>|72@ZKv
z78pisQ$O-K_;KMYHIEwUjyJ(c+fPEv-{3E=R5Y!KpVy2ONK7mEmsiL6U9>fE_N&l!
zMVa{}IZ$ea7@P8hM(&~#BOQ$^dowLqTleKeLBF8KUR3f|t@AR$(?`|G%W7U+fd73;
zOiDGS+$71wEgeMAHu^2UhjDSLtv==Yl?*76a>^f6GNc4hAD^H)tjq7w3+gJf{oGbM
z-KM*DDY-+*BdRi@oUy0Bp~rUUa#V?}_npd(DS3d(<=o12C0Jbccd2At$>;RYZsq=5
zxjoA5Rif3qVL2NeFKOJeR`Pdno5-)Ge(AyEp{X<~5TDB&tf9qvV#U!(@}l>@7L^gL
zaAdHaKT3O8XhZ1@5QbYW-clsBltx@he(qF4&9ju=0?dhnzJMJ1Aq}FFr{1ma*0IGG
zB~7)=B}_I=m(Y@?Iv!j6Oo#l`$Z3;Qz6HyDf@j{Xu@m=#9bQ2Py$<Yfl0sj1r!}a}
zf8u5Jul{b5Hv0{`J&0s5`&!c2Y_Z<GI_&jE!YkloD7ITnH^HaGihKRufHB=gW4gP+
zNW)PaF73h`I*XNBwZ(O>eWF-u{(E?`K36XxZ6Mms%nT7_YC&nnJupcO=K9L?WeX$K
zG65*5YIQlde9rXY?UpVr(L`OxTQnPxj!FCh%@7AGZnbJbR232DBB7w{@$XNqoie`j
zwEEz5yYAWu_sZ15()5-3>9hRjVk*SdrPJN~Naqu0&&DAbH0{Ek%@ZXNDb!m$rYL@~
z)+SR}B|~I<w0?FZDpon<(fr-e;ip}qqM-UQus3Og#H(`HPgG+00&<8%1&<n1JZGRJ
zFkSUONRA_UIXMK+XftrNYd?TiA!q=Rit!_e#j@^r(1+;6+5`c;T<EUmz?I@4OS8E4
zad7PA>hzUmplrb6mf+O^+Lkaj>TNz}1MWY-x{JDW#sY^NEY>-rQVf{R*AzA~|0^8f
zJIfR$qL;2XxagB>+BK^!|3Q-SkTuwPlUnv4*4;;x>{Yi6=zWXY(RXcS@+8QOv|WUT
z6W!7jOLf-#uZRN`t{5?UcVY8dZiz`QOluPLPkk=%GdanNbb)X<$RCYKL*447v24yb
zH7>@WM-d^&a1EQ|W@#h4`ZHk`Lir~fxmC>M$Q_g|IU*%&n>k2^SddsA=WZf>fqX;_
z`}V}uhhYbh0nanl82IIsqxHK9y<Av*kf~;y($4J&5+sDM7?RwVvY40<sG0Rdk-z!Z
zoqAV`gP=Q&!qsoKmbHX_IlzF`#16a@7K?}z3Pl84#IyYEGLt-_MusqvSMPR?Np&%W
zSW$aLg#jWkJuX-6Yx+YB3VP8K!cubbZhzQ)L%BW|Y=~U{>qNOj;cR!j0ho&<ZqZo`
zM!Q=FG+JPy_}+)OM_h^KJzLtt7yT!g5@!D=_~ZX1$y(_=(cffMM1c0|rEh4s_038o
z470pcPp>Jw2{9s?RgSIgK4vG4Xwj*VxD0@!xmf~BFO)d->-hiI+quQYag}l0_O9)<
z_v%ZW#BrQ;lE&VMxFl*LR3K?86q>d%5K<_nTx_qiUe`Bf*J)U7K%&s63iY8vLW)q@
zC(0WVk4Q!0Rvz*|0V!I6kPt6aLZUA`A;RzfotfEP@7i%L#g@ls&di>fojK?Hzsvu8
z-@?+|8lm4yIi{j8cGvnEDIRnAV%}04(O2`?)w%M*l3Fkq?;_4;O-Sx)(|Gi}U+M-^
zm&ndabc(#Yvr3O6c~z2RLSca|A15MKwvr*w7}2qaCBd1RT1@z_q%Qg=JxkRS*~-B4
z^|M?R#ub?yykBW!V$!cw`i(Y<Q|SFx8ykWeY%a0fIZCyDysw1GUMUi^-lUe_>r4}^
zB*~EC?3gOv(?+AtdrF%>Xk(W!x%E%%)`->evLB`R9UV5K-k(+ai#C!D+{VNLO#&Zi
ztUg8o_Zko_x3(bK959h#pjze`+;4Vt46es0rK9}mH%N2VI*yBLr2VKT!9B;gz3d68
z>vkI9seX?0r&(LdM(G7?THEu^uMg4Px7umC1I9V*={*|!isng43kxX0k=@UV`tIro
zy1R}9Z1iB?*=|N)v^rbOkoe_Xr_M&7_u1$%EHOXhh}h`ZG`ODl+I?uzKcF^&qol|C
zH`YM4jH|@{!$EIgqx(nlN$fBwED`k<WQR%Q(`JVe7rjjz`HbHtZMSJ*5PK0^YlibP
z!F@5vUClgGRz@st?svex-2hf5#oMTGT`Vywmu_&@5S9;{hNzcSZ18(TO;r$HJ~_xI
z#Z~vp!&q=e;XDUi^#fRNw5F4l`b>*gvAD?QvKI>uE;CyGyX}wUiD0IE?-%8L<^8Pl
zA2SQiFczFWJPkgDRfk00Sa1$v!5QP6af>LXXkv$Czkzo?!kOc++Ou}I@Wp+uIx9Ch
zf5I$2qvgZZ(fnTQHisx3f`hKokt+5e*9+jEQ^Wp{@yunxxsCpR02{)T*$}V*VI$d7
z=**AZAf|-RT0dyEob2_9Mk@l(EU?)O<;QtOvU1(#?jEU5&`XX~k8l;)c#=#gV#V{W
zZyuvfaml%k7~rzUxNBkc7|2g>ALqBf518b~@`uez0{h#Z!z5SARdZbF2-i8<VkH?u
zhJ2)M)i5&hAx4}tZDdt?ia~1jjR&_JkRl+EJzN;={Qu3C(BKumH@m_^VE?vL)N_;G
z&-HYfXCH`7x>b*^%Vn?Vy&2%#(lx%L&DXS%2<0&8EtMv8o@QIOJh)rC79YTPR^Qbw
zvu_KPRBCnZ_^x&4zp0%hGrf18+OwHr{;3-YRcLXwk!|lob&jDr8zI7!Xxf<VrAE=<
zwkpzFuu*DZrWZu|7IzY`EyoF49jMHKy4>GJUEbGuco!^7oE=R(OR3&O&H-<Aq^=|f
z!Qfj_wr+XC({x79+;pQYgd<as&vSIf5sq9DJ!6ZCd`u0ztJfzX-t{Pn2n`zZ-IV1|
z(!?JnQH=1pSTq{7MJk=c>C7RgFn4r%I%k2y;ZL_;vK8M=Ds#e=W%;Z7LO*rV>Tskq
z`l+oBiI{Mkjh^aM;Hkdg6yCbde9Kf-DP@Z!20&o+HrJ#XZ@|(c@N7VO_gRxu5;{e&
zqyt%1PI|La3LraPLr&ySxK5+wGVLu+7p9R+OE0Ne|DzHx%LGu(UM<dLty`WDo5n?r
zy?>26`*xR6`D+PXkC5D^%X`o0L2E?oM-N$L(~CbsA@HseKPW;-`e>*FqX#R*FchOH
z`Q8IBd$jIluVosqBZbt&R;F8%Q%#?IR-Gx-7~AWKMU5}biyhs=Xo!LDeCx{!mBN{3
zSF*H1eWg-{KFdur<B;_jElw;MQ|-;62lIhn>4nX%V@C=`4|}q#PxK<v#IL<hJ3WXn
zucAefh@^`6CTH?8VVe{V14nWK!9@C(*4Ta@s0|#?Nn-d(4<u$E$9tBE*GGU^+;*fN
zB6!AfM!}c>?i9QW3lXJ6kVFB@7<?=IDxb8^#&MZZCK$yGB2EAD2yPGhJ7?q4Mo_>(
zo$tTVMYxb|M5!3sY1+A@uaJUMdqxtD&s0vAl|$KVIO?D-zj7%|30rc}&!iC1o9mr^
zo5e~yxUK*3-#d5i@Q&q5aKuVKJX#|d?{#cnQ*gF;vrQ?oZeuW}m)h2^ZQ%8&T}0jM
z6_t!>8aK-E#`+4{rb{2RqIt}KO(dZm9N%tTmf3CFAdmNHE(BBPC33jR@iV`$^s;xF
zVr!2!YodCzbSov{QFK6;L}$$ODwq6V3@5kKo@Dq8EFmp@K6G@4P7a}YID(&Gr%doF
zyh&XetyN_MjF1zT0YH~UVpqOr@{ba?4^psY!bQsNLJAd6fhF)fl@6OCc6hG=9q?&A
zZJYbIdo=-{BpmTB3cFAAG;B4MQ;oc)hor&VL5Mh6p#!^b%R^|uw=VC3-Gm;!*??Ot
zyar!}T&9M9UL9ap7G8b3dEHZjTZ(l@BVw9j<K6G+$bxP(_}Lz+;s!wq>QRR(2VZ#p
zTcAqz-e#yu)`E2qW{thDI#*~g@t$;W6`+jb{;#gDtuJg6=G@tDvlus+_7!%}Sz*D!
zRhFx1y;l<ms2dpwh}rIgfEYzx#8QYVmVyP_W){Wlsb!LMB}&Y9bejl~6Z96;cbM%%
z>!hQD#?6rI*q*ht+rj=r9##;-2Zvb_fMaczg_3$^B~e1H&9LXRjx4>@ABI?gErOI(
zNBl+aZJt&;yikC_5HHZI9s*tnDR8T!;Fc~Kg2W6#WFQcN{DfGrh94nXyw(OSWD|sl
zgh9b@!z~^mpNtSB1Buun1?~c4WF|R2Kr2W@`Mv_fnE*)<>cExeI*c9!2qicYcw=h-
zNCHaW@WfssEasE&|C7}?YO6z;uY)p)WzxgZMyd&A(x40^22kb<xRPl?nT(-KW~TD^
zy`W6sYbb03GKL<uy0z9jUtHMa?>uxU1C~kNVJ$hgQY@^|bCe3WWQ!K?cZp|O>-AF(
z7dTHpPZs|3_6P4>od4`uzlP*rZE$JNZg6Rd_MAjq*P=eBp<r`O2t?jR)d`_$(K)j0
zYD;>tIt~pq=uu|23kg`ez#(r&tqX{1eg0VE5<!vcY!6I7((o}eUkFTE#Uje8Tgf6%
zA6_~BVv)oAA8Q1uAdr8%!PW{?T*|1OHp^_V6Ns7)r0DGcmZhvU>K-E?T8<K~AM&4~
zw$QI3&(p7?%PfCtuA^U*%PQZCP22*^(kuB-pN?7jT}*KXZ~Z>J$)s3cDjnpe_MYbn
z);kI3@bj;;wQhT+0iD#~&R+2)J#~A}Ur_0bY&gk8-Eh)!-rrdre?gls>7=&k_m19#
zkfsCSV3^-go^=>guH6%EKtZBOsNshUfS;kwH*{Mv!^R@fu2eVhm-STN9{w(KVw0xk
zv@%~)^O?oCH*fah&P*9J2F&7^tF^j#K4CD}XKv1Z5GGw4Z(>=r=)9E|4c@Z+gi4p-
z5Z2)d0%c}R^~p|2)6&|wb$kh^Zk$kBUnRH31_+6F!T37axy^#o)wuAFK5E)(xS23M
zk@|9on3ve``!Z3&tDPWbk3&q1dGK}RxdAgHrW?$Z>AU7>p=YCIoGJGhk-1lqcFfO&
zTd@&UE8`b1nsax>ia8_fD;dGsX`wLNL%>yP$<os1l`AWYgiFNx;S$&y@6&t@CQ+NH
zZ9Z>Nnu9)TX!{i^yhpV$wsY8zTl=iQ_m(<`+3Q&h>S8K7M?qzB=eHF?@rO@)_Uy~&
zxZC@tI^Pb)KnIhdPj!CmE?;mLl;=?Hc<w`#JCqdm8lh55_IMq9hs2B@G==LTrF5E<
zv=p@;f98(~(oLyMz{t`YCb{Sw^H{3hLOCkxk~G|RITC2NscPz`NYg21Y}Z<v*4{_P
zU@<GJT;6A9N6vUt^y(DSPn;k(?4ifzkuU}c_$Ub@CJBRkA@vDz#o<7sC6pygUazO|
zXtwIOpW0r>{n))oADgydfzfs1N_i&%q@7AsIXa(R$WCpX%r52lcZyWjF0z!W*j>nF
zm#FKib#0vF?^Gaav~al_b>iWBlRZrPvZ4DbUCDb>oqa+jQ;cl@*l;#^!w~Dim0w3%
z3MG#nWQ@OR_?+gR>dG(6VK8J$%8*OZt(A51hLS|;7w)$y$*96pcNuD}l17tCJjI2Z
zB@H+LP3mxi0vz=vgcCfyc5fS*(<a+c>Ae>Px04a?>{@&^$n3ehe6a?Ak9sdMDQlIG
zVSK!;jn<T0o7e3f?_SgWGHw4(Z%G_^fSy&6sea=#O?Wz%>%5#h=b0j{mTNKm<qJ#G
zYirKbVvN75Ixw*DO*@3oy0BDMkbAoP%IrROzAZ|hh8r8K#`fIm?4HHq8cE)rmqG8&
zWzoeTx>B6?q+6SpMA=%_Nae7;Qu4T2Y|`3j7M&fi3tTNKV8rxhxyTuP)Ym3k?#GS-
zr{pJt$3(LCdGbN$Z3QiPEms!{b5cINxIP_Q$*<b^p5!v;qSq6$h#k)FBK^@D)J7cB
z*|a_Z!C-TR_Tepg!_?L$8*9U>a;pvZpgotcZbS`@_#c7gFg>?fUiUNkxquy_(9|ZH
z8}%7g!L{`aJcR)AMzuMhHupEQ8B(yOEilHIj=S_O(^#$nFYlnLT;`eSZ}@>NJWoqB
zyq{`nF+}|+yS4Ol!OOmA-`qQ?le9|KR-qXe&aC@pH!<nadQ}7MAT$KkVxMqd;f&sJ
z>8os_toaV~5jj~@)*?H8yQbKxP7Cf5U5SW-Uc@5(J+bKUw>#YL$1_36{XaW+oSHa)
K>CCCj<NpJrSr;Dw

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_8.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_8.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..71489cbafb2ebb931633c478ea9d271512492c74
GIT binary patch
literal 58037
zcmeFad3c@IecyTa4f{e6+%I=R;;voMVp$YrSrnv5loek~xd!lFfB=CD@VysY;1^@X
zlx?|=5~oR*#C8HYD>G@6OzNi1(l+T5&ogPW^=UhA+nFxoxN4K9Nu8~XEzalrJMY~9
zP)?i9GyfDM4$gbF_q=EM?dNwEdwL2ne*Vk3k5?{zEFSy!dT9Si^Y9*iYrh(g#k^Q4
z=Ec3le7qF5zlnupDH(TZsZ!d~GNmkklk>TSd?{~vsrkY}SE*~EyVPyZ>G__8-cs*E
zU#ZWYGxPlm1Eqneox#!|^|SLs3qz$LOUum<FKj7ov3P!dWMOM*tHlfRqYK+g+brHS
zzkOjxX~)9O($0llrCkfVOS>2Ll=dtXOT~q~rM(OLO8cx__x%2a1Em8N@0mZiaHw=>
z;c)4&J@?KZSr{vgE!<VQ%bxq@?_M}sI=XPIbZp^x>A0o!&!1R0SvqO)f%#Jlr%R_T
zJ~)5R!o8(?i4S?h-WG4<bIH<u-cfJV+xAAhbicRV+wn%M^uW_GZ>P8GjhMHqlAU|7
za{79z^bl#gy*;GusXTN&S$de~qPLgly(&Y>H<7Z>+fT~=%ELTA!t(*|AkPQw`B9z^
zd53vETzQoI$9O*CjqyBY&yVwbmv=YMciZ!Oyu?efv11cUOljfD?7UZ8n7vwG_AAB3
z>B5D2eW~{Fi4!xk^$W|B$EOw-PRuV)&3gXCY_)d6t5g?fYqQmvW8Uoa^m1)>v3hLj
zT7A)<x^QB0esS`|0}q_Mf6|+tx_|n<(+^DF_n>#rJ<|`~JNdxmeWy=X?mtzzzjDv1
z`|q8+r`1xq*;2XDQu8ev^NufF+pLR=OO@)x?1|ZhC4cd9#cTD>^{VpFz4t!&(CG*7
znVfv^f%_+>PTl*!1Kxd}M{ggRoSgJddk;Ny^7K@t*+A>XCuSy=Yqi;lYAfjueRy(l
zVd(?2wG;J;1r1Q$pQzTR7yX5bUn?9tR^;#0Y~gKfuJN&Wm^(MIFzHR4dvPog=Kacq
zPw|;bnDHx1m5Dluy;F<TdgW?;VWK)QQ}M&D>hk=2lt%sR(u7~1otP(@s$cRJ_?ust
zxL7IA__H1ly|j0+yi)NO%Zt_dYorv;&DQEqE>@>!XU1}2@BHFSxqh`?o~V^)7iPjt
z<to{=FzZ#OCzj{yVS2W<FtHS7neGKTK<gC-VS0X|9(GSHE?4VidN8q23%9Z`>JwEu
zQ?Hb#{l$ecz0^zMVdmUy{i%fsePWf7n3ykLoDX-NnVpzftWL~79gXZ6UhL0KF4tL2
z6dinq`LEB`ubrLsYxS4vdU?+i%k{-db!yS8_|KjjONN8?t&^4dM7cI~q2eviSHdAi
z%b%F4muIW>#d5t;tu6XtK?}B0o}ax~p+j9~&OGznOA`xA^A(@R{%7=Q=N4xe;OAzq
z&Q?k5t1V4TRm%0*g-WenS*nG(DSgls-`z#bU#!(8>h)?EzeYDQQBJ;<o}BO&&QB#Y
z&(V+KY_-^v{MJ$gL5%$uUzvRVa>c*m&(<r&1wL&4IK7{(7O$}LiuDV#wc_-0bxMO$
ztS=Vzy~Ua~sg*iWxjZp%)e1#^yv0hbSY50ar!Ht26(_3K78d==@tZLQG*zq2PlqYJ
z!#~LDEx1)Zaba<xa)L^g$;HKsC%lRJ#EBVyae3*)OUu=BEMdmAdO|~f;tZSpz0bUO
zV%}RgVO<3hH1_`q&s$n3U#i)lmZxW{mHKkE;#F*}?!B*k&&l$G?3l3Eo2^Yu&R4v$
zeZXN2hDIip^vC!!rGaVuC@#!nBes)B5DVg?v3hJS9wfYYkerQwM|_yHts2S1c$o1j
zmuIIcd~Di)QJ7-+dt<3E;h*C7beOJPm{_X#chOdB4E!TJh6BshS?yEyN!c}XjAZNE
z&bWV)_!Pe#Kf9x++FxZ}D>1NPtVF-OgqQSEGchmiWxTAH^YUK7>mt_e^?1Er%Io7_
zKYs`KP4lmhUyDue{$y-A?hSfFZ{$nKd_3Rly&yaS?np&XBi>e?(v^(&%?_(vkC(EQ
z+*<xdtWvnnJEs$dW#XiFIV@A^c9^Bq<1k98*Wr;;pTi%eeup<o0}fw+dGD$W(azo8
z(Kk}1Vb{9%BJlM18)|n;l=6~y!aHdxBT>qEVCX%TvX%bb>)l6RN4@*K2f%vUyqCQP
zy@yEIPTs@bHxb+6J>oq|-cIi^?{WU_^4{Y;!Qb89d%Y+5yC-V*jQ7+VG2UGC&U#Oi
zzSn!7_Y8mcMfo4_&Uw$-d-r?e-t*)f_+%^<D;;EAyb`9)oO$87xBo+uS<cL}GAH~n
zF@Jje>`IQ^b7G#Yd>%L_81S}s=-U*Gz3mZv<=B%u|I@#_fBWNK`N;5JfBN69e9z-w
z`P$9j|IOznUw`~7jg=5Ze(&)x37D8FHq!a4b(P~+?bQNSc|CSLwids^OM<Ol{Bo?;
z^ZuBZxE}vlJlGn<%87b1h|Q(u(w~iaiY54)`h4no9D0JaU(N)*gxR2nFc%aE^FbG3
zA?POT3bqk;*L#9^FmgRXytm#*E4^BWZzXtVp0s{a`u$RUKsndrqzu-RUba3|A0}mR
zZOaYD`Ffl;=RTj*n}bm=wybg^<d683V3hdQfYRmBV2E&Au!V4YkW)E-Imq#}BiKo}
zGuZ9FUf&g@ug7bzzdy!Trmf89X=!(m4l+U3%fAKXNI9kT*S=e&=ubg!Ztr8T$Ny`j
zcUg~A^2e+_^>~ksXKbz*3<k*{5%dN9!GPEOR)$gQSuNK0*7uQGoQYp&q*wRW_s<=u
zALMKI>N|Nl<n_L#(K_t)t?rwNt?uXdK>Y~)8M_`o6MJp+)x_$-)kDF-VDIIa|6hXr
zYj=75!J%N^j@W4IhDPYn>b~H>+FijxJzWaf*Mo1wg8gS0$<_U|b9a6J+JRvIjd*=1
z*nd6l4cTm@c<0ezJK?eV@wpSh4oaS6_Yc37SUpVq*k|JY{`#r8)5Py#4!2k@@2O^j
z*xJ3`NO1Uu=5*I*;!DTt_g&ZexSx{uQ|l4F>VdTfZz$~{57?;Rv904{N&hjHVwkvk
zGK^ndIXSiL`{3C5Yemrppp~ViMW~7DOmQDec^~M*EAG2`GKyT@x6<=)@x|q8RVBbs
zEB(io{E9d0kdqL?c@Sq7Y&N-AtE}`tTd6Fm=u^)<QG=!|hVkP-=C0Wph=5A@V&z&b
zEa=5$!zH)MFIH-my1l-#Fd<~OBs^!C#VeKBnG5yW@#1-5ME)<=D$~pJ#Va7Eq7RB*
z^z|yy9dx@$1IO%}KweW9mlh!fj^8>{yu0;NeC3H3$DbL0nlF0lrKirHCt4hT`MENk
zEx!E1nJ3Oa^-}Tt^TpCrFFxN<Z7dUZ&CFh|RLi0|YGF?&@sV#44BZaYl%b0JdRT~_
zDpe0sViDrGQm(x%6mv5kroH+#(9)wky)6=9N<rhtup<vI@mt$YFvGmQmS2qr@i~Jg
zYl$1N?@C;Pnm3#VJ(3a@50g_%%VD~@ys&hQaZemS8K%_P^@y+V$UKx66#mj<pQiP%
z>V4Z|D|-brkC&mMX6xnh@kiztrzYlWj}d8<9!vO7Q`UcmfEN^|XXoe3K-A?5w8rd;
z2(2E!a%p*%`GFpt@PNLHGnXg)v4=s#%Y1N{YqTHs8yQ)tSNz$P2~j&?dWk8mhxu}O
zYJQ?tE0<NeI<Zh8)(^quX=G<AZuNvcQ}eS+Aa0}cys%4jZH1+Hty~ZLrbR+V#aVY@
z!GL{Pon$hgm8zBbS~$v@DlaiLdObv{5t^*3TG%hj$Jm@#{D~!XJZ;O;ALP@+tm?7i
z{9!%iEl1B`a=uaxQyQ-@JGneP4GR$VsBnW;^~3&VG$NP7uFL9!#sRW*m}NaPYn_>&
zTr9g?5cbW}brG+PgzUm1vqbqWqlcNW>Djv9qwe(D{5g93;!-^vU@|K8GOMDzyu>o8
z)RZ(sA1i*dUty9mzQ~&}QC$ktVoR=t83^fm#RFAV>-U@v`<hE#ls03Muh446r_Hc;
zVt(lYEX5Ry&tGKXHEOK<ts7p|s4Fa7)-DLU=RwfbvRf+t-ReY&tv(<2sMF1t(z9xq
ze?iQLR*nB573^!a@5Y;*0C)<!m#5%zL{>AMFjoM!#f7lzMf(eDL637PvgB81#*X-h
zXsXl&`C5-A#g@4LK9c=6^qM>{MT}4Vg8!t-!bMGZ#$c6I%U@Q>iwY_VrWMR65K-Yz
zDR{qvXBC`NU>m8Xn6IEt;1;<pMtUuH>3^&8E-AR6;Nt`(t#tO3?Nk4<lE0lm%%tX7
z%Wz9;%6V)|yZ=$uw&mCy)6yV24PH=_*O<FfzFN8BfL*D(QL-$aFg{!A=_qP*a5L_I
zSiLY7Llltz8o@pM)?Ozl?Cp<d;@kOc|JNPwwdd~m?sy?#ITq$D?2qT-{qa=1z~65E
z=7{l6dAC!^ep*Y}zd>>atz6>u-uP%dlgK2LHXQFx^d_v0+bcaCaBV6}pBg{CL?j2(
zoYsE9=(PYE!OsTq)oe9ihwFgCNQrXs;%}u_vtA;|j>cAV^~78f`Uk2X+?}r{)>7o;
zJ*e@IC$`0IBvuPSb}s#Dd^K6m1W;Ar<t*)pic1sbgZy0DOKt<M&t)$rYVUz6%!n$i
z7wTPe-Td}=Sz3nHTkG|5(0tkJal`(477dthdoJ({{nX6w3cBb~Uy#4Sw=`3`U8?#i
zDyA)`$8!2N$?0v?+0j|2uRX`>Z?y=e6q_3ujRo0ZDBYki8neH=bBDfmK<Qg!tI)4}
z$pAg?ZtKaQ>j}Ld47$Cc8;Op74r?6j%L+H(0$qQ1Kyx*}aZAwU-f5+Bo0Xay@<uKu
zpqZei=Z3wlLop9Zd~OR|Wmk~9u6FeWBkK2#*lJJpA)bc3QLD4h+oqQ5U29u4y4L>o
z#*@B~@_ViP4lDmSc~H;a_&qn`)J)a4c{`0a=<)hl8$Dj~5sg`|w~IKkLQ&4N2n9K}
z<3@aK=ZzTO@v7)zqxXvAH<RnZj3^9W#05l0Y64Nu+qe0-b+-f8#S0Uc5iBegCo4td
zKaeP15tg`EnV3Qp5lKS)ajJ)DqkjBvQ+%is*aJl?m&2jjB|m~BEi|yp5*HzTkc}<Z
z9H41s?-bI$W9~Mq>{q71^PO)DceJt=POZ!IUr-+hn>ZB;#Yc`eRS9>u()8j+I^@PI
z*vk9L{`daH?i-)`m2V1%TLn$}HMLloKDBA;X(O=T7767)PVg0e`S~|ZcJV#_dw9Op
zzp!j_nuz2Esdmp_o}Y~HdF?FIQ-I8<2`yLaW4jEO``f7Me}o{+Mu^q_n37H?DG3tx
zKdHwSY7uaA7&H^f!Lrk!DzLZz7gQ&uaf3`*oH4W$W*1l!3(E@*(R^05zo5XTIdk5L
z((=qu%_M2z(D-8g84)52^s?eT<@?YeNDQkJRsVn03z8G_^EZ19p*j?2I8+2<TQ3a(
z>Sd4#MmE^{{Ld*RClQ|H%B~}7bRGTbUn5PmzdA`Egi%N&Kn%G=FMvOnC>UZHjrS+`
z2O1ekZU<p(2hj{CGx5ajTs(EVKe-$9Fl@Cz8@IcYIiaZA-H8JroNnT|+qq;nwYSGt
zMht;835^B;_dhYo|HLH!<1tBn<OUMY|KBl*|ChD#n%KktD|-B11>dK@P(qiXgovIW
z?u5`TwX|(rsxU<(%Ru<LeB&!h`z`BEcp<94@7-`m)Tuo!G_ejzL<RP?3T(h5u86}X
zqb-Ec#2StVI?&4545vhm{1Af{Zf_NBVU?&Oo8_I3Docdp|EL0C82`r<+*I%$1wT#@
z4z?o`SM99-6Xa~bB>rDlr6v~ff0D=2PzR&s>cbkC<kX2XqO|^&P$TMG6XW<lrG|f6
zfl!VAzbW_`1;*g~4aJO|`8$gJvVy;>;EM|Wrh=bU@V6BFBL#n3!Otl$&Z5O$Tqo_f
zl>UnZD?2xb&aQhK^f3jhh#jlxjo&8fD(3GYjISXPoJ+2yZa4{_QlQl5k|ytizFLm^
zd!XedCyjp|xndlFi5L3-lHFHaE00<$*^X9Ry~m?^30JQnpVMDT$NUf8v1X~W=9QTK
zjU~rdwmn>Y34M_?U?Sfn!-@oED?1QynRrZD5eMAJ9LqT-b)Jt$A_JbS_;ZREn!*?~
zfm4f12vHXOvU0NUk}8MnM_jUUO+C48Qji6}Y1XJ(>4My>i4RP?E^e$Beylg1AauML
z{My%3iOf6V)B59yuO|{Ky*7o-2BuO}Zv1Q^lH=D{v`kVWb(BG7h5&<<@{;vLkXlPh
zh>fszC?I89Y&F$NT}=mRgt{ZEnIOU6fz@n~SxW`k8|ZlIsA1MJK}Ku_qSaK8c{Fxu
z&m{?Fvyj%0)^if2zY(kF*I*acx+Gv->t>12l7z`Hf;}7ayezC??yV$m`)x$Yd6>a`
zkh&cA%a=aDJM!Bg>0#{(gxz(RJA~3Xu|Y6`mrNXPPhS3F%KtcQ5NxHGN?4mHM!Aoh
zFdUXn2&nc*P+04`0rTNnU+s16d4-D!gy{W24=hC2HhN5tx>x&9D!{%eWh5pxBj|f1
zcIgMIpN?wfFQ)nS2iMixz1kP06rz@fSNq#rLI~gGkv9^XR&QxF=$}r|@;|ur^Yz#o
zte$dVbOzS7(@Q%8t=eh&<P{!3heF-|KuxiNuIAvnnmwI0dFx=%{|HRcrF1X|GtR${
zC)AtOLB4HJ>W^ro#W>v?46Y4%-SlXa+}c6DU~rel)JB!Qd>}|N)=#rGhFN7O_TEZV
zwcWS0o|)~T)nQgup8<+~-Zo^al>uVI790FLtUPR-<#=5eQ}pB#HHLzr&&C{paAl*s
zZzFFw7;fY>fXCRDasNfO1$;v{q8Wr|D5@nyEg&OBA1^7`t>T%=e8uGJ$dZa5S#{F#
zR)OOkVrTXi8nj64EXk&coilt>d^Hio+XT+G_t7n!r%;%Zgf@(q*8v3o17sc&C`b?)
zj;$QFP2MS-+)T*4vhVDyQwl6%p%Otg0Xd{njmMiPx4lYjN?PCg#Jod}V*8BMH9AEc
zRYvv&j6-h|Aey@1R>tYtZ}ngRgN+Aj1Svjl^%s%lU0te7!Pyk0tyt-MF3NwFr-uv0
zm7&7Rts+t4*hnO^luXdC+L)-eO@&{tNw4o<X-4YiiMkWoEvR4~BLKg?xQH-1jaYiN
z8un{AJ!fMfF((3KmlDa7eE~zq0H&wrp^?Hg(#;tp@Z!?Kv@AbrVSc$<yR=-XAnxun
zQDtk=4d9P$anL-~6lwZr)SpDn^Z&IX@rz+%_F|Z;M%EtwZzwfwXrh#7X3C~m$G!tu
zVO!J+gjT?!{`XjGbEQJHj36AD>AcsBQx^`@4RKg#4gUJ2bcHsyYbl9r>4mA2R4~WC
zf=Lk*uP45qOX<njyW5GRizQ$Sd%u>;BrFG+V0>qy0Nbb9gYgkq!CvswVEm9`$w64X
zUY^7_4pU<ohBB8(yuLkw9P;)7m`nWwd);0c+GHtDY5O#OGDNhM3~9ZWNL`p-(!}^&
z;(9Vb%7kPPN`llRva%>0w54*X=5yLfAE%SDc9_`B5SZ0Rk)Nk4b&N$2+aNJ$7||3j
z9l2Vr7lY!}^2xE|&7u$6OpwST@aqp-wGWLKuNI%=_xSPSuN9rFKq@=arp#ArsNAYc
z%jn<Mw-VKdV4SS@s?jtEDVo_#@rX%!il@eo7vER$D@ThbF@eG)3Y8;<ELhgGHR)qB
z;#VIoK2fYMBUGR+C65*xB=*r_J4rouw1{v)uZ+IraII)7uhpP2EMsY3N;CiS3M46U
z(uE%&HkOHqn5cy?UdKL$cZL`SI3ZD~yZxQ(R$|z9scfWsqx1*0Dzzhrv%^p`h4{*_
zt&MfXIw%?~j5NPm9A6ckv~h7*id5Tfxg>{CJii32AHSJ$-?m0I_iZ<bk!rMAlG*qF
zm;{u7vVL0M`_V{y8<F&;BeH@OKI5zUj6pU;;&$Tp$}aniJC(oV_xOLRj(v?_Eb8gs
zQ-aOHFDdqpf=+rKv7A#XZ|Ihz;V;rTAI*1cW&gW&Dx&x`yp5msp>B1pF^H`N^8HkZ
z2GPGw+y>FHSL^z^eUrYfltCZm%a`b)t%bo<G>Cid)ZeJa9f#4#=XJ}1MX+vJ07Ya6
zWy9^KG)id&>w4UrbTs_VOfP!-H5ImT?VRlO<vMyPcCWd3MSw)TWU)3NThz-&O2Xv*
z4agRyuBNS&;G5NhI5#beAe~!28Mn}XShHcv=SLL#f`T6;xS6#j@Bfh!np0Nl?PRfA
z<=2l&*QJ)y^`i{LS9@uwvHJT{ww3#%t=w$$jw9jcG?upPqkjJb^~?JG4;A}m1%F=w
zrqLUwXrQe#ZWX_&Qe6a*@Nq){(<yfJCr|5*nW;_l0IPzfT#!H;h307vmI-!1Ha~zB
zKnohHx00)ABVn=ICT~ZOf>}!X`-Aivnk_Hw^eW(^v}#~elzcV0npw@R=48ne<P0xm
z$4vx&GvPLk?5Wx%;lKBH4Jf0T=$tB`v^Y;E5`Z8^7L3+^VX5pFco@q&h!(%<TAiY3
zL(#stjCTEM8NGnWU8j5(P`1eqN1C%Gg>qBIzkdGvmO26Xck$BNAaCggh!65B1!H%7
z1Zi>v($`g((pWcsx`{N${$o5qI%BKx1zBvwzbpPC?8djoSJN81=J+SkC~7=&q~@^$
zqda4px}IDupu@tV@FH5p8vy<}7}%M(my#@C4kpe^NH_HbTK$d~Y$Ob%mk#p7v9;_C
z(u0D73S${r|6p7n0W*!D$+9ACkc<BqsR#vZq28quBdV~2SeQ1&f${%lLTU7B68d<o
z+ikthw!>~`!CW?k62>R}-&A|n=rE09xhq@O&PmZ?3A?L{RkK%Kn5bP`KTo-(va>M#
zJVk2-n%tfkfet&G7>!4o-A0KH3QWjMJK{-xYik6IJ9}W2arWX)K!64pi&Az$LL!2h
z_-dw}s;BE{dMTTg?U>o*Wzq49l|O8AhyE7bFYNquoY_M7Fjw%>Z)tyZX@Ag?X1IF<
z+734Tz4ny#&>6G8o<{J{6Qs>%tQaKxap{$(vEm8R(094<#>{BUZWYX0S-eTzm#~Zd
zWn=j~@v*owDG<jq*6T|b;xg|hX8eFxH9jHCA#a6;sCfRDRp^rn_GpSfp^_>1hZ)zD
zryQ+ZR)c5>TN2^0BhVa7f-RA$cg?+Or=~6`3-TTlgf??U@UNf6p2i$8{fkRKqN&_T
zF$ip?G7~RCdZ*C7>}0nkUYEK*6<-;2$TOOLtN3<5l{%(bo36R>_VJ);W|B44niPS$
z6i6mH4N=X8(*cMCX8681Q}&18q%+hQQ$pH7ox<b<qBXM?vVv{J4yu9$qTJvx&HVX`
z`XVh(3=sGtX`@y!ImevJ<#zz;22qlveC=TZP+`){fZ5t_<gO=R4YULS0&l_|Tu;8q
zwzd?$!KDNc17-r99M`_?_))V%{C-vaL3|Q_6QN6lRar!~L<%y9!Quwvc>5SkI1D<s
zJ}mbLz?gFikNqN_W)@<F*ovcPE=5z)&b+&_jGxfxdh%P;7382TD|?MdRNoMVy^vEi
zF=vvUYqb3f8c-uh0mjC2IV+Av{-y(JdtK9Fv;}(`cqwiR3dB-a22l!>lL+b-<<#h&
z!8Gfc^>N*7r)My*x%^A&uF*{en_^4s=unzI=+`=QfrwV6q2D3V^rL3P$NB_5BfG?y
zk9f)ngS`ZR;v_i->PZ;}Bb}s%p>RsLv^h&N<CCPTS<iZy+<q($(8tjElz%VpL@)!#
zHOBD}sOx1j*<9JPT)kLbyi$d)g2R*XRq^mgnub`bhuQpZsA_UzvgZE@F<UPFn@TG*
zP9;pthq(nEt)LB)**CULO0;p9fyrb@thyi!h1LyYP9vs_U(qa&(s%(>Yui3w_>&ZT
zR4Oz4wNxtiP6Xi_<vRwsfpfLqYhsNzZX<z#H^p5iy(F+1fPj>rO~i1fXJgQuY%edx
zfbR{G*!%3Z=QPh_Z^n2YvF8lW_fs=PORS|EDw#(pCpe#MSWRQ;9AjF5{<8glD>;|t
zG(guE;#egkBqlHCzv%T;<6iG-9xQ<U4>kKfp9pfo+Ot@aAbz$uWx?QxjKL^@1gSZP
z*}?RUdi~Vvit4BwVql}Op~|9`2h?(-Rfl872JZ|i{9{_uB42$@yG8KLcBRdnwv>O0
zjDN1o(*zMDkl3+StJF3^f(E(Q+Ubv}k;fIhhhSykg@zqpWc^uO88iclYsfX4MiOU%
z)LMo#CtT;`d=Drg9wxon1>}FyaD+*WP*=Lmus<@CT*<p9Y%Ff2EjrfYfYUwth&i?P
ztqM507ipdzQ%rBe2BB#==u9TfJkJaq{2B!s6i_EH6HzMS=bArAOJQDPV{%yk5w-j%
zL8%~`lx;FSdsTLP6mWK(rJOy`Ten%-Kw@K%{~7g-FC1ipuwBf~L4vLV&x47b7`%-l
zq#8tc)teYbbk!t}+$43V!UUZ%{EC{WJwhO@iV;3Q8#V<b5(EySD9EGKTG}9#7C6{+
z2D~KVbC&J!LmbBeqb42c;uq=SSk|GKtWZqTLZwAu9@JR8Uj-eqH4+umBp<8W%*8AE
z>q!1}KqHInlw9u*H4H+b`!MTR={5))Wpz*nDk9KeFj$LEEDVGnVz(eUhwjGUai0i4
zOEn-N+)Hk<4&;UGi?3z)a#Hxl^tf*eFL8YQ*5G9ELh(v*s(9?N;v|1~U_tgQR5*Ke
zt?Y!njDsj<M_Q{RVFERpVc!9%&}L_r7cq#6B*LQ_kVuM?HtNIw9o05`v~DEw0*5A@
z`%{gD;GoDT$v<Ta8EMm+g*ME)M#e_P(MxL@8D<}#hzc}->_y@)V2R<Qh*N_L5Dp*U
zWoF&ln7EgfU>(%b_McI(uB%<h=fDXFb}CEB1fA8I>t*%k;wz53ZY$6M-MjQvjgI~+
zB6KkZ6@HJPIGj=nC6VZO>1f@PrV&*hn%bDZ&4W*^$58^ZV5P1y)z`35@Ss)kwm}-X
zYkEdpZNk3~e%49`Srk2cFF`<}{e^<Kkzg=#z-%r4JSV!`zua3$T1W%h6Xv4S0!qJ(
zrRV*5S}5EQSAd$$>!xpl)LE}5D6A`;V{!G4hw&x<3?mpuKu+Kes!a05^EXpr4x{Rs
z@+9Y&WKpkUp;=C?X;{M?k;>BI)CG<ya1s|Yb|s%RW&&csFJ<QaGWZ;z_3bq37y(<F
z-^n9f_k~4H(D~n@mji`n9LV|^6`jzOE~!*uaS3xktmn~3belLC`+klL*2cOVU8ltm
z<{jNU>s<|dnia}Cy2Y%Iv5z#RfB{6eY9^JjZt3(iV6c5oPK)O*dLoIx)_Mx#tx*;d
zvj2DbQ)V@MyMPf_=5_{2{s`bPVjw!2>3|Sgi&|pZoV9;Lz?88UJa$5~1Kcy@f<3d;
z*5Wr}-^n5I&XS*MJVzxz6*p=g(&CNWn=IPv@lVA*8Rwu6Fol%3rNm!}dFTjIvFjLh
zR&%FfwI{AX1nA!f`DL+4M5E(`cBC*`a0w1nZAg0)<Kgw?-i&vqnSp3Ng40@D9jZS@
z*9cbQ8zejjxJM^WJQxiqANJaJ+liB#oJ_Q{51KE)!g>CLN%UP%ny8v0l&2EqlWtB$
zK7@rwq>;2UCx>~P{|bpKM+|y{yJtC<fb!C&x{a7<R=Hzm{W&cilMvhbndfn1M^3dm
z;xAI-NTUa5&-<^cM$(qQ|EqM$_mtrD1ipc|84^sGfkjO7`>KoX6E{h7ZyQVaRaIP4
zU_;Qi?n8@O)1rQ5rYIy18fR7S{MvAO*j9HWcv@GjV=T0G^=o)s6UVwiaHy546(q8<
zp)oc%&X-=Kls~UPL=Xwamz`5LpHpDtXo2;5kZG|!1oRP8%%P6s>bQ1KhAyKH+iKIe
zv0Mk{5!Pv9o})Z~hh;&LfU{-KMO-h~*kWK~iNFxeIP9Amzco^A;97@vs}Awr8l5ag
zQKcW@Y0Tw-l2cWG@d}@4;7;S@OD5Dnv{}cT*TUYaj>6Z<7~qzv5OznWitJ4N`T-d#
z3$;=#lBVml&uNWFfu`L-1u-)#qc$p=6n<Mk$4J>J4PxV`kBD-F`D$0037s7^(82)O
zWF}U#$O*IHFLon{8id{UoHXJkYgS7bv~YejA+r&1UKRk6cqCTOUgzWj90~-G<>WCm
zh|O`50K9r1_(>S55J|4i+Ed;`I|=`DvEEhhMs`)0fm87y4Ul;ur^M_;_?Ne>^+G5W
zC}XD%bZ%j;AEOseCYV&L8{TFhz|?}%lH#QX18no@71%aN`QJ~<kVndQLQPO}n5h>R
zC9vDG+PeWzen$tO?4g}5*~!ytueI9EQ`-Nz`j)kkI^VvwRmMTc-4br(jPU#EH{t&c
zOk>FL22wo%<(GhTVB<aqCc}(Vfg0;^D|5=09XNTXcFE%eVNX-{sr69Gy)eIMmJ8?E
z{CBIAq)h%5742#)DGGM%U{+2G^p5agpf{_-e(hklZ50Z}(GA1U&3$Tc?Nu5xS<Og8
zS0;RJN8j3cv?f>`8}&cU8(XAZm_&K&NY<CsbzxkyXlN53dP>tycCvw=Qp||@6RKg{
z)yI^!qDO<MM-}@BX&eRBxv?4+7N?w@!B}4#`4AsRZ>f>=RU_<*gpabNaG=SEO=)YK
zw5r8izE5cyXa98tHYy)e%s`kiU2YT}(nfudu4G`91aUDKk7qFS?oI53YH)B^Ldf1k
zD$xxqwG%V&UZ{!PZ0W%S+q%!Tb!%KZCbI*`Nwmjmj;+R>IL{fWiO8NyVW_6RNS!$g
zlc;CrvU52JdL*KZKwe^;Nfu=^mJfUEgq#kA>9C!f5kYVNOA0=&;6n=TCJ56;dWGpH
zyorTZ8iFk~6lRQatA$BQv@Iwbq+ys6oOf^lx7;M%YGn;qH;qcFnStEWp{$i{HUyiM
zznK~L?@>oiD|kqOM#29k1(Gj}CC+1*bC&%96%`|To(_IQGbQ`h^W$TO{0Smqp{!G$
zIQQV_XLq@LX?bGa9R#(rWp)C<zph5p3T&0xw`al2I(X`a=!!CLDFE=q{G<X1^q8Mj
zP*BiKz$sZQaGdJHgyWJ0+ZuGs*+c{;9eBSo#uXShB%0ZuRd7jxNN%_6Kc(1b2*RG|
zFzZx>qq6NbQh!O^(#ZR#2=rU~4uV{ag@*w7zwck~zfD}g+U0iQK^Jz%GribZK9=jw
zW%!rlH{YAjbd#P_9?GsvE?vkJ`f__LKb;xn%w10|mB}*M2d!^YbgSb>W3Hc?sPdK<
z4E$Fl^2vNEpXP5OpCOj=N(N)`iGhEMl%V<>G5NBnWY=;x9Q^I_x)FLq3Y7BnZNU4G
z)o>m$3g!!gBafcEV8lw@<^h94jNYr>4sR#^Bsd%8?Ix$s+v64a+wbl5_VIVX+wUFV
z@1S?kJH+21?=|yv5y!iQz1w>Ct@bYYNCFGu<-&5jGX9&6IWHH)k9#M0&(<jKq<88K
zFa<s?<o7~8F68qfhKGv}o7W3@xRB2adAN|r3wgM39xrx#6Xx?GUfRRE6W*lvo=Op)
z8&7!eeIs4M=aWW6y&`YQlm82i$-ZX#lNwdeGk+M$+272e200b;kfEFdQBK0zn)aUc
z&RJUrUAp(2H*Vi>D0-gop7&m`5{H{LHHT?$#yju5Y&l1~4|uQeHDghoZ}C28ZQNyE
zH|_>&UI_cn0i3G!=F#bu!h-zT*kSH*|2|d>^!B6tnd;H<iGIX$hz=6{Q<L7?d|OE2
z0i9!BEs%C{X43p)1icd1T#VPYm;o#Xv2go)f~>#GOCnuO!Ko)aNpR)O0dbiZBc+$q
znaBkYaA5jMoYhDi<5l8$iw{`5FrzYq_4IXd`#m0~ENA0EuY_FCCCGa@!->qeafWlH
zyIw$8#(B-T9*L8XC*XnOl~@(%qJN%zbo71@!`)ide0TI!kru4=dp-35Xcf*dI{o_E
z5CVR^A+`qP6UDY5dr%q_T6u&Mpm^YbZo>nI#kU20W|96=)Y*=q?Vwun<SS%b(C?Ev
zx1;tWW=g!%>oarHU8L={bEtcQ0soilJH39TTgjidv5d{_Su5U9zoCJQT^ty9QrBmg
z8GoOG{RH7KNBC<Tdj<h`jg#|b|1-+}iv(x=?^g6l1>dEBDMiU!!<hhXqHxo+=%4qW
zQpQ<=vAy9|WMDX6lA9}?zIJVb>r7l54t5?*P=Q8clh<Gvy)e17xa7a824B;gx+sDz
zta%9th1D8v1B{E2Cn9sP)kuZ9aO}Z%(9crN^QZwHEM?(!7Vw9~X=o0idniq6{%0xT
z4mSwd*nx+2Ua(qj2f~qd_gJlNgrn{0k-H?5qlbg7Qj!Zw+G)z!Qt#yCBBou77b{hX
zV8S8U<YUT*WUYD5-#27;QT-|_7*=33!n|Tv3C4zu*eVUSdq;>qS2D@QkVHw4O*IVq
zH5$IO@BSAQ+)%Kl;71hL4BNPTzaBrQ;G6>EC7X+)ZA}>BJE%Sx_S2kQ6q)fYM1+wt
zaKX8R&WFnY!~Gi}RsRw?<89I(&V3^Q*FTq&L6xk!Ix{ZZnc^s)J2y@{Utbwr2b!90
zE;`6Jx)GCslh$tSMFKl(03U6#@PxKFDKZ%V7?m+;TEmQKE#qAGH~=u0y}`y6sV54a
z?UmBDoy&ea@rc`0DYvO$N<YX8e0{Yq?%lQt`ZrrypHPA+>qS}1eF&2?>tWuD&`thC
zIEG{H%4}kwrVdd7ZM?TZ*XI1Cxyg<ErTwTzTM&0Qiud98%D(kuV=g5+s!WNnZTx72
zR6WUW?UxDoJ_I_106{>nj<YaU#6BdQJAYpzC0Q#IqEEmqVi(8VITX)vcO+s4o-(*e
z;?O-M1e(!?qR4_7#Wr6`yRINbO!BvGnNO_sFuDNZ4=Yd3juI%o`fdPPZ*Xu<ppbT@
z+)FjA<{R&6fVTC2$1q}IpxuGDmG?koU%<Fke$}v9E>10T0%k!@XMkZ1YjGgYu~rIN
z9AL#nwSr84{@Tip{rmS9Uw%n0W?njf<`@oVo_yk*o=7(+va+k;`9?n4j<*9xnYxb+
z_}WhRIB^bpoPgKKV}De;%XXaYa|3)$*}UIm+rmNTHV?-jO;;%4s6-cS2~CeDa->$D
z;Lz<~U=%{m4CCq!S0+im5=DkqS4$R#M>m|!$f_K1iG;gi%fbQoI=^LB7sglo&drW@
zlTKlhL9w6EdK~2yN3fgBAq$fUmUcgu*u;nbaOJ@I6}nk7cVs*|&gp9L=J>V6`ydhL
zp^K#_k(1dV-hf45@o4F?z>Mxl`kc9!HC7vTOhg+w?QqXE+OpjkZmC_6+mxyyv!;g?
z|0fuw##lbB!tYaHxb^*tJ*(iHg3duRX6F|*Mg=lQlHHh?AYi(^vUmNcY&@Y%t>%xY
zc73%0H%WjS0=OU{u{K}oypbi!(y%0?^g}K*<V4}929q3M+Q(LNe;KQC=F%XtAoo{t
zlk~W}NSgn@uuo=q;@j{S2`IUI_B`J%QZG!&gJC$(z6#AU%)9~mPil4lJIeVbf`(;*
z+${<X89w_B<^E>{hDBPa<Oiwyk2FyIWJ(ezB%#9sxkTz~D~Hz)S34eV>rIE$vSZN1
z%HDC%Ohh+lx>ANM+Z;9x&ybN*iSfo5*jGx*Xk+CJDaL4A0;SrNkD{eI0hzR69Q<F<
z=*S|+j$>>(2tQ2jZ)p(3wlfGi=FTtc<;c4YKzm<0MxaBa@gua!;)vBAlc)m7K2XPI
zDIz2hXq1!sQu1U@dVoAFCFkBQ?VQUQ5o@Jn&M5gW%P(uoFKOb)B+gO57E+!uj_K7T
zkUUxLx^#cka{OYlcE$)+y(L+Dzmd7kwi<l^qKAV<sJ2aQxx8}lWjpU<iJV5jofL-(
zYKz5*T2UC!B^~#_k5OF7pKlzWT-l)lxEbr9dL!?hXVI(~Rb=q^=QTT%jFbOi1rrLi
z?){%rhGyBV@h>T+HGVU)jIz141-{wZ{vm4rD^2S@s)z(~M__bNC70AnH>DVU45VAd
zX7$j!O}4}B`gp%YyB*kFz)?Wbz{CY2ILDLmwhVv~PBAC_VB(b1F9hi|v1MRqXGkfQ
z<x;LrJ~NxCXTa&g+$m<3bD0RJv&6FCe8uvxR2c-x#Dgs0BO`&=Kjn=ma0VluU@?Lr
zrg_u=))g!hj1|nn(QFpOLAQL{QCkZWI`Uq0m5#hlOkMN^2Aw7#GIli4p<Z3u)p#+U
z#oX>R7#10+n0T?2Sn+%b=zX)G?*u~Oc9ZzGNqZfME-k_DYyKqSm5CEa1Q687TZR6a
z#^A2?b9U#ZI<}}jN88k<><kf2R~_CNA2*z9hFD5ehb#@<H>vQOq!b{E;%*aSi<Hpz
zG%Kd8?B$paP8sjmfYWvQ#+i++Y&qPt*D=>*&LapOh6{2O%|PRE#n;z5Nt)5MU)1Bz
z65zX7{_jM_G&E3uLnHb{1)U@6|0HQ&*U$-~xxvZ+(+bIr5baLnO~-K2R0|QV@F0k0
zf`l<AsFm2k-)F8lQ5oP4Y9A#q#ul57Fc-ssdWI531{cUBI$ySXn3h?-o?JU>_j;ms
z+8~rvEfV>XVsOuDnG0Jt$XL9Kp$fBjQG@vM4TeDZbL2@ptN$g9%s*0KV|rMFnB{v_
zw!r|zr`C;~Blt{s46mqd-|#M$W6{`JESMTy5G0i;V01Ug!R%Cg^P%0Wp$#f^p-fwe
zA0Q(8Q)mYy9U30|<YGU^*#g<O%aRUb9Lq;>zD9~m1*AkJ1$a1p0Ih+YhV!+B-G)&b
zj$9WR<GfNl!N@=@*e=gtT4pCAFV9x4n5qKLi1^ly4u8cjgz+o>UR5>9HHovk_2ZVq
z3G@Yd)ZU_T5=<<@ESUme11s4m+A$KU(j1A`dC->%Y3j6}!Wkqzd^0I-Bv?SQM`t90
zB;G&Vmjg6qtMV2|P_`}_4MsCH3j5TSo>4l#j7^ZSk-~QvB^0Wj1FyN2Un~@(pXfy3
zg@=n5%+2Qdj4KawLe{HaXr(t!>N$7bfMYxh1Jbm~#ju!mr8v3h!xJ2B)+%1Xo%aO}
z*;ZSHALTsYL_I2M=WcO9WZ6Zsj^C$D*V>d^4EfE*+AILSm+_R2?Y&HTn7df1EMW^+
zb4$V&LZ(eO^=sNQ|5U*n3VxlSVdMTAO889zY4NW}`758J>=VC=f5G2UMz_AQJYSiv
z<3<lDvz}ZjiG)h|iibl=rLJE=P<JqdWj7={wTcA3idm_gMp;Yjv7PJTeS4J--LO`Z
zSZ>=_YeZNrhLT~Hv2$W?Ia^lgg-t-gL3=EqV2fEnT2**8OS1FuIma{mX_$OWi(#h4
zNXrU#8;T8Um0ViQTB~N`>ZPMrSxd5h4YF!=Lpzd<X0cr5t?bb3c&X1EbTavoldDF;
z#MLOlX0%kO&sNHlK5owa(onOC65ZMFMk3W(i7*qghjjZrrCO!I+R{9JFimIMIJ9hY
zFVtw%z=mlL6X&k+d)4%uVcxwD2EWvGE_%7dke%o#>=BY_7QIU|yOTFaU?np95J!w9
zh!TaCdS{el$;|Yw^)ubkhK)nVd=Ju~=G^9cm{%ZYj=h?4XRnZGqbrmR0I<YNHMfJm
z(Id*u1W+_e=Wmpb_{7Z0FJn9;^YY7}C(4PZ!^9$m6AHW;0$FmXp1UROUhT1zG@i^s
zdLT!{DY6(q8t(FuG|MpE*&*nS()3d@dn3J;<1fDPoSTX3@ik5}0oil-%S$11?~9(-
z)g6vaYc2MDCSL8e-lqINje7g1e}>)~aY`>(*HN#Cp<8j^s9m8PGxb!A9DVn|`mc=2
z>5H>2KAqs32W(8*zuCP<->uj;{&tn(%n-Y7;}~pct1%YUPuU2_LBz-03@Y~DFdFIY
zjnN?1F&fHoqv3MyG#V<`7!51A>1e2wt-C<6Z#)_*l~>tKMq@);&Cy8IyP!{damT?b
zn&vRXSbpW>P3+J(<Nh$nV@6b}T>WAVR6AOnI$FGNw0LE#Rkn%DtT0zNESbA%SB3NE
zQ7s@6Zi4L?(1%XGb`+Not>;s(wc2aZB3wMs)mAFZd9(A*EeF?jc$h=ui&n?=y-L`{
zF-8v9yF*4l%3gQ&#RhF)C{K{$j5CaIev`*A^Kx|>_X%O@ETR7|1^@0+F_Zb34B6nF
z=<BgTC$tlR<NqrK|2ILHSe_0O9>0?{yy;H)O$6xw3zhjl2})ghM+88_%yRTfCHkS-
z=xLAu_F<*&w&JJP!^%D*$km=}=V3uhaVWj%(WH6kgAXbu4q#|;K>qr88cn1y#K^pJ
zJb5&+GkzC@{${Bu#9%9XHiN3x9rw1WJa20JG}MgrAXfVVAJP*1f$rc%O`$e`kuRz~
zXC4*gL98H00FxAVc<v#FkrYNxNT$s;T!v1VZdl##Bc88)ADdQ6$$NMzz$wc>s_P=Y
zQb?DFq9;JdAk@|F>K1KpQr-tU)?Mvw_6I|y=6k!VxNDWO)VXe?=_!91XaC*xUbu>i
z#rohwF4CiZ8CT&#RBly?H|kqjBIw48jr+cDL#^5PJ}woQ8}PrE8iV3okXZHN;|B--
z`XYMR&ql)qDJs`g&^Mc;b>vv9)VKZ&_<yalrhEZ;>Ca)Z#pdn`dS;TV1Ds>}dz8X)
zD?J$u2G+KaHz?8<5o^x>l1q_M8P6I2*LWUrW@5nr8(c?NaX2-B?eSu@gbQ%!ONqwt
zMk$S0nB*QCT)}jihPe!%RGEj;moF8hkL9cG53bPt-_z>;eS)#APFiaM^QNra|Le*Z
zpXvXU0+Yg+X6ti$)M{~loo^^+OxUMcM$X9OA1mzwkEJa#Z(YXu4tpNj6}Kp})7-4H
z1rv}q&ddDZV+!*3si)g_DC)VBmrgzB1ns{=gW>kb0;*kv<-i6ui%w86<wW@MYAziS
z+RWym&88=D<@jIbbtR62%Hc-Y$?W|m%gqG=Q~s~&Lz*L1N=>P;q)uC>u5#FVy;xd9
zSk-=(1#mB>*HYC<a6gE+QKwRz1U91`jg&*~SB8n1nhxVD*%EQqGs(G6>WHnE>dq<s
z6oYEow%PtR`mRG_CHhXkPxAZK4-;^-c0as%>_&9vuFf$)o|8Na@!PR#xONz(xy&3|
zC;UXojkA11VWWLLp{rF^v7e?y0jI@s)?%g6L$Olv!;u|-TQ77Ri7^~;2e4nz5X!>g
zX16gXO@4F+Cdc2dTq*Q?2?@hoBw4DI1_dT1ZPJB2OlaM|W-_X<z|=)2MNC#AHI2$O
zWQG!#l)4)(DW;_V`(=BHdS`U&5yiCBeo#}T)0x<{0bu2qt(VwF?cYXG+8syR|5Lr}
zYYN^`kVLey9tG^&*j`y4k<3i%C!3vOJ+yV|==zy|w?^(P+!4pcBpT=4JOc|j(D0i;
z<)+h|Zg$MZV8Y*4|0FL0+iXyiZq%=BDgk(w>NRQ50EFV%?5GXr{_(XS-plEnu;xuU
zKHbdZ{8#8<gI#%<xLojim8+&z4-+$W*JInUcKD`~M=@*mUsT`4G;$tHyJ^SzF>OD3
zlcrqbr;CV~4^b4g|3aW4l*qzZLRbY6{NO*y_N9%@$nAhSZ{?EPVN2e%WRCqPN9s9a
z9HRD2!>%(|_~T>uNe7UdlX*|vpMg~6Z|!Vkrcy9QwMzyvm`VyW&Jcoiwh2dDpM<67
zdSP@6^n^2B(hOs`ETd)B7!XTLt|ZV%rf0H2Iy%voiB7a($7r%8hCXKJzU*w9S^Yv<
zeNw4Lm;JV4Mys`uiS6)bwZvL@#1MyJil0@+k14oE!3hPU3S?W@I8Bq)Ml#w>Q;a(;
zE+23U;GBxewt}ON_~Zx2)ywsXMLROb3A6PAKr>eSWvv8B4G%$2;osQ!QyEUk<<)`s
z%GUKOBQieeaB(4>Km*E}dyzUUgbA55)e(zhkX~cVaipLz7DFiw(TJD8FiVaRa4j+$
z<!St3t!TkmD+E|U2D2+J?Fv%0Q)mScM<7#`Ac1pt`!vXV6ciQk@j9Tf4h=WQnh}vN
zsfMcceHySorb?I+9t;Z+3KXndP~MP&&ixa1aeGDMnwnqN=*gM}!xDIVM9*NhVY^;{
zLq;2RRa?ysa8tYNNq#qT?*o<zFt%qT7(1B)>*)FpM!^AuuT!mI%X^!+eNw4Qpuc<-
zIh?;i05z&}rZV-$cpy5e;0pfR24IjMjm`9ecdFO{XaymfA}Q&yj8TNu)`<?2bOR?4
zb28I>S)A42PII&l)<qx?6$E}a5ob!!2@tFUN#v(-E$?M+#5DqA*@<srlA8S6(~;{l
zu8YA}{j=wAU*<2E-+DPjGuK@1LKlv?Bh@=LDVv$2@<OJ2^gKM9HJh58t=EpBh~VPL
zg_avJyK=}Hotmf~Mp(%WnbG+-+CdXITXz>Id0g69_2k;joT^3Yl2$_vt`s}9Z|Hf=
z9X4pI+u%|sigxlwOznrpU(@%ZKwswGzaz+1^#M3WlW#RXBPxm4M`>ePZM8U52ge?<
z)x@1PxXn3L%3{sVsGV}pryI|w&B-u&HiK^l?tV%u&!C?%p11V4MZtgqnfTkWga+pN
zBPHD3*np+%B|w%O_53F3r9PfyM&EFyR?1zRXXP=a>vv<pC6_0cyf16*7N}mx?F4*@
zHC)-g9(=WmcOYazFRfbJfo~95={KP~m_hU*crE@GB6?vFFi;Xh&XG6<A6OHKfRL>h
z)Zz^&z|Bg3P#!&#I!wb~G~ySb1D%KkCC9p)8vY&dCkq$E+D=Z0?`ST_r`%PRjVU&N
zl@rBN$IcWnLtMI6w2OIct+y=;Py<IeH~Q+Mu=Dn}u6}D_L<_WW)kE#!;zXM>z^Gj|
z3?hE*E%$aFq8m3iG&JdB7Br<RWy~|5nwyt92;X_?(;9vjVLz)=SIE{HcV1T9!flR1
z@eV)DnniAz;vUSXTim)dgPuXU$;R!SZ9Qx4BUmkNG&1RsTpATW<c_pS?sK$w>ZtRD
z$OU6QIt{+oK7lJTYOc?1<Lh?O^GqBVTRQ{q7`gz}=BQ2cEX2+o2Q^c6+V=Qq3bzi1
zi~n2VoIH^Ix3o=U4HWjR8^kb)o!R<5F<jo{($lYMcl6UnB3}ei9ZHB58@ydeY_y=-
zta^vML`YY=g(hRQI`oKk0jFPoh5Z2SW`^^quFPMJbenjo0mE}84m4IIEfK~?VIW!z
z2UR>3pzo5aU1k>`OF!fuSOMVq7@0vgvIc~I^4KfsNR-}-@81|^);(Tk7_<AXD3|MV
zxL%WTedP0(zI8X!phzXz;3(J{Hi1$|H2Vw1Y5nat9E(GZpx^pC5DZ}UjoieJ!aa{9
zL|%MFC-y;?%4HhTnYli$E=;X%AvVYbh`qrOYB;X9@w(t0M}QA0ybsH4yYa<(ioTc@
zhLDL21|v8GNkaT@T^(KB7HqSAcg-OkY4n-?ZmaWdBqdveZ7A~!H-ILq+k@?EBS59W
zOZ5%M;r>bcaX1c`$L6-ewc-+FRer2T*R};*78Av{TYM+ZPh=3j!~b2c7wW8UbvN=M
zE+Rs}%$LlK&TXIDIk#(7j?s5RtL-$}eowH6%S`*<l1Jd;>S(ZQtr+aabU@BVinC6J
zQw(;QL#sba8++T@7^IE8YkQlnQ1-fZ#)kg<PD{Fh`_5^}2K$PJYmK%e$C1|~qxAhA
z?a2q&05gSu{)0=87JlV%L!?urbo^+Kn2padRrnfD@9sxsTcb<<K2;E>F&1~Gz<o-P
z1i@Jeo#7E$H{ef+tInWc5g$8eJ~jV>%1#h)*PuY|{KPd~RA+iH|0&W}vL`&VE*snD
z*p<Jm8s?(On2$cO8F8fvHgV}1dNjKw%t5H1R5uLy8=Lcz3L4%w4(D~{enR!~x~$DP
zlj>F(Gf?<-#f~cYtO8@X|F&X(OTk$J6Iop@OTd(I`ZQ<SrZ4iJR`4gZPS+3VO~w)#
zLD8dLi=y}QiWL;-B$D$eC3?RU`I?JF?B<cK)aU%qMQL1!iQ<u~1)Zrtm-&&l^XIiv
zU)zJuJKKrRQpTD&T^i|V#~i*G2X<IJ{vM4-%KS1U>NP*3ZVf9#B#*yKed~7h-PvQ^
zwHE0tOVS85td7ht0_XM2Q;fe|&4|x0W!!u^Y;aNyA5p{N9kVPdGomkGM%xc*kMS(v
zU-pk9m;oLzaw(PQPml7f+!6H9vL8$Q$<F`SyI0b?)DV(5xI@Vc3b$o8sB6>|A0@ug
zh`zCZ?3BWk$g;*yL)!i%3ANuNz`>I6fCL~=K9Yq4OL1Ok9R7ZCz$I9Qj~I`i6aOHE
zB0PFx8dpt0-uT2Wj`fZOUBI6%lR;Y@{{d=C$(`h|v9M21RZ@9u58*x^54w29pUZV!
zMw{~AAM~IEN1Z7vaU{;)LP=Sv_1uU9l{suo*ef|W={ZzQQjum2V)w6tBY6Ksq~ZfW
z@@}c1ICxZxo_p-s^2{w}PG07EHF4(o$Uw>C420jSS+gCky^h(B?Lzq?vZXOQ&8ZeF
z5GrnCN}1#LtBUQQm>MjoUL-7T3JKhf5NOiz5R>vYC5k#ZFwl@Twyw;?F_@DEpXBm$
zy;7DB02xeAbroS0Ejy<?rIgmGb8RFaB=8-Mb=Jz}ws=Q^y2btlb$W|B-N{$`5Y(~n
z(K<a!iIlGDLlTVdKUP<PCp*l6j9AuiTBSMwqF)G@4Z5Cn?l^t1G#%&4wvr%GmMuZO
zcvGL8Ou0=q3_xT7Zk|n9eBCV3JW7y^zn6Q6LBcGcoPSr)j~;%2T7#>@)ih72IrLOF
zO*v<=(s-Q18|SK!f08R4o~oOendRijfyC+-yw70KgMZJw`S*mp!M|s=E)6<Fj)!NO
zrhm^7-oMC#(;2YAVY_?+cADGjw#I@XZV607l*A3?=KPP|q1>?LZ`uY7NMW2cxXCi|
z4{EI)QeYlC%xT6Lk1PEzxqE|)9-33`O*?Lb?|(=WE?nlg&8>R;F+Dy(3^$dANeq8z
zHp0O(6~G&ISug^xaQ~-FEq185ahh)`_K?bD^%g**lag&ytLDzZr0KWS^wu_A=dUW?
zkb|~@zeA6m9};%eD+}<-6RiD;w3#@{E;ruM{784o-=Ma-jJq9E*Vfzu7a;fV($LtJ
zhDVI~yX{qasV#bg%Q{W`Gg|LzV|xmBHTDmk$JbJ6EJIlRtb11fAP!Aa!01$bC;F1j
zAZ_GGQjB)vN28^alu3ST0yRN@P3lIe?f`(R19(P2aO&@l{tmv0!p7N0tP-D*e<aqm
zPM~wgbVAk~DM-vu#ZdW4g?UQu2H6#m@8%_S2xVmN>JSFO!?>4G%dgclZ^rQ}k?^1S
zq+HyXm0rumO?BAv5e)jT7{dVIZRA$dYU52K=sy*Y&NEKKDnoQ*9xT;iy`I;7nAoWd
zm^aoew|sN0N;T7-3PBF(58K@Xxc@|uN8mKT?#l;TFjZbz9pM{#gF(o65w>Zb92u`W
zAN&0;5)RA_a#2dMb|Vrm4+UEh0q}(qZbPDS!N;5NV6>iG8+j8V7ji#Ie<8AiA=1^(
zDCmqn1S8mk$oiq%>v^kSOvJWG?fUN>qO1Ludk4gQF4&6N8M5B)$?VmwqQQ`P@f)J2
z{gD@r|4e+q>oXr5e-I2q%5O7OY0AHa&}`&7v|EsYXqWvL^{#`|F1<<2cKld>n_IRi
zR$}vqrDD9Du=H2Y5`bZ2ukDc~YdywCu@EHLtuY9SH>RoDEtgs?&d=p)YMg5zOolsN
zqgQHe+H0IHrqNVGtf{Sn)wjS8e@s-=+C8U8{R+G1tLkreY~{V{1wC?#1#evAcJ2tP
z0Yut}`KF@_(wv9Vn@J}(uF@jTE+&@zdA&Mv*I|@p>-~#l{kGmO$|e&_#EZ9eSK!J+
z|K0uQ7#@99nqM0poqw|XR&tt!#xO;9t{6vQL*@J}IaTewG|&cXK0pgnOWJ6S#{S!T
z*}Ufa*2oL3fh|5=!5Gq@wD}=6mL<$wxHgFyi?K;M)ndT5g)8JC$UmW=1yGITF;-{0
z@6lH$N7O_(h`*oFcI-t<vTqifITbme+%GGJ03y<ui1~2(R8y+V>Cv?Ird~02#n_w^
zs$q^DteYRB+}I9>Gi=5UF+D-5Gq;}95dnQm*o`ewSr*tz;-U!or#67U+4pJjhTOn#
zKFpf5w|w4xrerz}n;3_1M8$Ht_Qj6)h+**eYoxjeN`<Q@WqMPem6KJyT}pg6$q%ZK
zuu8aP({=xw%)KdQQfVED)$Nv!1Hhge8eZCX;h)^?BT1b)NF)Yt%TKW`=F*K1-AOd|
zIMtR<Cpl0nP{&-d#tUD|rNtd!w@N*yw}%@z_-EP$Q{pxoKO)IAn@{p<HaI~~gs>#F
zK(4yl8=<E@4BpI<Uxb1<_K%dM9|R-=x_)#_TqJBI4pNhgnI$IzAst7XKwRfbpw%C#
zbKaz$2Z1!)d2{x}5J+1pRK3$Pz6QZY8@1;QNEhVT+nGHgS!%}NQC_qBLD0+W>Ja_w
zGAYqe&`TU?6nu>&Mm_$g341{|Igrbh2)XnJ!yp&*kMhk0i6~i8JlFzqF(&{Jb4jx`
z7>28Gk|$6Fu8`7vwd>`NBWc2UKrjsY7y&&%YRXXo=%e4zM?c`dCqf^8lX%waHuUlH
zW?R!AS=#)P5Qd?TF8@oQENU6@D1&x?O&pDpjQx;@J#9$D->aSUTg+a|6RQ(pJf_9@
zIJs!oTd<(D<+r-M@L5$AE!4K<O`WjWMz?MG=h@>CUD(>=ht&P|s;f_^505Km+tVP|
z6ZQ?tc#mRFDlkB09p0(B%^mEYQEunv^`BCjp$`j;0+cJ6LYn11PQ$5;vkSJ1$+@ar
zb~4(FyV-*&yKlD0vktqr_8@!o9o<CR*nM~xe_j02W>&==Ue?XS&XhQy#*a|2p7mSP
zz}kY3`FPM`A?nwx6Q)zEg^0epubzOu<KULRH=^%gxf>MRr{d6e%Ga5q6Gj$bN8ZR?
zN8hnYAr}fT>2}7}6z_QyetDKyrju-76r0qEe~PDfqbq#5UUMQh8}xhhxDGRljyTI4
z`1g^ya?g6W<tXcSB^&(v)vQPb-@tHdenlfZrMBf+YGs3j@y>1EkrePB*1MW0-~T2(
zuKOb4`uFLrkNnx*+C=Trit8&{n_!dmkHB$;7dWr?Eo5w{TI(6})I-X*jb$|3H*V8$
zf6wXNHh4{JW-`J@`Ig#F&AX0(w_z%|e2|5hIHr}6&1bH|Ou><AvDmXTR=Yp|K-(v!
zx(*~pD_k5+Vp=B|G6;^u;a~(djk$s+Ywxr~O;GRh`ij-M0YU^!@a#wkIQNkHTynEA
ztYU*I9CS{u9~pfyDjt1c+sGc(P!`q=CHLB)<O&N4ojB60JvfJA33l6n?*{XV22<q3
zI+(+~b5=~+)#o*^`*^QFhD3oNbx=MCUxyE~g|Ev~hQ*T`^{B08I`pU+n$xdg@!uxG
zun6~_3~*h>P&=!TKA2066h<k4kQ&+!U6=u|K(BzdMD{zW>JiS<p?M|sF0`VWB6O^>
zAZeBmi-qO}OW7xhfs2_4N;T~XtR>Feb;*|mJ|!@zvbRjvyRnfa&9piL9G~K7foyLz
zm2ykic$-pj`}?AN@T<)FeEFjQOyMKKXv*K!kQWI!#JK1!uJoRg;?OyzB;x-sWZ~D&
zHjY58GZX!^V*glG|91r~Vfq*JI7x1}TV{YACyYA`N8SCnUsgLN0D<0<Zosypb|Kg5
znH*kbu>PN@RTG97lF|xv%fztKvw)4)v@}$i)A8&$(TkF_nH8tW;)I#{by6ugavN-O
zeEshG2EF}=dhm6Ew@>quseXdS&oLs}sD_CC2K@;7d=p$$IkpOACktxb%9(HuFipB^
zN8i|@+Jp&oP-*GtR=2E<tZromZJEOh3+%UZirIR0PBFJ207>B#a)iIkv{p}Kv;Ei5
zGh*fG={zULpXLhKcn-7vwgy`%n|&l!Jx^-4mt*xUvYPZB?9A*hcaly^Y)F2TLXOvk
z4ig<;TMM!TM<hTC-?y@J^ESNRcWVJlgn<+*^V;`1dxQ_hs`)FidS9@Wv#DQb^c4N2
zo@JFifN(+XHT(G1Zo~sa^+8;Jjt1Co2hR29HnZ~{ww>{yi6IbkKs9o1zxGcMO5lXk
z>o>abkLn}!t!tyfc3EkAh-JJ%zE1kcZIs%M=5hym$?cTcDdB|a8FvNSXk&*rM9H1O
z=o%L*tnCSQ-Y^~ErN2YjVlYbC(fZzC)EkCc-$iVfx5Z-nf?dI=n@?{9s()9oi*MP=
zx9BL$w<8c4y&nHgv?{!RKib&@(s&L&dfVPg1Gaas?&;8w=1+0?QtfoRj?{XyoA+~5
zA}w#H<pM3ABmF?IhxEO{Ueb4vp3@ua;xP`@4`0_^yN#zK?up*)q&En=*Yzgu?J`a2
z=KWdO*q_FDHv6@k_xE^v?1tEXM&HKx(q5?keY{EQw`ls>9|`s$@*4K`viA0UPPw?g
zqermQojpREvClj3mONAJUp-)Bu%CAw3=Z<G+;;07`fTd+iAT83+CvPOa(=1Z=>0}V
zZ}Xo?_@6}h>~@^~YDefVJ0h7(Bv%iu9$r1dEZ;?s4+lp$Y8b7R)iLjIFowM~V!z<9
zcZ9uC2oA%+5F_PClyanzGM!`$qVm)lo7UaVV|?LVd?5nhwlCD>8(QDy^p-kz+n4TM
zJ+R?>619gqzDH|wccX{R(Lbs;Gcw2U+q5Sb8+GqJ{^xq{@}J?oCwOlcBm8rG<w@R#
z2?<|$w^<<^C3Z}|(-PyJLNAqAy=(REO}_BQJKr>SAh=r&3)xLagQI-ay}TvMTmD7;
z_}Ym&!r--&Hz1L?6+02_JRkx2zt(h<d!KhdZgR6{C>6QMePA5V^!^_R%>SW+KUVN%
z1^mWLXC3x->a6_)h$rlpA!g$sI<mL_q&mN&;73*Q0fM);5g-vXA6Y`UVb_J_g$b@K
zvO^wQI?gsUnsUmARW&3W>~#1w#p(){6?~K+%)KN630dLIbQ`JmG=1~mrw%4LR^M_b
zyRxI<*SGl1@@Hd*pU+fGao~Y~TV!yv1^*SbC9bkG?5HS?qnvc^a8%v&f46!#c~;Y5
z|LU;`kGo8!%KU+WE@~7$Nab6H8Xa|pQg+T|c6t_j#c97%Db6h8IGq#NH?w!_M7Xn)
zplDP<$A~|^()P~W_K67x8$F+zo^i#?x01(CPdjaoF#GY9{^rm;Q(eN@V;u%#-FVng
zXjq--Ok6f+b&W!;>9nqDN%x}+u7!zQC>Nhc-=eJ2P@}9gFj3)KTbkwCM3ptxk_ior
zCwWw_{3T6yU%R~Q-TW{6-=-zfqLqx{`&cv0NGCCTe+bFqUnjDWD3V2^b2YpfLpgTJ
z`KlH9yw3HqJDqObpHr?0GL2N6)T1_zJBTDMKFmE^x%QOrqoWmF<@YHqQ(r9mGn0<o
z{IZfwv|=I*(TYyy@>7bns82KHFaha3`p%Cj_wOtA4-_Qzm{;(O-v1AkW^}D7qb0NU
zCDZdoPx@j7{IiO+xJTKQ_#aobe@$%cn4?x-R>CI~d{TiJA-C~xzX;o--S?)_ObCl=
z*`_@lSYmB7wAbc<9Z9H?)=h6M9C9URr>B=?W@u%7Imf=^Zg60w2XAlY)*o#+8bae%
zf0cfRIc2DC&OwRM)uuANrEVInZGa`a;w*ig22yIlkR-Da!E0KxKXNSL+=BP4w0}^T
z=U!sBGt9RP=QXFouEzRj8~Ubf?`f7QE6pTRzoCx&rUDaVNl?W_9W60_ExJNP-nDIk
z+p@M5{VB>`W<}TjBcIxvK`15haP+VLYyD|+p#^hbtO2aTaeh_esHLP52?SI+?v9IS
zgjzBmY5CIDA+G}M>24ZpnK&x_+q-WUI2!MciH%^ZnBn#g<s{<W@9e%kNH3friac_=
zbhy;77t(|A-TyU{--nRwv@t1qn_3&R9>^6^dy2lqUC5_ad-cYITrZd-j}3Oit-aeJ
zoXIjQ+CRqcN-hHi=0uqD@uE{@aJdpGtWVP^h?IXC?gSf?xh$4)@z0|x1D_|XZqC(}
z`c*tLrCKSfo0sW8<XtO)y)frUO?oTa;V53m^(^X5xDexpQ_(ZlF!roB_xUU4g{%K^
ztX7s3x;{W^zr-Q5&aH9Sa}L4*XXb`<6^@+adil@k6o*~O%WY5Yib{B7i9#?JSUJdr
z-SU|g*~A09&3BfeVJv-}lkvSPdz+4coujPw@(lH5jEVn(7VIMgE5(;8<~qN)yi`1A
zy1H{O7Hg5mN*}fwr%-t=vT=Gmf1kk23(N%vMDxvJs7&NCME>RSAjnJ08023Y8}>h~
z*O>UqK<k&N=ig9TS+Nf*m{2gO;NuEDrr<*g?pE;aD*Tp$B|TnJAhn%OH-HX}4oJGD
z)Jc^zf3gKP7-g)y{O=^Sl+{rlJzdck*iN*-cH(C!@h94e;+UPwb?qtLfH=rHc1JAb
zwcZ_>#7k;Y`!w}+yq&URHd^_D&vBdA>C?UA6o>jRWMpewgX7Lky45YIOl`EDSi}6q
zo#24Sz-7b@yxi#|B)--%bvO~OL)Gf#3<vZj<|{nm_CcmC!)mYVhC+MZ!A&(#(&U>w
zMHYbbkaGC6717r)gxsT{xv2yx5Rf)P14m*Ln+;f-_Cr@6QSmob-0Vf<EAaAZ&+w^K
zyjT77CPS@J7~#vEn>@h8ZiZO)7aOIF4y}%HdCq;j<m~zJaqHLFv;KF{5$LrNf8^3$
z>(PBf=Iq&V{*5Il^cU4zfsL{BS(34~;ScGi9C?Ivl6ucmJZjg-D<1e6>pDBm>y)K>
z&SyR;O2D9|DPcM9pVKE=W9QGGJ@341<8JV%Ti4&CF8n2eFk8lz@f4?zUsGRyLg@v?
zY+TJc^lvK7oLT;$Vn3|lhZGp{|4GG6iuF$w`)3ONrGo#c-~$SbdPu21cB=YYls2xw
z?A-d5F{EId0vRs)e?`HF0#ni&2_ZSNBMgl4Fakq}%Dn3Pnmzxo>an$h%xk7?Bg7=J
z3zVq+IG>h_fmAXLbk_LW`v<v466i@mPyLB_?>o6XPVe)DEjY^WB9_hd7Wy;ys&xPD
z%)Q9Px-tc1WJ#Rm_vQw2De{M7AH_IQ`FmWe-S1>}H`?$1dgg()_~~Ye0T5CT{$WO2
zxv`IOebw&4xUYUs(dCXGz=8eLPV%SWY`<lLQA)~5O9h!_C4~r^d!TS8pEjrLIz5;7
zXQ|d9z7<U;bf&-2^|LtNjo*M;VVfo-1;+;!t{i32W*)=SAQ-l~ZSf>%WQWj*xwK4;
z!`~R3mBzZ<<Y&%GSh|lFn8$G6w1WqDyj5^^1LAl5Z()jnn6IjsEyo|=`|+N-uD!At
zB<DI5_e%|u&bbT~ksSGt!{a}h=Jng$hy(W=4Q=b91!MK5z1sL%_d0Mc3V~BZ+nU(w
z6`wPF6~E~sdc1Y^u)|4@=;9{qoTu*$kPNO+j38YB#>ttNbBAO(It_V1E@y%`5}V}K
z=rlTN2$ja|UPozXy$S_UyGIaM4q(X7nFd_vEv$WX!#dMyE|=^bx~%kDH+x5W8&;e5
zZK%x&0RIgqeXY$Jt*}XNH*bkrIW*s`p5$q3Eo$pKU0bcMX!U<Xy#n=qhOgz?Ji9ox
z@tu;XTI+Jp&rP@JrEP&3(0zi@x21;7?f=2FxE1$NEwF@3*sD~nIA80~BovfYDjdA=
zTHmHjgM+`W)?U}+&nn+tMej@`|5%mxC@3lrlkWdl6&DW>CZ;w$e|tpB?F{A3&*QGb
z|079zYJ1@;LgQFu$5X?F-&Gp1{=)A|*W1mKOm!9hNF+D0Y~hyaj8mC+xL2yN2^tGE
zI<AYF6tuv-o9bB;<h!k9#E(JYJ3>tXilTa|HJ!ts)y}K~9z1sZG$E2seKsJHz=oIn
zoU;xC8of~fG6@>RN~ua(XS%&XgIam60Kr_Zly#|GGtrap@rJy9Z{Q7GUct$tRIDVm
z@NgI@{9ax%VMX)1aoP3a_*jBlaTc#cf(f~*=yRr=^W1noFGm<tOo{yn(&?iF^~5RH
z?kAAD6*f@{)3>xVekNv1_`O=HY)Ce#5!-@%@<13Xqm+oO)>ttupd~so+|uq#*m`Bo
zRX?uf@d<)RO%wLFdhAFuzgeWsYyKt>>sd{VAQnc_#?#>9H8P1_KRQjg)j{a#+cgeP
z^83|$2~d+Y-0uSW`M%A(Kg<0tpvpWw^Yu3q<7aO=H_qnHRu}#?X21dDF<M^d-`4+o
zj8Ge%HWd9y2IVEb>8p}vNa5RXV7n5ze%SC;n~w0=4uITGTLQ|4*?yf!kgoMb^DMOy
zvyJK~h0rv^9nOQz9JIkwELnciR7>Pb|I3jD5w-}O5iF<dU|JVgQHn;6AOv05lfu{m
zg5$coMYo>ewjL&pH^}9V6iZT!z7N%8#_Y$t0sL8hd%W!DIB0;0(iElg9A1<Ao<5XA
z84i}Abv8TFzMyBCjo?zFLj`hsXwud%YgC`p0((KRZ`LwQoqzt!^O!hx0w^d&Tdqc(
zRTvhx=rl@B5M)5_yjmGm@a>IN>lBVAdCF=L_qR!Jen++cNX69=cX6nez#Ro7Vfrcx
z=__Gk<!ac~wo}7A_I0M`tV#T#1%=FbGiNg;?SUfiO(nMDy9DCxw(fwCCll5WbR2R8
z<O1vrI7m6**|3JBv9x2>ZKU(tU(%O|81FDu)0k+#b@5w!mE9Bcw5ceu@;4P_zf_bc
zEZqSVkhNBst+AZ!s9P#ZR9&bioM^qNq-;DpCFKwTe<>-4`70%*EPpfn-O68>@-<A^
z-2F+YEJwvoiLpYtjUkz?IiMtJMot81#r*<mY#mtaLalj0v9<m;Ip&PP*%n&sMy>e_
zsRO2z{1oiJdLk91{`NK1ngdY^IbJ`ingLL7tA!)h!XRqRo13@b4ZUmq94g2yC@g=v
z(Wm;bJ^uo`h_}FA4tT?Sl~i;4>v_g`TQJyEU~aFE@O~7YnBhHy($S7OMheXT%bvD+
zTcp6GcBa0qzI}}{rotRSg}Ie4;Te@>Y;G4S%-!tX9hBJ<pwL8R8JjCgff?-dMk$Fp
zb8RoZLNR&6sWAUBW%rx9b6fpDu+7^BJlIWax3}G52ZP<gHaFYe4w&NIJfkS&Yox+_
z6>IlhsCI@8=N!WNDuKzJRG7O>g}G<7xIv{8uf4ZZr9*EbRV6L&p=GHmU!|9agCgIs
zFOVXzNP3sI*HnoAmAoUosT(qKzpdK?Z^jtCS3ecU+H0SU)LV5_UT@0$J1_O7|3>{T
z_spr3Eh<s_gyF~iz_0^v#vehodTE~8`>pm5H*0U7yW7;GpI2$uLaThUlzYc7MV4g;
zyn}D#P&PN;sJVszXuZdJ)_P_X4>5}S*N-CRj0e2Krv5v)ddTMOAR}@ZRpfwogz*@=
zqoVT<I>uOv&ZE3V%0?g~s{+*-p;U0k^=oRlyG;#uWOZ!AoF!^IJKBrQN%_g$VyuX}
zg1hL=app&k&Q8FqC!Fr=GlIT=RHs1m$9XKJfQ$ahh(ZC!@_X9gXYE$}Sd-GBnuN|e
zB?h_%a^(oX^rp0PLWsC7_3(YOlRHSR!_0=1Pf_)C0a;bgH7(^ffYFwmv`xp+I5=zQ
z;(2|<3ksSVJ^!2a_@aW}<vspOik(+r6#O^V+W9Z5@COu_<<skGMR2=Cig%0y1&!j9
zoX)NnX_(hX>ycF+nb$XmfjKy@24ymHGvUl<4AEH$8zWiX{Q_g;Ebs2S1ESe{gzyq<
z7a;p3eZ+?pyr$q=73iu(|HBHJiYtGD$FUQC29;5mnp@;>-UF&`r2dQg46T4RA4GRj
zCbd*ZPOkE0Rr~)$b!1W;(;t}u(--xkPKD9f-EB0+CK@8&(<hp8Xj-uu1(NlcddHtt
zY)-**S{>5e_!sr~Gm3ptX#)ycYMbv;?q5>y-3p||@#hsYPEVW3U(jP!fvvkm#g-IY
zBe#@Uo+{TDmi$Xfkv!ORG`{wbGZ(HawyeNb(f2B5OG)CaC2co}sd_mMCf!LQK5|=*
zFr`n)#qWas>(sg&HCnk-+p-&T{l*Q_-&orsy-EgcNg;m7C@1Rjz9WR<ETn;H*R-tH
zpWyN+x_9PXz`{*9QQeBsCZPPw`fTmi@iy^fVN+gJyg{2l(#Td5{oRpuy-tnyYb1wx
zK@Nv!Ty%+}?aUS@WE+0RSBWVyJATs9Q>~McZQf&<@}R^Jt2w4t%xsVDuDl-ea-4J%
zBP%u*M@4pTV>h=gc6)<fpVx1UY{3}WE};8R7<t$_JB#Fhxmsy*vMb+FC^mlHJ5igh
z72)Oak1{=t50YxVIPEG-;KjDKJaxfb4>dLBMHQ_q)i0bjg>$Q=^NaGQu3G~d4M;^J
z>YQfqofNfNF;A58p)!v9Agg{ev2x59x#Q}CQ=YnKBb&>U^NUj#YsXLW&z;zhic4v_
zzDj$IlP9AC{<0bvZ(h*{lV0oK;6!mZx%Ah0z}GlxGrPMH;}hbj&aUg0#=IyPL}?lL
z_iV(!J6;`*t&B-2p%hrUvk^5|nFAkSY>Yus$YIxe9h0#4qs%aaQG50ugHkAn_qvg2
zyXA@N+CD=+>^j4hG2B5p$4`wREJhA`W;lYzO_vucQx}(j#C6=lNSf(@aZ^|j_B86*
z(EH!dstoh;Uylp<g&G$CiV@P~g+d^9iim^Q?P>Bxpd+7l+h4<YSk3r9Ku{XAwwf<d
zqy7!$enG)G_2~bt?QCNstI9Y&o$2;%rqk)mcG>RIon@D8he%;JVgyB4#f=-FF<DG>
z;Y&NyZKw2Q_s(>|$t<jENKipPxFG?7fQc__1PmsM?<V>|6N%9n1LF(E#1CMM!64xS
zT7UoN+<WKFOS><`Ha)%f+?R9iz2}_&d7kHgp2r5p@R6G3ENMq;>PfQgr;<4HL^~S%
zjxi!xz;|E1I@H{pL)F8fyuM=aqY7&xBHcr&OeJukf?CfcC4#*bAO%-DoiP+?Vi+l1
zG%rZJL9U#~F;l%01z?>2gJNz&4~8i0U8~S;5P7?vMw`JL4MXOQu@k#+B8EAYuYD4U
znl0{o634YDPSsnv#~3oY<a@OW9lcY=&}uJ6TAkm=|L!N8aa7^w6o{_p#6j+=mQ$eR
z@LvuEs`$i5jf#HY0Z7oglK;I%Tf?zn75VMNj%uQGnj<Nmt542iR#dnC`aU(modM;i
zsz`D~2TfO#pVn5IEnL#DHSh#3(J<87kwS)5<>xqmu$l}5+}W9S<OPo4o*+0Pt(EEO
zho?Q86}-q-@P4gCDw2xaV&HQ?BLv+sJ5kv&VNn^>l9b6he+HW3JbrhVvcy4}%rDHW
z;1;`-i~4a|T_NgSu23vkOc)wqA-g<N&M)W=W&)kVE~}x&l{R(A`@L%Ll9HE|)Rg=|
z3GE|RN#9;%Stl3Y$`Elwwles!6+9*3ngI80E(?aW=`H@KYi!Q*CvE*%iJ5g@(U$cU
zbq(*Y+I@<xS|8>{+ygHa@znjQau<~>ljQo{?Bg|MsI_{xDEYe*ySq&W|G{nzza97Z
zo?3sS!~azBhLV3PF{ro1p#C26URSrenGN~gkSSLnE&~K>98QlU6cb3K3begAY>J=G
zK%!t}-f8*6?pxn$v<lM&uLrT6TDxCyg#ZZVh>BsTZ18)B7b0&11^X&0qS5wN{z~Hc
zY)rvEHB5VzgN>+uV@p83hE}{){Z%d-*%G~;t$6&@c2;I@sg5EA>qtP%4)&eSV-(7!
z({T)oY)uKx#Oxk47@r_t#B6i|lqMQ;udl&K75eDA6*OeJFD3)~4;?!~%=V8I`q5yd
zk>=}nkOm`F7~l?%wZB6l<KK}xx4FKD{s<Kr3btP-DB0S7hEXOXf-^N8w5J+0v(R@b
z4-Ljztn>@c7)Ebk;|96XAPqebKW2_qxoU{3MBKinJc8b140>-s+}=^vqCr-EGlpG{
zrb4=kZRkBP7nNSKz0m@WsZ1UpDqmLKPJG~<ruP^@@3EEV=Nfb&u|??EPhF1QV<+c~
zn_DQkgxV#o2UP4)&K!s2owb@mmX<dB#LJw!+jJyj<twUVg>7g%cCob!O18H4RMDWg
zFAg1>9QKEeU@g7JG7|3&Wa;fDOQSbIi?9{X7ds!v4}nkOzT7k**;Bh)r0I^*kw$5n
zHyCI-h704oF=>jHsqdB5-L#ZF)jgCg%}75~X_e3We`+5ki9}85qCQKH;c9@UV6d>8
z+T}K`9Te()g)2<60KwVb%mgLnsyRy9Lpgg}H4D|@BSd#cbMkyc?a9quS(p0X5Hy?G
zR>Pl@XsgPT{O`zr7b8>SD?OMsIM2z2i;}XDed{XZb`p0R0X8>r0dndkNTc=X%g^*a
zY^492-q9sWs6W?T81GtmjS^0CS6#69xlbQvPRS(-qQ6o_8l_QDfSLVWf)v_=6l&IY
zlSjG7(Qb0&m#bMOZ9*mqRv8#>rH6PWj{G-{M;?)kgio7^pCj=r5Lq4JZ!-~;DI>ok
z;Ppw!H33#aXs3yv?x+c!Z})X!-g>LEZ|Kd=zQEshE6UZCE3Bz|LSMk7iY7)(2z%2F
zjF?oj$a$Y`kW!+#uWMg|>xcxpx$yZo*Zxk}p5-cIn{>w$@i?@sw_uj6XSN0Wka)+)
ze>|O?=fsAeE&g6fs<666)d`}7l-W&XQ|%3Hsr{CRK_F^NwLGSn_&AK(QPY7P^{^8i
z>(b|oCiF?DnZqMH6EtsaMKb1b7-Km74A|B_eQ#2eJTd)fUY3=U1S((dMEJUwi5Z7N
zT7St!0h^nepJFChx?8vPFU8*_K1S~?XHOPqvesy?7RAGP?7Vc@HOr`Vkhfb)sXW{w
zn^f$b(+gLB4}SErLmBe`JAoY#EJ*lnz&j%nB~qmvzQTLiCT(`z3Y@50foqv)nG){&
z$h9WodEcN85$UUqeYdDCNF$1kK}v#+;@ftczgQ?>O}AP;C1L3~OL7lJopG6moPtyY
z<rA4%&p<9zKK)9@YP57>*JO=0(oI%RNq|GMXD&+a`Vi8If^##klI3#|Sy~PeuzF&F
z!p<pI4ISPg@Gr2E_Ty#^S!>aPl0YrR%N7nkafd)4l)JckU@^m-ryoB<UC!aBV%6cO
zjcagjpt%ta*+AC*YLeg~Y%}-R$8}rL-qYwg7waqE3tMWT@9{Vuyta6sLed*>w$o$8
z2-KbtR+H0}_mmaL)f6IXqb|s6DGXm(a_+2X!P36e+x;C)_SIMayS~2eRM_gM=GV&m
z3g0$2>OHN|yvfXJ!iNdxhFS(X#=JMdrEYN5T-+^Ou!i!WcCY7twWQHmkFPFa&UE2P
zo>ehtbb?4Kn^<Jsx*XGZwgDRNtJE60%H7@>HjZ}A<`?eq&a&C84>0AM+NSE5PDohE
ztfvzi#3h~x9^7`u1Qa<AR2a0t0^8DlLknz!s$ShO?2>YqSH}MkMvJ%_Eef8vx+5!z
z#cP&7BBsfTCQHymT&ux4wK$wiZ{tk%5=^Y{8PD6nFsUaB&Et5bwh#0*<=U9O(HCJ^
z(>`5aSD2XoxT)Xj77yv&3~}x#&ku;Lvs9hMvR&(SJ!#83h~QV;`?@M)_52cB=UOzP
z22X}e)@UBl_rCy*sGe~)sGAQ;&>XvZ(TZ*#?eKIwPqiC<LKwP@>?jNsk_IjE?ofvy
z2TNF{TOdf9(MAaBuLWd#(^no|o|%s*LH9ct3eZE|)R$LRR`YAPj&>T(>|6<Fo#uqd
z=_$dtW!64v0nx^T`xKAi5ALF|XmP(ku%lXHU_>pbb#$~Qwiv`{If-dEVypQugom0t
z$3IM(y47(EYXfcvEgcac>m&YAViBX0mG+K&h8qzPZ4juprly9Z&--Wq03tA?J1Tqp
z)z)p^*G3>P$ZbLp5UHL4_Z@<Og9H;Ua|R3D5m-2a4^9Jx5CT5j1_Bf_3n~n%Rfcf_
z3noQUV9f$l7YGW_C`79fQ_T2-I!@H7`B<U%%VL!EJ)9s1sEFpET)=4~UT**p<CqWV
ze<8IkwgWL8ez{;{$cK$<xE|ljL>EXmz{UWufzjQ244}v~z(&SkBQssO?qXmg(8SLN
zsKM7T>u0TdwwPbz?+{E?ThBPbp<4gUQgMET2BKJ_CGTdw7ZaEd@~_tIM>I7K{QTSJ
zUjOawvmd?5**Xt$g2SAV^I(n?T5dX9U+wk0MjIPQLXh#ksl01crOp|Gbt;83e@0gu
z#tcS8BGZ4psVW%#$fVXyF~)mFcNfUidi~Br5y2+qYy>(_sz2O7CF-CfRkVm=4tBJ~
zVj&H>_y!s8mJ^GejUcpllRnfQh$z&Ug2oa&&Q35ltb-6unZd&%Uox9gARIK+gt#Y+
z48TOAWt61Rb}nLli005}iT8t!nT0ggXj!m)nHnq$B7iPwzXV-9K6I%)+~IvkZ||Hw
z#f>DjE%pin&a^?8ebI00H2bVCXzP0<oYW4&eyHQ$SMmcTO#_kc_!GVOG^Gs@l+V_k
z@LE|G5xDzmHy}k24L*-<EHLw)Q&N{Vt*}Ww%(pmjHYi8vl<w?em04F7WYV#Xk?G@H
z9z~|!L~o+tgdtx*@=2%I=@gv1kT&EBNT)q6*Yc8ul%Z6w+3EUt_XD-@277*s=vleT
z&?n1JDC-EVAgx`EA3~x@GUjI|CE#dn)G8`-BqmNMtu7Nz;5Yz=65Yrf*_pL@#pp0h
zbD<<{HPm!IKw+LVruG<4j6^-~+avCinHY6cyP|T6><<81!e-;SGF(Tq8PMXG)$h{q
zP225~+ZD5uU>OT-13v6k#t)-GrjEsmbnnD&;K7W0>qNj{1|`5!n7Puz+OcCxb9k&n
z)$bx%8hOz4?>752Tdp{+L3=bL{IMFwgG!A4x+yTesCMFJSkA4Ty1%Une7iES`R59O
zaziGp_Zd~<vr0av<fj_i>Mhn9s~+JcD&alMM&;e-De?kosv%C_LW%z`oc1F%N`(6e
zm<a6@vsxM#t7%2KV@o0*Tx{<M5f<1gLl-O^Hk{WvGK9%I)@wfP<QGR`US7^&eKKsi
znufZHF^fUa-7U@C%HT4TLwNjU8XV_5EkbOX*+LSvIC}_l`^&gpHB{~qDpL~-UEb#d
zp$>0;tU)|AfT^rm%zJWUG4H(>gV$?DDyNxWeXJaXY>5k%y|dYTcJlcC>_U!z2M8YP
z-1{ht-Fz;)Kwi8SKfa&815OhBf?AM?gTogL%_f0ZcQ9Jab9Ai8;%w_o)<w?ffQ|6)
zT<&vBmk^jYgy2`y^I9NS<|v?7Jy7cqQRuC+4v{{oFmqaaD`&=X8gFvo<axSFdl+tz
z2j5j6_M)<&55j>g7?J)k+4BwkRhasm%)4wNUzhow)u(C$ejW5nQ^O2FBnI?y=c#@K
z_&VM2%_2TwANKzl)!{6~+sZrdJ{4(#oQ;VWCC&YaYk#S);rK0ZbXfP9Uhhd7%VjsH
zsWZGU>RAuEl63!C51!of3sWmAZh55f!L%ZurSzY)Ll{8`crVx6G~=b(W$t`i*E<E*
zFj&`WI%EYaEv^tM*=f)9ja+<6RC!CqSx>m-Nrk0&5Gy#5UtKDBR66GUMTthd+MpG1
zvZz)0Q)}fSXNa(0n`o(z9R-VsY6O>skjb+IRxQ|KMZ)1N&&|(BwDOX=x@W1dZ0CC^
zm1Mx98H~9_aO^h8MxIc_1i7OFx&l-5+7kDNyW;7^)+VC4VX0I#QU~q1kXNHza&*+c
z1bg_@%vyQXj{xSpB8Wp#+ue9mpH&l-t#{yUSG~rR?9e^8NA4NYaHh=rz)l@^0VJl-
zxdJxc<;rrgEv8QR-UB>uTzC7I8edLZ*OJr*ZVNv4?e^r}ejR#LCl$3Nj+E)AO*>!{
zG1so@fM{Dn#ahq%-6MQ}%VVWklFnF9k27j(Arr1=83~;EAk{I+2%$VJY~nYr_AAGh
i7zDaTI%D^{-^}}hE%$%urp#5b`!h-Y1~LamuKOPnYIYU?

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_8.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_30_8.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9443706c604c60225364a3af2a10f889410574b3
GIT binary patch
literal 59715
zcmeFa37lltUEf=KRaaN<Jv}`=Tg^T_noXl!$XZ9^Mbbz}W6P8+m%95__jFHpb<eHp
zQBQLlj7GKvSh6tWK|u2B0Rqnn0Uk@9<FJ@K9!Yq9eDZ<!$miWp7PAB+LWnVt@CcCi
z{r%6aTh)tX$B<8w&nG=|Pn~=2x#ynszn}kkxTmL>;P1<if24Nx#bn~Qbkq9F5_yoT
z@%dyT;Uy{wFX^S`la-|XrWVqbbke0|Dp^a*Rr36%=L-v6l`eahnJ+GMSGpHUm6F|O
z=X(}<E4>SSl|H-A&G#=1R0g7Q1}lTSpPwIE7^)0eT48>8VMk?$#k=N57Is#4TD&+v
zy0ELV%i`VhV+*@0yBGFU_AKnJ>|NMb*|)I2vVWmmDK8wT99TG5IcWJx^M@7=R}Nde
zXa2~-(aO<<W0hle-#dSNVZ1WFaH4X;?)&C%SvXlaxp1m-YT<O{w59dWpIJCtIcxEO
z`CAvxRnA#_aQ?Q1+bg#dAM%F19p1?6>B=46NpIBK^-{8Or#I&9ekoD8>zxU2kGJ=w
zgtxbrpS!zu?nb6^4{7_n{iN-$-E$*dxtII0cYyl?%0tT6k#f*GM9QJsz1-i&{bBD2
z_ebph9o!%Fj&Xmi_70xk&;4<4ocnRRe}MZF-YwkUV)qYvsb>=7rzVyd(&7uV^Imyj
zcBQ%O*UF32#mmj+Qsdq;XJ%%bmzO6`Pc1H-nO~lo_56w1dgF{&t1r$rX6rMjyxHmL
z<;Lt{{nXO+=Au7!`OM_};^djT?mBztq&Gcv=ky)t?wY>iZtu3+rtiLe@~+7{&Yi2>
zd28*?+HJSqdHdvT?UJf-N!3<K@mID!JH2#$yDBa&)#?+oXJ!|c{KacEuU$J=tJ*!c
z-+uQ!=kC63a`NuG?wpvq_4d2&^6v0FYJ1P*<fM1byXT&>=ca0L0qqZ;nVDE_G-fC2
z?WDJ;!y}6eOYfU)oM}!hsDqmRM7=S+=r7d#M)B0CGQU%^#W%IEE{-R|!s8PQlitMR
zPmia<F26S6lYOQZ=KR`HZK6qH@6=+wSzBo?Ow=c4YJS*VU!I?j(s)0=G~qXAC+3M}
znpeFAe!CVXuGFeC{;bDMFXdgSuGaj;>SBHVIw{4+XB*8&7VFcqGvkG@cYblE+FWT?
zCmPk+g_$r{Tj6md%zL%ziRJlbn4N7bOe}?YhI@euP<o9{n4X_#hNY>+<$AM94JH;E
z;Z7z-bD~aVnzib*zqnAPmikCC%soEad~{(#O{~)s6Z6$8^WmQJvlBCm^@;g+Mm>9;
z5Bsx|%S~nzSqI<E_%~;p*DuWajpj2=eZ1$P<>q3oKDFr8{Kp<2Plton*2!9PqS~0c
zT=SOaYvB;R<xfmCtF!gyVzpVTHx~V{s0mxE&d*+{QK9bh=imLrGZPC-^EIE_{&%Zs
zk1x*9!B5Pt%+^WkYb;Gn)vC?eg<7LoTWW-bDK%({c6Sr=7aNU<X0sk9uTza&^rWku
zo}Ta)E={F0&iv_@WtMn^tC1lH60E)C>g1ExYW@rUY_nEgpi%RusrhWZ`~s`4+`K&7
zC{Hifr_>?k=3-fGEjP44E!UaawTXFqtyt#ITdXz8^~Gj+>awO#d7^%OVbQOh{&a%Q
z%rt8A(_u#6@Q?6$3;wKr=JMh~?F=u~CKng4obe`_6K7`p#pR_l&n(v;XZq5o^)u@3
zGv`_855N2AGxOfU8LKK#ptbhTc;3=N^=iX9v^qUouQiwJHLqslbo(9E+s;<+X0?R9
z-fUxHa=zwOtpUf>8P-E-f1LP~I!1r`6~?jN+CwBr1j*4vGclJ8QeHAh&n90@4wJT1
zy_mWf=Dgaq*{K>0%^DC2GfaDLJQJq;Te;4K*~aCGrJ8?&vfAzUk8>LiEZ1kXK3N}C
zSIjArUA>b2S>jV%@!!5E)_9HatR=vJi3)Y{QeN82%p|<5m-F&o!RzvhUN^Cl*W>ki
z8Ly9j{rnDaW%<{~WwGgfA4^Opy+Lp2rLIcadrEi%T#@m<-W&0D63f<d-qQ|)+(=gP
zwZeMW%|xwugKtfz45K7T?{*laQgT?N(&Mm4rPtw%N}t0Om41gKDgzETR0eB9-ig<f
z-YwqAmok-MSGxBMu=DgwDtAYea>+a6owbyaDCNDt%iAnvr}sYZcJB`AI_ll&-35l*
z<vr`&?cGDlnD>6~UhnIO?e^~T-odjy-u>PK{O<K0^d91OpZBo$2*3NIa?g8@zLen0
zW$%LbPSOu}@ABTw@4@K#H+YYGPuO=4c^ADWd2;w;Oqt3N=0YXRoIn566L0=*nvu=T
zGZQELFg1Vf;)T@$tL4l*%k~m5P3!$lt;jdYn0V78_~NNY_WYaQymRb<FMi|jkG}Ia
zR=?|kFaG(b|L|9znEdbqGp*5&B42tSOal_8%2B%JQGrWkHC|;#HxoA!>&csZAlT_8
zujCqI&6JnAk$f>3><kjsbTbnq=CX6S?@M@TVrhOeZ)9#Hp&OX%)qK!PSO|IuyMiKN
zG3X}j4oZZjU>9Levo}ZvBR5jS`<nffq8XKYCB-+pNE;w!!2hY{pq|`Fk}}lHc=_gV
za|bCy>mxVm;~Po7TzDg`F9)Mu;)>p+hI2dpUk*lzj|RhpyMiIYv4A(LyMu!A`M(zw
zxZ4x#A>146doAf_n)`z6jpS>|My8pfrCH@8{XFIE53)fn$a`I{Kqp#mr4EqytnyN`
zqUz9D+h9NMUgk-+)z0Pqkd>_Zm#udba|eRKARVNFzMwxC@Jg@b=)<11a`RyG5UJ&v
z<PG|L?LhPJ+>z!{+I~Q-=I)r+`-=MRxYxIKa3-;Ki0kmm!_9H(b>c?yeB!y=KbTrO
zvUW5$5*)aaqb|P~99qA{>kp0w2X`k%6F1doN7oJphu3cjj_B^S<keq%Ho*!Xd?^te
zI#2(u9ip_8&BN<Qf<rfx&EeqCjifhZW5f)oo(jeYPdCrZoeg%A^H$)%@GGgcW5iFt
zmhwN?JU4e6@!J{M9ah`h>&$@lJG_zL*iDV}-q%t~e)G;7nl*Qk^Df@{Hd=T0`aL(5
zcCRN0x%1+9+P{y<8KzdwhRJKIXQ!5ZAM84Ry)3E#q_VWM2pv(MDIa8F9|U=L<%26{
zqsX;`t3CIYpI)xll>_{=+J9=vuX(c$H3=PD0%hjGWs{4I+G_t}wc3)hKKjH%4XDX-
zm^=+^?w*|i{nx5jYS$ZKQ6H`vCRwdMU2D{u_W9bvgizg*u$(<CzfhZ<x!i1=E?*Kp
z<o|M`HoZJwegV`}_Cd~zzCI<2gKC#4;FPrq)HQWwX%Qme^y>NYE$zSZvkyIe@!c2S
zNsAtR=Fv-+h?XzD_lYW%Ex-4v^ABBm^qKOdC(D&bpMG+~YvZ}FduH}ptzH$q(Fl7s
zk{w-bQlaDkPZfH|Z-&L_u2%OTA{HT=Yt_b^LNfGO)@xp0s>N{SO%VxG3iK!3C|vjq
zS7VG|hB19^a4i`m=M28Ar*0;`BXu?NjKgYBA{pWEFg>-j9A@jw3rpAO_tfdLVMdkR
zg!=xhc;7rU7qtD-{ok%n=zF=u>H&e!(^crF*=DtR`o8(asfqc<{X|;1$5Z|z<n<pV
z-~+|!+4=b@@O8Ncr7^oI6066rU0t4Ke4s=pJYezS%(V%B{Odr)%V6TL&?-OdeW=l>
zG5HqjkNQ5OTE7vLwPwwqU7b*Jm|bFsn_*YAIyFDhXjH3uyFRf{Bi0Y;<*BD<YHl`#
zJyY|uOQ3S2_`I-N6mE^FdA-^U`=&)qM%kH%VbM@QRh4FNpq%Qp`9?U(9IGxdRQf#R
zs*#+`vPRf1I>;EFXZ?vKH6&}()E}fP!o1#N2KvLg>#`@h57YCtdYDoFh55<l>1mjO
zut%9&R4dJnqY>d8c3)E+)DQH^Fw=Zy)+#eUxmb0pBJ7)|>LO!X3HgOZMveU4Mj11J
z)3Z&#N7d=I@pP2=m8E7lz~I!HRc1+bd5LLLYba@kI@bKSUSXO%zKEVMRbL9TVo<J!
zIY{be%>$9vo41_{`{JoDnwvi9s!?jqr_8W-Vt(l|%*7Ow(O+aDsMlBtJGXqQ^{#Mq
zRjVQ_&4a${RX26~TU3b*OMgD>QKjRLQnPy4^_18U%^Lr`%Gei|@A{jS0k{fF%Tw?>
zBC;7y*en3u;zHQ{wEe<lP~(EKEcx}B@#FqcimG%&%r>Jzu_^DLC)xjIeWnW-BNnK6
z*?&lR;io1%W3{S!%U@E?X$8+Km{2gOpsL_Q3N9#kr-FAWu!U7uY(c>yft%zu8L744
zr~eIlHm_hx!G{Sdn(3@DTetpIC4Y=SY^BCo({M+7$hmC|yT7WpZ92w%S{Y=e!4+!o
zT4Pt~s@Gm{z^_tj<*bT3OwLw%He|Ii7*F~iP%Vt%5RK$NOK=-kBOpi|C?@-px#SpE
z=f6_2*X~QneW{{iDSMhKxLk?eq<gQFEF?>$DJ+t+l*&-vSTbY(hRHQZte?_GQz~PS
zoWrS7s+Zr@o((W=DzEzR*UKd-m<FNoGe*e;&<j4a)>^(k*o6Op+Q^8S@sh7(*YaK}
z$d4x03eD798p;SB0Q}w6Os!{l(&a&;e<ZalaWl164Dxf?4<^^r&0GMz1#Zq$j;OpW
zVOP*Km-W)SfbVnpE2+kZn?)}tda>EvEY0<B^?G?qh6-Ho^9oRd`5Q^20CeBQeOJf5
zs72_h_<48GO^y14uA8(aPA$1q)l<|=$CDm=GO*2)-u64YH@?%?`NZpQmk7m_m>V2T
z1o`0vB^E~$_RBYStF?nl-<enwUDM|cQ1enpO$J>}sQpk-@`i4vHq>)i{a`IC-h>}?
z_1z8K6}NFm(CxlyxeD9mnj7{;uB4!wpmgVUcsqv@o+#iE_{;8~a6{#)1v^#m-HElH
zy3gIPH)`*k_jak|W@&v?y=&!<weHkH@;_|(cU%5%;u+NYSN_h;BySd)W8NO)8hX5b
z=0=Z~zE6GD=j|nqv{2MEC5m?3b2GWV_hy22e5P4^5v~-vuPnN6BY24^4PQhCcsetI
zz~{~X#NX;I4pf&fPh3O9uvng~m5~BLu6Sh_<Z^9d3ZX<K4`FS;hXTT^(L!_y@u7{N
z9vE7?77jJ8`VkyyBZ0k^xCkMHjBL5#z)bt`9--{Fe7ehC_G?pM`Hf!;cefuc+`8$Z
z|Cs7H7-LlE6(4!tR4v@sPSb~5=@1*U;3{wC(8HfB-F*EYeO)-*&S+Awsm0p#t=p!a
zGjjV)5mf#=2%>KPLd;4&=-*HBYX8Er$!sD59F$s`zcxP^Ve`fX2B-*;(GXIuH^=uG
zCii!#*lPq~K0>Shnv%{aDGdttU({`T<lnDb!8nIRbCFOiKMl45f%_lPiy3tul*;0a
zVV*F*z|>e+UU0bPlY0AaDzLH6U2<}@E{3ULVzqGS;$rjNqD2;{YRz-R9OA?JMBV?C
zK9HW6pC9ivoa(Tgp;6J09aS|PsE<J@80BE!^S@Im1<CUyWOfx%p_@orze(EnaW$SM
zNC;<u84CRC1@;%f3Bo0#ss2<yn4v#C0>03{ed%GI_ow!ebBt?WO4y|gW;x7NN*?6h
z5^t=I7~Y6!ix$oHKM=?U!2CZD$X^+OG<V)a<oVw|0`XsCo%!EE@HOF&Zo?lD=|8*?
zoV(P@u1%>zE3HQc?>9X+7Nl~>9=#P3iQYf>c8Dgb)c!V3*@RJ|3<ug7wjdps#i5?j
zHui{7jANG$x1Vf>fLtMMG&I)E-bO=FWwv{EE~<qjK>iyF{_{}MU?)m)Wr3Ld|Cy4u
zU?hKC<;U2_zsYT7Xal?EN`S#I>8ulyL}~qPQAt$ijhM^-ewF?M3WT!!A5`$S6d14c
zLy8$^^rMRXgn}Pa@WTrJ7X_bI@V6EGJq3S<;HyVT{*UM#;XlJi{@+#1aN>5T=$m-@
zYh3@OP|=?4QIYGo4Vou$G-4Bt?SfWJ`1cVe*O7nDrPnhzoh(x+@FH_*lWamkBIUgg
zZbdTU<Qqs)lZn-$mv|piZKRx{Lm@(}ocxAz+?&^;H&gD-_wxo6Y5L;ou6xVRpx=_V
zPDG}pZjp#@bvGh86X@wt#G|(!ju#voElLU*5@bNluPI)P#YK%uerj<EG0&o3(UUyf
zuJR$RI%@Zbilj>2L^B|*fERA?YONchxG_TBgz<~w6nm2?xJ$=9uJ&4&$JwVOoYUV9
zt^${_jmOhc0HCTe;|Q3xjF)bvg3NkaBKIpf|7bwUuEZK*@>c3vcIECMi_rP(T5jb;
zkRo<)Eg$68GeQ1l(to&#vS>XQ<b-7r`(}dNI}%qPel~IS?q`i<gzR5$7Nh`pDbeg&
zM`f_yEdlU)iNU2b3BA_}uccOU&7Q5#y*#XI;g$4xMA*;$CSva{*wwBeb0zP8{_1D=
zTGy_`TrbRLk+9T+{X~Rb5TgdG`lV(cY^vRPT~{&)%J;yk!TO5vh5d?Rus_Mh9_2-e
z(an3G35x6eH(`)m`D?vV8Aavvn=2=q13?dLZ}%>0Pk)T9^`RDl%~Z-rLhMk`cb|mr
zPd`hDhO||OEWPnvEAQL%{`s}OmHVQ!VpQgdwf@dBQ6Y4DJR3<&tL~H^^iQYk8$YCP
zT>a#;jO;p0sdB-h4Xm8qRN_8L9B7x5rJi2#E;K_{%AeZ%VrAX?gK;ZcH5%LaKD8JO
z`gu2HBRz;ffqx%KseWsND~D<G;9PG!8e;#xEf`!M_Da-lkhd-~KYD_}z3OM{>H88_
zKSNsjwN&G$ShhQuizp~xN!3v~&5gA0nFT{@!_2!r0|5PeX~@(?1H^_cHuwfY2pCX%
z;&n6jX}`c5L&4DZCET)id822Cc{Usjx1Je5*m2Q+f`tR0Q$mo0s1F6ZM7~8Ny4Vp&
zX17{CUz@L)b^ys&`5V`qtiPS%bSyAu<+TOolEIq5yhK})YpEdFAu2Z*0C)T(nZk@@
z$zig(2{8Cy&%>hv3F6}rj32Y*zfp|3op^lp;DuSI&RE2%CW3APcF50Kw}{O%Uac`D
zjdgQk-r;4jsK%HZK_%`lCz}ZRpf?7Lrs!DBIU9l19*lOdOF;o5mCb5@8L8vSQf&$b
ztSmjtYTpyl^T)Wmw^&{sD!#X!B}yFM8A(?qz4T|TPxQ78gWqh(9>KxXoK)o#O($Nw
ztc+b4FZj*HMa1S=1n9H%uwUKjIlB-EWD#|{lt}XJ3sf4j>I=Q|+B|exm_-;ng91Su
zWtf%8N+axAt~ahO*J?=T`%JRh9(9BM<2xK^&%_c~|6vuGYIy!HDw4bsre?2%g?ePB
z;{Tj|(Qrnki{Yu7WdMdLh&DT-mLaqZ7WF@2rOj1}^(u0EM7;A}9OqrQW!KMPu{{Et
z=aWwvW13HML=p%h`#~@WFYRA1s78#c6K;xWO(h3mJ3&9Nti#FP)IRPaRyMg$3@!i8
zQp#|2HwIHXqB0S3j`933m;S8|Z8LADv{v+&;}RY-+@;QI3ZqESnVd`ANCybO5L!Y4
z2+1SHi_*bf%9n}nvraHNod(UPCic-xGfOG+_s&`qYZjzQh%s6gKIN;&SE|i&P+qB?
z9X}msz1Mo1L<ZV#-fORY;9_~D{0P_S)2E*+JK2S_il*S1uQkvL)|Zx15pFIedLQa@
zvgYfJ*kYw@Ha+FzCVnd4I)1wRF35?K<+IplVXuW&6iXY-emaWuF>3Ma_m&?jH<uA|
z@Gdz|mRqF!$#N&@K7O){utT4WT5_yWw%OG#FcJxU>4EXAFWsL1T?)=9Fbea3A~p`O
zz)zTuN(qxqjClBNh&6+gUR6q+-`zCh!oI6jV-Z@pzd^H4%P*1F>_MP0Y;$8%wwSQd
z#K>~l<Ty@b>*TN$Y2e*-Ne`nNX9^@Qj%QrEzMI!vyS|4Qsn(=4qwoJq63|o1G;4Fs
zM<WGu#2=X2$}-l}1S7msgUQvsR{gii``6lzACKGdzqZ@q|E5ya4*$0l9Ek(<7NSL`
zCg0KuN2Ooc(3V4Q+m?u~x0X}%U&qL_XWZYQ#AswXT6BNh7;BcbXl$DntyV!k)oWMd
z)?^r&{cqKpDEC)AHq59^a}1DVmS62T_O-Mmo?%!#ID^HgX4m6(Y@K17^Tw@^bTwWn
z5<W&Ng_Y3)qfskbDJj!?w7_VTx|X$E0;cxfS`P9UZIgjj+M9E0N+M3Tre0Ufc+l@9
zHlDZD<^Kz%##1xw-6%0==ifYEyRWuY&flzV5bKzr8%9iEy4oK>i@3c1O6~tCwa;MX
zPb+4e{!bFyzWx61DpTA^n;SjQ(P+2IUf!nZBb%G<vLXoi5T(cc(8WzW52m?+3Byt`
zNTCBn-!}(ScqQc@CkDg>YN986CB2q4@*8`3p4}N_));^PLXcfYpXg<sHVagn)f?D&
zr9YTn%dO?t3bLsR3N|<L7fs~=uiFMj9b%p@{d^nWq&H0u5Kv7WWRjELCWbxhJhq%w
z|4nYjyBxqyuDFub+Lmd<;xhIOD^+wcBH5j}gYb>@bvP2w5~=WGy~^h9?_1i4qrO9(
zKgd@SMjH)6|45}<N{zIz)utDw)Zh9mam9TvBq}=@uH<y`dtfR*lw8ZI-{QVcqr+8S
z7D(-CLh6tkt?Z5TS`n=w7M)kn<K6_H&1Hiuwpkg;H0EGOy_B?ppGv)&fGLI@_Oe0O
zaAG}wlhmLXL8qLwZ@*71+WlsNz)16E@n}%K@Brcso6aYc{|^=HCJ3{pp)pB@8L(bf
zVj2p6yyO=8#`43Gvm37(rVNvl{?Dr@EA-##)j>B#8-_!8D=gI)>t-dsFwwZOd4vi}
zRcG7#Mtfb1Kx+1;PP8=XtsF5~S)f#{1*05(kO;lb>|djoy`&R<pfSgmm-(M!zGr8W
zh(7AgOf%a=w@&W7EceVlGLKdr;ly(nZS2sPqish#F`Z<z5UtD=z3eMmTise4l%)|a
zjX<&@X85Sx<vq0YtgVL;sq_R{Nw_3F2@snYbY&+*kcDa~Tx^Yw`fRmmHsa#0Fva&j
zN0R>w1jMaN-=nx&|2CRP5<jUt_R!2##M5|nV|c;>a%-5On&*FBnO;=z|0(zu0*r$e
z>od+E#7sjZLS<5jR<fB94m$$WK^^!j+1WSTr~bVPNXvY`#{|mp(@*H_BXM8|yr(r(
zEXu{DAJT}9l0&*ONxO2%gJ{5t$vwm~q^u4)JRZ#ed+E)7^2Z~s6&J6%gWPDO8E1{Q
zMn!Ng13pVn!z{4kwCg0r-Ut6L$jcI$;K$j-$RHPBWb|QiasnBl8D?3=j%=G6lyh_Q
zr_~b8I?3*GVtcUi$BxZCjp1{PZc3A}aW4V*Hf<Kztm>BvH&QS=ntiPQmtlBrq+e!D
zTMD&zDFGCT(I4rHQ~F-(Pel%KHHrktiIOiPBXXp0%JJ7ln+~QfwPB$X3{zW+Y;+3u
z0eRUFa{tfLoIg^@5g~Oo8h}oDgUd;e$!uja6VXZ1*+<%VL7!`V+I_WE`p>8nbNVdi
zG7xdjOs4g8ODEdmRtZM5u$Tdl;u#@rEQLXjQXttju?5?cYzEJMixM5s+%(kL84Q`O
z{emi~aTH;)TC`EN2XXzfJkw?Mb4Ym3z=3~QjLN7PPck{v{v;Auj40PK_4_q#y%a#A
zLi+h;dM-1UMdnI=q3)}bE@jSA%+M#zTZLxc!{`~QCPtX!xr-Q^-|FAarx7Z_L?Xe#
z{K{K4L)q2+%k?Yu#TV+ZiZHD*aVsDD#@O6#?HKF*oAqjXVzS}?D`GZl{2NLuwswL{
z%!h>qZDpYhQ)jeROh%M>n1gSoGwgK%6*6t=#De-t5B`z*ahSprtoIVDU4n=VLWhx5
zAr%dMD{tJjC9<n)(_D=*C2U<kTKO+aP>}Z0z++$lng&)6-I$$CK+&?Uyo^EPG*(s8
zMWnqf_fNl^;QplD=eVEe%?u?mp>8VYAOb|wHMFeO>GlKz0>syTi&xTfdG=he<6<U_
zNE3lKPYeDJd42Vi*T2>U?8nZEcL&}`1qEU2SuA!Cr(2x7U~UB8;E_P$--5&FVEER1
zgS^)ry`y{xvW<p^4vR`2Qpv4Sq1pxOhW|66grA{Tw+9b?mPh|gk1=y^K?X@pv7l{i
z0~sPpuf6i`W7JIh=gdzAo@yCQMy9sq)j_jIxsJppwn>q)&t{{EC*M#JcM&kR)86a?
z0!*oD!Zh~3t0miE5!s}!cDXwYT2`|b9q)0VDz7HgRFKh}1>JsHw}Xo5TiANUMw-q#
z)hr>+ki?gDWQzdW2--xAiuSqg|63K(rJgY%vfr<g2M8)fQ9rDe>Dd*TOp?Kw&{hg|
zLv1&7z5|f0PX2XDXedblvfpeDC7yFve`*jRU2kd>%4iUKlHSyCYaO-TkM(j|Ljtr<
zay9NGkjlvD9smyO0U8PV1_2xNQD!|W5XBwzCv*?E=~Jw(SA|^C8z>g|AEP_+LNu`<
zP@A&+b3Wut1Q`JhqffB`^D%ACU3o#j+XCJ^Vqczmf1>6a=uIC?u9gh+Mvr28K#v6-
z4EWwpL|h9LA40?+J%=L7IG}x^{4CYrgzzp;*%^Z!WagFRdX5&8LJPAO-E>QxzPLI#
zS-xC;p*&SSb$@x1A8we2JqtCqx?Qh2*))A{S<SRot0G|vEueMFfHZ)!Gs}zEoJE4{
zFO#7qlgb*+;Y(@j2=`5pvrEv>WR3q({UA`7FrrCPTTqcx55}1`jJ0~jdc;vk>*^Io
z9GHj}Hh|JaNg-H?b*VuOc>R=rh|l=X^JZ(f4l2QLq^7dZrspm6T{m(@5CIaW+EO}2
zH61$d1;R({dq;*CtatN~LaU0uO=P7>^%4->b2jnNIikrD;Vu=iyVLNYlSENm@Nf5E
zNztz*(P=V8rC&3x-Ekw?!KGqkgDmpn?2K606zr`{W&mv-ozw+%nFR>)1S)9gh?^<8
zqX5X(l<#62o%>gK1pr7I;GVD$r54d$0uM>)^1na{#hYRU&=PtjYA8UR_j-cjrrZUl
zR`10yx#aJl7sJRhLLd%scZ8%%>>n&(vp!RuWD}V-Flhg6o^68a+OR;RwzN2PnOz}l
zd1jY@k}nwl02Sa@a`S!_L=K?(7==2b-zMg_aSMxmd66xPK6}ssb^=M#bva{(TQsa!
zl&iS7gvBMB6Hua*Of=4R3if9<#=9MBpa~InIa+zvTM2vO7pmO4sjS9WBTb26K+w&c
zlX|dej`y_St~IBi38Wb-UfgLYGBAFlW-N#90^_<~Q;v_MB1U1u>oH8w927UG<=FT<
z0mH-!@R<M6ym$A|2zJjTTW9~=C*GrB?acYDt@|kFw<e8thn{$;@G{f&M)F$|AA{dV
z6|R(|r6gsq@b!tO6YuA3DEo}<(X@S<(S%B|TTS;H(q2w%la}0=W=hO12ZN0VjCGOp
zoDoXK5lFy4%g42O2ZDxmMB`y1zSB1E0Kzi1ZA4?ivKLMW(<tD0fZ8iUVaicG>&8nM
zi0wO?GPZr^n8rz8L`!+xpfcEamQ5Y#Qf(kRh=cg0udS*+-VGx1#MRWlgM_U$y|rCa
z|1NSIZ&&dGOUpl|VvZ{qSAeetF@stT{y(50K8_Yz`VZg0-mQA{^E|?IDDIQ`qU)Az
z(p&}T&O7zu-3rugVc({%DXU`6=tD-K(!ccs&G(poZMY?K+VoP~cbd_<43Dc6W*Bsg
z=C9_O=u4X;!1Hw5g2gE>({2s#QPuV;(D=8zEgWR<Y&`*i#0;}<=Cmqn2pFnkb*ByW
z)?6{>QO+3ioa6=^EK`sa94)Ia;&#Er76TJY>V-VUVcyKe)scD&&pMP_cSv`2bg~>p
zm42MNarZ=Xx$ZB%Kx+-ysc(EO{T7nVJBGXw_SWTbpi#xXxylP+DcZ7St_wDI#ZXmf
zm28oyUR$ZosV~H93ki#N0mF@2k8G3q&4MZ-;&sXpB9T5W)l9grQK$d4ZPx${onwQS
zTFWCA%!9qOV%15hj{CF`EqODV!ZHVwsFZA8z<7B;N9w*rGk=4vAuuNpKK7(*Wd(}}
z@RZ@kIdGG(Q!x@%y=r$|9$HqIpSrewAgn6Rz@&H(1_-$jOky=749xe|dm)jE<muDa
z18ohNV~Yo6I5AiWwq_u}4g^tG-7xE5zy=6v2EGksr*oP~I^>b|iBZFz!wkOID1qRf
zwcZY(*~44i+IoQPORtqt;x0=WbIl#=BTd@7zEjpsh}KfBw?yRe{QWPZ3^WjpmB~wp
z?gWsZNho!^&w<S_=hU|bK38+M+C&2fORB{z+C<>@#5!EfhDzb7`9(9exWqzFX$}1a
zule^83%gqrij47kWzN`&FtDukYGApy4VE29ZqL669!F5lA!u8t0wxC8KB#%9asj?h
zhbMGy8w)#`vKa1;kNO|gyN6WW=ZS@B^w5r86-Vuadb7$RROuvzRmHTFolM~S6*Fqy
zxJ6^A)|B>u3NQegR_q#SY`oR3yarYww>q<k@xBh~A$buMtUA-z8)0`OXO#VigGxqk
zKCIF<j)l*(ab4{{qL^W?r0XGrw8m+<ez-?{B)N#Bb25DB2fP;2IY@|pK&{Lc1bBPl
znR3|P??D<V!lInaNXFA=tGeB*Q+&?;Ho!MY^H>ClwWO2aIfFaV*K;VHkw+lMpTTDY
zi#RuzpDReTBY9*5?Mx;ql4!gu?6IAP+H9vyjAjwz_|?zw4gbRm-mBmif-q||SD1as
zn^<_ZrA1XiVa`alMwli?dqz1lGz>3-dlWW0S7Vy3QB^mUC^TUrOuC0oS>0vL-!4BA
zC;v88<eY+g6sQ;cuOoQdXJRid!PsA5dIKR6L=nldi(6{@2T3qcdFkT#QU8OK6&9=7
zU5k4aw%nGg)vL=B^KJvK?dh{^7ydWt%UK0B|EzEFAZTp>c3tS};cqDTZ3X{Y!S5*e
zqJlRS{2l?jm9ZysY7mo<CsZMu&o<FbUSOgEv9i8wP26HK>ftus-lO1N1-2mGt=N+U
zVNbM)c&f&h<qnH40L};WeOuMH64TXqiJ*{4VTALw{p)`>vbDQ(rIMvytVr)KlnOci
z6}Y+zUA-lSxWL7;Qm&XS<cfX8{q{Ut813&ZWD0qn9Pw3~DX#cW$V-2XU!qlQKO}e%
zsTjD$2MPW}G6Lv$$iRn4Ex%s4=|C|uG9=y50hKQA1D4x)A1Rve5jK*V=aEFEWL`(u
z8tQ$}+wJYahexls*V{*Vecpbr%x}MUz&ptAfOp6{%<rIg#5>CGkoTPV7fIq##J=l&
z$i5^0Qy@RQin!D&zOha@uOi;_-f8a)-`E*FJL}#0637CdBJwjLpCa-xlE9<Lgn1c}
zM-llLkw+1E7?DR2=V4@@H)%dbl9m156lO;czMk|R@*aLETfvu>dcpHNc^Od;9E?(G
z=50hNhoY2e^E;xH!%@nN_n7y%y>%o?dBVGBbvzo~U-q8#p0XUrqLf+hlJ{OqIgXc<
zXQ}0Q^yK~CH`u!;%zMc#AjG+_?{UCQy%}#ZUo9@kACPT(zv$n=cwe}rwZh+*9+>KJ
zF;hB-sL`Bge3JQKenO<RfN(JnA;>M+Wh$LaD(IEm<w~+~qM1Q{ig_;V`~5-Qzt>A6
zC(Xcgr%W3AlwfJ&DK9}v@4EOS?PJ9VmpHSFIL5QYyDUCn@#2i~3^sE&V7l3($iCFs
zWYGKFsaGXuMRv=6SLDC!cZG<l7O#&syODOW!&OGC9)fvN`<_kIS={>Pu`9IN_YH}9
z3eK&&@dfjR(q9*KJ>c~;2iX<M&QGTwUmr$LudgK5cOWlzu@OWGipf{WPWH{>(L|Q(
z)m;`J3;H4-C3C#B8*|@5mF3C%%2?1Z@`gUxgW>Y~)K|^D9-dS%c-=?pe%r%a4)9^(
zzsi$-J&AI^WIdahE3Y4rmlC7@2QC_Y<MxjqWT5;*1mQ4l5E^V12e^AJZv$2T+x7ev
zf(!nuiaw;^;|dsDl(h{UGT>wjCs&L9rT<>sXB-T7B6Gv1mOPkgJH2Z#NY;eH;o!zy
zBFfNe>Ev~oSkGCDrk56%{P(FZw3{4fT;`O^d_RQ3sV!#^#yH8_l(|rAr9#y>zT^pd
zvQqFo-T(`;bKr7y0e@%gamTfQhmxk@zn?5_pNUYJ?LFD#<*e;YA{^;-f7Y%>INF&W
zxx_MoeK^?8C1Iq*tET_0^iEDLVjZ`5rB;{ZCmhl~2CN|w>c!q0d_$$vs#isU)Hsep
zkx7ByAQ&GqBCj&o>GL9Lt^}d2E{T#L)*5VrP;dAZwY;j}LkgZ#Aezg~LF<<f=vL;0
zew%%br)_)Iw$N&uV%Df#5t(y|NJ4lIQb&fa5-?%r5&at>RsT{5o0JD#G?bt-m9-~O
zhvN7YFOkcYqf%B!Hv!Sul}SuC^7%xb>k%&7o@4tp0FEYdPYDo_BGUvwW)%~yb<DEX
zb7qWSPg&^sn}8OPm~zL$eMW#~F8`6#eQv>L+=7RLd4{j}Pb+wj0{TpveVbF?qJ$R}
zh?bXo7=~rm!~Pj5pZvwJA<7)l*^of19aXk8IX07%&J&&+*2ojy5384j7ZMl<my@dp
zH}{LV2-)z;ln845sdx0M*!PBZQ_(gyP7wm8;Z7B!CoI<brN<gdM`eNG#l0x&aC89J
z%z5V8O5E^z!TPETB#}Y$S|oQjysK1%r65C0f<H6RFzyyq@w}eNwN(lwTmMY}|4nv9
z3QSY3REe3^euCC_4BTQ4rGD08;O|QhtUd@Ke;F%adC0?nyF9hb-k=3J5CWFA?B@Y8
zr`jo~g@6bX^&0Ae`Rl8@4;?yGe(y7K7xc`f^QUkg^vFYx>yC7T6{~w&UX0{z?{p`S
zmhu1ipsz)Pe;Vg}$w`lW6ZU#MaPUSJ@oq^Qj60+U&nvC0*a>3cp!0x;XP4NQO*ks?
zNJm2KvPQn~niFi9{)k>>$2VRJ@kNuSsV!r<gDP3lH5~NeJxF%;h-4%_7UvfZxX<}*
z%fK+Xx@BGCpk|}6%0SkSXbO%{N+NUI7C9<2maP{0s5qO$o2TS<g&2opF5TY7M_arD
zTQgU(9-F!1IGr;#=AFb_W=wh0X$v%OnO#N}rTw$avrLSxQ_<?pI#ko799x3nj>cvA
zzNs4mi{0e-vvgmpKLu5si<n0hyP)8m3f@JqrQ?iu`dRgpYMn@r!2TlIJFvNDw(dFS
z9sA4bH7<SEplVCV7o;Rar-hPQv0kg0mD}~Zp1p~v{JEn}^p4Uo%}%citT6XWFI96_
zk0NujU&(<?4JW6dQ}Q4g_Q|qOOd#$*fgjfp=;nP%RKkp$QicPaGto@W%)z5SteN&t
z$ma|Aw+t6@6d|lr(}SN=@aqbW#DVc-&-3mtshec%;v{~BluX#MRXDb}vpSJ(M_ppM
z=!Pznj`A(LEV&|T5204nX}xi$sdJ7;os+uQ>H%w`WSQ1e&PrpP4}=J=8dE;XocaV*
z+m>GN*VH%Pq~I;PV3Mc5rY?wto^ov<tH|ST*Zj_!#O>c8H~B%TM;jB3RmnpD>E$N2
zP!T<g)TEley3P~Xrk>}C^w4N&v7^X3&w)l!lSkH(k}Wgs@;2?#ChSaCyH+0LowO8l
zOuOP*8o-^dc3+)?(m!h7OgBE-EWwTFE9u536>pcO8bm-n95i~lV@Rvj)g$k<4N#WI
zt_?havH7C0Se|H<h1^`yY5!SzZ?)@EYhUT=Ze_q@+y>%1a>IH7O9dl+jJOIkD$mhN
zPDCJqh|PBY8}&e=>t?nIhcvT|Qf*0sY+!9}Z#S(cDB_=MOv}6{sgK(OgsURNV-jl@
znBpa1f*ew;9(}u!c9>l4_m5I)jLQXA1QiWF%oD-YVTPTHr~;fp9)nT_^JbjNBFL@_
zT7Zq6ai$oktAp%ZBnQKrYv$H5k`>lwc-hH{6g*EX50+P~3pOu@teSX`2X0^noAf`&
zSJ4a%Mhw7W1Uw*bzZMOH|8+EPyy*6kN$MF6x>??d0ow@E5&WXJv~hT1>WVLD>68+Y
zh2?!3$j^~l+1sij9^~8-HZT|2$C#kAl3Mlrcj%4#6>Nl4;g~5OIz-QoXqUp`4>iui
zd?f+?lYO0bd^wU9vh~E~v3=`;;&rTsXf4~&o+qL~ZNl^8&4~To5T~j7QA<MwPb#a8
z6m9E_))%WdrB!T}RCyWhF=Xu0-n|9sYcrEGSzFz4EH)A{=X1{C3*MxQatclt<5kBO
z)^{>HqjY~*x1T1!Pc?>bM(jWvi=OqbsvAYGZR$&Zp0vMEr#X?R)A8g|!|AOc@2%*c
zxT`3pjU}&fBammpkBd%REKvzNtqoJsS?szd2L|90D<D$FE&;Jd%72`W1-z{#ud+`I
z!!~<bKkUj-nKx36=bHrpm~0dzbSA|BBjAnBmAhisxy)d3Bw-%^cu-}&ff#rIc8-Z}
z^*^Q_`G^AR%VX*)k%wWP#_EL@KaiZ-)Nzg+G#NHbqu@fzd0c@R6B*YRA;E0(A=?Gb
z*mDaSQJ-#C$dsm?{=^{&mD==s9}!u-LQf!)QP1EM8N<DG_$c6vO!zPXvgau8*GX}y
zK$|G1_7t+s0r~?Q5$8$`n-QZujGA{6l<(j>fD{e|5Cvh&I)@pX?H#=~TYJH@B)GuD
zeR{NgE_pdjzTh9=Jx9@|@nX2S#|rrQzAUHdJJdr)iY3d*va}Cd_{mn*xF3`~?uQR^
zqjn0hA<%-#98Jt1kfOz;xPD;f$oQRp2-5g?aqR|{%1Y)H(4*{Z)ED$+W)vsE9qBox
zbI~(F&iV<*aO6;^_Z-wMtp04V9Q{RmDlgw#zHGi!H$QmcUN#eZ&CBid)?P>FtsN+h
zGil(L*0nepQ|Xi^7ky~Klkr>S7x318nf=K1cIJ1mqjI7dWwkBJIF+);WgeZrLl0eP
zQ*s*R$A!6}euJ*Gt<ci$X~4t6m0E2HtHXw4qoZk%>rjflss;0%3eGF|1VPJ=UM{=+
z?;w!1zzb44%e5@)!>{8u@{@W{QahRS(@nf1BIeee%O#mmr3-gp+Lg{e)=gj^iVt3~
z>N;hQ=8(axd>QmT%u0j*-m@9NcfJy9r8Sqdu53-M6Jf3xriN$EHRZI!8PG~2Z88cx
z5ny`-cg+0JoWfyUnpFq)%{^;rnCI9rv+1E_db=y(ziy@Slv49nsu|XL*{D=zl1y%c
zyh`0t4j5i;`}9>-XZ~WP&wQ>j^pWSXR>H(elwjjoDK=+o)kz;e_I_n3enp9HlYA?Y
z*V>7&7JownXDdCUw<?2;rFmSNn%=mzvD?O7xX=*)mLU&Qk6-6naV&I~`yPCMrTg*d
z;}S$}8q^*^z&Ps(jp`_0aK=!j<WUIl$O9!dpO9yJH;?d!_pJ}&aUP@`jkS$)FGV8=
zPkb=rHg+L$M?ER410aYQgN}iyQC-R}2aq#L=Qm34h7yzQ`tM@MWWV4e@`<;@pd*9V
z4jfd@Wu{RLyozf<yUR}Fc3fYT-w@d+AgV7(RKJD*T$b<7>Oo(VRRKBkH?xx9%*o!s
zym8z}uCrANNM67OA%jAoFS_6K?r>sS^RMr<RDH~9D*J;!c6C+kE7aELRBFLoj%q~=
z=l8CCS_yh$2BnJeL~S0}+{!4QTAX+B=@e}q*wSYA9kpAruiSR!VnY%uZEGKFDJ$-a
zB%A&;S{b}{{G}T|#lFhE$d0x8g4l+>&=c1e?#WyAh4Quf!g6lg7s_QbE>P?%_l0ty
z-Wl7jFSeA`>Wg}&*$oNWi;oiK(6sha6jnbHV}q7gfO|m}Ga^LguNkYO#>w*3$@1ls
zSOK^5#>mSub0Wu*`5<?%uwy@Z3mAm2XWL-(fwRw@#9K!D{?_N(<+Ujf>_NLba)kwN
zcHVg_;S?4Rb88$5%fW$H3%l7FiAw^veMwj>vSBVcE02$o;!xHNVqxyR^=bS`gqaJ3
z{`U&lNzmsw7?T+ra1-^65Yb3xN#^|TQ_fEjgsJ7}Fy(PgHt>c#<;R%L|8C{^9)e1@
zz7avoFt;3iQi=ZGR2e-jY;SF<lsdAX+YA`{1lTNtE9YK;#s{fWBHNoDm4ijlUT><L
zIt=nVY+@Rj#b=UxlBZKA2~Td5b|Rfgt?u6rgm!#ss!LztPvR!wNxqzD`~(eX%lW`*
zaG_G|`3I%o43Nr08L{~T><P{U5<#0o*$a*#gT|r?HKLiE%ZduqkVq}4ttlK~(72Cv
zDmCdG_eD5mSzvWvN&7!Sx;#346ub-OTq|Auaa++j|GQwcxQpwAT~z$-Qk^f$;q6?>
z|7?))|Df3ul$yQp4ZmWsKDdmKTJ_6%Ypv(%mtDN|zS1b6ghwB1Bk+@>BH~tZXoL<{
z@IBrd6nBEas&}ogIYc@wqn7=wF<i-_2+jn3vuR2@j*^!WtWMj!D<4T-_I0enSlQhH
z+iTVa@F0}*Kh8IB!b^<?g8|g8UBRFzT@;}Oo_@-u$$pKL94S9a%80W`3kF!@+9Hb&
z(+P~67wZ*VxkFM)K88zC>A=D?Cll%50Mmb0;fiYW5ccvkg8;I+;(l;(@BgS~`HvCc
z$wqvLGY5%9=Kd8u7dPoID5xpR#}$|c%QRij>(*pG#&X#bowa;=V~SX1hiq+^aZALq
zhg`*9%<MF0CGEhZq=Ut>^9?X9d6D;Yo)TF-^N1HI<eW9@uTfw)7TH>L%Cj8gz&z3k
zBBq=qU*74Z8$y2>?*xrK!t06ugM6;S9#eTKsXEbpo3ZBLhbjNt)sVQSDw!!2mewBb
z%nDn$w@Rj^e5h+>OL>(@OIjg0lx)67h)V(IBviT7XiD41hLKbaTdX7lF7ghhGpu^W
z#k}hcZ7~(mRK%`nyR~jaT`ap2K^F@+p67-wyBFZV6BuzwRC%<?jzjLz!$A#Tx+)ny
z8aX>&1rr3UGR-BoJd|M+ej}yBbJj58CPxtm)beg)x!0j@Y`13q;+sZAKE}sdY-ZBH
z_IA5ZX2``KUR+AXpv3&>42+Kd$2@XYdJ-Xqg-DcC3kL;1C0^3GON?;SzivXQu*g_O
zJ5)?WBK3{(wS<YnM(hNLyoyXo5%?_{){TU#&AlJc$eFO55h-C}hr9`IL0i>;TV-ee
zRlDE+`)bf%q5J*Ms(_7s@Bag({fq)34F6{pGe!2#5$o(-|L65+n@$b)Z0)G3JY$)G
z+mFS)&3y0Z?US45_}dk-y|7A*B-7PI3hfQWno%YqiljhpThOL?Ah`GURWnJy5~;0Y
zzg3mEMW~^XXo%<l*ozUhT}yaPz)OV~N~co8I64!F9%NeBwry=OVh5>-&WrGBD`r(1
zre>P1$_ADO)S69|KTO)ss`5hf34jZeclYM*?A*+hRX+Z@xx`wTO};24^`0s9#XiFb
ziKPMIOOpXxBE%0EfZ4Tt=`oltV~O7?X91ui&;JEujiOv;I-tF4IBEpM*aoFJSyy%W
zzXCbVZ{y?g%+5Eujr}gkHVNC)#9R+KaIh#hi>*3|ESc+reJ@gz{va*2H)h~60oNNK
zc2j3c*p3i>Iv-@&5ty?bftZA9$Q<43HfkC<xy6<meDtC+-lE|DQgB8=8*9oY!QZWs
zsVQa%QvyhbCO@tREA~`@S!QMw`?9i(5o>Ky`iDx@I2a1Ts~SgUxS6m=Pxcb9i<FZj
z!063tb7Il<AhK<5vs}|mJ^!Jmge0Pg<k92_s7v7_;S)LO9Ai_&<|z^x*Til@qzj=8
z$9Yc^X~Hj=(7aSXX>lz58+(E@E-BR6@Xdy6PI@T}<m6xi_bju~GYteAGTDrYbOtJ?
z36AkU6l5A7LJPt^I2=nLBZ)Mq{z1Ox?^jS(K*P1iWfQ>>cQ(BtS7R-`@~ArNU-C+r
z5ylUT5waH&&>kHB4^-yHMHF^(-b(ABrGKiPbNYv3DuwVue&q_dJqNR=wl1xXH#Z;z
zEwe|sY#%Ib1|Z0ky5;65VE=-pvYEuC?v_%)EQ`lf!|zsLQ|)_*wP5dWDpjqkbPK&z
zSCBXQTf}Um+PN!BdaNg+tspMrZ>ViT+zBw<R`tMJy%s~<4PrNG{fsq6w4l8*CdLP9
zAjqYu!+M5T%q+*-8x|voh~5(StO#r+$^|Aifh%&*x!&cmX~ViAKOqHHu^j!qGjfl|
zDMuW$@Yv(Hxbqk6$OSpgGuLhI93uP%AhkR-Dbt{n@>Hk8AU(X_#YIidHXEnVu5j|{
zLfh4xot$NbPEFL0A&%vU*JxiM<)8(fZMt*7JWd_0dvbqgj`t!RP`jWOAC5KNSM<*3
zHg$Bo+v3Y7%C?(HZ2Si<KBx9#GO*0ymB*1Bs{uIAlm9)M5#_}Fr&LVsvK(v^l!HPO
zs~nMGnoS(Yga@KqD|xmH(08}G`*W@PbFHioqJFVSXyEVrDa1iT<Gg=Ew>uOJD7dEJ
zZxOVx*}qUis<j>~`Kv%L`8~4jX?@&juR_b|UZrqlp83b%ZeE;4mt37(^8U5vZ;@gX
zg(Cd1HkYrCZH8p+>@k+sl#j)T0XNsnrhEWfh%nYp@>ftr2(y5d(rDbz%CZ*oDbW)$
zY|A0mgY?}lx3j!m*$45Q)}|$Xu?x78@Ryu`X)S&A5?hPV?91YoMV_7z6BbX$N8L%R
ztpPUIr4!{_Pn|DgWw~^{Y^O=tOz)T!U<NkEZuRC!X6Fxb)BEjd5lzq586}N-%M%^$
zBBOFyE~qfHvfS6Xi4Kh2QqZK2Db$qCK{DTsDsEohHGJpGQ8V~4lo`xToq*hME?HHw
zP4KuETI+oCZ~SJNqtZCaHL4azh0UNxk-D{Y_-seb+N%h@l4H0`8YbUR<qx<`f|7Hc
zEZ=(4dCTOKJe?OyovaC*?b2{{?&#lW4UMzG9*nK!fVUAH;A?}_K@SMAb1X;06u7n~
zet?wrZhLXNOE?|?q<>Ot<5A`B+tiC;8VkG43t_k#je*=7wGQq5BP|JIV7gMNGADVQ
zKxl-be2dZ8b}z?rBB5N#j3_M8m`^#~90%njKY^DwcB>a`V*+|PbN-{<dAR%m!*iGu
z#8xCH62?a#BBBgeU)&Ytl4`BnY$4c*4k1@U--}0VWF#e|D@Z!!r&%(gD7_c2&xvM{
zXSrdvHFQT$IrxcVV#(LXbADNtrB)iGsiZmVxLfZt5;Ygs7t+(}+pqc}lgS1BR^NeO
z0Kd?Gn*%|q3w9l!Li!^=&d7w2t0r}v=Ugv$92y03+CglPV?BF=A+(Gf&E$21)J6ac
z8T?M~=xDR<qBc{)5MLY&MmTOW4I{8~ZFFr{u*+)REql&Zts$y+ap0&7@^=Qi&{Y<1
z0zTHpf-xOGx;A+A_n$Ra@@{K3d-OSG37#CMDr-8@Xq2OpcGI&iK4$SfIFQ-Pv)xFO
ze#PsB9_w4%ha8MUvyh#a_14_j+@86;Yx{}sgF4$|Wc~hNKgaO)zmlUr%4?&+-t}^@
zFLDJ_jvTg>gS~RS@Ku#@prZ`Ag*mW(Aa*Bnz?EY(-_T#P^Oq9v-Z_8SVlL8hG19Ts
zJMty9mGbZO5-1j7ruff(VCfyjfAoN1)+tirKdl?dS`9e~X>9gt^_Be9>}pjRw$PC4
zdnm2N-%GO8s$NV1#zGZmR<#3I0+sv_KNXJkgZ4$r?p!K0{Br^{w-9j7rXcnF#B~f5
zv!)=7yw0rV&v<5tH-6CZI#25z0T#!!{IYI;TepH~95INB(aywkknN_P7^Pr5(kAxk
z8};;C^j?>aCUm~Neou*FQB3{ptYA#oU`+d}zIR>02th@Hy=zs8;c`yZ=&S(MCV!`b
ze?>u5T`ah7tdvn7|5%SiFZ;i!*#Dv6e^>BZN)xY8iM;(qx(M?LSm|>f0HZXHgGDFH
zL7L8fqT76to5O%crLW^gThDDnZY5{MoU4p%D93#680Yqv`uZmbI8RD$GE&Wk{|l<t
zztaQJNB(~xsFd9MZZo(}7L3%n6{${IwpZq+f?bs6tjGU<REz`#m7E(-!y;fW>yP6q
zT3lzI`PO)fyF>>0qx2BTet1OL(G~DxT1*~RYATZ*L@kdKL(<KoMREiQC%ohkhX$Hr
zSN<-<AL+<I4iVhO5Tv!_sKbZ&-jvaAL|VFuM~FARL=ZUKA@K>MlC%auUtC9;c;I0m
z4)_O~`IC0MaRG>~!{^Z5!drrG;4Qm@tbZ}+26}ayl-}~2J^~Fc`MM4I_tb&ilqKg{
zt3fxCeet6Sgbo@1wV;O`3}|{~J0IlGU7!y@o7;0U2@GeyH({?d1*CVOhm+<wZ}7c;
z4V=RFe-*vK0C2w~{Tv&{elxo7v3q+aSEL5_3bOk9CmOK(Xg(|7W~xoicaYFp;cw@a
zwCLNyG!IddI@pw%on|W#1{*cEV5JiC|BV7$StjIrliQZ4Jr+8+?`x2>oyKJ3vG`T{
zEe2Ym+4jM)IBs*Y^s`}~eNu)ZfFG=Pbs%RHtvatsm5k=E^ExG0F|aX?9oLKx^AYXo
z7Qx?CeKjjKQ9-uh^}95CPx3H<)K=SKdYQA@{2~uSaO`Z4mV8n6^CKo{<Zc_S*(MjM
zUsQqmi}wuf<ffa;rl=%s;&-`{v9DC7giQvx2$MCz)ta=46B?PmU8s?A{^L!!VzVCX
z$8=$UHwM>+>U)9$B9L~<aJ|SKN+Uhz>~npokF&zl@|wCP!XkYo<9}C^6VYv7wj(uC
zYddfmgfSFeSP|Oeg%#on&n|h}GYARg;Z7*>!a71-euv39!oI%2VLSE&E`5i*vjszQ
zkYkWLNki^ZOcmZD=dk72wv=dE7cb%tcbiG&|0_-PFDUp81;44_w+L4IpK)hvmpv5G
zoF;g>#ea%*J)&%nF$#uuNl-k>fKS-Sx~bP;AH$u0NQQ9md<~H3PRe+e6LMv_vRkFx
zqTq&t9!kbupKk(AokZ@Blzv%(iN+-)M_XoCWxE#riE?giR@mLFEx?vfur6xS7vs0R
z+WJP^WWy6}$Fq1$GF^DXO@0A?|F_j{!2yS?{#3Dzax3$U`jc7<YIY(s!kl+ou;rpq
zt0IFoqlkH4fmL#XHE|ee)}GW>W4)HU(<yQC`cqG7=WLp*A+Iz+e?x@56r%vowE>)6
z*x5(X@8Opd^~e}zZSZ9zKoH@~dHE6GH1?E?WHJ!5I1cIMy@GF!B3Uu83g+8(2xV>W
z+7L$9!!M(bLLU6tX71%A+o4kaCqE`{MKa&(@K#hmZqE@y`af#S2|&B$%cx#d87~|2
z@U2O=f7Smyd^zMj=GUlY{oieNHH*z|On(N<gKj<;(BVGyQfDfpKWqjZeUG{6K=zDe
zYk<{=x_hNJ*kMNjW!6S;df6Kc!Z--AWVv&UgHA}*{@82$4$KXD#Ve?T8yA~HCf^<o
zb|P=s@p4kqbqHk+mU%fDj5gEjJ6}f90F#lX?vU!i5a}vw7bubX1S6Q-$ixI6j;|Dr
zb=egwxqscEOTP2>+;>nV7lNHA`(Y^TblF~=#TyKnPsSl?-0vK#@Wz+98}Rz%ScO{u
zc`&>_7VI*$w2W_xNbE*#BqJp?uq5E^8=y&!{24*MF%%eN81LWUYj284Y=2B4v#&QQ
z`2siQQsjPYh7&Z8=^@#2<EV?b2CqIUUukXL(>>LI1<A-ORQh8434KcQ)jm-o5->;I
zEHKe(ixSd825eCEXh9iOl;!q|tR=@dJiJ+|Bp+w+<PFZYkFXyArGtxz9hZ<6xtvWq
znYHxeI4_x40>M8?+&M`}x$hf+YOC5RAs<k`qXn(*VuRB?Uvo9$UQtO|F6)(1B6i$Q
zzcysHddD64gYen)P*lxtr!HH&>FX&$x?fkr#D9JbW#ez|xK>A=YWHdRoi)r$&8UWM
z`Wzqee}SNr8wzumuTNq))Z&C<VoAG-M6`6EmN%)p|9c8zh})v39&lXMZg;r&i+WEZ
z<j}&ODP}mNt!_B4VhVcj9~FFF!AF&5Dx7c8t?7VPlvdO2#}$Yg^J_{oW~i<CF=riC
z-!-y~?{?V5CW0|g50L5%+$XgSL-;Z*VbWEV&AXB~-9-M!4K-}jP@7mGM~7Sr^Cp6>
zUUF@HR5kkd3jR{Tf3olBU?SUVC3eLB7pm_s5mbsRXJrP}oRuqGeO-nWt@&X~$7-xu
z8}8V4ar~-giFzTC&46~aX@{L?#$*hFiiDtK8Cpyohkz?`s;u@l7hxl0B`RBcu_0~4
z2(cZ*sXZd^5|~va5d9wcs#eF8Nyn&$s5y<JQFA{*BZHm@r%9-SY!{PG!&FdoY*YqI
zos&CbuoT`P5c%|jqhz|+kIHMnAa22b4{vh9xx`lBtPi5sAg(PwumUShw)`}^z%eZ^
z?QH)s9Mh2s)8O=<pM}Dc{_}fH|5-%f6&akGpeu^Q{(M%?SAIJfL~!-b*M_KJ$%Ixz
zK`-$h*pbYd39x|SW7Mv1uHP%LU+l|~UUeWC2JfI7mESe!N(rix!4B|_`G$Z7Oq;RC
z4j30F3Il`S@G48ot)`S)LHLDNiC`GKG6L3sLX{T_@Jheom42{6PmEWtahLZ>5nlN)
z=1n_;{@C2<RbdjtE8YI*z-qiJta9~pl9}_Z9~A>-G-p3hyk`S;c`M)Z?x3){80$^|
z;{RL5d<*m-M+`hR!H7|}Zazsl4k#GuXfVRy#&(dwmc6axZ_#VpL58maC`2S@8z}sR
z3j0Iq?+ZNmf1sGHXCtS-q+5dr|4y+lE2ze`-niubKho2UtKI)&r5({578ou{kXFpH
zh94(Ak<%J>M{6M#z^{N#jn)cg-SS&v7f2CiO9F(}%603zy=F5Bd8^va0=z|jB7nkg
z1vExts#_oBuf(OeuEs|RY?&1NG33vXqg-ZeF`Zd6qK`#5*Gxe;vg6KwI3gTPZv8Tn
zUL0R0AsqEw8@-&noer{c6XD3xhT<q{e*n^K$JlkDu<T+Y=Gt_JFl?UdziHCu0wd{2
zo7*-+O($OZs%aVj_Zc5n&670wE73BKYW}Eypk=nmV&D3^|0Ri+cg0mod=*uSsTF99
zO>6K`wSVLCLjXIe)CmP`3c~0MEvJ@0Zd=!eM89&Gj<1#Z$M)FQ1gGy%qeR>#GQyoP
zW@XK$z@ha1$o8Wi+j~eVm*aYjtD#1~B;79y1?^^yR+-qf)U<ZI<dD7zF=9J7jopWr
z!YV$i9Rjc;cqB%+fyRY>6bMDcw!NP2e5!pA0_bvDEojEZ__V*VOHhvO;YmktrqzAL
z4q8dw$LWyh2g9O5Cy+He88#?c0>|^PL(p~Led<7wJe$BAr#D(QDWjiMFY4G1iItov
zbS#ZXDYaF}+<qxmGOPD=8F}%$L}<H+g69LA=h2hSdS)MtF<VKa6rfb&oIMjG%jpW}
zlw|Ofx%w)5DN#$4dI{CBMgRr0ENxmj>ZeIV!)2D$NFx#a2omOFlM5^9nFLHO`!v~U
zi4j^;U-^}l_cTk`P?KufuY>HO>lK8<7c~1fbv#V#drHOcH0Qe3N0evtbNMG>9PuMk
z*6^hKA5ky=UB#Wr0sB}Ny~WkuTcv1q&QXc*P&f$(i?&7t+8qrp`0ks@)XOSJg*wD*
z#>da`G~6en&<(rMH`o<B5BZ;0IhPecgGw!ORB_E1j-xaMgsi0fzo+uW<~VeCM42ow
z{P$-(cuq6oFzGt^MaH|LVdc;n5jY{jM#F)F7+7IYb9(a{`wDe^P@k3|=bPvF!j#QC
zBB!{t91Z0OTm^k_-ypj>RyI(cGPCbOMwoJrSnY%$+XktvlT5l43cRconrl1OM%H#R
zk#@}CmItW;!;hs-``$Tc-GK}zgR9aJei?2}qS(0qvTe{UdfL?cDV`KD0Ul-|?hJO4
z8z-stpCGjag_-lez@*Zbu+Xz#j_bwfOOp~8<E<xNH%ebrq8%kHSqVK?Wc+^r^}Dw(
z#p`{$GBiOmiDvJWB4ym`Y&V=;2Ipb)1v_sfNq?$UXB59K<uc1@0O^PvzV^}b5^|3F
zn}ax{Me4$40q2l**UC$_F@OVop}DV_B!xqPp>Of_fZXKAk*p-<hQ0oX!u@!2q`7l_
zG#HanzlXHR8>GEb=<Xul7^>ghD0aulvq$n3Q{C<jc2UZ1Z-|_Gg3)#SX{_%L_S~cd
zldyao*8)3H$BvSJw0R&H^@bVCy~Or<J1llE*c*(xarQ>wR`v#aY1K|zr6Y$kNOneV
zBtNka`AuT(5Gw8zN(zV?Z`UhXfcCz%{T;4s_3eFpiIYbuV~jG2^vw5Cr^CU1(hmd&
zNZ(C*L1j0OQ0CF*u^Xw^(v8;Lad*de_wZeo@}}=5y}hPr-sa1#eu`_bk1zFj`|a@L
z7pcX#crtj8gOsVcT4o&5{@uYr<a5K`0RY*-*Yy;Si+mf<zwz5BZVq~fU&+(%Lu-ev
zrw{S1Bf$|`b%ZY+{l3f_srx$A?m^dU8UHg#$=y2AuV%S^ndRw3DzSES?bzCJ#)yN-
zgJZ#Qwv$Gqy*BP03&!1%)WI?DIP<?49D`#b_K{R@JW4s<N|{d6+9*G7jZf<s^l{p9
zg0`I4&=#HQqIqgAm3ikDYtz29!<)ylRhPI&PwE@A_7sk|_6Osm?hB{?8ozLcFLb|_
zl^(%6OSyR4U=(gK<BF5SPU)JKFYQ}VprzJMtliRC>$$_hE%Ir}Og<T$q^7r1UY_#K
zH&3shX(D4@KYJ76grnb6(X7{M%=+hIXUcbYcj8PrFS0*!rhFG3XfnQ7fBz>H{Ir4}
zRKSG?LBg<iqw?IB>3dj`EosX=0C?_i>z%)&;0?VfPV>!O1Yovj^wWU}!|uz=3lsHd
zW6q8Z`%zjYnn>y^d3FApoq+fcs|kf?MCTy1#2-ML*!uTs!++58H0~12CA7KkSl!+7
z8D0K*c`&l=?dNNz&G0~|+mzs^AoPtOt_(Z!h}}nLodYDj9y8agJ?uAZ$eTm`)P%>`
zI8#-A=(93i=cg%U^=PZI&Rop4HqB1YVzW8z*J|aNW!(0&&mN-uttt}k*+?X`UO_R6
zTgY&1<I~RJ4F_AbpPHU=*{iGR)90q622e-g?9;3LaTmS2zJv$T=GN+1m#BLkcP>7i
zec48EHncNaZ%Hww2de{dc~7DY(xb{yE3ec;QRdYhalQ@#YJ>hWv4g0bw-l`=O|*(@
zs<z}+POUfaTOyf^m6dJAU)}NdpNNEFOredp5=gQDHP;5)`vAYyIY|~(y~!K6^IKpI
zuIcH=6&TSdD$y3L^UWyvL0EXKcKuP`M;$F1%kNWK4ih?m27$6tj9`3GF%z4p3!Fgs
z`xR@`p(f7BXgp1xyQZfv>h^{LDId&Zj{m2+eMx~)xTbd(d*(|9>5H;-dkOwTu}P&(
zD7dEJQKkJNvGG%mZat`k3ku$;K%9_UvUui%@6oz_m(spKAT7F0lyG1Pr|&IAy&VdG
zoK{Jj#=a2_xtz1p)5|i^w7h<XA$!X;<Te@h;3CitmO$~3!qZsPe?Yy%f*z=D&R>mD
z*uSr)eMZ$Zs@p(Pe%0AXx9uK6lFUp7FMDx)<Z#3}tM6IuY*5(6iQjH5nUf(l_)dl0
ztrf%KHT`PX6X&Wb&7@(ksUqK@z(i+~hik)Ftm0`zhuX;D`w<OAvhA_me>d-4)2#kw
zo+NVG7oeRLPV*{T8rjgeoqgsyCV85t(#)PD>^Fu%47!}SYnNYtvW(!&xptSxTmN9;
zd8*hbVYG`77DG&jgyJJ<<xr`z%yDTr&n9Jz_YNXDKPTpd=Lbz3;Jjo-cUD_Dn3r9T
z>K-|e-(pGKUY#+3keo}a!<byH2Cz7r1KEy`Xh0hbVQnQQ*qo*<A{pO<_uyEDxjaUN
z$v05rK@-5!kXCT-N?$BSc|(fc?ef13e@xLEEti<QEvssCkBOP8c}Rx(!ao{RLngYH
z;~d@V%YFVIn`gHEE4jwcH3wjd@sd2(FA)c&b3`Gf48io~pyoM#NgkZNuGh82#ZH)>
zVfT<pqG5|ofKj}-dW6I7<wq^rD*!Mze`<!=v4M8J?GLOTh`m)i=V_h!>1M3pnPMk3
zA^Qnd%g@y8z=rbjQu%RHCqDjkxe>W6^^Jw*+%_MgVu5oo{eP|(CAndza95lSy^Y+(
z$k$*OX!GhaTNSR4@9^KF+y?dxbbf~SI{dhOK(`-MV372~iW!`Xefj&3ses27T+!2c
z1t#|v+VBA!fPhvlz=_6uTic%Col4R5Nfm20|DqD|+Ulgcy}Y_fJY)Xo|5S_ZG{v?Y
z{!5N54bh;<q_wq4i%wG-emt$ERz%OV)dm_gj}OI-rwPQHTU2jh8k^d4MBENqH@PK=
zX`F2`#o5wk8(Uy9@EURxmxtODi~GG?6IKMPFil>zwlL)|zTpn<A2Q4tR(ahw725r7
z4o-sX=DA5wv~_tN;t*fPq9)yFCjI}H4r<XI55^P}azs-;8p69p*-a^7=+KbW0rlh!
zJun*=Igh+{&NHMbRrCtPf{ltnWdXPjxknf8j?fX#H;;j<ef0F<ZDfcJg^qKylk>WM
z@uL4eDtzHWe5^u~TYp)>H3csa=-d40Pz6m{*V5dD3m5q}p5k3yNjqyHdx2!EiTJs2
z!HV{OhgV7Jy+rotb5R<%nC4%&c<G`&(tGaI22zw)o2z>L2UQKxZ<qZ4sHl~9>C%Nu
zm#lmkgp+ReKc%N%PY~v-_`05Av--2D?yS-!p6M`?Z78j&wC5F@R4}2yNPv08KCa+9
z75oha->cx$3JxhS8sc+`T~TmY0jzq$S@->)iv5a$|69SBf@c)S(8(Q-Y=lHYX+|Cx
z{b6K=kd-YWU*qau(rrB8z#LodChvMLk1ms^@o`E;+LfcY4gY%27P6`2$Uq|5n=kZm
zGQ*BysnAU<U+69L7jMszyZ<&&T6eM7pDAPuy~P2}X6P*p7r!~t%d>sTfnT6K?w6Ro
zV`*oiXMO#>lrS3S_@=@>Up1SeQt_WKkp2uPw9a*Cn~89~L%x=&vAgVby!2xil&xg#
z`~W#(D)=66#!}lfH!4oQ(NV{EU{2noQmoXJWZU>3!?~tR^30um78Jv3w^g17k=)tL
zzlbRXMQ1L;`52s)+PoTbITs`$J;(>loxQ0c{lna@7M(?dxS&l;5!HO3Ua%?mLE4W`
z-AyG|#XvcSvbf-C5pT|stcK7)GLm&Ibk3e2hxvauukt`aBbi$hz-ENa0)6m+R==jW
z&b*aV0`v#FVg&ZA&z{GcA3F@!PUijOHTEs`+|)%1#W-kMrvdbrOvcdMA#efe$*<cQ
zBER}m_C$UP-X&i|K@uTOdS~RR09}J}<8(2~Es+o6N^%FpGL31$WcdZ+dBGI)+MQzM
zqqf#o;9UI=>>VAJ`<d;&&{=}L%h%#}CEi5PX(myieN)l@(`FXZVVk;cUlMN>P|D}(
z(^giOm36->3-RGTS6aIjRtK4rauQVZrUWL4jqFX2)^cv69o*Y$rv$1w?3i<A$E<0N
zqTFVjQ3cLM(q<nc{#gx{bm5)zk3qEb13cU`7j3SZgoKy%9Gq?w?5H4$p`TAsntw#M
zMhARanYubq%)itp<4IyXzW=}J9Weu8YHHg(%*Qni@8(TSLs7YMWUR^9i+V=Xtu{Ih
za$y(}U1lUgT=tGR>>RgD+Gv|I-aA1E`Av{M0`RJ<0pGU)mVzN=)laYk^>^Mu6>j1`
znNRA^yiK=&l;DY%e%;x$p-*|K2(aQnMUFhl)Uw(X?+qFt>*7cnOb;vhD0Rr|_Xb|l
zK`ZRV$|Ne%K@W$KR`0z_Ce?^rix0P_S(cp8wD>|K!H^r1K4*+NU%Q4+ClZXR7vR%E
z)vp-ihIA2f$-^RI!rPf7(Dpu-iLzTn{bVv}1A?fi1PTrKfnD<*FzZ7aT}_5?N2i}@
ztB`r$Jz<lFAX3kS{q4#+I?a!>w2x2N9$-DDk#GPDdsmwyEgwmnyCw!(G2PdK{*S2f
znr>*yS}vNw12nMh8C+(XEzjU`(F~-L7jOwK8IKeMjC1jAPLFjgWvlxgSRSYJO<s^&
zJG<A^Uo}1dbp-~HEikBjl74!o-AJ*I(oe2Njx)A2a@#Jy5CdmPK?I-;>HSk8K^8FN
z##|aCMjkCv2B~b06U7h5PiVY!O!ho&S|*-L;mf8!k-bQ&kw#~1OJ6q#Q>sTcV+7q;
z#lrOgj*~iBNXPJT7C4*W_<~$Cp;t0h>1!w}b7m=D0tw`~uxEOmEf-i*{Svvl*!CxP
zPkrc!a%{*$k!+T*eb~ISAY7`pSBw`R<Wf$^`-AlC_EO*NeAO_)_6sIuE<JhvNwDWe
zpalVGXxWIi=jk{%N7au`N@URRfX31Ybe%-$-=aVu(TS!0q3YA`fa5{s{53uLDaBM(
zch0Tm%UeoM!t4s_>KDS)>Ppz%v8KZ=ECfv*+K_BSvkzhR)+M9bS2UaU@nz|{$FRDS
z>l+CkjEBNCSq2UYohIz3+&FESVNz|{?)46e`pth#abHxKcJflo?oNH8#o5i(IKgrY
zdOGx(>~S!CX20~AXg9T0Q*#OP0s%{AVL^7|ZG9&CFSH&`<{xV`TlY?*IfTwZ8qHyT
zrO}jKbB^Dg{KC_(W0&X7{6f1qDt1ha8R~E>;&f03Inh~SONX8nz2S?fzO@;&8y)Dc
zDz@IQ?S0tJ?V!98I?(SSm9v&igE@jsLA8;-Qont%7IYv=;fdFe7H9yZ-7e#jl`&{#
zynp*LyrH)(VG!BH4%2kbwQANJj#6G^#qli|)d6pq_DP@j{$>|FzAG4vHJxK<I!CAn
z+ELSV?hb5QX{_n|WxL<$?T9p;Kh)gS9OF!=-KOc}94T)nZQ>ryW@2tHn$CT!_TA*!
zAE5U{v&nf+(sKrTyiszZ^;|zd&Cp%mY-u_%HX#2Y(~RzF9u9VSyMQ74i0$*nEOsQ=
z7wmH5@9l=o-p4&!Mp`91^`FAte=qtbFgRa3ikVmn3qol+_nD@1|5~|I&&0PQjV5L6
zr;LJ?kyt+#l&SH-Kw88y>D}G|(?9+?^*v6RB?!(#HfIm7d{(C_Klc;gnq=R>L5Q@p
zKhi|I_Oi^{yQBy8bIlX(o*g+ml%r8>-U2>71dKcUa*{c;#Jh*=-ShFgV{<1>i+ZQ>
zy7!~p565|2`Kyts+hOm>O9k}L@z*qxur2NzrOq;5OI`hSY%Ke{qx6>3k#_VJmY9dV
zW2R3$vUb$Q=m`CB4Bh5{cbwiB-&RL@l=7sfJZimv%8c|*GY)5TnYQm1)Ak)-8}A&2
zIcZ2a<&M!j5uD%)r|EGyzdK`aTyW7pD=<pL>FNAG7ZA$;<^0tVg#vZe2Roo)<Ij?(
zjKbj+xH$kF*E3w^xEdsHdO=AAk~--{0=VtOAz5DiFu-zy@!S@0*cO#`*wIBB0UX%>
z3(5f{ZKFB#|5A^(Y7YJXM7pnPn~dbYtbWi>@}2EfgHEx>3`v-3FqT3Y!)LU7+|BA@
z(Nja}=23#yW`k(ru^`-LgEoOfpFBxMGr=Kl$^yfPZR$sR4t_@XO2wl_y5lWy()NQ8
z?>G3XD-}&E;%7A?IReuP{?*lSex9<%&wdrUt}rvV$Og*a*HFx=)1FXpQGt<;#+AL5
z7ObuNa-yKm>$MjZ{B`TROz>oiugfZ4T!8<>3QS5hsn`U;_$?ho&;_+=qkhXT(l1W6
zRZ^@+L9YTSr~Ezz{R(2ulRu!_zoicdQn>ZpRyy6Pr*|m0O~J#;GN_obr@y7whIBiu
zz~=i7#YPm|Pv%N)d8*o6Sn_u&WmLh>>!n?aNr&W*DYjdIX77gSY;?S&aZ8%X-^*j%
zz?%A{#EyrS(x`y5F56%YEmn#YM<<Dk-v3%uMwG(&!FvBF<zb->rPq&7?!icr6f2E*
z)>a9%4pw>_FeeWB0&?gF)rn4?dZ*fb7r{kIQ!R1{lTFhl#L`s9V~d~Zke?bkZIa5j
zVY#&n@jdE0aWB~66?E!r{|+Z9^!0Q~o!a^nFRQ=CdkEU>H|+KRlELijNn^9cdiUt$
z*Bc41z#c=f-D0}&L?u?->+|}J=`I@6-3>+>3M1cX7iQ5}EZ1u>*S-2`vE2H5c%m`e
zC?jpa_09A&?q%xD^0a$l0*B0v<*Cc&Qz|wAC@X7isd@RF>BHM4U0S4whR(WZ6(Ajx
z_ydZ;DOOZ!%{;2gJ<dh^LfPw|POY9YzVo!|;B>q0-U#>V<ox2)mB#6_{Bu4PqU_S?
zZhocnjk9MX7YrJ9Vb9j_l86)<F^?&VU#P%j3M)i#laJQV&PT;6hd7$Q8#)QKOH>q8
zKL++DZIF0X8}<{`NWOp^B2i^W4Iw*cpd>I|^`9rk`Mew>!p>+jaJ75C39W+D0NfU*
zl8!Cskil=~`~E`&tK)e0a(=v89aoM#F3qsDk5gnX*QTy40cD#^?hk5m8=#Gav0+c^
zJ?n7)9n8D1O9w7+;=w|L11rUVvF%s}#ZJA=M)=M&d5Ms0$iYRQSj(<iW%>6LR0gfU
z_)99;e^5^!Qm{wWGN2c?Yb%o{L2jt!A~c-nmZn%<Wn2G>Y@lK*M#SD-*gThGG0BA~
z4Wj;3=K|LwT#Xm00)BCjKN^#Uy46c#*_?4|T#P}FB0`Yi8aBtR)V1u&F9lf$<)2&2
ztzafc?4WGP5h-EY%&9UshQ#tXcN6Ig<Rfa>cO<Sp3_E}fc#fgQz%Qp5rQeMo=EBN@
z3^mJ?a&AYEAR&Y~l;pXT#rT9k&8#<y{N~?q>Rn9^{8+CQu6~=9tSR)XO$=B~?7%l+
zvG6z{zeKP_yvuKwndIR@GJuJ^dbit{)DTmM6}4BCKOnr+<1?szPkrz~K`lxFEG5VA
z_66NH6ze61*CAqkZ{YF{g|j{J`e81TY>UofFxnlT)@XU7XWzs#{7ST*nZ#-6qLBcM
z{?9Nf{?8Jume1n~PF6*DYrkIp#x*y;|Nq|3CN_?<jN`U<ZLhuddi@b6aU5suq_H<5
z3yIPSwN+A3DTE}ZN>eBXA#AU+UdLa|uG6q`;84M-LOoPSND*ppgcCxX0rAbD2dW;T
z1qlvZ=!tURgb2U?^UgcF>s>oe(z4c$XWp54XLja&-~aRRf1ih$fW&0hEY(~^;Z2AU
z;SZUv9GSMXBWiRiBrXHPQL#Xj>D3G=z-AVg=huk?U&=BRMS;9F*2(vn%@^~Q=ZLnN
z&#cXt7nfCoxp<dwK5Ia7SIgni3x1^=OkE{aE72+P?#?P*j^tJ8CoKvSY&o6qSeZ(i
zq+^82B9;V4YG^UxzmmEV9(0MKr!$qomupA4B8)3CIRwAf!g}v-wDwyq=Ct^o7S;vT
z*<6;n3#?Up@xBr>d$mXedy`sZ8*`$SBpLFZom9b}w9sf0d|HclwXjQ=-1;7yRbsWg
z8ip+XxptdT@6THMix!d&+`_~H4FVsiubyKC_ZkqbDOG}PbHKz#+MUH4_nYk<hwCxM
z+6lh;^EuC&`*Crd93ZufxPt3zfsndprxBj&bBeD<?Mr-1uf!O!cjcYm9;UhPw9|BY
z#yRZi1M2(=?a7)Z7O)0K_AoQ*JGV#C-L=QF(fziw-HgI$b+(#e@yoeRjg7t#ve6S*
zVt&RRvC*+<a6R$0d(fi4PiY^HlCI$2SOZlvt`hqX2fdDs9`4EaV~0s$iKx{eJ4|1G
z!0a&MqPM6cpAPG!?iMu+VK0JfO?Uphzb`tuD_LO3N{hwK{SMl%8^p?_cpDY2izP<u
zr8^uojOD|IA<AVH8~P3*R27kzGlw`+T=k$lf(2&`&a>yLAHjm7Ii1YZXIQ+6#YHxk
zgII8InNjoKZF@9NC^KybzbqdrA7-BaxLI&Uu;3ivX>b-+9kP03!8wWrXPjduEVP)y
ziXD^v2HyE7M^3<M&)D6<7Z17W%-rPoNwfHjm8Z7H@&~co9AoVm9CWQ6-^L!~S^@lX
zO4uGYp1CYI_t5{3U_+QO8v+&}Y$OK??fLOL#FPkG>qpI&les-vZ$;pld7I5}eu8Ht
zE7v{l?uqS5TFLS4<6K2Ho_+=tvEq5xZ=Ix0aml%k7~rzUxNBkc7|c&{AE+k$>X!Eb
zll)|U%B&=?zwJ3pbEWKdmMb0SIwzW}B*Vy%Pt>d$KNXgnk(m|eLp2_<**6~FazMTS
zFMGI{+WG&REuqdU{BU-K4}tyLN>R^F??c0W1U<SYmq{pcfb%%I^=qny(YJ|8Dz!R$
zYTr8Z-&9Xh0QZR6vzcPdBo?XA;%Xz?K7#6;Ky@}kgelR~G1pCrqQUJ{q<3JW)WUQ(
zi1aNUiDz3*5X0K5%wApYZJ{n7YM<H%%Mxct1J6>bH<5F|UhRoPCv(h79~r(GWowov
zJWbE)nVW93gK%UD^7HJSb%Z0gDm_DQyQU(aPzAr&>vN1F?Ujg7r!hZBS$>8p-jhTz
z%6W-+JZ_Uz+D8V`$DG34)*k4XMRrG@&T!3Ud^f1feWonSU)>Y=sr#)AM@qe)+RBiK
ziMv_vsZM!M^+l)f)^z4~OjVUqwnUZy1jb-%U7GPGEIp#odeVEynw*i)DS{;p$cnPk
zo0VSx+3^-~BD<n_8YP!@Zz)&EA(@tLsM_#U37BP~sb+2#=QGwUPm4|CLdZV4#+|v_
zVT9sppRPxIZqwz1&*?#{N9&`9taa0iKgNppt`a{eMtu5sqywV|E8Z{^qZ#?$y_Y>+
z^Ridd_1BR?YGNzHt;wmTPhM1GiZsU7dSX%ii`K&3ji4cV-}%m$6)A<Yjjm*Ah5AXQ
zG;Nlf#!5#DWVAT3WK6X;haSua{-hT+x{e(w7(MK{vd-v6q)Fa-gL=9UVZMtNMOKn+
z#5XyUR|wmr&=@$9`w&c|e`${G_kddOc<v{LpY%Xt_OZWfg?N2Lo5gKME+T?wtfUo;
z3E)n_yRZ;Z$_hyoz_h`)a`kYgotwmE#yY_$W)W$6S4MGr(B3&3mo|a|4(fdGoesi<
zbRtS6&`t-OOZqzbIJIRY;pA-Psj?C(n+-<|)FoRkMfqV%F8Y}iB3g5;(eE-@X%Dyc
zKmL3F{(au@OS}z7tPI1W)r0Xt$M!V^XNNc2lp<?3rZ?%Ow)AUDRdCZyta-hrO=FrS
zjdHxXv5L0oatYngIA#Q5k%abee7kvBX18sDJi%vZ$1sI%1k<d{#TOQrUk%Q&+T5be
zny3~n-AWmG6dllI)EPIu%4I?r$I0!qCmBAyC8Vj%M~?2u$svPC)PtX9qfGEByh$DE
zt=mcs7$ZS09e^f_#IEGeBqJqmAEaPQi;I-qffOp90#o35Dp<MsC40PAPX~N~!)*S6
z6A&_bwx;V&YVBXt<oP5~k9Q&2L!zf)tErp<kk|E))Op(o5hp9OxBE6dggSid2tKQu
z&<#5qaFd1C;LDK9ZtbdWZC@5%{kjF+Qy;gK=#0n2G{weyIMAL&-KhWU4pm8mAO-cP
zL6w6qJpWBlC3|lpRP|T=JP5PKUR;|m)R}nCIJojCqa*-p8|xd3Tf{tf_S+1`4TgOs
zBd}xdf`e;JSJQm21`tp)G7u26Jp=(UiaLm;5LYY(i@eP!irG`sBxy>lG2YQ^VzXv=
zw6?=&A6X|I9W-f%WXJZb4!8~MKO|)ZAwqDNBO5r@W?3jHXH*g;)Ovo@a+-UVUg{4b
ztiTpQeyZc)r1ut2s})`hQbmLpa{4?zUI;00tE1qSt{Q?Q3_)Zd5Q6-eSg?j45n8<7
z0xe_{gouPe!EnP(9wE7m5F`zW*dYb(0%K(Mvwx6Ukcjd<1-df<l46vBD~)9sJqQp=
za3t`?<^YfplnCN|2Z^wl?}z{2znw&Fbtv<7P^NE%TyfNqYCxF*PzDkMD03cMNjIQO
z+E6AvTY2(fP{#Wj3cG-ep@+?Gt#&UI7q|Gn03FJJWtw+b?Vn#Q7S?Gw$_rewNelS9
zhIyma?lAvwf#dXfuJE6`AG~{I;l&GK37NoJ;L?Hp;L<YnIUUOeKLesKXh5>DCIX`1
zlFB~L!PWRYDR;Fb*H{gQ2J7@FGuuT3tbO26a9Onrh^jqdU*i&fmFw&dOg~ijF*9ET
zOq#_a%BoKsmyp82HxZlS|9Cw}`KbP_I$JYPaVevA+9<Qhh8Hy*NYUB>EX!YO)IEAY
zyc{QXKjc3}ZJ}L5o~K>MSD60PSVp@hu~ohso45s-rB(8uHl46syqMw)-g-lLlS#2Q
zRoW-24qoC3)*1=N@cHM3YNtI@PbW21$6oP^dg}I`uW9XN793=vZZv2)@9!(x|3xjn
zq=Q<b-!Jqggfwjk2P6EA@vNhma;=_l9SXJwv#QvfpdTIpUZKu6bz3sSCSvi9R44G4
z2~^%1{;qLggQn)RvQSeCnZ>v_-|EJlnKEef%;JPAwX$Ts&tR~}+?>52%s_Rbfo0L8
z^H#oU@Rs2dDP4j?Scj(xlu2G$fF&~{O-pm*Ht;2&x^X~hV~s=`n;<0K1>@^v=C=yU
zUE{((hNx+?;bx=wL~6?sVqRq<Y|BIm-);vnyBuO-%!98h&kdLvGu>deOxrb23oRQh
z<9xZxh|JxJv}1lI+=}(6TA8?r(VV+0R?Hb;KgkHzP78(E9s#cMOqQ0nu3ukWB3vTg
z50}l>c%SBLsCoL#=PgRJZ?lHBU!fp4t%b3jqjucfW(B_QsBxIRp2?strlNBcRHpZS
zTM-n0^t9)|eted@ePpWhJ>Lh~7!*UQ^SKB7z<p4jBe~=9M<{nFL+mX=rI_sT2KWw%
z89!(W*F{RXHCfYC)L#6VKO#ssYh?mPmfkeUMaP)OQsoxPQBjwq;r@X=Uc*gor+oI1
zu(knedmkBt$*jzBd7qITIpb~7t5ZlnNrK$4g&vzn!WdF!5+sa-Bn<8aZzOWXl-Fnp
zWyzG++XHwsn|0jhch_-0`7qMQmQ7focb&do-b(;!r4m(6EMyilGn;2J%US-NC9k!M
zETtkg7qgjV$~vkYn`ii*^`b@-m%Cmk9(g#~L+W2(UDs$%Z)>wAk5~t=?ridgA=cw7
zzmBvNNgjL17=Kmwd5U|gDZfmI!H_8_LoP)(SJufJ${49XxZkEEqYBSFpsO`Y8VxG(
z3>R*cG~fU<sKa#%Fyst6W1Efkt=={=rwz6y1?aIqjDp)vk9T%0ei~%<>;ryS9l*zf
zIR<6T5;BZWwzScjl569-z2p6Bx?iU5Kk6-sBM;CsD>Bt<e5T1j`?BrVvY!e}kyg#s
z1pe~H<=pzZo78KJznj`IxcN=nh0nUMTvm{Kn*8eAA$Pn@O3%TK^|P@9H#@s)sklzY
zcV|S<yK^aZF^H}f7Xs<lfvkbb*0N42hmF-zz|CTl)<UD`+@M|HW>Enna$DshNA&27
z>8AU!y}&8S%HT1P`F(-p(0Q9dOOngArNX?F&mgI@6RY_(J3f$H=3Ml8LKd;Z@fW2(
z218m5vzQyu84wJ%R;eG}QZPbkb-KPZx+=F?cMsZf+3QAC(P(%IEQj3uR(T^#>*oS?
zh(c4HY^*e7RQYS`83YOe6pU$cM0Fmnt23ft<t#A9xc0mBE;%gMfLCx-MK1G9?i*oX
ziy)X#Z9k!l1cs=eV6!@K*}v>7_RE7aI!Lo*Z5Enw;mo>kb`z5xtyR_04kANPHSw(b
z3Fq~OOJ8FVXU@0RN91JDx+dB2*D=FVWtwo;*U=YK(2GQ@w<{4J`F5N8dnWDI-2V$h
OrzlDC9Y~)|Klwk2VjNHa

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_36_15.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_36_15.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..3263e0bd287d8f23156d2f0e83ba3ab2bb0bdcd5
GIT binary patch
literal 58025
zcmeFad3dD9ec##lh5Hx`=I*%%<lMX|iV{U~D2f!MkR#CoNwomH4Gb_q1M_xc4(LTp
zYev$@GG+N(AEJYiZP_DRzT^0i?X?phyT0wLk9YGVJ~mb;C(p*AlP94i&3?YWdiw;1
za^l@*|49Q~U2nZr_10U*@2+1h_4E{Cd_MflN6HHyipTy^7p<Q(7jNZT{jGQ`=EWvr
zUffG8#3$nRn^;UvB;zhEHIcTo%tV&o<U(#SKasb4sfEH~*F@K1_e8f{rx$t_dnbAq
z`zHGAI<wHfI505~)iXFTNcrr-(Bja<kfr4oh8MR?Y_WKLVPtXZ#8!(J7DgAhO>DDx
z*TVM29TPhicTVhF+%>UlareaT#XS>y7K;<b#k~`I7xzu<vwYnP`xg&P9I$xL!okHu
z6NeTLPaL-E-i0HJV-sVGcTC)2*L@3jE*_mYx_E5j*y8bt<CfOHaANV~#7T<}ESy?A
zJ#pINg9~>p-aT<Q@gZ;6+v1IUDmihFchnp8wq1))-0N-kc3g{1-1kJx+v)AP7V~zM
zv-9_tPhU+<JV4rRZx3mE$`4#kPCUqU(c8=QUgaU>O{DDe_LH)|{2<p4aecr$$n`<H
zelypHyu(}{F29-k4|9FQ8{>M+uHVA-9p0T>-)Yxx^%Bp<#*R%cGo*z}a|>Q^aqe<$
z#V;3^W(wzOwdLxACr-@H)y}O<9iLuWJh8AcJ?Ht8bCv1|uUuK0tIkzsk9l)5Gb`1(
zrOL78E43wm`rL`Bg{7$z_uY5$-YIWp`res)PTx0k&;8zAcg@^?_tbq;_nbamzV}r5
z-tt|i?!9~Ju4YZ8Mop!9O^v5)ymx&0%4SVmS}s>6=T6KmF8fOt%U-i}u2tm+?!No}
z2TtFA*VNSg_uV@=ed_M}?(^>PJX-s})YO!B+I!%Elc%T4jS89%J~2DFQmxKSR+>q-
zX~UySi_7nwtDdM$E~<lS{$!;(v*a(9{c7RZu_C|IbA?y6xW>ogVeaY4#VK#{>1W0g
zVcsuK`edIihZ(=TT%N3v*gL&csg*C+7AGr{vt>W*s;n$5L}`@IE>HTkxyc2hsoDi^
zk>C8{<oR-G)}QmZ=%v2%rPZ>(R9dPmTp^|K^jx*}=u%~7Zgwme_AV^VmTH%4rO9e(
zZgDoulrM9;8fLxn%;d^KElkf<7blm)EW^D>1E{@BC(JBN*23=TrIkvpL<=SttKn8A
zMs2b}V`}BnjK8#4qLq3`Jj^^jS9^SMQm<H{Cngt4=NH1AkIhZaE>$KMo``z(F&^yC
zO|8_JO=KN>lJT$2)vlbG^Q*OIYkGLkBP+F~a%FnSEBjA9J(dgy?OmtJwaHR-`dryt
zSty4?^p-z4T`SF1YD=YBxl&#7!-6JkxwJ5MzD$F<9((M`cRoA0xV%vIx$J*ZulDrP
zEFJvLxyy4E()y~)lhfr=ZEmq#t(BLnVQyM4G|hW=5%ZU-)yZ0|62`C4j7)SV-%L+V
zdW&bL6B_5}qc}?~_9)+KiXe!w{^F}s@48s_FZpw|a&eItTR2YZ=PJcZth{3F++4Ld
zvr?H>rxa^TMZIsaszqwKPLwZBE?BWbk&m}jt`;jxwc_+SO{3yu<;vocUp{^_Mu(=V
z<%O9rrDymDdAtR$FP%8Iv{*htq4Lzy()knKWNq@qtiQCfeB#-a%F|3?`n7UGU4G&*
z7W><teCEW0w|K&u3M8nn{}Z0KyjZ$WwN5R~%vH*@l}gzw+gROwPwB3EOQ-H($%MV$
zTy=74q3o6H4Gya})HSK3KgQ3rdZzwST&TyoY$uT*7Q{zmwb*<-NO<ueIT!y#e3-PY
z>dM4;nDNRN=cdcNZQ7twm}35WW2rFVpW=HuOjplME|>i~sH@ot{t+(2ftAXf7Agy+
z<f=JFvbAkz+&@Wtns3`@cXU<#4hFUy10BXDXqT7pl3r>y=B2%im-TX9-Ya-r#Jas6
zuh&a?ef;(FJHR*1UmstK&FubYY$omvdPCRp6Ulr$-|M|why>)3impbytz4zc8SfuC
zv~o2*kuB%e^EYDU!d0F*lQ1+BC%wy|nTc+PS|)lNN}1?&h-9MAA&`lFhd3q%9Krzg
z-ccT+o;$sx*HRP1u6FMk0P68;s&`A2@~n5lJ83B+QOa3>=v|hwmG<53-9uYPy?ed;
zKzrM~_jvbv50J8*dk=bVBDTYO$a^#Qc6twcZ{c^B_g3!_es_Cs^B(1QPgL(?-s9I|
zJh|wd@tz=kulIKENq+Z5_uuP1?Y+~UyWbo4-o>2*AC0AA69<_W&xNVS9((sYU;TGU
zMme*<)SUFg#KP(EGpjk4&xr-r@>u|#fWWI-qOX!M_Nqtlg=3HI{OiBCcl%qu@WJ69
zdE(Dkzx6F&_|nb4`Ga>(z4VqZ)Mr8z`Hyc2lfa4TVm+N-&8r+=)mIIe<+a$=*n0d1
z4+*w<@r$u)&pTpX;%fXu@nCBZD<x{lAU2<xPk(pJQ!K%6>Xp>hI0OZAzmy4j39~^D
zVJ;{T=7TQ6LeNdv6>KByuJr`*VB~6ocyFzbT6#4PUrzAMJZb%;^!pRF0o}P8CuOjf
z^s=>~+At}D>sxNn&sXC-IrmCZPYy=C*oyLvaDT)v2cyKd2IMY{21A6~f-QvGgPijD
zD?yH{9l=h*oxyJZrP{6_eKlTv=^ZiNGHrQYp{CtII>-cBFaI*6Bl(opU;P&4qCEvY
zxwVbK9{+QscUg;+^9QXywRn&9XKcP03<k*{5%dN9!GPEOa)w^(Su57|*7lKFoQ+?l
zr`Pt@_Rk-v9pr8I>OHwS<n_L+-a72{t?iqQt?lP~pmv1zj9raC7JG5@g~ZyywL`(d
zVDH74|DS^W>vwql!J%N^j@W4IhI;7G+P>hx`W?YRU0n!S*Mrw$!T!hS$+i8|b7yV;
z`hj5ojd*P+*nc(d4cTa<c;?YyJK?d~@%a<M4sxDk^$)+CSUXJo*eB!u{@SVe)5Py$
z47XS-@2X^i*!tbxNO1Uu#&p*w<IBfu_gvNdxR;#wQtBbz>b~{+Zz$~n57?;h*w*o}
zr2jBeF-%-O8OATJo}6CseUR+Jm7<6P@XGS?668c>wz!X}ybt{0757~}8AUGcTkUzU
z_{>VBq8y;8)&674e%YII*hv`SESNJ3I-6RmmRI|qDwmg)_3?K;QiZ52hVkP-=B~L}
zsDN_meECW>Ea<@{Lng16o+($$HG6z{aZ=cBS%}VV7B7|OX3y2C$BSo$68T@LmS<KL
zikHAnMIRi!<m*u)JZN^23Xa)3fxV{BFE2q09Dn_>;+@S;@wrEy8Gmy83Et@OXCFU%
zmS}POJ?|{h*y4NM{n#UCAAh!Z_Fct^$DesuTd}cB*fl$Mv0N#M?5Ku4?bJuUNilRg
zP*Z{|@@rusx++&ZXo)4L=W?m~s!+_$c$oHTS3pZ|=IT|^2-6DcAH$AZJj=JbonV%6
zeKEfl590F%P1X}PV&9aw06A|s4T2;kG#)0Wmsi4cWo2>s3jLlqelkp{v7Lyo@X!LJ
z7bO1j!yl*i*Yv#YvDLi-n#W6!QFFCY>G(qnOVg7J)rX1HbB`tbC&=qRNx%aNGjj_I
zB_Qfb8De8@Rg_kbU%s$1$M`^yPI^G!rP+&<{@8;c;uT&v%+>1;`;Cq)*UJ9f>ZHh>
zFulx>*1~+LG`%oct(Hp4U71`g6YGcK^3=1lWjA}mp6P|TWe~R!dS2Ki!nVv*yi%%#
zeKVpVqwLJPuwcNxq((9r5KEQvLNy#^PL-A!8a*B=)hJD7RW<Av>0^A(bN=MA8lJXk
z=@0ViVOHgsasIHb@^(kpVRE5d2~+B?Fgvv}GXoD0_9%0mSoOpHMl_<A!>)^JgZcrw
zb(m>AJ7<kqm|7~iRS@<q&~#C+^@Qx=5~D=^E+dE;u$j4<-=pUA+W0ww{QPn)9AGfY
zwGy+Uw6e@JDOZ&=OB>66qg`Q=Jif@AFi}|!)8b36gc&I5TG<0tR%&;h4*MEYU8FXB
zk}p$h*{9C1cXDC*96ZG|lh0pb;MHr){H+@vRWB<nT+%8CyB9#vm6Dq({+((>ilx2~
z_NdW~htjf2n18pp56v3?d1dTt)bIM6l>m4OyH}=Rb3|7&oN!kFwxz|e>lyonx1hy2
zWm)zsvtvj6L#nC^`n47ficN9<?IioxbT1D~5$97o=Rc}z*r-X*IINON`76qKUO`#G
zjDlGOA}ai81@BPsl!B)f*g~o*<}0WXxJho4kyZ;{`tR4h3kuFD_^@(mrn9DOo%$D*
z{0&Nmo76aK8g6M0IhXZe_rI2G6E+>2eLBcWgBjG|)yHlkUnyU5z;2?up0gyTFg_Rd
zv}Lt1xEc39pjH@<AqvQUk>D=A)t3kgd;8;=_;$Xnf8Ft3yY7zfju#Sk$HJV2{qbD9
zKc0#g`0eI5M~pw+yOm1zQ(MaZ2Dvk6`4TVn#z*6sL?)rM;dpnVH(_<$TJ33rYtvcU
zRR8o6kswTCTK&AyYXLNZpAF(`*-E|!*8zo*66NB>Urw)Oy+n{5jjiQsiTNb-4^%(6
zJ6}tzr?`{%pvFI(*cQ8ySStkC`Sc6%wPY<5KvjX4v(zIhE=`yZ^7Cmgxed5JpFN+b
zz7?u4BdV}gsCCVE^X>7n)C{e+-s|O{`Lb8zhW&Ld8Zh5-UEmt}sd2w6=%Ph^LH-8s
z(n#%gscNUFn3g*|cBg-nJH5>^JKD?iwchdin>9iy#pVY_V?lNpN;fEs#_X48?$Fx~
zD1B>e4f>Tg8KC9eEiD;zEur;;LAN(_Bhl8*VfBN(S>Xm;plk0AXs*UPZV9^FGc8wc
zvt08--pKg`G!xYH{IIumDCR+l&u@XN><V&MRj=M)MD5-YTkEMjz}1j9YGw9$+f;L{
zYkjMF*XrM1ztS6$f3M}=Vfo*}J*ele{+t_eN~UVtyq(4y^mzTujUF%gkov6G+eI8<
zp(tl+go2#kaU;II^G1yKctLcr(R;=5o5@ZvBMQS8aRJehnnc#~>McI6-|4`0@!aG^
zBnwN$sd5nk5G0CMge5MPC#R7`M4}LXoZ?~Hs2~696d!5__CV3{#c-&4!H*zG6AkRL
z#3e`{gkvjJ2WXnNcM56W_U<+-?3bs(^X*RzcQkJ;p6a;izguk_Y~WNV6dys}bUEDJ
zOw)tw>5v<9U@Pwn```A>-8Vk<>u(B&n;A{~HN8}xIkjo(X(O;-6$$0Ph2RT(`TSXf
zU3{zmR<2*~UtBRkO+<2oRJ#{0E=)!Eyn2S=DL`gag_bL|v0aAC{cRNWKS&T}BgE=|
zNJ%G@lmrR;AJt_OwFtO544R1qVc8i_71-PVdX-73-yl<#W)1Cx*+u5W;>w~!H0vt;
z0}5=IGiRMFEzby5O_UZ6jW5-n6d|%mE6d*Fz7Gw8$gnb5@qbJYNKP&+-0U@k>QJ2F
zP!WtRtuzF<uIw2%uJ%0thEj5p;YqOU8lp-&=vV&=Y2VHFwJ8E2j6xy-V#p<W0sOf{
z!4S)6yg$JoXk;Y09fYwRL^GVs#1pr2@zkyU<ZjTzu$2OB-0Du|graVBCk}vcx{2p*
z<&xc$-X32aF$B^eH0lK0|HLH!6O;VEk4b7HH;{P#|Bgv~EimIx4SV>X(&hIk_$LYs
zC3G1|i0Jv@b_nfKC$@E@3RBc?83^yVZ+ykXe!F!$ybzV&_eQuQYSf-4n&?0hQHH(E
z3>)x>%i^%fXcHkcu!iG-4m9s<hEt+SZqjMt_GZ>5R*4$2*}c<IVTo}3A66g?<9|-U
zk0^Mnf*&Oa2V0SeD|W{JaqeuuB>qpTPy>tjKgH$5P#dG=%1;coiZi0L{-#hPYFq>3
z`2SQD|Fi<38vmau_!$Mp;QXv&#?Jh6#eQAEzfkaV3VvR}=N0^dg5Ol|iwb^8fpHd1
z_TnmOf2j0dAz0nHIdpc-+n|sUwJ&z8s)l}@sH>R2hcLd5L~uU2p1R>Ad`f{*pHG^+
z5Bh2)?(cz?mz*^I3Ub9b0uwLxUL?EExmw<AwPf3Bapm3;l}osCb@`lr$sP0GciWN^
z?Ikb8^f#6qU)}a#@mUl`Qh|wllME{on62(Wz-8hw-HJHi`pvPNV^SA*c_cF6*|I;c
zc%dPTK@~W?w2TmC$uH?n7G6^Mko_2!%w5w@b_@!#065JmMax}~n^p0FiI>EU^}>(!
z#uJ2&H-lgMaw?Jeiukntc;d^6#A>e%VWWcS6ou=bEktsBjYZ2OB~nKjWOfKJNGUH_
zO9ZL)q=eWAYli|-w#C*`&D6DYkVdFGvX%)F{0^*TgUos=$lgH5Q$r22o(VEyGZ3w&
zg3Oy^7xr9`U^WYB{pMOuqV#LAT7DgNVZBQN*7a_t2sKHV3?taH@t&83HO#%7<Y|9}
zC^-)^m=97H<9_MFdwE8F8zeofU4gK>26Kl{Iwv*=M(~1(!|lq;pHKN8h7E$P6jKRn
z6UE5)VH1YK(g^`o9|{WVeK%k}T<vSUu0F4DK7kOuKj?vl=-Nh$X;JrDA4&z-H>Hfk
z#AXD255+G0lgh`VQu*^~-u?cLa=X|1qLe~Z)9_k<YfT8@yFBiV#Aei5Y7P2l64d-J
zF8q8gwhpVOyD&Nf9rg55&p@+unl^cb`_Q3K_TN!bte`77*io{ly(CW^4Ei5}DY}pj
z24Tke`*1?7SsUct2BrRpdRmOr-NE4ckk?I%Ho03p$QumqQlDC{(w6rIN&5N;=Eg9y
z48`8diHer{mgY61J+wB=tm-pB(a+O{OtmsVY}jIhufWQ~#@QXO>wJoqT%g2IF!bFq
z2OwPD=-$_HZ#WpP->U<Uu`T2NGb{`EhHgYN2+z<|ONd%PMv6jSQn1&HkChk7CSOOE
zRQ%wYla@C#9B&glv$oKnMPg@3HcjlD;ho}Zi6Gu0aJIFNZs9DM!jvSmVZ77<5d8OY
z^N>J6g2-@e^{_4ScH!h^Lgv+dXXczzU<nhI2&xIlA(g6M-bB0Ym8;Xz`qm~F9CCa&
z&vC?x=oE2O85tPR54{b5XzGI3Gfvt5dJh&b7<r&ZkmBR@{vxux%gg0yIGduh6{~&k
zjP5_h)q{oN>QLc5%`8#k*hnO^luXdC@a!<9(l!)+tt!R7gQXd%n<r~dXt$`0d8`2Z
z+R_rj=rm&Kxk}it?)03Ig~Xf)kX=e7Pxb{29Rrx2UVugl(?~aGk-&>f3)3?FsD}BK
zO7+4@xs15G&qS5YK{tRuw#7m7R70fcKc@C1s-FKBip0-{iMjJ(t`eDh_<y9-w4sSG
z&&ZTav5tWUvci^VO9qSj-(j`QV~|lo5RS}r!E3}R3kT}DI4m?fzjG>GqK@sFN+Mf&
zVd^9m%<)$+DPrQK#Fuj^T^W0KE0J`u1Z-jNmvWhe-9aW8-<c@D_Nnw>d<0go7yL9B
zKcrZ45LU03D>074lo*Dg%q0>pZBHPFymbKPQh#8tTdPBxOyy}UpZX_5L`%t#*4v2G
zgy|(sjL#>oCIh5QNCu%KNKGOui_$?`%9m<fr=9e1CMj!&$=!5;nSB)bJW;M;ErQqv
zi9y|prg-7V<x;H}6fc)fjva4geb7dNL?(e>d(etKKVH0Ce3b9;<HuhtI$42KcBV~P
zC|6OrRhC!Kzjf9Um4{%QD*GzYum~xd-AwU_NqUN>#*P=?UiQmJizl&x!X^rpBbF?f
z*0faVV>IGd9xOgmtgRqape#9$7V9MT(PAq}J$AH+a6ylZ-sEt#Xfv-_p)o9DX<te+
z|I-R2DRI(;?;|#riHMk}hA>{kK!#_A7zRW_qVCpbcFe@E??TB)_j>O4X;x}Q4rhm<
zW(x7uVVfHr+1e-?O^h^OO^(+@CvBV@mLk=*n=Z*=6wl8B>&I`V+`D~0#oW7nh8U?v
zn<W{2{}u@-0cHNw+4|8)d+R2GGupC@WnSYoy~ZF5B5^BmYju~s#_jUo_I>;mHTh|R
zv8biL+Gy#oDVEiJBjqj7>m!zPO63jBay0yT73m^a-T#J-iYR_{Z~fES)y<~WJFz)I
zzLx?~C;B-(PMzr3tB$sA-=wXpCD2Fd;spxYTo_D6ow(<A?Tt#@wi}In?wA%#f{tkc
z6p<a2b+_N6X7?)SXmMlE(eS$=kDh)-nXO;j2fH(0TPwxxrTN<A6#){plF3?!Y*8!g
zDG8JJ*CAV!x|X(Ff^Sw1;@q?>f^=^BWZXpmtOmjs@ee6x!pR>bb~9^J-v5eH8$%ZM
zwzJsH{GGkhb)l(r{aQNWwO%T!&;I_DE#>}bDL3l8ZBO`EF}Jt-e^a}x-T%8{#_s-x
zVuPCu(LhUM+$_GXT$nz`BH`nP45m|TYfqlq>myU0;R03#OSvF{HVVztJS-FJfNXvM
zD}WX>Rxc;l(ni8!xXryCK?-Im>F*EH>u9#Tw9~7AkJ2iEO;PfN<XUDeyOxtBPmnXb
zlpQw__|1e{G}wN(Ny7i*FBwoqGtoX&KxuKFP9#vE5Jr^?7!#KKVJ^n<4x+^`yIQBo
zT356$t)N}MTtY7(a@Q%}1(a>D!;!{lNuk`3@pq1Y-*P)3|0W(<9pow90P#V-QZRPM
zN025*Aibl&wEDXK=_b<X`wwve>5Q$$r)9Ab|EBmeup94>ucg&@js8!fQB;5CNX=sh
zMt=G<bv3zGK!=4z;d!)*Hvs(eFtD?6FC|&PJWQOIkZ$VJ)cT1SY$Ob%mk#p7vGwc?
z(u0D73S${r|Lo9c7%)>0nk*~Q2D!LoNeBgOqTZn#MHSdVEKHl?!1#YNp)`6m34J`)
z?Ur79-C?&gVJ;a$3FA}#AFDpARD=tryE1ld9~4cNu)DHUF@xpB$?Eyeamp>1oQdfx
zWUUsca(iL~I_zj-G#+Vo>p9veFd;Lo2*Ii9IsyI88d#&By|@z)puxqWlvR+Bh+sCp
zmZ_y`=^C0|@@8c_W;S_Qbi88a58K$GzeV>8J3kX=v=Bbb7rgY#T3=mSAJn7~?jC`*
zgH3;%U1dFV#;mU=5Ipn*X|owC1_^&$dgU3cc!D(aU2eQSGU~I}3udk?-lXRL2??w(
z>&tf$AB#Ja0&zTJz5Y5s?&h7uj34kS#wUb1<gM@!WzYYjGJRCRn1=Wx%9(<Hm~~Zo
zx}%xPY!EGBQz9I81e$|MuqCqfuDVBwwL=Ak33-nRLK}BQ@OO@4PkoFS{-x#5Xeedf
z3xUm0X5vLi?-bgXovgOROH%iz;;VxWc}Bx;1z+u_P}?wT(KQy{J}xxO46=q=gCbCu
z0?8z2AgWn#+5wTk4&P@h2ZFQ~LyfzXkakd~Fg1y2&8&qiV@tM!Jj1=6kuDiGIsd8N
zNRtx-1l~y6s8wvvF{g6(+W>W)C`ppP`XB+QFllDMEbVK#s|i>GO+kRbb=ZTe$?Gg@
zOW_?{N&qonCeX=ot?RarsvY9{n%aZ-Bz_&CON3RKL^eeVGKj(A2J3ix8B90~I@{aX
zi{%~y7_(8~;a}j@%tEXXTXpozg=k1xnRl0${u3JQB)`9xG{`}lR`wVJNQNlvfe-Qs
zXOf+(*Zp~QsMfIfGvm3O6-WL4rXBhy`CBO`!BlZuP#_JOq``lQpl(u5_2wB&vzA#K
zJ4QP_i-FC>UsiLCZYtOiTVh*-(lkV0YtT6&nw5rrheXqlnh_uKGwBZ?yTq7}c*+Tb
zy##>bBzF$fk}?cNI!Osb;pA{>^Ok1DCrMYbmh~{X{ZJgBkD>D^|8Aa%U<QtBjQt@{
z*DGeSxw>bia=x;3sRCaGhbQB!;^7ZA46)V@v-q#6XmWC@>dP|Pri*`FX@&Z>gvo_4
zx2U}p)L}CF`m#xhHV!i|nRJO27la{GM>powW4iI{8s$+cFMw)o+2;%YH3c7)$_#%g
zmCAi3g7EeHZJk`lxti})VVI4}NMPW)xa*{s1U3T@kn*#L=+5+944RYW<)!HGy+IOt
zpWSwy=6dXUjBD9HC%p{U*x9b9sEN6BLpk#Z<pk%Gb*pJCons6O&|kLyFDK`->;~xi
zbR4T>gv8v-`OkPgmAKctmIn)9|3k^XR}w)^SbGjj62#9ICodQrkuex0kRUbZFguvO
zUap^VT~Qh3Lkw&*HdI+u^MGou*XnT0*x>C!g@2!hT3nldzXBUCTa^|QK=9whjX&1S
zhX^7_AhBb$TCQ${1a)$+xzZn2C2vviR)W=mch~LsBJ0oM>Yy1&TtTkUFp@X}q}DW~
zI^jAe=ev&>!o8$7w}|{t8jdiD5$bBU8TLnplB;=lg^k7QX^W2aIN<bVy~Mm~d%uE*
zb!nQXw=1TnVS~`H9CRj=W}asT4t|wFbqc5*n29JA@pHxBsY>$d8<WHOXH@eON-T&b
zWtq&(U6vgm8Jt~bn6nF7+tT^!_YeZeb@Crk-k{qc3xwrjb`BDB6}TQu?8M-06d~0h
z!mHlIFrupldF0Bsslo)EGJHi%R39RcR>cS(pbe8A5(xqaQ557+YCUZbN)sGxItyMB
z@j1tG_&&DdfKii<bn)3g85_$w6q6N-X;`Q<3Ct7fi|3TlAzLF+F-`KZy3L%wq+ds{
zb^whmvQzGQ|5@E26uJ+yj@531z|pNX%0O8J8Vm+&@`;6k@I&ktB<IoH7(DJ10cfcP
zB!qjpo2&tOA^YO%8Qz=}-Z4Gy+r&#8AAfyts(7w=sW@Fc_Hc2EA1;`XJ&R>FUtK9V
zAus(PirJCYYDkzsjb_+)Kq|Dk*_9;>q9Td#K6OYW#Yr3W;s1?F8$RmjiM+s}N$37f
zeIYm~GD`AK*+NFzv__^4<F1~uUUBr&x_XAu2PmQf4Iq1w_zPHK_$cDk-~xn$bOmPJ
zS|7M6C0GMBwf$!mbTqXK`5ZU_!A^NOnV_+nbG^*oTzu7W*DV>^pnI3zs@~9lMT92C
zpu+DG6o*qvp(GL=FCDGB(lDaRLsJ{`w|VfXwKz&ZCaly|rurH-3Ldm7-Zn@hcTLZV
zt4;X#z|UIFAd8}B?*#}*w7+l%HxhJ44w$XUpJzvx`^&uyup|v=Pne5R3n=|EmY(+)
zsG)E}Tmfn}ubZ|BQfIxMpwN*!$K>iA597=JI6W9fKu+Kes%(a&vo}*f0Uy9I#Xcrk
z)N5a8mR)P=)-Xq;yu37hj%^BT#l?(W$!Cn2fEZ8-zXU!9Xng~fI!3^z<~MQ)*L`k@
z9dy1#fPzA^4rKj|vQ{*t%gR+)TE-j@>v{AM-6l@PzMnmV)v+!|*J(0@c}F+Td6&bU
zMu8HSZZhj->?KVpU;xq0nkC)nm^wXm7;LYT)8x5>mPq2Sxt_v!Ym|wE?EfqMDYF{B
zRltZVb1Q=+e+2LtF%WIlv_S~XMNKg+&g!2LFl4L+kChOu0Cx?!VAo8wb+&SSBT%kA
z=f~>TQO=LWjhctFxR$%lq`eydSnQ*5Hu?ZlNQqlY{JEHijvy7gicx1JcPdtW<Pt=H
z{@%xz$s!SrwiDWx!f3)J*if}0?K;N8o%ycE+tbWIG#|lfO|CZ8AH8b?tMLsIo&(&Y
z9VZ@)29ytL?OSZe34ktTdml7jfCbY5Cee36X`*V1P@Zy>PP#EQG*);>8cEx8a+s(2
z&ry2yh(T{~_Z<5YP+r<lw-6JJBDZa<e_2z<B$+mUKFH<9hMa0P#Q(?SI8txHnX~?!
zN+fOi`@c`4d@Y<vPv9GfTOq-O8Cb+L-mkLcUU8E&_q4Huzovo`E;{C>uj56Fn$x0w
zWu_=34jN}x@BGqmde~-nBzWp5*47uAyZRa)*T6AvU`{ksHG@P})-}dXteR{ZfH*~9
z(yd=9GW-bz){hq0zzs4iwuXQ{Vv5<+aa;}8>dDY#)L~m~7&qo?!#u(|4a{?t3-GWk
zC=zhCthtEm1shuoY%CENq8W#MQ{%6XRO-0aq1}o@yswW=6{D!qk8m~S?r5e~{H05b
zu7NxClP{T29noeTb6yR5E7}WREn$FLqCnUk?JBZO^_?9uR1#_>TO>`_ZlBZY5%Jq@
zqk@>3)lut}O)|eKpkt)$v^uf==_8^$!hBUtYwaC1&_V~<U?$eG$O*IHFIFRn8id_)
zoiySlYgS7bw6K3PA+r&1UKRk6cqmrOUS;P390~-G-N|EU5SwQw0eJNu@RKl9A(C93
zv8%j?b`t*Qe66e2jqIv03#Z~i8X)sRPKnux@GnnW?}bn*kjHi%Xy3woKSnR?Ofacf
zH@wY2fT;z$CB;h(2H57)DzI&k^1p|aA&-=AM2t?!VTN8@l)!G!TJHux`H40_*+V^D
zvXiIQUaPg6tF-@fwJqx-HQs%FtBiw?yCq!D8R7Ti*Wv#SOk>D#4XK`h^0PoXuyLOQ
zlVQfGK#lcyJ#)&Y9XR=rR>_kDVNXN%Y11b6?u8|@TsX_(zel+wW%56wtX=gf#l&e_
z!OWZ{=pEt0KyOx?{aV3p%PbU(qZx*w8|&2I+6z=>vYL^)u1xscw!XFUXbrGBHtK(z
zCpJmDFp2Wkk*pt3(}i)(qM=26=qXJr*~tdX>d%P!2UNnis}Je!$8>2B^<Kq3NE%x~
zwQsD7g~cgnXE4^+LO#UD(Oaq~eU%8iBH^QKDI91rV)GNKt)27opHZ6n+5e;h>y>X-
z%;BD9FFc@ydOuCcz$yviVlW=hVCdbO*a_9(;IM>{y@^z!8&+y3X5hV06T4Z`g9(;(
zpDpWVjctS32IM5#V>ZXu;!d3BjMPMA&!;d{(=SqI_QE7;nfdH|PJ$kZC?k-U7-x`0
z8I9$`9@`<OO<~$>=SIX>*xy!9e^|lu3XHu-8|f9MAMqv^pQ{VDR8g2Q%B>nE$<Z>Q
ztdoXeO7Pyn2HaAEbgP!sU1BgK)yzO{X;aqfHtT}T^54vi`**1!rxiS)K)v9<Nr7pF
z6K65ZIm7zEXO3)i&eFiI*GT;^S<a4+9r7oMgoToJdE(uJt)Jbc(uI}D1-B8@_LkWW
z0RO60s=#KMy?YkCtc|Cxi!SMAQngFL<l}}Y=JzS+S0Fu;KSIDRSuAjz>SIiak_Fow
zbko^H1STC2-|9cFK*R2chD(aQq~Plm*tGvn#nuVJo@g`cbeXNPtu|8MrM9W({Zj<`
zRzE?Ii!t#KApej3_5OL{9M&$k68F2XJD%yq&hp`0cP_(Uj&Hs<pXnw&r+X;7GP!gi
zQ|Qa>vHR)FD0}XDa;Z#~!9Hkho2FT9pLBFp6;<B!f`R{vL_V2M<<tBo@)=?&Z^B?K
zJ~8lbkrGtD7LzZFa&|p;!@=J!uN$E^q`*X;whegCTM6e8qhP);V)(U?7mV0Mw|T%|
z6QlQnx5L|sKMD3mdAqsO=k4)|{Puf$y?y)+c>BEr{0@2ty+iyCc`ur;i#Xmb?Ag}3
z_uI4NBnd2tmkYb&m2lv6%z3#We%w33bGAnJPI{-Vfhq8DA-@;$aUq`<F+5y+z`S0_
z!-afa$isy^UdY3R^LVk_n>3#n@rgY=JK;@vZ!H(`x$%hiwrlALoIa^X)GF>xd-8vw
zKH1kue^kBddFBs8clI~#P=Y&U^N^uC2ckO(t82!4%6r=CI_T29cY5RY4u_)aN$*|W
zyDi7zMoEof+MD&xdhfA2N4)oX&+#^6QJJst-e+~(VO}@x1Z$oP`<@0kRcei`)2oF=
zIkvIQ+~fW|%o^xzZAhE$X_-&rIYbAE{;5juZN4p}@PN)SuNFu<*)wVWF@j!+YtF~3
zTg(6!gIKuzJwevr<t34>rr^{Qo+P;P=76}&i;>dH?o8wY2ski(CC+Rlj`1q-yu}AB
zUYJ#$!CLyNxcwfFU6yn4pjSdJ=n~|;?BPUaT;IdF+FdIkEMvdse2>IQ$P@6u@m#Eu
zb!?m`A055#!*I7+G2b116{H30{a#OP09u7Tj84D4K7@c@Pl&BU`9!fT$R3mig;pA2
z2Phsmpxf}kVexH2pIM~;7-hC&XgjEyJoyUQ7WDh1&hMyx#!QKKdVOYYx{I{kwhwhr
zFyQ}6ZKv0-bj$hk)|avQJ?lkz)~GWKV*}$(>iQ%j<L^_jpCBA&3xAbu&maJ=zH`3h
zf0OQi8^Ia>+ZBCO!M7-2NKx`uaVCJDDEu@n`Dgvdb>j@d*xqm}GB7+Z$<LK`U%NWN
zbtbM22ivzMC_}xnsVgvyUYJ~7TJ~R1h1z$Fw<h!2WL^S7VYRy30OMlhiO76x)l;D^
z9DDFR0@pduqXc*mEBJGzMLEREs{tNLldAtiWO17tglufXLx&rzrrUvVq}4xGvl-!N
zYkK4_$wa*2U^AEGf|7QcayIOpnp(oNYw3KsA`whDB%6Fp`H-wN_WApU?9QuQB?ZF@
zj7C^c>@vaFkP%zqV5@h8=yfHNtanM2gsi%XLBD##|D4|ZClvg+f*(`xvkGj4tzUjb
zmp`na$xAjSMa!Hp#Ag(mVLy%8d65}kLqr%U0~ef2Xn(j2Fx=k=srpN3kGDyG*!PV9
zTz~FP234}kYR|ZEXNs+SZr?cde0g=W12i?<T(psIG$SShC(YfeGj76Y6F%Bx;R!8q
z<m56L02q}pX<EmOX+7gy_c#DBpS{7t6{#l*p0X(|+xhH=6A#(K8cVrF1ylMyp5p7R
zeR1!$OwhmC%=(BDOj$3=TJA#_oH-BkUW9J)C&D%y^H*jA12uJs0%+sC4Z1exFwG5a
z<S^|A)Z2o%yHUIk$5;1t_KmreXe%-;!nXcV52<*PZ}nFQcs~R>gaAQ6ueP($SHzx|
z&Yj;ENl8{qgy<76i`d0>cQ(bd-5rUTfu{^^lGt=l4uNL0p(wIoMzPJ8(yl8=5tIC_
zTjmq%J@hVs_yf8pXGaMXUwt<KtvA>>Cs0VeQtqYdR`d1e)InS4?-)i*bhO*>w)$3x
z>~k2m%CQ<Y%f;yxcEBvk=?pNeZY>VvIo3=;ivz5ftdx=IFI-vOv48*m;(MNzi<xK7
zK6VU;Gmk#<w5~`uD6+b%?)gSe+K#sZN13{h4ftA4_&9M6dz^sR$zy+BOUzcBt#bo>
zjvCbqG}yLq(D}{7F-XG|N;oRfMN2}%<B1%p)h5|=`}I^3vS%1ycepZ1@|GwvG`pIz
zFg&{9Y(`e)h)X2g6<Zb#xX1ZTv$`<8>bEa;yqmNOlMIUeu;$|^k2r$eWDZ%FOt7^2
zvBV}m{GV43bk5Mts<|!W(Y8-llQ+lLChz@3oQE!^o<vS&gLorjcu`ZA31)Oh(r3@T
ztg%|KV<K9}X@`5XDcM$IxTShdZc{3T%o-k6{C`d_)%)@ZWq!K?!>#X7>?s9LD`@XD
zV|IQ)eN^D)NU|Ff69i1RR`+)H%ElealxloLwd<`7xJd$B7r+GxiM4rC=Z!2;l7=NA
zr5|#kAtwq)HJD@r(>|7(`=zf+nG1u+g50m<Ch2jxku(Q@VV}(K#JAxv5>RsS%vs)D
zq+Xbk3&U`rbrzasn0W*A->KRC1@ie51a->-`CAkkGJK|;2>#zHFf7tUB|lBsUsFf*
zb5oKyAqj03$R$!=T0Pv^U9EVyr8R9*%eGE?kZNz+X@)1}O;<{{WgFe5?imui!S1Nu
z2lkedGFo35^0&8_K&d+IqiCs2LMCnK2mcS%JO5I_Z9Cy-xci6d1hMUOg0{KyWxgDF
zqYh|oOIr`Li8MY#olK5c^<jxBfb0V`Y?dNI5`ji3sW;`G%t;S$PgBXcw@W+caz?~j
zE}1h*{>${s+Vo4BI5LTI6tIPqXY^xw4FilIS?aoQZ&Y*qe6sqO5v+Pjvic4qa~pNl
z+W<rl2aQl|8QM~5_27GKzmFxd8v%DxY$~WO6(_4jVLX>~-2W`SxSBs(-#)pzLm6;0
z)<*S4?mN$*Su?80;PD@5bf)Mh{{spp6=>f3zpoo(irE}D$&Kdt&B!v!#?~hIW^?<W
zQTktLSocvxB#_$zqm3%Lq-MG)#qeVw-6}S#hu&zg9d6gl`?J*BhTR1m1tbkjoFjsB
zJQ;7x02tvEW6}>MPC5NTkX{#C26lFalww&f<Z9e!WHYr4I9-@K#mKTR69ILWSQebG
zSRR%tgCLoBkY#;jB=GuEo`?cxFyaZ8A{ZjF_e1Kys|v0Wj1|nn(QFpOK{tI{QCkxe
zI`Uq0mA1T2PM`M$2Aw7#GIli4p&lLXsy`UdVs3RB42z6ZOuQH-Rz3ftYKOSh4j>e6
zH;I3XwAYd7(h~fQ#!n(%nK-dU06~qsRp`$&26uFh+3l-pTcUa$EmIq^$B1aSYVgkZ
zxM5#2#8RR%WNGNWNrm4er2tVBcZ(2Pq=c5ISutg0FUN4O%Xr5IoUYwB&TMRT%i)H-
zj=3gt9zk$1T#%<|IvS5FzTVbJ(u}VCIbD990O!T>e<v~qO4#U&-&AM**2W(7e~q*+
ztLp^OTxVr~X@%rQh;}>jrmefEs)-0!xe!D%LBg05)Jp8&?=#n&s0?riwT~PaV~b5k
zn2TXREklkXoeSg=oiE!xOv~)Pu3SB;_iCbg+8~rvEfV>XVsOt&nG0Lj$ymILp$fBj
zQG@vM4TeDZv*k%VtN-WfnSZ0e`t-0mG0XcZZ=C^%Pj}8nBlt{s46mqVU-vGSW761M
zESMTy5G0i;V01Ug#_Uvl^RC^jq75o_p-h{J?<FGpQ)mYy9qJzZ<YGU^-U8XT%aRUb
z9J`O=e4P}R3P_1^3h=P|09pe(4d-hMyA7i>9Jwws#(AZ9ik^X5uvMPHw9IxyUYsjm
zGF1hh5pk{^ZT^a%3*(pky((&yYZ7O7o&A=>3-md;)ZU_g5=<<@ESUme11nk2+SU^)
z(&&koxX_ylX==Bh!dWCeyfZ1TCzwF8N2e!(B;G&Vn*%gutMW2PP_{1W4SF**3j5TO
zo>4kq`X<O&PvJX^912y=f!EyXFBFQ=C)yEs?!n?YbF<lb<I;odko9Wkn(6hOdd{CW
z;26)sfHW;~F)XHIDo!o=@B~L2rHYsE=Y5V%wv}e)H?tpbvKD2veYdzEvfD*&9lu96
zU2W5HG2}NYYq0?Q9{N){wx=2NFn7LOUd9%%>ZZg~nh2Q|-PFG$!T%2mzE#1ix*vi4
zFDl`$3Gkn>bV<rz`5a}P_!ayM{(If%)?1bq$}=_G=!JcD<#I_Rgx3QPhm=a~oIy}`
zFoY%7B|9~X1ip${shmccORTY-o$$W3NSkh0vq>zst*dn+%oan*Fw59Ev9_EotMtMq
zpkSjt7ErLo%plDwyqYChdH9^;n)NiyeN2mCro~9h3U(W&IaVuosWof0nvJWMj%sBt
z$@(?Os@4tlNH&_qa+RmDLLtBU%t0rEUw2v=B}`t95^O}nLT#>Gn)2~;j$o@%M2V5}
z%P5gz%|w`q*ED#x-cu?S4px^J@PlbO+xn(u8+)NfqXsq(d*bOUVdCk_rsoXv?s+iy
zVb{~q!zG41sPYyZAr|%s$uzRwp^@E5x&&4tvkxT-2(KK4mio#l+madSU7aJ{R)_UN
z+jwgulE&P|dzeQcXO6v)a(l0kXrn8X4FIsjOf|QIztJPg%>+<1O6NC9M|@&t<rgs?
zl6m<>&=dK@(_vzf!U+Z541p{;RL|X#cCYnVN*Yh*AU%*H;uKj7APsl-ku=LN-Ps}N
ziPH3wGkYVwp5qtac+Snl)%ZHQnSks${N<&Px%Wla9c71OGn$KipNv;}t+grt@1oZJ
z-QT0NMx4?L=5^F6V(3=fJ8D(vj+uHY#vQ%)K<8UV`SiwF7oSP+&I8sbt?%rfqxV+q
ztG~N)v1f=?x3Ld4)K%|`%Ezn+<RIchZUhzk@92&6_Ihs+YwHc&alPU0+^#p2uihJ$
zbJN~XE}M6OVqd*Clq;{ioAkzpx*ENasC7Y~_Tr9%Sv13Dh_U?Yha1?TzQ_GRkjJd3
zRJr=a8mM}-IDNEu?r8DSSTk<}nOSDea9B2X)vgHp&!bX6B-{ksHlXKEzIYUu5Y6jT
zFE;CI(jpu@(AAO)b;8_&bIZZG9UkV;_@dQzeXksLv5k=p_HL8W53|;teX#)=7|Ii*
zIO7Z>oUd~kX5Leok^6+y8AAV-;NKm}_5sCAhHUUo^!3=FBdSH<_%SG6KS>ZKR%XJ4
z$9JlVH{EH!fdKu#Q^qe5gk5?@1VF>gO7uu2`d`-lo;nHOUaY$%`{_<t*=Gc~>T~rx
zC}=4Tr8hmAG!K37LB+%Y3@r}GUmsVai4=wynXepA9!>0w-vOb&S!xO~*y^6mpsF>;
zJ#AXXUiFW7aG_9I5UYNg7ikLqKzHz>rcfKe$QM<gGmi@LAXbngfJurwJoga8ND8AT
zB-3UaE<-0wH>~V;5zkk@i$yD?<Xv19;FM(`)pZ_UDWuCo(IcQ^5b9cYWs8<KDQ|}z
z>#p=R+Jm7|<GI}x+_lPC>U=lS^pwAXv;Xc|FI+|0Vtw!-=V?*DjH_@VDz~b{>*Xyi
z5p?6l#=YOyq1LRwABPIe5BT3ni9vBLNUVDC@q>eZy%8<!XQAPO6qRc#=$lJYJ94Zw
z%3FH|{Li(QlrJDJ{V7bgSlnGf&unsSfPGB=E4gsoN=pWVf%Pri8x-k_h&AW`hD(uA
z8P^&Acex&MW@5nr3tU@RQN&GRd(2i}T!2GgN;HNyN@>KxBnR2x3Z~06%q94w@&c5;
ze5oLPEM0a#xI*{0Y6XlEjBRz&S`(Pty0I1XtD61dGyPvzU{V;<Y?(gmCzMuZ-Z(Lj
zF<~FqrF^~kf2F(Uh=p5Z-nxSG9o9UwD{fKdW;j}B3nm~foR|5*#}p)_oQ~fitLJuJ
zI`y0rwEvAN+#Xp#wTiGD*uZAd2`Z+Y2wz^!r6WR{+1Rw%@FcD~eqQUJ?VxhFQF1bS
z|2NfivD_RGFzx?aT{U_tOiin@q;^}UF0<LXQ!K46tY|&U0=O5`YpLoaxF1B^s9h;e
z0-Mo_M#>@gE5pQ0O^5N7Y>C+GnPlH5WyID?b?20Ria|AP+iZOseb*+j5`CxdqkLcc
z00CQT_rsgVZbW<TYHa%BI?1&Vziq3AYlmT)&&;EB!cUakILkK_HriJcI$C87`)P6%
za9S*9EtVTS6f+e+9NF<#^+3mwJVif@#T~$YQC%nthnwBToHY2+S(qIEfbL46=l=mg
zn2RJ!<#147Qqm@!$islv{3|A-3JVNfv{S@nB~sHUUtMNsW7l1GNii+`-=C`fc9PYm
z+V^Uxv^x{KHUO;rvh@<%sQue0O1o`u`v<kI4k<8`Y8eK1JUiUEvA&W#BAJ=ik2V^`
zd}wLZ(aw>7qe^Zs+!4pcAR6c0*aHhVQ1_ca<)+h|Znn+F7fkqBwU3vIfo(RZNjK`(
zHkAN8Me(X6TmVAxY_`>gegF7c5bx!5PFVA%9G~vC;dwl~&aO-mmkWNceA&z(!^CXO
zwb)kd6MC7B0WXvG3u?PC9{a(xns#*dY3tUTH2LbEE+S$+L{U`#6M?!=A`4>)VHHI1
zga0Jkmo_#dw*zXt<x6gdEqTM7Io6{bsppJwi0U&9yY^h+kB`+S9YAhg<~?zL7E+bp
z>Y2z)rC^L|mkeYul@w;2B?Rki6OOhv2}{rS!sr%g346Sx8OCr~M$0NOAeNZBl0YMw
zp3Mg7Xh&Nn+R=s`qsf-&$C&N=vb}9)^$ThBQKcGP_ScFTt=2>)w!*(gqte79rXMg&
z@jJTlOA5?X_5sEADX=|p(QcZ%w2+K;(-7lti^B)p1Zdb+u=Nq2{NT7+sW!P}Tjtnd
z)+qoqW5r+5Opw&@5cCxOjg3E*;e=dX9f+@P?VK5r@kyJD3+V*vP}baw)L<b@$egK+
zSR8}&Dt(S41@*BQN@<8jyaa|>a*Tj$k-6xe`XAPc8jQ6<fE8pgyW-HUAXPnuRseAX
zGF1r@*mt*2oxDduQ2{+72~P(aZuB)hB41K<RqHeA+glU?p@9d(LWBYZD+ArzrJ#NN
zgk79oQ9q{U_tbl`ropfT-X762m~GgO7hsdohE>&4as%AdDtnY~>n3$DQ{8b+fU!M2
zLEp&~SgAYc1qTo^w6bo?n`^vMmq35%GIBV7g8*t&`%Go(jqyOVRl#Na^EzOVAC1lQ
zg10N!251E#8X_s_v5ZlK)z*#<lXL?o5OXrp!@7sa$V{7?8|xwvhzbI~n}{<d=mZGX
zfF$zMxSsbiH{$AnvFzlVc;yEF_C)0RjN@YPRsYn}xG(b;&2PONqM2(hcc2T~+>z=X
zo084UQF$TLL3$pZ%^For&DE;MP(*NY<YLo}m>oG}l}=Aq4kN7Oh|FmJ8}*<GoU6G5
zlsryttax&5W=_>2bxE_LI#-IF+E?|w<~AF&l&y2AlSSKkBc}HG@fY>JD9~3p`0ogE
zRlNX?(d1i=*NAfB^-<cGW?i;s97MZAljkAE3tlxXw>hW6EY|Gw+9`K^x_*7y{}E*?
zIQVAZ&W!yH`We}rR9#!h<PRv2iN9@2sADd((zlHewg9si09kU>^T%Ywm71IYG(Gen
z$X#IOF{GWVvEY(RQ_J4J)!Z#mypY=o_!4Wly1f&8HM6%NWI->@T3dnFh^+RTP#(-6
z`VhPpe;E<Ium~6^i6JM(;resAu=Rpkya5HcS?>4Aqi0f^Y4}AWejfPlL@X#dI_^~Q
z?|?s9xFF_sa`H!8=l0|7EX(>3yW_2jr;a^V#0+uyO3@DHvANzdEkF%y;oRt}kIc^7
zUq|`o#E2$n{iuiPgT=`fXMj<?EEq)mT3hbvTtr81Zm4L=$1G@CN6MIIK2^6McM!hw
z)TcT86vBRHr_PYA)(^cbyNTOamXVrb!%w4RkyEBP2s3IHr!LK+XOM2Petc(3%bM#5
zR*MsjOgbc&M#bmdmNv<KjuuZHb-oZeV9ZCS!Q0v^a70GcwYjB#9X0R*1IPYXtqi<l
z=mb<7qZZAx5IbiaR884w%j2iX+}s#0{%@Im@<8@~ueBs=ps=r_6GL_<Zm{bcE^Tt^
z>1$dYGJZ+qiy*2)39({>w+e}k7F3%RZ<Ch@>1wr5WvsdkJ)%{>?pMfSwpHiMa30l_
z`Kys`6E8Jjc#gz@#)_mR!uTi*M2q2|imL+jU2?6<>;h!zhui}z06ZTfGw4RvfbdTq
zdnFx-(tGj!i!b3WcDQbj?r?MtTd2v?$31>&7lzs{4Qf=fk{*70jVjGF+6uL4ZS6Pw
ziZhL%-`YA53}Et2UA7JGT}VOX!B^YY`k+bWF0F?v4&4DaZ6P+m;fKA!5Gpv1w(+{a
zX}Af2?i?h&`L?=>-j*6T)`t5dI0H#S_-|bsUE3CHvvzmM#<AXJsOW7R%P4ut)?gco
zyuuA2$=dc{yAEkw8@#}-Sq~yC)!Z}uTdiB+*j&!RHE^uB16h*~>#ZC<!$C1oe4E90
z;`l_y;M@Jb_Ijbr`qp+M8{!Zmgv)x{t@GRFcg*ixlT-9v&}lo3uHPN(;ULrgm*ow(
zxHcN>THh1w#%w^2Mv8Mzep3u~nKP@uPaS((>KLSsy)r$I>e=h+85{bCJ1pr2<~xTa
z8*D4;ZZ%rA8%G|GjLP@7vL0U*!RE++u42=C+df(wX4`Y07H?X?{R)hl+{Ay%wt92?
zqsk@@1IdAz0B=)*BnQqqXq-!E+>i64BhH{)5gR)vK2`ss@>X=KV5fI2OkP2EmNxyD
z|1Q#3vnM>WDjVDD*p$zxgqVNFbi7IIML}tTNgTF@{>)AXYeFX9IGYZp<|Wne5f#bn
zd^YDlN;{~WO@dUb&e-FJ6f-XS_Z0hG1@9&>Db>Z2WJ(#QLUYz^dKv#8DFEHZ#`Zbk
zwQ)g4LulKc^9pj0V)rVzTR~oBMB-0G?&Tt>x;den=yNXTqBPFHMAyh^g3dai%Unm>
zmh<XFU(1EIINK)AiHucqdSaxl9`o^HOxGzr{X47+%@sqUR`s<N*x#)iqH_GhYFoD}
z@3tH3c(q7PIUyxL-O$KfA+SZyoWuCzsz$8+M8*we%cMG@ip4HwnNManU%+y<A8H<Z
zSs=bl9&urqIA9b~D$$=FMYy87BdDHbGM4zSJHObwR|31#5W+VYLWu|pw`3`(lhYI*
zCBD&azP@a1W5Tp3vHGVjV1JZ^>VG7_r;<RwWF8PZ5`P0i@mFX}{eJF%J}?F!F}6M@
zmO(m1*z&{--kE~Dv4&mj<Q)sT06SeKcD6G9eUz47JIVfHA)KCy1o4;~!f?JN=;9g|
zE?0H_Y|4K}(1YF^4W<mk5i)-bIc1mzk^mU9znHLBqHof3XqTiF%^INYUjsGp{PPIH
z2LRyR(mJu%s2W}O*tOj=PnbD5mU~bG_b{(-5X5;r8r5{L#jRzIrH?JZ_mc_d?MrMT
zW+Mwm2N*jjqx3iacExN3-JuG#S25bY(Uc3g6(Ojk9U&&zZ3q#ySzw?pU~Ha}i7zlG
z06xiq=X#`!9e^)bn@YYOMN7^P4@O!@9h^5KxgCMkaNM%d@7ahtbd&uD)aX5GbUQ0;
zyDGm`^Ykb=QaYUv;V<s~m|X>~Z2tygVHv<_mTJ$3zR)Zi)H@v>IE{;G+e)^TM|APE
zDJZzKsT)p)+y)z(KQj6@XQoWPZl-7+y+_91%hAK&UM5h^za!{J^*%tU!L{K^nkzIL
zx~iFSob6X>oXp{Xb4@ru$*B#G*G#g^baGTcVr>f!XUvtDWXrhpETA8Wa|pAP<}o&p
z;f$u?(sP98FEQb??Q3w@&YOUP=774bv0#XU0n^YTal^DZ|7&lPZ`kf{S_eEwVVpVG
z5HB^Skw2li_PhdZCHG%c@P2~T{%74$!9@>cDMzIpx6b#^X~3I;vc0<eoMysP#PCdM
zk}+cyizI}DkCg#$m|{WuKgW@uvaZ;n?8aPPSL{*c%jzkBMkg5CuUZ#%(*)>av<Eeg
zB&wUf!$D<lf03}OR$hcXo@DNqrNG2DcB%f1#!I@5{RXwwWIXJchPLV^xB$8T0d>e*
z^khRkN0inv-^WP%NzHfFu|0){8v8@d@ugH6V-RLP^Pbs1h!4{gFgg|AiHc-1NE`W%
z6mMPssJFCvQj|xvm(~RRRcRQdtpfnA4d4s`S*gD}`W?KEj>efptP!7;OC;vCHl1^<
zbV3FkDJaX2#nAFeYk5i@23ZwQ?B*DC2>oO4+7LFt!+4ZY%`et6*W<X7Ncc~FR9<b&
zFt6#=rZQ~zkp=qC8IJ(qt>3MrRmXLs<3ARUwlL1X9S;QPF-==otL1eBCZ;I^=72TJ
z!QPytQpvQYLWg(j0L)qs;Qk{)9vRaBt1lmH!8&<$ZG?B|4F;j!<-jk^Ri9DsI_9z8
z|EGil^Mjm?lC0i{B+5g<R=9rNQ1WajQ_l3b9uG!q$@P)z(6&(dN!klt8w`=IdPYHK
zv>_P51Vjc8-Cobj1>+sIMVi*Xbcn9{hwd3@K6AlVG|f=+b~I+M4iXK9%yHikE$xpS
zZ~WiH2fRLWyYaWd@H(7=X-8B3>&VGQuEvGLpw(saMY*dWwF}pY*@_?QZ}G%7%}i{3
zm@3B32}{3rh5!s3dvTA9S8Fj|iisd$Zk0|@ygp2oZh6saGJEc>hQ@gV!b-U9F?yut
zrai`~U+PuWC7D_;7<vo*@MF>;TS7+FNY3{KR?ll{Z+C3<ZJn|m`N4u6u5uuEgw+5d
zEi`<?w*_g=De29m6BSEYi+_vBW&eYEbmXDKD9h&a<sIm+^n6h^nOGuTyrts-S0DHv
zZbw`9=&jOxt$Rj^*byr^jY7SfqGMK!nXs;ME|#2jcAg3xz_H$vg{?G^)?1^o|Nd4s
zPWFC%<lW7VEk04k{?VYcxgIudCCr?=GKJ-eaYx$7BD$jmS9oe#SwRz^8p&f^&UW8Z
zb`u?HARNR$MsGV4mZmV@j5Kq~a$I-6sMwz=W=aw<A5Mk(J4&0^<+ms>jf$}=O|^<Z
zv1{gs$TGIW;S3vbLrjm5>MX3Mv@1Y(FYLyAs3fCnC2=-{Tv8jr-|YP~c|#6eI2&e7
zz*{=&Ueh*p9n#Z{LpZH$Ii2)kyL!Yh_|K`Q?$Z+smru&trZy+Ps(QNg_HK|X%A}DF
zw`@A^|A4VKT}&#iU9dXP(s2Noa6`jO0WbWMyLBX~jRuLt;4Qf()>&LS<e@u>k{&<W
za^obQX$8ud7uI;;OS!bT158yZ=T!A@0|)<Jt6*A;X8j|QOrvQe*{A7<peI6D5>_Be
zUF(g|Qy(^NLP6Ui6vR${ge(0ZAlc0IqgLYVV9RlklAOgX(GUpfI0^*f+EN0o{!ooA
zCbc{Wr0&6+?IwmmT2i6vooew#2sY}dzSDqoLB715rK1UyqBuOti*|ny^fI?LMEkl-
zI5ZUW5=RIHUt@wI|Kqf*7j%;Yxm=2nOMfs7azW)NcU+K&5(mYDEg%=$_YZ1Hnt8!6
zT!j-jfhzERl;*8nD?e+7Hicjq^f3Z@fYg+40?<dlp^tvRe@}!yewujJ>o)Z9^JY%d
z9~syDvJi%$k1qc=Kv|SB<Wa&%=Xb@?7|GZVdDzo}MEpB7jsJuZY&u_c(u<lV=abw;
zq27cA%_YCpt%Y?J6)n`V<P8jLw$N=^{zKMyL>D&K_(?VYvYPsc+OT4cR?Mi}OS-h>
zeNnM13Jg$LgWFfIzpA_Ki`Ng7X6U1dL+;{{1)zL&oadx6&J@_rBIk5+nZ{@>?q&_9
z?1<SS*V^CST!XC9ujugE`s%}>_)FrKHZvq{bF6Mob*9Au)jvYPy4JU<j<pFN^KqKR
zMAX*|5@u3sg^0epr<Q=eW3QIKH=^%gx$6|%$Kud;y02|SCyXq>bi97IgT7;tLM{|w
z(rsI<>Du$?`0`A#Ogq^?FE*$Xf0nB^qAUD!>huPJ^yhThff+?doM8-nxfWQxs}pWH
z%K8n-27f`-id66o3^&JDG{WPmdr5U~5G>xl?%R?AzU1pIZ*)PI9d9ID|8_mq|NA|)
zf!aaU9gD!wQ~Uf!);P-poTK|DGB#AL`HV&CA>FryWi;DYuhVgV6Y4?hyaqNi8R46G
zm#PfsbK3RYf~n-?K}KQXm{vzNAGtQmgaVah_&&wAdX4~qwoiI=?LmxYxHy``j5aT1
z5FCla!R~7Y3k6RG-f4-Npx))^6+?9cgb0}6*^v-<%<xYoH_O8;HmJhh=45Bj=#A0p
z=mlGP_Fi>mp`$A~$_~9&SWxK1k<QGg+a+)&3fsS2XMRMTDRQC%=5W-UWs^enUFz6<
zJXat?qCk*3C?ABc{f61Xm*pJ8;>nGw(`Geos#A5ZzJ|sBiU{2z+<P*>TNz#L3_<!}
zF4;vGr2s;zC^>XS2EYP!0!k8@?4&A3*g}W0mDD>>gldRTtIBAkkwYvN3LA`LA0>uz
zhk+nh!-T+UVw+u;+(+O>0xK$e%1o^rvslthnKQtSDRvXc+*U&=PlWZSDHTt@&&ch*
z@^s#p3ktv#ZXt}O9M*!A<{XjOlDD+ldrG=O=ZBJr|FXul&+!=owazm1<BI*2M(ZyW
zSWSjlenOX1+zofj@~>@Eaht8ElLdZJ^_TzzdQWPBLrO5>vtV9x%<|RWt497?Rc%Ph
z)|>^RApNV{eNj_Gh1n3#iW9vkNt+pNnk-J3sb8Cuk|VdkHpe?x-&bkv*X!YO@A~R#
z9x~leQ2!hwqJ?UR=%;8$(C3@rqQXv9C_5QV>mbgA^Lc5~UEA%(64fG1pn6J6HMh29
zZDegLGib{^j#ps6ogd8Rwey3y1p!D3KaeB*GSZqok(u`AP%R>W^R$JN9Z&NG)Q7{&
zzpcSm@@5~3RnC&y?d6z#OUx!c2h%e9<rvZ#i4D0QrEtgVLT!m^ucZbVe<Kp0hVNS4
zxp^I4?;F*C@xeffnR(?qoe9GGVwL=*7;4I`Y)Accy`>z7s%sfy4<KBSN6kLowHxuk
zP;C%zpQ8aL+kx};iE5EnaR@0??ENN&K+FNv$X5O8?;w=mp4V@5<A14*)TBDxF2igO
zBW`byx04!j8@aZlu-t(<ayxl;N;qNa#a+QR>e%58k#lD-y3Sb&s0er7FxB9NUnFla
z7$xs$ZErB@4MVN(BDTxhVzGU}u3*%Sr#AxCzbn|qyKLoMwEN{75QvOkjejFTaGt*(
zrECIeJbN9zZ7-(*+q>8Hv?)gOr#Nq^db(9LYOUGL^En)mnzvJPftsHt{XnpX^u57e
z(sz)a(-UjrF%H!ZU)Axt^{XT9iq`C;HCr~cChqMr1?lGPS>4#4dVe<BwVUVnczf(n
z*gv9eW4viERR2C&qxoAjHSK4DeTck<y}iu6eV@`@yxq~F{hcjBo3YP3@Uomy>|Z-z
zeXyTr9Sjcgt{iad9s2InD~X3V!Ma7Q?0WC_+gh9dWWxU_!e_VQ^s5!2UsgminMkf3
zT06XUgi*eO79S3du)8ptD{Euk;b06CYs7xRVebfQr4Ss3gCR!BktpRzJ!K|IA4K^n
zH8!K;oX2>>J9tC1A}upO=WS?yo1a_C+-YyRd+or6_eoSAXnP;c&E54DHhTZ4o=ndi
z!&TFsU~JSq_xL~PxhsE<=bqrXUG(tJ@s=lf8tNk6@=h~8I7;l8zB3Zzo<c8`Si58G
z&Q0F%v+YlsKM>q0ABC)@qrp+$>TaHr<tcwsJHCFRhA?>j<PAt94#G}ED-TG({oieP
z#=Xb87tgra$H*0V#=UPGXY~F~A(~wZ_9)QK0w2iFekH=Nw_RQBALDk|E!)fb-g9Jc
z11dA5;D=T4K7v=b5g-wKpAv*{!>)5Hi<6vEWcxa{v~6doSLKus-ypmr9PCv0A5!e9
zf@=ysK@jGi6@i4T@MgM&RQtzj!Pn|pNp{jVJ;<)^sJrql{v)}tvHi}El}&Nrfq<K2
zaH9tQld8*PNW+ebVh74e=K)8>P5<|L51Y-Z+TUL}HtBJE$#jVy80ey2`dg^s^+WZB
zI@>7Q+A=pYhl%2hUoIDCSMZh2rt6#8+cqNH*-lW@i=bn~1z)(m{ch{Pu!q*j$ZBbN
zX4Yjdy`DUNdd6vcgxQa;_BXob$;vYR9cwTc9sOZlp>B1&wYcnN)z~?zlypD3!AUTY
z*W}{6(6{JTI8<+jG%!)-*S9qCwTLRK%q0^V7*A5wBmYFh-PbBFdn1?2{x@lgG-)Mc
z_`bQ3W~7rCzW)r#;(rT~jYN@*7~jCh@+s$@R^)Ssw`I2>-I_md<FCL-#VN(KaGcLQ
zaq(g9sq&S_eIFgI=qkTYX_?wm$)BBa<ff5{CR#BOhG<16b1|#UCiQ9d=?~}?O;q|J
z<qLKBe=11oQXA*|uhaAYwbG0s`~$^Cbwe^eqbHq_$eQvsxkoc|`!;3%3-XN}bJXgT
z624u*|Dr&QkXv|oT!ih>>iZ6**#bhf>?Xp%GIOJ@y*8igNJ5p=F}&4q$mN`ynOTu#
zq2=|zotE8pH8`-+gM&Bo<c~HS4WV(XzfZfvoNlOX&L@e{)jQO}1A1MfwGFUjSDkUM
z)1pc(7?NZ*A~;HGv`4-qoG0*})z%jZ^BhI&7M;11VT<N;*j1nZEJJ53+tbLUrRg-4
zUsP-UT7ikNBx>etj;0vD8l4&<huSv5ZCcxmhN_HRWJXstXL>UTr6eAX{`$YvpEj>q
zFbBpOz$zT)s~kryC6!1Zppp+Fyh0<?lKDu}eYW;>6{t_g(OAvIQR&~>eXGE3c(*fb
z1oOlU2X^RABHsO#-M0p5g)>Bv6K<Cdm-^*GdN98Gzh(0K5R#oXCPhzEZG+YVc|U4R
z(VMt4_*84Jo|urg1@qmp!Bn`pcH4wAS-M5*$N1eZ5dj0U3Cy{5(MB@3T!|Dq)3os+
z<)4N-!Nz1hi-BDH6?A3b^MsYnxw2Bfif1NQGeu?dG98E<Y9%lcX6vX)Z)GkV#cOz*
zMZF0ZV%%^ly2cvDuJz<Tf7Kjs^<RutOOis@21v!<E50JIgi(8HgK)r^`5~QtBY(JF
z{!`kxVW;qNz>_<b5*}Hi5X=Qu4{}Dg++;;&@BnXfmt|-e<6h@ieDCVshVNhJ8>=-x
zU40Sz;qTIfJwvcse70=f@{23W#ivbI_w+NxYUHHShpomBRGzbIoF31AuhyL@lagV+
zkqwoJyoAW5Tpk3;nfI0HD`Ug{ck3}GzB17I+l*n0ldccwa#F#Rf)6YBkb>tG+^OJG
zs^Ap`%euUvKx#XQ-+&JF27vsG8(&h+?;)5dxXvhH<>mhqC1kaGM^`_kH?Wmxfvv=K
za{O1VL~+c{%evN-JN8hbW350rs&~62ag3VOI!%2U2dAu<jfTD8b3Em>yK!&ZxS{q7
z8QI)c=eVtsZgxv5Qyr}()-iu^n>OGv@DgzY$9CHMh<mk64Nim;P&IqmwgG*K`3hHf
zdXQ<$u<Gl&q0p{(V7dz>&3%)n$O3QyQVxG!BKqn!kk6`XenbgUARukrr~WitXrp18
zLc^4X$9_iHuPghb<l_t@JPe-p44+EHd)Ys=Syz)b!katKcz}uB04UjCY?LxOv^vIl
zIrs38GiS%gtzBo%_&0bS^x7;x@?x*~=-wf7=FB*MV+k^ST8)>0dMtg0WUOuYA>EWC
z4^f1q-m_$n>NWC;3qE6AXU2J)ZmFDelTV5gFsNxrSkC$%)+<_NXV0EF>l|z2VeqJ%
z*Wafmd_O^$E#bX*nhnP<s;&QA=>^5CUq7SE-&Sx%X+N#l&nWn33Jm%Gnqq&Y;6ErZ
zw=1_4>sJE=HXZeFQkTD~%d&#!6}(AlZ&vWQf~OS7aMAw(1w{p>q%{&ka%M*u808^>
zl`lkPj`aO?UH+ghn=9zM)Mi^oH@P9TK=rdE<zgU}OdXxoe|!Hn_fP^oDd?#`5%2v<
zE{~u4d|?Z|@w<p+bG?QB%-zb}e=Bo0a<Q&V0U22m|M<PRfn19F!?CZ$I8yibxLUiv
zlG$Cazx&IX`&#0s8#x9*NIkfO8ExK;eJv+e?H-K#YWFlvZu<ZZ>{C6-Pu;(M(*|QA
zDZeabWR~R=B5aO+!hd|){IF}oT;89fSey7(G@a0yexnm-@wXel5d$$fCscAfh4)5T
zw3(CeGzf;JZc97~8rdN<VqPp$<Kb}C;H)&(r3OE9M#9p4JiwfU`=%W{z~$=&XEz{z
z$Ny%A2#EQDve|U}QQjYisU7u|#2`6up?F-XlXPZ8Yuf(EWn4I*o95`-Jck4KoH2mS
zizbZKllE%iYu;;*xhMoS4sB~-tLJ>S?Unt8SLpHPcEdKmJfe#muycX7(?Li!&|3E(
zT>-|)f0y%sWOuXy@_@X|1aTxb$*s`_bkq>aji<e~+|GIx3Zi<KAh3MCa6e}naBZ=$
z`q2$*8>@L)vS;Xg()VxnjMh4=G|$^m8pD78nv=d(=Zsd^q_vyZM5!E_?`BK#)U_Vf
z^^LBs=36w|zoA@#azDx2a#Eh1m0EvK$yBX&dF1DCTlCU4!3^li!qBD~Hn+d0`D8Qh
z7c{{VE@7`yHRF8ELz7UDSE+FD&S*q84GzAl+I~@&pVxhN`n)qQ)Yi4;bTB5}->;nF
z0m8)erdw{0Xu3T{NptPEtMGqG(w^F0_=3<l7TNLCaN$2FjaYx-Z=~z(W=f{I3jak2
zpIEl=y6KElnXhn^RDBWDCu+1q7d0tpfulCn#s<iDOUsBKgTl9kngSF>wN!IDhYPEn
zSqD70Z2L4Ik~VtQA(E75Uh-4UIt*xZtpH>aG>T26%4u!m_67}V<vH~ObG?bIOXXyV
zo_vot<n?<4*K}S5n~74f38{sL!${%xo@EnOG~OF8UC)e<B{&FY=~5(^kavncXUe(2
zA?FM7eIaaQA3-|3l%SqC<?6iz^02}pN@4nzmd4M-Yze<tQ<VkDA~j-LkWWqrV<nUl
zk<}VArU|ruoTr3aTHOa*tIVJ3ztGJ1Rf0%O6ZSV->_{`ek)_2^{w5IXDGiJumQ48t
zvGD2|nM8N?P6KYW5qf%e^~0llU%Q(CHCf%`F0h~X+sxs!JnjOj%(*jfe={+D=BD%9
zY#wZN#$SB|96%nU<_?#({vT^O834AxP_#TVKg&D4CTWHgzIC6qtC6>d4R5t+51(lR
z$o<qMplp~;U?50W`=W7{+KACcb(BJAn&H^y!Nwi5!BQ;Q{iLauxG()LTNXsvB6LQu
zoU*-WU0_8i8rgynbYV{lV+RP1>$nyjbcUyTm^7XsFFR5!Niq6%RFfIAAMXb6XZiMc
z*-x?2028HYa^>0cC67IQD2FoaCqwIOcBFkl&kPH}rAC_y<mqsirj-d=Kcd)+>fiTk
z8m7*^>#=uX;<$%*Z%PrN6m7a1byn6bn{?_Wr-n43H>p~T(Ed$Ts#)t4jwX4!i_5Y8
z7U|93sPt}SQbXKXp_&4>6_AAK%P6ET;V+lNj$2l0n8&`(^qf_RKQy6`8E@uaW<qPA
z$a7PP?YJv}c)O)zALP7*`2!t?TmiWND+3NvK6f_EA!#gaXLSqd{OWJ$O+<{hnX0Kz
zwBEY-R$pNC1U)S(O04`%McFSEB??Qo0R?2OS!Q!ACp+q<iV{^9>Io-WZzw72*G@?}
zguq`)%3*$`q?F}vhTpCH!j!LL%I1ztLS;EBc1nyD%54nEbg}_CQ8Th5Kr`<1sIg)F
zy)M+6=M-D-zs}BP49>PtTQ_RWCrKSJrR2w8|J4$yAobhVP-_lEDcte;QPm89f}1rQ
zu^I+ZYu?<v25;yM%jZx*Zb4!B<MlSxhVA<EtRkKQdpY0@^Hx&L?XTtO=WW4YLxH)y
zCVOrao+vOMK<TI*j-!zR^IzN5R&R?Gn3T@cw$-+;lgCt;Bd9R9@+MrPvSeQ)D$L!i
z-W}xG6QIyUWf_|<N`V>d^hU{vI&*z5twJ$*!>KTTo4or?-MOuHAlT+@10L)qw%glo
zv4g?xV4E9lZwE~AZmv-j@-|XozJRs+E>t^Cg?R|;s{|%@Qeo~k73Q9`;s%vYy!y6w
zl@6_mRF%}chnl6Te1TRT4vM_PzCen=BI#Y;UQ;3dH|`zbN!^f<`)%GHxE`bTUierX
zYp;FQQ!m$0d0m(JcV6mE|61)1cg?PpEy_{-h~dZmz_0_?;}4-)y|6&({Z{(>8>P3;
z-)ZX6SCre;(9GY+<(~1&k!9Hd@8Go@%I3xsHMZ~{&F5Iln%DH=A$oDYVeXb*#GLVf
zci7Z_2iFeSxE-WN4x@@3@Q%<QW4Bdw{%u<yOVN3hr%2i8)NrWIm>W{T9oN^?aCe#-
z?#SBMhA~T2ced3Ro0syF<HVQ|cLaCPn&XU*e4CwsS5G*Q{=0&yz+$IB^FJ!+ngTBR
zt0M{p97}I)fuGga<Hs754#i;O*z(QyS-$417))<UJ12yQ>l_c?*~xVv<u_6EB>`C#
z&owONHh|HloV3MguD)N^(8YDV#K#mgG<yEm>+<6Yq|fueLBY!kjDr8_T08#}%KVKA
zJ|bN8Nd<!2O-$F;4`eioZ%F+hRYl#rKAMlr^2oft(G84&{Fj-@%*})|pD{$&BwS>9
z_j&rrS>D}q8$`2t58)Wt&OG+NRWI>v3cf?ZcPjXlf>#u%`%Q7>e-E*-6MqktQJ9)v
zVsqZ2N*byEj9x=Cpv5ZQ1ZY!AvoZP!^`FSErV7c)RlcaS|4&p$CdD!Rkr^=koE~J-
zo`%9`?9LV%V-pRLe?zZm%AxO7?E4gazXGXu{Ld)(0R;`+jsJtX{24v!xAew46dX{R
zscn8wv45%H4-`m?<KI-wIK8|sKdZ|hQeg8=C)D_#Q*fQTIK`YU)fSiiA5qGWD)?hP
z>&Fy3rQ$!X*iR_1S@frh84WCP*0Pq{XDQIjb};Eq67i8+@_i|NN)8WAVgThyuda1D
zDq6?>n%b7#nCsUMjsEJ|7U@+oXiEz5Lq<7KmggNI6lWm~Osl4)Q-6Y&qv)uaHvkJa
z;Y4*1MvH**|Ilk|wT`!lCkvbMs^Se=1d>L!n&|Jgtn05(;vMQqd1lJt<BYQ{@vWWN
z;)HC&U-6Y=nwxE(baYj1XJi{kSEf8DF~n+)sTDKZqvI;C#=IOm-NeX>jm0;S9o5*)
z0gK(<px5X18zWmVMz#y+J`_eywa&~T`CqA&8=UOwCkn;-=WUbKxoQz!9+xOHGq@e8
z)QU5%z$A`qt1Hvz%-c{yV_sC&@^bCmX;U~iYdX6m7wS4Fpk9GgG@{O_26stOtz~mQ
zDYq%(cnq@QHxsMJjFCI8HaO*}yEd}9G_|laeZG49B!6zlew1BG)6OETB~G4<Hu%eG
zV7zfEA540+je`@#-Qdz+;v&edqc*eS8Zka0j_T~H4ra`Yf<cs)fq&0N{JZ1T;n+%;
zloCpTrDGaVgO%9u0mjA{6onjiy;m^_dk4x4GZ?jN|6wSFf_SeRiI#(%IHB#EXosC(
zxH^UhDCg^`-i5`;C(kTf&^X-kTzUHZGLX22Cm2aH9WZVP3&NgyIqQ1=-_RRjUM}nL
z8oyZO%s(+gY>AcWuG1v55xX@_-Uxcso=XIE<6%|f|C#C<w7MD(QKkN0=<Z)Cc%NEi
zfU)sPoqU$0qoG!k5N3B5{yUL1)c>{vckGP!h`&11Ih`A#hcnX}#rj7L&heG(p<1I6
zcuqmAXPgq}yLlJ|2Pvlvg_{_HOP84!M!fY@`HO>6eI5Z|g5LpIZet#dp|F2ar@3Ls
zo4J8B14<A@=9^;|_TV=Rbt<EM5ssP|^<KnRErL_6R_fRek1qaRvqE>DmohZSqoh^-
z0>9Z;oN`niWpwsj?!qI~s+r?4bNH1*f#CnC?c8GHx~ed2&p7tjV~@v|ICp1qwI?Dj
zi6|<iqM(+Dki<etkU**g_KahX>s!u@+cKJ{A*d1(g(qI9N>z%4lm~b~NFecmc%ns=
zhf)wsX-m0Cv=Y4mC4v<5{cE4wc*f49!j`wsWuLS6IcM+v@3q#y*TM}xYE<+Ck3fRf
zmHh8E+8VwDYshaGcGZ&QOPuNDxq9RyW=VDHi}tAr?hGiuw}vD~K+sGr{UvQ>*uoPH
zTm9d}3mS%6J5$WEs{9K12Wsgsrk&-qGe6=C?g@f3(ps6WerVFOS;2FB1@Cz!zb2{5
zEe1XZG(ylFb7R$AV-|=(ElCGi=g&Y>n#I-bLXOZ!<Au4YB|Kr5^KsWr%S!~h%NI*U
z3j;#~Ean!cDup@S!Azht*kv`;xY8z0crU01FDiLS$;(RqpoI3(R?VC#v8<DePccN?
znX3+bdI{%9xF*13n>z%<+VmE$=o*{z{83wfQt~+^e^z3BMP0*%&w7QedJpDCJOVG2
zaL)a@GR-T|j3nRZW*=|pfLg0}uaY;F*xhY1_%^!)0-br#ch347o&JZCYf9c#5>6={
z)T_6?rEc|6HspIlrd+|e3=phwxHOzp7$A`<(Dve^DRC(aiGr1Rm*fJwXYHWTDohu=
z7KC$Z?tY~uVjh_9DTbkn!S6XQMBWAp_7N(g(e@O7W$}DAreI$bRag!-;`)s(0r?tQ
z@mBR$d1+)z^m?`uiHkc}nZ2bpf)uPX0Wmw+cRG&|D4R~lF(k4zC9EfAr_f-0mg6F3
zqZ6Ps(U`k^4MuOVhrU}yL#F#;GO+*9v17z+|4gwD4MqlOKDvW67`?@Q?(kUaI~242
z9l3Lh>r?bcsK`*T{W?L(*8X=HWwIhTQ_}%^ssS?#{gCp|V63*4pC-o;dIK9bI4TX&
z;3J9i=1Y~U2DwVa?b|BD=sk8q?+u9CyDC~V$jWcVun(cBkgj4IdJoJ+<=1R)q=-K%
zlg9@u+bcT=3B23%9>eH8w(|U3gD#|P9y<1ud(eCACeNsOgOVqxz0!I>#U3H&C?xNk
z)fBR{wBZ+iN8Wv=BiUKmU)x#ShPGobTYI5oYwJJ_4T}5X(6Kqf{*V!@rT17w;@yQT
zy~8AS^d@K#w&M9>_v82>@JZZzOaqd;xUWf?ZZDs0l%{!ufu>`qILaH7rf8A+9;)r5
zr5vaoploSI`k+dyeBS?)hbc)UYDyRNS$Ygt12hE##eLK+zj5uLP#-StH_ZYBXL~ba
zl$5XKDd_;^9BkGs)P~Ly&>hXm_Zw=@(fpxE>Nggxsckj<If?cwY?}P<<h7fTsbjMS
z=PbE+Q&Kjvuhk`|y0ntGTL`fEu^W(6Z$TQ3rZ0894;$%UQ@w6cLVZDZVZ3YNHA*<m
zAGu-i)0WIVC1klpLG&-GNTW0=3NW+3TaZFiNTFtZH+htM)Vnw$zg+8O(sjrr!72;G
zt^6pj#F76-@yH{Rk??6V@pCw_ACc7&{x%aqnKJS#0$z`VToYg=gm#$t>CT$ad99}t
z^VZdlp22rIdIA^Q%_tWsS6EZ`lD>dR73&x=A?!^yFk*Vwi=02!4en5)xvy(qg6o(B
zy1DTABG*1IY|n9(u}!)X$wUHL)>W7#YuRl9KP1sU{H~|m1u{1Lgo3D4VUb1E38KYb
zvzyAM+8x?b`|KD7fw(Qzjxoi=wPD<jnhETvC!OdRNuSS}&?lj01}E%H(7e?p$(SoJ
z#_;nQu&sUi-nb@tV*1g%>`-13sNA;`)axN8W}FIX{bd*UYi440f|+3Xe%;o;6c?Ac
z6@93Zd!RIxvqpPV6c6X5^VVh89HW+z%X?KGo{&u{_TJD7SAP$F^s+-4^Ef+!9S|%?
z`kuc#ViF}%rF^`?d)X#!c4P%kMpoc@HeRNLJ3m;hiFn?dYF?4_)xy48QWvBVwe5tI
z1RKS-?XLe~p@22nZ26RgrROa7(50z!o7aV$f>Z?M6Pa1hz)`4t`jw2;Xz9eR$r^2>
zo1C7Kn1*K0oR{2n57LR^aWk)y<#QfcTFw!$dSZdX&M8+7?cNFSFR+sK;bsk4Yu@6J
zKrO|~7Ct?3hd}I;`?$J)KFge^4?jbl&c~<1)8VL%YluIhxe*T8K-Rum8vmDUGxs>m
zbz9Ni)9A^I^_B01Ew$M5cmfYzTf8qK>GeC?>3PBeYR?F(>B;H`D~jD}3K6wY7uU5M
zMyxD5cUH7uX<wptf6i;PufF=<wY4>;!Zw}tDUINN<=f^*9G!HXnbm|36V45_40JxP
zYfNyN(v}%qH5Yda7p$RtS-baeww^XR>&o&1=1don#3*%_l|dwxO)Ro*U5;ryTL6vs
zDz%2La=-UeZFEc*<{tEZ#pZf_fGOY9HdV(=Qo>4hEtAwBF7ZV0;I^|SpvZBcBA*2o
z*rxUyT3{Pg_3Dmcmz2A_iUrN2snKXr@Fdh7SxGEjv-}Y;O;!L|;vC{y4c4iJ-eh{4
zV6vBFVuhP{-cFnK{m49yBWn9Vf75f?n7+{$VOi5*T_1H>(;u(vx4OlndN+gQ9pU)_
zu{D;evskusEz*-Vy@MEjCA`;F8I4ii?@6vSX+#a444JIae1|d}&C(uVLyf4OaXP4*
z4@%G+JG;?}?igwJbiP2f8-79<I?drY3>DJ`E%NRVL6CzbEYnR8B+Y0e1ohPevc2gm
zPcBZ)#+0DP91I2MAqVP<%S+3JRlG$z4QFnqjI&O2LS%WbwAWLMtbHnD&fj=&pWre4
z!Cf>KE$;URc2r9YjHm^*j*iyE7K0cqCpqaxY&9Q-@KAH-_=jmzw>pj?;UD!Z($W$1
zF&gm|emFYLa3dn34FXYXTGx>DdA|_=fCvnk_UZwDwRMa4wGjvmaGMYW?&1j;aNi>c
zI7l$zGH<ZZ6@!Jd_~0~92qEC(Eg(Q)vY^7CT4fj~mS9>W1=cJ;b%CG&jY6~<F~y8O
zsPhDxnr$m~KP5(4&sUQi0V?80P%hxK5wAA@h;hvO^}pD=t!)=#I{b3M#-I-yw{iWD
zC_4!MaoFewHZZz-j{y|f2H40NY-A^^x8DqG1e*9+A2kdiWc{pnO_vI*{GEWQYU>%t
zbhK}3p)|XMZdc(&%ibsXUQA$4^RM3J2Ql@&N5<J}SKfN=%=D-3cDBw4bB3>jIZ|l3
z>8xRnfy}=((AYo{f{gdJj)_fQ@0cP!r&2icmvyya%s@;eGW}P``nds&@&2y68_0C~
zorfZVP0HB_be>RuxRXjmpd(eZGk!Uf7}_!5N0LM5wb3p)vDn=RLI*eLL#=^`LX9bC
zEWs6ag1KP?LNsLt4-0n5Y)Y|j&{PxEp2#r(6OEQpl1AGxkM$vbghoq<A9Tztq_IZJ
zg5~XMuq=oGx}^OQbftXgQhPY(t?BKl71+nGZ?RVxaHb6b?2G<K+3d5n18H7@giNg<
zY>Un(m2@asHxTKLKhcZNP~;$S__Xq*lh?|!h``-fs{tv7XzkurRYqXuy`v<OH?6R7
zJ<Pv2@%5k_p;M-#lT~J&Igm-`3S?^d36beD9F8MXcd|QKbi$A?Ao-+I>~xC9UCbDA
z1*FrIJ8H+$#a=_HZnM+%@b3HTqYd`_CegEcpP^5VpHS8jT0vSniXXz9NgTOx2{>9C
zwT#LfiHQv5<wYV0tN>sr(T%*3n_8V!SPsK97d_HeLroTfXVnW(dkiN=q8|9|5%<YV
zjM{6RaXCfy2LLT;v+;ZduA|uuXmQNy_iFg2?RLuTirGo9jD@xVANH!FC($5N$5K_g
zcfvLBV8*?5B499s5@0E^TzPKw{P~3$Jl3J=cfl)-JZSoNoBf(CS9HUmJsJ{rsqOtt
zt;3jzn*!srYA0@n<=on7)@f@3uT{r3|6CzZ7APwG(fg!)vqpJOs}j#B`MHvfsH(rE
z65cP!QT@Pmio8LZYKYVK(&heNIPC{#lnD0;FcI1*X0;41Rx=80$CgAsxY*v=idGbM
zSD*`)Pa4iEj|^clk9C_*JNd<tm{*W<Sf31=sbzw21^(SK+^qsGLpg-UU!lQ~=UXDg
zW|%FcQH!&OFn6qi+f_s5lu((PVCeF`76^5C^J5L-#ePg>>&3h$HWu^lxEZ`&HBvdv
z{OI{g9J0lxtsb1t6>{S%$8vLd{v9V?taI<9L+lpvxjBv{>WP(O{2ezB$KEg1f=nD7
zx?yNmL-edW7%k=p+Ku4aI+Jyg%Q|5r{Ck)C9MdHP<_#hEW%ax!2$net2~Ryx>kv`s
zt+5V~KB>rZT6-&J#&Q~Oa^U!Nx=U*qZjcAx$D=pHux9XaSuiI3VY25N;;)C)q(A0e
zHjy{Lv^V}weX17V*G|7QHOwGHVn8o{o$AMcufq-BEaDUPVgH{}9in9zksx>T{FY8C
z{fxeb<F~-kVclnDEtNKw%RW$3IlOQ4!1<4PR+8FEJ%uOt!ra8tl3N~Wd@!xZuMezz
z!A@ZWCC0s6Z_|ty>f2p@Th}`Q*DzSuYGgNf0W+l~q9Z%)xxSGLNr@_Np)~CYw>+t^
z6c1tvC-Tb+WsgeZhDwP>y!wC@@IXnc@+Vd+C31+cUmt6#kDUdJ2xJ78g(%6>#8NHV
zVnrh0EzZnNNwo6Px;nK`T(ta-&St=)8H~9_aO^h8MxIc_Ur44aFh#E}aDTWfo?dKy
zEPgaBm8!<-pgkAmYDAZe_?KW0pO{*$Ec-#foL2;KDC+wfZ|bvZg0l4vJo!5Db}HGW
zD|W>08Paej%=^G@op%8wCeXP8Hr^f`a<MHYF8JO9Ja1HY+o@J=D>B}1nN9~E`;0xg
zcTAbCsF#$qC61Kor%gLx6EWAW4(V)^XbM+P-Q^zPBV68Ao+jyNOQoDq+pSU7wlg2}
ywvRJH$fTf!P5dU*e&yH_gFxqSN87{hH~ZmW%l#j`JA0(<;cS|}{_OGL+y4hz$gGk8

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_36_15.cpython-37.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_36_15.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..62bd45b846e3377aadc3293d9b84ec1811da44e5
GIT binary patch
literal 59703
zcmeFa33MdadEeW6HyZmg7z}3X*$2bfI9x?hT!*8D<d7zZqTCYI0#FSMFhB!ST^P>5
z1zF}$66Kf@t>`6=O*$ASwx1J6@pGI!#qlC%ao+QD{BoQ;=bSn@FN$SbrnAXg;wZ8G
ze*f=QS2q??vh1AXos%K^;?}KOw{G3#+rRHV(%V~3@vr#I$Lk9(rBh$lMeARl%tQQ|
zznD&?yi_&irM=8Tx|+7XnZ;~1n|5WnYTn8U)gphh3#G-bYM0&1EtD6#tKEw|)gHUf
zFZ3?<Rr?nEtNnIeSQuCwtPaNQ3{{7CzPK>FI9wgJveLrH;*RPL%Xck|F7B-Ew0wDC
zY;jk0m*u+`#us;2cQ5X#?pfSh-MhH2x^HoRb^l_eT3I|$J+OGNdeG|iEF4-qTs>_0
z-i0HJN2^B{k5!M^b>G7A#fj>~;)&`ByY64OZSiFF<l?F7sm0UP(^fXHaAxsr^{nLw
z7j9oXS3PI>p@lmZ@2uWQe%KrFc6g(&XRCL4C%rLm*URbZ-QKvj`{h*io_D9bJ>K4z
zQ{LWsasJ-=xtqD_J1E=d?Wb&i{T(;6)%&=vcn7#XpgNSilahnpAxaL_@8kM@t`B=h
zxISXn@8bHXcZ};}^>=aq0j`gG6I@T&^@CiW@NVP!HoJbv%RHN!I5oM<l$KwdTktB2
zbJv3vzg}6IDPInP<>q~7&dkmQmsh4vPcJQ=Sy-8#^Zd!VM)QnUZ!FC<=Nhx8yt$c~
zmFC=1<J9tvV9B4pd}eB4Y3j^9_nf_Z%A1+Kd*-fl_sraNuXo2CGxy#(b<fmY=g!sd
zzP)~T{f^u3zH{o1c1yLSrCO_{<Skq8onF4NT^E;@>y62|Gjof}{?fI&*Y2I`Rs9`z
z-g)mk&fR;*)YQHA+&wvc`<?gP<K5+X^!6Q7Q&ZkK?;Y<rdv3a(G|+zWnc2yeW^-<`
z(Jp$6K0La#xcvUP=9yq}Q3DkCla1!glD}B@o8?odD*T<EE5E6Yb#Wpcm7bVfobo20
zcxECKb@}y4pX#&qsNmO^>yrV6ebY;gpng4AoNP?a*8Qlvv9hobm+^dYdD0K&CKt%&
zf~(#lf4deZuheU^{+!1}AMIVKt=9de+EQcT1|{Vu=9<ByOO2Vi*@;rrx3Dx@3$6#X
z$!2YCaW*Q{uXDQ@6}|e*<jO)2<>#7<lgm+&>0YD*v|eWrW)>!csAqa<r4iKV!Q^5y
z+R4HQCL44nsMluvrNtV()JxJ);fcB6vBgP!VuO*GT&P`Hi1wVHo19&0OfI}T9@+D}
z*q@tP30O^39eOYGAIt?eF3kDO;MqVg?|pbBSgJRsm%O_F_!AS^Xvn^Gsvb<%n$wr-
z-pWEf8fLWo$?2dr*9ewsLA}vj@}sgAY`wNHcco5;y3e0~?~~6?E-o+BeJ%&yt517k
zX_f(ga_;(EgR=hS^5k^A7R)Wyn?ZfK8I`8>LDPJ9H#vW)*_;f5MwGrmHwy8cu6B8L
z(p$VVozXn=Pk-5FsYm%Wb0lGky_a5{dg@x;f6<={>Xk)4YT-0JpKDZJWY<-K%X7`j
z%t~Wg15yc=D*Dz+QybK3ovB}&T(HN=75;fk^=7596jY`!YY9~*8#fl0{QBu1NimqY
zW_@8M%IO{c5ngY}*Oty)URtc5;i3A}($bYP-efR&X4YR?Sw8dZO5+KZFJsy`qv1Yt
zo{j#<d!IS8;4Pl9t^x&Gd;g5*EicxtHf=y_Gjol4u+pe|b(^O<@2cH#ckTAO*e+3@
zH`kn;TBv(9`+{Q{4I84YKS6$4Bcp%%E0klywTDcY3e#h$AT^&3GhRB(&ZR${9-(Zf
zhB0$7DtPs4bJKOcG;dHS%CYXfiCmQNZ|8R|$~P}hF4z4NwACH}|2UV?;7Vgo8<Y)F
zbIqKh*!3&zpCv!dFZs7G&NW|SLhC6|V5&-=yo{Iia<eHf?-jhFSMs{Nve!+n$Lsa_
zyqwq1e*^p-<d^5aetwpl+4srROxhdrhF|WgX1%9{I6xLT?_0c4Zzs8Yz2H6LP{_@6
zwOB8$cil?W%Qtz~OvX@3n(}UkQmQ=;jZ}Lb`l$9fq*3j6$f7#nkVJLRA&2Tveb_tk
zdfL0qJNa_1I^tUQo&|WGep&7Ah)XVcXS}mkG8&gW2f)0;N_Kkh_wMxWqOW7#-QGQ*
zxLw}!-o4&CC>i%Y;N9oFliY6ae(zn}+v7dpJ;>j^-b3EQ{N3k0;yudW{c*eJy~kco
z@#czm!FxC52fX)q@8$2o`2M$gPk2w-dk=XRy{EWy_>(M|>JipLHOieo|MZh@{$`e$
zEiAAUC;ce1aPH!T)e_s~%mVxN5<pEG{!MMkH>sF<(<AxHsYmzxn_s<q{K2n$+sNnN
z{jXO)_uyCl>yP~QFF!f;u?J^cvma-^`e2j=CQMi2a;>8hKeg5TG%Gqt-At{gZ}Ea~
zr<cA`XpRRNFLN{fQaaojrfS(B7pCU(^M&tAd0BE<{^s7u-AqF`u-0qEu#dD9_L6pm
zWzurkP1+sykoJVTNPC06FddHG%#iO7253bqs_<%tcXm-WNXel86Ty(~+)Pt49OS%W
zFcR#bWO#k_7Gr!f&6`VaWcB87%u8L-lk{+Yr~eD#82PbqgmhOpOgbL&WNmj?Qa%5-
z!V*_|!abyW!+o!%{ammw%->AEmTu;P3}2d8J<8A1-u^Hj7Q&*}^(us-)mG^sb<e6U
zJu9mZoxKhB^Xz5rbX)IS?GIba>VL&XH#L7C9162xChQLf!a=X+)dFMKyH*Jf28SrE
z%%*QL?rR5v!}CXiqkQ`TeKl9dyuMd8cE`Q`wS%*%wL|<4Uw<T+pkF6$rq8Edxbs7q
zwIgds!z1B=D+T)U55q(2w|N8M(eU8z)L814#_Z_Y!SL|<ZQ&7Jy_UZE51&u5!-rl@
zg@?{FzH5hQ?PPFx{YZG|Ryr6758X_A!!}2(fZC~Wob+^XX8vrro0_)+2S#4atQ{kN
z`n8Pzq2S#79pvw1W_MU`?`*IF*6;F0!(+EJ(|cdbEc?OTH??Z+q2@h2^__g(z3cC|
zrLy}xLCBpKC$j$iEY2u%{cM!Jwt9AY#rHw43pXkv3&1MN%S#XvjoHdU7WP4~hgUgx
z{cN1Mc5t=#zREKzjfQG~pjHP?E&FwE&fzBEgG=DdBFJoNsaangc)VU;R@KLze7FfY
zS&7o8fz91>v*7=F?MnSdGb-!FHA5w<wP)(hdSI`wFHQ>AEep-r&B}}Qx!KD>^K|8s
z5F-DtH0v`f3zZkaO%)&PyyWXuB01=Gg$7R9H-WpRuPiS?1)N?zU%9RQuk!rE&s==(
z#dq^Xk3IX?rAuTh7oU5wMrSL}J$?S+OOHKUx%5<}`q(p1ZFp>=5OvSaU8^^0A~>2+
z??$?#t4%BPIKWebAo7E#9ADKN9#q5<baTDdd{an<G0S_wjpcd*SKbtrFs(%Ygd2qm
zpXJvaCz)kVUl>|Thv|8PFYB3GsqfBQ%{}X|8l*^0NIc3;FRw)T#>(RI4aPll`fQX_
zXE&j~ua)m#farp_Uw+`v=@oikA+>ryAoO$%0%|U()lT2Purxin(0qVQtM)|3f0VlZ
zV<fzwJTtejPy@cM)FCzIRz+j=`t_?TbIcE<=%fcMUYfl&=}){9RJ;Nvj!LceqrQim
z%{q&3sqvWaL#quKMOhE({@m)Milh88QyfHHwc7N;WV2bT>FLJgVx8Oow3nxmo~^sp
z6!lIo%q@e;jpXy9ZjrckmgbFG5cSW9nvAQn4x_T6f|@$b<Ul$#>I=<ij5StUW~%gh
z=vAXRS!K;=K!lKSJkR@+%leSKEmMDpp@@oljuq&S=&H-^=sL<S)EiMw;};dDR%T}4
z2BKb7Zjr5gc9M-L=cxOd`k--OR7P0lvvby&g{h^Q+Z9p&0$mp!+bSq7E-`D=?>5qy
z37nY={9bjZ&*syS=2w=3XpqUN2Q^knZDpBdQ*WwhmOj?~q+d~%I=+aWDAQPu^5Rf#
zL<MN-pzeW48^ImtqW)y*i|A%dy6UuA_h~cgn_O7F40kcjV)U0-2pTnZ!p<$PYCS6)
zUDK|JdKN(6jhb6J{%z_+j;+5C^{Ug!OX*o7>Uvsyh*pjNoGSJw?Yr@2X8^9Eo|S1B
z9#Pp$CwvxwZfP;<e#ZX7Wzgf2sx14B*@@%+QJSiDL(K;9q}Y=8&r|GwyI#`;j1dnM
zT=pMUT^OoK&v>nxp7NJfb4JNWluRm_Qc_d$VI>!oyj#h8l-R~<D7UC&iNq~(Ta5Ht
z@YDZR-CIyHt>j}QRjqXPnC)Bts)|2JBEC}dtYx^PJ>^`srrlrF)3zLwF|7`<(_jlV
zd9Astb~WlRI^b9BY1OQWJxtG4dpA_IIhaWMA5<@l;}DJHKTmQ8zh+31I#5mzqzmbB
zex3jIr2Fi;C%rFIRxV?AGbL9m)t7e9^`uMb9?F!Kso9gs(cXAEXa9{*Ylz$ct&L^W
z#t=0}GCi3-{;u|JfN|4B^@o3b{3Hg`Bs71@D7g@N!H3pbD>jA#_z$R!oTwQu{c3)#
z=w-s<SZb{lWahI_M(_aO@2()Tp5soJ2aW#m%&yd}%vw1t&gVarUdslB5PA#TT%;XQ
zd3n;Vuxmc=Wp@GJ=ZjY|&5s3TuONCc=ni`3d-?TwMOuamT<`ZvP=m#rX`=vi-NkiR
z$F-<M=&9s>ci2sj2Ewjed`nW=<4V;}Q8OKPdhO2OHh22k&+Oj#On>JcZ=l^G6jN$`
zXe<>LM^dy{9!uF@-nm;}JEZcRsWs6x{oWuw@9F5tkn0J(9}auG;aiyv{T$Ia*q4=W
z!4JCr?uPD4zHvv`?cQm%O54?%AMr-7WT2a%bmw<?JBL%ADBw}}%kHprQ|;;tcB<dI
zQ)|5qpQ{mX%$_;#?NZA@&-$1~*V-R%UFi#{|A^J!ZS_COJ*fAu|DIcEo-769-X7x`
zdc6VGMz5E>Ut`wq?In-6P}DRnigw&{E4{w=R*LWVLQsAQt`xejBD!xQcu6P?Uql9Y
zIyZ^L=gmLn-|B4+R97xfUPH#PRGF$*5CcK3coi7rN_}z~sYEOgVQ;^W2BN&tLJSG{
z;f<gk7+Sv;4L7g)F&t?lfxT9^1R;cwY^CYIO#Aj8q3pN3yUQN->(gNQjc<&0w{I=p
zzUij_xcWGhU{vT8AA#O<J=)hU(~Dc>5F2ygDsT4CBcJWL_4+@4XEf5TXkxJGrTWb6
z+m@a)a{EmYRQ|h2;$i<%!b(2mKS1&7z~YJtZDIl(l-jd!ZDA_L=FJOCP#GelDWu#8
zCiWR7_jjq;Ya~%IMyvjsiq5Dg3kvpM(q((*KcGv&IEO?Fu~e)$1GWNz`ybbXISn3^
z%F?W1o~XFU(pX$sbhzfTdipOZvAHf>a)Pxkrm1OiwP^U_Qt)2UB8zmj?m1!(`C((S
z;eWqgkeysunCLT{>ad)lQPGecT{Rr2mq95Q<zVmg|GY{{66i_F>^h=GH<7e{K-u^4
zYd%Ag63zfKl=!a?*k1xC2$zgy1~LO+hJoxT_(K2f%Z_k=AhVB}<NWqzgk36Nmc#sd
z(g%6AhbLA?4R0j0MT=(pp9o|FVE&&7<WG%2f}OVzdH#=&K>XL(Xa0ASd{g+N+wezB
z`j2b`=dQH6Yg4JvO6!)v`%U+a1*smgTW^I#;^z;(9ioXlwZDy1Her;w!hv>$El9^z
zaj0jkjXe?+<JhIc?K|5cAlFD64UM;}x6x4CneFbKi+dpnkpG5~KN(6I>O@JdEfACc
zmuYDWM)KFyeu9nsTU=I$H?V821sDvI&N?AUTsF`amBfACh`Ib9Q0sq1iBOjRSC#yr
z662MANIBz-KCj%5EBTvBept!>s^mwM{52(iN6B9&`NmO_|D$?F_|Nc>|2LE~oVXn-
z`Y3n*8NdHlsA$jjsK^c62Gx@|8nKDSc0ntq{QF7M>&QRnv+KEAPL`<>c#-+6Nj9M%
zk@DUTw;~yF`VFM0>C|f3OT8bdHd0Q}p%5X~PH{s!?#XNMlNtBq2Y3RCG<$J%*L{^|
zQE*9JCn8f)w^+otx*L(43G{R;=FwX>CrXZu7A1uY2{NGW*Of0P;v&?R(@V>Ud6xXD
z?iAs6RS#*^QM*P|BwgwzngMAAyl|68>)jB=%~9&6j9(O|*q6@0T{`Y@wa<n;sXi^?
zoc`_LSK?=E<MFH%0H~_WIs&FG=VgOTm|M?E<bI{#9}Owlm0Ckg-YQ+oU%xlZBXmBy
zR=9p5%#b^{RtyX4xv+RE?LQKrELtyw1z}mlzPYgQuGG~>o=;u9_jzL(A^X>Zk`w?h
zr-H6^R0iwa5&*CFFuAlQq4!$pwaoQG(7W}%SA=yfy_!9b2>XR!MeN-LyV@1zt`z-W
zy!r*+*0n1&-v{$qChZAeKM|pq#Hhilel_TaO|>hp>q-tm`5ss`SYI)|uwQWw_9xxk
zqq-<Dx_RykVR?Pv77UVWf2}WWqpX^KaQ$R381};UcJHG1jK}y|KWY)!OqGnL#14i1
z_e<#h%=4tENLzi#Ga8?}{{BtRpI__0et%q6j@vx3HqhB7Duiy2d!wlt^_}*^ftifG
z<A?N)tDk+InO%n|RV`Sw!Ru!?wYZNK2ixuB>8Dq|2hEVR@(1?3SXuY{Q1X?n9*u8&
zo?Z-v13a6tnI1x*z<(dlsDEoi*AMf}L-T#fY>55)&Twdb#OtAVLp*hv_0by+?bSHj
zNZ+5j`UT3euVtD)!M5GOT0}wlYNmn8X@0bQ%_<mP8)4n`8vq#KO~a-x8YDMjxuG`@
zLcoC99j}|Y&-x{v7!HTOFXgtqs~g`t%)OCtq;<~#!j6mnlWZLLoE`*8i2Bg5OXOQd
zqKhGcWOl2S^Yw+gX$O#eRlaS_$@<$BPA39$c3xXxE*Y!|%zOBj^jaoNcZkXj2EZLZ
zMWrYwS#p%FZ2}Dbw{Y{QK!W%<1mnkS`)?GZZYLgJJ$PZxsWX-^tBIkTfF1I))+J)|
zoL6s7OJf~OE;zg_7S$MYBdEmv6=W2_81yB8(G(r41!p9%+Kbf=hAAjOq_SBZs33K`
zzFeP%0jo&QvfBS-eE)H-?kiVThs)2ktHgy9J7ejpq?i7jjftMNY4C%l3=$knEl5>9
z894FcWmW9LdchBtmJplg5undCq5%!3=L|z6kVVw(N@B^kFHmXBsxS1;V++t}Q66FJ
zED8j1lu=$bE6u2DrO~{)Qm-SO?>EV6d)5v5Pwa4@J(oye{YTVfrs?_rNSXANC^L5@
zDm7v|75``LjfOL-T})5SECaAiLA2SCwF;3{u&n<XYi+(-Zq$(5BjR1~l047CExU1!
z%Iz80yq<j87}t85CzC=DIRJt|cxnIjfojC4I^m|6)=YW`wiEON%Q}+o%k1MSW@Xd+
z#L)8JSy~y1uf}0&N7W`m&T;M^=BNKwhqqa`)7mTgSKuc+X1Gg(7YL(B(3zgk+{}gu
zzz|wO0tm??#*53rUaFT%uJcYXI+F#>XD0XYnP!(#;orOK0p=`7lMrLHOnfR=k6*6^
zm9TQXc6Q=)QuRI?Z3>xaKe*2x`{2dO^~$6CPM<#gLdD50q*XKp&qBS4R<N<Wf{Ji+
zEAjJCpHp>TPb4NQ6*KCo95?Y(<@SlwmG?kSoUEM1KnsH{w4#{WVE5C}q>ojL-?*>x
za3xql$icJJJXvXx@+T{ur2E9l3c?P(GX9cd&5EtAc7w4<=&KJ-<bCP({Ld*lqr@o8
zzd&vRVu3$VF>WPF2Uzj&-Uw?3C%vlnbiQ}fii`TM){I4H)&5qkK5f5LQL6`m#)z$r
zP1O>@Mhhd)&lbmVGFumil}H2cmP>X7-8f4ieQ_e^zUwdYnES5pAxEiwQkL2G{}l!3
zDP^0rx#wfCf;r|7Ol@Tq>-q#EyfQ=S)xFmJx2pSR`W-)(e8>OR{to}kD%E%Rzoz6!
zl4!ILEjl&%malMB`jrjea_DWpC8q1`%W3-0VrJSa?th`hcxF1j=z-*8?6d5P#<%&R
z)f&jBcI|5NH92Nx|66@cT>BfI8&=e&H3mpB%dhqt`)0l*Sz%Z_IE%%&XE&1X*t)_t
z*G;}c($!?ANcb4<6m~`njK*KlD#@7MqXkCe(zU$R5-_#r)(VivXqybI^1TJOrzGNZ
zd+K%Nj0gQ*auY?{UH<>8(qw5yeH$eP?fRS7YxmW*%K6(h3}PKq48y1iOjieDXpyw{
z-|72*Lf>bw@^33=oc`Y;xBd6~vbTNfj~?v!Xt&E=+2+$nH-EaTiXh~}w4RJZ7Z=Go
znBf8@3`@l@gANdV-#kp=m5hI!91s(ziJtJ)>{{N)Zw%(ScXyauWB&aMVSXKbqL+8t
zEKqG;PhjJf{ZMwTuvT0v$)+kS+1e;xG?fFqZW|bNh<U#Hvu%8n-LyPFKy?j}Nlt#9
z90sim7<1PA-{NAT%K_~4b=R`Kwq@F|w1Pdu^%}Yuk?hXgLHNeTIvPz@iB$NBUS;!|
z>tEi8qrO{%Kg3&7MjH)5|460VlNoJct4$9~YrOTZhoAC>t%O8nC&Mp2ll~r<$`7a4
z@*1~f?6c@_HI^kxy8=iZYNM6CnO!TRHN>Lx3VPgI;IsLBn8!9NCz-}P?5LNKHt_p1
zpH9IP!w!4-uxli>Uc5zVSdO7nLE5)}M=#prW`e+2^Je*YQoi&c;tgBQCshA;l<X#n
z@}{9NNroA)URGfi3V))<ZS;-pM?KDPyk?j(N>BM;R8!XISM}(So1+cWA-om!G?p4>
zCcZe?ys~+QO3O88-1|m*UyMR(_GM1AH0iAx30Ya9RqX|vi@!>SQD^n9G0I-r2|v)7
zW6R6>&#>O}vuQ*hjUX4~19a=uF3NJx>?4b4)e%m-aM9)tjXBzO#1k`VW((2EeA&yt
zs=d{%y+K==;hs@QR>TaSu&bhnmY%)!2qKl<FfR$0#3vzQ6N9d}LWFs!meR%6>}bqZ
z%Vs1l?h0Fc|K}+3|2_$M8`AeF@AkjVCX&R@s*c?>a~1J4Uc(rksD#`aW~lD@|3sBu
zQt~w=-$8<L&{AX88HChSC=n`)LbQ^tjA+CWpbqN5U&+wE>0b4p)Ie7D`@JSmPVS0n
z**uxOt*K&DE-n9%W^{~d(v?ZtRY)I116EG&A)ljUb;#lIcm>!)Zw^pDnQ85~WY-<!
zLNm=gYqm8jf^#|WS#}0yfgLBelMK!e{vT17Ei%a;XA{G$qyR=nAC{*kkrA3<mR0Pi
zc0><YViZZit<Ar!FVU)#?5-fT2Qz={*xcO&KDX$mEESvgk$`WrW`WJFez|lr1GA&m
z$Nql>hUaGX75221(0f-BLXjB#k-a#r_qG11$zguYG6`~`^ef1S94VY}{B_xugQ-jH
z0~iFu)a{Qc-4Eo&Bjkaf=5u~mEysk^)p!Cr<qfVTBPO$z%}himNoOx<;|0C0^=cc!
zQmgf!(jXS}TFhl2;+&aG>+Y68w9Ty+jA&sq10KaQLfTjflOC5qvTb4uwkO#Po(Up3
zpt)(P^RpN-UHdt8Qu8RnWVLLwY&Vkr<+-Py_0J*Uc>@RjVKFLWW<1H_%=%MEU@@Xx
z%QYU*vh^|mi7MsigY0~6K99_m`a<1Tr(DUrm6)MVmZwTV(ZlE&sU}946NQTyo8Ru=
z$*U16!9*g#!TicwF+<ta{VR<tjinbGu!=CPGI6UM`?kc~ZS5HQ{oD0uc5<ree~Fx}
z8vmxs$}Lxr$%Ux2C|?%ZFm*<2$K*trM+Nw12E!f~P@&SMK`d#kbmJdu97kw8#eVN$
zw@VO_L+CJ?DP`iRZ`DnPwuNK0*qaqv*t&nT^IwsmAnRp;$G`wI4eTC<F+Z1rqGexs
zIfKSotg57o$a;CMpLr$4^+~%faJ|5jIa*>t-BQgV1c;_<Xj!c@>?tM$h%bMOSF`g)
zd@k^~n9U;6MBvTclK(?qe<R}!taSnVv9sdY!8bBtN!WT0iyg%2mZvV58^JetB$W8K
z<S;rIzV+M?&vnPos2+lBqv4^$;+BWia;sIScEP%lKOvOx^Ni~D;K9ebJ@qeiyGYVv
zv1~Hy&21n<OzE|E{{76FY5$!0$>7s1qsiFRwz4{8_9!=y*d#V7QucjG51M%L4dw13
zVQpu<xkUt+Qq@FR?0;8#%)=tKNnP!7R~WRc<}EwX>p)dWpHNprMswmI8f)fo^-Q^?
zTWTqQqJpzdHA_e{B=Kb(*&={8f;Lg3qJ3`o|DHNgmu)Tyk^MonJVa70i~3=&%*<Vv
z$t1?6&V;sFvI}~<VelP*lxP|=<>s%@LQ_QokOO9ODDj;84P=H8()DG=pp1sFC+W+K
zwDwW!`9v=#@FYO{6u;)(BvKg}-2=d3KR_cv-ymRvKFY1<1){iu{)FxUH+?_5>(fFm
z*$os6{EyKcMIoBR5U5RA{vj{&ck>M~&@lQG8!#W!=E9X1^><sqTSV;3UGMkx`38Ek
z2h*!P272RL2|b|Of({0JA0Q*H1&R+LVwjyr5oH|EK2d&FYH&h$m%DhzU<a9dHN9To
ziz(p?^B3K6%bdQrIy6<eTzRoFT{-nYWr{ysunc<_>-f6es5#j*V{loYX<x05L>aVz
zHY|hE0M5;>EMapN3$le{TQaG<(H#Ci*VBfGHbKrVK|_-@{uhmdKw-*=CP{5UMNU1K
zRN64t8WkH6M<K0iRG4vKB3jrGN*5)CU?tY21~uUIGd>>5&TzOjU58X)UDHzbB~0E7
z-n)@Af(Vc})t9pwy6Mn)9}qrf-#aQKV7;5}m0Df=H8R%&x|f3Jp0|aE&Jj(P2zRN7
z-Ib;fog|9ll7FWMONxFijZTv#D*c*i?T#DC4lWfV8|IN8=V!&jW?*k^F#~9e=%g;7
z%Pc{Vr%*vdN8HLV93?=umV6g(bnd^>tH4Ie0QaP&xU`J!62M4Fm;WVNDBlt*fR@nf
zp@#z0MXxt3Z>n8lY4u%<(#!rRqZq}O5dv|5J869`;XhcyW_`9cg%g<^803Ik#7$5`
z8<oh^mzSn5;}wF>GhPBJzF_<VRDfSCEci7LIe_YuH0p?cTbSR;B`o&kC0rK$BYHWI
zXV!FG&Y0n(ruB+ym6w*WxWqXDB}$Kp#&M^>L9;p0?N|dXh^Wia%5&cJs5g0_#-&@z
z`WX92Q(_npbSvkSZfsiPy)C$FpHtET(ux%??lcrR7{4(ymP2=eab2G&$44_Uqp;!e
z1SV(=iW}5+Z2lq%)5H$&SpV_5ch}GecFiJN$A9iK@6)t)*8EiKI<EPtw9)R+6EBxu
zVY%K+e=7A!_<dC2DoI;OTJ{Rxl6off0iK4k&zX;=`Dw-rDucJ0t~ZpulG>&$y|K)c
zm|YGA8-7f5k&K+tpvDnMz(2>!<-7wy!#?8UVIuy#Iqv|%a^^OoxnSE1Cq!8kaNI!c
z1%?s|Q;phLH($a)xbLXg+;@&?p7chvl*bJ!gN^5K>Ohxj6WKu=BoBRaUG?v3t(v%}
z{@oO8?dk39p8EGv<9NG^7uZ_<aWylcgg6UOx+E}z+7A9t@ge>h6*&EeZ(#47Zhetk
zm<}al(pYlCvQ3%m0G+%?557Z*hAryfG&B`;%o%;iC{+5l0igLF^dtH&){ty^C>cAg
zX#EV2s})ulgof6y)|%)`TO+{pY}<mxDKE2b5ARjiq%hmO7l)WU+fP6sF~j)HoK|NI
z0Yi1H?zE}iS}Vaks+nM(lU#s<WeSpkqh<HS+%A~da$sUfy^zN^%$vKoI@)OAS%-2P
z4(YCrO;zHo%8zq3;qGWHH~gg+`C0>Z8XI3*zlCIrjv;SGeGP>iXx6ZAuJJ(B6T56#
z?1IfhF<cW`rCKbimn+pdjfHq^Az|?@V7M_Gk!>o!SyD$tyiOZJB-77NJrnM0HW+`o
z?HYh#aBT82YemF@MX;B4tOhC5ah)}yrD#S|Smt07m66R07_SKE$lRX_iZ^i$fjNQj
zu{&MYuVWDbo-*7x4{j26D#xO#Puo?Ohn5xQry<u5gjMBPm=q7f03jEGNvvjsfqCD0
zA0$$lI{k7zkZZ_1E*`Yu#9%$JHG?5`Ac(pehFOO~93bc!_%@WC&S@6uut(Wv#teIo
zF!^Gm1cH0l`Z|DSFHd#L^#I$KK5L_gt2}MY2RqhB1HN~Cr>vV0t!3P3iOA#r2VOxL
zXdoIZla~?Q2_Qe4Qt3p$1DjF7sc#K@t`=^$g$53mREt%#iNNnobhug#)zZ@oOJ-_u
ziH)As9{M95^WQ}->TWG4Dkl4tHDf!%z_RwMf#tq7Sau+}z5X^rIftMHrwy1GWc!fT
zq1pxbIvt+Sxj7bgGGz(eofz{!p#~D_F3O^ZcJ%52eZmRdom0}56q-;_TiMA3en2^+
z=8ao4hH6b^52*nIpgHBPQHJBKyz-h@h1~AUA}0DfsE71LRInOMUr$8cv7Ax%9}X%R
zz4@qG+c+0KL3i9x`{$K2?3H#SWRTW4EjJGD&=^TBBI%qA9|i!erECEbVgOJpvjqX(
zK6s`Aw)cCGMvAbgq;rz-^xLj(k7_cT8{nIyd2E8zTG~nQoWY&w>v@#U$Rm*B&k{3&
zO<b5S&X*+GkvuYnb|#Y)Ni@+F^_u6QobBYq=;oqM^Zpl<d`!u6N^T>G@<ww-`G>v9
z#phdER5cV8jAUy@S!%RbltV)!@FK)V!O^*v&}_|`hRN|28YYH^L0R2ppTAvwBu@Sv
z>c}}I?@*#q@ZU-Dw)ezdT!OK`!1BhCJVh3fEMDAF+dr(?6sWv(apI`|Au>_9ChuC}
ztKf3mQ>$HFnOtxVxaQMmZWsPX_2#@1TYok-MG&+cfZY)Ky7{Y0eqG6LDEW$#-&FEj
zO8yNA-pbe$IW@@FR0s@9+3NIdp_{zGL<M4H{ryU8Ga2<j_$5X`$vc(UhJcPt`A?HX
zy|EMVbRC!F4vQ}U&Ik1>EnNS068)Mllaz89jBviW{|4TRZ0#QXGU=W^tVkay^^^+y
zSK`-I>gwxZic9>s*Hb9xONDZOdB5Gym&OMAO1V;zJ4dWX({v{JC*-An&7bG1+HpvT
zAW}7OOAHd?iR1*(iI71Ik$Q2xbjyKaWMoLXp#!R2-UqF=i#}4e*dsWSTI7*bwa224
z;2P?E$lLAhA%;hvx7XW8d;Q*iufpE}?|^rZzk}W(?=XLdyd&OG{tkODSbUK*5k>61
z-iPfy3P1(&BdUlitq~mSl#43jeZ)KMo#7oj<9lbl+g}D*5K}~PMif&-F-B5E6q&Rr
zBZ?@Z7$b@(q6j03DB>cF?DM89#z?xl-<!tl=%Lrs-oxG_FXyWS)6yt-o~I}y8i9jx
zN!_B2sN_&wGGlQ^RB|{jne`s`p0KBm#3fI97p;#+<Lk@bQ{K~7<5*lW=Uwuivy$UP
zNqL@LPQ-UU;C-t-d%~iZ+y+9NkNTeg+%$s3$$YiEsDMD`_I}a7i}}8ANgIZLUwv@8
z*X2y<AfiTVqWM|YgT)Dv)&j!CB7`8f#LHAVnM~LxxyzMw^F)wCeu{Z6?E3>@(ZA2j
zA}7tkbZ1N&`?O$b^Jy<dN#DBoBl)pngiD^)MIPf>@?Dl6w0wD1b%uh%O_*+c6!A-)
zONV`bG4pA*urzV_yCVO^-xVUJR=zP7bR+G;!&OGC9)fvF`=3uWhSDkh=iwDv>;KkN
zBLnBw-TacpLK$d?x*qg;gCV>^@%(i9@%0e|^?FNceFySlmm5Wtpqyfr?8I-Dh$gaJ
zukEt@c-SAuD4FM}-I)6hsVz^@SH{BukvELN9t@Y?udxdDdPGve;B_CR`^|^95)#A2
z|1@_7bSJL;vW;wNzOsHmQA&*dAH3+qr;jm5{y`;&NTLzKAT)6m2e^B!XahC>&*}cF
zBp3WoEBmmL?^43#qO5Hakbx*$M7dh>Fa7@|e8$0OCo(sJYAJ-7-09s1gJex891U&s
z5>bWLmrmV)iS?YdXm)vN*?+&r;(3ilH}$X?w%8AmaB3?UgfUKvHf1r?TBT4mjxV{$
zNLEXp#}i=TY7zEok$`9TxD#5yLrK&0Wi03XOoYnJ_heI;vvx3vXtXo_S-Tt2SZ8@0
z6U$`D(NMdVgpm@jn*O)iH#N0{b==aGdP9<*Xjpy>SVJPzOQJXUhDvADuc{KMahyb5
zCIx<zWMbHeyy{SA%!~MQB?xT|Nn8Z6*2D=yqv3y0U;Ys#H6<Tb;wiB;XyfuBT~?K}
z+1F&*wpVQntsYUY47Y05u87RJL?$J?2bm*7R|%Lf^N9Z&rBwfA5H_g}x@agtXDVxV
zpbo|HDc?ga*N$3Q9oqy%lTaoJ*~sftMShR+Gxr?x*8n)0$UP%KM2Soj0GTyRwAL}p
zS}&L}g571K7jFStL}Dr&3)eXTmigkxGxxg<pK}`?4(3_j;y<J0eM%TJY4&YReTNEO
zQX*Pj;bEATIS>11q<o4OgF}=>qO&Q1Ry(R{S!!%1CtWB!H?46fydTym3ooQF5U!+G
z4{jb83lXy6k!cas`lr#+qhjBipC{qlaGW9qOv9Zj$4FSN^_LNAsvMOCh8OpttRv6?
zV6)&^Xe)8U>m?hjE|5eH&1;$3-SDnb5%z>RauWQRfrfFnpo)*^o<dt?P_p4KvOahV
zuSkJu+LbCX*YYQ5y~n^U=1>}EZ3g~*^}*GLAmlG&1*{Nx7;smnSMUv5Q~)7hY0G{d
zAakl+f?5cOFxjZ1E?Bs+y8F<fLzU;ARk)yMFP%R{@SsN@enMB28?0E}+lpeOaC@ga
zfwYYOCx(1&8sgKq;7d+=?3=K+&J71|WD)PSw81zdYd@m01F9m<FdA|p5Q*%P#IlLT
zBp&G~NJ7>qHeN7^%k;<fC?4O0EhH9AmQQUN%N<h7=hU@6MGumlJt7&2kHz^#gYI>H
z+cGdpuWs4b1gP04tTK@GqgsNaw35mlw@r@9jBTrpJ|@oQ@a84CT_eWfSV*_GiP4sL
zU~A=;_}R)8$LWHxvF;?^GGoeH1w++Zx2!HBi?aS%)>$q=*O_?t<{hdDXvel-w4-@h
zv2PlNz>+XI{yf9i8c#tL7b3=x{RLgVTgiJ!whWx{PCu=2Qm<3lQP^KZdj~d;%vPUs
zp0R&<J;qP(HK^JW@`V`*(fLA2t=O-%-1R&4cRha#QTYo;o#-8<V-`=Z3+yoWmr<$}
zt{z3^WPc?GGBuo{f=(-hWYjOqJ~4sB{{((qL!evqB~giTib@#`cCJJ-IkNzd{tm6Q
zU!a~Z;NLP_DA0tkPF**CR>{9mawJI%Z%p#+&uf_aD0dRSQbs21*eV>`JXoDbx1%qK
zTy(>bJwSVJIb`YUqV^DK#e>$D44MY#csw|ni>(o`Z<H+4M#?a~eS3u9syXeW%xO$Q
zwQU&%|Dzfk+1$SM5D0($GYvu64S_-%$SU&q+kJj#Pm=H7AUFA|w8l54nyZqB0MaV~
zwooxWi`1l+y}HgF*`}W7j`YxIX|bcoyU2k?QBz0Ok&-R5?69rSb0+M}*1E1g#4}kb
z=2&*+wJd-;TkF0$52b(9-kEKFBItn|(Oa_3&nn+;O)bShJsL84xnoLewbdifnFEv+
z;<Z5}Fq|)%OO?rHMaaz+o%TP#=&g2LYWbC}?p6gN#%&<J<2bAruv9S8$B3&n&B_am
zk`oa~AY!ZC4|QWgm$tG^IHZ+rlxj;7WD{#^d%I;lMHBy0b6VlK6g%Jf01>JP@tDNg
zC6;&(FhKz+R*$~jOgl`j@Ar?<YJ$rJR|FLeJ}i(S)L{<KMN|RKAdg9@fO&IHWfA7r
z1uekF&Nx#H)YTz87pcMY7J|Y$MzX@%OfR0SNWqKbieP!=x?uAP$g0VQMc@WzuxT@$
zL^CiHGXP65@PNGiW;6``{e0d;+4+%4>KP5WRo;mK+X&MU{GzwyI6OIh#TT@6N{QIQ
zQu1Q!WOZ+=i$sufTiC!{Y#(ES&T3}W^B>R?4=ULRr=oFFK6Hql9nmg@#cyk#M|gJ%
z{3rW5d3-sN7P9rk=DB_AhLU}(57Az>sXb3dlN!MD6U_+!Ziv%N<EWLPf~S<-Mv2@y
z<Nd`h&S)2#B~?*|drTQ#+Pk+PeK|8ZleN_y#}XqU3qI!pz7S5jtf=4&F;R7VVSOjF
zGfMY|b@?MC1gggH&4?XnW6`s|NHxE?bu9fw%KlJ;=0u`S$5Y6RWVeF6x1xWNp`w{K
zmb}J=K%NOdE;@0sL?w7y8>VEi@Vcf32H+AaAX3II0kKBLe}aJpysf3L;-`gSo84_3
zc5SH5o0;ZEf)W5sHVP6tQ(}M-@W$XOT(KKmR<M|=s7QP~s50L`3_JjyW8z!=S2QA@
zP-0_wOhYB|Fe>t~dZ5J*q^CD8LL&!Fh7HpwxX=n7S7OD)#`R@LFmpb{UC@j@x1bS?
z>2{55Q0oho+VXop8CktTPau-f$Pg46!@YI*DBz4t_%H#o`zY_%DRHGho4BTY3UPCQ
z{=i1Wg;K+2#Apwr=A8uPyLk^FWrTr&g0O8}z>Lj&N3YG*QDbO~vIqrQ?sMtOQTj#y
z0L6}?%@W0M^N5uQ^nF=T(|2fuj1)^((iLeRw(yg!s>wL0dNK|l<3isl#D+i%DswC~
zi$IDmro@c{D@Vrfj6;|u#*6!IV5zKRUIjhM&PHRwSmwqE65LT<P&q$FCM?)E5g3jd
zO7)zBx~0{h-s}x|`M%0!i>12x#*6piOzZ`h+vP1^M;ERgC`~YF;FtEbI2lvvRHl}E
zXu*@oQ<WD9*M1p)<VL&lyYQ%-4C1QhqD)XJyItYd>AQ5(wKlDwQGU{xn`-GG%oSSN
zJxzF2x>Bz%V|CcH1<_gtg$||IXS88Hr{pmu3e(-PqnBri|5*}Q3%n?`vqH<VKl}z^
zBmcZ^^yoWT^fLj`5fO9i%GHugsM<w1FnOip$GQpZL-D~Y*4&`%(Hb(CRjh!%N7!lb
z-+ML#_|8WXt+dvXIB(li>tt9fhN<D13xR@mI0IT~q)kRaBm!)&;Eq{8T2ln9%d+d>
zzPV;Ejc^}3W}F^crnkFN{_EB%cWJd~t(sx2mycUzCCTJAEUMKl?SSDGwoh+mcNQ;J
z`z_`wQy)hzYZXjhj|*%*tL0#>UYqg>WbapplSfqOoa9@DJk~CRwfGB~INRwtJyjiQ
zE-w(;)bz$J$8MW<ZGA)hTc$k9JaL2Hb;m+?x%a{MSG%8xUoJu9rb+D;1Wc-)(5#N}
z24@V_lRgRo9!H?W<`eR4@8%ia@Vt#dGS5S_qq(+u?xSf0;i(VhoMRUvchr-zIsk&0
zG3Ypm8r7xZasWBwa{k8U-B4n(UH?rCnd~q4h<f7fFzCn;wSxea^SK#R1D__epk3u>
z2s^H~DsG7E6A;z+NL0Uu09=;u&gwyLlT`sVi?{NU;LOY3z@l;7Ot0fA1tc$FgOEcZ
z&>vrKdUhl=qxIMSTBb2>J(c~z@43Dz_jP(}bSk}IEyukgNAP?1ecB1SV+N(laYx@g
zxcMvNdivs`%g<!^=D{uB?B1jAR_^P6yK3Q(#7^5f23y)n#v+Z=pJppZ)Q&%L^QYW5
zI2QTw)>x3+Fc!Mw#=_lstFchM)>v50ZO1~jY{iAjef_aeEz~>X+l|GRwpwG+$OYYy
zpnb$BVGYg5m!h=#@dO*Rq5|9pvX~Vis(8&<9W_r@rcYKbpTr8dT{l5qR+%F?mMsRk
zdjyaD_$gozv7XJr=!0ipI7zgO_Vw*AwA*V_9@v9+chrhX-rRzVR>CnX9_H2r6jp!(
zuO4;d8A(V2=e{H?7TYlQI4h4&P~uS5O>$A;xyB6fB%<5}Qvdq|>?G)O9L)QaGvFrb
z86l#P%x07Ex@!Jqk|?t>6J<PpQ%#~FPx}d`^S?)R{t`*GTknXWB}Vn}D^=)!uiEHs
zVSD?gYEMV?bDIHUzW|$6aP8bD(D)GjOXd5rV+ybc+Uv_yGKWEahfPc)v-n(kPx^G`
zB<aa*%1&f+nbrN<fzXaOO?T-n{FAszc#=1#nm^75wB>x@G`LWy_WXlVaE3_bp^R|;
z0DFQnfke>eQ1*gj$e^*PLXBu9=dz;0G$d9F$~A=}44U_|Po*ZE=ei81EDNmeD_Q?X
zDOW_NPk?v9oNGN-f6R7t!T%hL7FS81u!~CG-P7RB3V1u;<Nsus^Z$L&8}<Z!@D0CY
zxqi5eh+YlIdTXus>hHLG>v@&YLJtvr>>Gif92Jp#C5J}nU<KdfsUdME2(0?n`h#K0
z`7(Mrz#b!%EQ;V<*gu!0wc{vxX~Fum&9jP;<mF$-DvX`o9pYZIHb?}awEtbagCM;0
zXeb;+?b;O%iPA;7TH@~ayE56YQBt7f^OTG_o3wC{JuVknf_6<}<h<0V63QKtQt~lE
zib@9-WjUHi2M3t`y9QTOUx2Vzq!|Q|wd?K=A@2RZsa5`CBt)_iAL7hG4ye?2(t>j0
zCVkoYIZ31GvP^?zny!!NF_ZZi%Vk@1PGzs^iD`1x9kR7uAuJKw9&(j<F>^B<mb3$t
zk`5Nj&Nslc<R;JQKqabr&YjaK<eW9@Kc~THJhrv!lxI1}fqA48L`*wLzM|7hH-!E&
z@dQmG!t0LzgS@VakEx=R)ST$P%~*5r!?gch`jBL#s<~-3mX!~8?m8~qTP4#{J~Xtm
zrMyaIC9RMgN;co4<fVXf5~@OGEF<@^5hPV3mg|uL7j=iSId;ABV&3(HvuaXqb=P)#
z-H5tGb|r#Nzeo8sKSF}b?gcpT6h<5pRUQrSIOG~V9Mk}&tCHcPk+b7fFhRg7(?W79
zLK#NkH#0gsXAL86YLp2;t>`vZdmZ}5yfup#-!v-nabDKqFVY6KxAQ)kr568);?iRb
zO2UuM!sz(F&n;)AClO*)ibYBFXh`r=;w2rt#0&@i4HHU5W#%&WP%#mS)HkZv5++JY
zf+s*6Rb*O<z?U?w8wpoidw)eUXTox3qz4l_<V{2i+NuWJDm(k{I{f}m>Vy6i!|(q+
zHL!8){hw0VPb*<qQvS~<XNv5fCD%E+{ugy?n?a5CY#peYB4e3>+uuw^oAur?+9x-!
z@waPadtsFrNtUaL6xt5OCX|+lD3$_k=dt7q?)_c$OwzAZX6xK<RV8i_YG@{!B02!}
zVnoeriKq!gsSrcyRB9MU=VH-=Obgq#tu01ON=BU*;nlC3RcVx&4P2KEEDdSUH!b2L
zl>M|iFEpP5xUhJ4ZywH0XQsUB@voboSSz#1|F>i`dZyGD`wSx_mIj0`O$Ka<5I<l5
zX4mRv$6>mRC4Q@#C4i11|Cfw4ifftafPB{o)Ch<%2c>ygS9STn1Ub&%=65MFJMZW=
z_Pa;6N!XsI=6k6@fJKE_Z1qWG$$UTTdzqdLgjuP*F$0$gxSjy9n>tg%c7*UV#W0UY
zV8J{BF$vX_IlA>8azHU5!7aAb7>K;2inl5GB_(H+w6W$N>T-`}rmmbJObH+znl!w4
zLw9GCm}O>Ox!+Zl-Q-$MO8-EmMzLzHiK@nt8Ez%))t!AR%wxhc2}TcU!Q_(pAmX;S
zS*~fOo_}9ULK4wb`e^zD)TMBe@QH$Sj^h-ud5OfvHA$Ed=|X73;glIN0sN8)&3hUr
zEsv#tb5EEhB!vbWzS(fiNiTzeoB~V`o@Fk+r-{HJ6K71MGf+7JIL7~Qm}`C*EeL*a
z1eQQX5^GSM;n#j%=~tBS;qr0WL~ta7&1fjpSWB<GOM@j}-Y6%GAC+TdFC-uz9REM4
z&5fHV>gK$a)<H}ER3qo~59Le-;f3PLl?ZzdX3uQhS{+YrKnU7qkMc7=SiTv6AXDmA
zn4=K?1uMmw#Fg%rQo$^X$JN8XsKl1q_mFGB-Y=_EUsvrGdaGSW-so=;vyI7f*Lg#=
zqBgmT|61QB#GL}uZB-Av)nf_7-5_?8*3VdDL<{njF)=<+13@mII-+}s#msWN?XVa_
zMD&(~XGLHuQ7$ks0In!N=X#fi(}oR2aq^u^SHiKrI}VS>DMuW$@c0vixbqk7$OQ$>
zv(Rns93tWdAhkR-CDWjjid3i5AU&erB~4Au1<g}vS2+4~u^np8PR_DMrzac75XW-H
zYwTA@J7_`Y0(TCW$FZXgPvP$@@LsF~YB$v4!?DKuy3zTZQ%A?MExvrRV%|(*<3D)u
z1${3j11p?fc^t{HK7inPioeHa#5D>3DHT(@EeG2K<zm<tlVMs-9LPijqT8!Q+yxlB
z+uilK*7dno)rU~O*djFW_XD)$prP>#zpBd}N(Pl&Q}Tl(Eo^3v5ZTs#tQM~Vy%hK8
zi`1$1b0uGeR?xj_>B<7@kICJ<Im@oNHnr^iy4G)*$5W*;{IQ(NSI0L)vUc?ZOKZs|
zV#JV(8x>PNfGtEA%ai<7R1v}~V5KY?_p`FB#e7QigbdpXi1i?Sx2x?eZ?Erzc+Sdc
ziNDwdTuJ%MPQbL5y?P1PA~gH5xMh=PC&gbR%keRH6l-gOEp+K*<@QtOD_B`B->BGW
zQnu1N76q6A$JnjW9I5Q$L2i1!y)5G8**c@7d0%C+BV1(MF53kahIW>FI~VbZv0ECN
z@-c;));UNPyHU+8D7uF4VmWFBKaMhkwW%YJo2_$O>u%X5qfmK^f60>-j!NS!*SK38
z6*h|=Me5ep>9ZX@YwsfXN{-<&X_#U|RX*sP1SRJ>S-JhBi<ZeLc{(tbueDF$Y?r3%
zbI16$=mQ+=!PwdkL>tiwzBWl6^negM$8t1Hfopr>uTawVwima%OyB`Px|At)R#w!%
zX%wR@7IvFA!bmNi1BEwgo!a}m+7iaVbY(IXPVzW`&<I8O7NfE49!}&$2kD|@Ml_ac
zUd}k(90%p3Kl3u103yyD6VS_9@E>{S5%LEN&tXmwTd|x-7$1Fzh%!QbaaC4Gs<m#j
zg}@UXLaqmWFA=ekk@O&4LDH!>&5{Yl<$XkbCKhuyPQ~N#9S(iME1f$1+~Y4>5`%eH
z23aZzQ7?b{j65wQeTD3_z7D9b$YToOfc14S93&?6Ukhl>{0qMy$&X?<%P+7W@^nP)
zan;8$oVm3f<OVs?vo9P*!^p8rUN;DB6p)Z3=Jbw^@6}cOy|h5|YwnM7)MgeYVCUM{
z+OBYy^}KstQna|&5Y)SZVai81@^lw^%JMBh$J%%}zP<xMG<5YhpSLjb&Mh1Ne4<ao
zj}uU3O~)DS<Upm}jH=7;viu%`$L!_aIFh7a@cN*|`q%a$1LMFf<mP+y)cmgb-Sd0a
z_LJWWWwysi`hDSkj^G`5wZM2(*2coU>-)ofaTu6N9IT}h?p4@@Z>Ws}9c?Hi%mE!n
z8n<)6wPW<&@SkzummXlf3;eRhRHPMNq{GcSjw7{|>hFjU_@+oXi>_f9b(*W>pZ1Gn
ztcH|?EH+28`g(C{cD1?-J7`GteYDo%>m^ZYb?wV47v+}-Q=B=K;xa%TKf<3X$N53?
zB3*Z(l$!n>0x%~@IA2o`dSUVghKYGo493xBR*PpmGr*fT;CP&8^^5?DV^;nzUH-Z*
z1<N>E5Y?ich1Z5?jxE~65q+B;`3^nOrPByqIIn-BLUARgadw6;CSNc<{oC}aRV6z~
zOqO@8CdpgDX&0U0pFYRO7uSWrnK<Y~{l+~R!SR3UmIliIN6P)Yl7Fb=*Hk9<pc)7G
zi`5Sn2C&-ig8s#2oB)d&ma{XR$waq>AGa6*&1!$gg}lwp2e(?VX3kYdH?(76c8pp3
zW4-+|Bpe~7$QYTR>HmFoOCbl1B=Y}&T(!qN@BF`Ym|(2Itx8+cGPknG6gW;=d>-Fo
z9S!Nj3T`S5ae%OFJdUe<F`7k|TJveHQaMD9Qaz*xU<+kBSHkSLoIb45OfEl!CZ4#3
zlv@ys^e942*vDbc2{c`<LR^R?(kX$Q8@P?_M{B3ihY$0<X(QRlXsbw0@hHEhK0I{j
zLJ||GBPj|1x<rdKiNGTO8c+^q@+a*?;}U>e=gpz6g`EV&z)p6DdH-VA4bbW~!MxQs
z6$DCK>UA5??`;6GX-o06R>N+D`C>s+$Qg3}Yhf==3n+JGE*}<9SD*qwfeTUrbmQVp
z+9%}z<z1-Wq%1BPSRYsemGJ&wMpZBfxbKll4j-{!jjwy{+U_a(sDZnZ4E=sV^L{^{
z&!)GPYD@E76tp(?+XW&m`?e=7D3k;Zwq)jJ@czL@qwe;qv|zr3I=<~Jlk5E!m$AD`
zTh8F#uSt>zjfuq*iK|>J23ykD_F1vSYIEZAvr)gjQnn#L9t?JM24|eDxj;$PoYt=k
zG^OY;a4wDm|Ds-5jLA*63H}mA{%<Qek|efze}~J@Y4x5YlR{7{PnbT!?aJRpL4{z_
z@rag4Q6}@FCS2rd8%5cssHk62gZh`a3GU>go1ZOF$@azK@=GV-QCSkU7+@SsyaXd_
z0wxY+WchZXIV$*11Tey8FgSp<!XQr!tqnK!ge9aN?UIp3nJe@}y3cXvdQ(5ggJ%^e
zbxrg{_DatGT)-jd=8EkojLg~&q6A_8MBvpf3%ptmdkN-JG&jK%E`mf5iUY5X(wE<0
zagO53H#A}=oxqsyP(ZeDcpeH2Dkp6?UAm~!Thtt}8r!xKFYEF}BH$*%tM;Pue^pEU
z%SwJt$*(K<4U*M?XWeny6%XAr2MC^Sv7F*okE+_^%z_jCq%9v;APj75*VgA)Q$w3t
zBGJ(KIv^37HW-cPITTmMDvA*65Zg`V`e+#weclN;bwaq`Q~6~jCJq04)IdQdS?;!9
z^dEI&<7Y+PL46TUe3E@pmx`EB?X}iBl20}~;UK-bOpqi~e>dIY7x4GLqVE<QaLDRE
zE4OJce2ucdrM;jwQn^vqyz{zNR6^~F914sw7I}i{C)g8*5n%1fY_-p8ML3-n^R9my
zsa}3reoY0U2?v^@<)sS+c&-f+-@=YNivJ#dCDn-SU)F|RLFfbR&6-!}01jZ!$Sx)a
z4NI($KAtQ278{Zs19xEIY=_a$_N@(LXFc)?8Yo1-KN%EWN#g;P@jv@X1t^lmUPpkU
z#&Nrk+|mEM@go53R`{YuS#7*xJj18b&SlmAMOg8{5KXZ(vHoucT|qhM#@c7lg60;(
zL7nH*=;<tl;_uNhu|cmz=0MzxFl&(Ah=%)mU%10g|H-Y566dln9D-jEV##yG0kfbk
z9f~TKSRMH^I6vf-ub=^LUJQm!qz#pih+)SoX$jP!kD=yYNrz)Wc75k72p8Zjvh+PQ
zKNb#CuC{i866sGkij|G5N(kHdYS}oJU5Q@%R~)+JJ%7i&2kmhw+=+f4KEe)^?bGqQ
z;jo2a9Hz$uE|v;U{0>)xUcX|g(Chyij;xP|yG#=;dz&&sy3w2IST7AM33>V^Xp&QY
zMiFTYhXxrY1~!D(n`RN)KbDZ0)Ej|(feVWw@&IPSDL#+oA#rl^sLQt|uQ8_ZXl-`W
z-PMGJ={OKn_G0o1y-MrVUeQA)WR1F2VA9kUB_z-U*vM*s*@iN_QsD9{>?Ox9JhEAk
zq_AeN;Z2USkFg&BrGt4$VwF%9N14q!F|}0U#4DLx2Eji?-bH_kW3L#2+U~@?bmBL(
zp*39iG`-`S?nW{yYAMgpMrDkQo$xatAI#S1xKn<RH@gvvyD81X)?xY<T9CTe^)U4(
z@1e~4Edtl-=+o^nt-QOAMX3R6hUw>c$(Fd$QQ`89DQtyWOi)5B$&pAzO9yKCxAfir
zcS;h7+oGl(bPUyQcb@o5dd`d%B*pRnixR^jpU}gf(B)&wm6ZFhO8$wGPpa&9mHQ4Q
zruH$#PF<I6wa<*|88g&Y_6TUZ{;rXm*zK^3Ed-OUK1ivvX`hnk#g}O!>cNVuCNp;x
zaiEDpjvH#&=0j~^N1PUNDJq)Oxpv8Y>tlNHe^O%n2NF%*qqB(2y-Mte|0VVJACpwe
z*U!rGDVS6AT)kbk6Rq`OTgQ5=trP9oc60o)R*7iRR6YmVk)ICEXU1d<fr^BnWDHu)
z9EX4_bC9gunai*dG7Ob@{V3#I+ZIA>$4F+6$h#Dl6e*;=N4}xmF>L}dx*=*#^JvuE
zkMogXZ;aC<OF^8A6{le;C^~$T!BXcG!5Az>00<;K1K=oGD-NLL8Z?Mo^54gk9BM8}
z6*%ids4~dQs|Qx#dVrTt&;^cZ1!l+N$8bzXDNKV?dHy66o>ZRSYbwt&@fc$pQ{-Do
z9`@%ab^rRWg+urt{mZpsde~#Ks^PGYd@t-sZcTElK4ADbfBWYLyb>;AzZ0uc2g4EY
z4(d>aT7#}k%@0Gn?f~ytI0!IT*6cfWz_>UG7#M^YS9!kNdP=)hWL^Z42uHvxqhJjv
zR0Xa8uM8Mo82}sfCV1r<S4FQU#w#DgqG@M1kXSf<TA0M}O1J+-uo}+_t6cq}#N@o|
z^J1Wk<{SWu_in&0ZxwFdoeg#mbG<>5@>S-13-q8y0z5Xsh%vWsK1(|eD70aO!Hw-8
zgBH9aMZZJ2?I6Q902E@9vkesfNR9m-{r3eP{NGm2_Orl)|GP>I9{i4SzptdGdIrii
zZh8Md>h8wf?*Av19nlk(7(xWqr^<Qu@Dr4$3fjZ&G%dsegc8s}(b~c6TmJUo_)&)0
zk^rH#bKU-K@7Y{R0jjn$`fib*$esvQ!R)7|#dhhR@VN5&H3g4to0R-<M9z?-{LH>$
zCbwopADeJK$Ur#aWamE;6OJaTekH9yYM)9&IO@K9yIh2w4zh9+;mFp8;wZ}{00FjR
z?7GlVcCio(ZMs7kw#e^)Xwv2aGwDd1J2pd2Ctmu7X&L`NFhAcQE%TVxkH!aDW{U{+
zt>63Kl6ZMf(xucl(4~Z0fyUUh2Os0J9LbSvKV+{{DxFl)rXY;I&~|Et;I@5jNc8Ks
z>41qWKDNicCOCbUK1#%0Dkt2TU{>~Q1{})hk8VHfiHS#<XAAru=hxInz$85&qXapy
z#=A`HT4qL0E(N4-LX4OT&S2`{WiW)#%P9bMgdoHSH&C?TI)P9`Y}@PE&bxAb5I|R)
zYC$u6;j{k6AweIuhdUjknbYu<J7^^h9|uCB5{$}9oj}%1WAI6`0#4+?DbNkzyEK3z
zc{YJL4s5h)(my|?Q5@jC5-T}T=y(>9Qf8}Ox&2V0SJvq1XXM3ilHuD$6nrEkULGUq
z3}*Jh7~?$}mjI=j=j@&sSq@Y{l_VRd+|^IxK8Yrp(o1NIH3R6JWn9y$(Kt;RiY+sy
zMjAO^X0)l5SXo$0&!%8<am~cJ61%fNZ~3L`?+bb`o2Jy1UI%fU>l1_{{4*|_IuWMz
zK9v$`nxkABqpGv{zCw~PkAw~pjSEcp8yeuxEAK3*a9v&UmR9?2mu}U?LnXsa;Ur=z
z+8z;TcQm-*yR#ntlo~R^*-+R=bol~zqkXai-Qa+}!Ian>W;-HBLxV~qa#Rg##c%?p
zDIjDR?f>s;Rcww!heuS&62pIg$c-1YA`Vlo!(L>sD;icYoDqSOB5X7qID}0VHZ-R<
z@3F7b*N63L*>1jhjyFu($|G}%pSGi+Jb|mQAMP7u*T@8-JY`wmg^V!cVz1hvK<4=<
zZ%Gzi2EAQgdd#&QYolvBSx7tPiNu4{fa%9Lr+w|>vF<>IlOsCmD1Vu5EuzGp|FZey
zmOc6Lem{50=vzlvh&#ib)Fv)!<HsrOfx;~KUt&?|O&I0bUry-7&P$7usN$_VUN?GP
zG@%_WEL#h`S7h)0z>RyiZ^i3-yEe2ybE%;3N|`q9bLJY(6oVr$`oo<!)098m>T@s<
zmt1B$4I&*;j8{%Z4ttPuJP-^KgBGa^x@Q-IcGvZn%@=?(e4)9onIwgCfT3>*@PORp
z$B?X~=107Nn8N+8U^I|6c3gJ+9?~Xni0_picNg`>(fIC0pF2*SJ(91O#&&PGi&l1f
z!_?dpj;#|y1D)%hTeM&jmhU2(z|QqbI7a=k;6OO$jWCyc$?f%aSngoBHym^G?2W>$
z><#zwRXh1Aoidz5vNLuw{Tc8GeK>?>JA;w}qQ=|xY964yZ*6}^^jf`pA8+EYQQ8=%
zjWQ$iz4YmDxS#R^;Q`8bQ(jWr!4cX#8XUWsc`e&)T^)B<ymt@pt!#O3+S_aT<!#>F
z8mFWO`*>5Yx8Kf7eu-X8h$n;hILLdoRx8Xy*1tDAh<t9uI{+X%_`2>AWRZ8@zxmxL
zZVq~fUoG<8ht>|;NFU-|N5Ug~)e+uw^!sveWbW_KwufD><@_%oC3pKsf3?c>msOrk
zWm0QL*N&|nXO1|FJUkX2$AdJU?X?N-SUBNMp$?CE$65d7@E9BuxsPYU<8jIHR>@43
zuZ`>T)WnQVKA+%QPVg-!HhhZ?a?v`qD9Su@n|;&1wZog|velPlL{I7+eC;V>Z|x5!
z#@riD{~5mF3~%UuEiXNScb0aKTF-AYn~Ian+bR8K6h8ZQ6lj^X6Kl71_Imzsc$>ml
zvXW1RC+X>(v{$6P^TFx$GXXN@^|QAiPB`s76R&#h#=L(niJyFzcQ^5qiz55u_{sMW
zbSCHjErH#iQ1W+_{GbwkDbt%reH-=WzO3D&9+{7}A_9Qt{<@y|Q6+EaL2;UI?jixR
zJ*z(*m@w+Tys|jih<$N(Y;ZwowP+%#uNTPqn~wY8KdMhCJu5l~p(P;z+Qim>);D}j
z_p?Muuqe<Lv14_2D@=6dTNJd&JlfCKO`G9?P`7Ep&qL@NL0lbi<Pi=>XI;!AJ)SVv
ztG&1wHWk33aca`zIGpJke;Bh0L+2N0W%X#Qv(7@yyqV@^<}lNo@$2=<><W?kajl0a
zf2)o}do~gYtw&Id5(zRI-*~rkdZVFM@26*GUG>^(_Vl@#_yg#paQ5ldfn<o@+gK*(
zXt1?AHY6He$DNB&z!lpF&ZeBP^^_D-y0JQ#wD%OsAl<4Cx9Ung6jxr|k<{xDpf(vl
z6CaAoc}vl1+C-~FqiRcD6~}s-XY3ziZ_Y!;%E~t5Pao_1_eDafU1EK2A&_MB>B3!a
z$9_)S?&2bgs@@a;+(~8agX_9$B3C0CMJ3v%b>SE#KZr_?*Ka)L`>3NuWBL6mD_}L}
z&mvG(i3xRIQqIICq8go^@CTG@)1iN;%RHmy>|xe)c~h4!E0OZSg5>!BMVB8}Vid0F
z-Nl~yJC%sCbapa-pxl(oCY8vl#=oGle?)HLl%rb@so>p8-lIgEklV5Z<Am?gzJ0IC
zzDy!5x-FDwaG5yoEk(VZ1Av@XMVscn84bIdb2Bq5vdXl&evv7A%RS`$3VVqPXlF^F
z_($PsEb33u@2I33>YEEuV-&WOF8<$DH;w9k)cU&W%%a<N4^eT|okm4qy`(>iG2-IW
z_pWw+P}IdC-)=8iTp@gVr=#xHj$!kfel_Y%YSmO`(y+F-ukr4GtX|_ImZ*4|@i{h%
z@qR>8k!}ZR_uog~uW42P0(VjcxdO<U!f9S*OCuW^=geo(V$!F%E6wal(g9-_#Gor~
zyJBz-q$>!{Tr_uyy!9U}Ja?5FBaIygVKKyXNGLv<RSmT&%N$omylhIwdF~*h^K)WO
zxPQ>Z0WL6Bd}X~=uyon=sPA#i_$|iM?cF)eC)+>CxwJcs$(23?ERH`Q9{7j`<WmT1
zD>1?5GI@pMd=K7(lNjcU*b=7SK#d1Y08c|%$vrE5u^8nIC3?2Y|4u?M#ZR=NVDhvK
zsVyQVmZ=sL8S0CWXiyEg_*$`Ybgeh{`@d(w*aof?nm-c^!W0u2d7ePTc%tNWDj}^5
z!}J!Q<~eamL7Tm<*X7k>hf2@lFl3Tw*kTi4lrOFx;e30APK(_G0Ol4#%`iJ=&n}$(
zfz<;^fNB>nt+PJEjNv;=?3A|0AjxXw*}9$4P+3{7JYnj@C!VP^<0z%RvCtgV=0j90
za`dJDuk@fKH#i4(CDqW|C_;?F3wD7vudd*oaARVJ|F~)!*fY@i1)l2&-S$CUen^Qy
z(vK-;aPB$fZd3Av8hBF46<sbUF}b(Uh7aHX1hjeqPBiA*zRf}im))4uux9h0qLFG*
z-c7pFxiXu?GZupW5471%)2yPs$s1O3WNC<o$U6^PZ93h9AJ1y5mC-Y8HG2ll6E<<<
z0Ro9Xi~220V@rFUjB}cGi(8VI=GlNHju)HxvcO~zDC8DV4&??*#JxfQD}q&+7BAi`
zEIEvCxFX<(40A@*UiU4fcD)<3X2@>tn*>D`jtdZngfA8~>1H$S|4Rm_MRz=uP*BJb
ze^-?Z;n|d!;?mHeA*;h0$(yQTHZF=6dF`BMNK>ll>ktbzD+ZMX;5y_UJv=+gK)7%`
z2Cnuo(ucQ^AvzQ~!D&u+(Z+>K7cctH)8Pvj5Rbc47FM)Mt|@s@$#oLmeT+XkM?p*0
zeQDvsg^T<*k>OeW(ssN;{sP5V6Y=N51#8-WlSe7)yF~T)b#WP&SmR%~NH26t&$)vd
zC{cm@sruYW4Wi#J`F}*}p1pMG!lg^rJ`BQ1xB9<Dcb_4NiZ#MpPvf8dyt+HDa*1a;
zEMx<f1uC0VPSDbyQeq^)qH?b)`J9r!pyd0M{52(qmE2VFb4nJJ98t1giCOpkqH_PY
zlK)4E>^J>ON@VEdPDD0BBCRqb4~+gWGDFD9Hj%G+^`FyaGU31++wLasdOx>h?cWqF
zg|w@Hv~<INeP>JgOnP)MmF_E+dO3_?N4cleO|Dq#D-D$IEKqyk4p3Tmxjc|7<x73#
zL5^eSD~*)DJ=Mp(eX2nypgry{IeW&+&cyfn2l{AXEUEF)(mr23o2FCAKVcyKGoa8q
z%Asv0!jTTedajN`+3N)6$1W&e&D)UyioI0wJ)TUYwplLJoc_kA8xwRneTz=9Q!|on
z6LL(EIWftz2=;kU47=TSc@{)+cTju@Qwo~SU4-*7I4iYzE#Y!5NJ4s$7gz*)UusDI
zFqf-kXVD-oXcJRJH{Y)ZY{|XA1BB7t)N)M>l#3xt)U6ir=3>VhIV_NjWL+Db<0hzK
zA>S>KJW$X`=GFqR6=ADDFFc^#uO+TyZWV_B{lTsTfj#fz=2-WW7~yhW-cMa)-;!WW
zU6jy_gQg8Wfbo*a7@9i-E<inna$7^>*Ld3QC^W%)6n-d7BgDz>jDr)PYfx?CEXK7Z
z@<Cil?T}oqIU|^?P(a)-nSx#pDb_w}YkB_`8vov&(fPPv*zOITE!eZXEqPYrO$420
z5(U~f75zUA@{kVO^mY4|c&da_zR;Mlwz{mX2V7f-5BIs&+FxOPkU1%bKgCZ<V1n4l
zp7i)yj%u{CdRyNqfocIe=7QNVYnh`cw-qN4%Pfy*%^h<2zsDb2gq`b;Nwo3<+}yMl
zZLOMwgobn<oNfzjLJ-B!&ueO4W`d3m_z_j=>Oe97O0P^7iShXU7xavnfhaS*t=sZ(
zEyH_xQp-?Ou3{H!G4`RJ5p^ryrXhY9hQvQJ5+N^p#{zbaTNZ8X*-XGr5JGViq<6c`
zC?f;DZv!j^Ln`W@U<V{f(Loh;w)L+!&RBry7LXD=@v^Twn>O?*FP8yUoT11mC%JlF
ze(>Iq0kSSmp~3X9T8vAFy#a6VWu2{p`&BMgl@59&inV&rEt^y$`C7uZJ;S!-P^P69
zV+n>Lm-IVh)P?#r!Z=Z2RJ{P7Hmd%LF>Xp1p(s3TA{M+ILjrB@W0@$sMbu9wlQtlT
zib|l+fFIbk*a5R@nq5JlXh&ygY3q;$+&!h$qQm)O^-MI-?yRHJ{G>|z#Dwhu*5jH9
z2e7brwKdWTi?n%Y60nufeJ$w!s6JlH4NX}qDl>S14{QemmzicO7`UP`1F01FTY^jG
zBLe~BqIz4LW1U3V8h!_sCun_B;NzAv_eV8eb>06LN(>-dVo>=E<MeF%BgH;SKe-ym
z%GmOe+Yb4K1UO3yA^@#9Z~jw~Fb^1Vb1n@MGmjQ2hg7z}A>xOVJAAx!Om;tOS|;vG
z;ftT2$X+DXNTYM+#n%nOl<E;@jIbN4ShzmGaaxB7>Eu0*|0W6&Z%|Yw^h%~GeGO%0
z!7Sx_KmtX6*fYJ3%LUd{KToYLeB~73sUICt0Y5wx$z}=LkIg$9!j;OsVxj~gmvKVg
zA7o^=m-=oOo`wZh5lqTmdg}aBV9$*}3j)&6vJq_`Vc^^v)i^pSkwL?QnoA?lZ*MJf
zr<eR$-S`LU&wvAthg9&-_4MCXPF;0J+iJbMrSv4qUq@a2Vw73E9(8x@>8J|}K~sk|
zB^%M|LzumF%c%ENt)_jvS-S3VtnL);MuG?9p@>G7gM&h+3HzzYO<Pu&RGa3o-a%2n
z`G3&dSJb8*w$!q_)0k*;cJphVV7rCA9r{ds988}%Ablp<O?hZ)En!|DV96>h$!@%@
z&qV))*2BsC6OCr;+G#Y0(K$$?Il^CQG-cOZ;O|cU!qcx~m*<Z8Lc2L8c1(;J>ToRL
zbT$Sx(OF_khn^L^;Y+B#ksWy5=s<s2x%B~g_F+4>gZ6sRfqoCA9Is>=%u!?t>W%c3
z`rDsqK?ma!?sx-efd)a^?KUo18$;H{2exm+8-CjshLBzCFiq!tt7pMTT=Eh-j(5SR
z4tgVepY(Yj2)Y>YUExro=^RJXIZ8j!mzt(?cPMYtu|(7P3wFKJ+YxIze<;`$jB^as
zZqsyfbd<M~Z{ixwCjO*oI`^^LcT;D7h~5*;CPzF;&l&FV#;A$bbNv84Lw9+rrRl`j
zfcl3_GrB7{9PaXV0YmnY+vkm2?nt;V+~wxq+YOt&k88Aye3k6fe*$~|z37{qrt>If
zVi_z5rRm&fn$G=el}<eq?~XN^w6UKyO4ded{a9F`#|J}c5i68;dk0Ma_$&1HIBoVo
za2~QXd-(cK>LBG8e*9Bu{0<I6q-Fil0PWf<GHdUW9@H-cC)_n2IXhINSq^RkpB@6n
z9eyRv8d~PrL-y?X<k|80lcq(zTXo&@aqUNvI<EcI*wpQ?cjV;~`sd_rnn~Ce_l};<
zHeSnI{S|C1`@N%#meY}Tj24!dhrMH_Pdu`A)aK|2<8ch#=Ad_+(U{m)M|za@q^CRz
z`-FbajPy>EJENaz`))IB-|@AH&RLk3hLnTtn9UR63Eps;5m&ssGX}>67yUZ~OUXDr
zo&U!IVmY9kzdEW^psw~%2Q+McEq%%;93Fw21JLn%mY>C{2FaUVP*Q=k4tJ3NZaZ-Z
zoM`G}0Lu-=b6db+TU6Si$cP;QoHjzQ2PJKzIrP)IwN-QIXY|EcC6CkX|F-%;Kc`M?
zs~U8Qy>~GVrW#D7P{!~XEuRcCZ5TH_tYID_X*nCj3y%%qoDJFp4rB5Z70m>PyeSI|
zBetm@`5gR|@RgcJjda&r;H2#*A@Mi((<>ECE8>?lBV8n>75vky<NPvhO`QD(bX`$y
zehCN4|DdTbHswi;+(jiuIvQ8@R$8#O?#qdS{)rxYNl8tK$lyOCotFuoKA}!tR-@tq
z{2x}bQQb9hTL%%ejeg7T)j*qSt6#YRC4)+&obrd13@b@APyUE5e^9S#>h-pt+e)W9
zb@yH+cPn{RRdy(6?CG!Ru~A*_RATFWOu1c39;9-$urgf>7MJ~TmF!mXMLo1fxj$BJ
zuX6j8X!UMb&PK;e8n>*K{Jq>J@~f#|dhmE?DUC|R=Q0OtXtADHadeWr=>2a-Wkf3+
z8La1z(jFGtP<jJ|;g*ZH6iJEFh%3p@ol2;6meSjRIdRYzkV8MLL3Hv|omlMOM{-fp
zRLfk#WYcsBi8R&m*y3k8<fleXo22q>Snk_+=6xDFaWB~6m2}YSzz!!V^mTVygWCEh
zURM8_?;~lm->}<*NCxw-r;W`P>)orvUT>zn5<Z4vyTx=9d`hgi*Y6D&(_J>EyBmx&
z9L3?%F3h2`SZUM~u6y;<<x1<{Ba_X!W(8>j(QamDh$zztDl_hZNn$WJSEesp7^%br
zprWev<>2x;(}%ZPy0k<SO&xF1YCt+B@dq?R9IUw2x&={HM4XF+g0jbdB(r+T_|DVn
zgVXK0Ya`riQwvMeSDL5K@}G;T5LcH@ck?5iPn<m)hhWgO3wyRsltiS^On6LD{DLVa
zQ@Bos$oOde>_}9sa>%3kyQ#xZyF^7n^<!Xf(gulF<*=Ws#qtH@5UCm-HKcgXKuKV_
z>VJeBNAhxV2%gbq;A+=?0IfpM03sFRM-Yo;-SMCg(}~py0(!a7U9Ewu#6gy3aqZ*a
z*vs|lE6YIHfW<Ars{^!&FgEIKJ!b>%Kg7C=x^%_@ha4<6Iipexn9kP}HZuP!9N|05
z6eXgUt~j{plWW;Et1bUwlIoB(n7m0X`;Y4Gc_n+*EdzRgPVMNswlaAV<VM;qLc^(U
zX^N#fYyDTofeKfQn7zBOc`YYmk_*$CMEz5rOZ-eu@)BJj91ikFW71H!dRZ)+b54zm
zG3Zf52r^v5=D1b5mcRb<VID&HXVwbWF_R;AP`2cVl(22)AQ@soVtHJ+h4cmT5jE_)
zQdb{=9Y6*=&s1aJS5S`D?<Mqd>H0%VHQSVS?nICvA%w+{<i3=}#DqZ2tS8F+ExzH@
zyILFs-B~ML{Z4CHOX!ya3|LL<z&l~Fh&Z88M6g9X%ik_D$s=lH2orhrUgwz96jO*5
zwO3XcAOh3la@D@3Kg6J*7d;^?B`5Fphut@o>vO?|$o0QLlsgp8cE=llxk%#{oyTCb
zJ2{}y0u#meKFU4fO0@3T(jLC(KgpCZ`#;Se|7S>6E9Z&+CaWR>wBM+F+nQV7tVF^v
z%S-k2s=}KPBcfI1*y`?McG8FzoeGJ|063bPC9w2Dg;T$d|9`!mU2Ggz702z}wY~PP
zch?_r632DcNg8`2Vw$KQsGy{+szTbtlr%swrNZ_)>vjFb>^cpr4M-IFRG~gpNJs&t
zZ<IF#k4Q!0t32d^0#dXB@c=JWLZUA`A;Rx}?#%42ckS3oiY<?4?#$k~J9F<n=YP)s
zp0lttw?^pqQjV!8jNP@qMvBK=zL>YvM)cKuc6F}2u%sHy#k+{}Srd}G+cX|M@0V&|
z>LS@$iB6GsEvs}pl2@glOeid{<>N%e%2qPu86!Ftu_U-sQ;RA8mDC0QqUR`jJX;y~
zeElj{gmFbC2k%$fn40u!?fpg@#VPcDtBnmo4K|lp?kV<aeRy99lf6<TXuWAIzt@$f
zT1k^3#o2KcyswQ$o%fVBf6&HmVS4MIIIIz?<yAjQ@w+;0M!i34?=RX&J8&CQ3p5FQ
zsImGmJGj>n(Q<1`M4JO9I*h26c?S2JogIVgG0NUC{`4E9Icpuq#Wm7?)RW*o$+*4b
zNmAGCG?J(KIl-T1ZGnx_i`cZb=bc|0qPuUk({u%lbJ)}SH24+GlRYggU=NP$K~~gv
zRY$<?Iuo$bgJWmA8G+I2Y&ApTmvf&w8-3nqqsOtt{ERbVqhr(He&TEQf}+1iX%a_C
zxA$+XfvOo-iT#Iz-oQrp&*b~D!=$i8)N7C(CYev09Y$RAHg)7Pex20ariMZ6MR2Vd
z&d&zzVvxI%d8Vw4SlqOCz`or8Rwl*UsBm8_G1@O(=c*wrA2tn9F00t!cZiy*AiR8X
zkWY%M?v;nJ;Ecj~4!G)vu;6G-CoA=t7O!A&k<Dd4793n=)ckil9?28IOvm0Y$_L5^
zS?51y7Mx)$IQw`Sd<v@$iM+AkJb(pfjBCa%qL`wI9hUtD-uVbuj>BrtT5I8p`&@Na
zZgTyES$syzN2;Uw{n%{|vv(K{y7rD%u?M+c0RNm4j)#nAE(^{r@c$uf2vcT5zygGg
zWM82(KX#p%5<YAF0kh>~uT3;s5qM^S&1NV+&NGsg>lUqhusT66Ia)o+U1a0wXF?Gx
zo_BrYICY9k&V9rHmp#Va3#-RKeu8#9vwb^Yk{{0>F)In|Z+i}t+$mShai^o)=U9uC
zWC$Jd!MauBrwmAEWM;?tP>qLd_KkbD9FQU)&^=rj?fn1EmeAl8-kV+FKCpk=0rlLZ
z_j5gB=Gl8<lWxV)b-nC0y-O1~H+7G1X|s21{cY94?At;m6|Byk*fnPUo9an1(|dP`
zp3M~VPc<Zj(Bf*N+uj9qjsu;I5MdyiI%az)Q8c)%KzfU86fDg2Ad$X7O9Hm#IAN;;
z$Q;1szBX|AK<ANNuq<(QH1RA2y@{NIyw#aHNDhL*w@TT%<q1#ISv_;ZMq3C+29Q6^
z*%?PTa=YM+Ef9H96}+d{CnetXD5(eyi1|)n`3zP3Q5wYvpNq$0F<Ye4Ih@WMc7VBq
zwN_lS!0GU(%ipsV-%XG?X<%9Y>fX>#-EU<$QX2i#R)$7QxXngSbt>>wpL2k>j+t*7
zR23*&BryOAqqn&xWV``OkHE75>D^~dPD$w$!IBPSMLFSSr4&GSyo#R4sc@Y}$z|GG
zoGwhGnU-Erwf;w?V3rA>n!Qq-%UZWQDK?FZ8hiJMJNr(=sQk60?ng*&!}8v9de9ot
z`q4wyzTx7Ju@iV#i60atBz-Kzz~Eqo7=~grCEt7CWslXp?6pkeeFR8NZDqPOJ=O5!
zd3B}`F}BBvMU5}biyd9VAjH6TzV&5=pm3(yl`K@KuT;v=XK6Ik4q1;uabn3Bv^R$y
z%m;pj3!7cXjuZ?Id!npQ^q|rtuD(e<-Ka3HgQ7@8Qbm1}GkKY?O$vvBBRPp;BK)N_
zw%<k629D=`V)zLM60?u<-OI%5Bfu<fJJJskJYzYdU`z<^6ub)y5v4?sL;=j0_*VE;
zK53s#;4)*MU=%Z`G=0k>xIO6aT#ZW`MF9tOzVCX3a3Ni&QgP5}+PS1JlY&!6MiNfU
zR30rWhqBpl)InW-<x-duw&bFp2@uhn>z#g^#Y#K4t^e`g+qZA?j^*ZX#7aLrTB9)D
z@7TTuaJG1}4HQ|oF)^l>+SacvsNj~HSoeBOhsHEb80C0leFe1X(g&?*9y0>5NJ2X}
zzTLVkv)i_jJl<#M$1sIn@lLQa8=GHPdewW3-PRs$)<pGaVJju!QFK6;L}$!!l}mmw
zhLhW&CmB8iOGrzf4;|g1lS2lPa0EZjL7CuHc#|R;tyN_MjFJ<V0YH~UV^_Xs@{ba?
z4^psY!bQuDpoNO3z!G?#3h#<<vcr1~=z!<A%+?<`1z|$Z*063!ISBY9;fQxp*nOg>
z$yS4$5+HBrA!+b-5F$=i=)msV@(>#2Tf}=-jZlM~4Y<X^YvRi!mo@DLHF8%LUVXcH
zwJAxH;$5+*n5Nix4|+PYpoRuN+l#7%i68~_sEaBmzVQ6Fh$`88n?+TBEm#L(*4T5a
zbA<*I?<ps)0x6@o|Eue3>kFHNId}HkEXEC{eT5x#R#@cVD$CWh-m6IngvL=Z+kGJ*
zMp1-V3Ngh}uwdKFqL@9kOp>m|9`haDCOTt=M@P!*stW{12ThnE*|9xqX?KAAhdiuE
z2wynNk^mfQvn-U9Gb@P_Y9l}FIjtj0FZKJOtiTpQN~)v&qW3mWtDU@1fWc5+fUNF=
zypW{8tq#B~T{H<2HwhvGfh5R}i3Mx&Ba{|zw2>CF2|`4|pkQ*tEgm7Cj3h_~60wsM
zxC@Mt+0Xd_YC$8)_ZAq=1W1ZfhFobb!{|YPP=X^NZ)^<!Nk9o4p4?A_#e6^f|Nd$M
z*y^OrSCKNwWzxe@N2*E6q>(a^7?3h&kSm!cDU&fNlbNYJvKJ{6_!<h^2pN+ewz{>}
zGhbZT<nI~iP$pO=d55+Bxs_sJjh>@az$II>fWJ#T(^`+8a=5^C`gx-8pLafd|HAz9
zXZ;eAf3=ZI`*tIjmZ;BZ#5VbvAiAgt$>y3+5P35y`#2ZZVsm8K)t2;PbvZQ9fTPT8
z7ZR{`A&0!rs&)yYTCYFWxJ9tbeYO`&Kh*FsGhZl}w2DQPRkxBwo<6*C{Kcb3_&?St
zq=G>H?FL(`pyF0W?KD+$gM&cTbV7>Wj=-{%wMN}z1jNcQ!u3P`Q`8pvHRO5vb!?gC
zPt9fYYjRoTd$5UHfLVGa|LN0lOTUXL&cs_^Al~%WexS``Y=Yd>-WPa+^-jVWe*Sf~
z)@9E$ppzQZu~+<pp1Qr~FKX{aHe6(?Za8T<@9&p&{!7}->Y}#N?_IqKNtzCngJFJ0
zdDdY}xpq&uK?-(ypHs!=1igO(@OkQdOU;rQHXe;dQeB9@tf%t!;_ng{HX${K%6z2e
zGmCL=-sr)dnKIEBFpJ}^)XEb1q=~^^b943~VbZnnCYD7D=B>PB;w{Th2)ZN=VI7_%
zP-fPkPj*U3OKaoS@g)G=xS+JYN^Xq}BqZJi<LhMSHVaBu<HA4sQqxw$&4lrZ)R#kv
zd65IZFB2ua+KI&Mb`ld~9(-MSZjhN#!v-^D`mT9e=-Hr*Gv#h0GWRIbj`^8zD>h2i
z%J?~q=CrO@F=vE*B_min6biGw?hYL0yDn{BzPz$XxJ0}kE`hD_KF!x)8nvm~=JOV%
zIq0({ZC|Fqdq^8&JBR(awa-fU-c;u>dp(OmT}(yiD5y;C{I)_-{L#~%J^S)GT6@=2
z=Uc%T=wLGRsm>4I;S279@*L_N=RZQdLrGz;5-P=Xk2jI;keKm<rf^-Pluna9Ek*6a
zpZOz#bhB3`U}Whn(_D0oc`Q|Kp&S);NeK5n&IAZIRZZOxX*$J>?P^O?+Xv_vEM{eu
z%LmNt=oxQ|UY$bwNf6|QBXDdU4P%gikJ2#W(lBTXsZY==js%F7P?juty_UwK*@|(W
z-X7zAVlUdqrY%@tbe+6h-bn#z2Z<`j=Ccdgsf|<Fr5yiGlgipfmQoRi3%TqPWszEB
z;}n0V168Ai%iV~Hhxev?2>y9CbYB;`d0U4Q+B1E`27nD`(>F|F-8=Yoq@_^v*g?nm
ztA@{`w5bk$Sq_6CQ(A^xif*l}(>IhPQonG&4I~4Fr|vM+S~ZO(NIb=jn>7tM08MbX
z0RYE*4MFRJ+r4dcPMd5^3(#ZV3&8DU#5=ndUkx&Q?haq9LBPknuP`ZVm5|B!L|YrJ
zfn1x{_lR^~qVC`6Er}zKpl4NNs?YdL6Q0iHIxpp(^bAO==2{$o`NGol+L~L`YmC1u
zIx(>EbvuR6y0BDMkbAoP%IpDmy)8<gh8r8K#`fIm?C!<l8cE)r5kc?HWzoeTx>B6?
zgssg>qHHZ|q;gnaDS0#&o3u8XMP~=>23Lv-7%{zBE^<XL<+aI{_Sjj(Df!9ZF_G+j
zo_x@GTR}@+%hkogoWQ4-(5K@o`Bl5#lV0Xr^m;-TvBUXYgde>@ZNxF1P3sd73^rG&
zAKsETOlfVhu{69ZO>Jm{j$FdJ5mhwee*~7p^xS57-OuFb0(OW(Q=4e6)Mr!$_trD;
z6avT_)#i}uJlIfYNWq%6z!+mX@6x+WW4T6nc@L<_WuBS-njhH0^Rz_6`-!F&lc*1I
zSWBM|UiJn1=H4k?q*b!E3eC81X5BZtiD{44s~Ttrp&_Ui|CIX*XY__kFSCiU<~zV6
za<XV&i|qJ~OtDp&7ThHx$*6)}#G`%P@!0S;JKXOlGr^wwe|GQ$B?<o0nbVm^{s+E7
B6|VpQ

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_lr.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_finetunedecoder45_lr.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6b3bb10bbe19f2233e2a8308153e4fdf7b6faad4
GIT binary patch
literal 58088
zcmeFad30pQec#*rhJ6_fX7AYt<m}vOiWJ4+P!vgsCP$*xBGm$L8yH}K2Ih8S7U&|P
zG?X-oOiPaKB(^0RBipjaj<@G6-eYIeIrfVez2xQHEGKbn%XIusoF~Vz#}em#zQ4M?
z0T{|z-g*C|fv&Dwx2kU4T7G-|YO$xM5aaWsUs$gE%17g||DcQZPnwIj^R4}6JQnj}
zrI;7@67%s=+<p@a$x<@z(o&_grDaN4ev|XLg?uS*_fqqPg|1T9LU*a#uG8~93%#Y@
zg}zdsU1#R|7Y0fLQ9XmDLCR<6hZcrPLzb4CA70o}+G6qi{K&%A(pHNX=0_K{m9|;D
zYkvE}j?#{Wou!=%yGpwjc9(W9>?!S8D3*!~drNy4_LcTozV7+`3kOOEEZ#GJaN$tt
z(8A%;VY}{~Ke8}Z8e6!lbeCQC&ELInv~+agSn1fp@zQZi>z_ZdaI$pL;sf)i7EYH=
zTYPZ-o`ri$_YxoShP^G`$Y+zK`@EyxsJHF4c<Fv`ySL-DSm}YMV%|<~*K09vS0y|5
zVCD3+ROunoc6)nB+f#YyTC(&o*F|qH*L#(Rly4wqpSPcs{gsEgeuV1--a)Pp+V$JG
zKI9$d`f%lK+<%noBi<O-V|M)**LQh$bA7j6zuil`7#lk_vBZ!TF3-+;#f8}`^<}?O
zT%0bPuh*Aq51%+OGh08uJb8R-apA=L^3<&7Ps~<pC%j5^ake&FojK;sPERk_W*4i+
zmaf(p{i*XOCg&F?PdxCz$@?e0>8bmt?>qg#^nDL{_uMo6;JuR%Ox}0;bmjh2mHR9A
zoVx$s$$MHgm76t{8#Ohbvhm*WrK_7YadD|qotQl_yRhUhUaEMl*11+y9=iA52Om29
z;60O*4?b}J#MG&KA9%pK&+};QLz9z}-f8clhfbcJsx&KTJ@~}T#B!}RJ5g;V-JuOn
zEG{g)ceZw-KCz$<s{0ew+VrBoQ1NSpW5<g8PR$nH(Bc{&i-);qCl)5XiDzFJON4p9
zGU1bbrV?iS%2H*bPGaxWVzpkmQeT*;PRvyNu&cT}KOd!0KD#vG*Jmf@iKgloy#;>r
z3lkSA<r#n0<D!@PE|ga){$hEtI)9av!n3or`V))Q>DigFT-ZCmI8&})sh1~e<=KUq
zFjKk0?OK@iD$^6o^Yt)2TU(e|3bPFN0u7+{3Y{=LKT!|6rxus1^)fA(Sg3_tnHcqn
zDvhaE%G3VhLYY?TA@MNt>}>tXg$cc4m7bWGFJG7scb=J@m|3h&%s&<N>=_>H&rUAa
znN4IJe46pE&(^P=o%L(=7wdX>&*RJW#Y%N*(X05+JUf;Q2kl)aEA@$TZR&i*Tb{3k
zL-dwEF;y?mR_lx9dZk)h^uvNCY^6Lud!a&uy3U+=`neY;7MA8KK9~JZ>(!oJoS}oC
zo4qnyC9SWvG%-~v*Jl?hwR&Z#7Urh(LQ}kV7cqaaR-35Tt6}^q&B#P|@~!mbgtu^R
zDxq<XK8mx{Vo&g`r3iu;>o2}C`R+><|FS<@uM`(}vH9b)ezsb?%*rd)&(GG1)63N<
zbxN_mSk(I#Yg(k1>qO<!#Jm+N6#00Im0Gd7ST9bU*EA|lRIe^9`jz81V{~Y$R+*m;
zQ+kGfkjGnaYyQOf#f8cV3RNZ-7cZRfCh8L>X8gtFr4uhMSD$4H)34PN>hcq3SnTh3
z`h^qo-ogoMDv+SD{!e(`(n9%S%{sL_JzK5Rm#Y=8Vq<mhedT$d^%C}av$cuI`HEMz
z<{ws9s9#b^e~h0gbxY%;xbTkk*G?iqEQpWB>an?aknrL`ayI@g@nO=osvi^MVaBUm
znw_ffs%Zm7VT#G`jithbe~RzvFkL%8u~hNzqOMjC_(!-52bQa|TBWR#va9A8$=0@=
zasMRoDZU+_-O*L;?=q^D7`QN2qFr9XOM0o9n3wi4Ue?Qbd9UDg5$pDPyk0Nm_3_uw
z?*QL4e|>x{Hof~3vFW%s=ncJ=FD3Kwe6RPsun3qV6<v*ZTe(VCGTt{joN_H*%2sk~
z`5UoH;Tq4JP8g1fliuZUOsU)9l~RwxC#7D8MM`}RdzAVe)+h})YysZAt1?7AcY8-)
zOO=LQ?cNJO)8ntH-Yrqei{1(Eq@|2RDd&Ko_gKnS+IO#aA8j4=?)M%5=WX-e<2~p-
zM9OyVJ?wo0u^rwc-rKmh(|go=jNe_}+r7v6-R-@@dxGCRQN3rpCtr*4<f3=hdy4eE
z-aEag`P~=Yf3Nqf_nbX<zc=o^n>z<S5lh8N2bmWyg{d=Vo`3F*KT0ynnR({rgdZm6
zPmiBn$+31$%(IZs0p$b%-q8AdgN(5^Jc2JDdt&E*{Js0PKlbGh4gc6v|9<7$AN%rG
zZ~o`seQxsA$G+T{2~p%v9}ANJiK${EonOtX9ADK}3z+5g*tOVN{00vRwtDePv0Bf&
zVqW4}{G;(;YY;0Z>d7EBmzqm|cg#~P!EfqwscUhl3Fdw|6Z8^hgC4?MP$0|)U4(_8
zo3Ja`M%Z2N3F5)XwFL3rdLOm)Y979x;F)>S`bp{cOZ5TWxfUm7u%7g?^`ZJODT8ZU
zZqU!y;ygL`xul*PjC!$U<s0Gth+hdtiEj<aT^<dF2)6}W2)740<@1+=99KJnorF7s
z-TtffT|xR<y!Pt5V!UPA@_de(b_eMo6J)*o>(GtlQ(AxR+mws;6!hfwHU@kAA0@rZ
zTBMvmWc8`Vd#pcWbH!jVNCt_ZFX#^jyzbXC^jgnqvA(yykJRE!{2D#Iy0^Z6?m+z@
zZ@X9T$<-mR_jUEwVXtp>-%M<EKi>oOBeZAiTKr7xmC=_Is|QyP1qXw@mty|^6YO8R
z%j*vg1^afyMq@YBLx)!P1qasd3J&V(V#vB4d@UC2KSNKh?x&u+>-*OZ1p9Bq>qEi*
zYjJPLMkB>Dj|STbkJXRQod|Z2^CYW(`1Qo<VdBR=6ZiMmPtBbseh*`~#aek!H50_v
z?)653!#6ahyFL?NI$po;n&!v-<h-9!kMLFxtUY)`X%Bh8Mt#S&j*lh%N12LY;>yV|
zere_8)Uxk`Vdt+FMHzrpmX;QwC8{&UeN5$jAP=v&@5;$2a%ta6&%?zRmaA3e06(qt
zA6xP(-mF7SLI>wSn^|z#<YKL|(*I1QvZSm}KKFPHin18Sj{}*zW@jJ)D&-56tF^G8
z2bT?#+$z6Nsa5Ls_{zeBP~DQSoZT#5uFTGyuh)(j&j}y$zg(+KFV7b*gPMvy2zt@i
zqeOMk>>?E$vv&e@O<h=8gb+A>>rC<P)~EQ=<1dUqJ^mDL^yG_Ao;ydhIR2jJ$~3n4
zp6Aaze(uQ^i|5{5EIs+cyE}@FWx}qR*-MpbSu{s2?CB&u@-1?q+ku)gG?8Bq3(-}j
z>On{>LONH<wKs%fZpOp3SHB8cdK*`7h(MT9(D)d3<l;rXwe18mjO#1;)p!t}Gib7w
zxDor-#6@U%!)Z_>DdF%iIkmJLrmM>fOIPXl#PO41N{wBQ_zI8ALw7;nFFpDxYX6F!
zw>`G9S3vW485(M~UM?ShWPWjKV!rk$kw)&Zg#Q$I{ig|dKw)}ze!dJuU9LcB%&v&c
z>hUWVmuDFtsL=@z=({*`X~G|S7(~3x3x~Nz{b9cmk(GMIpIw;{trMn~7}9!}FPEq0
zCu+5FS-Gnd3l(DhkXxR5cBbNHPuMdxKf46tHY(2xyF}Gin2J}+^{{VR1Z0$*c^4K8
z*q7Bv1_Mf|TA8nfqs*!D5<{cML!ug)$*ih{{i1t}$$812SW?5&HZA=@UOmjJ95c=z
z)>Yo_=sHZ!SE^x3{S{^>m#3#;0Ky(+ZcwUz*x!ss#B$hmNo`O+K(r1st!HMfG4qp)
zWw#2#zImE1(zTJ0U07t4$lqnuFatI{Tlag^oL(D0M~z=ts)qv%Mx|b6R+N{Qm?o8)
zl4fXQ#c#GNOp?bJc@rk8OJQ0}$<;6e8C|b<pvr3fp3`ApbE=EZrcd$}YOVOx8TL-h
zFP(>>m}2tziwwMajhVl7!=oByg@wyn1!4C*2)bH!Q^mhqjYzT7=ffT~y7^FARt@vd
zi}lc~@!zkEea-q^f3p$*Pht1+6g-ZIYK9Zm3c$9w5O%#_zc3cGIHxR2esyN-h<}Kx
zN?j1I^=MFRiu>;**?&!s$pcfw_SDb&Pbe=u)P!g3RavF{W#zn}prT+}!HfbC75<ch
zcPV&A!Ltf%A=MP~71RmbB)7>(s|7Fp59r=S1?LrfjG&~M&YH4y>R(dwHxr1J)HrJz
zZfOlUmyKcfKdjO=9h-ex8f2xx32N{fV^_*oE0-OxD|I(=mc<dqXG=XDS#1n%#{Cbf
z6~<tQ0`gxWxQB1;Rf59a{&*(7op1YJcf8lGyW_j#g@oO)FlS+ZJQweer{V>EyZOx#
z<4^Z)r;`2Dma@M=?hIPK#H+pW(Re12NhobN-ks=8SRJ=ldOG0RRF*b1KD|UF1=E<;
ze#z*y02;y12JzKwHD8D8fWk<La`ED?r&qIHBFK)$R&({lToU>Rsvq2)uP4@0+{t@T
z;~z_Gi`_`97J}?t`sMg)vYrW`s=&)x>Jb%}Cd>!<xwMzu23()ZUP#p54po>DRah_7
zyXLz2_IO!phSpo_^>WaB*=upb{<;<om~Xo-a1H&`yx$dc(W1T}e}i{vrgpnjwNq3~
z+npY})4$1`-d33%on`vk?|A*K8ljY8a|5HXAUh1D8x%%k_RBMO=xqm-zBRTA{mPpR
z(DLrKmJGU<(E7ok+Z(!(=xFD#`oZ3;a04#TwRZ<JSMwdW1YPc#mMgbeuDKy^<U#_P
z32J(7*xNc3^Pt4%w!l?(1-WagS8p()cJGL-_EaC@YRDV4GW)!3s=3~^wpG1r^>1%n
z=?%%h*YfYM{Eu-D>iO$G=SG~8sroi=r|||oUO#iA$4fq<KI`>%5l2!e%9$FWAm?`6
zh_CIu5#v2x7F}%gUUB?pay^(4h2e|1fapj~An1AHHlJH}J8)e*KXD1M!eVi<QbhU#
ziQ*MuiHntqDFhLb9K;`|c$hZo$Nz}pL!H1LC|bD`4%IID5hQ7$fnAok2<d}lY`NwD
zP3!hfA?-Wf-DZXT$`p9M^NHb()~$t8>u&nbtBr$AoC<~FBgLDlgu7d5dT=8ha$^>3
z<$ZboJN{PpjnDqmZwQB58BO*zwOE-xwQ1^UBe35P3FSXV@MXSy{#}z@e7pa4u5a}(
zESr=jBDq1T-Sd~`CnJ1bJInACATw%0%hmeWF2m*iHj4TmA_%h)V)Z|&q!UU?f`t7~
z=(2@c1l$}3%|ueL>@=tf?Ct+el}V}JAX64+4DE#31?I%U@`6J&->uT0S75`OIp+jv
zc}A#aa<p)0e6jwt2$2O^S@E9qeP|Hmh1H3w|08-pa$<h|X0IVshvE!}iePMOr6E8)
z3^KvU278|WS*7G8z>`$jHAI!JqhI~2q^a~*CJBTv3W)@WA(!X{@aGZ*LoB25{se!Z
zk&)zf5XN>8&2TakPu$MMQ@8t*yFm}bRtmInyE~Z^in`sMH~_-wCZ4;UOLkLwdwgZY
z5J;2IXb^D!3zPgWO!7Y;lhj9UAo2YF9h3NfTMMs=J^a6;%kNb1T?z~(bQwyB==tGJ
z2<=i!+t#HDQ#5WF2w!*K_=?hgyLBhL5S8EeR=6W-)SecaScfE{40~G{HsBGL#bJ}t
z7D8xZ4aWl=Xx-Tir$m)}Kb;nCZ)I&^m8c<`-8&r>mI%lHK?TAv{tqd*so?DjewZK}
zY)2-p*jfMYac2W2@qbK(npnjDaV|?k9gLPM4{KnOQzy=d()wFMji_-=jN|`=D*j0Y
zLN)&Xrr@U(7=!cEiWxieD~kQPf?rkeGYWoI!Oto9`wISvf}dCL3kr<0Xt5X9Nc)n~
zf0<xq=jPDaHE)A9ra%#~V>LbTBciTi{vN{k8WO>|<XY;6lkh18N_{SA@;>OR<+#5G
zT3&L}_~(!-#u1o!vG*d`eaY4GHmfDuQHv}0SX3_I$~EM3`XzVFf8QNTmO4vbj_GeK
zIli*(;o^&Ei=_V&`6d}wBrsdqfq=`zW4aY_z>S+@Ime{V^YTb!z_S&9PVquh7=s>g
zYH<l6%A#M^oh-bh@*(>Xm#n*{oLo03$O7OrYZR??L2lN>2PR$>H`WV3)*DX{I^GO^
z?Q5w-=1uWw{qe-t5{Z>w8^UG<Qz;5JK3j<7_!^6rNlK)SGRTY&V31N?vYrT1Ye@;Q
z5!Ma`q-=|=rdp}1=^%|zcVsmaB={Xz%?6pZRFJ)aj;D?qW-SwB#AYB`O$C{^#V+o-
zD8XzN()w-noJ8rb#p?Mr*oCz&30T*<nIhCAVKR(h&&GRR7S=HLdXlI8A)@3w%wRr9
zU5fkVi|^$b`E8K&uyzH)?mEmJLg}2?AQ-`mCJwhNFMlEBe+)JVwo*(btW6Xn-^WZC
z4ofElRC^>Sto7Z1`Ea$b_PYAK!i5Av^!}g+7NTn#Ev7}?t9>XHVBeH75)+#d^gR;0
z_`TInMWyl=(!BeF>&oq3?Tb<hQBA|E{p~d&gzxgWHxipxYpFHppH5KoKf3sf_1GG$
zp6<fv46LiCmwE<TrPH*@D?ES>g|dG{NwI>i<lwrJJ)I?a>R{0S2u#t%bT9}r&fmup
zYR&2(?=~p)N7U0|obC+<*M_`qTC~aC+Cko6aF_bjdX=`kH%QXgPcb)!nPn*UUQblD
z+_$u@8SSCfVP;jI0g8T}He{-m0b;`z8~hxsJZzlZ@wzUgXvsxN3<X2q9diJ}<&Exr
zgnPrmaN}MBc#Lfs_g`RHz&CUwnn8Gmnp#5C0y0vx@sfhwDxRs#S4_T+EUEaRRVOWP
zWjNj;c4lp%L5sxBl5CpTIm0`}R}(?JP2g;6AKk(^GKDEgXv27U9YFBk%gsXq1qmX<
zv6aKN$UB9Tn+cg$_MM$|N`Xb}QzED)Acs_{ad{KfwpXc5N$Xpmn0LrgY@e~ZMyH6Q
z%E-Eae&}riL{k^s$~aB?tscx@u<$^QAjQY6{vxuxD@&CrIGduh6)SzuMfacK>fu6h
zWvK9;R+cDnY$TFdN+#%6txr_ihQhDcq}6w@G$VEMMBNGP7L+lM34mW;Ttpb1Ml3yB
z4g1xdp0lu!m=giAONr#kzJQ@)0Mk?R&`4n#>E;X)cyVcAT6Q0`Fuz=_U0kkI5O?>P
zsIoQa2JpwWIB1?~iZuN*YEPo(`Ts_d_=PYrdm+qKBU=yuca)koG*QYkGG$Y&W8HzQ
zuq|pCLd#%L|J$v$xl*B8Mi7q7blz*mDGLYcx;QMfI)D9Cx=bD0HI+oR^up9hDwyN1
zU{b`ytBJ4WQo1ts?sg*SVhPy7-mm5|3A=+#FupTUfbCQ1!T1QQU@!P-Fn&m}<RGkG
zFIQq5hbb`(Lzzn?UfrHR4te_k%%%RoUbk0<Hkry(T0V_WhKQDuA+2{1sSDFfni!u;
zTuTN>nUD-ZNsyXERu-j$wv;c`yiPmm<8)Hi4imfS0<-xj@_DLK$5aHd4HAQf2~F|h
zkt^kTF(_UspBy{h%=)m61c^)nzy7cld;fUxO7RK4$B!R>rRZb@QrVd{Wxi5F<yKu<
zM*p_HmZ&@g<7CBGiKaP7(Trw_M@-UFJT-Q__|A%7Ia)l40Tc#Ns2nk4!LFvQN*{|6
zzxr_T@nU@$p#o*ed9>Idv5ywpN$RnqMT84_Wb`J7Yek!RtqP4{8B6<8n)#noAW4an
zE_^Ssu}nn7L^Xu*I@U2fGsG~!35iPG?ay2{6T`lXWh31ix!<Q*sTDb#9fq1I#8-xG
zZmi4JLD6Vpr1@%cd_{E9#>rtRQf<5Gk{m|y{35V^{AS9%+Zx5(yWJp0s?laiM&JLh
zB%lP8?bG_!k4D<th@>|ik!7s#8eh?C46+~+w-dKlcG+v(DgPbc$Nwue?5hN0QA_`U
z5^NNHO|dr>bkg&P<(yJ^L$e$We}TsNXue}B``@xr5yh|WZG76hy4AEsC$=WY_fR0}
zME^E%>qN(1t!wM{P1?Fr27Qz-U8IRN7Y0*NC+@jZd!rI}>_#J>*G&s1!MbSy6p<a2
zb+?~VFQpZ%YjJbX(eOJXz3AyzmD&2WbFkOv>u9Cez2@W<0TQ*6$=ZNyQ7aoM36u9X
zAX}8WnzmemZ&nWC+_Wr$bZ+`&+(iEYjfPF1FDUkT1>Z+-Giy`c|8pfYhpg1w$zr$i
zukV$vi!G(=hv|r~^iok{_V=f3DfdT9xmo8Od&1ADFKyaK?fyq<m$myJEB5OO{-FX4
zqc;rEKwD$nEWWN>T?CQvaYF{vDR#6cPwkD7sZDbMtAeFmkU$%S=4lR=33fm>KY$fL
z3mU7}ldEYXVX@lg-i{yzvy}Ar2kA95TVC4fRlrASmB6Mb`EqhKvzlGa$&x3?8D7ec
zn+W`7!YvxvQ?*ON|MXW4D5II^94er+I8P@MfFMQ|jMjf)sq7cH7|T0|7Qf<Zog!;P
z(Z0BhcKu2jy@1GFr+gPsw!scZnxiF!a#O~?e*F8EIsy5&^3d8KPw57T5Av0Qu{%D3
zG&ut4>k3S%uN$9kBF(=4C>M~<*lK)U78~(zjlTf9@qzeiT7B2-|0Ehk^=FRMJa%B@
zr%zMYlB)%DSXdNZK&yBIz&{5AI}`U(k_F7c#CZwnran)t-x7n3gn{(ZL4G*4mc2oG
zP;gLTEF<e5OzR_HrXDm|R-_Gb@gE`;p@2=)yOd)@1$Gb%)228u{@+X}jb2ScACGmr
zrPo<^*zN3>%Z5<G_@w{4s?RDNrc&&7W$D^ED4Hx`cXhF9*2)VLwF~RVDYsO1_NAXA
zYpp<)+Y=+uVMi0A@kq1V$k9Q837Kg{Ji)iNMnJ!_23F~3FYW{cXmGJ8WfdeOBAAJ<
zX6mVWx{juoyjj_fnN4069j{pV!!~y4Z_)k2&QHe~Erbtq1uy-&)>oI-2Q_JgyGNky
zVAJ1WS6L68G3)Co1P?ty+HA&(LBb!GUU?cTo*)f<mm6=4jQZ?W!EBYqo78;?yI5b=
zm+vM%7I!8E;&{e-ed$8n&AW*iKj2l3PY83!Tj3!pp8p5R^a%xfG{hfQ&J_H^jH}Ai
z9nD;3gJ=nx65+5T&>T#HEs>#j%{^+ThAt`l@g5U|Ht&kyUq6aHjWJ^Q7ni=Eq1;I}
z2yBKj6E8w~r_jFaWVIz;mAXF_Um0}BGa7y?_(ng4I)+(`uDS5`aiL*mkTujA6oI-F
zNG3TAQO$zW_J;&U_`Wz()`#GvGt?MULfS!{!sG;^HM177jBVKts(=Ne+~6?H`1xn_
zMw*-$An-=gMy+6QjyaXP-vOu_L`jnTwTB5ng-J65W@*2cyOw}8&=dp+T!%flmb}ig
zwiMpMr34TIW&)iY*ShZbsM;aEUr~DypTw^tbcwJklgOq>K?X5c++Z4SFM|n(LFd+s
z<sJbTvrFO8U*^@!LaY#5arDf^Xh_<bcbAv`6B=Dlev7(-9JFa=j}eLL8=|lWvZp5I
zOtN#0x_?O>Y6L03*my2y#nHIGX@}Zc*RU9E!P*91iraz$u@t63lmg`>g1SXHHJWEI
z&01z{TsPY384PSL{hFF<bW_2G*b+M$l%@^(T7%9L(X2G|J0zNZ)QtF;pTK8iml*RA
zPdQ<*mjF<l<j#S5Qij1uCn;eloE$D~&eF{IB<V`lvmPe5AB_X_F?2rV-^()*%)oJt
zu{i|ldf7}iSN1GdFH{#VSK+JR@ML^dJp7@iA=c_)7XNE1nw*%d`F}yori*`FX@$m~
zgo*hux1fy`)L}CF#<EF?HV!i|nRJO27la|xx^B#=$8_VLYLrK*ya1}TWuGtnMG8JD
zl^OnODwTUPg7A&}9i803xti}avBn#hk-)%pao0&N32X)+AmwKf(Vgkp7&Ir#%S+MW
zdxIqQKD+HY&Gp#z7}q0qo#Fa^N~WlZxpYH0^9bbx=aUVqX)K*%3=7a-w*Rju=d$bu
z==yvdt7L@4+{^hdcs<p)*Sner3t<04$-d7ef}F7SES4mQpDj*aFgPM(FiIdnYR+MH
zFnyz3KjpfjGRlV-*l291vZ&?()!eAn;h3?(JA(@Un5MMISD(FZ5qz^%X(Oj4<)7ll
z|EZg&2_i@!v16@PscnP=4RWuw(jQeNk12RN!OFn%4LiQb`m?w)Xa*8jk!v)KB+de<
zH4UjwxX#J>9#BF&OnS2m$p56_2$LA0u5_DWe`F}Rl6P0wSlmimbgah#r+f4gbE@qF
z3fQ<8X`UWcOi#lGp=mkjOeW1d&kP*=8W|cCP$w`GQ7Yo+sy|3gVP1V>a#;Tn)%-Ss
zQb9B+%Vc`?itPBv;Osg}IlG{>Zn3n1#6~Co3(6Z`ILHEFxtN`U1YHHL2NOFncpF7X
zHHh%4H!+Ops!1NXLF!P22|8u?ikhfBLLjY*5k5d076l{{1P-Do$fML++8~rBIM{Rs
zyd>gtmgVsMY{vnkCLQVG7ir>H)}ffJP)yT8rA1&KRA0PH86C1U5*5=VAFJEUh0FSN
zB>y^~kwtdOUGKlB8-zmlVb-zIZ4fxR)j=63i$H_HU`;-;Fc5x--GbyCx*LPXeIfuY
z)qsR>FL#r5ATMNJd@aM9lfpZu$9<c4iR0t91}BT>i<gU2#bb{aC;8!m3E8tyVdvG=
zvJ>*s52BbIX|0BY3DjtYeFvmMo1Ix+#2_k?2#=~mA}LPVs1N^7RNC;-x}L}j9GZ0Q
zuhbWUgCe6O|CB9cq)lsP+A!|w8S52CFRiI(7=3^uD$oG37m2@sC5De8P7N-g{d+GD
zGwasIz`aKa)<8{d|7iv5n%aeY4xE5sr?Qkx&{)m6US@ADzT&v+whSH6y-RP^Xz0Ht
zLK9<9;r9rN!zrau5{ZtNj@Dgi7*XY+sg3#DJowal93>zVR_ZEKeGMB04_Xy(8>ErD
zrf0;}Cj9&0XDw%tMbWeOA_OGbUnqzh2|6PO%+}=3v!l!X<z7e9LK@JXFc+m3Q2J#o
zJ@3y`L*a(F0@Q3?H*FK7&U!sTVO{PVldE?;j4%0T=)o`oasqcyW#2^7xtpmlhf(!R
zd6IohvZ&Xl&@8*w)U9ETNM&hp>O9*N*oTW5yOPfuGXXK+mooEy8GH`V`erJ1jDSte
zPjd;^eSVQ0bpA_vI8bQDfvlfW)(>h(mz1lpxP&<%*7N8ix=oyneLoupYhzuGuG3@)
z^Nwzw^{#|H%>rdE-DK9w*h`vHzyP9~H50nAZtC<jV6eSTPLt;@S|W+R=6VX_tx+Zt
zvi~>xQ)V@MyMPf_=5_{2{s`bPVjw!I>3|TLi<)9uoV9;Pz>u*PJXS)q0^Bv^f?YG!
z*5Wr}pJt1AXU<PHuA`ivj2kr%Y4KX_I+ONV{FAXy#M$TrOd%z1De;$L9y)?l>>5U$
z)!eCA?eWVH0s4C%UnYx0G}=yRTMDBImtaHHhP3M#53kR6J>HpS2BP^0PHS>?sQ&0(
zBUp`ZknkMf9_={sU^Jk7SZm*AJ5Fx0Gtu@wXuRMj&hZl_(RV><qH2mzo^q5=x-k{`
z5EdShM$-129Oh~MOC+uwG3X8Mo@HMG%1aySHe#Y#<c^K?XEk-^n)64p;*AYC)oO@e
zCC8CQ3(lVN->(u$oBsZ9(kNfgiu44&fw*ZBOqhX1O!NJ!i|!RSNpnvFVlJrQq5|uJ
zzI88J)SMReD>FqQanLxkdgoV%)5A8qBf-<UVjX><xvQ_?aSa^v2En0bs%DVL%7(_+
z!SQ6v0K_Q*lWzS!t7bl@VBI<#WLRts0e!?2v#H~_8m`rop~<Mjw%Ram%-4Z=gms#j
z=O`E8VOdZl;A~lQ5!VYgwiwu0A}~ZV4*RCYZ;ezNxYnWFszbcDMkk9=ROv^!8gq9*
z$*HQpc$v{PaHoFqB@=2O+N@*FYhiCy8{unZ3~<X72)m<QMYf}UeTNK{g<8oLNz=94
z=d^l6{I=VuAZBJ|)Oux;%x?(j7%4lYPHcSoi0F<mU+oGbp}nI9TIe7f%*1LIIbjz3
z#cBjmgRtAKlSaH`&1wmQ7WR)OWHtiM%K{)0kHqTPYwTQrLxBLYJ9!KZVsq>y0I%K$
zeiDW%M3Sqsc9r+gPQw3OsCU)7kzEyL;8Z+F17u#vDKR?{{^e<Fy%0(T^4P8e?OT}Z
z$LNKf2__ZmhPN3AFtuQ}q<E>p0NZ?81-1=R{`ZhF<dO1es0m6AGxXx31a^B?dp7{e
zZ|MM(J=D`BJ9%pDwOYHmO8dW1-?BDR=iS$~$~XwQTf+665q>{;9sb|IG=?0nA=ML5
zei29qHtus^GR!y?sIeZmGN)|Xfs-Zq@*g7zdz!jW&4*I%`T0e&TsX(#zgxK^W%4g6
zYgc1Rk+EY1Gjm#?cZ3TAy;*JcYX`e+vrsUOW*COH#VY7xaP4I(Gg-|@LsuqzZd>2l
zd9)^29UJvOrLs-ZE=;1lbtLPHYPvA4Sv0hX4?U%6B|F)`Pbg+Y{Rx#Y?&_mTGxd@|
z)T6rbA=20is(oWMEG$krJA<*lHu51pj^0u|>8nK86$u|@OW{D15u1NQwRLh{{yUYX
ze)gryaeeg>#SHKo)8%^MAuZGgX-WoGNe~x<@puM9@7}~ts0IgzC4}rvq!Qh*QadpN
z?}eJ!&5|BWu&n!RS-1MNV=y~_oJ4!f=Gbc7iSwM1nuzSV6ozX0Me596m_$7@mz~Q=
z&?6CL1o9H&4DtjoF_sT|Y=@jSg=w>$8xdn+zo?%6n1c5!F!myCq*s`J+?!Z<sUg@>
zMPbG$w_2DaN85z5P8x<O!F2~4aLY~7tyWfdbyKsXni<F~ZOU5NW?isZ{+pR`{~k5u
zw1S5es2BWiP$2ojSmGRpIcHfPP*E|W=V;)HMoRXr=f=kl`4dFKLRq^!aqYp@&+c;h
z;_}42+X!lV%WMaLe@&I971%7Zch7>Cwei$-(PiDdr2xPa^OFi7&|`j9K|w(`0lQ?e
zz;UV%6OKz3Y;({}XA=>abl_dOF|NRtjA&;6yn-bK7ZupFH}&18xeR-v&8$-uw#v5K
zNPUNzWh?g-F@0;_LXe9w@em;Y_x<($`^0&yU2Z2HbYXWq(~F(uqq*)}hQA!&d~ZI}
zO?poEP<CZ<=|ZN^m)m3a)0t8B-1X#AnJj~S&{w;r_;!5M=law{mAAZL;J+e~Pv%qk
zG{1>_hFHog8H~jz2L3Hlg6glu<jbOxUCZ5Y@VCqBM(7PGP|DM`0q^}*!g<6fm@f>r
zJbLni5i50@2MjhbdM|rByq)-yU~iPSn>&5p9<Ru6zqi-h$M1l*-#ftXpm)$a#P5*z
ziut;T<K4oZZLRx&Jxd;vz=C+WusdEE?@h;?mkZ*@y%RiVYjp3Vcj`4T1wJn1_d-4{
z<ntnihl>xI*9&>Lkk1QwxRA#SdAM*MFLrwq=JO(6+QYLG-lX^TN)ewMk9+TUEnULn
zlX^t0;@*@e{}<|$ea-YI)T^Fn{xEcBfAbC{xKl9?8M<>Ix|6WFroCspXRWS-F5P?1
z8@G2j6kShv@AjUz9EY1FHHK+##yjV|$L<{Q-s`=@+l)nJzR7!^)p3`3-MAa9c|PoW
z7T{E^H@8l&6c*&w#x`@0`}Z+ppts+~&s2}5PxKMbAv#F(PfZKQd|OE20i9!BEs%Dy
zXVUy*1icd1T!`1Um;o#Xv2go)f~>#GOCnuO!Ko)aNpR)O0dbiZBc+$!naBkYaA5jM
zoY_bm<5l8$iw{`5Frz$!_4GAy`#m1JENA0EuY_FCCCGc(!->qev4?Y|yIw$8#(vGY
z9*L8XC*XnOrC1f{qCZbQI(pxS;cl&JzB~G=NDJ2by`K62v<iC|oqm072m!yI5L<)t
ziDFxjJtz$dtvtdGP&{xzx8Z@q;@g5gvq=95%52Bbc2G5W@)fcz==VvT+fn<1nG)~x
z`pn#P7iqg~AL^c9!2k97POo3-mh%^_FJp6i)`~aOZfGE57YD|j)b(jb#^0x4KS4Om
z7XBLBo<RU!W9NL?|BUYcErPTDw<-FBf^Sv8kfP+R;Y<K0Q8;N@^w0TE>c&}uvAy9|
zWMH^ml9Ma#zIJtj>r7l54t8!$P=-cjlUHFFy)e17xa7aA3SZHay2yepta%9th1D8v
z1B{E2Cn9sO)kuZ9aO}bBw6m1+JW7BEOIdiG1-xOg8=6h%9!is%|6OEpn;V2|Y{SDk
zA6PB71K~)!bF5Y~!qN8h$XSxf(Zj)3F3ANY?KI_VsdsX65!0^43ze!wFyWAF@-gK@
zvew+^?;EnapmvoN3@b1iVP3H-1Y<)+Y?TJvy(2`gE16`YOQIymrWywQ>J9(1dh;6!
z))aiNf*(*|BW(R51-<{>3R=8mb5gX;2}67b)h5G!8nX)`Groz4Fj59CIG51=a2a5@
zzY$XPm(U(>lm4*p8v(ff+?@=nWR=yPapBGsTlw6+aq9Wn%IG@K)O2&vLB7$9m<*gW
zcWW;Y*j@wpXp@B}w8Tk~$pFBpj7ifPW=v}t=eoxMfVu1q7OqG=QSdCUl$PyW_G5`h
z+@eakMFms(J|5uft$lItwoK5!+06R55=>by%3AJ27@S!T^In8*@+ZPJ9CKD?0|Pa6
zhyrNiy$!lH=Pk_*ZsaZP2i4nxxVurj564&bt?wIiDbZ16N`!6WqaIT6B;VSv6Yzcr
zbO-^0fL?89p|6O&UpjYwUm+z~D-)tmz${`H+uhj|&vthtVg{ZvxJhEuJvjuL(T1YP
zf*Hj&UrM{KAVo~_w{DqFto6{l0OAkoo}3*eP<-{>0JPp<<D5Vt^-8&yYFN!Tp3?wr
z>wm{EVxptnhPRcsLu8-FxK&=&uvsonEwck=K~86YVGV0>AkVQ@3R)as#YDA&On?6B
z%8vc}_ZQ#uqFl_pc<#(G9L_xP__Mkq-Jr<Iu7>9ud1yP{4jg6bJ~rTMIpO2PIqY!)
zUMG+Jf|i)AI9uli_?iry-(=guLFY6N#~@8tDB-9?7i|emk0)}ZR-a(g?cbyqLiP;f
z=nhvVN!}7ghGthw7KTSRoXyCp9C3++yJE}20rxn+WmXr)SNzV!j(3wzVUj_yzo+>)
z$|H_oH<?2gCKD{Jek`$x5C4~y1M6q#X4Tx0@#xs6tI3<=Ym@guBF;k>Q%@o%vq8K8
zi@@a3)MbJh-I4U!b1!SGHtd*)7INC*o~zVlt1;YCJ1@5>RYPV?4=etU&`XWJd`g+$
zslagSyA*py!LtfFJI$D#UsfL#xH*#S#>4~x)9sbL>w9J64rNL;Kcd?8)&|@p0d5H3
zf`r7{ys7g>mMBZZl9195xzLakg`*lwvVmzIOU?b#SLMvbL1aPhS8|i|xO_;O_rI`D
zW_aS;@D>Ruxpej%?=Dg=Ov!&?IM6-|%`(iq0s2p9cK;sv{1QRKvOvxjg@z2D{f_Sb
zCk2K@TBzjnl>H~_sD5rr5+@{~%>ubZ>Z>b<*LPPt9&T$*ht#s8)5OZ&vC~XMH)pz1
zx-Hx6Hg(UCkyDBBMjzN)O3G+`<qRpts9XZ2+LVu?r8)tbw4opTUsUh>l7c&S!uNCc
zOX>u%?R0{+x$|Yd9C@n_Xm3kL4|Iq$zCfK!j#%wci7J5X19fbcB0>^@MmecB<(|w*
z4{%RY$+@>nJLhsn#9A(yGfMu;^uq)~WlbEJ#5oGsLdrAxF}<1uk|)bu7w?a1j$cUD
z&KSX}rzC6dGBUSWSECI;^l;D!)wZE6msbwH$M*YJBD)cAC&i|M+G25{Rusl_Nyq)~
zq!(B6=Nj85S9T}^ZpJ#O-pF_7Su|@#6&XDK1&z)m{p5d8!Gr?Md;jNkL!<2G_!kw^
z9KRV^M%mcf1mA3Ke;=j)wT5*cMMMI*Eil@sl1pl)n^Fuv2GXr!vwG;Q2HW9wy}bWE
z>g~Yp0*(Tb1}4rE!8x9cw`Bl~aEdYM2NS29ej!M&i7f*=J3~sbEEjWi?lZEPdIp>>
z%$;In*_VlcI!i1I&Q~lCOO-*8OgzW}J~9$`{U1CL1<qi^6D&qBL}afpv|m$DCm1W3
zg`?RlhJ$YUwxhNdCUoSz=qhb_otV1d3k*6<KxFJ_qC-8pw5#!8Jd3&2X)r7@QZeyj
zDY4@Dm(>nQ>DK|FaJxzT+oZjYM3<J}KWqFX;+2ULTLci)$XkW}Ok;4@`Z2q6RUJ!I
zucKvZLw1IUhN}+mjE@`kHA5^Vsza8B?weHjO;QRFMRB(Xu|-N~d72edR`zlX2fK`S
zY{2Q-edEl=R<;~&+UuBWGUpKl|H1{iiKe6RxZ>+=og~fZ+Mm(o&k^9cSpM%s#y|-h
zeeqN3%%9%aqyCSP_BC~#Ae!r}3^1*b+z8R`MBa3C7ge<o;R+XmXeLM)bAnol9sGUf
zniG`)?x6OO17mEl=?HT%45(+wQKWN$T%z-3yN79+-Pe_?NA+Gy)J_|OlBz`_Us4S2
zSt)a2>pB^WcQI697B6ZLU%tT*D1Wv*iD&g?E9d`X1=gpB)rnc&S9u!@KzwTbY&3$;
zgvaoTTJ{a^VmT&_&BcPL(FH+LnF2<4gKW%B#W(NT%_`cUQWwg!nfP8JvOk4(K+>V^
z!A~ysbL=gUeY-5_FvhX_D9+bNajAfmD5n4qyAPl>u+wn9wy@hUO2d)sB4eCaipS|0
zs0CZ)8BEJ;N93j1%4Jhk;29Cm+R^5(`1vq?+25<8M!6<&cDKIYa=3s#FMrxw)K7wm
zMVKX10Bm3-8(BMgLPeTA@hTU3Qz1?5_ER{6q=$DV#q|UeNcQOTM3BV$hkJ8?rfgMS
z2MNm7MZH09rbc0(+R`&h=S$xN8S5#0hmk{}$~o|wTlwWeG5SP10?$8OJa29`*WbAO
zFgs+u`uSFRW2c^T<_$Q;voIh{i(CwgX;+Gqi#|NT(PpXQWt@4RXOnHUmHBP#2b`!!
zS#94fE{N=Qkz2>_(@j^~lw1t?&C1#=0Kb#|l#cBk40@QmP^m0o3s`eg!X`qdO*i$=
zv}V4f;F^NpB52sS|F#l-hd^5V%ToTz=P2vMuYzuVS2w!#mgV`%bR9Q(NSSrza!Djq
z%2zxbQYv-*41&6YAuPKt*{NA1@Kwx8<uuA%VvX%w5AWNHbm)dPo5XV4x>_T`Y%!D!
zvy7b+Ys=ZPN-t~z3O3qf0R>yk4AQK^t67qjhtD~#Sx>{<$Fvw`T8y-;V7H;zuv)oG
zty!zpY+Su`R4a2y)~`WUwQi_Kve7J-t2~tznjJ6onS)LSKXP)_NSL@1CD@3T3ia7a
zdD6$pxnCM;7Ez+x``t*SSSt}`;-72qY`v#csx(+zn#T{O>1-RDmTl~X8jTv*FzjLC
z*{gi7n4UAtyXV2+m%5&f9xgHDL6x`Q2(hq7NT!+fE{*I?(j~AGnSF>Y#u7w{LQB0l
z%C=-ide{1q?x@51p<}!UsZe8X<2}qHkTb_#PPx5TNVL%v$_4;fVy2qg!Qbc+<z@mX
z8m03ar6WEuv+_$A56Qgz66lG1;^{E4Na2J6Z-zjY9IEGTNxN5jEG3O6bC4d$5pjwv
z29Sok`$(E)nC|Ql^h9a;$(g;8Ud!=|Z#?H_;#z!--Aq9C9RBiB$lUv)>vd&^W7C?8
zeV>U}d#$x8|F5Ff{?%WjwMLxM3g&gxDq`qX+&gMj=#H6sD#jhX_rUtMjPmJ?vo1cJ
z;GGAoPukzvJxA}Y*w=q|<zmket8QZ-Y^bZz7u8Q%56D5pN8Jc2_TSJO>Ftf)AlA_v
zy5oAo-MLe5C|{#DEa#@Zp<Fia0>!?5Zzxw@c{k~e4RtkpBT?^yKJCRF2eW9J%@AYx
zm5(*CLt~Ho!yu0tQK@qEi#1T~XmRRj@%+)^<*`=YCNi_k9N@5I?y6l8_Mb<kfJnFr
zwrxP~Kl#d0Ttc+2PrcHruSJV+@IY5vt}y4#&O5gp9NOVw4vjBbZP)iIVHevN*<kNB
z8T}w@-Pso#pn;)0L5efZFv9sdmtp2T)oI)(gsHQHekk~Nm$JR6n8}b0-if{*8+1gq
z2ps=kDfqVpVPbhYOn7`JYk1S0@|y_IH`D%qNuE-do)H1iFtZ#zQi=Zmr8;^VB!InG
zsk<%v>GiO(&j@nW=jwS_&{7;qZ+bLo9{S*eiiralS{#tSKCVU+DGV_(Zyrw`P3(-{
z1);xLY6>yf%AU=jsx`+wZ7R=`8lQ%mkru>ik`%QBf1o>fQB$Z5VC0Lc&zVOBc@Qhe
z5x^wH9iDrLVI+mo6Ow7O4VR%4rW;oFyNKs&-^HSpQt}?I3UJCYkm|aCuN2bdq3ChY
zF$i_FyShcoo0NCLj&)aio9)3+srlUQD(+h4EOo9MX?n_E#@T;&y%(;cVzEB>kPEb^
zU&d9q5S3e1;*Ii_mI%7>V&mTLBd9eS@5iA6a|8Z&Qese?3lghdeEi_xUvESU`&nqX
zAVuYx3i@V~)Q%i$mGah}0slukOUf6Jm;Nj!TP*Ibpl2qzI>0`re?TrAx6+cqU|?+v
z_Xb7!B4W+?Uvw!lD&soi{}$IH&P*&AV1a83D-Ne7usvR^mT&<MeJRlx-YBIJ3zM8;
zgDaRW(=eCelPdF2`tqfM^s#)!{oo4S|4)p!|EC0FTb;Dl1m?DGYz6(O?u*a#e_Vk{
zVNA1S`m7sDt1xezn8%o~Pw7&=Ui{zG-SfmsTV&q4jPo7VJhUrrQD&z(S7!?*AZ?tN
z`N78&<ku;uvv<hqxt*6zJ?8}Ne?o=f_Q(RNU4-Sp1~!XMP%-61`0{El9TD2h=BCZ2
zCvn~Jf1k&d*bXX(8)YZ6_kT)F7t7570aN~O=&IRMrPP!vOKP`u>I$2!*Ndgqg;lL*
zSpfH9dM#C*1owl88?`INNnkTt(MUPuer1@Lsp&Ajk}VN?J(KMFq>R{lsqUQ8Pcf*b
zZJVucqwhK-R-*6peS+^-K1jgU+Wqk6u^Z8zyE@wdxke`kL4ZxE1o?37FidlqIkZmr
ziIN*<`G&$q`&vRrtE^%_O^yOii{-4va-)Z0rs9VqJN||q=s1$c>4&kn1K7{23uWPO
zv)h=HCO<j@ljCpK-Gd4wBn)$rWT{db6quB>NyqUppmqPM$*95tLl^B7F<FV!G|JbI
z8EWiG-3^x%Q_}zanmt6VYm%%s)xJ+drQMm>wE<w|m#vrBM(y85QQ94Q+rOoU{gr~R
zDM%n%S&ssCZmh2?k4R>w_2bP(F(29*b#(p6zf~o77Ve1SVi1k<Ztj5v9H{$EpmNh`
zPB+_T;|nJIyxJ#u5!hyfnslRntwRD(pqO~JXZYF{D*&N*Hrr~$zJGizi1&)pv~>X9
zO5On2gd;tad@Llcr9nOU#wMI(o291A9P0ie=W1bkUudc5<gDMTTrq>mFfmhi?`CWJ
zcWH@P6uPY!Ha}XEM8<5ttoIe;Ap>-?_1m$2B-%IZ==B;O^|kP==#bi<5vb!O=rBeS
z219HDSW_ASNPS^X)>TdhF62vYhnacHoH%1=`;45di|mf-GrhcaYT?C?H7f(=+?<Sp
z;{FUIGrzU7rYm428++?C$!bG4Ibw5|G;b4Tw>FvCGVBLV$m19?r=(BD+*zj1$}u2D
z+LVL^u`_HQVu+%haM@@l9QKbUlcFzUwqMNl#F@1)B-tmFYV_P6DrPiZ3kexyct%sR
zg?~)1U>N3Sb>jyW+@s)xf>8ysNrYfLq#CnYfJXC8=$1wuUtFAA;O0crFoS)QIP?cW
z*30#YMcZ%3_O$iVL$h@Jhnf-+IUa%-!#T1sx3apBzpO*?m96V1NMyOv;U7Z^g$9T=
z-z0VT50g-5sv{Q1n!VN=B=N!^g@U+Ish38)1Xf-0s(@dT+322z0lthHj4wm_6=bm4
z;>52YRXc@B0a*qDSxF(-w6{<FvPVHt0X?GK9qVv-v%~2TIihOlY2T?{{I3)VQ$nI)
zA;O;mwDY<*q@Z&lg<Twd(KzGgchq~sRDe|qf&t{#pv7T3n}8ik8x~hv$qhg^O)XFG
z-OM)-C@R2GpPr!aWJj!h?>p!PGrIM^MzMw=@NMGuN~JF0jq(+Qc>V_I)u=Y9%1#{1
zgJ_?F%lLbGA+38k;n+;6c&CCLP+A~HmDu}831(a-e7d%Sn3x=JfsB-q=FQ^F2DcTY
zwJR@zpXe`e+KDVwQqmMsE2L9$^0=1wGB@JtfwAnwH!w(rV)S_`@|(t4GdQ|`=2?87
z`3vUsUf$Hqua`RwhCT9#1dmP1;O3}YnCS>U54UK|swQXawPR>8ICFBL<%7)5E3!(b
zCaQ;#baJ?6w0VwtP$SOP-RVmn=RH<E`9(92Y>}>|RZ)Yb#mMdJx@~hi723)+Slo%C
zZQBt?{QmJ*^uB1{mpL-<2trxC0AAJPn2pzna^f;ls+?9`E%w#{vqu;&7~Qmd?wl%R
zF@>ksPPyyTjqB6qp%}zuK+a&^Pf*T*z`GQ?rC^JK0R`G0V0#)GQo27^!rhGpSjt`m
zWXa2quvn>&E7{;T{IHdB7v`CH4C(sSSa8YZ$tCX(HFt%U3%Q+uFY%Qt+t-7yR`w2T
zF6gCMYcuvWA}jqS(FZ|@xCFh$Uq`+#WC9{eVg))9kLl0l!cYuS^9JPOX1U)dcc4k_
z#Niip`UQwcC!aw>vhGd|=MXrfg@s~nCnv<OG$-Vf?i9?%5Sz2hiQ=haXNuS`E?q6!
zsXjK>+olDmf&HHw9sQBnxe8oYzBMtT3EDU(qV{lcqRo?FR4)q#{eae%dpZ}<VVxT)
zn)I<Ln$meR=GITu&C8dB@7(-p4nKpYfZ3_jXKRh)GAnN4HhZD5(%1hsOBOkziX$|m
zW^rWG3`z<qHX8?jwzaIaj^NWc4#`AI@`qG>zuONdq0!OesiV%3BBzu2C`5Q$dj-zd
zsJS+`^{-n+?`GiGJgb#~YYm;9YGc%<wib5hfP|W9N^N=kB$->g%Ede`vB@6DE>(q@
z5`}&1Ix$RQq_%!d43{@KFZIu~I{JA+B3}ei9ZHCg8@ydeY&5^xtoW1`R^zjuh*k?#
z#%fCtC0Ye+odxZ(Cp&Le_vpRM*^QZdKvrpgwY+plz>-3W2xLdwAOQ>F7OuJ=^ODG_
zWmF(DK}o6e0hZURk_lo4h_u@rB{$RiOxV%eh5UMZbO)!)1=ER1w1GZ@cF7U5r9qTR
z(9*+iuMwx5H50WJ!qeKSGZd59r)N2DP;DIy23LnTbCFhH81ZgID8s8;(8M7+fi&fm
z#JL`>IKc<p1f2s3IfAwJF#0>r!_f}6RnDH_ZHGv2y{)dIx1|Q$A9H`J*Ja+qMpw73
zZV$FwySrr|*=RGw^mfj4l(1zq*p3!3e*;Lex+B=36CPKGF0$p;g9uBtHXQ$E`#w5Z
ztIv(V?siiXXF1Mo<>Z>3^s<X@v-mE&rpQamcK<&{-c0u3>KYFzNT%gkZtL8(xgB#m
zSBu1VL8t99@_u(v#9eaV>vAQ$cXeB^du>m!2b%-=G1)uogg<+O-R24FuT#gqwmRhL
zWS{KTqk8tadj4m4JLv)DJ8vf&3^N)&IokFrNA8l0)c1R|9v@)x8}6JU((%!<Fq@!b
z%J)^S-rDcV_C`zmeJUUh1Bru~4)-ZR5(j5hbcRdF+Yj+m;%GF8R>aND`A^NCQQi*{
za5kZ!?EJ)49c*WMGyh4_SF$HOvql@+>)4d<PziJCWK2h&*o(N*1dceQ4ZWM4D%Ju@
zzHv6|n3`ABzK^R&Ugx?wzf;|+$7~sXOR=L0zDt2|=YL<ZpHpy_z@%H3%92H8oa)V4
z!RclErxg4JwU6y@<AOv!;8leq?u@?eQPV}q`+3C*3bcL6xuFuBUy2;mMRIv_V^`{P
zPV1sHPSHeL$$5m%zM#vTPur&TTB)z?LYt&*n`kLx)toMkbkt*BVvO@TthRr<dM0Jg
z91``KpHZ`hb;B6;U20pmEAKWe>wvaM(OHrjp<$|I&J@_>XYOVE?W*Q&Dx7hH>af5`
zReVGpA;vMs%reXS0=BpPkn~vG0`_Gzi4(-c0V9)AiT?B`;u+l?L3u5sv&3KQ{Nvue
z68WWu5Z}QSN_1DaEn7pKy{7mm@r{=EjkRN28#=H?Lqz`st*iYh0bZHJ(o1Rr#UlYa
za1@V+#^CSg4mbpJ@)2Y4bK)DMg#?!*rt$9-<c&Y<VlVJm&;{)2GQqW#@gJbHG~-Di
ztsr?%RhR~YMYzt#f-bIc0&`6#*rxn<1wCla(Sgde95MAbkyBU*Oo7aX9m<5g5}uQu
z0~blJnl)(Mzd8VpNF$<`i+dos^j7Rvszuj5c5U~}mu60`>3TJA<}S&=$YUC-UX7Y9
zb1ismg={Ixfssv(*=e@IV75?kt5X^vzh6acE!|eEpmLFf*(3zNp;TKDf=}90Vxr-u
z7*g8}1{xyD);XJa6LTWulbn~XN6K6RNQ1qq&M%CjW#`GKl+rwP{*dIj1b)M@%$nI8
zFz-lDH`zb0MsHE0I{9fILOS*xnx{v}k<$5oh>vj?$m}X`WjjC+EX#yWvsAlA^o4NQ
zpdec3!_yZl)3F(CD+yhEZ3^nan>zAj$ZfD8{3D5wxkF|0bu&ftXihTzUJfq?0W*Pe
z{#`*o%KHII4XzGX(_En&(N*2l>yQL#+|=R1b9D>uVUnEt@MPU&(@ZC&NIJa*mo=CH
z;S@A)PC*N3SmGSxEVX-#?Ps{NX*vZR;rWY9IBgso9JUiHV52$WZfh(U;%LD%hcL$t
z-RArc-yz?y-QTnh7>~j@b8wSs<R8@JJ*2?ge3++=F)mm7Uv!5J7d_Oi9JY4cI^Qp8
zz=hBpm$_A!KcvgYiQ(JQ*hk|RH5%dInF`<yV=fqgmpC+3_8B{r-8jtaian%!Sv>{N
z=tN}ORIB-b*sj={1So<IDfedYutS+TUnK0RR~F!rCz$&csX+0rU2Z(1`I62xfI)3F
z86Q2S<E^;~E<o<zrLM6h4R09pciW@Xkd`1;TabTR^IdgpPvPUn{!o;BHI>GkgxSx$
zXZ8=`1vLeXPQ`bkWZ4YTM&2jIU^hPMEp5LP4N?=R3Hoc&QA*zj09+lwodR-Ge|Pjd
zcpWW|Gpbl6J|m|}%xi61=V0oDOh8hcukpzkB)IgRr{trMRRP&<u2qN7VD_#KVMRQQ
zPa4(yN<DKujx&mc|MVy17spKdT7Ge=!*(Bmp#PFF3IN{5-D+BOTsLC=lksR1<TNbu
zK!7IH^o{j;UWaI6+%jM;Tt%vL7E3kLo(hrPts^w+J%Ia<2YCcd1FXJ$umv)H1yVgZ
z*BcB%!jI@ZxawoYq(rQ9l={yQ4$KX5x=ONkBa$vdpu_s}hLUZs>d40}*W<xxJ-Ie=
z9RgRxJMD$I4TeZpJ)@vA+7OH&&6deTx7YJ}0VFj9aqD!o|G^=;+Lzoj(4FRjt>~&D
z|Lt(iUL8dm44Lacu?+o@>yH0Ke8B57#~ptb46khqwwZo4<=;YBHgXLDF33Qz%V>;p
z*Fb6)uM@KsKi1#o%WaC8*!(b6jHeTp{>oVbFl_9VJu-i-$9O3wf+V^%IzjQqFjc$d
zhpWZ$xw{$~=T8Vb<BrGZk(!(K7^lE#RMik?YN=o<uBFS*-Tbtzo>N@@5v%7bYHxRJ
z<sIt<Jo2OkPh8_j?+B{_MA``Xrgsd|oLkeINhdFsx)%=_6HERDJv#F7VU%TS^;Ws{
zWj$Z~a3+?B7jNr;!Ig*pyW7#xJ$kD&U+bPxB6dhiPNUH1rs$v+V=JtyoRcP}-@TU#
z+Is7~)bPiR-WrYlx3{u+Hu$ZP=UW|He5!&4r9o+PUTl0zm^puS65AJJk+c;?bVnPm
zkXs@Dgn|}8HIm2po$bD-?5097y}LpD{Th5j4=suQHKj@V;2%)z4;1?+ikZ4ZY=~3B
z{;bmGbZM&oAJnBWF2?Je(4~2Uux5Ugd}BKt&ae?T#Pm3+&PIDu`wE2j!fp(Z$}-zl
z5~oYZskH(8&E8LwH{{rcb79u_l=3<Enop})JJkf^5RNEYPG`l~z92CS{(kjT7eT3T
z<)rL&>a+69s;5iy?*{onWfHX?ZrOC+e~Ga-g-j}~eY83f({TV8fkVSfg)jV*yL}|7
ztqO_6;B7fK)@fckCZaounjcTxas(x>YX!=fAJ=%{tGTqe1B_iM=al_$0|)<Ft6)kj
zXX7K1Or!Y(Uo+eZdLo1+0~X||$bSqCL6S-3hxfk-1>q|I!l@qwB&)lAbW5B*Y&i~6
zlGB_4x<W|D(IODnrW9!PN9t@+spml;4IkiaSTO|BmI_tx)Qqn{o>52bIRny?Bt^E8
zobcR?!=t=n_Xj~Qvq+4R_;&3By$l7t#Jk}ZQmc}m^!T45>;>KAKrWXf<kBAugIrKQ
z%Apq|qGUoymq0G&IRMm@Gy{cUxQck7dO#IOm(skoYvm6kWy0e?Fbw(_0X;x!%DVyR
zqu<a+KP~Br(8tdb&wAa4K7P>*a{43loL>{dF!a&oe-V^LDMKD*%zS=J9F38T{g8(}
zZAir5t7-ftMz7_*)yXj))#QAPyQt(_u%NZ%x4O0P-70EpytU*F3~aX0ZCU=0tnr90
zY_0J_YW_Ra)W_A3#}u>WX;kjx_71x7cEz4hV1UXR+_{4NGrHTkc>O1pX6VBLtyN$C
zXS&i%_i?UL8D~svr<rrUxr}SH7I(7-Q+7ygk!$UcZ>>Sr=$ksewz2wfN&c$%rOix{
zJ6yAyTb?O#K#h-3u&(v3sbg)z$9&vwG12rjlZNTkY9XTU?yD!D@7V3-?~Uj?SndV|
z_sKZ)o$hNR(g`CAFg|bGT}R)sNFf&rrgLX&1B(>BJBmIGBCrkHfENWNgcto&^yypC
z6~6qnIk}s4`aQZ_hZ#jjJjEFJ_i=OOp7n6cQPyusHu(3eT9FFAf#KHpibgoAy5){)
zWrK9_&UJ5iWh!y6{IH(YMEU+V=yKf~3D-ZZr#|x6dukK4OB=4YXf1+G=08%$8TI|N
zf)+A1RIT~kFj26DWi;E@uhVgV&*|CLc};9)(!xi17c;%jY2SDorjnlsnU#rSS{d1V
z<T}g~3bZ-H_Zhym^8^62eNv`t7h^QT#nB|Dwf!N3;7A+}mSfYHDR?pwPfOGU^)6Sj
z%uxe`2$<m6kq~&y@XsbU%fl=-sKV~(q`$6b^v0-mcJkBKvl5dw%zc^#I1CTvR#;Hz
z#F1wH!{!xBup<h5*O`~pnIb3F!5j{&vux6;KCh15$8!ZTBnkwngYrT6+Oe1|d`<2(
zES}t`KW$ahp+8mk>T6j14~ft%!o4R0{GHL&&Kjf-=8}DgQ3@cWw$1K|8RV2!G$k^|
zO;wMu2@j1csdu3b)exaumE}k?hghtfnb$r+3|!1WkgI7&U^TH(4?T-HqQK^=uBS}b
zyD_XK%``d#?75Ob5C*#%O8GKuJWZ+i3VuP3`;}+?eL1lJOyO9<Xv#lPmlp}x`MBsU
zuJoRg!q9oLB;v1f3ukz?a0F_dZRw{J`%M-7qJoxK{e~_lxf|}5?O?~Y<PK|7cc|_k
zs2&r5K<`OGa7YPSg`CA_5_#Fq`u|n6{)PfWQksEoniy7k2C(q~O$`-hOFb)2^r9qf
zX8LKeIANxKZEs4B+y>hmU%&dkPHR{6@UIfQahiur^%FEc$B1a58Y22>+7a~mCb+1w
z*A>c6X4g8(GvT~unsnFp%dteY2ova^(mI}Vb<66=>Q-jZmN{Iv$e0YuFVS}GJZWx0
z0FuI!<Osivv}R9a=>4PU9Wf~Lv<Z|wQ*#A$h{Mdkt-)6EW*>=F&ym{g<uLAEWH#wJ
z9Qta%n43*YY{>m6g*#psy31~;xK<4^A4eoW4d1o0bMrd9-nXg&3xt6bGxO?qIy;2-
z#j5$sF?5w%+0gp=MoT$RRo61*9zeJtpPYTXYd7M7q52^HLPrCPz60kkbekE658KLk
z(8Lf>nox~w_OJajgc5k}^!kl%{PX%qUHZfAGVS&-1NR1bI|LU9BhVhAwcLSDayxl;
zN;rX*Qu@Me)Um@GBInLvbPYWtdO|diqCPMF3VDmcD0xTgdxKGLm@(T$Y?rsiV*AiF
zj=J&mMxgq41-p2ct-Oo&&wMiik<n}MPoq`gN&BIZ6G-DBDZFj3rvcl$SNC*iNAssR
zv8i^tT|a8A+0FAgUXhx&Q*(iupC$c3u!r=$!CumLke<^M>*6sE)em3O0lkf@BkqdU
z?4&gayVtcQ?(H%y>E`WO+1Q>&e>U5-o9Fj<d+b=*e?i;Ec+*~}{(U@2^S5Za+AjqA
z5P1!IdzpLtKC8R<%cDj6*S835#y;=B>vCtYfAxU%!G4}~FgVD&azw6o=(|&&OFY6E
z*loIH*L%Oy(c1iH68<L;KD!mCU#$rJvLcenL~`}e>fzNRjPhNy_;7H9{fW_BSsn8Z
z2V)p%BlZgpdq-F+h2St83^7uUL@7raDbq>%Aj(gvv1uI$J;oc}#Tz03ZtHWM*rEAt
zo^mO3x4r4^)dL&eCsBK-<9#$YcQ;zt?ERy9GCgw)XHR>Au~GNj<Nrg?UH)r4_XN-F
zqKAKhw>-(ykaX~tcN;Hwl-MzSrzOTcg<dMLde`dRo4nx<cRp$EKybIb9I~2@21j|T
zdwEKhr~J$M@wF3mgu!bkZ$Kh(ly)Loc|Zd0|5nqt?tR|<_}0yyAy?#E_knTT+52A+
znEx{cf384#6P)~CitDhqQ*rIbK|Eo%3@{tJ*O9&bvC8~M1wRN<FlGB2+lV0%G{;&(
zxMA1%<%J2(IkKH0TRJvGG^%pShd>nx2Rr?}8q;z$;r|WQm3vVH60*XZ={8dBDca`0
zQw>bA7r*6$c4bGy8F29%<pjreOrNQk;=ls|H)%J&S%d$Q>JnF38g^6^dst37A37>-
z`oCK}Y*(wvLqPS|gvS9WQ)PZ&po{8-_fh!Pp+-ZUfs}2InVp`+PI20=REjgpcw1-N
z_RZ`a8xij8Bq$n1&@tk~ue81MZu`K5gN>F?P0zUO<y*<)r>C8^N0|NiN`JF!o~|z8
zA+ipGv93R?E7Yw{bS5sF4ZcR9*0jY}rKJ1O4bGB@{4N*YjlM;<N<)pj(!fNSZ*6Jj
zYZF!0m`lz^QDgo#J@OxDxcl1WWpCxQ*?(09TeQ*zU4EpQW~7rCzCVX#@jpsrBT*zv
z#<%d&e9Aev75Ti*U$fiNZq1+5T@z#)sW_=iEgauSMRD<A?wQKfCw(6st>`MhPidL@
zV%eXWbmZnADA`0SCc+S{=oD)|p;(LhJgB!b0qH%8eN=aUU9o?tAgN1j1@NEN^Z$|3
zjIR9!#U!)#CDZdoPdamvvx>F2M;Vv+uPO7tAvSi*QLFD!!p9VRT!9!NxA5@U2-~C8
z_X(w$5Ej+44SP7S#N23Tugz;al29eB8{S$t<Z{kVPcO^7(DM57sC~!P;J```F5>Jg
z0JPy~2#s6)P1+sibVF@(UQLXyHg)4)s+mS>8(_(<L^dj>jF(z4B*|<<aJANKkGxSh
zU*bJ0?JpGOIn3Ct40A}sCeNv`t1<sshQ2A=dz!h*N;ApSZ>u4{qd@D(M9rM;(Gug=
zqH{;&lG`S@O>3Lc*C}<G8D0BhUbQ!aP)g$A=&%2){b}>N1#@7m0j$DtzRGdbQc{To
z0xIp1$FDR(Et!wB9B^xASb_R<ppDf`9F_j<-M0(ur+0hHMzB%LaO8*XB;ws~?!G-p
zE1V&U+<3cmxYVx|(u48cf0@beLr8Yom=rxtwGCPe<UgrBMQ`Fx@l&n6dSXKU8qE90
z27BSw+U*d|Wa$>IALDl=mjMH_jm$ZQ(N;9LT!|Fcr)l>?%0CS^f{n>s7Av{<=g^gb
z&l6TQ=gLa`DxR5KtrV5b%XA=e$(6uPn9Zjqy_L;y6tCm27WF1vh;hTI=o)JnyVjHY
z{1tQE)qg2gD@zJpm*pcKWAQeDC5+lrAA|$W%nj-M9C_6B@}Jf24Liq|Bcj~7mhi|D
zg<vkQa*$KL<!CFig$H<>!!1L@Sok{6<a<~4HoXTs?^^Bo>FP@u5r0w>_7Q@W;)@mY
zw_jXdDn4tvx@TV~)*?5TK5R9fr1F@6IX#}gPm}z4#)4g?`DQj$Ch{{Pr*wG`<i%wS
z@vn{z`+rl9G4Yjw)-O`7&CS;bbvdD6Qo+X*d{n{v72K`hQ>x&b6)frUq5>07(hQ(O
zqXCkx>DGjDzD`goxXvhJ<>h~i60+LDqpQn$16zp}*h>5~IsQ^BQ5>`L)2=n8V-p9N
z$8KMyT;02Umbg|;YMrLOhKp2I%tq^8@HxKrIvv4x?A}oOg^X-&t8?6DOE<eEm8p%^
z6Kj~ixSbpD82FjEfonW%pTt31rVb~<8LFDS?A?IA#C(M-d_~B#1yhyyTGtJQcD;k+
zZJ?yNZ}Jq`0?tFq;SuZsp53s9+@r3!sRStykT#mO&v2oQ2CPl{j;jwT`*meEdl5Mp
zymZ<#d@2?16+g8}SF0Dc^5)LB9$;cOT`c>HjZ#L3R>wGT=RO{C_T2cmwd?Fz|1&fM
zdhLDu$Pd2eqkD(U*|X#PjU~wRt!k~n##s6+$ynR)L%JzP9wD8i-g9J+>NWC;3qE6A
zXUBP*ZmFDev`>l>FsNxtSkC#+>J_cBbLY;UbFR7ZQFzqN>u*yNzMUY<mhpc)#g^q)
z)YczSdO<Pk*YD8f&nPetn4eed`xShj0z>{krkF{wzNFYcR}d<AL&19$81;};dwx`x
zFDV#TV0Lc(x-q0+n*tdw`hQ!&hyqj68VMmevm*?Q@-PBJh|1R*@ilt>-%+f!g3Q&X
zEu+87jq~KF{V=bVi-A-!4RqG{?ftXdBMJ1Rpr`&sy!Xvq9#8rC!WO*icM;3xdJFxT
zdzHKYcIIB>VqKX6GO}bYQ|QeN<Wk%pj(r&8NZsG#YVCeAv%67$_t!EHw8c+1a}0ox
zdT=T;+PWM2FlS!v9*p~H_Y_U;_y7*<Q#;8|!$W_|2BVaeXO{{x%Ss9nHit*yfj(`X
z+_h^i@6S@KLwqZmPUuX((V4Y)?2X@mT49+cBn8KDm3%m2(PnPO(;yg@x-Ib}Xk>@b
zi21=xjl<s<oR!A9+~jA@N?5v&2bi02-?W1VxV%+xb_3#f{BL52fS4~Un@z{>;r(&B
zx~{&m7$oO!6rW8ElFs=V6_Fe{o#U7QZknrb^F0pSb2PNgix!O4llE!hYu;;@x+ny8
z4{d8=tCxKC?p6GzU+MAIp2H5$KB9}8uydZa(?K%0LZOCq1sEp}U|><k!R}}m<N^7q
z3F1g>l3SxK?5H7>8()1Lxt;YY6h!SFL13iPx}P%*xHj5a{pg0ZrM2}8opAbr&7RR-
zhn41e8%k#>{hE`$)@F@X*rc_a*F-7pjA*qaPhD$KU7vP!wceuD{te{{l=~^(mb3Qk
zwAIFYN~UVH%cnob<D!?g31&d|aeCjT8aB7T!LYa)_hC)2giF|~RLwYF^Ux#|<W(vh
zyzyEe(M^Mczopt<)#cCWzB|9(nHL&ezehC`6^Kdq|3cZt1B8jGP50f7X}X;uzd488
zRrq~L+Ed#LUltn2B0HWMF8tq0Bi3K|&(igFGbK}9g+CX`O)OiuWjf<j=1mTxYAk}r
zM2)uTq9z3`a9F2W)&%)(X&LcjQ235eQ-ETxTB<pn!wJ^TtOFigc6^!;NxMB85J_Oe
zOMcc_hXIXVD*%}Ujbf!#C9N&p-k?FPJm-dBu2;&sRL-X8$@h3eUcWc+nohjP8SIi;
zcsPs{e(zZ_VMX)3@$>b<_*jCYbQUj1f(iMz=yRr=^Bj|ITj`rbyda%kN>ERna_xQs
z`DkGgr7(R<OXFu^wuIlSsmg+6ks7fr$S1dju`)`D#^(J@8_-(Qz;9`HKy0ltkE|cl
z^!PYIq^1e`TP=2^ncvLP=4yWvi1n-nMi5J;{DN5cg^f(2*Y{2nZgmiPdUy516MVmN
zF9FK9hR<GLKkv7h%V_!R1yq@PXx{#2V*Kn)=lj`w;OdmX#t1loJVwpyoZ|Yws`X(2
z*aAb*pP*CD@lIcnG(!sChS%Gb$X~>Ux7xIa&vpRhe(DlXHq7?hM1pj!FB)g5jTmiI
zM=6A+84i9PY~Dc|EX9)DPnv3p`_liiWkG~3LT3caDf1-M1y+=zku3;87xttuc7Wix
z4t&uDH+<d0r11p#>5*bdiqUtXn#`E}csGDQ%eTkNewK{}m?%w=E6>I+`TXfaIh4Wo
zjyw{`j+D)4(<}s+8f_|&uSAo!{=RzkIZd$V75hd_!_>KVpLsVXj-3DsO3`qvQD+sp
z#Z5Z(k`n|O&^xDEjL`n(#;kP;N0U5dHHrJ%q&I(}(*H$e)DU;tsHVUj1tekm3JU4V
zVPfS<*wwaD!#wtNrsu3l{Gkbj%y=^oHSgCNDDvD?Vml5`Al`24;0U=tVg5kJAy+^y
zz{-Gwl-HgOb4VIX+g{y9I=}Hny@`nN4pTMtiPl>e-`dNpo}i~qMTwQasVMuUqC{co
zHlTp4HOp*{<zz?QQc<GnLOtO`>rEwP<Ju`HhY<KnNjc1~l$5gk&G5UGUzqYWOxfI_
zOQ<YI#ZHN_Lb;70na)NaCu&A^1Zc+n5^8K%f3FL*=6S`|`meLS8H2Me)Ygq!^J!8C
zOey(E*nhP|DoFkIHPxB}Q3`jwepEFBpx{;wN34cH)S5RpufZF7%knu?kXuk#{$!&~
z^<lgIB~}qnfxR5?hIuQg=Jwa~^z*i0u&Kb@ULWE4C_FL4dq`U&xf+cWnE%$Uwt8En
zz@&7hzOBA}jXb8p96^P-l{e9LN^OYTg$i>wt9J)^_5>(2ZJT6K3d~@qH%d;_nQMD#
z6^h9lPKEiulXt(VJGa#j1lzoAz=Peyc6-|`b}-l-Y;&XS?SLuX%{7Wb-bO0Sm$7!=
zg=%NmaLysDuM(KtNrk!FRG52KiyKrr@!C5&RXVgLQdLs(9%`1V@?~0iI4JTC`vNHf
zi==mXdrgJ-m)twTle!@z_uITZa6Lxvz5K~I)?WLpr(Un4^13eb@4VET{%iHS+%>yW
zwkSvK<Axvi1H%qnk3WKH_2N9G_gm>7XqMhSceklWKd0QThF1P&F87RIi!93ycn4q0
zp=@qGQDY1L(Rz-xtaVK<9-<eK1~7_ky$Iwy;2k#g-@(;GHf{&$k;AAW2fQQn$JiYe
zoqyKR$5M12<tb7&0vUM<sxv~V;EwBSYPh>i4R>U9Y{Qr(YCAjXi_J;-$$@0dh`WNj
zXw7lPN8Zs+z^f;m?(8#yzJOFVW(vBFb6H9O7yXqHg#wP{x3|I1+O7DpCZ$6$37umU
z0$l^Sas*&{Q`$KpL|i9<_`cc6twYMEDEbirSryMUE#)?V(UzRF&1kN%Y1YuiyY&*!
zD`;x;{BP9d3kv>3FZQBh=M)$P|Mj(Y{(F@9y$UX=f)6VY+-{NL9sNK?qxdAJvolE=
z=JnBhWR^$f_04Wz49=-SnatdTanKVC(X|K{S>F9JedH|f?z;n`*}R8vEo`SI`z5`^
z`xU&R-~$SDKBNCZ1x>}3Kf&eLiNA)*C``>QvN`Vo6*p4<1-*u5K$}(kos>x}6_S&y
z{DF%7f1)}vDURun%z)`<^q@|K(b(N>G{z<xBHz<1nsR7bu^9!D^_Y6cpH*y5K~s0*
zU(n@G=}|we@f=XlQrmp1?tX`YZ&M&Gjz6zhQ-k9#=(4K7=G~%Va)e+ino?$Ys$5@K
z@-HgIS0EjYuQlWh{_Bb@E3jGgor)O^EOFLFEjNj&df5&p-AN)oa$DXrrBBJ>rfI#_
z<*3oho!XY&nCmx=o&NgT7U@+oXiEyXd=lkES)O-<P@IJ{FzuR__4*TB9Yu%Lyaiae
z2`8$fG};7|e_OAu)jHlLo-Az2tBN;h6G$4_YNEe8vaZ)C@r-(Mm<QzWc*g0Nc<0V+
zaYDA?H+|)p;%3Ju9bMHr8QJFHmnjcQ46&MHYQ@a<=)lWsF)znXH!-qeV{uSqhdp+4
zL}Rx%==FL1#>f_ok?jJy4~3Cut+TU8{+FwjCMUb{Ernv^^NxwyY^?|{k6)DOX?%}V
z>&0nTU;@{-wdJYv=5na1F)u1>WvPDtv?-ihHJw|O6LuXn(5OHv8d2v|gU_U>){42W
zl<$;r+y+_kn~9ZU#>gF48=UggT^re4o}6Etx==fQl0Ub1Kguqp>G~q=B~G4<w))Fz
zV7z&*A6$E_gM$;r-Q?0=<pKxgsLkxaM~qL1qdL2$qaE|2U=XEc;NP<m|L%BoIJPn-
zrG!#o>7Yl{U}ZLZfUz+KMInb>?=?)q-i0#53`XtRe-ui=yusZ_v>gq_8E>DV9d-ud
z${22-oWrL^7ZxMOJTq)T<9N*Tm8lC$K;k;?U?k0Sz_=;l40{^otn2;nVOEBDIlag4
z{6dXW1H}ljC03@pPLs?=?DjNyBhZ#lJ0h@QJgjQ`?<FV=T3yYDs8aui?tWguvucq6
z#^x)n=d&ao{a0E^((QM5;Q<s`L*s8d5D5|0?FN5!cKvj2iXP5RX%rhDHCTj*WDm7E
zjlg{hVm;%OI3dnUVk(?+%22q8A-Hszd11ucNR`t#DAoVBwsVb*<ErAgy}Pm3-t~I@
zh||QWvuV=UTT!=(DB{sV1EnIOwUAOIa4R8u*Rj|2Bbi;NWwmWXP^I!HNEJvZw9q2)
z2%;1f3CdGQd_hPBfdrz<t3u*Kix4U=^+Uk?{^!oj&OYosDjenR%$=D#cV_Or|9j5)
z-*fIp0T|<VNX%{M!4QSL8x;=@B5%)AXfv3jVaU9t>(oA+hha|TYoA7<W{Yx9<Lws3
zsd_7A>_JACe6Loaqj$>~TI<6|tMmK$?R~--N6oDa^T*_=qm-(aQ>5kaD~AF#9M7Oe
zML+N`Bxqg9Z=ccDh?7%8emgZ@i<i%EB+Ya6vwAePrn(I@_NfW(47h)~h9pNw(M&D%
zVQppDO54`p6SzXdP-{nuSyq*w=lr2sDvXL}XWEhHID&hE;E1$VrmG*B@oZM`JYT{4
zxe}>Js&b2g&jF1PbjRFeb$rr7IjAKmlXd<KG^ItH^{(Uyo-|chnp?w{b~zvQ-?Y9)
zsJwi!RJ1@cG{9nRb*@raQVC`PUBE7@p~l^9`gZTvs=;rRyr876<hM#_A6?bVoh6oa
zaxpv?oKaPW-n)i}BwQ2Vw#{Y1ur|HL?{tmLd48|0KPdTtl0Pc3zM`(-{Ykq|vQ_WL
z+=x%%l@ji~UsCR}l2wxYfSY~1tPHhQ?=~fWQDUXrWbm)-*74hMpYPuF7dremCI3|N
zFC_-`mKfCEN8aDnt!`yQzBgpb6^P3K!5W7%qj3ci5~%`hFOHgGXR?qeSebW5PPF?s
z4;rn)bitcJK&bZaS6U-#f;pgK7^)cj-sOeJ+eE>>j*4iseK-Hh;`wY%!G2(b_9_P(
zQT@h_fP4+Dc&GZSTs5*KdOcgQ7|xA$qVpI-3f7T;m>ujpoyQoIO{e1+5!sqMv=XzE
zXfWPSzKGfA1h|`M%zeHFBVFvL?^e-}sa#A3_6r?5Ld^D$6bH~?WRT_?B}jvjE)G({
zxy~gNvwn$`+~N8p{ShiM6l}jvP_niEG^0#b1ZQeGWKT6@W}$C$KQtH{UFDx}#t3=?
z8#l<625I==*a`Ei%2mT$CF1rCl~ME_d!Y9Q#O?8l77eoUn=$N7Xey+^*oEE$b5Z$a
z+Z!v6_<E1w%2k!!1Qp(EdXG``9y@t{u0a>lwG18m$*a+O?B$$^+Bol4fra-;>j4#e
zj58-7dFQOAkfn{f{Q~FiHyz2I$~Cn;#a(DS_OZ1OO18ER)X<=~FAg1>9QH?yU@g7J
zDiZHFvh;3~rO}(9Mc9ewi@lHIhrlOsUu_zY+yndDr0MSR@n&h7HyCI-Mv4=>F=>ic
zsqeM5{XETq+5zq@&By>$Y2Baq|KuU=BoZ}u7xh_s3|9j*1w+OC)DFut@1bMupimzw
zUSpaC2+sCqCb?6-mgi0fxX;0M%|dPDI3eNDoIKl9dvf#FHl+T42%1f8tKrW{w4wa(
z$bXk3Q{!Mgm^C<e%f-u*vXOloD&$TQcLxDBKY0mq>J><%jp>UK3v8tSv)<7aN~k~9
zn`XRg<26b+&0lxP;^!fO%7T(B6huwC&@7FL0?h313Z&2^Qm9$qO&+C;*<Nzwp{rFU
zZ9ygpRzon{${**IIP%{F9(hDE5<YDwevZbjL1cA=zwJa&IiE%~4*{=VLM|xh$bAVD
zKiyFiIxqD1VBUH)(Lem}M1SC<yB*~k$`#hsJ)tjPQbh|RCWO72CPqxURpfk71*DZ|
z?(1d}g6oI`y1nrEIM@D4*q-AmW1IBG;;|UCtXE-{Y-V=_{E%4p=zlz&UEsu~za9Qw
zS*oyxMb!zS#kAQ?WmD}7ZK(s6he04}OSL?vnD{r0+EFus9rc(K9UIc;^Ct93s9C^`
zI}<c-V@)#VNf=`|`wZCDK7DUWlRPo~XkM0;mjo&Y@Pq}spNScVLRx>>g$`SoUYuqo
zSiV=q`j_J865pctS8}IIb2)3Y*NfueJat~V?3!cLI?CIvr*%KvA)8d}o!1Lje-Hlj
zvO^j32s?ot5G;uMKEM+Zi4v(&j$Yxt?2<OSVFivitibhb^q%I&eTN%KOg8%}b%;n`
z9qhZy>Vh<)u04>FV59i9-PSJ_3RpAkmQP7odd{-kgHdN(h$E*U6+!t#X4W&13zbj*
zBx5yOI<afAMjPoSr>7(;qS-T-C3n38=|mB|nODj3xr{6=hX`0bu|Q$xl&gkr?{@Gn
zu#yhoW(`?u*&>-hEyc?gt}k(iK=hP*xO#9o%bZ89K6{*_PX)NcQ5)Cb+(2_99I}C|
z1GN-RFxh6=*dealiT0jG&$(D%`Ciyki~Wzsa2K(|`w)`eptGHxAh4kJjIf%TslK72
z=&z;_Q5$uUVas96%d&H4MGKbprP1#1YO$}r^xw_RO{c<EM>W4z-WU0{`7uW)Z85W&
z@L>(WP|HB)xZuYGmxjSrb8)wD!5YdV+BGU;%IK^o*H<uSx|k+LsXMC^L{iztBJ0-W
zn8vdM(0E^>*3eb%_0F;}o0u;w-RFIZ%~pMYly7CVP1P|Im#~uE%)~W_OFR)gxb3V7
zC~_RA*lB@=psoFe7T6|Ly}D!ACFL%!VqrIFYBX9DJTY}gRuYTXEPq5ylNExNNQt;s
zgLP`bJDJ|bnC!)wSm7(4dSjT>6NTn+yiwZ+`m%CuOyBB@u&n8ju5TzzOn=<cZ&k#j
zdN;$IJ7&7u*d|NWSuESL+0c`=y@LpT#k{YmGFH#evUR>qBWmzu$YhPiD4Lg`5!Evm
zg1Y&j1kJIh53T6#v2M?fQ|+d|5Qgp`I|@U^ltGKUJ2W83!4j6~HVBetv>Ad1>H*o_
z^p(d}=N2PM(8CUf0`!m<_0{#Y^}+@&qn(B`w@}7er#T_AbI*fstE_#}0-~)4_h}x(
zAKXP_(c*r8U`MsYz=&E<>*y4O0+Np@N#Zl6lym$rKODTh{SnHlr%c`IIEM8>w}Y=A
z!bYMx@dKRC62KUpth9IJGu()XXoEnbHMKM(-QZjR03tACx~m8L)z%%}*H$1fL@^-<
z+)U$Sz<rG%;2^<-%e=utZv+;O<Ac*cA%uW;c7Ol{+JXv$YL#Ief6gh96j-wW)dhkA
zGz!sb#1u3BppN6uc@eM5Ul60L|Is)(Kt(hM<pNF{@p=P*7{`22zs2;fu5rY4_~n9)
zVIMYb;QA3!b`bufurUa1V08D+0TkIL*vJ}eWM`^3T@Gvnn)t;4HTW84{jB%SmkJyF
zza6Hkt!JE6TN;>KDJ`ziKop>~?A^-uVgmDae(Jq`7*ylHFJ3tR$}jJnf6v>Tt@9`+
zILsNn2<Awk<)*WSIR-M%YP7L|Bm^1nYs$MpRZ7ee;ZrG``Lnv(FlHzs5}E!hWc^$M
z#&}Pwbb(C0&+k0kBG}|UTY=8!)gRtWB^sb3RkUOW2Rqtgu@L|9kaJ<8S57SUHiOW?
zZTe7WAmXOR6f~CLNp^y{VFQF{$_ySBE|b}m;^3gEPT<%)K@*M9Xc;AGw25V`578VN
zEkT0NF|&}y8Z8T!uTq0$K?Kkx?U$e{=|h*=!(HCD^!Co{Q`|yQ-(jyX;7l8Y*%$q$
zPP5PYoVLD0!bzPV?E5<YT_xXB(lQXK#2@L!XSmxik^F4k39pr95rMm}P6JW|(ctr_
zV1b$UjFN`DX*ZkF!+e7SpAPOLbjl=pSY_6e1DSNJz)4MiAu_#>>?ktz#rxt#Ck*)l
zl21CtPN#_8#f%|WKsrsjT+2%p(}q&Guurg1hj%|%pJ=k@w~3zBdklSY{DrcP&<fJp
z_4pyonZ%Ksl7OSNQR}G8k(fB4yuL~pfs+6jN^~P{<mNUO6`R8_&BdRz)lf5q;92zo
z)E>i$k*EiLd&GS*6Qk~0PgG8k{Q*FW+iW~vf$L~C16mxj`h6O{X}djgyJB_{EMuW<
zz=yr+#4$9;)Ui~R?wyzoJeYBBod_7rpafWoO;=vpIB{ZS0grX4`d#=-BM+MX-DbaL
z%N55pXpe@3KUBkbM2XQ~w*|(ZshzkPmUC;T?r&=XFH|SD|6CzZZpno8KB7u|RLRGb
zY(-Ulfl7Ffaz^#_7rEsn(o{p7zKt&T|H5fMyrV?8cY%q}PBE)xaIs2+1=O_}6Zzm`
zd&etUQP@|3E?7QhIInYL2$OlN&wSd+FOI~#f}F$pWY|nC6T~d=OUqDN1zd)52#>!)
zgX5g1M2O8WTS%c6XAfcSa0R!khRR8yGBv@_<$WR$>hR{r8pH<%F_pE7d7s-_%zNYI
z;Pr-)%4z1;pQuD3TjE0X;C!x-n>u+ox0L7S2oYzUdmm-7Tgc~@$cxouClB-gh?4|A
zsTO47;K(IIvq>P<9gG(93>_=7INLgtb&<0=U?co%m-`&kB?RUzA^1i0yfz4yISS}i
z57atD6ndMiL!?iRGHun`TRAh9(|D5ur!LZ6I>T_2JoviCuosmDeGm?0!HD#S$)0cM
zufo*lWZq?hUqj~mx;|A0@QcweO${>)kr>d+U!?jG;7hpSn?-!WKJ5Q<szYNL#)G=0
zi4ry@UY0cX1Frq4zJ}wsz|mpdXJ#{*GM3AJP*Z1kkLg*DdUhw3Ur*x6y|6UBw&s>c
z9#@UDBA=!7AGAXlK^K=Qa=lG6UYWhhop0-UrxkA@Sbr;ClOmj!)`+j{wCDOpE~q7{
zyp__tC*1O+!ct_2HJr$=uarG1-Q~TgL?hnpklo;PNvrawH!3C05MjSQ*;XGr3KkK~
z2rdh8mgkAaTC~NAguz=~Se%n+<)w6Wa;3Ox=X+`0$$&>Q7;}r@*lm)HJfVm!B%2-7
z6_}zoRwy6til-M_pN!^)ccrS4I%v;@x*Fq>*)jhT?BUaM8<llG@|W|9APz-+e{-Qe
zt0uU&-hsDU_1dFkTxIT#lo`@+rp^1nULAJ<B&N~10yf^&%5t$Srce3a13Yg+rM*IX
z@bcQak)%F&NAR)lv?uot>(FO(Qb}9lNSS`xv;#H~bM2}zg|jV9;p)ko-6On}%e%_+
zB#EwM(iye2kO|kbj0DbnknWyhgixLqHt{c}_AAGh7zBDo6I~Cvf7!PLTkd!G*6ej%
O4`oyQAIu&Zz3G3FGJ;V6

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_qsample_lr.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_qsample_lr.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..49b6993ef5b16471c676bcf3337f321f5fb257ce
GIT binary patch
literal 54488
zcmb@v34CSQUElfkO;7t$s{Peo>aM-_b}#P5)o#1HY<E|?ZMoYIN%u)5mGq>#@2QsO
zC)~K&xEm-NFv$e49hE;oD8PJzA&?LtEa4N+B)|kRL*|BI!ZMf+$p<n90)^c*-{1e-
zw@TGE5LM~u+_T+t&-UN{XSugG7vblZzh|{{=mXKnALya|C&|Ox`K|n7G!pS5lMyfK
z#TKKJQTrQPiciL)E-f*cw6xS@n!oYI%u;qTYk7&q+)~eE&r*IeZ_mla-le|DzNP-j
zetS+W4lE5$4u<UvO%72%y*RuyJUMJ>nZ=Q%(aBMZXBWqowoGoZcy4igY3t-xi}x&U
zTiQOkeQC$!j-{QGJC}A%?poSCxqGQFSy<XLxo2tb<X$V6U);B}e{#RYdlwHZ9h^M4
zbZGLBJ@+jhUYeMkSh{2K4twrjymRTu<dLPLlSh}1O&+tffyLuXCnir=d~osP(y7T)
z#D~0LZ^Rq@OnmY#?@n)vH~wmL@@{Xdx9!!)<ULPCyzSnOS0mnzQhMRu(y6P7$@@s#
z>FpwISLwd1@yYvn-t855E~pGC50J9Q+e^yc()~O?$n!pLKhOK^`E5KO@DB2Pu=F<a
zAL99tcbMnH_WUr<6W$#>-(k;>c(Lar6Gx_2n8w_t`9-g=G=I6e>X!=3v$=EC>PqFo
z<HzUbtLIjykIgJE9ba6XnfLsu`EupBS1K>hSLVxeN4@#k+11MYa{1`WmFluTbME-`
z;_~$I`|dk&@3c2NbMNdur|z4*=YH?5yJqjdd-}fVdrqAy-FvchZ|Sa+_uf5ySF5FB
zv!!CArRG~U<{ewPvRN0GS4!on`Q!6TEB^AulGo~;>s9H2yYIgLfm8S2H9dX*efLhy
zoV@$K`@DNRkKR5oJw5H6@*a5L#HpE5vw_x&kIzl5Rx0yT<yO*d`tZc^(#reiE61x-
zOB$f6KUJ>GF8fO*zmhw8w7}n)`P>`YP-iBhLFSpMrD<>Kndc^ALDnx#`4pck1u4I@
zQktrg*f+CWu9hxWm!`^7b0t6MDX%UrhH2DKuT1&X`Kd*siRuM!iND#Usq>}coImgJ
z&_{dci?x!!TwE?MULhs-%zUN##BzCder_TY^eryW6|0x4#i>eherYa9l`fNA3DRC^
zc4~F88YJf{OH(UBn(1Dm1GHXZ5M~#rszH8cd9_?E(u1j`O0b26QJpH&nQEyx>n|@A
z>7`y04N}j{SKqNTrB5s~5>t!C^NYcbC+DZ;mdjI%PlY4<Bro>or&p`2CW;O{&HPvA
zt5;6X`<3eRRlU6TvDNBwsXVjnmHc-<GZ7Dl>|3Wx)v01-=3L2JT`UE|jFvw&Q!UPy
ztINe|sa#q1gPay@skk_QzC?$5o_zA@XP=*1T3IalJPtgqPkUy0jsbpl{_=b|$jqqs
zGkjzZF@L#InW|RHLG%jmOochwR(gEOTRJ-v)1-z!inFgGPw-nw5Y!`VwrFkoJr_&<
zC4atJDlG98i^u5de7SImZC0qBo39jRSIaXRfkJh;ppPq5v?Hz5@zTYqMXQ!8@Z&9)
zDuwcLwJ>u|Yo;(&zOuCJmyX?tFnWnfX>m44=pFukUT?w8$>ZmimrBQ}RGMC1K7ZVs
zs!koB^Osjwjz7OzeunkRAeN77e2+iL&VKyq=Z-IWOUJFNZ2QJ$Kkj)eOT`Nn8^PkD
z&sGTfy!p!1^kT^?T4xXF!!-(txPO>GGa7`(kKzI&Haa_q)Fbuic%&Lxh}L6Xv>u<2
zelj{j+7^vk>`ajIN*Cv6N_<$-PMILVy7nd#LCinN@2MbJIXAV!WYSjad;LQ^27{~R
zd2K*8K+!dGC&|{g9Z~-{@fm(Qes+aVl^SzciU8Uoll04rd2uf>7x9u_%1e70FYD#J
z9%6Z~*X#2VUO)c^_&dmNl7IdDT5NXLCnB>^0fkqylksdc+vmMY-~phJC?&mjdt=^~
zSEE;>lc`dAJ##%$%3kFyvoUYn+iK}KZ=1Ir(9z@V@OJVy@9pw-^S9S4czgKU=k4|O
z@wXpfFhD&4g2d#YYt4Jc0D;;c3R9jnP@t6IFy)MQ#5-#58lgYOyyNt3)H~swq}4I+
zJsggAk+OxnyS;mejeGZc_mQ{NyWe|&zuUYAy|?jqyZ4ayFn<AU`V#LE@6lHyyl1EP
zcJDFLcX^L{Pw;nlnEzhywD**~x8S|gdzzd*pNJ$PlY5!@_XUY3pM3ALZ~Sqbc}y+N
zmpLGV*y5=(r)wGZ!tq5m=2?y?ZQnPv-`=2L<PDGD%SWHs@h883@3u$3{E?9#f9m&Y
z-}C5~zk1`p|K_vPuRQwY#$1PyKYKKYmsTn>g+@Am_wbwHSM5~-TCPT}M%JU(c}abX
z7rhv%^u8<N#jZv_7OiioM~bm(ydGIdEF`}_;wcv6Z{o9wt5Hw^R(CN~?;}jtdkHi3
z9AUQJLzt`Q347{W3G>z7dbB=vHAcLz+D|KeT4~o}yfaJM04W3hWOYzESEHm1RpVZ|
zI$RwgWoUi$I^%pb%9}HvjqA<zaWAr}a%1F=`K9_e@hx>q7su<vgj?&Qgxl&FmGf8Y
z8J@P+cM$HV@A6-%?yM)TMk}wpE5cVMt;}a>X;(d2Pu0_2_8N#D<&-v1`Ls&WpPb&@
z-pBfG{|lt|SdUclhpj#Jc(;vbWT8+Ws>kcGdVhVOKIrAIrD!)nFZR^85$>(-Ti9RU
z&RY(!hkCEY)(XV;`v0MNaN!WK!;DhjHI3Tga;hF#pYZzYh3gu%oh!dvz2mCJ|4z!?
z>HkUf$imU}W7m~(+~aJ~@5I2FiManT6CK1Zp9rECYbRz_eIMYnc%>kq$q~G=vJ99l
z&lUEvQ1)`vdWF50PlS<+duzQ97M@!zmsNt(yf$!j#V>jD?jYCUc@`*>=H#DVu9Rv6
z?=F>ARP-IsK2`zQDg@DEoNPVwb3n&Z@qFn@CCKT;MLTzI7N09sN>zJ(X=zG;VMQ>(
zG7FbV^K<8_m1Bjof*$-|t(0b07Ymm-$P0d{!f~lr2_Vz$0u3CsZ{nbzIlr<Dj68Pp
z$-<qjpTdieJ$L5mGf(kF?|A+lXU`HXoO$8dBAqR~@ZKjMJNu633uoU`n0&``@9C&E
zkqUa|<}a4YMS<f=(Ax=>v(@<}dc*JqdAE^^pbLIA$c0a(vIp2*2CFO;D-1_&c7AcO
z$Sz+k0nO%Xg4?}*>B8zfUj=$I<#8e`&t07ICr)tuuX6eanIb@czFI5>1BR89swIEE
zHYIo(Bv%-SYLG1!XBMX_70OV$JhfC(X+XNC{>+u!m<PQxi}Ndh3qu>cphsv@i3z+?
ztOotF!brm6%vO-I1E{D@@(DoWa%r&=j58C(6+TF>2j?-og}JB%1A?uFHNWUjt*GNk
zn@oR5i!iNv%#%N&r>y1ZIfyTo%0XfgG$}|=ug=bbQ3ky#+(3Q&V4xWdv9F-#qWYk5
z01FynG0e?dXBMZIi*D@%{fl&6xKJY@y|m0IQohH~D+Y9SzUud?JAF334!t|SQVj-~
zj8e5&UR^4#u7FOIDoUE8k0rm^uOO>~x>&ukQeyHs!^=fCbN(G_Gr@XY40_c@Ymvmt
z<skcBp_ZCN|2-<$-)z}^evr@}3-YTo;IP7@7~LRNUIBe(hb}J#J<r)+kjE-Bqhc$5
zd2V8lFYF}fSuN9bho$=OB*lM9L0NfOHiuB#>N)>$J%cAtd4@t4Rm-1O$)bV}DwtI8
zAq5{$@S=jJ6}(HqyA_;PFsEQv!8roA?reR~b8US8IpviUyrf`RrL_21=C-u`1tnip
zGUS4WUMpj?H48j8X2J&TM$|v6Dux6eQtW*MC;6=iOy>3sL{rgi{I>tgNBitKAKewr
z#Vp6djD-WyOmrZci01g4=Wm7>|CDzt6;H*6s5=tP$MU35?pCc=K>S!^Z_cE7Z{w$r
zh}aEJ@LTyg@Wm=v-_@v}u1D9><!lvv5Xg`a#POonl51%%R!@&d)-u)DLL9gSJ_bC<
zR%7c4a<U#+)5l|5BiCbVxq5mb`BHQ(UQN}3hMf3mgDDx>67<b3B)#}nAi+ZVe5~?z
zpm|Eryqc@_Eads^_0qHqj9%~aGQjHe)u_P~Jqvbc+n#eg1NWNwJ@p=X)L+kD=UbYo
zd6%kwjz>CjdM#&QlbpU*o$Z}<`rC87fmVy2@PxtfNIg9QhEmUsN9-@}+^(-3RQi_4
z8h8g^GDy$!ZG9YaJ)!qQ^}IKHJ=W3B5sia=S?)Ucx$EzCu#V<8j@EnJJFQe^vr-Gg
z-q`urcSe1%poI}{%W%X4gIX8`-|wksuBu&q!I=8JJ+jtYet@T8Z`|ta^|q?zYR~!>
zjjpx7t?{HUr2HN$zun3|N*-9x*MHCTC^Zw+t=<m9>3Y2Z)<&-ve@J82=j|j8#x0D8
z7QspuwqK8~@3<b}J6;k#W0*uim_*mUe~6XsgFr(41&JxB>o;!kQ@e9&apl}prEqTQ
zBIMn2VY*a+<OcY81&EbGX=(<7Fa(*@OnP>DCLL_;bCruTtKxKc!Eom@r(Ee`FkHFd
zL)Y7Dnk9Ey;xaH6?!;=PJ9~%V+--BWT4ld918D4OcY7;q>10>t8~XZSs0oGu>OP#5
znNqN;m8KUr(t*(PfJJX^-{aq%zy6tj^R{55RnYhkGs~sflbfcVGED9bQJ(%I1YhQt
zpa0y%`rqz9%5!aCY1N455VtFNK&Fd}(;*0~oMy~(K<0`dS-Co~)4-p<m8$+Ff*{=(
z`;;Jf+4GMnH4cj5Ur?deD$sG~09`6%Mx<wfCe^!6`Bl|PXy^d$%X0=9g7gxrWNCHD
z!Gdp9?N2HAgbJt5I(|l$Nvaq-BN#rjTzy*Ldx?&gym$CMkQ=s5d8+LHL)D2-EiO*<
z8O(D~=qWX40I$_cgHd`}kaAON@AF?%N=AGXG51_Y)Myux?Jtq`?fh1z2?Qo`u^8YW
z6YB#MWMVl3C*#q982<nYWASYOoo#@Zk$5T^yOjY<48(WEGL%-Wal(8&BfxYkAKMR@
z$rI1q%Ea^3-WIKm8Axg30}Zri0}n>9Wk&t_X@YOjUXJ*?38U-K3k&h}#C1+pC<Ube
z7Z&37#BN@V`n!P~qNbvsh0cpcYJ0rM`=S3{bgevWb<!QJxO$I<^<u7GgU--jN`o=n
zw&rAK%}Wven~0yOZGEutJUk+?xO7+u7Y`Bt+IA3iL-UmtB0-JJiHyThzKM?qM`jL7
z{)ZLMHPK^OeKX4|ptZ~X^U6tsN~qjK%ppnS8vn6t!h@_3jS5vuze2UjVS?Bzg8hBq
z1AWmLp`J3)p=jc3iCF4Q#RsCXuf<}uKAXH|10AfPQ5wqdYY0IqE`$K4kjWMR#|bZ9
zjnxzDagn+gBmQuml&z7qL@RYISx-J0d3kItRgdv^a4lU=ttaZ~>rwwi6?V*es-6;1
zW6~4#)WeYryDx~eOmmPwT+N8!el=3ft^=IbdqgO%=UF1OB+?qRqjzJTmj)5ZT#NIz
zKY9sd1k@v2Ph5=p#S8D}9oek_5D<wRVZI6~aWzT^&;{+dU_`S$dD-&`|Kk8%kQkvb
z0Mal<xsMyk3<4pbR(YtNTkpRPsB^8a^||)E-1!)!_CUQCfYq~=9@C@zT0g7@kOifT
zMFiC9{SQSh{NwVshqbcjlYIOAUG;XY^@l0Bu%(f;f%cXl)q6bh#v-%oEv?pJJkj#M
zzVI)rk#!IY<pNd*yV~iaoxxV^Bz^L7_rYnQ?jKWAh=;2=)K#;$vnFpHst-H_2)>Z4
z4}tpe@8dD`W^IUX8xoTw9BF~+yX!;i!(N^qZIW9#z!wbd)R@|+(wFzw<BauFtc?*?
z84SK_v9h-NXzQ8T9$p(^RrT9JF~Hk~jR7-AY{X(ip9OIQX|Nox=X`>mT%g8qefaw$
z?tpM*!@O@HZ=^ob$ZH%P6QgH>_{_@cjcAaE9}oQmb6R9=4yqM~y9lA1g(pjkCF2u7
zN)$e_=BWNwfnzg*#El>J7JRk{rxN>DwB{VE{pA$j6kUtequ&*~fPAR6?csdKj-3b+
zBAJ6|v6bPttht9om?9@ydE|H1paAelq;|-*dMD9%>>>4jvGNF!MxEN;)ANqSv5X)p
zJlk~Cl}nczk2hdzd!@>Zc)-=EMgP;h!vCZKeU|?!!D+@PB{2e{(bqf{jrnmi<+$uO
zdy#V?%7G~&HqOm~0tDUVmC_8bvmo9}t^e6D|J^)2m@CwVb1$@tgozVlA@M50&==GT
z5~^(z;#Vu;=(`g&CFb)~)lqI!Dwu`+?^l<XVG<-^7a(sK(5QM&lp=Z;!rrBXbiS`6
z(V-a0nMGiKkc7mYgEJyLAV^BUQwg%G<;sQCQVBLhztO9$skdW)V$_}Ki6&j^>zEB<
z70>@gMWW|}*!=k*Qx1hX{!c13X}}`LGBZVEvm?eSU22cE6>!n$d=Ong=u(6|0olCh
zHRIF-!*K%~<XS`DwVwR*y!kizt$d8252Qh~W`=(`qfKM4#J-kE=*iHBTd}x{#Xvjy
zzM4tJEC(tyx+9hYtx)Zu=oqL*A5dv1dQh?W5U4~SPeM#as4)WSk%`4#*%pJky|o|Y
zMgM?cw`#+iEaVw&p~g>&h_;hPwDLFsl&nA(TWKM7HC~6VfffauAT<tk7N!HbR4&nc
zPC9ybHZIA>)Gn5`NqY+XJXNY9B!W<2*EA$$g$su-7psMO;d1fB#Ia`42W=imWC{4y
z2d&zN&J->ep5XV`v12b694;#Mqw%;FOBL9M<&{-<iQTP)^<mIVmweS|3W5s8(JLG_
zVxn+z;#lFGCBJl}Z~{3ea!=T&2(V!KwKeINkTI7ZEId}Iu7b2umy$;c4cz}op`8+#
zI8p$4)hoj<IaDdw5^Oc-i@=&l`rG*9zpg--xWitBkxrySL?&z@h*pt{@y-AdLP&??
z+uzx>5`+E=MWZ<yrO`@=RJ0#Q(j#CrxoB;~)<##c4y;BCqapD6ig2Zki^EdH1b53N
zJ`x${Xuc6Wb0gut?K`OEzU>;ZiG=&CIJ58T#0`=*kKMf=4>{3pAvnRT72L^Zd_|uz
z#D<97iruR1w9mL*`P+Vv{|D;W9}+Zr+M0!LbE#st`G2H9Jv9QXyPw-Od2g-A30=In
zuQA|5iLj@;Z`aeX#%=p*cxu<!GtymS&sLS-(Z0$5FY581DA?Z&4iO8QBMVR4jcfS!
zJL#|g21U9?w7XnKFGVFbR!Bu$z_6FBh{o{|_Og)@Gd51+_y|+il2%HGh}8oNHs+a*
z6Z;;vgYA^>;uUT+eYavpVSXpEiL|XN|NmBMbIO9gP6E7DzI#-9F0}Z~CpB^6@HLj$
zK*ILRK)7F;Z5qS0V>q(ZY7WON4<O=j2J106dq{>BAP7KqCHG>(uxa4_U5l?J4M9WI
zh`jCf1f)RR-&aqr!zJ{Rjz`5wn^X<tYVnuiYpJ#LT1L{Wdd5!N^cmwH+=#iA`elv6
zBLs@@^Jo9cj+-0tuJPgg`7qgzWdCDgXbCJL9WVNS%ELs~?bYaI*Q(gV4aWZRD%||b
zMVKl=<{YzM$Bm6}FxDI*am<?xlI}t8U+Fxe9@M}O@s>Qt(-6N2{^g@%(Ba`>)m33e
zW8L`46KRgUz%PhlbRqJ2$gqXzcSWBAnf`EeEvd0<j(;5Ptj04#Y8F{2<r&k&)%aQt
zejsx5^KfXdbEGXmp3OzQgectwNE9z7KHukQ^^*~ZB*-H#S<j9{*3;KXujkxWoJfhY
z_UH7aHG(bF4<5Inf!{+uB(I(S6Ds@H3bqpjNn^Pfb!&<OQ%a13_f6#8dhcvI$U6m`
zq8*h%blU$ZHD!(d4^<s<Gu1IES|dTeyj(VQo29AB`R;kjtQ4Kj#A!{CcoW-VV_+CZ
zV0eUF@<xdc+(O5+c7)*;>qHoL_P`qB>_r{<3pX6S33fqD1n^vREmcialU2CpluciS
zgJE*%H2iWAs)uau;Dy6ShghABGFy<=3pp=&P5Y}y`-7G=!}&3=7KqTt?J4cSe`kL^
z1zFu&Pl{i1p-_+cXT+nQg=DBF!CW$D3@K-3G-fw*rfegeyXuP*!~U|d{5bK6sFRmy
z3Sm=yD}P+(9mEXJ^~#3-1{v6z;Kn7-|B?zBpA9i@#9vX#1bFeBYs!Q$-%BE^;ed*2
zNdzMf^^jgr34JV2no1S-s$0sBO9864GxxVuyX)ZZZOjpSa(U%Fn#vs%19fC7Q_%uQ
zWdgp{4t87Y6^Ta@(b|xMSmE?r#Wx11)G^K4bj^*omj_KVldP%Mr0A$iaI(i|K^NI@
z(%*}rMdqVa37e=*3WZjsa3Xzskei-@ab(g-E7(@7>%&rRaei1|q{W$nk0(}6>Q)Vn
zDMUCfzvG}zgnbq&TzQZH_!Bo}7q<4RnX567UM)cmf!9EMuf|_vTU!d>;8N;9U{myo
zpP8Y7jvqBU%<os!AD9Et*I+4xpo~RiOC%?0A0kC`yKJghbOycb>kd&b$UMYhjGEme
zvwYfDVkG4vHHSo82&bf-D0O8SzdlNKBbMJrT8MnP*BDA;fWlsgf@;K(#F<9hB@L*d
zS{%kkO*n~IBY)EYwY`3rml;53Z*yLXjsWai3d<l&0bLQQ*g{tt-Lo^zdS-p>n(gEq
z(!+}j>h2r7rk1lQw#1GOb*##Bl+>y;@H;GQb==fqSf6o!5YAi-zEq-~aI8)*#zAp{
zoc-0flw;t%QNutuC0yEqrI{cv?rK)k9%?5ai*o3r>~qq;i+93zfR=zMkJQs!HN}eB
z?$z@7^75rJ6a%!2<RgVcA8AUO*AB7y%c>fmny&c&4>4OVzOS@gL(gt%F~}@QGmtj?
z&y&^IHVGl5K??GTA+hQ@VJOr!j2VrXGNeaa8K;FDpw_m1HuqNvD1=06<g1B9=FRYg
zZ<OyC<Oa;OZ6u;^<2KSU@S5nxxEJSa=0Jdl%_d?vlk*V}J+_yZV8Hj(<H$~T*>jTT
ziPs`LO9v?KrFg!Vnh9EBEnQd1EG#FT^YMn}5ZXQwriIgA8gAF(3u$!TdOjaTGY1wZ
zc^UsXueTia`qr{Q0c5Gv?Eh@6o)OfZM=A$%)#8)|g2NC6qSVDE%s9vnq;J$4pk7Z{
zN9AC08e$7p7PdU7mK&`)7&9?+`=i1i(31W`&9_K3H($0ZZRT_!_#<SzqD)lC8z+I7
zYL!Z<vhgHnpnI*IE^Q&h3!N}&@VyO1g;3k1P#ZGo(iJH7ri2ORl&vG06RdOmt9yyT
zw2FK4OYpVCzYpRlz1+;3a$BgFaWm_l5K!GrT6Ch<9ZqBVh!*OJ7JS53vA;(#EqbIk
z4b_O0Q_(Qp0}~CIK0*TnvNN;!i9$++{9N(>LX9{L4sjX%-D<f&fa(r7DcfXr{<4%5
zQ1NhjBtga==xy8JyMwqf$RDM?5o1Ga5Vng+h{VFl@jMjU5$lVM!+sfpEz=hpfhp5O
zkK8162zqovrTBdhzZJ0qBr`C?htq~d4~hhW1Cs^xD6yWj6G{slXgUX667o6EcK895
zPJyU#hr0NPdPNNxh)D~?G-Z3jCLG?7(OBH2f)3akii()mN6RO5{*wMWgthC?NW&8(
z*ZZ7?AwhF{5s20Db^?c49hiZN=x8vY@CXr+Fd%*~l<M&XBolTX_X`2ER68UDd&!Mg
zIe9_*qU$NXoD{w>dB)emn}{7db8~3AaISEvFjF}CNMV{kJg^{pmr7_8Unx4~CF3wf
zd52o7BS8!ykqynD1Vr<5tIJ4}Lr&-4sOgYDku>bX{|(hPc+@o#SsjPQclrU1h0Z~t
zQR3_YDI2n)W}yx9u92}(arn}@MuyqvP(;922knKQt-}%}5+SFK29EiU@G?`~X-wS1
zO0W)UY5T_&bak}{&J!>JPD*Lz3*=X{=K5H@nP|<Cv~2}CPWK*tRimT7Lxe6yz{2m+
zDGsHSKm;N@UJ^+6q-jJ@g&Z>CZ}p&4t5HN;ELaI+Ot3Rz7(8HAv~4{JM=?1ksy61|
z13hac>uJOodoF-LBHILmxE^CLGMw33{8{vt-M`GWq)9+I?a9f6sX0Vi(!-1_Wc@{2
z$XypzfDp;c(>I;eX|K1Q>nfdLarK=EqAUJkMkojmIURQ#es3q~?2SZ_L1|&GIE^l^
zH27rIkVZFK!y06Wlvb8!&Y?_*Aqi@8N<MAG1jvAD_(kA3ht@?Jb%el_Vl@J&?sLoN
ziTjW0<(xuuCK&NQrvmRJ&G%F)x4eRC1=@P>+4IIIMV}BQ_{v0&BNwz7f~>=v=e^58
zZ?i&?N4J>uG4_$Z(Fm<o^RzO$mQHWu47ShtA|)!?JmO_*ttSwBkF${AYQH&<FfFTF
zImF-6rxMqiec%X)j%GSe2(3jeF>TJue<onc*b5#zA>0A(8Fay(S!(O3Gk*%Zrp}VD
zH=e_iuSX4=2eo)L^BRlxYV`HUC!#3ga;A_HwUp?K5fAQnB65{evz$2@sXTTGBtZW@
z$S;dUEE*}zo8o-9;9@8hZ%BI$MXc^}uSGl4OxY$Ip3_=fGmL2SNBF%!g3%3Pc5%2z
zUe2ybj0~Xx=EGk5w28Mk=n|%DsriDQ_#%ITIC2~?O++X_G=V&oD4uY0YG5q)kdjT<
z_3wGJQ54%EwZnFP1HI?bghV)Jlih|+G%MY<yZ&=pJkKkz74!nJja@m}>WYtrUgS4b
z{(}Uk&-zcOM%*@l|6_cO|BMnG58SsyP6Vs5GbV>xUszstAGt}Id)q|Je^C`D71$v3
zcYWypMvmdXQt=QN2avO_cYJjuIb!QOWIc6N>lh2IUi}&**VM6cBEFRHjD{3q*^r33
zlRVy1aBw27xZA<^sGIjH=-P=xOpEOz&L5G*sP6nnbzJ)=R9D<$)3~u*2kc2uUcUx=
zjt~J7OG6+AZOg6;>0aQm#em0Rg29||@HcVh=2*D_Z5`k(JJ5S`e7X=um429~i6&4^
zl>OyP%<dHL^EFPsxIzt3n|9=RCFm>5ma9@kf2l}?ARp>6n^9QzfD9J}TqzbZu_coI
z9gRqirk!f1$eG%>jmjp4-_WsR$n1;;vGLPSL^*<emCH0J3ko}27$6-8fao-w!ZZ+!
z{RpTAWVb!X4Sh+QRt^dqSW(2J76H^tb40`*id55A(F27>0R^(0EP`~Pl|jdQfKY-_
zxsZ2t+Mco=k~b)z^VObe9`03c4qC+nIe_a0rxLXaL15mt-Uq6bqm1crN(*#h0HqG}
zJdH1whq@W8qe=kpOVrd*9Zg4i1-z{%{Lhgx?2+;*=nra+F!iFNbny1B^=&vRKiP3q
z_R>y|v<_*t&sxp%l=OeLI=Vhq<=fY{NcjcMTg;7|p?*K{8XN#Ss8LjS6~3O1<>xu;
zfXDsrpbSz@yl2G6&D2R-c0gtEzWlvRP|(|y>)Eo&ymxWgWb$X({CyN~?4v0a?P)A2
z7EZ?w?l^qMXxa|nv<~0)gSTxJ>Wrfs2BU4W8il5FEk$E?wvI(=`)v6MqMNdPJC)Wv
zRwu^&ReekoxeMY5CLPlHHuc|lj?*ggB?ZPnG+v>0vf~bXzhZ{g52=RHSK=i4o*wN)
z?N{s)X_#z?l~_S{;G~nFPxQAT5YaQp<TR4LY6LwY`%zi~?r1V}b6%}=(q2A>#SvfQ
z?4MU)qq3x!gFdZM_yZCv_tTXW#F9>2l<lG^l;rwiJHQ*<IV?70Uo3$>Hm5Nvbba6x
zyV%l0F}8Jo*SK~}riMkocF2jj$7+tOMIA%WDPalCUPz#brN8i<(JYQtQw!;Zj95Kl
zQiexf><p7E+-M>j^qLWXbl!b<u%<oXh}a+I9j5edy6cX>NE-4LBp>spmR@YIw$xCN
zGTf~a#3|9XplpyvoL)bMGQ}qHR>`uUjKmOMGX=sWHTK$88-mTs-$<SDPpL~M6=<CO
z2NY;1LQ3K+RQ+l8M=u>BdX^4~-Dl_F*)tRS{1+%2<chK@!8#RVlYFswVRdTJnVy(7
zzG+YTLRs7bb0jAqSynD?bj~aD_Z9rHg8!=EzbW`rf*^r4=c1E#7(ZO>Ut4r;{Tf$b
zkJ3b{`u8cYJt8#Of1iR6DR@bNt@Em4A0Y^OLwlK-64p5FT0bAt+ik0!Aa<PJ%BuvK
z2rCVq@!#!V-|xjHQQf!|yFVXI^`SEONG6|2WipvmUpAE|Eu(a#TB%Gjm&)~Lc9WaU
zBva!UBlKnxsWfwXz*moE`0e=7`0A(fSA18?s1tL{ST>$bWRv`jWmCiw-uvw+#l#bH
z&4do`S0j_jQfQoMc|Bep?lusAGMnwo4tg(Isx$D+nQ5ntHD%T*W6ckk<$n}&O|$#Q
zT+?j*G1uHJ8~@1xXW2hF=q&p2ckn^)pmzu}%g{X2dNpb9IP6Wl3Xq!|2@NPe<lW)j
zX$F*|-iMqK=j2$JCo@g}0cM-dkaOJ3IAyf?QfS7xEi~iY9-49P@Lo1UP7F4AXUzMs
z_fTmU2Bi;skGz^}7;UOI<Q2`NQ)7US0vUUf`lv?H`zAB;RL-7e4mHS`@}BhGVZGfO
z=ESV6Y40iTotC~Y?EBN+yR7f~!{?VVJ$=SX9B9_moF%;(@4eppEa#y2ocBEAc_^%N
z)_cL)IBX`R697Xm=zoSYr(A8?VAXO<GQl>hurvPMtPs#>agk>7S}x&_q^Ux0gw|BP
znH7YwLZm232K*&6*~HX>6f^$BSO8Sze6%uZdYdw@l-hC1-|59+n<hY8VxCyDhjpM<
z4tqjmdNE;v;0G=0vpB1dc)!K77Duy_GPyaG8LTF+ij2>D2u<gs^<HtWKsR6mW9R}J
zgPAGJg<`grujb%_VbrqFD?Sm70?bQajFi)kKJ#RN+V??3t(CHwnf8}q1FR2tz12Zb
z5E)r;rU<2hbFVi<)<I}22Gh7Wig64Ere@^2#p2`jK2x9mQA%z_V{}mUJ(<6b*ZX}^
z7q(Tdo8IeouitbdF*se=X(l?m>I44ISGRiuO1F|fZ6k=$sWTRBVE+??XB-ptoy>>7
zN5Ni#U<50p3MN{d=3c|_uISg5{|dor|63G&T){UhU?P#(R!~R8X#q|Hmi@E-lgfAp
z!9*e00!wSAgnb*QU$adBmvLW%p$?lhE6`|c`U=#a7sOYVSN!*?!T0G+J+y)1rWqLq
z0#OaCLnBFKK5q^m8mXWP4%fdzKZA_tQ3IC>sBE7rF3CGXRvqw=fmHl&rHC06wD6cN
zpA0Rl#bB)6SwpKE!B8tMW}aA-CQ%Lgrl*(DG+I7iDvNIt3`<QM%`(_wO=CdQ*>m*R
z^ZMcu1s_)MO$rtXCWf3$a;V*eGW;kp5*r`xlKj{7P1h8Bhk|cc@M#4$clO1v>+ur`
zS|nd{y4z;kz}0^B!(f(XV>HHAem{{^FHi+yH4~HJn=~cdzcEtvF9vfaWo#PM+2K?>
zYX)+V_^PH%;RNpz*eW=4I9mByZM^HCX<5M0-R2LSh@|;79P2qErtl7ZWGw5LHY_QU
z>~pjfQTJI##b-U`tjyhZUPxcZZbA5%&}_C%Lfdg6{qfjCMlMez+-8Ate3lpZ&nWn~
z0(As2nGg;WqC1pW=&x*XnT~l6?NWGSGS0<Nx3V$?cTLj*%5#8P2+h3>A~ENc!%b)C
zmE%1cYn`9FkdKc<YkRxL#w<-cs>}%UZ2V}1R9!@G<(CQgKI|+BInp>tWmCae5&Mut
z<NSS@lz62`2-_d+g`HUVVF!qXAB-zIc2ZdMV`oSS9kd87kvBmxq988WT2DPeOl-2e
zv>euZ8C?$7Z&IG@@?*%SP{LmyxQ^ACjyT$t3@*`7-EO?6ab$J>oj}(}<8QF0_I5Dg
zb7)t}p$AH{g_%{%FqUM&&DqpYR^~)G+Dbv#!Wl4CF2Q<Vyi(h~Z{NPc3(w15`}wm^
z9>se5iN~JN6X|w3)OI$^$mJa5So@(TP3wt4Uz-WjXJ`HH_^^&``VYu(CyecEJJ6a`
zmfxhUf+6Re1$*J9g>*13{z6+q(-2x-J*wEbIWs7XqVTbTg>#&*Y-pi18CJL#!B>ft
zn&D}P2P@hs7<8}mTPkNkwB~<IgWBBc7(I7#eC*shtMw>h?O}8ZGcc^ADW$!SDlqYp
zzo_l+UZI;cb6e`5!^l;OH^Z+jUQrRwnFLEud?C}%yY5)+TDpST!Xn~6W`<IlYVo|#
z_Asc2(88ubR#&zggVD-4JQQGj-?4@K&uFKH8hxi#_$dVjGoM!MT?*c<pmWfSA}MK%
z#046Q%aj)`(5>2@?ooMb6D)0EeYFm^uL!kjaLMX1ai958=Qtx)<f|8AlFcC&Nv?rY
z)#Js~1*|&IfsLIx%}0ox8zkfdCm3vBPNs=vK4yHJc18RjrksC<;6^BdlEWULrUOO4
zsa#te%@yK*7uCvq>{o>FiRI-aP<-y8?$2$9n{A!AJ!|lXXp&EhR2~sE%(*mJMKLBs
zZ=jiraeX3r(plV3o(97C5lK2P9)?<3De26K#mL~LZSdm8@rq;3%yN;0CLBC0o{K#f
z?hRXxo{v|aG&D+YiC5lb$XBziMjtqjf+0hn+CHpUtQ~m4j3q1)Wo~>Vpp#u$E?~T>
z#paTZ`5$KxYuU363$)sH6~N;}2QC&mJvxo}#vr<#CO@iK6C8Gy>LQqJvi%=X#$i2n
zPOSf9%IKQJ?<Q?kD|Rmx1g)LHfYkak@z#{)VUB2uYX{$IYTYi<xB8FKZU-pRPHqD1
zb3Dw+{){OU;JFrK3I+fh3CFRnC)c%?0VR^nUNgM;LZ(VS6Pc=}06+pW2__OtHdvl%
zVrc-BVp$aKQm`h8*VE80DX~y~hBqP-7zz>7<*K*|NMOOrUe>UDSV56sB4=82zPMMx
zkcWb)2ael1m|MrT!!eEPfRW(T%=tFna;QKzrB?OoU}xjSc*ihDT#fge$io;`L97Pb
zLN&xq>YAuvoAHp__=gUI6yN#VG=E|mN^J&RZ#V(69o0YMLEq6mXSZ*vV@K);(ov*i
zsPZI%rmG6sh7CHFpg{ImdDzmBr;-Y(LrM+^9(8*T{g#+^qiM#ZrGv$EV3)al0|1h(
zf>TARjUH<1fS7v?XP=AHwVZtCG0>Pm`}$hPsxc&NU5}q9z?~ZkO$Kv8+k|TQKcK<<
z!Hpy8|6|hrp@vT9mm4fRFvN24+Trd2yB#d<7%rM>p4ycf4?4h%hMjP%KT*8?UUPW_
z6B>67dnti}uP8*=g9w|eDM}O=Tn;KBrP2pOUCHwG<l0fYS7Vh^b|8s)FD4=>cFb8V
zbN=N987*qz{y`eEc`yavjttOL3`9gY`Ts&A^D_!;Ob^oUAkFuxYy(@3&UCLv!%mDH
z2`Qs(-*Crqy|ET^CYsd=l1SxH<Ql@1H4)u>Xg6zUgP1l+&uk^?%#%(dzz!l?!-I!0
zbh6etDmYA}orQvj<s&UwC&i_5q=Y4P@SxES5`paB`NBXKfVGK&2O|-h$Gxi90iMAd
zY?r4{S2F9di}R&RzIF!2g}56GO=Y6zg6Ji`pjw7z#c@m6J#HDi`J9uJ!5`N+=}atu
z*&9oF1Krao+A$KU(j19bc+i&$M#?ry;T(r0-%N@d2^NraoEeFF96t~4%Q-Zq|8Nb>
zHtA1jG#JgqIJkaWdP?d1GB)*;jTF8;D4|gG-0_;J{d}$v{)F~Z=N>GaGY3oE8J8Z!
zn$WABYo#~rnw*D2j$=HtainRJ3%fR{MPYi`2Z24(tW~&#hr@FiB$iu+AI5rWsu~tG
z8$}!fSayM|WA`Z2wKgMzf4|vSn^u`-TvD5ZOnQ(xUn;GjUsiET!WKfR#U=26QG4cB
z6)Y?GC4z?D^e-#nR|o=R#g`;bl(#ANiC@Mk&%ahiUSC;UEX`K&I}`NVlPe{@1a{&c
zPL<4JZ9koI%jgg>NhP6Pif%~$S>yOBq$R`{XDzYEc61+K?NvIYjap5j4{cwq6JfO&
zNQSaQcZR*?^dlu{G-?{tWVC9a;aEXhRd^kVv-9u?!!!G7gnZOhq5MSkNwab*Sf#Z}
zF0H1mRnwdFl3}Z?C23UF(`t1?I{<U}UeR0Gp<rA6=46V=Z#Z5G6Q(YQ2{xlat~y^T
zPWw26gzwU<qQuaFQ<zA#Rw4-Vzt!Z~eov@YFjQGt#Fvvv0UHLMHupLJ3>(-u?XhRB
z1hHo>o8&CWy7xhi20hP&FBdPYYhrr^WST{PMKimDH;A_)KMw~n$Q0K^7+T`Zaf~>b
z>7Cs(-O+}PL&toJ9jiIF`5uWg-&m8plyLSuXgG<%%3O1nKuT<lYy*BHnUY~Xr)ZeY
z-!L6o-&B<^qL3n$<%@tP%83X;XrI7rA8$jPEE$BEc`;hodMzc14@iI>z!BPCGH4jF
zE+2NMl%1VqTW^%?n3C!1$@L6>VXHdB{HxJ*EMz#@Gq73{7`XR`&s}v#BC}eH{offa
z_gQZf{@;YX{hM!~w}zb33)XemD`K$q-8X7iC<nOXV&v$%2fM#AET=C{H%ju&gEl7Z
z-|XHKw)PEtyUJsS!mism1{>OHj79l%8v&Wrf6UFGVt<R#NN#J42C<IOP>vf7mvg((
zP`SovSjkOCL#1rp)fM~tD?z2QD!a*OY-p=F8nJ2*_$k5zoGbLtqF_Ift$n-+9U2zF
z4+1>qgr&+|9r`7eBZZkGg>y#=mnK?eo50Kp<LGw992L4MSnGzhIFWF5XQm||I`Q%m
z%;#IrCtq&0*TO|`+1k@q3ZC}-qH|<`Cu0v4TU-#!xY;WOJ($K}j_eFqKFwZtI;(ct
zKtGxiu8I2%arP+?q+Te`%Gf+{n$Z7koqs0sHwM+mDdAw<=ZS@!Xs7Zi4IuwN5d^W-
z*&yceJ6*w(+>GA@fc|f(%zsds9=#(xK!en3_(~=E|A+E>8z_MLuzXwbQ{87}zaivm
z&$aVgIxU~1e|^dExcSS04l2a<Bf8%&?>Rh;#}bJ4Q*R!NABpXV-T|V&S!x2FO>OsP
zr>b?wy=_M7LyaHtR|G;iOe2+bKBUE^0^C81np9^H#ZgTAoWe~#3t$B}axh6|gs&4~
zNWoB}fsJbV#!{U@8Nlj(AMtGE``EOShuy_f4oX=nH$CU${#nxH)8#S1F#vTfUmn%=
zCgq)wWBGDlvp=Z5G~b&q<4#S^%og&neiQyGPE_;NKB$V4#rmN`&eNj-sq5h4OK#7I
zH|kqjte(g3hx@*7L3Y>pKAg2K4Eo<ojUiDksPFXQ6$Qts`XYKbz(&JG81k+}y?;JV
z>%&O$sBirl^uN$qQ{Gd&<Y!O<0u=PlA;ZHm=3h|;w`%lbs6Mzp%C;X8;*0bv<Nq6%
zA|)lBQ~s~>Jm%Dn>Vs@=8AT!snnIfu<58Sgf?eX>5v#VM#5F-2$KE*2=rPe|5h|&)
z2%aym6PT42FS|ciKokE|Yx~a#Cbl@nqVaZ{Y+e6z$`_UCYh5@-nu%ac@L~d*53*_;
zPsd0w+nCcz`vui_iC8cyg~?T%m9XQ%TyY>VKZ~pJQ8WzN=q~eril)IAsV8^j6!naU
zF)`fItuYncU|XnB)5g>&x<dmUILD@#aolG4<&-!Bfm+iHu;~j^IsWhIbr=}RIYH5}
zms<okj$vo~pVx;pM=D6nsIj>0cN3Q}-|gn0Y6#QprOH{py$@wx$&|!n9ztp)I|s*T
zOld#E<B+S05n?7$LYYUpDOeB2u_UC9QJ+>@(kaQCz?rsgw!00x>)=-jyVI{wrLTMw
z0mhp9pv)uJLo2Q-hGje>J^~#;M^QXyXm-e@1r&K8$#7*14@$mt(E7a^ThA)rgdI6t
z3d*sCl}2R3N`+sWKJ$iN=;#q6H=^#izF$N55Lq|!Mw&FK(K$#Q|7)rv*^~dH1VN_3
z*9wMoNQ&1ar{zp&)xTmKsvyVIg|-LADUn1$<wDM{F-y?!MrmJKyfYG2{e;@zuQypB
zO2n4hLz=2Sp3sxxc$FWkK4KdMRT~*;w;gT&YkJw63L>2AzVz9;-W}}N*j`b-V@&tw
zoy|@$>usGn(mnHU)yVBxJ7O0x%|>@OttvSVH2lU+x#5I-4oG$ln$Cph)IUBd0<75}
z(c35r+LZp6^r3RE^~5MPGv45+mG!=(;b04y(t1iMh)ZOd7Q32oRC_}WSli7EI!wl1
z^_RJ105$HAeJME-KJm(4>2jOuo$Y98l9@SE*UXp-_51aKqBCT-De+Kkd-u?{+kA+h
z(fE<xt7^W7-^%|-pph0|!{|ntC4n6SB*pIsvlp;rBSAY_x%f6{m$xjLVcW_ZYQ}K(
zusxF*%h(4uXKYf*1u_c&Y`odcfzR@{aynF~$Qd=;gH?#^$OS&<2z%wOtkEZd@`XOA
z;T%1|xJa^2^kt<hs~Ur%o5>ZM8us-ZYA2y7RVp;4LVZr!`40Wl9Gg*<DVu_ST~LA{
zZ~v=ehS0TOjvWMdYjRA5z`%{k41Pgr|3tw*QeYaILyF1l$QfgWW*lN(naZJ^-8eA9
zrLS89_o)261lZgaIeYY{I<;)JT9`I<a|uml?j@}Su@(=4bm5ZFh*+rt$cxbaXl+aP
z$_N!GI(#cgaG<mC34SZH1aO@(kOoOfmB%cO9(08<$AaI4fZ}ICk&byW6o+Jyk57yF
zFi+z@8-pe_(>3@UAPJ}iaVJ7OQ8|g|0G<VmS8*HQ2kzl*{w@W(74Y%03+RH@&9P=g
z<esM?JT#5e-=zw04{#aeLP)72_5B*mzf_x@M?}zrYs`ih^97CG2!&9UfB_&6!8m)w
z%n$NEwsBXr)!cBrYM0sg9VWszb7Dx>wBaVDj*hpbqH^z2w=)WMPWb;XH5w|Ue?;6(
zc#j}N@iNRhe}f+BxC|T(d^Fv+xPl+i2MJ%60|!Kf8(8hPtJrZK>x57v_C1qCjGBZV
zmmz_%!#OA5h4O_X%7d?D8nrEpE{xmcivu|YCnG#haYH#9tDGcqf483XQrDvzfr<2#
zc2|?gdn)uMg<o_W&c6E@JVN<P=6qH@rp()s^IVR7HLSm*)6xPtBF9d;lElMVRI{n+
z`D*1T(h4rzS!#J^GLQ7u=*(345FAZz777ioXb0)weAW5a_qgn)?8#e{`KbzJDy@bZ
zq$?U#U)PzH({XRB+aP793TE&j%J)NOUe@;_NnYg&mBTP!^#S;rlDjHCBP@wCJ&9~u
zZMk#IPPDkz{M(o>2->9lwVVvnsAV&1C*AX@#`7uvAE;Q)oo{yBnU1-geny@CKKcHr
zf<Xn+EH`cV2JkZFaMQoH4Vb>bktGi?KSe>jQ_1?g<r)TV?;<OYDec~kIhS0VUhzJv
zwVR`QF0+H<OY~!HTlewRD&9dRij}7wNE4CD9|8O!oOD9{Dtj(t3aTKH=J-JZ9Vz`M
zfKhlIF%l7Cjb}KmYdLJzMI7Rx$4hdyrHp7wczEOXk7BXjTbAEucy2u3Q_t``V9!~e
z9|to?(lStomU_$h_`a$;Dj>|@uJr0wmTU3#3|8y;&r9Tl7j1NV@<{P|u_f=5s%mIY
z4hbroN>EoIASH2@S?yhiP}82N_ro*PTl@WQYqrow3j=K}&@%2HX=kB5RnXC%f$3_Y
ztzQtsd2awqdfnqOq_@`xM<QyGltw>g^FA2%QT2ya=03I6C^4jdG)wfzxV1T=|E@8j
z%{dtk=VXYsbXv+`X8W7~RTxQ;wj<&^)w#Tk85?QV9}VksUkLR_z`;hXK3)YNrN?+y
zefjYov8OFoUo)7lk6_W#JtJE!XA7-#=g7Bqx<0;1o$bwUALHA%cFuv#?iSLwZ<0Pv
z`nI-otLsW`iMSk_>BiV@-8i-r+nhvghGpOJqkSOO$v5hD;8Rp^ZU;6<vm3_oZ6<@E
z-x5QaoEF4koji}|g%2xul;E_#Pl@{#Oc2z1oV$RzsU@zF5CUq_KObAqh<_)+-}@te
zrtuN}pV9i?ySbllG*`>Go?kvu8*Id!dH~p2ZNw#+Po`FC=o^7s8Srl@3eKT~!y>kh
z=!>6I+7`t=uh<^N<Qc&~s+bK=3+=Q6+9=Uwir{b*wX-gV31)&ZvBPmAbU06DFiLjV
z!=T=(W7_XktWQB+!IzaLEDHavP^gO`qrwc$F=R0(3eL5{NfnaqGQb2=#ezC{ieNHj
zAAV{w^gR`_M=f!(zrE4o<r87yyC%mvGR&icFkS9NZ0m({bJZNu$?=ZN%O_en06C0e
zX@A%vol`M65au?^I~X*qPcTaRE@Ic1?1~sd@(n#8<Oz0%Fv5oWhg57E{;=dTF2!Ha
zD_lXylGM2pPhh_^1j`e9r<naEzAYwa;^W~zIUL`peB0Pu&exoPU*o62_<Vwdiq<p!
zQ$+cIZvjnYq#OmKLe1XjsD0!hkw&9*%qXdhur$ebpoC(xcw(t%O&ZdJ&E7=42THcb
zAe+_k@1wTljd3gq<;A79EXae_7f|iddXK@fgu$|e|1Q-1b&FU^+NJfB{{c!$VW<~Y
zBsJIi2>Zm6BpveswTeVLPU^tgAne2>j8C~*#-XyL0rj+B37>oI+49WmT}Do%hO|uW
zOwky)Bj&#=CWehK3Kk6v+f~|7wv6Xz(Q-!^Tyj=PpI5=JC@>B}Ob!02>V=#;qsNTs
zwjJ`Knn97=SZ+=Bjx>%38?2w!4Ltb$ajdx$LBG9HLTK1-XzYvg8AglF=LX6%5PHsY
ziCheT-8llqP^L6AAWi}*{{K){Kc(QW6m(KAW={3%TB}FMOt1pw%pTWXP%Jr~ObHx@
zo!D_&rH(NNi*FOQdzZI8LmX#%X}0mRg@_miwgg4dZt8N6DYwantYKM7bJQFcviR~W
z(JaV8%HP9PTc{U;458LtA3!8JNUfo@k#dr!xO=La&=QZ`NgPVyY;SE8M-FlPkH4d8
zJU^C`QpEKe#kqvJ4HDN6_kEyOxbI6>CCtRK$iwkO(|sRkfWORw8^aD{2$UvoF8oI2
z&Zj=ig(^v|42znJxQzeN+msuz{7u_{gvp&@4Q{fG{C!$$`xP8ea8SV^g4)3I?((jJ
zhY*W9C63wP`;Y5LYVjs+GWP!WC@^+qpF`FSJO67^gP|u&9MP!Az-WJw`->#?-mc;X
zE_}sy&?47n@)ize$BX!?(x(&{K#9^H80-g>u-P}fsf^AK33{rfB`A_9);`x?y5qL-
zj^;-?6JD)tC#O$`OmMfrBl^?{4Txb428fU+Y*_CcGLiqL^{#K)mcSc@{euDh)kIQM
zx=v{^uZQqgn27CX#qWT5z1c|{`iT=c)A-S7X~pvb_^kA%K2VY5$Dzj{!ytTOQ|$`>
z4!ssB%Vu7rF!4FLKVe;irp37A&C(O(ydE)YBP=xKO^jUuGtHbJ4kH)pTN_3<a|G`Z
zYWd}A>a{3F<1zo~PsoFX6oA@1Sd>RBANGsyZ235N8~!cINwx8s`Q?8-iWEof&B6~K
ztRr7CIZm~j)zv|$3=EpH#WWY$aVJ1I)t<UGTFvX?plUD2{kPV$koBPT5JsbDjn>x2
z_=djv5KKHd>q_#}&yGnLT0P+ZGtveZhVb$qclJ{}57)Oqyzqr$A+8DC9(^rZAFsyO
z$6kZ*5g9~(BMV#V!=wv69|xS#hx!;~hXm1FU3o1BkQ!dw8cM+azyZ3-Z#hOGGO0{`
z3lhL>T*im`3YJ_rfUFOjldfTUIuM%J|0(f7uip&p|GYke?blY5TP6IPoK$0`@|;3I
zFI61sU1hypc#W9tcrM{{)U^of8Gf5TEEVIU3QB+FGyxDc^73ve6;>mB6bnTBvkHTt
zcw?H%d3nq+dKAd3iBztp#(Dlgd$Z#;dZpH;y~Z)d8%;HsM}iM1;A-pg#|Ux*8HU-&
zqu8DVT~vSbk=o<k<g|R=V7pP_io6i2afq~G`b{f#(wqar8*#^-o2Er9*{4?g7kCLg
zd3&{CnyvSLfUJVvFU%$tiA4*ybcsssfxmM<I)+DIC92<sM|Kl-SwKd!&={uh!hk;{
z(?BU5{i5Oq@b97l2ViZq{uvqj8lyEH`P=uhd2ztavG=wHw(wL5y)>tY{L^}Ol2Hj#
z=dMhnQD8LGSs0qaJKA8yQFZN@g4Tg*D38Fy#5Q-Mn{jKVRnbm-tTjT_f)@YOl;_0H
z^V7=xl4Ac{vA<DlMS=0OjdlHDJ$|zSV`3Zq(c)z5D0bbH^7h1b2QzHOkEzbv37qEZ
zw6wJa_kuiX9Yra_DhUtgazASa{$}5&6%^n){A`dmx}|v5edasWx;cc%sS7<SmeFkm
zQ{Y+V=m_q3#qZY(bI=7C;Oq8j%J6atRNWlkr_t0z2cw%V`#e)`-0wtE_I0}T2df(c
zEeW(6BpDU>$=o^|m;QGwHgro~NOh@<JeB9;$U`9w<gZ2Uk#f{Ahl|nNS2Ib;S<xV*
zo|B>Co$J6iv<YTJJT`uWk!dmwBl|q#Q11=lm9zq2&B9c2(5WBQBXiRgLZCi_KLfxY
zsWJ_~$mhy6D{+9DTw%wVEBJ8?mMw8<`h%)JR7LZ@ngxC|ob_S_VDO_Y6{Ow?b6y5J
zqm9b5c1Gvqw$~}3ajji54vF%z<qrW}=GTVlUk{2wQO?{x;(3UL#F|`a_xj&X*ax)9
z09!7(b03E%Bfu7fJc!?-V2EN?MeC!$7PAD_^{*!P9f7D2O9<El<j8SJAFfjD3x5RW
z&=62&1n4mabO6)DTA7RJ>H`Kn2Ixs|2t9t9I2S1!^!S&i=P(dzJ^Yg3g+Y%V|4TqD
zY8mV(qCN2&B4`X_8~{7)ZG$0xLJQaqflo7y=I+jp1OH(y&b{O!>1-VZtsTF`ZG~@D
zEyDm?JKj!#&GxzN%D>MR57ENb7SF2tkEyF~S4SRE%yy?Cxo_8_?e0evdtAW>njP-k
z!2T1;{cYtwshBpee_FwQrCDHya-LT{shxR-^hC;OXxdeeF<H9u%%0lC9!%IhTm_zG
zA=}!6?9n%MLsw(>;XC*hkxQHDcXs$tHt!iTB7hn{g1~y#Z$(pK3m&ctTwhV<Y(f3b
zCf0HxdUsDX26~4HjnDnjhTes6U!HJYkAmJQU)CMR4J&9v{<_dRHYwOb&Il~@#Z9`I
zMXHiziKRNR21c=oocJerdMmWTKdC|gL#ntzKc&Ymz$i5073RRdo6Oo>-RG7=tltuA
z@b6Kx!W4WvhFkM16yc+4`+l{(f%|y-wr`6G_z&n^tuJ~|k6m9RQ2#N#^=;qSTbqy_
zNZqvwHd+5jFh8ULmFTQnmZ|lOqM*@Rwy_Lv`uc4;;!o1yE@)<UI+ysC$_`$ak>zL`
zq>`Puv~@%<)y6iTxei&d!_*>?chgwq903Q~UO4WoSJOGy3Ku~Wo0U#^3cisD929b9
zkqLV-G)<G@G=aR!rz85qb`VNe-Q94&3ja)evofq=J5^4f#k8nvWcJe!B&7Plwvp}E
zQ0BUZ66UD9*5(w#{zN|NMCy5$!1V`M<+#C|(qIak=sI(_Ey;>Wg!Q@xb}#SMk)iv3
z0P28z0KUw?(z&n6z~18Vjl!x{GaX7M60Pdjp!gpV0cH#Ko~YxNh@p0(_kL(gjJv`V
zj*!Y$%M&fgWr3(eNJ%15K8$5k+-#rjKp8@20o*DlF@3XyC@d5*&{=$f7_b<~NU5gK
z)>^{)sYfR0n4%*btLiPYRbBp0nmI2U#E=CgZ)rhlD&=yl@iwL6)blxc*jAbDe3`LJ
z)O*HR_x)E5d7J=_*s`}=>pQ8-f1JZPBK|qD@L6R@td3fLTEVIUQ?vM4)o(G!uj%pq
z<OaK>lGS0_*`dVeIPG6jJ5vgnS}7aYRy5o*XC5_7CDH$DwQ5v?0V%D(HmLM0$HxDn
zrE!q-449sD;iN4mD*}aL#R)PE$Yvxyb_-~8tb6x;o!*wzgIfe|oZ=-j0|bqqqeQe(
z4G{elBoIDKjBioKfDp`1x~#fUF6Nxu8FSZ+kl3Pd>A(e3FC}$L+}h~c*xD9W(C7lr
z8=$_KcZB?{XXo&46#hp7hi_y2Wu~=yLhacvU=IrajcbZ^*#a?ymLsgcE%hywO+OSV
zpCvW#WmtX7tR}q&Wgh#>eR{Lv8j>HTkmL1W{*VWYYqcPqTsQ)>@O`x%o44Why;TeN
zl^IO1GOv8E^DgtjNI82cQthvA!AkJ+jh=F^pPr>3I|y$<uJii&);!#S;pz}>{l*bM
zvO7)7c&lmLa+@Al?ES`t0Mi6(#9+Ge>+mEHIqLdjI4XZx9jk8Py5DWmnB{8YxHrVt
z!MA`Hf!7(7thOU$-A0)m;!UjS-p8Hwt+cV-8>Zxr`uI9-)7N*`cU-3h_!GZC*+PAs
zvg6e~^>J^6IonBWr#EV`z4e{-aW|jd7~ImG^__gn7QRJB3Eu%fWc+IMQvh||v=1CP
z1~VR%!rOW+$+5j_ZFh$=!0bt`R;iroa0jT~?Bf00@I}kpXgNpA&(O>L_1&cJsqZ0u
zJLws{u__wlVD->d-4oe(I_#e44VN2#YIH+yqTWtc>FxSc+t{DRcsBdBi}&|>yX|(v
z|C7E=@TGlV{d?(+)^EWsWB#G~UN~MO-X7N8-p?o(e^>NqUw4lXX6*I$UrW=ceQWz|
zM)vWp1N8%ZD>pTI2fsh@+1Nu|#p<C=m2h5X-RS-Mj^1X!Gv<E+-m}|r`l}tGzwC&3
zJQiO&xOQmmFtdCIJw8-FjL}!PR@NrGL-h%CU*Y=I4|#{#E4lh1C>Ua-91c?sH&SNf
zj6qnQS`)Lnmvn+Jyn`>?*71c>VAT3HhnCd2)4p`q+WrmS6RSMX@jbR08a-@|{t>;I
zkvWRHm)-RVu1D8<kNv&gyZQ~h_c-tEVT6B{x1Hc^<Mx$z8ZCH)nLn!ES#fbsB9@A+
z-LZD(CSUj?oo`y$U%yjc%Ggau>PPshyLn5RxBSQIvGwCsc){x@u7e_RqiZbOd7K2~
z|3=dV<~`oMxWG&c@ef^K-ggGS)Bd09nEwj}|3ks=Dd0Ea&k_cGb_Wy2C+>O)KMCLo
z^3rx|m;}SodP8;oL_w3Pd}Aw3zz}>^EeYTTJ?B=Jrnr#A6z)em^voMgIoZRU8WIe4
z^8J?;`>+Dp@B8xvLFRcONU#cTB->E6chWciX>~A;fpg1MWNmxHXJ_Gkp;rKpTR}~3
z-~oVJ$X>Gre^PB33u(k*QIhXAjQY82N0jl+UTlCWGP^Dxo$|P;V5Z0)2<YAV=$C2X
z=D|is-8GwLb1*+Uk5bsIUn&*mR_8t0!rn;VwiCgQPK2US1ra0exq@w-x$P5!X``W$
zmEz3ooGV_u89#Pv)(LwA*^ku*nnUw+c?FMeRS1l(@vx!LusYgWRCcRrG|N;=q90|{
z2Al1@2SX)gapij04GCbv!Z$~o<=Qxv71ok54U8svNU!{-n(qE~e%V|3R`nOPL|V9#
z5qwL{G$ZIm@Er$b@ykRuB1KXgdkbf(C!I4&q0e1zRh^x%b-%^OG*nTzp=})JwM$fd
zka>6Mid{<X3$OC~m6qa0aer>wp__&#8f(Q^7{V1Dm-YJ<Yhj<pJsr{~8mCmOQB&Rb
ze^O~;-kUpQ|J@qRpH`aTwI+)e%i0%D&lf)Fo5Jx^inXZ6OUl(HF8(hPn>gyQ)rXaE
zL4l46=c&Oq9^Tj>d$jv1%CHTDWZ5l*!4+KRHiXyaTp3QNlDek15)8YN^Ru(76}b$T
zBWC}b=-F*|gF9AwaS~-NyX_8cr_}yq^gGBXLw$3uA`GuKY2#PbO~bYAU`f|Pm6hMs
zcYj$`NiqcyoGmr`BNKl+UxMCR`v(PC?$dOu-MoBYIX4sZG}b@c(5WZ(HcM%1I@a<p
zt0TXnz}Q&gG|Nb%DITeW7X-+2t}Sp|*0!Rnl>Mh72L22E?MuOw5_>rOH}KVgr07Ox
zY9nKtT=bZQvFH&<1tj&c7z`?v!2K&sEtw>@{N&2uCP#a^dBR#Ij#U5Fu3I^bl%2`W
z7<wuxeDf<O7R|r8>(&sx5Gf-6yQ-^nsMN3Jl0(s5f1S$ig-LeGh!mAkYeUupIi+b&
z(U-XEw$y5$-WZb+yt!1^pq$y-yB*A#G{d6(WAyG<i2#AI&T-y0+8p+Fr%Ba3;h%yU
zLB(VtjiO|D>3uzBbu+H6q_3ixDb-3*-K;bRLbpaS-JQOGMH|=vPKg@vD*h*tZbF3^
zHJk{aQHHT+y}947nfs%Gi;+rET<EG)BnQk>4vH}10asC&gcUNZi!Qh&Ame2}BUO33
z<&wKx+y!*d$YO<{El@kaMYD2N6e=Ebc$?EA1H&l#ItRjgYI~aAK%Ix8_VNt%MI2zP
zYQgR%s1=?sxvQ#IR|?OVsP38P3YE~&oex=!d!8&BVNQ$}B*Ypqc#pMSwpk9DiM%(+
zomCbH$-T|xnJW_`{zbiKnyGVk4?nEfcPQ<7#m*{tL4h5$(or+xxR$?U|B?z{R^TbO
zq(EXjpF@S?q0s@LpONEFtK=1e$($RFBARr5O$likOz7#OIw@@{T3}o8Ba}D=aH)u7
zcK(#LsbnaF;#Rl`<&rwIUyJ4A+8Bwi;U$qRv(djA039z~oerBjbmi6mCNpP&2WQ0Q
zmbch4mGNq99rYKdCl8T<_5O7({E=B2UanGAND*#3)bd4B9t0Eh6`rubmukz1+UvQl
z(4Mz*5e29;`NmL@1>qvN9DbJW<=t*VxozAzIxr<~LzUn-n!wNC;T<Yz!a{?V2FtE$
zFk31k@=kW~6gnM@r=+}>{p=>Ay+GP~HQEBHjt@VC(Fwec4V*>?S|_+3<8I2HK6~bj
z_3QL$zs@Vb*&gGMTvmUw`5CFxr_b<jB1WM#a-C6C@-)dP-0+8Vp@D`@k<@pV;$gdC
z8joitdQP9=b;?pb=M$S0C2+H2lgo0}mpe#T@$A{tXPvv<GiN96a7+86>cT4oLAr={
z(;2Km->0{KM(JY4IP~CK^!P&xmX!8gihZ{NgZ=+lF=Jx=s$&0E!EY=0V+D^W_@aW0
zy7D>29#wFkf?=hND%hz&T8sY26i6A!k11xzLtZiQMjg&!sD(h3xz6@AXZ}?^wzkg4
z)PU`uSILmdUF8Qz%0vJg$p#>6{O$XT%!4r`qyVM?t~GxXPuJ<(C?2tU@O7Qe^yLOp
zcdP8ct<>FsyPi}IPFOsX%JpRiGl^VZ>@myR?b^$~ncCH8IsdiPeQohm%@TtEA$da`
zZ{<cl%7rw$hN8ZDJVR$Yepr+CQ#rw(hF9&D+QnpCE;35+DN701*4*rZ7wM#VW0qM`
z)_<959V}bnaS49)H@r|1Kb=vzL1bsf#QDYp5PtKd#AZ%|u?E!cwp|?ubZi&skncPE
zo(3ZoJ6a{BE;eZx+zE#EP<wfclL+o3b~EQx1w`ZcFEKeBlJ8XsTYlfh_v2`>tG%L#
zBIleC=Qj;B&J<`(z#n?=3I>&Fj(^RQupAP8vzoRQ(K=T3rVnYOYn{tnSXcq(8C#nW
z>O~*psFL6GCVQr3)Yai4B})U%<5JWD0|Yk%%llq9DjaKeb)Y$;u^bt3?3crjdK4~A
zd`oBzjOan7ae4+nr&Zc1tb#RE?$V(v_Z#GAOzLfhA8H?Yu#5#;wRHX9hc|midmHq|
z-nXH)OhaCE9M{Uc;Rl=acJr2~l|kCw>WR!q*2A_w<=Sd}h4sMtzoA}^dOyk6auJqY
zg=phQPsK;I*5#L*yDO1J@6q(yA$x$)_qEhev;9$}nhg4NEikNYl~7Qi73XUmique9
zr9!paI{9{G8a(+WwI)}2#${&@9Z(g+@eP;XrN<^=?*EO73-1qNGn*PVPHVY6NhR~L
zw=?&L;;tvQ<-RP)oj|=hF_QalN+ULq`|pzP=2?=7p4?vwsV0`r-84CIBK0Qs#56WR
zW1)t2kcdVB1L>|K-fgXEb=i*I+n$#?C??fYt?3Njgm$Fe;lX3a4`uA9BAH<0MAEV0
z#XsW|!Z?jy&2cj6G>S|nN=X^ZdP8<<W%2cohTde_rQ%<{H{0tCdjsC!tMX@$iBlpn
zDN*o95Q5$pR*XT>{B9g3J$Ggz#w{1imqKQQTm|$y?af8*FgDxdCgQG>P9LRHPt>tl
zL(AC(+aQ6qTT)U!W38e3rKQS-WRn_#t&>mw-y%f>5smBZQf-HpP{&}j-EFD$%3NE1
zr`FJS5rm?dV4&4whmHBoB5m%qKSm!NNHajD6YJxe7@b%MV~mc$!Di@uq<eIl=T--9
z*EsQ?P}@JO6?=o=hV%MrjzD!ES7SKc>3UdY52>sn$!GNVV|ujH)B@|^ck_kU=%D}O
zWY$94`3>FLbcjxOoT%boWcbyDRsJ18Jz41wCrYv&rVDvb0!}AaqnC%8If!s24YK^W
z$%E|L`L!%&20}*R6vGWpn6#q@uqNpsW&`yebU>lTI76dm9)_s^W^7Payg?pIBn^@@
z^PR{OQ>OEs=j2ZF+v}x2gK+?QMbnhZ;=<Y)x25nwf}q$mKe<_8mh1XZaOg8?DFXr)
zbL`RI(<r`ItK)vf9v}!3XW#SWdr(R2JYBel)BvR6Qy*kV+=|f%InIrpSP!Wc!;1fb
z8r70>5<>A<CkEO$Ex)0Pf2rc?gS)~{vwd4;Mv%OW@c0rgN4bNvZHEO}RN+h}SrMy3
zYv(pjI=)A1x4?T7SdsKW-w)i9lXuzsORzYA$Wb@vK(avX$#Kbsr6cL0*`2nbxNm$(
zUm~QrL#^yi60}cDocI!}zTVp=PDCN!#EAnECn8*QwjuyoEiPMH8L2$C#EFQd5G6P^
zcvGO*cy<CsEC)@XIKp2E6s4q_;_nvzLO`#hcIGZPLA*FF%11O1f@qY!<jI_phy+pI
z*FyU_M9z@uUJoM3bBe7G=ss?g!$xT>j|lQ<QU^_d_&TJzdLnV5{`NOTkb_|gIo<$b
zl|evftA)eX!Vn_J8=JS_4Zmgm4C2C3go{7X=u>sXo_~(b!&{&P2fY!#N@BEq)hy$@
z6<f%LuyI>es@Di75jH-6fKVdB@le?KyY{rj8<nt;+NtW+>b7;tn7DBaapM-gglEKy
z2o@1H?n2zSoie-Y2p<tIMivSZHr97|<CH}7xW0#8Aw<0H#Erj3*?lIO+*;jV-|B7U
z7}!N@m$%Jg2kN`(TitAX+o4}~@r+Q9uaUU%B^1-SMk5qA9z;_L3sDm{?lN)X?zO@O
z@k_Mwc&GS<-h^UBTHZ~|5-Yw$FAvoVe8b+lgmwked%Qg+?)z)<4)dlw=-)nDxBFj<
zFnTY&9z`>2uZ`5TD&ncvq^X^iDAIqmdWU<)fMir9D&J~wZ68P1{@0=pAqKs$NbP-A
z`v;q~w=LXhqRh{#v}>VNzFEq><Cj7$u>IbFS2GBPn{U+ILN~PDV?Ap<Gl~Zp#eD{G
z+eQ)9!~NbN6V)A9J81KEfDt)_*l^H0%y>-PRw()B9b+t^<PqK?!JreFA?9LjNL+SI
zza}!f(?n*6*Csa1S*)_7qrJ$21eHg;qpXNK>UYqaW6X~nL>;#y-A?m=q%)Tz)d{=&
z_wk6|qT-@o8&jymvH143^Rsd@dbEk5P)$}O5h{UXa9j!sqP$`PHpeXB3Ts4X{wckt
z3;rVhj{K|+Syj(8wc0iuqb;6hn}S=z09KH~e@?;k3O-F!jt?VG{{9OJ{wweC->=w<
z3Jml9`Vuz(11kJM1ygF^yaHo(G*Re|aiE}KYQy}s8w47f@Zox7m4}+}&0zqneo_6E
zddrQNQ*|*w*TPYxCHI%YQA|tjp4$ML%|{6Tt9BKs|6zSZQNgqVAz1#bf;k0EA(MZO
z$BE<LfOscJEG%P3cdx2X>kGa>z@>JpODk8}+_KzG^wSdfI9`Z>#0zTYf3~2<I4?h=
znVVAc>k2x>I}>-dVHumqbNuuANE78ODOOgntU$sTe?`Fs1^4PLlC=1~9zUnp4=Sxo
z$nw3){j7rTS0I^-uN~m5gTJh5RXwgMu$6aFu}ccH;P5OsQ>-qn_?MM(MZr(0QcW?_
z&i{yFbp^JLKChS|ykbH9h&J4hP@xaoElE0J(Xm@{ASmfa1`jqO-P&a&M}+Ozzm{xd
z7s~aGyIjA%Y(!F!6taz6^q}ER)a8AL2}Kih%SF1S9e9BWFPC_WgW(1=q%IF=V<mo0
zpRN6Rrj3!9+my-_A<)J;G@RAgbyqbx&F$(~GB%Q*q{gcn$q`<ViGT%WQaC71jXL(&
z$eX@O%#hjfLk;`sB0-y1%_Tf=F&JWw92FVbtLx>iM!XDWX(B;Ifa1Hwu2jo&L0aA$
z^7_32BSCXUg7$E}4+o*sq0{rQ=~v68CJkEqWUkQod3>reUnxL^<3xIPw&a7t7G_<A
zDZGDHR%gzcH<qS+yP%?_mFl@uCLV6Jbaq*OsO4?F(SQUh!p~_2cQRqCCG+7Yuh(br
zsAAP`#A-*4YCEPrI8mp2HpIC&y|_GczH;mY{~SX-EG`jeca`=UCr*U?_ghTindbdk
z(A$*`(oHybf?uPrUm;RYuOsBL>()>eA&#KysxDT`3WtFSmV)w5hm^OYvLV%qD1aKT
z5wBg1Yt~FH57kY?XK3MGMS<&GNFGeR)1LiDz!h>LyRO4I7L^|ICf&7UkJFjj1Ripn
zFPz2@7IdvvVX82Py%?9%oh!|pU*RmSaz(lLknT8c@+O1cMm-yP{|2iv$ja|GuD6#e
z-0LM$NH>oOgqX<>R@3ci^0L@VSL_n6hN`cc@qa{Z4Ov^wm#9(y$CUd81^1{&b}-h}
z#N!G+rl1Aj#3dMk%V8ih--8W5e9gwcZJawWmu@vEtkd1gxydv*J)>D{{HVi1MV{cd
zqMeM7cNlU;E8#)Ti=)_^aH31-iDAfdDP^H@+eno^EpXKj!Mcm_cTjrPDD$G$>z~m5
zO{m?buEX{L7ews%aOA>n{3;>*WVSCtI1{7Yi*|h<q`BR{vJK)~guVG*BlknJldNYE
zsj7UAzv<69QKr1k=(efMg~w=BE5~Ey@D~mbZ-dIg<ABu9piZdC->fN_>0&=9xC`6s
z@yaXY<oPanQHhn+x8Bw>J;C<?@6Xqv<+#ufKR8zv>m!!8SpR475`|dRa=Zadz;93<
zKQzr-tgNKv{Bv^fCD6!euH@k}h8Orh&Mf%<RKZUW)Z~%^2LU=E*p7umwe5$@o4r<&
z5;4l>pqZb;W9edn>tYVioSQ1+PPj6`A%o*3sYA#o6jxhKc=Mj!rNIL93QM}UTP>Kp
zvV;&-M~!>ivD5y)P!Im4f}d0H^9p`J0qY}D>w9({8%!#rl~Q@IP#b)qT)Mo19|1gC
zxg_+VjcM_(RA*YjFDm#Y1wwxPUshmyMO(xFRXzSFu^@+R5XY*E^LRKGd&r?kQ^a7C
z98kiqDR^9gmaPBl3Vu<6p@aWBF|PD0T?&0%{-kn#Q^9{y@H+}xP|5(*dD8CGwt71e
zsUWypZv=#O!)@%9k+|+sk&*;_+FQ9;?3Dpj9<ZZ(MSg^HxAvRrf^l(gHSWaOxc%nK
zIF4^!^e6Cp)^9@I4uxrrole_#^IMTt&g~VZ&EJr`VYG8Nk_}a9lz{O%1dl)-l5j)d
zVzJoTR?MiA^)cvR<uuf&8_$mGF@}Pr<8lm3HJUect5Ij*VSJEusZqlf;B8Vm&W1dU
zyq9C|)=+^~yD<ChFUr$xYSdwl*9#A$4|=}Uf_NBtub&paZ}S$sfv`nd-lTqp{fOc)
z3e#bqP@vZH7l0%KQb(qzgZ5Q}CeZmA-Ukn3HB$K!C5GV~*ttQf`I!26Y}R~DQfr7>
zQnNl@9f9+)4Q1R0=VN<SCVrS{8|Pz3UEGQ-a6S<9RQ`)Sjd{Z%=VPckUfsIBt-ixJ
zA0u!+M)`i~zzvBkpgg_46VAsD%IvCdXKZ(px?6k?6sN~1xeJx-g7p;DW>d8OzbL!M
zxFp-Ed+Xc0E$};b6WfgfwPO3~@KD^0qdZLt&%>q;EzZXhH1Bq(>8(by!#RPEFp8Us
z9iPEzK&W!P(|90-wLKf?>1gHoPI{UVZ16gUy<Lo$ctuO}_dtCQU$d{ik9UhV(u-oV
z-p}~2ALLC^dFJg>R2C=0^#ESMptpzKP29eBC_EqZ_8M;ibz~cvL%eCCKEa#z@t*w~
zc?<R7=Od;t{>6^olR9yrMfd+XTh;MwiG*Sysk<5nz^Ur}ZK&!ttik!5_5N0LHq>v6
zf4mvZy$K0Bap)VMRR0dtXbXKYJ%DKq&)<ha{Yi}~=iLS^L=mT`HW&c7ojPF~`Dq58
zi`4P|j)dqYDt}Vg1^fFuP@x&9P?MV*JxUuvnv_}Y!|kR?yC9R$v3ChFeFsD0RP}b@
zK1UisA~WBL{2Ynxg=BT=d2dC6>h^Y+^Qg<^MC2L)D<U*u<fqFOp_cn<E`?y~uM)YT
zzf9yBo{Bd@xfXi`3%DQF3>a0>rMuDuduKXySMuE?=O@&_fCAyZ4wDeLZlgdoMz_}x
z`6txv(+C!*Wf~&sSUeU(hw87;2Dvq`B^7FR#G)f#^Hq9=f*n6iLR6GOOBU(~Q7>;2
zPf0_$<-FvipuLucNT6K`YH6UDy79bC0@~M*fIjU=#}@VZYew{ms43yg8U)Q>EsMrn
zLA?vdp$*;WP}%OVkUZ(-vAis4LKLXHU~@ClWe_v6no9T;cY9Ll*xWHN!OD3x8&;hH
zIjrI_zHo7Vs$h%usFX9D1JA!Z?ONc}lBOyyum1nq&h0g>B97xtcAM;Wlg%crX;W+6
zVrdewsR>p@#Xfk!LK}qksq`hu=4x)tZWKecH27K(stBbBUJwdC>4O&(t*9Vsy}ii!
z4}8=&p)`KJzd5&LvuQ(d!=9X-bDJ~soB7V~&hv;AYh1DWObxE_9(~l<;!cdm9AYO@
z0D=XXFfwvB#ZkglD)2fME4wAnPV~2#M1LEU($BPz*4zC|V!YYs)FH)vt?<xwJlA_7
zAimr7Vn~D?NvM@w-(xX@HN1(+Q&K1DOn?jc6u2S`G~t=`4wPbdr$6ynYgL>mH1Vj0
zyD94}6{*wWev0pU67ED9cv)9T<vByB1BEX@-H8qg1*ZZlG`VNMzeqjWPE52M_p@XU
z7=jWK>l`T^W{M<`ILp-?GbPqM?F0oPmAVwF%JD^1HLf8Pf!0PuPXk#=!9dyt_Hpt$
z#C3NezlZ2K7qu(>g%Y*0<BHO+Z+92q^g6uSbb{P!`eh2!!tnCLao%pDQJ)L58mIlM
z`q)!Yu%s^u-Jf2kT>b0j>guW&Vw=?3>_tAJZ~MEw?WZyNiH_Bn4<mq4^|*nK*%lh(
zl4x5ke8GTDF+y?L+eaq`1+zU}T$n|n=@UDc1)Q1GU*m*Lbh6%y#w4B_0FC>YT4QMP
zyt}BK?C99ki|z`$>*WD@nGZ#qs&bK-e$u-ZGn&N3o(LZNe#saV#a>tXs7OLkN58QS
zc8#!J!!aI`Pw=BguprbH%_3R^Jk1)8`2ETX8Nt0msVfU7$>NBEhKZ=<FJyV!%wjLY
z!YWFur?H>t#tGDG9rTGxZBB30BRC=1G9VF{nEZG}9b@<z%WuC|+jLRh&HDplt2mb9
z+S;<3$VuzkAcbGe?o(C9>iGeCSL-CA22X}eMl=`b4{sH@IVkBJuS9hVK?&2umON6?
zo^F=H9B&LmCwQSW3{?sSEsDjEK#+$ebklVZB+2Mn2x<=^(|D6t9-kYTObI~;Jq$(Y
zp#al!3-b$OOS2=PgtJT<d!}>6b@Ot{rXV+mr&1M#e&fkqLn@ryGxGAq6+fKV5iKz>
zA{><08`Q)elNgkf88)Gux9kejH(TpVj}#P4-0H0Z$)Trrk(7?id5BKP+9;w?4%eg{
z$dd`7O#%rut#Dq7ko15*)LDFqQ!bgn@_nIi^akzK2n0I0T4FX)byc$q6Yd>?fQJNQ
zF8d7@T2ruaih!903NZvcy#WL$*AO-^h*p`#Nyt`^nF1aL?Z5(gcEaf&;ksG=={VVX
z2nn5kSGHg~E@mhJD$*qg7l^P(j2ZyM48jQqO)AN@-K2Q0v||h|*ysvj;|SO9l&ucT
ze;PJAfDP2{?lM48S_2y;gN@Sg^3kopMkI-!3{k@n;u2jEw2n@XE%AK@8`F~%xI-#!
zCtv)e5`im6W7VPpvw%57w*{?XCP=eZpS*GP@83_3K6lJ(I}^;=bqmZH(W6GSG0ZWL
z`Icet7q&6TxGz+up%P?ANOGtx*6j(#bf#D$lYhl_KevG~?n~X>Kqem!9v%_x@|;GX
z^QXqcqtq$^9f_j7^2@8lkd6UA;v8CT>~B@biG6E9D9R1L(HPnoh+27|2BEDD7uku{
zh6xBk%1j=XG0;rHAnI(5OyFeJ0GJ?JW=V)PJA<=Jx&)#n{y#Ei++K`maWCE@g2j6Q
z=#unH(3J~=BOQbgPP-q}dT&w>>-=E5HDP4!YzSaI`d!?uKWgtMHk`B(g#Du9KWlSc
zn-$%9Eu#IahA&dyg*BBqiO{S#Fhs$*HfqciqBXg1bYp>;`&t{jyKU@YhMd<pa6NjC
z(5aYh!6&Sx3^M6hV`P#>I*m;EOg>Zb%uvW6h32B@>6FyCQZ(d>%tdp))XEB#Hbber
z>FIXR_8r0gH5&ao)^qu+p--7l%<BlPFk3rJ&_J^Afd%D0F*x#MT0mqD$HWQMg}L!^
zbrAqVh;HT^<&mXHr7bW_^C|eucWHP`{7Li?J!bhDJ<63Rq(@>MvoLBp*OKN_q<;X=
zGPWA;uVLh9Is;2$UcFb-H?*q=@J*&CK{FQX213|d-hUhkGIg9@mh64RUd*g!JqsAE
zpafV-&{dsUnwXd!C%zn_exCu*EC)^gZmVC@<-SgrM)YV%_!|Z8k~TWDIcL;L8-yz?
zAB0Y;PV?ZqvD~|Pcf~+?8#b)&4b`NPU1}GV@OG&W-Qtm3wNHKOf5ho9!<|^TXTe15
zO;M{AiHBNLelYgL@?k~oPSxc9m#<+kSUqkyuXCgblX@&~5$F`fMr_`gBDscEutkzI
zh^-T{`@0pnTMb;s=nxD18U)8V@5mCi$ZDa0Se#!lbAvTvq#7#cgv!(eMVEUmV(N&!
zhsWT{9jMCIvw3efX7e7|3SKXnIXGl~c%qhuY?;z79~>=@mHQS4%TxXQ8zSkZ55}V+
zevb8*rzmR<nimK84*9tKZ;6789o%`_(5$tZv*BQ7F_-l-fomH~Hbe}LjRUww{qBtK
z^Oi0#FgJwY9*w*@2$nhu4o@RcPJ;;aR&&6v+zpftPuw3ixtwKuV3=+vv=ynFPzN@K
zVNH4QwD$l%>S0(b_%tt=;{LKBgqq8wx|4O6F@6cp_m#S8MU8F9)T=!meN3;c1wa?>
zz<^%=Ek=6^_{eF=4e*^X>_h)=?w{u=GaosZFmg60Zi$=wmFi|xHE-X7jSfzl#noKF
ztXvL&nmWV1qIW&!pQQT-Il_UCO%2S?dq>9ib&~w&gU-d5><~&&5``;HHe@_|W{*GL
zyl@AUEFtoibzG@5kJIy{nDv4T^^ec&hbV7$dejNGoZ1dYSz<uc78Yi!j!NT1u8la8
zGoAK;7pA2&9$2bPbB6q%g5J9N*io=(yoz0kQq7K%lColMg=7eu8=o8zYvp91jrn<H
z&dzsjdXfQ;RxlRCz+1P8H*(_BLl-bzfhu}wmiuG7;?%G~Z@M&os`T4Q9rVj*0qWKz
z-QgwZ!v{u|Y71dfBp<^-(F}qEYi}A_)kM!$8@L|TYp*tU>x!P#J!3PR0gDx|Pse>`
ziUDM<fQ`FH*Zm_h>(ky5ozZb3>eiF?j_yGndQ~UMKgX=0OcHH!09$f-3bN6VY&0%{
z+~dBo5(W8;5ZP_HTpQzDNGwHx_PwH@t*MW>KxJ}1VsUQH#4R@w{I~4NZac5P(uDu-
Sw_U@f`zb5%?I;cHI{F{(4c!v}

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/__pycache__/ddpm_truncted_diffusion_finetune_x0_qsample.cpython-310.pyc b/ldm/models/diffusion/__pycache__/ddpm_truncted_diffusion_finetune_x0_qsample.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..83849db5f19a01ac81ac14a95af43a7df2d39a6f
GIT binary patch
literal 88062
zcmb@v34CPNUFTc-rqZrftM}?|FD<)w?~dcO-HsDmF?J^*6`Qi8TT)A1DyeUkdQnPX
z(l*_GGlopUyvIN?sFPu!S!N!OodgI>ATtbWGVq3a%;&=ZGr`OPJOcww;(p)X|J+-P
zBs*c|Ngo}Zd(OG%p7sC#pZ|6idwX*cey-g5c)9PpW0617L+4L|hllyAJr#>Yyht(P
zMZH)hT8!HN*h0J*kGix(F==V3Vw(T)N@gKj%vxTel3VC0_AKO!d3#P)dKdbNeHKqu
z`WFU@0~SwL1{Vg4gBH(Jh8Bj4!xqn0Mi#adw^%$^8C}>~+`2GU99!5{+_td2xP4(q
zamT{W;?9L##a#=#i@U8{Pi4<Sp;)kZzOr{=Uvb~U{^EXn?ycOkaC7nIg<FcZEZkbW
z)zbPZ2NuSQ;|sSHZ(F#%c)O+bR}L;5Dju@<K;`hlk>Zhsqs624JXkrlaJ+cj;zN}?
z7Va$Gxo}tUE_)uX+`Vv5@t%cyi}x<vSG;fG{^I=$?<l^*@?w<-iVxU0y>lTRi$o&}
z4;CM?oDo{^aPi@w)FZ`5tkk2$NA3S(#mDUbiG|0DC#}qu$`cFkD!$9&qm?HY-d%h*
z@vYvNx6RxBiFoln-VSf)%aP(!-o4&#Z_mro;?rKi+e^xXcb~W4yNQ(ddN+Hwkn)Uo
zt9O9^-{9Tv-R9j+&a>V@?+_`^JrVH^dq-Z5ct^_V`BRtT#rKhN)H_DXvGV(Pem~F0
zy*qfm!=69D^PS#ZJl_?(@r|V1?cGDlJ>~cF#s|IFbCL0POfE$uk=*&YidR^eyHH>D
z%Z0_6-06CKsdoR7BeQe$)5}wbrxzEFRF<dbJb!YoT07#EtBZ5Bx$5j8Z*FF0xi+_0
zJ+ySOzUWV%J~CBVoH}yPJxA}J@@A&*p1JGzJu`RR>)mn3%)NI`-7|I9@#E#YkCpE(
z-*N2jJE!hw*HmiNR0?ZqePv_b;iZe4HF0sNT%DXdGPkhgFP<xV?bf+gmG8Ur&U^1W
ze(xPqQ}^C;_vG}kJMX#2yUX)v?R`^IQ;gqz_Z>YxU2av-{_v66$>myYZnD}=dW$wZ
zwz#nL{<+$b`s9KJsP0czYcq@fLfNn74jn4+e|j$Wx;FR3c(j>$YI0%9n|$io@mMqK
zmnVIS&z74hzr0kQtW%(Gda+tBU#KrkRwrl6J}K#?NxwcfSs|LJpY;~_pIw+dQ!dT=
zbDrLyiZi8^vcFhbtX3|LXPSML#o1E*LcKIuE6pv;HdEybbG3S{nfA&vlgpKQGdWjV
zm|SY6nbQTnLVaa=I#ZdfH}lhr%hh^`Z%!`Mnp>Iu`ec>Q*UP0De{rEi^JeB|`E2T`
zx%%S^lj=v6?oL)pXDZE|C*~$+7ps$%CxYQS!H501spUFLW<1^;c>gy%z37!IPtr$!
zW^%gR9D0&zug}#lo}BY*_2=qpPVXbj^~G{^nv(u|p5nD3>+e*#K3S?wpDugLm2z{K
zLGmZ3>!rDBeX&$8S8I!YGpBW0E>-5vlxbqmi4#vg{oLfjQiXwxH8UqzyzhSI@h6&@
z_dWOYbEhZ$B?@I`<|;Ji_z6l3Eltk(Wv|4Ddg`mcxYQh`{ZvrzOdL-(cLrr9J+IVS
zEAHz{_2!O_+$EY+TdbBg&8#hzr_WT%?abXBZ`LL&^`-LCQrU0k4^7RToAhly!ZD^3
z!!|&kH@8r#&n=W|_3~1U<oGj>J;J~A6OT+hwOFe)Gt-*CX{Nb{m`~DVy<TlbFEYld
zASc^Sk576F(?E;X56CEz<?k{6YR?liKtkIhtI<YuE#k$zxEH+|*VBA-Ep|1s7QbqF
zUSey+i|-&_Ppl=qWFzW<)Lx0TLofA8d?vmv;-yC-S7QLw2%su6;p^kCTcA&WnLmF1
z{GpZH;Ujf=d3dQh`}z%jW-a3n79So@v0uvm!g9UbOn8&^$!2_dX}OuKE-x%yB!2EV
ztA63UKUXg^-^sb^(sI3-vVWy(5&s?3bN$?r(~Aq`Bh|%vd1`U-%n?;{WY%9?UOMvJ
za`h>8Icu<bM9cGtmeiw9K6|9%EgV_U$_7}5Ao7UkEioVda&@}SqH3j=X6CBp`f{~g
zx^T2~wr1-S;L^+*Kt0ZmJxVibq7X9CK7RAjgHivTJiZ<$nAT8)KWQRK{%l6RiAX)d
z=tf_QuErXXwdmEz^YPXAYGO6XQ_N;4!3?DuskL|`(MXO&8u6<!o?~l?MruB}mSWb@
zS0h)W?J|w%Ox!&)CrJUZe3fOiu(ViZ=hplqeAQQrZMMcU{;kBDiP{_>=?+EHez|sf
za*2ICGgoIk0s!gXsdw&Da5uqt+P_cnhX{)CrRm9Lbh;Tm-He_uW_T)T7Hoc8@Bqb{
zL(_mAWwx7h59zf&f}9QAKy)NJ5KTn=lO%PGoJcH<o{ij#MCy@wt=yPav=`gKI1<;I
zBt8@K5*ky=#98I6VZzjFiKyO6Fyd)aGG5lpy_Q^!*CT6;_tmI!HKM_K?wPQ$dR-Ii
z#)_>8Ftw37M&fv)nW&U!fY1Ki?CE+lUI+f9rWWh<#f2}EZ$HKOe7W9COadZR{v23G
z;pBy4Vp>2wH7S@<A5S;qOMv<m9{|TUlQsp#{@U`?Qf0ATD@}Uy%Qdb2h1%)G^UZW^
z@?4p+?2*b06=(Gf8mONxrf15NhDYMckIgNtpN8IZbG7Bk3KL$U=)*MA7Vk~bAx1yP
z^7TJN+L!<0^dDXMiNE>d2fL;rLr#i6n}(N(uuvm_jaWU}h+SeSpNZA_-pwRjihhI?
zhCT60Y+K~{n3lB`b%TZMC9lSrm;_V86mE^M+>>hwma@xDd+DnQ7IgTm(n(jkt3<ky
zZlq>JC1qBA@3G0s^fDN+3<{sD7QhIFTHPyv+6$B3<WhZ(;c?h;THsq4aBgK%c<ykn
zz|SN8Z0&xF>4|Zv7Y>{<EH!?A;m#8Uhp}8)l>lNF4$M>*p#ttNJW>EbS1N^bla=K%
z-}A~|p}ttCpDvRG^si2r^*niQan5UfL$!Ew=jOcfp{a|9L=YS<oNyTb{=%{Hp}XA2
zUf0TBx#bjf8QinBi40^NEJGDel{=a@;db_MUbzN70m0ZB9#8oXlhlk?%36JM<@3!r
z4P)1z^;w)6G3bzldfklEGgmq~<!hQjxulm=-o+T#X4ca5;)d15CZ8@Dg+jq6sKEAj
zCas;lofWn(nu+>nNZ+uq)Kgnn&k*TYSkX&Sw}4n=5sN3aJXkcbM(k>IECPJA7&uFd
zBH^9{EnRG@@Kbb>161UZU@<V5B9)k!%@%`gHk&`cJX7`|Zpz`f*EEx-*=?2$MK)KN
z^b6CA6|FQ@-CU)z%qF*`RkO*S<dxHlHO-&B<CSN9)+9?+gUD3YHd}&5SVG#VOvh5x
zI@r*r_Oh(iFG8Md^3Gd*bUbd8Z4)BObBR8%7N9Gk`GU_kbM0!~(g+eX%YYvHrdCDk
zEz9T$>rhsIILx3{HE;d0+4D7a>HD>i498~jfT%Hk{rdVvBt+6e5>GMPLa+7crI`OH
zBoC<KYSh2C9w#l~v8cS5VACZb3J?TKD|RCC!YBwK!86dR5x*3_1Qa_H^N-Y1ooPCg
zJ{LK=tqvk-M7}5Dr9Mf?j|j13Ugn!?AeXDr6RMjfTL<M_if)ex0=oDpFwX<V&9Y$Y
zS>^KO_t+xOfav3m<e3O-Bwme*7+cLYvW*NWKHtkVvV5<nk@=p;Cldx`GhX~m)W7J(
zS98@&BkRT45TuPo*bX@_zuLoB^7UjRV_)fC?QP`Oa*b5Ir_pl>=qtq6=xOBciLCaV
zeU<H#CkA9~Wci;u`~G^Wk-ijV1NBaxVu#w2evGpa>vytH;lRLww9q1BwL6ws>pFd~
z9}0`$Ms`C5P!}9U3!MI@%eC4;t7#U%?6w6o-&EP}uIzldaGvvw0aMOUuGaEduvg{c
zAt7R~u(Zf=Yq9FKnp(mJX?-;K)Zv_c>m*yWe8C8~gM~RO>X&b=74A5Cw4(#s#HyqQ
zTBc&QHYcoFVG@x0p<_o69=q${@uM$TeUx%dpV48caICu;foKjR0WJVcAM5za!SEA{
ze2o{aK7C@{D|fXUcF@3!S_VPNfa-G_-#C6W=<JhjC|rGoIXYl<S7-UB6DpDOa1RRm
zbPU;N?Fl<`fnl}xC}a(@$<&v1LMtpm9Qvw}dG=hZbdoJVV73NL->Xy>&zHTRIrrNU
zwZJ5|k39F=AhhSby+rIp6vE4JHmxbbnj3$DdDc2{D7CP><bY!E-C#1IJT;m37s_;<
zPP7-WYenrKYo|R@0$^}d!eQ9>T}y$QU6?#nE31c`9Mne^GF8o0y*W;A8fp!mI^|Z=
z+w`QMu?<>#xbTQg8w4T@g~Ie?t$ffnm(AJ4zQTF>P%ccoU3E6`?ICsp8|+cnD3+*N
z7|!wARo0&VU~qj7Mt=2vjab>>!1^f+Yu&IV132O8;q1pEbo927+AE{elT}W1g=q(9
z*tOGjme|TJ8^xMix_}qB31#>$Zn>5>oPI@7JSWUCLNE(W@j{AIEdcsb^dH5}LOL26
z($SlWuEm<M!$+IZ3+r)H?g4>09a<i|L_KflL^l#y*<V|NCDLBbhaadcLj2SoBodat
z9zD3SE$B<IGDElmtXjzj-4Roy;6F{h<N0PzygNsnG~?_;p<i}vP21H!M;ZS`0!|&q
zi;6Sy&D{G%c|Pv@FjbPZrOI5rnJqKxWvxig7%(NO9&fPvz45+gFInw&7yG(*N3qu{
z*Nm}ZyQLYOD`o_1ODxT1()c(+`ZTFD=Ur%K+2|!(SH<40ii}h!_RPTcAzzTNnVw$u
z^-{5?4mhv6*P7AN`VsG6H?$rdu02eU8_hvN_VG6ojopaFa?!+%k+>*=1iv~yW}-3j
z6E`yPRCFMw)I0~99O?S=znw}>by1$0P4O%q<L?bEn1)e{D^u@1SN4tdP%u`(VUEI3
za0WE%r`cCC;uHu&Fn40;6lx;Nt<;e+M*(}ywv)G5u0dzAoKI`c0OS`J7X9*JxB6nx
z;pNH<bhtQZ;%T_R&dhz@Ty1iy0@Y*{e-F7ewKx&?KgPdlH8}iHyv1GE$pa(|)UlnE
zA$}V1x#+h?p^8EJp!3*7Gv$@f%}pC?#h4k*gxFaeCu9CG{*L>nsKo!cf_2RrT&~WE
zZ3+*w<SJRCl(lSU)c-Ex)BJV)><FG}4Mw{hK{ygA(kjR0@RG9-Q6isoDkrUnSl;XP
z`Z%fg^E<%*LH<(w_Rs9N9GQuFL*DSq*<w5!&Gvauc_ZEySn`SBY1G@wQ?i`$o))wJ
zWp;cqUCvzMD{u$4dppR@dON*cu<vum`i~TQ0@i=t+vn{kz1O>`+(&7#{}aXj^1#~Q
z)rdD?VgS`Q6r{Y@8~1LrFAN7M&zN99Wk&eEL<7;{7Ge?!M2n+N6j0pieS--BqQx=t
z#QcvIw;_nQk8f{B$nXyS@9-Y*-pT)+-h<vl{NELP;bHHQmmx2UyS+!f$4KAfo$wy#
ze<8?!)_a%tq_t_U_ipb!<m|f~NkoeKnfd3Mi4!NDdHVG~i8HYeGA+$i1s)d%yjbP<
z#L1Nmi}y$ccH=2nZsIt<t_||KHsO~KJ+|||{Ndf(ANul#NB+eV|8eCzANulFuKmRq
zpPqW@Ap>8&7$g1+!SpVI@Mk{}u@3lPr1rZ;m|rq}7o_Ky7d;oL^|qJ>V~t2D29XY@
zW<Ck2t5}Tx36Z<v&5)mJ^b@8VeS{f^*GMVb$P?xoy@WlD?S%PyZzI~+dI`>2sjp7?
zd99_?Yw#-@InoA58SsnsLFGW^M@mEWxR<UE*GEVhTHA7!@x2t~%NY^(ytU1XEUVlo
z`J;Zhv5olF#umb{2Je)%HOMJ#hXkjbzYNhHDeY*a?CD#1XD38;8qz%bTAbP<^Sc_u
z5c{#lKx42m1TQKD2WoeH4=M3k4lNuWc6mLkyCL8A@K>l8cxSH>@nakn_O9-0>}~8i
z7x90pv1e_cmv8KA?A{R>a|Hdq)!mK48YM*DpZ#9#Bn~Ef>`1bQdiK|ETD!Ti=W4V*
z+}OiWsNcqg!%XRx#ty<;>j&n?8#^g^8#`^_HI3r!Jl*QQUOza0h}iAgkuF{QU*iD}
zO>myn>^EZ<z@+C^jsoyGJ_~mnjaXY=GJ+j~V;_u$eMZynyKpp!oZGk3dw=2C<*H~d
zkb7Ze;Ls8*{y7I|1t3p}Lk?t~;zYeN@SbvcNkt!j8d3?2+Ke8CQQR{(%W<<@I#a$_
zYv%Ofl7Vp8%{_MDC`mX`YxZ`_D6*|nOf&B`1PBT=*USY^LZX@JMbLA(RC7yR?9{g;
zJKoq%u8|#CW3$_m7puL4__!8d$637=^Dk%w6KdQ#xa+Tw^WXV18f1H9Wv@2!;S#JZ
zc(tX?D3I}(e~lOXA0l9Aa>!3AFmoUm3_mlPy*?)@@O=#t)ucCm7z$!ps6G=`+Z-_B
zy^P!?;F&d(OFlG9Gg~T6S0-z<Qc0z&lM7{H1CY(0MsBw3hP~N4U71?~Pnb-_YxYRs
zRb~z^mg>!ZJ3Ix&nJ~y7ft->k4h9>9S%ppC9OJN2T4LyhabY8yyos5wH3v?FC(94`
zB6OfpY_k0UdfZIwz4F4OKcuIu<=FF!dQON?Yo@1`XW;U%9#l9KS^3R@R#b8raZ`KF
zsSO$j2(6LuU{WeqrWQ-*%f<!nhZsV_2^S_vNG~ihIh60SNfJd|nm&WnY0xjv@_EEX
zt?zs*jWQC&e_74zo2)FIMk+XsIK*G{{8yB?Wy5Qw`s89WR$Xc)C7?tyA=q82UxbIB
zuK?DoB?cV+o&TNI=b7b7#lNg*%BAxzY@BB#L)09P@Ap4SZm|cMM?L7dEuUg<YlG;M
z{tr^h{~cA7RfmlI?0>&rMWi|DnS8UPO8nnY$zN9R3kp7`;1?DAyn>%o@WTpzM8UsM
zU~Bf*6#KUdex1NAEL#9fyf(N0)5`l*1;3==cL|DGU3AvAgqzV;tD8eCR^+T2)o@Ue
zhnCL=@E{+SEWw)huTWdFx2y2AsQ<J2kWoXT&-|Yv_&)w>FB0T-4@6TiIa6J~dFWB4
z<fFTzxtQfym?6wSvs&7Xfw*YZ1hi^~zXa53?4>@qT&Y+pHp1JZ(6lSPUE6(Hysq%4
zkBInVkMZ}$XCV>YF~?6gqO0j@R{VMK-Wy2}QWUXA+KV;PW0BQNJvJW~zX*{@Y&Ba)
zv_uXl%u9+phPWx$NY5vqkHQyA!4VSzgvCmjp)MibEZlhnFQ9}p-2aC`<SFr`>bZK)
ze4f8vFHOxL>9syj$Qo{vgogH<<vH8&oD(Oi;~5?mE$VN;P3uU_yHvGP=&mED*K!6n
z$?0pqv!nZ+{>~h4pk1RU5@8T-8e&dr%#CsCb2TGwB1h?P=O8@j*&r>?ceG^4wS?9W
zHS*r@RT$MBxg#0}>sjt9BzahCM~N+s9`{WvmD#M+{IEBACic;&4>>bG;%yy9vxV3e
zNPPH%msGEMFsgRL3+%1l$J4MkX7B9twyEZN&)QavuGPOid@|l)Z*{Mg-(lq+A`jB+
zYkvpfV{bK4-{$Q!?qshwz}o2b;ty!d`n+Al(X<f>M2!$?^E<9a*LGfw(2wWe9f60c
z!%1NRytob|7-<RZDA7h@5*-O&a*pe_J0KbG<QF9nL2&{U?W8~D$?4NACK_)hJp+tM
zqX7KhriSKlH<&F!ahz)oOE%&_MGF8ocUdBgt`aQ%Wu$Ul@VU8Di0v(Nx7o`+%$%k5
zFYah(EgV~yX%ZZC9YVY#NsNh<n!DR+`f!*IN}dA`dSBlE=*RL`Kk<ckG)LM6O)xXP
zSe`kyY3gyK-kcQa(~QlR`Qzsg9tt9nhy8!X^Ywv+WjoLY^jsO<f8|_dD!_TQlMHkY
zR9q8&tJcSN`8~Ylk16;Pffxa8l;)0^#@V>6676}HsL^4qloP-qftAedj^oY#78-Qe
z+n-jwaR$OLMl+3!b#7sK!QqHMRYiZHz-Byk$|;~^nXH=WpEQRj7VA$6Z!gf`vKR1D
z=4w?$Tz_j-!{@lxXQ<I(CqsWCt2!ENxJw^{>0)cczUTiJrDUY|A|))>5LLRKy!$QE
zl>WvPf$&2v7K^5CWMblYWZ-rPdyK)?;1|p>8s84K*$!SAf!h(gkpV{x#COLsl-65d
zj~n@TMi}WvK6VorC{H{CZzRwA+oLO^hEG}}8PXwdiCFwUqFroG|Bw0a2w-hU|34|v
zZuGTI{68URW^IQVZ8FroPn1^Nx-Lzd&Sh+}@hoHS+ipH9Qh2)!XeHbZ-fR<E_MW%d
zh$`SV<7j(7wKkn-uzl_H%{HV}@P@`_tX;6ZA(g#p&T)HD$Ex<m^#59q|3iT`r2jVr
z&4JE6sFzOq*Gb>7{rs=!6;YGzt=AmvVhZZ*;y~vq$Wr>+Cn0N8YlHe<Rc&u7xS>G1
z(HGmm(Qh0=BW@@5DAub0J=cg2ArSGw1rZ<ZjEJ99fO2cZ=P(@c(+av-PyQFF`MA<^
z1S{J&U%_hJ1}>k-BekpclLX%?bQJM*7DhBk7?%oy3B!~EX)})=4P*)kX*ufef%p+G
zAo@wp<xvh^UgZ6B?*p!ucUmp!u3FrC4+ZbV+<W0kS^p^=@jv*MZx*}XJRi~Tczj}I
z+x>;-;6h5XPOL|cz6E#GsIBba@Qk7rXJ4|!uGHB$$VAo0DDYpQ^BnP1$p5_Jxz_0p
zox|zHB~E3F{?8~!sz55o8B!+#<pLW~<m;Xic+Cq^^+Uh&40BRDK-t(!sn`&NuT-L9
zgnBZ<_N$3l>P^K5qOq^WVk><%fvpO<7@C?ev5g+%ulD~EIHeyk4WTF#Uc4Sd*pJ%U
zC0Iy4%oNd~tBH21G)qjTi^>%L2PM};bt;V<_^9N$Yp7O<E@PTesCsAQ?4Gkb8u8jY
zm=tJS$1bXBS~KluST7mq<fI233k`!Jku8iYa$IOGmPn2mG%c~7Pr5Re<K@pJ{L_^0
zHQB8#BUjRL&Xdz8*>3kj8SkGkr7O<qc@8C<<RnpAteQ{&El;sVR?r(k<xz9Zp{VfT
zRVU2h1RfMPEPR|251zt19_aXau={7FueE$ka<h_h{O*-rw2x7+<qHyzt_&17)?JX|
z5(`aS7ALjodpaoe9+K`atPH-dUCJem=Ym714v792n9^p#-v1V&ZqcR0S4I(~-_DTj
zPS|4gU!)wB*m6ur-^mV~{tu|8Sk3cMY>zafXV7~(0}0@o)6BJ7yKd?FzsZX$TC}$j
z>0?o5z>yh#rH}?reJS?Uu>_|vu&MB;wEH-n-B=m6W!rtCpN>;U_#?=e;Lq0XqeM85
zNo6f2MYT)lAdEz$dInn_K*Jd$NN4q`oY;a^jumXDwwbtC_LJ&fOG5VqGB?Ta5D_r7
zPHVSt_P~V_oWjC|($R4zT`TYth(RwnhL#h&erTd_q3{@ghYufqAy9A<56!qV6(sa9
z;Lx`Wv{O22sediO?E#~X3&+L}7v6=0@?hbpX=TBolZY5GXh*|B5#l3-`ZCm4;eZh0
z!Is|CK||>wmK{GRv9cN&^aSa>Y09-LG^%7g>2G7m{r{@Kgcu_G{QpjXE|i01K@H7l
z-S<du!u57^bur(Gyw@$5X8&2I+!U66O^ZiBYQ!|0a?zC$TN>+%b-^{QizI)xHUup<
zt_@3xhpV6wAAzy^94DWNYYEr2KjSsmwLd2|o^ZX2Gx`1j?SZ6C<NDT*1>97(4xEm-
z75pc>^M<-H1fWK;c4K9ib>pqdcl9UDdtraxc$SBtKeRtw>p_3Ql$f2I!?hlyt|qON
zR=mB3l9%x-wD$emX&`fF)B2Zu!u9>H6f?SB=;2!0M#0BgCE{a}60xBu_H|QR?egnK
zr{`>&z4yHu5HVH40U1cZs)?=)1hd<!^DRfhKc*IRw!2>mwj8C3=BOYnd?1|_tV60S
zwAMF7&isJN+)2<{U|re9MeAydRKLSnsfl`&N*yAR7(5;LH}j^g=}XuFtYM|W_jxV8
znlyZeZ4z2fI~r0&j{Ex?$u;R~CLP}f=!8-@+MTJ_;wUVqR@18)C)zM5m7cIeBz$Zw
zME|gYPxE;l4fy%<U$8C8`M!Ia*s~%>9F=&4m(X#qpsZ8!@8DrP>!$Xiidenk8iowr
zMbyI67fO)z!i<i|D>!0P-RvOO-*#s&b?@vS<+HURzLIC&B|;W^E*~A`7#2*yx)-K3
z+Tl;0NNe;z!UJ-E=zQc;rf&a_qR+zb_@?M8ij2@C;S8XVZzGx^H7h-}d3aBa#3eMl
z;02*gdj{^@RUpAUbj>WfVImRdp?<uWcs!q?)^A6Ps{xk{y|nCzDTj|B?$T}pf21Uy
z=^jBl&=~fHw<%$U9W0;~ZJC<r_d!jdBpJ=59lrfXRP041#32I4^PA30-r2}FQwV=n
zue_$fs(gSdF%yvnzIf9u)`cw=6xJ83W+1aLSv#|STV$3>P6z+zG(Dmax5q}o%Lh5_
z2K>>mL>E35Y-<@qtnt98vj<igWiyI^YJ?ApklbiX8{gHGGy&`IODUU{K+c4bX*i{E
zsL5Mx=HP<zX10--i85DEr1Lp1`I`1ukM;-qNAsH>1=GVFebk=Na2|`Wzn*{^?QJAY
zL|16U{0VVqrP16-GFyQ`1@odYyPh-C7?EJoR7s$tG2Kko7weN1cAAaq>l7Q0I(E9I
za=g#~Q~tTk?<8hABVN@Pyl!SVaKoV~d;Uu*bXmdUn(F^sxrxeRb=Fm7Jo;BiWK{@t
z+R|u_IK(Y`9t=1T(Lrq1+^00zXw+cJ)Z3kVMQ^VIbG_kgu@DxQenpci)rAa93=v=j
z*c1ua7dzQ?v6m7MW{K#^kV8$uLa-NKAK;~~3D@RpZNPmzXu_FpO}eH>Ku@Zc@fnCM
zHl7s8W7tdh@1+gwlh5&r6H+{@grxPonVXvA<Ym$+E7(zNUB^;xnf{J?q-6>tpB{;w
zvx21)QahL51+qiJK!jxNegY6;+;kn-;xA_|p++wqXxbjT0@-*eeuZssDfGdmh*Yr6
z9-o+|g03G`JIvo3Y7fQ^(JLHy1Ki3YvL%v}3J4M-Y_e>sSbhelt#=Z6G&2tXma&R^
z@ITNs6GY}BD|-U8=sMLYh;(HcKS9LxWZ8csEudZ8XWA&kKKMaC;k&P%4W|3|HK0bR
z0F;ducT$lse_E7x_@lTMfeq+MA`ED<Ye16>pVY*YI_}#b0&jz1l_Ggpx=e7s*5LZ@
zSDS_Q{SPQ;_oFq0evynnROW{WR&MPYJFu8(KfV1JHe(#jZP%ZoR{>GmF@#3>YIGg`
zZ8yROb08e7@MHgxb_9be)oKnnTQ+UfrmOWaK4XV|Bi|TKkdlr#5P6ou9t%=9^oWdg
zL{UJ<wrTXv#b8*<NC#UH*8bM`HIuVQCeQt6jo?v@V9pwAiCr!1f&r&#reJ^oG0Xz?
z|CkBnfd+AZ5Jo}_LzYA%;W#s%)V`0Bb5lJoO%zxKykVphC0yFPrJ3kV<}FG`AzWe)
zP%X+d$NW$5&1M38CId$-NWEp#WLVj=Ts>1=JYVH>!#PV5iNdWPZi#JIZw1tvU^PBD
zRkP!XZBGC9R9!AqJe#aEGYhhxqz+@Bg+MAHs=b*)eSjfxryJloh1LyYMkA(-|ElR1
zcV6Z{24u6jzmYy&A~o`rL?ZKMfLX)xU4tA_I=08e#oGw91az<HjEgxII1%v6&H`jH
zoXI(~%*ltoWWe_};z%o{l7rb6&*N9nMMrvxt&q~-sid8zJx}21hAE7_7N19x-phR|
z>h)mul_4+fKkntxp6^9>KDCBcJa6`WGS<k*z-ta^7aSjpQx-Z28E0fZ*FfIc$Urqw
z1}Z9iub=mNf_GF7nWUXwpm~Fu2UK%dYhaT!_}1X}>xB?s(fT$B;h?LyKeZa>A{~mR
z4!VZKbQ^lkFquYmuMC)AUpfoU-l?!MWTKmka1UDI8yG~kPN>L;z@vIys$u>&5kO~%
znJA;7%Jsadh6PFv*R$>k+0XT)MIi?b&_A!nw+VXV?#?OAQ-zt3JK0PvFUc0hB>X05
z@qdj1AyH|lr`0Rb_&Tj#^zWypK+i!OFn?AxpC%~gL{GB*XXY+Q(F8>Zr;gFg*aHpi
zfXEKBgyZ^mykEPCuMe^E*#{<Xk<4W%wlmfj8$${+ge<-<HWJIkTC}E{lP+enz>YA_
zd-<#B8{%6U6$uDoWn;1kwFs97S`!mn13M9b2?WV3s9kjY9Eks)p{@y05_g8R{z;mL
z*`;CWw6HY9l_T4P0vs}!&^TOBK}W9`wU4wE>$Ya<%z6EHr>Aw0APv)vTyHE!i$#U(
zL$tJ#H?SIHbrDx8qWx~5U+YBYL3polv>NewII0F@`$bJ#s_lG{FXUohBYVErqPEXS
zp^wQ4TW!~3hbOKNO%+ZT&KIT&haN0U@s9`Cm%R&RbY?G>9Cwg$c#*HVQ;Hhmzo4Gm
z*bItGjTz-40$Nwve^?_Dut<_d!TI}D-muKNp~woz8Ru)a21AQkq^LN8(#r@K)~!Mt
zCSF5h!{TVpH4P2Z52!=RAR|d<%LYzNh}1V+2=R4E?U8m&ICCFqHBjr?UskZLsXg$%
zfb=kP%1gJXv08LuQD#CuMMr@yn5`3lXeeuFl}NxFh+mD-JTpLsS1Dg5?l<-tk{P<e
z8I+I}Y`<E+DN9fIcToc?c|PSOF~UtT>w?%Z503k~(x6GsHm9y7ni;g3W=m5jo=Ou5
zJ^p4IePZ>dnITeMTAV(O#vU#!(D+jFNjs^6S^Z+F;+H^<EQD7W8SU;_V}?7S(?8H?
z{Tup)`X9!h3n-JJ-<3B>2WoB@hwDu{&1jqvRvK9^efHkg3({0KpyCGqA8s-Oy&)X4
z=H0G280D)200>02H*yGHQ#VraA*t`gf^$pPD_y%PoE0tTnn0YJ73|hDD-Lh6oILQR
zJtL&j^77oQ$xz$~n(x4iO81+uhR?y9ud*aN-+VP@G%*~Wmorx)QWAYNayg3PC$mXP
z)Ka3CW28Xfkb*RYGO-UtJeb^x$fbBRQq3HT)Qac9tor>Rf0}=|N>b@IweD~Yv930z
zr8cHrLG^b1yH}!{q{X_^OoIgTZ3ZyXMrYG{41bs?I~Uk#NW2Z;ZR7>h()6J4p;ds<
zA2$<`YbceQ#-KJI8}$PJnsI~{P;|&BKrq5UDp5M>HmG5h+yhEBqmdJQ&Hpi8TRC6=
z6)ZXj&O?f0tEK~;w_bV6#`+)ED*R3b1`Pf&v5gHm)^3RZU6gn{Y{CCxy~o$pdorx|
z<SGB#)RMUEMt@B06KA}0lldRDRK}q=)E>9$q8qwR(vEGC=32ma{ts0quC@Q&3T({#
z*NyBuXkKj}A9U&xIncU(b!;4o?z}FJz59iA<hb_hD(D(6?MnR_R@a(geMfxhwb_we
zG-W8>bqAbyTi|Mx5Un<dv!VunSiw4YKE#SJ=m1O<mjQj_33WkW9cvaH%eLAI+E}g&
z4G9Ic(9l61z{=ocv(g802W$jzv&Ezr2wX>^>Tq#l;`(Sc#Jmm(V+AQRe0^-H5JZ)J
zfT!^mf=*QZ#q)IWMZV{2oO~ge5bNTwK*}|>W?xm#32HJXD)B-yAE*VJf1LFLGF%d#
zr5Ko8db3j3eO4nP28J-QJFl#a*{Ez%_;t0xv?jXP0{wcOKWQb^L@w1Mmm&rvVb2O|
zfvwXpVAGIj00OW!*t_F7Zlp}wv;+_!W0PQ}#4rTY0Ik>qoC7bRTFLPQa!6{Fuyb`R
zK`ad8@Gg!UcHjt@f+y|C40k#5oT>NJ^RNYTvm8?}L*^jFk%>3LudVe#Fy<&@YLQZ`
zoFBl}0M$ujIOaLn3^ww{RJ@zTIn+RTkXEHpz)Sf5l$2qQl<z>xf;UH)dilW+Lg`)Y
z+W^qNy$hiCQcsT*`KYzeYR&VM^nbR#Wo@)h-`BQE7!OM`=0?uQz#qNBnconMtPo(B
z3VA#SkcALHyv^{#>vyt71J~D6$86bgCJ`&qpQl8#w`E#r%O>*-ZZRbCJ~bZmPiZfS
zZRcNA(VlQgv2eO}Fe|4G^@XPm>Zhff-wE+MR-wQv%>bDNn^kbeKx%`^jQKT+C7>-o
zQ6Q$f-g!W4LG<x4c(yvKTvA;k^qO(Rq>h3$;?=l0Ur_xoDY&e_c(2;Z?)+~`Hb&!r
zqiV4;)+@^WXL>ZK`##0KL>g8QV$I4V@0gQ|kN0;_Jkbe6gc?d;Z!~)XcCHEI4K^8V
zd`-1=AHw{)(lpTi4=FGh_9Kcpyw@Iv4{E30OH)#uoCJ7fotMH|uP?R}f=J>t{h{X^
zOT_Y=tahTAhV5Q#H+1$;3|hN?-N1IuW*3wbvy|m5r$x4bomPlQ?)ii<d*Bz$%GXG&
zo|;e3XT;1A(>H*5u?c2b1m1YI*=suUEG>LUF@QEBpVuIdsFS~|*uPg`hmE9Bi_Ii{
zB^Evqaxzs>GiAhPtr@38$BME+8gcdn*bA0g)Mo8lG+a7vh)*naae2a7*=9qqS@~<J
z3IBW4kndKorr@fAPbm=paXfa4bNxv+$a6dqJw*e*kEj8}QxoGi`JbU|Ggp$G241=_
z-pH3qGOu^09H!mw|CZ|eq5_?3-72#|`At3ErurnVGclci7cop%78fhN$&K7;6(ld)
z5+`4HR)uYu+WP$_J(d(0dqB|GZz!;hBI4fvn1b(7@CgOB*ndK?&k{6y10$5_GL|i!
zrlfzU_SuI0C&WI@U+p6VnFz}bM*QFRx9^W*r($S=#O`w;6V#U(&g3(xOtvqZ%IBjg
zd{d<Q5ErI$$xM>4KeLP6OeUEc!1{D3lSrj`>t<i=ndYzSN29Bs+Mm*^wt*xDj<IY!
zo5&{lAIqk)3Ge*|Oyz4LTTBRizZ{XNV>!K+x#|FIkC#VW0M0LFY1^Rp0eiz4Oy<l~
zQs#~_l8hAd-Z#n{BF2Zkfj307w7uQr^m}`}0{;iRz1}|l51L0r$T#nU-Ywp(yfN&)
zN$X->(Y|rO8-E!zR~&IU-iMs8L{hd8m%l^|AcH*fnMlr7=PMC2Nb`}15v2J>#0c{F
zz$|h{U>3PEFpJ#fy<mos7(m*$zsY;Bya!XwhrEYhPGSUUBSDVZRWcJvjm6$p`Xd@k
z??p43RL;Iu4sVb%>7DeRF!Ra%K~Bu-n)06X-fihOxpeP6-cv7w&Wbm;$|bW&Z`ymt
z`vxm_i}$Se9AkHDQ2vznJ}ZB~%rVD7fjCEb3P4k>w+xxC=N4pYY^GHc{wG-z(AUTK
zH!XrC{1Fi&%25Ang8kyXh`PfnBWfm&f__n~4<JC!L~BFRuE)|NvK;X-MfA5j|00kn
zG2^xF^F*@bk@RD-2mvHNAs9r?D&AL5ToT6>vs`gH8+q}jARst#VvH#}kiaHnB`<#_
z^SOC!Kg0z>vnf*Q{XnFecIQG*Mv=WAL?&H(-V7xBs-&#-qcw+ltyv<V5^4m7<I+N|
z$jJ|B$Wcopcs_}gwm?>4hH3Gw*b_)W^MB)=G0Z>vq@#c_B&L}mH9p=OxE3&FZPCO2
zA%f-z25vPBlmM<?XjfM9{~zUlzk=Uaz_>=5`5K1(_|?G+!J>ci)c7{zaW;p#Rc2LB
zSaj+lXFCt4UW-fqPpTdn3gO4k{P~$#TvJFOw39QcUv}Q+Q8G*g89Q?NX9ednp2r(9
zLrZfuSdeGPtQz1UN2vM#nj%j1Rv_2Zao72Sl#>z~C6|q(osJ^g&A>xjyGa4t&t$;O
zzNx82RErid+7~OOIV{akOe5vZCp2g?UAlnAXbyFn^##oqV=nBPOY;AXTKyXeeqF(@
zDfo8^Z0xO1zpBUfk*PgB2HxMH&Kiu@0KXbzE9fbohii+*yjWpU+CXgsl&asDY^p33
zbyPUKPN$w61kO_QNW8f+wr)GN>?^y-4?fj$HWVm5vDr8~7;87Cr4QRz>IAUNN~k)m
zq29EXa`wz_W6Y<oB6Jj~A`(=jiWc^K`jyxNc9gu9a0{1H)x&(v|8WJ9^4qGsmT^R(
z<n@kP6lEv_RYrNvtNGtX28=Q0T8=7gtl?ki2#R!`d>gyg00wS5-5dpOJ3m7w-w=%1
zjUav`y0UNm$e1I`u2-f-mW4lQB6?lN!8d-1fZib%B?L5Z){*uzBSq{(lEL%;1ybU*
z5+Te*6ef0IPG+u^P>(hH?|KT$ag5q1kpVCvbz1{~%wtAnVx^u&f|wX%c_~(`^)j{q
zsTU=3lKpoKp-n%o<_4}}5+zVUy<$`*LdDzgJ0b8`|Nk25O$@WMepq=J5>bw1<lF({
z(ZV!N`fz;2?f_tf>aqZWL+upA5p3PbY8n1?<>JbY{rmS9-uK*-6HgSLJ9XmFvyVUb
z*dtHriFDi4E4%RUVamPo5^=Z_5=s60+Mur;f{m@Sw02xm-x#2>w0D5vXVhYYL#^Y7
z<CyEv(Ht^sTHsp;h#1q!yCb1xaVxJ8b&RSE*vK9Sr=+-*!&8ZOt%Id%+@5Bi{6I>P
zm(#wusXA0N2i@oVwtibPy5j$#n%G{ABkOn^25vs5^(Z;zfrNIqq%_jgO~@vO_{d+a
z+_Zj$ZdT1(lFeO)FIv1}64>I^F-A-Rcs)ou#JrcT&Q-_#*3xBxm5ev;W78%rtu{C4
zTFh<ZYe_c+Mo<g#HMiiz4<ITBoq~+PuJiwZyl^b9@nL6aYuM|D_4p$S%+#oR(2P|7
zZH<xmBcoXP=okUlXz%(_*=X{^Tdf~$VfFek{@%EgAmkt6cr<!dxYs#oh?Pu{MtpzR
z8o1I##uYE6&JJnQ(Cd?QVvyIH3Hd{54t6dlQ~oj^CjNq!sU5C<g2LAVIgLErg6xcQ
zhP}D0lh;8k|BD(!byue*=io(x8ajOI`VMy@ypCqPHQVf;QzdXAQhQJbMIh5)9mczv
z3W%-EWUG^6zLR$oc^V4mu_EdGMHrD{rKF}O_8Y^Ow&9B#Hz<zrEYNN(*BU+yY)d_7
z?+$8?o{85^7}2J$#B1+Xyj@q=hVk@7bI3@Uj*gW|*KdBGDePNf!SFuH(>2^s*6>1w
z7QZDO_P@X&UeBHitwpZyPyt*kb<stE3!Ib4WergqJo!7#+h-XkXB7Pz#Wdaib!B`^
zvF@q$zoLw8wBvt?wExH?8lzU2*=gX9u(LC9`Rs91nul+oDPFnd?WWf4BcpM;edA~&
zYeU<TK1~FV8d5)IGm8zt@*MzcBpfTZk;LCFvx!2GsRJ4>ErSa^Fc3<CDe#Z5OoCa&
z<OE(|niy(<7Ry4YA~u1w05_0rkP>U*NjixTa|oi!{Eyak%0@~FE%ix_$v;&fmE&tU
z)6R3I{zD!LtzH*gZbNNH8QS?8Yo5vJGi|0Tl!36kW76o;#a-ct@xow^y2J0A7{VAi
z#n_7He@gB6w1VzgYHl~?Y6oMj&5-?T?Z8pKCxsD|$IVzmgjK+ty>0!3y>(Sx+ffim
zkjSR&1QAV_d2`Ikn<I!lRvor96m;<MIE<xxteOLX2b%<)k624vBrSC$rUL`29UJhE
zc^xz*oRuxNwzMP6je|1>WkgqLfj^?aco!OYXT2uX27JBqy5ZVu0hj;3X%zqa#-a4T
zOj?Z28Psy4l!}f>tN6hBk$Ni}+%-;u+!`wzrxhXsTXxDDcf2s23H^QMQU|u|rKrD;
z5@-^m&V+wTmNnx=i2|bvcoHcp6(}@xEMHHq9@Tp(Ry%HBNbGE#FG(@rW^V!WaUyE2
z;ICzkkZ-1OU;%~T8=!zV#ePHQ9ABCS9b@{BXm>MBe^oZ1QqBGB`h{q8iE-#SduY3d
z4k3Cpq*{hKr~%OhAc9m5?WG~i%@S`hxSLhfC6X2OD#nZr?)!;A!-Ef`38n$^@X~!X
z&M{f)*~Fv(Vfjc9<iXpe0!@OFf;p(qLq5zRaxs?+s4=lDQGj4s1meo$ib?SXiorH`
z3hamDFn>}R+7tMFo~&PUrYyephX1&N+Z7m%702Dj`k~9<YW}pYAsEw82_hCC>y6R6
z;m8vf?HUZd(i)7Hcu;4B9%a9zaF&z0{H*e8gTX42PBnwkkVARbc|fLY30{NGfMA3}
zkwgWJ#2AErM|w)>{7K6zWrKwm2}&r`dk(l}R(>v52z~+^r_=WrPMa&1^%>{y$C%Bl
zpKhmzwnuH>v}Q{c`fE03eqn0ShkAA1O6^PMx&Poa_Hfm9;df%;gqvPh)J){CKDX=w
zS%>dZrmJmQ*5!Vy32pJTnNLfYQDR)1nKR|`5<U@~VVzqGsSX){EreUOh5k>Ljeird
z5a_p7gQTbDC1sOe8n%jGE&IOIMBw-#M=Qy9d1-kGE}x!UDe)4D*|LWN9<wT2hXoMU
zadauUe(upK66IA$i?u(-T4I~+To1!KU+EIUX*Fqk+16Sk!fG*u%&{(2mr<?`X3&%f
z;OrocHxlOzpB1E4g@21Un@`T&rDB@kGJ-g2njHH=jlyjabm~<rxzw7rTBVPVP%WsH
zwM4C~Cj6dktOrysFCWyGVw0vPiv8xwh{+FJ9fS#!7lH(v(PFMXS1wKYcm?!}!>w18
z=&W(WL|$trLM<QGl-ib0=&k0EIdgRCNWm>1Z6IK@z=mmW#-6&!-vv_$Yi8Z|IEof~
zo(ewB!+OxrD@@ZWdPp<7lNpi0qv+{dVmVkTjyOxaIhGJpWo6g;neM8?#-VGzWlEzt
zxA`976YvQm&*KGI_Cx4tiVw+L150eqw}ZWr*WmA1Kr~3_e~`{W-&AzZp-m!H-E-h5
z%IO4xEIWb6W9&JAEE!~c#;siKwG`=|fce0X9Q-Ak4Y`@CIFpl-uM_y{ixR0)GJQ3P
zo@aC(FDa}2xSZ?16kU^xLUcaiS0+{m8vViZx_3t+Gg^!NAB|S~thEXM?}FC;-Ph4t
zBTZ=q>pEx^F?igrkJ=T=0rj{TIqLV|`d$X*)Z=tml0FZLgWJ_-_nn}&ucPlOkKGHq
zZsQnis4E<c>Z>*a3B0|0#Lb{$e~-~fZVyL;Sl4JM$Bl-|d8^S-xo|YB<ffycQnv0I
zihb>spi)_t-DEU2)YTe|SiJ|Hd>?#kR?!Taao4gdue7j3XtjGk$YWMCs+^*sc7bo<
z>4SyS2Mg!N+htqG%nIu=9P^0hUcsU^cngSxLp3vC`Owi94r0CCem?d>yS_F#f<Mrn
zj#BWUad~ex=-hRL<K8M)76txYdoaYqWZ0R<9A&RNwN8UJ*YZx*{V#>EjC%e(3N%yi
ztIjM|yk_Dgp?{y?p9#WEByPM;!*wFEuMN3ljR+k7ohtb-K{K{I(~NoiP1W#ZHr+uE
z-K{eB5a5ND+QVy-Qp>?7mFVBAI(kDAz&cjUcN7nFo(vO;wcPACs$BKCdfq8$DXiO<
z9E+QC7EVEh*i8tuZ<4DPp2lJcF`(Z(96uP_8NCf^f3wsC44aibn?Y4;j{DlQKxp_Q
zkQ(w=Yhs=YEkYb+Ae<2OfapY`$JiziWN49OI0R&L1PIJK0fj0}8fI?usP>_ho8bsx
znkxuZ4oh6rfFJEjnFfl$AT{Gc9xW2Uc4Q3~UXXox-jh~KFHW8$kojpgx<pM+lh)7K
z8lLR@z?r!3lb*s|LkZCmZCo3aHVP=Vr@DolETDXbgIrIwuRa93jYFNZNm^g-sRDLQ
z=Vipdh(625>p04)k8&)zWHJ1G`mfTau|^NA%U9)`i+K3HrJ>G`(=K##Xyx}J9t(R&
z+4*gF(~RJjildLyh^gZZQ+;9t`tdrN#{qvYf~UT@IJFNWGUI(~&o=+xcE2fMl$ZZR
zjBn){`PmqVBJTeO%Is{2TRqqqSlh)M4T@++yp{6*n3UaWH%|iy274!PO*l8b%pWJB
zhnZ_%$0Vnhzp8EYEd)4qOi$~E6HH%my$H3c+b3{qB$17u&=#PSWKZU)k<(YX0zYlB
zR=VK+U8M3;D)v5t@jcDx$v|~vX_0+i_CKd@zNp}o0^69jf4@~tcvyk$*#E4U4!8cF
zDEPbz|Gol~??~gt^`Xs%Hdd0Is|_f8K}Do=ZQlHvTckY6%}^z^f~zQMr{`w47HtdS
z{*GgxIr&8F-@|*lz=ERAQQ~+2Ek2^g>?de$4|H5Q*<=o2p_b+tB-4(~EzgtU$RdDi
z89KJSJu1h4mzrsSsCCZA=jpUNvbIK(OMX;ET=vO{3z#Rb<8qsh!&%@*&7<#`cmj7%
z34Z0;I`x&lv_w=AVmUT<9uXBTp`ge}B?eT1YcLW;b%vPANzq1nBUnntv6_T>vtux!
znB)kz&?ngpzL@e=m*gfS)@WPg4P9$_CF<|L64zxv#tj{=W8=j$vMUG@ln2G~el>n2
zHlLgal3aoBsnhMNa5{-6FU8if%J))~$nA(+iI4D9@4144_$9fp>mxnyWgyu9+G1Ju
zTf$;FVo8hj5KCQ&&-X$!8p#N}NqBjPn(UcG?E@kwi05##GFKPDWjU8(7Pp*BKtc|o
z{a&x}ag(igmqnBW?_Q>SkYn!#IreUnckvY*jnJNfD}fLI-?<4|1Wb_zT_b-5c@-lO
zUmH|g^$pzcxz<<<EfuGDA`TgYA?ixjhk;w4zZ|)YlJ#XQQt@>?^m3*?5|)+TAiT%^
zYQN>jyy4CBnUyUKdCMMw|2P0q(FbJZKce-4R$CO)_mMr4*V7oF=0SWA_i{w%KkS}X
zo0u7ty_wU|EAcDQk4Bf0XaH`Jt*lblNR<I;SNTF$`CXfpA0M6YU(jYhqQDp(Q3qVU
zS)28kQt#A$9>K#4vhJ3H6Iedy&NgFbxS>ed@pETq47gqIH`6)^+}>mKOh<wvP2j<r
zvhs22-XO-!k~qxY8xIngzuq0IQMv&VwT0Yu`CWJs9v7`~hP)D8+D;c?Av>SGlJ#V0
z8!%PuaS6-BZ{_g<b(>F71RPnlJ|Ftln3gcs7`AP8p(qXW_WSQdT}Yghio{~KyvhE*
zv2yqNJ-m@m+@Vz7x{P2$zg4aE`)(rE@2&J3o*tILD7AWT)9<U{1q!!o#un(0zv#OC
zRvr00bwtDp9f=;k;e5u~P;E7yM~l-n3-82#MAb^3?S?TEvdfD@qP{ewy6g{VUH2~<
z<GYyyE(c~mPUg8?jnXo5OUoD~Kg0L^r<FFTnBkr+ioK*DuZ$`|NNL)9oK>o3%%4%9
zmFL!y;eaRg*y6eP@6qGC6$ofL;xeR)&Dn5tQ5Sov1D-pCFR7eOw@}!R3mH5^W6+3%
zWRwf7K6E1;J6&LWEV>f}Fa{ZkUq68Ufml9{>@${mGoJ(j+~~vPWGCr5qW9q$As_3D
zE8Hl%-?(Vr5(D^RIB(pwzNfZ#?^(aBwDw5N>XCWGJWsJX1@j8tb|?C0RHUMywYGd4
zsTLh>yJH|h37=vsk8ju$xP{i$UEGEWzd4Y5bV{`UUR%W1-3F#r+0mf7(z+Y8MGab3
z`hQZNwZ8px3RYTWTNB>gE1l-IZR1~P)7WwVumAtL9FG0H$E!=|FhK8g5}|#OH$KMv
z*3suSLBqcSfN<IXlmDE><dRLk36PKy!?d0lfD>&EC<v*V>2zVK(oiQ2TxxDnIAvhi
z16EFh-R~w2O~5r+3IAQ3VuBnnwBnXy7&?ZjFb;&FpyRD~PZ1B_ouPb?WA6qz_AZ8r
z9#>+4-Nkn1Bqg??mpAO5UCjx@L&+n_eHHMymb#+$(>@%!xz<=aEVVbl?onZMPG5aq
zKiE7S<^|Zix1PaqNH4zm@vjf%0PO&_=09wxzE<6esXk$A9K<!CKkOvZ1GeTwB0c)~
z9A_|2EZy<(!3hWcHrbJ{zsWVM&EXbJ+xEHc_NpP~)`kUnKB48i>6Q(VNPC9_xX+1r
zgcbZvHp-*nM%jqXoB-)-ZB^S~FcrQIF#i&Z)Ug~4!kgS$@~3M99Snsw$g;kEoxt7y
zu=*!s0Do11K?CcqfpndT4${*%gkf_nFs>_X)&cS;fLs*31K@E7z!7?H0=PGhPuJEI
zj^&e)_wZLM5OCxaV}Om<wj<QiMqF*eCkNJErm3MH2|F`=m;`skd5lV&A*6M$4e_L%
zeTr|Hg_3iTw6bG8VCl3$6}Sk0B#zhLd-<z<7lD`-@NdPm*xX85ECjJBbZ~!=O~-#J
zh`~Wa8O*<#e!?qen~3R=ql9er0X3+CnK0j^0&G_9zC#%dWegO-FtL$G0=+R{TRu4p
zfCwzG6M+RbYIt-MtgjK7Cm0`e%ju#LE-TB9#&!%gqUG~Sv%{~QaqUq1mn!`G3hW5|
zykcg?VMl1`y^ja_Q+BSjL2v6l9v~m5AJcAywWpv#`Uf=yLws@{SVDYyGg`-U&d%t{
z*7eg7=yi4p_B1Ww-y6GmFja1#_{*7Bqh<(-Xb#IAJB7fRNKSOol1bjsiZfbLbS_eP
zkb%Iu<LL7rAiqOLn5gA}6i6c+(|-_y<;5gtR>`ou<UC_Yn4Q}jl?zLggU>5+KqC%N
z8}~F6qaLLZ?sOrN9)C$mxF2|&77=(YiGhN6BcCR#8M}DY$xctGU=$V#oKpzQ79XXJ
z;G6iga_V<_y-bgdFyF~A%L{Wpx^n;e!Ipm&QysSJ%#69%VI34^s=BTL1(LJ}bc6>Z
z&bhlxW=dUmms#t=)5WZH!T)#*5VcMb6&mY*nCz8%|9(wuUuEXL<dhqVDI0;B|FQ5>
zG{NHbZG9H)>acqqefKnI{2^HyN^exw+w7HTt*-D#V-b=awI5)~LUyofI6SASqZY?h
zuLd(=Hu`GpO3oxeQr;T%Vt50QH!fHSbL7>NS8`I=;eh*@I!p;zqWBxB-Dc1v;UCt2
z-cL)`{A$`_NKamsJSc+%$%qyhgvhS0<bXQSE9omeSL2W_NU}sKfT%?%^j_&*%U$W^
z9S9IFcO|X=ggQAY73PMq0eUa@Vq0xi8!$zJ97iU++Q-+q3JaF$FSs80zfkY3(?`KC
z*?7Df4W3bIwdV*uB4$1Mxb-BW?_%JXRr;&!B3cE^+@w`X4O;~m2*?+#NoZG)7qly0
z{Y7h8zugbudjlFJu5$VfRCWjD1c$4Y(yqts@*aI_2U54KYC%S2#e{f1_G<Fg6i-9*
z!`v;9<VK3T$BhZYkyo>qlb2JMGnccMduEa_JX{;P#%8tG%A_x6VVR7mM5EUjC4HB3
zjOP|h>2s_T*u(v5!#s8|sQm`(=A+TtEwp!3?S&ZyM6xm2%IP=r3hW}V$0)7a3achj
z-*P1pygjzI?J^$UxLQo#?AI5>OsQ{g40zjsN&|dtz?Efv>}U*FZY#~oZ4c+!+W`jZ
zc!O)n@-M>zR;{nbt9@p7ze``*ur}hC<Ck-n^OyVR|L)C7AP<Z*2GytfPFwBYqxDc9
zo8Lxij(Zqlt3wU$s(|q1KTQ3ha8+1LD?x-OMx>1)3_D@ntqxanj0lkVN`59b$2giK
zahUH62jBQ&1NmO<i;ZD@7hR%ZPzFxp&GKTL9qiwxz>eldXtqhOZxuWNM!yBEsjY8F
z<}eY1iNk`4Z_la-E?Y82h2c$OR75*;vsz&C+XS8jd=E!fj;sfiTrLz|B(>Sm5w^}J
zd9#>#>RD$7iYZ8WR_bF;rj!xLl<o5dNZ(>V-5K_3lwN(!e$7Vbx|lHSzQx?d-N|XY
zB&SU)NIC3dVHfX=_|IyzKSFT*_@j%<lH@(Qc;VTW?mBlIhF+gHJ17XpZs?1MPn#i)
zsm;)K`Vzl_Wy3an)v;(|qx`=T6%UwHhy<6qU8w+wnHD3h<0PUk1W2x!aXi7Q4<5%A
z7B+v3Y3X!F{5-zwbr})r)NMQSz7QK#Oq$@>INd{YRYMc9+D6|T<)(zL{)s|rt-$tN
zfd)i!LWa;Ih*I?^=wHnB&-#@J*5UVS(5*vp;|UfM=zvs;3EZw%SVKRk>>pBKyW&T}
z(vC%>8T{vJru1%77}h4<gh3R|yxG-I-Z9GBCPI77CSp*umwf)A`^WCR%C8!`mLLUI
zbjLP9vFtl}tC@QWCy=M?sto^k*p<cJl2J$Q0OC73_T!Wc36SsL%N!OzN>ZC_GDgi0
ztI;9%Z{LmUw@8XoSioGuE^NUtTE+}hx)Yr{)O8)IDTp85(68146ParL>QTSUzgQGK
zF(}S9FSdih?A#kVO{u_%M|25#FEfq2+C)0ti{Zz^eX^U|PAUaCK9I`;V&jGC^^qV_
z@RaJZJ!*?qHZ+)efPwUWG;A;i=g1tR?fuJ;8AR+@=FRB(t3YiSKTB=?zhEjhd`64Y
z=3AT@8<&JYV=$}MHwzr_YbG}7@-Kv49@tpD4r*5Yfw1}k!Kit5v0SBG0abzu=+(A}
z&_Ty%>D=de$FYYKTooq4npmN3nb^4|wc=8`r{I^u_8f5Bv}?N@!?rCfw2FUUwHp`U
zMa2xQoKmdK=Y2_!zrnEKiw_ZDGcku;)A@=|VVSUu=hAa3Vl4Y~u)K@?vaSPcELN5!
zU9Fja`u3vl;3#PDO>&8+G<h)}A90<l^GWA#>(|&B&0HXW6u<v}SJnx=_qc*h_{RQk
ztHC4nii2gHXA*fbl3=8QXCp+SL!D2h`kCqqc8HT6Lg7G1wu4LY4t>`aTE7!kep(Pl
z&GThB;>*NlJrMd~r5Pl$ooyU&&6`7^UsSG<{5>5gvdZMuUZ!hYVH498XEOfRwKEw`
zr?}FJjT-ZMr(J&*yWu>vp!^>@fbK*dC*(VE1t4bjPy#cSHwThPMBmJ%<Udso*b3A!
z*1=D~l|XD~d>`edf9{yzl#5GJisz*bOo@(`c5%IFbz8+%YaOx%;KC6TZl818-@skJ
zrE&yysXN%<j)`)RQ3(uKIvkMqNYYcfU?xlsF18kWb4<?a<QM|Ot-#BPEBHG!#UQ4<
z8)MS4K{@=bUhHB93110hKE~f0MFLuEhjKZCG3^jkrcm6ahr+`s%sEFeSSy0j<X8%~
zQ{u;Bcfu7JDCkOc7x`-)mB~xddS)%_rLKa_r5-t{#nsYCdLp=<fU6?8RQ)|qam9eY
zP=Nhlmlm9pw|#9ZE=o2%)k9Odo%*0|3eZK$!UC<TrdZ{N(62v)N^INJ|0&(VY?X4u
z_N|6ULT?JXm!A6zxRYEr(|cvzf352W?9;*iZdVlQS;;!q4{kSjtJ@Ja+=&qCT}>9u
zX^+&jJ~Z)y`s*M|;Q*S#>HwD)=>7w`5tQW84v~lLc63lY#%g++`9k|DDOcvlifKG+
zFlxu#^YQTcxZMJT-H5?F;ca&pkFlkMoSlmX0dK!w!Iu<7v`_;xkz2KuPbtu*E~d}c
z%@Ml!>n`^5B$pmGU)AYyF>?lhq$yp$l5;M(G_~aYp4P4?sa$3!GAt=Ot!!UUPq&MA
z9r5LJUq^ny)Y@qp#9m9Z`BBF5ap+#$c`@%`3!*G+*9J&I7>A;!;_L765*+%I(pF+!
zr(grfooKcyk2wjC%L@HtjReo}dh(KEBBJKY)rTC5UyH9{_%4^C9DcX}5e3i;<|I9z
z<QhO3Nz9Rsi~)~64UUW)n_x9klFJLtcMox{2<&l}1J3uMsT9_br-pn-4WuJKpq}~u
zj#M!ttrW*HF)usX#j!c(_2K4JZw>0}jouM+^iE3n6}c?!4ce&phg9Ys)fJYI%hG(S
zL|UE*krzxe=3PIU1Y0>@r&HrUVd9<3(hXKWp_k3hvy<~)YAjV}Dpu7h{4^i6D+EsZ
zXY^cAut;#d$9Wr_om}8X*MF~K-=ttk1<n$T4|g~p{R818Z7Fx9fHVl5ixE%n)bwao
z7&q27ifx`pl<~VNa;Jh{SMawAevJT^>V^Vs%NWu$Ja|~|2-$N<50c$V$*C2Y-#{F}
zm^V`Aqx75#$yp)$3144Y-~M?8h7nrsuFZ$%ml@33QNF4jB)>SQ^MFXrVm-DUfA_kY
zBE^3&1l~#I)Xd!zx6y(<9DH|6gmz~5quC7+MD5Rczyq(AGspm87uyPY8s!%A8-WTC
zIk1J{l#LpvEF%;mEotNlu^D{%HnPSIfbZ^)H+nc4NoW?l<KIIJa*?t<wsCuHD<T+V
zD?Wr$UN2iQVOufbzq`?kW<RQeGSWsl?i+bmhIhSJqo5=dZ3z4L)y3X$eN-#9A#qX%
zRtMo%B>gk)V(ms=oZ&{=uLaM&_H37K>%NL>85!s`wOD58zJIr7!<OJZikULXy~HqR
zvK95y6tZcao59BeW))@U{>IkVeR@Syq<=vb-a(8tX-lTf)VC$`bE*b63iV~~=WXdT
zKp8d%LtTgVjmdb9wsKsm-EW`d4er&JHwx+e1<{go@QN)O1_RDTuKa3(dVHh$UehPj
zbC^s`L-0PJR{n*0{7%Je{k6yvbKa8!4A;c8NwBtb2O+NbA*XUYxiu&q1=(KX+!F4$
z{&Jfru&%o$!|sWn?xouBXDg8&{vPA6HcDeSzCfj{yDEbzx5>7Xw!~L5&44Y%wW}=A
zESgLye=j$<!^)ML6aTix0PNU7-WpmRsU~@fyQjLT|D(&9gr1Z0=PivK`mhQA@w(}*
zv7D5m>u|SljRrbvT)Tnx8rN>1`^mK%>AKj!Fz3XsY+buC%J&!1q8i0T$`BA4eOM?e
zE@Rya0}PETbX#9T7R*x2_}}st<wh)j(>j39xe0DT*<=~{f2Fl1Gfe+~DEJ!%e@k$E
z;5m2Sdci}e!IhwgZSYy{Kd&Ojdt<mG{zF;;f329{DYsl2EnzV6L9&`dC&~aZck|c>
zxKLUaB->ToAm;BW=BZp--vRbHt<Xo5{-Oedspt9T_|~`F4F02PM0Z!3J@xVe1k)s|
zzbqx~IozFx-)MDn-R&Ycku@#!V2MAbYX46CwM}iH`h?Qft@y`D8`p~0QG9zM&p|@J
zs8N0;k<=kV&{xNaA?_MU#BL&ICr6OYU~O=-!ypxfKRbwM&Bytx_0gKfKuzLIY3BhR
zR|mz#T#ov?ga1QUB2{TNaGqWzJ{#pyJ65(Fm{kKKl>|Ak^rdYLn36|&z%{HKG`(Xm
zX=9DP)nTlhN2JO{%`YGnh~k$p=06#*^-ZRC#bkP~Myn&1kNCuQ&Y}Rlq3hRbQgvK0
zp~tIH*yyTn2Ak(WjA3Hv495L>J*x|!5#J8lMJZ|6%bC~W)l_Hd>Xv$5SL4-t0r=n6
z$a3l!WdCIwTd==gSskSheT^afr;Mr}Ji)LB`E2EB!2b!t!TBLChb8<~Lq1$)+}H~B
zMGwWmTGjQ$TdqVKWA!*pd$G5AUW?P-$o$sEFzKpiOuC&2XGW#%X|SXBwH#<_cy(K#
z5&lOG)z!Y_zOl9i7_$}QvF(lRb`f-+uFY-?n+y11S~?JTar`slgI>Q~?DCh5k+p4&
zZAK#|{OiCf1PnrBh?`{A#(S4QYiF+zvpqjP&>>>pz}wLwPI&r_lLX+{$P0U9rHc%M
z5Y9gGuNs4(csNbfe4QoJCYxQZrp9@-#jAMNXIT8UHtjQFE+U8xW;{IGYQF$F1c3Nw
z=Y=-DC{n=->>lHu=OZhRuBZ0p>KR^2ZSI_Ov9yx`A|13~%TXd}EvNRTpZ5hmh_6I$
zWdFx}1XiPc+K8|A_rHs*=k<NjHqtOG+|cc}EBF0Rx1(!#*syjevEdox!8mLg%|bX#
z9sE-REg?@-Sk2F{?;XHt$r?)GXpKew{;h23bY35QraiEQC(77(+Ns)q7X#wIhoG4{
zeQ^q-U!$^5VPh!DqXS#S)U>#QHb6C^N9f_&cISf94wIS)2l1a$<%S+?hKwO+`ju%F
zd0M$IDK;#yYw8W=q-a7Jf2rW}$~DH_W(pAk#jcsJP;PvO!x=W?hM2yORA)*)CBFy4
ze9gS10J1JuQZr#!M}1mf*{pwBK~3(;JJn3vk-v1xb<<p|%<8RY)P!$Pv5a_ylTccw
zyo(A(&A*}#<}MtSK@sYOT*IM{OGe`6IIqz(j@Xt>m;Jk#db^P=k(5g+UDs?XoS>Ej
zmO|o?3H@Yl96%2U2br;<8{&KE9#vh<mye^>ugjh7U_A<2lFa}pDw_LB1|?7&;TUhs
zSpwh-{^qZ16HM#m9sY<W6WDo-zuKo5hemIJuXOH&bA&FM!%qF2!i9mh1sJH$@XrAF
zM`nQov(hb>%1%TII`c`+=2*;tKMuoDA`bpYL##hg$A?Be3;qZ%+Oli5z#ko{ko8%6
z_XX%P>Zm<!Fgm9zx12F&BioL1NO{5Xhrlj#tHZRf2P3;EFt?9*p5sDdRoCD3`u~8i
z4{VbGx10}f%RplU+%gEICzYZtlo;ZWu?5_MO%k{T9s%259(1KRq<X*<c+ySMYuCyj
zgfB4!mKgzijDj7YHL)DZf;|QddkoN$-T-_2B=NMDH|+75#^4%wC9n$mMd1s>9zFiQ
zmbN}m!W|{B?r-WyV>F|)7jg_F;~hPr1#Cb-CrD>AByivlX>pnoQ>+6Nw01lc+Ry8S
zKUH8`+-N-8@dgDp+vm0`|5vtnKo+*Q_<Pk<`6=+V+5Ihw+3qwlcdH&ht&CB{wkfby
zTZ6kdu)kfoS5$b1#RQXfDQ4KCjY6K~lP_s!PEa<Ha@Nker<fayA^Uh{FYeYp(*^2S
zyJt|Z?%ady(KnS7?mlk&dPxVS&1|K+Zap*qtX;kxg@JV>(qHY{32edRUI!72&gME%
zN3l2!{?8;<a{-xmUp)qyhf^B==77vYt05%hjLcKM{H5Gs)PeK)Z$#iaGLOv))sVBJ
z7Csp4um!XCwJb*woS=Ol!?giNhLnn*WPsm}#_&^`f!j69HUnuruEUa|DL%`D_*pVn
z?pP1RHe5Bfk%aJbs#au!Z!oz%*`gOdqq=)l_XcLwTi3mz%T(f6WPSQpyGQ+cT-PIs
zZ+~829r*gb+CuYa2Cmx?n`{Qew!g>+{Ld+9BjzvYQHwjgRLAWy4a(Q9)13rAtim>U
zE!1XSaDI%w)OJ#TCb6B3(MDVHvnT^Jot{=kH=nsKuCF880;Sc{)ChFjCn*amgK`5Q
zNUPILY(~x+Qm~P9+QAPXiZ3vvFyT&$@dPFBTnS_iDw?AQpfg**OMD`}Ss8BXG7y7T
z&$ttV8F^_VqaKeZ{2dy=_Q<|ZL+P%d?s#FDLd_iJ&<jWr5apPIf4tjUg89Dp-B@1L
zSc<Y(2Y<NH+lq<X{>vKIebkshpe2zR2p;qg@|Sa^bndH+u8YSv^3~hbIA_l<P|xfC
znxKp6atjF>bb~MjIE@^PK&g0o4P4<f<TmIGFTV^ScfCBHT<v%KHKEhSpu|NSq(b-j
z;j?p(uOn>&m41$|UXuTL{~wFd4$9>K*>G<0w?G8wZcX6N3-=n5gqsIS#Hm{C89N4|
zfoDkZf)1FdxsT2Y`Et{yj+c7Fw0q$Snp^x)%Jet3fUxb*2qR~-vDHxZ7)LVh6$2rU
zT{Y(Y*@r31@r$w}^^wL1W&^R+ZN#==1fbZK#<s?Y>mfz}q%^kClU{lv?**Twu2Bys
z_}?*Nf`F}{VvgMMR)axcl7+Z^b;kyG`aj;~PG7Cr&i8jVc2e^IHRt?ar{1x~4pMhD
zWIQ)WYKHIi*0(WI+v_`|#Z?QROsA^`Nk&_SXv>xjZHaor%-iOz`MZs+2}iTlsu8}Q
zwf^=8v#tMIn05W<7#)wt;;Xw?_pBBe&s~hco<;#bEaLhyE313Gt&P1-L94OH8^e3Y
zXk!nA5V2QcjY5!82vcU_d@m@^Tia$>U#okqe~i?3Z13uyJP&Iv+2sqowd1w4^>km?
zxMz=X*KY0X*U{IV`Zi;}i~kzY-TzBp{K2o|i+lJ|52O3%_Pqk%8?zqnW;I3T_wZNH
z-;7v6d*N=!R`;*ov`O#&p!-{}Id1aydHY$lH#cskM>p}E^haZUUw!Y|K6GCrYx|`u
z+DO|)4UDmN9Qot~S2f)1-GXLgS~`fqRSmaJ$bbCvBA;GRP*PxK5&*M^|IZ1VXj!4(
zfo_Mp=g_ejzhA)a2Na0xT<7r4zoxVwB;ew!Hf9ny^$TE0?ifZWYe4gdD0F>Cc=1Ew
z8v+k{Ug3oC6UrXu3!Bv7-z-dQhr|&F8!>%98alwk46jWfKdr9{ULL+a(E84k)g`=f
z*Eu<_tK8~U8yzVsq7;Wc?w%95?9p4|lq%!;V5`3O!ktuBb2uz3eq~Vj`j%F?j)Q$|
z9Il7jX1HlkpG-6Q&Hm2g<=fqM;s3Bk*)X^TMiTu<E6tb^B8g5Q3NrnomBZ_iwuHcD
zCe)$8-oE3Ay9EBFs=K7XZ2j$=V2w5c<RvxV4iHADgKz>~!dDe*%ixSh@Qtd)3?yDr
zt_&lcaV9;7s8YZ_iTL|865^&ha>oQOI=T9z1kD8E_L{$2WqwhKpHuK91wX2ncN2`?
z;&=-mQNsVF;KvjQ#k<vq(`_)imf(*o%~la26gPo`OI$({N@?txn&uWuS~o4V=CCV?
zx!-aPod+xHD1;1dgG>5|Q>J-L%`+bVl)Ckz0_~7ydc|ocICUa%v6W=EA>w7XRk7T=
z*=}j}u5`AnnY|F)pIdYO9qo{GLSK7af!3(Ipj(^|*(SHxb+zE0Hr*9si^tZDEt#L7
z?C*$H`2p(SmIE;>g5QC!3?xOlK)Vd5kQ4R6Ef4Sfkb?qe424_?38w5@ksjUHc_Wt)
zts|!eqewba#_GTwLNx#8&KpB1mEFgo>o^i9*c@_paAWtIL%f4(k?s}I^Nk^BCVkOe
zx14`-_l-VEq)-_OZWY;}iO^o!UDQ1tGb;?uN$tl80L8dRl#z;5C<Q=QMUy7{JvdP`
z(<F2RqMwAx3J{N3xr{3(zC9(f{z2SR2Kk}Z0Ltijc^Cn8${0yklHf7L|3>pDJAgv{
zgW^b_Ob-{FFLEWh?wIARAsx-;hdGkrML}8%pNPY4!Mh-Qb-P>8u2#%VtlWHm;SqU$
z4YXr`N#+68JSU=!;w~!Ldwp-~0s^Oe*;$?;I)_WapJc*KkbS-IT;P>@d8zP}F>IfD
zw$SomeE9HTv>~(H(dU@C&4h?zLkHL5&1|b2oHO~ulXu|kj9)&xjBd_FwC8?G-!Zsh
zoI*&fz(f4cFby5<W<RS(6V@A`FnRuWD($EB#?L7DH+uY41(L`3Y$$koVFSX;7XP72
z{;Yz}D0ob}p$#o<t$&t`CjgIj)3LjvMHh)VzwraM+QgR2$4zSV)x?clii7_Kf7GBy
zyc&1k{L-bdr3UgD<Gj5wPXxD|5xM0w<0!HE$Lg#;+>e>ImQXit@VG`)Tq%44r|Mj9
zgI<bODf(KhC*+ho;j~OLOlf(o^;}hG&zNLDU^}OuIJlGJlM0+n9Me9?w{6D8qs^4_
z4QHxGCKAW+()hZjcWEPT9HLWt$AI~h8p5_@P~M!+9rw%tOU(QWu*chC@$RAK=HkVY
zKRfjz!(fWej`1>tJ0(VR!)ekL=1R%nri_URYvD=%=a?}*BNsR)`ETM*1>?z+Cnru$
zT#KDN>HoZ@TZeWfo}56!%0JSLA0~2OV&YogDM|;mngEpte#U!FPMn&sEWPJ0*&#(`
z$9vjmuT%bhjfHBQICb*msZ-;(*}Xgd=k)DgBxt5f+~P2e7n`5b`2B&>Wfb6;?f*)T
zf2QEKl=el%<TTkKzAr0g_Jy|!SC|7q^P*+iuZE5deSWiY|Fa6eqM)JRyOm~24S%HA
ze^GEj8NZ}Jl3#aYkD*X;5FBDOtSiuO_wV@cRqsBp$M!V-p;~I&=9kI1OvSZNkd%o4
z%o8DsaR2-MI&(Y*9~`7G5R3M`naOg%-NLmx*-URHo$1RBr0!J7fg7nuQ80KQ^$xb#
zt(jz|Czs32W)iu+*nZ31<Lb-5nc5wGA^+9XJst7mtrCMMF!YB{nX{>FLs4HXou-jp
zKRO=krzS1ex4V?bE}OB-W4J&jkxhGx#$E#|+a*0Yujl3Um-JxN)w-moKRe_t*&D;&
z2yf&tUK-_%9&f8R#{axs(Srtl>x!PU82s$vGM#>};O(WvfCzs64|+FwH}ikUuIPyr
zhdo`%6Df{(zIVHKkT<pjSJ52u4!;ZzD~_V+cH~o0%$bgQ$H?Cr<i||gQ=$HLq4@x_
za3*J$^k8;Y91F_T?7AMMZ1ZZtg+1GYynDU->>E3Ri+gt3wLLhb_8#;eqE)-Sy7#d6
z2w&afJ?cG1O2IqfJ<k8V-m<%dr?}6q-Z|&JtGs_EW*6`zi#J)?`5<k=doS1igjet+
zEbl`53Le)MQr+b~7wt+OwKuqwhxD{v`?KQddLHE*Soh9{z4v<`u$=LAISsB)`k>|9
z<`umU(em57N<8hYa*fdoR^nig<1Q2WX09fB(b5lfrGE?84^3J6;jZ+TxK_xs^dl}E
zA0jh0x<@<8CA@F-=Dc|;cg#EERcO)ip!|YYweqlj6OrPbFmgZATyG%!+_R@pk^jIc
z|5w-#Axv$<!|4&XAuQZSC|uRgU+n;JK)RY9I0`%zMOMK@0Sy=dT#NvVg3AJMu~|x@
zHK0p_h6vM0Jt8IPXChgV;V;(|-DL6?Nl<`hsBIk#t|^imP+bXsC;ryZ7?9qvT~jn_
zad%D8)_RU>ih5w?;F=Sj;1=-&p~ZA{P#zXGR}aB5h?M%crl=qMIUtP!^mydJI8qv3
zcTLgo2RWSAs<6Phc*w3P+QKzOqib8`zwIhA7!K`Y!nzjQ7F<&#?{S#Skw(8tJM5C8
zoiHgPrCk=^jfRlqU*F5yd&HhnO?F|?ZpI;AgRwMUSlg=$BrLY?Y6Rw3KS%$u)otjC
zj-A7|9yM(%?LYg&yl4OGH#xsl=zDPe1n;}I^7ht_#*X>@{+Alt*KQW`3)#^Z{=e$Z
z1U{~-zW16%qtR%W#@1p<mgQAmWLr)GWFx_eLxH#@G21{;WQ{CKu_gP?II)$h5^w;g
zZCDCzX-mO|Qs7YtP|8xEl%<cQ>`)35N?F@N*|$CtV!!Y2f9|ZBk)5S|ebPrq=bn4+
zIp>~xmjC&0zwA1{Pc)?R^DHh+t+owSp8HYBbKk0X4tms{+Za68QmhZAMsvxN=SrV+
z8YJY{fw?4o0efb(eRY)i#y5%eF+PF^EKx9<dt@}~zR#Aqi1p(Z()sqsVC!JyZ{vn1
zV%e?PZTYL&*fgUMYPmn%u{gk64)V7;JE{_m+1N)*hhQuX`Te)5oWuU$>GjmO#vfP(
zm1N&<4J^3T^cifK5P6R7OAA1)&L6x+l)t{#_Uw++JAK$hjHW@@DJ#;*3<+V6rB~h^
zg+?Fq`$?}G-$m|Sr!UIGf{b)mg`_ethH3~@=Fw=+zEnL>KIy52kt-{HOzX5F%`O)%
zT9Rr8IEcOR@m%Y%g-a`+ax9$-c=SqJD$34L?3{Lvx{?3hJqnCac&lP>D_mNdkqnZO
z+{?1!Fq1E;U_|8CrvfTdi4sXcUgG_w%6&5dEUPyvZt{|1>cvh`_~Rm5K0%Oc!M&z8
zzW^4C4%6UaPRyl2oR$03Tm$YDv*Xj*Tr*JOQ<wxDIRv0mu0colW0}Lb#=|qm1r~`?
z{6DC`cd9}o<vyuPa{=*b#ctLOJ7#=Fm!DPeIRzh6a87rPmNuS9fzZqKm!g+cehG({
z^?p73k1FL$3O=JtJ38IRWv*R#j)DKvDb8CNPe?1yp2-eQ@8dGJva)D5J~;l5jho7K
zE=uec?{mr=$;^4)b5)AOlir;Qj6O96%NKR2^BZvL-V8pj@h2~-2<JcmZ$^e4?f*||
z>X!*p-Ak%==cEe;2z#1{(5EYShJu?FOejbxc&379DOgaICeFH7O?<tY7*+3hXStJ0
z*jd;aquF=*XXW~uf`3u)bp@vsd_%!E6%+@qKKU&S+>FLq^GfM<6n~{cdm(USrdq4#
zkQ|GNA*PDd4rLn^55vMBO`xe_lt>w|{7OPNcQ1kclxjyt&7m}5O1jb$k@Btp5n#F8
z(aMoZPQ?~D!pp~IfdQ~&k}a9g4ZPxRz5Z9!qd5ilD)^cL>yl!>=s@#zP3UhBSt4vL
z)b$|1#Jq2+nr|uiwu1jqQ1LZ%WcrS3{V|b+t+n3pkpYPEexerqRKd>_{6axRgET#U
zsRsR$n+qeg8g${<Ox}O0mj9#RR|<ZuprY@!ikwrW=ZGvZYA^JC=?YWgAGM4#QI#K-
zxk>;K&}EM(J`%fSM4pJDd}Ks`4H_tI5h$N45dB5R8;_2aU?lA^+Ce^zBbg@|udIN#
z#LjD=T;g>&KrZnK>e!LIkb;k@$HIz52vTha@SQ93rukV@uHVBs-d1m8I~jXfD`g1-
zszaArW4&i<d9h{nFqe78(gV6>3-OfhYmuq^LORRGm`<4ul;yNev`*v!keshqG)c}j
zqm&FIQ}eZwJ74uX7I*8i&qzDe0XBOx{X`(zIr6}ubYwfkxX^W6cRAij#PQC?^$PZO
z0me=7?J7pyP2?QVJ2CP==CU@|+{jDEUgbVe)^e>m8ic#~%iKXw(Q?BcK{_n*8f9W3
z+z5^248MyJ%)$hw&LM+Am0Sch!mr1t$_~FCI};AJLY;8;<ugj=1((+y?Pnl{Fu!8>
z33<ebT|>z!45kz^m??wrl&$S8U_JNg8>_1pmhNiKsNGmn`@B_T9En@kR@ne8>sT&=
zze`6H43It75D*{UkZQ4S5jc%1d<#4z1fI^lVP(?51M8bTZiy@Nf#te>Y*-&q%C7DM
z3XWO>(5V&5B5C2`T5H&{<=X2{MkzS|BZ-bau0E_+ujv?5kWkQ~pkjDQ<Z}@Zcx^-$
z#%i_VAplKz?W(RrL8pSGf{NPzjb6P<wXfjj!j4+Co16F|qGAC+v1~|Fg6Un`o3B)5
z-3odY3@E6mTb#b%P~EGEEMND7!Zz}TRM{E@!wS|bsHj^E#;2<8O+;#w4CjRo@<vqn
zW(8XnY*SECm-v`BsjlrrmT=Jv%iRTm%Xm9f!A=EZ3NBVqQJ+|*x2V2Li7Zi=E!3xP
zT7XG0j&e<i6YnxrbGd@YDY#ZaMXln|-mY5r5?KOTSXOH(1OVf&@FPAkCAc+FgO?Qx
zFCS-wtiqbEt<yvShK4c=1BJt`jJY{fBTA4;uDjDP8^sh=`s^5jgV$6y-e|ed<;mPk
zE(8kb<zT6oxg38<5_dKJDWU^E1aK;V0tl$2ouho_HuFqW22hkDb>Y^aJ@u|VHFCB|
z;59D22mJ{M;M3%k*h`QD(i3X&@x~RD*%riQ|JNc!7yL3eq#Me^19*%(-HGr-2Ei(S
z`CMW|S^nkAly(U3M7hhCDxC{j>li_;X}|zW%O~#M6{DW=R=f8&5G}9g0`sGcjKLZG
zsQUCb5}%NE#czN>g|+;_sUh2}Bm+c<jqq=N{*W2v&j6i)>Ah|!R~kbNFEf)pfgS!(
zVB5ZRC<~&|+>jujhi;gfIJE8B8*jMo>b;jHhJv4i+3b<brQ5e-ba?RC{!u*mZ$CaW
zhYO_n?IrmWh5V0`8;c?6Ei4kPYUk0Sq6I=eIWb48vNMwy^G|#DwlI$zC2boW+Jk3-
zspF=8v2`d=t*{J*+MI#@<ovNiX#@E?IB~09H1NSI)L!AM$YGy)!J&O;Mu+x+@H>Uy
zf++x3riL&`J2ZcS7OFe*sF{VXEYyVuo1jPOarI}h*;V_t=mj8`Kn!|p%2*HImCMx}
z-mjH#$gG*`7MkqRfx#@S9RfpEFSOsj8D=$jJ09A9V#vL$z0uy0u3Y%gQST}h*5FRo
z8E4*Xhg{PtF-e{X3U3;M!MjC)m=U>pWP$UX+$WEvC%nTX=fbc;f;JIrbghH($;HJr
z!CvpM{4Jt=fmG}z6yXf?b}M#`f@>ArO&}cV<GIA4qo^uaDY?Y76e=uUu~_0zFt{;J
zCG4h_=^oZTXCs}^!4`KS9BHwP=1<tS(h6t}Z3u6`z3L`AO7Dp5kV^NrYaa-!RWlO(
z?p3t+yLS+HuOgTc3B81d5gnt~%zTwV&X&Xp7VKRRLk{Z*(sZ~_pklfrhh3q|nyxFl
zT!&kY+@%j*WIm>4f&~Tr0ZR>HO4Txl1ULM!o=2V>v?h8>eP+U)se;xjW!9zm;c0AG
z@c`<XueKe$sQCr<Zf9hSL~=35ILpN*56)v}Y-0{0*M`9ySv+}b&o%C!fOn?BnMwd$
zZL#r#CGp!*v6;D<ENSk4pn78a=jGIyYg0W~PGfLtE;+uefgm5K@-8EvSXc2Y0X_CF
zm1w|)jkThTNvYlEe>M^+Y^-CPV>oQMO%u~PBw~@f=`XgHW=lt*KEsjYB%UZp!yH{M
z9xaZ{0}YY2LWALug?HxwYbVC@aq_m37uS-O7ny`SN?V+TT<-HM-hof(#xwY%NH@t_
zHYegv6mS#BQ)GXc%r>2FMuyjn#Wb5C_&fNEK7-e3M#T!a(Ku-I4v`03h~Rz){Arvo
z$}VNDiyWQgSX7RZ`#ha)w;bVgJGPlkr`xHc>oqz|b!U;AVDlp7+9YbYW<^82@Vys%
z$SM5-*Af&&Z>jVCGK->Zkb=b~ZWF_PPr4UdndDrHly8H3eQfVNx{E3&_fXkg!C#})
zY|n&dWfiKo>+h|5Q5_1pF(ffbd~3Jk^&^R#NvgX8=sT!CiY*wl7JQrD!cwPdZ>_HE
zEl8xOdZW-^LnZx{bp1sh^+~?Rl;Z+=EZZZw7jMyve~^_)<Orl711n8A5T;jA_AK=3
z@RD9#t9POWPcP}$b<6bY`tp8t{gd8cWxSrNrC&B^B_<|8>jbuzZ<qBYcD^;g(8lw9
zp}9+IeB1Lr`EXiW7%`g|F_M#MG>y=^t8mEKihZxcdg?tHWrnqy#cyj^q%i5@{V3{&
zJvZ{}W|gy_k-1qqHx}D+MxNfc>dWZ~*T+E}+w6OVGh1lE0BNXqpB_9@M|s=zwAB!Q
zNjP&USF7Fs(?fv-aMfvKfvyd+tI`P+r8k+h(ppe*Emj53dN(6i99A7FSq{J5v*Z2t
z_N1pr)4YGaM3m3Thoq|^y&Xwh*t^5;Pj3$}zxw@wLVUpQHx#Y4%zx4gao*?jI=`P(
zBw|R*kcOS!z`BO=GU)@P4_P`=G)rHT-oB{zZKSUEW;gM+_t?%-cX~v!RC*p2XMZgU
z%sXC*`f_%&EXI+;*($(@4SQem8&7Y+#%>GqHiV3ANsi3lSENOG>bpF#JztI?nl)DB
zLIFDS3v0`2Y}4K7-r}z@SzzQP5s$fkdaJG%H!ym)`POs(hG{45jgf=(h?Wi86G?W-
zoeKxADWUt0%p945w1kd?xifQM1}9e!o0QH|-nrQ9ktra8_n@t1`k+Nz31nLtBv}V1
zJltU6_f9*oL5JlU9IoHad}J)w1Qjl~p#lYTiqSW*53;X^j~&TH1rM8x3Y!)InNiE)
zmG1X+AoX6TVx?bYezMux5s3-3_!Q|B5S`$7k&Ex2Q4gR>_?r+=s5P79>!o_C4Z8k>
zX|zsRskyG{iDR4!Cgw^Lo17B}D;j-tH5kB{psHPz2?ZiYVwqf{m8bThev7?%22Q_d
z`=DL+Z1K{2nVPQxyx%95OQ<_tvGiyIK3NWN7|OM0q~M+&-+u_I#mbgbaJ{T9#&Rve
zm9=lo`z<x7z2e2ja>*)hHkNBHc{TD>Y4eyl>Buz|+csw6R*<s;&_yH4bZD^tO?A!9
zL(UcsOzvwviqEyzd#d_>)HzNR7uXXHnKz)AiOX)%QNW~;&sFRy3v_o-!J_)8!O?gb
zZ(dhTB^4nXotxpvblCfCmHISw>(dpi(Zj<8xyHGv+s56|A|7NO_wH7nTXj2}IpD2U
z<^s{^)zgGrWMb}w6Z9>p<P&<}4#mv$(-hpqpK};=A5sB`;gJd|-_5!-d8etwJ(IMN
zjgB;Zo-(afP&9COww@_kIm~f4)tE&&#{7k*@=kfCvx?dn;EeTLwdEv1svgDh>8y9B
z5}vQ%1q8Wz$wLpif46xPRFP{sc0?0vToCfPdK;n+=<h{(z~=S%5i5@JPpc$T5Z|J=
z`GK1JL$$0caBDXXv>zf~WYOa%a?u4f;by(aJ_S!vuv-tBdU*vq>LlfSOLVkU{}OSb
z$6?ZkBDUm=I!IyL8csy9hb29HHXMmQ;0*ud7^szF6-abw>S3g2A@=I`BeE=9#qJ+$
z5I$WpitMqZj0DDm33Dzal)A7{?G6F2xnbq9<+Zl8A`w*$)M%J_36%dt{SPAjL?L|1
z-PTAma7ai-xg!~<doKR#O<?WGYot6>qPA|%wPMyr_udYHFG1Pn0*Mkp-s38HE{QUF
z4^JSsm86zqr`~R1&2nsEJF%3V9(8ij>5;En_dgbG>2tKup)Flm>+P0;N>x+Om}Gnv
zei)KwiP**eB8bfLFiZcamK@Vy{*IoCD|V}b1qC*&k0_?2qM47pjRzhot`=gqWpK4B
zZHEho%cWiM|KAsHfxYa$`1Livc%Fau--@%<sfJzX0FdtU4>e+*Lu84wmEs3G-<y<4
zZrT}Il<8*d9TdDA-jZXMJWoSXL*MD5AJnA*qMcMsyY_`lj60QCZk(&ZDm^yeFAFuN
zsM{wl2wus1zN&eF0-IA=#cWP#qa9IQ((SIT(y_FzNv&}CAN8LEx=-ZK#NSZZ2%|Y2
zpkIl<HuF5|OBgE$1mTnTo%La9J1hG;)r5evr0qLbKQot}I_|W4Y=^Xs&Zymk+^^s#
znjZX<c?FYE%XMNjw6M0;C@%#uQj7CHlK+cqWE!y2f`N+$mxwb$18&vf6Jc9{?@-0%
zcwN`x-YquDTME9qLinO;NJku~J=PjHUqo;nL+7;7PxJtWgKu8^kD&L6XoHh#A}bG`
z!Vhi&)fhOBS>ob$EH>F4%Ek1yL5RxaY2>6-nx|2D$kE~}EaW36Z7j<PTG{DV(?>Ty
zV17b6kK}8{FDbPN@<cT+Hq(N(S<~Y5?hywPuPk@!3`Kf}cQ=Nnoeq)-=MqA5Hh3oP
zCB1vHE5L2${Ki$c54LN*{>2tjdJrSE&<3<>Pxl#26V4qCM{bq(_mtF+?Nc+{sn#GK
zC*Bss8;Q5iVRlZgP)RPl0v2-7@5)6@1C%7g?KbpI@g+#VgLC9Zr*UB!z(B>MMJz!|
z;G;`t96KA0gp9qpWYS8st=%jmGeq+gh0AxUVsP}9dSvH$O$Sav$H|zIYsM)cBD{AM
zCY;SR=|djWTzd=SYl$C>!tw>VecObWuD0Jge=md9T4#jWsR?1KdZOgL6I_B2JwJz!
zzbU~-JJx?f$J24#pdUF=X`;DZkC~d^12nJ1q0^^Svc%Nh`*h>yx|I4%E|hWTS7z7w
zbS2xF!u0h_vqYvRAZr<`xP1QR8ZlkNq*l{=K~LUJkPFS>ZhahEBK<?e<h@00+N4_~
zKxdnyrW7uoFh?GrOvQu9SRuB{tqIoO*3~kLd8*a~T|pf}&-R4}`QL7s?S1_3;m?rT
zpfBZ)RkU%R6E?7h=uR^Z0+-;pSAM3lb&z)3KBE8LUG&{@2wUFosB?c;!S5+}k%E^h
zC~(IO0_wRMGA}26i8HoBTdLtL6~_;U8ap2^3s2?!L7`@={ttEeM}@S)06wMG+|K(g
zTvcn0UA{FHNL1b{3U&OkVpfq1n(k0Ve@bK-2~YmbN|33_<oGLSXMS(byp^wlt3B*F
zahh{sFjM$zz5tl1`U{4c0`0m2W~v;I@_!o46yQ?-tuRyZhX6Cx0?bs)1;R|VE`gb9
zH5-7o5}2trOEDNH$!-1%V5STZiaoajZgq#jOm*O-q@x68s>ALU(kx#&%v8w}0nAi~
z+WoL$raF;=bykI$!eghwOm!hi5WXq@v3%|-z)UHYhnebP_r=$lFKpA~92oAW!s(&Q
z>#D#_C8;kNz)gM0ug8MqOUQy8+>|)4YaU6sDesk9VYX=Fu{F+!?>|#oamCdp{TeMj
zuhP=<s9I}W3GPI3p)Fb3Ud_`>=n0B-E~d^@X~%1pw&N{o$Db?x1_gHPc%5Q9C}>IP
zQqi=t_g^T>>lN5*y+N@z7A^}b`TJG+TSzY}Nn4keOe^UgJyk5}t-5^M(vm);lHOif
z64xepwUYiyPg&2uL$P-%s6rlnOeMWrH{U}rTGlcW%t+13`y1tXuY&g}c)x-VC@BAu
zpH+DuRQiV$d|1Ip2$mEqQtw9oy#K8%A5~x_eXI~GFZj#Wd5okbC8-0giJvHc)CSu>
z6w(Tx;a-*WDbkmeB&oHP^y$J=pHa+8`fMSsP}0{_(ifC2W(pj2bKhc2PVb8(kF0Ps
zSymH3<dt)T@-ABJ{HFh*(u$fLPgGG)Q1B!Lk5*t-F83<-RR#a7;A;wQSKuqSOr=<F
z{fjQGd0!{?J2Wrz0$TR}5=kZ=NtzV~oV}W7#FDvP#_7=*(V1~Kf8ICrZZ@E|>V3ay
zZ>}4~aq=zFeyVTsZ6dXGJYnWn1&%4>QfNwQf?M6+QDy(C;JXUGr@*Q$*8Lx({Ze&*
zpU4u-_#)wG2q1#`D03>dE~rgoIU~k5j{dO?KvFFj!e%>9qB()pbJ*}JI23WNM}3d8
zo%AHa<c@lmaW;*Z20<9zfy)X{qOxsIf_j~fCGkkAFu>SwB{k<uLiHOz&7A8ZXYfLb
z>bKMg>%AwW8$bk!<405$gi<lw2O>z<oIFRz@GKqnp5?c2-EP-ST+hIf;3UtvFow*y
z3xI;S0Ti^GbiNUt0)Q#gEq6ywC(w>)HSGxYUsNGGQI8Z3#|qgV6!1cv@6$I>elw@=
z6|-nIAc@5q)pQM1Z?n|*W;?pDd!%o=uunBNXYwR(UP!zA*y-dM0k@fIB|CmMUP)Kr
zF%*zWp6RAX^gi9lN%QUKpdCFW?I0cM2mjf#rKxp*LfSPosf3cZ!>ej5j>JMZLBk1|
zsb2Qt<DNGoC0|RDT6%*w)$^--@;pjq<#Vi1I|@0>b7}tDFh+XwU#IZJR|a2PvU<6q
zZ1nrezOa-{qkLhLb@Rm={Z;91#!`Qwgr1-TN!>$8Y~x6@HQFSrj{1XK193qKa<$cO
z*Q>1rWY)aD+_F`E-8FV+HRV*^kyN|UAFA@q+CmTX(gSPCzl4pHKGN4#Nnb_!a7ns7
z>vA$q?2di!d=K|8?ctFD$G4bdQkMN_mfAk?RINm|COxgN20x<^vxPXtWek6T4&0@t
z#(f@Vq?EQP0eN}YO=!wD!J!qHy5gK#`dZ$_;dlHwu}a_=$8Y_C);n_%^K9ilqUVkh
zJnU9oQE1gAk3@01^TjR19p26H9Sa@#*toG|;C3t|U6K<T7g8I8)T!*qKnWO*89y4U
zM_XzMEY7%k!9=7pilx;v8O8Q9JW7~8y}Dr*l{f3gK3!U0UaXkOiLcgU=8WeerF}}l
zuM~Vz!FKNACCs=x8<q2uYKVvkfKrPT$5^WLYL}xo-l;18TEROMyjj896^v3)DrQ}<
z(*Sr<jjlfsE_**z1;<p?j})6$-WD^f3GjKwy<2tn2`cHw%3^Hd+lpn5<vPpqTiLm`
zvb5tnf(Go$tt`8t%z%1M%%@f@NuMfw##B$)O&rtZKcxV^GICXy>avJ?vBKT<;BFzm
z+sbX;+LRSLEy_^J9HILqnQVt&)oTBdafVCbgh0=eMASfq5|MWKHpT&$oEt3#7m9Mf
z)1;*Bz+tGQc6ek0K<^2DLHY~VHV?@!5e|x|RO3=ymx9J6_Gh9V<OKqYzr|+!i90AE
z$p@OXv*TzpgqC1S;B8x80pPb|pfG4UJdFlTBw!#s#TqnG!k~#TA^`VGo(e#e$v*@9
zkb@?QLRAP;tgQr^=sIYEwgNO!y?2A(Rsv1bAdM@7CTc@_i6<9@G}A6OsiXrAtri5(
zL}+b7H^hJ<>f|(>^nq<QJeb>q>o&W#dx9qNexgD7GlDz`USx|J^=-uX`!&0Mp`5=|
z@Sh6)kAhzj<QlVidv)(VC0?o#f4^czZ%Yf?`?W4F*Nva(@(RT&=t5}(JXM29N60vA
z&nT?X3&sShseoIF4J0Mv?+&REQahS#C&>ToBBIf<o4<^3#-IX2NE>4IwV^W^DJ}9j
zZzBcmS$WOE2&O}DP>Wq@^tt#$AJ!x^=xjk-UW!jGG)F~C`pG168GN(hv4kBJz0-6Q
z+68UTV*bo9B>*K4#no-8ppH<=O_Z`SC<UF)l2W?y^`TOr&5HRm*W!z!hreF_`oMKb
z$|J=~LN_%pt|FzM6yzcpDV~)JSe1YS+`7zfqn${b6!%_EjjMthSLJJ593*uhNF6Ao
zt}Z+m_Fg3(4AQ5;Lb+k@4U9ddt@a1e)El(7U*oS{9LABWdSWd;F^0G=oxEZGuSYML
zUrbgk^oRYma^&i-3G~o6Kto3S4Nhxr*_<2wUUF_U9logF1K+2W+559%+n@>Ogvo?X
za9CdM&@I+gZ*T=I3Lsc7K19HLLF$HVNis6zI-6|f^;F(LK2?^}-^3cyz#1~b8p2qj
zzP&sN0-(+;nO(s;g8vtP#P5-3o6>qJODkDpMhxlYDgFqn%*f(qzUYW+pMBSs#VzPk
zMko^vMksOGtv<lR$g}bz<LD85r$nSxiMN(%-ntoi8*3T*s*UsmUIZ&rw&lmsP0vA_
zH9vxGD;+_+(V4<nkoNGq9@;44$qy|ZN2Gr0yd!CA?eS30iAEZ@gaiDw>EsZpyS&cd
zcAjV2Tt0g5rTqP_zt!J1P`9|%-^y3bf6K79k+@rlL-}u^vEq+_*brG9os+xb#xy5B
zG(~s#Bcw#K?RL_uC*^W~^mK>l{WCHFmLuIKlD`vq=BTatql+D^`C4hD&ztl|ZOsqy
zRAE$t*%%?uj>VmeV~mclcfjA_@ASv~&2A4MWj;t5wG_LnU0|eb;Tegc;OFXTP;{ov
z`f2D!**eKm3GxY%-Si2w<C^dQ8}%l<6Wh%8$6U4!WsV6w6|qNvWn3sP5NVpmpvg6#
ztG&SJ<DV*afYsL9qYXi%ts}FgKj<7nJ(?I>(^blJy{M==6nnmcn-v&YxK8DmfX_(J
zck9N_Ny{}Jo6BItGldf15j|`)<;!$iw4x(QAEVg)3XZFoU8-<brHJqX_d@S|)WU{1
zg9@nnM($jQ&SU6y2y`J?K?O<9#53b|Oxq%_QM0ZkNTF6blXlKt>t(ufJV?QBVg|D(
zQcmO&rr_n=h<uKArvj~2acnZV2%b~{mo_jnOHpaL2nD3-=MO;fxql~8Ev^^`QFA=i
z<Sv1-B}>`86Q%b~q}p9Zd%N3a(%FNlj*^svQ!~>Cv#Fqs%9L`T+=}KpN@#I8F$T=7
zKr$dVbuQ{>Qi)>oQ_aYYv);`9V}RrYP=n{EC~_++_I|*7=5C$H92b%CPCB}|-jNxi
zRHH~%1u)QhHX1FD>q=(n(5X0vZ8o1Nct$bQ4LqjA<Iu7R`|C5kz`pPfljdFmZHwd<
z`c`^?;X6Q_j-ca+^cFX@l+?=IR%*<bbWwC`DEfo=3Lp=0B<uXLI483`p>>c$N-_Au
znw}Y|zwV?INAf>9W3}*SBu*d0u__RvdEjT6!C=eCU@6B42(b~e7$YNy;wBhe4c^)U
z7$R<G{Z2@sE{M`3CO`FD!OYTC0EXD%#{fgbxB$`yY)+VP*J-Pr#g!<eOA{;;h#std
z@q<tThS*jU46%p$kFrVh;D`;vDgZ<5HZa5$WnhSn-g9f@TWJsuRm)&e1Prms>A|YE
z>>TONip!G2wgh+&Xeqf~kDXP_);)T=5(m*c$K8?M^Q<<H<H<_$JFMj_qM$&CyPBs}
zi!CU&&PA2Q2=Co0#mMs4DvQpw-Zcs;=<sVvyI)geGuis0K+M=Lj()146xOd}a2rZt
zO$R@DAeu>Wg7QBDM5NiYo4?Fvs;oK?Im;MFvc^C(jhYUmf}y+Hpnw~^jU@wehrf~m
z*-Z&Oi+v2pl?+I`3I^ng!hl>A49F)I2BcCnAXnLd#6n3sdD+11syQ(GX~zkMbUy>L
zr!X*~q2Z?WRt(I$YUJy)fmyW-H!#~C>cISf#>3w!Fq7O5Dt3W`6=dS#{`oOID2<Y8
zD4kwNH%#~DYQC@Asjt>pyG`SL;E^2f)n4e~&H=pM2d>P6En~2WQM#2^>8mf!2Yi;?
zAH`#MDQ@RQY$b&mvCNq?*yY@YG%3~W{7*j-!KBfHb(wa$%kKza(RA1k398e)&y>S>
zMhwQYD}eEg7>p;-n0T8(9}F1JE?_(bs+ouJl*=Lq<C)|z+qRhGSOBo4mHf!Tcsiv$
z!FYC>VPvNtErIdu3}8Gv%V0b$6=Tq5KT&}3OyF)Dz*|tGS8`~*9pg~Jcy<A_CJdG$
zjHg5bwO~AZI0)roJSAcfj3<C{OJF>Oa)}6m96cP}XoUbkO>M6QA&gu_B{AEJTTgx4
zF@`c<v}0TyrlgIyzhv3q5ERFZuGNoYNXPEPAvZ}|I|3-S9~18kfINSJ_@-AFEs*{7
z-u>*f-RAPLf#XhuU6<zqAWs+tS{5!0$nz0{$J)oxckozGsHJg$UDAUD-dr9v>qb>+
zs99G(>?)Utt|cbbM&z|>jWUshZGg4Fw4SQQ7?{>WZcNc<q~U5jUJKz96?Cczv&8m!
zU5f5hyNp&yayioP*2w}F+bzGZQzk>JGiIE@({=0V*Xl+!y1-U?L%J)%*u6=aZJiVv
zgF}m|24mN+k_Qx6FBQSL<gxp)^vxrPGT96jzAVb*wg)vF%9hm)ye~wJqc;Y{+iMM$
zRpZctHVM{&TSN_}Dvvf9VWJA{Sj(i@i-Uh46h<wja&1$OUUSUbmBlB1R?cmZJLBy)
zLJnkrU`v*yLZKL@q_+?1LR!q%IKCoWfjqWkq8NuywNLk1nBnoU?B$Sh+7h1G&;JG_
z$46KO(Q{%M#O9lL6ANh+55DmH<{-_h-ODzcZgw5ZCkvthT_J<T4julG@hQL{Htp+n
zzZ1Eyq@o?(+X*{CfoQ_}-K*SE2M3~;1r;8sVOL5z3K!E~r3F<sw!m2lyRy%g$&bVF
zmcz6Uk<04!TOj{GboA+P*ScA3`&csj=h}i?pXY6K|MP|Wy}FM+n#%Zw-;GCP<62P~
z@LAN_4~I@NW65q??rvbIbsNw991dFHW!|5KujwrF-4|eKMGnq#jgMo>77tWb&u!9&
zFD%zReydxq;c5-irg*wlFkw8qmL(goFI%#;OmAkHhPTK9-pth)homIe@NtelJ=?G4
zTsjAEUt(Cd3qTpa2(HKBr^e;gm#_P`i>vw4L;S4)kJO&Bba6?CPWNO7*fNIQvp(Jg
z(y^29^r&%7$vu27l-`3^bz^oA575{}1D&PV8e;HTi4Q3iUMv5(B2AKguGZK+@^NpC
zT@l(n@-RAS|KeI&z1CVcvA8ZfJiC^^b=YNhq&u@4{M8~p(p}y150Ol-kW0Jj_tPuM
zjTLqe+JQWQRss4C>$4ku{2}}J8c+|`D81)ArE`Co(%0lmr-xp{-*v#Ph2iW68>}zw
zzTOhFZ0|ZPk+y70HsN3^{zNfb#b2+xZ%|NJ&jI?gL!zzF;t9D`UF&-;e91+*@UF33
z^uW}_crG%2$wm7BGKo(epTPt0{4JPg;>LFTzUb(YxoK|i8XMVk9=udZv`Qe6jY$6|
zrR`OtSnDA16-Xi6anj84_UF3$5#9Kz1-dM1&Ko<;_-r~5y4T4n-^~j4DENYcFDiJe
zvWPvG3**NpE;p-b{L*=ANY6!Q=V#`;-&Ve76667qDz>%XR1pszj#bWgbB%awnwcA?
zLvt{#FfCBm9+}7<bSAr1wz*uwoEJI%*(o(`l{&{h)hV6WKdMXZdsq=4J3MZ=a}DmH
znOw6)F1g5_53Y8NxvP5RpMeQDpT*H97iKg7>WR#`)uxK`_6@xFZQ4jr(~N4Mu!CYX
z1%R+Hix7qi6n15J81_*f92UpDTmf!@#0&^$7)3G{bsH*>v7i%~p)D>y`$?A}ZTJoe
zYMYHY)y@fCOf2pg3k%azm{99NTgj;qXTJBvV02-=R0;H4(AO%Qm=4X)ADK+@F8QBb
zM0h{9^?o1v%Wn_1UMV99=s6i|y(<D3X9x5gK^Ud2cO_eIgK!wxdb?R5djh0dG2H7p
z6F7yZtKIgI*=ZY83(8giavB4^xfo|ty*k)bwPgTv#4^X0r#OENx;En}ZD=Jah|F*n
z%l6iYV)04@Dxov{ngJBFz9A8^9G`AMSxH-9-0!cF6AZ?(oUY8iDmk?YOI(mIvrhm4
zc-D)sz4Y_;YPD@zWo;g`ng^?tH;C)`((+u+fhsvEwJfJxZwNL@TeRG5v>}G<W6NEQ
zLemhnw{txJ#$t@^ehs3l4(H4Q;sZ~p!S|EWQYRw5L4R06YQ%kO`I<tUT#@a{CbKIb
zn+9!TtnX&gXS>AwOo*D0Pwp<bnK(9D=dWX<{0ZVgkphU9GMkB^Qb$RTrbb{IAcfYW
zv>k!X(gZ2A0a9qAkwR7OWjRH%jf;rDd5w`m&G~ODb-Trl{ziRK`a=kw8f|^l<nw>k
zY(InrUn6rcAcqG04Q%QXq4CX3I4CIZP1G=G<dEJw<Zl!SLm7W$gq&h8$=C0<K_YO?
zb|l3y_4!8nd}OY(koUE`o6Ggul3W{<3l%|E@?RnF?C?j7G!VKY#27-r4b(Ek$+C$S
zwA~%&47(DghV|+80)g{6%3h^!V7+;-eMzu2AuzfrZ>7aLgF3#)llA7<=X=>T)WC_I
z?`gA;H#kS@F?&Ef8(!S(Z<eq)-68R>oDp#j2Bh3bp+<kbkxZ>z;j)Ty)@#Gpn`MXn
z-U89|ErhWmnmXxUn|F&_FiF@F5KUWI54Q0)im9JH6%b9^{cWc^vTGp>x4UQkt^CJ@
zmc0q0w!*zGh^CTzF9{#r!Ms@KZ{s^|vls%n#Zh1J^-5*#ZL_Nlx`Jrh=8xJv@^Np}
zt~N+)$3327^%2pugI4bdh^C#{jkBBh8#%q1FPF@2@wXZYw8F@mm8b`b^r?D3-9v5!
zcLCAVYedsnb}M4OZT`-HXsTX%-+4;s{xYSH=1ZrCp2Od|^eX&*ig5Zq->%yjmJ($C
zTvvXN#ESk(2&w*3LaInU*(uiqemeh_P*}C-Do_@Tq%!4`B3<=rHWqho{)Beye^BsA
z1)n0==Y3ky&nozwg3lA|8QFFodddhWGhdWpmLqe_81X7qX0(%0Rz-=UNfV7&G6JfC
zATwI5f()xhe3`{!O=_#0(DJ^ZrhQR?h%E0P6)VzMfdI1_jinBBM2(S350U2bzN8BO
zNx{q1E9Ep+Dsp^B#{07FnmU~n>ryc@p~N&&c9t@#bkdY}E7yAz{EY&WOT1UH3U)*l
z+RXck^5qoVtKh2&{#n7-6qvE*zbIBta^>2~h!ynT?iKwdUBgQHOY8^?FzcJ4kD_6W
zHBA)_bu=g_{UHuSiOsIK<6L`)ijpVmO-)9(4JJ$IHJNgDaxH91<q@TYGY=E&IY%hC
z3_06hFIr4@^ou)_P-~Rj50YYw#$`Hc&O$;9n`*L7@|<qbZFI$?25SCCy8Rl=^kH4*
z?F48_gb&pn4&4<!9KNgWaOAFf+?2!bI3(BSp}TPKeis;CcVU|-A|)MLRd-k9g`l&&
zF!I8vEZh#D4U>-F8P8&OrNxkKy&kr6K3}LLA38FY56AJA4{w!t1Ot$njin(QJIx=W
zL)_oaHljP8xLf*qAT!Mx(j^>~`UBB3VwW~8Ty~wDQlY<)8JZZ%`$ZZ$JdwF&2*0m^
zB;GU|B%c;=rIR%Sqna^i3UW?$WX8K%6K#@S*gJiD@-5eWW8u!nx^tbDwIz&3*+huZ
z__N0mn2Gj)Otk&bDJ<@rg8#|7Q=wB3k{}S(ovPRWlc7^l;t}p}eUeS&rQxHz#~Z;i
zq98YvZzBd2u^N<XJ|u%`oJLCw&)I7xjaw();F}uwYOS}w@SJIF%8u+6=($PjukLYm
z3s=+{8bGoH-m!u{D%H8>w2JTDBUjdZZ|_@bz_%4>`UNKF-WJ7!rndx5clrX}zbfyt
z&HfXb?fsjwNnvE9&AWus+#`jLV;UlOKRU)a(9}b6QGHf-1pc1#+kz#BQttHoH~Op}
zRI{9`hg3rOU4N#UEFLH{#OLhX2juS;$ewCAZr-jksfH6Sl55DTb=Vs>xDoU3s_8ol
zs=c5-Nv>l_SHFliHRJxIPJd#IqBtO>SyThUn?tp+J(iR-DW%u^&oB{5mv-})*-D#W
zZ$dc+Dvqrg`PY`iP<JjvMG<<fo*bNo1DYOcU6OiZhoYp&nM|xP<&ZLz*rATuH->L-
z;?Ra-kX|K5XtoijoRtjHl2i;m9D{TsV30o9sOu)CcacF_h;x>kaMZU<VR(-QbGg`S
zijHfL!8m5=G5(`$=bfbe(D$4WR$<Jd6G9^=1Ypw^2j+&V-^;c_zsvn*yOhT4gn)}r
zP8JcJzU*w!;je%?l<7yOu{XPBlj&yIq;Cn%1uJ}9i9$iUa{-h+rxZ>*l#9A1Y*L*w
zLXNcw?R;^zF$>F*Cs05<V+-F+X+XvLW^o^h3&;U-wq*wc7F)Z~b1>bETC93M-9_$I
zMq4J0QXI;{4s><+tJC2Fh_#K)R=oyCkJhe%`!#F8Eo*5FM?cHE4?!79x+b@VYhc$7
z@ZF@O+y}+m1e@E~w)DeOIEysc3F9eH^`tY3fW!;VAG4e=R&tiX_2eSVNn_o@6X1kl
z+8^XxWvp8iwK<_3mn<&b5U_3+IoYh0851WPv2O2V1EGJo66^N&VBPYxdgSOVv~9?H
zxzVyLzkau43YwxC3q~o)T0zy;qF^01y^*QQQ>aG8jx<!V5UWYL#ZkoM;cwNT;@b9d
zT^d^M3l%f&Tm?;PH17AcDAnW?UI3HOxMWwUj%CSF@8hb-$oH%=i&7p55W1?7s1J<|
zSVEFkbX64!)lrwG{!oqmXEddejz!s*5s|iBC(G$aC&hoYR&2dqb4;%1#4U@9#tTag
zjBBX5L8S(qyjG~mJSVRoIxx@4OQ0IIF0=_6QDINO!=r=;5{@b~da%gFQ|a*KYJyKy
zYpMYWm#MyY(C&8r#HY)=g1}Z_U`ilZ*b(?ch=Hr#;QxW>VpE_k@R)4)R0y}!VQ)Qz
zo*dMg`Q<4i+U^R?CG9>ZZx2XN0ARtvxijwW;;;vegHyN{@2$H_E}6Jyb!ZBlF*%Jh
z_<PbhV-=6x-O3zn*z3KH`L~4Z5UDD1>k$)t|6H_2BLjEgvRVX^)8Z(hiqbQajw&K8
zFiJ~k=>LI8jAq3OgUEXYX~Ld%lJ1)JZnf)J(gm<qT^C%SX#1C{-9+0LUVA~7jc!~8
z%l=L3eLI;K#nCCfbMdHX^5@m0<(=@9%zkyz1(awDSG8hqR{_?&*3m}lK83XJX<iP}
zws^f1JawMJZREL4k(lYVMVWc1R^GobxcpZxlO}7RNomKX%PWZ((X-riY56akE?A_%
z+KaJ~NkKEg^l8>~!Pis=Q>OjDbIQa~Ep=0-rfaxB@#*-EhcRm!RE_C})I^6`wwi}M
zaT02PO`HeSfa2U~DH-X;(E7UO$?F(M4n_3*vxp<q2bKNr6f7#ZN8|saihV;9tZRw#
zi0&6B*acMkV5VBGvC!pbHQ{v7s5RlFP->IzSG9eZ7D_AA8ckBFwkn%<FDH`UrtIur
z*q+L^DNfet4UDl3)!8<+uIMA+RGoRIshexHyV|67*apR>q&st**HN6)##E<`=|H4n
zt_53@<{Af(*ehGji`)6OsLZ9be1>vw>V^0Z;rR(}KyRJAKV|S2r}$hV8_MGAE*-iv
zxGNFJ%Zw8+5-uCSd{$X9Iv*h~CQ_o*)SzKyYhrQVNs)Pw0i4nCKe~+TuHX|m>gGbH
zYlRyTZE^#uOu9F*x^SrHToPBOwLN5>x6-nA*DLUf_U;+cFUDs9mdXz~u1U$un}SzG
zUx*blq{%EVCl9_+#R=V7GP5pp8dnghqT!dD)j!azPM&90*Q~qRTrM;q)!_u&=3&}>
z0e7S*?y2@&jreUT`$S27+##BtpX_qX96#3|?5^RnQDQ)(++SOATvlG0lQd0pom-yB
zUuKj5{ZXK9o+nySaGhqIpb7g*uoPu_i)0J8W!eT!Q!<!{wy^HOm3z@3+-X1CYN#Or
zmZlk~uOeQuu8Mqs&UuMaD=3v?N*mJjvOZd)nDy})o^k4Q6=Q9fESuC@8mG|+Jl}>e
zrvY3`b0Vni2PNp8L#SAa?)BDEOqoZNtrVw@vNiE9;NBX;1{|1d-AT84F&pTxFnn5m
zWmAY)z2R96vLkkutJ1R1qUlDbMmV-(p}H6@xrVr3G=*Vrmye2Z8rU$jwWNq#kM*2V
z0J7vz9`mtJ=BmN28vJ%xBSKWE_pbLljAhiu;k?;<hdi7(Z8kq5;vb}yM)INXi-7mh
zfk3_q;ba6R&%BA}6KPawVdMM}8f-0yNBaWZwMap0%@{QZ;X4q*_mQU)-L8=N?dU=b
zh~1wtZu(&WfCWQa(IFEX;OIDcR|vHp+s~C~j5YJN1H=Z>VqOhoyJ1uGxSqkn&a}mN
z$2Q(k41)WtCf*ytKf+}A*%IIOf*2Ee@fz3)2#sz3`HW(VtLMU?#Un_^bBV-cVOS*k
zHhnO0VhrA)#Whc3zl!pEG7_qri|?$<eD^k5so!VtXDydL*RVfKE4pY!f>!W#@I<<X
zbcE}qCrOupeGhtLb+d!n)hERY$zKh*D{Ck0s~2}$J0;3hi&{lZrbF3b@Z}{cu544|
zPKn+xTG}edDpRipt?Q;0jn+qNd19}z#M*)}rvGRl`udM1VI&fcEUsHzzqo<^Tt_R`
zBR~$AZ<Mz&-6!F>u&WS0ufmrtLVE<u#7Nl?q-@Bi9EebJke{de4=}1Fz~{Sw7V={3
zp6C_5uih4kf>?!q*px*!rU#iDdh1PPU&6ihJs0rSt5rAOV2J;!k+1SFYW(*LsBxH@
zgdDTW>H|IH#RK++>+Ory%}V^gF}==?@Ae-$CY)Y>w=ih19bWRD;mq3dTJc-6IlUph
zk<q`!-@?0XqDI-MJ~_MT^hg%p!>2c&VeI1-I2_CrW-j+1e~PWy=&h0T<~xzrh^o2~
ziiT6kmc9E}cAv$@>)oVaT*2cN%xRtD&mnnvf1%5DEPJWACuoduFbYh3TvWU>8S<Yg
z->Ve7TEQC?{4WJ!$g#6ltl)1_1-1j&Y0}p1H!1rDszMY{k-fO!ujjry=Ng}Cqb_%-
zuqzc@uiz$vQE!v(j40Twz)r<m6x*s`n*uHAC=0$zaoaxjsRgDc-k`1+<;D_<r*O{y
zOJy!@Gsb0jb0N)6z#_o@Tole*h%8)E>x|3xGH!OfnQDaI$QF0HIH|HfpcgcjqBZy#
zYI>8}`aa#T1IQhUy-<NMCQbDHa$P>4!1#*Bk~1l~Nxe;M{(RkSR#B9R%Y6m5BloCY
z`%2o8y;fzhGtd3H>{h;&3be&KdA9B5CbO5GddZgl1?7929vdVW+3Gg;=PBXs3jRug
z)nc3b4n6%2rP+~er!L17>{4)%0^9#DR_ur51H@3{^Ub<8R~mYkRP+QR*MVg0Rs-yN
z+W@^p88@kdQ3a3E<)amdqDZw!X)rr;7;9pbLOmO!kI~&r6}(nYOsl+W6kMyo2HD+;
znO^^8%6GW}8zwe5Y+U>`&pt-fh`vy~UceI0&Mz8D{<^&$XDx6_oK~S+5X&2fIGF%m
zpewi%zNX~Y(js(HYJuQU2E0hYSiq4C4~F~iO8|Lc<)E!d8RumDDi#0cvxDb4(MeRT
z!|*%DkfGz$7~R9>bAxAfhjPGGwCi)7fM%#>9LCT_J=<!}B9X_u7T)LZ`A2ox91=;X
zIkcO<Op=aZeFWVQUj<M}@(af<jwIWzqwX5TrF=?#F{SK@h##qVBA-%xLR3`DyUxNZ
zv>#?00Lf*4#|OEfVS%AV@h2fEHhYukv@qC??m7}&OD(a*djU2EXw%<X=Y(KcQ9nZe
z!{PunCzaC9Sv5A>k*>cR&lZxpzY>1~+S|+1K%qji?$=L;{g^57Se*5_aj$oU2F^Qe
zlqtAUmycC&wE|m%A4hDY(|ech+uZ*Mo$9Gm-Fh&toL^Gfdlc9dH>Xav9=u!m-lO2-
zy88(QI^TQDaAdqHJHQ_QJ!O)lHkCLs;Y}RQjL(5u{$3?C$aA-@cB`)9a%kiHedKt(
z#`&WuQ6TZszi4H=IT#g(HLlSvxRrRgIigt*eV{q!{zva?MvJ31cb)C%z|8b94;%j@
z%u`JMy-b;RM7@YK0e0(!u2Taaui%La_A01o?{AX!ajn<CMWi-QGCM!+P4AZ}*q>4G
zo_%}wniA9=e)jI&yT|Q9?^mBaS(V(N;6??1tF$cyBQckEB;o!Wx~BJj<<ai7*V6aw
zak^F4D`oHAkyi3nq*XrW?V|`X?xjs?i~WqWt6gO&9pyPHiWzA!8HMuX3qjg{BkkTp
zLP=tKWxk4nWW#OeCcV@?J3ZMe?cJw08M)YdFX`UjsaHOz;3?{WcPZ`by2MXq;A7&$
zy8MxX&noTnihV)BO?qNhf$dBGrZjU-zfs%NsDj5RxL(16^4+1p_EOt7e_J>HRgXPi
zfv-Sr#yxXpAy=K=e<(Pv8y`|2S4o~(;9jekN!M(jyFr&?|9Rh6AeMpK%zmiYj}+J*
zsny@p)bKRkypJeW9Eji43)<{?DtBiXzL{T>6tBaLLA~20YK5&oh#w2rOO_RGM;rZY
zyfNMo?~d0eI^)g6TexaXw8t)D_t+F`QoiPMb<c_IW!Jbob{R8pW4u1TIgyATj7Jl#
z;TyuC_H*H3h2)C5TAR<tND1N(#D+@Zk1B~z7UOl#iSLSa6t81XaB)iND#TV4VqE|&
z(R%$hpN$`m4R?k-^~5BdQ1+uuL_g-TYN{@UG%FQ0a+jt1&Uh-4#tdu<{=`)D0Q`z{
z!=3f1`WJGzOU1OndLQQ5NeV9fD9)!go|f4g>kNN7B%H=9`vDYo-MFwM>c#;whtdwQ
z{wKAMq(xDO_X;#u`~Yv8f=-I_#5RG>ey0*_Tx)&JwF#5@SOyKJadQr5CLj=^Vs)fC
zbkn8FMK24Svk<XyVZ+1*hf4c`mR1pJQdAS8XQB6+e92Gqa92C!jgF)42hQ&JUT=hq
z-e!VHZNd2;u?BU$o4?Fi0$xo_TRGbjtT8MLseB2R#qQwl)mxVZQ{E3@7aTzc0f{3U
zB^xNYakp8P8$5aM30byvfgN+Fk;ARZzLK)6Y`8@zi=hjWuc44i&KRzn^64-(C~Kx#
z&1L1QTsi$jfer1B;?r^l>(dHck<*Go?Z%d{T-m7H0gW^V-_Z8gXVV|{HPN5qobE_R
zK2#x|i)TO)<TIu1djt4kxAfWV3`N2g;tYYWD(NR06Snu7(;U&sBMC<Y8@XzvXyH3K
zLiMM5W<pEQ%>=pd<YdWWVM9S{)>kwXbPke)Et*XHAgTWyi64t{!-AwG`HN^%;+INA
zD&CN|&!pqg*x7}iB||0v8X1Xd_kaYbSfp(n8?|AJWzFU5!?fn}YL<53cpaWsVs6fH
zS@xq7ihj)JMb(KMD+sR}Q&F43^~|`2T-TE(PM~Tw_8W(%3vW?I0H|pFwdISKb2Hdl
z@h32DPlg}?nc|EX#;_lnVLXeEiBhE|babz|R~OLIl1hf==MFh{O#1em4<@FkK-D~P
z#qi|(yqBI4#7bsdz#7Ber}dS?#Wes<s3y9l7tv0<D)aki>OD1I>aGj@RoYRQ-cBsf
zP$~TItkFL9>aXD<cvW6hT*z1AS=am}ilPmB9qBNhlOs^0_(>FUk$eE(1+M^qY?$US
zX&O-Uj~Bi}oJk~#ojZB%4cx;{*uREfBs6PzwPhO0uW}RUL*N-1$8TpA9Z@L-F+sz^
z8)5-ygrlJ^iqn^{S3C~N{cA39E6AK_^N3Y;j^Li1BP0zrt}c8K+RGwTcLpw4fFd;#
zPIXLAq5BN1BM_hChbHz<9ZIz+Nl-Y`^WKT^ne<2~RnO(zJhp(*bOvg+xX?;sbB@k*
zQi>hbK0?*=d!T~A3%!aKx<G9<=Ub_S#)doe<JrUZ2O|r#YUCR5?*)jemrL;86KoO3
zaNHDEoTg<`NsYOgEZYY!wEx)5p)_=_4X0f6$f5bH!^XT=bs5EP3yx0q&~-NQiaoqQ
zbq;JN%1dOtKUDRx!?ty#8iO`YXDn&_z#Ph%x6Cboj!NTG4GL!tPh(uCIY7*f;a_UO
zvV}=+^q(5TJ+!6A%-dL$I83hOSLltVJcH<e$^05Eb}37!@o8dJ*CFe)noH2nY06L3
zfM5v<uZ((1qs%L5691m1Uitv-97#FqQ=OHI<LPSh;ON|(_Z8j0M1go0xsD?z>=wp`
z_}H4mIUBmkhjKkGznwxx4*_>Sd2mV}&7k$@(5acZ`6C%IBb-HOE<u-RNFJKm?}?7d
zH6A!NH<_Ks$C#(KdfJk69aikEGolnmkD&{jF}6%@<C0d`U53O|W;Bg%usQ3RN*e*4
ziyulKw)>v;dQXxH=Qvm^geQGePiutNu0Zrjs!dP0VV4=5Iy~V?+akBIG__ER)U=Cr
z>dMoM^pYxQW$?(<q0Hz6?!y8IQ%_2<j%n1blGl`%9MDGj2I$2-jUcaAK_5Y?x1?av
zOGZqMrqeTrjc1i=**}q)nap@Wul4J}0X;Zac4L%5E9K$IDID=yJ%h^BozJ8xH_Bw3
z%LwZ>;{a=il%=;UO9?V=##_UEZ&<-v1?v>7SFp(U&voU8)~J1Po5Wt40q8(<H;@r7
z);xe*rquT+U3VOw$ZBFUT1JnYP%Ia7|GgWypIVW*<q(j|bEAT!JPbapEe_sB-D{B>
zVTOhK=WQa*OR*MsS17he!7~-yq9Ci_MGAgj!7B)IZBNP%_3LdR^8Q%2{$4E>S#Ddp
zO{x;Iq_BGC;+MP6s6$g-MypS>E*=+A;tv0L^ws9jCe{T|xcKG2#rutPk<T}mTTCi`
pI2g#N7WP&ne{CFT9u=>L+J26HLAtm9`4DV?vg`7z&fAi!|1Syz2eJSF

literal 0
HcmV?d00001

diff --git a/ldm/models/diffusion/ddim.py b/ldm/models/diffusion/ddim.py
new file mode 100644
index 0000000..27ead0e
--- /dev/null
+++ b/ldm/models/diffusion/ddim.py
@@ -0,0 +1,336 @@
+"""SAMPLING ONLY."""
+
+import torch
+import numpy as np
+from tqdm import tqdm
+
+from ldm.modules.diffusionmodules.util import make_ddim_sampling_parameters, make_ddim_timesteps, noise_like, extract_into_tensor
+
+
+class DDIMSampler(object):
+    def __init__(self, model, schedule="linear", **kwargs):
+        super().__init__()
+        self.model = model
+        self.ddpm_num_timesteps = model.num_timesteps
+        self.schedule = schedule
+
+    def register_buffer(self, name, attr):
+        if type(attr) == torch.Tensor:
+            if attr.device != torch.device("cuda"):
+                attr = attr.to(torch.device("cuda"))
+        setattr(self, name, attr)
+
+    def make_schedule(self, ddim_num_steps, ddim_discretize="uniform", ddim_eta=0., verbose=True):
+        self.ddim_timesteps = make_ddim_timesteps(ddim_discr_method=ddim_discretize, num_ddim_timesteps=ddim_num_steps,
+                                                  num_ddpm_timesteps=self.ddpm_num_timesteps,verbose=verbose)
+        alphas_cumprod = self.model.alphas_cumprod
+        assert alphas_cumprod.shape[0] == self.ddpm_num_timesteps, 'alphas have to be defined for each timestep'
+        to_torch = lambda x: x.clone().detach().to(torch.float32).to(self.model.device)
+
+        self.register_buffer('betas', to_torch(self.model.betas))
+        self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
+        self.register_buffer('alphas_cumprod_prev', to_torch(self.model.alphas_cumprod_prev))
+
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod.cpu())))
+        self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod.cpu())))
+        self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod.cpu())))
+        self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod.cpu())))
+        self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod.cpu() - 1)))
+
+        # ddim sampling parameters
+        ddim_sigmas, ddim_alphas, ddim_alphas_prev = make_ddim_sampling_parameters(alphacums=alphas_cumprod.cpu(),
+                                                                                   ddim_timesteps=self.ddim_timesteps,
+                                                                                   eta=ddim_eta,verbose=verbose)
+        self.register_buffer('ddim_sigmas', ddim_sigmas)
+        self.register_buffer('ddim_alphas', ddim_alphas)
+        self.register_buffer('ddim_alphas_prev', ddim_alphas_prev)
+        self.register_buffer('ddim_sqrt_one_minus_alphas', np.sqrt(1. - ddim_alphas))
+        sigmas_for_original_sampling_steps = ddim_eta * torch.sqrt(
+            (1 - self.alphas_cumprod_prev) / (1 - self.alphas_cumprod) * (
+                        1 - self.alphas_cumprod / self.alphas_cumprod_prev))
+        self.register_buffer('ddim_sigmas_for_original_num_steps', sigmas_for_original_sampling_steps)
+
+    @torch.no_grad()
+    def sample(self,
+               S,
+               batch_size,
+               shape,
+               conditioning=None,
+               callback=None,
+               normals_sequence=None,
+               img_callback=None,
+               quantize_x0=False,
+               eta=0.,
+               mask=None,
+               x0=None,
+               temperature=1.,
+               noise_dropout=0.,
+               score_corrector=None,
+               corrector_kwargs=None,
+               verbose=True,
+               x_T=None,
+               log_every_t=100,
+               unconditional_guidance_scale=1.,
+               unconditional_conditioning=None, # this has to come in the same format as the conditioning, # e.g. as encoded tokens, ...
+               dynamic_threshold=None,
+               ucg_schedule=None,
+               **kwargs
+               ):
+        if conditioning is not None:
+            if isinstance(conditioning, dict):
+                ctmp = conditioning[list(conditioning.keys())[0]]
+                while isinstance(ctmp, list): ctmp = ctmp[0]
+                cbs = ctmp.shape[0]
+                if cbs != batch_size:
+                    print(f"Warning: Got {cbs} conditionings but batch-size is {batch_size}")
+
+            elif isinstance(conditioning, list):
+                for ctmp in conditioning:
+                    if ctmp.shape[0] != batch_size:
+                        print(f"Warning: Got {cbs} conditionings but batch-size is {batch_size}")
+
+            else:
+                if conditioning.shape[0] != batch_size:
+                    print(f"Warning: Got {conditioning.shape[0]} conditionings but batch-size is {batch_size}")
+
+        self.make_schedule(ddim_num_steps=S, ddim_eta=eta, verbose=verbose)
+        # sampling
+        C, H, W = shape
+        size = (batch_size, C, H, W)
+        print(f'Data shape for DDIM sampling is {size}, eta {eta}')
+
+        samples, intermediates = self.ddim_sampling(conditioning, size,
+                                                    callback=callback,
+                                                    img_callback=img_callback,
+                                                    quantize_denoised=quantize_x0,
+                                                    mask=mask, x0=x0,
+                                                    ddim_use_original_steps=False,
+                                                    noise_dropout=noise_dropout,
+                                                    temperature=temperature,
+                                                    score_corrector=score_corrector,
+                                                    corrector_kwargs=corrector_kwargs,
+                                                    x_T=x_T,
+                                                    log_every_t=log_every_t,
+                                                    unconditional_guidance_scale=unconditional_guidance_scale,
+                                                    unconditional_conditioning=unconditional_conditioning,
+                                                    dynamic_threshold=dynamic_threshold,
+                                                    ucg_schedule=ucg_schedule
+                                                    )
+        return samples, intermediates
+
+    @torch.no_grad()
+    def ddim_sampling(self, cond, shape,
+                      x_T=None, ddim_use_original_steps=False,
+                      callback=None, timesteps=None, quantize_denoised=False,
+                      mask=None, x0=None, img_callback=None, log_every_t=100,
+                      temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None,
+                      unconditional_guidance_scale=1., unconditional_conditioning=None, dynamic_threshold=None,
+                      ucg_schedule=None):
+        device = self.model.betas.device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+
+        if timesteps is None:
+            timesteps = self.ddpm_num_timesteps if ddim_use_original_steps else self.ddim_timesteps
+        elif timesteps is not None and not ddim_use_original_steps:
+            subset_end = int(min(timesteps / self.ddim_timesteps.shape[0], 1) * self.ddim_timesteps.shape[0]) - 1
+            timesteps = self.ddim_timesteps[:subset_end]
+
+        intermediates = {'x_inter': [img], 'pred_x0': [img]}
+        time_range = reversed(range(0,timesteps)) if ddim_use_original_steps else np.flip(timesteps)
+        total_steps = timesteps if ddim_use_original_steps else timesteps.shape[0]
+        print(f"Running DDIM Sampling with {total_steps} timesteps")
+
+        iterator = tqdm(time_range, desc='DDIM Sampler', total=total_steps)
+
+        for i, step in enumerate(iterator):
+            index = total_steps - i - 1
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+
+            if mask is not None:
+                assert x0 is not None
+                img_orig = self.model.q_sample(x0, ts)  # TODO: deterministic forward pass?
+                img = img_orig * mask + (1. - mask) * img
+
+            if ucg_schedule is not None:
+                assert len(ucg_schedule) == len(time_range)
+                unconditional_guidance_scale = ucg_schedule[i]
+
+            outs = self.p_sample_ddim(img, cond, ts, index=index, use_original_steps=ddim_use_original_steps,
+                                      quantize_denoised=quantize_denoised, temperature=temperature,
+                                      noise_dropout=noise_dropout, score_corrector=score_corrector,
+                                      corrector_kwargs=corrector_kwargs,
+                                      unconditional_guidance_scale=unconditional_guidance_scale,
+                                      unconditional_conditioning=unconditional_conditioning,
+                                      dynamic_threshold=dynamic_threshold)
+            img, pred_x0 = outs
+            if callback: callback(i)
+            if img_callback: img_callback(pred_x0, i)
+
+            if index % log_every_t == 0 or index == total_steps - 1:
+                intermediates['x_inter'].append(img)
+                intermediates['pred_x0'].append(pred_x0)
+
+        return img, intermediates
+
+    @torch.no_grad()
+    def p_sample_ddim(self, x, c, t, index, repeat_noise=False, use_original_steps=False, quantize_denoised=False,
+                      temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None,
+                      unconditional_guidance_scale=1., unconditional_conditioning=None,
+                      dynamic_threshold=None):
+        b, *_, device = *x.shape, x.device
+
+        if unconditional_conditioning is None or unconditional_guidance_scale == 1.:
+            model_output = self.model.apply_model(x, t, c)
+        else:
+            x_in = torch.cat([x] * 2)
+            t_in = torch.cat([t] * 2)
+            if isinstance(c, dict):
+                assert isinstance(unconditional_conditioning, dict)
+                c_in = dict()
+                for k in c:
+                    if isinstance(c[k], list):
+                        c_in[k] = [torch.cat([
+                            unconditional_conditioning[k][i],
+                            c[k][i]]) for i in range(len(c[k]))]
+                    else:
+                        c_in[k] = torch.cat([
+                                unconditional_conditioning[k],
+                                c[k]])
+            elif isinstance(c, list):
+                c_in = list()
+                assert isinstance(unconditional_conditioning, list)
+                for i in range(len(c)):
+                    c_in.append(torch.cat([unconditional_conditioning[i], c[i]]))
+            else:
+                c_in = torch.cat([unconditional_conditioning, c])
+            model_uncond, model_t = self.model.apply_model(x_in, t_in, c_in).chunk(2)
+            model_output = model_uncond + unconditional_guidance_scale * (model_t - model_uncond)
+
+        if self.model.parameterization == "v":
+            e_t = self.model.predict_eps_from_z_and_v(x, t, model_output)
+        else:
+            e_t = model_output
+
+        if score_corrector is not None:
+            assert self.model.parameterization == "eps", 'not implemented'
+            e_t = score_corrector.modify_score(self.model, e_t, x, t, c, **corrector_kwargs)
+
+        alphas = self.model.alphas_cumprod if use_original_steps else self.ddim_alphas
+        alphas_prev = self.model.alphas_cumprod_prev if use_original_steps else self.ddim_alphas_prev
+        sqrt_one_minus_alphas = self.model.sqrt_one_minus_alphas_cumprod if use_original_steps else self.ddim_sqrt_one_minus_alphas
+        sigmas = self.model.ddim_sigmas_for_original_num_steps if use_original_steps else self.ddim_sigmas
+        # select parameters corresponding to the currently considered timestep
+        a_t = torch.full((b, 1, 1, 1), alphas[index], device=device)
+        a_prev = torch.full((b, 1, 1, 1), alphas_prev[index], device=device)
+        sigma_t = torch.full((b, 1, 1, 1), sigmas[index], device=device)
+        sqrt_one_minus_at = torch.full((b, 1, 1, 1), sqrt_one_minus_alphas[index],device=device)
+
+        # current prediction for x_0
+        if self.model.parameterization != "v":
+            pred_x0 = (x - sqrt_one_minus_at * e_t) / a_t.sqrt()
+        else:
+            pred_x0 = self.model.predict_start_from_z_and_v(x, t, model_output)
+
+        if quantize_denoised:
+            pred_x0, _, *_ = self.model.first_stage_model.quantize(pred_x0)
+
+        if dynamic_threshold is not None:
+            raise NotImplementedError()
+
+        # direction pointing to x_t
+        dir_xt = (1. - a_prev - sigma_t**2).sqrt() * e_t
+        noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature
+        if noise_dropout > 0.:
+            noise = torch.nn.functional.dropout(noise, p=noise_dropout)
+        x_prev = a_prev.sqrt() * pred_x0 + dir_xt + noise
+        return x_prev, pred_x0
+
+    @torch.no_grad()
+    def encode(self, x0, c, t_enc, use_original_steps=False, return_intermediates=None,
+               unconditional_guidance_scale=1.0, unconditional_conditioning=None, callback=None):
+        num_reference_steps = self.ddpm_num_timesteps if use_original_steps else self.ddim_timesteps.shape[0]
+
+        assert t_enc <= num_reference_steps
+        num_steps = t_enc
+
+        if use_original_steps:
+            alphas_next = self.alphas_cumprod[:num_steps]
+            alphas = self.alphas_cumprod_prev[:num_steps]
+        else:
+            alphas_next = self.ddim_alphas[:num_steps]
+            alphas = torch.tensor(self.ddim_alphas_prev[:num_steps])
+
+        x_next = x0
+        intermediates = []
+        inter_steps = []
+        for i in tqdm(range(num_steps), desc='Encoding Image'):
+            t = torch.full((x0.shape[0],), i, device=self.model.device, dtype=torch.long)
+            if unconditional_guidance_scale == 1.:
+                noise_pred = self.model.apply_model(x_next, t, c)
+            else:
+                assert unconditional_conditioning is not None
+                e_t_uncond, noise_pred = torch.chunk(
+                    self.model.apply_model(torch.cat((x_next, x_next)), torch.cat((t, t)),
+                                           torch.cat((unconditional_conditioning, c))), 2)
+                noise_pred = e_t_uncond + unconditional_guidance_scale * (noise_pred - e_t_uncond)
+
+            xt_weighted = (alphas_next[i] / alphas[i]).sqrt() * x_next
+            weighted_noise_pred = alphas_next[i].sqrt() * (
+                    (1 / alphas_next[i] - 1).sqrt() - (1 / alphas[i] - 1).sqrt()) * noise_pred
+            x_next = xt_weighted + weighted_noise_pred
+            if return_intermediates and i % (
+                    num_steps // return_intermediates) == 0 and i < num_steps - 1:
+                intermediates.append(x_next)
+                inter_steps.append(i)
+            elif return_intermediates and i >= num_steps - 2:
+                intermediates.append(x_next)
+                inter_steps.append(i)
+            if callback: callback(i)
+
+        out = {'x_encoded': x_next, 'intermediate_steps': inter_steps}
+        if return_intermediates:
+            out.update({'intermediates': intermediates})
+        return x_next, out
+
+    @torch.no_grad()
+    def stochastic_encode(self, x0, t, use_original_steps=False, noise=None):
+        # fast, but does not allow for exact reconstruction
+        # t serves as an index to gather the correct alphas
+        if use_original_steps:
+            sqrt_alphas_cumprod = self.sqrt_alphas_cumprod
+            sqrt_one_minus_alphas_cumprod = self.sqrt_one_minus_alphas_cumprod
+        else:
+            sqrt_alphas_cumprod = torch.sqrt(self.ddim_alphas)
+            sqrt_one_minus_alphas_cumprod = self.ddim_sqrt_one_minus_alphas
+
+        if noise is None:
+            noise = torch.randn_like(x0)
+        return (extract_into_tensor(sqrt_alphas_cumprod, t, x0.shape) * x0 +
+                extract_into_tensor(sqrt_one_minus_alphas_cumprod, t, x0.shape) * noise)
+
+    @torch.no_grad()
+    def decode(self, x_latent, cond, t_start, unconditional_guidance_scale=1.0, unconditional_conditioning=None,
+               use_original_steps=False, callback=None):
+
+        timesteps = np.arange(self.ddpm_num_timesteps) if use_original_steps else self.ddim_timesteps
+        timesteps = timesteps[:t_start]
+
+        time_range = np.flip(timesteps)
+        total_steps = timesteps.shape[0]
+        print(f"Running DDIM Sampling with {total_steps} timesteps")
+
+        iterator = tqdm(time_range, desc='Decoding image', total=total_steps)
+        x_dec = x_latent
+        for i, step in enumerate(iterator):
+            index = total_steps - i - 1
+            ts = torch.full((x_latent.shape[0],), step, device=x_latent.device, dtype=torch.long)
+            x_dec, _ = self.p_sample_ddim(x_dec, cond, ts, index=index, use_original_steps=use_original_steps,
+                                          unconditional_guidance_scale=unconditional_guidance_scale,
+                                          unconditional_conditioning=unconditional_conditioning)
+            if callback: callback(i)
+        return x_dec
\ No newline at end of file
diff --git a/ldm/models/diffusion/ddpm.py b/ldm/models/diffusion/ddpm.py
new file mode 100644
index 0000000..8d20a81
--- /dev/null
+++ b/ldm/models/diffusion/ddpm.py
@@ -0,0 +1,1811 @@
+"""
+wild mixture of
+https://github.com/lucidrains/denoising-diffusion-pytorch/blob/7706bdfc6f527f58d33f84b7b522e61e6e3164b3/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
+https://github.com/openai/improved-diffusion/blob/e94489283bb876ac1477d5dd7709bbbd2d9902ce/improved_diffusion/gaussian_diffusion.py
+https://github.com/CompVis/taming-transformers
+-- merci
+"""
+
+import torch
+import torch.nn as nn
+import numpy as np
+import pytorch_lightning as pl
+from torch.optim.lr_scheduler import LambdaLR
+from einops import rearrange, repeat
+from contextlib import contextmanager, nullcontext
+from functools import partial
+import itertools
+from tqdm import tqdm
+from torchvision.utils import make_grid
+from pytorch_lightning.utilities.distributed import rank_zero_only
+from omegaconf import ListConfig
+
+from ldm.util import log_txt_as_img, exists, default, ismap, isimage, mean_flat, count_params, instantiate_from_config
+from ldm.modules.ema import LitEma
+from ldm.modules.distributions.distributions import normal_kl, DiagonalGaussianDistribution
+from ldm.models.autoencoder import IdentityFirstStage, AutoencoderKL
+from ldm.modules.diffusionmodules.util import make_beta_schedule, extract_into_tensor, noise_like
+from ldm.models.diffusion.ddim import DDIMSampler
+from model.mixins import ImageLoggerMixin
+
+
+__conditioning_keys__ = {'concat': 'c_concat',
+                         'crossattn': 'c_crossattn',
+                         'adm': 'y'}
+
+
+def disabled_train(self, mode=True):
+    """Overwrite model.train with this function to make sure train/eval mode
+    does not change anymore."""
+    return self
+
+
+def uniform_on_device(r1, r2, shape, device):
+    return (r1 - r2) * torch.rand(*shape, device=device) + r2
+
+
+class DDPM(pl.LightningModule, ImageLoggerMixin):
+    # classic DDPM with Gaussian diffusion, in image space
+    def __init__(self,
+                 unet_config,
+                 timesteps=1000,
+                 beta_schedule="linear",
+                 loss_type="l2",
+                 ckpt_path=None,
+                 ignore_keys=[],
+                 load_only_unet=False,
+                 monitor="val/loss",
+                 use_ema=True,
+                 first_stage_key="image",
+                 image_size=256,
+                 channels=3,
+                 log_every_t=100,
+                 clip_denoised=True,
+                 linear_start=1e-4,
+                 linear_end=2e-2,
+                 cosine_s=8e-3,
+                 given_betas=None,
+                 original_elbo_weight=0.,
+                 v_posterior=0.,  # weight for choosing posterior variance as sigma = (1-v) * beta_tilde + v * beta
+                 l_simple_weight=1.,
+                 conditioning_key=None,
+                 parameterization="eps",  # all assuming fixed variance schedules
+                 scheduler_config=None,
+                 use_positional_encodings=False,
+                 learn_logvar=False,
+                 logvar_init=0.,
+                 make_it_fit=False,
+                 ucg_training=None,
+                 reset_ema=False,
+                 reset_num_ema_updates=False,
+                 ):
+        super().__init__()
+        assert parameterization in ["eps", "x0", "v"], 'currently only supporting "eps" and "x0" and "v"'
+        self.parameterization = parameterization
+        print(f"{self.__class__.__name__}: Running in {self.parameterization}-prediction mode")
+        self.cond_stage_model = None
+        self.clip_denoised = clip_denoised
+        self.log_every_t = log_every_t
+        self.first_stage_key = first_stage_key
+        self.image_size = image_size  # try conv?
+        self.channels = channels
+        self.use_positional_encodings = use_positional_encodings
+        self.model = DiffusionWrapper(unet_config, conditioning_key)
+        count_params(self.model, verbose=True)
+        self.use_ema = use_ema
+        if self.use_ema:
+            self.model_ema = LitEma(self.model)
+            print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.")
+
+        self.use_scheduler = scheduler_config is not None
+        if self.use_scheduler:
+            self.scheduler_config = scheduler_config
+
+        self.v_posterior = v_posterior
+        self.original_elbo_weight = original_elbo_weight
+        self.l_simple_weight = l_simple_weight
+
+        if monitor is not None:
+            self.monitor = monitor
+        self.make_it_fit = make_it_fit
+        if reset_ema: assert exists(ckpt_path)
+        if ckpt_path is not None:
+            self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys, only_model=load_only_unet)
+            if reset_ema:
+                assert self.use_ema
+                print(f"Resetting ema to pure model weights. This is useful when restoring from an ema-only checkpoint.")
+                self.model_ema = LitEma(self.model)
+        if reset_num_ema_updates:
+            print(" +++++++++++ WARNING: RESETTING NUM_EMA UPDATES TO ZERO +++++++++++ ")
+            assert self.use_ema
+            self.model_ema.reset_num_updates()
+
+        self.register_schedule(given_betas=given_betas, beta_schedule=beta_schedule, timesteps=timesteps,
+                               linear_start=linear_start, linear_end=linear_end, cosine_s=cosine_s)
+
+        self.loss_type = loss_type
+
+        self.learn_logvar = learn_logvar
+        logvar = torch.full(fill_value=logvar_init, size=(self.num_timesteps,))
+        if self.learn_logvar:
+            self.logvar = nn.Parameter(self.logvar, requires_grad=True)
+        else:
+            self.register_buffer('logvar', logvar)
+
+        self.ucg_training = ucg_training or dict()
+        if self.ucg_training:
+            self.ucg_prng = np.random.RandomState()
+
+    def register_schedule(self, given_betas=None, beta_schedule="linear", timesteps=1000,
+                          linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+        if exists(given_betas):
+            betas = given_betas
+        else:
+            betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end,
+                                       cosine_s=cosine_s)
+        alphas = 1. - betas
+        alphas_cumprod = np.cumprod(alphas, axis=0)
+        alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])
+
+        timesteps, = betas.shape
+        self.num_timesteps = int(timesteps)
+        self.linear_start = linear_start
+        self.linear_end = linear_end
+        assert alphas_cumprod.shape[0] == self.num_timesteps, 'alphas have to be defined for each timestep'
+
+        to_torch = partial(torch.tensor, dtype=torch.float32)
+
+        self.register_buffer('betas', to_torch(betas))
+        self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
+        self.register_buffer('alphas_cumprod_prev', to_torch(alphas_cumprod_prev))
+
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
+        self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
+        self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod)))
+        self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod)))
+        self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod - 1)))
+
+        # calculations for posterior q(x_{t-1} | x_t, x_0)
+        posterior_variance = (1 - self.v_posterior) * betas * (1. - alphas_cumprod_prev) / (
+                1. - alphas_cumprod) + self.v_posterior * betas
+        # above: equal to 1. / (1. / (1. - alpha_cumprod_tm1) + alpha_t / beta_t)
+        self.register_buffer('posterior_variance', to_torch(posterior_variance))
+        # below: log calculation clipped because the posterior variance is 0 at the beginning of the diffusion chain
+        self.register_buffer('posterior_log_variance_clipped', to_torch(np.log(np.maximum(posterior_variance, 1e-20))))
+        self.register_buffer('posterior_mean_coef1', to_torch(
+            betas * np.sqrt(alphas_cumprod_prev) / (1. - alphas_cumprod)))
+        self.register_buffer('posterior_mean_coef2', to_torch(
+            (1. - alphas_cumprod_prev) * np.sqrt(alphas) / (1. - alphas_cumprod)))
+
+        if self.parameterization == "eps":
+            lvlb_weights = self.betas ** 2 / (
+                    2 * self.posterior_variance * to_torch(alphas) * (1 - self.alphas_cumprod))
+        elif self.parameterization == "x0":
+            lvlb_weights = 0.5 * np.sqrt(torch.Tensor(alphas_cumprod)) / (2. * 1 - torch.Tensor(alphas_cumprod))
+        elif self.parameterization == "v":
+            lvlb_weights = torch.ones_like(self.betas ** 2 / (
+                    2 * self.posterior_variance * to_torch(alphas) * (1 - self.alphas_cumprod)))
+        else:
+            raise NotImplementedError("mu not supported")
+        lvlb_weights[0] = lvlb_weights[1]
+        self.register_buffer('lvlb_weights', lvlb_weights, persistent=False)
+        assert not torch.isnan(self.lvlb_weights).all()
+
+    @contextmanager
+    def ema_scope(self, context=None):
+        if self.use_ema:
+            self.model_ema.store(self.model.parameters())
+            self.model_ema.copy_to(self.model)
+            if context is not None:
+                print(f"{context}: Switched to EMA weights")
+        try:
+            yield None
+        finally:
+            if self.use_ema:
+                self.model_ema.restore(self.model.parameters())
+                if context is not None:
+                    print(f"{context}: Restored training weights")
+
+    @torch.no_grad()
+    def init_from_ckpt(self, path, ignore_keys=list(), only_model=False):
+        sd = torch.load(path, map_location="cpu")
+        if "state_dict" in list(sd.keys()):
+            sd = sd["state_dict"]
+        keys = list(sd.keys())
+        for k in keys:
+            for ik in ignore_keys:
+                if k.startswith(ik):
+                    print("Deleting key {} from state_dict.".format(k))
+                    del sd[k]
+        if self.make_it_fit:
+            n_params = len([name for name, _ in
+                            itertools.chain(self.named_parameters(),
+                                            self.named_buffers())])
+            for name, param in tqdm(
+                    itertools.chain(self.named_parameters(),
+                                    self.named_buffers()),
+                    desc="Fitting old weights to new weights",
+                    total=n_params
+            ):
+                if not name in sd:
+                    continue
+                old_shape = sd[name].shape
+                new_shape = param.shape
+                assert len(old_shape) == len(new_shape)
+                if len(new_shape) > 2:
+                    # we only modify first two axes
+                    assert new_shape[2:] == old_shape[2:]
+                # assumes first axis corresponds to output dim
+                if not new_shape == old_shape:
+                    new_param = param.clone()
+                    old_param = sd[name]
+                    if len(new_shape) == 1:
+                        for i in range(new_param.shape[0]):
+                            new_param[i] = old_param[i % old_shape[0]]
+                    elif len(new_shape) >= 2:
+                        for i in range(new_param.shape[0]):
+                            for j in range(new_param.shape[1]):
+                                new_param[i, j] = old_param[i % old_shape[0], j % old_shape[1]]
+
+                        n_used_old = torch.ones(old_shape[1])
+                        for j in range(new_param.shape[1]):
+                            n_used_old[j % old_shape[1]] += 1
+                        n_used_new = torch.zeros(new_shape[1])
+                        for j in range(new_param.shape[1]):
+                            n_used_new[j] = n_used_old[j % old_shape[1]]
+
+                        n_used_new = n_used_new[None, :]
+                        while len(n_used_new.shape) < len(new_shape):
+                            n_used_new = n_used_new.unsqueeze(-1)
+                        new_param /= n_used_new
+
+                    sd[name] = new_param
+
+        missing, unexpected = self.load_state_dict(sd, strict=False) if not only_model else self.model.load_state_dict(
+            sd, strict=False)
+        print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys")
+        if len(missing) > 0:
+            print(f"Missing Keys:\n {missing}")
+        if len(unexpected) > 0:
+            print(f"\nUnexpected Keys:\n {unexpected}")
+
+    def q_mean_variance(self, x_start, t):
+        """
+        Get the distribution q(x_t | x_0).
+        :param x_start: the [N x C x ...] tensor of noiseless inputs.
+        :param t: the number of diffusion steps (minus 1). Here, 0 means one step.
+        :return: A tuple (mean, variance, log_variance), all of x_start's shape.
+        """
+        mean = (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start)
+        variance = extract_into_tensor(1.0 - self.alphas_cumprod, t, x_start.shape)
+        log_variance = extract_into_tensor(self.log_one_minus_alphas_cumprod, t, x_start.shape)
+        return mean, variance, log_variance
+
+    def predict_start_from_noise(self, x_t, t, noise):
+        return (
+                extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t -
+                extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * noise
+        )
+
+    def predict_start_from_z_and_v(self, x_t, t, v):
+        # self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
+        # self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x_t.shape) * x_t -
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_t.shape) * v
+        )
+
+    def predict_eps_from_z_and_v(self, x_t, t, v):
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x_t.shape) * v +
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_t.shape) * x_t
+        )
+
+    def q_posterior(self, x_start, x_t, t):
+        posterior_mean = (
+                extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start +
+                extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t
+        )
+        posterior_variance = extract_into_tensor(self.posterior_variance, t, x_t.shape)
+        posterior_log_variance_clipped = extract_into_tensor(self.posterior_log_variance_clipped, t, x_t.shape)
+        return posterior_mean, posterior_variance, posterior_log_variance_clipped
+
+    def p_mean_variance(self, x, t, clip_denoised: bool):
+        model_out = self.model(x, t)
+        if self.parameterization == "eps":
+            x_recon = self.predict_start_from_noise(x, t=t, noise=model_out)
+        elif self.parameterization == "x0":
+            x_recon = model_out
+        if clip_denoised:
+            x_recon.clamp_(-1., 1.)
+
+        model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
+        return model_mean, posterior_variance, posterior_log_variance
+
+    @torch.no_grad()
+    def p_sample(self, x, t, clip_denoised=True, repeat_noise=False):
+        b, *_, device = *x.shape, x.device
+        model_mean, _, model_log_variance = self.p_mean_variance(x=x, t=t, clip_denoised=clip_denoised)
+        noise = noise_like(x.shape, device, repeat_noise)
+        # no noise when t == 0
+        nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1)))
+        return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise
+
+    @torch.no_grad()
+    def p_sample_loop(self, shape, return_intermediates=False):
+        device = self.betas.device
+        b = shape[0]
+        img = torch.randn(shape, device=device)
+        intermediates = [img]
+        for i in tqdm(reversed(range(0, self.num_timesteps)), desc='Sampling t', total=self.num_timesteps):
+            img = self.p_sample(img, torch.full((b,), i, device=device, dtype=torch.long),
+                                clip_denoised=self.clip_denoised)
+            if i % self.log_every_t == 0 or i == self.num_timesteps - 1:
+                intermediates.append(img)
+        if return_intermediates:
+            return img, intermediates
+        return img
+
+    @torch.no_grad()
+    def sample(self, batch_size=16, return_intermediates=False):
+        image_size = self.image_size
+        channels = self.channels
+        return self.p_sample_loop((batch_size, channels, image_size, image_size),
+                                  return_intermediates=return_intermediates)
+
+    def q_sample(self, x_start, t, noise=None):
+        noise = default(noise, lambda: torch.randn_like(x_start))
+        return (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start +
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise)
+
+    def get_v(self, x, noise, t):
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x.shape) * noise -
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x.shape) * x
+        )
+
+    def get_loss(self, pred, target, mean=True):
+        if self.loss_type == 'l1':
+            loss = (target - pred).abs()
+            if mean:
+                loss = loss.mean()
+        elif self.loss_type == 'l2':
+            if mean:
+                loss = torch.nn.functional.mse_loss(target, pred)
+            else:
+                loss = torch.nn.functional.mse_loss(target, pred, reduction='none')
+        else:
+            raise NotImplementedError("unknown loss type '{loss_type}'")
+
+        return loss
+
+    def p_losses(self, x_start, t, noise=None):
+        noise = default(noise, lambda: torch.randn_like(x_start))
+        x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+        model_out = self.model(x_noisy, t)
+
+        loss_dict = {}
+        if self.parameterization == "eps":
+            target = noise
+        elif self.parameterization == "x0":
+            target = x_start
+        elif self.parameterization == "v":
+            target = self.get_v(x_start, noise, t)
+        else:
+            raise NotImplementedError(f"Parameterization {self.parameterization} not yet supported")
+
+        loss = self.get_loss(model_out, target, mean=False).mean(dim=[1, 2, 3])
+
+        log_prefix = 'train' if self.training else 'val'
+
+        loss_dict.update({f'{log_prefix}/loss_simple': loss.mean()})
+        loss_simple = loss.mean() * self.l_simple_weight
+
+        loss_vlb = (self.lvlb_weights[t] * loss).mean()
+        loss_dict.update({f'{log_prefix}/loss_vlb': loss_vlb})
+
+        loss = loss_simple + self.original_elbo_weight * loss_vlb
+
+        loss_dict.update({f'{log_prefix}/loss': loss})
+
+        return loss, loss_dict
+
+    def forward(self, x, *args, **kwargs):
+        # b, c, h, w, device, img_size, = *x.shape, x.device, self.image_size
+        # assert h == img_size and w == img_size, f'height and width of image must be {img_size}'
+        t = torch.randint(0, self.num_timesteps, (x.shape[0],), device=self.device).long()
+        return self.p_losses(x, t, *args, **kwargs)
+
+    def get_input(self, batch, k):
+        x = batch[k]
+        if len(x.shape) == 3:
+            x = x[..., None]
+        x = rearrange(x, 'b h w c -> b c h w')
+        x = x.to(memory_format=torch.contiguous_format).float()
+        return x
+
+    def shared_step(self, batch):
+        x = self.get_input(batch, self.first_stage_key)
+        loss, loss_dict = self(x)
+        return loss, loss_dict
+
+    def training_step(self, batch, batch_idx):
+        for k in self.ucg_training:
+            p = self.ucg_training[k]["p"]
+            val = self.ucg_training[k]["val"]
+            if val is None:
+                val = ""
+            for i in range(len(batch[k])):
+                if self.ucg_prng.choice(2, p=[1 - p, p]):
+                    batch[k][i] = val
+
+        loss, loss_dict = self.shared_step(batch)
+
+        self.log_dict(loss_dict, prog_bar=True,
+                      logger=True, on_step=True, on_epoch=True)
+
+        self.log("global_step", self.global_step,
+                 prog_bar=True, logger=True, on_step=True, on_epoch=False)
+
+        if self.use_scheduler:
+            lr = self.optimizers().param_groups[0]['lr']
+            self.log('lr_abs', lr, prog_bar=True, logger=True, on_step=True, on_epoch=False)
+
+        return loss
+
+    @torch.no_grad()
+    def validation_step(self, batch, batch_idx):
+        _, loss_dict_no_ema = self.shared_step(batch)
+        with self.ema_scope():
+            _, loss_dict_ema = self.shared_step(batch)
+            loss_dict_ema = {key + '_ema': loss_dict_ema[key] for key in loss_dict_ema}
+        self.log_dict(loss_dict_no_ema, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+        self.log_dict(loss_dict_ema, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+
+    def on_train_batch_end(self, *args, **kwargs):
+        if self.use_ema:
+            self.model_ema(self.model)
+
+    def _get_rows_from_list(self, samples):
+        n_imgs_per_row = len(samples)
+        denoise_grid = rearrange(samples, 'n b c h w -> b n c h w')
+        denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w')
+        denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row)
+        return denoise_grid
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=2, sample=True, return_keys=None, **kwargs):
+        log = dict()
+        x = self.get_input(batch, self.first_stage_key)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        x = x.to(self.device)[:N]
+        log["inputs"] = x
+
+        # get diffusion row
+        diffusion_row = list()
+        x_start = x[:n_row]
+
+        for t in range(self.num_timesteps):
+            if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                t = t.to(self.device).long()
+                noise = torch.randn_like(x_start)
+                x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+                diffusion_row.append(x_noisy)
+
+        log["diffusion_row"] = self._get_rows_from_list(diffusion_row)
+
+        if sample:
+            # get denoise row
+            with self.ema_scope("Plotting"):
+                samples, denoise_row = self.sample(batch_size=N, return_intermediates=True)
+
+            log["samples"] = samples
+            log["denoise_row"] = self._get_rows_from_list(denoise_row)
+
+        if return_keys:
+            if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0:
+                return log
+            else:
+                return {key: log[key] for key in return_keys}
+        return log
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params = list(self.model.parameters())
+        if self.learn_logvar:
+            params = params + [self.logvar]
+        opt = torch.optim.AdamW(params, lr=lr)
+        return opt
+
+
+class LatentDiffusion(DDPM):
+    """main class"""
+
+    def __init__(self,
+                 first_stage_config,
+                 cond_stage_config,
+                 num_timesteps_cond=None,
+                 cond_stage_key="image",
+                 cond_stage_trainable=False,
+                 concat_mode=True,
+                 cond_stage_forward=None,
+                 conditioning_key=None,
+                 scale_factor=1.0,
+                 scale_by_std=False,
+                 force_null_conditioning=False,
+                 *args, **kwargs):
+        self.force_null_conditioning = force_null_conditioning
+        self.num_timesteps_cond = default(num_timesteps_cond, 1)
+        self.scale_by_std = scale_by_std
+        assert self.num_timesteps_cond <= kwargs['timesteps']
+        # for backwards compatibility after implementation of DiffusionWrapper
+        if conditioning_key is None:
+            conditioning_key = 'concat' if concat_mode else 'crossattn'
+        if cond_stage_config == '__is_unconditional__' and not self.force_null_conditioning:
+            conditioning_key = None
+        ckpt_path = kwargs.pop("ckpt_path", None)
+        reset_ema = kwargs.pop("reset_ema", False)
+        reset_num_ema_updates = kwargs.pop("reset_num_ema_updates", False)
+        ignore_keys = kwargs.pop("ignore_keys", [])
+        super().__init__(conditioning_key=conditioning_key, *args, **kwargs)
+        self.concat_mode = concat_mode
+        self.cond_stage_trainable = cond_stage_trainable
+        self.cond_stage_key = cond_stage_key
+        try:
+            self.num_downs = len(first_stage_config.params.ddconfig.ch_mult) - 1
+        except:
+            self.num_downs = 0
+        if not scale_by_std:
+            self.scale_factor = scale_factor
+        else:
+            self.register_buffer('scale_factor', torch.tensor(scale_factor))
+        self.instantiate_first_stage(first_stage_config)
+        self.instantiate_cond_stage(cond_stage_config)
+        self.cond_stage_forward = cond_stage_forward
+        self.clip_denoised = False
+        self.bbox_tokenizer = None
+
+        self.restarted_from_ckpt = False
+        if ckpt_path is not None:
+            self.init_from_ckpt(ckpt_path, ignore_keys)
+            self.restarted_from_ckpt = True
+            if reset_ema:
+                assert self.use_ema
+                print(
+                    f"Resetting ema to pure model weights. This is useful when restoring from an ema-only checkpoint.")
+                self.model_ema = LitEma(self.model)
+        if reset_num_ema_updates:
+            print(" +++++++++++ WARNING: RESETTING NUM_EMA UPDATES TO ZERO +++++++++++ ")
+            assert self.use_ema
+            self.model_ema.reset_num_updates()
+
+    def make_cond_schedule(self, ):
+        self.cond_ids = torch.full(size=(self.num_timesteps,), fill_value=self.num_timesteps - 1, dtype=torch.long)
+        ids = torch.round(torch.linspace(0, self.num_timesteps - 1, self.num_timesteps_cond)).long()
+        self.cond_ids[:self.num_timesteps_cond] = ids
+
+    @rank_zero_only
+    @torch.no_grad()
+    def on_train_batch_start(self, batch, batch_idx, dataloader_idx):
+        # only for very first batch
+        if self.scale_by_std and self.current_epoch == 0 and self.global_step == 0 and batch_idx == 0 and not self.restarted_from_ckpt:
+            assert self.scale_factor == 1., 'rather not use custom rescaling and std-rescaling simultaneously'
+            # set rescale weight to 1./std of encodings
+            print("### USING STD-RESCALING ###")
+            x = super().get_input(batch, self.first_stage_key)
+            x = x.to(self.device)
+            encoder_posterior = self.encode_first_stage(x)
+            z = self.get_first_stage_encoding(encoder_posterior).detach()
+            del self.scale_factor
+            self.register_buffer('scale_factor', 1. / z.flatten().std())
+            print(f"setting self.scale_factor to {self.scale_factor}")
+            print("### USING STD-RESCALING ###")
+
+    def register_schedule(self,
+                          given_betas=None, beta_schedule="linear", timesteps=1000,
+                          linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+        super().register_schedule(given_betas, beta_schedule, timesteps, linear_start, linear_end, cosine_s)
+
+        self.shorten_cond_schedule = self.num_timesteps_cond > 1
+        if self.shorten_cond_schedule:
+            self.make_cond_schedule()
+
+    def instantiate_first_stage(self, config):
+        model = instantiate_from_config(config)
+        self.first_stage_model = model.eval()
+        self.first_stage_model.train = disabled_train
+        for param in self.first_stage_model.parameters():
+            param.requires_grad = False
+
+    def instantiate_cond_stage(self, config):
+        if not self.cond_stage_trainable:
+            if config == "__is_first_stage__":
+                print("Using first stage also as cond stage.")
+                self.cond_stage_model = self.first_stage_model
+            elif config == "__is_unconditional__":
+                print(f"Training {self.__class__.__name__} as an unconditional model.")
+                self.cond_stage_model = None
+                # self.be_unconditional = True
+            else:
+                model = instantiate_from_config(config)
+                self.cond_stage_model = model.eval()
+                self.cond_stage_model.train = disabled_train
+                for param in self.cond_stage_model.parameters():
+                    param.requires_grad = False
+        else:
+            assert config != '__is_first_stage__'
+            assert config != '__is_unconditional__'
+            model = instantiate_from_config(config)
+            self.cond_stage_model = model
+
+    def _get_denoise_row_from_list(self, samples, desc='', force_no_decoder_quantization=False):
+        denoise_row = []
+        for zd in tqdm(samples, desc=desc):
+            denoise_row.append(self.decode_first_stage(zd.to(self.device),
+                                                       force_not_quantize=force_no_decoder_quantization))
+        n_imgs_per_row = len(denoise_row)
+        denoise_row = torch.stack(denoise_row)  # n_log_step, n_row, C, H, W
+        denoise_grid = rearrange(denoise_row, 'n b c h w -> b n c h w')
+        denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w')
+        denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row)
+        return denoise_grid
+
+    def get_first_stage_encoding(self, encoder_posterior):
+        if isinstance(encoder_posterior, DiagonalGaussianDistribution):
+            z = encoder_posterior.sample()
+        elif isinstance(encoder_posterior, torch.Tensor):
+            z = encoder_posterior
+        else:
+            raise NotImplementedError(f"encoder_posterior of type '{type(encoder_posterior)}' not yet implemented")
+        # print(self.scale_factor)
+        # exit()
+        return self.scale_factor * z
+
+    def get_learned_conditioning(self, c):
+        if self.cond_stage_forward is None:
+            if hasattr(self.cond_stage_model, 'encode') and callable(self.cond_stage_model.encode):
+                c = self.cond_stage_model.encode(c)
+                if isinstance(c, DiagonalGaussianDistribution):
+                    c = c.mode()
+            else:
+                c = self.cond_stage_model(c)
+        else:
+            assert hasattr(self.cond_stage_model, self.cond_stage_forward)
+            c = getattr(self.cond_stage_model, self.cond_stage_forward)(c)
+        return c
+
+    def meshgrid(self, h, w):
+        y = torch.arange(0, h).view(h, 1, 1).repeat(1, w, 1)
+        x = torch.arange(0, w).view(1, w, 1).repeat(h, 1, 1)
+
+        arr = torch.cat([y, x], dim=-1)
+        return arr
+
+    def delta_border(self, h, w):
+        """
+        :param h: height
+        :param w: width
+        :return: normalized distance to image border,
+         wtith min distance = 0 at border and max dist = 0.5 at image center
+        """
+        lower_right_corner = torch.tensor([h - 1, w - 1]).view(1, 1, 2)
+        arr = self.meshgrid(h, w) / lower_right_corner
+        dist_left_up = torch.min(arr, dim=-1, keepdims=True)[0]
+        dist_right_down = torch.min(1 - arr, dim=-1, keepdims=True)[0]
+        edge_dist = torch.min(torch.cat([dist_left_up, dist_right_down], dim=-1), dim=-1)[0]
+        return edge_dist
+
+    def get_weighting(self, h, w, Ly, Lx, device):
+        weighting = self.delta_border(h, w)
+        weighting = torch.clip(weighting, self.split_input_params["clip_min_weight"],
+                               self.split_input_params["clip_max_weight"], )
+        weighting = weighting.view(1, h * w, 1).repeat(1, 1, Ly * Lx).to(device)
+
+        if self.split_input_params["tie_braker"]:
+            L_weighting = self.delta_border(Ly, Lx)
+            L_weighting = torch.clip(L_weighting,
+                                     self.split_input_params["clip_min_tie_weight"],
+                                     self.split_input_params["clip_max_tie_weight"])
+
+            L_weighting = L_weighting.view(1, 1, Ly * Lx).to(device)
+            weighting = weighting * L_weighting
+        return weighting
+
+    def get_fold_unfold(self, x, kernel_size, stride, uf=1, df=1):  # todo load once not every time, shorten code
+        """
+        :param x: img of size (bs, c, h, w)
+        :return: n img crops of size (n, bs, c, kernel_size[0], kernel_size[1])
+        """
+        bs, nc, h, w = x.shape
+
+        # number of crops in image
+        Ly = (h - kernel_size[0]) // stride[0] + 1
+        Lx = (w - kernel_size[1]) // stride[1] + 1
+
+        if uf == 1 and df == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold = torch.nn.Fold(output_size=x.shape[2:], **fold_params)
+
+            weighting = self.get_weighting(kernel_size[0], kernel_size[1], Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h, w)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0], kernel_size[1], Ly * Lx))
+
+        elif uf > 1 and df == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold_params2 = dict(kernel_size=(kernel_size[0] * uf, kernel_size[0] * uf),
+                                dilation=1, padding=0,
+                                stride=(stride[0] * uf, stride[1] * uf))
+            fold = torch.nn.Fold(output_size=(x.shape[2] * uf, x.shape[3] * uf), **fold_params2)
+
+            weighting = self.get_weighting(kernel_size[0] * uf, kernel_size[1] * uf, Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h * uf, w * uf)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0] * uf, kernel_size[1] * uf, Ly * Lx))
+
+        elif df > 1 and uf == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold_params2 = dict(kernel_size=(kernel_size[0] // df, kernel_size[0] // df),
+                                dilation=1, padding=0,
+                                stride=(stride[0] // df, stride[1] // df))
+            fold = torch.nn.Fold(output_size=(x.shape[2] // df, x.shape[3] // df), **fold_params2)
+
+            weighting = self.get_weighting(kernel_size[0] // df, kernel_size[1] // df, Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h // df, w // df)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0] // df, kernel_size[1] // df, Ly * Lx))
+
+        else:
+            raise NotImplementedError
+
+        return fold, unfold, normalization, weighting
+
+    @torch.no_grad()
+    def get_input(self, batch, k, return_first_stage_outputs=False, force_c_encode=False,
+                  cond_key=None, return_original_cond=False, bs=None, return_x=False):
+        x = super().get_input(batch, k)
+        if bs is not None:
+            x = x[:bs]
+        x = x.to(self.device)
+        encoder_posterior = self.encode_first_stage(x)
+        z = self.get_first_stage_encoding(encoder_posterior).detach()
+
+        if self.model.conditioning_key is not None and not self.force_null_conditioning:
+            if cond_key is None:
+                cond_key = self.cond_stage_key
+            if cond_key != self.first_stage_key:
+                if cond_key in ['caption', 'coordinates_bbox', "txt"]:
+                    xc = batch[cond_key]
+                elif cond_key in ['class_label', 'cls']:
+                    xc = batch
+                else:
+                    xc = super().get_input(batch, cond_key).to(self.device)
+            else:
+                xc = x
+            if not self.cond_stage_trainable or force_c_encode:
+                if isinstance(xc, dict) or isinstance(xc, list):
+                    c = self.get_learned_conditioning(xc)
+                else:
+                    c = self.get_learned_conditioning(xc.to(self.device))
+            else:
+                c = xc
+            if bs is not None:
+                c = c[:bs]
+
+            if self.use_positional_encodings:
+                pos_x, pos_y = self.compute_latent_shifts(batch)
+                ckey = __conditioning_keys__[self.model.conditioning_key]
+                c = {ckey: c, 'pos_x': pos_x, 'pos_y': pos_y}
+
+        else:
+            c = None
+            xc = None
+            if self.use_positional_encodings:
+                pos_x, pos_y = self.compute_latent_shifts(batch)
+                c = {'pos_x': pos_x, 'pos_y': pos_y}
+        out = [z, c]
+        if return_first_stage_outputs:
+            xrec = self.decode_first_stage(z)
+            out.extend([x, xrec])
+        if return_x:
+            out.extend([x])
+        if return_original_cond:
+            out.append(xc)
+        return out
+
+    @torch.no_grad()
+    def decode_first_stage(self, z, predict_cids=False, force_not_quantize=False):
+        if predict_cids:
+            if z.dim() == 4:
+                z = torch.argmax(z.exp(), dim=1).long()
+            z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None)
+            z = rearrange(z, 'b h w c -> b c h w').contiguous()
+
+        z = 1. / self.scale_factor * z
+        return self.first_stage_model.decode(z)
+    
+    # 2023-04-08
+    def decode_first_stage_with_grad(self, z, predict_cids=False, force_not_quantize=False):
+        if predict_cids:
+            if z.dim() == 4:
+                z = torch.argmax(z.exp(), dim=1).long()
+            z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None)
+            z = rearrange(z, 'b h w c -> b c h w').contiguous()
+
+        z = 1. / self.scale_factor * z
+        return self.first_stage_model.decode(z)
+
+    @torch.no_grad()
+    def encode_first_stage(self, x):
+        return self.first_stage_model.encode(x)
+
+    def shared_step(self, batch, **kwargs):
+        x, c = self.get_input(batch, self.first_stage_key)
+        loss = self(x, c)
+        return loss
+
+    def forward(self, x, c, *args, **kwargs):
+        t = torch.randint(0, self.num_timesteps, (x.shape[0],), device=self.device).long()
+        if self.model.conditioning_key is not None:
+            assert c is not None
+            if self.cond_stage_trainable:
+                c = self.get_learned_conditioning(c)
+            if self.shorten_cond_schedule:  # TODO: drop this option
+                tc = self.cond_ids[t].to(self.device)
+                c = self.q_sample(x_start=c, t=tc, noise=torch.randn_like(c.float()))
+        return self.p_losses(x, c, t, *args, **kwargs)
+
+    def apply_model(self, x_noisy, t, cond, return_ids=False):
+        if isinstance(cond, dict):
+            # hybrid case, cond is expected to be a dict
+            pass
+        else:
+            if not isinstance(cond, list):
+                cond = [cond]
+            key = 'c_concat' if self.model.conditioning_key == 'concat' else 'c_crossattn'
+            cond = {key: cond}
+
+        x_recon = self.model(x_noisy, t, **cond)
+
+        if isinstance(x_recon, tuple) and not return_ids:
+            return x_recon[0]
+        else:
+            return x_recon
+
+    def _predict_eps_from_xstart(self, x_t, t, pred_xstart):
+        return (extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t - pred_xstart) / \
+               extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape)
+
+    def _prior_bpd(self, x_start):
+        """
+        Get the prior KL term for the variational lower-bound, measured in
+        bits-per-dim.
+        This term can't be optimized, as it only depends on the encoder.
+        :param x_start: the [N x C x ...] tensor of inputs.
+        :return: a batch of [N] KL values (in bits), one per batch element.
+        """
+        batch_size = x_start.shape[0]
+        t = torch.tensor([self.num_timesteps - 1] * batch_size, device=x_start.device)
+        qt_mean, _, qt_log_variance = self.q_mean_variance(x_start, t)
+        kl_prior = normal_kl(mean1=qt_mean, logvar1=qt_log_variance, mean2=0.0, logvar2=0.0)
+        return mean_flat(kl_prior) / np.log(2.0)
+
+    def p_losses(self, x_start, cond, t, noise=None):
+        noise = default(noise, lambda: torch.randn_like(x_start))
+        x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+        model_output = self.apply_model(x_noisy, t, cond)
+
+        loss_dict = {}
+        prefix = 'train' if self.training else 'val'
+
+        if self.parameterization == "x0":
+            target = x_start
+        elif self.parameterization == "eps":
+            target = noise
+        elif self.parameterization == "v":
+            target = self.get_v(x_start, noise, t)
+        else:
+            raise NotImplementedError()
+
+        loss_simple = self.get_loss(model_output, target, mean=False).mean([1, 2, 3])
+        loss_dict.update({f'{prefix}/loss_simple': loss_simple.mean()})
+
+        logvar_t = self.logvar[t].to(self.device)
+        loss = loss_simple / torch.exp(logvar_t) + logvar_t
+        # loss = loss_simple / torch.exp(self.logvar) + self.logvar
+        if self.learn_logvar:
+            loss_dict.update({f'{prefix}/loss_gamma': loss.mean()})
+            loss_dict.update({'logvar': self.logvar.data.mean()})
+
+        loss = self.l_simple_weight * loss.mean()
+
+        loss_vlb = self.get_loss(model_output, target, mean=False).mean(dim=(1, 2, 3))
+        loss_vlb = (self.lvlb_weights[t] * loss_vlb).mean()
+        loss_dict.update({f'{prefix}/loss_vlb': loss_vlb})
+        loss += (self.original_elbo_weight * loss_vlb)
+        loss_dict.update({f'{prefix}/loss': loss})
+
+        return loss, loss_dict
+
+    def p_mean_variance(self, x, c, t, clip_denoised: bool, return_codebook_ids=False, quantize_denoised=False,
+                        return_x0=False, score_corrector=None, corrector_kwargs=None):
+        t_in = t
+        model_out = self.apply_model(x, t_in, c, return_ids=return_codebook_ids)
+
+        if score_corrector is not None:
+            assert self.parameterization == "eps"
+            model_out = score_corrector.modify_score(self, model_out, x, t, c, **corrector_kwargs)
+
+        if return_codebook_ids:
+            model_out, logits = model_out
+
+        if self.parameterization == "eps":
+            x_recon = self.predict_start_from_noise(x, t=t, noise=model_out)
+        elif self.parameterization == "x0":
+            x_recon = model_out
+        else:
+            raise NotImplementedError()
+
+        if clip_denoised:
+            x_recon.clamp_(-1., 1.)
+        if quantize_denoised:
+            x_recon, _, [_, _, indices] = self.first_stage_model.quantize(x_recon)
+        model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
+        if return_codebook_ids:
+            return model_mean, posterior_variance, posterior_log_variance, logits
+        elif return_x0:
+            return model_mean, posterior_variance, posterior_log_variance, x_recon
+        else:
+            return model_mean, posterior_variance, posterior_log_variance
+
+    @torch.no_grad()
+    def p_sample(self, x, c, t, clip_denoised=False, repeat_noise=False,
+                 return_codebook_ids=False, quantize_denoised=False, return_x0=False,
+                 temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None):
+        b, *_, device = *x.shape, x.device
+        outputs = self.p_mean_variance(x=x, c=c, t=t, clip_denoised=clip_denoised,
+                                       return_codebook_ids=return_codebook_ids,
+                                       quantize_denoised=quantize_denoised,
+                                       return_x0=return_x0,
+                                       score_corrector=score_corrector, corrector_kwargs=corrector_kwargs)
+        if return_codebook_ids:
+            raise DeprecationWarning("Support dropped.")
+            model_mean, _, model_log_variance, logits = outputs
+        elif return_x0:
+            model_mean, _, model_log_variance, x0 = outputs
+        else:
+            model_mean, _, model_log_variance = outputs
+
+        noise = noise_like(x.shape, device, repeat_noise) * temperature
+        if noise_dropout > 0.:
+            noise = torch.nn.functional.dropout(noise, p=noise_dropout)
+        # no noise when t == 0
+        nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1)))
+
+        if return_codebook_ids:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, logits.argmax(dim=1)
+        if return_x0:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, x0
+        else:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise
+
+    @torch.no_grad()
+    def progressive_denoising(self, cond, shape, verbose=True, callback=None, quantize_denoised=False,
+                              img_callback=None, mask=None, x0=None, temperature=1., noise_dropout=0.,
+                              score_corrector=None, corrector_kwargs=None, batch_size=None, x_T=None, start_T=None,
+                              log_every_t=None):
+        if not log_every_t:
+            log_every_t = self.log_every_t
+        timesteps = self.num_timesteps
+        if batch_size is not None:
+            b = batch_size if batch_size is not None else shape[0]
+            shape = [batch_size] + list(shape)
+        else:
+            b = batch_size = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=self.device)
+        else:
+            img = x_T
+        intermediates = []
+        if cond is not None:
+            if isinstance(cond, dict):
+                cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else
+                list(map(lambda x: x[:batch_size], cond[key])) for key in cond}
+            else:
+                cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size]
+
+        if start_T is not None:
+            timesteps = min(timesteps, start_T)
+        iterator = tqdm(reversed(range(0, timesteps)), desc='Progressive Generation',
+                        total=timesteps) if verbose else reversed(
+            range(0, timesteps))
+        if type(temperature) == float:
+            temperature = [temperature] * timesteps
+
+        for i in iterator:
+            ts = torch.full((b,), i, device=self.device, dtype=torch.long)
+            if self.shorten_cond_schedule:
+                assert self.model.conditioning_key != 'hybrid'
+                tc = self.cond_ids[ts].to(cond.device)
+                cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond))
+
+            img, x0_partial = self.p_sample(img, cond, ts,
+                                            clip_denoised=self.clip_denoised,
+                                            quantize_denoised=quantize_denoised, return_x0=True,
+                                            temperature=temperature[i], noise_dropout=noise_dropout,
+                                            score_corrector=score_corrector, corrector_kwargs=corrector_kwargs)
+            if mask is not None:
+                assert x0 is not None
+                img_orig = self.q_sample(x0, ts)
+                img = img_orig * mask + (1. - mask) * img
+
+            if i % log_every_t == 0 or i == timesteps - 1:
+                intermediates.append(x0_partial)
+            if callback: callback(i)
+            if img_callback: img_callback(img, i)
+        return img, intermediates
+
+    @torch.no_grad()
+    def p_sample_loop(self, cond, shape, return_intermediates=False,
+                      x_T=None, verbose=True, callback=None, timesteps=None, quantize_denoised=False,
+                      mask=None, x0=None, img_callback=None, start_T=None,
+                      log_every_t=None):
+
+        if not log_every_t:
+            log_every_t = self.log_every_t
+        device = self.betas.device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+
+        intermediates = [img]
+        if timesteps is None:
+            timesteps = self.num_timesteps
+
+        if start_T is not None:
+            timesteps = min(timesteps, start_T)
+        iterator = tqdm(reversed(range(0, timesteps)), desc='Sampling t', total=timesteps) if verbose else reversed(
+            range(0, timesteps))
+
+        if mask is not None:
+            assert x0 is not None
+            assert x0.shape[2:3] == mask.shape[2:3]  # spatial size has to match
+
+        for i in iterator:
+            ts = torch.full((b,), i, device=device, dtype=torch.long)
+            if self.shorten_cond_schedule:
+                assert self.model.conditioning_key != 'hybrid'
+                tc = self.cond_ids[ts].to(cond.device)
+                cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond))
+
+            img = self.p_sample(img, cond, ts,
+                                clip_denoised=self.clip_denoised,
+                                quantize_denoised=quantize_denoised)
+            if mask is not None:
+                img_orig = self.q_sample(x0, ts)
+                img = img_orig * mask + (1. - mask) * img
+
+            if i % log_every_t == 0 or i == timesteps - 1:
+                intermediates.append(img)
+            if callback: callback(i)
+            if img_callback: img_callback(img, i)
+
+        if return_intermediates:
+            return img, intermediates
+        return img
+
+    @torch.no_grad()
+    def sample(self, cond, batch_size=16, return_intermediates=False, x_T=None,
+               verbose=True, timesteps=None, quantize_denoised=False,
+               mask=None, x0=None, shape=None, **kwargs):
+        if shape is None:
+            shape = (batch_size, self.channels, self.image_size, self.image_size)
+        if cond is not None:
+            if isinstance(cond, dict):
+                cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else
+                list(map(lambda x: x[:batch_size], cond[key])) for key in cond}
+            else:
+                cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size]
+        return self.p_sample_loop(cond,
+                                  shape,
+                                  return_intermediates=return_intermediates, x_T=x_T,
+                                  verbose=verbose, timesteps=timesteps, quantize_denoised=quantize_denoised,
+                                  mask=mask, x0=x0)
+
+    @torch.no_grad()
+    def sample_log(self, cond, batch_size, ddim, ddim_steps, **kwargs):
+        if ddim:
+            ddim_sampler = DDIMSampler(self)
+            shape = (self.channels, self.image_size, self.image_size)
+            samples, intermediates = ddim_sampler.sample(ddim_steps, batch_size,
+                                                         shape, cond, verbose=False, **kwargs)
+
+        else:
+            samples, intermediates = self.sample(cond=cond, batch_size=batch_size,
+                                                 return_intermediates=True, **kwargs)
+
+        return samples, intermediates
+
+    @torch.no_grad()
+    def get_unconditional_conditioning(self, batch_size, null_label=None):
+        if null_label is not None:
+            xc = null_label
+            if isinstance(xc, ListConfig):
+                xc = list(xc)
+            if isinstance(xc, dict) or isinstance(xc, list):
+                c = self.get_learned_conditioning(xc)
+            else:
+                if hasattr(xc, "to"):
+                    xc = xc.to(self.device)
+                c = self.get_learned_conditioning(xc)
+        else:
+            if self.cond_stage_key in ["class_label", "cls"]:
+                xc = self.cond_stage_model.get_unconditional_conditioning(batch_size, device=self.device)
+                return self.get_learned_conditioning(xc)
+            else:
+                raise NotImplementedError("todo")
+        if isinstance(c, list):  # in case the encoder gives us a list
+            for i in range(len(c)):
+                c[i] = repeat(c[i], '1 ... -> b ...', b=batch_size).to(self.device)
+        else:
+            c = repeat(c, '1 ... -> b ...', b=batch_size).to(self.device)
+        return c
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=50, ddim_eta=0., return_keys=None,
+                   quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True,
+                   plot_diffusion_rows=True, unconditional_guidance_scale=1., unconditional_guidance_label=None,
+                   use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc = self.get_input(batch, self.first_stage_key,
+                                           return_first_stage_outputs=True,
+                                           force_c_encode=True,
+                                           return_original_cond=True,
+                                           bs=N)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', "cls"]:
+                try:
+                    xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                    log['conditioning'] = xc
+                except KeyError:
+                    # probably no "human_label" in batch
+                    pass
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+            if quantize_denoised and not isinstance(self.first_stage_model, AutoencoderKL) and not isinstance(
+                    self.first_stage_model, IdentityFirstStage):
+                # also display when quantizing x0 while sampling
+                with ema_scope("Plotting Quantized Denoised"):
+                    samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                             ddim_steps=ddim_steps, eta=ddim_eta,
+                                                             quantize_denoised=True)
+                    # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True,
+                    #                                      quantize_denoised=True)
+                x_samples = self.decode_first_stage(samples.to(self.device))
+                log["samples_x0_quantized"] = x_samples
+
+        if unconditional_guidance_scale > 1.0:
+            uc = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            if self.model.conditioning_key == "crossattn-adm":
+                uc = {"c_crossattn": [uc], "c_adm": c["c_adm"]}
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        if inpaint:
+            # make a simple center square
+            b, h, w = z.shape[0], z.shape[2], z.shape[3]
+            mask = torch.ones(N, h, w).to(self.device)
+            # zeros will be filled in
+            mask[:, h // 4:3 * h // 4, w // 4:3 * w // 4] = 0.
+            mask = mask[:, None, ...]
+            with ema_scope("Plotting Inpaint"):
+                samples, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim, eta=ddim_eta,
+                                             ddim_steps=ddim_steps, x0=z[:N], mask=mask)
+            x_samples = self.decode_first_stage(samples.to(self.device))
+            log["samples_inpainting"] = x_samples
+            log["mask"] = mask
+
+            # outpaint
+            mask = 1. - mask
+            with ema_scope("Plotting Outpaint"):
+                samples, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim, eta=ddim_eta,
+                                             ddim_steps=ddim_steps, x0=z[:N], mask=mask)
+            x_samples = self.decode_first_stage(samples.to(self.device))
+            log["samples_outpainting"] = x_samples
+
+        if plot_progressive_rows:
+            with ema_scope("Plotting Progressives"):
+                img, progressives = self.progressive_denoising(c,
+                                                               shape=(self.channels, self.image_size, self.image_size),
+                                                               batch_size=N)
+            prog_row = self._get_denoise_row_from_list(progressives, desc="Progressive Generation")
+            log["progressive_row"] = prog_row
+
+        if return_keys:
+            if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0:
+                return log
+            else:
+                return {key: log[key] for key in return_keys}
+        return log
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params = list(self.model.parameters())
+        if self.cond_stage_trainable:
+            print(f"{self.__class__.__name__}: Also optimizing conditioner params!")
+            params = params + list(self.cond_stage_model.parameters())
+        if self.learn_logvar:
+            print('Diffusion model optimizing logvar')
+            params.append(self.logvar)
+        opt = torch.optim.AdamW(params, lr=lr)
+        if self.use_scheduler:
+            assert 'target' in self.scheduler_config
+            scheduler = instantiate_from_config(self.scheduler_config)
+
+            print("Setting up LambdaLR scheduler...")
+            scheduler = [
+                {
+                    'scheduler': LambdaLR(opt, lr_lambda=scheduler.schedule),
+                    'interval': 'step',
+                    'frequency': 1
+                }]
+            return [opt], scheduler
+        return opt
+
+    @torch.no_grad()
+    def to_rgb(self, x):
+        x = x.float()
+        if not hasattr(self, "colorize"):
+            self.colorize = torch.randn(3, x.shape[1], 1, 1).to(x)
+        x = nn.functional.conv2d(x, weight=self.colorize)
+        x = 2. * (x - x.min()) / (x.max() - x.min()) - 1.
+        return x
+
+
+class DiffusionWrapper(pl.LightningModule):
+    def __init__(self, diff_model_config, conditioning_key):
+        super().__init__()
+        self.sequential_cross_attn = diff_model_config.pop("sequential_crossattn", False)
+        self.diffusion_model = instantiate_from_config(diff_model_config)
+        self.conditioning_key = conditioning_key
+        assert self.conditioning_key in [None, 'concat', 'crossattn', 'hybrid', 'adm', 'hybrid-adm', 'crossattn-adm']
+
+    def forward(self, x, t, c_concat: list = None, c_crossattn: list = None, c_adm=None):
+        if self.conditioning_key is None:
+            out = self.diffusion_model(x, t)
+        elif self.conditioning_key == 'concat':
+            xc = torch.cat([x] + c_concat, dim=1)
+            out = self.diffusion_model(xc, t)
+        elif self.conditioning_key == 'crossattn':
+            if not self.sequential_cross_attn:
+                cc = torch.cat(c_crossattn, 1)
+            else:
+                cc = c_crossattn
+            out = self.diffusion_model(x, t, context=cc)
+        elif self.conditioning_key == 'hybrid':
+            xc = torch.cat([x] + c_concat, dim=1)
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(xc, t, context=cc)
+        elif self.conditioning_key == 'hybrid-adm':
+            assert c_adm is not None
+            xc = torch.cat([x] + c_concat, dim=1)
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(xc, t, context=cc, y=c_adm)
+        elif self.conditioning_key == 'crossattn-adm':
+            assert c_adm is not None
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(x, t, context=cc, y=c_adm)
+        elif self.conditioning_key == 'adm':
+            cc = c_crossattn[0]
+            out = self.diffusion_model(x, t, y=cc)
+        else:
+            raise NotImplementedError()
+
+        return out
+
+
+class LatentUpscaleDiffusion(LatentDiffusion):
+    def __init__(self, *args, low_scale_config, low_scale_key="LR", noise_level_key=None, **kwargs):
+        super().__init__(*args, **kwargs)
+        # assumes that neither the cond_stage nor the low_scale_model contain trainable params
+        assert not self.cond_stage_trainable
+        self.instantiate_low_stage(low_scale_config)
+        self.low_scale_key = low_scale_key
+        self.noise_level_key = noise_level_key
+
+    def instantiate_low_stage(self, config):
+        model = instantiate_from_config(config)
+        self.low_scale_model = model.eval()
+        self.low_scale_model.train = disabled_train
+        for param in self.low_scale_model.parameters():
+            param.requires_grad = False
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, log_mode=False):
+        if not log_mode:
+            z, c = super().get_input(batch, k, force_c_encode=True, bs=bs)
+        else:
+            z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                                  force_c_encode=True, return_original_cond=True, bs=bs)
+        x_low = batch[self.low_scale_key][:bs]
+        x_low = rearrange(x_low, 'b h w c -> b c h w')
+        x_low = x_low.to(memory_format=torch.contiguous_format).float()
+        zx, noise_level = self.low_scale_model(x_low)
+        if self.noise_level_key is not None:
+            # get noise level from batch instead, e.g. when extracting a custom noise level for bsr
+            raise NotImplementedError('TODO')
+
+        all_conds = {"c_concat": [zx], "c_crossattn": [c], "c_adm": noise_level}
+        if log_mode:
+            # TODO: maybe disable if too expensive
+            x_low_rec = self.low_scale_model.decode(zx)
+            return z, all_conds, x, xrec, xc, x_low, x_low_rec, noise_level
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=200, ddim_eta=1., return_keys=None,
+                   plot_denoise_rows=False, plot_progressive_rows=True, plot_diffusion_rows=True,
+                   unconditional_guidance_scale=1., unconditional_guidance_label=None, use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc, x_low, x_low_rec, noise_level = self.get_input(batch, self.first_stage_key, bs=N,
+                                                                          log_mode=True)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        log["x_lr"] = x_low
+        log[f"x_lr_rec_@noise_levels{'-'.join(map(lambda x: str(x), list(noise_level.cpu().numpy())))}"] = x_low_rec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', 'cls']:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                log['conditioning'] = xc
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+        if unconditional_guidance_scale > 1.0:
+            uc_tmp = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            # TODO explore better "unconditional" choices for the other keys
+            # maybe guide away from empty text label and highest noise level and maximally degraded zx?
+            uc = dict()
+            for k in c:
+                if k == "c_crossattn":
+                    assert isinstance(c[k], list) and len(c[k]) == 1
+                    uc[k] = [uc_tmp]
+                elif k == "c_adm":  # todo: only run with text-based guidance?
+                    assert isinstance(c[k], torch.Tensor)
+                    #uc[k] = torch.ones_like(c[k]) * self.low_scale_model.max_noise_level
+                    uc[k] = c[k]
+                elif isinstance(c[k], list):
+                    uc[k] = [c[k][i] for i in range(len(c[k]))]
+                else:
+                    uc[k] = c[k]
+
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        if plot_progressive_rows:
+            with ema_scope("Plotting Progressives"):
+                img, progressives = self.progressive_denoising(c,
+                                                               shape=(self.channels, self.image_size, self.image_size),
+                                                               batch_size=N)
+            prog_row = self._get_denoise_row_from_list(progressives, desc="Progressive Generation")
+            log["progressive_row"] = prog_row
+
+        return log
+
+
+class LatentFinetuneDiffusion(LatentDiffusion):
+    """
+         Basis for different finetunas, such as inpainting or depth2image
+         To disable finetuning mode, set finetune_keys to None
+    """
+
+    def __init__(self,
+                 concat_keys: tuple,
+                 finetune_keys=("model.diffusion_model.input_blocks.0.0.weight",
+                                "model_ema.diffusion_modelinput_blocks00weight"
+                                ),
+                 keep_finetune_dims=4,
+                 # if model was trained without concat mode before and we would like to keep these channels
+                 c_concat_log_start=None,  # to log reconstruction of c_concat codes
+                 c_concat_log_end=None,
+                 *args, **kwargs
+                 ):
+        ckpt_path = kwargs.pop("ckpt_path", None)
+        ignore_keys = kwargs.pop("ignore_keys", list())
+        super().__init__(*args, **kwargs)
+        self.finetune_keys = finetune_keys
+        self.concat_keys = concat_keys
+        self.keep_dims = keep_finetune_dims
+        self.c_concat_log_start = c_concat_log_start
+        self.c_concat_log_end = c_concat_log_end
+        if exists(self.finetune_keys): assert exists(ckpt_path), 'can only finetune from a given checkpoint'
+        if exists(ckpt_path):
+            self.init_from_ckpt(ckpt_path, ignore_keys)
+
+    def init_from_ckpt(self, path, ignore_keys=list(), only_model=False):
+        sd = torch.load(path, map_location="cpu")
+        if "state_dict" in list(sd.keys()):
+            sd = sd["state_dict"]
+        keys = list(sd.keys())
+        for k in keys:
+            for ik in ignore_keys:
+                if k.startswith(ik):
+                    print("Deleting key {} from state_dict.".format(k))
+                    del sd[k]
+
+            # make it explicit, finetune by including extra input channels
+            if exists(self.finetune_keys) and k in self.finetune_keys:
+                new_entry = None
+                for name, param in self.named_parameters():
+                    if name in self.finetune_keys:
+                        print(
+                            f"modifying key '{name}' and keeping its original {self.keep_dims} (channels) dimensions only")
+                        new_entry = torch.zeros_like(param)  # zero init
+                assert exists(new_entry), 'did not find matching parameter to modify'
+                new_entry[:, :self.keep_dims, ...] = sd[k]
+                sd[k] = new_entry
+
+        missing, unexpected = self.load_state_dict(sd, strict=False) if not only_model else self.model.load_state_dict(
+            sd, strict=False)
+        print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys")
+        if len(missing) > 0:
+            print(f"Missing Keys: {missing}")
+        if len(unexpected) > 0:
+            print(f"Unexpected Keys: {unexpected}")
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=200, ddim_eta=1., return_keys=None,
+                   quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True,
+                   plot_diffusion_rows=True, unconditional_guidance_scale=1., unconditional_guidance_label=None,
+                   use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc = self.get_input(batch, self.first_stage_key, bs=N, return_first_stage_outputs=True)
+        c_cat, c = c["c_concat"][0], c["c_crossattn"][0]
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', 'cls']:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                log['conditioning'] = xc
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if not (self.c_concat_log_start is None and self.c_concat_log_end is None):
+            log["c_concat_decoded"] = self.decode_first_stage(c_cat[:, self.c_concat_log_start:self.c_concat_log_end])
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]},
+                                                         batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+        if unconditional_guidance_scale > 1.0:
+            uc_cross = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            uc_cat = c_cat
+            uc_full = {"c_concat": [uc_cat], "c_crossattn": [uc_cross]}
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]},
+                                                 batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc_full,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        return log
+
+
+class LatentInpaintDiffusion(LatentFinetuneDiffusion):
+    """
+    can either run as pure inpainting model (only concat mode) or with mixed conditionings,
+    e.g. mask as concat and text via cross-attn.
+    To disable finetuning mode, set finetune_keys to None
+     """
+
+    def __init__(self,
+                 concat_keys=("mask", "masked_image"),
+                 masked_image_key="masked_image",
+                 *args, **kwargs
+                 ):
+        super().__init__(concat_keys, *args, **kwargs)
+        self.masked_image_key = masked_image_key
+        assert self.masked_image_key in concat_keys
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for inpainting'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        c_cat = list()
+        for ck in self.concat_keys:
+            cc = rearrange(batch[ck], 'b h w c -> b c h w').to(memory_format=torch.contiguous_format).float()
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            bchw = z.shape
+            if ck != self.masked_image_key:
+                cc = torch.nn.functional.interpolate(cc, size=bchw[-2:])
+            else:
+                cc = self.get_first_stage_encoding(self.encode_first_stage(cc))
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super(LatentInpaintDiffusion, self).log_images(*args, **kwargs)
+        log["masked_image"] = rearrange(args[0]["masked_image"],
+                                        'b h w c -> b c h w').to(memory_format=torch.contiguous_format).float()
+        return log
+
+
+class LatentDepth2ImageDiffusion(LatentFinetuneDiffusion):
+    """
+    condition on monocular depth estimation
+    """
+
+    def __init__(self, depth_stage_config, concat_keys=("midas_in",), *args, **kwargs):
+        super().__init__(concat_keys=concat_keys, *args, **kwargs)
+        self.depth_model = instantiate_from_config(depth_stage_config)
+        self.depth_stage_key = concat_keys[0]
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for depth2img'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        assert len(self.concat_keys) == 1
+        c_cat = list()
+        for ck in self.concat_keys:
+            cc = batch[ck]
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            cc = self.depth_model(cc)
+            cc = torch.nn.functional.interpolate(
+                cc,
+                size=z.shape[2:],
+                mode="bicubic",
+                align_corners=False,
+            )
+
+            depth_min, depth_max = torch.amin(cc, dim=[1, 2, 3], keepdim=True), torch.amax(cc, dim=[1, 2, 3],
+                                                                                           keepdim=True)
+            cc = 2. * (cc - depth_min) / (depth_max - depth_min + 0.001) - 1.
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super().log_images(*args, **kwargs)
+        depth = self.depth_model(args[0][self.depth_stage_key])
+        depth_min, depth_max = torch.amin(depth, dim=[1, 2, 3], keepdim=True), \
+                               torch.amax(depth, dim=[1, 2, 3], keepdim=True)
+        log["depth"] = 2. * (depth - depth_min) / (depth_max - depth_min) - 1.
+        return log
+
+
+class LatentUpscaleFinetuneDiffusion(LatentFinetuneDiffusion):
+    """
+        condition on low-res image (and optionally on some spatial noise augmentation)
+    """
+    def __init__(self, concat_keys=("lr",), reshuffle_patch_size=None,
+                 low_scale_config=None, low_scale_key=None, *args, **kwargs):
+        super().__init__(concat_keys=concat_keys, *args, **kwargs)
+        self.reshuffle_patch_size = reshuffle_patch_size
+        self.low_scale_model = None
+        if low_scale_config is not None:
+            print("Initializing a low-scale model")
+            assert exists(low_scale_key)
+            self.instantiate_low_stage(low_scale_config)
+            self.low_scale_key = low_scale_key
+
+    def instantiate_low_stage(self, config):
+        model = instantiate_from_config(config)
+        self.low_scale_model = model.eval()
+        self.low_scale_model.train = disabled_train
+        for param in self.low_scale_model.parameters():
+            param.requires_grad = False
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for upscaling-ft'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        assert len(self.concat_keys) == 1
+        # optionally make spatial noise_level here
+        c_cat = list()
+        noise_level = None
+        for ck in self.concat_keys:
+            cc = batch[ck]
+            cc = rearrange(cc, 'b h w c -> b c h w')
+            if exists(self.reshuffle_patch_size):
+                assert isinstance(self.reshuffle_patch_size, int)
+                cc = rearrange(cc, 'b c (p1 h) (p2 w) -> b (p1 p2 c) h w',
+                               p1=self.reshuffle_patch_size, p2=self.reshuffle_patch_size)
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            if exists(self.low_scale_model) and ck == self.low_scale_key:
+                cc, noise_level = self.low_scale_model(cc)
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        if exists(noise_level):
+            all_conds = {"c_concat": [c_cat], "c_crossattn": [c], "c_adm": noise_level}
+        else:
+            all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super().log_images(*args, **kwargs)
+        log["lr"] = rearrange(args[0]["lr"], 'b h w c -> b c h w')
+        return log
diff --git a/ldm/models/diffusion/ddpm_ccsr_stage1.py b/ldm/models/diffusion/ddpm_ccsr_stage1.py
new file mode 100644
index 0000000..597faf3
--- /dev/null
+++ b/ldm/models/diffusion/ddpm_ccsr_stage1.py
@@ -0,0 +1,1858 @@
+"""
+wild mixture of
+https://github.com/lucidrains/denoising-diffusion-pytorch/blob/7706bdfc6f527f58d33f84b7b522e61e6e3164b3/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
+https://github.com/openai/improved-diffusion/blob/e94489283bb876ac1477d5dd7709bbbd2d9902ce/improved_diffusion/gaussian_diffusion.py
+https://github.com/CompVis/taming-transformers
+-- merci
+"""
+
+import torch
+import torch.nn as nn
+import numpy as np
+import pytorch_lightning as pl
+from torch.optim.lr_scheduler import LambdaLR
+from einops import rearrange, repeat
+from contextlib import contextmanager, nullcontext
+from functools import partial
+import itertools
+from tqdm import tqdm
+from torchvision.utils import make_grid
+from pytorch_lightning.utilities.distributed import rank_zero_only
+from omegaconf import ListConfig
+
+from ldm.util import log_txt_as_img, exists, default, ismap, isimage, mean_flat, count_params, instantiate_from_config
+from ldm.modules.ema import LitEma
+from ldm.modules.distributions.distributions import normal_kl, DiagonalGaussianDistribution
+from ldm.models.autoencoder import IdentityFirstStage, AutoencoderKL
+from ldm.modules.diffusionmodules.util import make_beta_schedule, extract_into_tensor, noise_like
+from ldm.models.diffusion.ddim import DDIMSampler
+from model.mixins import ImageLoggerMixin
+
+
+__conditioning_keys__ = {'concat': 'c_concat',
+                         'crossattn': 'c_crossattn',
+                         'adm': 'y'}
+
+
+def disabled_train(self, mode=True):
+    """Overwrite model.train with this function to make sure train/eval mode
+    does not change anymore."""
+    return self
+
+
+def uniform_on_device(r1, r2, shape, device):
+    return (r1 - r2) * torch.rand(*shape, device=device) + r2
+
+
+class DDPM(pl.LightningModule, ImageLoggerMixin):
+    # classic DDPM with Gaussian diffusion, in image space
+    def __init__(self,
+                 unet_config,
+                 timesteps=1000,
+                 t_max=0.6667,
+                 t_min=0.3333,
+                 beta_schedule="linear",
+                 loss_type="l2",
+                 ckpt_path=None,
+                 ignore_keys=[],
+                 load_only_unet=False,
+                 monitor="val/loss",
+                 use_ema=True,
+                 first_stage_key="image",
+                 image_size=256,
+                 channels=3,
+                 log_every_t=100,
+                 clip_denoised=True,
+                 linear_start=1e-4,
+                 linear_end=2e-2,
+                 cosine_s=8e-3,
+                 given_betas=None,
+                 original_elbo_weight=0.,
+                 v_posterior=0.,  # weight for choosing posterior variance as sigma = (1-v) * beta_tilde + v * beta
+                 l_simple_weight=1.,
+                 conditioning_key=None,
+                 parameterization="eps",  # all assuming fixed variance schedules
+                 scheduler_config=None,
+                 use_positional_encodings=False,
+                 learn_logvar=False,
+                 logvar_init=0.,
+                 make_it_fit=False,
+                 ucg_training=None,
+                 reset_ema=False,
+                 reset_num_ema_updates=False,
+                 ):
+        super().__init__()
+        assert parameterization in ["eps", "x0", "v"], 'currently only supporting "eps" and "x0" and "v"'
+        self.parameterization = parameterization
+        print(f"{self.__class__.__name__}: Running in {self.parameterization}-prediction mode")
+        self.cond_stage_model = None
+        self.clip_denoised = clip_denoised
+        self.log_every_t = log_every_t
+        self.first_stage_key = first_stage_key
+        self.image_size = image_size  # try conv?
+        self.channels = channels
+        self.use_positional_encodings = use_positional_encodings
+        self.model = DiffusionWrapper(unet_config, conditioning_key)
+        count_params(self.model, verbose=True)
+        self.use_ema = use_ema
+        if self.use_ema:
+            self.model_ema = LitEma(self.model)
+            print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.")
+
+        self.use_scheduler = scheduler_config is not None
+        if self.use_scheduler:
+            self.scheduler_config = scheduler_config
+
+        self.v_posterior = v_posterior
+        self.original_elbo_weight = original_elbo_weight
+        self.l_simple_weight = l_simple_weight
+
+        if monitor is not None:
+            self.monitor = monitor
+        self.make_it_fit = make_it_fit
+        if reset_ema: assert exists(ckpt_path)
+        if ckpt_path is not None:
+            self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys, only_model=load_only_unet)
+            if reset_ema:
+                assert self.use_ema
+                print(f"Resetting ema to pure model weights. This is useful when restoring from an ema-only checkpoint.")
+                self.model_ema = LitEma(self.model)
+        if reset_num_ema_updates:
+            print(" +++++++++++ WARNING: RESETTING NUM_EMA UPDATES TO ZERO +++++++++++ ")
+            assert self.use_ema
+            self.model_ema.reset_num_updates()
+
+        self.register_schedule(given_betas=given_betas, beta_schedule=beta_schedule, timesteps=timesteps,
+                               linear_start=linear_start, linear_end=linear_end, cosine_s=cosine_s)
+
+        self.t_max = t_max
+        self.t_min = t_min
+        
+        self.loss_type = loss_type
+
+        self.learn_logvar = learn_logvar
+        logvar = torch.full(fill_value=logvar_init, size=(self.num_timesteps,))
+        if self.learn_logvar:
+            self.logvar = nn.Parameter(self.logvar, requires_grad=True)
+        else:
+            self.register_buffer('logvar', logvar)
+
+        self.ucg_training = ucg_training or dict()
+        if self.ucg_training:
+            self.ucg_prng = np.random.RandomState()
+
+    def register_schedule(self, given_betas=None, beta_schedule="linear", timesteps=1000,
+                          linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+        if exists(given_betas):
+            betas = given_betas
+        else:
+            betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end,
+                                       cosine_s=cosine_s)
+        alphas = 1. - betas
+        alphas_cumprod = np.cumprod(alphas, axis=0)
+        alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])
+
+        timesteps, = betas.shape
+        self.num_timesteps = int(timesteps)
+        self.linear_start = linear_start
+        self.linear_end = linear_end
+        assert alphas_cumprod.shape[0] == self.num_timesteps, 'alphas have to be defined for each timestep'
+
+        to_torch = partial(torch.tensor, dtype=torch.float32)
+
+        self.register_buffer('betas', to_torch(betas))
+        self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
+        self.register_buffer('alphas_cumprod_prev', to_torch(alphas_cumprod_prev))
+
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
+        self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
+        self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod)))
+        self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod)))
+        self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod - 1)))
+
+        # calculations for posterior q(x_{t-1} | x_t, x_0)
+        posterior_variance = (1 - self.v_posterior) * betas * (1. - alphas_cumprod_prev) / (
+                1. - alphas_cumprod) + self.v_posterior * betas
+        # above: equal to 1. / (1. / (1. - alpha_cumprod_tm1) + alpha_t / beta_t)
+        self.register_buffer('posterior_variance', to_torch(posterior_variance))
+        # below: log calculation clipped because the posterior variance is 0 at the beginning of the diffusion chain
+        self.register_buffer('posterior_log_variance_clipped', to_torch(np.log(np.maximum(posterior_variance, 1e-20))))
+        self.register_buffer('posterior_mean_coef1', to_torch(
+            betas * np.sqrt(alphas_cumprod_prev) / (1. - alphas_cumprod)))
+        self.register_buffer('posterior_mean_coef2', to_torch(
+            (1. - alphas_cumprod_prev) * np.sqrt(alphas) / (1. - alphas_cumprod)))
+
+        if self.parameterization == "eps":
+            lvlb_weights = self.betas ** 2 / (
+                    2 * self.posterior_variance * to_torch(alphas) * (1 - self.alphas_cumprod))
+        elif self.parameterization == "x0":
+            lvlb_weights = 0.5 * np.sqrt(torch.Tensor(alphas_cumprod)) / (2. * 1 - torch.Tensor(alphas_cumprod))
+        elif self.parameterization == "v":
+            lvlb_weights = torch.ones_like(self.betas ** 2 / (
+                    2 * self.posterior_variance * to_torch(alphas) * (1 - self.alphas_cumprod)))
+        else:
+            raise NotImplementedError("mu not supported")
+        lvlb_weights[0] = lvlb_weights[1]
+        self.register_buffer('lvlb_weights', lvlb_weights, persistent=False)
+        assert not torch.isnan(self.lvlb_weights).all()
+
+    @contextmanager
+    def ema_scope(self, context=None):
+        if self.use_ema:
+            self.model_ema.store(self.model.parameters())
+            self.model_ema.copy_to(self.model)
+            if context is not None:
+                print(f"{context}: Switched to EMA weights")
+        try:
+            yield None
+        finally:
+            if self.use_ema:
+                self.model_ema.restore(self.model.parameters())
+                if context is not None:
+                    print(f"{context}: Restored training weights")
+
+    @torch.no_grad()
+    def init_from_ckpt(self, path, ignore_keys=list(), only_model=False):
+        sd = torch.load(path, map_location="cpu")
+        if "state_dict" in list(sd.keys()):
+            sd = sd["state_dict"]
+        keys = list(sd.keys())
+        for k in keys:
+            for ik in ignore_keys:
+                if k.startswith(ik):
+                    print("Deleting key {} from state_dict.".format(k))
+                    del sd[k]
+        if self.make_it_fit:
+            n_params = len([name for name, _ in
+                            itertools.chain(self.named_parameters(),
+                                            self.named_buffers())])
+            for name, param in tqdm(
+                    itertools.chain(self.named_parameters(),
+                                    self.named_buffers()),
+                    desc="Fitting old weights to new weights",
+                    total=n_params
+            ):
+                if not name in sd:
+                    continue
+                old_shape = sd[name].shape
+                new_shape = param.shape
+                assert len(old_shape) == len(new_shape)
+                if len(new_shape) > 2:
+                    # we only modify first two axes
+                    assert new_shape[2:] == old_shape[2:]
+                # assumes first axis corresponds to output dim
+                if not new_shape == old_shape:
+                    new_param = param.clone()
+                    old_param = sd[name]
+                    if len(new_shape) == 1:
+                        for i in range(new_param.shape[0]):
+                            new_param[i] = old_param[i % old_shape[0]]
+                    elif len(new_shape) >= 2:
+                        for i in range(new_param.shape[0]):
+                            for j in range(new_param.shape[1]):
+                                new_param[i, j] = old_param[i % old_shape[0], j % old_shape[1]]
+
+                        n_used_old = torch.ones(old_shape[1])
+                        for j in range(new_param.shape[1]):
+                            n_used_old[j % old_shape[1]] += 1
+                        n_used_new = torch.zeros(new_shape[1])
+                        for j in range(new_param.shape[1]):
+                            n_used_new[j] = n_used_old[j % old_shape[1]]
+
+                        n_used_new = n_used_new[None, :]
+                        while len(n_used_new.shape) < len(new_shape):
+                            n_used_new = n_used_new.unsqueeze(-1)
+                        new_param /= n_used_new
+
+                    sd[name] = new_param
+
+        missing, unexpected = self.load_state_dict(sd, strict=False) if not only_model else self.model.load_state_dict(
+            sd, strict=False)
+        print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys")
+        if len(missing) > 0:
+            print(f"Missing Keys:\n {missing}")
+        if len(unexpected) > 0:
+            print(f"\nUnexpected Keys:\n {unexpected}")
+
+    def q_mean_variance(self, x_start, t):
+        """
+        Get the distribution q(x_t | x_0).
+        :param x_start: the [N x C x ...] tensor of noiseless inputs.
+        :param t: the number of diffusion steps (minus 1). Here, 0 means one step.
+        :return: A tuple (mean, variance, log_variance), all of x_start's shape.
+        """
+        mean = (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start)
+        variance = extract_into_tensor(1.0 - self.alphas_cumprod, t, x_start.shape)
+        log_variance = extract_into_tensor(self.log_one_minus_alphas_cumprod, t, x_start.shape)
+        return mean, variance, log_variance
+
+    def predict_start_from_noise(self, x_t, t, noise):
+        return (
+                extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t -
+                extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * noise
+        )
+
+    def predict_start_from_z_and_v(self, x_t, t, v):
+        # self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
+        # self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x_t.shape) * x_t -
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_t.shape) * v
+        )
+
+    def predict_eps_from_z_and_v(self, x_t, t, v):
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x_t.shape) * v +
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_t.shape) * x_t
+        )
+
+    def q_posterior(self, x_start, x_t, t):
+        posterior_mean = (
+                extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start +
+                extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t
+        )
+        posterior_variance = extract_into_tensor(self.posterior_variance, t, x_t.shape)
+        posterior_log_variance_clipped = extract_into_tensor(self.posterior_log_variance_clipped, t, x_t.shape)
+        return posterior_mean, posterior_variance, posterior_log_variance_clipped
+
+    def p_mean_variance(self, x, t, clip_denoised: bool):
+        model_out = self.model(x, t)
+        if self.parameterization == "eps":
+            x_recon = self.predict_start_from_noise(x, t=t, noise=model_out)
+        elif self.parameterization == "x0":
+            x_recon = model_out
+        if clip_denoised:
+            x_recon.clamp_(-1., 1.)
+
+        model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
+        return model_mean, posterior_variance, posterior_log_variance
+
+    @torch.no_grad()
+    def p_sample(self, x, t, clip_denoised=True, repeat_noise=False):
+        b, *_, device = *x.shape, x.device
+        model_mean, _, model_log_variance = self.p_mean_variance(x=x, t=t, clip_denoised=clip_denoised)
+        noise = noise_like(x.shape, device, repeat_noise)
+        # no noise when t == 0
+        nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1)))
+        return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise
+
+    @torch.no_grad()
+    def p_sample_loop(self, shape, return_intermediates=False):
+        device = self.betas.device
+        b = shape[0]
+        img = torch.randn(shape, device=device)
+        intermediates = [img]
+        for i in tqdm(reversed(range(0, self.num_timesteps)), desc='Sampling t', total=self.num_timesteps):
+            img = self.p_sample(img, torch.full((b,), i, device=device, dtype=torch.long),
+                                clip_denoised=self.clip_denoised)
+            if i % self.log_every_t == 0 or i == self.num_timesteps - 1:
+                intermediates.append(img)
+        if return_intermediates:
+            return img, intermediates
+        return img
+
+    @torch.no_grad()
+    def sample(self, batch_size=16, return_intermediates=False):
+        image_size = self.image_size
+        channels = self.channels
+        return self.p_sample_loop((batch_size, channels, image_size, image_size),
+                                  return_intermediates=return_intermediates)
+
+    def q_sample(self, x_start, t, noise=None):
+        noise = default(noise, lambda: torch.randn_like(x_start))
+        return (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start +
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise)
+
+    def get_v(self, x, noise, t):
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x.shape) * noise -
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x.shape) * x
+        )
+
+    def get_loss(self, pred, target, mean=True):
+        if self.loss_type == 'l1':
+            loss = (target - pred).abs()
+            if mean:
+                loss = loss.mean()
+        elif self.loss_type == 'l2':
+            if mean:
+                loss = torch.nn.functional.mse_loss(target, pred)
+            else:
+                loss = torch.nn.functional.mse_loss(target, pred, reduction='none')
+        else:
+            raise NotImplementedError("unknown loss type '{loss_type}'")
+
+        return loss
+
+    def p_losses(self, x_start, t, noise=None):
+        noise = default(noise, lambda: torch.randn_like(x_start))
+        x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+        model_out = self.model(x_noisy, t)
+
+        loss_dict = {}
+        if self.parameterization == "eps":
+            target = noise
+        elif self.parameterization == "x0":
+            target = x_start
+        elif self.parameterization == "v":
+            target = self.get_v(x_start, noise, t)
+        else:
+            raise NotImplementedError(f"Parameterization {self.parameterization} not yet supported")
+
+        loss = self.get_loss(model_out, target, mean=False).mean(dim=[1, 2, 3])
+
+        log_prefix = 'train' if self.training else 'val'
+
+        loss_dict.update({f'{log_prefix}/loss_simple': loss.mean()})
+        loss_simple = loss.mean() * self.l_simple_weight
+
+        loss_vlb = (self.lvlb_weights[t] * loss).mean()
+        loss_dict.update({f'{log_prefix}/loss_vlb': loss_vlb})
+
+        loss = loss_simple + self.original_elbo_weight * loss_vlb
+
+        loss_dict.update({f'{log_prefix}/loss': loss})
+
+        return loss, loss_dict
+
+    def forward(self, x, *args, **kwargs):
+        # b, c, h, w, device, img_size, = *x.shape, x.device, self.image_size
+        # assert h == img_size and w == img_size, f'height and width of image must be {img_size}'
+        t = torch.randint(0, self.num_timesteps, (x.shape[0],), device=self.device).long()
+        return self.p_losses(x, t, *args, **kwargs)
+
+    def get_input(self, batch, k):
+        x = batch[k]
+        if len(x.shape) == 3:
+            x = x[..., None]
+        x = rearrange(x, 'b h w c -> b c h w')
+        x = x.to(memory_format=torch.contiguous_format).float()
+        return x
+
+    def shared_step(self, batch):
+        x = self.get_input(batch, self.first_stage_key)
+        loss, loss_dict = self(x)
+        return loss, loss_dict
+
+    def training_step(self, batch, batch_idx, optimizer_idx):
+        for k in self.ucg_training:
+            p = self.ucg_training[k]["p"]
+            val = self.ucg_training[k]["val"]
+            if val is None:
+                val = ""
+            for i in range(len(batch[k])):
+                if self.ucg_prng.choice(2, p=[1 - p, p]):
+                    batch[k][i] = val
+
+        loss, loss_dict = self.shared_step(batch, optimizer_idx)
+
+        self.log_dict(loss_dict, prog_bar=True,
+                      logger=True, on_step=True, on_epoch=True)
+
+        self.log("global_step", self.global_step,
+                 prog_bar=True, logger=True, on_step=True, on_epoch=False)
+
+        if self.use_scheduler:
+            lr = self.optimizers().param_groups[0]['lr']
+            self.log('lr_abs', lr, prog_bar=True, logger=True, on_step=True, on_epoch=False)
+
+        return loss
+
+    @torch.no_grad()
+    def validation_step(self, batch, batch_idx):
+        _, loss_dict_no_ema_0 = self.shared_step(batch,0)
+        _, loss_dict_no_ema_1 = self.shared_step(batch,1)
+        with self.ema_scope():
+            _, loss_dict_ema_0 = self.shared_step(batch,0)
+            _, loss_dict_ema_1 = self.shared_step(batch,1)
+            loss_dict_ema_0 = {key + '_ema': loss_dict_ema_0[key] for key in loss_dict_ema_0}
+            loss_dict_ema_1 = {key + '_ema': loss_dict_ema_1[key] for key in loss_dict_ema_1}
+        self.log_dict(loss_dict_no_ema_0, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+        self.log_dict(loss_dict_no_ema_1, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+        self.log_dict(loss_dict_ema_0, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+        self.log_dict(loss_dict_ema_1, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+
+    def on_train_batch_end(self, *args, **kwargs):
+        if self.use_ema:
+            self.model_ema(self.model)
+
+    def _get_rows_from_list(self, samples):
+        n_imgs_per_row = len(samples)
+        denoise_grid = rearrange(samples, 'n b c h w -> b n c h w')
+        denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w')
+        denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row)
+        return denoise_grid
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=2, sample=True, return_keys=None, **kwargs):
+        log = dict()
+        x = self.get_input(batch, self.first_stage_key)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        x = x.to(self.device)[:N]
+        log["inputs"] = x
+
+        # get diffusion row
+        diffusion_row = list()
+        x_start = x[:n_row]
+
+        for t in range(self.num_timesteps):
+            if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                t = t.to(self.device).long()
+                noise = torch.randn_like(x_start)
+                x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+                diffusion_row.append(x_noisy)
+
+        log["diffusion_row"] = self._get_rows_from_list(diffusion_row)
+
+        if sample:
+            # get denoise row
+            with self.ema_scope("Plotting"):
+                samples, denoise_row = self.sample(batch_size=N, return_intermediates=True)
+
+            log["samples"] = samples
+            log["denoise_row"] = self._get_rows_from_list(denoise_row)
+
+        if return_keys:
+            if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0:
+                return log
+            else:
+                return {key: log[key] for key in return_keys}
+        return log
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params = list(self.model.parameters())
+        if self.learn_logvar:
+            params = params + [self.logvar]
+        opt = torch.optim.AdamW(params, lr=lr)
+        return opt
+
+
+class LatentDiffusion(DDPM):
+    """main class"""
+
+    def __init__(self,
+                 first_stage_config,
+                 cond_stage_config,
+                 num_timesteps_cond=None,
+                 cond_stage_key="image",
+                 cond_stage_trainable=False,
+                 concat_mode=True,
+                 cond_stage_forward=None,
+                 conditioning_key=None,
+                 scale_factor=1.0,
+                 scale_by_std=False,
+                 force_null_conditioning=False,
+                 *args, **kwargs):
+        self.force_null_conditioning = force_null_conditioning
+        self.num_timesteps_cond = default(num_timesteps_cond, 1)
+        self.scale_by_std = scale_by_std
+        assert self.num_timesteps_cond <= kwargs['timesteps']
+        # for backwards compatibility after implementation of DiffusionWrapper
+        if conditioning_key is None:
+            conditioning_key = 'concat' if concat_mode else 'crossattn'
+        if cond_stage_config == '__is_unconditional__' and not self.force_null_conditioning:
+            conditioning_key = None
+        ckpt_path = kwargs.pop("ckpt_path", None)
+        reset_ema = kwargs.pop("reset_ema", False)
+        reset_num_ema_updates = kwargs.pop("reset_num_ema_updates", False)
+        ignore_keys = kwargs.pop("ignore_keys", [])
+        super().__init__(conditioning_key=conditioning_key, *args, **kwargs)
+        self.concat_mode = concat_mode
+        self.cond_stage_trainable = cond_stage_trainable
+        self.cond_stage_key = cond_stage_key
+        try:
+            self.num_downs = len(first_stage_config.params.ddconfig.ch_mult) - 1
+        except:
+            self.num_downs = 0
+        if not scale_by_std:
+            self.scale_factor = scale_factor
+        else:
+            self.register_buffer('scale_factor', torch.tensor(scale_factor))
+        self.instantiate_first_stage(first_stage_config)
+        self.instantiate_cond_stage(cond_stage_config)
+        self.cond_stage_forward = cond_stage_forward
+        self.clip_denoised = False
+        self.bbox_tokenizer = None
+
+        self.restarted_from_ckpt = False
+        if ckpt_path is not None:
+            self.init_from_ckpt(ckpt_path, ignore_keys)
+            self.restarted_from_ckpt = True
+            if reset_ema:
+                assert self.use_ema
+                print(
+                    f"Resetting ema to pure model weights. This is useful when restoring from an ema-only checkpoint.")
+                self.model_ema = LitEma(self.model)
+        if reset_num_ema_updates:
+            print(" +++++++++++ WARNING: RESETTING NUM_EMA UPDATES TO ZERO +++++++++++ ")
+            assert self.use_ema
+            self.model_ema.reset_num_updates()
+
+    def make_cond_schedule(self, ):
+        self.cond_ids = torch.full(size=(self.num_timesteps,), fill_value=self.num_timesteps - 1, dtype=torch.long)
+        ids = torch.round(torch.linspace(0, self.num_timesteps - 1, self.num_timesteps_cond)).long()
+        self.cond_ids[:self.num_timesteps_cond] = ids
+
+    @rank_zero_only
+    @torch.no_grad()
+    def on_train_batch_start(self, batch, batch_idx, dataloader_idx):
+        # only for very first batch
+        if self.scale_by_std and self.current_epoch == 0 and self.global_step == 0 and batch_idx == 0 and not self.restarted_from_ckpt:
+            assert self.scale_factor == 1., 'rather not use custom rescaling and std-rescaling simultaneously'
+            # set rescale weight to 1./std of encodings
+            print("### USING STD-RESCALING ###")
+            x = super().get_input(batch, self.first_stage_key)
+            x = x.to(self.device)
+            encoder_posterior = self.encode_first_stage(x)
+            z = self.get_first_stage_encoding(encoder_posterior).detach()
+            del self.scale_factor
+            self.register_buffer('scale_factor', 1. / z.flatten().std())
+            print(f"setting self.scale_factor to {self.scale_factor}")
+            print("### USING STD-RESCALING ###")
+
+    def register_schedule(self,
+                          given_betas=None, beta_schedule="linear", timesteps=1000,
+                          linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+        super().register_schedule(given_betas, beta_schedule, timesteps, linear_start, linear_end, cosine_s)
+
+        self.shorten_cond_schedule = self.num_timesteps_cond > 1
+        if self.shorten_cond_schedule:
+            self.make_cond_schedule()
+
+    def instantiate_first_stage(self, config):
+        model = instantiate_from_config(config)
+        self.first_stage_model = model.eval()
+        self.first_stage_model.train = disabled_train
+        for param in self.first_stage_model.parameters():
+            param.requires_grad = False
+
+    def instantiate_cond_stage(self, config):
+        if not self.cond_stage_trainable:
+            if config == "__is_first_stage__":
+                print("Using first stage also as cond stage.")
+                self.cond_stage_model = self.first_stage_model
+            elif config == "__is_unconditional__":
+                print(f"Training {self.__class__.__name__} as an unconditional model.")
+                self.cond_stage_model = None
+                # self.be_unconditional = True
+            else:
+                model = instantiate_from_config(config)
+                self.cond_stage_model = model.eval()
+                self.cond_stage_model.train = disabled_train
+                for param in self.cond_stage_model.parameters():
+                    param.requires_grad = False
+        else:
+            assert config != '__is_first_stage__'
+            assert config != '__is_unconditional__'
+            model = instantiate_from_config(config)
+            self.cond_stage_model = model
+
+    def _get_denoise_row_from_list(self, samples, desc='', force_no_decoder_quantization=False):
+        denoise_row = []
+        for zd in tqdm(samples, desc=desc):
+            denoise_row.append(self.decode_first_stage(zd.to(self.device),
+                                                       force_not_quantize=force_no_decoder_quantization))
+        n_imgs_per_row = len(denoise_row)
+        denoise_row = torch.stack(denoise_row)  # n_log_step, n_row, C, H, W
+        denoise_grid = rearrange(denoise_row, 'n b c h w -> b n c h w')
+        denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w')
+        denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row)
+        return denoise_grid
+
+    def get_first_stage_encoding(self, encoder_posterior):
+        if isinstance(encoder_posterior, DiagonalGaussianDistribution):
+            z = encoder_posterior.sample()
+        elif isinstance(encoder_posterior, torch.Tensor):
+            z = encoder_posterior
+        else:
+            raise NotImplementedError(f"encoder_posterior of type '{type(encoder_posterior)}' not yet implemented")
+        # print(self.scale_factor)
+        # exit()
+        return self.scale_factor * z
+
+    def get_learned_conditioning(self, c):
+        if self.cond_stage_forward is None:
+            if hasattr(self.cond_stage_model, 'encode') and callable(self.cond_stage_model.encode):
+                c = self.cond_stage_model.encode(c)
+                if isinstance(c, DiagonalGaussianDistribution):
+                    c = c.mode()
+            else:
+                c = self.cond_stage_model(c)
+        else:
+            assert hasattr(self.cond_stage_model, self.cond_stage_forward)
+            c = getattr(self.cond_stage_model, self.cond_stage_forward)(c)
+        return c
+
+    def meshgrid(self, h, w):
+        y = torch.arange(0, h).view(h, 1, 1).repeat(1, w, 1)
+        x = torch.arange(0, w).view(1, w, 1).repeat(h, 1, 1)
+
+        arr = torch.cat([y, x], dim=-1)
+        return arr
+
+    def delta_border(self, h, w):
+        """
+        :param h: height
+        :param w: width
+        :return: normalized distance to image border,
+         wtith min distance = 0 at border and max dist = 0.5 at image center
+        """
+        lower_right_corner = torch.tensor([h - 1, w - 1]).view(1, 1, 2)
+        arr = self.meshgrid(h, w) / lower_right_corner
+        dist_left_up = torch.min(arr, dim=-1, keepdims=True)[0]
+        dist_right_down = torch.min(1 - arr, dim=-1, keepdims=True)[0]
+        edge_dist = torch.min(torch.cat([dist_left_up, dist_right_down], dim=-1), dim=-1)[0]
+        return edge_dist
+
+    def get_weighting(self, h, w, Ly, Lx, device):
+        weighting = self.delta_border(h, w)
+        weighting = torch.clip(weighting, self.split_input_params["clip_min_weight"],
+                               self.split_input_params["clip_max_weight"], )
+        weighting = weighting.view(1, h * w, 1).repeat(1, 1, Ly * Lx).to(device)
+
+        if self.split_input_params["tie_braker"]:
+            L_weighting = self.delta_border(Ly, Lx)
+            L_weighting = torch.clip(L_weighting,
+                                     self.split_input_params["clip_min_tie_weight"],
+                                     self.split_input_params["clip_max_tie_weight"])
+
+            L_weighting = L_weighting.view(1, 1, Ly * Lx).to(device)
+            weighting = weighting * L_weighting
+        return weighting
+
+    def get_fold_unfold(self, x, kernel_size, stride, uf=1, df=1):  # todo load once not every time, shorten code
+        """
+        :param x: img of size (bs, c, h, w)
+        :return: n img crops of size (n, bs, c, kernel_size[0], kernel_size[1])
+        """
+        bs, nc, h, w = x.shape
+
+        # number of crops in image
+        Ly = (h - kernel_size[0]) // stride[0] + 1
+        Lx = (w - kernel_size[1]) // stride[1] + 1
+
+        if uf == 1 and df == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold = torch.nn.Fold(output_size=x.shape[2:], **fold_params)
+
+            weighting = self.get_weighting(kernel_size[0], kernel_size[1], Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h, w)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0], kernel_size[1], Ly * Lx))
+
+        elif uf > 1 and df == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold_params2 = dict(kernel_size=(kernel_size[0] * uf, kernel_size[0] * uf),
+                                dilation=1, padding=0,
+                                stride=(stride[0] * uf, stride[1] * uf))
+            fold = torch.nn.Fold(output_size=(x.shape[2] * uf, x.shape[3] * uf), **fold_params2)
+
+            weighting = self.get_weighting(kernel_size[0] * uf, kernel_size[1] * uf, Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h * uf, w * uf)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0] * uf, kernel_size[1] * uf, Ly * Lx))
+
+        elif df > 1 and uf == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold_params2 = dict(kernel_size=(kernel_size[0] // df, kernel_size[0] // df),
+                                dilation=1, padding=0,
+                                stride=(stride[0] // df, stride[1] // df))
+            fold = torch.nn.Fold(output_size=(x.shape[2] // df, x.shape[3] // df), **fold_params2)
+
+            weighting = self.get_weighting(kernel_size[0] // df, kernel_size[1] // df, Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h // df, w // df)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0] // df, kernel_size[1] // df, Ly * Lx))
+
+        else:
+            raise NotImplementedError
+
+        return fold, unfold, normalization, weighting
+
+    @torch.no_grad()
+    def get_input(self, batch, k, return_first_stage_outputs=False, force_c_encode=False,
+                  cond_key=None, return_original_cond=False, bs=None, return_x=False):
+        x = super().get_input(batch, k)
+        if bs is not None:
+            x = x[:bs]
+        x = x.to(self.device)
+        encoder_posterior = self.encode_first_stage(x)
+        z = self.get_first_stage_encoding(encoder_posterior).detach()
+
+        if self.model.conditioning_key is not None and not self.force_null_conditioning:
+            if cond_key is None:
+                cond_key = self.cond_stage_key
+            if cond_key != self.first_stage_key:
+                if cond_key in ['caption', 'coordinates_bbox', "txt"]:
+                    xc = batch[cond_key]
+                elif cond_key in ['class_label', 'cls']:
+                    xc = batch
+                else:
+                    xc = super().get_input(batch, cond_key).to(self.device)
+            else:
+                xc = x
+            if not self.cond_stage_trainable or force_c_encode:
+                if isinstance(xc, dict) or isinstance(xc, list):
+                    c = self.get_learned_conditioning(xc)
+                else:
+                    c = self.get_learned_conditioning(xc.to(self.device))
+            else:
+                c = xc
+            if bs is not None:
+                c = c[:bs]
+
+            if self.use_positional_encodings:
+                pos_x, pos_y = self.compute_latent_shifts(batch)
+                ckey = __conditioning_keys__[self.model.conditioning_key]
+                c = {ckey: c, 'pos_x': pos_x, 'pos_y': pos_y}
+
+        else:
+            c = None
+            xc = None
+            if self.use_positional_encodings:
+                pos_x, pos_y = self.compute_latent_shifts(batch)
+                c = {'pos_x': pos_x, 'pos_y': pos_y}
+        out = [z, c]
+        if return_first_stage_outputs:
+            xrec = self.decode_first_stage(z)
+            out.extend([x, xrec])
+        if return_x:
+            out.extend([x])
+        if return_original_cond:
+            out.append(xc)
+        return out
+
+    @torch.no_grad()
+    def decode_first_stage(self, z, predict_cids=False, force_not_quantize=False):
+        if predict_cids:
+            if z.dim() == 4:
+                z = torch.argmax(z.exp(), dim=1).long()
+            z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None)
+            z = rearrange(z, 'b h w c -> b c h w').contiguous()
+
+        z = 1. / self.scale_factor * z
+        return self.first_stage_model.decode(z)
+    
+    # 2023-04-08
+    def decode_first_stage_with_grad(self, z, predict_cids=False, force_not_quantize=False):
+        if predict_cids:
+            if z.dim() == 4:
+                z = torch.argmax(z.exp(), dim=1).long()
+            z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None)
+            z = rearrange(z, 'b h w c -> b c h w').contiguous()
+
+        z = 1. / self.scale_factor * z
+        return self.first_stage_model.decode(z)
+
+    @torch.no_grad()
+    def encode_first_stage(self, x):
+        return self.first_stage_model.encode(x)
+
+    def shared_step(self, batch, index, **kwargs):
+        x, c = self.get_input(batch, self.first_stage_key)
+        loss = self(x, c,index)
+        return loss
+
+    def forward(self, x, c, opt_index, *args, **kwargs):
+        T = (self.num_timesteps-1) * torch.ones(x.shape[0]).to(self.device).long()
+        t_max = round(self.num_timesteps*self.t_max) * torch.ones(x.shape[0]).to(self.device).long()
+        t = torch.randint(0, t_max[0], (x.shape[0],), device=self.device).long()
+        
+        if self.model.conditioning_key is not None:
+            assert c is not None
+            if self.cond_stage_trainable:
+                c = self.get_learned_conditioning(c)
+            if self.shorten_cond_schedule:  # TODO: drop this option
+                tc = self.cond_ids[t].to(self.device)
+                c = self.q_sample(x_start=c, t=tc, noise=torch.randn_like(c.float()))
+        return self.p_losses(x, c, t, T, t_max, opt_index, *args, **kwargs)
+
+    def apply_model(self, x_noisy, t, cond, return_ids=False):
+        if isinstance(cond, dict):
+            # hybrid case, cond is expected to be a dict
+            pass
+        else:
+            if not isinstance(cond, list):
+                cond = [cond]
+            key = 'c_concat' if self.model.conditioning_key == 'concat' else 'c_crossattn'
+            cond = {key: cond}
+
+        x_recon = self.model(x_noisy, t, **cond)
+
+        if isinstance(x_recon, tuple) and not return_ids:
+            return x_recon[0]
+        else:
+            return x_recon
+
+    def _predict_eps_from_xstart(self, x_t, t, pred_xstart):
+        return (extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t - pred_xstart) / \
+               extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape)
+
+    def _prior_bpd(self, x_start):
+        """
+        Get the prior KL term for the variational lower-bound, measured in
+        bits-per-dim.
+        This term can't be optimized, as it only depends on the encoder.
+        :param x_start: the [N x C x ...] tensor of inputs.
+        :return: a batch of [N] KL values (in bits), one per batch element.
+        """
+        batch_size = x_start.shape[0]
+        t = torch.tensor([self.num_timesteps - 1] * batch_size, device=x_start.device)
+        qt_mean, _, qt_log_variance = self.q_mean_variance(x_start, t)
+        kl_prior = normal_kl(mean1=qt_mean, logvar1=qt_log_variance, mean2=0.0, logvar2=0.0)
+        return mean_flat(kl_prior) / np.log(2.0)
+
+    def p_losses(self, x_start, cond, t, t_max, t_tao, opt_index, noise=None):
+        if opt_index == 0:
+            noise = default(noise, lambda: torch.randn_like(x_start))
+            x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+            model_output = self.apply_model(x_noisy, t, cond)
+
+            loss_dict = {}
+            prefix = 'train' if self.training else 'val'
+
+            if self.parameterization == "x0":
+                target = x_start
+            elif self.parameterization == "eps":
+                target = noise
+            elif self.parameterization == "v":
+                target = self.get_v(x_start, noise, t)
+            else:
+                raise NotImplementedError()
+
+            loss_simple = self.get_loss(model_output, target, mean=False).mean([1, 2, 3])
+            loss_dict.update({f'{prefix}/loss_simple': loss_simple.mean()})
+
+            logvar_t = self.logvar[t].to(self.device)
+            loss = loss_simple / torch.exp(logvar_t) + logvar_t
+            # loss = loss_simple / torch.exp(self.logvar) + self.logvar
+            if self.learn_logvar:
+                loss_dict.update({f'{prefix}/loss_gamma': loss.mean()})
+                loss_dict.update({'logvar': self.logvar.data.mean()})
+
+            loss = self.l_simple_weight * loss.mean()
+
+            loss_vlb = self.get_loss(model_output, target, mean=False).mean(dim=(1, 2, 3))
+            loss_vlb = (self.lvlb_weights[t] * loss_vlb).mean()
+            loss_dict.update({f'{prefix}/loss_vlb': loss_vlb})
+            loss += (self.original_elbo_weight * loss_vlb)
+            loss_dict.update({f'{prefix}/loss': loss})
+
+            return loss, loss_dict
+        else:
+            # train net
+            loss_dict = {}
+            prefix = 'train' if self.training else 'val'
+
+            noise2 = default(noise, lambda: torch.randn_like(x_start))
+            x_noisy = self.q_sample(x_start=x_start, t=t_max, noise=noise2)
+            model_output_noise = self.apply_model(x_noisy, t_max, cond)
+            
+            model_output_x0 = self.predict_start_from_noise(x_noisy, t=t_max, noise=model_output_noise)
+            
+            noise3 = default(noise, lambda: torch.randn_like(x_start))
+            model_output_noisy_tao = self.q_sample(x_start=model_output_x0, t=t_tao, noise=noise3)
+
+            model_output_noise_from_tao = self.apply_model(model_output_noisy_tao, t_tao, cond)
+            model_output_x0_from_tao = self.predict_start_from_noise(model_output_noisy_tao, t=t_tao, noise=model_output_noise_from_tao) # predict x0 from tao
+
+            loss_x0 = self.get_loss(model_output_x0, x_start, mean=False).mean([1, 2, 3])
+            loss_noise_from_tao = self.get_loss(model_output_noise_from_tao, noise3, mean=False).mean([1, 2, 3])
+            loss_x0_from_tao = self.get_loss(model_output_x0_from_tao, x_start, mean=False).mean([1, 2, 3])
+            loss_dict.update({f'{prefix}/loss_x0': loss_x0.mean()})
+            loss_dict.update({f'{prefix}/loss_x0_from_tao': loss_x0_from_tao.mean()})
+            loss_dict.update({f'{prefix}/loss_noise_from_tao': loss_noise_from_tao.mean()})
+            loss = loss_x0.mean() + loss_x0_from_tao.mean() + loss_noise_from_tao.mean()
+            loss_dict.update({f'{prefix}/loss_net': loss})
+
+            return loss, loss_dict
+
+    def p_mean_variance(self, x, c, t, clip_denoised: bool, return_codebook_ids=False, quantize_denoised=False,
+                        return_x0=False, score_corrector=None, corrector_kwargs=None):
+        t_in = t
+        model_out = self.apply_model(x, t_in, c, return_ids=return_codebook_ids)
+
+        if score_corrector is not None:
+            assert self.parameterization == "eps"
+            model_out = score_corrector.modify_score(self, model_out, x, t, c, **corrector_kwargs)
+
+        if return_codebook_ids:
+            model_out, logits = model_out
+
+        if self.parameterization == "eps":
+            x_recon = self.predict_start_from_noise(x, t=t, noise=model_out)
+        elif self.parameterization == "x0":
+            x_recon = model_out
+        else:
+            raise NotImplementedError()
+
+        if clip_denoised:
+            x_recon.clamp_(-1., 1.)
+        if quantize_denoised:
+            x_recon, _, [_, _, indices] = self.first_stage_model.quantize(x_recon)
+        model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
+        if return_codebook_ids:
+            return model_mean, posterior_variance, posterior_log_variance, logits
+        elif return_x0:
+            return model_mean, posterior_variance, posterior_log_variance, x_recon
+        else:
+            return model_mean, posterior_variance, posterior_log_variance
+
+    @torch.no_grad()
+    def p_sample(self, x, c, t, clip_denoised=False, repeat_noise=False,
+                 return_codebook_ids=False, quantize_denoised=False, return_x0=False,
+                 temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None):
+        b, *_, device = *x.shape, x.device
+        outputs = self.p_mean_variance(x=x, c=c, t=t, clip_denoised=clip_denoised,
+                                       return_codebook_ids=return_codebook_ids,
+                                       quantize_denoised=quantize_denoised,
+                                       return_x0=return_x0,
+                                       score_corrector=score_corrector, corrector_kwargs=corrector_kwargs)
+        if return_codebook_ids:
+            raise DeprecationWarning("Support dropped.")
+            model_mean, _, model_log_variance, logits = outputs
+        elif return_x0:
+            model_mean, _, model_log_variance, x0 = outputs
+        else:
+            model_mean, _, model_log_variance = outputs
+
+        noise = noise_like(x.shape, device, repeat_noise) * temperature
+        if noise_dropout > 0.:
+            noise = torch.nn.functional.dropout(noise, p=noise_dropout)
+        # no noise when t == 0
+        nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1)))
+
+        if return_codebook_ids:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, logits.argmax(dim=1)
+        if return_x0:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, x0
+        else:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise
+
+    @torch.no_grad()
+    def progressive_denoising(self, cond, shape, verbose=True, callback=None, quantize_denoised=False,
+                              img_callback=None, mask=None, x0=None, temperature=1., noise_dropout=0.,
+                              score_corrector=None, corrector_kwargs=None, batch_size=None, x_T=None, start_T=None,
+                              log_every_t=None):
+        if not log_every_t:
+            log_every_t = self.log_every_t
+        timesteps = self.num_timesteps
+        if batch_size is not None:
+            b = batch_size if batch_size is not None else shape[0]
+            shape = [batch_size] + list(shape)
+        else:
+            b = batch_size = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=self.device)
+        else:
+            img = x_T
+        intermediates = []
+        if cond is not None:
+            if isinstance(cond, dict):
+                cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else
+                list(map(lambda x: x[:batch_size], cond[key])) for key in cond}
+            else:
+                cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size]
+
+        if start_T is not None:
+            timesteps = min(timesteps, start_T)
+        iterator = tqdm(reversed(range(0, timesteps)), desc='Progressive Generation',
+                        total=timesteps) if verbose else reversed(
+            range(0, timesteps))
+        if type(temperature) == float:
+            temperature = [temperature] * timesteps
+
+        for i in iterator:
+            ts = torch.full((b,), i, device=self.device, dtype=torch.long)
+            if self.shorten_cond_schedule:
+                assert self.model.conditioning_key != 'hybrid'
+                tc = self.cond_ids[ts].to(cond.device)
+                cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond))
+
+            img, x0_partial = self.p_sample(img, cond, ts,
+                                            clip_denoised=self.clip_denoised,
+                                            quantize_denoised=quantize_denoised, return_x0=True,
+                                            temperature=temperature[i], noise_dropout=noise_dropout,
+                                            score_corrector=score_corrector, corrector_kwargs=corrector_kwargs)
+            if mask is not None:
+                assert x0 is not None
+                img_orig = self.q_sample(x0, ts)
+                img = img_orig * mask + (1. - mask) * img
+
+            if i % log_every_t == 0 or i == timesteps - 1:
+                intermediates.append(x0_partial)
+            if callback: callback(i)
+            if img_callback: img_callback(img, i)
+        return img, intermediates
+
+    @torch.no_grad()
+    def p_sample_loop(self, cond, shape, return_intermediates=False,
+                      x_T=None, verbose=True, callback=None, timesteps=None, quantize_denoised=False,
+                      mask=None, x0=None, img_callback=None, start_T=None,
+                      log_every_t=None):
+
+        if not log_every_t:
+            log_every_t = self.log_every_t
+        device = self.betas.device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+
+        intermediates = [img]
+        if timesteps is None:
+            timesteps = self.num_timesteps
+
+        if start_T is not None:
+            timesteps = min(timesteps, start_T)
+        iterator = tqdm(reversed(range(0, timesteps)), desc='Sampling t', total=timesteps) if verbose else reversed(
+            range(0, timesteps))
+
+        if mask is not None:
+            assert x0 is not None
+            assert x0.shape[2:3] == mask.shape[2:3]  # spatial size has to match
+
+        for i in iterator:
+            ts = torch.full((b,), i, device=device, dtype=torch.long)
+            if self.shorten_cond_schedule:
+                assert self.model.conditioning_key != 'hybrid'
+                tc = self.cond_ids[ts].to(cond.device)
+                cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond))
+
+            img = self.p_sample(img, cond, ts,
+                                clip_denoised=self.clip_denoised,
+                                quantize_denoised=quantize_denoised)
+            if mask is not None:
+                img_orig = self.q_sample(x0, ts)
+                img = img_orig * mask + (1. - mask) * img
+
+            if i % log_every_t == 0 or i == timesteps - 1:
+                intermediates.append(img)
+            if callback: callback(i)
+            if img_callback: img_callback(img, i)
+
+        if return_intermediates:
+            return img, intermediates
+        return img
+
+    @torch.no_grad()
+    def sample(self, cond, batch_size=16, return_intermediates=False, x_T=None,
+               verbose=True, timesteps=None, quantize_denoised=False,
+               mask=None, x0=None, shape=None, **kwargs):
+        if shape is None:
+            shape = (batch_size, self.channels, self.image_size, self.image_size)
+        if cond is not None:
+            if isinstance(cond, dict):
+                cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else
+                list(map(lambda x: x[:batch_size], cond[key])) for key in cond}
+            else:
+                cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size]
+        return self.p_sample_loop(cond,
+                                  shape,
+                                  return_intermediates=return_intermediates, x_T=x_T,
+                                  verbose=verbose, timesteps=timesteps, quantize_denoised=quantize_denoised,
+                                  mask=mask, x0=x0)
+
+    @torch.no_grad()
+    def sample_log(self, cond, batch_size, ddim, ddim_steps, **kwargs):
+        if ddim:
+            ddim_sampler = DDIMSampler(self)
+            shape = (self.channels, self.image_size, self.image_size)
+            samples, intermediates = ddim_sampler.sample(ddim_steps, batch_size,
+                                                         shape, cond, verbose=False, **kwargs)
+
+        else:
+            samples, intermediates = self.sample(cond=cond, batch_size=batch_size,
+                                                 return_intermediates=True, **kwargs)
+
+        return samples, intermediates
+
+    @torch.no_grad()
+    def get_unconditional_conditioning(self, batch_size, null_label=None):
+        if null_label is not None:
+            xc = null_label
+            if isinstance(xc, ListConfig):
+                xc = list(xc)
+            if isinstance(xc, dict) or isinstance(xc, list):
+                c = self.get_learned_conditioning(xc)
+            else:
+                if hasattr(xc, "to"):
+                    xc = xc.to(self.device)
+                c = self.get_learned_conditioning(xc)
+        else:
+            if self.cond_stage_key in ["class_label", "cls"]:
+                xc = self.cond_stage_model.get_unconditional_conditioning(batch_size, device=self.device)
+                return self.get_learned_conditioning(xc)
+            else:
+                raise NotImplementedError("todo")
+        if isinstance(c, list):  # in case the encoder gives us a list
+            for i in range(len(c)):
+                c[i] = repeat(c[i], '1 ... -> b ...', b=batch_size).to(self.device)
+        else:
+            c = repeat(c, '1 ... -> b ...', b=batch_size).to(self.device)
+        return c
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=50, ddim_eta=0., return_keys=None,
+                   quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True,
+                   plot_diffusion_rows=True, unconditional_guidance_scale=1., unconditional_guidance_label=None,
+                   use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc = self.get_input(batch, self.first_stage_key,
+                                           return_first_stage_outputs=True,
+                                           force_c_encode=True,
+                                           return_original_cond=True,
+                                           bs=N)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', "cls"]:
+                try:
+                    xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                    log['conditioning'] = xc
+                except KeyError:
+                    # probably no "human_label" in batch
+                    pass
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+            if quantize_denoised and not isinstance(self.first_stage_model, AutoencoderKL) and not isinstance(
+                    self.first_stage_model, IdentityFirstStage):
+                # also display when quantizing x0 while sampling
+                with ema_scope("Plotting Quantized Denoised"):
+                    samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                             ddim_steps=ddim_steps, eta=ddim_eta,
+                                                             quantize_denoised=True)
+                    # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True,
+                    #                                      quantize_denoised=True)
+                x_samples = self.decode_first_stage(samples.to(self.device))
+                log["samples_x0_quantized"] = x_samples
+
+        if unconditional_guidance_scale > 1.0:
+            uc = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            if self.model.conditioning_key == "crossattn-adm":
+                uc = {"c_crossattn": [uc], "c_adm": c["c_adm"]}
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        if inpaint:
+            # make a simple center square
+            b, h, w = z.shape[0], z.shape[2], z.shape[3]
+            mask = torch.ones(N, h, w).to(self.device)
+            # zeros will be filled in
+            mask[:, h // 4:3 * h // 4, w // 4:3 * w // 4] = 0.
+            mask = mask[:, None, ...]
+            with ema_scope("Plotting Inpaint"):
+                samples, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim, eta=ddim_eta,
+                                             ddim_steps=ddim_steps, x0=z[:N], mask=mask)
+            x_samples = self.decode_first_stage(samples.to(self.device))
+            log["samples_inpainting"] = x_samples
+            log["mask"] = mask
+
+            # outpaint
+            mask = 1. - mask
+            with ema_scope("Plotting Outpaint"):
+                samples, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim, eta=ddim_eta,
+                                             ddim_steps=ddim_steps, x0=z[:N], mask=mask)
+            x_samples = self.decode_first_stage(samples.to(self.device))
+            log["samples_outpainting"] = x_samples
+
+        if plot_progressive_rows:
+            with ema_scope("Plotting Progressives"):
+                img, progressives = self.progressive_denoising(c,
+                                                               shape=(self.channels, self.image_size, self.image_size),
+                                                               batch_size=N)
+            prog_row = self._get_denoise_row_from_list(progressives, desc="Progressive Generation")
+            log["progressive_row"] = prog_row
+
+        if return_keys:
+            if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0:
+                return log
+            else:
+                return {key: log[key] for key in return_keys}
+        return log
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params = list(self.model.parameters())
+        lr = self.learning_rate
+        params_net = list(self.model.parameters())
+
+        if self.cond_stage_trainable:
+            print(f"{self.__class__.__name__}: Also optimizing conditioner params!")
+            params = params + list(self.cond_stage_model.parameters())
+            params_net = params_net + list(self.cond_stage_model.parameters())
+        if self.learn_logvar:
+            print('Diffusion model optimizing logvar')
+            params.append(self.logvar)
+            params_net.append(self.logvar)
+        opt = torch.optim.AdamW(params, lr=lr)
+        opt_net = torch.optim.AdamW(params_net, lr=lr)
+        if self.use_scheduler:
+            assert 'target' in self.scheduler_config
+            scheduler = instantiate_from_config(self.scheduler_config)
+
+            print("Setting up LambdaLR scheduler...")
+            scheduler = [
+                {
+                    'scheduler': LambdaLR(opt, lr_lambda=scheduler.schedule),
+                    'interval': 'step',
+                    'frequency': 1
+                }]
+            return [opt, opt_net], scheduler
+        return opt, opt_net
+
+    @torch.no_grad()
+    def to_rgb(self, x):
+        x = x.float()
+        if not hasattr(self, "colorize"):
+            self.colorize = torch.randn(3, x.shape[1], 1, 1).to(x)
+        x = nn.functional.conv2d(x, weight=self.colorize)
+        x = 2. * (x - x.min()) / (x.max() - x.min()) - 1.
+        return x
+
+
+class DiffusionWrapper(pl.LightningModule):
+    def __init__(self, diff_model_config, conditioning_key):
+        super().__init__()
+        self.sequential_cross_attn = diff_model_config.pop("sequential_crossattn", False)
+        self.diffusion_model = instantiate_from_config(diff_model_config)
+        self.conditioning_key = conditioning_key
+        assert self.conditioning_key in [None, 'concat', 'crossattn', 'hybrid', 'adm', 'hybrid-adm', 'crossattn-adm']
+
+    def forward(self, x, t, c_concat: list = None, c_crossattn: list = None, c_adm=None):
+        if self.conditioning_key is None:
+            out = self.diffusion_model(x, t)
+        elif self.conditioning_key == 'concat':
+            xc = torch.cat([x] + c_concat, dim=1)
+            out = self.diffusion_model(xc, t)
+        elif self.conditioning_key == 'crossattn':
+            if not self.sequential_cross_attn:
+                cc = torch.cat(c_crossattn, 1)
+            else:
+                cc = c_crossattn
+            out = self.diffusion_model(x, t, context=cc)
+        elif self.conditioning_key == 'hybrid':
+            xc = torch.cat([x] + c_concat, dim=1)
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(xc, t, context=cc)
+        elif self.conditioning_key == 'hybrid-adm':
+            assert c_adm is not None
+            xc = torch.cat([x] + c_concat, dim=1)
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(xc, t, context=cc, y=c_adm)
+        elif self.conditioning_key == 'crossattn-adm':
+            assert c_adm is not None
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(x, t, context=cc, y=c_adm)
+        elif self.conditioning_key == 'adm':
+            cc = c_crossattn[0]
+            out = self.diffusion_model(x, t, y=cc)
+        else:
+            raise NotImplementedError()
+
+        return out
+
+
+class LatentUpscaleDiffusion(LatentDiffusion):
+    def __init__(self, *args, low_scale_config, low_scale_key="LR", noise_level_key=None, **kwargs):
+        super().__init__(*args, **kwargs)
+        # assumes that neither the cond_stage nor the low_scale_model contain trainable params
+        assert not self.cond_stage_trainable
+        self.instantiate_low_stage(low_scale_config)
+        self.low_scale_key = low_scale_key
+        self.noise_level_key = noise_level_key
+
+    def instantiate_low_stage(self, config):
+        model = instantiate_from_config(config)
+        self.low_scale_model = model.eval()
+        self.low_scale_model.train = disabled_train
+        for param in self.low_scale_model.parameters():
+            param.requires_grad = False
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, log_mode=False):
+        if not log_mode:
+            z, c = super().get_input(batch, k, force_c_encode=True, bs=bs)
+        else:
+            z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                                  force_c_encode=True, return_original_cond=True, bs=bs)
+        x_low = batch[self.low_scale_key][:bs]
+        x_low = rearrange(x_low, 'b h w c -> b c h w')
+        x_low = x_low.to(memory_format=torch.contiguous_format).float()
+        zx, noise_level = self.low_scale_model(x_low)
+        if self.noise_level_key is not None:
+            # get noise level from batch instead, e.g. when extracting a custom noise level for bsr
+            raise NotImplementedError('TODO')
+
+        all_conds = {"c_concat": [zx], "c_crossattn": [c], "c_adm": noise_level}
+        if log_mode:
+            # TODO: maybe disable if too expensive
+            x_low_rec = self.low_scale_model.decode(zx)
+            return z, all_conds, x, xrec, xc, x_low, x_low_rec, noise_level
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=200, ddim_eta=1., return_keys=None,
+                   plot_denoise_rows=False, plot_progressive_rows=True, plot_diffusion_rows=True,
+                   unconditional_guidance_scale=1., unconditional_guidance_label=None, use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc, x_low, x_low_rec, noise_level = self.get_input(batch, self.first_stage_key, bs=N,
+                                                                          log_mode=True)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        log["x_lr"] = x_low
+        log[f"x_lr_rec_@noise_levels{'-'.join(map(lambda x: str(x), list(noise_level.cpu().numpy())))}"] = x_low_rec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', 'cls']:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                log['conditioning'] = xc
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+        if unconditional_guidance_scale > 1.0:
+            uc_tmp = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            # TODO explore better "unconditional" choices for the other keys
+            # maybe guide away from empty text label and highest noise level and maximally degraded zx?
+            uc = dict()
+            for k in c:
+                if k == "c_crossattn":
+                    assert isinstance(c[k], list) and len(c[k]) == 1
+                    uc[k] = [uc_tmp]
+                elif k == "c_adm":  # todo: only run with text-based guidance?
+                    assert isinstance(c[k], torch.Tensor)
+                    #uc[k] = torch.ones_like(c[k]) * self.low_scale_model.max_noise_level
+                    uc[k] = c[k]
+                elif isinstance(c[k], list):
+                    uc[k] = [c[k][i] for i in range(len(c[k]))]
+                else:
+                    uc[k] = c[k]
+
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        if plot_progressive_rows:
+            with ema_scope("Plotting Progressives"):
+                img, progressives = self.progressive_denoising(c,
+                                                               shape=(self.channels, self.image_size, self.image_size),
+                                                               batch_size=N)
+            prog_row = self._get_denoise_row_from_list(progressives, desc="Progressive Generation")
+            log["progressive_row"] = prog_row
+
+        return log
+
+
+class LatentFinetuneDiffusion(LatentDiffusion):
+    """
+         Basis for different finetunas, such as inpainting or depth2image
+         To disable finetuning mode, set finetune_keys to None
+    """
+
+    def __init__(self,
+                 concat_keys: tuple,
+                 finetune_keys=("model.diffusion_model.input_blocks.0.0.weight",
+                                "model_ema.diffusion_modelinput_blocks00weight"
+                                ),
+                 keep_finetune_dims=4,
+                 # if model was trained without concat mode before and we would like to keep these channels
+                 c_concat_log_start=None,  # to log reconstruction of c_concat codes
+                 c_concat_log_end=None,
+                 *args, **kwargs
+                 ):
+        ckpt_path = kwargs.pop("ckpt_path", None)
+        ignore_keys = kwargs.pop("ignore_keys", list())
+        super().__init__(*args, **kwargs)
+        self.finetune_keys = finetune_keys
+        self.concat_keys = concat_keys
+        self.keep_dims = keep_finetune_dims
+        self.c_concat_log_start = c_concat_log_start
+        self.c_concat_log_end = c_concat_log_end
+        if exists(self.finetune_keys): assert exists(ckpt_path), 'can only finetune from a given checkpoint'
+        if exists(ckpt_path):
+            self.init_from_ckpt(ckpt_path, ignore_keys)
+
+    def init_from_ckpt(self, path, ignore_keys=list(), only_model=False):
+        sd = torch.load(path, map_location="cpu")
+        if "state_dict" in list(sd.keys()):
+            sd = sd["state_dict"]
+        keys = list(sd.keys())
+        for k in keys:
+            for ik in ignore_keys:
+                if k.startswith(ik):
+                    print("Deleting key {} from state_dict.".format(k))
+                    del sd[k]
+
+            # make it explicit, finetune by including extra input channels
+            if exists(self.finetune_keys) and k in self.finetune_keys:
+                new_entry = None
+                for name, param in self.named_parameters():
+                    if name in self.finetune_keys:
+                        print(
+                            f"modifying key '{name}' and keeping its original {self.keep_dims} (channels) dimensions only")
+                        new_entry = torch.zeros_like(param)  # zero init
+                assert exists(new_entry), 'did not find matching parameter to modify'
+                new_entry[:, :self.keep_dims, ...] = sd[k]
+                sd[k] = new_entry
+
+        missing, unexpected = self.load_state_dict(sd, strict=False) if not only_model else self.model.load_state_dict(
+            sd, strict=False)
+        print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys")
+        if len(missing) > 0:
+            print(f"Missing Keys: {missing}")
+        if len(unexpected) > 0:
+            print(f"Unexpected Keys: {unexpected}")
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=200, ddim_eta=1., return_keys=None,
+                   quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True,
+                   plot_diffusion_rows=True, unconditional_guidance_scale=1., unconditional_guidance_label=None,
+                   use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc = self.get_input(batch, self.first_stage_key, bs=N, return_first_stage_outputs=True)
+        c_cat, c = c["c_concat"][0], c["c_crossattn"][0]
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', 'cls']:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                log['conditioning'] = xc
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if not (self.c_concat_log_start is None and self.c_concat_log_end is None):
+            log["c_concat_decoded"] = self.decode_first_stage(c_cat[:, self.c_concat_log_start:self.c_concat_log_end])
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]},
+                                                         batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+        if unconditional_guidance_scale > 1.0:
+            uc_cross = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            uc_cat = c_cat
+            uc_full = {"c_concat": [uc_cat], "c_crossattn": [uc_cross]}
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]},
+                                                 batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc_full,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        return log
+
+
+class LatentInpaintDiffusion(LatentFinetuneDiffusion):
+    """
+    can either run as pure inpainting model (only concat mode) or with mixed conditionings,
+    e.g. mask as concat and text via cross-attn.
+    To disable finetuning mode, set finetune_keys to None
+     """
+
+    def __init__(self,
+                 concat_keys=("mask", "masked_image"),
+                 masked_image_key="masked_image",
+                 *args, **kwargs
+                 ):
+        super().__init__(concat_keys, *args, **kwargs)
+        self.masked_image_key = masked_image_key
+        assert self.masked_image_key in concat_keys
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for inpainting'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        c_cat = list()
+        for ck in self.concat_keys:
+            cc = rearrange(batch[ck], 'b h w c -> b c h w').to(memory_format=torch.contiguous_format).float()
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            bchw = z.shape
+            if ck != self.masked_image_key:
+                cc = torch.nn.functional.interpolate(cc, size=bchw[-2:])
+            else:
+                cc = self.get_first_stage_encoding(self.encode_first_stage(cc))
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super(LatentInpaintDiffusion, self).log_images(*args, **kwargs)
+        log["masked_image"] = rearrange(args[0]["masked_image"],
+                                        'b h w c -> b c h w').to(memory_format=torch.contiguous_format).float()
+        return log
+
+
+class LatentDepth2ImageDiffusion(LatentFinetuneDiffusion):
+    """
+    condition on monocular depth estimation
+    """
+
+    def __init__(self, depth_stage_config, concat_keys=("midas_in",), *args, **kwargs):
+        super().__init__(concat_keys=concat_keys, *args, **kwargs)
+        self.depth_model = instantiate_from_config(depth_stage_config)
+        self.depth_stage_key = concat_keys[0]
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for depth2img'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        assert len(self.concat_keys) == 1
+        c_cat = list()
+        for ck in self.concat_keys:
+            cc = batch[ck]
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            cc = self.depth_model(cc)
+            cc = torch.nn.functional.interpolate(
+                cc,
+                size=z.shape[2:],
+                mode="bicubic",
+                align_corners=False,
+            )
+
+            depth_min, depth_max = torch.amin(cc, dim=[1, 2, 3], keepdim=True), torch.amax(cc, dim=[1, 2, 3],
+                                                                                           keepdim=True)
+            cc = 2. * (cc - depth_min) / (depth_max - depth_min + 0.001) - 1.
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super().log_images(*args, **kwargs)
+        depth = self.depth_model(args[0][self.depth_stage_key])
+        depth_min, depth_max = torch.amin(depth, dim=[1, 2, 3], keepdim=True), \
+                               torch.amax(depth, dim=[1, 2, 3], keepdim=True)
+        log["depth"] = 2. * (depth - depth_min) / (depth_max - depth_min) - 1.
+        return log
+
+
+class LatentUpscaleFinetuneDiffusion(LatentFinetuneDiffusion):
+    """
+        condition on low-res image (and optionally on some spatial noise augmentation)
+    """
+    def __init__(self, concat_keys=("lr",), reshuffle_patch_size=None,
+                 low_scale_config=None, low_scale_key=None, *args, **kwargs):
+        super().__init__(concat_keys=concat_keys, *args, **kwargs)
+        self.reshuffle_patch_size = reshuffle_patch_size
+        self.low_scale_model = None
+        if low_scale_config is not None:
+            print("Initializing a low-scale model")
+            assert exists(low_scale_key)
+            self.instantiate_low_stage(low_scale_config)
+            self.low_scale_key = low_scale_key
+
+    def instantiate_low_stage(self, config):
+        model = instantiate_from_config(config)
+        self.low_scale_model = model.eval()
+        self.low_scale_model.train = disabled_train
+        for param in self.low_scale_model.parameters():
+            param.requires_grad = False
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for upscaling-ft'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        assert len(self.concat_keys) == 1
+        # optionally make spatial noise_level here
+        c_cat = list()
+        noise_level = None
+        for ck in self.concat_keys:
+            cc = batch[ck]
+            cc = rearrange(cc, 'b h w c -> b c h w')
+            if exists(self.reshuffle_patch_size):
+                assert isinstance(self.reshuffle_patch_size, int)
+                cc = rearrange(cc, 'b c (p1 h) (p2 w) -> b (p1 p2 c) h w',
+                               p1=self.reshuffle_patch_size, p2=self.reshuffle_patch_size)
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            if exists(self.low_scale_model) and ck == self.low_scale_key:
+                cc, noise_level = self.low_scale_model(cc)
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        if exists(noise_level):
+            all_conds = {"c_concat": [c_cat], "c_crossattn": [c], "c_adm": noise_level}
+        else:
+            all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super().log_images(*args, **kwargs)
+        log["lr"] = rearrange(args[0]["lr"], 'b h w c -> b c h w')
+        return log
diff --git a/ldm/models/diffusion/ddpm_ccsr_stage2.py b/ldm/models/diffusion/ddpm_ccsr_stage2.py
new file mode 100644
index 0000000..aa766f6
--- /dev/null
+++ b/ldm/models/diffusion/ddpm_ccsr_stage2.py
@@ -0,0 +1,2058 @@
+"""
+wild mixture of
+https://github.com/lucidrains/denoising-diffusion-pytorch/blob/7706bdfc6f527f58d33f84b7b522e61e6e3164b3/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
+https://github.com/openai/improved-diffusion/blob/e94489283bb876ac1477d5dd7709bbbd2d9902ce/improved_diffusion/gaussian_diffusion.py
+https://github.com/CompVis/taming-transformers
+-- merci
+"""
+
+import torch
+import torch.nn as nn
+import numpy as np
+import pytorch_lightning as pl
+from torch.optim.lr_scheduler import LambdaLR
+from einops import rearrange, repeat
+from contextlib import contextmanager, nullcontext
+from functools import partial
+import itertools
+from tqdm import tqdm
+from torchvision.utils import make_grid
+from pytorch_lightning.utilities.distributed import rank_zero_only
+from omegaconf import ListConfig
+
+from ldm.util import log_txt_as_img, exists, default, ismap, isimage, mean_flat, count_params, instantiate_from_config
+from ldm.modules.ema import LitEma
+from ldm.modules.distributions.distributions import normal_kl, DiagonalGaussianDistribution
+from ldm.models.autoencoder import IdentityFirstStage, AutoencoderKL
+from ldm.modules.diffusionmodules.util import make_beta_schedule, extract_into_tensor, noise_like
+from ldm.models.diffusion.ddim import DDIMSampler
+from model.mixins import ImageLoggerMixin
+
+from model.q_sampler import space_timesteps
+
+
+__conditioning_keys__ = {'concat': 'c_concat',
+                         'crossattn': 'c_crossattn',
+                         'adm': 'y'}
+
+
+def disabled_train(self, mode=True):
+    """Overwrite model.train with this function to make sure train/eval mode
+    does not change anymore."""
+    return self
+
+
+def uniform_on_device(r1, r2, shape, device):
+    return (r1 - r2) * torch.rand(*shape, device=device) + r2
+
+
+class DDPM(pl.LightningModule, ImageLoggerMixin):
+    # classic DDPM with Gaussian diffusion, in image space
+    def __init__(self,
+                 unet_config,
+                 timesteps=1000,
+                 t_max=0.6667,
+                 t_min=0.3333,
+                 beta_schedule="linear",
+                 loss_type="l2",
+                 ckpt_path=None,
+                 ignore_keys=[],
+                 load_only_unet=False,
+                 monitor="val/loss",
+                 use_ema=True,
+                 first_stage_key="image",
+                 image_size=256,
+                 channels=3,
+                 log_every_t=100,
+                 clip_denoised=True,
+                 linear_start=1e-4,
+                 linear_end=2e-2,
+                 cosine_s=8e-3,
+                 given_betas=None,
+                 original_elbo_weight=0.,
+                 v_posterior=0.,  # weight for choosing posterior variance as sigma = (1-v) * beta_tilde + v * beta
+                 l_simple_weight=1.,
+                 conditioning_key=None,
+                 parameterization="eps",  # all assuming fixed variance schedules
+                 scheduler_config=None,
+                 use_positional_encodings=False,
+                 learn_logvar=False,
+                 logvar_init=0.,
+                 make_it_fit=False,
+                 ucg_training=None,
+                 reset_ema=False,
+                 reset_num_ema_updates=False,
+                 ):
+        super().__init__()
+        assert parameterization in ["eps", "x0", "v"], 'currently only supporting "eps" and "x0" and "v"'
+        self.parameterization = parameterization
+        print(f"{self.__class__.__name__}: Running in {self.parameterization}-prediction mode")
+        self.cond_stage_model = None
+        self.clip_denoised = clip_denoised
+        self.log_every_t = log_every_t
+        self.first_stage_key = first_stage_key
+        self.image_size = image_size  # try conv?
+        self.channels = channels
+        self.use_positional_encodings = use_positional_encodings
+        self.model = DiffusionWrapper(unet_config, conditioning_key)
+        count_params(self.model, verbose=True)
+        self.use_ema = use_ema
+        if self.use_ema:
+            self.model_ema = LitEma(self.model)
+            print(f"Keeping EMAs of {len(list(self.model_ema.buffers()))}.")
+
+        self.use_scheduler = scheduler_config is not None
+        if self.use_scheduler:
+            self.scheduler_config = scheduler_config
+
+        self.v_posterior = v_posterior
+        self.original_elbo_weight = original_elbo_weight
+        self.l_simple_weight = l_simple_weight
+
+        if monitor is not None:
+            self.monitor = monitor
+        self.make_it_fit = make_it_fit
+        if reset_ema: assert exists(ckpt_path)
+        if ckpt_path is not None:
+            self.init_from_ckpt(ckpt_path, ignore_keys=ignore_keys, only_model=load_only_unet)
+            if reset_ema:
+                assert self.use_ema
+                print(f"Resetting ema to pure model weights. This is useful when restoring from an ema-only checkpoint.")
+                self.model_ema = LitEma(self.model)
+        if reset_num_ema_updates:
+            print(" +++++++++++ WARNING: RESETTING NUM_EMA UPDATES TO ZERO +++++++++++ ")
+            assert self.use_ema
+            self.model_ema.reset_num_updates()
+
+        self.register_schedule(given_betas=given_betas, beta_schedule=beta_schedule, timesteps=timesteps,
+                               linear_start=linear_start, linear_end=linear_end, cosine_s=cosine_s)
+        
+        self.t_max = t_max
+        self.t_min = t_min
+
+        timesteps_inter = 45
+        use_timesteps = set(space_timesteps(timesteps, [timesteps_inter]))
+        self.timesteps_inter = np.array(sorted(list(use_timesteps)), dtype=np.int32)
+        last_alpha_cumprod = 1.0
+        new_betas = []
+        timestep_map = []
+        for i, alpha_cumprod in enumerate(self.alphas_cumprod):
+            if i in use_timesteps:
+                new_betas.append(1 - alpha_cumprod / last_alpha_cumprod)
+                last_alpha_cumprod = alpha_cumprod
+                timestep_map.append(i)
+        new_betas = [beta.data.cpu().numpy() for beta in new_betas]
+        self.register_schedule_inter(given_betas=np.array(new_betas), timesteps=len(new_betas))
+
+        self.loss_type = loss_type
+
+        self.learn_logvar = learn_logvar
+        logvar = torch.full(fill_value=logvar_init, size=(self.num_timesteps,))
+        if self.learn_logvar:
+            self.logvar = nn.Parameter(self.logvar, requires_grad=True)
+        else:
+            self.register_buffer('logvar', logvar)
+
+        self.ucg_training = ucg_training or dict()
+        if self.ucg_training:
+            self.ucg_prng = np.random.RandomState()
+
+    def register_schedule(self, given_betas=None, beta_schedule="linear", timesteps=1000,
+                          linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+        if exists(given_betas):
+            betas = given_betas
+        else:
+            betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end,
+                                       cosine_s=cosine_s)
+        alphas = 1. - betas
+        alphas_cumprod = np.cumprod(alphas, axis=0)
+        alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])
+
+        timesteps, = betas.shape
+        self.num_timesteps = int(timesteps)
+        self.linear_start = linear_start
+        self.linear_end = linear_end
+        assert alphas_cumprod.shape[0] == self.num_timesteps, 'alphas have to be defined for each timestep'
+
+        to_torch = partial(torch.tensor, dtype=torch.float32)
+
+        self.register_buffer('betas', to_torch(betas))
+        self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
+        self.register_buffer('alphas_cumprod_prev', to_torch(alphas_cumprod_prev))
+
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
+        self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
+        self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod)))
+        self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod)))
+        self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod - 1)))
+
+        # calculations for posterior q(x_{t-1} | x_t, x_0)
+        posterior_variance = (1 - self.v_posterior) * betas * (1. - alphas_cumprod_prev) / (
+                1. - alphas_cumprod) + self.v_posterior * betas
+        # above: equal to 1. / (1. / (1. - alpha_cumprod_tm1) + alpha_t / beta_t)
+        self.register_buffer('posterior_variance', to_torch(posterior_variance))
+        # below: log calculation clipped because the posterior variance is 0 at the beginning of the diffusion chain
+        self.register_buffer('posterior_log_variance_clipped', to_torch(np.log(np.maximum(posterior_variance, 1e-20))))
+        self.register_buffer('posterior_mean_coef1', to_torch(
+            betas * np.sqrt(alphas_cumprod_prev) / (1. - alphas_cumprod)))
+        self.register_buffer('posterior_mean_coef2', to_torch(
+            (1. - alphas_cumprod_prev) * np.sqrt(alphas) / (1. - alphas_cumprod)))
+
+        if self.parameterization == "eps":
+            lvlb_weights = self.betas ** 2 / (
+                    2 * self.posterior_variance * to_torch(alphas) * (1 - self.alphas_cumprod))
+        elif self.parameterization == "x0":
+            lvlb_weights = 0.5 * np.sqrt(torch.Tensor(alphas_cumprod)) / (2. * 1 - torch.Tensor(alphas_cumprod))
+        elif self.parameterization == "v":
+            lvlb_weights = torch.ones_like(self.betas ** 2 / (
+                    2 * self.posterior_variance * to_torch(alphas) * (1 - self.alphas_cumprod)))
+        else:
+            raise NotImplementedError("mu not supported")
+        lvlb_weights[0] = lvlb_weights[1]
+        self.register_buffer('lvlb_weights', lvlb_weights, persistent=False)
+        assert not torch.isnan(self.lvlb_weights).all()
+
+    def register_schedule_inter(self, given_betas=None, beta_schedule="linear", timesteps=1000,
+                          linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+        if exists(given_betas):
+            betas = given_betas
+        else:
+            betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end,
+                                       cosine_s=cosine_s)
+        alphas = 1. - betas
+        alphas_cumprod = np.cumprod(alphas, axis=0)
+        alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])
+
+        timesteps, = betas.shape
+        self.num_timesteps_inter = int(timesteps)
+        self.linear_start_inter = linear_start
+        self.linear_end_inter = linear_end
+        assert alphas_cumprod.shape[0] == self.num_timesteps_inter, 'alphas have to be defined for each timestep'
+
+        to_torch = partial(torch.tensor, dtype=torch.float32)
+
+        self.register_buffer('betas_inter', to_torch(betas))
+        self.register_buffer('alphas_cumprod_inter', to_torch(alphas_cumprod))
+        self.register_buffer('alphas_cumprod_prev_inter', to_torch(alphas_cumprod_prev))
+
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.register_buffer('sqrt_alphas_cumprod_inter', to_torch(np.sqrt(alphas_cumprod)))
+        self.register_buffer('sqrt_one_minus_alphas_cumprod_inter', to_torch(np.sqrt(1. - alphas_cumprod)))
+        self.register_buffer('log_one_minus_alphas_cumprod_inter', to_torch(np.log(1. - alphas_cumprod)))
+        self.register_buffer('sqrt_recip_alphas_cumprod_inter', to_torch(np.sqrt(1. / alphas_cumprod)))
+        self.register_buffer('sqrt_recipm1_alphas_cumprod_inter', to_torch(np.sqrt(1. / alphas_cumprod - 1)))
+
+        # calculations for posterior q(x_{t-1} | x_t, x_0)
+        posterior_variance = (1 - self.v_posterior) * betas * (1. - alphas_cumprod_prev) / (
+                1. - alphas_cumprod) + self.v_posterior * betas
+        # above: equal to 1. / (1. / (1. - alpha_cumprod_tm1) + alpha_t / beta_t)
+        self.register_buffer('posterior_variance_inter', to_torch(posterior_variance))
+        # below: log calculation clipped because the posterior variance is 0 at the beginning of the diffusion chain
+        self.register_buffer('posterior_log_variance_clipped_inter', to_torch(np.log(np.maximum(posterior_variance, 1e-20))))
+        self.register_buffer('posterior_mean_coef1_inter', to_torch(
+            betas * np.sqrt(alphas_cumprod_prev) / (1. - alphas_cumprod)))
+        self.register_buffer('posterior_mean_coef2_inter', to_torch(
+            (1. - alphas_cumprod_prev) * np.sqrt(alphas) / (1. - alphas_cumprod)))
+
+        if self.parameterization == "eps":
+            lvlb_weights = self.betas_inter ** 2 / (
+                    2 * self.posterior_variance_inter * to_torch(alphas) * (1 - self.alphas_cumprod_inter))
+        elif self.parameterization == "x0":
+            lvlb_weights = 0.5 * np.sqrt(torch.Tensor(alphas_cumprod)) / (2. * 1 - torch.Tensor(alphas_cumprod_inter))
+        elif self.parameterization == "v":
+            lvlb_weights = torch.ones_like(self.betas_inter ** 2 / (
+                    2 * self.posterior_variance_inter * to_torch(alphas) * (1 - self.alphas_cumprod_inter)))
+        else:
+            raise NotImplementedError("mu not supported")
+        lvlb_weights[0] = lvlb_weights[1]
+        self.register_buffer('lvlb_weights_inter', lvlb_weights, persistent=False)
+        assert not torch.isnan(self.lvlb_weights_inter).all()
+
+    @contextmanager
+    def ema_scope(self, context=None):
+        if self.use_ema:
+            self.model_ema.store(self.model.parameters())
+            self.model_ema.copy_to(self.model)
+            if context is not None:
+                print(f"{context}: Switched to EMA weights")
+        try:
+            yield None
+        finally:
+            if self.use_ema:
+                self.model_ema.restore(self.model.parameters())
+                if context is not None:
+                    print(f"{context}: Restored training weights")
+
+    @torch.no_grad()
+    def init_from_ckpt(self, path, ignore_keys=list(), only_model=False):
+        sd = torch.load(path, map_location="cpu")
+        if "state_dict" in list(sd.keys()):
+            sd = sd["state_dict"]
+        keys = list(sd.keys())
+        for k in keys:
+            for ik in ignore_keys:
+                if k.startswith(ik):
+                    print("Deleting key {} from state_dict.".format(k))
+                    del sd[k]
+        if self.make_it_fit:
+            n_params = len([name for name, _ in
+                            itertools.chain(self.named_parameters(),
+                                            self.named_buffers())])
+            for name, param in tqdm(
+                    itertools.chain(self.named_parameters(),
+                                    self.named_buffers()),
+                    desc="Fitting old weights to new weights",
+                    total=n_params
+            ):
+                if not name in sd:
+                    continue
+                old_shape = sd[name].shape
+                new_shape = param.shape
+                assert len(old_shape) == len(new_shape)
+                if len(new_shape) > 2:
+                    # we only modify first two axes
+                    assert new_shape[2:] == old_shape[2:]
+                # assumes first axis corresponds to output dim
+                if not new_shape == old_shape:
+                    new_param = param.clone()
+                    old_param = sd[name]
+                    if len(new_shape) == 1:
+                        for i in range(new_param.shape[0]):
+                            new_param[i] = old_param[i % old_shape[0]]
+                    elif len(new_shape) >= 2:
+                        for i in range(new_param.shape[0]):
+                            for j in range(new_param.shape[1]):
+                                new_param[i, j] = old_param[i % old_shape[0], j % old_shape[1]]
+
+                        n_used_old = torch.ones(old_shape[1])
+                        for j in range(new_param.shape[1]):
+                            n_used_old[j % old_shape[1]] += 1
+                        n_used_new = torch.zeros(new_shape[1])
+                        for j in range(new_param.shape[1]):
+                            n_used_new[j] = n_used_old[j % old_shape[1]]
+
+                        n_used_new = n_used_new[None, :]
+                        while len(n_used_new.shape) < len(new_shape):
+                            n_used_new = n_used_new.unsqueeze(-1)
+                        new_param /= n_used_new
+
+                    sd[name] = new_param
+
+        missing, unexpected = self.load_state_dict(sd, strict=False) if not only_model else self.model.load_state_dict(
+            sd, strict=False)
+        print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys")
+        if len(missing) > 0:
+            print(f"Missing Keys:\n {missing}")
+        if len(unexpected) > 0:
+            print(f"\nUnexpected Keys:\n {unexpected}")
+
+    def q_mean_variance(self, x_start, t):
+        """
+        Get the distribution q(x_t | x_0).
+        :param x_start: the [N x C x ...] tensor of noiseless inputs.
+        :param t: the number of diffusion steps (minus 1). Here, 0 means one step.
+        :return: A tuple (mean, variance, log_variance), all of x_start's shape.
+        """
+        mean = (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start)
+        variance = extract_into_tensor(1.0 - self.alphas_cumprod, t, x_start.shape)
+        log_variance = extract_into_tensor(self.log_one_minus_alphas_cumprod, t, x_start.shape)
+        return mean, variance, log_variance
+
+    def predict_start_from_noise(self, x_t, t, noise):
+        return (
+                extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t -
+                extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * noise
+        )
+    
+    def predict_start_from_noise_inter(self, x_t, t, noise):
+        return (
+                extract_into_tensor(self.sqrt_recip_alphas_cumprod_inter, t, x_t.shape) * x_t -
+                extract_into_tensor(self.sqrt_recipm1_alphas_cumprod_inter, t, x_t.shape) * noise
+        )
+
+    def predict_start_from_z_and_v(self, x_t, t, v):
+        # self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
+        # self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x_t.shape) * x_t -
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_t.shape) * v
+        )
+
+    def predict_eps_from_z_and_v(self, x_t, t, v):
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x_t.shape) * v +
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_t.shape) * x_t
+        )
+
+    def q_posterior(self, x_start, x_t, t):
+        posterior_mean = (
+                extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start +
+                extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t
+        )
+        posterior_variance = extract_into_tensor(self.posterior_variance, t, x_t.shape)
+        posterior_log_variance_clipped = extract_into_tensor(self.posterior_log_variance_clipped, t, x_t.shape)
+        return posterior_mean, posterior_variance, posterior_log_variance_clipped
+    
+    def q_posterior_inter(self, x_start, x_t, t):
+        posterior_mean = (
+                extract_into_tensor(self.posterior_mean_coef1_inter, t, x_t.shape) * x_start +
+                extract_into_tensor(self.posterior_mean_coef2_inter, t, x_t.shape) * x_t
+        )
+        posterior_variance = extract_into_tensor(self.posterior_variance_inter, t, x_t.shape)
+        posterior_log_variance_clipped = extract_into_tensor(self.posterior_log_variance_clipped_inter, t, x_t.shape)
+        return posterior_mean, posterior_variance, posterior_log_variance_clipped
+
+    def p_mean_variance(self, x, t, clip_denoised: bool):
+        model_out = self.model(x, t)
+        if self.parameterization == "eps":
+            x_recon = self.predict_start_from_noise(x, t=t, noise=model_out)
+        elif self.parameterization == "x0":
+            x_recon = model_out
+        if clip_denoised:
+            x_recon.clamp_(-1., 1.)
+
+        model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
+        return model_mean, posterior_variance, posterior_log_variance
+
+    @torch.no_grad()
+    def p_sample(self, x, t, clip_denoised=True, repeat_noise=False):
+        b, *_, device = *x.shape, x.device
+        model_mean, _, model_log_variance = self.p_mean_variance(x=x, t=t, clip_denoised=clip_denoised)
+        noise = noise_like(x.shape, device, repeat_noise)
+        # no noise when t == 0
+        nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1)))
+        return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise
+
+    @torch.no_grad()
+    def p_sample_loop(self, shape, return_intermediates=False):
+        device = self.betas.device
+        b = shape[0]
+        img = torch.randn(shape, device=device)
+        intermediates = [img]
+        for i in tqdm(reversed(range(0, self.num_timesteps)), desc='Sampling t', total=self.num_timesteps):
+            img = self.p_sample(img, torch.full((b,), i, device=device, dtype=torch.long),
+                                clip_denoised=self.clip_denoised)
+            if i % self.log_every_t == 0 or i == self.num_timesteps - 1:
+                intermediates.append(img)
+        if return_intermediates:
+            return img, intermediates
+        return img
+
+    @torch.no_grad()
+    def sample(self, batch_size=16, return_intermediates=False):
+        image_size = self.image_size
+        channels = self.channels
+        return self.p_sample_loop((batch_size, channels, image_size, image_size),
+                                  return_intermediates=return_intermediates)
+
+    def q_sample(self, x_start, t, noise=None):
+        noise = default(noise, lambda: torch.randn_like(x_start))
+        return (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start +
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise)
+
+    def get_v(self, x, noise, t):
+        return (
+                extract_into_tensor(self.sqrt_alphas_cumprod, t, x.shape) * noise -
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x.shape) * x
+        )
+
+    def get_loss(self, pred, target, mean=True):
+        if self.loss_type == 'l1':
+            loss = (target - pred).abs()
+            if mean:
+                loss = loss.mean()
+        elif self.loss_type == 'l2':
+            if mean:
+                loss = torch.nn.functional.mse_loss(target, pred)
+            else:
+                loss = torch.nn.functional.mse_loss(target, pred, reduction='none')
+        else:
+            raise NotImplementedError("unknown loss type '{loss_type}'")
+
+        return loss
+
+    def p_losses(self, x_start, t, noise=None):
+        noise = default(noise, lambda: torch.randn_like(x_start))
+        x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+        model_out = self.model(x_noisy, t)
+
+        loss_dict = {}
+        if self.parameterization == "eps":
+            target = noise
+        elif self.parameterization == "x0":
+            target = x_start
+        elif self.parameterization == "v":
+            target = self.get_v(x_start, noise, t)
+        else:
+            raise NotImplementedError(f"Parameterization {self.parameterization} not yet supported")
+
+        loss = self.get_loss(model_out, target, mean=False).mean(dim=[1, 2, 3])
+
+        log_prefix = 'train' if self.training else 'val'
+
+        loss_dict.update({f'{log_prefix}/loss_simple': loss.mean()})
+        loss_simple = loss.mean() * self.l_simple_weight
+
+        loss_vlb = (self.lvlb_weights[t] * loss).mean()
+        loss_dict.update({f'{log_prefix}/loss_vlb': loss_vlb})
+
+        loss = loss_simple + self.original_elbo_weight * loss_vlb
+
+        loss_dict.update({f'{log_prefix}/loss': loss})
+
+        return loss, loss_dict
+
+    def forward(self, x, *args, **kwargs):
+        # b, c, h, w, device, img_size, = *x.shape, x.device, self.image_size
+        # assert h == img_size and w == img_size, f'height and width of image must be {img_size}'
+        t = torch.randint(0, self.num_timesteps, (x.shape[0],), device=self.device).long()
+        return self.p_losses(x, t, *args, **kwargs)
+
+    def get_input(self, batch, k):
+        x = batch[k]
+        if len(x.shape) == 3:
+            x = x[..., None]
+        x = rearrange(x, 'b h w c -> b c h w')
+        x = x.to(memory_format=torch.contiguous_format).float()
+        return x
+
+    def shared_step(self, batch):
+        x = self.get_input(batch, self.first_stage_key)
+        loss, loss_dict = self(x)
+        return loss, loss_dict
+
+    def training_step(self, batch, batch_idx, optimizer_idx):
+        for k in self.ucg_training:
+            p = self.ucg_training[k]["p"]
+            val = self.ucg_training[k]["val"]
+            if val is None:
+                val = ""
+            for i in range(len(batch[k])):
+                if self.ucg_prng.choice(2, p=[1 - p, p]):
+                    batch[k][i] = val
+
+        loss, loss_dict = self.shared_step(batch, optimizer_idx)
+
+        self.log_dict(loss_dict, prog_bar=True,
+                      logger=True, on_step=True, on_epoch=True)
+
+        self.log("global_step", self.global_step,
+                 prog_bar=True, logger=True, on_step=True, on_epoch=False)
+
+        if self.use_scheduler:
+            lr = self.optimizers().param_groups[0]['lr']
+            self.log('lr_abs', lr, prog_bar=True, logger=True, on_step=True, on_epoch=False)
+
+        return loss
+
+    @torch.no_grad()
+    def validation_step(self, batch, batch_idx):
+        _, loss_dict_no_ema_0 = self.shared_step(batch,0)
+        _, loss_dict_no_ema_1 = self.shared_step(batch,1)
+        with self.ema_scope():
+            _, loss_dict_ema_0 = self.shared_step(batch,0)
+            _, loss_dict_ema_1 = self.shared_step(batch,1)
+            loss_dict_ema_0 = {key + '_ema': loss_dict_ema_0[key] for key in loss_dict_ema_0}
+            loss_dict_ema_1 = {key + '_ema': loss_dict_ema_1[key] for key in loss_dict_ema_1}
+        self.log_dict(loss_dict_no_ema_0, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+        self.log_dict(loss_dict_no_ema_1, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+        self.log_dict(loss_dict_ema_0, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+        self.log_dict(loss_dict_ema_1, prog_bar=False, logger=True, on_step=False, on_epoch=True)
+
+    def on_train_batch_end(self, *args, **kwargs):
+        if self.use_ema:
+            self.model_ema(self.model)
+
+    def _get_rows_from_list(self, samples):
+        n_imgs_per_row = len(samples)
+        denoise_grid = rearrange(samples, 'n b c h w -> b n c h w')
+        denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w')
+        denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row)
+        return denoise_grid
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=2, sample=True, return_keys=None, **kwargs):
+        log = dict()
+        x = self.get_input(batch, self.first_stage_key)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        x = x.to(self.device)[:N]
+        log["inputs"] = x
+
+        # get diffusion row
+        diffusion_row = list()
+        x_start = x[:n_row]
+
+        for t in range(self.num_timesteps):
+            if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                t = t.to(self.device).long()
+                noise = torch.randn_like(x_start)
+                x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+                diffusion_row.append(x_noisy)
+
+        log["diffusion_row"] = self._get_rows_from_list(diffusion_row)
+
+        if sample:
+            # get denoise row
+            with self.ema_scope("Plotting"):
+                samples, denoise_row = self.sample(batch_size=N, return_intermediates=True)
+
+            log["samples"] = samples
+            log["denoise_row"] = self._get_rows_from_list(denoise_row)
+
+        if return_keys:
+            if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0:
+                return log
+            else:
+                return {key: log[key] for key in return_keys}
+        return log
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params = list(self.model.parameters())
+        if self.learn_logvar:
+            params = params + [self.logvar]
+        opt = torch.optim.AdamW(params, lr=lr)
+        return opt
+
+
+class LatentDiffusion(DDPM):
+    """main class"""
+
+    def __init__(self,
+                 first_stage_config,
+                 cond_stage_config,
+                 lossconfig,
+                 num_timesteps_cond=None,
+                 cond_stage_key="image",
+                 cond_stage_trainable=False,
+                 concat_mode=True,
+                 cond_stage_forward=None,
+                 conditioning_key=None,
+                 scale_factor=1.0,
+                 scale_by_std=False,
+                 force_null_conditioning=False,
+                 *args, **kwargs):
+        self.force_null_conditioning = force_null_conditioning
+        self.num_timesteps_cond = default(num_timesteps_cond, 1)
+        self.scale_by_std = scale_by_std
+        assert self.num_timesteps_cond <= kwargs['timesteps']
+        # for backwards compatibility after implementation of DiffusionWrapper
+        if conditioning_key is None:
+            conditioning_key = 'concat' if concat_mode else 'crossattn'
+        if cond_stage_config == '__is_unconditional__' and not self.force_null_conditioning:
+            conditioning_key = None
+        ckpt_path = kwargs.pop("ckpt_path", None)
+        reset_ema = kwargs.pop("reset_ema", False)
+        reset_num_ema_updates = kwargs.pop("reset_num_ema_updates", False)
+        ignore_keys = kwargs.pop("ignore_keys", [])
+        super().__init__(conditioning_key=conditioning_key, *args, **kwargs)
+        self.concat_mode = concat_mode
+        self.cond_stage_trainable = cond_stage_trainable
+        self.cond_stage_key = cond_stage_key
+        try:
+            self.num_downs = len(first_stage_config.params.ddconfig.ch_mult) - 1
+        except:
+            self.num_downs = 0
+        if not scale_by_std:
+            self.scale_factor = scale_factor
+        else:
+            self.register_buffer('scale_factor', torch.tensor(scale_factor))
+        self.instantiate_first_stage(first_stage_config)
+        self.instantiate_cond_stage(cond_stage_config)
+        self.instantiate_decoder_loss(lossconfig) 
+        self.cond_stage_forward = cond_stage_forward
+        self.clip_denoised = False
+        self.bbox_tokenizer = None
+
+        self.restarted_from_ckpt = False
+        if ckpt_path is not None:
+            self.init_from_ckpt(ckpt_path, ignore_keys)
+            self.restarted_from_ckpt = True
+            if reset_ema:
+                assert self.use_ema
+                print(
+                    f"Resetting ema to pure model weights. This is useful when restoring from an ema-only checkpoint.")
+                self.model_ema = LitEma(self.model)
+        if reset_num_ema_updates:
+            print(" +++++++++++ WARNING: RESETTING NUM_EMA UPDATES TO ZERO +++++++++++ ")
+            assert self.use_ema
+            self.model_ema.reset_num_updates()
+
+    def make_cond_schedule(self, ):
+        self.cond_ids = torch.full(size=(self.num_timesteps,), fill_value=self.num_timesteps - 1, dtype=torch.long)
+        ids = torch.round(torch.linspace(0, self.num_timesteps - 1, self.num_timesteps_cond)).long()
+        self.cond_ids[:self.num_timesteps_cond] = ids
+
+    @rank_zero_only
+    @torch.no_grad()
+    def on_train_batch_start(self, batch, batch_idx, dataloader_idx):
+        # only for very first batch
+        if self.scale_by_std and self.current_epoch == 0 and self.global_step == 0 and batch_idx == 0 and not self.restarted_from_ckpt:
+            assert self.scale_factor == 1., 'rather not use custom rescaling and std-rescaling simultaneously'
+            # set rescale weight to 1./std of encodings
+            print("### USING STD-RESCALING ###")
+            x = super().get_input(batch, self.first_stage_key)
+            x = x.to(self.device)
+            encoder_posterior = self.encode_first_stage(x)
+            z = self.get_first_stage_encoding(encoder_posterior).detach()
+            del self.scale_factor
+            self.register_buffer('scale_factor', 1. / z.flatten().std())
+            print(f"setting self.scale_factor to {self.scale_factor}")
+            print("### USING STD-RESCALING ###")
+
+    def register_schedule(self,
+                          given_betas=None, beta_schedule="linear", timesteps=1000,
+                          linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+        super().register_schedule(given_betas, beta_schedule, timesteps, linear_start, linear_end, cosine_s)
+
+        self.shorten_cond_schedule = self.num_timesteps_cond > 1
+        if self.shorten_cond_schedule:
+            self.make_cond_schedule()
+
+    def instantiate_first_stage(self, config):
+        model = instantiate_from_config(config)
+        self.first_stage_model = model.eval()
+        self.first_stage_model.train = disabled_train
+        # train decode
+        for name, param in self.first_stage_model.named_parameters():
+            if 'decoder' not in name and 'post_quant_conv' not in name:
+                param.requires_grad = False
+            else:
+                param.requires_grad = True
+        # for param in self.first_stage_model.parameters():
+        #     param.requires_grad = False
+    
+    def instantiate_decoder_loss(self, config):
+        decoder_loss = instantiate_from_config(config)
+        
+        self.decoder_loss = decoder_loss.train()
+        self.decoder_loss.train = disabled_train
+        # train discriminator
+        for name, param in self.decoder_loss.discriminator.named_parameters():
+            param.requires_grad = True
+
+    def instantiate_cond_stage(self, config):
+        if not self.cond_stage_trainable:
+            if config == "__is_first_stage__":
+                print("Using first stage also as cond stage.")
+                self.cond_stage_model = self.first_stage_model
+            elif config == "__is_unconditional__":
+                print(f"Training {self.__class__.__name__} as an unconditional model.")
+                self.cond_stage_model = None
+                # self.be_unconditional = True
+            else:
+                model = instantiate_from_config(config)
+                self.cond_stage_model = model.eval()
+                self.cond_stage_model.train = disabled_train
+                for param in self.cond_stage_model.parameters():
+                    param.requires_grad = False
+        else:
+            assert config != '__is_first_stage__'
+            assert config != '__is_unconditional__'
+            model = instantiate_from_config(config)
+            self.cond_stage_model = model
+
+    def _get_denoise_row_from_list(self, samples, desc='', force_no_decoder_quantization=False):
+        denoise_row = []
+        for zd in tqdm(samples, desc=desc):
+            denoise_row.append(self.decode_first_stage(zd.to(self.device),
+                                                       force_not_quantize=force_no_decoder_quantization))
+        n_imgs_per_row = len(denoise_row)
+        denoise_row = torch.stack(denoise_row)  # n_log_step, n_row, C, H, W
+        denoise_grid = rearrange(denoise_row, 'n b c h w -> b n c h w')
+        denoise_grid = rearrange(denoise_grid, 'b n c h w -> (b n) c h w')
+        denoise_grid = make_grid(denoise_grid, nrow=n_imgs_per_row)
+        return denoise_grid
+
+    def get_first_stage_encoding(self, encoder_posterior):
+        if isinstance(encoder_posterior, DiagonalGaussianDistribution):
+            z = encoder_posterior.sample()
+        elif isinstance(encoder_posterior, torch.Tensor):
+            z = encoder_posterior
+        else:
+            raise NotImplementedError(f"encoder_posterior of type '{type(encoder_posterior)}' not yet implemented")
+        # print(self.scale_factor)
+        # exit()
+        return self.scale_factor * z
+
+    def get_learned_conditioning(self, c):
+        if self.cond_stage_forward is None:
+            if hasattr(self.cond_stage_model, 'encode') and callable(self.cond_stage_model.encode):
+                c = self.cond_stage_model.encode(c)
+                if isinstance(c, DiagonalGaussianDistribution):
+                    c = c.mode()
+            else:
+                c = self.cond_stage_model(c)
+        else:
+            assert hasattr(self.cond_stage_model, self.cond_stage_forward)
+            c = getattr(self.cond_stage_model, self.cond_stage_forward)(c)
+        return c
+
+    def meshgrid(self, h, w):
+        y = torch.arange(0, h).view(h, 1, 1).repeat(1, w, 1)
+        x = torch.arange(0, w).view(1, w, 1).repeat(h, 1, 1)
+
+        arr = torch.cat([y, x], dim=-1)
+        return arr
+
+    def delta_border(self, h, w):
+        """
+        :param h: height
+        :param w: width
+        :return: normalized distance to image border,
+         wtith min distance = 0 at border and max dist = 0.5 at image center
+        """
+        lower_right_corner = torch.tensor([h - 1, w - 1]).view(1, 1, 2)
+        arr = self.meshgrid(h, w) / lower_right_corner
+        dist_left_up = torch.min(arr, dim=-1, keepdims=True)[0]
+        dist_right_down = torch.min(1 - arr, dim=-1, keepdims=True)[0]
+        edge_dist = torch.min(torch.cat([dist_left_up, dist_right_down], dim=-1), dim=-1)[0]
+        return edge_dist
+
+    def get_weighting(self, h, w, Ly, Lx, device):
+        weighting = self.delta_border(h, w)
+        weighting = torch.clip(weighting, self.split_input_params["clip_min_weight"],
+                               self.split_input_params["clip_max_weight"], )
+        weighting = weighting.view(1, h * w, 1).repeat(1, 1, Ly * Lx).to(device)
+
+        if self.split_input_params["tie_braker"]:
+            L_weighting = self.delta_border(Ly, Lx)
+            L_weighting = torch.clip(L_weighting,
+                                     self.split_input_params["clip_min_tie_weight"],
+                                     self.split_input_params["clip_max_tie_weight"])
+
+            L_weighting = L_weighting.view(1, 1, Ly * Lx).to(device)
+            weighting = weighting * L_weighting
+        return weighting
+
+    def get_fold_unfold(self, x, kernel_size, stride, uf=1, df=1):  # todo load once not every time, shorten code
+        """
+        :param x: img of size (bs, c, h, w)
+        :return: n img crops of size (n, bs, c, kernel_size[0], kernel_size[1])
+        """
+        bs, nc, h, w = x.shape
+
+        # number of crops in image
+        Ly = (h - kernel_size[0]) // stride[0] + 1
+        Lx = (w - kernel_size[1]) // stride[1] + 1
+
+        if uf == 1 and df == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold = torch.nn.Fold(output_size=x.shape[2:], **fold_params)
+
+            weighting = self.get_weighting(kernel_size[0], kernel_size[1], Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h, w)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0], kernel_size[1], Ly * Lx))
+
+        elif uf > 1 and df == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold_params2 = dict(kernel_size=(kernel_size[0] * uf, kernel_size[0] * uf),
+                                dilation=1, padding=0,
+                                stride=(stride[0] * uf, stride[1] * uf))
+            fold = torch.nn.Fold(output_size=(x.shape[2] * uf, x.shape[3] * uf), **fold_params2)
+
+            weighting = self.get_weighting(kernel_size[0] * uf, kernel_size[1] * uf, Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h * uf, w * uf)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0] * uf, kernel_size[1] * uf, Ly * Lx))
+
+        elif df > 1 and uf == 1:
+            fold_params = dict(kernel_size=kernel_size, dilation=1, padding=0, stride=stride)
+            unfold = torch.nn.Unfold(**fold_params)
+
+            fold_params2 = dict(kernel_size=(kernel_size[0] // df, kernel_size[0] // df),
+                                dilation=1, padding=0,
+                                stride=(stride[0] // df, stride[1] // df))
+            fold = torch.nn.Fold(output_size=(x.shape[2] // df, x.shape[3] // df), **fold_params2)
+
+            weighting = self.get_weighting(kernel_size[0] // df, kernel_size[1] // df, Ly, Lx, x.device).to(x.dtype)
+            normalization = fold(weighting).view(1, 1, h // df, w // df)  # normalizes the overlap
+            weighting = weighting.view((1, 1, kernel_size[0] // df, kernel_size[1] // df, Ly * Lx))
+
+        else:
+            raise NotImplementedError
+
+        return fold, unfold, normalization, weighting
+
+    @torch.no_grad()
+    def get_input(self, batch, k, return_first_stage_outputs=False, force_c_encode=False,
+                  cond_key=None, return_original_cond=False, bs=None, return_x=False):
+        x = super().get_input(batch, k)
+        if bs is not None:
+            x = x[:bs]
+        x = x.to(self.device)
+        encoder_posterior = self.encode_first_stage(x)
+        z = self.get_first_stage_encoding(encoder_posterior).detach()
+
+        if self.model.conditioning_key is not None and not self.force_null_conditioning:
+            if cond_key is None:
+                cond_key = self.cond_stage_key
+            if cond_key != self.first_stage_key:
+                if cond_key in ['caption', 'coordinates_bbox', "txt"]:
+                    xc = batch[cond_key]
+                elif cond_key in ['class_label', 'cls']:
+                    xc = batch
+                else:
+                    xc = super().get_input(batch, cond_key).to(self.device)
+            else:
+                xc = x
+            if not self.cond_stage_trainable or force_c_encode:
+                if isinstance(xc, dict) or isinstance(xc, list):
+                    c = self.get_learned_conditioning(xc)
+                else:
+                    c = self.get_learned_conditioning(xc.to(self.device))
+            else:
+                c = xc
+            if bs is not None:
+                c = c[:bs]
+
+            if self.use_positional_encodings:
+                pos_x, pos_y = self.compute_latent_shifts(batch)
+                ckey = __conditioning_keys__[self.model.conditioning_key]
+                c = {ckey: c, 'pos_x': pos_x, 'pos_y': pos_y}
+
+        else:
+            c = None
+            xc = None
+            if self.use_positional_encodings:
+                pos_x, pos_y = self.compute_latent_shifts(batch)
+                c = {'pos_x': pos_x, 'pos_y': pos_y}
+        out = [x, z, c]
+        if return_first_stage_outputs:
+            xrec = self.decode_first_stage(z)
+            out.extend([x, xrec])
+        if return_x:
+            out.extend([x])
+        if return_original_cond:
+            out.append(xc)
+        return out
+
+    @torch.no_grad()
+    def decode_first_stage(self, z, predict_cids=False, force_not_quantize=False):
+        if predict_cids:
+            if z.dim() == 4:
+                z = torch.argmax(z.exp(), dim=1).long()
+            z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None)
+            z = rearrange(z, 'b h w c -> b c h w').contiguous()
+
+        z = 1. / self.scale_factor * z
+        return self.first_stage_model.decode(z)
+    
+    # 2023-04-08
+    def decode_first_stage_with_grad(self, z, predict_cids=False, force_not_quantize=False):
+        if predict_cids:
+            if z.dim() == 4:
+                z = torch.argmax(z.exp(), dim=1).long()
+            z = self.first_stage_model.quantize.get_codebook_entry(z, shape=None)
+            z = rearrange(z, 'b h w c -> b c h w').contiguous()
+
+        z = 1. / self.scale_factor * z
+        return self.first_stage_model.decode(z)
+
+    @torch.no_grad()
+    def encode_first_stage(self, x):
+        return self.first_stage_model.encode(x)
+
+    def shared_step(self, batch,index, **kwargs):
+        gt, x, c = self.get_input(batch, self.first_stage_key)
+        loss = self(gt, x, c,index)
+        return loss
+
+    def forward(self, gt, x, c, opt_index, *args, **kwargs):
+        T = (self.num_timesteps-1) * torch.ones(x.shape[0]).to(self.device).long()
+        t_max = round(self.num_timesteps*self.t_max) * torch.ones(x.shape[0]).to(self.device).long()
+        t = torch.randint(0, t_max[0], (x.shape[0],), device=self.device).long()
+        
+        if self.model.conditioning_key is not None:
+            assert c is not None
+            if self.cond_stage_trainable:
+                c = self.get_learned_conditioning(c)
+            if self.shorten_cond_schedule:  # TODO: drop this option
+                tc = self.cond_ids[t].to(self.device)
+                c = self.q_sample(x_start=c, t=tc, noise=torch.randn_like(c.float()))
+        return self.p_losses(gt, x, c, t, T, t_max, opt_index, *args, **kwargs)
+
+    def apply_model(self, x_noisy, t, cond, return_ids=False):
+        if isinstance(cond, dict):
+            # hybrid case, cond is expected to be a dict
+            pass
+        else:
+            if not isinstance(cond, list):
+                cond = [cond]
+            key = 'c_concat' if self.model.conditioning_key == 'concat' else 'c_crossattn'
+            cond = {key: cond}
+
+        x_recon = self.model(x_noisy, t, **cond)
+
+        if isinstance(x_recon, tuple) and not return_ids:
+            return x_recon[0]
+        else:
+            return x_recon
+
+    def _predict_eps_from_xstart(self, x_t, t, pred_xstart):
+        return (extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t - pred_xstart) / \
+               extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape)
+
+    def _prior_bpd(self, x_start):
+        """
+        Get the prior KL term for the variational lower-bound, measured in
+        bits-per-dim.
+        This term can't be optimized, as it only depends on the encoder.
+        :param x_start: the [N x C x ...] tensor of inputs.
+        :return: a batch of [N] KL values (in bits), one per batch element.
+        """
+        batch_size = x_start.shape[0]
+        t = torch.tensor([self.num_timesteps - 1] * batch_size, device=x_start.device)
+        qt_mean, _, qt_log_variance = self.q_mean_variance(x_start, t)
+        kl_prior = normal_kl(mean1=qt_mean, logvar1=qt_log_variance, mean2=0.0, logvar2=0.0)
+        return mean_flat(kl_prior) / np.log(2.0)
+    
+    def _extract_into_tensor(arr, timesteps, broadcast_shape):
+        """
+        Extract values from a 1-D numpy array for a batch of indices.
+        :param arr: the 1-D numpy array.
+        :param timesteps: a tensor of indices into the array to extract.
+        :param broadcast_shape: a larger shape of K dimensions with the batch
+                                dimension equal to the length of timesteps.
+        :return: a tensor of shape [batch_size, 1, ...] where the shape has K dims.
+        """
+        try:
+            # float64 as default. float64 is not supported by mps device.
+            res = torch.from_numpy(arr).to(device=timesteps.device)[timesteps].float()
+        except:
+            # to be compatible with mps
+            res = torch.from_numpy(arr.astype(np.float32)).to(device=timesteps.device)[timesteps].float()
+            
+        while len(res.shape) < len(broadcast_shape):
+            res = res[..., None]
+        return res.expand(broadcast_shape)
+
+    def p_losses(self, gt, x_start, cond, t, t_max, t_tao, opt_index, noise=None):
+        
+        # obtain the input of decoder
+
+        with torch.no_grad():
+            noise = default(noise, lambda: torch.randn_like(x_start))
+            x_noisy = self.q_sample(x_start=x_start, t=t_max, noise=noise)
+            model_output_noise = self.apply_model(x_noisy, t_max, cond)
+            model_output_x0 = self.predict_start_from_noise(x_noisy, t=t_max, noise=model_output_noise)
+
+            time_range = np.flip(self.timesteps_inter) # [1000, 950, 900, ...]
+            total_steps = len(time_range)
+            time_range = time_range[round(total_steps*(1-self.t_max)):]
+            total_steps_end = len(time_range)
+            time_range = time_range[:-round(total_steps*self.t_min)]
+
+            b = x_noisy.shape[0]
+            t_tao = time_range[0] * torch.ones(x_noisy.shape[0]).to(self.device).long()
+            noise2 = default(noise, lambda: torch.randn_like(x_start))
+            model_output_noisy_tao = self.q_sample(x_start=model_output_x0, t=t_tao, noise=noise2)
+
+            img = model_output_noisy_tao
+
+            # iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+
+            
+            for i, step in enumerate(time_range):
+                ts = torch.full((b,), step, device=self.device, dtype=torch.long)
+                index = torch.full_like(ts, fill_value=total_steps_end - i - 1)
+                model_output_noise = self.apply_model(img, ts, cond)
+                model_output_x0 = self.predict_start_from_noise(img, t=ts, noise=model_output_noise)
+                model_mean, _, model_log_variance = self.q_posterior_inter(x_start=model_output_x0, x_t=img, t=index)
+                model_log_variance = model_log_variance * torch.ones(model_output_x0.shape).to(self.device)
+                model_sample = DiagonalGaussianDistribution(torch.cat([model_mean, model_log_variance], 1))
+                img = model_sample.sample()
+        
+        # train decoder
+        net_input = 1. / self.scale_factor * model_output_x0
+        net_input = self.first_stage_model.post_quant_conv(net_input)
+        out_decoder = self.first_stage_model.decoder(net_input)
+
+        # # save the paired data for the next training
+        # save_path = '/home/notebook/data/group/SunLingchen/dataset/decoder_dataset45'
+        # if True:
+        #     gt_npy_path = save_path + '/gt_npy'
+        #     gt_png_path = save_path + '/gt_png'
+        #     input_npy_path = opt.outdir_lr + '/input_npy'
+        #     os.makedirs(gt_npy_path, exist_ok=True)
+        #     os.makedirs(gt_png_path, exist_ok=True)
+        #     os.makedirs(input_npy_path, exist_ok=True)
+
+        #     for i in range(b):
+        #         # save hr png images
+        #         gt_sub_image_save = (gt[i] + 1.0) / 2.0
+        #         gt_sub_image = torch.clamp(gt_sub_image_save, min=0.0, max=1.0)
+        #         gt_sub_sample_show = 255. * rearrange(gt_sub_image.cpu().numpy(), 'c h w -> h w c')
+        #         Image.fromarray(gt_sub_sample_show.astype(np.uint8)).save(
+        #         os.path.join(gt_png_path, basename + '_' + str(opt.seed) + '_' + str(num_sample)) + '.png')
+        loss_dict = {}
+        prefix = 'train' if self.training else 'val'
+
+        if opt_index == 0:
+            last_layer = self.first_stage_model.decoder.conv_out.weight
+            loss,_ = self.decoder_loss(inputs=gt, reconstructions=out_decoder, optimizer_idx=opt_index, global_step=self.global_step,
+            last_layer=last_layer, split="train")
+            loss_dict.update({f'{prefix}/loss': loss})
+            return loss, loss_dict
+        else:
+            last_layer = self.first_stage_model.decoder.conv_out.weight
+            # train the discriminator net
+            d_loss, _= self.decoder_loss(inputs=gt, reconstructions=out_decoder, optimizer_idx=opt_index, global_step=self.global_step,
+            last_layer=last_layer, split="train")
+            loss_dict.update({f'{prefix}/d_loss': d_loss.mean()})
+            return d_loss, loss_dict
+
+
+
+        # if opt_index == 0:
+        #     noise = default(noise, lambda: torch.randn_like(x_start))
+        #     x_noisy = self.q_sample(x_start=x_start, t=t, noise=noise)
+        #     model_output = self.apply_model(x_noisy, t, cond)
+
+        #     loss_dict = {}
+        #     prefix = 'train' if self.training else 'val'
+
+        #     if self.parameterization == "x0":
+        #         target = x_start
+        #     elif self.parameterization == "eps":
+        #         target = noise
+        #     elif self.parameterization == "v":
+        #         target = self.get_v(x_start, noise, t)
+        #     else:
+        #         raise NotImplementedError()
+
+        #     loss_simple = self.get_loss(model_output, target, mean=False).mean([1, 2, 3])
+        #     loss_dict.update({f'{prefix}/loss_simple': loss_simple.mean()})
+
+        #     logvar_t = self.logvar[t].to(self.device)
+        #     loss = loss_simple / torch.exp(logvar_t) + logvar_t
+        #     # loss = loss_simple / torch.exp(self.logvar) + self.logvar
+        #     if self.learn_logvar:
+        #         loss_dict.update({f'{prefix}/loss_gamma': loss.mean()})
+        #         loss_dict.update({'logvar': self.logvar.data.mean()})
+
+        #     loss = self.l_simple_weight * loss.mean()
+
+        #     loss_vlb = self.get_loss(model_output, target, mean=False).mean(dim=(1, 2, 3))
+        #     loss_vlb = (self.lvlb_weights[t] * loss_vlb).mean()
+        #     loss_dict.update({f'{prefix}/loss_vlb': loss_vlb})
+        #     loss += (self.original_elbo_weight * loss_vlb)
+        #     loss_dict.update({f'{prefix}/loss': loss})
+
+        #     return loss, loss_dict
+        # else:
+        #     # train net
+        #     loss_dict = {}
+        #     prefix = 'train' if self.training else 'val'
+
+        #     noise2 = default(noise, lambda: torch.randn_like(x_start))
+        #     x_noisy = self.q_sample(x_start=x_start, t=t_max, noise=noise2)
+        #     model_output_noise = self.apply_model(x_noisy, t_max, cond)
+            
+            # model_output_x0 = self.predict_start_from_noise(x_noisy, t=t_max, noise=model_output_noise)
+            
+            # noise3 = default(noise, lambda: torch.randn_like(x_start))
+            # model_output_noisy_tao = self.q_sample(x_start=model_output_x0, t=t_tao, noise=noise3)
+
+        #     model_output_noise_from_tao = self.apply_model(model_output_noisy_tao, t_tao, cond)
+        #     model_output_x0_from_tao = self.predict_start_from_noise(model_output_noisy_tao, t=t_tao, noise=model_output_noise_from_tao) # predict x0 from tao
+
+        #     loss_x0 = self.get_loss(model_output_x0, x_start, mean=False).mean([1, 2, 3])
+        #     loss_noise_from_tao = self.get_loss(model_output_noise_from_tao, noise3, mean=False).mean([1, 2, 3])
+        #     loss_x0_from_tao = self.get_loss(model_output_x0_from_tao, x_start, mean=False).mean([1, 2, 3])
+        #     loss_dict.update({f'{prefix}/loss_x0': loss_x0.mean()})
+        #     loss_dict.update({f'{prefix}/loss_x0_from_tao': loss_x0_from_tao.mean()})
+        #     loss_dict.update({f'{prefix}/loss_noise_from_tao': loss_noise_from_tao.mean()})
+        #     loss = loss_x0.mean() + loss_x0_from_tao.mean() + loss_noise_from_tao.mean()
+        #     loss_dict.update({f'{prefix}/loss_net': loss})
+
+        #     return loss, loss_dict
+
+    def p_mean_variance(self, x, c, t, clip_denoised: bool, return_codebook_ids=False, quantize_denoised=False,
+                        return_x0=False, score_corrector=None, corrector_kwargs=None):
+        t_in = t
+        model_out = self.apply_model(x, t_in, c, return_ids=return_codebook_ids)
+
+        if score_corrector is not None:
+            assert self.parameterization == "eps"
+            model_out = score_corrector.modify_score(self, model_out, x, t, c, **corrector_kwargs)
+
+        if return_codebook_ids:
+            model_out, logits = model_out
+
+        if self.parameterization == "eps":
+            x_recon = self.predict_start_from_noise(x, t=t, noise=model_out)
+        elif self.parameterization == "x0":
+            x_recon = model_out
+        else:
+            raise NotImplementedError()
+
+        if clip_denoised:
+            x_recon.clamp_(-1., 1.)
+        if quantize_denoised:
+            x_recon, _, [_, _, indices] = self.first_stage_model.quantize(x_recon)
+        model_mean, posterior_variance, posterior_log_variance = self.q_posterior(x_start=x_recon, x_t=x, t=t)
+        if return_codebook_ids:
+            return model_mean, posterior_variance, posterior_log_variance, logits
+        elif return_x0:
+            return model_mean, posterior_variance, posterior_log_variance, x_recon
+        else:
+            return model_mean, posterior_variance, posterior_log_variance
+
+    @torch.no_grad()
+    def p_sample(self, x, c, t, clip_denoised=False, repeat_noise=False,
+                 return_codebook_ids=False, quantize_denoised=False, return_x0=False,
+                 temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None):
+        b, *_, device = *x.shape, x.device
+        outputs = self.p_mean_variance(x=x, c=c, t=t, clip_denoised=clip_denoised,
+                                       return_codebook_ids=return_codebook_ids,
+                                       quantize_denoised=quantize_denoised,
+                                       return_x0=return_x0,
+                                       score_corrector=score_corrector, corrector_kwargs=corrector_kwargs)
+        if return_codebook_ids:
+            raise DeprecationWarning("Support dropped.")
+            model_mean, _, model_log_variance, logits = outputs
+        elif return_x0:
+            model_mean, _, model_log_variance, x0 = outputs
+        else:
+            model_mean, _, model_log_variance = outputs
+
+        noise = noise_like(x.shape, device, repeat_noise) * temperature
+        if noise_dropout > 0.:
+            noise = torch.nn.functional.dropout(noise, p=noise_dropout)
+        # no noise when t == 0
+        nonzero_mask = (1 - (t == 0).float()).reshape(b, *((1,) * (len(x.shape) - 1)))
+
+        if return_codebook_ids:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, logits.argmax(dim=1)
+        if return_x0:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise, x0
+        else:
+            return model_mean + nonzero_mask * (0.5 * model_log_variance).exp() * noise
+
+    @torch.no_grad()
+    def progressive_denoising(self, cond, shape, verbose=True, callback=None, quantize_denoised=False,
+                              img_callback=None, mask=None, x0=None, temperature=1., noise_dropout=0.,
+                              score_corrector=None, corrector_kwargs=None, batch_size=None, x_T=None, start_T=None,
+                              log_every_t=None):
+        if not log_every_t:
+            log_every_t = self.log_every_t
+        timesteps = self.num_timesteps
+        if batch_size is not None:
+            b = batch_size if batch_size is not None else shape[0]
+            shape = [batch_size] + list(shape)
+        else:
+            b = batch_size = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=self.device)
+        else:
+            img = x_T
+        intermediates = []
+        if cond is not None:
+            if isinstance(cond, dict):
+                cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else
+                list(map(lambda x: x[:batch_size], cond[key])) for key in cond}
+            else:
+                cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size]
+
+        if start_T is not None:
+            timesteps = min(timesteps, start_T)
+        iterator = tqdm(reversed(range(0, timesteps)), desc='Progressive Generation',
+                        total=timesteps) if verbose else reversed(
+            range(0, timesteps))
+        if type(temperature) == float:
+            temperature = [temperature] * timesteps
+
+        for i in iterator:
+            ts = torch.full((b,), i, device=self.device, dtype=torch.long)
+            if self.shorten_cond_schedule:
+                assert self.model.conditioning_key != 'hybrid'
+                tc = self.cond_ids[ts].to(cond.device)
+                cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond))
+
+            img, x0_partial = self.p_sample(img, cond, ts,
+                                            clip_denoised=self.clip_denoised,
+                                            quantize_denoised=quantize_denoised, return_x0=True,
+                                            temperature=temperature[i], noise_dropout=noise_dropout,
+                                            score_corrector=score_corrector, corrector_kwargs=corrector_kwargs)
+            if mask is not None:
+                assert x0 is not None
+                img_orig = self.q_sample(x0, ts)
+                img = img_orig * mask + (1. - mask) * img
+
+            if i % log_every_t == 0 or i == timesteps - 1:
+                intermediates.append(x0_partial)
+            if callback: callback(i)
+            if img_callback: img_callback(img, i)
+        return img, intermediates
+
+    @torch.no_grad()
+    def p_sample_loop(self, cond, shape, return_intermediates=False,
+                      x_T=None, verbose=True, callback=None, timesteps=None, quantize_denoised=False,
+                      mask=None, x0=None, img_callback=None, start_T=None,
+                      log_every_t=None):
+
+        if not log_every_t:
+            log_every_t = self.log_every_t
+        device = self.betas.device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+
+        intermediates = [img]
+        if timesteps is None:
+            timesteps = self.num_timesteps
+
+        if start_T is not None:
+            timesteps = min(timesteps, start_T)
+        iterator = tqdm(reversed(range(0, timesteps)), desc='Sampling t', total=timesteps) if verbose else reversed(
+            range(0, timesteps))
+
+        if mask is not None:
+            assert x0 is not None
+            assert x0.shape[2:3] == mask.shape[2:3]  # spatial size has to match
+
+        for i in iterator:
+            ts = torch.full((b,), i, device=device, dtype=torch.long)
+            if self.shorten_cond_schedule:
+                assert self.model.conditioning_key != 'hybrid'
+                tc = self.cond_ids[ts].to(cond.device)
+                cond = self.q_sample(x_start=cond, t=tc, noise=torch.randn_like(cond))
+
+            img = self.p_sample(img, cond, ts,
+                                clip_denoised=self.clip_denoised,
+                                quantize_denoised=quantize_denoised)
+            if mask is not None:
+                img_orig = self.q_sample(x0, ts)
+                img = img_orig * mask + (1. - mask) * img
+
+            if i % log_every_t == 0 or i == timesteps - 1:
+                intermediates.append(img)
+            if callback: callback(i)
+            if img_callback: img_callback(img, i)
+
+        if return_intermediates:
+            return img, intermediates
+        return img
+
+    @torch.no_grad()
+    def sample(self, cond, batch_size=16, return_intermediates=False, x_T=None,
+               verbose=True, timesteps=None, quantize_denoised=False,
+               mask=None, x0=None, shape=None, **kwargs):
+        if shape is None:
+            shape = (batch_size, self.channels, self.image_size, self.image_size)
+        if cond is not None:
+            if isinstance(cond, dict):
+                cond = {key: cond[key][:batch_size] if not isinstance(cond[key], list) else
+                list(map(lambda x: x[:batch_size], cond[key])) for key in cond}
+            else:
+                cond = [c[:batch_size] for c in cond] if isinstance(cond, list) else cond[:batch_size]
+        return self.p_sample_loop(cond,
+                                  shape,
+                                  return_intermediates=return_intermediates, x_T=x_T,
+                                  verbose=verbose, timesteps=timesteps, quantize_denoised=quantize_denoised,
+                                  mask=mask, x0=x0)
+
+    @torch.no_grad()
+    def sample_log(self, cond, batch_size, ddim, ddim_steps, **kwargs):
+        if ddim:
+            ddim_sampler = DDIMSampler(self)
+            shape = (self.channels, self.image_size, self.image_size)
+            samples, intermediates = ddim_sampler.sample(ddim_steps, batch_size,
+                                                         shape, cond, verbose=False, **kwargs)
+
+        else:
+            samples, intermediates = self.sample(cond=cond, batch_size=batch_size,
+                                                 return_intermediates=True, **kwargs)
+
+        return samples, intermediates
+
+    @torch.no_grad()
+    def get_unconditional_conditioning(self, batch_size, null_label=None):
+        if null_label is not None:
+            xc = null_label
+            if isinstance(xc, ListConfig):
+                xc = list(xc)
+            if isinstance(xc, dict) or isinstance(xc, list):
+                c = self.get_learned_conditioning(xc)
+            else:
+                if hasattr(xc, "to"):
+                    xc = xc.to(self.device)
+                c = self.get_learned_conditioning(xc)
+        else:
+            if self.cond_stage_key in ["class_label", "cls"]:
+                xc = self.cond_stage_model.get_unconditional_conditioning(batch_size, device=self.device)
+                return self.get_learned_conditioning(xc)
+            else:
+                raise NotImplementedError("todo")
+        if isinstance(c, list):  # in case the encoder gives us a list
+            for i in range(len(c)):
+                c[i] = repeat(c[i], '1 ... -> b ...', b=batch_size).to(self.device)
+        else:
+            c = repeat(c, '1 ... -> b ...', b=batch_size).to(self.device)
+        return c
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=50, ddim_eta=0., return_keys=None,
+                   quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True,
+                   plot_diffusion_rows=True, unconditional_guidance_scale=1., unconditional_guidance_label=None,
+                   use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc = self.get_input(batch, self.first_stage_key,
+                                           return_first_stage_outputs=True,
+                                           force_c_encode=True,
+                                           return_original_cond=True,
+                                           bs=N)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', "cls"]:
+                try:
+                    xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                    log['conditioning'] = xc
+                except KeyError:
+                    # probably no "human_label" in batch
+                    pass
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+            if quantize_denoised and not isinstance(self.first_stage_model, AutoencoderKL) and not isinstance(
+                    self.first_stage_model, IdentityFirstStage):
+                # also display when quantizing x0 while sampling
+                with ema_scope("Plotting Quantized Denoised"):
+                    samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                             ddim_steps=ddim_steps, eta=ddim_eta,
+                                                             quantize_denoised=True)
+                    # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True,
+                    #                                      quantize_denoised=True)
+                x_samples = self.decode_first_stage(samples.to(self.device))
+                log["samples_x0_quantized"] = x_samples
+
+        if unconditional_guidance_scale > 1.0:
+            uc = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            if self.model.conditioning_key == "crossattn-adm":
+                uc = {"c_crossattn": [uc], "c_adm": c["c_adm"]}
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        if inpaint:
+            # make a simple center square
+            b, h, w = z.shape[0], z.shape[2], z.shape[3]
+            mask = torch.ones(N, h, w).to(self.device)
+            # zeros will be filled in
+            mask[:, h // 4:3 * h // 4, w // 4:3 * w // 4] = 0.
+            mask = mask[:, None, ...]
+            with ema_scope("Plotting Inpaint"):
+                samples, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim, eta=ddim_eta,
+                                             ddim_steps=ddim_steps, x0=z[:N], mask=mask)
+            x_samples = self.decode_first_stage(samples.to(self.device))
+            log["samples_inpainting"] = x_samples
+            log["mask"] = mask
+
+            # outpaint
+            mask = 1. - mask
+            with ema_scope("Plotting Outpaint"):
+                samples, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim, eta=ddim_eta,
+                                             ddim_steps=ddim_steps, x0=z[:N], mask=mask)
+            x_samples = self.decode_first_stage(samples.to(self.device))
+            log["samples_outpainting"] = x_samples
+
+        if plot_progressive_rows:
+            with ema_scope("Plotting Progressives"):
+                img, progressives = self.progressive_denoising(c,
+                                                               shape=(self.channels, self.image_size, self.image_size),
+                                                               batch_size=N)
+            prog_row = self._get_denoise_row_from_list(progressives, desc="Progressive Generation")
+            log["progressive_row"] = prog_row
+
+        if return_keys:
+            if np.intersect1d(list(log.keys()), return_keys).shape[0] == 0:
+                return log
+            else:
+                return {key: log[key] for key in return_keys}
+        return log
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params = list(self.model.parameters())
+        lr = self.learning_rate
+        params_net = list(self.model.parameters())
+
+        if self.cond_stage_trainable:
+            print(f"{self.__class__.__name__}: Also optimizing conditioner params!")
+            params = params + list(self.cond_stage_model.parameters())
+            params_net = params_net + list(self.cond_stage_model.parameters())
+        if self.learn_logvar:
+            print('Diffusion model optimizing logvar')
+            params.append(self.logvar)
+            params_net.append(self.logvar)
+        opt = torch.optim.AdamW(params, lr=lr)
+        opt_net = torch.optim.AdamW(params_net, lr=lr)
+        if self.use_scheduler:
+            assert 'target' in self.scheduler_config
+            scheduler = instantiate_from_config(self.scheduler_config)
+
+            print("Setting up LambdaLR scheduler...")
+            scheduler = [
+                {
+                    'scheduler': LambdaLR(opt, lr_lambda=scheduler.schedule),
+                    'interval': 'step',
+                    'frequency': 1
+                }]
+            return [opt, opt_net], scheduler
+        return opt, opt_net
+
+    @torch.no_grad()
+    def to_rgb(self, x):
+        x = x.float()
+        if not hasattr(self, "colorize"):
+            self.colorize = torch.randn(3, x.shape[1], 1, 1).to(x)
+        x = nn.functional.conv2d(x, weight=self.colorize)
+        x = 2. * (x - x.min()) / (x.max() - x.min()) - 1.
+        return x
+
+
+class DiffusionWrapper(pl.LightningModule):
+    def __init__(self, diff_model_config, conditioning_key):
+        super().__init__()
+        self.sequential_cross_attn = diff_model_config.pop("sequential_crossattn", False)
+        self.diffusion_model = instantiate_from_config(diff_model_config)
+        self.conditioning_key = conditioning_key
+        assert self.conditioning_key in [None, 'concat', 'crossattn', 'hybrid', 'adm', 'hybrid-adm', 'crossattn-adm']
+
+    def forward(self, x, t, c_concat: list = None, c_crossattn: list = None, c_adm=None):
+        if self.conditioning_key is None:
+            out = self.diffusion_model(x, t)
+        elif self.conditioning_key == 'concat':
+            xc = torch.cat([x] + c_concat, dim=1)
+            out = self.diffusion_model(xc, t)
+        elif self.conditioning_key == 'crossattn':
+            if not self.sequential_cross_attn:
+                cc = torch.cat(c_crossattn, 1)
+            else:
+                cc = c_crossattn
+            out = self.diffusion_model(x, t, context=cc)
+        elif self.conditioning_key == 'hybrid':
+            xc = torch.cat([x] + c_concat, dim=1)
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(xc, t, context=cc)
+        elif self.conditioning_key == 'hybrid-adm':
+            assert c_adm is not None
+            xc = torch.cat([x] + c_concat, dim=1)
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(xc, t, context=cc, y=c_adm)
+        elif self.conditioning_key == 'crossattn-adm':
+            assert c_adm is not None
+            cc = torch.cat(c_crossattn, 1)
+            out = self.diffusion_model(x, t, context=cc, y=c_adm)
+        elif self.conditioning_key == 'adm':
+            cc = c_crossattn[0]
+            out = self.diffusion_model(x, t, y=cc)
+        else:
+            raise NotImplementedError()
+
+        return out
+
+
+class LatentUpscaleDiffusion(LatentDiffusion):
+    def __init__(self, *args, low_scale_config, low_scale_key="LR", noise_level_key=None, **kwargs):
+        super().__init__(*args, **kwargs)
+        # assumes that neither the cond_stage nor the low_scale_model contain trainable params
+        assert not self.cond_stage_trainable
+        self.instantiate_low_stage(low_scale_config)
+        self.low_scale_key = low_scale_key
+        self.noise_level_key = noise_level_key
+
+    def instantiate_low_stage(self, config):
+        model = instantiate_from_config(config)
+        self.low_scale_model = model.eval()
+        self.low_scale_model.train = disabled_train
+        for param in self.low_scale_model.parameters():
+            param.requires_grad = False
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, log_mode=False):
+        if not log_mode:
+            z, c = super().get_input(batch, k, force_c_encode=True, bs=bs)
+        else:
+            z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                                  force_c_encode=True, return_original_cond=True, bs=bs)
+        x_low = batch[self.low_scale_key][:bs]
+        x_low = rearrange(x_low, 'b h w c -> b c h w')
+        x_low = x_low.to(memory_format=torch.contiguous_format).float()
+        zx, noise_level = self.low_scale_model(x_low)
+        if self.noise_level_key is not None:
+            # get noise level from batch instead, e.g. when extracting a custom noise level for bsr
+            raise NotImplementedError('TODO')
+
+        all_conds = {"c_concat": [zx], "c_crossattn": [c], "c_adm": noise_level}
+        if log_mode:
+            # TODO: maybe disable if too expensive
+            x_low_rec = self.low_scale_model.decode(zx)
+            return gt, z, all_conds, x, xrec, xc, x_low, x_low_rec, noise_level
+        return gt, z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=200, ddim_eta=1., return_keys=None,
+                   plot_denoise_rows=False, plot_progressive_rows=True, plot_diffusion_rows=True,
+                   unconditional_guidance_scale=1., unconditional_guidance_label=None, use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc, x_low, x_low_rec, noise_level = self.get_input(batch, self.first_stage_key, bs=N,
+                                                                          log_mode=True)
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        log["x_lr"] = x_low
+        log[f"x_lr_rec_@noise_levels{'-'.join(map(lambda x: str(x), list(noise_level.cpu().numpy())))}"] = x_low_rec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', 'cls']:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                log['conditioning'] = xc
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+        if unconditional_guidance_scale > 1.0:
+            uc_tmp = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            # TODO explore better "unconditional" choices for the other keys
+            # maybe guide away from empty text label and highest noise level and maximally degraded zx?
+            uc = dict()
+            for k in c:
+                if k == "c_crossattn":
+                    assert isinstance(c[k], list) and len(c[k]) == 1
+                    uc[k] = [uc_tmp]
+                elif k == "c_adm":  # todo: only run with text-based guidance?
+                    assert isinstance(c[k], torch.Tensor)
+                    #uc[k] = torch.ones_like(c[k]) * self.low_scale_model.max_noise_level
+                    uc[k] = c[k]
+                elif isinstance(c[k], list):
+                    uc[k] = [c[k][i] for i in range(len(c[k]))]
+                else:
+                    uc[k] = c[k]
+
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond=c, batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        if plot_progressive_rows:
+            with ema_scope("Plotting Progressives"):
+                img, progressives = self.progressive_denoising(c,
+                                                               shape=(self.channels, self.image_size, self.image_size),
+                                                               batch_size=N)
+            prog_row = self._get_denoise_row_from_list(progressives, desc="Progressive Generation")
+            log["progressive_row"] = prog_row
+
+        return log
+
+
+class LatentFinetuneDiffusion(LatentDiffusion):
+    """
+         Basis for different finetunas, such as inpainting or depth2image
+         To disable finetuning mode, set finetune_keys to None
+    """
+
+    def __init__(self,
+                 concat_keys: tuple,
+                 finetune_keys=("model.diffusion_model.input_blocks.0.0.weight",
+                                "model_ema.diffusion_modelinput_blocks00weight"
+                                ),
+                 keep_finetune_dims=4,
+                 # if model was trained without concat mode before and we would like to keep these channels
+                 c_concat_log_start=None,  # to log reconstruction of c_concat codes
+                 c_concat_log_end=None,
+                 *args, **kwargs
+                 ):
+        ckpt_path = kwargs.pop("ckpt_path", None)
+        ignore_keys = kwargs.pop("ignore_keys", list())
+        super().__init__(*args, **kwargs)
+        self.finetune_keys = finetune_keys
+        self.concat_keys = concat_keys
+        self.keep_dims = keep_finetune_dims
+        self.c_concat_log_start = c_concat_log_start
+        self.c_concat_log_end = c_concat_log_end
+        if exists(self.finetune_keys): assert exists(ckpt_path), 'can only finetune from a given checkpoint'
+        if exists(ckpt_path):
+            self.init_from_ckpt(ckpt_path, ignore_keys)
+
+    def init_from_ckpt(self, path, ignore_keys=list(), only_model=False):
+        sd = torch.load(path, map_location="cpu")
+        if "state_dict" in list(sd.keys()):
+            sd = sd["state_dict"]
+        keys = list(sd.keys())
+        for k in keys:
+            for ik in ignore_keys:
+                if k.startswith(ik):
+                    print("Deleting key {} from state_dict.".format(k))
+                    del sd[k]
+
+            # make it explicit, finetune by including extra input channels
+            if exists(self.finetune_keys) and k in self.finetune_keys:
+                new_entry = None
+                for name, param in self.named_parameters():
+                    if name in self.finetune_keys:
+                        print(
+                            f"modifying key '{name}' and keeping its original {self.keep_dims} (channels) dimensions only")
+                        new_entry = torch.zeros_like(param)  # zero init
+                assert exists(new_entry), 'did not find matching parameter to modify'
+                new_entry[:, :self.keep_dims, ...] = sd[k]
+                sd[k] = new_entry
+
+        missing, unexpected = self.load_state_dict(sd, strict=False) if not only_model else self.model.load_state_dict(
+            sd, strict=False)
+        print(f"Restored from {path} with {len(missing)} missing and {len(unexpected)} unexpected keys")
+        if len(missing) > 0:
+            print(f"Missing Keys: {missing}")
+        if len(unexpected) > 0:
+            print(f"Unexpected Keys: {unexpected}")
+
+    @torch.no_grad()
+    def log_images(self, batch, N=8, n_row=4, sample=True, ddim_steps=200, ddim_eta=1., return_keys=None,
+                   quantize_denoised=True, inpaint=True, plot_denoise_rows=False, plot_progressive_rows=True,
+                   plot_diffusion_rows=True, unconditional_guidance_scale=1., unconditional_guidance_label=None,
+                   use_ema_scope=True,
+                   **kwargs):
+        ema_scope = self.ema_scope if use_ema_scope else nullcontext
+        use_ddim = ddim_steps is not None
+
+        log = dict()
+        z, c, x, xrec, xc = self.get_input(batch, self.first_stage_key, bs=N, return_first_stage_outputs=True)
+        c_cat, c = c["c_concat"][0], c["c_crossattn"][0]
+        N = min(x.shape[0], N)
+        n_row = min(x.shape[0], n_row)
+        log["inputs"] = x
+        log["reconstruction"] = xrec
+        if self.model.conditioning_key is not None:
+            if hasattr(self.cond_stage_model, "decode"):
+                xc = self.cond_stage_model.decode(c)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ["caption", "txt"]:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch[self.cond_stage_key], size=x.shape[2] // 25)
+                log["conditioning"] = xc
+            elif self.cond_stage_key in ['class_label', 'cls']:
+                xc = log_txt_as_img((x.shape[2], x.shape[3]), batch["human_label"], size=x.shape[2] // 25)
+                log['conditioning'] = xc
+            elif isimage(xc):
+                log["conditioning"] = xc
+            if ismap(xc):
+                log["original_conditioning"] = self.to_rgb(xc)
+
+        if not (self.c_concat_log_start is None and self.c_concat_log_end is None):
+            log["c_concat_decoded"] = self.decode_first_stage(c_cat[:, self.c_concat_log_start:self.c_concat_log_end])
+
+        if plot_diffusion_rows:
+            # get diffusion row
+            diffusion_row = list()
+            z_start = z[:n_row]
+            for t in range(self.num_timesteps):
+                if t % self.log_every_t == 0 or t == self.num_timesteps - 1:
+                    t = repeat(torch.tensor([t]), '1 -> b', b=n_row)
+                    t = t.to(self.device).long()
+                    noise = torch.randn_like(z_start)
+                    z_noisy = self.q_sample(x_start=z_start, t=t, noise=noise)
+                    diffusion_row.append(self.decode_first_stage(z_noisy))
+
+            diffusion_row = torch.stack(diffusion_row)  # n_log_step, n_row, C, H, W
+            diffusion_grid = rearrange(diffusion_row, 'n b c h w -> b n c h w')
+            diffusion_grid = rearrange(diffusion_grid, 'b n c h w -> (b n) c h w')
+            diffusion_grid = make_grid(diffusion_grid, nrow=diffusion_row.shape[0])
+            log["diffusion_row"] = diffusion_grid
+
+        if sample:
+            # get denoise row
+            with ema_scope("Sampling"):
+                samples, z_denoise_row = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]},
+                                                         batch_size=N, ddim=use_ddim,
+                                                         ddim_steps=ddim_steps, eta=ddim_eta)
+                # samples, z_denoise_row = self.sample(cond=c, batch_size=N, return_intermediates=True)
+            x_samples = self.decode_first_stage(samples)
+            log["samples"] = x_samples
+            if plot_denoise_rows:
+                denoise_grid = self._get_denoise_row_from_list(z_denoise_row)
+                log["denoise_row"] = denoise_grid
+
+        if unconditional_guidance_scale > 1.0:
+            uc_cross = self.get_unconditional_conditioning(N, unconditional_guidance_label)
+            uc_cat = c_cat
+            uc_full = {"c_concat": [uc_cat], "c_crossattn": [uc_cross]}
+            with ema_scope("Sampling with classifier-free guidance"):
+                samples_cfg, _ = self.sample_log(cond={"c_concat": [c_cat], "c_crossattn": [c]},
+                                                 batch_size=N, ddim=use_ddim,
+                                                 ddim_steps=ddim_steps, eta=ddim_eta,
+                                                 unconditional_guidance_scale=unconditional_guidance_scale,
+                                                 unconditional_conditioning=uc_full,
+                                                 )
+                x_samples_cfg = self.decode_first_stage(samples_cfg)
+                log[f"samples_cfg_scale_{unconditional_guidance_scale:.2f}"] = x_samples_cfg
+
+        return log
+
+
+class LatentInpaintDiffusion(LatentFinetuneDiffusion):
+    """
+    can either run as pure inpainting model (only concat mode) or with mixed conditionings,
+    e.g. mask as concat and text via cross-attn.
+    To disable finetuning mode, set finetune_keys to None
+     """
+
+    def __init__(self,
+                 concat_keys=("mask", "masked_image"),
+                 masked_image_key="masked_image",
+                 *args, **kwargs
+                 ):
+        super().__init__(concat_keys, *args, **kwargs)
+        self.masked_image_key = masked_image_key
+        assert self.masked_image_key in concat_keys
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for inpainting'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        c_cat = list()
+        for ck in self.concat_keys:
+            cc = rearrange(batch[ck], 'b h w c -> b c h w').to(memory_format=torch.contiguous_format).float()
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            bchw = z.shape
+            if ck != self.masked_image_key:
+                cc = torch.nn.functional.interpolate(cc, size=bchw[-2:])
+            else:
+                cc = self.get_first_stage_encoding(self.encode_first_stage(cc))
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super(LatentInpaintDiffusion, self).log_images(*args, **kwargs)
+        log["masked_image"] = rearrange(args[0]["masked_image"],
+                                        'b h w c -> b c h w').to(memory_format=torch.contiguous_format).float()
+        return log
+
+
+class LatentDepth2ImageDiffusion(LatentFinetuneDiffusion):
+    """
+    condition on monocular depth estimation
+    """
+
+    def __init__(self, depth_stage_config, concat_keys=("midas_in",), *args, **kwargs):
+        super().__init__(concat_keys=concat_keys, *args, **kwargs)
+        self.depth_model = instantiate_from_config(depth_stage_config)
+        self.depth_stage_key = concat_keys[0]
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for depth2img'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        assert len(self.concat_keys) == 1
+        c_cat = list()
+        for ck in self.concat_keys:
+            cc = batch[ck]
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            cc = self.depth_model(cc)
+            cc = torch.nn.functional.interpolate(
+                cc,
+                size=z.shape[2:],
+                mode="bicubic",
+                align_corners=False,
+            )
+
+            depth_min, depth_max = torch.amin(cc, dim=[1, 2, 3], keepdim=True), torch.amax(cc, dim=[1, 2, 3],
+                                                                                           keepdim=True)
+            cc = 2. * (cc - depth_min) / (depth_max - depth_min + 0.001) - 1.
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super().log_images(*args, **kwargs)
+        depth = self.depth_model(args[0][self.depth_stage_key])
+        depth_min, depth_max = torch.amin(depth, dim=[1, 2, 3], keepdim=True), \
+                               torch.amax(depth, dim=[1, 2, 3], keepdim=True)
+        log["depth"] = 2. * (depth - depth_min) / (depth_max - depth_min) - 1.
+        return log
+
+
+class LatentUpscaleFinetuneDiffusion(LatentFinetuneDiffusion):
+    """
+        condition on low-res image (and optionally on some spatial noise augmentation)
+    """
+    def __init__(self, concat_keys=("lr",), reshuffle_patch_size=None,
+                 low_scale_config=None, low_scale_key=None, *args, **kwargs):
+        super().__init__(concat_keys=concat_keys, *args, **kwargs)
+        self.reshuffle_patch_size = reshuffle_patch_size
+        self.low_scale_model = None
+        if low_scale_config is not None:
+            print("Initializing a low-scale model")
+            assert exists(low_scale_key)
+            self.instantiate_low_stage(low_scale_config)
+            self.low_scale_key = low_scale_key
+
+    def instantiate_low_stage(self, config):
+        model = instantiate_from_config(config)
+        self.low_scale_model = model.eval()
+        self.low_scale_model.train = disabled_train
+        for param in self.low_scale_model.parameters():
+            param.requires_grad = False
+
+    @torch.no_grad()
+    def get_input(self, batch, k, cond_key=None, bs=None, return_first_stage_outputs=False):
+        # note: restricted to non-trainable encoders currently
+        assert not self.cond_stage_trainable, 'trainable cond stages not yet supported for upscaling-ft'
+        z, c, x, xrec, xc = super().get_input(batch, self.first_stage_key, return_first_stage_outputs=True,
+                                              force_c_encode=True, return_original_cond=True, bs=bs)
+
+        assert exists(self.concat_keys)
+        assert len(self.concat_keys) == 1
+        # optionally make spatial noise_level here
+        c_cat = list()
+        noise_level = None
+        for ck in self.concat_keys:
+            cc = batch[ck]
+            cc = rearrange(cc, 'b h w c -> b c h w')
+            if exists(self.reshuffle_patch_size):
+                assert isinstance(self.reshuffle_patch_size, int)
+                cc = rearrange(cc, 'b c (p1 h) (p2 w) -> b (p1 p2 c) h w',
+                               p1=self.reshuffle_patch_size, p2=self.reshuffle_patch_size)
+            if bs is not None:
+                cc = cc[:bs]
+                cc = cc.to(self.device)
+            if exists(self.low_scale_model) and ck == self.low_scale_key:
+                cc, noise_level = self.low_scale_model(cc)
+            c_cat.append(cc)
+        c_cat = torch.cat(c_cat, dim=1)
+        if exists(noise_level):
+            all_conds = {"c_concat": [c_cat], "c_crossattn": [c], "c_adm": noise_level}
+        else:
+            all_conds = {"c_concat": [c_cat], "c_crossattn": [c]}
+        if return_first_stage_outputs:
+            return z, all_conds, x, xrec, xc
+        return z, all_conds
+
+    @torch.no_grad()
+    def log_images(self, *args, **kwargs):
+        log = super().log_images(*args, **kwargs)
+        log["lr"] = rearrange(args[0]["lr"], 'b h w c -> b c h w')
+        return log
diff --git a/ldm/models/diffusion/dpm_solver/__init__.py b/ldm/models/diffusion/dpm_solver/__init__.py
new file mode 100644
index 0000000..7427f38
--- /dev/null
+++ b/ldm/models/diffusion/dpm_solver/__init__.py
@@ -0,0 +1 @@
+from .sampler import DPMSolverSampler
\ No newline at end of file
diff --git a/ldm/models/diffusion/dpm_solver/dpm_solver.py b/ldm/models/diffusion/dpm_solver/dpm_solver.py
new file mode 100644
index 0000000..095e5ba
--- /dev/null
+++ b/ldm/models/diffusion/dpm_solver/dpm_solver.py
@@ -0,0 +1,1154 @@
+import torch
+import torch.nn.functional as F
+import math
+from tqdm import tqdm
+
+
+class NoiseScheduleVP:
+    def __init__(
+            self,
+            schedule='discrete',
+            betas=None,
+            alphas_cumprod=None,
+            continuous_beta_0=0.1,
+            continuous_beta_1=20.,
+    ):
+        """Create a wrapper class for the forward SDE (VP type).
+        ***
+        Update: We support discrete-time diffusion models by implementing a picewise linear interpolation for log_alpha_t.
+                We recommend to use schedule='discrete' for the discrete-time diffusion models, especially for high-resolution images.
+        ***
+        The forward SDE ensures that the condition distribution q_{t|0}(x_t | x_0) = N ( alpha_t * x_0, sigma_t^2 * I ).
+        We further define lambda_t = log(alpha_t) - log(sigma_t), which is the half-logSNR (described in the DPM-Solver paper).
+        Therefore, we implement the functions for computing alpha_t, sigma_t and lambda_t. For t in [0, T], we have:
+            log_alpha_t = self.marginal_log_mean_coeff(t)
+            sigma_t = self.marginal_std(t)
+            lambda_t = self.marginal_lambda(t)
+        Moreover, as lambda(t) is an invertible function, we also support its inverse function:
+            t = self.inverse_lambda(lambda_t)
+        ===============================================================
+        We support both discrete-time DPMs (trained on n = 0, 1, ..., N-1) and continuous-time DPMs (trained on t in [t_0, T]).
+        1. For discrete-time DPMs:
+            For discrete-time DPMs trained on n = 0, 1, ..., N-1, we convert the discrete steps to continuous time steps by:
+                t_i = (i + 1) / N
+            e.g. for N = 1000, we have t_0 = 1e-3 and T = t_{N-1} = 1.
+            We solve the corresponding diffusion ODE from time T = 1 to time t_0 = 1e-3.
+            Args:
+                betas: A `torch.Tensor`. The beta array for the discrete-time DPM. (See the original DDPM paper for details)
+                alphas_cumprod: A `torch.Tensor`. The cumprod alphas for the discrete-time DPM. (See the original DDPM paper for details)
+            Note that we always have alphas_cumprod = cumprod(betas). Therefore, we only need to set one of `betas` and `alphas_cumprod`.
+            **Important**:  Please pay special attention for the args for `alphas_cumprod`:
+                The `alphas_cumprod` is the \hat{alpha_n} arrays in the notations of DDPM. Specifically, DDPMs assume that
+                    q_{t_n | 0}(x_{t_n} | x_0) = N ( \sqrt{\hat{alpha_n}} * x_0, (1 - \hat{alpha_n}) * I ).
+                Therefore, the notation \hat{alpha_n} is different from the notation alpha_t in DPM-Solver. In fact, we have
+                    alpha_{t_n} = \sqrt{\hat{alpha_n}},
+                and
+                    log(alpha_{t_n}) = 0.5 * log(\hat{alpha_n}).
+        2. For continuous-time DPMs:
+            We support two types of VPSDEs: linear (DDPM) and cosine (improved-DDPM). The hyperparameters for the noise
+            schedule are the default settings in DDPM and improved-DDPM:
+            Args:
+                beta_min: A `float` number. The smallest beta for the linear schedule.
+                beta_max: A `float` number. The largest beta for the linear schedule.
+                cosine_s: A `float` number. The hyperparameter in the cosine schedule.
+                cosine_beta_max: A `float` number. The hyperparameter in the cosine schedule.
+                T: A `float` number. The ending time of the forward process.
+        ===============================================================
+        Args:
+            schedule: A `str`. The noise schedule of the forward SDE. 'discrete' for discrete-time DPMs,
+                    'linear' or 'cosine' for continuous-time DPMs.
+        Returns:
+            A wrapper object of the forward SDE (VP type).
+
+        ===============================================================
+        Example:
+        # For discrete-time DPMs, given betas (the beta array for n = 0, 1, ..., N - 1):
+        >>> ns = NoiseScheduleVP('discrete', betas=betas)
+        # For discrete-time DPMs, given alphas_cumprod (the \hat{alpha_n} array for n = 0, 1, ..., N - 1):
+        >>> ns = NoiseScheduleVP('discrete', alphas_cumprod=alphas_cumprod)
+        # For continuous-time DPMs (VPSDE), linear schedule:
+        >>> ns = NoiseScheduleVP('linear', continuous_beta_0=0.1, continuous_beta_1=20.)
+        """
+
+        if schedule not in ['discrete', 'linear', 'cosine']:
+            raise ValueError(
+                "Unsupported noise schedule {}. The schedule needs to be 'discrete' or 'linear' or 'cosine'".format(
+                    schedule))
+
+        self.schedule = schedule
+        if schedule == 'discrete':
+            if betas is not None:
+                log_alphas = 0.5 * torch.log(1 - betas).cumsum(dim=0)
+            else:
+                assert alphas_cumprod is not None
+                log_alphas = 0.5 * torch.log(alphas_cumprod)
+            self.total_N = len(log_alphas)
+            self.T = 1.
+            self.t_array = torch.linspace(0., 1., self.total_N + 1)[1:].reshape((1, -1))
+            self.log_alpha_array = log_alphas.reshape((1, -1,))
+        else:
+            self.total_N = 1000
+            self.beta_0 = continuous_beta_0
+            self.beta_1 = continuous_beta_1
+            self.cosine_s = 0.008
+            self.cosine_beta_max = 999.
+            self.cosine_t_max = math.atan(self.cosine_beta_max * (1. + self.cosine_s) / math.pi) * 2. * (
+                        1. + self.cosine_s) / math.pi - self.cosine_s
+            self.cosine_log_alpha_0 = math.log(math.cos(self.cosine_s / (1. + self.cosine_s) * math.pi / 2.))
+            self.schedule = schedule
+            if schedule == 'cosine':
+                # For the cosine schedule, T = 1 will have numerical issues. So we manually set the ending time T.
+                # Note that T = 0.9946 may be not the optimal setting. However, we find it works well.
+                self.T = 0.9946
+            else:
+                self.T = 1.
+
+    def marginal_log_mean_coeff(self, t):
+        """
+        Compute log(alpha_t) of a given continuous-time label t in [0, T].
+        """
+        if self.schedule == 'discrete':
+            return interpolate_fn(t.reshape((-1, 1)), self.t_array.to(t.device),
+                                  self.log_alpha_array.to(t.device)).reshape((-1))
+        elif self.schedule == 'linear':
+            return -0.25 * t ** 2 * (self.beta_1 - self.beta_0) - 0.5 * t * self.beta_0
+        elif self.schedule == 'cosine':
+            log_alpha_fn = lambda s: torch.log(torch.cos((s + self.cosine_s) / (1. + self.cosine_s) * math.pi / 2.))
+            log_alpha_t = log_alpha_fn(t) - self.cosine_log_alpha_0
+            return log_alpha_t
+
+    def marginal_alpha(self, t):
+        """
+        Compute alpha_t of a given continuous-time label t in [0, T].
+        """
+        return torch.exp(self.marginal_log_mean_coeff(t))
+
+    def marginal_std(self, t):
+        """
+        Compute sigma_t of a given continuous-time label t in [0, T].
+        """
+        return torch.sqrt(1. - torch.exp(2. * self.marginal_log_mean_coeff(t)))
+
+    def marginal_lambda(self, t):
+        """
+        Compute lambda_t = log(alpha_t) - log(sigma_t) of a given continuous-time label t in [0, T].
+        """
+        log_mean_coeff = self.marginal_log_mean_coeff(t)
+        log_std = 0.5 * torch.log(1. - torch.exp(2. * log_mean_coeff))
+        return log_mean_coeff - log_std
+
+    def inverse_lambda(self, lamb):
+        """
+        Compute the continuous-time label t in [0, T] of a given half-logSNR lambda_t.
+        """
+        if self.schedule == 'linear':
+            tmp = 2. * (self.beta_1 - self.beta_0) * torch.logaddexp(-2. * lamb, torch.zeros((1,)).to(lamb))
+            Delta = self.beta_0 ** 2 + tmp
+            return tmp / (torch.sqrt(Delta) + self.beta_0) / (self.beta_1 - self.beta_0)
+        elif self.schedule == 'discrete':
+            log_alpha = -0.5 * torch.logaddexp(torch.zeros((1,)).to(lamb.device), -2. * lamb)
+            t = interpolate_fn(log_alpha.reshape((-1, 1)), torch.flip(self.log_alpha_array.to(lamb.device), [1]),
+                               torch.flip(self.t_array.to(lamb.device), [1]))
+            return t.reshape((-1,))
+        else:
+            log_alpha = -0.5 * torch.logaddexp(-2. * lamb, torch.zeros((1,)).to(lamb))
+            t_fn = lambda log_alpha_t: torch.arccos(torch.exp(log_alpha_t + self.cosine_log_alpha_0)) * 2. * (
+                        1. + self.cosine_s) / math.pi - self.cosine_s
+            t = t_fn(log_alpha)
+            return t
+
+
+def model_wrapper(
+        model,
+        noise_schedule,
+        model_type="noise",
+        model_kwargs={},
+        guidance_type="uncond",
+        condition=None,
+        unconditional_condition=None,
+        guidance_scale=1.,
+        classifier_fn=None,
+        classifier_kwargs={},
+):
+    """Create a wrapper function for the noise prediction model.
+    DPM-Solver needs to solve the continuous-time diffusion ODEs. For DPMs trained on discrete-time labels, we need to
+    firstly wrap the model function to a noise prediction model that accepts the continuous time as the input.
+    We support four types of the diffusion model by setting `model_type`:
+        1. "noise": noise prediction model. (Trained by predicting noise).
+        2. "x_start": data prediction model. (Trained by predicting the data x_0 at time 0).
+        3. "v": velocity prediction model. (Trained by predicting the velocity).
+            The "v" prediction is derivation detailed in Appendix D of [1], and is used in Imagen-Video [2].
+            [1] Salimans, Tim, and Jonathan Ho. "Progressive distillation for fast sampling of diffusion models."
+                arXiv preprint arXiv:2202.00512 (2022).
+            [2] Ho, Jonathan, et al. "Imagen Video: High Definition Video Generation with Diffusion Models."
+                arXiv preprint arXiv:2210.02303 (2022).
+
+        4. "score": marginal score function. (Trained by denoising score matching).
+            Note that the score function and the noise prediction model follows a simple relationship:
+            ```
+                noise(x_t, t) = -sigma_t * score(x_t, t)
+            ```
+    We support three types of guided sampling by DPMs by setting `guidance_type`:
+        1. "uncond": unconditional sampling by DPMs.
+            The input `model` has the following format:
+            ``
+                model(x, t_input, **model_kwargs) -> noise | x_start | v | score
+            ``
+        2. "classifier": classifier guidance sampling [3] by DPMs and another classifier.
+            The input `model` has the following format:
+            ``
+                model(x, t_input, **model_kwargs) -> noise | x_start | v | score
+            ``
+            The input `classifier_fn` has the following format:
+            ``
+                classifier_fn(x, t_input, cond, **classifier_kwargs) -> logits(x, t_input, cond)
+            ``
+            [3] P. Dhariwal and A. Q. Nichol, "Diffusion models beat GANs on image synthesis,"
+                in Advances in Neural Information Processing Systems, vol. 34, 2021, pp. 8780-8794.
+        3. "classifier-free": classifier-free guidance sampling by conditional DPMs.
+            The input `model` has the following format:
+            ``
+                model(x, t_input, cond, **model_kwargs) -> noise | x_start | v | score
+            ``
+            And if cond == `unconditional_condition`, the model output is the unconditional DPM output.
+            [4] Ho, Jonathan, and Tim Salimans. "Classifier-free diffusion guidance."
+                arXiv preprint arXiv:2207.12598 (2022).
+
+    The `t_input` is the time label of the model, which may be discrete-time labels (i.e. 0 to 999)
+    or continuous-time labels (i.e. epsilon to T).
+    We wrap the model function to accept only `x` and `t_continuous` as inputs, and outputs the predicted noise:
+    ``
+        def model_fn(x, t_continuous) -> noise:
+            t_input = get_model_input_time(t_continuous)
+            return noise_pred(model, x, t_input, **model_kwargs)
+    ``
+    where `t_continuous` is the continuous time labels (i.e. epsilon to T). And we use `model_fn` for DPM-Solver.
+    ===============================================================
+    Args:
+        model: A diffusion model with the corresponding format described above.
+        noise_schedule: A noise schedule object, such as NoiseScheduleVP.
+        model_type: A `str`. The parameterization type of the diffusion model.
+                    "noise" or "x_start" or "v" or "score".
+        model_kwargs: A `dict`. A dict for the other inputs of the model function.
+        guidance_type: A `str`. The type of the guidance for sampling.
+                    "uncond" or "classifier" or "classifier-free".
+        condition: A pytorch tensor. The condition for the guided sampling.
+                    Only used for "classifier" or "classifier-free" guidance type.
+        unconditional_condition: A pytorch tensor. The condition for the unconditional sampling.
+                    Only used for "classifier-free" guidance type.
+        guidance_scale: A `float`. The scale for the guided sampling.
+        classifier_fn: A classifier function. Only used for the classifier guidance.
+        classifier_kwargs: A `dict`. A dict for the other inputs of the classifier function.
+    Returns:
+        A noise prediction model that accepts the noised data and the continuous time as the inputs.
+    """
+
+    def get_model_input_time(t_continuous):
+        """
+        Convert the continuous-time `t_continuous` (in [epsilon, T]) to the model input time.
+        For discrete-time DPMs, we convert `t_continuous` in [1 / N, 1] to `t_input` in [0, 1000 * (N - 1) / N].
+        For continuous-time DPMs, we just use `t_continuous`.
+        """
+        if noise_schedule.schedule == 'discrete':
+            return (t_continuous - 1. / noise_schedule.total_N) * 1000.
+        else:
+            return t_continuous
+
+    def noise_pred_fn(x, t_continuous, cond=None):
+        if t_continuous.reshape((-1,)).shape[0] == 1:
+            t_continuous = t_continuous.expand((x.shape[0]))
+        t_input = get_model_input_time(t_continuous)
+        if cond is None:
+            output = model(x, t_input, **model_kwargs)
+        else:
+            output = model(x, t_input, cond, **model_kwargs)
+        if model_type == "noise":
+            return output
+        elif model_type == "x_start":
+            alpha_t, sigma_t = noise_schedule.marginal_alpha(t_continuous), noise_schedule.marginal_std(t_continuous)
+            dims = x.dim()
+            return (x - expand_dims(alpha_t, dims) * output) / expand_dims(sigma_t, dims)
+        elif model_type == "v":
+            alpha_t, sigma_t = noise_schedule.marginal_alpha(t_continuous), noise_schedule.marginal_std(t_continuous)
+            dims = x.dim()
+            return expand_dims(alpha_t, dims) * output + expand_dims(sigma_t, dims) * x
+        elif model_type == "score":
+            sigma_t = noise_schedule.marginal_std(t_continuous)
+            dims = x.dim()
+            return -expand_dims(sigma_t, dims) * output
+
+    def cond_grad_fn(x, t_input):
+        """
+        Compute the gradient of the classifier, i.e. nabla_{x} log p_t(cond | x_t).
+        """
+        with torch.enable_grad():
+            x_in = x.detach().requires_grad_(True)
+            log_prob = classifier_fn(x_in, t_input, condition, **classifier_kwargs)
+            return torch.autograd.grad(log_prob.sum(), x_in)[0]
+
+    def model_fn(x, t_continuous):
+        """
+        The noise predicition model function that is used for DPM-Solver.
+        """
+        if t_continuous.reshape((-1,)).shape[0] == 1:
+            t_continuous = t_continuous.expand((x.shape[0]))
+        if guidance_type == "uncond":
+            return noise_pred_fn(x, t_continuous)
+        elif guidance_type == "classifier":
+            assert classifier_fn is not None
+            t_input = get_model_input_time(t_continuous)
+            cond_grad = cond_grad_fn(x, t_input)
+            sigma_t = noise_schedule.marginal_std(t_continuous)
+            noise = noise_pred_fn(x, t_continuous)
+            return noise - guidance_scale * expand_dims(sigma_t, dims=cond_grad.dim()) * cond_grad
+        elif guidance_type == "classifier-free":
+            if guidance_scale == 1. or unconditional_condition is None:
+                return noise_pred_fn(x, t_continuous, cond=condition)
+            else:
+                x_in = torch.cat([x] * 2)
+                t_in = torch.cat([t_continuous] * 2)
+                c_in = torch.cat([unconditional_condition, condition])
+                noise_uncond, noise = noise_pred_fn(x_in, t_in, cond=c_in).chunk(2)
+                return noise_uncond + guidance_scale * (noise - noise_uncond)
+
+    assert model_type in ["noise", "x_start", "v"]
+    assert guidance_type in ["uncond", "classifier", "classifier-free"]
+    return model_fn
+
+
+class DPM_Solver:
+    def __init__(self, model_fn, noise_schedule, predict_x0=False, thresholding=False, max_val=1.):
+        """Construct a DPM-Solver.
+        We support both the noise prediction model ("predicting epsilon") and the data prediction model ("predicting x0").
+        If `predict_x0` is False, we use the solver for the noise prediction model (DPM-Solver).
+        If `predict_x0` is True, we use the solver for the data prediction model (DPM-Solver++).
+            In such case, we further support the "dynamic thresholding" in [1] when `thresholding` is True.
+            The "dynamic thresholding" can greatly improve the sample quality for pixel-space DPMs with large guidance scales.
+        Args:
+            model_fn: A noise prediction model function which accepts the continuous-time input (t in [epsilon, T]):
+                ``
+                def model_fn(x, t_continuous):
+                    return noise
+                ``
+            noise_schedule: A noise schedule object, such as NoiseScheduleVP.
+            predict_x0: A `bool`. If true, use the data prediction model; else, use the noise prediction model.
+            thresholding: A `bool`. Valid when `predict_x0` is True. Whether to use the "dynamic thresholding" in [1].
+            max_val: A `float`. Valid when both `predict_x0` and `thresholding` are True. The max value for thresholding.
+
+        [1] Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily Denton, Seyed Kamyar Seyed Ghasemipour, Burcu Karagol Ayan, S Sara Mahdavi, Rapha Gontijo Lopes, et al. Photorealistic text-to-image diffusion models with deep language understanding. arXiv preprint arXiv:2205.11487, 2022b.
+        """
+        self.model = model_fn
+        self.noise_schedule = noise_schedule
+        self.predict_x0 = predict_x0
+        self.thresholding = thresholding
+        self.max_val = max_val
+
+    def noise_prediction_fn(self, x, t):
+        """
+        Return the noise prediction model.
+        """
+        return self.model(x, t)
+
+    def data_prediction_fn(self, x, t):
+        """
+        Return the data prediction model (with thresholding).
+        """
+        noise = self.noise_prediction_fn(x, t)
+        dims = x.dim()
+        alpha_t, sigma_t = self.noise_schedule.marginal_alpha(t), self.noise_schedule.marginal_std(t)
+        x0 = (x - expand_dims(sigma_t, dims) * noise) / expand_dims(alpha_t, dims)
+        if self.thresholding:
+            p = 0.995  # A hyperparameter in the paper of "Imagen" [1].
+            s = torch.quantile(torch.abs(x0).reshape((x0.shape[0], -1)), p, dim=1)
+            s = expand_dims(torch.maximum(s, self.max_val * torch.ones_like(s).to(s.device)), dims)
+            x0 = torch.clamp(x0, -s, s) / s
+        return x0
+
+    def model_fn(self, x, t):
+        """
+        Convert the model to the noise prediction model or the data prediction model.
+        """
+        if self.predict_x0:
+            return self.data_prediction_fn(x, t)
+        else:
+            return self.noise_prediction_fn(x, t)
+
+    def get_time_steps(self, skip_type, t_T, t_0, N, device):
+        """Compute the intermediate time steps for sampling.
+        Args:
+            skip_type: A `str`. The type for the spacing of the time steps. We support three types:
+                - 'logSNR': uniform logSNR for the time steps.
+                - 'time_uniform': uniform time for the time steps. (**Recommended for high-resolutional data**.)
+                - 'time_quadratic': quadratic time for the time steps. (Used in DDIM for low-resolutional data.)
+            t_T: A `float`. The starting time of the sampling (default is T).
+            t_0: A `float`. The ending time of the sampling (default is epsilon).
+            N: A `int`. The total number of the spacing of the time steps.
+            device: A torch device.
+        Returns:
+            A pytorch tensor of the time steps, with the shape (N + 1,).
+        """
+        if skip_type == 'logSNR':
+            lambda_T = self.noise_schedule.marginal_lambda(torch.tensor(t_T).to(device))
+            lambda_0 = self.noise_schedule.marginal_lambda(torch.tensor(t_0).to(device))
+            logSNR_steps = torch.linspace(lambda_T.cpu().item(), lambda_0.cpu().item(), N + 1).to(device)
+            return self.noise_schedule.inverse_lambda(logSNR_steps)
+        elif skip_type == 'time_uniform':
+            return torch.linspace(t_T, t_0, N + 1).to(device)
+        elif skip_type == 'time_quadratic':
+            t_order = 2
+            t = torch.linspace(t_T ** (1. / t_order), t_0 ** (1. / t_order), N + 1).pow(t_order).to(device)
+            return t
+        else:
+            raise ValueError(
+                "Unsupported skip_type {}, need to be 'logSNR' or 'time_uniform' or 'time_quadratic'".format(skip_type))
+
+    def get_orders_and_timesteps_for_singlestep_solver(self, steps, order, skip_type, t_T, t_0, device):
+        """
+        Get the order of each step for sampling by the singlestep DPM-Solver.
+        We combine both DPM-Solver-1,2,3 to use all the function evaluations, which is named as "DPM-Solver-fast".
+        Given a fixed number of function evaluations by `steps`, the sampling procedure by DPM-Solver-fast is:
+            - If order == 1:
+                We take `steps` of DPM-Solver-1 (i.e. DDIM).
+            - If order == 2:
+                - Denote K = (steps // 2). We take K or (K + 1) intermediate time steps for sampling.
+                - If steps % 2 == 0, we use K steps of DPM-Solver-2.
+                - If steps % 2 == 1, we use K steps of DPM-Solver-2 and 1 step of DPM-Solver-1.
+            - If order == 3:
+                - Denote K = (steps // 3 + 1). We take K intermediate time steps for sampling.
+                - If steps % 3 == 0, we use (K - 2) steps of DPM-Solver-3, and 1 step of DPM-Solver-2 and 1 step of DPM-Solver-1.
+                - If steps % 3 == 1, we use (K - 1) steps of DPM-Solver-3 and 1 step of DPM-Solver-1.
+                - If steps % 3 == 2, we use (K - 1) steps of DPM-Solver-3 and 1 step of DPM-Solver-2.
+        ============================================
+        Args:
+            order: A `int`. The max order for the solver (2 or 3).
+            steps: A `int`. The total number of function evaluations (NFE).
+            skip_type: A `str`. The type for the spacing of the time steps. We support three types:
+                - 'logSNR': uniform logSNR for the time steps.
+                - 'time_uniform': uniform time for the time steps. (**Recommended for high-resolutional data**.)
+                - 'time_quadratic': quadratic time for the time steps. (Used in DDIM for low-resolutional data.)
+            t_T: A `float`. The starting time of the sampling (default is T).
+            t_0: A `float`. The ending time of the sampling (default is epsilon).
+            device: A torch device.
+        Returns:
+            orders: A list of the solver order of each step.
+        """
+        if order == 3:
+            K = steps // 3 + 1
+            if steps % 3 == 0:
+                orders = [3, ] * (K - 2) + [2, 1]
+            elif steps % 3 == 1:
+                orders = [3, ] * (K - 1) + [1]
+            else:
+                orders = [3, ] * (K - 1) + [2]
+        elif order == 2:
+            if steps % 2 == 0:
+                K = steps // 2
+                orders = [2, ] * K
+            else:
+                K = steps // 2 + 1
+                orders = [2, ] * (K - 1) + [1]
+        elif order == 1:
+            K = 1
+            orders = [1, ] * steps
+        else:
+            raise ValueError("'order' must be '1' or '2' or '3'.")
+        if skip_type == 'logSNR':
+            # To reproduce the results in DPM-Solver paper
+            timesteps_outer = self.get_time_steps(skip_type, t_T, t_0, K, device)
+        else:
+            timesteps_outer = self.get_time_steps(skip_type, t_T, t_0, steps, device)[
+                torch.cumsum(torch.tensor([0, ] + orders)).to(device)]
+        return timesteps_outer, orders
+
+    def denoise_to_zero_fn(self, x, s):
+        """
+        Denoise at the final step, which is equivalent to solve the ODE from lambda_s to infty by first-order discretization.
+        """
+        return self.data_prediction_fn(x, s)
+
+    def dpm_solver_first_update(self, x, s, t, model_s=None, return_intermediate=False):
+        """
+        DPM-Solver-1 (equivalent to DDIM) from time `s` to time `t`.
+        Args:
+            x: A pytorch tensor. The initial value at time `s`.
+            s: A pytorch tensor. The starting time, with the shape (x.shape[0],).
+            t: A pytorch tensor. The ending time, with the shape (x.shape[0],).
+            model_s: A pytorch tensor. The model function evaluated at time `s`.
+                If `model_s` is None, we evaluate the model by `x` and `s`; otherwise we directly use it.
+            return_intermediate: A `bool`. If true, also return the model value at time `s`.
+        Returns:
+            x_t: A pytorch tensor. The approximated solution at time `t`.
+        """
+        ns = self.noise_schedule
+        dims = x.dim()
+        lambda_s, lambda_t = ns.marginal_lambda(s), ns.marginal_lambda(t)
+        h = lambda_t - lambda_s
+        log_alpha_s, log_alpha_t = ns.marginal_log_mean_coeff(s), ns.marginal_log_mean_coeff(t)
+        sigma_s, sigma_t = ns.marginal_std(s), ns.marginal_std(t)
+        alpha_t = torch.exp(log_alpha_t)
+
+        if self.predict_x0:
+            phi_1 = torch.expm1(-h)
+            if model_s is None:
+                model_s = self.model_fn(x, s)
+            x_t = (
+                    expand_dims(sigma_t / sigma_s, dims) * x
+                    - expand_dims(alpha_t * phi_1, dims) * model_s
+            )
+            if return_intermediate:
+                return x_t, {'model_s': model_s}
+            else:
+                return x_t
+        else:
+            phi_1 = torch.expm1(h)
+            if model_s is None:
+                model_s = self.model_fn(x, s)
+            x_t = (
+                    expand_dims(torch.exp(log_alpha_t - log_alpha_s), dims) * x
+                    - expand_dims(sigma_t * phi_1, dims) * model_s
+            )
+            if return_intermediate:
+                return x_t, {'model_s': model_s}
+            else:
+                return x_t
+
+    def singlestep_dpm_solver_second_update(self, x, s, t, r1=0.5, model_s=None, return_intermediate=False,
+                                            solver_type='dpm_solver'):
+        """
+        Singlestep solver DPM-Solver-2 from time `s` to time `t`.
+        Args:
+            x: A pytorch tensor. The initial value at time `s`.
+            s: A pytorch tensor. The starting time, with the shape (x.shape[0],).
+            t: A pytorch tensor. The ending time, with the shape (x.shape[0],).
+            r1: A `float`. The hyperparameter of the second-order solver.
+            model_s: A pytorch tensor. The model function evaluated at time `s`.
+                If `model_s` is None, we evaluate the model by `x` and `s`; otherwise we directly use it.
+            return_intermediate: A `bool`. If true, also return the model value at time `s` and `s1` (the intermediate time).
+            solver_type: either 'dpm_solver' or 'taylor'. The type for the high-order solvers.
+                The type slightly impacts the performance. We recommend to use 'dpm_solver' type.
+        Returns:
+            x_t: A pytorch tensor. The approximated solution at time `t`.
+        """
+        if solver_type not in ['dpm_solver', 'taylor']:
+            raise ValueError("'solver_type' must be either 'dpm_solver' or 'taylor', got {}".format(solver_type))
+        if r1 is None:
+            r1 = 0.5
+        ns = self.noise_schedule
+        dims = x.dim()
+        lambda_s, lambda_t = ns.marginal_lambda(s), ns.marginal_lambda(t)
+        h = lambda_t - lambda_s
+        lambda_s1 = lambda_s + r1 * h
+        s1 = ns.inverse_lambda(lambda_s1)
+        log_alpha_s, log_alpha_s1, log_alpha_t = ns.marginal_log_mean_coeff(s), ns.marginal_log_mean_coeff(
+            s1), ns.marginal_log_mean_coeff(t)
+        sigma_s, sigma_s1, sigma_t = ns.marginal_std(s), ns.marginal_std(s1), ns.marginal_std(t)
+        alpha_s1, alpha_t = torch.exp(log_alpha_s1), torch.exp(log_alpha_t)
+
+        if self.predict_x0:
+            phi_11 = torch.expm1(-r1 * h)
+            phi_1 = torch.expm1(-h)
+
+            if model_s is None:
+                model_s = self.model_fn(x, s)
+            x_s1 = (
+                    expand_dims(sigma_s1 / sigma_s, dims) * x
+                    - expand_dims(alpha_s1 * phi_11, dims) * model_s
+            )
+            model_s1 = self.model_fn(x_s1, s1)
+            if solver_type == 'dpm_solver':
+                x_t = (
+                        expand_dims(sigma_t / sigma_s, dims) * x
+                        - expand_dims(alpha_t * phi_1, dims) * model_s
+                        - (0.5 / r1) * expand_dims(alpha_t * phi_1, dims) * (model_s1 - model_s)
+                )
+            elif solver_type == 'taylor':
+                x_t = (
+                        expand_dims(sigma_t / sigma_s, dims) * x
+                        - expand_dims(alpha_t * phi_1, dims) * model_s
+                        + (1. / r1) * expand_dims(alpha_t * ((torch.exp(-h) - 1.) / h + 1.), dims) * (
+                                    model_s1 - model_s)
+                )
+        else:
+            phi_11 = torch.expm1(r1 * h)
+            phi_1 = torch.expm1(h)
+
+            if model_s is None:
+                model_s = self.model_fn(x, s)
+            x_s1 = (
+                    expand_dims(torch.exp(log_alpha_s1 - log_alpha_s), dims) * x
+                    - expand_dims(sigma_s1 * phi_11, dims) * model_s
+            )
+            model_s1 = self.model_fn(x_s1, s1)
+            if solver_type == 'dpm_solver':
+                x_t = (
+                        expand_dims(torch.exp(log_alpha_t - log_alpha_s), dims) * x
+                        - expand_dims(sigma_t * phi_1, dims) * model_s
+                        - (0.5 / r1) * expand_dims(sigma_t * phi_1, dims) * (model_s1 - model_s)
+                )
+            elif solver_type == 'taylor':
+                x_t = (
+                        expand_dims(torch.exp(log_alpha_t - log_alpha_s), dims) * x
+                        - expand_dims(sigma_t * phi_1, dims) * model_s
+                        - (1. / r1) * expand_dims(sigma_t * ((torch.exp(h) - 1.) / h - 1.), dims) * (model_s1 - model_s)
+                )
+        if return_intermediate:
+            return x_t, {'model_s': model_s, 'model_s1': model_s1}
+        else:
+            return x_t
+
+    def singlestep_dpm_solver_third_update(self, x, s, t, r1=1. / 3., r2=2. / 3., model_s=None, model_s1=None,
+                                           return_intermediate=False, solver_type='dpm_solver'):
+        """
+        Singlestep solver DPM-Solver-3 from time `s` to time `t`.
+        Args:
+            x: A pytorch tensor. The initial value at time `s`.
+            s: A pytorch tensor. The starting time, with the shape (x.shape[0],).
+            t: A pytorch tensor. The ending time, with the shape (x.shape[0],).
+            r1: A `float`. The hyperparameter of the third-order solver.
+            r2: A `float`. The hyperparameter of the third-order solver.
+            model_s: A pytorch tensor. The model function evaluated at time `s`.
+                If `model_s` is None, we evaluate the model by `x` and `s`; otherwise we directly use it.
+            model_s1: A pytorch tensor. The model function evaluated at time `s1` (the intermediate time given by `r1`).
+                If `model_s1` is None, we evaluate the model at `s1`; otherwise we directly use it.
+            return_intermediate: A `bool`. If true, also return the model value at time `s`, `s1` and `s2` (the intermediate times).
+            solver_type: either 'dpm_solver' or 'taylor'. The type for the high-order solvers.
+                The type slightly impacts the performance. We recommend to use 'dpm_solver' type.
+        Returns:
+            x_t: A pytorch tensor. The approximated solution at time `t`.
+        """
+        if solver_type not in ['dpm_solver', 'taylor']:
+            raise ValueError("'solver_type' must be either 'dpm_solver' or 'taylor', got {}".format(solver_type))
+        if r1 is None:
+            r1 = 1. / 3.
+        if r2 is None:
+            r2 = 2. / 3.
+        ns = self.noise_schedule
+        dims = x.dim()
+        lambda_s, lambda_t = ns.marginal_lambda(s), ns.marginal_lambda(t)
+        h = lambda_t - lambda_s
+        lambda_s1 = lambda_s + r1 * h
+        lambda_s2 = lambda_s + r2 * h
+        s1 = ns.inverse_lambda(lambda_s1)
+        s2 = ns.inverse_lambda(lambda_s2)
+        log_alpha_s, log_alpha_s1, log_alpha_s2, log_alpha_t = ns.marginal_log_mean_coeff(
+            s), ns.marginal_log_mean_coeff(s1), ns.marginal_log_mean_coeff(s2), ns.marginal_log_mean_coeff(t)
+        sigma_s, sigma_s1, sigma_s2, sigma_t = ns.marginal_std(s), ns.marginal_std(s1), ns.marginal_std(
+            s2), ns.marginal_std(t)
+        alpha_s1, alpha_s2, alpha_t = torch.exp(log_alpha_s1), torch.exp(log_alpha_s2), torch.exp(log_alpha_t)
+
+        if self.predict_x0:
+            phi_11 = torch.expm1(-r1 * h)
+            phi_12 = torch.expm1(-r2 * h)
+            phi_1 = torch.expm1(-h)
+            phi_22 = torch.expm1(-r2 * h) / (r2 * h) + 1.
+            phi_2 = phi_1 / h + 1.
+            phi_3 = phi_2 / h - 0.5
+
+            if model_s is None:
+                model_s = self.model_fn(x, s)
+            if model_s1 is None:
+                x_s1 = (
+                        expand_dims(sigma_s1 / sigma_s, dims) * x
+                        - expand_dims(alpha_s1 * phi_11, dims) * model_s
+                )
+                model_s1 = self.model_fn(x_s1, s1)
+            x_s2 = (
+                    expand_dims(sigma_s2 / sigma_s, dims) * x
+                    - expand_dims(alpha_s2 * phi_12, dims) * model_s
+                    + r2 / r1 * expand_dims(alpha_s2 * phi_22, dims) * (model_s1 - model_s)
+            )
+            model_s2 = self.model_fn(x_s2, s2)
+            if solver_type == 'dpm_solver':
+                x_t = (
+                        expand_dims(sigma_t / sigma_s, dims) * x
+                        - expand_dims(alpha_t * phi_1, dims) * model_s
+                        + (1. / r2) * expand_dims(alpha_t * phi_2, dims) * (model_s2 - model_s)
+                )
+            elif solver_type == 'taylor':
+                D1_0 = (1. / r1) * (model_s1 - model_s)
+                D1_1 = (1. / r2) * (model_s2 - model_s)
+                D1 = (r2 * D1_0 - r1 * D1_1) / (r2 - r1)
+                D2 = 2. * (D1_1 - D1_0) / (r2 - r1)
+                x_t = (
+                        expand_dims(sigma_t / sigma_s, dims) * x
+                        - expand_dims(alpha_t * phi_1, dims) * model_s
+                        + expand_dims(alpha_t * phi_2, dims) * D1
+                        - expand_dims(alpha_t * phi_3, dims) * D2
+                )
+        else:
+            phi_11 = torch.expm1(r1 * h)
+            phi_12 = torch.expm1(r2 * h)
+            phi_1 = torch.expm1(h)
+            phi_22 = torch.expm1(r2 * h) / (r2 * h) - 1.
+            phi_2 = phi_1 / h - 1.
+            phi_3 = phi_2 / h - 0.5
+
+            if model_s is None:
+                model_s = self.model_fn(x, s)
+            if model_s1 is None:
+                x_s1 = (
+                        expand_dims(torch.exp(log_alpha_s1 - log_alpha_s), dims) * x
+                        - expand_dims(sigma_s1 * phi_11, dims) * model_s
+                )
+                model_s1 = self.model_fn(x_s1, s1)
+            x_s2 = (
+                    expand_dims(torch.exp(log_alpha_s2 - log_alpha_s), dims) * x
+                    - expand_dims(sigma_s2 * phi_12, dims) * model_s
+                    - r2 / r1 * expand_dims(sigma_s2 * phi_22, dims) * (model_s1 - model_s)
+            )
+            model_s2 = self.model_fn(x_s2, s2)
+            if solver_type == 'dpm_solver':
+                x_t = (
+                        expand_dims(torch.exp(log_alpha_t - log_alpha_s), dims) * x
+                        - expand_dims(sigma_t * phi_1, dims) * model_s
+                        - (1. / r2) * expand_dims(sigma_t * phi_2, dims) * (model_s2 - model_s)
+                )
+            elif solver_type == 'taylor':
+                D1_0 = (1. / r1) * (model_s1 - model_s)
+                D1_1 = (1. / r2) * (model_s2 - model_s)
+                D1 = (r2 * D1_0 - r1 * D1_1) / (r2 - r1)
+                D2 = 2. * (D1_1 - D1_0) / (r2 - r1)
+                x_t = (
+                        expand_dims(torch.exp(log_alpha_t - log_alpha_s), dims) * x
+                        - expand_dims(sigma_t * phi_1, dims) * model_s
+                        - expand_dims(sigma_t * phi_2, dims) * D1
+                        - expand_dims(sigma_t * phi_3, dims) * D2
+                )
+
+        if return_intermediate:
+            return x_t, {'model_s': model_s, 'model_s1': model_s1, 'model_s2': model_s2}
+        else:
+            return x_t
+
+    def multistep_dpm_solver_second_update(self, x, model_prev_list, t_prev_list, t, solver_type="dpm_solver"):
+        """
+        Multistep solver DPM-Solver-2 from time `t_prev_list[-1]` to time `t`.
+        Args:
+            x: A pytorch tensor. The initial value at time `s`.
+            model_prev_list: A list of pytorch tensor. The previous computed model values.
+            t_prev_list: A list of pytorch tensor. The previous times, each time has the shape (x.shape[0],)
+            t: A pytorch tensor. The ending time, with the shape (x.shape[0],).
+            solver_type: either 'dpm_solver' or 'taylor'. The type for the high-order solvers.
+                The type slightly impacts the performance. We recommend to use 'dpm_solver' type.
+        Returns:
+            x_t: A pytorch tensor. The approximated solution at time `t`.
+        """
+        if solver_type not in ['dpm_solver', 'taylor']:
+            raise ValueError("'solver_type' must be either 'dpm_solver' or 'taylor', got {}".format(solver_type))
+        ns = self.noise_schedule
+        dims = x.dim()
+        model_prev_1, model_prev_0 = model_prev_list
+        t_prev_1, t_prev_0 = t_prev_list
+        lambda_prev_1, lambda_prev_0, lambda_t = ns.marginal_lambda(t_prev_1), ns.marginal_lambda(
+            t_prev_0), ns.marginal_lambda(t)
+        log_alpha_prev_0, log_alpha_t = ns.marginal_log_mean_coeff(t_prev_0), ns.marginal_log_mean_coeff(t)
+        sigma_prev_0, sigma_t = ns.marginal_std(t_prev_0), ns.marginal_std(t)
+        alpha_t = torch.exp(log_alpha_t)
+
+        h_0 = lambda_prev_0 - lambda_prev_1
+        h = lambda_t - lambda_prev_0
+        r0 = h_0 / h
+        D1_0 = expand_dims(1. / r0, dims) * (model_prev_0 - model_prev_1)
+        if self.predict_x0:
+            if solver_type == 'dpm_solver':
+                x_t = (
+                        expand_dims(sigma_t / sigma_prev_0, dims) * x
+                        - expand_dims(alpha_t * (torch.exp(-h) - 1.), dims) * model_prev_0
+                        - 0.5 * expand_dims(alpha_t * (torch.exp(-h) - 1.), dims) * D1_0
+                )
+            elif solver_type == 'taylor':
+                x_t = (
+                        expand_dims(sigma_t / sigma_prev_0, dims) * x
+                        - expand_dims(alpha_t * (torch.exp(-h) - 1.), dims) * model_prev_0
+                        + expand_dims(alpha_t * ((torch.exp(-h) - 1.) / h + 1.), dims) * D1_0
+                )
+        else:
+            if solver_type == 'dpm_solver':
+                x_t = (
+                        expand_dims(torch.exp(log_alpha_t - log_alpha_prev_0), dims) * x
+                        - expand_dims(sigma_t * (torch.exp(h) - 1.), dims) * model_prev_0
+                        - 0.5 * expand_dims(sigma_t * (torch.exp(h) - 1.), dims) * D1_0
+                )
+            elif solver_type == 'taylor':
+                x_t = (
+                        expand_dims(torch.exp(log_alpha_t - log_alpha_prev_0), dims) * x
+                        - expand_dims(sigma_t * (torch.exp(h) - 1.), dims) * model_prev_0
+                        - expand_dims(sigma_t * ((torch.exp(h) - 1.) / h - 1.), dims) * D1_0
+                )
+        return x_t
+
+    def multistep_dpm_solver_third_update(self, x, model_prev_list, t_prev_list, t, solver_type='dpm_solver'):
+        """
+        Multistep solver DPM-Solver-3 from time `t_prev_list[-1]` to time `t`.
+        Args:
+            x: A pytorch tensor. The initial value at time `s`.
+            model_prev_list: A list of pytorch tensor. The previous computed model values.
+            t_prev_list: A list of pytorch tensor. The previous times, each time has the shape (x.shape[0],)
+            t: A pytorch tensor. The ending time, with the shape (x.shape[0],).
+            solver_type: either 'dpm_solver' or 'taylor'. The type for the high-order solvers.
+                The type slightly impacts the performance. We recommend to use 'dpm_solver' type.
+        Returns:
+            x_t: A pytorch tensor. The approximated solution at time `t`.
+        """
+        ns = self.noise_schedule
+        dims = x.dim()
+        model_prev_2, model_prev_1, model_prev_0 = model_prev_list
+        t_prev_2, t_prev_1, t_prev_0 = t_prev_list
+        lambda_prev_2, lambda_prev_1, lambda_prev_0, lambda_t = ns.marginal_lambda(t_prev_2), ns.marginal_lambda(
+            t_prev_1), ns.marginal_lambda(t_prev_0), ns.marginal_lambda(t)
+        log_alpha_prev_0, log_alpha_t = ns.marginal_log_mean_coeff(t_prev_0), ns.marginal_log_mean_coeff(t)
+        sigma_prev_0, sigma_t = ns.marginal_std(t_prev_0), ns.marginal_std(t)
+        alpha_t = torch.exp(log_alpha_t)
+
+        h_1 = lambda_prev_1 - lambda_prev_2
+        h_0 = lambda_prev_0 - lambda_prev_1
+        h = lambda_t - lambda_prev_0
+        r0, r1 = h_0 / h, h_1 / h
+        D1_0 = expand_dims(1. / r0, dims) * (model_prev_0 - model_prev_1)
+        D1_1 = expand_dims(1. / r1, dims) * (model_prev_1 - model_prev_2)
+        D1 = D1_0 + expand_dims(r0 / (r0 + r1), dims) * (D1_0 - D1_1)
+        D2 = expand_dims(1. / (r0 + r1), dims) * (D1_0 - D1_1)
+        if self.predict_x0:
+            x_t = (
+                    expand_dims(sigma_t / sigma_prev_0, dims) * x
+                    - expand_dims(alpha_t * (torch.exp(-h) - 1.), dims) * model_prev_0
+                    + expand_dims(alpha_t * ((torch.exp(-h) - 1.) / h + 1.), dims) * D1
+                    - expand_dims(alpha_t * ((torch.exp(-h) - 1. + h) / h ** 2 - 0.5), dims) * D2
+            )
+        else:
+            x_t = (
+                    expand_dims(torch.exp(log_alpha_t - log_alpha_prev_0), dims) * x
+                    - expand_dims(sigma_t * (torch.exp(h) - 1.), dims) * model_prev_0
+                    - expand_dims(sigma_t * ((torch.exp(h) - 1.) / h - 1.), dims) * D1
+                    - expand_dims(sigma_t * ((torch.exp(h) - 1. - h) / h ** 2 - 0.5), dims) * D2
+            )
+        return x_t
+
+    def singlestep_dpm_solver_update(self, x, s, t, order, return_intermediate=False, solver_type='dpm_solver', r1=None,
+                                     r2=None):
+        """
+        Singlestep DPM-Solver with the order `order` from time `s` to time `t`.
+        Args:
+            x: A pytorch tensor. The initial value at time `s`.
+            s: A pytorch tensor. The starting time, with the shape (x.shape[0],).
+            t: A pytorch tensor. The ending time, with the shape (x.shape[0],).
+            order: A `int`. The order of DPM-Solver. We only support order == 1 or 2 or 3.
+            return_intermediate: A `bool`. If true, also return the model value at time `s`, `s1` and `s2` (the intermediate times).
+            solver_type: either 'dpm_solver' or 'taylor'. The type for the high-order solvers.
+                The type slightly impacts the performance. We recommend to use 'dpm_solver' type.
+            r1: A `float`. The hyperparameter of the second-order or third-order solver.
+            r2: A `float`. The hyperparameter of the third-order solver.
+        Returns:
+            x_t: A pytorch tensor. The approximated solution at time `t`.
+        """
+        if order == 1:
+            return self.dpm_solver_first_update(x, s, t, return_intermediate=return_intermediate)
+        elif order == 2:
+            return self.singlestep_dpm_solver_second_update(x, s, t, return_intermediate=return_intermediate,
+                                                            solver_type=solver_type, r1=r1)
+        elif order == 3:
+            return self.singlestep_dpm_solver_third_update(x, s, t, return_intermediate=return_intermediate,
+                                                           solver_type=solver_type, r1=r1, r2=r2)
+        else:
+            raise ValueError("Solver order must be 1 or 2 or 3, got {}".format(order))
+
+    def multistep_dpm_solver_update(self, x, model_prev_list, t_prev_list, t, order, solver_type='dpm_solver'):
+        """
+        Multistep DPM-Solver with the order `order` from time `t_prev_list[-1]` to time `t`.
+        Args:
+            x: A pytorch tensor. The initial value at time `s`.
+            model_prev_list: A list of pytorch tensor. The previous computed model values.
+            t_prev_list: A list of pytorch tensor. The previous times, each time has the shape (x.shape[0],)
+            t: A pytorch tensor. The ending time, with the shape (x.shape[0],).
+            order: A `int`. The order of DPM-Solver. We only support order == 1 or 2 or 3.
+            solver_type: either 'dpm_solver' or 'taylor'. The type for the high-order solvers.
+                The type slightly impacts the performance. We recommend to use 'dpm_solver' type.
+        Returns:
+            x_t: A pytorch tensor. The approximated solution at time `t`.
+        """
+        if order == 1:
+            return self.dpm_solver_first_update(x, t_prev_list[-1], t, model_s=model_prev_list[-1])
+        elif order == 2:
+            return self.multistep_dpm_solver_second_update(x, model_prev_list, t_prev_list, t, solver_type=solver_type)
+        elif order == 3:
+            return self.multistep_dpm_solver_third_update(x, model_prev_list, t_prev_list, t, solver_type=solver_type)
+        else:
+            raise ValueError("Solver order must be 1 or 2 or 3, got {}".format(order))
+
+    def dpm_solver_adaptive(self, x, order, t_T, t_0, h_init=0.05, atol=0.0078, rtol=0.05, theta=0.9, t_err=1e-5,
+                            solver_type='dpm_solver'):
+        """
+        The adaptive step size solver based on singlestep DPM-Solver.
+        Args:
+            x: A pytorch tensor. The initial value at time `t_T`.
+            order: A `int`. The (higher) order of the solver. We only support order == 2 or 3.
+            t_T: A `float`. The starting time of the sampling (default is T).
+            t_0: A `float`. The ending time of the sampling (default is epsilon).
+            h_init: A `float`. The initial step size (for logSNR).
+            atol: A `float`. The absolute tolerance of the solver. For image data, the default setting is 0.0078, followed [1].
+            rtol: A `float`. The relative tolerance of the solver. The default setting is 0.05.
+            theta: A `float`. The safety hyperparameter for adapting the step size. The default setting is 0.9, followed [1].
+            t_err: A `float`. The tolerance for the time. We solve the diffusion ODE until the absolute error between the
+                current time and `t_0` is less than `t_err`. The default setting is 1e-5.
+            solver_type: either 'dpm_solver' or 'taylor'. The type for the high-order solvers.
+                The type slightly impacts the performance. We recommend to use 'dpm_solver' type.
+        Returns:
+            x_0: A pytorch tensor. The approximated solution at time `t_0`.
+        [1] A. Jolicoeur-Martineau, K. Li, R. Piché-Taillefer, T. Kachman, and I. Mitliagkas, "Gotta go fast when generating data with score-based models," arXiv preprint arXiv:2105.14080, 2021.
+        """
+        ns = self.noise_schedule
+        s = t_T * torch.ones((x.shape[0],)).to(x)
+        lambda_s = ns.marginal_lambda(s)
+        lambda_0 = ns.marginal_lambda(t_0 * torch.ones_like(s).to(x))
+        h = h_init * torch.ones_like(s).to(x)
+        x_prev = x
+        nfe = 0
+        if order == 2:
+            r1 = 0.5
+            lower_update = lambda x, s, t: self.dpm_solver_first_update(x, s, t, return_intermediate=True)
+            higher_update = lambda x, s, t, **kwargs: self.singlestep_dpm_solver_second_update(x, s, t, r1=r1,
+                                                                                               solver_type=solver_type,
+                                                                                               **kwargs)
+        elif order == 3:
+            r1, r2 = 1. / 3., 2. / 3.
+            lower_update = lambda x, s, t: self.singlestep_dpm_solver_second_update(x, s, t, r1=r1,
+                                                                                    return_intermediate=True,
+                                                                                    solver_type=solver_type)
+            higher_update = lambda x, s, t, **kwargs: self.singlestep_dpm_solver_third_update(x, s, t, r1=r1, r2=r2,
+                                                                                              solver_type=solver_type,
+                                                                                              **kwargs)
+        else:
+            raise ValueError("For adaptive step size solver, order must be 2 or 3, got {}".format(order))
+        while torch.abs((s - t_0)).mean() > t_err:
+            t = ns.inverse_lambda(lambda_s + h)
+            x_lower, lower_noise_kwargs = lower_update(x, s, t)
+            x_higher = higher_update(x, s, t, **lower_noise_kwargs)
+            delta = torch.max(torch.ones_like(x).to(x) * atol, rtol * torch.max(torch.abs(x_lower), torch.abs(x_prev)))
+            norm_fn = lambda v: torch.sqrt(torch.square(v.reshape((v.shape[0], -1))).mean(dim=-1, keepdim=True))
+            E = norm_fn((x_higher - x_lower) / delta).max()
+            if torch.all(E <= 1.):
+                x = x_higher
+                s = t
+                x_prev = x_lower
+                lambda_s = ns.marginal_lambda(s)
+            h = torch.min(theta * h * torch.float_power(E, -1. / order).float(), lambda_0 - lambda_s)
+            nfe += order
+        print('adaptive solver nfe', nfe)
+        return x
+
+    def sample(self, x, steps=20, t_start=None, t_end=None, order=3, skip_type='time_uniform',
+               method='singlestep', lower_order_final=True, denoise_to_zero=False, solver_type='dpm_solver',
+               atol=0.0078, rtol=0.05,
+               ):
+        """
+        Compute the sample at time `t_end` by DPM-Solver, given the initial `x` at time `t_start`.
+        =====================================================
+        We support the following algorithms for both noise prediction model and data prediction model:
+            - 'singlestep':
+                Singlestep DPM-Solver (i.e. "DPM-Solver-fast" in the paper), which combines different orders of singlestep DPM-Solver.
+                We combine all the singlestep solvers with order <= `order` to use up all the function evaluations (steps).
+                The total number of function evaluations (NFE) == `steps`.
+                Given a fixed NFE == `steps`, the sampling procedure is:
+                    - If `order` == 1:
+                        - Denote K = steps. We use K steps of DPM-Solver-1 (i.e. DDIM).
+                    - If `order` == 2:
+                        - Denote K = (steps // 2) + (steps % 2). We take K intermediate time steps for sampling.
+                        - If steps % 2 == 0, we use K steps of singlestep DPM-Solver-2.
+                        - If steps % 2 == 1, we use (K - 1) steps of singlestep DPM-Solver-2 and 1 step of DPM-Solver-1.
+                    - If `order` == 3:
+                        - Denote K = (steps // 3 + 1). We take K intermediate time steps for sampling.
+                        - If steps % 3 == 0, we use (K - 2) steps of singlestep DPM-Solver-3, and 1 step of singlestep DPM-Solver-2 and 1 step of DPM-Solver-1.
+                        - If steps % 3 == 1, we use (K - 1) steps of singlestep DPM-Solver-3 and 1 step of DPM-Solver-1.
+                        - If steps % 3 == 2, we use (K - 1) steps of singlestep DPM-Solver-3 and 1 step of singlestep DPM-Solver-2.
+            - 'multistep':
+                Multistep DPM-Solver with the order of `order`. The total number of function evaluations (NFE) == `steps`.
+                We initialize the first `order` values by lower order multistep solvers.
+                Given a fixed NFE == `steps`, the sampling procedure is:
+                    Denote K = steps.
+                    - If `order` == 1:
+                        - We use K steps of DPM-Solver-1 (i.e. DDIM).
+                    - If `order` == 2:
+                        - We firstly use 1 step of DPM-Solver-1, then use (K - 1) step of multistep DPM-Solver-2.
+                    - If `order` == 3:
+                        - We firstly use 1 step of DPM-Solver-1, then 1 step of multistep DPM-Solver-2, then (K - 2) step of multistep DPM-Solver-3.
+            - 'singlestep_fixed':
+                Fixed order singlestep DPM-Solver (i.e. DPM-Solver-1 or singlestep DPM-Solver-2 or singlestep DPM-Solver-3).
+                We use singlestep DPM-Solver-`order` for `order`=1 or 2 or 3, with total [`steps` // `order`] * `order` NFE.
+            - 'adaptive':
+                Adaptive step size DPM-Solver (i.e. "DPM-Solver-12" and "DPM-Solver-23" in the paper).
+                We ignore `steps` and use adaptive step size DPM-Solver with a higher order of `order`.
+                You can adjust the absolute tolerance `atol` and the relative tolerance `rtol` to balance the computatation costs
+                (NFE) and the sample quality.
+                    - If `order` == 2, we use DPM-Solver-12 which combines DPM-Solver-1 and singlestep DPM-Solver-2.
+                    - If `order` == 3, we use DPM-Solver-23 which combines singlestep DPM-Solver-2 and singlestep DPM-Solver-3.
+        =====================================================
+        Some advices for choosing the algorithm:
+            - For **unconditional sampling** or **guided sampling with small guidance scale** by DPMs:
+                Use singlestep DPM-Solver ("DPM-Solver-fast" in the paper) with `order = 3`.
+                e.g.
+                    >>> dpm_solver = DPM_Solver(model_fn, noise_schedule, predict_x0=False)
+                    >>> x_sample = dpm_solver.sample(x, steps=steps, t_start=t_start, t_end=t_end, order=3,
+                            skip_type='time_uniform', method='singlestep')
+            - For **guided sampling with large guidance scale** by DPMs:
+                Use multistep DPM-Solver with `predict_x0 = True` and `order = 2`.
+                e.g.
+                    >>> dpm_solver = DPM_Solver(model_fn, noise_schedule, predict_x0=True)
+                    >>> x_sample = dpm_solver.sample(x, steps=steps, t_start=t_start, t_end=t_end, order=2,
+                            skip_type='time_uniform', method='multistep')
+        We support three types of `skip_type`:
+            - 'logSNR': uniform logSNR for the time steps. **Recommended for low-resolutional images**
+            - 'time_uniform': uniform time for the time steps. **Recommended for high-resolutional images**.
+            - 'time_quadratic': quadratic time for the time steps.
+        =====================================================
+        Args:
+            x: A pytorch tensor. The initial value at time `t_start`
+                e.g. if `t_start` == T, then `x` is a sample from the standard normal distribution.
+            steps: A `int`. The total number of function evaluations (NFE).
+            t_start: A `float`. The starting time of the sampling.
+                If `T` is None, we use self.noise_schedule.T (default is 1.0).
+            t_end: A `float`. The ending time of the sampling.
+                If `t_end` is None, we use 1. / self.noise_schedule.total_N.
+                e.g. if total_N == 1000, we have `t_end` == 1e-3.
+                For discrete-time DPMs:
+                    - We recommend `t_end` == 1. / self.noise_schedule.total_N.
+                For continuous-time DPMs:
+                    - We recommend `t_end` == 1e-3 when `steps` <= 15; and `t_end` == 1e-4 when `steps` > 15.
+            order: A `int`. The order of DPM-Solver.
+            skip_type: A `str`. The type for the spacing of the time steps. 'time_uniform' or 'logSNR' or 'time_quadratic'.
+            method: A `str`. The method for sampling. 'singlestep' or 'multistep' or 'singlestep_fixed' or 'adaptive'.
+            denoise_to_zero: A `bool`. Whether to denoise to time 0 at the final step.
+                Default is `False`. If `denoise_to_zero` is `True`, the total NFE is (`steps` + 1).
+                This trick is firstly proposed by DDPM (https://arxiv.org/abs/2006.11239) and
+                score_sde (https://arxiv.org/abs/2011.13456). Such trick can improve the FID
+                for diffusion models sampling by diffusion SDEs for low-resolutional images
+                (such as CIFAR-10). However, we observed that such trick does not matter for
+                high-resolutional images. As it needs an additional NFE, we do not recommend
+                it for high-resolutional images.
+            lower_order_final: A `bool`. Whether to use lower order solvers at the final steps.
+                Only valid for `method=multistep` and `steps < 15`. We empirically find that
+                this trick is a key to stabilizing the sampling by DPM-Solver with very few steps
+                (especially for steps <= 10). So we recommend to set it to be `True`.
+            solver_type: A `str`. The taylor expansion type for the solver. `dpm_solver` or `taylor`. We recommend `dpm_solver`.
+            atol: A `float`. The absolute tolerance of the adaptive step size solver. Valid when `method` == 'adaptive'.
+            rtol: A `float`. The relative tolerance of the adaptive step size solver. Valid when `method` == 'adaptive'.
+        Returns:
+            x_end: A pytorch tensor. The approximated solution at time `t_end`.
+        """
+        t_0 = 1. / self.noise_schedule.total_N if t_end is None else t_end
+        t_T = self.noise_schedule.T if t_start is None else t_start
+        device = x.device
+        if method == 'adaptive':
+            with torch.no_grad():
+                x = self.dpm_solver_adaptive(x, order=order, t_T=t_T, t_0=t_0, atol=atol, rtol=rtol,
+                                             solver_type=solver_type)
+        elif method == 'multistep':
+            assert steps >= order
+            timesteps = self.get_time_steps(skip_type=skip_type, t_T=t_T, t_0=t_0, N=steps, device=device)
+            assert timesteps.shape[0] - 1 == steps
+            with torch.no_grad():
+                vec_t = timesteps[0].expand((x.shape[0]))
+                model_prev_list = [self.model_fn(x, vec_t)]
+                t_prev_list = [vec_t]
+                # Init the first `order` values by lower order multistep DPM-Solver.
+                for init_order in tqdm(range(1, order), desc="DPM init order"):
+                    vec_t = timesteps[init_order].expand(x.shape[0])
+                    x = self.multistep_dpm_solver_update(x, model_prev_list, t_prev_list, vec_t, init_order,
+                                                         solver_type=solver_type)
+                    model_prev_list.append(self.model_fn(x, vec_t))
+                    t_prev_list.append(vec_t)
+                # Compute the remaining values by `order`-th order multistep DPM-Solver.
+                for step in tqdm(range(order, steps + 1), desc="DPM multistep"):
+                    vec_t = timesteps[step].expand(x.shape[0])
+                    if lower_order_final and steps < 15:
+                        step_order = min(order, steps + 1 - step)
+                    else:
+                        step_order = order
+                    x = self.multistep_dpm_solver_update(x, model_prev_list, t_prev_list, vec_t, step_order,
+                                                         solver_type=solver_type)
+                    for i in range(order - 1):
+                        t_prev_list[i] = t_prev_list[i + 1]
+                        model_prev_list[i] = model_prev_list[i + 1]
+                    t_prev_list[-1] = vec_t
+                    # We do not need to evaluate the final model value.
+                    if step < steps:
+                        model_prev_list[-1] = self.model_fn(x, vec_t)
+        elif method in ['singlestep', 'singlestep_fixed']:
+            if method == 'singlestep':
+                timesteps_outer, orders = self.get_orders_and_timesteps_for_singlestep_solver(steps=steps, order=order,
+                                                                                              skip_type=skip_type,
+                                                                                              t_T=t_T, t_0=t_0,
+                                                                                              device=device)
+            elif method == 'singlestep_fixed':
+                K = steps // order
+                orders = [order, ] * K
+                timesteps_outer = self.get_time_steps(skip_type=skip_type, t_T=t_T, t_0=t_0, N=K, device=device)
+            for i, order in enumerate(orders):
+                t_T_inner, t_0_inner = timesteps_outer[i], timesteps_outer[i + 1]
+                timesteps_inner = self.get_time_steps(skip_type=skip_type, t_T=t_T_inner.item(), t_0=t_0_inner.item(),
+                                                      N=order, device=device)
+                lambda_inner = self.noise_schedule.marginal_lambda(timesteps_inner)
+                vec_s, vec_t = t_T_inner.tile(x.shape[0]), t_0_inner.tile(x.shape[0])
+                h = lambda_inner[-1] - lambda_inner[0]
+                r1 = None if order <= 1 else (lambda_inner[1] - lambda_inner[0]) / h
+                r2 = None if order <= 2 else (lambda_inner[2] - lambda_inner[0]) / h
+                x = self.singlestep_dpm_solver_update(x, vec_s, vec_t, order, solver_type=solver_type, r1=r1, r2=r2)
+        if denoise_to_zero:
+            x = self.denoise_to_zero_fn(x, torch.ones((x.shape[0],)).to(device) * t_0)
+        return x
+
+
+#############################################################
+# other utility functions
+#############################################################
+
+def interpolate_fn(x, xp, yp):
+    """
+    A piecewise linear function y = f(x), using xp and yp as keypoints.
+    We implement f(x) in a differentiable way (i.e. applicable for autograd).
+    The function f(x) is well-defined for all x-axis. (For x beyond the bounds of xp, we use the outmost points of xp to define the linear function.)
+    Args:
+        x: PyTorch tensor with shape [N, C], where N is the batch size, C is the number of channels (we use C = 1 for DPM-Solver).
+        xp: PyTorch tensor with shape [C, K], where K is the number of keypoints.
+        yp: PyTorch tensor with shape [C, K].
+    Returns:
+        The function values f(x), with shape [N, C].
+    """
+    N, K = x.shape[0], xp.shape[1]
+    all_x = torch.cat([x.unsqueeze(2), xp.unsqueeze(0).repeat((N, 1, 1))], dim=2)
+    sorted_all_x, x_indices = torch.sort(all_x, dim=2)
+    x_idx = torch.argmin(x_indices, dim=2)
+    cand_start_idx = x_idx - 1
+    start_idx = torch.where(
+        torch.eq(x_idx, 0),
+        torch.tensor(1, device=x.device),
+        torch.where(
+            torch.eq(x_idx, K), torch.tensor(K - 2, device=x.device), cand_start_idx,
+        ),
+    )
+    end_idx = torch.where(torch.eq(start_idx, cand_start_idx), start_idx + 2, start_idx + 1)
+    start_x = torch.gather(sorted_all_x, dim=2, index=start_idx.unsqueeze(2)).squeeze(2)
+    end_x = torch.gather(sorted_all_x, dim=2, index=end_idx.unsqueeze(2)).squeeze(2)
+    start_idx2 = torch.where(
+        torch.eq(x_idx, 0),
+        torch.tensor(0, device=x.device),
+        torch.where(
+            torch.eq(x_idx, K), torch.tensor(K - 2, device=x.device), cand_start_idx,
+        ),
+    )
+    y_positions_expanded = yp.unsqueeze(0).expand(N, -1, -1)
+    start_y = torch.gather(y_positions_expanded, dim=2, index=start_idx2.unsqueeze(2)).squeeze(2)
+    end_y = torch.gather(y_positions_expanded, dim=2, index=(start_idx2 + 1).unsqueeze(2)).squeeze(2)
+    cand = start_y + (x - start_x) * (end_y - start_y) / (end_x - start_x)
+    return cand
+
+
+def expand_dims(v, dims):
+    """
+    Expand the tensor `v` to the dim `dims`.
+    Args:
+        `v`: a PyTorch tensor with shape [N].
+        `dim`: a `int`.
+    Returns:
+        a PyTorch tensor with shape [N, 1, 1, ..., 1] and the total dimension is `dims`.
+    """
+    return v[(...,) + (None,) * (dims - 1)]
\ No newline at end of file
diff --git a/ldm/models/diffusion/dpm_solver/sampler.py b/ldm/models/diffusion/dpm_solver/sampler.py
new file mode 100644
index 0000000..7d137b8
--- /dev/null
+++ b/ldm/models/diffusion/dpm_solver/sampler.py
@@ -0,0 +1,87 @@
+"""SAMPLING ONLY."""
+import torch
+
+from .dpm_solver import NoiseScheduleVP, model_wrapper, DPM_Solver
+
+
+MODEL_TYPES = {
+    "eps": "noise",
+    "v": "v"
+}
+
+
+class DPMSolverSampler(object):
+    def __init__(self, model, **kwargs):
+        super().__init__()
+        self.model = model
+        to_torch = lambda x: x.clone().detach().to(torch.float32).to(model.device)
+        self.register_buffer('alphas_cumprod', to_torch(model.alphas_cumprod))
+
+    def register_buffer(self, name, attr):
+        if type(attr) == torch.Tensor:
+            if attr.device != torch.device("cuda"):
+                attr = attr.to(torch.device("cuda"))
+        setattr(self, name, attr)
+
+    @torch.no_grad()
+    def sample(self,
+               S,
+               batch_size,
+               shape,
+               conditioning=None,
+               callback=None,
+               normals_sequence=None,
+               img_callback=None,
+               quantize_x0=False,
+               eta=0.,
+               mask=None,
+               x0=None,
+               temperature=1.,
+               noise_dropout=0.,
+               score_corrector=None,
+               corrector_kwargs=None,
+               verbose=True,
+               x_T=None,
+               log_every_t=100,
+               unconditional_guidance_scale=1.,
+               unconditional_conditioning=None,
+               # this has to come in the same format as the conditioning, # e.g. as encoded tokens, ...
+               **kwargs
+               ):
+        if conditioning is not None:
+            if isinstance(conditioning, dict):
+                cbs = conditioning[list(conditioning.keys())[0]].shape[0]
+                if cbs != batch_size:
+                    print(f"Warning: Got {cbs} conditionings but batch-size is {batch_size}")
+            else:
+                if conditioning.shape[0] != batch_size:
+                    print(f"Warning: Got {conditioning.shape[0]} conditionings but batch-size is {batch_size}")
+
+        # sampling
+        C, H, W = shape
+        size = (batch_size, C, H, W)
+
+        print(f'Data shape for DPM-Solver sampling is {size}, sampling steps {S}')
+
+        device = self.model.betas.device
+        if x_T is None:
+            img = torch.randn(size, device=device)
+        else:
+            img = x_T
+
+        ns = NoiseScheduleVP('discrete', alphas_cumprod=self.alphas_cumprod)
+
+        model_fn = model_wrapper(
+            lambda x, t, c: self.model.apply_model(x, t, c),
+            ns,
+            model_type=MODEL_TYPES[self.model.parameterization],
+            guidance_type="classifier-free",
+            condition=conditioning,
+            unconditional_condition=unconditional_conditioning,
+            guidance_scale=unconditional_guidance_scale,
+        )
+
+        dpm_solver = DPM_Solver(model_fn, ns, predict_x0=True, thresholding=False)
+        x = dpm_solver.sample(img, steps=S, skip_type="time_uniform", method="multistep", order=2, lower_order_final=True)
+
+        return x.to(device), None
\ No newline at end of file
diff --git a/ldm/models/diffusion/plms.py b/ldm/models/diffusion/plms.py
new file mode 100644
index 0000000..7002a36
--- /dev/null
+++ b/ldm/models/diffusion/plms.py
@@ -0,0 +1,244 @@
+"""SAMPLING ONLY."""
+
+import torch
+import numpy as np
+from tqdm import tqdm
+from functools import partial
+
+from ldm.modules.diffusionmodules.util import make_ddim_sampling_parameters, make_ddim_timesteps, noise_like
+from ldm.models.diffusion.sampling_util import norm_thresholding
+
+
+class PLMSSampler(object):
+    def __init__(self, model, schedule="linear", **kwargs):
+        super().__init__()
+        self.model = model
+        self.ddpm_num_timesteps = model.num_timesteps
+        self.schedule = schedule
+
+    def register_buffer(self, name, attr):
+        if type(attr) == torch.Tensor:
+            if attr.device != torch.device("cuda"):
+                attr = attr.to(torch.device("cuda"))
+        setattr(self, name, attr)
+
+    def make_schedule(self, ddim_num_steps, ddim_discretize="uniform", ddim_eta=0., verbose=True):
+        if ddim_eta != 0:
+            raise ValueError('ddim_eta must be 0 for PLMS')
+        self.ddim_timesteps = make_ddim_timesteps(ddim_discr_method=ddim_discretize, num_ddim_timesteps=ddim_num_steps,
+                                                  num_ddpm_timesteps=self.ddpm_num_timesteps,verbose=verbose)
+        alphas_cumprod = self.model.alphas_cumprod
+        assert alphas_cumprod.shape[0] == self.ddpm_num_timesteps, 'alphas have to be defined for each timestep'
+        to_torch = lambda x: x.clone().detach().to(torch.float32).to(self.model.device)
+
+        self.register_buffer('betas', to_torch(self.model.betas))
+        self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
+        self.register_buffer('alphas_cumprod_prev', to_torch(self.model.alphas_cumprod_prev))
+
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod.cpu())))
+        self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod.cpu())))
+        self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod.cpu())))
+        self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod.cpu())))
+        self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod.cpu() - 1)))
+
+        # ddim sampling parameters
+        ddim_sigmas, ddim_alphas, ddim_alphas_prev = make_ddim_sampling_parameters(alphacums=alphas_cumprod.cpu(),
+                                                                                   ddim_timesteps=self.ddim_timesteps,
+                                                                                   eta=ddim_eta,verbose=verbose)
+        self.register_buffer('ddim_sigmas', ddim_sigmas)
+        self.register_buffer('ddim_alphas', ddim_alphas)
+        self.register_buffer('ddim_alphas_prev', ddim_alphas_prev)
+        self.register_buffer('ddim_sqrt_one_minus_alphas', np.sqrt(1. - ddim_alphas))
+        sigmas_for_original_sampling_steps = ddim_eta * torch.sqrt(
+            (1 - self.alphas_cumprod_prev) / (1 - self.alphas_cumprod) * (
+                        1 - self.alphas_cumprod / self.alphas_cumprod_prev))
+        self.register_buffer('ddim_sigmas_for_original_num_steps', sigmas_for_original_sampling_steps)
+
+    @torch.no_grad()
+    def sample(self,
+               S,
+               batch_size,
+               shape,
+               conditioning=None,
+               callback=None,
+               normals_sequence=None,
+               img_callback=None,
+               quantize_x0=False,
+               eta=0.,
+               mask=None,
+               x0=None,
+               temperature=1.,
+               noise_dropout=0.,
+               score_corrector=None,
+               corrector_kwargs=None,
+               verbose=True,
+               x_T=None,
+               log_every_t=100,
+               unconditional_guidance_scale=1.,
+               unconditional_conditioning=None,
+               # this has to come in the same format as the conditioning, # e.g. as encoded tokens, ...
+               dynamic_threshold=None,
+               **kwargs
+               ):
+        if conditioning is not None:
+            if isinstance(conditioning, dict):
+                cbs = conditioning[list(conditioning.keys())[0]].shape[0]
+                if cbs != batch_size:
+                    print(f"Warning: Got {cbs} conditionings but batch-size is {batch_size}")
+            else:
+                if conditioning.shape[0] != batch_size:
+                    print(f"Warning: Got {conditioning.shape[0]} conditionings but batch-size is {batch_size}")
+
+        self.make_schedule(ddim_num_steps=S, ddim_eta=eta, verbose=verbose)
+        # sampling
+        C, H, W = shape
+        size = (batch_size, C, H, W)
+        print(f'Data shape for PLMS sampling is {size}')
+
+        samples, intermediates = self.plms_sampling(conditioning, size,
+                                                    callback=callback,
+                                                    img_callback=img_callback,
+                                                    quantize_denoised=quantize_x0,
+                                                    mask=mask, x0=x0,
+                                                    ddim_use_original_steps=False,
+                                                    noise_dropout=noise_dropout,
+                                                    temperature=temperature,
+                                                    score_corrector=score_corrector,
+                                                    corrector_kwargs=corrector_kwargs,
+                                                    x_T=x_T,
+                                                    log_every_t=log_every_t,
+                                                    unconditional_guidance_scale=unconditional_guidance_scale,
+                                                    unconditional_conditioning=unconditional_conditioning,
+                                                    dynamic_threshold=dynamic_threshold,
+                                                    )
+        return samples, intermediates
+
+    @torch.no_grad()
+    def plms_sampling(self, cond, shape,
+                      x_T=None, ddim_use_original_steps=False,
+                      callback=None, timesteps=None, quantize_denoised=False,
+                      mask=None, x0=None, img_callback=None, log_every_t=100,
+                      temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None,
+                      unconditional_guidance_scale=1., unconditional_conditioning=None,
+                      dynamic_threshold=None):
+        device = self.model.betas.device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+
+        if timesteps is None:
+            timesteps = self.ddpm_num_timesteps if ddim_use_original_steps else self.ddim_timesteps
+        elif timesteps is not None and not ddim_use_original_steps:
+            subset_end = int(min(timesteps / self.ddim_timesteps.shape[0], 1) * self.ddim_timesteps.shape[0]) - 1
+            timesteps = self.ddim_timesteps[:subset_end]
+
+        intermediates = {'x_inter': [img], 'pred_x0': [img]}
+        time_range = list(reversed(range(0,timesteps))) if ddim_use_original_steps else np.flip(timesteps)
+        total_steps = timesteps if ddim_use_original_steps else timesteps.shape[0]
+        print(f"Running PLMS Sampling with {total_steps} timesteps")
+
+        iterator = tqdm(time_range, desc='PLMS Sampler', total=total_steps)
+        old_eps = []
+
+        for i, step in enumerate(iterator):
+            index = total_steps - i - 1
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+            ts_next = torch.full((b,), time_range[min(i + 1, len(time_range) - 1)], device=device, dtype=torch.long)
+
+            if mask is not None:
+                assert x0 is not None
+                img_orig = self.model.q_sample(x0, ts)  # TODO: deterministic forward pass?
+                img = img_orig * mask + (1. - mask) * img
+
+            outs = self.p_sample_plms(img, cond, ts, index=index, use_original_steps=ddim_use_original_steps,
+                                      quantize_denoised=quantize_denoised, temperature=temperature,
+                                      noise_dropout=noise_dropout, score_corrector=score_corrector,
+                                      corrector_kwargs=corrector_kwargs,
+                                      unconditional_guidance_scale=unconditional_guidance_scale,
+                                      unconditional_conditioning=unconditional_conditioning,
+                                      old_eps=old_eps, t_next=ts_next,
+                                      dynamic_threshold=dynamic_threshold)
+            img, pred_x0, e_t = outs
+            old_eps.append(e_t)
+            if len(old_eps) >= 4:
+                old_eps.pop(0)
+            if callback: callback(i)
+            if img_callback: img_callback(pred_x0, i)
+
+            if index % log_every_t == 0 or index == total_steps - 1:
+                intermediates['x_inter'].append(img)
+                intermediates['pred_x0'].append(pred_x0)
+
+        return img, intermediates
+
+    @torch.no_grad()
+    def p_sample_plms(self, x, c, t, index, repeat_noise=False, use_original_steps=False, quantize_denoised=False,
+                      temperature=1., noise_dropout=0., score_corrector=None, corrector_kwargs=None,
+                      unconditional_guidance_scale=1., unconditional_conditioning=None, old_eps=None, t_next=None,
+                      dynamic_threshold=None):
+        b, *_, device = *x.shape, x.device
+
+        def get_model_output(x, t):
+            if unconditional_conditioning is None or unconditional_guidance_scale == 1.:
+                e_t = self.model.apply_model(x, t, c)
+            else:
+                x_in = torch.cat([x] * 2)
+                t_in = torch.cat([t] * 2)
+                c_in = torch.cat([unconditional_conditioning, c])
+                e_t_uncond, e_t = self.model.apply_model(x_in, t_in, c_in).chunk(2)
+                e_t = e_t_uncond + unconditional_guidance_scale * (e_t - e_t_uncond)
+
+            if score_corrector is not None:
+                assert self.model.parameterization == "eps"
+                e_t = score_corrector.modify_score(self.model, e_t, x, t, c, **corrector_kwargs)
+
+            return e_t
+
+        alphas = self.model.alphas_cumprod if use_original_steps else self.ddim_alphas
+        alphas_prev = self.model.alphas_cumprod_prev if use_original_steps else self.ddim_alphas_prev
+        sqrt_one_minus_alphas = self.model.sqrt_one_minus_alphas_cumprod if use_original_steps else self.ddim_sqrt_one_minus_alphas
+        sigmas = self.model.ddim_sigmas_for_original_num_steps if use_original_steps else self.ddim_sigmas
+
+        def get_x_prev_and_pred_x0(e_t, index):
+            # select parameters corresponding to the currently considered timestep
+            a_t = torch.full((b, 1, 1, 1), alphas[index], device=device)
+            a_prev = torch.full((b, 1, 1, 1), alphas_prev[index], device=device)
+            sigma_t = torch.full((b, 1, 1, 1), sigmas[index], device=device)
+            sqrt_one_minus_at = torch.full((b, 1, 1, 1), sqrt_one_minus_alphas[index],device=device)
+
+            # current prediction for x_0
+            pred_x0 = (x - sqrt_one_minus_at * e_t) / a_t.sqrt()
+            if quantize_denoised:
+                pred_x0, _, *_ = self.model.first_stage_model.quantize(pred_x0)
+            if dynamic_threshold is not None:
+                pred_x0 = norm_thresholding(pred_x0, dynamic_threshold)
+            # direction pointing to x_t
+            dir_xt = (1. - a_prev - sigma_t**2).sqrt() * e_t
+            noise = sigma_t * noise_like(x.shape, device, repeat_noise) * temperature
+            if noise_dropout > 0.:
+                noise = torch.nn.functional.dropout(noise, p=noise_dropout)
+            x_prev = a_prev.sqrt() * pred_x0 + dir_xt + noise
+            return x_prev, pred_x0
+
+        e_t = get_model_output(x, t)
+        if len(old_eps) == 0:
+            # Pseudo Improved Euler (2nd order)
+            x_prev, pred_x0 = get_x_prev_and_pred_x0(e_t, index)
+            e_t_next = get_model_output(x_prev, t_next)
+            e_t_prime = (e_t + e_t_next) / 2
+        elif len(old_eps) == 1:
+            # 2nd order Pseudo Linear Multistep (Adams-Bashforth)
+            e_t_prime = (3 * e_t - old_eps[-1]) / 2
+        elif len(old_eps) == 2:
+            # 3nd order Pseudo Linear Multistep (Adams-Bashforth)
+            e_t_prime = (23 * e_t - 16 * old_eps[-1] + 5 * old_eps[-2]) / 12
+        elif len(old_eps) >= 3:
+            # 4nd order Pseudo Linear Multistep (Adams-Bashforth)
+            e_t_prime = (55 * e_t - 59 * old_eps[-1] + 37 * old_eps[-2] - 9 * old_eps[-3]) / 24
+
+        x_prev, pred_x0 = get_x_prev_and_pred_x0(e_t_prime, index)
+
+        return x_prev, pred_x0, e_t
diff --git a/ldm/models/diffusion/sampling_util.py b/ldm/models/diffusion/sampling_util.py
new file mode 100644
index 0000000..7eff02b
--- /dev/null
+++ b/ldm/models/diffusion/sampling_util.py
@@ -0,0 +1,22 @@
+import torch
+import numpy as np
+
+
+def append_dims(x, target_dims):
+    """Appends dimensions to the end of a tensor until it has target_dims dimensions.
+    From https://github.com/crowsonkb/k-diffusion/blob/master/k_diffusion/utils.py"""
+    dims_to_append = target_dims - x.ndim
+    if dims_to_append < 0:
+        raise ValueError(f'input has {x.ndim} dims but target_dims is {target_dims}, which is less')
+    return x[(...,) + (None,) * dims_to_append]
+
+
+def norm_thresholding(x0, value):
+    s = append_dims(x0.pow(2).flatten(1).mean(1).sqrt().clamp(min=value), x0.ndim)
+    return x0 * (value / s)
+
+
+def spatial_norm_thresholding(x0, value):
+    # b c h w
+    s = x0.pow(2).mean(1, keepdim=True).sqrt().clamp(min=value)
+    return x0 * (value / s)
\ No newline at end of file
diff --git a/ldm/modules/__pycache__/attention.cpython-310.pyc b/ldm/modules/__pycache__/attention.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6bdf92a74601a56f9a79948a79d3dbd648b35ec5
GIT binary patch
literal 10381
zcma)C+ix4$d7m>k4u=#aQSx<d?|5%nHnto)+2&Hm-Hq4UY_qFvV+ETqyCEo^kwl6&
zpBehXQfV45VCBJ0)3oR%2_T_pQMW~#UeXqAf$URI^grl4wdh0Phd$*Yz+!jZ-|q}5
zQnD13#C&t+n{)rZ^ZUN@o#{+X`38RPJo>G0;euiOHy<W{HXdHY6)g(GkcJdZqb15h
zFl{!?vZ?P@+0u8r<(P&rT5dU`b68EQ<&>S4SN8O?-ORT9vfs*;bNbn7=37(csaBy}
z(9dpjx;0at(dkTcwpA<_b=qvsmFL2_#{BjX>B;PlSw1R#nR~}5ACq}Gh5K<?kkh!I
zkTY@?_mi?H=Wu`G6+_O;BkvgUNa!?9-L=Z6!8t0AfpcugIU|qD6Uh6dJSm^RJwT3A
z@-%Xs9_Ba;&Kdb6I8P2aPk|H2v*4T^a-Ig~Dfu)wPY*fIfb)!e7My2?oM)wZ-6%cx
zkTYBoiC>Rad+l0W@3fanW@5J6i5u42QLlxktO~12RoiP};;OJ4R%0-|H@eJSZ6?-B
z?K`ZZwjS0tx}ADE##7<;YDcv~6;-0R8i(b=OUuhkm0$hhwX3gQfAx)}#9r-w{89~)
z-2E|a;<}0}@<0ZVY)@P_9*Cs}?4xzF+Ei0`p~-unzp&nEg%{eLI9%y;HZIgUGQ7}5
zi|o{e>!1GECq8xQ(=Vj#t5>go@j_F!F0?wb*9@Zz)i@5@?Egad4n}&r9>viyHq0J)
zLe1cIc<+jMD}^-QK_6c>?uu<=$G~h_OM8m2Dk|-DToH@Zk^OCFQw)SOzH8o~0K~E$
z45xhIV~H6~3{Xv@NaA0X^;%r(w7OSLu^v`hH2TMT?fQ*{%gs)$+KjF&j5A@L?2YhF
z#QB2I#`Bd{XS3WByA=s_6zLi(9sQW*rNwXti5R`UE0jCHKw|VA=_yb=4ImtqN0ySQ
zCrq&%(goZSA@{ePk>P5!*NmUU>&ON1ge{6f9Y=D{#8DG0T)GRCL~Lw}fzdE`M8nFV
z;j~dZapI1uttZZEy}jB=92wu~hKbdx-hLnw5$|t*3NI?{aIFG8>V?muUOk&LLa|4C
zdcf8xI^Ba`Kmx5p;wOwvNIe$Y=0NO-CyY&VU~F3hGiMCU9aA8e)`EwR@O#Knl+ed$
zy^3Dht!liU*wGCYCmw{r)ln%~i3>f9Ix4Yby_Hx|{Mf8H^>#h3lu(&o6uv0nZwp0E
z4cDF@`_11^p8*MtG8C<B-ZNH()>aSLXh|EmV&B8?$_{$#B+e_>UU}_HwNX(PXWALw
zqX3E6xC;^N08nES0~P~Qig{xP!>`zqgEp5QYT={0sE0+Ec$JFoctx?Y#C@&a28bkf
zS9KZ~wjG7d)x^c1kuAya-s!~<TeecEHLFonsiflLB2Wt;mvBXti7$?denEG7VbsVT
z4Q7M8!g~%65E671%A==2TKCPj%uNB|2|XI9)jY>wFR63Ltpbn-LVX0clC3_5TjJE#
zd+m)xEGG6^*zA=ob&+gw8_z(u{hF46bZ=Fa{6%EaeLfDE^{3K458pj-IY;ekE38x!
zzfwt&3F&;La-&ymrf>An)nzs|HJq7R*ldnwN2B2<@K&#G^--kQv2=AAc`;*aUJblG
zKTZ#S5vOOQCo-!w#gXniS!hZ|dV4g*PjZW4C>PVAEj?r%i~qw~_^WY2kA6gNUX(<%
z4l=1f&>m|Zh_dO%SUNxw6Oyo_GqJtx;BxWa-7%5M>_EfXcI=UZTp8vVkY)$Yj?j5s
zF1#HJdgxORQ((xc^Jr5~U1Gr;X@p7j(7ztu=n?s0O5kJ|iT%p8*S@4DDapL7I$daC
zVztA#<Y^IWEl?lVIa|Q4#9C|ips&N6*?PMjstVVxPt6W%G4xq!iD=pEsxL(t5l(%P
z7rroU_XA(J!v4^+%nxkG6dt(#2mSepIX+muxz0GTS6L%nrjdG&hA-5s-~xxK3hFZ?
zyH+#R6L#hcxHSIxqCYp;=ukho9N5I7dYLtRmSkMbSm!>67r%sSsB;A|Gp)$mgQt7?
z%voTccVIi;_Gk_XE!!sH<%X>>RP-<V0e<CBbpnVah}XlQTUA&sSZz_zSq-W|x?+NA
zTLvnOd#WAO;{{y|0OOBy17^awN;}a<brslJH!Yfo?!=I^xy|V;>FYN}qsrSFahLax
ziU?1&*OHh=A~qVL0ceEhViIZZ+eFx<hXLNL-T&VoZU5{iKfJOGqu{rDt;(9}^tw@!
zXS%jtg(fwl#0tT3tE(_+Az<bNN~W5deTCVVU^ZadbM^Lak^Mu+GEcQx?}uMTVJ+f$
zq25BehHKBy8a^qSIGBR@>)k4s#&v-6OCt|y54KW}8C)zGeH~=Hvaw2yiitH14}o?e
zb{cLY11?b)W)ryUfbSi`&mO|}58>yu-GP#xMK2UZ@xV%Q8==B9Rw9f(aidt(Wte2T
zRVnN3wNieJ3+k&Z#Ysu*Hs)($qTQ>V_RUK&5jPS6$iJC*T-h{Fqa~%w9<avWMs*P}
zpDWupJz)<2-uoU_c=|sqoqxZ7`k?7Q7zA=?g%(<?-n&Q)XwPWpV9j(2HX11)Bb*hc
zu(0T=Amt3KhW(bk>278=y-m#fz>-8^3n^QgNZGNwosrhS8rVD5ra$n3!I`^4+H=Od
zvFXQdBiryBxq&ZXEbg1&I0HwQbdd6826Z5pt^?1g!vx=y-T+&s`*=cM^245@pT}r@
z{iXNjSAtrw9^4AfUkO%%=aDJ}wf<giE#>#~EA_^?mBz;TD=YPlp%wbiAEySKBIp-Z
zf_l(Mt7-)Gw3@U|UA=CJe&Tv*O0A%Tk|ZI?D#=qMoa0nCXztWC66${9kG4zL0A-@i
zY7FO6t>H~6H^$wvxn9{<Z6a0@QBxbNL)=QtTSGr*Y%K`Rw{S)5fK98Q{{@S8z^hpl
zKGJjpec_q?QwM6z5XBG2Z7M7Ef9xxqROHog?fG%w{QdNKkVAY0h#cO6nD8reN#UyM
zMA0scEIp)0@bE<tZBh0(41^RSsfLwrA~69gHr)j{3sB>cvmnJBEJ%e4Is_WJ2pVF@
z$q+J-d(6?Gvxft<--g=4-GIXZcf;Av(W%fbZGHqTb@XHrI%2QXs}a=XJf2gJjC&Ti
zy4Zb%iL)M7Wu(<&=u^a<%8kSZ-O%)=T4ce`vy2NaJ>5*|L1y8Ns5{z$$kjUSIJ_P6
zLET|qk8Uw5SFf@B>m&!j&hLO9u|rS?s*34~B9x>kY*TYS>>t}JiZP^&WQ)QET?vXv
z%iu)-D)&6!qNN6e6F?>;FoqIi+tv%&jh*p6V_*-QfxBwmHQ%x}?M(;LbEUOms;|YF
zh9~VfI}i;Y`{ta)oU2btr!Ax_GxyDJpTsW8-p=nB1BcWE7;QT76rclj-i34<1!zO&
zO~eLl{O?-AXnPlp=*zbd3F7}NxZ2srOMKn_x)V<iENox5;u%0*RzqH9dv=g%6tOv&
z%Nd(qJdZd2mLYS1DQl3~J~FVTje)0kcbUdf#(3@vxS~_CFv!3ywE&<qgRGn$mdRqi
zjt#Oq*cz8+_QnHfqlhPSxFYV6xwDo61<m^wvRh02qbot104GZXuIE>R^%59THITa`
zQZzb^v}UA+U%M^d<z}_DBCA*af_0l91*8A`KB*5OVIpcUETm{nu(hf6WHqsB-5y+|
zOxUijG((x>WO%b)3oG1P_m2=%<vE~pBakV!0%ZF$E5X9T!VtERD)ncXV|NJ8mx4yx
z<p3=yjnBHO#=fd#|LE?Z)9QxPQ97=dgzB?AbzS{B$pC~K{%$f|*&ATepq+TtUW^b)
z1iu~503AV&CGioHQgL*vjxD*f+U!(grQ!WH>tw(oo#~TIJwbAcgaDP~Th(YIl$F&w
zc44YT{w7GtAESPf8Lnh*`lF5g2EdB2z8P^yEC4|&M9O#Un*TALBJO$VMWpC|#`LUb
zu&C$7sSnL>6m8q|@McC^fxMMBX$$&C4;J_XI7u-Z=#jh(^1xhLI@~fu$nhOqyB>XU
zR*`q8Wl#!lOS$Cq>tU;-?p%ZOTB~D|wYNL_JMowSAOLCb`Ew-31|ZfBbpt-j_pnk1
z0u+7>C=3Ir^MmYNEPcR3Y9`=k49x^hfeStjg4Y1#z#tH0$cIM*4o$>}lY5MlkG<_F
zT-l+C@U{!PCSqFB0L^Ga;Vt!Z*TXo5iy8E~L4SH7_%)(>0Qe5-QP7_~7ffKgX65v@
zr5CYGKNDyRvaqObaw2EPb|!JUiW}G$z*kpDwC>Y7sNW>vd}*VTu8iG9F}65bJ(K)y
zz_ZfnDmv~5S(1MPKcd$A;w0Qv{}GkU-L)sD@YH|q=L-2?(~?D%ySoN9vL)cDrbY$+
zUn(!~Qcu-ED)m$$J$S0tw*dHnTo{;pVinqez^gU~`%O#(#syN<#z&%!hR;0*Ks~2n
z|Fd!an_>W&N_$6a7q*yVdf+u|>6~DCX2;&1l{PfOkr||sW8&>Jb0FqrZF=Zm5hj5U
z?>!Fjgb5AtwD+(EfOs?qm}dl2nz|_qmDy;*;Vuz=i$EGVz=aY|<b(UY?dsSvc9B}v
zXHlu%?PyoGlu5F^b_6FP#D*uMLLF>OUDsVD!3HQ)tzHb=qZ-!NdYxXRXht5Gs?GXz
zpkVBQ(gFJ~Tspeo8AlPZIF21n|C2vo^oQCPn3F8@uU(-Vm6%XKn!<GBrXy~DfXC9@
zLExSa<sP#W9ujYKtN`%DM)_n*BTgk8K~7~?&(-g<5Wzf|Tk6EGwz|!*1yBm*HKjUA
z{UKlf5y|dEsqgVo!+CZL=jz+!{x*pu2}xE#5>Zj~y;2DO0aG-_I$HN1NJP(or0|WP
zk7;5WJ;#6Vr0<zF9QSlu&G&t`|Lp&3V%2w9`|px`mP89o6YUi0wto*Veu8U+-6pV`
zyhE`&dJ%6&UVUhVwiMqt%TCtLIx??4Ishc%fv$uPCbRJIGf45oOO!nwMb2Uw6p|yK
zt48(OGERvi=m8FXK8Lf34eXd{BHx=C?vT!p#HY>B((+=B&@lSZ4cUc>Kf}YM5!kkN
z43(1to}#&#9l+0O*f))6{#OhvYNQ8QMCbj12Uy0I6JT$0`-(S?dTv1f5@tgOAl#Ti
z2^-LFYg6Fq+%y}=FCCpH8~fM~VBMm&{6u<cB2E9#sK5|l{O*$aQ}kC4w&bYaV0x@p
z+I#;7c$f~I{sPm(Y~=6DTbdbLfK=^|)T7F1>MGTn)q0c8Suzcu=i1UT&eJNdzwz?5
z>&dLFM|@LZc%5sZoj9Zybx1BTS69>LOX(&y`&#u*s2DFw97aJdrngJ!?c<8V8SQ63
zf^?E;>-`8KRB~ct!XcDB2EySNt?E+h0eR`FM>H)gdl~?0&FCPz^FA_10f=wTXjcY~
z4_C{h8Q2~d_q>_cVblJZ1J!J7b%>FpAJg==uj0~HjsA86o0);R$_NAkj(9fiB1o6U
zjV;}Egylz@=QmLDK^p5%<ER=6#_Mi(x#P?@b-PcB{^^6eIIOdFXb-=L#@L_Hh6`y^
z<p!o!U2z{)@)(Yvz=i*UO&66DhGWb}Y}{b<e1~t+ytGi}63#N%?@~t5u~L6ULh-1-
zB>8I+ZX47ON&cQ>-w6K>yjhOWL*gmef_Y(!Lq#zhDD<1I=F;-=VmiBi>cIRm8+eg~
zKBWdH#kD>@=}>G>y+ktZ$Qa`O1TTJuOCPCv*ddz2{m`^cQF!DRi<aX18m>J*d-xPx
zLSh7WQj{$MUDlO34Rs7+WXm=-d0FWL;NWrKn9Rj2P9?=*m}LUusx~G?2b?FE!3u#d
zz$^syR&^~r->h$hfvm>WG&a3hS5bT}7@ijBz;?aeMTq<P>P9sPTPtB2tkm_Uv2AVC
zgSZo{oC{D?rWGx(hwY%+?KbZO*v_;i&YOaGjK4uupXQGm#hG~X&N&3icPrDq)2%FB
zd-F3vFA6J7eMqPK4o5x=MrO5kC&0G=F(MI+yfYSm@%Bafv^OJNL2AmixEGgeG}NOX
z(?k(!ee~}jlaRi}bLH9C+A?DMt{Kg48EXOo2ZFA%1ut;k_!c6O!+3IL;6)!z{rD{y
zsfmcP--drrxZbjL_y)fHHcZz&anHPG-LvmG_uSPC+$`rEFP<9McLk5^HMAE{=bP~~
zYu&<%8rP~L*fR(t;N7U^l&1~M><IMFVnnGk^EIR%JMy4Go5Q7#+19YK-=5wSv9V=y
zbXyqr07uGO=9Vjs`<d?yzX-sY<F3I-*QthKp%ohzWR`96TLr|BzQOe0lIR7k%e~IW
zzX3@Kl<D}ZjC3^-v+z|x6JI>&uL@q^!lsy9ETx$b5paCezwo%O??UWk_+V5=sW@S<
zw9AF>NvJW^Hg?-K+ZqSl(_kwVLUEbGJ7stsr)C#)&tBJGXuO6ZiIaYFQk>B0ie6oj
zT4wPzU0|%ZqfPi%`I0lBrv~2zgxz?3|D?Xl_y0tqq2(W$;@oH)I?GA<6rOy$2m=HY
z1cPMbK8oV|?-yJfE}U!HAG(ihXB_L#zh88*O?u+zOyM3IkWT&3O8qmMgq6Z423QZ?
zz+BU()D08S`e7gqaelZRJcCC_PCGu)89Y1bGgk{^`8+#K->UNyZ{Zv@uv^BlTOKFi
zi2cKjqNC)UU<?Xp;x{fKhpk>g8MshvLDy(8oX6v=alO%3#(7XjM#u1-CAH2e{1FHP
z?!yp%8p+WCcj)g1b96!AS#Z4SUq~42PBr6Qrv8<L##?<KB*}(2BnuUf+J8(YB|nn(
zXUO<AiQda}nA&ZlY~SVw2BMxgT5~2WiWc@|<i1K_YhNAS22c9*@eNY+LKsKoeO^;f
z^iLij*pbO|CxLKrIkkJ4eReO&R4TGl!x271JbJfqxUu^OWI4|M6_D_79AIXtf`t!?
zPbt1DJ(py#(7HHy)?eSKr&;F-kn)`VIA)>UUKk%$r163#^H95W)JuQph0Rnt-Q2w*
zNPVY^$b_OZu4wd=v-o&wAw3q27G!;O6`_kxd&KC)^=9fjS;&{VtY#--5b<VRb=o*J
z4`a2(+K3MNqoXvE!8n-ylqGS~g|Cm|wYQqC@wb@aZIZFOKnt%&KOs%+DNv+|cRB6W
u6-=9<?R_3!8hMc8jQ-P?^!x&LV4jV?>)A!G=zH^CwrCe$!2jaWv;PNR{Q*b-

literal 0
HcmV?d00001

diff --git a/ldm/modules/__pycache__/attention.cpython-37.pyc b/ldm/modules/__pycache__/attention.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..62be67f6537cdabfc4c21523f6e2306ab1c1b4bd
GIT binary patch
literal 10534
zcmb_iTaz3|R<3(jSNHT>b-$HlOQ?6{8Osi{V1!s()(FSSZPpsEO=DBkbXL!FPj_`I
zt6G+tntj;uqA?(dT|;o&g&vlGg(V0g2zUgZHVFO!^#s8Kp$DFL*ocjdfbX2@+l)pw
zIOvYd%F4{j%FD^`JLhD*Iy2)a_`Ul#i^1}WqWnAEjD7|ZUqpyjRYehsP+Lk{t*R=e
zwU$=Z_^MZRz8Y;)Q&gpGRdZZJZ|QBbYPRjF&G$wt-*&1_yHG9gz1b?ZXR0&pQnkeQ
zR%^CBSDoW@t~K9Ys4j3?Yb{n6gT?03;E1qAeow0&6^<x;s8o-MqL{(;xG0HPTu+EO
zF^}s>u^<+4J#|YFOXA3fiZ~LO&C@%2^$c>3ietz*Hq1FIj*Anhdrq7br*L&q<Fq(~
z8fS(z&LiinIES2b!<-Aqam9J$oFC>qgPaTE8RR@O%y||$&x(u4xj4+ZD6~6D`O;&W
z-LjfGji}xWYjL9!R?1qcg<)z1jWFu9ahI2YFQp%D1gRy1Zs5nrwBPJfZNHW3uY?b%
ziP~mR+v;{2VT`-d{dz~X0~vWy?8ia1^vc>=#rwIpZr*tP&g*YhQls8|@r4>zl7Dp7
z5N;quHlhTpx|K%{)ykh!|MYu)OBPYU(GS10ve{_|D`6)N);pc872(JJ%7*Opx+`~j
z;q6AafzE_0wT=i@UTf6rufG1)N=vj?+8xnr1<{Hh$3aNlU+zA@K<_uAI66k}rcNlT
zoJD+i=N3>RRH1!{F21Jhr~_qBK{xfvxH>gKvU&$0qUsG)ec#zug_5WrYj+JA)f$h3
zW_tPgR12oYCQEqaht4(8sKvETyL<gKwL^_sO8<B-Y}{SG*6P&!R&;%NQVCOJYy}S@
znkuZ-WU9Ei)S%~1jT*uLk04#+!6HFd3OxH~k%-ab9aR<+3?fz&73q_RahIsPCS{GM
zNYYbC&)|}(;^13zBB=YlR(uY{k%dT6ZPidE&GdM>Xz6tXe)${{u`*B-rK#<yO}#*K
zt5(WJYQ`N|+f2=RBdm8)Q^XItL8`a?`xvnrAAIu+3cN7b@E|t5;38Tbt?U`3$M2^1
z(o*reKZgX~i`6}$V6n7V9cYQVr=C)_wL}@{iB?b&ZBJ8Ci&y;df8+O<`cM6fHhuKS
zX#4SIYD9NsoZ6Ul+5}!%Pc6t>)RCz!8tqh%;wR?HY=n*2E2A-%e?@iF68`eXkW)h#
z|ESOOi=IUkC{>76RePw^RhCK{@TUs{&|y5p(5fbSYNqC`o40PiT^l`0R}+D*k&7r+
zb};FCz{^;}K-EMO>XNdDVS9$4!D^Mqywqqa>U9H|+MdT9_9QiuTG(>Hh}7uHP7}j5
zqM%hzEesab(i~koJ%WY2=ha$%6nS2@WF+kJ^N3zRh-fWn<oyzNczN`av82+=#tU`{
zH&__RBt(UWFZ4&+d)l^&1ygw>5TYgOzfqAiwUUIptV$AyvLT<tB{gfCy>Kg4S5spn
zX!XjvT&8UGe%06ja2=Ga99HxLUy2_`rD1=MtNod*zr(V!Me`H-?ZER=$MZ7OLAvO9
zcYA&-E8$^FYA2l;&Ppw4wMH{TV27?=#fA5<yo?mx@mp=RXmCyqIpZJCAMPTWnvraX
zJWGcutPk^$j-0UfNr#gbR)avSX2YsHrZ!gpW7K4gw=M1nlqs;xY$7o&Gmsd435c<6
z#n>i53Joh@MAu?tU?Ny3xAruoa(j@f(2i}&L9HCsNGQ%H=AO!RE!t#zI%Lmb*~9iP
zNZOn{VW|#&VFbw}&$$!a?GcuPAFiNMYTUYc`)!_@H20e9bRl=C9tLsQ=7r_8m(O#}
zHn1txH(EVN=&)wK5r%>EAh1=3WI3qCkl@ieFaeE4_8J<A2(d_Hi>hN>(hOAt#Pye^
zX7ouRrkzC7cmp9KeP<-Z5L66Sm9HWXs7i7uU!rJCe3>MW%zO!9h-v-B>DP^95F$?<
zT$MkKxNONU;39v5qMxLQ_8beJe3???H;jhLqx>lpQ5cF`X;#wR7(Vw6<N`|WpQfSg
zff(PP+XinCrVKQ~mAlS68EmTkYmSRwHIN+_Yvab7f!p;ZHVZae<aX+=?`9ju^+Vyx
zAnwV~ZN$s`7_h~e)CR%?u}`$nM{^$N)XE-B)$U|S83h)&N>=<Pjfo}OD5-O*ygfZC
z0(2_l<yk@^R+?%PhzOlTBXE603~p8)yC~au^q+q__`&!8`uZ9;zzKV8kF}5}EmFF+
z=|h}aQK|>sD7E}L*ed|S9LJNH_~vhs(oz9S02~XAaQw)@?&qnF-)i)Ock%Fu&DM~v
zA&h_eF!g_7how7RpZ3okVD>8`#b^w<w8%LGYB~BUqRGa`CRRKRTOJAmQAKPvt!56n
z1jb-cz_E#Z`ziVPr{p_N$uBS*K(0xjkj%S!TG$FCrr3)xGpQBDvLS*r*Y$;f5?C%y
zfJ2f#CTYS_Bg9-zErNBU6W)74r0U&N1ryv$ZQA<GSi@aN&yIG+eh_#cB@yjZ0s%{N
zu>Cdue{!mS=CDbB5-SR5n5deC@gI>$h}?+uKy91?StA9wgtkH{11W$oQf8tzjrWXg
zYdg1XZ)3I-T@W<uNEt#y%80E&PUwlA7<>A*lQ;n8+>R=YMP*6Zc4DiUZ#vCF;s~(E
zBMmubV)BzFQjW->4b<W`aE~@L<ZHrCU>ZHb9r{un_7wd*MpW!n`b+C>&E0fAaG$&G
zuDh3!D!a7<xwS06UtDiAudFw>p1Z!@*xJAZ;P<cpz-*2MAtLXW*4>8N%o=ODjjW-p
zRek}#i`2x;@{Ig>JRyI9qW37G<&!Q&G|QPx5Ch8V6p_rQ&d4x=sgsL3^%%OP{6&<2
zmc|fW)i%8Y&8F&ls@9}OZKxll+6P0GXCgm@;~yYIgq;dBV<-FTCBBxlIdws&JkkUm
zt$+GZnHd58XM^~+@n)Kdkyde1QcewF{2OZ(myz`ptpc(Dy+WPRItyv(hU`Sq7*Q&Z
zNq>0!MMO`k7lbebx{NYkLt+YL3{ozjT0p)>s)bM&K_?zb?htoKp<u+JoE#wtYEN<)
zgZAZMKMWwxP)?wpKshl71yVs+=`9YCW-M>35FcZ`;YX05u`0q^2oz%2VnJ$d2EK?`
zVupH1-0|+F2I4J_@5!IRu;eR<U=4H_U2>VOnTG}=ACS_K7HXX^4(`WvBU@D0CgqtL
zmpAG8RYZp)(Kk^N9YsV)WJ6vEj}}z?p_V|WKDK{3CRj6CUs6T#2)qVYfUjf?eMH&`
z5S@@^Y6T`3qYMn*;Z|%;tdPV=%*3keJKB5twy|wu^(>)pY4W>qu4xM+&L?Wqf%RMv
zm~$BlGt`77a*wp%{2c5nV^G{v5|d(0q37cnU<%sY!Rj_kkcQkla1eZ1c}sa$2^XF$
z{i^X*GoDR!7_L8x=YV^8#=YENKFKu~U@j~clx;g+LW#4hhyt)mPjZ7Ji9V|&Hrsf)
z=23F!JW`>g&WKWygTkr<Mdy;dm>oWo$J`uC@_S?y&rLi4L`?)IiwF@ZQe;mH6%T2T
zbX3(V{iEw{NC+p&gj<)_-OVyG1Xh;t0vCE7m($sWGBb^UnoI3#Ex)}k{OjMt-G3n{
z{mTbdcZdp8zJlQ+k|+hnPOT^WRIhb=(3)~V=&!c|krqU7uTcv;vab6_2v@}wKy=d;
z8N32#`*Z8=^78T!yOApQ=c&ec7?;a#b2yG^dn;?Md7wGbKRO<K*5q(n%Ewv!WgETC
z)cufB9YkdIchgyKzekC&JGK2@4ChD$y&k#%>4!9))PWO9#?c2280Tia)$wD=D8EH*
zlCvj+{}YrtNzo~a2$0fZ+mE&a;nf?k&7@EHT|{MP0{&@kxNFJ$kG2l-fF5BlliwlL
zq44rDj^>9}leO^gAQ2I=D6B*<5ol+1TYDB8_^gI4tYJ$-R|OU5MXiWy|LEaMPR1?)
zmn17qHW0yluY5)eppM7CfiSWF<|XC)ztF>(H-dIYKDY^`xYmFLw|_KKV>p~G2(1{Y
z{sM^!Tg0rWH=+6b0k&JBB8E1E7%YO={C0i^E@NOx#wpNNhMYpYVIiN`!)^kfz#fE1
zln=cKIm9QUoWc`1#n>LqAmoR9Vh>7VKADB<Px0oMRqRT?a3_dksHbkP>-J}t-7gdT
zyTEz35xM>OEAA8ta8`j$p<YDK^m8s#%`!MiHZk_{{DhKHvn$CyX0HKTeB^7(c2Yzs
znla|5Y-5f0)P$T^y3^v=L+W+9<&3A~U3ysVZ}3w{>7P)0OiGtXhBa$UOQ+TTrB4)s
zCpn7viF5^Cd1`!Dg)W@26Ew_hEs>Y$!X{FgE{v6hF06k9<^Z09sUE6zNDBP8%qs`E
zstK0DiZhD|yrb!mEd$0VFsJn6;@8v!>n)5ubx_)+8ncPrG=zDA(sO&pU|txI7E|Pq
zMvbY`S*ii|n!atLe+%Fn;tzX2e-QRCfBY;Gpc(K7UIh*z*Qn6=H98tH(7Z`jDU=@-
zY1BY2gnOzUl>33jZfaPqECJ`xC|gjOBArX~y)c3n62OSc$$)*C(xpnIgGdq`7*&C6
z_hNu4iDzS@*XhCCoH3Lp*QwqzMHBrt#_hi$u&M2+N8u#tf8moC{He56)~b2PWayC=
zmTH}D)_oobyY_w+FDfq{#uwQ@`qVc9jjBC5ngAYmqw2{v<B%7OfXExmzWj(DB7R60
zE1mfDcDEI@0abyxDP>2>U!&q*r)WG)@)6xKL*yq6AxWi`I~47=JyO+^U#D!EHTZlc
z(>SJve?bX*X-Tikh@l7zvuG|DC$XUn=y^1=$F=^&|JB^?=ZA;<-p|Yrzk<iVPaPyW
zwYAc`q?{VU_(z%y{i17#M&?7H2Zm6;rBzK)WVI2%3iqTXpy-J_42c|0(db-Ft=jC!
zrn8{g^vJ7z)TpiDEGmMe;CSg(9EEIQ-iVj`b3^OwxzPfTnWM6{S|jM7AL&112>C8<
zrpaTV?<w-AP@#@$+xY|vUenmJqRX*~4Ucq^$KimJ*i04<NJ1my86}pyosbd>x?!gW
zNCc$@D8pb1$tP5s=mzzL$#tB#0P`688m-MtrAt%kSty<(>aZ$L4v|M*HZyFVNFl>A
zK{e)v)2~r_SnUu9pC*xPmmFzC-e?{@|DNAyk&2kkLhHI&S;G;Y_r{yA-Mo{QUWuYW
z(pehsiF95xBC6@p5ddwaFf}Q@%D%%?tJkyp7c%QQf7^c$NOEqaCb?y<W|tST%Zrj$
zZDeR)L^{odYz@JkCZ^0H8dB9Jhdng4IzJ_Uhw?JefZH#q+KeBy7NHgsB<;D(qxyNI
zeuNN_zO85uEN`;GErcZ~T^8<&{O_MVR01bNMes^Lx{~DIK!`}CBpaX!Ln_hg<Q9Qf
z6!*#wp9WUo6QqIS*kA~N@)mA~XAzkXm=T0~Jn1$$5O7>QatHR$9NxKMo3%sx_7>Wp
zo{Y>qq&0ao(O8z%N1)VWI1WQD6d{;;B&}eYNk07d3VHbVs3faPbf%FboB9mp;dgP7
z-=c_?ME*WS-=^sIDEdQ0pE}Y#RGO!e+DIts1@$-<=BbM@95Celw@|yXwgz1Z{i#4{
zXQgQ32Ct)>Axr)?rAQ7kuQ<K%lU(B;qx=U55v_);!R?RF4NCKlE~%u3F#cJ?RCEQ2
z5daDmXE6naKJuL8s|Kviyl?<`&`xl2XJHa&(uJYRXA1Is1M|Sn@Tp99oq*8Az}-gM
z-w2*-HMRnmj>R*-`D#N(@fCM?q`^-2M%ab>{j$I1yFq(B$Q-8JZsvDqCBTh4?)nuM
zkBY3JHJq!vez)6t;KHT~1&+epc!I~S&)*P?UW(5HS`V(ki9c?Pd&h67+<fO{w-*Io
zi_Z?Z?=T1gcSOXs2QI!qh~fI6Qvo_BxO)F8>BjdWZh-y36Nq13s}a4^kFK<zNZ_M?
zMl=oayL7aDA=Y=5*x1pc3%kmO3NHtNu(=D3aY^|Ie$k;%JD1qeS2BBG7lbwC)po)X
zwEaS&?i%bff{ic$l|57+Y7g~?#zXU=RnI{!Ge5NBnZ(#p>1zRo`V!jwt9X`L-NlBQ
zw8~!eIruYBHflM`W5Std`@!1h|CV3D=;S*T(Ye6BqZuN)MFc(#+`tBZ|LnFJE4v1b
zau?%IaMHc2?OH;4l>5f;!vq|Qj=3Y-u8hDci#5oE&buUenXs>!M+TI^`<kCyrQ1J3
zl$L01C*Ooj^g?{x(8327{Bgrew5K&xx3vDsld?X>)oDM*=-KkZ#PG?KG<?NE661$3
z`3-u<#Oa?oen}u?m3o(B7k?wnoOR9{{9(v#Jd&E(mo5uaQtk2Xiujy}UB9U_5!{g{
z;10D-BbNhufhmG+ym@dk|B}l81ksR#K5Z7c?u4Vx({w$L#}&hYV*sWUdHPfX*X-1x
zsFmRDJF3kg9-Ty#;05fT`h>|H69erVLaQ4H(cdEiY2iBv?2@@e+u&pMriK%Rp-&HI
z5l~L(g8_I4Sv`ub;oi*dX%EqP10DKirLsCHy?+S`lG21Vj>E#^IJCzJ1=KB4sO%Hu
zD8n)S-8vl_HVyerJO}lO-omZw<XxnLxJld5-%aYEl^mZnm`^!t7*Z3R6bw%vGT&2n
zQlPO9P6~`4n#Z5wBL9pcmm)qt;8i5mF>@RJ22yE0z=>ZV>63x)Q3;<AOgV`%6M*%5
zVu#5)s>Xf#NJ4F-CW~3NQHWNubtjU)LF@i363|Up8zGM_X(uI)mq3P@u0nDD<RNQ4
zqC_h@J^K!ts;sWA5h>;lkYbwiJkhD)Lxc$(zKr^lj*S#3wo(2VMaQXsBt5om&}VmQ
z>71ne@})F~9ofY(I)5u9&r>_c5mguYgP!FuT%Me-WbTPyqB<{9WFX@2&|s%!Q%%wd
z_xXb^Tp^OQcu5LDdLAElEoW!D(Xzk?U+|T5!cj&qZnRjxs_Ll6>MbnDh}@p{8nP4O
z8-O5|zf7&MFYpMZj!{HjJN^VGwK9a@4+~h6&A?!dD$t1u`^7R%;$ym{(N8990)Ft_
vX5G4uDa-b95j)k!Zl0q*Tzf^wR!f>~;BO(=3(mrlowtpJm+-&3aQ=S)>r)v$

literal 0
HcmV?d00001

diff --git a/ldm/modules/__pycache__/ema.cpython-310.pyc b/ldm/modules/__pycache__/ema.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..df8712092916aa22955bc628ae9bb83b52b50465
GIT binary patch
literal 3212
zcmcImTW=f36`q;BQIts0jpIuiyNh1ZuuUl2b=t&nojSJD3mTz_(iUYCtX4ZCX{qHd
zHM_JSVF?tGfjqTE{{i__<VO_f@7UKqr0Glh929WBGfPpl+Se|*XU@*$%(;B$%w(-r
zli_*#><{XdHOBr%AIp!8k00U9zXKwf<O%CjD-t&3ea_jBnY5(+lu29hfjF`HmgHNk
z>wJOvY_rQNk){#x%MVGVCw~t_u|AiqFXZ~CY?m8pE0E$y_bKb!(vuCej$DyVw60v0
zYiK>$l5Mn&Pnpg7E7-A9xj&DKN5k;*>wo_HPkPV(`Ob&Y={~fVH-9a(`CmZ_cEYB7
z#yBg5<OeKop%n+VzEB8ho$yb1%t{tmC7asPE}Vf|3Q$ivzZWxsRn+g6mVTYaqzBrd
zv6MJhHcr3ITs~cyv4`wq|06N6Cr;@Mnz~=w(^ax-CMIsl2G(?~v`aVUGx(LHmu~4y
zTQfFo&)8s{<Q(4FBb_>u!Q$1;J(yvh&+F!2?;KxysG{)5ACB|F-%<D%Pn1gi8-AEd
z|7PV#8vcxfDUL>}T205pU_6pxp>oY3<%?%{D&DJD_d@05<B`%;BM9O&E`p$H{whqy
z>XFu&u3Ez&4TtJhZklRctGzfcln!>rySrHB6q%0pE4NT-j_KZRl7+?hZdF#C7FCCq
z%3u^~v{;l^Ev*j6u~vDor$bqJT8)x0Qk7e4S2<CVrTCPp2qD&n%3;-svLsPaaVjeB
zdX=54WVdpLnN*2(VN$<6&Q(wwN_JK@ZC>$F<<`2x9V<Z)C1IWi!DsBtcQ^L4q1s5Z
zLhWSP!A1m88*oIPrD3wM_12BsZ{B+Ajk^26gRKuX5;@$!e&a;t8)_K#Mn}4hjcN1z
zI!Kduc!Rs*3V(^Wx&5qTxqMA@_$4lW*A&O=HE?^UZjLArrw{Rj-3NGcAEad4gbprw
zAqJKdptiJXOd4F+n|cNFOgJ?y?K*YG1?v3@kr#-(NJCtcg8Ewv@(Q@>0^*>Gxz7wZ
zuA)1qe6sn={P@aT<Nul#2~0VC8er%(s{aEFK+Yxt6v`hs`dW$P1uuTU1PfpejY~6r
z8D)j$U1ncpCt@O`U5YDgVinfZp0Pi^$64W&eCp2F37;T2;SkU8PH7{}dPpdw&s8>P
z=-)!l3h@)KbV~0kn>OpQ8K>{NY~H(CS~XXXUSSs*CF0?2q#-0c=K0Rg*eCq(_ODR@
zZTwCw4sMIxWO@1oGwJsmye0_FW}QwZr>Mqi2c)}Z5NX0uS;IIr;i$Yct5I*uI4bld
z%&b~<hSbTF>&qmfLjxWqZV&>-`FJPjHtGO=jm8by*O4XxM33{pBzxrz>pZJk`F<#~
zC-qleSqJI}DNg&GkH5E@=_jF<*C5q^&gD(fM&ui!Dc1NJwY&-Z`n)5=mtX()aeI!4
z`NHRLK$eqg1_vL3DHvjh(u>#$@N6m!Z<PYEyEZ|rOB=DX^f%|6J#`Qrh7yc^jMyeJ
zZA1!vh|G6QWZK5_e6#E6*RYH6qkf$@<Gq)`p^Rjsqri9)X<@K$Tz7+}-6V30$hmy*
z%&H=zv7A!d_y8Wb4*Tsh>*v#+vz{`7USs`5FcOG=3J*Ml{V3)@I*Pe{{FkObf40KM
z$}jenAC0wEY2lxhNgw|8k#(r8ak}RxsAsyjSa@IW<#!fe{?fWT{!cM8+)0$5?fToM
zH22c9_lwg4y#2P>acQR_!|Aak%=Zg59A!GxadPC721=f7{ano=bE#Hr;myfpI<-D_
zCmdmCGxjHOD7I=JqyD^v@vf*u??&aA?aw>Ec{jnq5@o~Dy?03<Nx#UB{W;={CcV2!
z7KKTE56tOg+&(k`xGPMYtLB+;=eKZ-khd4oz2miI>CeT}6hX{1@q85wA$S1OP6)lL
zIN-&^Dy@N-0?z|skm{OCkq9f4gM~{!Fk!4CeWu#nv;LE~*f$~BHo9&5I4AeBC6E6y
zSG(iHN0=UmNlZoBOgHCVz53Dn_m?1G`m%B1hd7?2g5U#;Ewllk_F|lGel{xNVSF4G
zahCccokc3od;S*1`#eN;6yRuj7h<;ODX`GjzrPFQo$5$3Uj>9~E%Wd()abtE<A8?j
zIEOV0Nj1vrJ+Tdn)RHl0!23UAzP^3Pvv8Pb$p5LzX4kH)<9K9Hr>WfNH$b{}U1#oq
z?|V^Qe)XF)e7?-gZGMQcoGRwU)ef!;i{d&T2A~OqA9kB`HKY1XSF<3fuOQ&pg5Yo*
zCiRH9is*0CcDIRqp9n!+{}7~V&2K)DN|GR0(&!sBg>D1d0-+jiDqn-OdE<qK-FUIF
q(pfQMFwop8C^0MdeRB~q15473r6jFt+^zMzM@OTvvVn4X$@&jGG#d;6

literal 0
HcmV?d00001

diff --git a/ldm/modules/__pycache__/ema.cpython-37.pyc b/ldm/modules/__pycache__/ema.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..ba3800b3683c1d2ee66bd4a7be78ec84cbccc6f1
GIT binary patch
literal 3186
zcmcImPj4H?6`z^i<&qRh$#LSyP9khvAP9?;OdJ#`6vK&KIc*9@Xsxskhy<(E&QMxv
zxl7FqWs6u|A_F-UMZW=gDthjvm)!Cpin;dC=G2^XN#C0#DM9JAOYWPu^XI)c@BQAJ
zdC+Xu2|S<u?YelQLC8PxWBItid<d`lDG*LLO-T=1mXaauQA&PAxWnD&gu8+c*tyel
zINc*{?^~EpcH7jjEQ^s}J}40%^$`$3dX$qM<Lh6MKDCl+Ai?96=cMOypI4#vc#YSg
zt?)J8fY#?t-h#IJ6>&+g2HG}E<x!#^4x@i<{pH`k?)>nd4?c~TdVElWuY)?j0iww{
znbA3+q+pyL6Lkw(cI?XAnsMixen}HjkkBc}%;m2329<&V^||*)HfOL3`zwVbdpO2@
zpj8~pTK2oT{KmP@9u;&}o0A>##fP7>sXO%wZ%~(~g*#irUCr6FQjmc&YZPu#N$4Cr
zit>v};mw+JGHcDrU>zl$oIW9kXx9cOu;aDeL)eOa7R+}aOkUd&adZ|8$4UnW0{;4`
z5Lxhc5M?}g$9NovAtWBxXCq<Ovhgq+k9eeok`!cp`vW|Nb_{7>GoBicgf!JKOtM6W
zp{f5NO2^`%l({s`VVFfjaYxz4tV?m2C@o}oFz)wZm8WwVADN04nS$wlKg}ck?j7SK
znKm1^l!v28LJNzOX-aW2PNY!bp^Uikr5L4AEKH@e&UkT}XYk2|jzFvnCP!u?&eK%H
zdMQe^ed8*T_Ki2pxk#l4Tk5sON`&P`ai7LT<%W)WmC|-_M=cEFG*T)I|4P1p*geXJ
zqMPMf9OU_N7gco+Wj-Et_r}?yBs+|cMAnT#Lw6_X_aA)xxSR4}7gUZ@p}Jxibw+0r
z@`zO6Rkwf;TBjSd$|`h|-l8wC7ifhwpyw8wte3d$EWI3K9Pq&wVf#yXRRE+Q`v?un
zsb&L*GoUVaN=RBrxVsV;*hsdiMD1;_>{`eE4J6l)T*o0QUq$})6}b(ZxduTHdf_mO
zjGNG{Y%)=oPBs_z{%BgvU>-hIcx&)l!~X#c;G0YtP%!=2liLMkE%59;!bk|4!EtWK
zUxr$O^Ij&ekaIR=+%4E9nL65;xpVTT=ags<@~JW>=X4603GVPM?>*y%3pw^VWE14h
zO){v;zk{S2@>9R?3jZdV)yuIt#qYah(YsbSC0CAak{1bP;mH7U5F{M?YM>im(v!gx
z@&qd1@5p2F8MIjAcFU}>se>hSm5J@91)p`kafV4|op1asF9GiIB-ZjpSYn!GGL(6v
z<V(;eTR7ljj)oB+R*err)4H#ekUGgTn}O|W8NSzWz6JU^WJn0A#wxUVZv0`HVy3B%
zBA!1jzm#!~#TjHb>b)9$zn{yek>n^<Av_2mwN6{~RaS+zK^s8vcY_9O(pq3)v1o+=
zGzZaV5wH)8CJ;5KxDYi4Jee`eJI@{z45GF@g*X>3M9q=ES#a*mgSZf=z0eOMu8qkS
z#0dHzCR^Hvm~3H0AQE&J{P8AevObk<<g6z#`=EN{qqER@RlWrskQEmC)^|U{xg8{3
zBv;bGw@Y<?2gVfUG(jg=7i?WPU@>Xs089mZtphO19N_*OoUj8ffD#Tc2czA|*Y#lW
z>_w*{&_^PO$5M(+2N%UL0QUxvZctI7_yj3bG1*xyye|*c_UbFRv~D~27-mKXsR;6Z
zux|@;C(Am&U6$Sb_pQcDnsg3l#wB4vpv7>M%Sa~aS%4M@exZ6*s&K1VAok!@xLrJ_
zzGzG-M4imZpV^7KS9%tT&L)hvnPHu`jb~L~aryW4yD1zcaXuW~dw>E_dWB4a1=y?=
zox5osN2$68%yKf^HRLN+xQ(u<Uu^7(3IIQD<^82};!hfvrLW?Mo5I(|5s~xk4TSFi
zHg=A{y#Yr$n>vLvU^Ay+17<<m0#M==h(!h}!^)kHZ3xRKUzoP=YVb7CM>ZV$*0%is
zPR67BlDB`aM1PzH5Tes4O|UZC>Gq^6S3msZ{-yBSzI?2M2oB^}BXj^?D{BBuhe;Nt
zK|a#SFquSHS%Q(wW1&<h*h2?hg~x#aDB9kYnEeIxSM~<?`+&NojVSX?fNyD~icZ0d
z_FDlQ$sjvXu$z^n65*vzPy@B;(q=9|_kYHFdEej{;jq!*|I?J+wriY8GP0<XSZO4d
ziMCr-m-m40`Eglw<<D{WYH?ZYc?ZVI6L#$eT~*b^V4F#V;HT|6Ud6Cz<5esS%liiK
zjW9eJM`<}??;Y|dsO%ji?;`mHl5HTSxwz@XB2B~a!X8MS04Lo>;sC)CZtGo@wysy*
s>PyvHwPr_Pd-fuMX=y5-*jtYsxR6$ID9cpumUiC5Gt%y{N*m690nwEqPyhe`

literal 0
HcmV?d00001

diff --git a/ldm/modules/attention.py b/ldm/modules/attention.py
new file mode 100644
index 0000000..ac45171
--- /dev/null
+++ b/ldm/modules/attention.py
@@ -0,0 +1,343 @@
+from inspect import isfunction
+import math
+import torch
+import torch.nn.functional as F
+from torch import nn, einsum
+from einops import rearrange, repeat
+from typing import Optional, Any
+
+from ldm.modules.diffusionmodules.util import checkpoint
+from ldm import xformers_state
+
+# try:
+#     import xformers
+#     import xformers.ops
+#     XFORMERS_IS_AVAILBLE = True
+# except:
+#     XFORMERS_IS_AVAILBLE = False
+
+# CrossAttn precision handling
+import os
+_ATTN_PRECISION = os.environ.get("ATTN_PRECISION", "fp32")
+
+def exists(val):
+    return val is not None
+
+
+def uniq(arr):
+    return{el: True for el in arr}.keys()
+
+
+def default(val, d):
+    if exists(val):
+        return val
+    return d() if isfunction(d) else d
+
+
+def max_neg_value(t):
+    return -torch.finfo(t.dtype).max
+
+
+def init_(tensor):
+    dim = tensor.shape[-1]
+    std = 1 / math.sqrt(dim)
+    tensor.uniform_(-std, std)
+    return tensor
+
+
+# feedforward
+class GEGLU(nn.Module):
+    def __init__(self, dim_in, dim_out):
+        super().__init__()
+        self.proj = nn.Linear(dim_in, dim_out * 2)
+
+    def forward(self, x):
+        x, gate = self.proj(x).chunk(2, dim=-1)
+        return x * F.gelu(gate)
+
+
+class FeedForward(nn.Module):
+    def __init__(self, dim, dim_out=None, mult=4, glu=False, dropout=0.):
+        super().__init__()
+        inner_dim = int(dim * mult)
+        dim_out = default(dim_out, dim)
+        project_in = nn.Sequential(
+            nn.Linear(dim, inner_dim),
+            nn.GELU()
+        ) if not glu else GEGLU(dim, inner_dim)
+
+        self.net = nn.Sequential(
+            project_in,
+            nn.Dropout(dropout),
+            nn.Linear(inner_dim, dim_out)
+        )
+
+    def forward(self, x):
+        return self.net(x)
+
+
+def zero_module(module):
+    """
+    Zero out the parameters of a module and return it.
+    """
+    for p in module.parameters():
+        p.detach().zero_()
+    return module
+
+
+def Normalize(in_channels):
+    return torch.nn.GroupNorm(num_groups=32, num_channels=in_channels, eps=1e-6, affine=True)
+
+
+class SpatialSelfAttention(nn.Module):
+    def __init__(self, in_channels):
+        super().__init__()
+        self.in_channels = in_channels
+
+        self.norm = Normalize(in_channels)
+        self.q = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.k = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.v = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.proj_out = torch.nn.Conv2d(in_channels,
+                                        in_channels,
+                                        kernel_size=1,
+                                        stride=1,
+                                        padding=0)
+
+    def forward(self, x):
+        h_ = x
+        h_ = self.norm(h_)
+        q = self.q(h_)
+        k = self.k(h_)
+        v = self.v(h_)
+
+        # compute attention
+        b,c,h,w = q.shape
+        q = rearrange(q, 'b c h w -> b (h w) c')
+        k = rearrange(k, 'b c h w -> b c (h w)')
+        w_ = torch.einsum('bij,bjk->bik', q, k)
+
+        w_ = w_ * (int(c)**(-0.5))
+        w_ = torch.nn.functional.softmax(w_, dim=2)
+
+        # attend to values
+        v = rearrange(v, 'b c h w -> b c (h w)')
+        w_ = rearrange(w_, 'b i j -> b j i')
+        h_ = torch.einsum('bij,bjk->bik', v, w_)
+        h_ = rearrange(h_, 'b c (h w) -> b c h w', h=h)
+        h_ = self.proj_out(h_)
+
+        return x+h_
+
+
+class CrossAttention(nn.Module):
+    def __init__(self, query_dim, context_dim=None, heads=8, dim_head=64, dropout=0.):
+        super().__init__()
+        inner_dim = dim_head * heads
+        context_dim = default(context_dim, query_dim)
+
+        self.scale = dim_head ** -0.5
+        self.heads = heads
+
+        self.to_q = nn.Linear(query_dim, inner_dim, bias=False)
+        self.to_k = nn.Linear(context_dim, inner_dim, bias=False)
+        self.to_v = nn.Linear(context_dim, inner_dim, bias=False)
+
+        self.to_out = nn.Sequential(
+            nn.Linear(inner_dim, query_dim),
+            nn.Dropout(dropout)
+        )
+
+    def forward(self, x, context=None, mask=None):
+        h = self.heads
+
+        q = self.to_q(x)
+        context = default(context, x)
+        k = self.to_k(context)
+        v = self.to_v(context)
+
+        q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> (b h) n d', h=h), (q, k, v))
+
+        # force cast to fp32 to avoid overflowing
+        if _ATTN_PRECISION =="fp32":
+            # with torch.autocast(enabled=False, device_type = 'cuda'):
+            with torch.autocast(enabled=False, device_type="cuda" if str(x.device).startswith("cuda") else "cpu"):
+                q, k = q.float(), k.float()
+                sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
+        else:
+            sim = einsum('b i d, b j d -> b i j', q, k) * self.scale
+        
+        del q, k
+    
+        if exists(mask):
+            mask = rearrange(mask, 'b ... -> b (...)')
+            max_neg_value = -torch.finfo(sim.dtype).max
+            mask = repeat(mask, 'b j -> (b h) () j', h=h)
+            sim.masked_fill_(~mask, max_neg_value)
+
+        # attention, what we cannot get enough of
+        sim = sim.softmax(dim=-1)
+
+        out = einsum('b i j, b j d -> b i d', sim, v)
+        out = rearrange(out, '(b h) n d -> b n (h d)', h=h)
+        return self.to_out(out)
+
+
+class MemoryEfficientCrossAttention(nn.Module):
+    # https://github.com/MatthieuTPHR/diffusers/blob/d80b531ff8060ec1ea982b65a1b8df70f73aa67c/src/diffusers/models/attention.py#L223
+    def __init__(self, query_dim, context_dim=None, heads=8, dim_head=64, dropout=0.0):
+        super().__init__()
+        print(f"Setting up {self.__class__.__name__}. Query dim is {query_dim}, context_dim is {context_dim} and using "
+              f"{heads} heads.")
+        inner_dim = dim_head * heads
+        context_dim = default(context_dim, query_dim)
+
+        self.heads = heads
+        self.dim_head = dim_head
+
+        self.to_q = nn.Linear(query_dim, inner_dim, bias=False)
+        self.to_k = nn.Linear(context_dim, inner_dim, bias=False)
+        self.to_v = nn.Linear(context_dim, inner_dim, bias=False)
+
+        self.to_out = nn.Sequential(nn.Linear(inner_dim, query_dim), nn.Dropout(dropout))
+        self.attention_op: Optional[Any] = None
+
+    def forward(self, x, context=None, mask=None):
+        q = self.to_q(x)
+        context = default(context, x)
+        k = self.to_k(context)
+        v = self.to_v(context)
+
+        b, _, _ = q.shape
+        q, k, v = map(
+            lambda t: t.unsqueeze(3)
+            .reshape(b, t.shape[1], self.heads, self.dim_head)
+            .permute(0, 2, 1, 3)
+            .reshape(b * self.heads, t.shape[1], self.dim_head)
+            .contiguous(),
+            (q, k, v),
+        )
+
+        # actually compute the attention, what we cannot get enough of
+        out = xformers_state.xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=self.attention_op)
+
+        if exists(mask):
+            raise NotImplementedError
+        out = (
+            out.unsqueeze(0)
+            .reshape(b, self.heads, out.shape[1], self.dim_head)
+            .permute(0, 2, 1, 3)
+            .reshape(b, out.shape[1], self.heads * self.dim_head)
+        )
+        return self.to_out(out)
+
+
+class BasicTransformerBlock(nn.Module):
+    ATTENTION_MODES = {
+        "softmax": CrossAttention,  # vanilla attention
+        "softmax-xformers": MemoryEfficientCrossAttention
+    }
+    def __init__(self, dim, n_heads, d_head, dropout=0., context_dim=None, gated_ff=True, checkpoint=True,
+                 disable_self_attn=False):
+        super().__init__()
+        # attn_mode = "softmax-xformers" if XFORMERS_IS_AVAILBLE else "softmax"
+        attn_mode = "softmax-xformers" if xformers_state.is_xformers_available() else "softmax"
+        assert attn_mode in self.ATTENTION_MODES
+        attn_cls = self.ATTENTION_MODES[attn_mode]
+        self.disable_self_attn = disable_self_attn
+        self.attn1 = attn_cls(query_dim=dim, heads=n_heads, dim_head=d_head, dropout=dropout,
+                              context_dim=context_dim if self.disable_self_attn else None)  # is a self-attention if not self.disable_self_attn
+        self.ff = FeedForward(dim, dropout=dropout, glu=gated_ff)
+        self.attn2 = attn_cls(query_dim=dim, context_dim=context_dim,
+                              heads=n_heads, dim_head=d_head, dropout=dropout)  # is self-attn if context is none
+        self.norm1 = nn.LayerNorm(dim)
+        self.norm2 = nn.LayerNorm(dim)
+        self.norm3 = nn.LayerNorm(dim)
+        self.checkpoint = checkpoint
+
+    def forward(self, x, context=None):
+        return checkpoint(self._forward, (x, context), self.parameters(), self.checkpoint)
+
+    def _forward(self, x, context=None):
+        x = self.attn1(self.norm1(x), context=context if self.disable_self_attn else None) + x
+        x = self.attn2(self.norm2(x), context=context) + x
+        x = self.ff(self.norm3(x)) + x
+        return x
+
+
+class SpatialTransformer(nn.Module):
+    """
+    Transformer block for image-like data.
+    First, project the input (aka embedding)
+    and reshape to b, t, d.
+    Then apply standard transformer action.
+    Finally, reshape to image
+    NEW: use_linear for more efficiency instead of the 1x1 convs
+    """
+    def __init__(self, in_channels, n_heads, d_head,
+                 depth=1, dropout=0., context_dim=None,
+                 disable_self_attn=False, use_linear=False,
+                 use_checkpoint=True):
+        super().__init__()
+        if exists(context_dim) and not isinstance(context_dim, list):
+            context_dim = [context_dim]
+        self.in_channels = in_channels
+        inner_dim = n_heads * d_head
+        self.norm = Normalize(in_channels)
+        if not use_linear:
+            self.proj_in = nn.Conv2d(in_channels,
+                                     inner_dim,
+                                     kernel_size=1,
+                                     stride=1,
+                                     padding=0)
+        else:
+            self.proj_in = nn.Linear(in_channels, inner_dim)
+
+        self.transformer_blocks = nn.ModuleList(
+            [BasicTransformerBlock(inner_dim, n_heads, d_head, dropout=dropout, context_dim=context_dim[d],
+                                   disable_self_attn=disable_self_attn, checkpoint=use_checkpoint)
+                for d in range(depth)]
+        )
+        if not use_linear:
+            self.proj_out = zero_module(nn.Conv2d(inner_dim,
+                                                  in_channels,
+                                                  kernel_size=1,
+                                                  stride=1,
+                                                  padding=0))
+        else:
+            self.proj_out = zero_module(nn.Linear(in_channels, inner_dim))
+        self.use_linear = use_linear
+
+    def forward(self, x, context=None):
+        # note: if no context is given, cross-attention defaults to self-attention
+        if not isinstance(context, list):
+            context = [context]
+        b, c, h, w = x.shape
+        x_in = x
+        x = self.norm(x)
+        if not self.use_linear:
+            x = self.proj_in(x)
+        x = rearrange(x, 'b c h w -> b (h w) c').contiguous()
+        if self.use_linear:
+            x = self.proj_in(x)
+        for i, block in enumerate(self.transformer_blocks):
+            x = block(x, context=context[i])
+        if self.use_linear:
+            x = self.proj_out(x)
+        x = rearrange(x, 'b (h w) c -> b c h w', h=h, w=w).contiguous()
+        if not self.use_linear:
+            x = self.proj_out(x)
+        return x + x_in
+
diff --git a/ldm/modules/diffusionmodules/__init__.py b/ldm/modules/diffusionmodules/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ldm/modules/diffusionmodules/__pycache__/__init__.cpython-310.pyc b/ldm/modules/diffusionmodules/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..3efbeded0a780ba679cc47823ca6f8f13fe9b7de
GIT binary patch
literal 181
zcmd1j<>g`kg0}yCsUZ3>h(HF6K#l_t7qb9~6oz01O-8?!3`HPe1o10EKO;XkRX;Dk
zBsD2NKU+UJKP6SaAhoDCKQA#yKiJa1#KOqZ6e8^G92}&dlai~So1apelUl5wl9`rP
jTAZ1m2NQ^o&&<m#iI3MSsJz8t0}-_Y*;vd3Bv=>#dvYyb

literal 0
HcmV?d00001

diff --git a/ldm/modules/diffusionmodules/__pycache__/__init__.cpython-37.pyc b/ldm/modules/diffusionmodules/__pycache__/__init__.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..db54141262ad3e5ecd861f580e934211005a9843
GIT binary patch
literal 178
zcmZ?b<>g`kf{4w+sUZ3>h=2h`Aj1KOi&=m~3PUi1CZpd<h9ZzKg7_7!pOK%Ns-Kr%
zlA4sCpRJ#gSdyrpUX)*2pdVbC=aZS2o}7`Ir=OgklB(~LnU?0{8Kj?+lB=JapHiBW
nTCAS}6e}&x%+G@f#K&jmWtPOp>lIYq;;?~;+JP+n48#loP|Ghv

literal 0
HcmV?d00001

diff --git a/ldm/modules/diffusionmodules/__pycache__/model.cpython-310.pyc b/ldm/modules/diffusionmodules/__pycache__/model.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..225745b5c028b4092283af4f039df4c915d3ef89
GIT binary patch
literal 20783
zcmc(H3v?XUdEU(I?C$Jhu~;lX5aL5xQW7OdCP7k`^{`AqBK5E<A+{t}4UU}EVrKv>
zxYz~n3@8yDG&-hg%W3Gk@j0m<aSFJ3m`>WdNu5WV#!lVZ?WvQxJ!$LA={ZRg*S1dL
z#;NKiRYkGi_urY>T@bL89-q?%=HEO2+_`h_fB*Nrv)$q0qJhr`u7242y$=}1AMvFB
zW8-87hyT;2VYr5Cwv9RcZO)l;ZmneIGN#7ab6Mc5c5Wp<mls^7?W`2$3M<9AqMY09
zp_S5HNy^#w@Je~EjB?J+zi7;jxQ<)Ecf~#K4!NZl&AHJ>40qTqzi7B+FSA^|n3)>`
zX2h)kQ;9L-z>K<8V5%`@4=`iyI56WeW&)T!?gTIsF=j6?d)-N3CcP5copevQ``rC_
z=5@Goz&(g72jjZ?fVtki0hk+N%zj{M?jc|f#h3%Y9CmL6=EfLv5SW|Xn}NAG##|4~
z5%(5gZiz8B0CTH51<X{8sR47Fdpj_<$CyLFyv02V%+VNg7?@-3G%(XK=0?|i(wI8=
zGG}h~vKbYW*HEg_S@fo?$oZkwpw;a(+EM0#&V?y6I{3J^(p488T3Bc`TV5wPt-8Md
zKoFpadZ<u-exa*YJmuH@pb>aYi|xiAGDHr?X&nBq0R%>9ESnci*F0<ag?Ab6H$&64
z_8A*iU~OjH%$6Bit~qJ!H#RbXy_pT{<y>e1=3V=e6=sHvabwHe$b^|?2Q3xc>?NKV
zo5d|MM(%G~Zf?Q4&ImI+<2ld9rM2ZDv{(x9415psdlAoa^Or2NfL=_1Vb5A>4o@pD
zk8D=ZhO@<M8IY^cH~^k5*6?|1spZ#J8bNc(^J~G9S8J`Tw!IY)q(PF@TAkV%uhVV$
zt<GZYOlx6b%_sWZsypA9Z?#)~&}!Bm@48;wzgHT3Sanxw4>wxM_s$&?%!6yKwp&93
zYks%oHrh3BW!`gLH0bLJZ=w&FA-}6m>aVxd=GTIni_T!8YJR)5xD>Q6$g>vsbPmJy
z{HAKj^m7bPdQBP59mnphbr))f(>brb)2}_yZr6UeyH=a^Jom8f(#xQa@iJ$>Rk_75
zz!_OLxUlM>kaNBBt)@q2%ebwTDO<6watNeEFf|%w+FmEh`b&*fFS0@5C1ARXQ8wtR
zCd%IPt5L2YoEqg9+TBKQ$H~aT#Ub1tWd##unvEdJfOIHz{mANcqT<3@N5(>~UTwHj
zc2o#jE1n;Ct9~?`(zcGi%{fbr_5w?I-pc%CLmfinufBJBsk`D$ce;T$-|e2AZeseT
zS3ygP%Jh@>9DnQGC-1paqfeiH@`>rTyE47fb=TUSKTWa=4<rbDd+lSZ7ov$pFQ_M7
zsiz&8z$5)}fU4tIRr7#ZHg7VoH%n&4v`uv*E;LE8<ij#0O{A=hLI^qAG($+rxbdWM
z*_@rSR2BFr@3$6Lz)_e8^LZJW28VgE)aiEGtqznyE4XkwFg_=<Y^huE^>PhhK(=`?
zm2Gx5fMhRQA)X7VvfRui8**dME?)Z1=QscS%bz>-6zE>;tgX}+Rd)@9FR|QQYIHh~
z1W+5OTmx#R<4t9xyJ|mLjS7#j`Ydz|)YDL_lis3k!{zORD6nFq-CFmK<Eo5c+*B+#
zaSVJi@!2akq2L*`P;=G?jRhzt*Lu;JgF13;SQz_*pzd52gULqDdsqDi#l4x_%|=+x
z$KU?jPza2Rn1C(AG(syh>Nu7ybzP8IwwJRYkUT+1l*QGYi%H@EV!h0!L@+LAqM@^%
z0&VI(Mib@yK($;i%7cNa2rlO#DQjTC$f?&`omNn<M}_BF!BV~1?VRtUU+y%{Pr8v!
zJX*DVuf3qA`JA@{fE5b$db8c|{d&E}4eDN8yAy}-VbRRt&oQm|uk@OIb$v9U#j&IV
z*YHOc4V-2pWFgHT%ha`+KQbHOm1SF1p}#>E=yi}r^9-u<GZW>>kC0T%eaUY&+FpI3
z(S*oDKR}44a=n471Gs?<5_oF0+lDEU@u_zJec4j?<4Zv=UBfYXn6Kv=%4G$gDrp=;
zEke-NE9n5^DwlI4cBir8f!M`*UCT4drF#9`TBEJ62<g-#Xek<wDckhg?S#72Cu)X(
zQm0M=ycLH}S#k`>cu{aooDO`*8T_ad0AR)9neKC)9~d)&)h`V)V|G^E!N(N{sA3em
zsqQ8?MKD7kqmhA-VL0f1hOzn>=JoZlUXOQX84fX3DG%cW>|v<At_k-5s%07S1L?6S
zIq0rU_UUJwxK0iXmc?K05*dsPm7l%V<Zz}V2QmS#r4z^ip(W1{n?2uXuX!?XpeO{2
zLOqq&3`I>Us!2Fb9povQEGjn6FV<JP-S$Zr8mz}~QbP#UqZTRk1P-72ItOl>AkSsS
zuY&KYeRPUBPd$oieY$Rjr$Y9}Q6i~&{4=D8yM}*gA^7o+0VL{9OQLK4R7MmZi1uO;
ziZAav1B$N@4L#xc9WQvW-EE$oeYj;n`%h(}Qg<y#^>H*z4O91*x+-X{1<`N-gO*-R
zEf#eGCpnLU6YhTlpig09QlT&{*H|-DIW#+%@C}>;3r`Ht4xlS8vQ{q0FFVTxVF_v!
z+8pw5G0eL5SzFz~>zhMiHng^Qe+hS5a_6vIDTg-vO`|>%x5KCQw^O0o#*sOjQLCb<
z=68AGDAjl0HTK1hds9u-dm>R}JuOTWtEn+{4-u3ft(M6b(de4*i7HPCDpoAYc3|L6
z^r^l3SUXG7pNMjgiT#FqL(;FR?y|0bGRmJ(-Brw1ly~ET+;UPq$=gdE*wtj>pj3r-
z#YC_#Qb#?*(o+O81U*(}$KQv;p9LsdIe00w7DcOPFPzt$EYEURi&nYMjySKCvV(TT
z$@Qs>zgQpdW4oTtNK_%7Cyr05Gd>GMr~?9o01OlYGt6An7=#ZJI*_$Ybqkqklh?r#
zTZ~Q(1P>`(u&*=rB5Xh%`<j-raZAQ#E-5YN!5|J8<vsW=sK;<$F)?XjW-r(Sn4v+8
zqT?}DQhya+;$*3(S^8muIRYU$N!#NK^<JKG($tRtK(D2GCbBu}k-5~TCqzY1*IdNm
z*8ptj8Ha~;M5d^SatV_Lj9q-)vDU}>N&bfF=RIhW-Sn9maL~GCR<2f#DB`bCLCL%&
zCai#%q9ZEh8qXos4u1m;ZPyL~d<c<a=z`@ONUvQg?Nv|%IJ@8ryWoqv;D;!?m+jfU
zP=!VibGA)n>0m}SMRU7kn&%?(Y-FB~9ExKdBA7TMB40DOahoiC5-9N^>2+<FqC`3<
zQ9G4OudP>8s`exX>N@=}T1qA6+c<&5K+SE!SA~4A1ep*?F*czRQL=%{NC}+3Y(Haf
z<Tmmf&PE~3!2agKeCUJ)1m+nA%bPiEROAdH{J?cP$ZtB3JR5;^F)Rj!Fpp$_J;o5d
z2nAYP9$GGiMX1j#!g46`ZTDxxVqg2X1FZpv*3i>l-7Q<jF&vKw<Dmpj<z}@bBQ6cK
z@S-NlAACWTP?{<$vaE6mJc0#+5r8NUCsg9nJY4RTH8_`5qBHX=2o&LS_8bBPS>1(T
zrSZIC>r+F00%|SQw>2J_5ZI=AmTj5OMb>k%3*OVN6yDF{@Tn{)y!x-4sUj+@N=#Ta
ztC_Ntz%ADLXm9@CP<$I`^b>5I3?bg<K1JNFqAQ+!&wH5DegiBo0wx<aFAKjRD*;oE
z1dk@SrH|6StzVJs0~0Fm^^r(PvPd;A$uw=ttGp$x?|WWtDF{~md#9%t5%H|eA8U43
zrXOzv!BWdxd+ObfJ`u;<2+*hJ+uiwT_wM8KciwSgVd3uMZ#(WaPk4=c?mjvHwmTaq
z=I?eF?mE73*By<<+wN*k`>NS@uLSD;G#b|FF2w9NK6dit9cfo)kr22S*%M@h@ZVs{
zX0h!ZUGf@8-yqI9n%-LLf^j54BscG_@LmM@|9>hxI!aV~Y91Y#8s16MX)&PyXtuwP
zrEO9}=8sD0@ax@GwZv*O1U>eq^YEW=_>3$5UlgA6S~0tRXg4m~Sp!o3DgLxa{+|H+
z<u%}kj1QUAf<czr?nQVTp&e!yG8e69panPVjcfoh_@KGKSOwXH9j?Q0iTf&dx{P&-
z48k^?5KgO`-7+^zFCnQJI?J}3yUuvY+(OzCaZMhHO$TLM?YoxQaDsA(oH(%1$S;|0
z@lp<L8Ji=a10C+bNIIeghh|e>979KrLKA{Mg;;UsXkGK71^KmRUmc&XHET<?=W0h!
z)#hurqBK=&D%ulugg~Y;ozKX*uiaRgcN?c(Wpk`nG}ce<Iz6$1I(8r8ibmT%g@V5Q
znl4g7jj_!c09q>-UfZUL(ZWf0k>qtPp^#*4U_VGLTy?GSv<9wcHtj81f&o{Y6wuLL
zHY%)j{Bvubx9+JH(Nz9w86Yac#SU7FYuz;;DNPe*JTf1Q%+r0$d>-|#;E0_eYG)=B
zkh>^A5IXhwRs$JHNP;|(9H^t~J5VB_W9n*B3)SJ@by@9Mf^!6`08^FS4OM2Pv8wW@
z6gi1XjWS*2*7mJPBEO#M*Lo_PDqzdc5y+r9UPa;1CalmiVX8yypr}XYqmlVE9<ml_
zTCB6qVFDVjKf@tbqu+S#GcBuRI@#i@`y2<RjHylQwecK%)AiaLu?r&cce43+5&RW^
zC||F;-6m2j)F<*-^%zS%?wsMNbjnBR-{FvC4_($$uCiYddyUnzf#k=(9w4=PNC+ZH
zNSjxn{j-rGBuhm%ZR_{oQpwY*u?iC~+e{i|!_-8aEYU<~PzcRs1O65ChiDgZY%f|H
zW?)@JAj(h|%Munc8xA5D$IXa$g>YB&Mk3+h?lrUpzRQzoz2^g_)BUsRM~TgB=O$Zo
zxlFkDwVbx8{#pMJ5bDPWgd53&eT{s8r!xep9*`-hg3Uz~%5YlXq;PKE@@8msELIDb
zt=ZYB?6$dngV}uv{TStgNoL2GFcmRT{(Pg;YPTCrjxzbMOucd%hyPCin3_Cq_@#G2
zZ-nM24L9>CM33(`nNHs_)C8{ESR$Hmvz!s=X4XOE2k3k%d(lL`SJ#-<2@L;sE^#4T
zF)ktx#k>^mz8`lM<enm1H;n-PHLt^#4B`502v6KC-Dd<jx7<ScXUq@lyN7Z0Y4LfH
z+DHAuMYKLOLK^)zdjD!A9^}zP@?YJrh5l$N^hcXqzoT2yiE=vI&$y-ItrtZ_amA*#
z?$p*gXFJF*O1tV38eiYH(m2Z%tyF7pRY<DevYx-C#?*6dy@ZvcH085?%RJYyc+<97
ze;|qb;@d|0Z##DEnEEiD7!{z1)djA_sE^=WxNEAYmD`?{Ur$C^Z=7$m+Ku_P7iCu!
zo$&^%Q=$<QV?e?JdNsj$r%P8ZxDbuOcY9~Da_6}ZDb-bwjcIvuN{hK6DV{}!c+u}u
zC)6ZS0VMILMmzh+aW&v3&5Ffjo&%NU;IAUkfv?V>J7=knp;8maz=tY;AD@ZURL_Wx
zVXBkK%%94M)?sFj8M?Atbce72ZM!9Rc);f#iL#G#ed)5TqgJYzmnjS{)8KpyhkTxT
zSmqWkdcL_-Us-Dhl5U792z0Q*QBRi=q7pMv$s$!W!ZXb{GII5I(Ohb+8AeUwDB_@W
z`M(PnpwF8fQUzplOVDHSs))AFbu9KWPXlZgeiTxBAta>ExDvs9SOmm5@+uDo#m%8D
zSac-wikrhhdASJxCQj3BS?M_>ZaK4MEh2hFsZ_7<%JL}NgSSAEf~b-#tcHb+VK*0y
z@%hO3l-aI4VLaG_+K}oq#{1?rMl^n-5`T|k4CQb{7MH6iXX5gh+?A8N4sMKxmCcD@
zZ&(e-*rO187cL>`2t6^jQ4NQ~?0G|dJRC)xNz^I1&L#BNG|q*13dXsyC#-~f>eq#%
zfcrv@6uN%WK%VLI!G7G4zi8pgfpFL@U=e*7*T**}!adSPL$?819h4qmnI-7~N(bxL
z_ns-|H|$uerpJp`4h4r3ipc2B?nslA4t+G4&@^%Fz0h-@=26gP9M8<1wLW4}tEQBk
zNa;1P7c}bic#vke>#f&P?IY$nC#Kt8&~2~KZ7=EO7SCpI&kqGR#x&bozlk(M)`2uT
zw0W~z4EJtXp!^>PN5ZOb;w>@HRH2RgIp$VIM^i(y>TiR=6|*&Zp+x*SriU(@B6$!%
zDiiet!BYfGl;I*JrnjszHPs8TA`9!>`8aQ`%BUBOaFrA?*=?_JIo^-TF`|yZwt3c%
zioJ{ad7ErrmQHg}1XzJqbv9JgCy!zOfzDC=HDa?+hrXv6!HKhQ9&3qMa9_g`XJosu
z08v&ixaa7-bX-a7i4u#l(t*gP{*xt#sK^`il2nvgX}M8es??E~7A<-rL15W_Z8dU~
zcODVC=gNXt<X}-))>flji+u`o!rNh1jd?R!tvS!q)TH)Ef0n(@0RfYvei|q0GsINn
zVLh_=LU{G}SV3mwGE18TTLhdBG`^-32^|&1go#Rw@~wIsoouULV#QyUih4{SQ$d&5
zHPn+6^^Z_xF7Kaht%8v1*Ll|r!65>1Kk1<!r@8nI6p9%K0YZtULH{um3j@OSM+?bv
zfma-dc~$;7uh}`)ra7UBDM`s<ZPU48+YY=e>jk!%!xfq!_+A-a<x#v+v3uoWjvkqA
zg&sKndM(jov*VVKAx6ClTI!|rk!UQ7ok>INojqIet-guz!f$2XeVEI@)D4J?Vc?ME
zqlUKB3GtXre<JpnxbR}b^My57Mv~XfU{jBcOg=J*`a8y(3BH5uy<{p8>0)b-i3prF
zqbM5Z%tlVmhUDzl4QA_0A)@N#;V_RDpjD+EOQPuH+6Z!!87w3%rL{+-wu62k?}#lz
zh^(PA9)cIn=p7rEpqbw<SK>A{M|Im<=o4z5(r1jzm#dp&ut&^<>a+1UPUn?9$(3Lt
z8HMzGCK-ndTP2v=WGs7oV-fnYcavL&Y;XdNL9t|Hx@I!6|A>CXnBK_F{Q~S17rp*4
zxw?6s?%TOfBl6y-%a{jE2V--8+;c}q)}jdM2}qp#@>P7M4rrZ2-#|;=jVw`sMLVfq
zWa(Cde?ag#f}ba#M2l2u5u{EOIaL3cB|7bj=~4A71iwn~YXtw4fZ~T8KM1)08BhN?
z!EXW##5Na*l>kT7q_yO$FA?=+f@cXtb_Iw4GdlVx7}|df1?<m($d(Yy($j;;SS93Q
z%jP~r{^JlEmLQOlpGCe_vbGbZeRbR#7ePhX`vS#^0>3`21uBkyQl|=!kt_|KsY4YK
z7*hj5LYnR-oM4%n<BD?6dED7GWorF6*zu^$-A35xL2$uSU*Z_xo*4+s#`Tj2E2zK4
zrnkr1eIfS0#r56e?2qH>G@rpaa}25$tPQ?iY*vmdUTfkQ_@rJijp&>5f(vdDzAxf&
z#NgbZGZf_?>TqwCQlCH@>QexzqpK~%GsM#rM2_2CL(EWL?^%JA<in_t__16s@qZBm
z>GNYt;>Qj{kHL>ctcb<w@8`h=1+HxD;C+_heU8BU#CAmSJ}dBBj5>YL5%EDQ!6+s=
z@j*v5exn+HkKy^1a8!KIaq85#yhrXze9(!ox><#PJ08-X?e{_Vzz-V{AN09!40Y(0
z4W~Zn9G<cVJ}AA{z4ZxrV|(j+;feNq(2v4{92XyyzSfBNp!88Io47;t-Tk@^jA@_r
zfL>qR1C;jF_xGMD=LdGIbuhR-92cMLhM1D$;gI?&qUhqdfeix10spKf{#h~h&oFP>
z{WJAx;{HCt&k!&|nmW|;!n9iYIbITV)6+xZIEV|eE%_}<>mTxte?jmI1at+oA`?~h
z9-e-T;Ijm&Lm{ry?uzU<EB7>4pXZ~#z)PvyqrS+~l)ZbNPE=e^Lvr=oth7ZSUe9oA
z5$=^&kK^0l;sqIjsK8zy|0OT{D*{oAGPD_<9wInQK)>c+gW5i`phb8hvI#&SmiWa7
z>3J|*Z~npfF!XS=Zv>B|5=(#TlMF1gP`~gvjl=(+0B{%}x|>MMh{M3m0B}WOhryOJ
zh$RA+TQ{((jLk=I60m_U3s)iI(($lhR&xQ3zrM!BObZUfT|JD{hA3`es{j@&b0QYw
z400KWio@_8xe~XriA7bm9XkwA3_Qc#a4@ie0K1{#Fz7QlFi7m{@^J6UW?AEdkx)O^
zfTN;oa!n4a*<ZofdpYQ-(I1e_6)qJUE=D%rmo}zpRsV)S80-7m)`ErUIfA<1-<GCU
zb0nGcH(c%fDjNL?S(MI6&mvVGP`uK_G4P?8<wq9dx7i~m5OiK9KaiK%!6KdUcCkqL
z?G~xD*gEfF+m2U9q!l^Xn6|`lrzfweeWG5v4hQorVxPW&VfERkVX;rxKK178Q#gcr
zg|Gw&T^bCf)`&JG*7ss%Mni15PcUN!at>A3Mt-ImRu`DJDFw)-fC1tC2HZlAZ!j`M
z3R%w4<?8Yn>W*Rsi^Ovz|1uuH+Z9{KJjtlo7g~tIC5p&IeKKJ71L-b!u{__mW3ByY
z<$$yTTM^uVny{H7x=(G4>1!K&66zqQ!cmRMcqXK7F3n6FgtZ&rtjRMb&`R7B3T-&C
zF^X|AWAmHg9@I(uB_h40$FV_+I9WdgTRT}l9F9R9WYm8~-eFYi<RtPARX2;g1G6y)
zk$14d$t~s`ehhhs(Kzp*DK(0SsGm}Uc?Zn5p$!??RBYJ#?N4`=<Y;v=mI+$fv{3u&
z8?|>3AeD-6pJ{|+I%h#ot7i%OQxR{*2>yHC{T~Qu`sVWM*h7ZB`|6`Oi*gW0?BT}_
z{ME?D-WPY<&KRw9#1@JJFZNNqcrlVaS6+<iuk$J5)Jtd}7PV)dPV%bw_`ku@Hk)}}
zFI(_tZKnDiw$mHm7Eft0=StdQ#s*Vtr5MNGVvTK9vu7}+`!lR^h#)nXN2u#w0BXCz
zr12~vTHr1ZhVr?qj;+lclK{nhFO1(gq-SW~3+DUN0y;B&O)y`H)zr?WGlwo_afs>s
zUnp>$3i6J9b7DfVTB3IhEn|fh(nTXBIaut;(Yx*-OWZA|uLSvkF>fK3(Yp}U^BfU$
z9Mv8YgF7S!x1<fO7~2dyWdoix4RJX(w(ss?bWQH}J2xHLB75&kt|Ct?Jvny-7%ZUj
zwH_G%QH>3%)OoVs51HV2EaU1zum|~6ea_c_GJRXYruNFZ*<49Ns)YB&6Wlc--4@ph
zTL>Fdq3gI-*jPAk(4J~RtD+tjHDDRyZJLp}t|RDY*+S2{3ZM0u?Gaq<klBu*;8=F1
zm?>u%o-!nzkYJU0BZi|n_-fpCCc7~hmX0K>@rDNWYiRstK83?c3@mm8Ua35)c<mY!
zn*oxaWJloelkzfE{J>^%nN`SsG<735{7rzKx#PQIIJh|+3wT=@?hxhz-Z&QUXr(V<
zpJ05BCA$HvJ|C=$mUQj7HEt-!y2~YN5ndFSp3JNF9~Oclwx#P$=~&pnUr~&j>>Ff?
zIMvPC-C{hd)$J!ydJ1v++lje@fU0<4W8r>cMfu&wlITY&x+uHC^_@N~N2bt;jjUF6
z5~4i8x;EFXqFk)5W>~Su6m;1?%ITuQEMoE*q1&eYx|13Fu4J$<|GnbyK>BZDV6Jx7
z8=N#7$7vk?y8)6H0O3?upTVv@NSwtrR|Tm9nGLRWpM%aIfu+Pfh}11MR#7GY2bqPl
z4Gbn55^2q1(XK~kk`hbt1rJ$r?zi&P?*XF&s9RYQ*EA;6EkHzOE6PefL_N#vJIKvt
zlBj}<RI13nT3R|F8G8mv@P_j7bu{u5Y>oO<w%bi8;;xYk(u^OUeNJ4@0{j7SHo?@~
zdcbADf%JzHjU=zE#1e$Pri^YiAon#%s^mi{O;_q(#Ys=3T=5)1bA%9Jc~Be>V#T~C
zgag3}gBS4t3qfu(FTR(PmLM3cBOciO8^uLq13o1}l%jZDSi8g3tc!ioQlo-Zn4!&~
zpo9zwd{At<hu;&gla)4{b8iO@d!pESIJ701RIUl?KES_fp<KZAVeE*=UX&Fy_^H_W
zvr#yA8+S#?7)F8$HVbnN5G(gk40z2!uK-HEz}dbxROIxjKH2Qwr~h*x*7oE;DzCC6
zlPu4SB&!-RVtYn_61`oNMc#!`%iH8xdu5;DJ(7!%=wD*~K__GmFZKHIl0?r7nc;nd
z;{U7UMm`T4_bqCC8!iX8n3hp0q1<yLa<<-b3XG&I6L$o0_vALV+gTY5I1w-4ncQ|4
z7)jJEtZ)QM0o&(v?47y;rC$6kCLXDNC<^Ob5=ZOA`!v`9szc&(w0sWff4Cgnz-%E|
zjUYr^6-+-dq2Avnsss7p|3cDyj^O7Bgu?%or9UJXw8Y5^5>y{rX?QGsiQvlwJLr(F
zlg<;kXj>DY>ICSEyu}OL?<A?;z1t}m>k?#k|A0-3<#=?k@fn^TCis1TU*Qud-LSh>
z`D*2mV0YCluN*<uM6<Yhb3weNi6(PEujHfhV^F1?;0e4UjCb&q=#wVrGx@Nb$C1I|
zzaOBdN^}s)mmfn58}hiT0S^+<%LZ@?MWykP4rWPGofjeEP&ZD1&5R<t#k5{-hk8S7
z8yf*(wIR}QJM{+EDTuUKRuV<bYZ^WS!8e#dD!r-jl>;a&!X12H5gU7qm+*&ILAX3+
z{~hZbIOTF?n*ql%E)6m^0b0k7M`j(bQI@dc*-NygVipze<2&Ejp}czPyLjj;)Vevt
z%Gg)51QoBW-#$p*RDAmPMJ8?RItcaO(6IU@!M6y$P4LGAX<yWLaQZ!TLj89X;C|f6
zuf9FLh}z`!%D#QQ;?Rm4Lh<O+Z71|Ln_nKpp41jLeuwB!aZzU?|AZxRNwDO+8Z4<9
zBBXnVS)woUE%Zz>Kinyki_?0HcgopZrg)`jBM!oMPTqVdy}rJ87qSgx{m7Bz%+omh
z9|uS>ezNe$>|UPQAay>RKwdQDi@E{cFhii@NUjeiW*ppjv4`dF45U~rXJ+(bX8Pg}
z0+nl8M`jSeK|qH_yD!MbiGdcs#l*|Gc-I#bB1Hc#I@=dClSkD<1Z@DU%XinX9#3~z
z37I(&>dy$obo9T0_)Em@pp~wY;8k3t#xI$tMDuff8NSWLHsZz#35+itesr+O8J@~S
zegma%vR$eK$0(bXNk!bQn#jKmR})QCunJkXC@O)!^Fc31|ISDMBG35ao?7(WSQ{lP
zK2(uU$+{2OAD{2uFWS9X(x(E40eOg?nnj^cPg!spkZXn8Pzt~gd~XojkV<fO=(=FJ
z6FMPtiK#L!_7vf&l;9Q(aq}geBXJ|J_3|rd^<ljCAx`*baH5U2nCjo*>2DK=`TinH
z&k%4$N|F<ovwAk=?o?>h|G;(FoqJe~q9Pf~DDTCqT_iB~Q)u-J2D+bhq>Z0tNuKrv
zmZU>pWU2RDb}cH$cPB4ck0<AdHn54LE>b&Jk@u@tJhugd0BzUNa@^A}>|6=d)vo~~
z#ME0GxIH0fBd2k79lt1Y?d+v0Z$NgzMlw^B-B%_7`ehfdDW!6htb+b;)KT<g)x!ik
zu>2(Y_6k{(EDtt6uw!o$>zma}h<Y{g&)K9yuq*R3H8RA})2*f<w#@SDm69U9iDTen
z$7h2AVE&~AAolh9mONG3)oYla!hjCy)$2ayH{o;lnBM|xXW=Oq@Lqxq)@L5ny2O2W
zn)OX#{-MN-3+u-V3)rD4j3%u9MLemG(Rv~w(jZ(Z{P`ECH^AOs0|M`=#;J3%w@hoo
z+0wW?M%w5xwX`OU_t^XY@b1?D=JEk{Htxc-H*nfS7h*AAGp&O_7+&ypmW}EC#w@F@
z@X0p;APy6*<U0PYtT}}qsMk?@8&f9F>oMgg*sj)h=}hm=lvh#vdX6q;N*Gr7hebtv
z6UV?OVM_H!Ku_hPLgJY2;Pg=zhH#*5!~ZluA3s5Xp3VA0vYW?(`v&JPH1}ay3t2nv
zvjjg8Gr_lP`4$bcg?J+dJE@UQ0fh1egxdY2@hQY_SU+&chuCeM!5ba03S1wkGtOko
zz%RmO3-#-0T!eA3iXkiDc;3(@WEW$nuoNd{NVoNZ{Ll_JD6fy`t;l>28YcMO+IqQm
z6tBE+a~Tob+T7^4HSHeN#0KK!j6X#y+x(kr{4-K#42Pr;#L2`j7rbO>$FfgFeH(RP
zU4?mj`t&=%-is!D>cWb#ek7fbl>c_}5~mz3g(4mzEj=0OnFT(p-~AKMLj5kVk&R#8
zIVqWZ^#IQGCj3+mC7Y+?C9*W0c0X^*$K;Eggl?KRPp}0(O#c9kY?brp3YwVpwN-6I
zG8B%(tV$ztoP(y$r)M&O<0c$1BiI#}<JOwj*Q?i(Vpr?Y6Ga7H+Fb_k!&9{kray0@
zlGhQ`_t>{b2@a)Th9&7VR&k7Jb~<%@N{YB9Y7Bf55gSPExP>$hqSI4k#*zEG`}Vfv
z;P8MIXCICn{*DxX8Ks%D#Rss=XQK~>KLaozj@)p~*ocB=sp|uqDIRXEgP7q8U`D^f
z5ajt<L9Ei(bNUs>^*d|W`0dx)cxcDN@0le#J>``E2*z`GyP~GoMr!;usfE__sJ5Q1
zF<Ci~*OZj!8Fa75U30^AJvRO8D(WYBmAbeu4tWPoA#vn0VXC=&+!y)Lj-A|))0~Vl
z>T7MD!&WThH(p;K(a%cdN33O&R7*ZAlai-#$Sg5QiI*>s3C3#|@;gUZ8YKOgDW>Po
zA0uHp5=M}k<zXZxImBofzeTi&S0cKgn(`GJ#M>IZp&bB1JN2xNyJUQ~5X-uKj3zwx
z2yjoJcRJ9A|1ya!qU+^&K#76K%s#9N;5-C)Vjgy7SqGIKWOE72R!0<ZxN(ZzvE&_r
z*{PdUg`Huyu`JDB<2u&q=sm_+BwAN5Xq}PsS{=M<Ca;b~IlOrR=1}BHmF1m#@S|AA
zv^VJ=!yaAzPOajz@k3wkioDq&KX0XX0v_j`pCvd+@Hhd3?%wMi`o)fq5%+O|uM&KX
z;OhjxPw))_`d?!2L?UI3eWX3atE4YpU)23O1tJ(e)sWsH%e-$zw)r_FXHPL(wX4<Y
kMD<kl&dP(uLUoVvqt)xGKT=((-c&tay}3G89i15azm|`LY5)KL

literal 0
HcmV?d00001

diff --git a/ldm/modules/diffusionmodules/__pycache__/model.cpython-37.pyc b/ldm/modules/diffusionmodules/__pycache__/model.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..822ccb994f26d730789b5203a2543c4c75fe99a2
GIT binary patch
literal 21736
zcmc(H3veXIdEU(Id$9l(i#-6u9glnD9Z86H!0~84%+SZjQAg4tVHSB3Wkk|?u`>V`
zSnT3@21nd_R(A1dlZtgREjhNMR{%$eL?v<TvMe)7rBZehM|s(KI4ODJIEtM_X5yrj
zN>Zs($#%Z)@0r;LaK|I5O6G2Qd%Amiy8r&){Wl&P9nBlKKJd$v&hPxVVf+CfqAQ8a
zBe>nqnTBB-w%Ih6^sl*O%CohZSW1|>EV-0Inbk~hW|lHimS|=-b4$6+{8C<?lg*LM
z!csx<spjbB*wPsCX*=_>u{3UH?HqoK_Gx>>F1&0mO*~=PqxRU#hCSva)=O6sOOq%W
zw~Ht#h9zZ`OxPuql){ojD4Dd&C@F^}Qz$uPPoZQgESW~hv^|588K;1LXY9M}S^F@)
zc@t`m*f*o*=CJK7N{-sMpyZaY<S<Gq_A!(k3rmim<W~DOl-w4U+>Dar_6d}n2uqHl
z<aYZGl-v=P+=7xj?Kzapg(Ve~yvM!^C3l4-$53+8K82D~Vacs1nYR~EvJjTsW}D9&
zb7$V<sx8i$K~6a}rE0BJXU+<;Kd|LB+O1kMNIcrQjM|%@aW>oP@_ScS8uf<L^3JKY
z>ptpv$f6w@jP0zn)uyA|s_WG}r*5%dTttR6?sK@^H;{OSZ>*bFOxxVB+>&R#WcsFU
z%^F?HOYA1?#GdI}wmD-QHo6HfwVU=*>lxodnzfTxEk7}0l#M;JoA49sIrNmbQ&;(9
z?2haaDbhb?+36MQCc{tgiO+m0OK<B1^f>C{8~7dL?<&4!XRcc40i&1#vTj)F%lO*(
zdU1CGePs7omjI}8wIjgid<EBqwT4^Sta<e{$E|p4PNlKA)pRxij2gjFX|yVjJFRxZ
zZM0S^k2h9Uwp|uKtJ=%8<wmpNdX0MJnYQgT-3O(^^QygBIbUli*SR<^C68@4nsx;p
zY`g7-U29gH&1J{7(V?pw+`t&HI&NE?(LX0E%iCVX#$d2W6}Q<~UGtik<y$NKI;UYf
zZe2BG={bkzow`itp80z#?Ul-{@p@ODax0HEo0T7IZ&wx_$G%k$=}o}Lc#|vNDBfWh
zz>1XZUEXq#N!!k)M%^KzCG5uLTvD;E$}o`}letomXgaMR<*wDXoM7xx*L4&}@Lr|b
zDo6t2YiQA44N_iP)sc60wt{p`ur|o7G}|@ro-=`kN^++lB_%<kUh{$kK#5%24Xjox
z$gga-WNyUpt(rZT402v$({Vj#%MC_jSXVLDrR-X*xxyU2x4Hbfp(^P7+aFk1Yi~LW
zt+wYZx7!;FcFn6Ttg80**244Kt*5b2z>d~J9jm-R6nN~Z=N6jw=E7#%-flYX0#`!N
zA}YYI)12SB989e`UNss`H6F_pzUr2d7}=7QHB06ZbId$$9yJSQ(NwphyiOqU;x7qL
zh;9rSA7r;{`XHRL0aP{@=PWgW@*v|jRyToQSQm3gCZ>~PtSq$Jt!ASIzR>V4--Qxa
zsM=B|kbkp+BqnoK#AMDvdLZ(3%f~lCV3wV@n#6aLi>p`v+pp~YpWpb!vljsJd~18N
zDj;+5E#&KKwN?v60YGCJ(lv0KmNS=-p{iN*8swfJgcreTz@<hSt@srMgwre2H-Q`*
z)|!oua~idJ^rkw6d>wauCBmG092v);kXo`{F;>8%Z0qIhQo>HyNeG$bD}d@!3KK~M
z+4pU^HBx;&dYiw*Xa2fpka)%w0DTXu;ak2@#l3E+c`vb^Tu%Wo(z}2rh1#^ug-I^H
z$=8K0<`TikhNA$Fs*4!}Y1dN?+X*tjSMq_m3<za=%TYnLT5Yr%UbPzJUTk=4)q1;i
zsSjr9b9g>u2T7Kr*QDz-SJX*<=UyaO`dqbIZ`NG5TI~T(eGlqb?-r35X5LKWU)Hqb
zUuPnk|9tfE0bEGom2-UgP9z!^*2!WuE+o3Zi}j>B4z>+YfSm&z8b8pS*CRoOcn5;S
zIv3n}t?5)(YIRUH*Z>G+F5R1;I*b>H44$L5+D%9wnU}gB#lZ6i@T0i0!|0uX#Cox&
z?1xe3W|0_X8B@%gonkz>u*qDSKyKAG9RN9Bt!kP@zEG`R+^#isjety@MMuGC2+_LJ
zY(`M6pkmHc--{nb`cm^q2xXVFnl%e~DXHTjzK9LH-VSQ!A8)_d`mP`+(Ee%<Ia7;j
z0Ygy)l`w5kCh1Du&4d^~mr?gH&q?VuAd?ZwRrfI`OOnt_(uaT=CKa|2!0`PiaQCAg
zSYC*FlfC)vM?L~9W<-zU0f=U(leP){0t{&#)CbzLNOzcAl86|gV5=OqNCQz*_?Ny)
zL?vQo7WYdm(t3~uO+n#lc^V*=dWfZ|OSR^<BY*+|#xMeSNKiwVtB$aYkO}g&ORLqb
zcDs4T25ftXD=jK)CW*j^X!aa#w}8X|-X_f{vtX6YPN@$*G5#uIxq5`jqfBIp*oS(I
zNhL}K5R?>r9(RPGBZ?(;-0?-!<>eBYBi^aW+)jQbAzT!Yd^ry;nz6G3Tr?MqJm<JA
z$9t^Vu5T=!Zx|3GU{Hnjwih$@V3cgD>aMkwSKs!6Q4bO^t`(6-X2Ug)x$p1cp^y7w
zjlq2_+t|peBfi-}zKa{B(Cm;VJusmwY?byh>sjcPqJsiQD7Z)fc|T<*H<D_F^}8c}
z%D49TKJ-c7k~c@CX3S4Qe>JM(VL$vTSzgpnkum^xVk4mfVS&GXjW15LX8LN_19W2<
zOGzX}M2wvCK04IMIN3*xLTpc>Z7{LzIwB!rkcvPIQZ2~JyZdmH1RJCX{JVoR^ck>Y
zC`knVmTIr-_Gg03<Ep)d^$Id}n2}e`sHgaPp#>2gEgg8apmaPIX`_h5>idyIs8?Y@
z%m+S%Czm*CSQJ4;+-c+@ZonM><)IAq$_i)|Jr#PKquhsXJ=Q^%!HcLIvyRW>!PkZb
zObHqsp6Ms9=n|MAVuBE*a%z!Cw99&+!X7PL14c-kt|V_VreS`dO>$pPsjw$wHy!0*
z?ilM?AjoI%n^T{_dwDG*TY2P1B6;a2rbAt>pq|EPMJrR!F!uvYo@FAyeV(}<)c-6W
zxjgCu5-_%y(F94ZZ(y$V_<>#sGKwp>-J?hhuo{9r55V3oOOZ#OATL;Xt25aT<aft^
zC<8vo&vIF^_-yHfV%Y!(ik9!~h{zmFnU-*!0j(uwARoa;h8c(l{Y6L}>|#BQRnjV)
z)@aBivezijU86jIjq;ID73^6@`v`olFihK2YJ((cc$kFEi-EZjn3sYq>9z`5jg%sy
z#KA=OY19Zu7rk$Y+_V@TB*9QLmtwf~$cv@{uMyladHo?CKwjX5yU>n7Ow56RgshBR
z@J-~BC`(8VWU`+8V6vO;X1dvK&QCyQr~Qnd^>gqBBxs@UrnS_Or{qw5(#z~-L8eJq
z^Lan-<@^jh5XnjU1@hpt`Sp?Yf}aPMO~DrcZa?&X%Fp-pk2la8%Fr7|+H1R)lzEJX
zGs1j?MAM7f#udehIJSx#*Ejwypii*Ojj0ddNBtm^B_@B4$taQ_10`6@;S5y$&21>R
zVz?!iH(`%L=hfPksK6Pwz2a@wb`)u5ZlsT{t+nb<=Ya`2udDa-6XuJ7^<t>{S5VNS
z`Cq}4(3Pfn(7BmUjKTKgO}n#{U>-SA{zPx}-yPk*fX{x4pC(I&Vmzx@axm53iJG3?
zv7|K)$XnQLNyy9;G>(+mZE5In>0oBakrfZQ^T*-~p?l|jF5EhiKda(IKCdA!&&jyG
z>o}D)&)aezTv%9z0k*w7UvF<NJX7<$wT82O;n^pj3(bAFUKW;{?d1jgfz!+P-gEcL
z$^)nGJMGl(c52`Iz?tRy?ycRu{D8f3|LK+c@2S=9yT88Rs(RnM;v;Yu(6M&B!Ci3M
z(`U}y6Axt(?uZBB%YlCi|7&z0tu~!gYfcT$B$$1t;#Vte;FGwk<mH3;;L8B+|4%+h
z73&7L;pjo!P?OY<0Ww5kB$S{K!=&2YdTZYJZ}5&wo&N3dMwpeyu7!L1nPm>$lU-1_
zxqpP@?K8`d8Xq;uFufER<IB)u{G^{+NnEi$2!`2Bc2gb*;}vs-rV!j52h>B#i8?HA
z+BBu|v<SOdA4;&D+B0_xufpl-XV;T<`X=L5b1%6&3d1P_$7>dO)b`aSx>;|`hrb@B
zVBW8qcK&J_eHpvseirOL3yGQ)hB>l0;L0KpsKdDQ>d}8}n^!DQv$fbMFIVc7waSZ?
zQ)es7l{=A}tJJ|&DRl6296v+4Optx3S=(H;YiGZW%KwSm=$yIc!i0Qm{vlW>wWfO(
z8U6CW04dfcal4Nop-8i#2kx4fC)9l#?q}N)*oY$uLWXcct!<U#7O02&dyg7jziQtT
z9Pl~iwp;GSZO7?2idwPA2lZj(f;`l8ud%w_-gfE9qtnEEEHKaMxr;8{K)XN2ExJ@*
z*f(0`fcvfLa-#;nXnRY4jS@;5`84+l;t;MH$O3O@jbz6nKa|6qjbyHPZ8?_MtZgX*
zD#%8xHAuAK=9}G=t%YjLXR9$isx|h!!9*s^xysUN376|hIJL~GLFx|7Cj;{Zd}P}b
zIDoWUY}IBG+tB|5PX~y&S<5mDsggCD(6B1PvIYrP-U?G-;}2TeL8e-@+jTgBZph@O
zYVcbelBm3?q*XYqSVm~a7sVnkQO9H92lpggmK1$Cib1$4>2pm5$A`u|xOek{s%=3M
zEY_n=$$E$a!g?M@#y8gu=w4tL!cIi(y<&Au&$<FjmbNnXNi6tEvM_<Nc0x2W*vrBw
zB9cyBi&m{5L*#XKJUaN@jcedrVi?tu8vRM=n){zcZmu6CewbB06G2rXXkRBgY#w^!
zDy)hzqGSTHvb_n_hO1vPJ11`hhVzluV=XSurH1f8{DRkWwsbIv34WSW%9@JLLFQ7e
z)o3<rb*=_4{>EDLm+?TS>PFg~@xV@e^V5c%_$<uSmrT#xwf124p+1Q{t&?^ND~3HW
zX#*1zEwi6ZT`}RE)-67)JtXe?Jp*q;_r3zp7d>Hk`y+T0yKQ(U&)!V~eh;bl()I}I
zf6j*%Zx<diyvz!6AJ=b=qUK9pb{B8%p~TByLC+%2#|WP;M&Buh6Fe0$h0bBk4^GAW
z;8dNPjnrIPL0a!e(9&r+oxE@_ZZTCl*~)fnqXp0BwfdditZhIWs>Ixdds_1D9i7Y_
z6?)DqodWi~V&_ulj%Dt`@ui{Ve>5^5!`H_9Uz?wwzy2h0VB+dBca+qR00z+1xksa!
z<DrIIjX+VYU8*&jwdJN0q_z}weX7C24q&Nc)`BB?V<h{OO_k8Q985w7d@9<HbZpr>
zg8@lnFJ$+8b!kLkw*d$Es;k9Bg`JZMf*e5QkPWwHiIA)?%o(c)3w#QOe*t$E%75f}
zp({YWf;ObH_#zMC<&rRCCM67tzE}EJKa&=gMZX~ZgJX8y9>I2c(k|Ge1A6;-ka~vO
zV-P#+R;1ILWLR%fNm|7%*M;-4_)zmR^|k8ecGD9FNKk|gh0U63ygL#U=mm@R*n)9B
zY0Q#Ys~<;qBH;_j%-|+FAnCb(2<q(9h(}ZjuH^!lV7S4fWqLIM0s^6+GzF~-jzu3t
zP^G~NKY<TFPL-S2IrXxa=f*Xj3&@RnW9xZnL!py!&x)VnJz7ufS*tLMkt<Y-tXZF6
zf6y=pP5@N92mM@k)J}Vo{673nW9(PHP~Q2PcL?o4^p6`K`f7Jv*K~{F?*wKs=8wx}
zc?tPMn4jEvQr=6;dk=QYesOoooAyioBuC{V65}e|o?s-C-I71*r#8~+JN^XP%%Dxd
z&R)f+P2=KsUH}W=Y`cg2qJOA*lRtrU*5`a7E@q64#A~Kf-eJ6zxniO2h(BuQ5E?Lw
z7UkV3|B&?a_oIFQwVP!erQSG@ySaL__hos$<-k^z7&6Cp{=0W;1RI&<?;Z%97$SYp
znc8V;=-A0=utY$12@opdyXg(<HA_9%2i2(<a#PcQR&UgU(8F6t_rvcs>*CR9=G|#P
za9SWZO$gd~z|K;?;N2F&Zn}D$u!DPpusgPU!p{5CdzNXuW?uY#^m4mj670Mq#G?|}
zcRv>0DQtRVQGEnRuRh9z%3!!L+1t|4+dif)Pxd((>ChG8gN!9L_3awX^{lnP!Y+O$
z^yn)RTQJUzV9;y3xy`MAHy8^GsxX}E8*Y&ARqF3uW%sjuHw`9%&2{xYKIHYo(}=dv
zF0n<Hrobi<RUpo7Djt`xb&7C;ZS3R*NgLuLNa=8bG}WS(E&fJfZb3>05G2VLWg{fW
z^98-r79=(sc94-KRXEIryWSlYAShvbE66J65<CEoE&F^yc01fc4$=*d$<vPbES-3C
z)5+%2C2`ozXifJgWs<0Zh*Cd+2lWY-6y;+*u(;ao`!O~k+gG1tB4GMBb6gK}zO57i
z9ppvi3A+q3jcOBvY^u+&;b*0xo)f^7(>V?e?c_oIMKoE;xEqZv08;&BJndJisQ8|y
zAo(IP1{CU~k)*d6$`$vN)0TGW)2qzw*J6^cUV0SsR#IANQB$Cw2|~w0E&4!?!iYT-
zMI;s3v(&O(m_u1v1Qp*d<L1AphrBrx_K-(u)-0oSv~AEC6NwX>=KB@obC`_#Czu!1
zU+x3Ua-W-=0TwmH92|+{)s|Wo#VY5X^b(?2aRVp?1sZaJeq=j=V5KB>7Ls08|Ayse
zRNhM3hB!m9F@=yNvzyxalm{8IJ|g_)W8L)5hvm5-&(C(bgD~pDB3~c#Gw21}Rr<1^
zqLbU|x*}Yl32bWpcCSQQXE7GIHxd2>s~mj$<1f%&$JZE^g?)*SNL|>=Zb|pPhmoP}
zN6=QjM!SA}a<>c#MW?HNIuzRZteJ{xc0T7#N3)6Nv##f3LxM@!2V_n&y*UZ2Im(;%
z3<=@_yaByvhPq`qLr9fnj3zAW%@MqCnR5daD0vp6HoLQWoTy(!y|H^(KVy9~Y>eF_
z;Y_mHUZ+w5yr(Z$Z0I-t6xZC%n(s`rRH#qXyj%H5N)|SyeulXdOn#Qhr<nW{6Oy)&
zm!?HBM<GZ>=e+tHlfT4dm&qQJpJ(!UCZt0IHo>C*D}4L{6B6HmeYD9^F_MH|T5GQQ
z8!Y-oCUguBsfA*}e3EobtI)8rRtdBlQDw<WCP;Tha~5<~hUqkB9!+G$J3N-qC9|N|
zS<vp3&>76>QB7l^-8A4+!FO~72{k;d3}_3mmdGW}BuKyLFoX?Im+h+x%Jg`Tv386x
zR0wHda@kRTkE5ntWtg3^{^no<_>va)neKfSJ00QIGu~cB{Q^HiMNzY;sapQDVtE~R
zd=XQ4xnznrq!s7v{J^gHNRWANi(zG|UmMmk`p=`&{o1e=6|xmbnGj^{_BM=)YDdN5
zKtMl=%1GHID!XrBMt#a|L6qH5usA5Yur{%OemynU%W+@j02Q?W6?GgcDk3pOMJ+;^
zF{)Hl$3;;sLQ$Q-l1GZ_gs$n9!rw`Zpy*GCqFN@y4fBVfsPer?QJwNjyCo>XWuIzq
zzoL2wD%!Xxs&Dv{XhVf<G*(pK5H;%%6jds@)72@cYSYzesH^>o>O7RpvM8!l#KuKY
zC37u;B~A-7{L!c%%x+f3LA5X(2XeF3!@V!d^N|Bv-RvFp%c8E`62i3XkAPu)9yWQt
zY#^Q?KcKu-M0v}H${W`3yP~|!9qZ|Bnt#2<o`o;<7?!9SqI68KmTH*}{Sof=bA0^k
zO#T8BDi50D2~T^1kFPNKX(q7>BFfgaIo>HY{sI#+hPggnNxja}SW!}6;$ut)J>4kC
zcVeSo{VE&9gt0W*ScO97R6|Soud!BUAiT0S#$RB?-(({ElT$lb6Z#t2L=)9UGSY;o
z^@t_}e%Y`2h&B{!FYiS6p_xuis#6SkMy$`o8$KC~gtw?O-$4e-26toO`w?X$0cArJ
znNT<<p=|IOv<qYB*)BXmh^c~tfvCn5)Qp5p#l?asPJ2`vbRG9wEhrms$VDa68noI)
z=mYjs(?T;RA#kY}!5k4~<A~IS@4%BJy@$%i5ww+O^nk4ighr&X^NR1vQ-M8(c_=(l
z4dT9~9Ey<swy|3bTXKgmxu|{#Fz9*0=O(^Of>|(B1iheXrq6jz`Kx}J$y*EH9K@LR
zS6^4)N`Qvh8IP7da!sa(-IA0;Z|R%-3L5=G;%ye0P`o{?SUMowSj)@356O^lqlZE}
zeKG@1p993(<IdX=Z<%57R#<IZauB5CRAIjbSw!cp(TM8Nv<N%uVmINYz!qWmzXpZf
zs0h2U8x?l`2$+1%FCYao_C{jSM)4Gi4G~xqe$FPzh{t2nfOn&0C*ct)`K1+liV7aQ
zH$YN&dAxC;CVqkLh%AEBbz;4=K8dyy*kL0yv*Ph7d!=0nE4)ju%!G(33Ypwh64F$4
z#-pbN&NXPre4ah9)nW8<M0&Y-=Qq5g-YsYg0W3uR<?f`e?ebe_1rHAAH7PTi+Id!5
zr(+MuSqRSZZbiN`g?_@(kih+^?gZuv9}xuDdH)bv#iL_Abnf#In=+#=1?)`q7{vEX
z^;Ulp41{|6SK;ND5WzbGFGtBv!OKAp(9Q62SpLkO_Hs;#mt!LIa%jj+z!K|++@P0(
zf)Xs7ys+nG?A-N2TZ!LTJBaBtmUxWH4|gly#{_O%xD@EBm=_<({$QQ!`9<dXydgIt
zRlmy7{cR?crAwI(!s~EYLcN5iAPp+VAqgCD*b0(}_OXXzfix!*;VP!X6YOV~$tz6u
zn1~S}(wRntHYda=5fR;!ck`?j>*B96Hzb5_>NzZKR0yj#*iUbMJrP{cA~=-5B78;K
ziYy)y!#xSi;U6r4Zzl`&P<?;|rW}SH!%d$t<jauAVZ0M;mg{*J=w-Mezd|u&Z~G;2
z8Y4>K7D@ak$Z+QkWQ{!&k;vHJ(eY91*eVA3QPzoL7!gQ1AQ7a-Ktf%!^RkzL$qF}f
zNXDw0+_@kn`Pi_2Oyu{7$nS!d-y*vc(9U7<OZiwRyO(<<x+TLQ#e`0hG{u0HH{imS
z@o4jTy{LYZC#axP;6xQ?;FZW8B_tAV>TphF^=<DE+yeTUC$18Unhu39r)9rysUSvO
zg!Q6SUegK|wC$;;g3p4$1;AF@(t^N(bA#eplNeQ4?6g3Dgt4oE+0j<Jn8!U~U15ik
z9Al0<9xWo-dys*loJb}ZHdQdSAv^{1n2x7%0#Vyx0HvHe7>`in-AU`eg^!-#=PBDX
zR;JC;dlk#;xZ^7k=H5=|<mmHq*ex7sJ?esZOEEf*+ucRdvl~4N!Lb%FR(6i}aH<cx
ziooSE_A03oTtz@-_>A4jfl~cE*cLtM*60O66VNenf{(V_D)c%~3vpftYUaE=BK~!}
ze-17=2AnWZQwLbWxh<@oN{gs_tH{U&yZR(D>PaTw&qQnndciM%quk4q1tj}9eScWy
zESm`<r8cDf$Yh#9YLi=8eH>FPr7yC>lxCap*Wql&_FRzGtnXlM_DQbv?RcB<Nc4L$
z^qGFqC`FWBUj%1oYG48Pb0MxIx7N<#b}t~=ITM}PguZ|LgNS(q<yzczR(FEefIBf7
z;9`M!P&IPmG}GKFiLhC5Zi4z@34qpt1)yzf5zy<wU%-h77YELMhNU^`*YRa-)tz9j
zA3VoV5SWc1CBB_&x~g6$V2WtVb3NF#v=E+*U$~tP4T9j^QQNQK1E1ih2sC_VO0i^s
z&RD}smMPYYwH^!-V7vymLkHYE*<d_6g%r}->z3X`A;NxB`a@FZ6R$Dcq<0de3l*r3
z0Bv#DU?E`e0R6EHNI``3@`99LT+!GPT{P#VcQc{|X5(CL=QZ)hVi|z|FLm>)Mi=@m
zOsc%-e+3{2z0oNf5$V#Vf?cPP-4U+<=MOYu9C?7&748uhy4i~zA_xys@EiUJ^jm6%
z>_bKZQSyG|bEqFhKnZ;1QlI1AYBzWB{m|?XM<#O__loeLa~~APe!)&~$fWO(M0Pmv
z_F&4sa1CniKR|JiuOp^#*^`R)r^cggov>io!9gA{O#H$j1jWaOIDpm(#TP8Dm7n0-
z!j1-2pe3B&cEdfHSO+?oVf-qQm0!e(3T(a5^?=BF{A%{WZIVcp;W1MQ^%aM0800-O
zgqQnS$bcgADn7*M#hqo(h#U#pbpTd^IK-XzXm>)alFam63ZiCM!eFe6J=cKVhoCbJ
zijO!O(Er7<upJuuc=B*)9LxyV4cCH{_&Q`zm=65l|A2rT(h&X$%l;{oFEHsbp77)f
z8INW+OUzwY{(qfNE~A-&O{^J__=IH0fu@IH81W!W{Udh$kC8y<c$oP|m^{kldzetU
zxSlKeYpDN9%#W<GiekwP@FN|J5yp>EFc0vfOfdE|_;Sm84u`~X98(E1s&lD%@t5pe
z*w0JI^yI%bp?U7rw-EHm2oQWk$TWpB|N681To#KaK3^WA0gK9d2o)7F&&Tc32ecbu
z-w3vZUI6lkMjQqWgZwF1^okFkRg5<q04fQ!oy4zu1j#Cx(}&84mNH(&KOE%YW?u5A
ztcwh%OWcU60J?hu9U;Ebnm-+wRm5Z#usz)KF;-A3p*G*nKg2%=fN=4xv)|&PaWv^f
z=dM9m$0XJth#I~3|70fWH#y|rLW0TtbLJ~djxqTcNd9+k@Mv%vC3AgI*|1FI@b0ib
zSkxmDVIK#*TSqy`-ZS_5HB&pIHGq&1|0O>wT9ORqN%nx9{Vi`TWO>mbWqc%?Z*qJ`
zk%5-P{pQH3gGycmQ6fbt{G`*@0DSB&C4v*t&*64|8cF0T?V39?;7F8_W8gbbe?rBm
z1bHO~PkHRy!x2N6by;zeLdKPW;8%L3{KFM-DI>G>E^>4NOujI(?+c!AK13nK>J{9W
zS&U5!HWNp_ae^ERBKFe1Vuk3OBESyw@%xzk5E6txw6_uVFv<b}GZ#YrHj^(R0RcR~
zys$zsrYd}tM1W29^rr-N2^p%SuzJk1;V*QrsUZMIjL(z-zeAq)&+w`L%#UgvY-~oc
zWPs;Ut>rbSgXP5+$-gB-9Zdf%nWV(d@|gcfP=3ZytBxJ=ywI8YP2Cf$f!F)u(Rw>@
zVg0ur7S2;Igv|+3-ohdmk?CWR7QBLRn?t`Tc)%PUFh#^z0oob$EX0k04JEG9$H*<W
zJhZ9;w73xl*;1F24Ijjf{}OsVk9a~MjZfe~s}G_yh<imgH$KUVFEM$J$;X&*rv&DQ
ztoR^dsejF?bl>LBS5OdS9K8uCkY;F$mI^>J!4K)<C)u-nO*Ud=Kttc<(1Nk>?dXVq
zIeLbrgm|8+Q1(($PWW#+b^~$<ycB<S=AeV)1;AIU{ZLw>7!xWv9sCiS{fim1)|kS9
zN3upj;WK#XQ8;oiUt%hg-RR%tB>x?f=;(-$Munx4#8ffEfr-2XHGmS195|ea9#ZOe
z(8UE6u%my=gk0wx(#O~F>PYnYtYw&EdBw6i?)ds{P)KB|T#qg=tLsom0pr0)4fkmz
zeH3EChUig<1^iDz>(Aj-N)q%z53_ZZag~}rOremGNX-!X2)B0T^4cT$2KpbS0>N5A
zQz02aT_W*~QU*ONI0>!yFN88QW`sl`pA=G%o_q8lgd(4Oj-&b*liy?V?~yEJJjAYE
z6W?#)x{je84@hf#=TYb(Y!4UaU1K{DUHwNSZ-wo15xP+Z?8EkGI6Z9t5A68&kz7!J
z!0fxg_HW?*qtWNHCa^uHSXRd!Uz8NQ2Iaw9>A<;2Zy!>BDQrh@<Ew`I^GNzI7yNu9
z<xcTWU3tJxIBQWdi;#G@RT-fR%w@?GPl;y`P*0aG&Rrqc9<D2-Xpa<9;M2xuVfA7k
z2*+2k@rfJfhak1M&r{7~&j|`P%U7Ya`uZRwch|D8!zCmF-y69Khj4XN=%Jt;V5=A{
zu5Ufu$;ls`Vz_%}TnD!E^gc9)2P>Vi$|)SNV(1|Y7${$B*%htTRkY|%9wa>mlv6@y
z61T8tQM0pMoP*M;wdR%;R0Fm8wN0T5miM&H2f4cB(`ln~d%Pquav!86&t)~)3*`wR
z_Q+4jKBhsO#=VDM6KjivC5n;yJIDn|{Pn3bve}|Vj~HmwHH6N^!>Wx+A(5=eLIg`{
zFb3HOetrH$&QAX&BnhZ=mvS1Y)$OgGE-N62b=lLLeHrg*^$LIX!#pmOFH)|wA0F>i
zwB0bVKg1FBiMIH!1F85-%)8R>{uAc9z;BRh5mQ!BEUDv;ub7_iL{aQx%ZYwR<C6Ma
zo{nN3FzD0p($MXI(TnGUEdMkc|I9@_>Wjae6~1h^Hj)9VX2d^@P138Se#lGG&$f=g
z4FGCI4bp@@VB=+Y@P+okM$UcG%YqN+&?uH+Z17xZua!RQnqBOq^*P?oy%n5Ha4Ss=
zs^#F%2S(w>a*hS`_ae>)RrKyyg@3?prLlu%om7P`X04El86eDb*aGf1(6m$gAxyD{
z+)JZX4IBNJkJOK`R_LK;Jc#4!epJA-_5kZiN=un=z=|CntfigfYR;gVI4rR*hz#Uk
zkcjr%I^+7|F;T}lL!HaPi@#i+bGThDyLhE71w#m@jbMtw`_J=fbSr|zLK)!S`#~N5
z5E+`|`qsneVbO4y!6^Pv*eVYHv;k`6sy~A=i2dGFdv=n{s^1P`+aCMUY>IZ!bDSg>
zz<@wbS~F&6ESyv%_Y#Ze)qh1jfb>WFd1o%4|AJQ^<F|;gAdhjy67+dHe=Io^vp9EL
zjpI>fgs2Gq0loQFO9#%U;CA1-tXW*zX|!-IQ_fiiX`B@TvMGXEvG?E*{)pYY*2~>_
z1fJ`I=!(kYQUY~~E23^?a)QYzCbW|!=v{|$J;BEpm~a{NDW+GLdzA^z3(ZM?jk&Ke
z5f=M(=Dxv%)Gu@FgL;K!<dHbxtH*bi62c`~aw~)Z-uy!a2v{v-4`oyNWT{k|DxEFe
gTY9XR%O5KJh;dWt!==sA@zUwiiPB_gqBQw`0UBxSCIA2c

literal 0
HcmV?d00001

diff --git a/ldm/modules/diffusionmodules/__pycache__/openaimodel.cpython-310.pyc b/ldm/modules/diffusionmodules/__pycache__/openaimodel.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..e61d899fe52dbc60332de7833d87121dfed812d1
GIT binary patch
literal 20651
zcmd^nX>c4@e&2LY&xyg{AP5q?)YL_emIOpnqD1OgX-QFgP}qai+KXN7urb{L2AGSl
zdq`r^Sgj2ytJtQZ*e*Nq?v?>p;)Hf3mBSxgsY+5wIY(8Ja#bqM(VtRrs!~ej^2$ls
z+Lf5!@4fDx9$+A8Yg74>2KvqGclA5|?|;8;kB$~oxZb$`w)4GPsnj3xA$eu+@B(i4
zhL%d%DO+o#n%aV<F|IfC1zrB87YzBES;)%YTr<Cr56cvqc|E12n#F|?nUZdlnxhM&
z&GJIIIkqs?9A6l3PAp6`Cl@B0`xf@eGDc&nxqo55jAt6t%>xSu5YO7VH&P1+?Yv#U
z@65|7yJ(NRk+Mge{OY02^ul4Jl<ZNYjD{&kkW#kCkTQmpqt1~{W8oOm#_b8DO@yT$
zK+2@O4=MY?6cZ^^_I{-74^xgKW!gS~lmlobz3?EO585+$o(W4lWFN8*qhAjr=ZJk2
zIY+~sN04&NegG*Cgei|A#k7wj<#?EK!q%^)Di6NPxvywJ*;;fxVb#2*<E^yq`ARM*
z)>fR_TBluac|oq$Zr!T3>>%5yw;W3ZBi60uYNy?9Fma^kh<3Hvwz~}{7;Uvh(`wXv
zmRE1Lf=RF5bX?ErRGsFcW83xCGF~@%t%Gc<aa~v~cL^07fz+&Xr|x=g4J#jCjPt&P
z`#l4Jow}B~r_J9(5NLNq8A&pD|Lb!r?WQx=YJ1LNyS+A7YunCT#}RJ3Wi{rmU3mP7
zr_Ww^QYBxybnPp14ZAs~I_=Kc^`)h*iw;K#?T*v3>UeS*vz_%|KPSKuUO39C*RC#|
zIe&ni(h(S%n8eSXJr^Z*bCiMX@K_v^j}v$t`igUB53%Gv22%)XxCgF6_~bs1h?82-
z>=d?$wvaC9_K2N!4BNn+!JT!CrTlkNOPZat^KXnUWbA@n1o&jXlQL2ZIRHZ`7<<w4
zoEF#ZOIX^ocJErz#Py<Wbv(y5mqfd1Ub_0ZFI_ZOJg?(koSR#&dn?_=*;>0fr)F%9
z`E!en_TpUAs<-B9je3W(ET>VnHXn?*9jk`buiEuyP>^sTqm!*}v%2C~wp*>OSgn@R
zaD!sI>vg&w^J~Nn(ZzTUcN(|bLExn}Q(M|LfW_0d(w?!E@pWIbwSB-F2U6R*uiNQ;
zsl%yl%}-UcUe0EU;p<b#-!zbxS82Y{Ie{FNRzMr+Ln-7Iy%9gPTHIvq5@E2CnSYl{
z|1P0(88@!~@acEipUP;Eak~J$AYZN4TXnBm4fIwkD16Bh7B(0%bZ;fd0Ql`zFx6?h
zb<Q~^B;F%}pdRFJuHCA3M0+*J*3@qyo}!@i<u80SUaz2FMN3!71cvK0mc%rg5Cl<i
zkiiTB?3qHfT5DLYTdgWkB<zV}2p-1m5>AV`hqWpE6}54#cQAg10*TpZL^TdRy0{uy
z+}x$^+X#GZHT6w>L*Gbm7{0cpaUN}LJH4TK`f3`vhOOCHk-J!<WdN*g+|@TUAG5of
zv5D(bZxYiZTuRTRwz8;^+tv=GP$spN_qA<J#tLz)7{;&x`gQFLYh$gl^Y7lrb-kho
zX#!nt&2c)2+|$Hq^j2^UD<fF76|3U}IpN5NIF95XU9-HvXgXF)Jj7?+YX=$ITgUox
z%EU<qk07XIM1}Ej1c9~~XtjlW6`)Q;*T+*rZuT6I-fc_RXHnK=g_2g*j%W{K()l;7
z^$zTu_^^h=B9%Nrs|8>JXcemwR;mW2YV~H<YJ@qtYSnJnP$Uf`E*Dum&CZR6Fi>+E
zjR+1TB6<wDvV&qjVh`YUiGcDzaC*V`p|0y6>S_H0-OzueXUabuFUxW@6dAa<30}bM
z9zl@AO29){^fywQ8t{+7O%#k?4++F)8tvLz@4CV`t#wmjV)OQjBOEgv{K<;xS!<4K
zMpVU&cZ_RV_%~e#nA$civD|F}mCefLYV1&SF&<GAS5g!W9npKAmqjJM>vS6Hrnllm
zeel}mjXO8Y<@znBW!}ImZw%{zdKVj*m&<)Goq&flYV805xfY*6Z(MFABh}kKFcq`m
znCB~5aRCLyMF!6>c$NVx$tDub#?(Z78oAHmcJm12`Wa<Srq+-!aB)Ijz%7?LxmtF{
zPTR&C8kj9I3)$d6)WMfX%da^%yQCIYqvxqr!|X>%@zkwYp4kDu*<;Z#wVdeMO1)tV
z%&v#>vN0ggPFr|p-IEL3GMzhMP|SJ@)IiPje2uFYT|_~AzJ%N5too_Tsn?Ed7+%UZ
zJZ)1KV?I!kuWtg;+|*?%;^L4@O{-MKH-TDi0xexr1iO+BigmZ%ay_e6bL0#PvareB
zE?z`}WRg7eGRQQnbw?zYMdDcQyf5N*xk!0z@{+oBL+d@T$HYc3pf-JGC>*fe#Jl<+
zz)AQVUr5YM5GZE^({=$sG>z|Z`wsv>K{5;(JNrffKtRSJwUChvLlyv04DzpbT&vk>
zIM$2UQVIxKW>+PG)idQxn76?tm{!YdcNCp7$xXEzT|TL`yx1XW1Q;l?e=&|>!6ksm
zb;kzB-c}reJ5aIfI#o7$(X>o(e+|<nnQemjq7(Hcq$XZz*3l65c-^rF8UaCYFUB<i
zu&7AjbnD9?&a8Fj@+tEy?zS+`U7j^Rw}hWlEYpgS2lv0c60^{*)w%-BMMG&h;<R_W
zZQ_Mai(F?o7(rw^FAj5&xM<PEc+Bj49uIwHha|JAgJ~I&u`~#6s%rRYgkXTgLtd8g
zRiv(_ZOzZv`rQoja%5p78<U<F#M0mlave+Y=Xr4yDUx0U`DjUk66U?nm*t}w4-8J5
z;#Uk#GSHPIgY2ic%HS0SUtmCR6ofJburcpJkd$PwsAaJ$viLJV4vqW0i3q}mGafoB
z#*BxD2QnVEHiO;2s;%j6@$;#lB~t;0ZcP^_z4R(^0Ly?ekiYs2;!`SasQ6MEm@j)5
z%t|Y-Y5)&r>>N|jRz^GxCIl>rWI{+h@@mcGM9MXMms={xQB;EP1JTAH3k>d~<kqZ)
zQ(c0j)fVE*C{@Xc$MGZNbaAEv?Mg5LdhLi#yJ6u+&fz?A?&*DklGyPvc4?rDOF4wR
zo1s0b<?oLhoKIkw@pv9XvI6SVhqimUp{Rvyp&~I9FA}emODNy_GBOr$$8@HgA1?@|
z?ulkx)*e7JN-lw%BGHTzyZdOS5KKxr5qz>^N5kknaut)?f>d($P7>h|)c{aIawIIz
zze~C>Kq$C01er8$_dJ4Kn@Nv0lV_}Av#8BO{uP^tk^*!NSafV0PT)N~7+G`BlSb98
z_naULIl7Lu&h<BuT3fjR$WjGkm#|IZA)U1Yz1yi6iCx2@fxhMr+bTWe$;nn*<%r%p
z95)ru&fc4aqrSu@iU<J40Q~Kk)M?_INENR#kaVO^Oh{#Z2Dju@idyf`ke7y#Rnh0g
z*Vx2Q4Ytz<nx96-9X1v*CuMCsCz!e?(1<I@1;r^|Zr^T&H0N?mb4)u*{r{Zk49vq%
zN_I$cxI%KL_=xP<DZ|MQcFI~>%-DKMv(p<{kfBpM$WRu9$he!02vM}hfUut+!uIkY
zT3J79XYP`aNn!;Vu8&_UX-OpI29!ZX7vDe!6@ei>suYIqzdk*Z7+P<41aSd&>qpZp
z*|-GY{+1wid>glW5J8`CjRW245S>dvzEfInUmsvU!LDSR!%5f{SW@m(?k+M2n1Vjt
zu#m9lw*5XbrJzHy(3?#3C)}{&tE|4~ZoPrrH`zwKTl3>%f~kA%R`&(uCQ*DrFQoBj
z$S`e>4C5<@D5UaVaa<`ETK^f8J`^%QQtLK^ToYQR+M1LY%vyrGgzQG9W!aI;8x$ss
zjxgIxF-tkX^F{2<#T|tuzd5`LB)OOm+F5P4TspiBTeLeMQ>wZEm1hUT#DV@LnXZdu
zNZpQ8t1p3dm3*h!A0DY*G@5}XB2G<Kzb?8?#6`!WAZwdw7Zpqk2~u+rGqwW^G{hn3
z>DXx$gu1tZDL!-O%m9}S(XPJ^{boWIKjMbnwR(ruT25`3IPl{#<nkq<C?Q83Yp~<7
zpE5bRWntNM$gb#Mq~j*xXp=$5TqxDU;3Rf-G$9Kr2*fr@upk4~h>t(IViIGks`FPQ
zN~?`v4@DQ_R3A%J+yLF%g^J1drzPKSKpL2k(m=*1ub=hvM4~{bRB{%!sLF(ti&$wp
z<&F9|mWA3)DdaNSM7d?Z&^hN95gYRh;)*v8RjT3VHkmgu@HE+)@r&Z8ynO>Dl24Q1
zq6=G7+YrxL8~3E_g?j&fNGw^=w8>~?rcy0RDw$0k3fok5I_xb|qW%_E4}cZR-qpN=
zvhT=C_I^{_CW|)1G#?Ux_?~(<(hiYr^E2YNcaJ&BP2?yAu&h(bg`|ggS9}(MS|RZD
zTBmY&2cZc{Ndyt(!d<NdAkv?pJP4b?6rYt~;I7n{yeds7Lb8g*zLG@{{RYOh`qfvJ
zv{k6Ls#3mogKW1$IFfvKkXfnXG05EEkHphK{&O}pHQu@)I~)|*!?0@ZvO-0I)LyFF
zqnt5rO-j|_)~t*qh@_3u!eJo!T1i)z`6z|GCC120LONu>hEUGGja-+6r#MP+FB1D8
z`00F_)UqEGvxW}N5cHJcIJAygP*xpz`iI#Mja**V(xIx>?`MBhER-eXErIe*8@-8$
zbo8k%H&1ji&bhdRTPnVw5Q4~vt)MRf^tSYEKBqUSq|_|f78UiOWdCBU9t;Uk+qBH<
z5aiq96fv1CwfT@tfKIF-ibA!^*;0tA?m%nNGQU1=-Z3xXo}Hb2Eo_encFD24E@<Aa
zYFNa6#U#unYwspV$%cf3rh%Y7;`Jaa(6}WkK~mx5K>;1I4?$5XZ>Vr_g<Sa;kUva-
z_yH%I{Fa`ZqHsSRuTj`kjV(tPabCPc237zIk`fPbnV{s?-h|{X<B;}sWoUq50ops@
z129V&Nau_pP0Hzp4;-?QX)z7U9Ol9Rw5`EZlv&Lp1!f^XyP5~C$sq;5Yv3rshT=`b
z^4w?e{(=w)i&CG!Ul^v@GxR98k@rScOJrk_3Vp&i^o_zs5%?B80X8ihNI~yVgi#{{
z46S-s0LISBl+;!+9>Xa6>t_y7>;x`4g^8)?v-8qykzj=9m|hlZ_z|lN8Vs5Y+6V$e
z8ZDk=G~IAoihq=w_9n|5XGW&B(rv8?fq1mvX`YZ8&V?K>4j|E!5%L3pwz7~T=mAO=
z^5j*Bx>Sq>VNs#JgN7C0wVfbAea(ls-A55*wL==UqVi9l#HP+`Q}?s_q0knU*NZ>{
z$!#6N3})nt!pB#xT)CJ37)1u(DV}H07jXN6=^`?I1Gi-9rPh3k2q*e0|F8TD<0ZjT
zHQWOi<q2NyMFhK8U!!1rVy3qoTYePlTaJ*=i@+z5l`qNsg8gdWCI=@OHZ!s(!b$?n
z#0oZJj#XPR+t5`Cg-O0B>dSTLHNh@Z(H_}%X>K`XE`ph~LbGqEqMtu)gUxmwReOj5
zc%{7xBPJC1t(A7)pssAeXPn3FnzMLr@!S&^o?KjX&RjTi=DhW!b7t|}lNZj{?8l#a
z{Cw@nC(iZjRlOx;H+JW`Eysfq{-LYq&#RX!nCCXQvP(}&^g=A+{$&KaxFm|Ex)_&Q
z(sXM-(0zIu24D!D=I5aY^tJD%urt%&O?}_k)Sx*;zBK=)VHtp4)dNo<O~5oDBMzP?
z3j=aS8IZwSyx<jp4brkqv6YoZ<M7EV#dT0#mKc?aq;N;e-P{|yd1~#{t(cKFdnd@8
z!>Iwv;#;soEH)f-ah+_xw3PQ*cr_<f<pKr7`dg)X;{H;YS@95(zmNiO8Wv`8g4rh#
z1d~ad@>OTqs;w)&AWthaclm&Jm+b}<yZRzu6&xT}({gTCW0<YBg$=vN5N6;AK-{J5
zTg+#%@AFzY4Pt)32!db4{s#k~m(mCIgi%y(jXr6IbQoWZ$8>{6@QAc8fzKsjw>3%D
z_4(@%q6sTPVudbad+EL|u!n3JrrGJi4H_9~1}x&_rs7VJ&7o93+#)W?RY<5}vPAL)
zh8?6y5+;Be$O8Bi*wL4v?*9$#E6_bCZGa>PS%uBda15_A_!a{$s`wcMk%B|w^7Bl6
z3PF&I%+s>UI@9}Tae|MWnddn65+XEGk4p@FFlnF;*CN8lMq+zj0dX8?w2x~ykO7Y?
zrLk5PW`$?V@8^$>8-n=*`~YD$p3<v$fSU$rznoUM8Ej-cI2VB4W9mTZLACuf=qrBr
zk<J=Zsg0Zu8Dtf^7~1>XGHB{kDIL4oPGces>m)i8su*hRE*0=JJ4?O54Jeg6H=tCm
zty$D7K(b{CnFqfY2=FxGP^7h5vWXbxVe<;J73z8gegvr3HWwXAqPE$nuQ?ZE>9}dt
z+*5cDjA5?%9NKSpX0i2_>J8^)T!^eA?Fo%j14+t6_S{MJ3-y3BtiwF#Rt5{!x~|u5
zR@)u-xj&QK@E6VG6P$W5#OhK5<u>{aROCAq>p}fDT*VaVJ%<OlxpV>p9kzpl*Y@Di
zg35}qC|z!MfxHx-4F83U96_L0?*!Vqh(S`kfo2v)h*Lt?Tqrgzuh~^i4MJxyp6Da-
zOzR2(ihwdcQr4!7Lz?&nJnSOY-1NNOdjdg{SWCSWT@>=LchVpMd#96Co`7&#$VuC0
z9^jG}-$vEmikyuXBWw5Vl{$zlRvdtX1;ljrDfX=SiseDy;F=AISO+HC{1ku(eD++s
zi_nx#P$r#Fpq7xY7(gz>L%cbR_ofi+q;sSz(q;@RPfQD7k5b-MksawB)EtJ5tC&RT
zYe!zJcQ^?YiY5%lyTlNrzL#*jzl(q<?;GjR2?&&z-1<~1q_-2&G8d;FA{VTIZ>Ri^
zQcw^xG`7=l7m|JmcXhOtC0A`r_b%z;g}uJ59Zqc&yzCY!a@H@PR*oT@jfx()dB3<-
z0M5_S#RxogadiZ!6*(xMgN7rIawGl--)v|3J}z$Wx8t3qq`kqFj59+44n(vsL!JDr
z`IvbQ_oWJix$Cc)V89?6Qk45L_VyQW(~AAogh+L9ryM5R*iqkFB1Vc4i$W(v1xhWB
zHA^YPOEKX6(qhd!wODgcKfma?Qer@o=Y~l&uToOGR>A<O26%K>#Odh-77iX{H-)D*
zZq|YGLl-_%ZhOgVT6cmBP=e)&TYMpfiF+lb7a$D@^hGzw+^*YbBMTX08NYgMCD3jL
z+FGF95;R{0+1pNidBqDfYbUvR8rk%sbf*z1Wv4-KPAUcf?7$%~1rPxa4FSlI_6*Pq
z?hZ0Jgva*?oF3#w5K?^;YpiyMLJ!y^e-7E==NSAvgB@sQB!OnS+#K9|CB1iSNKX<)
z?-f2;7=bm|1LdeGpIh<Q_~t!DlW!w;fiuIAkX)9GyrEA{Y7!CEkT`I0@Vvxb1i)70
z@p@nS<k(8WsDubb*1&!)EkQH@@seDe0>3@6l6af%rp8kIc?N%h!4Sspb&N>-A~JjX
zlXJN@f+?UZ7u5aN2!4D714jVZq=*U<$RxdGc(}<pgvbO8LWE4Va($FU^q`Q-&QytP
zpfcD4*|0r{Y_>{+t%b-3ssERd4-pR$&<^DB{}lO%U&7i*oFjf2vC3!y)!bp$uQ2#k
z2FDrvMFfFytM1$ue~Hh(#vn#7Lh`iVW!m7jl%5Hj+!EwN6tbY-D$Q@x_%l8@lsQI(
zF*N{MABQT$U*W4H=>8^Se|5;4xmACapnDII-ob(P69_>3)?nQ}^DDO_yFRd+r?8uu
zvuigKjq!3<5G45(X;(7}nL@UZ!(U$hF~+|FQnRqRF~qYyDY+Y-VW2mcJh=cjOnF`b
z=P35cQwm8QXDTSXI`4R|kkz!l3(_vp$8~r_Ep;0WlX;2{4IR8<-2~~|ZpEWQ86nAp
zCMhyj?cwYdsz3*7NuDuLaS{56NLCJ6@`z=ZMF%QVJ+=2ckxR}?@uvgz$csXmRh8Oa
z_hQ22hSY>5(k0{9_e2O4CE<8uy3M9S*Eh6fq1P~;Nz<1!0Y-idj!CN^{M0;nz=Ela
zS&q0NOga{U?yzR3VbvV+m+KTOEVyBMryy9k2r0mH;h?8RM0Vi{lAJrzYwwiuaD##Y
z$~+J5`mXtea%bavpSaT>4A@YxEzttUosnG#C7FRa8?bMN4<_V-MQgDRE@OS5aJVQ{
zF!`{EhEFD+>U7dmVwhB4HJn>cgY!hUJi^~#JJ3jSdga)5Sl!^FD7}TKwTz};igyV&
zJ%DXIvfXC6a3OyZn<ODgI^R23`lKz_JW1!}7Gwag5-vbwh{SlqS&sWAn>Za?8ej>R
z4wg3(&EGJ?_Y6$Xpxx=>N0~w)p$f<D*5Cu}G}~ewWg+<vw8qy-L1m{AJcg4{j)7P#
zSFfKgZBN4+OKj2P3;|S^p|)~j&6n~8rC%-$(HL1gJp}S0mt`*1??B*=;7c;8a?AGn
z#o_GHtq@L}i{|aNSaa2s$5TLw9=i^Eug@nT6P^bj*DV0ElJ<zW`oK7AUUoXd!P$cy
zPO4nPowh=?l9qX;e)W~pR48yiD>D?$+D(UJuDi{Fp<$y)j7BS=v{IvGtoa~DQ=OXI
zb2qRzLv(ymhx(JzpJQmLA;`4k=Ih|Fmh>QRtbg1*$rY`)YQh1|w@qhhsa}H#zZRkr
zsgSrAs-DOXh%X{|m#Tp4(wq*f`Y9$$=UU3d(e)`r$XKLtyB$!ugfKk?r}1e!)z#cF
zsE46ef=XCEi;6$x&LB-^8ac{_-b`;C#pxr%pN4YxDAddqsEZNbnD)o*44&uw^2WGd
zmhor&G5n_8XV6Agw##Q-ye#w6?#pqy_Lc?{zK$^GrxE6T17X2~dB_`q1H0vwe3+{I
z%rXRK<czk)kV1pGOot~{(<|FK`Hl?p{TYr(d<*$AaVg#X76tOR;1rC!-||Xt;gAMG
zZFLl8Fx|Z?{j$cq@$aO*qb=$3Ol01hP~|q^Ucafizb?y6+Qqk0P@|9dS%jtUqbC~F
zaCq}RKZkJ2&m-LL7Z6VSBM1-pcu)19KPvvAH?w-Er6W9ylp}r_;Zc7K;W2+4;RF5w
z_75==P5`4&=^gWqZ$0Qe1jmPCn>ykT!=>RN{}`XPIa`ll1hURMni%(re3G;FLzI2g
zAMuZEb7oJp_RDhRH67(fy-9nFy+)an_IO+-?Ugs_opyYy;yvbl3d#eXYMKDwG<g@J
zq(TvGosz3`8hTo7;{iXtVcO#x$Gusu5e_8$9*lqwZam~oaAy4nXE4i1-Pm~8oAXWd
zJLh98P+=7Pl0RYZy9+ht#v>|i!}O2)*;^^}a9ctT`}z&kDnohlXe*1@7-D6AY~xY?
zAo5!N6vBoo(bAaiJ<c^p>K%WEZ{Il4((S2@6W-%oZ>0ZO|KYb%%vn?Kd1UJhTXBEc
zJBxNoe%XK2-)HZ?n`w=B=h!ly_xscS3IB*+vZwFnw=r)R+o`QH-edkFZ{xfjpP_a=
z?VD2FdfI;gZJJDfN4t5!e^9}xJ$^8?mPQZ$$Tv4m#w|=D<qzVyIT_B&0sG+Hd@JWY
zfgTi)cG8}?3zG`l^iN`*;y3um?L!w*8^;HGfcMVEGx&j;pCL2IBO_Dq=weir2xsuI
zxHoEz4#STTrQAn8W`+(&Gqk(@CL#67!9EhomN9$CIqzi@{J?CC{b^_3g!4L}=683(
zl0AMv!PC8s3g+Oc?bL?q?PCX061LL*L%U#0rp2(Oi{B4tMi)Q#5i|2WIWxaDFf$6a
zbn%-*VM`ak<5yzX;(HWq>Fz%YXKwtHz?T0YR^)xO6z(7FSUv-0WP6i%`fu?J_u=J;
zgqNq$-xyvbv;aPiNcdPizc<9bqu;b}U&op}=w0wlU`B;$E=nu_<6s{PNUZY=dS||q
z7UvT+h%eH>U*nsG#5vDWNh=M2N9TKAxq?G0$B&y|Yj;T)mtpS%M*(uz=-SGUF+M9o
z_IJJA6=qGeUH5dKuqJ6fd`Nm<8`>tRJh|t{TI}_YU`vwxNt%|#7T!<AtB^4+wOdQ|
zWl3Cn=U#&QJs9>rUj(KmdWbnYa|PV4GJ$|+>b(xR;1*owWkE0?p`c|ts;!taHcu|&
zSg|x%W2iij<y9!($a8;`cIKp{W1fT<c3Mgb6~(2=;o__$6z^0>vO-ei*mY`~oB)qp
z%d-?%NvI8HQh;29YnkX+7|zUdVEz8m8?27g08>nkS0x;b(vV911<h(f{@I4rT(qs{
ze?%n!)rpg--ovr99qYelpM{UB)o`CjL{;cLg$c(QgDy{y5EkM}VBTsDjwsr4B6s%S
zx!yCg=5=~Es#92U5w5;;?Ha_gdRssy16V)=>DR;1Vo9mj0|P$m%pz<_F90yW@?hq$
z!G68|n)$r>xO5^o``BZ3X}+dyERr9SGH^nyRDzD{QLd%M8-gX^V^(r<_#6R5)OxGi
z?z+7bNJe}a{4&H`p0Y7(Rm*PnPlKw1XOh`d>haHFis8BR1^gsj)Rysh1<6ZHew4w_
zGT20bqp2C&=|F!C?JtahcjOuLV8X7u7VKD69<i1tr=UnCSam}-bR9<nhrCC)Y*kO3
zh9a8^3-l{|rR)K_vq$A`GV8g7F=q-NJAi#2n{Sq|&j@zm-(*`<odpI{adfc>u}z&!
z!J(pf3kKyuRf_8HCDwk9!EFYAi|rJ@YBjpfOG30okjCZ<(oL%)&#o`SuLrP$&qM4V
zy0X$eFeoLrS}=-!_GR=y+WA6BRZYQhzB4B+3}?=Vk#lFIouR)#LE=jRkz9sz^blC<
zHRyPtui_UVs&(6yUX+4bGwJy#>T*cbO{uf?LE)w7vlKz0VJ+em58Ro8>{a!diz0Mi
z_^1WZoIE<I94|}KR2S}S9HtIE^wcaD9Bc`id>!d0Bu4ngERGw(4M*Cv<>^szgUO|4
z-L?TsQVt77l?p0qH{HSEgqKlFI({lQj5LY_rMiue0m!koMdecpwfsF6{QC_40Rstt
zTtD%T88{6734?#iKtkL<V~k)I$g_M)oipeAGc5i&a|w^Z81O(+P?Wnvd3S^~q?({1
z!9?PrDmRdzEs(}X%&fyth)DBiaA4<7q9h2^CVq#-4@SA+6!o{;On<xK;|j@lOXv^E
z{neFUm6%AZmV|#-{5s#D+pttgG69<xbWN~b3v`Z0cikWlax0Y^!ARH<yXsm&R_PE0
zomUsqP|GQs#^1rCIyQd{vA;`{{+|#OA28r9k=Nmoia)ru;HYn2XUe#im0qrNZ<&N=
zD?C|oPvM@{_zeJEE8fp$lwWM<4IBE%$~F*3jSouM6540WY|p^y{Bd}~!WZ^_Hpc^$
zd^>z)b>&u}d}npd_*3J4endIiu{HSG!nF>E1oeY>(sSAOBX>7l%l=T$^7tTXePI0Q
z_$cjflX#YGjb#6KzhwB1Kha0>u+-^)Y}_{{i|85Lc)ndk`_Y?r)-c}J%X__1&m>D5
z?`QvO`DjT$hB?Y(RBAjw`KYExIaB)3(UeDwKhBpqi^lI6&vPZFFz@*cX8^~yKZFnK
zWK4f!qCJ$MAZ~r{Bm&sC;eqPG?-arce8m!RBZ}Mdpil4FSSfM|XcM2Gk(?XozT1wJ
zpc0LfnF>*+Qt_C$=0QpR72X$Qc-DDcexOGD8zhK-ivVA~s4zZgQVGV0pZT2|e&7Ze
z;Xi^kc`u8CC4;cmo9T~sN69@0E&e;I<NFFf_F&GP-8~rbKiFxyJoJYm{wE(lav<l<
z{uc~H`~U^_9KyM?HF_6C7x6ZqiAmi*K>#LS>(Yk-il8o4BBPP|XPMr?bj{*9V;UFJ
z^fZDe5<Xe7fwLE?hIqm^@Y$1$Ja;UA(Z*@%1%Wbf^fKfJPVjLPcqgU!xswh066PW3
zjlyfV2*Cx%;x^Mr&2AO^0^#T8bEp~Sz?8`HBPhQQ{tYFl7g(~qox%rJV1C30Rw(+!
zt=xRoFGL?!ku}G}ZL~JdmT^uZwKbt)mAv(3A~V7iufF?x40}>@P3y@bJzV;$7OKoY
z+P_`7EA6Tj*^i`=<jm>Is%;7O;Ri#K4iw?%0~DhM8$we3*=cl_m%$Z-1HDy;!G&TI
z7?xH%qOhCmp%2Gvp%IAhk}Lr{HlT{$rFBnqd8hxz5I25B+!KAC?>*auBNhZP9_U&(
z*ZC6peS;QUauKK4qNbPz8h*1Usq?=Ad1FL49h?rRwoyZQWXTgMbO5PL3Q6E;FIOzV
zzhvxj2EWgsZ(k*n5P!fR#tzCZdzNYcj={fR5HR@n2$GKB@bik6LtJKopGOd6*YI(h
zWjD}QKw~`;e(69i=wTRSF(@!qJ9ezJ=KcV?)SX44YykxBzOEY|8focqs2>C=DS?oX
zl$JrR=o(T28V2do@s)tcP8m64R;Ja2{0%e~DS$%_uv7yxfWP<^DI(0zU!R8)IgWJC
z&*R|e1?27g+6oD~_#+1IF?gE6^9=fUr*GZ;>&W;XxqxR8k*<x8Bq`ZP$IE}1dOh`e
z{;_<1GB4Ot<s=yI#hUuy%da3I$biM{tcy2g9OM9e$Z^K=GCtdCNw#-kU-US;)NM%<
zqtytC{kXWsO2_dXv{|(z?(EL*pv~eN9S!9pOB=LeqDc=VKdLxWf>^LW*-9)9NM1L{
zGfy@|q^bxJQFxGjMM1)TCV!fN<kUaInB;tanz0)U78z_X_yq>vW^jT5XDG1<aycfD
z1`%5g>Uk`L0;*DqJXjuJGpRkSmkQvVjC^`BGkI*PG&wPOY;tt`=%kxJGdVSBPCxj6
E0570Y00000

literal 0
HcmV?d00001

diff --git a/ldm/modules/diffusionmodules/__pycache__/openaimodel.cpython-37.pyc b/ldm/modules/diffusionmodules/__pycache__/openaimodel.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..120994ffbd313aad79a6401c12f6870d39ab9534
GIT binary patch
literal 20785
zcmd^nTWlOxnqF0P^_9&gMNuS0$#O}SZBdp;OR{BKV|#2%R@O5z*JEouJDl;fnq5V*
z$-e1SwM6w6<IK=rJDZUsB%5r2nF+|vK4h~=uvr91u(@K9$Gqk8P)|V;AaIa}U<2&K
zVz9~g{inLRo1`pHyg>q_$U1%M)VZJk{P$CDjE)uzTst$bdOy8m82^b6y(^8#Yq<Se
zreU~-Yu1g1xn!D*TXk#6lHb&lEx+ldjQnOBxusm3C*R0fhG{ekOCyq!suvrhOQVg_
zQmHYvG}ahj8gCq0I@Fk0nrIweIxKnY`eftC(h-TL>r;*CrD?=7ZuTQ%X~xaDdHf#z
zbB0@RM?Ny#5ihqsyPaA(hLoZ^ij>hf<v3DG?if<Wkn)&!eA`|+fwXb=5Yi6CxgJN#
zgnJk%hvO6nDU<FIq#TJ;P9kN>okq$uYDq1f!t;!K6wgQF98b8j?lJW1G~OI{AH$o+
z;x|ts<%Ih<QXY>}o<fS_o<z#YIOS>Ade4|U^*QH#&WuWxWj|1rYS8e4wU)a$myHV5
zHLtqSZq=GWl&!X!cgjsS%G7I3ucD%n%AM76yVa^Qair_1R=Lq~J9RG_ZMIaSQm=I@
zL9Nw{CW2bS^8>G4_8QBc>(-j9Xl~-YHeOfiH&vzSub_aZkecx}YkuHYvGQ{9m-kiN
zzKy^&&}Vb;ci3ZdQx%aU!JqtKVXf8h7MiWVTW+;B7TiiuSy)xAPJ7|KPV)vD#bBBX
z)t2imT&u0Dyz$n%3w5`#p!@7ExJc{x=x;Bf)%Kc|8lJrReEVK>gyZ*A5D&5(w8|?N
zE>E*x76QXmV+bGFap~-wsYa2x-+NWU<AF_Fx;Tqhy(Y%l1n%U@#4-OeBA&5ix(4=#
zxs-Bqo~{4f)KBeCj1|+h-SkJJOX(P<GoKi?v6RJZ=A*IKgTQNYvEIRYUUa+f6&zfz
zyOnm}xz37eHJqzA-g@VXvlawx|H{I`YAsmnEYDY4jRieV3w*z@TyHHeG%7XBbG_E)
z+{)xhZHseuG~%}_RV;ehtu>;&fP+LQnw>^@&8xV6xw=+qHodwZ6<VF3-3j=<N|eBh
zzggTV+<qHDU~C&Z<}P3)uy#^`y^{{D&~(kiKo!%*t`%Bt>acOl*fm3=oC&fnQ|!>1
z#QSX<X*r!1+U-TW(P??qk(xE|wh)Yj#(H6!rHcZ3>BY~v+@BLDS8?O|4`2VB{h1q$
z(tfAysVG-2*P6AUT#l?}Gs?eHQ5Ebdys(0`C=Gabo6%&u<<~gln2cmwC_;FYyS;Iz
z+*YmiC{xwH`DA*c;<vu}ezICozS3K`xpZXvUVTMPpbB-E!6btt2(VT1<#M%N@%?gH
z10CT_5wcF>_PL9UY{5KjPU5d%c4v|%G(gPvMpPA8M{ceJfloH(XNZL6y72>R%i2n9
z*`c{(at>W{H??I3)_MwWZP#?M8uzh2s{m2gzHe=rA!c?x?Gm*c9}}@7TuB`@b}}fD
z-8H8T<S}+~p}A{HET6;*aSR~eFYBgR8bF;{{QQ67x;bY>DFRz|!}HpRVC$VhZxz>X
zZUif~R%v@tR(TRpk0CiqRVzVcH@r$yJ<ew<Xhmr^xQDfM@T@pz>Lh}>w0e?pZsy2b
zj?C&(t_%<-Ufa95v%qn8E6Tlyygv7_VHVAjc^s2okiYKq-bs&30Nu^y2vkjg5rC>t
z?%|+vR4kWocPjPxO}1QiTUBI90b|SZDmE95#;afT>h&H3aLq9pPvJ+fk2-`HhwF1$
zb16GJo0qg9d_zEgjhT)k7{oW8&Bgu5*fxP-Y%D-J8oe3g{Wt2Z>PGjb#wpEvj>fRg
z-8D~nPCUpnb52m%@O-C7ESzLl_)Z1?j_&~jyG})|b{ard^HN=v9pc5`VM1Y|6IS;d
zl1;E#yIsHM1Z!Te{h;OC+Pvkg*6w&s=N6i~bx`BF9agZYad&E9%~L3~XQ5@r-#~9P
z&bGTphNfXY9`E8@MqNTK^&Ere84!gl?k}lgpjshvvSVMu+gET4Da0CCiYZmx$;F9y
z4Y$wTJ-9+{+D*CkM<)0t()~;{-Rsz!B+Ku4w>zW(m3lYOD}@>FB_&g~RtcOo@W?|J
z3RB7nuCCSUuEMMa$S;)vF}7POaB6`pX4COD!D~3RCNR97;l(PEA1|U7#;@Y`uOrxe
zEi|qfAH29_2S#WI=C-Ar5J)Dpwt+%!XCxJIby-qVI+gKlAdTCZ?*KVr#Mnlus8I83
zO+TnKtDelAA|E=qkJS|<2)7fUk5Rf_x#y{Y#gKL2uKH!%KA#NPzD0yZt7vu~f5^1<
zR-XMxAGr3cb)cpGvVRRP-omYy-Nv#rWoX%Zljonso1taT3-Jfp|0s{8&$>A<jA<eM
z8B9tc%6+@-R~qfQS9u*9NH2ZU>F7kTO^(cfa~BlfsWhEdTT?BE{7|dj;gep$D{T@w
zEWRe-SCSZ(S-?V88(SH>SkrcYs9?wU%B=K?Q*pq;)g6~avjI|yPSjSAI?$w3Lq*up
zHP0QY1SsCWl9U9Pq98%iudRaAveboZ=bVeUTgtg~ZQgln1wZGQr<tG%?smBbW}#KB
zb`+}X4W;R+^TFMgg9g1O`NnuKY6=UmsK^*riAj0!H(^@{(S5ds1h8y@{g{$i3S`qL
z+hGbJSQ+(3kYRissp}x8VcNCsr|~W;%t&fc&7d~W_iVeO8S9)niWI^9QLeWXQ4#aq
z=d5zQnT~8um*zTbPA-s=5VY4>2lt=4#(?Ndy@sGiF#4cLEJHv{c0vaGAcH^K>>lbt
z)_(SZXpXbOHv9kq*$3A=3Z`q_+_3z~!1@tcgOqD+Sn8d?UQc6wm<QxteyyX3PwKd>
z<0~oPxXgX99L;fE0vIptW|@MzGO7($BHLpd0Q@<<PBMwIZl7}-MOn%>kXSIUHd$Ek
z5JkUQse9!Wh(|4@ehInea*A6`ajK%y>p+Q$Tj5Qmkb$!nnb)Hc&|y!tTlETlWIo?y
zQL9fv2KIY`S?b6Wvk2tIY;!&}ju{1CHlECAk2gpSz}sB*K(xXo(j?_Y#$@dn^cGXv
z{W=nsaBJ$HwXITKF^y}RTn|SwxkSPslF?FgAIapS2@&(a^LlP?2;C=dV49ndFD`Bl
zq8lO_fGfz1K=k70qz4IcU>68DDcmCT?AuCKZz~B1#9N2FD7Fqcey9tu^w>Jws94&O
z4G+Djm;GATi!u;_YgqAYe;dgP>n(5$C8Dvb*eJ<xF1nG`Y0udM`-ZzEvNrtQ{0?uH
zE<2t>K(c{FwYtZW%96P`c*Aha1R|d_3!n_(-;D{Isoq7}qX6rNEX=KEnBCa}TI~Uq
zwWt<Z-B*C8w^4o*w}vOca4D;p_9fxzQADe7t#!8<ldWqB*>c=o>i_3-YiJ(63gsfz
zBCyJKBk1niZU<8?Y`2Y+y69R>NCjINP_p;;P_hhYnSDP)TGn)corS^y^REQCADKZO
zR4)@|-1L1?LQU;5n%e1on-U5u@pvPvchEger7T6}OcLe_2kvBUBTd@`<6Et^`UaDT
zS^)N6NCb7gvFJ3sxj#d3pLiwS+(HS*0sWRa3mo1(+y~o7Z1EGQ<6xS&17TXWJ$5O%
z4B(!G5a~gDo2ieWF+W2dVJ8hp{N0Ixe)W*RpA}`fiIzTQYn<C0Da)8*+7KonobmF>
z0}rvUBL}gsZr&|`jqU5Oic#)e&lf?n@?U`AVhILRyyikacA%K5ZiqPIR0o(?2!!O3
zRz2Ynp=DWyh~HXCIM*T8vd2wb*^^n=+JlQgJWSZGy~Xwmx(C<cs#Y8HP#0HVB-p_?
zaj1WTJl+-Zy?)!P)>gnF3w!JIhez_-8_iG^J;qRq-&7s1$E+u#AO}5AFA6vnB#6o*
z;h%^2bVyrJ?6K3x2(@q>Q+#3b!VohLxv{?v{c2)KqQ`Xm8?`ozHNEOSspA*r`=yas
zVv$iNn(k!m=Nyi1RaM*?1Y>ltrwb=>X;BWqTxea!@FcePR3RBP1Y&ay@H#`qh_k=A
z;2>7lMHjCN@~e)><}nQZQW=%P?bksU_n~Ww7by`hYzQ5vMCeF|6e%)cj;I!Bms-*M
z4t1Xppb;zX8o_7?#u;cj9c%H@yF|>TFyCGY3y6({dG+I99J*IK%x?4Dp`oXV=Ebm}
z{(f+HD97N_1eo^x&g3p+f0o8QA-&M;AAw+$QIU(DuBRuBCOEnE^tJ^JuTh?gd&`tw
zfAj0p;Q2E5&0t3Qj(3B--!^y23m#=!2!Z2)sb18rBXzd<Y?wyM`};>NZzl%QwZ?c>
zc*<-{pad(MLW=bI&{ol&JGO@sMa4n%5M|>%ty#^hSiz_?44Kg+pS2|7uhmw9G7T+C
zxY9yjf1^mSBm2GDjc;qwE?;YwMTqyKOs7pK64pLSua)r_r8oH_I62C_<x&Y0+*9Pb
zqXK&v7tLPNh)5Avi*~;EW{ew?qI<kC=SBty=&vBR`Yr?V#aamZ9%Gb|6h#hUT_K91
zUk9_yzmH5l$+I!4Wy_w7nXzo4ursLwVkMK57NIpR>6TDk+sujVWSPh9B2=y!q)w%}
zhkE3qPlvf_c=4B`=Msn(4f-WWr2v0x1>m-0?eaObO)aNcfz44<8!8N5Pjrki!D%@a
z=O*+6Ep?8V%N3PBL>{1z$^qI&y{lPMjG8u~3~4&wTXZ&^tGMUq=Rb(+BSKyADnSP{
zZeKAhTfg8Slns{NPk;t1QXZ-XLi&Q{K}Mjy8z=-;gp&tdb<9LW1<?snkK@Z)&kbl#
zMbp+7N3|B3^#6}I@gi$xEy%pdO2w5{*#<9SyJVsM7!Oz$LtW=$!6smS3>AUIp;oXA
zu}gq;f=UpWV8;s6uqZGl>=YCPb_ndTm2Lu?!FoY^1K+|{v}?kulwQvuC9~-Sz>w=X
z;F&Da@VjBCAAwOtTU(jiE>w-idS3l|gauJm@E3+^_62>&ZsmfJ^&%Ntq(W8keQPVf
zRRC^9+rXUpX#*;a0<0csU}N2%31H`pq!>Geq}QYDVVIt#EDOBzG)6yX%`b{wqt6;+
zKb_&5k}Bgz{eZzO1{DTX1d%OvjRi(ib+4&;N!g$t^N2wtU0v%mH`EFedV8BxMiy`>
z3+w_I3?xFzADL@QSpps4V<|_jglI~~m=Q)7DpP2C0S?;>0BflDPjUOF5g6F6v)I65
zA+ZYPggGT!cvc&#EWR1s%01pSyLeq=-s{(|Ls4^(qJ0*%Ypn=dR9~PPc^~h83%5Us
zAXl=pFf=58rRQ@+#T=Z><PzY;)Bip@TS|;UnyD@I(9G+=96h7oAX^K@s&$vVnXqC`
z&%6kO3UCk$xQuy~>YCGnvQufC@yn{VT7zN~j4<`{J!3B|Gv}OTu!)t}1{`bl7tgz3
zpnXr5KEN8h)>?;66MFyVTB~nU*M{K>-m`AiTfVe>>A9DlUtabuymaBh<;wHkh2=}n
zzjV3kKKsJ6m#fb|cd1{l9IR*~vcJ%2dI9Y4Pu#eCSvRhsoPd9QNYR>ThkhHc_VGUy
zOldw4KB##XNIf+LTQCIA*8<xGF8GOoeVF=*@iTkdgbEVx#S)wXet=3ALM$){V%Z=E
zUIg#m3~b41u_fojwE7DH@CS$k^Nci$L9zc7j3#A}UviXsd;^Ced$&7w``pI4I|+O4
zbe|^64Yvkpn(x4Pv0V3@<$Gl2#bDlN%=LWCprX04J)DL*#p*@C6@sf71YsTssZk0R
zXY~a0aMMK-gEr<H-fE?KPZRwdt<v1qLwaIX9v#}(A;I_2G+CCWcek7XZMmgf`g*9R
z*x<1Id>l|NrbQ!1N}n@xsS-B5D9W(qAv>mu)_|c@_C}wATd0v2e-pw$alA*tSFkLk
z->xat+giK{f<$tn1PRm-dz9e(61yl3ncb=3P1>{YBt~*_N=ZK`^HQQe*htQP;TB?)
zkRFmKu;U;<LVN&jAOzq<U|I(!V*gF+GopX!0R;VLH12+e!%+-wGvJD<06|Z!A(;DF
zrd~i0WqbB&DbirNtVw@iI5n?u)J)P=>v6%z`xl~zf9-+I`PC6{8_mKoOWPQ7X*;da
zZ)rMbE51t@0>X7N-#738mktnrnRB=$Y^4J@BVdj(uORgx+F=Uh7Qcr{cI`<J=n&$^
zx*eq9fRJ4UiG2aUR0d+eJlmBs=tQhpsJ1#(?bCcM>VR8NTDNaOX<gl@P@MpQmO11d
zJY*oZ(?mo`w$hX;lKlXSTKrn$wYTASfO0Ko*`omJI`!IycO?;u8<nbm4(-6y<~y&T
z{#JV)+j6B=_s%4l$U)L_Q9n17qz!PdoYB9~3W$Lo#y@{<IAgWr2dzfA)%IWc)j?!%
z#Tk5p`wxa#Ua2GBR=<KdX;X6|+<fpP(~$WB9^ewwj%;+;UCIZo0NyVsthtQhwN?k%
zO|#4Jcu30#BCEU^nfFuzvFa|WSsEd}i9vU%(5M8Bj;?ALc%$)wJ`x8tuM?06D8|UR
zS+XZh_2&^EqTWOZy!_`73@W2U6-76P9E==vr-CKb%V={493Qe`Rm}lZa>_^1?wZWR
z>pkoF-L)F1FIF2sgT=#C_DT7?^L8bGiotj4f^df>+Ia?m16F$h-fW=B;wa_NVFj89
zX(GWmL`=Ln1mRcq5<nq1w1oFb1ngbf_o|2bK1AVz@pGAzN0LGPX7|`2Flail9~W~y
zNRqGO_Wv;g;=u2xVn-p`G=b43jhG0Z61!iLI!i{_#Eu1SWG?888QQz4b+GI7XSi>n
zt_&G#H^p5mI~V(V*F0uy<%8@F$#f>nqg0k59Fht_c0C^!b_$>a8P>2f5*F5rz_oaT
z{8=b8a>zFlj<97n!}hqS-QP`ilNR)bS1@@Q%Wfd5eevncFF9wOOSrGjL5{omfdh^V
zQX!?dZ(%=w6F2SR?+-{*SN4ivQpcY1l@(&41es_Q!tJk(npiJGaB(aNbc@T?;M{W6
zKmY1-;0xVAQsBo))nIN^?^VKp<`_s%z^lX5PNyiac5pPitpdGgGakGtI&iP@TPs1M
zvKgg;4Jv_JXK@M@b49HSAa#kXWj{*at+}Wp16gAgzgBfEGH*xbMr7VmO=im6^=hkY
zL1b3;5~7QEO$SQ>Y$H>IFi<kjkS{PNtXagnvj;$AOp_8K!L30M7Jwvl29Vi6TCHs^
zGPZ`s2-qRNikB*6u*F~xW*Hg4EFJbrm;$SN;($I3WSz@@u{Z)#a{yvd)n2&jFSF$@
z5Kf3^me@K`8a;xFc5XtjPZbHt#X<A>Dk$v<l<&Ut<x!ddw7`v~dthK!Rv=h_x(N%Y
zVf8DaE+wccHg!B4ZGFfIO3>Dk!38=PVdXJYfscQS$S;nt=&G>~Il(to{8RhLDYKL9
zAt!h^CMb^RMkA}ese-&R4@F+APmtHn=x}W@GDGU05i%npBSPDQjQ*cPW~IHztd8|m
zKV<MD2Gnk-Z3K~hr{>*NAMyEP1_=UF#D9ZWO#K?)f5PBX27i&kuQT{d41Uajh)msQ
zK#SlW%pqEbZGt3)wH_X$=Pm7oQ|2si*$J~dIRvKvhp?F7m)~SxV(c|D)aoyQx6}iU
z^ik+aa0r_lKv$C@dR93Xy^CxkFaKo(gB+MUAv_olVbIY;j$XjwR1RF=Ajn}kW--W#
zO-A`|FM7e-WQHq00V5*#>n1$HRyy^%!*`lfjvd?*)e5QcehsO{Suuf!;;Uyxdx*1J
ztg9U=C0y-5!DXmZdg6X8H1vcD$vRY++O&taiMQkrH3vOZj@&YIZ)H(S!$A(uv8fZO
z113_R>}lVPBG&W7<4w2>Ru#Mcp{@#g4Xd8m<is}FbHeZ(nk(U_7r@sStZ%{~B^6=P
zu?Tehb=vhx)gu>tkAhDH&ZogSh(10-idY>uDC!ZBxw(!cZ&SP$&uI@o=t!V@3~=<o
zcb?Pkd~El*&HiA(EQ87FEpXBq=|U`P4b9n*;Wd76Ah|48mTO>!?hR#*7o`j_1g6^f
z$>CFtPKsJk6zyBxyW`b4PjpKp{0-Lw1uCak#<s^C2p3K9m!gujH~k`j4!9`-Y?G1g
zw-?3>`BjXigd{mi;9=<pjm*v&x@k8dsRVQJ0`x4T7;ikwN#CT3^NC3WCVFu&y){t%
zEhlbgXo7}KVpqP%Mhh8Qd2Xi)k8iKhQumM-viVSLY)-_fz1H~zPGV^bg1f9<KV8gX
z2Uj+*MF(dHpt=gBo0ll>v@b0Ek!c~v$dc(HkRNbWoR!)pR02Kt5?)%iY`<R|&O^Et
z!-;prx!Y13zMk@A3aE3yuEXBzGirDVJn^V*0hqN+O2pL%#(C$O*H#`*Q|xil<r?la
zW9pTxIM-`8-ab#22lunKm(kwd@Hpn0-xwMiHi}?08ZgCtj+(LN!x&9>YT==~fxS6E
zL)q(4e=_=W3?)0nuBL3hHcp(0-+cYvqvlCgwAQRD4>;d-yp@$&6}|=47?p_1<ac7-
z8A+hJj^J~W%9~>6hq+%A#WXNi$-D3(e^SDo!tMVCDBOU^KdDZHQ?9X*_fG^+g^5O6
zp4GeIq<;}<7SqV9)>938>oJ_8N{3U>O+N-L`Mc0|Bfd2ij=O0*{~#=FjfW+P-wDU?
zoAU3Vj*QgHXG?u1?^FI~NxJ!|39kYRVJ=J|%!f8YdKaXFBAoXB5Y}|q(Zcj9M0>m`
zHBTUgmV8Nv8(9>Lxmjr=&G+F^j!69+?~f+AEdS?J9ej$jKpT13?w5kmPjMIpp}Agy
zl@1nu&H-G@#)Ctj7@wFcmK^HL1+eBbC;l)O$aPI}9d-+!8n6Y7gc*dzpP@e{RE21N
zGRz`866O$2g?WV2;RwQ+u!!(zIEvPPJ(yiT*0d2GN7`dy3E_!w4B_M9I6^0!W<L=-
z2``0FC>~D)r*@tQPQx?f#I}X_lkmAX9iHIRE@$m2jAKT|W~s9Q+FPID>^-oQf!t4r
zBjJf%&heS%S;<+#h%I$GIP8wG=g2eXjwg9C!PvHexj40RHh3mD2ZaYuYaIgjG;tpz
zrQ#EHotO2R4^yk=*5hGn%W=oIPChU<e;O=s#jY6-%nv?;1@qL_6G$xv6P)Yt)KSbP
zj%)nM*6HBc(7_n8;S-{0EriAJkbC$(6sucL>a;B<JQ-#-G7nPfeSs+U`d_0|396r`
znnlFM5G#daTTg{Ec=zAKNreAR=V)3?4=xZ6P=g&FW$Rl{H*I%v>*?SE*A?kbc=}U=
zZ#E2R=gFOms3qmU99%*@#jq4U6&`kv+)p>Bg6CK>o{xl6;nU&qu;@<R&+TFcF}CwN
z7lUWQCqKh!LOw%LJRdruAf68&N1YDSA6U144OZY&8hYLF8Dk@b9<PVa)|sS^3H0M%
zCbM$}^Us-?c4zMAnx)`*^Z^R0@Qi!(KI}xSH9UivO4<ldy0b4CTPKJ6fR@+Mvh2eL
zwwlm$bigdyI4CoT9OyCUIiAI5lK$x49D{Qva{0@Tn5ARAS=wKIn~-~XxR->vRm>pX
zya1^AjUc0;3HER7T{rzP9Q*~cHWSEN??aY5J}uA%$a{Bd4m0xNuCaAiT0T2%2&|>T
zC!p*bhBZk`U@fiwFrKBfdhwyNlvdxz_|xiFhGt2_S{g?Q4uZ9``f)gyz#7}pu$K1!
zc07yYUk28~Qx7s~1~tX|3458(fFP-F0=@dvWVZXzb6lY3CG<Ce9)TS|%W;90^%ouv
z!4K@)|M@!x!hiK0Vv8q&E1?7YsWH*l1V_LI+Q$)sm0m^<Poi%R21*c{q=4VXw{5{l
zuZYU|vBmDY*Kw%l<Vokdtquw2DomQ-I6w~T9alSGCTD5L9^Y$qlv7nL-#_0cyn{Re
zoMF1(J+Mx40pzZG#_*tj##~7Dm#JEaF&wuFHz0#tZ8cYFt6E)p=}jC^07KvBo51w+
z9um&aSpzq$jX2<`x*tMrxPyZQk`W9_EQmRtt}Ef3oinRALM`@q43+29f;q}Rasral
z(1J)p&KZbl=S7m3(_EYk7w1D^KA>Ke1qMZup>yfv6b_SHm_UFPg#PoW0jNc|k&bns
zQHew9_mnPdHJp~1WOA||@i4TCSM;OgM55fw^-5#et-Shki2DBnDMt5nB84X!xA~Xh
zlv}C$uOgzecVECn;{-;Br)&rVNg*(GRSyR>U75hW?Z49f+Prg<?wI;?noRzUH{W{?
z;##exAgci!Ae!{cVPJ^_)$M{YpZDewHpRICjIkU|J!q)kt9{_S>O3pH5f{&%t%+rx
zR=u8FIVc+sh?`mf@&n4dH0DF#1YFDuhllShKy<A&JFSl2eHzJ#uY!+;_{)<>PPJTd
z8~yXJ`as@*xnH{)!ujeA%=Ca?9ldz2BUx<zr}(bJ;0^*D&rQ2t8wz-+v0=#Elymmc
zA-CpNV1q03D7l!+q5|1t-4UtiCJrwi(2iPVt(I<G-jjia68##Rlpe4<4=MaEU%lEl
zNr~kln#2bk;Cz6+H&4)K1e@@$u|lfnBAcl=(b<4hr;qR8yj8LTqtdXZMfaKtB}GG|
z+F<b4Sx@2pO1<OB$*m}by%(h#m9`wzUxk+vKnS0|IC9_ui(_L{9NcNqDEiqK;Uh6r
z#)7P#2QuzaR_r7fF2|8e7sXnl9W}U0(ZDAWdSV+s8UT5=R)q=(>MVZcqFi%*@&8oc
z=V)mARv)l*fvPM0VSy<BCeQcsvl~&qURlObCpc$C8GI50z!eps8pB6Bh~?x|sdgDH
zioG%3-Z(!Vdm8Fl&U@Gyv{`%N`J_hJVg{!%;bbHRc{x<89ZHLhn(G3fL`;iDwT7x!
zZ>o)R7hh+mDdT7#lN7Q<#hQyRAIMl+YVH}0Uw(^u{uYDZW*|Vu^;3U`fw-9dE@OX}
zfq>oLV~nsD$&tsE_Jzy+8CJi?w*<;)4Cr7`j+8B;T~}ffQXU{7(V>Bdvg{)DGgbpT
zDrVN>CsM?Q8%^)sP81Y@;nd$?_L<(>c#8Ua?r49{;Y$*OZ42~ArT*&5=Q9ostd_vP
zFIwrya$#>1DgnzKR8cU#D|C)0n*AsTYAd>rXe924TlOnaMynXrZ=#D!DQN9tOEC5}
zd##s(SYFPU{sDEl?;|iC=Rw7Bc;wN^wrJ)MvvD(J+~R>y{=tLK8pk`h_E<dHIAcC#
zPNjQq`8(Om$?|Xj%9wEZ5!YVS$(k(N!eJVld2RennNxVye{}7&%tET?FUT4)7Tt&7
z9&F8+_@n{iHgdr74`<;CzBjFr6DT!dBONVF!ppEAZL?&r=D7ZxE}uzSWO^}G!m;nu
z_>w~o>ADYJz9&=YgN+_caGs$0*AC6Ix(7IzlTg5fZ0;lFS26qk83eGA!*w-)!zu&@
zIH3*1?Oxo~pG9~%QB_<8w!+a>;g>*l{Z>ykDcF-NOpK|t;>E$m^XvZz?L}#xz`iFR
zzES@K3F@CB0K7cOxUd6#o?wh<k>AhZhjXwI|D%OoLt6o3Ea9blv_H~4Isfkr_|nDw
z16Kcr0~=@Xiw*bE#sB&xhkNN_l~|k?(JJHL#O>SIV_?6{jXYPGzi?Tc4RENHijh?~
z9|!of4Od+Zk12XG!Fd^<P}#z<4qZa6g*Zp;rsV{+{GyJffNqY+1CJ+LKDdI9t>Bcj
zu@13!E5CUfu{^~2S-2?|AXq?&vz^+?hZ#5{=Lu=ILpnXfheF=L)R^CV7G;Vk^AtQP
zMxhd5?y+41pMAl}V0`w4;z?4|?cYQBJkMq87EyLw{V{4h#QNW3-4i+{u$!xV3saGR
zZ{OoSfmhK%(+;vskDC4}#>($6cB$8{QTqz@EJ0%KxN!cOu3NTB{E?GE$Bp<y2%52i
zHDFMY*{*k1SHZ`EXS`E`>4X9fc#CE-qPUy)V&9PuVsj4L66y_45h#`S>CXqcyw`8!
zfRh874iEHsvHNlZPG9lau6xcsHX>hYXu>xbaS9lEifL-$w}b|@_tzmGj3}>-0}168
zN@(vd>M3K#lexo6sPQ+Dq-b8(MEv&|JIml7GFWFow-LcF>JJ!8cAB=<o@3fSWAG0c
z{0j#E9KoPtdX!JrOTEGjTL_}e2EM<u>POZZsA51;txq7(=P}A)Y+!@-U@bA#{t>q>
z`)K4rjfC#N{}FUz#w6zFUkQJ7M;S*bXSxsdu^B{Ne)nf?O3%q(MLj)jZmiA~oh}1-
zt6vwsesS^UB3#&UN_}zhiyTJio}m661Nz2%k)Oz4!TX<*Skspzcib-YUDzKLD}88u
zXpH7?Db_M~22ApDRe!1Ghe(LhU<cdx)TYEiL#k`qIVdOb`DRmC)TP6{$N7~`Q>=xR
zdQ|Ag6%k!@5+7xo*NfrL@BJv-JU%y4*ABZjGHZsHL_~N;%|#Npq9cQ~B&@serBROW
zq#|NiO}&WIqs-eH1`aWqAJ)-a_172^F7<83ZZdeEL7M@+2gTn;J5Y0m27stf|8g(u
vLA{DaFnmI>W|GJ-O<1S%03$n>%B3exOcW;%Asn4}Y}}u?kei%vCQkh?ZTVg5

literal 0
HcmV?d00001

diff --git a/ldm/modules/diffusionmodules/__pycache__/util.cpython-310.pyc b/ldm/modules/diffusionmodules/__pycache__/util.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6c68efe7b2cc7fa7a422dbc36fba13a422cd5ba9
GIT binary patch
literal 9710
zcmb7KTWlQHd7d-39quk^iliw~Qsl9u$X-VxMOiUyMU~Xbk=wv#15<U<O`Y*@&Tu)@
z?#%kkEG06_APwc%C?dNC3>1lNlx%v_LvJs=r8l&1edt5cKI8xedU-I^KD2=!3K)jf
z?>}dk<dSmHE-`1$oa_Jp=f59j!^33*zuWiUjK6!zF#e4`cK>XAT)~r1GQ;2oXHBE7
z->lBmw^=vU+iKZ$o9RBM?xM|XmRerjQ*Bmrs8z0)QJ>`{?%ihf3O~ooyn<R4wJIM*
zZJ0mFAL65^jqn3}?6y&Vh>!CLyhooi_``hiw!tT3bNRsX*oIXf=jZtpKa9~6{0Kja
zQ4jOa@MCyS@<+Ij_d)(BKaTey%s#<SV)n_{UY^3-!)W;|KZTZ4XgPwGqx>=cIL3a4
z*Z327ALF0nr}2ISBhT<@jGXSx_4()d3_pu~ALVS`s6BO$C^lDP(iL$#4hz%{CTU)T
zX_16Q94v~g6+~IOm^3~(bM@$d{OePPFMsfjiQj$hU)J8a{K3a}fBg3^EPU&7<j^v`
zpNF@L=PDjW`z~mI!xV=KW0`eLZjKvWOO*^QCb#a`+#Wa9ovyQ4-ZJJ5jCU~pW%OEI
z#@&NPZ=Bg>M~tq0&{%i7E_bPR#3<}#XTsq25o5hn7@Kalw8b_{+}>jAr7r$>X$qs(
zOBj!b8?TvD25JWPsCL9Sgq^X5*L85ZL%e*)8i1q%NHPt17tr(}9vD-_n2r5o2*;gq
z#SkCv8t)o><b7*{eVf%Dnv-rbN#js7H0A#JGN|wHVznLLGtng7D9bV69{AS~RNT9d
z-v^(2eey?t_uv0{xn@cy3|dL5Ug52JC4W^EfnHPE^s!dXqotU4nz8@bTKO@*lio<P
zo9T3|BAp@=(USBqJa30lEK7^cEG(WnFWpdtX(N_q+LmTJkru#^Ry(^X?WmcwW%;GB
z*@?d>L?)z@c3SP#nkT*fM)h(U6iF-2i?}VT8khhKB??*YeZ?uSdm6VOm(D_5gt;8*
zPs#y6v%M7NcMWkINPl#FW+`jMGig@D3t4t!Cdzm`(~d=+rD1bs{+YAqpFa1@=XLYd
ztMf0;G<j>Ll@ZqY3{MspJ9(0&{f17FG^g9Ea;z2Jhy&U?=!5<+cFP|};jv?;dEcAx
z{_8N_lh!D!vZHL2d7k&t1oIs4<3r4|M(>N07<@e<P1FCCP@zY0|F==;asNaY+<Vwq
zx51;VFbhZkTI>z*0yx!4k71kzK6(QEHaAf^kQIBv*kWC_WpQU5yj_5|G1e<Qa5M2z
zN%InT*HcnqiOb-#p|1O`@xHmin9+41<>fh9>ZHkHCR&P(r2T3q<c;rtkAC+qiy?ry
zHgkO`7O|h?ewz85CoP{Rc_iW@Sqlr$&u_)WQpWu#Y&K)=AODCQ7ZsZL+4;B`M+Mrp
zPteC!dW}46wVScH;MdB6$SX@>9tshzO1qilg>)3Fi7|AEQR;WYoCH8xNm|J9xvcmZ
zTA&rD*r3O9ZGpa8seXW<3>1`s2ChB^b_`VWHZ1+Nc59n0b6Hx6#X^?Hl0|Z4&-fwY
z+&suSg<?NloFLf770V0`pGHSMfx=@R16Qz<Y}E8@n~gA=iF4@qboO`yy<MZrrh05a
ztTJUZ7_~t`nWUhmj7@XPI0!a?Dj~M8xYczIfmyotA*e5UcO6n=#3~b{8i@&@QoA-7
z1WFaGU=u5#&Fz}RG>Sb`TNg^S>nvJe5^qlNhNuE4eXx&hAFP(g2b^jc6c_xp$`|}F
zc)U0>Tf@hxYdHI2DfXi!5IFT$NLl#Vq7O+s?cZEVq9tOowVyhTQK*PG?=*|tPf|5&
zF>5xnn@QU6^Q6(j`hH(J_igXj9K}<DL{_>*MjWNMNYJT(LxvRKQKyxQSsG&j1~Eej
zK8m7V(Rc-I5wA#B?U(O8^Uv&@x#DF&7PO%t!En$(m$6Jv%<Gva*(u9r6RgZk=G`xw
z;wjo^7)4~E=>JGJkcCwG;mfErjCHF38*H#=jhBzE<Jh`nubi&U*+I0Ujw2<V!J!Gd
z?go|41}xzrqib%MTR7TR-NM?myVe%jbTd2w>rwXcyDDJPV?Xps&Fv7fSPF|B>5g+n
z_(rmV(^3=(ieMoW{$eMMNb*mswe-pHco7y!&j4LJP3VS|*l!D{+o+&peth-}?N;+M
z$e2<IJ~<J7;s1pHwX>&ZU!GRWT~I{ymAv*<^p1dDP#~dm0c&7UWg`3{>unSdtf65B
z#20{RGi)vJ5YvHT0S%w~A+QpS4yh0Tp<p>nOU<H(D_DqBgFlVW{|(Phn^vO@KW#DC
zK0<<oV%r=MCQ+Q{1DLeJTgoD7^zoInIxT&8*rJoiU!X#ZRPIAgh{8)ia26mucz5D$
z!W@o71E(dDB8hvODk+ByJpnF+-fg#Gm!$=NLThSA>1jyO!B06Psl|(UrSA1X7mv_t
zERkk0i|5>4c^X$o8NfgU+voNIrsSk|>`E0Hw91Y{io_)}42TXnmE>qD(K(Mwmn|D#
zXX_S(21b>{W6LCmYuW5t4F)4>K|mO{NH_i(d(n6k&SlM>Q!4H*Q|N0N*wV?DLRcNQ
z5#p4DEI})&2CEsTHCM7wvO<<_1-g|jNES%zK_();RfsSu0+^^QDB?8F#47+bC*aFK
z?HChZLUq6>5K`nr>>y@|N$3Q&0e_QFfNi%62DA*Igw_T;hdx4NEEHla{2_cq#RXp^
zrK=^sFU0=M1m?R_z<x(dq+S7FELt%qH{S=u$%-=b`Dy5Yo(9GtspSMb$O!hm6}K|6
z>XT%FmE&7r7;q!0c0$^ZvR1o;#d?~2A&hR^3<dYw;T|?8m^6!TLkISbdq59JM3Q33
zTo2xs;yiw}6Q@zE6ap5f_bas2D+6<thQVBDzWxNjx}&^%b9|1-C*`>hA7=ps;u+)~
z&gO&U2{<>CC@Gq&O1bzU)Jspd3`+81Gi+S&lSThp*v#Y8q;10mc;O8Z@`U6OZh3U0
zo#7O47#i#v*s$op+mT~ew|lSklrVh!cDuPMh(~IUv<Y45YVt@|ql#8+D{X+Z&yJOM
zHeAFKIoTk)JW4K$DdzDJ;TiZ@__~58Cnq;HC}yskvBe=+w_&a7HtH6)Ro&qZM9US=
zW9Wd)5^K{|&*90xi2}yU2#kUu1O#i7hXaFV1SYk>;Z@<8YOxA?)7gU8vgxX_v<2_z
zJqu$!v<+<;Z#@L})>x*;d=mol6@x>HY-7VhyFG8r8u`;V88hhb=kZWm&77-QvJ|Iu
z7+er!a+o9{4tj01vaE_u8uU3+)YKS7=$<_1vZ_Bd)x`lCF`nc>*eS9I{%e0NIoXJd
zoz_OX6R7A;PVF8%u)Q3oQQHfV05#yQrddFogM+uC;+A-cc66bd8pV=$kv>Ks+K`k!
zgnPAa38GB0cnf3m!zf&HjEyl@{du31tr1qfZ=0i^*yh^d{meJrn;uaR+W*nfQskhH
z_K#7~oIywOtUgp6GdZtsorUwsmaSz(>b#WDfw{lV-ehk*jGPMMpxf#gY0f${+;XW)
z$7$_8&eX$U&h9?UeZ5z9_hL71c?b-p*`o|UigBAecWg2huVBAd%HOo$4kLpArtszl
z5QF!m><meTB31(vULIP9ec3d&$kM>me%pN2oHy>W>xjZ#4u=P^^oW3-NnkqEVFh<h
zX-=P&?AG4XC|@K`mDJ^>@6aB!;bBBz4<LMBiB4Z^g0-6YB~<!zKON{vEU17kX`o>k
zdjKm?kY*^YB=OB4%tef$d*BmGX!|Z61=yr<9pkIYC<L7p{s)H?&FO_I;wv;?NkBzQ
zkis3Af({R*h3HRueQp-NKw}l}sy5<iX;Ue_iXM0(U&l*Ud#p?yd3^*Hm+TReP;GcF
z%9zTEmOnLO5Af3Klk7GY%TJ*gGt2PH?fat+qWF)@Zyhz<lBxW3ler(;u1y|0TRZZ<
zkbe$7fENTQ-2Nbt<si`61k|fRfMB}W>!~8qQ6!O$Mn#CtMNY->!QupclCHdoC#N)~
zXDEm&BTTj58-2_$t_?T|q^sz;N>ly<Mc2w-0$cA&Dr`c7Zm|lb%aB6oBGmwav+h_E
z(0xb1v&%>tEF(2`hhj;@hTM70<jx{=@TRwAAc^2i!BbjyF>eTwX1VKb!2#K<(7RzG
zc~-h(BB_QwCk=6(Ms+FWRz*6kw-zPUXuW~)rFo-3>I-QJrqXLj%0cKV;aB_aN-s=-
zi`vPvgon-TOoAUH0zkl>HU`b>c63kHefG?={&nSPkuNd_|AQxRtxv<CBQk-Dybu-$
zI+3NI7(m<FL3f|Nt}n%KW?JDY{B1-D<PlMRX~3gEzCpVps9;0NiuC+?&9ei8$s@+@
zNyab0HA)MSHFews4_o>5i=-~#$E@}j*@rB$FCCEr$JGDYoPP^mMC;{UsJ4+F)KKoD
zlgjC+fFAbF(djn}UXC((caiSp@d}94_l2c}rIjzd<}LVVF8K>ShtxA(8_~jT!Sj)J
z3*ND`nps0o#-o=peDWOp{7iJI#Zfs#7^$!j2dTlrp~~`FEV3Ml$s4gCU03sjkCz~S
zpkCggQ@x^-&_R|qksyNLFJeVdDc44BXq!O01d3!yj`bNL*dFqIwC6`r*h3ytF+MZz
zyQcfeQR^6l{c&bnBlnS@AvL#8k6LI{Vs;D_gceZxh|N8vOBoCRY#L&2SQ4-?Tf3+n
z#X3;*$-gGoc$bd}{s?`oEhyeZx`U3?PkMZa(<$$a77;d-4v{M$yQ1ew*4~S6MGev2
ze}FkEyr|kL=kzwdBO-JEN9fpi6iPlS+b9je;F_BXB{C7aka|WTH4l>mFZ=(Ak+_3B
zG%7YS1VPObZ=nTjNXv;|rrsTec!IjBs;j3}{}e4fghnZe_71)W&WOAail-Dpbi`!m
z3zZ>3d<5mU!8R!HGw)D7#9DhwA#)ws0EJ5{hCHe41S|_Y{q&4h5Zgk%J3bccSf3nj
znx&jU6)!74TwI|(2$PQU_98VzQv$f;#ou6DPDgUo7QcemK5Q&BDs0FW6!4^w4}yD9
zGuM7e!Ttil9{O}I!zrXnk;)C5{mg7LT#dzm?un2TNaH4gpPd|vvYCiGbWd;iE)A2W
z(=Vu`xk|JR?4SqV^8p;ttvn~PPJ0d@Jb4bEUVq8G_zw@P@wb?)Xi#>=yQoKa27W}J
zD|l2kwI6S2pR$ff1JtmkJ)b=Pa|6sxv$mOg1Rr!qvq`A{BtywTo2!{xg_S752lE2A
z%+BNHqPlk31G6&O|42a4x+83@N*ngw-?^GC3bcz36)RLI6Mvj)JH-k5d=JGx;K>O%
z&!{rhvJYtsz4TL>s*m*R#v*Ji9AmF;y$<Z^6uDU9dl>gX+-Z^g5Q@UsP`HzZ!Yvqb
z1<SYxn>WP!=mZlxc?T9r1N|DBq)oYYWfR1&;`=U>LqQNV5mN_&#z-MUSV}yKDu>wG
z_%06U69?|w185_Md+SAUv)Nx$k?IFnPpH)iuOHFsnp6`^V@guq{}G1N`?GjHBoC5O
z<+v@W!U@4yBc?DZqEL5-Cfez)5%(<v?R2|x678VMPP<B!<0ge}OK>OWtesc#ww*Yg
zefG3}4i9qLPd-a#d?jn*28pgil&omU(5WN}hVBuhA06tXpmaMEbStu>R&WdA;8rb_
z?o})_%k}$Q@BJhf6q((%q>U*(ljrIPa1VFM5<wEA{GTx4L*jbR#_bf{P~m|O`TJ;!
z@NEBxo%;mLNAaNuXMhk)6Ap&#(LvPpW!(Ut5T(1AE8XYSy(K;=_!PfK#UB1Iq3z%C
zD3%$u1$_^wz1>GDeiOYQ{?mgH3=%&?J~?c_xMoQ7BZ2kc0}#K3rNnQm^)#?UKe^KQ
zGUkBZ{VGV_ulA@eeut(F;vg&G3X=Vp3QGT&`pMLaKSVJ=TYMXY4gCg^CA;Mqfv$}!
zd0WXdB}*K6;HOs?M8d%fT)A<vi2BRWQdFm7qjye_Y3U~4qX2LBk|Z$Ts%CR&yPlkj
zEsUmk1&(F`p`#V#7+3~w8C~Z0x04S?@XPYBDAL`{P&J#FQGY0+*4`Za13-NY4ZA0x
zP5b~OHQi27jq(5h(D;z-a~VMvbm-xIcux00+lMfLdKHgCnEVt{<Z!%5ts%C7zrt`o
zW+J=G-ctq=>KzuS2dJ#<f)L#KCs;|4JJCZ){24X;ITeFPtyH1|xWy0Axi+;Q;$8*>
zIj%igjxg35^*09BRg_Ug`6G0Ef=4+uum$F*BdBcvdj}1KJw5qp6t5evvqfC`kR4cu
z9Z+ACF7c4WFvWP++Q5<H3YpwcdoIG1fgde6jwio}g2bJJbZ?MhQAsw+AHRW1#yjS+
ziB^j)9=TmuZ-BEn8%UWub6UZD0}JU(Lb~^YF+6~my29nDa_qHNM5MiIs(?UwNIUEy
zy^H#O=28{3&WXD(8f!<>EXiXP9_&tYsEeFS`_LjSrqCj_M`*plfCjsxxj_(&IC+9h
zP?W4U|I6*#399`Sis3#4BYF(%*A?pPQbCZ_tG#GMfrtNeASib&M=Qzl4(t_Lkl20%
zOh3xRU!w}XBX|X!g`hsMK&?#O3j)kr__|hO*J{V5ndRaaX@nxGr~W%*I!%=h6Gx~+
z-JiR0l4k8(L+If@9j56Y8kC5u|3*+(&id~05>5IN74uYND5TfVC#y7r3V+1g)cXz<
z<eMtzOuE_?R8ET=+BWyfGArw%6uQ!dij@_%nmEU45mhJ(c^9>!z6A0Z`FRvxneNK<
oO;6P<Tji<pf%1qq?zs~O$|t;o-VpvuW92c&v%O)|hDV0~2kY#P5dZ)H

literal 0
HcmV?d00001

diff --git a/ldm/modules/diffusionmodules/__pycache__/util.cpython-37.pyc b/ldm/modules/diffusionmodules/__pycache__/util.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8b3e1ed11b4b54209cbdc52a83e3d570dabe61a7
GIT binary patch
literal 9626
zcmcIqU5p%8R<8f<>FycZGn28$cASdiIO$|O6WhToj<G}Rb#|rAu*8N$y<{sjb!&QR
zTwT?<RXv`W?OCCW7hAAS_}A`c7iM07ED(fvLBJbt2=RcHhrJRKKM#>X0<CyJiV*Of
zd#k#q$Ic2K=yBb;b?g3|bMN{2&MCh#K3>!Cd-IQ{!mpp!w11+9@@Jv)GOlb!*EFUv
zJ=R+4uD5jgY_tq{H#=6#($zb=<)F=oE1hbqD%;F>tW#^%P(RBmta?Xp)!8{#V|COT
z>^y6*an#1yi|hb9h}r}@#3t`(tpjX|P2+y>RgFEyX6|TgCNx$Ltxj&6tts{+Y?d8G
z?`d|79Y?Rn*yHR3?lbHO=Hh;sJ;_euegvbRVy7_rRA{ZvV(d}0Jk3s{<uqE3q2)Mx
zhCPeEkFzFw4)+u6c{Ydp6X<z{&7<dhG1g@l*aCY2>psc!8(Q;6AJA?sHg)0fup9b0
zYKNmF%l#ye{5<rQdD`)UG+B<?AD_8;{9pd@!qF=qe`WfkSN~!2*RFj0pCA6<Z(dva
zxhsK9)8JRd<>0!CD|4{#eeC-!gFlyRt9svH#+24KWvQdZVCH>`SyS4UoqP{HZS?#I
z?PgzR&S9<SVf6K5THiXXZ8?31IaE8Q<<_b_tugDEwpGcs9j9N})psh)+SRuzef+V?
zEP8EK&>t7mzGlp7sA;T9wPV^5tcf|QeH&*w#%lLX(C|gjPzMcvrKs(Lj*?2eJ%_gq
zy#3dQ-x5dYaWEfa<9+P|jZJ*V+}7XK`}R(4SHnIZSQJhiC85vT*ctr({t9;5X8A@p
zd|=>NI6<1B!vpM@#;JJl-}tqkzdQ5&zy6PZyV5j-?R%XlkvD(6RnOkxxu@n77Cmg%
zvS1}-y*PBA*{nU|_L5skdOMkK)`gv?JXjG`bkDke5Q@rjocj5N7lq?<KWT@;NV>x4
zM#2Oc!tAEEg%!k6SJZySk9*-~IZwH;lU}F0(X5JUxlpT?czM(bvpnp!8j2VX!o%}c
zt#}F()~YJryiC|jVeV&QtQ?eqfVjKjXCG?(2{8T18w)FGCtOI<JX}iCTMNw3{e?D9
zd)<W_z2pl~(gu&oLXfg>;aar3{L1HFUx-;}p_3Bh*#bjbFN@NoZ0O}tJm1|AlO6w7
z=+W|CN&RD3Fgt-l)2qfwvuYeSX7q!4LqD!h8dZHpuj)sQgF1f_J#GeMPUu%DE-B*g
z;-L`nXZsM;quQ1Qany4ohnS<q+J+!NDDC7D`k4@)n|N<A1Emdxu%@+Ly|3?@%$|Z6
z=MZBlNL2_f2~wpLBvzFQVDhg*$j1842ikXxZC%&;4#carC@Q@qT26UK%8sz!>G`bv
zt#8ur!4*CRGB+1)u7o^vqs&cGm$9hhvM399m`9s_jvaG5VZM?wH}K;)WbVmN^pm_!
z1E0PT#$k}7ZE$QZmQr(Mey1CU{G!{e@oDrH6+iPi_cw$UM_DdxDPQ~$UhxC;-tjY%
zdSOOME~XaK{PQ$HCrq$FA!l8Vcg;%c5K-w#Dm_J9YZ5Z($>sqqWm~s1ki|??)<V9N
zW}(mnF|n`z7>R4<rM+CroGM(RZAuEsX!m)%$cQsY3I2}YI%OQRs=B35=zIa~j~0!4
zc+=PV`fMQ*Bqp<3Tc<YIA%pDDthQt9YKI{LTM!0_gvrdleFTEiw~oL{(Y<ezwIVT@
zCJRUc0GrvjARKnzh8S2R2558o1__K54cXR*t?b*&CIqCqDCL4TK$JSvX9kC=6z~xz
z>U;S`ceDPI>wC}UXU;b9aQZsVJzojkU<I3-xNBq`+;rK6Ud_3;SE68r1Z?w_IrKt>
zhgmPqGdD`)sO2<{)7wyDH;dXG%<q=E_|Woh)0XnYp9aao$y1UhDM9>sv_Lw>B;i4?
zlW{sDVS)xeM+}lqx9W;7ugk+V;mGx}{pbDo@VuoQgEFrRiwI#0d3FWUWD_VfSc6mM
zX`J}9K5g*xX!*D_%z;TA@FTlG?oZl<o2az4Epz=^4k6gquW4_6W($YbC#PihEv6qv
zJL))4vK9<Bpzmx`X>Y^b9nt#6wy}$YeP`kgEjM?pzPU>d+wh-)yQsPNUFGoCq3gS3
z)P@_GulV^;v%^eod^=jhX~`Y&a&O7!?s6{)NbBe1T<YXlxC|GhMqpoiG3>%x=yo}b
zYmn11KYR8Jt=4o?=$AAHE+GUrcYoad(hGBE-<p@xU6h;XN?k1t@=!T1N|Nw$33Fgl
z=@Z;MEfxw#=1??);)~!k_B%_=$8fNi!=7iZ53YE-N9F@WNLpT?sm|gJleExQjs6rr
z|2sZ=U7C#+{Itp7?1%|cih=VXPJ%GYMmXvC>(c)y_VE;TdL4CmSfZDOFHxaXDs!PH
zw8JZ4@GMAp^y(zr#5s^d8>hviJPM0NrItg7o`EF7=61Vq#KHuCP*&QOHX2%V_)`w4
zP%n~|R<$IZAE(*$NErE9T<7-d)08|)56;0GoZD*{l9Aof>SQsi@UN$I{t})CszZp8
z5TsO{bEx$7Rqf0AmI;-CGbQENHJ~>3s?j%>bvREG>Y+1}?BZ`4?-9bX7Nv1(>XLEA
zy(R2y#fPtP+CI*Qiwf+KY%t?6X*xppg`SJbx~E$8C82x5e3WAd*K_U%xd+FTdU=>+
zDSsP;W<<KC*T59Nj@rN#5IckzBuf`if#C)S?(5rtltetdxs~f+2-;1p%fJeX6><>7
zk_706|7QS$^7KDT##U*3sl4v(2rj#q!)FI8WL805$U7k;?CyfzXid8IY!13#SY8~C
zoQ?x32Jq>fu#@r)moy3@9Iiu9AcbVoiDfrPJKY{8D=heuAKbd_bLMvaed0|tDbXHK
z2eA$dqDM3#N-$-nIuEor3*YI5Nf1gyfXV6pHcj>R$XIDvFcucC8~|E}M!Oi}GK4b;
z%Us}^C2SB&q3=MIkJ2OX97jQv#~ae9xIWBEVN*smcscgl7u{&tz3#_ZI7e32Ujp22
zbDu?|hrp)6t!|1_z+ot|7n_j6ya$LQCa(@&UoWgL(0sQWZ*X!oO<P#RuW;1<2uJaX
zR;(*5(DaZ4s|{CN#uOPjAgyKr#m*S~Cs7Zmq5PSsyo@V5g(B1_3T_#p$)Hhp;G|j>
z>L#;f-DWnl%Hf|w*MXvvR1=nb6<4-`0{%+#v|L9}2dO4V10SY&2DJd)%CJken7OrM
z?*gXmII^tlLcYIkqHh&#W4qe>lfbgtDqT?ME$a=9X+NdCuDyxSkNn<Z(-f63q5Wo@
z*c9U=^r2U@HEW{5dx=*HA3rVo$gM0CHWLl?sP>f~q#jdI=J~xm4FF5axx`F6%!jS*
zZqJjEnwZ_&d1QGpMZE?Sk&d*1LX*@Z;lR<FLB7skrxhL8o_4;%7wKUF+6!eUDco;R
zE6`R_w{`T*CQ)dHqfg?`!QTYz&Wyf!^kK1@FNOsaNYalEjxr_k;J%KEI&<3aYwE~w
zh=fJozYHv?ubQifub1gCocHwiUF1Fx&fJj)ME$p5wpJ^BIw*5^P{SH#sIHDd1k{1n
z0Csd{G5fwn7X59^^LFj4CJ-xPc^qzKaX=`=!bI#1T-i|+ZFGKXatkhCN82T503iI!
z#ycbv`b|W|4g(%RY&*eW4kFlhx#1q(81pX(eSP1Kep0(cK$5K5<)5cDXh8?T(IYmo
zv~}~BVu(kaT}GuG`)Gn^N`fa8!SGuJv62jg{FSvRyzTi}(~$my`xvp0M81SJI-`QN
zn;6!R=G*HfKnM)Vu-L3~+9D;xb)^H6bHwdE<_i;HnW&cX$ls*CQu<^Y$)vET<b+fK
z6I!^n8ig>@i>x&Pr$|l%>7OoO1^-#<Qdb%zd+eiIRLc#!gPs{d1<ja*j80nz^%Ia+
z19bzTwgn(<=$prW$d)g{IafU|0TTB-QS&?%&ZFM&JjAwf@uq=fLmmYxIOK0(MShD4
z$|dY6TzVu+b{khlei<;NAzL0C1VXw#@Y`fr=(<WB{~Sf%%)SfZ-4ogGz_RV?bxKbm
zj_)JhU$v3QGN)mwjzKzC5xK7-6?LD2Is|FV{+hw;W!T1@>aK=_lRXRAvgKgh7(%{U
z-`NEw*s0UKZ6H}yxo;qeg*9h1{wLI{Pbsbj(pSY?l*FR>_V!OEa-^P+8PH`i3n?&Y
zNZo(4L=Q!@0HoN<(un!-AeeVU0CB{;Nqf}W4FYU(wr9^=b8kx6L(pImaDYWXxO4EY
z2oQi&m;4;D9})!w#g*$DMcwptb0q|}==d9ey#VV3P}0PYM;tsteuWHB!Ge^SDB!h{
z&XLXpOtE^DvWq|}NzT()g)acQ5?(KpSpclqC?|P{StLMuJOPi1`=v#99ZtRT)*e>7
zNXscIAKE9GR*`Wve3LEHUI;=H<({ClFimUNq!O446H`lIx>;Rv&s=twT!vmcYfdON
zHUUJ0)d5r$W}LS96vVH{)SW*ExSjHhT5OrSgG2FiP6kJqI8;&F40)O%A$Tj~WYd~e
z{y1(tN71Sc_es=MqSs547&#iK_%de1E@jFgjSV(vpZ<u+QXsrRmv3S6jA+otEFgOe
z$i9Z`*$LF2GmstG#+wmKX`(@@)=5;>j|1Jqs3SN&DD9;V0}mgBpttr0%s!d;%}Wxo
zwZNQ9K$kG;9<T?=@LgpaDBwfBfezD+3JJuy)P}wJm$9g@X`{$LIBJ~GyZbq<%P#8m
zA7P9_`PKwr6~Bk)fHs(Z#12^#$<irQ)=vyrnuL=yb|g#W21c0rG0D^oI2FL%|0747
ztgB+hLWu4)O-_ys0!Y*GyYzNw{#<(1kgp0G`lo0aFf^(2U&Qm^j0i6{xFkbl!O4T?
zGC_eb1qN?h-=@gRxKGIlbMt~^<R(IK$ww!I7Rl5AJO|)*az>ejfhvy#r@9nQg1t0O
z0yY`0N;J$bVRcd>6>RNiX^e(=0LJ-Wpr7=Vn#JEkeS{Y}Aju1XpB$(M$ruBa8=Jo%
zDSwS9_g(tl!6>RpkU{n1G9eoK8zCRDHv!`DB#aS%4AU3VBLe!XH%0e7RwYU1Uz6Es
znd2H+L3MuMf;g~Rf0gpwiy-0rIXqU&Dfja}J~GE&VzA=h;lF}<fNSt0L46sQOkkFX
zUtzx%us(Le64s6A3qL&&X&NyQX-a<4H;^5Qz>$9>ja&qDk;u71IW@QvsFIb1G0Zu7
z@1s?^<i8~Xq{`@<4O*;(r-+SsOoKX9BveQjJ``VBbLqveqWB&zIZ;)Y&m*jwcu7AA
zJW5vYYDgDQcvEeecfnVSaHGk83H=`8m?p@Mpvbjt$uZd?d{shtzKZWkigz1&)`Q1U
zgnk9Rg+=LX8IoXUejP1xBXEFYo)^T3WIa#qic9Z^L4m8>kf}X~UhWI<*dw+@j2H6-
zVH}rp(gq>t@eP{VrGi-dKCW_~B&(#>^dlDOqR23#_=A$t_$Dz8ltH;i<2|s6HUn+X
zq77d(25t21;u6}h9jZ;Hq4AA^zC*mLt8j7iMX6SU)a2P~bM84@$Qqx&Mjm=Cjqyc>
zJ~l{|Pzs<@HWXMsAnG=c^b%}xH|6xrVQ4!px{ELtMv@5UDrP#%)css>KhO9qM#JV?
z7*n`2M@3@$#8Om<6fa?ahXE?Zi-dkdCa37mchMf;l0OpN!LBv%Aa{$hG*T!wpB&{`
zefSz-5Q6am#t8=>u<#8KkF@!Ghl-)FyYyH=@z1!Fd|2{%VA6<+0VV%0=l7XWUPjN)
zN7C1C!vUwr#v|MG=riPm3V!Grs?8YtiM93@F%!<Qtl}uk>fki_Z(+Jm-7iu1*O0Tv
zG&s#4(7sKe&_+-5x6v{<O?pjxNWYOp$X;8a(jYBM+6+pB*y7OVHkNqAAbCsz!jK2$
zG;}Y8Bs+-*WkCqZ<WQj^0RlK6YZfc)sLg)^eJRcXSk0l)G>@2sXad^NNB43e@mK(C
zm-%^~>}`f2BXiR_5KwC|1|RBMlW5pG0Bxm>7#h^y#ZVA9OkviLN_dFJ`I5L1W{J?N
zxFoZLc*s2f!jaiTYy{+^<3pFxhVA{fbhmIC@Mi^?qP~Yd=)#AXiGLSG!4UsFYWRIB
z_9+J$PkSHDvk%i;WIvF-+Gj?IXJxs+IWl{>J1({U4vPQ6rQ$@RVat{Q|CF5|{^`o*
zP~6q-B73X6%oe<ie4=zo70C{RpYEI6IB<Lkren7j0}L7bXflHN*##72R+xQ{9F|O+
zeZ#m1Q+(f8HPB|xk}I=v^B&}h>3cHY&Z5e8>oPo0@rcTT7s+sVS?RFpQ!Gnlbf6#>
z?JoA8q6!JOhHgq!FR5=WmmApbe@6#x^LUa*Stuiny%`btN#^nhSA-uVSEP%?)jb`Y
z==TdzP=sP(qKf|k6`xKC{|N8LK?MK9K}TCrE9oY^Yc-0PMpB6Xr$B^0Jg1AUqBi8B
zK9r;h`j$&jA0kN9b2=DKNSLxr^pC$t4+ihzcD<=zZ=MuJnsG`fsNJo~|6?(qB+|O^
zqx51Lg>b?sNxPZccl;L&)|sdOlOUB{{?`Ke4OxCvlwU>UcLbTc<q-|9mbqe?u8<*>
z3_>Jumxx!wKH(^kDS?p~8=wja2&6xgu7r~<lmac;-J@Uw%Yd_2hW&CKeY0~Y=)-s6
nAB~1to2?zHP1L5UPW90AQ?<j@G5l4kwMl!@s*a;JJ~{qBuNaJR

literal 0
HcmV?d00001

diff --git a/ldm/modules/diffusionmodules/model.py b/ldm/modules/diffusionmodules/model.py
new file mode 100644
index 0000000..fb61fed
--- /dev/null
+++ b/ldm/modules/diffusionmodules/model.py
@@ -0,0 +1,855 @@
+# pytorch_diffusion + derived encoder decoder
+import math
+import torch
+import torch.nn as nn
+import numpy as np
+from einops import rearrange
+from typing import Optional, Any
+
+from ldm.modules.attention import MemoryEfficientCrossAttention
+from ldm import xformers_state
+
+
+# try:
+#     import xformers
+#     import xformers.ops
+#     XFORMERS_IS_AVAILBLE = True
+# except:
+#     XFORMERS_IS_AVAILBLE = False
+#     print("No module 'xformers'. Proceeding without it.")
+
+
+def get_timestep_embedding(timesteps, embedding_dim):
+    """
+    This matches the implementation in Denoising Diffusion Probabilistic Models:
+    From Fairseq.
+    Build sinusoidal embeddings.
+    This matches the implementation in tensor2tensor, but differs slightly
+    from the description in Section 3.5 of "Attention Is All You Need".
+    """
+    assert len(timesteps.shape) == 1
+
+    half_dim = embedding_dim // 2
+    emb = math.log(10000) / (half_dim - 1)
+    emb = torch.exp(torch.arange(half_dim, dtype=torch.float32) * -emb)
+    emb = emb.to(device=timesteps.device)
+    emb = timesteps.float()[:, None] * emb[None, :]
+    emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1)
+    if embedding_dim % 2 == 1:  # zero pad
+        emb = torch.nn.functional.pad(emb, (0,1,0,0))
+    return emb
+
+
+def nonlinearity(x):
+    # swish
+    return x*torch.sigmoid(x)
+
+
+def Normalize(in_channels, num_groups=32):
+    return torch.nn.GroupNorm(num_groups=num_groups, num_channels=in_channels, eps=1e-6, affine=True)
+
+
+class Upsample(nn.Module):
+    def __init__(self, in_channels, with_conv):
+        super().__init__()
+        self.with_conv = with_conv
+        if self.with_conv:
+            self.conv = torch.nn.Conv2d(in_channels,
+                                        in_channels,
+                                        kernel_size=3,
+                                        stride=1,
+                                        padding=1)
+
+    def forward(self, x):
+        x = torch.nn.functional.interpolate(x, scale_factor=2.0, mode="nearest")
+        if self.with_conv:
+            x = self.conv(x)
+        return x
+
+
+class Downsample(nn.Module):
+    def __init__(self, in_channels, with_conv):
+        super().__init__()
+        self.with_conv = with_conv
+        if self.with_conv:
+            # no asymmetric padding in torch conv, must do it ourselves
+            self.conv = torch.nn.Conv2d(in_channels,
+                                        in_channels,
+                                        kernel_size=3,
+                                        stride=2,
+                                        padding=0)
+
+    def forward(self, x):
+        if self.with_conv:
+            pad = (0,1,0,1)
+            x = torch.nn.functional.pad(x, pad, mode="constant", value=0)
+            x = self.conv(x)
+        else:
+            x = torch.nn.functional.avg_pool2d(x, kernel_size=2, stride=2)
+        return x
+
+
+class ResnetBlock(nn.Module):
+    def __init__(self, *, in_channels, out_channels=None, conv_shortcut=False,
+                 dropout, temb_channels=512):
+        super().__init__()
+        self.in_channels = in_channels
+        out_channels = in_channels if out_channels is None else out_channels
+        self.out_channels = out_channels
+        self.use_conv_shortcut = conv_shortcut
+
+        self.norm1 = Normalize(in_channels)
+        self.conv1 = torch.nn.Conv2d(in_channels,
+                                     out_channels,
+                                     kernel_size=3,
+                                     stride=1,
+                                     padding=1)
+        if temb_channels > 0:
+            self.temb_proj = torch.nn.Linear(temb_channels,
+                                             out_channels)
+        self.norm2 = Normalize(out_channels)
+        self.dropout = torch.nn.Dropout(dropout)
+        self.conv2 = torch.nn.Conv2d(out_channels,
+                                     out_channels,
+                                     kernel_size=3,
+                                     stride=1,
+                                     padding=1)
+        if self.in_channels != self.out_channels:
+            if self.use_conv_shortcut:
+                self.conv_shortcut = torch.nn.Conv2d(in_channels,
+                                                     out_channels,
+                                                     kernel_size=3,
+                                                     stride=1,
+                                                     padding=1)
+            else:
+                self.nin_shortcut = torch.nn.Conv2d(in_channels,
+                                                    out_channels,
+                                                    kernel_size=1,
+                                                    stride=1,
+                                                    padding=0)
+
+    def forward(self, x, temb):
+        h = x
+        h = self.norm1(h)
+        h = nonlinearity(h)
+        h = self.conv1(h)
+
+        if temb is not None:
+            h = h + self.temb_proj(nonlinearity(temb))[:,:,None,None]
+
+        h = self.norm2(h)
+        h = nonlinearity(h)
+        h = self.dropout(h)
+        h = self.conv2(h)
+
+        if self.in_channels != self.out_channels:
+            if self.use_conv_shortcut:
+                x = self.conv_shortcut(x)
+            else:
+                x = self.nin_shortcut(x)
+
+        return x+h
+
+
+class AttnBlock(nn.Module):
+    def __init__(self, in_channels):
+        super().__init__()
+        self.in_channels = in_channels
+
+        self.norm = Normalize(in_channels)
+        self.q = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.k = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.v = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.proj_out = torch.nn.Conv2d(in_channels,
+                                        in_channels,
+                                        kernel_size=1,
+                                        stride=1,
+                                        padding=0)
+
+    def forward(self, x):
+        h_ = x
+        h_ = self.norm(h_)
+        q = self.q(h_)
+        k = self.k(h_)
+        v = self.v(h_)
+
+        # compute attention
+        b,c,h,w = q.shape
+        q = q.reshape(b,c,h*w)
+        q = q.permute(0,2,1)   # b,hw,c
+        k = k.reshape(b,c,h*w) # b,c,hw
+        w_ = torch.bmm(q,k)     # b,hw,hw    w[b,i,j]=sum_c q[b,i,c]k[b,c,j]
+        w_ = w_ * (int(c)**(-0.5))
+        w_ = torch.nn.functional.softmax(w_, dim=2)
+
+        # attend to values
+        v = v.reshape(b,c,h*w)
+        w_ = w_.permute(0,2,1)   # b,hw,hw (first hw of k, second of q)
+        h_ = torch.bmm(v,w_)     # b, c,hw (hw of q) h_[b,c,j] = sum_i v[b,c,i] w_[b,i,j]
+        h_ = h_.reshape(b,c,h,w)
+
+        h_ = self.proj_out(h_)
+
+        return x+h_
+
+class MemoryEfficientAttnBlock(nn.Module):
+    """
+        Uses xformers efficient implementation,
+        see https://github.com/MatthieuTPHR/diffusers/blob/d80b531ff8060ec1ea982b65a1b8df70f73aa67c/src/diffusers/models/attention.py#L223
+        Note: this is a single-head self-attention operation
+    """
+    #
+    def __init__(self, in_channels):
+        super().__init__()
+        self.in_channels = in_channels
+
+        self.norm = Normalize(in_channels)
+        self.q = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.k = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.v = torch.nn.Conv2d(in_channels,
+                                 in_channels,
+                                 kernel_size=1,
+                                 stride=1,
+                                 padding=0)
+        self.proj_out = torch.nn.Conv2d(in_channels,
+                                        in_channels,
+                                        kernel_size=1,
+                                        stride=1,
+                                        padding=0)
+        self.attention_op: Optional[Any] = None
+
+    def forward(self, x):
+        h_ = x
+        h_ = self.norm(h_)
+        q = self.q(h_)
+        k = self.k(h_)
+        v = self.v(h_)
+
+        # compute attention
+        B, C, H, W = q.shape
+        q, k, v = map(lambda x: rearrange(x, 'b c h w -> b (h w) c'), (q, k, v))
+
+        q, k, v = map(
+            lambda t: t.unsqueeze(3)
+            .reshape(B, t.shape[1], 1, C)
+            .permute(0, 2, 1, 3)
+            .reshape(B * 1, t.shape[1], C)
+            .contiguous(),
+            (q, k, v),
+        )
+        out = xformers_state.xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=self.attention_op)
+
+        out = (
+            out.unsqueeze(0)
+            .reshape(B, 1, out.shape[1], C)
+            .permute(0, 2, 1, 3)
+            .reshape(B, out.shape[1], C)
+        )
+        out = rearrange(out, 'b (h w) c -> b c h w', b=B, h=H, w=W, c=C)
+        out = self.proj_out(out)
+        return x+out
+
+
+class MemoryEfficientCrossAttentionWrapper(MemoryEfficientCrossAttention):
+    def forward(self, x, context=None, mask=None):
+        b, c, h, w = x.shape
+        x = rearrange(x, 'b c h w -> b (h w) c')
+        out = super().forward(x, context=context, mask=mask)
+        out = rearrange(out, 'b (h w) c -> b c h w', h=h, w=w, c=c)
+        return x + out
+
+
+def make_attn(in_channels, attn_type="vanilla", attn_kwargs=None):
+    assert attn_type in ["vanilla", "vanilla-xformers", "memory-efficient-cross-attn", "linear", "none"], f'attn_type {attn_type} unknown'
+    # if XFORMERS_IS_AVAILBLE and attn_type == "vanilla":
+    if xformers_state.is_xformers_available() and attn_type == "vanilla":
+        attn_type = "vanilla-xformers"
+    print(f"making attention of type '{attn_type}' with {in_channels} in_channels")
+    if attn_type == "vanilla":
+        assert attn_kwargs is None
+        return AttnBlock(in_channels)
+    elif attn_type == "vanilla-xformers":
+        print(f"building MemoryEfficientAttnBlock with {in_channels} in_channels...")
+        return MemoryEfficientAttnBlock(in_channels)
+    elif type == "memory-efficient-cross-attn":
+        attn_kwargs["query_dim"] = in_channels
+        return MemoryEfficientCrossAttentionWrapper(**attn_kwargs)
+    elif attn_type == "none":
+        return nn.Identity(in_channels)
+    else:
+        raise NotImplementedError()
+
+
+class Model(nn.Module):
+    def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+                 attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+                 resolution, use_timestep=True, use_linear_attn=False, attn_type="vanilla"):
+        super().__init__()
+        if use_linear_attn: attn_type = "linear"
+        self.ch = ch
+        self.temb_ch = self.ch*4
+        self.num_resolutions = len(ch_mult)
+        self.num_res_blocks = num_res_blocks
+        self.resolution = resolution
+        self.in_channels = in_channels
+
+        self.use_timestep = use_timestep
+        if self.use_timestep:
+            # timestep embedding
+            self.temb = nn.Module()
+            self.temb.dense = nn.ModuleList([
+                torch.nn.Linear(self.ch,
+                                self.temb_ch),
+                torch.nn.Linear(self.temb_ch,
+                                self.temb_ch),
+            ])
+
+        # downsampling
+        self.conv_in = torch.nn.Conv2d(in_channels,
+                                       self.ch,
+                                       kernel_size=3,
+                                       stride=1,
+                                       padding=1)
+
+        curr_res = resolution
+        in_ch_mult = (1,)+tuple(ch_mult)
+        self.down = nn.ModuleList()
+        for i_level in range(self.num_resolutions):
+            block = nn.ModuleList()
+            attn = nn.ModuleList()
+            block_in = ch*in_ch_mult[i_level]
+            block_out = ch*ch_mult[i_level]
+            for i_block in range(self.num_res_blocks):
+                block.append(ResnetBlock(in_channels=block_in,
+                                         out_channels=block_out,
+                                         temb_channels=self.temb_ch,
+                                         dropout=dropout))
+                block_in = block_out
+                if curr_res in attn_resolutions:
+                    attn.append(make_attn(block_in, attn_type=attn_type))
+            down = nn.Module()
+            down.block = block
+            down.attn = attn
+            if i_level != self.num_resolutions-1:
+                down.downsample = Downsample(block_in, resamp_with_conv)
+                curr_res = curr_res // 2
+            self.down.append(down)
+
+        # middle
+        self.mid = nn.Module()
+        self.mid.block_1 = ResnetBlock(in_channels=block_in,
+                                       out_channels=block_in,
+                                       temb_channels=self.temb_ch,
+                                       dropout=dropout)
+        self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
+        self.mid.block_2 = ResnetBlock(in_channels=block_in,
+                                       out_channels=block_in,
+                                       temb_channels=self.temb_ch,
+                                       dropout=dropout)
+
+        # upsampling
+        self.up = nn.ModuleList()
+        for i_level in reversed(range(self.num_resolutions)):
+            block = nn.ModuleList()
+            attn = nn.ModuleList()
+            block_out = ch*ch_mult[i_level]
+            skip_in = ch*ch_mult[i_level]
+            for i_block in range(self.num_res_blocks+1):
+                if i_block == self.num_res_blocks:
+                    skip_in = ch*in_ch_mult[i_level]
+                block.append(ResnetBlock(in_channels=block_in+skip_in,
+                                         out_channels=block_out,
+                                         temb_channels=self.temb_ch,
+                                         dropout=dropout))
+                block_in = block_out
+                if curr_res in attn_resolutions:
+                    attn.append(make_attn(block_in, attn_type=attn_type))
+            up = nn.Module()
+            up.block = block
+            up.attn = attn
+            if i_level != 0:
+                up.upsample = Upsample(block_in, resamp_with_conv)
+                curr_res = curr_res * 2
+            self.up.insert(0, up) # prepend to get consistent order
+
+        # end
+        self.norm_out = Normalize(block_in)
+        self.conv_out = torch.nn.Conv2d(block_in,
+                                        out_ch,
+                                        kernel_size=3,
+                                        stride=1,
+                                        padding=1)
+
+    def forward(self, x, t=None, context=None):
+        #assert x.shape[2] == x.shape[3] == self.resolution
+        if context is not None:
+            # assume aligned context, cat along channel axis
+            x = torch.cat((x, context), dim=1)
+        if self.use_timestep:
+            # timestep embedding
+            assert t is not None
+            temb = get_timestep_embedding(t, self.ch)
+            temb = self.temb.dense[0](temb)
+            temb = nonlinearity(temb)
+            temb = self.temb.dense[1](temb)
+        else:
+            temb = None
+
+        # downsampling
+        hs = [self.conv_in(x)]
+        for i_level in range(self.num_resolutions):
+            for i_block in range(self.num_res_blocks):
+                h = self.down[i_level].block[i_block](hs[-1], temb)
+                if len(self.down[i_level].attn) > 0:
+                    h = self.down[i_level].attn[i_block](h)
+                hs.append(h)
+            if i_level != self.num_resolutions-1:
+                hs.append(self.down[i_level].downsample(hs[-1]))
+
+        # middle
+        h = hs[-1]
+        h = self.mid.block_1(h, temb)
+        h = self.mid.attn_1(h)
+        h = self.mid.block_2(h, temb)
+
+        # upsampling
+        for i_level in reversed(range(self.num_resolutions)):
+            for i_block in range(self.num_res_blocks+1):
+                h = self.up[i_level].block[i_block](
+                    torch.cat([h, hs.pop()], dim=1), temb)
+                if len(self.up[i_level].attn) > 0:
+                    h = self.up[i_level].attn[i_block](h)
+            if i_level != 0:
+                h = self.up[i_level].upsample(h)
+
+        # end
+        h = self.norm_out(h)
+        h = nonlinearity(h)
+        h = self.conv_out(h)
+        return h
+
+    def get_last_layer(self):
+        return self.conv_out.weight
+
+
+class Encoder(nn.Module):
+    def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+                 attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+                 resolution, z_channels, double_z=True, use_linear_attn=False, attn_type="vanilla",
+                 **ignore_kwargs):
+        super().__init__()
+        if use_linear_attn: attn_type = "linear"
+        self.ch = ch
+        self.temb_ch = 0
+        self.num_resolutions = len(ch_mult)
+        self.num_res_blocks = num_res_blocks
+        self.resolution = resolution
+        self.in_channels = in_channels
+
+        # downsampling
+        self.conv_in = torch.nn.Conv2d(in_channels,
+                                       self.ch,
+                                       kernel_size=3,
+                                       stride=1,
+                                       padding=1)
+
+        curr_res = resolution
+        in_ch_mult = (1,)+tuple(ch_mult)
+        self.in_ch_mult = in_ch_mult
+        self.down = nn.ModuleList()
+        for i_level in range(self.num_resolutions):
+            block = nn.ModuleList()
+            attn = nn.ModuleList()
+            block_in = ch*in_ch_mult[i_level]
+            block_out = ch*ch_mult[i_level]
+            for i_block in range(self.num_res_blocks):
+                block.append(ResnetBlock(in_channels=block_in,
+                                         out_channels=block_out,
+                                         temb_channels=self.temb_ch,
+                                         dropout=dropout))
+                block_in = block_out
+                if curr_res in attn_resolutions:
+                    attn.append(make_attn(block_in, attn_type=attn_type))
+            down = nn.Module()
+            down.block = block
+            down.attn = attn
+            if i_level != self.num_resolutions-1:
+                down.downsample = Downsample(block_in, resamp_with_conv)
+                curr_res = curr_res // 2
+            self.down.append(down)
+
+        # middle
+        self.mid = nn.Module()
+        self.mid.block_1 = ResnetBlock(in_channels=block_in,
+                                       out_channels=block_in,
+                                       temb_channels=self.temb_ch,
+                                       dropout=dropout)
+        self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
+        self.mid.block_2 = ResnetBlock(in_channels=block_in,
+                                       out_channels=block_in,
+                                       temb_channels=self.temb_ch,
+                                       dropout=dropout)
+
+        # end
+        self.norm_out = Normalize(block_in)
+        self.conv_out = torch.nn.Conv2d(block_in,
+                                        2*z_channels if double_z else z_channels,
+                                        kernel_size=3,
+                                        stride=1,
+                                        padding=1)
+
+    def forward(self, x):
+        # timestep embedding
+        temb = None
+
+        # downsampling
+        hs = [self.conv_in(x)]
+        for i_level in range(self.num_resolutions):
+            for i_block in range(self.num_res_blocks):
+                h = self.down[i_level].block[i_block](hs[-1], temb)
+                if len(self.down[i_level].attn) > 0:
+                    h = self.down[i_level].attn[i_block](h)
+                hs.append(h)
+            if i_level != self.num_resolutions-1:
+                hs.append(self.down[i_level].downsample(hs[-1]))
+
+        # middle
+        h = hs[-1]
+        h = self.mid.block_1(h, temb)
+        h = self.mid.attn_1(h)
+        h = self.mid.block_2(h, temb)
+
+        # end
+        h = self.norm_out(h)
+        h = nonlinearity(h)
+        h = self.conv_out(h)
+        return h
+
+
+class Decoder(nn.Module):
+    def __init__(self, *, ch, out_ch, ch_mult=(1,2,4,8), num_res_blocks,
+                 attn_resolutions, dropout=0.0, resamp_with_conv=True, in_channels,
+                 resolution, z_channels, give_pre_end=False, tanh_out=False, use_linear_attn=False,
+                 attn_type="vanilla", **ignorekwargs):
+        super().__init__()
+        if use_linear_attn: attn_type = "linear"
+        self.ch = ch
+        self.temb_ch = 0
+        self.num_resolutions = len(ch_mult)
+        self.num_res_blocks = num_res_blocks
+        self.resolution = resolution
+        self.in_channels = in_channels
+        self.give_pre_end = give_pre_end
+        self.tanh_out = tanh_out
+
+        # compute in_ch_mult, block_in and curr_res at lowest res
+        in_ch_mult = (1,)+tuple(ch_mult)
+        block_in = ch*ch_mult[self.num_resolutions-1]
+        curr_res = resolution // 2**(self.num_resolutions-1)
+        self.z_shape = (1,z_channels,curr_res,curr_res)
+        print("Working with z of shape {} = {} dimensions.".format(
+            self.z_shape, np.prod(self.z_shape)))
+
+        # z to block_in
+        self.conv_in = torch.nn.Conv2d(z_channels,
+                                       block_in,
+                                       kernel_size=3,
+                                       stride=1,
+                                       padding=1)
+
+        # middle
+        self.mid = nn.Module()
+        self.mid.block_1 = ResnetBlock(in_channels=block_in,
+                                       out_channels=block_in,
+                                       temb_channels=self.temb_ch,
+                                       dropout=dropout)
+        self.mid.attn_1 = make_attn(block_in, attn_type=attn_type)
+        self.mid.block_2 = ResnetBlock(in_channels=block_in,
+                                       out_channels=block_in,
+                                       temb_channels=self.temb_ch,
+                                       dropout=dropout)
+
+        # upsampling
+        self.up = nn.ModuleList()
+        for i_level in reversed(range(self.num_resolutions)):
+            block = nn.ModuleList()
+            attn = nn.ModuleList()
+            block_out = ch*ch_mult[i_level]
+            for i_block in range(self.num_res_blocks+1):
+                block.append(ResnetBlock(in_channels=block_in,
+                                         out_channels=block_out,
+                                         temb_channels=self.temb_ch,
+                                         dropout=dropout))
+                block_in = block_out
+                if curr_res in attn_resolutions:
+                    attn.append(make_attn(block_in, attn_type=attn_type))
+            up = nn.Module()
+            up.block = block
+            up.attn = attn
+            if i_level != 0:
+                up.upsample = Upsample(block_in, resamp_with_conv)
+                curr_res = curr_res * 2
+            self.up.insert(0, up) # prepend to get consistent order
+
+        # end
+        self.norm_out = Normalize(block_in)
+        self.conv_out = torch.nn.Conv2d(block_in,
+                                        out_ch,
+                                        kernel_size=3,
+                                        stride=1,
+                                        padding=1)
+
+    def forward(self, z):
+        #assert z.shape[1:] == self.z_shape[1:]
+        self.last_z_shape = z.shape
+
+        # timestep embedding
+        temb = None
+
+        # z to block_in
+        h = self.conv_in(z)
+
+        # middle
+        h = self.mid.block_1(h, temb)
+        h = self.mid.attn_1(h)
+        h = self.mid.block_2(h, temb)
+
+        # upsampling
+        for i_level in reversed(range(self.num_resolutions)):
+            for i_block in range(self.num_res_blocks+1):
+                h = self.up[i_level].block[i_block](h, temb)
+                if len(self.up[i_level].attn) > 0:
+                    h = self.up[i_level].attn[i_block](h)
+            if i_level != 0:
+                h = self.up[i_level].upsample(h)
+
+        # end
+        if self.give_pre_end:
+            return h
+
+        h = self.norm_out(h)
+        h = nonlinearity(h)
+        h = self.conv_out(h)
+        if self.tanh_out:
+            h = torch.tanh(h)
+        return h
+
+
+class SimpleDecoder(nn.Module):
+    def __init__(self, in_channels, out_channels, *args, **kwargs):
+        super().__init__()
+        self.model = nn.ModuleList([nn.Conv2d(in_channels, in_channels, 1),
+                                     ResnetBlock(in_channels=in_channels,
+                                                 out_channels=2 * in_channels,
+                                                 temb_channels=0, dropout=0.0),
+                                     ResnetBlock(in_channels=2 * in_channels,
+                                                out_channels=4 * in_channels,
+                                                temb_channels=0, dropout=0.0),
+                                     ResnetBlock(in_channels=4 * in_channels,
+                                                out_channels=2 * in_channels,
+                                                temb_channels=0, dropout=0.0),
+                                     nn.Conv2d(2*in_channels, in_channels, 1),
+                                     Upsample(in_channels, with_conv=True)])
+        # end
+        self.norm_out = Normalize(in_channels)
+        self.conv_out = torch.nn.Conv2d(in_channels,
+                                        out_channels,
+                                        kernel_size=3,
+                                        stride=1,
+                                        padding=1)
+
+    def forward(self, x):
+        for i, layer in enumerate(self.model):
+            if i in [1,2,3]:
+                x = layer(x, None)
+            else:
+                x = layer(x)
+
+        h = self.norm_out(x)
+        h = nonlinearity(h)
+        x = self.conv_out(h)
+        return x
+
+
+class UpsampleDecoder(nn.Module):
+    def __init__(self, in_channels, out_channels, ch, num_res_blocks, resolution,
+                 ch_mult=(2,2), dropout=0.0):
+        super().__init__()
+        # upsampling
+        self.temb_ch = 0
+        self.num_resolutions = len(ch_mult)
+        self.num_res_blocks = num_res_blocks
+        block_in = in_channels
+        curr_res = resolution // 2 ** (self.num_resolutions - 1)
+        self.res_blocks = nn.ModuleList()
+        self.upsample_blocks = nn.ModuleList()
+        for i_level in range(self.num_resolutions):
+            res_block = []
+            block_out = ch * ch_mult[i_level]
+            for i_block in range(self.num_res_blocks + 1):
+                res_block.append(ResnetBlock(in_channels=block_in,
+                                         out_channels=block_out,
+                                         temb_channels=self.temb_ch,
+                                         dropout=dropout))
+                block_in = block_out
+            self.res_blocks.append(nn.ModuleList(res_block))
+            if i_level != self.num_resolutions - 1:
+                self.upsample_blocks.append(Upsample(block_in, True))
+                curr_res = curr_res * 2
+
+        # end
+        self.norm_out = Normalize(block_in)
+        self.conv_out = torch.nn.Conv2d(block_in,
+                                        out_channels,
+                                        kernel_size=3,
+                                        stride=1,
+                                        padding=1)
+
+    def forward(self, x):
+        # upsampling
+        h = x
+        for k, i_level in enumerate(range(self.num_resolutions)):
+            for i_block in range(self.num_res_blocks + 1):
+                h = self.res_blocks[i_level][i_block](h, None)
+            if i_level != self.num_resolutions - 1:
+                h = self.upsample_blocks[k](h)
+        h = self.norm_out(h)
+        h = nonlinearity(h)
+        h = self.conv_out(h)
+        return h
+
+
+class LatentRescaler(nn.Module):
+    def __init__(self, factor, in_channels, mid_channels, out_channels, depth=2):
+        super().__init__()
+        # residual block, interpolate, residual block
+        self.factor = factor
+        self.conv_in = nn.Conv2d(in_channels,
+                                 mid_channels,
+                                 kernel_size=3,
+                                 stride=1,
+                                 padding=1)
+        self.res_block1 = nn.ModuleList([ResnetBlock(in_channels=mid_channels,
+                                                     out_channels=mid_channels,
+                                                     temb_channels=0,
+                                                     dropout=0.0) for _ in range(depth)])
+        self.attn = AttnBlock(mid_channels)
+        self.res_block2 = nn.ModuleList([ResnetBlock(in_channels=mid_channels,
+                                                     out_channels=mid_channels,
+                                                     temb_channels=0,
+                                                     dropout=0.0) for _ in range(depth)])
+
+        self.conv_out = nn.Conv2d(mid_channels,
+                                  out_channels,
+                                  kernel_size=1,
+                                  )
+
+    def forward(self, x):
+        x = self.conv_in(x)
+        for block in self.res_block1:
+            x = block(x, None)
+        x = torch.nn.functional.interpolate(x, size=(int(round(x.shape[2]*self.factor)), int(round(x.shape[3]*self.factor))))
+        x = self.attn(x)
+        for block in self.res_block2:
+            x = block(x, None)
+        x = self.conv_out(x)
+        return x
+
+
+class MergedRescaleEncoder(nn.Module):
+    def __init__(self, in_channels, ch, resolution, out_ch, num_res_blocks,
+                 attn_resolutions, dropout=0.0, resamp_with_conv=True,
+                 ch_mult=(1,2,4,8), rescale_factor=1.0, rescale_module_depth=1):
+        super().__init__()
+        intermediate_chn = ch * ch_mult[-1]
+        self.encoder = Encoder(in_channels=in_channels, num_res_blocks=num_res_blocks, ch=ch, ch_mult=ch_mult,
+                               z_channels=intermediate_chn, double_z=False, resolution=resolution,
+                               attn_resolutions=attn_resolutions, dropout=dropout, resamp_with_conv=resamp_with_conv,
+                               out_ch=None)
+        self.rescaler = LatentRescaler(factor=rescale_factor, in_channels=intermediate_chn,
+                                       mid_channels=intermediate_chn, out_channels=out_ch, depth=rescale_module_depth)
+
+    def forward(self, x):
+        x = self.encoder(x)
+        x = self.rescaler(x)
+        return x
+
+
+class MergedRescaleDecoder(nn.Module):
+    def __init__(self, z_channels, out_ch, resolution, num_res_blocks, attn_resolutions, ch, ch_mult=(1,2,4,8),
+                 dropout=0.0, resamp_with_conv=True, rescale_factor=1.0, rescale_module_depth=1):
+        super().__init__()
+        tmp_chn = z_channels*ch_mult[-1]
+        self.decoder = Decoder(out_ch=out_ch, z_channels=tmp_chn, attn_resolutions=attn_resolutions, dropout=dropout,
+                               resamp_with_conv=resamp_with_conv, in_channels=None, num_res_blocks=num_res_blocks,
+                               ch_mult=ch_mult, resolution=resolution, ch=ch)
+        self.rescaler = LatentRescaler(factor=rescale_factor, in_channels=z_channels, mid_channels=tmp_chn,
+                                       out_channels=tmp_chn, depth=rescale_module_depth)
+
+    def forward(self, x):
+        x = self.rescaler(x)
+        x = self.decoder(x)
+        return x
+
+
+class Upsampler(nn.Module):
+    def __init__(self, in_size, out_size, in_channels, out_channels, ch_mult=2):
+        super().__init__()
+        assert out_size >= in_size
+        num_blocks = int(np.log2(out_size//in_size))+1
+        factor_up = 1.+ (out_size % in_size)
+        print(f"Building {self.__class__.__name__} with in_size: {in_size} --> out_size {out_size} and factor {factor_up}")
+        self.rescaler = LatentRescaler(factor=factor_up, in_channels=in_channels, mid_channels=2*in_channels,
+                                       out_channels=in_channels)
+        self.decoder = Decoder(out_ch=out_channels, resolution=out_size, z_channels=in_channels, num_res_blocks=2,
+                               attn_resolutions=[], in_channels=None, ch=in_channels,
+                               ch_mult=[ch_mult for _ in range(num_blocks)])
+
+    def forward(self, x):
+        x = self.rescaler(x)
+        x = self.decoder(x)
+        return x
+
+
+class Resize(nn.Module):
+    def __init__(self, in_channels=None, learned=False, mode="bilinear"):
+        super().__init__()
+        self.with_conv = learned
+        self.mode = mode
+        if self.with_conv:
+            print(f"Note: {self.__class__.__name} uses learned downsampling and will ignore the fixed {mode} mode")
+            raise NotImplementedError()
+            assert in_channels is not None
+            # no asymmetric padding in torch conv, must do it ourselves
+            self.conv = torch.nn.Conv2d(in_channels,
+                                        in_channels,
+                                        kernel_size=4,
+                                        stride=2,
+                                        padding=1)
+
+    def forward(self, x, scale_factor=1.0):
+        if scale_factor==1.0:
+            return x
+        else:
+            x = torch.nn.functional.interpolate(x, mode=self.mode, align_corners=False, scale_factor=scale_factor)
+        return x
diff --git a/ldm/modules/diffusionmodules/openaimodel.py b/ldm/modules/diffusionmodules/openaimodel.py
new file mode 100644
index 0000000..7df6b5a
--- /dev/null
+++ b/ldm/modules/diffusionmodules/openaimodel.py
@@ -0,0 +1,786 @@
+from abc import abstractmethod
+import math
+
+import numpy as np
+import torch as th
+import torch.nn as nn
+import torch.nn.functional as F
+
+from ldm.modules.diffusionmodules.util import (
+    checkpoint,
+    conv_nd,
+    linear,
+    avg_pool_nd,
+    zero_module,
+    normalization,
+    timestep_embedding,
+)
+from ldm.modules.attention import SpatialTransformer
+from ldm.util import exists
+
+
+# dummy replace
+def convert_module_to_f16(x):
+    pass
+
+def convert_module_to_f32(x):
+    pass
+
+
+## go
+class AttentionPool2d(nn.Module):
+    """
+    Adapted from CLIP: https://github.com/openai/CLIP/blob/main/clip/model.py
+    """
+
+    def __init__(
+        self,
+        spacial_dim: int,
+        embed_dim: int,
+        num_heads_channels: int,
+        output_dim: int = None,
+    ):
+        super().__init__()
+        self.positional_embedding = nn.Parameter(th.randn(embed_dim, spacial_dim ** 2 + 1) / embed_dim ** 0.5)
+        self.qkv_proj = conv_nd(1, embed_dim, 3 * embed_dim, 1)
+        self.c_proj = conv_nd(1, embed_dim, output_dim or embed_dim, 1)
+        self.num_heads = embed_dim // num_heads_channels
+        self.attention = QKVAttention(self.num_heads)
+
+    def forward(self, x):
+        b, c, *_spatial = x.shape
+        x = x.reshape(b, c, -1)  # NC(HW)
+        x = th.cat([x.mean(dim=-1, keepdim=True), x], dim=-1)  # NC(HW+1)
+        x = x + self.positional_embedding[None, :, :].to(x.dtype)  # NC(HW+1)
+        x = self.qkv_proj(x)
+        x = self.attention(x)
+        x = self.c_proj(x)
+        return x[:, :, 0]
+
+
+class TimestepBlock(nn.Module):
+    """
+    Any module where forward() takes timestep embeddings as a second argument.
+    """
+
+    @abstractmethod
+    def forward(self, x, emb):
+        """
+        Apply the module to `x` given `emb` timestep embeddings.
+        """
+
+
+class TimestepEmbedSequential(nn.Sequential, TimestepBlock):
+    """
+    A sequential module that passes timestep embeddings to the children that
+    support it as an extra input.
+    """
+
+    def forward(self, x, emb, context=None):
+        for layer in self:
+            if isinstance(layer, TimestepBlock):
+                x = layer(x, emb)
+            elif isinstance(layer, SpatialTransformer):
+                x = layer(x, context)
+            else:
+                x = layer(x)
+        return x
+
+
+class Upsample(nn.Module):
+    """
+    An upsampling layer with an optional convolution.
+    :param channels: channels in the inputs and outputs.
+    :param use_conv: a bool determining if a convolution is applied.
+    :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
+                 upsampling occurs in the inner-two dimensions.
+    """
+
+    def __init__(self, channels, use_conv, dims=2, out_channels=None, padding=1):
+        super().__init__()
+        self.channels = channels
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.dims = dims
+        if use_conv:
+            self.conv = conv_nd(dims, self.channels, self.out_channels, 3, padding=padding)
+
+    def forward(self, x):
+        assert x.shape[1] == self.channels
+        if self.dims == 3:
+            x = F.interpolate(
+                x, (x.shape[2], x.shape[3] * 2, x.shape[4] * 2), mode="nearest"
+            )
+        else:
+            x = F.interpolate(x, scale_factor=2, mode="nearest")
+        if self.use_conv:
+            x = self.conv(x)
+        return x
+
+class TransposedUpsample(nn.Module):
+    'Learned 2x upsampling without padding'
+    def __init__(self, channels, out_channels=None, ks=5):
+        super().__init__()
+        self.channels = channels
+        self.out_channels = out_channels or channels
+
+        self.up = nn.ConvTranspose2d(self.channels,self.out_channels,kernel_size=ks,stride=2)
+
+    def forward(self,x):
+        return self.up(x)
+
+
+class Downsample(nn.Module):
+    """
+    A downsampling layer with an optional convolution.
+    :param channels: channels in the inputs and outputs.
+    :param use_conv: a bool determining if a convolution is applied.
+    :param dims: determines if the signal is 1D, 2D, or 3D. If 3D, then
+                 downsampling occurs in the inner-two dimensions.
+    """
+
+    def __init__(self, channels, use_conv, dims=2, out_channels=None,padding=1):
+        super().__init__()
+        self.channels = channels
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.dims = dims
+        stride = 2 if dims != 3 else (1, 2, 2)
+        if use_conv:
+            self.op = conv_nd(
+                dims, self.channels, self.out_channels, 3, stride=stride, padding=padding
+            )
+        else:
+            assert self.channels == self.out_channels
+            self.op = avg_pool_nd(dims, kernel_size=stride, stride=stride)
+
+    def forward(self, x):
+        assert x.shape[1] == self.channels
+        return self.op(x)
+
+
+class ResBlock(TimestepBlock):
+    """
+    A residual block that can optionally change the number of channels.
+    :param channels: the number of input channels.
+    :param emb_channels: the number of timestep embedding channels.
+    :param dropout: the rate of dropout.
+    :param out_channels: if specified, the number of out channels.
+    :param use_conv: if True and out_channels is specified, use a spatial
+        convolution instead of a smaller 1x1 convolution to change the
+        channels in the skip connection.
+    :param dims: determines if the signal is 1D, 2D, or 3D.
+    :param use_checkpoint: if True, use gradient checkpointing on this module.
+    :param up: if True, use this block for upsampling.
+    :param down: if True, use this block for downsampling.
+    """
+
+    def __init__(
+        self,
+        channels,
+        emb_channels,
+        dropout,
+        out_channels=None,
+        use_conv=False,
+        use_scale_shift_norm=False,
+        dims=2,
+        use_checkpoint=False,
+        up=False,
+        down=False,
+    ):
+        super().__init__()
+        self.channels = channels
+        self.emb_channels = emb_channels
+        self.dropout = dropout
+        self.out_channels = out_channels or channels
+        self.use_conv = use_conv
+        self.use_checkpoint = use_checkpoint
+        self.use_scale_shift_norm = use_scale_shift_norm
+
+        self.in_layers = nn.Sequential(
+            normalization(channels),
+            nn.SiLU(),
+            conv_nd(dims, channels, self.out_channels, 3, padding=1),
+        )
+
+        self.updown = up or down
+
+        if up:
+            self.h_upd = Upsample(channels, False, dims)
+            self.x_upd = Upsample(channels, False, dims)
+        elif down:
+            self.h_upd = Downsample(channels, False, dims)
+            self.x_upd = Downsample(channels, False, dims)
+        else:
+            self.h_upd = self.x_upd = nn.Identity()
+
+        self.emb_layers = nn.Sequential(
+            nn.SiLU(),
+            linear(
+                emb_channels,
+                2 * self.out_channels if use_scale_shift_norm else self.out_channels,
+            ),
+        )
+        self.out_layers = nn.Sequential(
+            normalization(self.out_channels),
+            nn.SiLU(),
+            nn.Dropout(p=dropout),
+            zero_module(
+                conv_nd(dims, self.out_channels, self.out_channels, 3, padding=1)
+            ),
+        )
+
+        if self.out_channels == channels:
+            self.skip_connection = nn.Identity()
+        elif use_conv:
+            self.skip_connection = conv_nd(
+                dims, channels, self.out_channels, 3, padding=1
+            )
+        else:
+            self.skip_connection = conv_nd(dims, channels, self.out_channels, 1)
+
+    def forward(self, x, emb):
+        """
+        Apply the block to a Tensor, conditioned on a timestep embedding.
+        :param x: an [N x C x ...] Tensor of features.
+        :param emb: an [N x emb_channels] Tensor of timestep embeddings.
+        :return: an [N x C x ...] Tensor of outputs.
+        """
+        return checkpoint(
+            self._forward, (x, emb), self.parameters(), self.use_checkpoint
+        )
+
+
+    def _forward(self, x, emb):
+        if self.updown:
+            in_rest, in_conv = self.in_layers[:-1], self.in_layers[-1]
+            h = in_rest(x)
+            h = self.h_upd(h)
+            x = self.x_upd(x)
+            h = in_conv(h)
+        else:
+            h = self.in_layers(x)
+        emb_out = self.emb_layers(emb).type(h.dtype)
+        while len(emb_out.shape) < len(h.shape):
+            emb_out = emb_out[..., None]
+        if self.use_scale_shift_norm:
+            out_norm, out_rest = self.out_layers[0], self.out_layers[1:]
+            scale, shift = th.chunk(emb_out, 2, dim=1)
+            h = out_norm(h) * (1 + scale) + shift
+            h = out_rest(h)
+        else:
+            h = h + emb_out
+            h = self.out_layers(h)
+        return self.skip_connection(x) + h
+
+
+class AttentionBlock(nn.Module):
+    """
+    An attention block that allows spatial positions to attend to each other.
+    Originally ported from here, but adapted to the N-d case.
+    https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/models/unet.py#L66.
+    """
+
+    def __init__(
+        self,
+        channels,
+        num_heads=1,
+        num_head_channels=-1,
+        use_checkpoint=False,
+        use_new_attention_order=False,
+    ):
+        super().__init__()
+        self.channels = channels
+        if num_head_channels == -1:
+            self.num_heads = num_heads
+        else:
+            assert (
+                channels % num_head_channels == 0
+            ), f"q,k,v channels {channels} is not divisible by num_head_channels {num_head_channels}"
+            self.num_heads = channels // num_head_channels
+        self.use_checkpoint = use_checkpoint
+        self.norm = normalization(channels)
+        self.qkv = conv_nd(1, channels, channels * 3, 1)
+        if use_new_attention_order:
+            # split qkv before split heads
+            self.attention = QKVAttention(self.num_heads)
+        else:
+            # split heads before split qkv
+            self.attention = QKVAttentionLegacy(self.num_heads)
+
+        self.proj_out = zero_module(conv_nd(1, channels, channels, 1))
+
+    def forward(self, x):
+        return checkpoint(self._forward, (x,), self.parameters(), True)   # TODO: check checkpoint usage, is True # TODO: fix the .half call!!!
+        #return pt_checkpoint(self._forward, x)  # pytorch
+
+    def _forward(self, x):
+        b, c, *spatial = x.shape
+        x = x.reshape(b, c, -1)
+        qkv = self.qkv(self.norm(x))
+        h = self.attention(qkv)
+        h = self.proj_out(h)
+        return (x + h).reshape(b, c, *spatial)
+
+
+def count_flops_attn(model, _x, y):
+    """
+    A counter for the `thop` package to count the operations in an
+    attention operation.
+    Meant to be used like:
+        macs, params = thop.profile(
+            model,
+            inputs=(inputs, timestamps),
+            custom_ops={QKVAttention: QKVAttention.count_flops},
+        )
+    """
+    b, c, *spatial = y[0].shape
+    num_spatial = int(np.prod(spatial))
+    # We perform two matmuls with the same number of ops.
+    # The first computes the weight matrix, the second computes
+    # the combination of the value vectors.
+    matmul_ops = 2 * b * (num_spatial ** 2) * c
+    model.total_ops += th.DoubleTensor([matmul_ops])
+
+
+class QKVAttentionLegacy(nn.Module):
+    """
+    A module which performs QKV attention. Matches legacy QKVAttention + input/ouput heads shaping
+    """
+
+    def __init__(self, n_heads):
+        super().__init__()
+        self.n_heads = n_heads
+
+    def forward(self, qkv):
+        """
+        Apply QKV attention.
+        :param qkv: an [N x (H * 3 * C) x T] tensor of Qs, Ks, and Vs.
+        :return: an [N x (H * C) x T] tensor after attention.
+        """
+        bs, width, length = qkv.shape
+        assert width % (3 * self.n_heads) == 0
+        ch = width // (3 * self.n_heads)
+        q, k, v = qkv.reshape(bs * self.n_heads, ch * 3, length).split(ch, dim=1)
+        scale = 1 / math.sqrt(math.sqrt(ch))
+        weight = th.einsum(
+            "bct,bcs->bts", q * scale, k * scale
+        )  # More stable with f16 than dividing afterwards
+        weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
+        a = th.einsum("bts,bcs->bct", weight, v)
+        return a.reshape(bs, -1, length)
+
+    @staticmethod
+    def count_flops(model, _x, y):
+        return count_flops_attn(model, _x, y)
+
+
+class QKVAttention(nn.Module):
+    """
+    A module which performs QKV attention and splits in a different order.
+    """
+
+    def __init__(self, n_heads):
+        super().__init__()
+        self.n_heads = n_heads
+
+    def forward(self, qkv):
+        """
+        Apply QKV attention.
+        :param qkv: an [N x (3 * H * C) x T] tensor of Qs, Ks, and Vs.
+        :return: an [N x (H * C) x T] tensor after attention.
+        """
+        bs, width, length = qkv.shape
+        assert width % (3 * self.n_heads) == 0
+        ch = width // (3 * self.n_heads)
+        q, k, v = qkv.chunk(3, dim=1)
+        scale = 1 / math.sqrt(math.sqrt(ch))
+        weight = th.einsum(
+            "bct,bcs->bts",
+            (q * scale).view(bs * self.n_heads, ch, length),
+            (k * scale).view(bs * self.n_heads, ch, length),
+        )  # More stable with f16 than dividing afterwards
+        weight = th.softmax(weight.float(), dim=-1).type(weight.dtype)
+        a = th.einsum("bts,bcs->bct", weight, v.reshape(bs * self.n_heads, ch, length))
+        return a.reshape(bs, -1, length)
+
+    @staticmethod
+    def count_flops(model, _x, y):
+        return count_flops_attn(model, _x, y)
+
+
+class UNetModel(nn.Module):
+    """
+    The full UNet model with attention and timestep embedding.
+    :param in_channels: channels in the input Tensor.
+    :param model_channels: base channel count for the model.
+    :param out_channels: channels in the output Tensor.
+    :param num_res_blocks: number of residual blocks per downsample.
+    :param attention_resolutions: a collection of downsample rates at which
+        attention will take place. May be a set, list, or tuple.
+        For example, if this contains 4, then at 4x downsampling, attention
+        will be used.
+    :param dropout: the dropout probability.
+    :param channel_mult: channel multiplier for each level of the UNet.
+    :param conv_resample: if True, use learned convolutions for upsampling and
+        downsampling.
+    :param dims: determines if the signal is 1D, 2D, or 3D.
+    :param num_classes: if specified (as an int), then this model will be
+        class-conditional with `num_classes` classes.
+    :param use_checkpoint: use gradient checkpointing to reduce memory usage.
+    :param num_heads: the number of attention heads in each attention layer.
+    :param num_heads_channels: if specified, ignore num_heads and instead use
+                               a fixed channel width per attention head.
+    :param num_heads_upsample: works with num_heads to set a different number
+                               of heads for upsampling. Deprecated.
+    :param use_scale_shift_norm: use a FiLM-like conditioning mechanism.
+    :param resblock_updown: use residual blocks for up/downsampling.
+    :param use_new_attention_order: use a different attention pattern for potentially
+                                    increased efficiency.
+    """
+
+    def __init__(
+        self,
+        image_size,
+        in_channels,
+        model_channels,
+        out_channels,
+        num_res_blocks,
+        attention_resolutions,
+        dropout=0,
+        channel_mult=(1, 2, 4, 8),
+        conv_resample=True,
+        dims=2,
+        num_classes=None,
+        use_checkpoint=False,
+        use_fp16=False,
+        num_heads=-1,
+        num_head_channels=-1,
+        num_heads_upsample=-1,
+        use_scale_shift_norm=False,
+        resblock_updown=False,
+        use_new_attention_order=False,
+        use_spatial_transformer=False,    # custom transformer support
+        transformer_depth=1,              # custom transformer support
+        context_dim=None,                 # custom transformer support
+        n_embed=None,                     # custom support for prediction of discrete ids into codebook of first stage vq model
+        legacy=True,
+        disable_self_attentions=None,
+        num_attention_blocks=None,
+        disable_middle_self_attn=False,
+        use_linear_in_transformer=False,
+    ):
+        super().__init__()
+        if use_spatial_transformer:
+            assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...'
+
+        if context_dim is not None:
+            assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...'
+            from omegaconf.listconfig import ListConfig
+            if type(context_dim) == ListConfig:
+                context_dim = list(context_dim)
+
+        if num_heads_upsample == -1:
+            num_heads_upsample = num_heads
+
+        if num_heads == -1:
+            assert num_head_channels != -1, 'Either num_heads or num_head_channels has to be set'
+
+        if num_head_channels == -1:
+            assert num_heads != -1, 'Either num_heads or num_head_channels has to be set'
+
+        self.image_size = image_size
+        self.in_channels = in_channels
+        self.model_channels = model_channels
+        self.out_channels = out_channels
+        if isinstance(num_res_blocks, int):
+            self.num_res_blocks = len(channel_mult) * [num_res_blocks]
+        else:
+            if len(num_res_blocks) != len(channel_mult):
+                raise ValueError("provide num_res_blocks either as an int (globally constant) or "
+                                 "as a list/tuple (per-level) with the same length as channel_mult")
+            self.num_res_blocks = num_res_blocks
+        if disable_self_attentions is not None:
+            # should be a list of booleans, indicating whether to disable self-attention in TransformerBlocks or not
+            assert len(disable_self_attentions) == len(channel_mult)
+        if num_attention_blocks is not None:
+            assert len(num_attention_blocks) == len(self.num_res_blocks)
+            assert all(map(lambda i: self.num_res_blocks[i] >= num_attention_blocks[i], range(len(num_attention_blocks))))
+            print(f"Constructor of UNetModel received num_attention_blocks={num_attention_blocks}. "
+                  f"This option has LESS priority than attention_resolutions {attention_resolutions}, "
+                  f"i.e., in cases where num_attention_blocks[i] > 0 but 2**i not in attention_resolutions, "
+                  f"attention will still not be set.")
+
+        self.attention_resolutions = attention_resolutions
+        self.dropout = dropout
+        self.channel_mult = channel_mult
+        self.conv_resample = conv_resample
+        self.num_classes = num_classes
+        self.use_checkpoint = use_checkpoint
+        self.dtype = th.float16 if use_fp16 else th.float32
+        self.num_heads = num_heads
+        self.num_head_channels = num_head_channels
+        self.num_heads_upsample = num_heads_upsample
+        self.predict_codebook_ids = n_embed is not None
+
+        time_embed_dim = model_channels * 4
+        self.time_embed = nn.Sequential(
+            linear(model_channels, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+        if self.num_classes is not None:
+            if isinstance(self.num_classes, int):
+                self.label_emb = nn.Embedding(num_classes, time_embed_dim)
+            elif self.num_classes == "continuous":
+                print("setting up linear c_adm embedding layer")
+                self.label_emb = nn.Linear(1, time_embed_dim)
+            else:
+                raise ValueError()
+
+        self.input_blocks = nn.ModuleList(
+            [
+                TimestepEmbedSequential(
+                    conv_nd(dims, in_channels, model_channels, 3, padding=1)
+                )
+            ]
+        )
+        self._feature_size = model_channels
+        input_block_chans = [model_channels]
+        ch = model_channels
+        ds = 1
+        for level, mult in enumerate(channel_mult):
+            for nr in range(self.num_res_blocks[level]):
+                layers = [
+                    ResBlock(
+                        ch,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=mult * model_channels,
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = mult * model_channels
+                if ds in attention_resolutions:
+                    if num_head_channels == -1:
+                        dim_head = ch // num_heads
+                    else:
+                        num_heads = ch // num_head_channels
+                        dim_head = num_head_channels
+                    if legacy:
+                        #num_heads = 1
+                        dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+                    if exists(disable_self_attentions):
+                        disabled_sa = disable_self_attentions[level]
+                    else:
+                        disabled_sa = False
+
+                    if not exists(num_attention_blocks) or nr < num_attention_blocks[level]:
+                        layers.append(
+                            AttentionBlock(
+                                ch,
+                                use_checkpoint=use_checkpoint,
+                                num_heads=num_heads,
+                                num_head_channels=dim_head,
+                                use_new_attention_order=use_new_attention_order,
+                            ) if not use_spatial_transformer else SpatialTransformer(
+                                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                                disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer,
+                                use_checkpoint=use_checkpoint
+                            )
+                        )
+                self.input_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+                input_block_chans.append(ch)
+            if level != len(channel_mult) - 1:
+                out_ch = ch
+                self.input_blocks.append(
+                    TimestepEmbedSequential(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            down=True,
+                        )
+                        if resblock_updown
+                        else Downsample(
+                            ch, conv_resample, dims=dims, out_channels=out_ch
+                        )
+                    )
+                )
+                ch = out_ch
+                input_block_chans.append(ch)
+                ds *= 2
+                self._feature_size += ch
+
+        if num_head_channels == -1:
+            dim_head = ch // num_heads
+        else:
+            num_heads = ch // num_head_channels
+            dim_head = num_head_channels
+        if legacy:
+            #num_heads = 1
+            dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+        self.middle_block = TimestepEmbedSequential(
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+            AttentionBlock(
+                ch,
+                use_checkpoint=use_checkpoint,
+                num_heads=num_heads,
+                num_head_channels=dim_head,
+                use_new_attention_order=use_new_attention_order,
+            ) if not use_spatial_transformer else SpatialTransformer(  # always uses a self-attn
+                            ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                            disable_self_attn=disable_middle_self_attn, use_linear=use_linear_in_transformer,
+                            use_checkpoint=use_checkpoint
+                        ),
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+        )
+        self._feature_size += ch
+
+        self.output_blocks = nn.ModuleList([])
+        for level, mult in list(enumerate(channel_mult))[::-1]:
+            for i in range(self.num_res_blocks[level] + 1):
+                ich = input_block_chans.pop()
+                layers = [
+                    ResBlock(
+                        ch + ich,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=model_channels * mult,
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = model_channels * mult
+                if ds in attention_resolutions:
+                    if num_head_channels == -1:
+                        dim_head = ch // num_heads
+                    else:
+                        num_heads = ch // num_head_channels
+                        dim_head = num_head_channels
+                    if legacy:
+                        #num_heads = 1
+                        dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+                    if exists(disable_self_attentions):
+                        disabled_sa = disable_self_attentions[level]
+                    else:
+                        disabled_sa = False
+
+                    if not exists(num_attention_blocks) or i < num_attention_blocks[level]:
+                        layers.append(
+                            AttentionBlock(
+                                ch,
+                                use_checkpoint=use_checkpoint,
+                                num_heads=num_heads_upsample,
+                                num_head_channels=dim_head,
+                                use_new_attention_order=use_new_attention_order,
+                            ) if not use_spatial_transformer else SpatialTransformer(
+                                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                                disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer,
+                                use_checkpoint=use_checkpoint
+                            )
+                        )
+                if level and i == self.num_res_blocks[level]:
+                    out_ch = ch
+                    layers.append(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            up=True,
+                        )
+                        if resblock_updown
+                        else Upsample(ch, conv_resample, dims=dims, out_channels=out_ch)
+                    )
+                    ds //= 2
+                self.output_blocks.append(TimestepEmbedSequential(*layers))
+                self._feature_size += ch
+
+        self.out = nn.Sequential(
+            normalization(ch),
+            nn.SiLU(),
+            zero_module(conv_nd(dims, model_channels, out_channels, 3, padding=1)),
+        )
+        if self.predict_codebook_ids:
+            self.id_predictor = nn.Sequential(
+            normalization(ch),
+            conv_nd(dims, model_channels, n_embed, 1),
+            #nn.LogSoftmax(dim=1)  # change to cross_entropy and produce non-normalized logits
+        )
+
+    def convert_to_fp16(self):
+        """
+        Convert the torso of the model to float16.
+        """
+        self.input_blocks.apply(convert_module_to_f16)
+        self.middle_block.apply(convert_module_to_f16)
+        self.output_blocks.apply(convert_module_to_f16)
+
+    def convert_to_fp32(self):
+        """
+        Convert the torso of the model to float32.
+        """
+        self.input_blocks.apply(convert_module_to_f32)
+        self.middle_block.apply(convert_module_to_f32)
+        self.output_blocks.apply(convert_module_to_f32)
+
+    def forward(self, x, timesteps=None, context=None, y=None,**kwargs):
+        """
+        Apply the model to an input batch.
+        :param x: an [N x C x ...] Tensor of inputs.
+        :param timesteps: a 1-D batch of timesteps.
+        :param context: conditioning plugged in via crossattn
+        :param y: an [N] Tensor of labels, if class-conditional.
+        :return: an [N x C x ...] Tensor of outputs.
+        """
+        assert (y is not None) == (
+            self.num_classes is not None
+        ), "must specify y if and only if the model is class-conditional"
+        hs = []
+        t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
+        emb = self.time_embed(t_emb)
+
+        if self.num_classes is not None:
+            assert y.shape[0] == x.shape[0]
+            emb = emb + self.label_emb(y)
+
+        h = x.type(self.dtype)
+        for module in self.input_blocks:
+            h = module(h, emb, context)
+            hs.append(h)
+        h = self.middle_block(h, emb, context)
+        for module in self.output_blocks:
+            h = th.cat([h, hs.pop()], dim=1)
+            h = module(h, emb, context)
+        h = h.type(x.dtype)
+        if self.predict_codebook_ids:
+            return self.id_predictor(h)
+        else:
+            return self.out(h)
diff --git a/ldm/modules/diffusionmodules/upscaling.py b/ldm/modules/diffusionmodules/upscaling.py
new file mode 100644
index 0000000..0381666
--- /dev/null
+++ b/ldm/modules/diffusionmodules/upscaling.py
@@ -0,0 +1,81 @@
+import torch
+import torch.nn as nn
+import numpy as np
+from functools import partial
+
+from ldm.modules.diffusionmodules.util import extract_into_tensor, make_beta_schedule
+from ldm.util import default
+
+
+class AbstractLowScaleModel(nn.Module):
+    # for concatenating a downsampled image to the latent representation
+    def __init__(self, noise_schedule_config=None):
+        super(AbstractLowScaleModel, self).__init__()
+        if noise_schedule_config is not None:
+            self.register_schedule(**noise_schedule_config)
+
+    def register_schedule(self, beta_schedule="linear", timesteps=1000,
+                          linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+        betas = make_beta_schedule(beta_schedule, timesteps, linear_start=linear_start, linear_end=linear_end,
+                                   cosine_s=cosine_s)
+        alphas = 1. - betas
+        alphas_cumprod = np.cumprod(alphas, axis=0)
+        alphas_cumprod_prev = np.append(1., alphas_cumprod[:-1])
+
+        timesteps, = betas.shape
+        self.num_timesteps = int(timesteps)
+        self.linear_start = linear_start
+        self.linear_end = linear_end
+        assert alphas_cumprod.shape[0] == self.num_timesteps, 'alphas have to be defined for each timestep'
+
+        to_torch = partial(torch.tensor, dtype=torch.float32)
+
+        self.register_buffer('betas', to_torch(betas))
+        self.register_buffer('alphas_cumprod', to_torch(alphas_cumprod))
+        self.register_buffer('alphas_cumprod_prev', to_torch(alphas_cumprod_prev))
+
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.register_buffer('sqrt_alphas_cumprod', to_torch(np.sqrt(alphas_cumprod)))
+        self.register_buffer('sqrt_one_minus_alphas_cumprod', to_torch(np.sqrt(1. - alphas_cumprod)))
+        self.register_buffer('log_one_minus_alphas_cumprod', to_torch(np.log(1. - alphas_cumprod)))
+        self.register_buffer('sqrt_recip_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod)))
+        self.register_buffer('sqrt_recipm1_alphas_cumprod', to_torch(np.sqrt(1. / alphas_cumprod - 1)))
+
+    def q_sample(self, x_start, t, noise=None):
+        noise = default(noise, lambda: torch.randn_like(x_start))
+        return (extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start +
+                extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape) * noise)
+
+    def forward(self, x):
+        return x, None
+
+    def decode(self, x):
+        return x
+
+
+class SimpleImageConcat(AbstractLowScaleModel):
+    # no noise level conditioning
+    def __init__(self):
+        super(SimpleImageConcat, self).__init__(noise_schedule_config=None)
+        self.max_noise_level = 0
+
+    def forward(self, x):
+        # fix to constant noise level
+        return x, torch.zeros(x.shape[0], device=x.device).long()
+
+
+class ImageConcatWithNoiseAugmentation(AbstractLowScaleModel):
+    def __init__(self, noise_schedule_config, max_noise_level=1000, to_cuda=False):
+        super().__init__(noise_schedule_config=noise_schedule_config)
+        self.max_noise_level = max_noise_level
+
+    def forward(self, x, noise_level=None):
+        if noise_level is None:
+            noise_level = torch.randint(0, self.max_noise_level, (x.shape[0],), device=x.device).long()
+        else:
+            assert isinstance(noise_level, torch.Tensor)
+        z = self.q_sample(x, noise_level)
+        return z, noise_level
+
+
+
diff --git a/ldm/modules/diffusionmodules/util.py b/ldm/modules/diffusionmodules/util.py
new file mode 100644
index 0000000..637363d
--- /dev/null
+++ b/ldm/modules/diffusionmodules/util.py
@@ -0,0 +1,270 @@
+# adopted from
+# https://github.com/openai/improved-diffusion/blob/main/improved_diffusion/gaussian_diffusion.py
+# and
+# https://github.com/lucidrains/denoising-diffusion-pytorch/blob/7706bdfc6f527f58d33f84b7b522e61e6e3164b3/denoising_diffusion_pytorch/denoising_diffusion_pytorch.py
+# and
+# https://github.com/openai/guided-diffusion/blob/0ba878e517b276c45d1195eb29f6f5f72659a05b/guided_diffusion/nn.py
+#
+# thanks!
+
+
+import os
+import math
+import torch
+import torch.nn as nn
+import numpy as np
+from einops import repeat
+
+from ldm.util import instantiate_from_config
+
+
+def make_beta_schedule(schedule, n_timestep, linear_start=1e-4, linear_end=2e-2, cosine_s=8e-3):
+    if schedule == "linear":
+        betas = (
+                torch.linspace(linear_start ** 0.5, linear_end ** 0.5, n_timestep, dtype=torch.float64) ** 2
+        )
+
+    elif schedule == "cosine":
+        timesteps = (
+                torch.arange(n_timestep + 1, dtype=torch.float64) / n_timestep + cosine_s
+        )
+        alphas = timesteps / (1 + cosine_s) * np.pi / 2
+        alphas = torch.cos(alphas).pow(2)
+        alphas = alphas / alphas[0]
+        betas = 1 - alphas[1:] / alphas[:-1]
+        betas = np.clip(betas, a_min=0, a_max=0.999)
+
+    elif schedule == "sqrt_linear":
+        betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64)
+    elif schedule == "sqrt":
+        betas = torch.linspace(linear_start, linear_end, n_timestep, dtype=torch.float64) ** 0.5
+    else:
+        raise ValueError(f"schedule '{schedule}' unknown.")
+    return betas.numpy()
+
+
+def make_ddim_timesteps(ddim_discr_method, num_ddim_timesteps, num_ddpm_timesteps, verbose=True):
+    if ddim_discr_method == 'uniform':
+        c = num_ddpm_timesteps // num_ddim_timesteps
+        ddim_timesteps = np.asarray(list(range(0, num_ddpm_timesteps, c)))
+    elif ddim_discr_method == 'quad':
+        ddim_timesteps = ((np.linspace(0, np.sqrt(num_ddpm_timesteps * .8), num_ddim_timesteps)) ** 2).astype(int)
+    else:
+        raise NotImplementedError(f'There is no ddim discretization method called "{ddim_discr_method}"')
+
+    # assert ddim_timesteps.shape[0] == num_ddim_timesteps
+    # add one to get the final alpha values right (the ones from first scale to data during sampling)
+    steps_out = ddim_timesteps + 1
+    if verbose:
+        print(f'Selected timesteps for ddim sampler: {steps_out}')
+    return steps_out
+
+
+def make_ddim_sampling_parameters(alphacums, ddim_timesteps, eta, verbose=True):
+    # select alphas for computing the variance schedule
+    alphas = alphacums[ddim_timesteps]
+    alphas_prev = np.asarray([alphacums[0]] + alphacums[ddim_timesteps[:-1]].tolist())
+
+    # according the the formula provided in https://arxiv.org/abs/2010.02502
+    sigmas = eta * np.sqrt((1 - alphas_prev) / (1 - alphas) * (1 - alphas / alphas_prev))
+    if verbose:
+        print(f'Selected alphas for ddim sampler: a_t: {alphas}; a_(t-1): {alphas_prev}')
+        print(f'For the chosen value of eta, which is {eta}, '
+              f'this results in the following sigma_t schedule for ddim sampler {sigmas}')
+    return sigmas, alphas, alphas_prev
+
+
+def betas_for_alpha_bar(num_diffusion_timesteps, alpha_bar, max_beta=0.999):
+    """
+    Create a beta schedule that discretizes the given alpha_t_bar function,
+    which defines the cumulative product of (1-beta) over time from t = [0,1].
+    :param num_diffusion_timesteps: the number of betas to produce.
+    :param alpha_bar: a lambda that takes an argument t from 0 to 1 and
+                      produces the cumulative product of (1-beta) up to that
+                      part of the diffusion process.
+    :param max_beta: the maximum beta to use; use values lower than 1 to
+                     prevent singularities.
+    """
+    betas = []
+    for i in range(num_diffusion_timesteps):
+        t1 = i / num_diffusion_timesteps
+        t2 = (i + 1) / num_diffusion_timesteps
+        betas.append(min(1 - alpha_bar(t2) / alpha_bar(t1), max_beta))
+    return np.array(betas)
+
+
+def extract_into_tensor(a, t, x_shape):
+    b, *_ = t.shape
+    out = a.gather(-1, t)
+    return out.reshape(b, *((1,) * (len(x_shape) - 1)))
+
+
+def checkpoint(func, inputs, params, flag):
+    """
+    Evaluate a function without caching intermediate activations, allowing for
+    reduced memory at the expense of extra compute in the backward pass.
+    :param func: the function to evaluate.
+    :param inputs: the argument sequence to pass to `func`.
+    :param params: a sequence of parameters `func` depends on but does not
+                   explicitly take as arguments.
+    :param flag: if False, disable gradient checkpointing.
+    """
+    if flag:
+        args = tuple(inputs) + tuple(params)
+        return CheckpointFunction.apply(func, len(inputs), *args)
+    else:
+        return func(*inputs)
+
+
+class CheckpointFunction(torch.autograd.Function):
+    @staticmethod
+    def forward(ctx, run_function, length, *args):
+        ctx.run_function = run_function
+        ctx.input_tensors = list(args[:length])
+        ctx.input_params = list(args[length:])
+        ctx.gpu_autocast_kwargs = {"enabled": torch.is_autocast_enabled(),
+                                   "dtype": torch.get_autocast_gpu_dtype(),
+                                   "cache_enabled": torch.is_autocast_cache_enabled()}
+        with torch.no_grad():
+            output_tensors = ctx.run_function(*ctx.input_tensors)
+        return output_tensors
+
+    @staticmethod
+    def backward(ctx, *output_grads):
+        ctx.input_tensors = [x.detach().requires_grad_(True) for x in ctx.input_tensors]
+        with torch.enable_grad(), \
+                torch.cuda.amp.autocast(**ctx.gpu_autocast_kwargs):
+            # Fixes a bug where the first op in run_function modifies the
+            # Tensor storage in place, which is not allowed for detach()'d
+            # Tensors.
+            shallow_copies = [x.view_as(x) for x in ctx.input_tensors]
+            output_tensors = ctx.run_function(*shallow_copies)
+        input_grads = torch.autograd.grad(
+            output_tensors,
+            ctx.input_tensors + ctx.input_params,
+            output_grads,
+            allow_unused=True,
+        )
+        del ctx.input_tensors
+        del ctx.input_params
+        del output_tensors
+        return (None, None) + input_grads
+
+
+def timestep_embedding(timesteps, dim, max_period=10000, repeat_only=False):
+    """
+    Create sinusoidal timestep embeddings.
+    :param timesteps: a 1-D Tensor of N indices, one per batch element.
+                      These may be fractional.
+    :param dim: the dimension of the output.
+    :param max_period: controls the minimum frequency of the embeddings.
+    :return: an [N x dim] Tensor of positional embeddings.
+    """
+    if not repeat_only:
+        half = dim // 2
+        freqs = torch.exp(
+            -math.log(max_period) * torch.arange(start=0, end=half, dtype=torch.float32) / half
+        ).to(device=timesteps.device)
+        args = timesteps[:, None].float() * freqs[None]
+        embedding = torch.cat([torch.cos(args), torch.sin(args)], dim=-1)
+        if dim % 2:
+            embedding = torch.cat([embedding, torch.zeros_like(embedding[:, :1])], dim=-1)
+    else:
+        embedding = repeat(timesteps, 'b -> b d', d=dim)
+    return embedding
+
+
+def zero_module(module):
+    """
+    Zero out the parameters of a module and return it.
+    """
+    for p in module.parameters():
+        p.detach().zero_()
+    return module
+
+
+def scale_module(module, scale):
+    """
+    Scale the parameters of a module and return it.
+    """
+    for p in module.parameters():
+        p.detach().mul_(scale)
+    return module
+
+
+def mean_flat(tensor):
+    """
+    Take the mean over all non-batch dimensions.
+    """
+    return tensor.mean(dim=list(range(1, len(tensor.shape))))
+
+
+def normalization(channels):
+    """
+    Make a standard normalization layer.
+    :param channels: number of input channels.
+    :return: an nn.Module for normalization.
+    """
+    return GroupNorm32(32, channels)
+
+
+# PyTorch 1.7 has SiLU, but we support PyTorch 1.5.
+class SiLU(nn.Module):
+    def forward(self, x):
+        return x * torch.sigmoid(x)
+
+
+class GroupNorm32(nn.GroupNorm):
+    def forward(self, x):
+        return super().forward(x.float()).type(x.dtype)
+
+def conv_nd(dims, *args, **kwargs):
+    """
+    Create a 1D, 2D, or 3D convolution module.
+    """
+    if dims == 1:
+        return nn.Conv1d(*args, **kwargs)
+    elif dims == 2:
+        return nn.Conv2d(*args, **kwargs)
+    elif dims == 3:
+        return nn.Conv3d(*args, **kwargs)
+    raise ValueError(f"unsupported dimensions: {dims}")
+
+
+def linear(*args, **kwargs):
+    """
+    Create a linear module.
+    """
+    return nn.Linear(*args, **kwargs)
+
+
+def avg_pool_nd(dims, *args, **kwargs):
+    """
+    Create a 1D, 2D, or 3D average pooling module.
+    """
+    if dims == 1:
+        return nn.AvgPool1d(*args, **kwargs)
+    elif dims == 2:
+        return nn.AvgPool2d(*args, **kwargs)
+    elif dims == 3:
+        return nn.AvgPool3d(*args, **kwargs)
+    raise ValueError(f"unsupported dimensions: {dims}")
+
+
+class HybridConditioner(nn.Module):
+
+    def __init__(self, c_concat_config, c_crossattn_config):
+        super().__init__()
+        self.concat_conditioner = instantiate_from_config(c_concat_config)
+        self.crossattn_conditioner = instantiate_from_config(c_crossattn_config)
+
+    def forward(self, c_concat, c_crossattn):
+        c_concat = self.concat_conditioner(c_concat)
+        c_crossattn = self.crossattn_conditioner(c_crossattn)
+        return {'c_concat': [c_concat], 'c_crossattn': [c_crossattn]}
+
+
+def noise_like(shape, device, repeat=False):
+    repeat_noise = lambda: torch.randn((1, *shape[1:]), device=device).repeat(shape[0], *((1,) * (len(shape) - 1)))
+    noise = lambda: torch.randn(shape, device=device)
+    return repeat_noise() if repeat else noise()
\ No newline at end of file
diff --git a/ldm/modules/distributions/__init__.py b/ldm/modules/distributions/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ldm/modules/distributions/__pycache__/__init__.cpython-310.pyc b/ldm/modules/distributions/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..12659b1b6906d4d00373399937bf10804738f88b
GIT binary patch
literal 178
zcmd1j<>g`kf|mb%sUZ3>h(HF6K#l_t7qb9~6oz01O-8?!3`HPe1o10WKO;XkRX;Dk
zBsD2NKU+UJKP6SaAhoDCKQA#yKiJa1#KOqZ6e8^G92}&dlai~So1apelUl5wl383*
nl$lgol9``ZtREkrnU`4-AFo$Xd5gmaqQ(wnT`?1oU||3N=4dP}

literal 0
HcmV?d00001

diff --git a/ldm/modules/distributions/__pycache__/__init__.cpython-37.pyc b/ldm/modules/distributions/__pycache__/__init__.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..a4dac7ba25217bcb6d823e569f4cbbc56226c3c2
GIT binary patch
literal 175
zcmZ?b<>g`kf{4w+sUZ3>h=2h`Aj1KOi&=m~3PUi1CZpd<h9ZzKg7_7opOK%Ns-Kr%
zlA4sCpRJ#gSdyrpUX)*2pdVbC=aZS2o}7`Ir=OgklB(~LnU?0{8Kj?+lB=JapHiBW
sTCAUvSzJ<-nN(VmnV(mzA0MBYmst`YuUAlci^B$@#tvlTXCP((0J|<O6951J

literal 0
HcmV?d00001

diff --git a/ldm/modules/distributions/__pycache__/distributions.cpython-310.pyc b/ldm/modules/distributions/__pycache__/distributions.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..a14e79de601e7a2668cfdb5203d0d6810a16f293
GIT binary patch
literal 3776
zcmai1-)|eo5x(8~!4pL}imlp89VeGMcEP|AB~?x=*L4&~O^`Go(5O#D0^)dkqK-P=
z(Yr@GCa9N4K$8M4&_4Fb%A5X*=I^+t0(~g_79ej$BjuYtQkG;(Jz$5sv$H$9GvCZC
z)2XSlfp7Qlui-x?4ddT5IR04}te~k+L4+Y#Y&42r*5Dr-f(vuk5ateVm>5~Y#>fuM
z9kXF!?1&P^rO?LM5w4inH5w(s){ScU8TK>k0ny_h$6y6by$%u@4JM2R7ff)hG=(WF
z{6Y(MaPpZCTPl-&kgY}djavOIN|S(L;rOFIwN<nq9gxh}F}`N|tiI3kOZ7DSq}z+b
zZkS}D_`Q^=1lv+!ys>Y{35fr<QQJ(rVJ%6su$87;wICH?trtp_CVpI7U!GsOv$%Y_
zV6U#Oe_o44x7JNXKMqw*96Q#j(OmCw?kJzM)3JV}YddUk^<_7jE9B0+m^86ilB1+e
zq(p>HUl8(dZsAO!=2b8py!KK(BvY^2x!Y_eem873^K!FUTn7D0v-zO!$Hj~URt!0H
zq~xSJN!OvXD7u_$xU6Do-r4Sc8-i!MTR3pLElO~^o4>ji!RaS_PWr|l^{J816<vPF
zhDH;ZT`!CT7uyg0xF1%zbTKV25jlAQGC`C?S+jY8cA@RcZ1Bq$6wi&t&UB7O-O%Wa
zwabg7?IkYK7JZc_wTEV3)b;;h@C#b3VQ{7na@fp~j>|9VC?McSy0(0U$TShX=6sja
z(P?`{HYJz4CY(;(7o1LbT|rYsj+}l$=Msh}*I5Kn{)#9Gcb7Hnwjn0IV;fAA#U#d#
zsE8>5q9hyuVmg0)E%MtGkRSVfr6NB$j*X83QaApnPrzJ7Q@bDqI5O7ZLq=&vm`}_v
zH~?d!Ss8n3!+91RX-{*8H0OX1<uo&6>=C?N;?P)+CeU-_3V3Do%stkb#IMn*h|&{g
z8IXrW(<Mk!tGe}P1hHq7ssK+@MBO&J`0jr{>MX7Nr{MhS14uX`d>92`bt*6Se2G}i
zkdkscOJ%T`+ref(*+R&Nev&(J+J5LuFk@t!d`ggL7uZ%=6y#?3RWCPHCUTR6%7aj*
zsu@RHVa~DHHK?qbMU=dOzN9EU6)GAVuYs=!bq<?l7ny^00rYKV9S&}s=be`a5E<jh
zOYdMn)<a_%e+n<alCX1!JmYjwPM)wC<BGA(hE!O1T~g_&y7Ddj$eTpo1j%jbCnCvh
zwdo^)SBbChq%T>df}DMI`gR2R>u8Ds>n5|<;QG1uDT1^f&=tlX^(hf*(ES}9x`Lbl
zJW$!%Y!Al<_>8_GZwL-3nwYU5V+(T+aO*f%P+4bbRcY2hEZE;M;XL6p1fLV~Ll|Wi
z)flT-$GJ)?oT7Jp2XIPPPVG}I)kpY(S+3%U@(RcZdp~@zZ_3#*^BKjEB#*<!8W7sN
z&Y!|Wr4nfBcXP9sJ}T`0YZCqy36D%{r`cvGPnkLATQNzhc@T$Hn9DE1*8Gs8(XYeZ
z1IL<W<zu`N|EIy5=Xzu-mUGzkL>eGxme90EK1T<ZI^~hfd}_hEX2+%w-^SYmo^gf*
zKUfbQaw&CmCSY$%d#QXCARKaG7oMp)b@?uK9@)@Xd2ZIic$3eKC+XTbPY{a;*PQnv
zxd4u~n_Q%x#xhx`Y8Bx~xXdji3e|A2CVP}((WxS4CUJZla*DD!S#=sQ=`qgSgMoLh
z&9LyyiYAZhOqCVqB1@jqt)B+JBfhR?Z=I`+icXb%q-z&2b!P2ZqWzYMocZv$j9Ege
zBDF4xwp%V+(-$hPq?v&HKK>}dP*f6})n7ma`xZEWSNYJ~u-1$(E8A9P3{eC+e22@K
zq4g~ulL>p^egoM)Q`bI4{Oz!>*`E#6CVHC^Cd>`KX6&#p_%@-&Mwu{y3pPubF=>7l
zcmfAl0d`843cfiBE~PPw>Z<F%gIZAb@Ts&fgYd4mnPok7w^nOM*=D~r7o^=<+6xmu
zs<r!(2*tZ1+SuqTynt)1IBnGy-wzh&{W}|eYdL5wy}u|nZj1L?LcAB;5eo~!osFg6
z%(s>omqyE*&zHxkQlY{N_4;QEi~0<!X}8zULJyJa{oylDL=QvR4wE4CT4DAm3=<Fd
z^V*~8tLF5gb;`5K^JVDy{VeVJC^&u`KlWNO^+n(-yt!}TdS1)Vf=x&UeuDix+zd4>
z6ke7V2EZy`s#~7lO_R2#(}&kr5cKJoTBbg@>1SRXVpA_opis8^^s4u|eQ>a;?-l9k
z1gghnh8}?G7DyXj`SRj6gV))}>y;r(N`>e4kd?oIm;S~daJ(fc*zi~`M=DBG=EEAg
z&J?liWI7w_GE?O_pPQGYV4um~;-?4}{V*@zi&2W{&iDU_8U1>449?#;bPfmCl2ms6
zxVaV2-HTJ$LET4ZEUy#FD82mfPo1!l9q!Wpl#%s8<PG$2N}C=73wfz{4J;Jn#qoGC
zH`7+9IF$}seYA-q?H|&~s194DX`~U0QuD#IS+*QrKHxTzG^J(E4=N`A>9WQDXC2Bl
zQtXaOZ_@Af9!tta`fq@K7Um9ghC!wyQ;Q$VAv%-kQ@Hnve<1e}5aY0R*#&djot~^z
Fe+KpbET8}Y

literal 0
HcmV?d00001

diff --git a/ldm/modules/distributions/__pycache__/distributions.cpython-37.pyc b/ldm/modules/distributions/__pycache__/distributions.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..7755b689570eaf1fdfaf9edf7c8d2aca53875c90
GIT binary patch
literal 3732
zcmb7HTW=gi7Ov{sw8s}5$4d+eqL)B+2L(^;H4sq9)eC}+tQA>lR~FHzr>kb%<DTwG
zRd+7dSX$vofJ7d6<tcvTnI8bZqo0uWDQ`UUwBM<oaeRxHZK<lOs!!FqednC&o72;E
zhUZ^Do|fNFGWH*uj2{Ps>uCBjbb<+<u=VQA*R4;Ou!Q}X346y{w=r{si<v9!9edru
z+!Hm-YtqHs6TX;u%+_my-(ijVGn~iPBI?JFg~4?+eGZ*u>s+vPOK@Rfmo04J;0+Gg
z?UnO4+B#QZl&{8k$L%7Ivozw^NDsZK-9po(T+VjbmpBi5`C5aQr`NLl)BYfleVOJ`
z{8gz;VcF3#>HNTyi!Hz1ZEj?J*-W!swzF)rDZ)H#c2!mkns<uycAR#j4VgBhOvvVH
z-09r>^z&vS`pteOibU$B*eBVc+2Y_)>FJOJ@JK|`obj!*$0@Z~(fw}Sl#R0*T#@ID
zB2`CxN(ZvZ35>=ZUpbZ|QLQ!H(r>lWurFJ!vfgS{YGFLpYTYZsr23+2#Ku3!i_E8p
zp(4d-rb!x~`?gs*-s^9I<i%bW9=z|08a(Kir&eS5{($R^32_3wsqMLb$cL;2$c|m_
z-Vc*PHmu4((;z);C8!;zao%dp;N#d*uO4B!I3_#ZHKhNL^~QV^aXQ{D#)hZ%@me^_
z>m(5TADqZh^ZPixe3Vn=d`}Tf6vd80PGE{`@jskB1rpLylR)HbxSQ4vxSP;<9Zlav
zx99FzlSnwi+@uhM`6*Eo{$mc8Pl&ph?AjeJro=Q>JTW6)0vKw-gHPwmbE|RK&C)RW
zBrLR!!*o9`K8}bXJ@h6h-a^y=MppqN=RIr4Da#1^iT#BIu-Ir$&Y!w)pF;xenw24a
z^RN!a^ktsy6103;p}Cn&VC1PG*6JA9ySz7vH|tG_+7s>o5G_!&d;>>+?HM8U8D%OY
zA6vxzE{1rn{yZA2Tpta7xdsYP$op|58`EWd5Gq7wE|o6bJX6s|=|&qxx`_~%VOn}g
z*1aDpEGEb@<+LEuK2WXmI4W)Va8TMh7o|-^^)0C~-Adw3Sz0*kG$d%)ReTUw)T`8)
zpne|Ngojtq^ej5YJ-)!_(9iK1^jmM8<ddU;LuT0X&D)re&CqymJ%#OIKUlX%{_uLJ
zA5Zu^JHxj5kSd9Vd|rv{t25ZDsCuYBpetP!rXnp}y%7$idY#sdK1RVR0hIjVVa+I`
zokue{i}4G5>->q@RH3;RkrMRKn<Bsf?RyMJ0W}9mpqO>}E`$c)*ru;;3k&eG@x=j+
zE9_lBtmmCUAzh$dwFMMY6h5zO13uO~a>>CRLl^}#w2$|^v$V&nM*GjeN)7YFCsAhd
z*GBkaY!7IPJUW{FdQDMcdPczmGtbe>ukc_j?EOQgI#U2s(J$>m_MmdXYoJpLL_D&x
zo8=o)9kO(3Z6<#Lk)~q_LQ|HH{v4mRJZpt-y?&zEHxnb8kyeqv_9C5XO%2UB;ByRM
zmO~E6?WYb*Y4==;;%(e6;0SL>Sc9qH5TCL`Zvw`3j8p1ofw-Y1+{z6N?}*{p^H4Q1
zjm<Ur++2!ETxq2UKE!8f4Px~cR*a?81sWLull2--6@R2h=^!)cbsu~32N^aU48_t;
zljL2{X-c-nj(iq&4ER=Xq6M+-_mwV^>IJ2X3d%=l<1+cj6FE|CY4DHf4(7kO!2B8K
zsduP*;l1J|d?kFD%bm~bb>FGa%=wDeBl15zWF`tk0<ZoSo#5ZFhZlyF`K_VdaaP&g
zm$#jq4Uz48){dp-hR!$a@9f^(KiSuy-ki`^a$8tC{7e27Ljj^YC?Uh{SgUM@e_?GC
zc3jj08|dH*gdB&yF920o0b+ol+UaWDp2R9;E(+v^AHI#mUk~u;tWc5sFxbfRf&Q@B
z?8f;<(O!(Qelr`$G>n_wA{J7-BXG?uG;X@hc9ON5%kM|am%_`Pu>C>QUU`36blw&3
zwS{;ux-6EKqRX9?zg%j6u)H$b-g>@0$+VUl*U<B~mzIqTx3d1A$Yp>y4gPjJ5b=Gf
zx-yMq(3bfFDboPh3%a8cs~63tJCs;;5GomjMV|FTRF^PG9tCZcg(3<y?$#HfUeFG6
zT$XhZg(=Pppcx976hWR<2EeXR>5D<w&(dySl0{Hx5QZeCp;_Cy5avN5acGdGV5qtU
z-QI(KffXDY230cJ!}PeSP`cL_(RJaGyKioDc$|+szA)tJ47_F!k(i%4U&BM+S@)bf
zRlUO<xE|{`)p-a@n4(g}u$P;>XKG5rDy_v!>J2bJkTmhC0+gWlTm4FcnoIg#rA%bP
z1oiQ?G*kUBX>BHpSCS0c>#G=yDM!iomusWJuRm28#Z`-RI%VEUjF}s7qoQ9Q^9Q5O
z;tOPU=?<_|)~ZXuQZ-*5&zDO(Yxj(dV|hu`a$--W8<=XUl&t2UqEB%L>15We^I6++
z)n%+Be5(qZ7X85^)xeUu*_-=b>EV1C<t7>p`M!*0a+l%fU#Wg6uG0Ny6v)0k<Ihgc
GHvS9z<RW(f

literal 0
HcmV?d00001

diff --git a/ldm/modules/distributions/distributions.py b/ldm/modules/distributions/distributions.py
new file mode 100644
index 0000000..f2b8ef9
--- /dev/null
+++ b/ldm/modules/distributions/distributions.py
@@ -0,0 +1,92 @@
+import torch
+import numpy as np
+
+
+class AbstractDistribution:
+    def sample(self):
+        raise NotImplementedError()
+
+    def mode(self):
+        raise NotImplementedError()
+
+
+class DiracDistribution(AbstractDistribution):
+    def __init__(self, value):
+        self.value = value
+
+    def sample(self):
+        return self.value
+
+    def mode(self):
+        return self.value
+
+
+class DiagonalGaussianDistribution(object):
+    def __init__(self, parameters, deterministic=False):
+        self.parameters = parameters
+        self.mean, self.logvar = torch.chunk(parameters, 2, dim=1)
+        self.logvar = torch.clamp(self.logvar, -30.0, 20.0)
+        self.deterministic = deterministic
+        self.std = torch.exp(0.5 * self.logvar)
+        self.var = torch.exp(self.logvar)
+        if self.deterministic:
+            self.var = self.std = torch.zeros_like(self.mean).to(device=self.parameters.device)
+
+    def sample(self):
+        x = self.mean + self.std * torch.randn(self.mean.shape).to(device=self.parameters.device)
+        return x
+
+    def kl(self, other=None):
+        if self.deterministic:
+            return torch.Tensor([0.])
+        else:
+            if other is None:
+                return 0.5 * torch.sum(torch.pow(self.mean, 2)
+                                       + self.var - 1.0 - self.logvar,
+                                       dim=[1, 2, 3])
+            else:
+                return 0.5 * torch.sum(
+                    torch.pow(self.mean - other.mean, 2) / other.var
+                    + self.var / other.var - 1.0 - self.logvar + other.logvar,
+                    dim=[1, 2, 3])
+
+    def nll(self, sample, dims=[1,2,3]):
+        if self.deterministic:
+            return torch.Tensor([0.])
+        logtwopi = np.log(2.0 * np.pi)
+        return 0.5 * torch.sum(
+            logtwopi + self.logvar + torch.pow(sample - self.mean, 2) / self.var,
+            dim=dims)
+
+    def mode(self):
+        return self.mean
+
+
+def normal_kl(mean1, logvar1, mean2, logvar2):
+    """
+    source: https://github.com/openai/guided-diffusion/blob/27c20a8fab9cb472df5d6bdd6c8d11c8f430b924/guided_diffusion/losses.py#L12
+    Compute the KL divergence between two gaussians.
+    Shapes are automatically broadcasted, so batches can be compared to
+    scalars, among other use cases.
+    """
+    tensor = None
+    for obj in (mean1, logvar1, mean2, logvar2):
+        if isinstance(obj, torch.Tensor):
+            tensor = obj
+            break
+    assert tensor is not None, "at least one argument must be a Tensor"
+
+    # Force variances to be Tensors. Broadcasting helps convert scalars to
+    # Tensors, but it does not work for torch.exp().
+    logvar1, logvar2 = [
+        x if isinstance(x, torch.Tensor) else torch.tensor(x).to(tensor)
+        for x in (logvar1, logvar2)
+    ]
+
+    return 0.5 * (
+        -1.0
+        + logvar2
+        - logvar1
+        + torch.exp(logvar1 - logvar2)
+        + ((mean1 - mean2) ** 2) * torch.exp(-logvar2)
+    )
diff --git a/ldm/modules/ema.py b/ldm/modules/ema.py
new file mode 100644
index 0000000..bded250
--- /dev/null
+++ b/ldm/modules/ema.py
@@ -0,0 +1,80 @@
+import torch
+from torch import nn
+
+
+class LitEma(nn.Module):
+    def __init__(self, model, decay=0.9999, use_num_upates=True):
+        super().__init__()
+        if decay < 0.0 or decay > 1.0:
+            raise ValueError('Decay must be between 0 and 1')
+
+        self.m_name2s_name = {}
+        self.register_buffer('decay', torch.tensor(decay, dtype=torch.float32))
+        self.register_buffer('num_updates', torch.tensor(0, dtype=torch.int) if use_num_upates
+        else torch.tensor(-1, dtype=torch.int))
+
+        for name, p in model.named_parameters():
+            if p.requires_grad:
+                # remove as '.'-character is not allowed in buffers
+                s_name = name.replace('.', '')
+                self.m_name2s_name.update({name: s_name})
+                self.register_buffer(s_name, p.clone().detach().data)
+
+        self.collected_params = []
+
+    def reset_num_updates(self):
+        del self.num_updates
+        self.register_buffer('num_updates', torch.tensor(0, dtype=torch.int))
+
+    def forward(self, model):
+        decay = self.decay
+
+        if self.num_updates >= 0:
+            self.num_updates += 1
+            decay = min(self.decay, (1 + self.num_updates) / (10 + self.num_updates))
+
+        one_minus_decay = 1.0 - decay
+
+        with torch.no_grad():
+            m_param = dict(model.named_parameters())
+            shadow_params = dict(self.named_buffers())
+
+            for key in m_param:
+                if m_param[key].requires_grad:
+                    sname = self.m_name2s_name[key]
+                    shadow_params[sname] = shadow_params[sname].type_as(m_param[key])
+                    shadow_params[sname].sub_(one_minus_decay * (shadow_params[sname] - m_param[key]))
+                else:
+                    assert not key in self.m_name2s_name
+
+    def copy_to(self, model):
+        m_param = dict(model.named_parameters())
+        shadow_params = dict(self.named_buffers())
+        for key in m_param:
+            if m_param[key].requires_grad:
+                m_param[key].data.copy_(shadow_params[self.m_name2s_name[key]].data)
+            else:
+                assert not key in self.m_name2s_name
+
+    def store(self, parameters):
+        """
+        Save the current parameters for restoring later.
+        Args:
+          parameters: Iterable of `torch.nn.Parameter`; the parameters to be
+            temporarily stored.
+        """
+        self.collected_params = [param.clone() for param in parameters]
+
+    def restore(self, parameters):
+        """
+        Restore the parameters stored with the `store` method.
+        Useful to validate the model with EMA parameters without affecting the
+        original optimization process. Store the parameters before the
+        `copy_to` method. After validation (or model saving), use this to
+        restore the former parameters.
+        Args:
+          parameters: Iterable of `torch.nn.Parameter`; the parameters to be
+            updated with the stored parameters.
+        """
+        for c_param, param in zip(self.collected_params, parameters):
+            param.data.copy_(c_param.data)
diff --git a/ldm/modules/encoders/__init__.py b/ldm/modules/encoders/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ldm/modules/encoders/__pycache__/__init__.cpython-310.pyc b/ldm/modules/encoders/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..ccdc342557e7756411222297f1dbd46672b84121
GIT binary patch
literal 173
zcmd1j<>g`kg0}yCsUZ3>h(HF6K#l_t7qb9~6oz01O-8?!3`HPe1o10CKO;XkRX;Dk
zBsD2NKU+UJKP6SaAhoDCKQA#yKiJa1#KOqZ6e8^G92}&dlai~So1apelUl5wng>!|
gRIDE#pP83g5+AQuP<e~P1|n?-vZ<H}NU$&f0Pcq=>Hq)$

literal 0
HcmV?d00001

diff --git a/ldm/modules/encoders/__pycache__/__init__.cpython-37.pyc b/ldm/modules/encoders/__pycache__/__init__.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..b7372fe72ce79f5db35a107755f9fc6418992bc7
GIT binary patch
literal 170
zcmZ?b<>g`kf{4w+sUZ3>h=2h`Aj1KOi&=m~3PUi1CZpd<h9ZzKg7_7zpOK%Ns-Kr%
zlA4sCpRJ#gSdyrpUX)*2pdVbC=aZS2o}7`Ir=OgklB(~LnU?0{8Kj?+lB=JapHiBW
kTCAU%2a+l(){l?R%*!l^kJl@xyv1Pyk+uU__8Ev70KZ}@&Hw-a

literal 0
HcmV?d00001

diff --git a/ldm/modules/encoders/__pycache__/modules.cpython-310.pyc b/ldm/modules/encoders/__pycache__/modules.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..bc3f3b9946977324080cb45d12ced6b867d73b5b
GIT binary patch
literal 8440
zcmbtZO>7*=b?*P^84iadilV6fofw-yW3R~7MvA@J^{y1HWUnJ}lNIg6lXZHBQ_W$M
zJ>8?~9{o&)Ihb<R1}wmr91_C-86XGc<Vy}d2S{=Zl3V4NL!v`~06qi|oG9OWHGdqD
zyEf8;s;RD@?y6U>zW2S?>HK`j!1cp_{qu0`Rm1o<x|zNl++0D*{u!Atgc%zR{cm=x
zhGkOOitUcmaJX#8ZpUkQokF9~@f&`p*eG(n8J8L*^mXF7PPtL;%s1vu<7<X+h4;`9
zUTE!9_U*<3Y6`+fjUO)HzKZ*zDB->|>~Tsgi?W!1Xf_tnRuKzmTj0K@abFduaDR&L
zmvFx*PUHSG-=D$#k~o9=Gkm`+%-cr&>=UfZSg)I^)Y=YPyWKQO^17quHZE_ZyI~Ub
zL#Zkom#-(SRD|*tvRKVuyZOQIO`7#>ct7XXx}^#t+-mmXT$NjCFUf;$Q#LzUi{3PS
z*~6RcH55XlVG5&ReQIo(!W7m+zhMhoI1i16^Qqw&4HvU@)xy=yESJqzKAcd?oO&Xa
zEhKtyRza2<`^JHJXwVkiHXfVnn4FvSx>%DR1W^*@L4c(@Sr~6UHdQeQT5&VWg5a_7
z^ux97v=go+X&!E->Fye>e+?tEG-<|bx8Hi>(pTPm>$0xCcJ1~@Yq998byCrb!)z_o
z;>t!9tKA2(j0sW7Xo?lHzcRCw)zOTOtclh_>-vyJk~ZeXcTGu3i;16JPxB8t-8k%o
zNgj&pQl_$Q%XzdarzzW+@^<ghzpSE4XQ!yq(-T}k`R6np$Luf9PDgjCdnBr)*$IO{
zm4ZO85an_Z-03xA-NK8LXXw%MLm{@pI1YlT7t@AFdb=zln?uSxWWG^0t2Wm?ry(@8
zE7Wlb*+@fJI}FkeXoRB{J`h-Zls_101PvNpR3_<NLt^O-p!6&~o><+|vNG>a&6B4{
zOQ9r}k@c5n-m$7>vtD=PS?aVx**VI1%tY8+;s?Hn*3(El3#{3aRP)@5X@}{`-b6ND
z@jRajN4O7*Sa(JEq6h?V#DbU;Wt3b#As*yW#o;6szTVjk1rF?!|9}E1m(}{;|MD;B
z`q_Iem*1f?D$`rh&&XDhO%Bb#9vFd@+Xv3T8kmO`?p&0Nz#BNo3v{ToD*BT4leIwE
zyWs<6C5bBPNkmCo6+@m67ZZKR!c6(SRy&Z*JgmEVA9S}sL>>9-^t@lA>^CT5<(Wv6
z&e01<86AX@>6sOz5|sGJ;>Ray{YGmfSoAkY?I^KUJVaq&38QDo!ob)u_OR4}b&i1D
z_#W!FP~O9GcK}Pfj(mGyXKxEDw+?J1XW$%~)V5?Cnk&YhJ8%cik|FGYyJC#mmkji9
zg}q`d8E5HP&VhHx&pas5c&Z^)UT<S6d<_f#_1+Uw-nu^#gmUv#wzgGqGn7f%i{nTY
z(<ID-INA+aXv&Gxq%B{gfu0ttyV=ZJ+x(Emf{3!_W*i1WrrorcE3>7XPBYt`IRUD$
zl}ccTpgqVIk@*nul2tKZFkdr$2pgDj=BT)bBb3wXiDruwYIKd1j3xd$YDv<9BxdL4
z4kW$>nG5TYHL#bowC#ZtjU;2Q%kvlm!1F|SH)@4AcdQ?E)GCM8&x;$Wk+qpgIzJ64
zm{vAr^;)wFe5D=cK@UKkh$xQ;)q`;vih2PL$b4iab4joE^`oo(?BRgnS1?$DR2Z^`
z?6I|eqi)KJDAo(JutEAP-=OTbC?nOLn(hQSUq%Dr3{T956)6=K%aUq0TAz)3II4BC
zzxm<aP~MYK9@YTlVY~`k8YQ)RQNCTvx1+4K1%Z$-YI$1gG<U;V){|k4d##0cn=ub7
z)o_V4%xbWXwbphsX@|9D@}QH-aFyph0#=9KGPM$dLRA^6pa}T6muNjSdB>J-;je{s
z<Qg5_?6*-p#zd2lka5uy1>mB?s0f}QN@5PI!DF6eGzyH3K9;|rsyAfX4-?RtAw2fa
ze>1}}^6jv;ahaBxWKiNxC~F$wWevBr9NgyVQaerCak!RWzHs+qaPdM6yBhL#KDmW?
zZ9E66r2mw%-$6D3RYDGcDNrMDa(7(z7+4Dox&FYBuMUiZ;)uCSfVMONT1gLA3tKYn
z1YH@jEJLA+`G~hk5}cZwS{0uPRq8bF2XUCR^KIpA$uR7X0k{D5$gus~QC`7#d6lwb
z;P)NWXN1Tlvy7jIpM!trS^wOul*c%HgxS!R(-N*B{hZ*6FoD<_B!hHd9UAW&-*NYV
zy8|<@?|?qT+<X}wA3O3r{3(YtQk9rqhB;=g=4E)N7fIk@TQ<exp?P<?$BY{C!miW1
z>5yXymQRyE``qNUgpM39G9E-?XzJM&WHj{wO_a@qu!y+|_fY}H#1h^)ECww$pdJ==
z#~Ik7@QA@}$0d^i-SpQtZm_(S)y*G!s+3Ed#3mg@U5%CHJ(-O5?cDG-#fELg={*2e
zz#E^bLboXdS=!35PaY;&3gHx^q}$7bNMyjFk1%W4hj;Ned@l?Ofu<JiDUgUf*p9#<
zk|4`rqaP2Ai9@HMQDKu3@F}5pd83!Zq~Il)GW)ageghBxT_k4aWUswvRSSw^!H}#x
zFLiBwj)<K!h7xNG?b-lB-#^0WhFdPbh1S^#5I5UEV#e#6$85De^sp{(({O@nNou}P
zFU|nADg=Q@Ti~~h+<Zo8eVYHB2;T%?lrkFSgLPL-zvz3_qNLgu(vfQnd^b=%4Sa1l
zoqq)aSagfNIK?1HZekJOM}&~TZ(daJUjRj{;@@WwEJD#1)uNU)`EX+x{OTA6sroa4
zuy%e6h*R6{wcFVEttPB--3IA^z9_GorZF(c(~9%-<+K|n&1kI^N8Jl|qg;dLg)Y&N
zi<hPZECZ}?y;Yc6K9%iZ%6^2x1VTZwj2-JP#P|(3L(ES#pt`vSN~X&N22`H`wFp3c
z4UkeIK&8>a+`vO#9=ITNmVAHUfdHD>C3+sn-Tc7YHyylw*OXU<y=!JSsKqq69WX>t
z1i~hpn3pt>J)s?~S0?DJyqj0Qef^``lBkdTDrJ1u2s-$P%|NVj<K_eSrA7Y3A-%8c
z=4J+}K;E1r8!|IOhE%sj*)i};0?mkllt9@g08J#$1AQx-HPiWD$LhcMyf%D3TuVBE
z48`(&Dt(=@>Nq1S&+wX3;Ty<CHep)nRzD9|zmI{~=wAs~(ccw?T}$|nEN1PlC}M-_
z182w6P|H0^sCx%>aA7pd*Z*5KxIP0i+I;jZj_&JoBY>8KVe%$rgrTegpgUAnOB(%-
z;AY0UEk8kfy)?mFd7B200atG(p_7{DLnH(JEHyt3!kOOhZ_?Nqc@zBfkTOD_5xf#}
zg<W<_zgRIHyZj7L`@%6BJ|ylZf~{=O3(42gZ~dcji6_DPP3Zbvs{a_-vHLwlS5IzI
zw+XuH!w=!i_|o5@Zh|s8T$+f`DnAEg{}^5V5Gh+kMj-2(mTQ@nXMPn>J1?pIWT^cF
zt<z9Tw!CctYQ3T*mKkcv`zENJ;h8!b*YL$zMBs`$r7er&Z*F5w$K4ub2H|-i$BJ6k
zyYM0Kar#X3zrx&ljK-t(&qQP%!a`JYcpd_8Q3nS5ssF8L<HFZ3T)d=U%|RsI+zhhc
zM7H_mw|2E%YY`faE;6VDRs_1)Lu4ldE+~u%?&=9ZEeYt6{jt60?G^S41Md?gK<fl}
z&(8p_u`?$eoWIM0fVfOhijy`cJk-4>;DhVBe10F03u2_~R=3yx{h4r_Xy(newoH57
zwVUa^o8jFszIpT3+U=a2TSVj7y~)<d0Ko-}x<U5=d(ITVg5R?lB{;Nn{J<Z_<XIn8
z!gCo@b{^R@sIxZ0;T}D-@nTDc5Vd}2&`j8+(LO{b$vf4r@a{$#Vr9zZm&+^Ez?@+M
z;iRl^^hcpPxa$o0GhH+EQf6=*;Qt6s*-OX>@Lw_=3qWs^@>5_$|2(Vz>hXH_t3W>J
z5lus0r;M11{2FB^*?_-BBfG0)gqUK#e8L=#u?NI}hzO9q$-amzhlfB09>khRHAP`4
zHUnKeH1^G%!XdTU1B3=gfvGzfrTmdoq|Ry-3drRp|8wN<F&--?0lC%7<CT$uV7+XD
znNt1%D&)5*W8~wE;qML9{S6YwG)v}7#A;tTX`#n(TRL@g`go=EDy(X<R&pOec@RC~
zHB$&w*oQVr0cTOSy263Yd%oQ`fN#kXvj9D=j-_bQV&Od$_ipBS5>OBbbc6P`vB2^Q
zCQ-YkZkmmvB$L_#;^CcM9;y{`3Z}&w3{8ss6MCE<C&5;P2+J@yt?4-_>#P{>q3)+h
z>?DxJRLn&r_?2dV^#pOAu%qjeUhrHPYAob!wB{sc1fX{8eG|^+7eBW4Y|sX+RKt4o
zKrj#B^a8jeoQ>fKPkt#c?D*uQXJ5)oh=?MDwQuj4cM#O0vM{k7bcZo@YhCv7zWN+O
zp-B$`2B3!4GxB5`MYyb+aoXC=lnu|Et$bsVeAr;}Sq5~QF-6%s5x%WNS&+44)Xk}$
zc5)d0;$^ekl^H2h&WiX?(e-CY?7T0VE2ag4u=I<HSs|TWJ3+`NZSen_GyEN<qB+A!
zxG>}muIy0PgtB*#ZD_P6RsxLvWw`?>n3kfq!G5tMd3+9~*hQB=Lt<`4l!6F_^FNly
zwBn5K`rcx(T9VYSg>>X145RBvNWW+o9Wlprf<6IKSf4-qA&kWz{H8+sWp!Hb2@!?y
zmpCR-PfyJRrS8WQUUkH{lcA@G1;I=vkxh(yP62O(TcNKM?<d_y7?wOEzK2f*jkkB8
z>pzCunHTA9-xM~&+|G{?<)HI1Z1uR+ooX%o#NUVC#Qu_x&y4a4-raX=+s&-je)OMz
zd+^_X|I>H-_UfBk{e{}C+9Z_odaYlsweSTl)6vO(Y1Evpf-278F`5<<ox_JlGVxkb
zBs<fHpj;kRD7P!|SyR&IF0ClUb%xRLh8{Z%d^Kp;bP%kuz>l7U%o_9PF}Xd?6#pZh
zu0vy1)vUtPBD06{^{oHWF*`Z2y>y627nLcIJ@{qm_dikJkC9E2{ISLiKUy6*u%w%$
z67-g=h76lIu{y9gLu+f3O0;1=v{kj`_t94$njfGvZ6X^U3#{L8Ag5!4K>vafY$i1$
zn07%YlO5U*VA>|V>Ql<bE(4$c6Qbzkhv-l`IYLw5H;JOtQT_*{VHm~&0>V4+@fDI<
z>gGoMtcD2C27GK;O%nM9G^k~L+smUkTb<xCgRd%2k%^3BvmAfb{y!Z*shJ2q6R*+}
zB=H(aim1F>{Mnelf%0cTj_7kRoOPL976uod;MAI`^bF!3%6|vk65pQRCHBfZkUp%I
Z%@xas5-!?Rf6*(wSoNx}R9Ba-{vWFtN)G@4

literal 0
HcmV?d00001

diff --git a/ldm/modules/encoders/__pycache__/modules.cpython-37.pyc b/ldm/modules/encoders/__pycache__/modules.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..bc5a031c90b435bca4e6e86d16f3f961c1d0827c
GIT binary patch
literal 8516
zcmcIpO^h4aaqfRMIULStG#ZVx+Fd5Gjb*SJX4k9uC-$ziyOLLlG%HJEJML`I=J54!
zdZyV;y>9A<;Q+yoyaoaY$ThJIj1D>2Ctq^#IY8p0k=z=&<&fwQAh!TI<*Rz#B!|){
zSzyydKX$*@@2|S*tFNlxUtFwdIL`cQGiblAY5zty<zeCCGS2j$Q3y@wk>;s?y>EDi
zPS-|c_ASriYcsO@j_33%UZwAPZolePxn7TIUJcJ$(L%rO)%%OyqON^M6Si<3X~GGN
zy~d&GEup3&T-3P167H9AUlldn*NP`ji8WCdi;r|~1#JzngtjGq_B8I7#VOpM;`>$H
zuZYvQKh5`NaK9?f;QkEXuL=E*);jwX`_eXBdS2`92HpKZ62@7}$``gSZ6*6b9F78+
zH?}Tajk}2m<V_S&zWBk7>$j%O>Na?gacj%SD<atG45KWscave9`Gbz^^wTcQR32t=
zk$wpmf#&H#^NgQpJG#(?@yPW|VG8S!=2<_{EX}jAQafL|vYlqK)6I&-boF8o`ZsXW
zplJg|rX6ZW`V);rkFNSA7G<Zy0rukhei(<D?_<kW8bmwzuljyB>ZGaf|3Z6qyS<zA
zgLa%`!FH1Dw?!xGw0klc4%&By@r^L<K~Qm<#L@mR+}U~m`t5ck`t5!qhEb5V10|`n
ztk@VllnYo>ofb!{)b!E%+)g&iMNPIuTcK@zj5BpnWZI{Yk=l%&pWaNe>-|9#^n*AH
z#8oL1*)k=~ky{<vOLJ%cKK)A)hboq+(NW8z>vAz`^NUfBv>cLE-026tpVxd}?GM*=
z-@iNTM5=|iCQsAQi-i=sK@|D^bmEpS7ikeRr>vqNX{2OWu3k5}=9p@bj4sn7t0+p<
zV71UlEg%9bU%D=^@i2Q>ssalZ&(lB2?E{=V$C$KS-{j@A3?(7`!E|+MS+s#O)P={<
zM{9HAZz!2;wk%1jlJu-xqk?+QjK<>Tovx#S8hJ?9kaT}+pCkb~(zj7e_PNMLU<vzC
z6}xQ+S5#r_mRJ%CqK+$@4Z^{umSKmawO9Mwfq-p3{SPQjZnGx;```Wz9lw6BYx5kM
zM*lP;p3%|`6jK`+n`6y4GV{n98)N;6fjb*ln(vG)l$BELTuC^|Z#*~mgNL~h$9Yvv
z53|Ut0k4EFV`Wi6n!Cep&zGGnXxU2o>M>u0efhgI-fL9+9*R;?rZOZ=Igc}CfYxie
zqc?Dl){bp_Vm+JJx}{JZ^rG_v>ir1C*bv&jD;r~NPdmUi$HqCrZtcgY-@)|(wz>!O
z*|+4qv6<czMrItDIIXeuM5ne@?TNmw9oS=gY^`d-9NX(!*}kgb30s)!+NySz#<Gr_
zCp_j+h5AzsDeh(uOW~uq_^<b#lDf9ssTguQOJsL9uWko2PKHqw=G7z)Qa=j!0~T3s
zMM>O~ze61zB~*L6lXZ7_5N}C@X=ghMd?AxTGR$(ln_K-(x<8|*3<i~*L;@uQ4V13X
zvUFE>j8zEvy#6Iz*Yweu<DxEl<W_emI$csGI!fhYX}^gIlCmIanVG%^Y41SZ!gy?q
z%~d67b8LmBRLo8JW%L0MJ0iFjb^}-$YsObln_IMZ-dL$X#&#;(X!4+1O3iX(*qz_p
zYrP=zhXCVPgjtxxoyebDLq#uQ00veKyIiH6zIlA7FYgYpeFY=P3sk&{qGfDeg94Lf
zwJP%<L3%7{M)_4L&QifzPDQEqj2fiv=WwQ^ysqA;)asJzwz`Z=xT62#2*+MwwEgkD
zK;D;O7Bm6XL9_u^8OF`~VYb`McEhx}17VOLnpx89clLv3I+Q_^pK1s9IuUoPHF1a}
zNSkng&F*d|?gh<G{IH+MV1w6Of~ZBaOn1RJnl}p76F!go4%W)}r<w8+uDduVM`^0*
zH&HdgKb=sIG0+heV4%ea2nrx-VgZD}Vfv$a6~;jqMzxeLUz5owh`~|{C>*`<y%bx>
zc7x{DCE8+~LUH?nY$_a=P24s!P?l$_y(H;HK|8y2@!nhhTNfjE&46Fy)0<e=)-eDj
zeWx>h7sV7n2`K=aKndT<?8&vmAX;H~b;p+c=2$zb&XSVa6i_7%n=kFir0)-8z|spu
zUd>7pCP`fR!pvUS2J%|J^T3aSxR>qb&W;R%(FABKP>2*SoFA0mLVx*f6s5i$hgoW7
z?nL%b$Irpf(nsg!WjjIJ30EPFq~(7=L%e~4te@x?<e@!S!>)}d+K1Yo*A9TMW8F9J
zuKhp}GqM>={t>$57O7xfW2zWrSa`lDgS*2}0{D8eBTBWi`H4Bb<9%JF>B(#~y>2x0
z(b?x#uEaC3Bzl(k2(SAxE@<6jS|fV`VGsuu_Tvitg&~}C*auonpgA1oo;5Z_<uPN}
zo=si>s_1TRU1LG##vlWMs$~*J*CDG@)!0)ulyUjO&K0)_-qcQ%+y^fC{KnI~GUy0F
zuC!4+lLc{_KrGcT9t<--6e(Whb*vgLfGnt(7rukAQ3>@F_(A6HhM)(rpJwEQ3x8pe
z^5+fq8a^8Z&E=Gn!)suYG}lM-k|q@U6Pzh=O%0x+t}jF0aLVh?%UijZiNHxw=wyKt
zsU1VyhtCr<c>`_p6BC!(!m04&MXy)cHWcb>)SZx6ew{kCs&japSA1V2U651;W-<ze
z&P?WBg3f=4CrQ@KM_iakqv|?tRZ@Ky=j32D#9jI!ieuQuO4T!gYb%U$glbiw8wGWQ
zZXD8i6&1>PE~8u(s&zVFQSznOYl3s%p5Po+zZB`3Z|neEn!CeZ53hQs12+pE0sY9G
zL6SrO8%HV1vsaTr5O>0MHwp(A?}eGd#ES!B7;n8ZBiPMurPwk3o5h*_6)vXW2)d;0
z8T(GUo|&KOnRR3rQ0Ziq$+gQ+S!JlK0V+QNXe<y^Qg6RLc2F*kZO}7Ael~VM<*Ml}
zjR%_67(0i$h3WTo`Io}n*VE5p1m@gIuF=|*-@}R|InAe}wXLNo?&i*oD?hw?`%d1t
zlBR(qOXXln5~Y#fpn{EyUMSnt9GK>I)Om=ovdVw3$_KgG*-k+LXeLR1Vy=husP4~D
zycmdmw4~=zXhx0AtAt`uGyD)U>7!SkSB3u{SeEZ$9r7|2S5OonEZ?V!34u`|6j@P%
za~6UN=P|3Y8!rQ4ChDIPgrQxwsO%fUeQYp^w?!3Na%*huIR|$92i#smO@<o8GK%u6
z{~Z&=-;V2Oaq(Q0?ik1yN+6enRRB%0V@6m>Zbn|C;yYA)L<PYxD;Wet*ZHaf&|}b>
z$~*KJVOXm+g>3o9R6D_Jr5=Q^-Xhzj^wn7T;w+A*N`n-dE4(DSA~&h{n2K9e%)KK5
z?;%bOtxNDG@=H|zyaCNR{{qb_IR6(1<Db$J$PpkF@*))}fZUPa$0IX%!vfrd@K&jQ
z8iRkAZvT{u&xz-^(TDnR@B=(I4BbEg1j#G`*B2$VAH#J*1v=7CP|V|c&j6~gF{)G4
zPN+V|SqpS`#o3^`R}-gxrgMtU_7+rX&Z{%tF9NmJ^ELG<J_Z=i8qnzL%+V*nU$(y#
z{J9H@FJHiT$jya)uz_dp_rtA=A6<Owog7pF*!T9fpMDj^_Se3#ubgMMbY<kPY5N58
z98sSP@q2*y%sHwEaAIOnkIhfa1LvS}P#HTvav990P=9L<>b1SPuwbscf}npfUWZ}U
zg@c-rsBn#wH4e*cn7f&q-NA75-IKF+H<J4|f_p)9<HpVQos0tGJGUu07er~BJRL&F
zaNrLfa@0*}F2qjTVGL6yqXZ?L&~atJ<VWO>sQ5C9S!bej@Q0{H#9of<z0#EdWUWT_
zJ2A&bDj-SaS7;ob$GI1#NUP;GuRvi9Gn0wGkjpfms{(#x`4I23f0;0Qg&mYx$aUm5
zRoaaqV2P$7<M?Zw=>-(ZyIe3V-7$&ilXp=#;2NXXU#xmxypN%^eoi0rceDdmN_axh
z{8vv{>ztpV@QGLo;d%P|C^Cd7#6&>uiP%%Jhf+W`C2kJ&y~-16Gsl!~ME=;70s7=_
z8IR>>6-fCMUsJGKMwkrexfO$6OWxgO^y4=SV#=I=$t~mbNf7-@bSbu2GcM?(*G}5v
z9KR;hBopO55_V!g(_K>V0w#v^6z`TYs=|C?(*6<5s8(B8@S@MRYexuXSyl$*VXK6V
zVpN3Vltb@iS?p7m3EYP^?k%yb{Ap5fb&#ZGDrQ<+K~}y$%z}KKLXufg`i0|>F%9QN
zvA+`{jZ@^gmD8hK$-I}_N8PV*7U3)fvkiTPxzuRm1W}&wHt}IZ(iF!sT@w-SqUFJM
zMq);mYtKB?5#`?c#5gcPM3g!e^>_l=LSWe2x5}p-xs_G+T*?TfTNy|%<s=Wy1O4u@
zx)wTK1@3-;APGV@Xe}g()&erE@etVzpo-ElilDtJLgno!>F%ewiG%<X1aF06XMg%>
z2h^W2CpY&(e4hzZKkdqJkWoFqNtLAKP4jHI9_G577x7(8_-mYnh}ZRXI8NjZjb(k*
zK0(AMz1}a5SNsF5sNfYR!9u|;Y}vuXa+?Yc5*gdc(i{J0c0qckgz0Q?a4tzK>Wjcj
zf=D9xpW$Q@=9-oX%3*?XthzP@lPfhzPjqok4$?R}(hU?7+Ofb?gT5frrvm<N2m-?2
z3uitP)brEog%F9De8~G>g1l#^=F(yhqA3rXPy#|>N}v$#q!PuH5@eKON0JvxOkWqs
z7?Av>2#)XV3=!R)xd#LIDL$ZPRk}OWg^9$v^;0BW$Y6?Ahg<EL*2>S_L%sO6=aN~F
z2OQbW-A>x<J^s(XKl<-~{MEZ7bK~uu(NgnfbDC3nvpK3ayZF49s=Vl^RyL;_pp`WY
zqbMX1KYYC;VxSa8a*U6J&ZV*<w+9m6N|lpkLP`v&$gg;+Z;^oWsM~BZ&RgQ4r>Q${
z5l>8SU+lYogYi|$8j;~LexnO7c(YR{&6a>Z^iL<d1zP0iO}qRSeM-BU2@odARD2zq
z7&7@<(g~ihi3_~VsULz5Dm<@AI?*GK0=qNi7*8teIl$GdyYw8E*T}OVrWdB9{slcI
ztl)TrJY8ntPw*XKL{t3*6=YG209>4qy63zQ1y3(a-9JS8KjGw8<m#~Kifc-$ZRuOB
zvkECdAi&!;;#j_phI~!k4zn;yH>N1fP@LCiXhq5iVon09)CAdH!AbD-d4oof#BU_M
z;M}>%U!(csDt~$8j|iNDXGP}7hH-`$I0Nqu8iVL!?tTkQ6rZ%;CBn-Y3CcaxjdjD-
Vm(3M-*;#p|<}AOqys>=ce*h}<Xr2H7

literal 0
HcmV?d00001

diff --git a/ldm/modules/encoders/modules.py b/ldm/modules/encoders/modules.py
new file mode 100644
index 0000000..ab8b31d
--- /dev/null
+++ b/ldm/modules/encoders/modules.py
@@ -0,0 +1,215 @@
+import torch
+import torch.nn as nn
+from torch.utils.checkpoint import checkpoint
+
+from transformers import T5Tokenizer, T5EncoderModel, CLIPTokenizer, CLIPTextModel
+
+import open_clip
+from ldm.util import default, count_params
+
+
+class AbstractEncoder(nn.Module):
+    def __init__(self):
+        super().__init__()
+
+    def encode(self, *args, **kwargs):
+        raise NotImplementedError
+
+
+class IdentityEncoder(AbstractEncoder):
+
+    def encode(self, x):
+        return x
+
+
+class ClassEmbedder(nn.Module):
+    def __init__(self, embed_dim, n_classes=1000, key='class', ucg_rate=0.1):
+        super().__init__()
+        self.key = key
+        self.embedding = nn.Embedding(n_classes, embed_dim)
+        self.n_classes = n_classes
+        self.ucg_rate = ucg_rate
+
+    def forward(self, batch, key=None, disable_dropout=False):
+        if key is None:
+            key = self.key
+        # this is for use in crossattn
+        c = batch[key][:, None]
+        if self.ucg_rate > 0. and not disable_dropout:
+            mask = 1. - torch.bernoulli(torch.ones_like(c) * self.ucg_rate)
+            c = mask * c + (1-mask) * torch.ones_like(c)*(self.n_classes-1)
+            c = c.long()
+        c = self.embedding(c)
+        return c
+
+    def get_unconditional_conditioning(self, bs, device="cuda"):
+        uc_class = self.n_classes - 1  # 1000 classes --> 0 ... 999, one extra class for ucg (class 1000)
+        uc = torch.ones((bs,), device=device) * uc_class
+        uc = {self.key: uc}
+        return uc
+
+
+def disabled_train(self, mode=True):
+    """Overwrite model.train with this function to make sure train/eval mode
+    does not change anymore."""
+    return self
+
+
+class FrozenT5Embedder(AbstractEncoder):
+    """Uses the T5 transformer encoder for text"""
+    def __init__(self, version="google/t5-v1_1-large", device="cuda", max_length=77, freeze=True):  # others are google/t5-v1_1-xl and google/t5-v1_1-xxl
+        super().__init__()
+        self.tokenizer = T5Tokenizer.from_pretrained(version)
+        self.transformer = T5EncoderModel.from_pretrained(version)
+        self.device = device
+        self.max_length = max_length   # TODO: typical value?
+        if freeze:
+            self.freeze()
+
+    def freeze(self):
+        self.transformer = self.transformer.eval()
+        #self.train = disabled_train
+        for param in self.parameters():
+            param.requires_grad = False
+
+    def forward(self, text):
+        batch_encoding = self.tokenizer(text, truncation=True, max_length=self.max_length, return_length=True,
+                                        return_overflowing_tokens=False, padding="max_length", return_tensors="pt")
+        tokens = batch_encoding["input_ids"].to(self.device)
+        outputs = self.transformer(input_ids=tokens)
+
+        z = outputs.last_hidden_state
+        return z
+
+    def encode(self, text):
+        return self(text)
+
+
+class FrozenCLIPEmbedder(AbstractEncoder):
+    """Uses the CLIP transformer encoder for text (from huggingface)"""
+    LAYERS = [
+        "last",
+        "pooled",
+        "hidden"
+    ]
+    def __init__(self, version="openai/clip-vit-large-patch14", device="cuda", max_length=77,
+                 freeze=True, layer="last", layer_idx=None):  # clip-vit-base-patch32
+        super().__init__()
+        assert layer in self.LAYERS
+        self.tokenizer = CLIPTokenizer.from_pretrained(version)
+        self.transformer = CLIPTextModel.from_pretrained(version)
+        self.device = device
+        self.max_length = max_length
+        if freeze:
+            self.freeze()
+        self.layer = layer
+        self.layer_idx = layer_idx
+        if layer == "hidden":
+            assert layer_idx is not None
+            assert 0 <= abs(layer_idx) <= 12
+
+    def freeze(self):
+        self.transformer = self.transformer.eval()
+        #self.train = disabled_train
+        for param in self.parameters():
+            param.requires_grad = False
+
+    def forward(self, text):
+        batch_encoding = self.tokenizer(text, truncation=True, max_length=self.max_length, return_length=True,
+                                        return_overflowing_tokens=False, padding="max_length", return_tensors="pt")
+        tokens = batch_encoding["input_ids"].to(self.device)
+        outputs = self.transformer(input_ids=tokens, output_hidden_states=self.layer=="hidden")
+        if self.layer == "last":
+            z = outputs.last_hidden_state
+        elif self.layer == "pooled":
+            z = outputs.pooler_output[:, None, :]
+        else:
+            z = outputs.hidden_states[self.layer_idx]
+        return z
+
+    def encode(self, text):
+        return self(text)
+
+
+class FrozenOpenCLIPEmbedder(AbstractEncoder):
+    """
+    Uses the OpenCLIP transformer encoder for text
+    """
+    LAYERS = [
+        #"pooled",
+        "last",
+        "penultimate"
+    ]
+    # def __init__(self, arch="ViT-H-14", version="laion2b_s32b_b79k", device="cuda", max_length=77,
+    def __init__(self, arch="ViT-H-14", version="laion2b_s32b_b79k", max_length=77,
+                 freeze=True, layer="last"):
+        super().__init__()
+        assert layer in self.LAYERS
+        model, _, _ = open_clip.create_model_and_transforms(arch, device=torch.device('cpu'), pretrained=version)
+        del model.visual
+        self.model = model
+
+        # self.device = device
+        self.max_length = max_length
+        if freeze:
+            self.freeze()
+        self.layer = layer
+        if self.layer == "last":
+            self.layer_idx = 0
+        elif self.layer == "penultimate":
+            self.layer_idx = 1
+        else:
+            raise NotImplementedError()
+
+    def freeze(self):
+        self.model = self.model.eval()
+        for param in self.parameters():
+            param.requires_grad = False
+
+    def forward(self, text):
+        tokens = open_clip.tokenize(text)
+        # z = self.encode_with_transformer(tokens.to(self.device))
+        z = self.encode_with_transformer(tokens.to(next(self.model.parameters()).device))
+        return z
+
+    def encode_with_transformer(self, text):
+        x = self.model.token_embedding(text)  # [batch_size, n_ctx, d_model]
+        x = x + self.model.positional_embedding
+        x = x.permute(1, 0, 2)  # NLD -> LND
+        x = self.text_transformer_forward(x, attn_mask=self.model.attn_mask)
+        x = x.permute(1, 0, 2)  # LND -> NLD
+        x = self.model.ln_final(x)
+        return x
+
+    def text_transformer_forward(self, x: torch.Tensor, attn_mask = None):
+        for i, r in enumerate(self.model.transformer.resblocks):
+            if i == len(self.model.transformer.resblocks) - self.layer_idx:
+                break
+            if self.model.transformer.grad_checkpointing and not torch.jit.is_scripting():
+                x = checkpoint(r, x, attn_mask)
+            else:
+                x = r(x, attn_mask=attn_mask)
+        return x
+
+    def encode(self, text):
+        return self(text)
+
+
+class FrozenCLIPT5Encoder(AbstractEncoder):
+    def __init__(self, clip_version="openai/clip-vit-large-patch14", t5_version="google/t5-v1_1-xl", device="cuda",
+                 clip_max_length=77, t5_max_length=77):
+        super().__init__()
+        self.clip_encoder = FrozenCLIPEmbedder(clip_version, device, max_length=clip_max_length)
+        self.t5_encoder = FrozenT5Embedder(t5_version, device, max_length=t5_max_length)
+        print(f"{self.clip_encoder.__class__.__name__} has {count_params(self.clip_encoder)*1.e-6:.2f} M parameters, "
+              f"{self.t5_encoder.__class__.__name__} comes with {count_params(self.t5_encoder)*1.e-6:.2f} M params.")
+
+    def encode(self, text):
+        return self(text)
+
+    def forward(self, text):
+        clip_z = self.clip_encoder.encode(text)
+        t5_z = self.t5_encoder.encode(text)
+        return [clip_z, t5_z]
+
+
diff --git a/ldm/modules/losses/__init__.py b/ldm/modules/losses/__init__.py
new file mode 100644
index 0000000..876d7c5
--- /dev/null
+++ b/ldm/modules/losses/__init__.py
@@ -0,0 +1 @@
+from ldm.modules.losses.contperceptual import LPIPSWithDiscriminator
\ No newline at end of file
diff --git a/ldm/modules/losses/__pycache__/__init__.cpython-310.pyc b/ldm/modules/losses/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..0cbd09bc40911b0318a68be8191a141c5c75874b
GIT binary patch
literal 251
zcmd1j<>g`kg0}yCsV+eJF^Gc<n1CDyATAaF5-AKRj5!Rsj8Tk?3@J>(44TX@fuanW
zjJL#m0z3nP!!t`VTr!K3i!yUF^Abz)i~Ka1q7-vda`kfaQ%ZAEi}iByi;IC&a(-S(
zL26NQYC%b9Vonjrl9dca%s>iE{PNS!$j?pH&&w}KP0G*D)=$n)N!15xD9+DI%+U|F
zG%&F+vNVMVJ39vl=>zT1huWbJu|q#TJ~J<~BtBlRpz;=n4Mft85okp*$T|TA0E7}l
A`Tzg`

literal 0
HcmV?d00001

diff --git a/ldm/modules/losses/__pycache__/__init__.cpython-37.pyc b/ldm/modules/losses/__pycache__/__init__.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8d2d86d4caa685bca01c7646d729bff26f841055
GIT binary patch
literal 248
zcmXwyy^g{#5QOcdK}eCVqNhxgeSy;np)?T)0V$AZqMU7rb+$LMoq|W=m1wD`>8Mza
zyOCyQe~NFOWn)IlW(!qPUti0=5*lXY@IfJ&<c3+k=bXu;W_fy}E6e$1vRqf|dZ%Mo
z>L9$f+O-k=Do^%PBW+<(4n~E-U<gzKUF?+?s*gi!{0nksKBfAN#|ApI3LHlDjraqy
zjV&BK4n5R^TWWU{opK;hDk$~gFt4^i82H1$8v{+FosLaY^ry?{^ZoqdM7fjM6O`9Q
AOaK4?

literal 0
HcmV?d00001

diff --git a/ldm/modules/losses/__pycache__/__init__.cpython-38.pyc b/ldm/modules/losses/__pycache__/__init__.cpython-38.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..daa879d25406c155759c947413303583a77560c1
GIT binary patch
literal 258
zcmYj~F^d8*5QUSqkpscX-f|0fz*@w@Mi3FhA=oT232rdSEJ?QT$M^&MOSV>5Uw74w
z=M_G9@9}1e?{hHtV5I5hs+ZJ{=Xh5^V?j<s3e7Zk%<}{1Oc#Qcy)U}5lHW(`?P@Fb
zHaAO~RJ7g(nK5q4;xKmFR~~ieOsX8FM5Qp~)<k96+({RokQ<9Vs=xTzLxbKxz-*3)
zXV5ZBI7RGQ5M5Z?a8iv4pimnSSstAcyNQ=J0KwoN1}_-YwGB4cb=6+({n_g|=fwIZ
FvtRc=NM!&3

literal 0
HcmV?d00001

diff --git a/ldm/modules/losses/__pycache__/contperceptual.cpython-310.pyc b/ldm/modules/losses/__pycache__/contperceptual.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..ffdaaff11de380387e56ddeb466e98af387d539c
GIT binary patch
literal 4174
zcmcInUys|y5#J>#N}~RrPS)wpZsG<(P_@mS5i~{<7)b!PE?T&UAT|($Kta%2-W_!m
zsVu3ptx`bI7)@VX<Sh>c3hAXlfEMU?=(pfw`p_DEX&>9dZhy0+bf?CCDNqs|?##~4
z&d%)Y%tnod#qj*%-@o#H^D<+9BIM#@0C^8z{04wvf(Ojb|9q_38mG1v=y%x<n9zms
zhzY~fM#d+)T@w5b8<d`496KEFjDK@T&5I8Srzd^~z+*NS%+{*(8r70$e#(xxFh%)M
z)z(EzRE713+eWokHAGv~MFUvrQ)V#R#6(+J@AmsYeE-geesc0XKORbd?1yd=$v>+p
z0E))n-hHA2yOotr{P5TtHdg)kQ@ezZ7~;<W67~sO@Fn9ctqXp};@Xec&v~k+%+XQ<
zP)`hSl$H!_Gpz!alggr+SfiTIg#MVPT9++3==G$LnrSJugz;EkG?SL1wo~fcg7ng<
zCf`mvAy2kJ>854COKAnLm(~Dx1j91T!+|Et&@3yTxuG8f?iUQkHDZ6}4RpKWhm&dI
zgu|>7I)VGplW|s?#-8I2lQ5EFHFVI=O7V2!$*kfye&{ETlT}Xx=f3A3pCopD;y>_K
zSXNS2%?$6%4YC#n4ZTS+bpt1e;yBy7K*)z@S{P>4_g(3Zy#!N~;DX^vW<;SEXJ!x`
zpSjZBKD<r3ZTKlGxsyro(5~llVx?Ku5Ep%!^*2z*#|%!KxDr|`Z#qKhWj71s6nIvp
zak<Jh)jWbhBAInJ<rPb4E0$pt3P16q&<zF+SlA1WY)kd18M9ix7^SCd!6kNES&j^|
zvSmy0htqN43_BO?htBZC4MQ)8<=1Fru2ep~+y+I~2_PbwyaMnA&#I0y3|xqIWEGTu
z-aR;p#@;~~CEmR#Iz1Rh!aIOFVp90v&Rf^M^W8VzdNU{Aym{w?gFuWA#*vr?UVK10
z;)f+B=N{ajJd{m{uHcJb1+YqWZsH}_*9=YL8aJMqy7AOBI;zLwvzMPAcKfTXbvRsw
zIV3XLx)fnI@x{LdNOi%cTwYGth>tWxgD$wx9_vEiWpksTY8%v2cxBDaMvC*fW4_u~
zO4!0kjU`wA(vk~fhh1iK6L={#MM;=TePJ#^L9DV-xvN@;yYk^(gjrpBi3{f)OLuZ2
zIn+OX{pOXwetCWGT}i7v#>dz%|M=dZB710M6?d9M;1-#opal;m`CiBj3dpPkr;aC?
zDIj@xB45I&L0Mi^gTtT*GxJ#OUHJ`=?JAK;T5Ok(^F~$?#g0>ZBM0QcY<K7e!)f3q
zI0A$_N&GXf;G`+y4nT$Xc%NU<uE56~ZgGRpuD!sgFBFC_moUt*26?Mz7mY1^%GGbf
z=`b=+`9i~CsS)hsV4UM<gA4hoA-|rIb1w8RwRxS6zLB2RyKIMnx27D$M&^=V#*sTR
zL2aasc{8o5ZFC(+a05<jg0Gx5al|%96~TAO$vLe7T4POXKt@%~{*$z{u;BQX8cBz9
zYt7Z7?RTnu#^uwrjCq)hF|H<0RWe|UdcGd4rXAw&9@XLps~&9BM&6s2k2IAMG~!dU
zw3ocb=35xi6b3Bt`<aDyD{Z7(v>6}k^UjAX>=`U+Kl~j>65?dTEo@RGjZp{d?0m#R
z^L|dnww`w=d5oHrnnqnwh6KXZ3ZcE7HVI2xWX-gbZg*L_r7YVOl?9Rw<aQUmC0p#^
zD=(R-7F1Z8Xu>M+)|#VMP#SAWy`a?Altw}6C>hNhqm`uF2*Fm-yS3u?EAvaK1${sA
z*}S)@QQ8yj8_M&<4Qvb8S{vXU;A>rgZv$Uz0=x@c>4Em88_IH|1+X4^);a*+xxvEE
zVX^<kPX7<9lnbl;{kc{t{D(CZzQPI$?}4^qk=)*(ZrEi_U0Y{GS$T?Vw4!ZVYE31t
z{UhzBm(t#meuq)(NW)(j8sbNjzs}<atvy70J&$(sB=+u(qNtOhptQrQL{l-O&IRT3
zi*cmJoX6N#`rFTm2Smi8pYEjXB}%h-U-WbB)BeWJ-z}&+BvZx0UO{QED88;dqRG#d
zM>fX8R)t*Hc`XCfLN2IlIiMD@LG7<?hZfmRm3@~FC1sl@1dxYx;g40q8E};oB&Af)
ztI3J<;*%&4v&K*72T7C^h1#qNG_O}?7MiH`s05hR&|YO3g-9<ilN7Q_DzhfoR3(#_
zO-fpxk}~Uj64{3qS-FrczfD+WB~b1IRmF~c9j#e?B~s}@6F@ybi5=;=fts+ukK9vl
zu$!k--MtrQ7F8|&@idyod8tH6EYrOQlUxB~?71N-+!Gh24b^3tb{gcR)SluTx=F6c
z0ipE|rVuywzAq&TLNoEgIFc$4+jUn&lVp`yFB7+Ugrw~zl(tdP%5Q<D3N&bVRrRX^
zJkDBJcNE5noDQjkjI#zL;Yu<?!Rm_#S?xH8?!h^6;!SL865%iXNXE*YnL)KDmM{tY
zBr6vMv2CIHn99%*{!m`WWVSiU%L1~ry_K_h!dW>~zDaZ{Q^~GmM`ikRPh<ueKhwcs
zSExy%*8|8As7vxGsN}p*R+M3e`>qt^%J>feJ>BA6-sC;a;1#@Gyq5Y~dPTE%n_DO_
z>VO)0>b$FI&#Y2MtMfhn8h@2{Ft!b7{L3sGN>Uqm8=OSfe%97rL+>Wn9{75AJL+#@
zmI~xsC}O(&s&<ukv^MmiQ7BqV24*5|aB0S0W-q;<#3>dwsHyl;ft3T3ET8}d+j+;W
zTluAmAL4R$cqp%-_wd%BOqZ(=mmplOEXT<UZ?x-<^D(YZd5<c$C2d1_8z5^GHU7{;
ziI>;;xzS%C7Aoi@T}`Qoj=KOAMkP7D&8JphYihjnOmPg%qW)C3ul+ET#BRTmxVRM`
v?-$pAee$^%@1K2qehJ9TAFDfp5|dB<U%s#0$cMi}hfz$Ig$k>~s6G8Zjtgkn

literal 0
HcmV?d00001

diff --git a/ldm/modules/losses/__pycache__/contperceptual.cpython-37.pyc b/ldm/modules/losses/__pycache__/contperceptual.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..ff6e749627f1a404132ca90560a41e3b11658c96
GIT binary patch
literal 4166
zcmcInPmdc(6|d@Ux9#@d@pwF*N!SgriZHO_6%x@NqD{z#pq0WX+N>1l1Fep$$8Jx%
z+gshvCPVweMp`bB<`bZ?gpl|O9JnCy5jgY_Bv1s0oH!s3{9bk26Kgjdkg(O&@4b5U
z>eYMox_))1(XbePfBdKK{pxMT{zA<9XMnhkC;lE9!2}POoxk}^vo%h%7U-X_A2Fc|
z<2e(Cr%jE|b-N_^LpCVA#5i_5;2Hn+h{*M?RuIat$80W`t%>Fr?1T$bl%H2^U9?11
zSkJj_h_<MU2B^{(%wV<&!#i2;{)3MmJbdIQle>OAlK#vO-6WEKThCAsuWr56A-b29
zCVn{fMvbDM-X%QL8K0v`*yrqmuNY@(UGP&g?j_nUc&ew&(NhDhk(8iet{CWYT18t)
zsuxyLo7ROU^cOtUx@^TEZzRpsOiQUHj2HSvD`_j~PD*_pSTCJwGD&trE$KqCmzL4?
z(hA!Bw1&1X7?x-r4>VbZXIc4)8~Q=u{+1z-M(m$?1KqCp;e3%e;V7$wPT)TCWSrF&
zvFEs>B#h)t4ILb0rFb#-WL9w;KlBsF$(lciV^1c26#h`kNM_a3!1>Jc$CJda&;6&~
zn#)Qm!kOVcaf7S{BO`B~EZo2eqBzcWUK8ZQGc63W>H}B0GcUoEC8T3C$&4uU;>--9
z@e^0tyT|v**R7moC3ij#p4s)>POLo58shc5X8kSJ@iBuFC$5CoDzlDIe%Z|;SOuR|
zX<Y7dO_3)E5SHEDwpT2ntyD%)D13^L8w?tVxEGw*mg-S6X0?1V%1_xwitM(E90g__
z%a+m)7qcQcc5d1aozcV%LobNsTQo9vDxY3<A(3@zu2J(fG)Q^XaYlg)(~hh{@~=n3
zNi_3@VU&1}qv&)f+{7J@Wwe+NA1=cC*yp2(7Y;{}@P>E&lao6ieLM`rY&eU=BJkoN
zxrrB6nOr9EX#Pw#VZDMUzJ`XCEMDg({%EznW^j#H@Hcr!NzdN?>Iyts@3i9)eLm<%
zyO@gYLp<?s(4@Lxr&W0)VN*WUP#C)4LVKYL{eUfvLaJ>^OX2$)Z8lY!UpeN>eWiq5
z7^$)1>RnoKVeGRjY-xforKTtebEQ*ZFHMv!o0hwZLJ5|SKSB8G(o0<AZ7kjSL~^)4
ze)q$xfB*B1!&_r8>X#G0{KM@*MeakARoq1qVT8yG)mlhkl5d5~p!&>8$mwjJnF7{F
z6NyhRCX`=SgTtVxGIOl9uKX6nc9q0rEw;<!oX9Go*l=oV<jp%|<6z_lqeb8*_#y~*
zp7>9^Lh~BB;|>~zZ1#9xyUMR>ePq?*XYaisr*AZdDwhb(X#GJyfY}{9D%E$9bOf2F
z{6fQbQ=_(zZ{!kRIi!%E8uCs`$$71Rr7i396`bm6z039)bZaV6Y-+Cf6?}cCCZvtD
zv23PQwT%LNjT=Z>6ME&eiLY&QS`qw!lAOyLkTo{425eN-?7vG(7ZwuVQX^Z?YHhSy
zwEa%Cb3+<w8S_-+UooyG|CH2mF0hM6z8<Wm9bTuN_|>`x8?~ADrsWe&oemo5sabFu
z)Yx(dBU-{ZHO{$w{0!&7w4FB69omjB^kwG}3$Gh2**Ue)|1*wLiBlvy*s7R&x{K9z
ze#%1aCoHs-9NT-@q4Q+gqSI=+C(5w13rel@T$8^~o5ZCpifY<PVI$p95$=l01<oH>
z?q2M#*hL>tdBsGvkRswF6IQ{uHX5};(%4Aqg`~ETGzv*a*=Xh*`JC>e6kA2_&RXBE
zEca6DT$lgw*|N9oQQ8yjn`#G;Hn<&d8*jjOz;Ap3zYBij3HUB}<p=1!n<}E@1-Kr1
zHa>vgzsb&dc!Ze!FOm9x#HIXS<5FZ0t0=OCH5B=SY%3;t#39{^%tpG2&swt16-8++
z+m6*nN_qQt+D-S;-iiK@5p|;BtqTp6RFb#zN<vM2pejh!J>?R+cmJ}c3Q0S>O)^zY
z8_lhXQe!TwP3?;9iu-a!pe8O3(*3l(0$yA8ML+jG?QiY<gF?DbHdQqo7LxW_;_E6S
zCHcPEsaxX_tHLhgys-gkVHeVk9Y_n?koGsR14Xe@Sn$fRI(J^u;e!)N7yfK>^bEK<
z9SZ){EdmV5M0)Wg3dC9C=gVOdB?W3btAfqJ%b5id(2u~tSq=31%u$T=a?GTd^$~N{
zgqngiIfhc!@&hWfE+>(624v;Jw)`$}m9;^61Xdv*`7Wrl`r4%OgC>A`eiA#<a|1PD
z!JoLN-rz7lTy^(xoLPil{P7}M#5r!-rHOR!={$G9n0amp_&ad{atNel+G&vEsuuZh
zMX8V6B==@WZ2jW}Y|gwK8Jme0#*tJ<vt4&ZG*8xt?N!n?PjHUA37zD?xbiz>RYA6c
z{1`VCM31u;79WLiA{Qe9m2uX9C0tm}0AYRcG^>q+=rJ-BC*It)<`Le~k7TU!ni+&f
zv5tA*Ct10G$hHMgvXG%8{E@tYiEVS9BLxb#y_2hX!dYikexKw7Tq(E|OcnpjbYuo)
zAk(2?SBNCjn*o#!K$Uz70->)VjWWXUnJWe5GyXLiruTp)y1c3NxS=p(7n}vEt6N$H
zR9mz7RbJQXz?5BZeaJh&nTPy3e+NjU$=kfAnHUW#py^)+)z#Y2=xk{mLZ^p+hu#>2
zc{B_5!E4v}HP~&#j}GtYWbaVvTH4uLZ$LW5Vg@xy=v5MI%^Q;_)m+Mmx_2+XYVl)S
z*^ZCpdypO98<gqF72+y{E0^UsIrau!cbrdgiOPEv;+C}a<b5<*qd@v24@fUZ`zp|=
z@o$n6Av;M|)-s-0!Sx&)@aCS?WL)(P%mRX{``1wz$~|b?-$-2Ck;g~HE#Qc9?8Qe<
mKE1pHWaf|66+xMK4N$6tD^yvznGgSfzC<xaqk=S6^#1_!oNes@

literal 0
HcmV?d00001

diff --git a/ldm/modules/losses/__pycache__/contperceptual.cpython-38.pyc b/ldm/modules/losses/__pycache__/contperceptual.cpython-38.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..11b9439891da98781049c7ac856f5ec69c0a0e8d
GIT binary patch
literal 4235
zcmcIn&5t8T6|d?qx7*(ykH<5+N|peN4BBxBQrN>PSx8ufv>B~3L4lT_)#K{1+tcoL
zs@t1g)4q_E!eym7azPqjkn%5Z;DW@Tz@d*p0<GYXGmsU(s&0G6B+CH_TU}l6SFc{Z
zdaqu;S3j=REP~%_N55geewmOz3*zce192Bm{2CfY7!8P>zv)!56)O5lpuR_bNSMmB
zCxmI9GSNO#ZJp6$($}A19y#jMjDB+{x>tXdf{?;JVpB$Jh1EYMr<572^rUR7tij66
zdO~fDHCdI_(9=IA8nF!sZ)M&44}SFE_<cVazwO6E?oa*DO(Oo+^(sZ@vp1iqfbM1b
z*bhhEu(lq@Tl952f+0RflaP<d1zizJ(ki28M%+!5U(!@fiKC_(S}oDRVXO%HrL>IJ
zOv)ElQkhhl!qlfURXSuvf!C6HYNUE<G3}{((MXySZKYzY4e9!s!johtRFV#m-L!<Z
zo0@3%(hAxhBiN>K)K_>3nq{SjZs-Sr`w4+Twb);HeKj-va5hhzaG05)6S$8(9%q$#
z>^bf*2_rt0Q~Udw9?xeU&rHYhLqBnxtp2??_ITn);SV^Ecve0OoDV&JG)}VW%zxx<
zC|z1P)4YdnkToD==*^P38#qA}$64zNP)t|CFe^WBxjXd|EUCjK!*QlXp%-UH5RD$X
zJli?CFLd4VTc*3SS@77d=5k`GX;x!beVg^RNXN$tPMo+rv8&R%4wHJ>^}<*Mot4GB
zT;+=Fp29%bc4u2&v4y5&8Ac)Vg@xRpUxUTH;MBI{NWNkv-;C6gx8NeXDJ>@qQ`pjO
zNd9m>Eu3TLqW#brj@>Zyf|!3r%*>U_m*-s|TpUgQRngo)^9jw$jx!8gh<11x$UlBI
z7)MiY5Jri25=CbN<|gi7#H0Caa6Aw1<E#(IUN{&=%o`jh?n&Sszk6%y`r#m8)4?=i
z^T3M-LRq}9*W}WbhqK4Lfi+A#@z>B0-J(@$;Ez`6DH>I%^4w6B&kfC>Z8@@d`OD++
zaDC>EhT?;={)sa!Lhmg+@o&+jDkEoQ{(3?tbfO?kR7RQdRAuTuS!xBUY@vF1d&5m8
zlJiUFd~r-q$c2__D=OdmiZX4FTq8>Zw4NGFXU0kuVZJmF#AH(H$Q}Z+bo3tLugbjy
zrgTQ!osBt#`lDNKUH{vsuOGb0#V$v92>I0??)FW-k6vcF^CSW{%QV?qaAlGoicAxc
zn(6TFbe0(ml80md3TE|7Tv#8m9Tf3qjN~Ea-vVrx1uUe+c4?G%GLscYPacxI`G%0O
zKXilPJa7|y8<;yw{D)q_c>}|78x5f*?b4odonBXZ@Tx@@ufD*iFBFD&=P=B0{dpH+
zg#ko-NLRlNr^CoJr56gmqKasH_+l>c&4UZ+na01L3g=uIUn$F~_$E%&wAvwi1iTgL
zC^9it^cud!69Z^1tu5<mSstSR-{u;eRtH}xt>fEUpO}p93n%AT4Onf1)gYrRSN~nA
zURZE^L(Xh~tFhr~q#AgXbB$|h3G0~rFPK;1e@d#z66B(m?+3eShBw7X{MmX02eqD$
zrlnIw<_bmd$yJaWDrDKhj0V%rv~$W&7RU;dW?D;I;y6B4m+kjScvB-u>&(LV&na>g
z6_#w{sABEO4tCrADG8OIkkFDij`y-Hvdg4_JVYkDtOPkb=*gX)EByCqT~KL5ST${@
zkdd~eg*(i=KsJKp&c)t}T=ejiR)m!c6c!g4vkJPg;iwcyZG%(`q_RP31=5x>>NzEJ
zPInNBjbgO5=J(CzZfc#Y{O>+lcDFT3yR3Ofo&doOss(DJ4d^!LjV_>fKyNew-2pB2
zK!5j+w5ZSmR2L%~9YF8hA?Gwagw6gJOZ`9WQu?oUDLjZ>6yCxf3jYDyvPo`npj(#N
zpd0(Fk+rXcmDbp{tv0Ce+rQFIx|?=S)ng)hrwZO0QxHiyzne!AV(JYUK{D=zFL8SJ
zE@P@dn&Dl6$!OYeZbg)wa~W;&RBT7w7b5~Oaj~E7rOg$pw`Gs@a_!UJ*4f`L&^;kj
zM#DjYG}nl)N{{ILwmhj@^I@w(F6_LK0kn_{bR!37Asc9KV>|SO?PO(m?MUXGXCnC^
zBdN@v%GA@RGCOdQT!EL9G56we6tG3@7t28sB}MJFD1*#Pn1zKd3P4c|EGp=)^Nl39
zmsd`bSf`go9c;3=$tx)-El*Cib-4(?1}(BuA)5~bWv&JCZ-bQe4}TTCMRhGw>LC_@
zdVUf++;aoDU_qa{XI}pxPpqnY5@(jEU;NQLn#Xzdv?~@;y+^ZL0d4BJA&TI!iy}v)
zqD(mp^6IK7ba+X!58Wi!M%;I@>brBuoO*eEY$RS7M_i_5yXvxNmaG#ku1+K~PLadi
zL}YPPy8KW`y@95`pQms`7V2@<z~-YcPWXH%O3T=`W>JhP){juR*B}+wm<5VjpFPSd
zqaZqgw=h1FKFu^ylVTsUz)!MLQ7GFMijg@F9p(@DZLDb<v%E?W=C)foo5!5BSNV4Z
zF3MM7TwzRU|4Vmdn(#oTg2OgNSBTzB5Z<tjywLv7Dx(B5eCTp^3oFEbL_^dr>WB`l
zD_yFgzA(}60Kr?e6cfFsV$tigs#H-^c0l!jw^3&v(3{jG7K%`C%g4YQ5{UIrtz<wN
zq|&DRSJCSz&3wiSxDUYJ#lJ1yiXn7REUXP$xj}DW%_ej#R(r1LVub@a!%`M6y`T;%
zw%f1B(33HjHwL!OWh3J7a^T**{7OcT`ufok{|*L^?)6LJW)|YcgqxY=IC*i7e${b4
zz+EdJktI17XOq8!CaV<}f}w{pFuxS!?)bW35k(&t_gYbC$D*1vMZu8&@;4=)*Lzl7
zLIQa%`TIsuvC50vVHol~jM=XwE-u)k!{Xv_D2{~}A3prx^4gFY@5oz)6m#Wdi0g{f
UCo}Kl)87!w#lk!$Ja4N11{=V3FaQ7m

literal 0
HcmV?d00001

diff --git a/ldm/modules/losses/contperceptual.py b/ldm/modules/losses/contperceptual.py
new file mode 100644
index 0000000..0af8318
--- /dev/null
+++ b/ldm/modules/losses/contperceptual.py
@@ -0,0 +1,152 @@
+import torch
+import torch.nn as nn
+
+from taming.modules.losses.vqperceptual import *  # TODO: taming dependency yes/no?
+
+
+class LPIPSWithDiscriminator(nn.Module):
+    def __init__(self, disc_start, logvar_init=0.0, kl_weight=1.0, pixelloss_weight=1.0,
+                 disc_num_layers=3, disc_in_channels=3, disc_factor=1.0, disc_weight=1.0,
+                 perceptual_weight=1.0, use_actnorm=False, disc_conditional=False,
+                 disc_loss="hinge"):
+
+        super().__init__()
+        assert disc_loss in ["hinge", "vanilla"]
+        self.kl_weight = kl_weight
+        self.pixel_weight = pixelloss_weight
+        self.perceptual_loss = LPIPS().eval()
+        self.perceptual_weight = perceptual_weight
+        # output log variance
+        self.logvar = nn.Parameter(torch.ones(size=()) * logvar_init)
+
+        self.discriminator = NLayerDiscriminator(input_nc=disc_in_channels,
+                                                 n_layers=disc_num_layers,
+                                                 use_actnorm=use_actnorm
+                                                 ).apply(weights_init)
+        self.discriminator_iter_start = disc_start
+        self.disc_loss = hinge_d_loss if disc_loss == "hinge" else vanilla_d_loss
+        self.disc_factor = disc_factor
+        self.discriminator_weight = disc_weight
+        self.disc_conditional = disc_conditional
+
+    def calculate_adaptive_weight(self, nll_loss, g_loss, last_layer=None):
+        if last_layer is not None:
+            nll_grads = torch.autograd.grad(nll_loss, last_layer, retain_graph=True)[0]
+            g_grads = torch.autograd.grad(g_loss, last_layer, retain_graph=True)[0]
+        else:
+            nll_grads = torch.autograd.grad(nll_loss, self.last_layer[0], retain_graph=True)[0]
+            g_grads = torch.autograd.grad(g_loss, self.last_layer[0], retain_graph=True)[0]
+
+        d_weight = torch.norm(nll_grads) / (torch.norm(g_grads) + 1e-4)
+        d_weight = torch.clamp(d_weight, 0.0, 1e4).detach()
+        d_weight = d_weight * self.discriminator_weight
+        return d_weight
+
+    def forward(self, inputs, reconstructions, optimizer_idx,
+                global_step, posteriors=None, last_layer=None, cond=None, split="train",
+                weights=None, return_dic=False):
+        rec_loss = torch.abs(inputs.contiguous() - reconstructions.contiguous())
+        if self.perceptual_weight > 0:
+            p_loss = self.perceptual_loss(inputs.contiguous(), reconstructions.contiguous())
+            rec_loss = rec_loss + self.perceptual_weight * p_loss
+
+        nll_loss = rec_loss / torch.exp(self.logvar) + self.logvar
+        weighted_nll_loss = nll_loss
+        if weights is not None:
+            weighted_nll_loss = weights*nll_loss
+        weighted_nll_loss = torch.mean(weighted_nll_loss) / weighted_nll_loss.shape[0]
+        nll_loss = torch.mean(nll_loss) / nll_loss.shape[0]
+        if self.kl_weight>0:
+            kl_loss = posteriors.kl()
+            kl_loss = torch.mean(kl_loss) / kl_loss.shape[0]
+
+        # now the GAN part
+        if optimizer_idx == 0:
+            # generator update
+            if cond is None:
+                assert not self.disc_conditional
+                logits_fake = self.discriminator(reconstructions.contiguous())
+            else:
+                assert self.disc_conditional
+                logits_fake = self.discriminator(torch.cat((reconstructions.contiguous(), cond), dim=1))
+            g_loss = -torch.mean(logits_fake)
+
+            if self.disc_factor > 0.0:
+                try:
+                    d_weight = self.calculate_adaptive_weight(nll_loss, g_loss, last_layer=last_layer)
+                except RuntimeError:
+                    # assert not self.training
+                    d_weight = torch.tensor(1.0) * self.discriminator_weight
+            else:
+                # d_weight = torch.tensor(0.0)
+                d_weight = torch.tensor(0.0)
+
+            disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+            if self.kl_weight>0:
+                loss = weighted_nll_loss + self.kl_weight * kl_loss + d_weight * disc_factor * g_loss
+                log = {"{}/total_loss".format(split): loss.clone().detach().mean(), "{}/logvar".format(split): self.logvar.detach(),
+                       "{}/kl_loss".format(split): kl_loss.detach().mean(), "{}/nll_loss".format(split): nll_loss.detach().mean(),
+                       "{}/rec_loss".format(split): rec_loss.detach().mean(),
+                       "{}/d_weight".format(split): d_weight.detach(),
+                       "{}/disc_factor".format(split): torch.tensor(disc_factor),
+                       "{}/g_loss".format(split): g_loss.detach().mean(),
+                       }
+                if return_dic:
+                    loss_dic = {}
+                    loss_dic['total_loss'] = loss.clone().detach().mean()
+                    loss_dic['logvar'] = self.logvar.detach()
+                    loss_dic['kl_loss'] = kl_loss.detach().mean()
+                    loss_dic['nll_loss'] = nll_loss.detach().mean()
+                    loss_dic['rec_loss'] = rec_loss.detach().mean()
+                    loss_dic['d_weight'] = d_weight.detach()
+                    loss_dic['disc_factor'] = torch.tensor(disc_factor)
+                    loss_dic['g_loss'] = g_loss.detach().mean()
+            else:
+                loss = weighted_nll_loss + d_weight * disc_factor * g_loss
+                log = {"{}/total_loss".format(split): loss.clone().detach().mean(), "{}/logvar".format(split): self.logvar.detach(),
+                       "{}/nll_loss".format(split): nll_loss.detach().mean(),
+                       "{}/rec_loss".format(split): rec_loss.detach().mean(),
+                       "{}/d_weight".format(split): d_weight.detach(),
+                       "{}/disc_factor".format(split): torch.tensor(disc_factor),
+                       "{}/g_loss".format(split): g_loss.detach().mean(),
+                       }
+                if return_dic:
+                    loss_dic = {}
+                    loss_dic["{}/total_loss".format(split)] = loss.clone().detach().mean()
+                    loss_dic["{}/logvar".format(split)] = self.logvar.detach()
+                    loss_dic['nll_loss'.format(split)] = nll_loss.detach().mean()
+                    loss_dic['rec_loss'.format(split)] = rec_loss.detach().mean()
+                    loss_dic['d_weight'.format(split)] = d_weight.detach()
+                    loss_dic['disc_factor'.format(split)] = torch.tensor(disc_factor)
+                    loss_dic['g_loss'.format(split)] = g_loss.detach().mean()
+
+            if return_dic:
+                return loss, log, loss_dic
+            return loss, log
+
+        if optimizer_idx == 1:
+            # second pass for discriminator update
+            if cond is None:
+                logits_real = self.discriminator(inputs.contiguous().detach())
+                logits_fake = self.discriminator(reconstructions.contiguous().detach())
+            else:
+                logits_real = self.discriminator(torch.cat((inputs.contiguous().detach(), cond), dim=1))
+                logits_fake = self.discriminator(torch.cat((reconstructions.contiguous().detach(), cond), dim=1))
+
+            disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+            d_loss = disc_factor * self.disc_loss(logits_real, logits_fake)
+
+            log = {"{}/disc_loss".format(split): d_loss.clone().detach().mean(),
+                   "{}/logits_real".format(split): logits_real.detach().mean(),
+                   "{}/logits_fake".format(split): logits_fake.detach().mean()
+                   }
+
+            if return_dic:
+                loss_dic = {}
+                loss_dic["{}/disc_loss".format(split)] = d_loss.clone().detach().mean()
+                loss_dic["{}/logits_real".format(split)] = logits_real.detach().mean()
+                loss_dic["{}/logits_fake".format(split)] = logits_fake.detach().mean()
+                return d_loss, log, loss_dic
+
+            return d_loss, log
+
diff --git a/ldm/modules/losses/vqperceptual.py b/ldm/modules/losses/vqperceptual.py
new file mode 100644
index 0000000..bdeb624
--- /dev/null
+++ b/ldm/modules/losses/vqperceptual.py
@@ -0,0 +1,180 @@
+import torch
+from torch import nn
+import torch.nn.functional as F
+from einops import repeat
+
+from taming.modules.discriminator.model import NLayerDiscriminator, weights_init
+from taming.modules.losses.lpips import LPIPS
+from taming.modules.losses.vqperceptual import hinge_d_loss, vanilla_d_loss
+
+
+def hinge_d_loss_with_exemplar_weights(logits_real, logits_fake, weights):
+    assert weights.shape[0] == logits_real.shape[0] == logits_fake.shape[0]
+    loss_real = torch.mean(F.relu(1. - logits_real), dim=[1,2,3])
+    loss_fake = torch.mean(F.relu(1. + logits_fake), dim=[1,2,3])
+    loss_real = (weights * loss_real).sum() / weights.sum()
+    loss_fake = (weights * loss_fake).sum() / weights.sum()
+    d_loss = 0.5 * (loss_real + loss_fake)
+    return d_loss
+
+def adopt_weight(weight, global_step, threshold=0, value=0.):
+    if global_step < threshold:
+        weight = value
+    return weight
+
+
+def measure_perplexity(predicted_indices, n_embed):
+    # src: https://github.com/karpathy/deep-vector-quantization/blob/main/model.py
+    # eval cluster perplexity. when perplexity == num_embeddings then all clusters are used exactly equally
+    encodings = F.one_hot(predicted_indices, n_embed).float().reshape(-1, n_embed)
+    avg_probs = encodings.mean(0)
+    perplexity = (-(avg_probs * torch.log(avg_probs + 1e-10)).sum()).exp()
+    cluster_use = torch.sum(avg_probs > 0)
+    return perplexity, cluster_use
+
+def l1(x, y):
+    return torch.abs(x-y)
+
+
+def l2(x, y):
+    return torch.pow((x-y), 2)
+
+
+class VQLPIPSWithDiscriminator(nn.Module):
+    def __init__(self, disc_start, codebook_weight=1.0, pixelloss_weight=1.0,
+                 disc_num_layers=3, disc_in_channels=3, disc_factor=1.0, disc_weight=1.0,
+                 perceptual_weight=1.0, use_actnorm=False, disc_conditional=False,
+                 disc_ndf=64, disc_loss="hinge", n_classes=None, perceptual_loss="lpips",
+                 pixel_loss="l1"):
+        super().__init__()
+        assert disc_loss in ["hinge", "vanilla"]
+        assert perceptual_loss in ["lpips", "clips", "dists"]
+        assert pixel_loss in ["l1", "l2"]
+        self.codebook_weight = codebook_weight
+        self.pixel_weight = pixelloss_weight
+        if perceptual_loss == "lpips":
+            print(f"{self.__class__.__name__}: Running with LPIPS.")
+            self.perceptual_loss = LPIPS().eval()
+        else:
+            raise ValueError(f"Unknown perceptual loss: >> {perceptual_loss} <<")
+        self.perceptual_weight = perceptual_weight
+
+        if pixel_loss == "l1":
+            self.pixel_loss = l1
+        else:
+            self.pixel_loss = l2
+
+        self.discriminator = NLayerDiscriminator(input_nc=disc_in_channels,
+                                                 n_layers=disc_num_layers,
+                                                 use_actnorm=use_actnorm,
+                                                 ndf=disc_ndf
+                                                 ).apply(weights_init)
+        self.discriminator_iter_start = disc_start
+        if disc_loss == "hinge":
+            self.disc_loss = hinge_d_loss
+        elif disc_loss == "vanilla":
+            self.disc_loss = vanilla_d_loss
+        else:
+            raise ValueError(f"Unknown GAN loss '{disc_loss}'.")
+        print(f"VQLPIPSWithDiscriminator running with {disc_loss} loss.")
+        self.disc_factor = disc_factor
+        self.discriminator_weight = disc_weight
+        self.disc_conditional = disc_conditional
+        self.n_classes = n_classes
+
+    # def calculate_adaptive_weight(self, nll_loss, g_loss, last_layer=None):
+    #     if last_layer is not None:
+    #         nll_grads = torch.autograd.grad(nll_loss, last_layer, retain_graph=True)[0]
+    #         g_grads = torch.autograd.grad(g_loss, last_layer, retain_graph=True)[0]
+    #     else:
+    #         nll_grads = torch.autograd.grad(nll_loss, self.last_layer[0], retain_graph=True)[0]
+    #         g_grads = torch.autograd.grad(g_loss, self.last_layer[0], retain_graph=True)[0]
+
+    #     d_weight = torch.norm(nll_grads) / (torch.norm(g_grads) + 1e-4)
+    #     d_weight = torch.clamp(d_weight, 0.0, 1e4).detach()
+    #     d_weight = d_weight * self.discriminator_weight
+    #     return d_weight
+
+    def calculate_adaptive_weight(self, nll_loss, g_loss, last_layer=None):
+        # if last_layer is not None:
+        #     nll_grads = torch.autograd.grad(nll_loss, last_layer, retain_graph=True)[0]
+        #     g_grads = torch.autograd.grad(g_loss, last_layer, retain_graph=True)[0]
+        # else:
+        #     nll_grads = torch.autograd.grad(nll_loss, self.last_layer[0], retain_graph=True)[0]
+        #     g_grads = torch.autograd.grad(g_loss, self.last_layer[0], retain_graph=True)[0]
+
+        # d_weight = torch.norm(nll_grads) / (torch.norm(g_grads) + 1e-4)
+        # d_weight = torch.clamp(d_weight, 0.0, 1e4).detach()
+        d_weight = 1.0 * self.discriminator_weight
+        return d_weight
+
+    def forward(self, codebook_loss, inputs, reconstructions, optimizer_idx,
+                global_step, last_layer=None, cond=None, split="train", predicted_indices=None):
+        if not exists(codebook_loss):
+            codebook_loss = torch.tensor([0.]).to(inputs.device)
+        #rec_loss = torch.abs(inputs.contiguous() - reconstructions.contiguous())
+        rec_loss = self.pixel_loss(inputs.contiguous(), reconstructions.contiguous())
+        if self.perceptual_weight > 0:
+            p_loss = self.perceptual_loss(inputs.contiguous(), reconstructions.contiguous())
+            rec_loss = rec_loss + self.perceptual_weight * p_loss
+        else:
+            p_loss = torch.tensor([0.0])
+
+        nll_loss = rec_loss
+        #nll_loss = torch.sum(nll_loss) / nll_loss.shape[0]
+        nll_loss = torch.mean(nll_loss)
+
+        # now the GAN part
+        if optimizer_idx == 0:
+            # generator update
+            if cond is None:
+                assert not self.disc_conditional
+                logits_fake = self.discriminator(reconstructions.contiguous())
+            else:
+                assert self.disc_conditional
+                logits_fake = self.discriminator(torch.cat((reconstructions.contiguous(), cond), dim=1))
+            g_loss = -torch.mean(logits_fake)
+
+            try:
+                d_weight = self.calculate_adaptive_weight(nll_loss, g_loss, last_layer=last_layer)
+            except RuntimeError:
+                assert not self.training
+                d_weight = torch.tensor(0.0)
+
+            disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+            loss = nll_loss + d_weight * disc_factor * g_loss + self.codebook_weight * codebook_loss.mean()
+
+            log = {"{}/total_loss".format(split): loss.clone().detach().mean(),
+                   "{}/quant_loss".format(split): codebook_loss.detach().mean(),
+                   "{}/nll_loss".format(split): nll_loss.detach().mean(),
+                   "{}/rec_loss".format(split): rec_loss.detach().mean(),
+                   "{}/p_loss".format(split): p_loss.detach().mean(),
+                   "{}/d_weight".format(split): d_weight.detach(),
+                   "{}/disc_factor".format(split): torch.tensor(disc_factor),
+                   "{}/g_loss".format(split): g_loss.detach().mean(),
+                   }
+            if predicted_indices is not None:
+                assert self.n_classes is not None
+                with torch.no_grad():
+                    perplexity, cluster_usage = measure_perplexity(predicted_indices, self.n_classes)
+                log[f"{split}/perplexity"] = perplexity
+                log[f"{split}/cluster_usage"] = cluster_usage
+            return loss, log
+
+        if optimizer_idx == 1:
+            # second pass for discriminator update
+            if cond is None:
+                logits_real = self.discriminator(inputs.contiguous().detach())
+                logits_fake = self.discriminator(reconstructions.contiguous().detach())
+            else:
+                logits_real = self.discriminator(torch.cat((inputs.contiguous().detach(), cond), dim=1))
+                logits_fake = self.discriminator(torch.cat((reconstructions.contiguous().detach(), cond), dim=1))
+
+            disc_factor = adopt_weight(self.disc_factor, global_step, threshold=self.discriminator_iter_start)
+            d_loss = disc_factor * self.disc_loss(logits_real, logits_fake)
+
+            log = {"{}/disc_loss".format(split): d_loss.clone().detach().mean(),
+                   "{}/logits_real".format(split): logits_real.detach().mean(),
+                   "{}/logits_fake".format(split): logits_fake.detach().mean()
+                   }
+            return d_loss, log
diff --git a/ldm/modules/midas/__init__.py b/ldm/modules/midas/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ldm/modules/midas/api.py b/ldm/modules/midas/api.py
new file mode 100644
index 0000000..b58ebbf
--- /dev/null
+++ b/ldm/modules/midas/api.py
@@ -0,0 +1,170 @@
+# based on https://github.com/isl-org/MiDaS
+
+import cv2
+import torch
+import torch.nn as nn
+from torchvision.transforms import Compose
+
+from ldm.modules.midas.midas.dpt_depth import DPTDepthModel
+from ldm.modules.midas.midas.midas_net import MidasNet
+from ldm.modules.midas.midas.midas_net_custom import MidasNet_small
+from ldm.modules.midas.midas.transforms import Resize, NormalizeImage, PrepareForNet
+
+
+ISL_PATHS = {
+    "dpt_large": "midas_models/dpt_large-midas-2f21e586.pt",
+    "dpt_hybrid": "midas_models/dpt_hybrid-midas-501f0c75.pt",
+    "midas_v21": "",
+    "midas_v21_small": "",
+}
+
+
+def disabled_train(self, mode=True):
+    """Overwrite model.train with this function to make sure train/eval mode
+    does not change anymore."""
+    return self
+
+
+def load_midas_transform(model_type):
+    # https://github.com/isl-org/MiDaS/blob/master/run.py
+    # load transform only
+    if model_type == "dpt_large":  # DPT-Large
+        net_w, net_h = 384, 384
+        resize_mode = "minimal"
+        normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
+
+    elif model_type == "dpt_hybrid":  # DPT-Hybrid
+        net_w, net_h = 384, 384
+        resize_mode = "minimal"
+        normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
+
+    elif model_type == "midas_v21":
+        net_w, net_h = 384, 384
+        resize_mode = "upper_bound"
+        normalization = NormalizeImage(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
+
+    elif model_type == "midas_v21_small":
+        net_w, net_h = 256, 256
+        resize_mode = "upper_bound"
+        normalization = NormalizeImage(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
+
+    else:
+        assert False, f"model_type '{model_type}' not implemented, use: --model_type large"
+
+    transform = Compose(
+        [
+            Resize(
+                net_w,
+                net_h,
+                resize_target=None,
+                keep_aspect_ratio=True,
+                ensure_multiple_of=32,
+                resize_method=resize_mode,
+                image_interpolation_method=cv2.INTER_CUBIC,
+            ),
+            normalization,
+            PrepareForNet(),
+        ]
+    )
+
+    return transform
+
+
+def load_model(model_type):
+    # https://github.com/isl-org/MiDaS/blob/master/run.py
+    # load network
+    model_path = ISL_PATHS[model_type]
+    if model_type == "dpt_large":  # DPT-Large
+        model = DPTDepthModel(
+            path=model_path,
+            backbone="vitl16_384",
+            non_negative=True,
+        )
+        net_w, net_h = 384, 384
+        resize_mode = "minimal"
+        normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
+
+    elif model_type == "dpt_hybrid":  # DPT-Hybrid
+        model = DPTDepthModel(
+            path=model_path,
+            backbone="vitb_rn50_384",
+            non_negative=True,
+        )
+        net_w, net_h = 384, 384
+        resize_mode = "minimal"
+        normalization = NormalizeImage(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
+
+    elif model_type == "midas_v21":
+        model = MidasNet(model_path, non_negative=True)
+        net_w, net_h = 384, 384
+        resize_mode = "upper_bound"
+        normalization = NormalizeImage(
+            mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
+        )
+
+    elif model_type == "midas_v21_small":
+        model = MidasNet_small(model_path, features=64, backbone="efficientnet_lite3", exportable=True,
+                               non_negative=True, blocks={'expand': True})
+        net_w, net_h = 256, 256
+        resize_mode = "upper_bound"
+        normalization = NormalizeImage(
+            mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
+        )
+
+    else:
+        print(f"model_type '{model_type}' not implemented, use: --model_type large")
+        assert False
+
+    transform = Compose(
+        [
+            Resize(
+                net_w,
+                net_h,
+                resize_target=None,
+                keep_aspect_ratio=True,
+                ensure_multiple_of=32,
+                resize_method=resize_mode,
+                image_interpolation_method=cv2.INTER_CUBIC,
+            ),
+            normalization,
+            PrepareForNet(),
+        ]
+    )
+
+    return model.eval(), transform
+
+
+class MiDaSInference(nn.Module):
+    MODEL_TYPES_TORCH_HUB = [
+        "DPT_Large",
+        "DPT_Hybrid",
+        "MiDaS_small"
+    ]
+    MODEL_TYPES_ISL = [
+        "dpt_large",
+        "dpt_hybrid",
+        "midas_v21",
+        "midas_v21_small",
+    ]
+
+    def __init__(self, model_type):
+        super().__init__()
+        assert (model_type in self.MODEL_TYPES_ISL)
+        model, _ = load_model(model_type)
+        self.model = model
+        self.model.train = disabled_train
+
+    def forward(self, x):
+        # x in 0..1 as produced by calling self.transform on a 0..1 float64 numpy array
+        # NOTE: we expect that the correct transform has been called during dataloading.
+        with torch.no_grad():
+            prediction = self.model(x)
+            prediction = torch.nn.functional.interpolate(
+                prediction.unsqueeze(1),
+                size=x.shape[2:],
+                mode="bicubic",
+                align_corners=False,
+            )
+        assert prediction.shape == (x.shape[0], 1, x.shape[2], x.shape[3])
+        return prediction
+
diff --git a/ldm/modules/midas/midas/__init__.py b/ldm/modules/midas/midas/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/ldm/modules/midas/midas/base_model.py b/ldm/modules/midas/midas/base_model.py
new file mode 100644
index 0000000..5cf4302
--- /dev/null
+++ b/ldm/modules/midas/midas/base_model.py
@@ -0,0 +1,16 @@
+import torch
+
+
+class BaseModel(torch.nn.Module):
+    def load(self, path):
+        """Load model from file.
+
+        Args:
+            path (str): file path
+        """
+        parameters = torch.load(path, map_location=torch.device('cpu'))
+
+        if "optimizer" in parameters:
+            parameters = parameters["model"]
+
+        self.load_state_dict(parameters)
diff --git a/ldm/modules/midas/midas/blocks.py b/ldm/modules/midas/midas/blocks.py
new file mode 100644
index 0000000..2145d18
--- /dev/null
+++ b/ldm/modules/midas/midas/blocks.py
@@ -0,0 +1,342 @@
+import torch
+import torch.nn as nn
+
+from .vit import (
+    _make_pretrained_vitb_rn50_384,
+    _make_pretrained_vitl16_384,
+    _make_pretrained_vitb16_384,
+    forward_vit,
+)
+
+def _make_encoder(backbone, features, use_pretrained, groups=1, expand=False, exportable=True, hooks=None, use_vit_only=False, use_readout="ignore",):
+    if backbone == "vitl16_384":
+        pretrained = _make_pretrained_vitl16_384(
+            use_pretrained, hooks=hooks, use_readout=use_readout
+        )
+        scratch = _make_scratch(
+            [256, 512, 1024, 1024], features, groups=groups, expand=expand
+        )  # ViT-L/16 - 85.0% Top1 (backbone)
+    elif backbone == "vitb_rn50_384":
+        pretrained = _make_pretrained_vitb_rn50_384(
+            use_pretrained,
+            hooks=hooks,
+            use_vit_only=use_vit_only,
+            use_readout=use_readout,
+        )
+        scratch = _make_scratch(
+            [256, 512, 768, 768], features, groups=groups, expand=expand
+        )  # ViT-H/16 - 85.0% Top1 (backbone)
+    elif backbone == "vitb16_384":
+        pretrained = _make_pretrained_vitb16_384(
+            use_pretrained, hooks=hooks, use_readout=use_readout
+        )
+        scratch = _make_scratch(
+            [96, 192, 384, 768], features, groups=groups, expand=expand
+        )  # ViT-B/16 - 84.6% Top1 (backbone)
+    elif backbone == "resnext101_wsl":
+        pretrained = _make_pretrained_resnext101_wsl(use_pretrained)
+        scratch = _make_scratch([256, 512, 1024, 2048], features, groups=groups, expand=expand)     # efficientnet_lite3  
+    elif backbone == "efficientnet_lite3":
+        pretrained = _make_pretrained_efficientnet_lite3(use_pretrained, exportable=exportable)
+        scratch = _make_scratch([32, 48, 136, 384], features, groups=groups, expand=expand)  # efficientnet_lite3     
+    else:
+        print(f"Backbone '{backbone}' not implemented")
+        assert False
+        
+    return pretrained, scratch
+
+
+def _make_scratch(in_shape, out_shape, groups=1, expand=False):
+    scratch = nn.Module()
+
+    out_shape1 = out_shape
+    out_shape2 = out_shape
+    out_shape3 = out_shape
+    out_shape4 = out_shape
+    if expand==True:
+        out_shape1 = out_shape
+        out_shape2 = out_shape*2
+        out_shape3 = out_shape*4
+        out_shape4 = out_shape*8
+
+    scratch.layer1_rn = nn.Conv2d(
+        in_shape[0], out_shape1, kernel_size=3, stride=1, padding=1, bias=False, groups=groups
+    )
+    scratch.layer2_rn = nn.Conv2d(
+        in_shape[1], out_shape2, kernel_size=3, stride=1, padding=1, bias=False, groups=groups
+    )
+    scratch.layer3_rn = nn.Conv2d(
+        in_shape[2], out_shape3, kernel_size=3, stride=1, padding=1, bias=False, groups=groups
+    )
+    scratch.layer4_rn = nn.Conv2d(
+        in_shape[3], out_shape4, kernel_size=3, stride=1, padding=1, bias=False, groups=groups
+    )
+
+    return scratch
+
+
+def _make_pretrained_efficientnet_lite3(use_pretrained, exportable=False):
+    efficientnet = torch.hub.load(
+        "rwightman/gen-efficientnet-pytorch",
+        "tf_efficientnet_lite3",
+        pretrained=use_pretrained,
+        exportable=exportable
+    )
+    return _make_efficientnet_backbone(efficientnet)
+
+
+def _make_efficientnet_backbone(effnet):
+    pretrained = nn.Module()
+
+    pretrained.layer1 = nn.Sequential(
+        effnet.conv_stem, effnet.bn1, effnet.act1, *effnet.blocks[0:2]
+    )
+    pretrained.layer2 = nn.Sequential(*effnet.blocks[2:3])
+    pretrained.layer3 = nn.Sequential(*effnet.blocks[3:5])
+    pretrained.layer4 = nn.Sequential(*effnet.blocks[5:9])
+
+    return pretrained
+    
+
+def _make_resnet_backbone(resnet):
+    pretrained = nn.Module()
+    pretrained.layer1 = nn.Sequential(
+        resnet.conv1, resnet.bn1, resnet.relu, resnet.maxpool, resnet.layer1
+    )
+
+    pretrained.layer2 = resnet.layer2
+    pretrained.layer3 = resnet.layer3
+    pretrained.layer4 = resnet.layer4
+
+    return pretrained
+
+
+def _make_pretrained_resnext101_wsl(use_pretrained):
+    resnet = torch.hub.load("facebookresearch/WSL-Images", "resnext101_32x8d_wsl")
+    return _make_resnet_backbone(resnet)
+
+
+
+class Interpolate(nn.Module):
+    """Interpolation module.
+    """
+
+    def __init__(self, scale_factor, mode, align_corners=False):
+        """Init.
+
+        Args:
+            scale_factor (float): scaling
+            mode (str): interpolation mode
+        """
+        super(Interpolate, self).__init__()
+
+        self.interp = nn.functional.interpolate
+        self.scale_factor = scale_factor
+        self.mode = mode
+        self.align_corners = align_corners
+
+    def forward(self, x):
+        """Forward pass.
+
+        Args:
+            x (tensor): input
+
+        Returns:
+            tensor: interpolated data
+        """
+
+        x = self.interp(
+            x, scale_factor=self.scale_factor, mode=self.mode, align_corners=self.align_corners
+        )
+
+        return x
+
+
+class ResidualConvUnit(nn.Module):
+    """Residual convolution module.
+    """
+
+    def __init__(self, features):
+        """Init.
+
+        Args:
+            features (int): number of features
+        """
+        super().__init__()
+
+        self.conv1 = nn.Conv2d(
+            features, features, kernel_size=3, stride=1, padding=1, bias=True
+        )
+
+        self.conv2 = nn.Conv2d(
+            features, features, kernel_size=3, stride=1, padding=1, bias=True
+        )
+
+        self.relu = nn.ReLU(inplace=True)
+
+    def forward(self, x):
+        """Forward pass.
+
+        Args:
+            x (tensor): input
+
+        Returns:
+            tensor: output
+        """
+        out = self.relu(x)
+        out = self.conv1(out)
+        out = self.relu(out)
+        out = self.conv2(out)
+
+        return out + x
+
+
+class FeatureFusionBlock(nn.Module):
+    """Feature fusion block.
+    """
+
+    def __init__(self, features):
+        """Init.
+
+        Args:
+            features (int): number of features
+        """
+        super(FeatureFusionBlock, self).__init__()
+
+        self.resConfUnit1 = ResidualConvUnit(features)
+        self.resConfUnit2 = ResidualConvUnit(features)
+
+    def forward(self, *xs):
+        """Forward pass.
+
+        Returns:
+            tensor: output
+        """
+        output = xs[0]
+
+        if len(xs) == 2:
+            output += self.resConfUnit1(xs[1])
+
+        output = self.resConfUnit2(output)
+
+        output = nn.functional.interpolate(
+            output, scale_factor=2, mode="bilinear", align_corners=True
+        )
+
+        return output
+
+
+
+
+class ResidualConvUnit_custom(nn.Module):
+    """Residual convolution module.
+    """
+
+    def __init__(self, features, activation, bn):
+        """Init.
+
+        Args:
+            features (int): number of features
+        """
+        super().__init__()
+
+        self.bn = bn
+
+        self.groups=1
+
+        self.conv1 = nn.Conv2d(
+            features, features, kernel_size=3, stride=1, padding=1, bias=True, groups=self.groups
+        )
+        
+        self.conv2 = nn.Conv2d(
+            features, features, kernel_size=3, stride=1, padding=1, bias=True, groups=self.groups
+        )
+
+        if self.bn==True:
+            self.bn1 = nn.BatchNorm2d(features)
+            self.bn2 = nn.BatchNorm2d(features)
+
+        self.activation = activation
+
+        self.skip_add = nn.quantized.FloatFunctional()
+
+    def forward(self, x):
+        """Forward pass.
+
+        Args:
+            x (tensor): input
+
+        Returns:
+            tensor: output
+        """
+        
+        out = self.activation(x)
+        out = self.conv1(out)
+        if self.bn==True:
+            out = self.bn1(out)
+       
+        out = self.activation(out)
+        out = self.conv2(out)
+        if self.bn==True:
+            out = self.bn2(out)
+
+        if self.groups > 1:
+            out = self.conv_merge(out)
+
+        return self.skip_add.add(out, x)
+
+        # return out + x
+
+
+class FeatureFusionBlock_custom(nn.Module):
+    """Feature fusion block.
+    """
+
+    def __init__(self, features, activation, deconv=False, bn=False, expand=False, align_corners=True):
+        """Init.
+
+        Args:
+            features (int): number of features
+        """
+        super(FeatureFusionBlock_custom, self).__init__()
+
+        self.deconv = deconv
+        self.align_corners = align_corners
+
+        self.groups=1
+
+        self.expand = expand
+        out_features = features
+        if self.expand==True:
+            out_features = features//2
+        
+        self.out_conv = nn.Conv2d(features, out_features, kernel_size=1, stride=1, padding=0, bias=True, groups=1)
+
+        self.resConfUnit1 = ResidualConvUnit_custom(features, activation, bn)
+        self.resConfUnit2 = ResidualConvUnit_custom(features, activation, bn)
+        
+        self.skip_add = nn.quantized.FloatFunctional()
+
+    def forward(self, *xs):
+        """Forward pass.
+
+        Returns:
+            tensor: output
+        """
+        output = xs[0]
+
+        if len(xs) == 2:
+            res = self.resConfUnit1(xs[1])
+            output = self.skip_add.add(output, res)
+            # output += res
+
+        output = self.resConfUnit2(output)
+
+        output = nn.functional.interpolate(
+            output, scale_factor=2, mode="bilinear", align_corners=self.align_corners
+        )
+
+        output = self.out_conv(output)
+
+        return output
+
diff --git a/ldm/modules/midas/midas/dpt_depth.py b/ldm/modules/midas/midas/dpt_depth.py
new file mode 100644
index 0000000..4e9aab5
--- /dev/null
+++ b/ldm/modules/midas/midas/dpt_depth.py
@@ -0,0 +1,109 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from .base_model import BaseModel
+from .blocks import (
+    FeatureFusionBlock,
+    FeatureFusionBlock_custom,
+    Interpolate,
+    _make_encoder,
+    forward_vit,
+)
+
+
+def _make_fusion_block(features, use_bn):
+    return FeatureFusionBlock_custom(
+        features,
+        nn.ReLU(False),
+        deconv=False,
+        bn=use_bn,
+        expand=False,
+        align_corners=True,
+    )
+
+
+class DPT(BaseModel):
+    def __init__(
+        self,
+        head,
+        features=256,
+        backbone="vitb_rn50_384",
+        readout="project",
+        channels_last=False,
+        use_bn=False,
+    ):
+
+        super(DPT, self).__init__()
+
+        self.channels_last = channels_last
+
+        hooks = {
+            "vitb_rn50_384": [0, 1, 8, 11],
+            "vitb16_384": [2, 5, 8, 11],
+            "vitl16_384": [5, 11, 17, 23],
+        }
+
+        # Instantiate backbone and reassemble blocks
+        self.pretrained, self.scratch = _make_encoder(
+            backbone,
+            features,
+            False, # Set to true of you want to train from scratch, uses ImageNet weights
+            groups=1,
+            expand=False,
+            exportable=False,
+            hooks=hooks[backbone],
+            use_readout=readout,
+        )
+
+        self.scratch.refinenet1 = _make_fusion_block(features, use_bn)
+        self.scratch.refinenet2 = _make_fusion_block(features, use_bn)
+        self.scratch.refinenet3 = _make_fusion_block(features, use_bn)
+        self.scratch.refinenet4 = _make_fusion_block(features, use_bn)
+
+        self.scratch.output_conv = head
+
+
+    def forward(self, x):
+        if self.channels_last == True:
+            x.contiguous(memory_format=torch.channels_last)
+
+        layer_1, layer_2, layer_3, layer_4 = forward_vit(self.pretrained, x)
+
+        layer_1_rn = self.scratch.layer1_rn(layer_1)
+        layer_2_rn = self.scratch.layer2_rn(layer_2)
+        layer_3_rn = self.scratch.layer3_rn(layer_3)
+        layer_4_rn = self.scratch.layer4_rn(layer_4)
+
+        path_4 = self.scratch.refinenet4(layer_4_rn)
+        path_3 = self.scratch.refinenet3(path_4, layer_3_rn)
+        path_2 = self.scratch.refinenet2(path_3, layer_2_rn)
+        path_1 = self.scratch.refinenet1(path_2, layer_1_rn)
+
+        out = self.scratch.output_conv(path_1)
+
+        return out
+
+
+class DPTDepthModel(DPT):
+    def __init__(self, path=None, non_negative=True, **kwargs):
+        features = kwargs["features"] if "features" in kwargs else 256
+
+        head = nn.Sequential(
+            nn.Conv2d(features, features // 2, kernel_size=3, stride=1, padding=1),
+            Interpolate(scale_factor=2, mode="bilinear", align_corners=True),
+            nn.Conv2d(features // 2, 32, kernel_size=3, stride=1, padding=1),
+            nn.ReLU(True),
+            nn.Conv2d(32, 1, kernel_size=1, stride=1, padding=0),
+            nn.ReLU(True) if non_negative else nn.Identity(),
+            nn.Identity(),
+        )
+
+        super().__init__(head, **kwargs)
+
+        if path is not None:
+           self.load(path)
+
+    def forward(self, x):
+        return super().forward(x).squeeze(dim=1)
+
diff --git a/ldm/modules/midas/midas/midas_net.py b/ldm/modules/midas/midas/midas_net.py
new file mode 100644
index 0000000..8a95497
--- /dev/null
+++ b/ldm/modules/midas/midas/midas_net.py
@@ -0,0 +1,76 @@
+"""MidashNet: Network for monocular depth estimation trained by mixing several datasets.
+This file contains code that is adapted from
+https://github.com/thomasjpfan/pytorch_refinenet/blob/master/pytorch_refinenet/refinenet/refinenet_4cascade.py
+"""
+import torch
+import torch.nn as nn
+
+from .base_model import BaseModel
+from .blocks import FeatureFusionBlock, Interpolate, _make_encoder
+
+
+class MidasNet(BaseModel):
+    """Network for monocular depth estimation.
+    """
+
+    def __init__(self, path=None, features=256, non_negative=True):
+        """Init.
+
+        Args:
+            path (str, optional): Path to saved model. Defaults to None.
+            features (int, optional): Number of features. Defaults to 256.
+            backbone (str, optional): Backbone network for encoder. Defaults to resnet50
+        """
+        print("Loading weights: ", path)
+
+        super(MidasNet, self).__init__()
+
+        use_pretrained = False if path is None else True
+
+        self.pretrained, self.scratch = _make_encoder(backbone="resnext101_wsl", features=features, use_pretrained=use_pretrained)
+
+        self.scratch.refinenet4 = FeatureFusionBlock(features)
+        self.scratch.refinenet3 = FeatureFusionBlock(features)
+        self.scratch.refinenet2 = FeatureFusionBlock(features)
+        self.scratch.refinenet1 = FeatureFusionBlock(features)
+
+        self.scratch.output_conv = nn.Sequential(
+            nn.Conv2d(features, 128, kernel_size=3, stride=1, padding=1),
+            Interpolate(scale_factor=2, mode="bilinear"),
+            nn.Conv2d(128, 32, kernel_size=3, stride=1, padding=1),
+            nn.ReLU(True),
+            nn.Conv2d(32, 1, kernel_size=1, stride=1, padding=0),
+            nn.ReLU(True) if non_negative else nn.Identity(),
+        )
+
+        if path:
+            self.load(path)
+
+    def forward(self, x):
+        """Forward pass.
+
+        Args:
+            x (tensor): input data (image)
+
+        Returns:
+            tensor: depth
+        """
+
+        layer_1 = self.pretrained.layer1(x)
+        layer_2 = self.pretrained.layer2(layer_1)
+        layer_3 = self.pretrained.layer3(layer_2)
+        layer_4 = self.pretrained.layer4(layer_3)
+
+        layer_1_rn = self.scratch.layer1_rn(layer_1)
+        layer_2_rn = self.scratch.layer2_rn(layer_2)
+        layer_3_rn = self.scratch.layer3_rn(layer_3)
+        layer_4_rn = self.scratch.layer4_rn(layer_4)
+
+        path_4 = self.scratch.refinenet4(layer_4_rn)
+        path_3 = self.scratch.refinenet3(path_4, layer_3_rn)
+        path_2 = self.scratch.refinenet2(path_3, layer_2_rn)
+        path_1 = self.scratch.refinenet1(path_2, layer_1_rn)
+
+        out = self.scratch.output_conv(path_1)
+
+        return torch.squeeze(out, dim=1)
diff --git a/ldm/modules/midas/midas/midas_net_custom.py b/ldm/modules/midas/midas/midas_net_custom.py
new file mode 100644
index 0000000..50e4acb
--- /dev/null
+++ b/ldm/modules/midas/midas/midas_net_custom.py
@@ -0,0 +1,128 @@
+"""MidashNet: Network for monocular depth estimation trained by mixing several datasets.
+This file contains code that is adapted from
+https://github.com/thomasjpfan/pytorch_refinenet/blob/master/pytorch_refinenet/refinenet/refinenet_4cascade.py
+"""
+import torch
+import torch.nn as nn
+
+from .base_model import BaseModel
+from .blocks import FeatureFusionBlock, FeatureFusionBlock_custom, Interpolate, _make_encoder
+
+
+class MidasNet_small(BaseModel):
+    """Network for monocular depth estimation.
+    """
+
+    def __init__(self, path=None, features=64, backbone="efficientnet_lite3", non_negative=True, exportable=True, channels_last=False, align_corners=True,
+        blocks={'expand': True}):
+        """Init.
+
+        Args:
+            path (str, optional): Path to saved model. Defaults to None.
+            features (int, optional): Number of features. Defaults to 256.
+            backbone (str, optional): Backbone network for encoder. Defaults to resnet50
+        """
+        print("Loading weights: ", path)
+
+        super(MidasNet_small, self).__init__()
+
+        use_pretrained = False if path else True
+                
+        self.channels_last = channels_last
+        self.blocks = blocks
+        self.backbone = backbone
+
+        self.groups = 1
+
+        features1=features
+        features2=features
+        features3=features
+        features4=features
+        self.expand = False
+        if "expand" in self.blocks and self.blocks['expand'] == True:
+            self.expand = True
+            features1=features
+            features2=features*2
+            features3=features*4
+            features4=features*8
+
+        self.pretrained, self.scratch = _make_encoder(self.backbone, features, use_pretrained, groups=self.groups, expand=self.expand, exportable=exportable)
+  
+        self.scratch.activation = nn.ReLU(False)    
+
+        self.scratch.refinenet4 = FeatureFusionBlock_custom(features4, self.scratch.activation, deconv=False, bn=False, expand=self.expand, align_corners=align_corners)
+        self.scratch.refinenet3 = FeatureFusionBlock_custom(features3, self.scratch.activation, deconv=False, bn=False, expand=self.expand, align_corners=align_corners)
+        self.scratch.refinenet2 = FeatureFusionBlock_custom(features2, self.scratch.activation, deconv=False, bn=False, expand=self.expand, align_corners=align_corners)
+        self.scratch.refinenet1 = FeatureFusionBlock_custom(features1, self.scratch.activation, deconv=False, bn=False, align_corners=align_corners)
+
+        
+        self.scratch.output_conv = nn.Sequential(
+            nn.Conv2d(features, features//2, kernel_size=3, stride=1, padding=1, groups=self.groups),
+            Interpolate(scale_factor=2, mode="bilinear"),
+            nn.Conv2d(features//2, 32, kernel_size=3, stride=1, padding=1),
+            self.scratch.activation,
+            nn.Conv2d(32, 1, kernel_size=1, stride=1, padding=0),
+            nn.ReLU(True) if non_negative else nn.Identity(),
+            nn.Identity(),
+        )
+        
+        if path:
+            self.load(path)
+
+
+    def forward(self, x):
+        """Forward pass.
+
+        Args:
+            x (tensor): input data (image)
+
+        Returns:
+            tensor: depth
+        """
+        if self.channels_last==True:
+            print("self.channels_last = ", self.channels_last)
+            x.contiguous(memory_format=torch.channels_last)
+
+
+        layer_1 = self.pretrained.layer1(x)
+        layer_2 = self.pretrained.layer2(layer_1)
+        layer_3 = self.pretrained.layer3(layer_2)
+        layer_4 = self.pretrained.layer4(layer_3)
+        
+        layer_1_rn = self.scratch.layer1_rn(layer_1)
+        layer_2_rn = self.scratch.layer2_rn(layer_2)
+        layer_3_rn = self.scratch.layer3_rn(layer_3)
+        layer_4_rn = self.scratch.layer4_rn(layer_4)
+
+
+        path_4 = self.scratch.refinenet4(layer_4_rn)
+        path_3 = self.scratch.refinenet3(path_4, layer_3_rn)
+        path_2 = self.scratch.refinenet2(path_3, layer_2_rn)
+        path_1 = self.scratch.refinenet1(path_2, layer_1_rn)
+        
+        out = self.scratch.output_conv(path_1)
+
+        return torch.squeeze(out, dim=1)
+
+
+
+def fuse_model(m):
+    prev_previous_type = nn.Identity()
+    prev_previous_name = ''
+    previous_type = nn.Identity()
+    previous_name = ''
+    for name, module in m.named_modules():
+        if prev_previous_type == nn.Conv2d and previous_type == nn.BatchNorm2d and type(module) == nn.ReLU:
+            # print("FUSED ", prev_previous_name, previous_name, name)
+            torch.quantization.fuse_modules(m, [prev_previous_name, previous_name, name], inplace=True)
+        elif prev_previous_type == nn.Conv2d and previous_type == nn.BatchNorm2d:
+            # print("FUSED ", prev_previous_name, previous_name)
+            torch.quantization.fuse_modules(m, [prev_previous_name, previous_name], inplace=True)
+        # elif previous_type == nn.Conv2d and type(module) == nn.ReLU:
+        #    print("FUSED ", previous_name, name)
+        #    torch.quantization.fuse_modules(m, [previous_name, name], inplace=True)
+
+        prev_previous_type = previous_type
+        prev_previous_name = previous_name
+        previous_type = type(module)
+        previous_name = name
\ No newline at end of file
diff --git a/ldm/modules/midas/midas/transforms.py b/ldm/modules/midas/midas/transforms.py
new file mode 100644
index 0000000..350cbc1
--- /dev/null
+++ b/ldm/modules/midas/midas/transforms.py
@@ -0,0 +1,234 @@
+import numpy as np
+import cv2
+import math
+
+
+def apply_min_size(sample, size, image_interpolation_method=cv2.INTER_AREA):
+    """Rezise the sample to ensure the given size. Keeps aspect ratio.
+
+    Args:
+        sample (dict): sample
+        size (tuple): image size
+
+    Returns:
+        tuple: new size
+    """
+    shape = list(sample["disparity"].shape)
+
+    if shape[0] >= size[0] and shape[1] >= size[1]:
+        return sample
+
+    scale = [0, 0]
+    scale[0] = size[0] / shape[0]
+    scale[1] = size[1] / shape[1]
+
+    scale = max(scale)
+
+    shape[0] = math.ceil(scale * shape[0])
+    shape[1] = math.ceil(scale * shape[1])
+
+    # resize
+    sample["image"] = cv2.resize(
+        sample["image"], tuple(shape[::-1]), interpolation=image_interpolation_method
+    )
+
+    sample["disparity"] = cv2.resize(
+        sample["disparity"], tuple(shape[::-1]), interpolation=cv2.INTER_NEAREST
+    )
+    sample["mask"] = cv2.resize(
+        sample["mask"].astype(np.float32),
+        tuple(shape[::-1]),
+        interpolation=cv2.INTER_NEAREST,
+    )
+    sample["mask"] = sample["mask"].astype(bool)
+
+    return tuple(shape)
+
+
+class Resize(object):
+    """Resize sample to given size (width, height).
+    """
+
+    def __init__(
+        self,
+        width,
+        height,
+        resize_target=True,
+        keep_aspect_ratio=False,
+        ensure_multiple_of=1,
+        resize_method="lower_bound",
+        image_interpolation_method=cv2.INTER_AREA,
+    ):
+        """Init.
+
+        Args:
+            width (int): desired output width
+            height (int): desired output height
+            resize_target (bool, optional):
+                True: Resize the full sample (image, mask, target).
+                False: Resize image only.
+                Defaults to True.
+            keep_aspect_ratio (bool, optional):
+                True: Keep the aspect ratio of the input sample.
+                Output sample might not have the given width and height, and
+                resize behaviour depends on the parameter 'resize_method'.
+                Defaults to False.
+            ensure_multiple_of (int, optional):
+                Output width and height is constrained to be multiple of this parameter.
+                Defaults to 1.
+            resize_method (str, optional):
+                "lower_bound": Output will be at least as large as the given size.
+                "upper_bound": Output will be at max as large as the given size. (Output size might be smaller than given size.)
+                "minimal": Scale as least as possible.  (Output size might be smaller than given size.)
+                Defaults to "lower_bound".
+        """
+        self.__width = width
+        self.__height = height
+
+        self.__resize_target = resize_target
+        self.__keep_aspect_ratio = keep_aspect_ratio
+        self.__multiple_of = ensure_multiple_of
+        self.__resize_method = resize_method
+        self.__image_interpolation_method = image_interpolation_method
+
+    def constrain_to_multiple_of(self, x, min_val=0, max_val=None):
+        y = (np.round(x / self.__multiple_of) * self.__multiple_of).astype(int)
+
+        if max_val is not None and y > max_val:
+            y = (np.floor(x / self.__multiple_of) * self.__multiple_of).astype(int)
+
+        if y < min_val:
+            y = (np.ceil(x / self.__multiple_of) * self.__multiple_of).astype(int)
+
+        return y
+
+    def get_size(self, width, height):
+        # determine new height and width
+        scale_height = self.__height / height
+        scale_width = self.__width / width
+
+        if self.__keep_aspect_ratio:
+            if self.__resize_method == "lower_bound":
+                # scale such that output size is lower bound
+                if scale_width > scale_height:
+                    # fit width
+                    scale_height = scale_width
+                else:
+                    # fit height
+                    scale_width = scale_height
+            elif self.__resize_method == "upper_bound":
+                # scale such that output size is upper bound
+                if scale_width < scale_height:
+                    # fit width
+                    scale_height = scale_width
+                else:
+                    # fit height
+                    scale_width = scale_height
+            elif self.__resize_method == "minimal":
+                # scale as least as possbile
+                if abs(1 - scale_width) < abs(1 - scale_height):
+                    # fit width
+                    scale_height = scale_width
+                else:
+                    # fit height
+                    scale_width = scale_height
+            else:
+                raise ValueError(
+                    f"resize_method {self.__resize_method} not implemented"
+                )
+
+        if self.__resize_method == "lower_bound":
+            new_height = self.constrain_to_multiple_of(
+                scale_height * height, min_val=self.__height
+            )
+            new_width = self.constrain_to_multiple_of(
+                scale_width * width, min_val=self.__width
+            )
+        elif self.__resize_method == "upper_bound":
+            new_height = self.constrain_to_multiple_of(
+                scale_height * height, max_val=self.__height
+            )
+            new_width = self.constrain_to_multiple_of(
+                scale_width * width, max_val=self.__width
+            )
+        elif self.__resize_method == "minimal":
+            new_height = self.constrain_to_multiple_of(scale_height * height)
+            new_width = self.constrain_to_multiple_of(scale_width * width)
+        else:
+            raise ValueError(f"resize_method {self.__resize_method} not implemented")
+
+        return (new_width, new_height)
+
+    def __call__(self, sample):
+        width, height = self.get_size(
+            sample["image"].shape[1], sample["image"].shape[0]
+        )
+
+        # resize sample
+        sample["image"] = cv2.resize(
+            sample["image"],
+            (width, height),
+            interpolation=self.__image_interpolation_method,
+        )
+
+        if self.__resize_target:
+            if "disparity" in sample:
+                sample["disparity"] = cv2.resize(
+                    sample["disparity"],
+                    (width, height),
+                    interpolation=cv2.INTER_NEAREST,
+                )
+
+            if "depth" in sample:
+                sample["depth"] = cv2.resize(
+                    sample["depth"], (width, height), interpolation=cv2.INTER_NEAREST
+                )
+
+            sample["mask"] = cv2.resize(
+                sample["mask"].astype(np.float32),
+                (width, height),
+                interpolation=cv2.INTER_NEAREST,
+            )
+            sample["mask"] = sample["mask"].astype(bool)
+
+        return sample
+
+
+class NormalizeImage(object):
+    """Normlize image by given mean and std.
+    """
+
+    def __init__(self, mean, std):
+        self.__mean = mean
+        self.__std = std
+
+    def __call__(self, sample):
+        sample["image"] = (sample["image"] - self.__mean) / self.__std
+
+        return sample
+
+
+class PrepareForNet(object):
+    """Prepare sample for usage as network input.
+    """
+
+    def __init__(self):
+        pass
+
+    def __call__(self, sample):
+        image = np.transpose(sample["image"], (2, 0, 1))
+        sample["image"] = np.ascontiguousarray(image).astype(np.float32)
+
+        if "mask" in sample:
+            sample["mask"] = sample["mask"].astype(np.float32)
+            sample["mask"] = np.ascontiguousarray(sample["mask"])
+
+        if "disparity" in sample:
+            disparity = sample["disparity"].astype(np.float32)
+            sample["disparity"] = np.ascontiguousarray(disparity)
+
+        if "depth" in sample:
+            depth = sample["depth"].astype(np.float32)
+            sample["depth"] = np.ascontiguousarray(depth)
+
+        return sample
diff --git a/ldm/modules/midas/midas/vit.py b/ldm/modules/midas/midas/vit.py
new file mode 100644
index 0000000..ea46b1b
--- /dev/null
+++ b/ldm/modules/midas/midas/vit.py
@@ -0,0 +1,491 @@
+import torch
+import torch.nn as nn
+import timm
+import types
+import math
+import torch.nn.functional as F
+
+
+class Slice(nn.Module):
+    def __init__(self, start_index=1):
+        super(Slice, self).__init__()
+        self.start_index = start_index
+
+    def forward(self, x):
+        return x[:, self.start_index :]
+
+
+class AddReadout(nn.Module):
+    def __init__(self, start_index=1):
+        super(AddReadout, self).__init__()
+        self.start_index = start_index
+
+    def forward(self, x):
+        if self.start_index == 2:
+            readout = (x[:, 0] + x[:, 1]) / 2
+        else:
+            readout = x[:, 0]
+        return x[:, self.start_index :] + readout.unsqueeze(1)
+
+
+class ProjectReadout(nn.Module):
+    def __init__(self, in_features, start_index=1):
+        super(ProjectReadout, self).__init__()
+        self.start_index = start_index
+
+        self.project = nn.Sequential(nn.Linear(2 * in_features, in_features), nn.GELU())
+
+    def forward(self, x):
+        readout = x[:, 0].unsqueeze(1).expand_as(x[:, self.start_index :])
+        features = torch.cat((x[:, self.start_index :], readout), -1)
+
+        return self.project(features)
+
+
+class Transpose(nn.Module):
+    def __init__(self, dim0, dim1):
+        super(Transpose, self).__init__()
+        self.dim0 = dim0
+        self.dim1 = dim1
+
+    def forward(self, x):
+        x = x.transpose(self.dim0, self.dim1)
+        return x
+
+
+def forward_vit(pretrained, x):
+    b, c, h, w = x.shape
+
+    glob = pretrained.model.forward_flex(x)
+
+    layer_1 = pretrained.activations["1"]
+    layer_2 = pretrained.activations["2"]
+    layer_3 = pretrained.activations["3"]
+    layer_4 = pretrained.activations["4"]
+
+    layer_1 = pretrained.act_postprocess1[0:2](layer_1)
+    layer_2 = pretrained.act_postprocess2[0:2](layer_2)
+    layer_3 = pretrained.act_postprocess3[0:2](layer_3)
+    layer_4 = pretrained.act_postprocess4[0:2](layer_4)
+
+    unflatten = nn.Sequential(
+        nn.Unflatten(
+            2,
+            torch.Size(
+                [
+                    h // pretrained.model.patch_size[1],
+                    w // pretrained.model.patch_size[0],
+                ]
+            ),
+        )
+    )
+
+    if layer_1.ndim == 3:
+        layer_1 = unflatten(layer_1)
+    if layer_2.ndim == 3:
+        layer_2 = unflatten(layer_2)
+    if layer_3.ndim == 3:
+        layer_3 = unflatten(layer_3)
+    if layer_4.ndim == 3:
+        layer_4 = unflatten(layer_4)
+
+    layer_1 = pretrained.act_postprocess1[3 : len(pretrained.act_postprocess1)](layer_1)
+    layer_2 = pretrained.act_postprocess2[3 : len(pretrained.act_postprocess2)](layer_2)
+    layer_3 = pretrained.act_postprocess3[3 : len(pretrained.act_postprocess3)](layer_3)
+    layer_4 = pretrained.act_postprocess4[3 : len(pretrained.act_postprocess4)](layer_4)
+
+    return layer_1, layer_2, layer_3, layer_4
+
+
+def _resize_pos_embed(self, posemb, gs_h, gs_w):
+    posemb_tok, posemb_grid = (
+        posemb[:, : self.start_index],
+        posemb[0, self.start_index :],
+    )
+
+    gs_old = int(math.sqrt(len(posemb_grid)))
+
+    posemb_grid = posemb_grid.reshape(1, gs_old, gs_old, -1).permute(0, 3, 1, 2)
+    posemb_grid = F.interpolate(posemb_grid, size=(gs_h, gs_w), mode="bilinear")
+    posemb_grid = posemb_grid.permute(0, 2, 3, 1).reshape(1, gs_h * gs_w, -1)
+
+    posemb = torch.cat([posemb_tok, posemb_grid], dim=1)
+
+    return posemb
+
+
+def forward_flex(self, x):
+    b, c, h, w = x.shape
+
+    pos_embed = self._resize_pos_embed(
+        self.pos_embed, h // self.patch_size[1], w // self.patch_size[0]
+    )
+
+    B = x.shape[0]
+
+    if hasattr(self.patch_embed, "backbone"):
+        x = self.patch_embed.backbone(x)
+        if isinstance(x, (list, tuple)):
+            x = x[-1]  # last feature if backbone outputs list/tuple of features
+
+    x = self.patch_embed.proj(x).flatten(2).transpose(1, 2)
+
+    if getattr(self, "dist_token", None) is not None:
+        cls_tokens = self.cls_token.expand(
+            B, -1, -1
+        )  # stole cls_tokens impl from Phil Wang, thanks
+        dist_token = self.dist_token.expand(B, -1, -1)
+        x = torch.cat((cls_tokens, dist_token, x), dim=1)
+    else:
+        cls_tokens = self.cls_token.expand(
+            B, -1, -1
+        )  # stole cls_tokens impl from Phil Wang, thanks
+        x = torch.cat((cls_tokens, x), dim=1)
+
+    x = x + pos_embed
+    x = self.pos_drop(x)
+
+    for blk in self.blocks:
+        x = blk(x)
+
+    x = self.norm(x)
+
+    return x
+
+
+activations = {}
+
+
+def get_activation(name):
+    def hook(model, input, output):
+        activations[name] = output
+
+    return hook
+
+
+def get_readout_oper(vit_features, features, use_readout, start_index=1):
+    if use_readout == "ignore":
+        readout_oper = [Slice(start_index)] * len(features)
+    elif use_readout == "add":
+        readout_oper = [AddReadout(start_index)] * len(features)
+    elif use_readout == "project":
+        readout_oper = [
+            ProjectReadout(vit_features, start_index) for out_feat in features
+        ]
+    else:
+        assert (
+            False
+        ), "wrong operation for readout token, use_readout can be 'ignore', 'add', or 'project'"
+
+    return readout_oper
+
+
+def _make_vit_b16_backbone(
+    model,
+    features=[96, 192, 384, 768],
+    size=[384, 384],
+    hooks=[2, 5, 8, 11],
+    vit_features=768,
+    use_readout="ignore",
+    start_index=1,
+):
+    pretrained = nn.Module()
+
+    pretrained.model = model
+    pretrained.model.blocks[hooks[0]].register_forward_hook(get_activation("1"))
+    pretrained.model.blocks[hooks[1]].register_forward_hook(get_activation("2"))
+    pretrained.model.blocks[hooks[2]].register_forward_hook(get_activation("3"))
+    pretrained.model.blocks[hooks[3]].register_forward_hook(get_activation("4"))
+
+    pretrained.activations = activations
+
+    readout_oper = get_readout_oper(vit_features, features, use_readout, start_index)
+
+    # 32, 48, 136, 384
+    pretrained.act_postprocess1 = nn.Sequential(
+        readout_oper[0],
+        Transpose(1, 2),
+        nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])),
+        nn.Conv2d(
+            in_channels=vit_features,
+            out_channels=features[0],
+            kernel_size=1,
+            stride=1,
+            padding=0,
+        ),
+        nn.ConvTranspose2d(
+            in_channels=features[0],
+            out_channels=features[0],
+            kernel_size=4,
+            stride=4,
+            padding=0,
+            bias=True,
+            dilation=1,
+            groups=1,
+        ),
+    )
+
+    pretrained.act_postprocess2 = nn.Sequential(
+        readout_oper[1],
+        Transpose(1, 2),
+        nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])),
+        nn.Conv2d(
+            in_channels=vit_features,
+            out_channels=features[1],
+            kernel_size=1,
+            stride=1,
+            padding=0,
+        ),
+        nn.ConvTranspose2d(
+            in_channels=features[1],
+            out_channels=features[1],
+            kernel_size=2,
+            stride=2,
+            padding=0,
+            bias=True,
+            dilation=1,
+            groups=1,
+        ),
+    )
+
+    pretrained.act_postprocess3 = nn.Sequential(
+        readout_oper[2],
+        Transpose(1, 2),
+        nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])),
+        nn.Conv2d(
+            in_channels=vit_features,
+            out_channels=features[2],
+            kernel_size=1,
+            stride=1,
+            padding=0,
+        ),
+    )
+
+    pretrained.act_postprocess4 = nn.Sequential(
+        readout_oper[3],
+        Transpose(1, 2),
+        nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])),
+        nn.Conv2d(
+            in_channels=vit_features,
+            out_channels=features[3],
+            kernel_size=1,
+            stride=1,
+            padding=0,
+        ),
+        nn.Conv2d(
+            in_channels=features[3],
+            out_channels=features[3],
+            kernel_size=3,
+            stride=2,
+            padding=1,
+        ),
+    )
+
+    pretrained.model.start_index = start_index
+    pretrained.model.patch_size = [16, 16]
+
+    # We inject this function into the VisionTransformer instances so that
+    # we can use it with interpolated position embeddings without modifying the library source.
+    pretrained.model.forward_flex = types.MethodType(forward_flex, pretrained.model)
+    pretrained.model._resize_pos_embed = types.MethodType(
+        _resize_pos_embed, pretrained.model
+    )
+
+    return pretrained
+
+
+def _make_pretrained_vitl16_384(pretrained, use_readout="ignore", hooks=None):
+    model = timm.create_model("vit_large_patch16_384", pretrained=pretrained)
+
+    hooks = [5, 11, 17, 23] if hooks == None else hooks
+    return _make_vit_b16_backbone(
+        model,
+        features=[256, 512, 1024, 1024],
+        hooks=hooks,
+        vit_features=1024,
+        use_readout=use_readout,
+    )
+
+
+def _make_pretrained_vitb16_384(pretrained, use_readout="ignore", hooks=None):
+    model = timm.create_model("vit_base_patch16_384", pretrained=pretrained)
+
+    hooks = [2, 5, 8, 11] if hooks == None else hooks
+    return _make_vit_b16_backbone(
+        model, features=[96, 192, 384, 768], hooks=hooks, use_readout=use_readout
+    )
+
+
+def _make_pretrained_deitb16_384(pretrained, use_readout="ignore", hooks=None):
+    model = timm.create_model("vit_deit_base_patch16_384", pretrained=pretrained)
+
+    hooks = [2, 5, 8, 11] if hooks == None else hooks
+    return _make_vit_b16_backbone(
+        model, features=[96, 192, 384, 768], hooks=hooks, use_readout=use_readout
+    )
+
+
+def _make_pretrained_deitb16_distil_384(pretrained, use_readout="ignore", hooks=None):
+    model = timm.create_model(
+        "vit_deit_base_distilled_patch16_384", pretrained=pretrained
+    )
+
+    hooks = [2, 5, 8, 11] if hooks == None else hooks
+    return _make_vit_b16_backbone(
+        model,
+        features=[96, 192, 384, 768],
+        hooks=hooks,
+        use_readout=use_readout,
+        start_index=2,
+    )
+
+
+def _make_vit_b_rn50_backbone(
+    model,
+    features=[256, 512, 768, 768],
+    size=[384, 384],
+    hooks=[0, 1, 8, 11],
+    vit_features=768,
+    use_vit_only=False,
+    use_readout="ignore",
+    start_index=1,
+):
+    pretrained = nn.Module()
+
+    pretrained.model = model
+
+    if use_vit_only == True:
+        pretrained.model.blocks[hooks[0]].register_forward_hook(get_activation("1"))
+        pretrained.model.blocks[hooks[1]].register_forward_hook(get_activation("2"))
+    else:
+        pretrained.model.patch_embed.backbone.stages[0].register_forward_hook(
+            get_activation("1")
+        )
+        pretrained.model.patch_embed.backbone.stages[1].register_forward_hook(
+            get_activation("2")
+        )
+
+    pretrained.model.blocks[hooks[2]].register_forward_hook(get_activation("3"))
+    pretrained.model.blocks[hooks[3]].register_forward_hook(get_activation("4"))
+
+    pretrained.activations = activations
+
+    readout_oper = get_readout_oper(vit_features, features, use_readout, start_index)
+
+    if use_vit_only == True:
+        pretrained.act_postprocess1 = nn.Sequential(
+            readout_oper[0],
+            Transpose(1, 2),
+            nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])),
+            nn.Conv2d(
+                in_channels=vit_features,
+                out_channels=features[0],
+                kernel_size=1,
+                stride=1,
+                padding=0,
+            ),
+            nn.ConvTranspose2d(
+                in_channels=features[0],
+                out_channels=features[0],
+                kernel_size=4,
+                stride=4,
+                padding=0,
+                bias=True,
+                dilation=1,
+                groups=1,
+            ),
+        )
+
+        pretrained.act_postprocess2 = nn.Sequential(
+            readout_oper[1],
+            Transpose(1, 2),
+            nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])),
+            nn.Conv2d(
+                in_channels=vit_features,
+                out_channels=features[1],
+                kernel_size=1,
+                stride=1,
+                padding=0,
+            ),
+            nn.ConvTranspose2d(
+                in_channels=features[1],
+                out_channels=features[1],
+                kernel_size=2,
+                stride=2,
+                padding=0,
+                bias=True,
+                dilation=1,
+                groups=1,
+            ),
+        )
+    else:
+        pretrained.act_postprocess1 = nn.Sequential(
+            nn.Identity(), nn.Identity(), nn.Identity()
+        )
+        pretrained.act_postprocess2 = nn.Sequential(
+            nn.Identity(), nn.Identity(), nn.Identity()
+        )
+
+    pretrained.act_postprocess3 = nn.Sequential(
+        readout_oper[2],
+        Transpose(1, 2),
+        nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])),
+        nn.Conv2d(
+            in_channels=vit_features,
+            out_channels=features[2],
+            kernel_size=1,
+            stride=1,
+            padding=0,
+        ),
+    )
+
+    pretrained.act_postprocess4 = nn.Sequential(
+        readout_oper[3],
+        Transpose(1, 2),
+        nn.Unflatten(2, torch.Size([size[0] // 16, size[1] // 16])),
+        nn.Conv2d(
+            in_channels=vit_features,
+            out_channels=features[3],
+            kernel_size=1,
+            stride=1,
+            padding=0,
+        ),
+        nn.Conv2d(
+            in_channels=features[3],
+            out_channels=features[3],
+            kernel_size=3,
+            stride=2,
+            padding=1,
+        ),
+    )
+
+    pretrained.model.start_index = start_index
+    pretrained.model.patch_size = [16, 16]
+
+    # We inject this function into the VisionTransformer instances so that
+    # we can use it with interpolated position embeddings without modifying the library source.
+    pretrained.model.forward_flex = types.MethodType(forward_flex, pretrained.model)
+
+    # We inject this function into the VisionTransformer instances so that
+    # we can use it with interpolated position embeddings without modifying the library source.
+    pretrained.model._resize_pos_embed = types.MethodType(
+        _resize_pos_embed, pretrained.model
+    )
+
+    return pretrained
+
+
+def _make_pretrained_vitb_rn50_384(
+    pretrained, use_readout="ignore", hooks=None, use_vit_only=False
+):
+    model = timm.create_model("vit_base_resnet50_384", pretrained=pretrained)
+
+    hooks = [0, 1, 8, 11] if hooks == None else hooks
+    return _make_vit_b_rn50_backbone(
+        model,
+        features=[256, 512, 768, 768],
+        size=[384, 384],
+        hooks=hooks,
+        use_vit_only=use_vit_only,
+        use_readout=use_readout,
+    )
diff --git a/ldm/modules/midas/utils.py b/ldm/modules/midas/utils.py
new file mode 100644
index 0000000..9a9d3b5
--- /dev/null
+++ b/ldm/modules/midas/utils.py
@@ -0,0 +1,189 @@
+"""Utils for monoDepth."""
+import sys
+import re
+import numpy as np
+import cv2
+import torch
+
+
+def read_pfm(path):
+    """Read pfm file.
+
+    Args:
+        path (str): path to file
+
+    Returns:
+        tuple: (data, scale)
+    """
+    with open(path, "rb") as file:
+
+        color = None
+        width = None
+        height = None
+        scale = None
+        endian = None
+
+        header = file.readline().rstrip()
+        if header.decode("ascii") == "PF":
+            color = True
+        elif header.decode("ascii") == "Pf":
+            color = False
+        else:
+            raise Exception("Not a PFM file: " + path)
+
+        dim_match = re.match(r"^(\d+)\s(\d+)\s$", file.readline().decode("ascii"))
+        if dim_match:
+            width, height = list(map(int, dim_match.groups()))
+        else:
+            raise Exception("Malformed PFM header.")
+
+        scale = float(file.readline().decode("ascii").rstrip())
+        if scale < 0:
+            # little-endian
+            endian = "<"
+            scale = -scale
+        else:
+            # big-endian
+            endian = ">"
+
+        data = np.fromfile(file, endian + "f")
+        shape = (height, width, 3) if color else (height, width)
+
+        data = np.reshape(data, shape)
+        data = np.flipud(data)
+
+        return data, scale
+
+
+def write_pfm(path, image, scale=1):
+    """Write pfm file.
+
+    Args:
+        path (str): pathto file
+        image (array): data
+        scale (int, optional): Scale. Defaults to 1.
+    """
+
+    with open(path, "wb") as file:
+        color = None
+
+        if image.dtype.name != "float32":
+            raise Exception("Image dtype must be float32.")
+
+        image = np.flipud(image)
+
+        if len(image.shape) == 3 and image.shape[2] == 3:  # color image
+            color = True
+        elif (
+            len(image.shape) == 2 or len(image.shape) == 3 and image.shape[2] == 1
+        ):  # greyscale
+            color = False
+        else:
+            raise Exception("Image must have H x W x 3, H x W x 1 or H x W dimensions.")
+
+        file.write("PF\n" if color else "Pf\n".encode())
+        file.write("%d %d\n".encode() % (image.shape[1], image.shape[0]))
+
+        endian = image.dtype.byteorder
+
+        if endian == "<" or endian == "=" and sys.byteorder == "little":
+            scale = -scale
+
+        file.write("%f\n".encode() % scale)
+
+        image.tofile(file)
+
+
+def read_image(path):
+    """Read image and output RGB image (0-1).
+
+    Args:
+        path (str): path to file
+
+    Returns:
+        array: RGB image (0-1)
+    """
+    img = cv2.imread(path)
+
+    if img.ndim == 2:
+        img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
+
+    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) / 255.0
+
+    return img
+
+
+def resize_image(img):
+    """Resize image and make it fit for network.
+
+    Args:
+        img (array): image
+
+    Returns:
+        tensor: data ready for network
+    """
+    height_orig = img.shape[0]
+    width_orig = img.shape[1]
+
+    if width_orig > height_orig:
+        scale = width_orig / 384
+    else:
+        scale = height_orig / 384
+
+    height = (np.ceil(height_orig / scale / 32) * 32).astype(int)
+    width = (np.ceil(width_orig / scale / 32) * 32).astype(int)
+
+    img_resized = cv2.resize(img, (width, height), interpolation=cv2.INTER_AREA)
+
+    img_resized = (
+        torch.from_numpy(np.transpose(img_resized, (2, 0, 1))).contiguous().float()
+    )
+    img_resized = img_resized.unsqueeze(0)
+
+    return img_resized
+
+
+def resize_depth(depth, width, height):
+    """Resize depth map and bring to CPU (numpy).
+
+    Args:
+        depth (tensor): depth
+        width (int): image width
+        height (int): image height
+
+    Returns:
+        array: processed depth
+    """
+    depth = torch.squeeze(depth[0, :, :, :]).to("cpu")
+
+    depth_resized = cv2.resize(
+        depth.numpy(), (width, height), interpolation=cv2.INTER_CUBIC
+    )
+
+    return depth_resized
+
+def write_depth(path, depth, bits=1):
+    """Write depth map to pfm and png file.
+
+    Args:
+        path (str): filepath without extension
+        depth (array): depth
+    """
+    write_pfm(path + ".pfm", depth.astype(np.float32))
+
+    depth_min = depth.min()
+    depth_max = depth.max()
+
+    max_val = (2**(8*bits))-1
+
+    if depth_max - depth_min > np.finfo("float").eps:
+        out = max_val * (depth - depth_min) / (depth_max - depth_min)
+    else:
+        out = np.zeros(depth.shape, dtype=depth.type)
+
+    if bits == 1:
+        cv2.imwrite(path + ".png", out.astype("uint8"))
+    elif bits == 2:
+        cv2.imwrite(path + ".png", out.astype("uint16"))
+
+    return
diff --git a/ldm/util.py b/ldm/util.py
new file mode 100644
index 0000000..72d4f84
--- /dev/null
+++ b/ldm/util.py
@@ -0,0 +1,198 @@
+import importlib
+
+import torch
+from torch import optim
+import numpy as np
+
+from inspect import isfunction
+from PIL import Image, ImageDraw, ImageFont
+
+
+def log_txt_as_img(wh, xc, size=10):
+    # wh a tuple of (width, height)
+    # xc a list of captions to plot
+    b = len(xc)
+    txts = list()
+    for bi in range(b):
+        txt = Image.new("RGB", wh, color="white")
+        draw = ImageDraw.Draw(txt)
+        # font = ImageFont.truetype('font/DejaVuSans.ttf', size=size)
+        font = ImageFont.load_default()
+        nc = int(40 * (wh[0] / 256))
+        lines = "\n".join(xc[bi][start:start + nc] for start in range(0, len(xc[bi]), nc))
+
+        try:
+            draw.text((0, 0), lines, fill="black", font=font)
+        except UnicodeEncodeError:
+            print("Cant encode string for logging. Skipping.")
+
+        txt = np.array(txt).transpose(2, 0, 1) / 127.5 - 1.0
+        txts.append(txt)
+    txts = np.stack(txts)
+    txts = torch.tensor(txts)
+    return txts
+
+
+def ismap(x):
+    if not isinstance(x, torch.Tensor):
+        return False
+    return (len(x.shape) == 4) and (x.shape[1] > 3)
+
+
+def isimage(x):
+    if not isinstance(x,torch.Tensor):
+        return False
+    return (len(x.shape) == 4) and (x.shape[1] == 3 or x.shape[1] == 1)
+
+
+def exists(x):
+    return x is not None
+
+
+def default(val, d):
+    if exists(val):
+        return val
+    return d() if isfunction(d) else d
+
+
+def mean_flat(tensor):
+    """
+    https://github.com/openai/guided-diffusion/blob/27c20a8fab9cb472df5d6bdd6c8d11c8f430b924/guided_diffusion/nn.py#L86
+    Take the mean over all non-batch dimensions.
+    """
+    return tensor.mean(dim=list(range(1, len(tensor.shape))))
+
+
+def count_params(model, verbose=False):
+    total_params = sum(p.numel() for p in model.parameters())
+    if verbose:
+        print(f"{model.__class__.__name__} has {total_params*1.e-6:.2f} M params.")
+    return total_params
+
+
+def instantiate_from_config(config):
+    if not "target" in config:
+        if config == '__is_first_stage__':
+            return None
+        elif config == "__is_unconditional__":
+            return None
+        raise KeyError("Expected key `target` to instantiate.")
+    return get_obj_from_str(config["target"])(**config.get("params", dict()))
+
+
+def get_obj_from_str(string, reload=False):
+    module, cls = string.rsplit(".", 1)
+    if reload:
+        module_imp = importlib.import_module(module)
+        importlib.reload(module_imp)
+    return getattr(importlib.import_module(module, package=None), cls)
+
+
+class AdamWwithEMAandWings(optim.Optimizer):
+    # credit to https://gist.github.com/crowsonkb/65f7265353f403714fce3b2595e0b298
+    def __init__(self, params, lr=1.e-3, betas=(0.9, 0.999), eps=1.e-8,  # TODO: check hyperparameters before using
+                 weight_decay=1.e-2, amsgrad=False, ema_decay=0.9999,   # ema decay to match previous code
+                 ema_power=1., param_names=()):
+        """AdamW that saves EMA versions of the parameters."""
+        if not 0.0 <= lr:
+            raise ValueError("Invalid learning rate: {}".format(lr))
+        if not 0.0 <= eps:
+            raise ValueError("Invalid epsilon value: {}".format(eps))
+        if not 0.0 <= betas[0] < 1.0:
+            raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
+        if not 0.0 <= betas[1] < 1.0:
+            raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
+        if not 0.0 <= weight_decay:
+            raise ValueError("Invalid weight_decay value: {}".format(weight_decay))
+        if not 0.0 <= ema_decay <= 1.0:
+            raise ValueError("Invalid ema_decay value: {}".format(ema_decay))
+        defaults = dict(lr=lr, betas=betas, eps=eps,
+                        weight_decay=weight_decay, amsgrad=amsgrad, ema_decay=ema_decay,
+                        ema_power=ema_power, param_names=param_names)
+        super().__init__(params, defaults)
+
+    def __setstate__(self, state):
+        super().__setstate__(state)
+        for group in self.param_groups:
+            group.setdefault('amsgrad', False)
+
+    @torch.no_grad()
+    def step(self, closure=None):
+        """Performs a single optimization step.
+        Args:
+            closure (callable, optional): A closure that reevaluates the model
+                and returns the loss.
+        """
+        loss = None
+        if closure is not None:
+            with torch.enable_grad():
+                loss = closure()
+
+        for group in self.param_groups:
+            params_with_grad = []
+            grads = []
+            exp_avgs = []
+            exp_avg_sqs = []
+            ema_params_with_grad = []
+            state_sums = []
+            max_exp_avg_sqs = []
+            state_steps = []
+            amsgrad = group['amsgrad']
+            beta1, beta2 = group['betas']
+            ema_decay = group['ema_decay']
+            ema_power = group['ema_power']
+
+            for p in group['params']:
+                if p.grad is None:
+                    continue
+                params_with_grad.append(p)
+                if p.grad.is_sparse:
+                    raise RuntimeError('AdamW does not support sparse gradients')
+                grads.append(p.grad)
+
+                state = self.state[p]
+
+                # State initialization
+                if len(state) == 0:
+                    state['step'] = 0
+                    # Exponential moving average of gradient values
+                    state['exp_avg'] = torch.zeros_like(p, memory_format=torch.preserve_format)
+                    # Exponential moving average of squared gradient values
+                    state['exp_avg_sq'] = torch.zeros_like(p, memory_format=torch.preserve_format)
+                    if amsgrad:
+                        # Maintains max of all exp. moving avg. of sq. grad. values
+                        state['max_exp_avg_sq'] = torch.zeros_like(p, memory_format=torch.preserve_format)
+                    # Exponential moving average of parameter values
+                    state['param_exp_avg'] = p.detach().float().clone()
+
+                exp_avgs.append(state['exp_avg'])
+                exp_avg_sqs.append(state['exp_avg_sq'])
+                ema_params_with_grad.append(state['param_exp_avg'])
+
+                if amsgrad:
+                    max_exp_avg_sqs.append(state['max_exp_avg_sq'])
+
+                # update the steps for each param group update
+                state['step'] += 1
+                # record the step after step update
+                state_steps.append(state['step'])
+
+            optim._functional.adamw(params_with_grad,
+                    grads,
+                    exp_avgs,
+                    exp_avg_sqs,
+                    max_exp_avg_sqs,
+                    state_steps,
+                    amsgrad=amsgrad,
+                    beta1=beta1,
+                    beta2=beta2,
+                    lr=group['lr'],
+                    weight_decay=group['weight_decay'],
+                    eps=group['eps'],
+                    maximize=False)
+
+            cur_ema_decay = min(ema_decay, 1 - state['step'] ** -ema_power)
+            for param, ema_param in zip(params_with_grad, ema_params_with_grad):
+                ema_param.mul_(cur_ema_decay).add_(param.float(), alpha=1 - cur_ema_decay)
+
+        return loss
\ No newline at end of file
diff --git a/ldm/xformers_state.py b/ldm/xformers_state.py
new file mode 100644
index 0000000..aeb9f11
--- /dev/null
+++ b/ldm/xformers_state.py
@@ -0,0 +1,30 @@
+try:
+    import xformers
+    import xformers.ops
+    XFORMERS_IS_AVAILBLE = True
+except:
+    XFORMERS_IS_AVAILBLE = False
+    print("No module 'xformers'. Proceeding without it.")
+
+
+def is_xformers_available() -> bool:
+    global XFORMERS_IS_AVAILBLE
+    return XFORMERS_IS_AVAILBLE
+
+def disable_xformers() -> None:
+    print("DISABLE XFORMERS!")
+    global XFORMERS_IS_AVAILBLE
+    XFORMERS_IS_AVAILBLE = False
+
+def enable_xformers() -> None:
+    print("ENABLE XFORMERS!")
+    global XFORMERS_IS_AVAILBLE
+    XFORMERS_IS_AVAILBLE = True
+
+def auto_xformers_status(device):
+    if 'cuda' in str(device):
+        enable_xformers()
+    elif str(device) == 'cpu':
+        disable_xformers()
+    else:
+        raise ValueError(f"Unknown device {device}")
diff --git a/model/__pycache__/callbacks.cpython-310.pyc b/model/__pycache__/callbacks.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..b89cfcde975a470690434f67dc80da7daf64bd54
GIT binary patch
literal 2674
zcmaJ@ZI2tb5hl5-)vi`=cb3mS+i~O+Xsf>5xv#K;B4~Ur3J1>F$X!x&(Sq1pQo5D)
z&E(4VT~=RuE&>!t^9Nc~0{X@OCZGAWK!N)UZIa87(mB@{C<P9O!{O|3IP*|>tJOfz
z?x4^3AAtLpo?L7;Onwbr{tXC23@0eX0ZvIku%45|3M@0)LCuU#P>0b<+|&y^qhlwH
z)DQfXelutqy;jhIv6i&cPS7!H4(#a$UBlOtwRAmL2i|2~aw+WvJ>VP6ACchlZxL%U
z{{><`@19t}1|xfD&^iSV(SRuXejJs`dN(`T9bnaYKW9AIIpoo#$m6U8k-xY9o5%Fg
zhx?B|+=prHK^l%YFz!y6BnM$M0lqE5Y(kH@$Z4J>N2gfZ#;R*{ALgSG7w^Z<<4ouP
zDgS~v(r#RAoz+DC1sIM3%uqlW4lIv&mS?letj2BTK(F%}cX)kBHubU5(ES7rG4ojC
zMJsUOu->@Anm~P6wTBLCvGxl)X!6$otaW&sb$OkwfyH&`TDQ$EvEB>&1P7f@kd1;a
z1Z+e2;L!0$jXpFuQa{#lkbO24aW?9gB8)T5<f8vHOkx(6FxxWPhU`DO|LE=h-n1xk
zQNE^f`gt9pQyraMwU*=~%Aay^L^CQ&UP#qT!{>DQLCVAE(1@3cE9vA}C`M8_f|pZ~
zMcUGN(>#H0L6_eJQlb+y#|wl}g)2l!MSxnBRdh=G1OX4Jxrbij-GQTOa$0bqTuR}1
zC8eSb286yOWi!Ye(FJy3D_inp_|2jbzr@O;G)h7#DSe6lb8q`FPx*G1m;4~lC)*L^
z+BR61c@`$ydv|aD{LWAB{$eTK+1dMGJJrd!9j)?AZWTwO2cLTbx^#gw90xnNi;sI3
zGHUDGRaG~a0^WN?Lc!KF;gJr)dDFb^JJ97ikP=m-#3w}7-$Q?b#M?tci=j#yv5YfW
zhFQdghJb-3Hee9fHMyb5<|}dyIMvKEI)s419#MWDBvN0&IzGO7(bbDes-92FVp>X7
zI|$3@P<dvcF?+7skWOKehl~q7k04oIHGKg(80HlBpz<m*hD#jd4ZxoX5r2m{nboVh
zSgXj~($`+v^V$M^26zE$PGt{qtasG^iT{dc?(YDpZ1|m67-jBHko>_jjA0i>zt?-+
zu{UqPK7ZU~gjow~=7FqLdCX?DPXXsD_dfdL2eU@yR*g+GZ!gfivp_Sy?7}zJ%Js@W
zff&!4<)v|Ne7W)&s+xb+MeMNpr)1WuTJsHtnY$n`0-nKmg*6s<eif|1D)W12z6lyW
ztXwOvk8dpSti_u1H!AIyKGV&rwy<aIm9JjaUXWR*d~@EfT43i^)mgxap4@|&`~zb0
zmX3+O?ejip?FtQHfoH0VLP4sUNhXbeE`J`&lI9c90v!<8naH0B4<@Jjl26Z|ot^=A
zeCy+xDb*i;{Pyi%&M04v2g`A*$VP+jD1VgX2Vr6qO$90U98Hun6r3M(_<AlNCF)sD
zM<QgZp)q1v`li$r$)&Q3usl@u7=T+jpa7_=Tp2#)l*PcM;e-PNAi;PEAfl|Om?|p;
zST^-+B|KBU`Os5H_dHXzX&EO{d3q;&H)hHM&oWu$lB>FWGF|$XP1E8?*-;V~$_Zt8
zRB%PI0=B~oatF?ki!>}%E1C+yvy$>6j}C<c0aD&j<Y_2G2>Z3C%AID*BM-XYuCO;@
zgXn8=OOv;NC^weW1WP%qX#7lzSNAsIN2+Gz3hyf0oWLE%aLzF0;<mQ(t+HK*iZu4<
zviNw$eh6w(>vn4nwy}#FWP{xJs#j|R--X{*e3M*-l_s{&+SWR@32^_mH;->V&`-oe
za}isw8wLz%Y8P}C?GDy7#%jMPRSmj=E8v@yKADEeYNbvo%Oen55L%(LO<A%OLc^HB
zMn#H5)%W#l1h?OL7>^E1{Vu_~4)7;#YCVk>qNd3sAgZ;(=!ho?0Q@%*xuR7xnwh6h
z<DGm5h>L7YT!IO<zaWHsMJ)1V!zN#lTKm7(e3(Hj$eHe-b7nNOmeK5eIk=)6$PjSJ
zxWDv#TYp5`7ViR4YsHaCLz<ksFFw-p>#xbTUOf)WzZSP%*U&i}mob-HIx*$);6Kve
zx~pwkj~_fVxhOQKs_*?j0}Se3#0S%oGZRGRr21ctG*5zpyMdPqBB=i;TnM8HsCg9J
hUoBsucQm-jK9DYQu}fUYohD?{as-|b8=?p~_BAD6&N2W1

literal 0
HcmV?d00001

diff --git a/model/__pycache__/callbacks.cpython-37.pyc b/model/__pycache__/callbacks.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..d85ad13d6ba0bee15bfe5f5ff43b4f3714d22758
GIT binary patch
literal 2593
zcmaJ@UvC_@5hr)|cK2@YpX9SFJBcDUMUqQVUD1ak2-+fvZ4?F+*^T8SSTrD3OWtWu
z`_ClLk(9lc%0-^i_5=9g4(LmPzV(arnXi4)SIA2`vv;zM7$^@M&JKse;c({1`?S+(
z5olli`<nO@aR0@RtL6anF?97D2q&DTB%>kCtk9zPZKZZ-*P|0Q>d_57812;0nqjln
zane>6guz<B9ky$|PS}C5k#@6Q*sJGUSkn*tHSeVx*=D#2yw98IwQLX$fN$|&VujcL
zKzN%6&j=4h|H2Npxb=t(JD1=g8Cu4<o5b4KpX6r;Luz`T6<nlujzxS@7D=u_6g)b-
z_mDk!eE9J3A$&IOXVF9e<KKzW^eBo?fbYsEKVj!W7OcqAvrCG$sp;3c`^97;<Y&og
zl1mJb30@J$=*CrRt{IgBBS=U&2`x?`WL@FFuOVFFjjfv)%Eoi&J2Ix+<=(SS=!@pG
z1wG&n_n+Im$y?8yuq`@Qa?uqY+JO1sxd#F1JaaB+*!zw+B<w>#dNP1b&%bExLIam~
z6FjHd<ujS&lU*&NBuA!@yHBGu;gN=~y;{2_hY#*P_-OahtSk$uUsJjKk4MNQp8jCy
zn2ofUF!5B#GnO-@MXAgni%!{UOD3ZDxE8Mz*V2=3A~{jUl|s*C9-}R6I9kwc=;~b{
znp}_tT@p$vS`lVd7Ersg%N^}p5a1#4;Jox;=$eL_l|mYyF*uxNOm@M5>;W+jhrEz|
zV8LNWiFEvOawVpw$yl65N-_3N^25E+agm8pUTASt6elByP&AszVpfhG&GLPSR(ve-
zQ49$-x|@v0pWgp`lwr1w;&mdaz4A<6gB{<3u6!WGZBZ9~=Yv;LX79?|p>4eSUf>AL
zGG$<GmI}E6>tYNCHXO0vfvz@zXi{04URbKDtv^G8twLd&R-%$z=_rqd!~!?nPm~f;
zCq;fw%0dopc^$;^CX!o7ZoDAdz?pWDvoQn_Y)Jh+NEBWRq37GL`u?iY(0ggkc*RVY
zGp$VHDAMt<Y1RWv_^Ih)DAOY1LgM%8fRImh!yCf7sSo%86kcUb;g+X#3*hI(`a6~H
zg2bA8l_!5*SqmF8LpzJclKgE8a>Satl{1D+nR<VvbN@@Itv`~_$yY@C)8?WD^TD*u
zEp9LEc@t!vstGmJ_#Oc5^k0>Kmwf$?d8_iP)=jeLF3F;|B=bP`!QO`6tbz*&^}MaG
zO$XEKRlrHr{tMPGRPFcHyi;`+Tby!#X~7841>+6gTGGWfSb<p{49MapXaq2GOTRIF
zb4lkN-d?;_p=ap*PSsdC^X}T)X4PF<^PYZtv0HV(&h4tVq?9}<za*8nipAn*7=ilY
zI|mYxDs0vz#-J)qqfRIUI$xY7O0(ibc0dONPA-dYB*yv@Z}##>XqP_%cYgcZd0nr+
z{Pv??{btU@dOTW<du2WuzGH$(S{y}btyq_m@vlI|xML~Axq#gZ33u=11)Io-n-&7a
zs{ZSmGghIDQ%3sOI8(r9<AMT!tMOIzR4|?ZmqjN67(faxG$4tw<8o%~4A8eOY&#XX
z3F?iWLe3YtY0PwzD$~T3uy4Xm6FkdRStwyV^<=j4EuUrOnQ`JYDUBN`eO3x%<s~eK
zhvGvxLm{(Bn@&8FQskP6vWSnR3js3Cu`IGk$_Un@r^cV<t0NElKP{^rSRmg<vWw(4
z5aTC`)xk3EIvT%5@%r9Q#F=T-azhV{Q=h;eCveUv6LN%Bep0KOP=>WVwyH9b^Iw3P
z!m=feO+Xxn`c})jWe;eVwrC%I+tv=<26^A=(t!n-v~wR{i2eEk_Fgv#fYHXSF?$Y%
z8wj%K4`Zgq)=&j}o3SUeC|%EZjPW7{u?-=Wm}SORT1rH)8X1jMrp7zO#|Un{x1UUo
zH9kx5Y$Nm`a7qMD*+6mt#B|mm9g8#tgnto<A(dOGgy2=befXsYK>UW|K#9PaH{C7U
zvJiuZH;fB64qMf3tvuSprN2Y+F%Yv+p4Dl^(kr**W0b$~ntbmC<gfk_+<sjHb28D1
zP<xnYYIPvIB>5rQv>)EzuX9Tx5SsV?p8*baER&;|7Q7CQaWnigqE<K$F^BgDB7^V$
m%kZqY8rl8du3nbkBUq?6f%G9eeCpFS{I15f)pjiE+W!M|Yr*~i

literal 0
HcmV?d00001

diff --git a/model/__pycache__/ccsr_stage1.cpython-310.pyc b/model/__pycache__/ccsr_stage1.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..1409dfb219b409b1e249ee4224535d9ec3f5560c
GIT binary patch
literal 10337
zcma)CS&Ss-U9Y>VtIz4_x%TX}y+^O@8GD_?Hr{OF?BS5C$H8OAaS^Lj(^WImGu2hS
zU)Ai<s<so)25*2#jEFcDF%toj^8gZv2d;=GUU)-@hwn8=XeA_gK!6SL^80_)*NiuT
zZq>iO`}lwV`>Uu_vLyWYci#8E_@X5Jm@<<;4Vlm5i4|FrJjs&-sUzERN3l_Wyb`Fk
zinkhQ9o^OkWrl5tx}2RGyiM`WcP!fyZM2}!DcZ$O$u4!ucDYlrE23Nvs+}o&N~DdT
z)|s}aJ9WF>nXzX&v-WIf&Ylx>xnRDtU@wTY864>xwU2g=*~dD^?c<^>ADrl%v`>n(
z6*M}h>{Cb=ykhW3=TZAnS$awGN?!S{<dyxo9o2pe`HELXzUo_>r%`eSB~xAvB{hEm
z`Lmv~A~mP)6Ly*k&%Nw+yX|nDt1pIkmYXs!yuv)6`QDXwD`{#x*NVcMPUvwXXotSb
zc%kRB$mvAhcHr|`((d?i;&&atv+8>u2Eq`vm9Cq#-QYFmhVfd&IzDUaeD1YD+pE-c
z#lNxbht!Ijuln&9f~d8{tt-*3Fm^j#3|)RPNvJ^-W_9`3m;L1B$nyitP`U0><(2l@
z+IEa4O_i5}Xx&L}CypCC?an$k{M+p~iTPYRj1!k;N_=OHMI8q_T5Hofk6mfI>rv<i
zFS*-s+;+n&=)u~n+cYccOK9Hn!@*|ngK{J)#qL_zfLJDL5gsRhDl(tP6VC$pk}Z3Z
zt$4bpdfHtXG$sT42A{qhg$av-!1rDc2m4yS)-up%_$wh{;`uV3xQxlyrG74v_T+u(
zyU)l{qNGYHZDK%A*;V^yD(z{hxi9zgsl2D7#z5M_J9kBTtJE)~h15#(sk)=0G`DA_
z*1jxDH;$uUKC9VB%^P3_4gVcgmO}GsDZY3Mt?2(Io>cRsZJB);?X^8?UwT&RTS)<P
z6nA9I^7z-4ei1Y2u`OCkY%P&COK6>=*6Ru(VOQx}VP&#B8*6T&esZlhzLHwlk38JZ
zUYS<DB7LAQU&Cn>nBVo?#EHV-4x0h0P^NnAPP4-GBx0=%o(m&qow*((Wb!g7*AJZ5
zh8u={5Oa$tIHO*!C3m_$*FEtnw!`jr;;a&3#@qmx_~3$Kr|o%xpA~bp8+Cc!54S<#
zATM5uwv&lYYRgTyin3;jYq1}!arrjS4_PJ_yp{O3lPm#h;8og&gE@#ui9Wcob<1Vz
zF;_NXY={=hRRZz`HwH{7w?B~BXRx`Ct}Jat9e*i|5`Q&{ww7Ah>r&TeF%kLF%JWZs
z>bVQge>y9^e0k;7C4r2kRx4%>m}K35da-+l=fJX<)LS6JI|HD}mR!YC6i*T9V;`FN
zqFhso^1OUPDa%F0l6yxVLeq<bv6|ZQGR{lmrsISzFz#^6888;oMaQ|Z?FLy5dk%Ao
z4NZV&sca5l9#2f-+?2|)DER~nQZ4=BuK^5MFfZqoyqeeYI{t?EQ;Pn%yotXmAo5=2
zJ!wt$@}6}!4~B)Xy(`&TCT4B@J#dF@fQhTz8f+MXiIH&lKob;rUqSjlE#f|r+-vL-
z;I;b?@DugRtz#${{>~v$#zS%=?x9yA?Fc3ZyG^r7IzucMSH1+DwIjbPXL;64r{e{b
zDO5&m6(}9Ge+*(lOJ{cE^w^fdnu$*7{!Ci&bmW)PQooXxMEZrajJF!UfH4d)F69;W
zlBieXmqz8*J=rl*88DYBfM%)!<`XN`lfsSyx-X_CX4Kc^x`dh%n7f$jJ91XOrLs;^
z_Htqlo$AwhAw=Fr{rsqv62JW=>0S9A#190_50m0OiR$jjo24CjONn=>HmM}l_oVma
zHN`W%{9QAd%G&M7JHSKyd!kLvv+ha13musnfW_})CBUSCnWxhnU_CVfXVN_2Y+3-E
zON;DB$^7O*r~n>8$<ed~cq}ag9#1QPC(>D38B!-x)hpGd{_&)-cPe=#Rny};3eu16
zOUNCk+&*FGF_fK7&Ln3c{*+XklhChBx_K8XT9@}}taH@np1jvg)%{`Fd4b=@Q*~YL
zpGejINw3mxBu@~IA!7d#BKB1Okz|T6lb%`tmQeaj{YR50)03DzmtsYA33C?GsyDTp
z!zhnsW&M+BBQ<VH@5}6);`9}yE}~cIh}1tF8c3CqDy8NA>2wZt%jq=W^{hoGQ+cvT
zXNJ<RrSmj<|4gWOwf>o8k<JU{Yw4r+B&ylU=6P&yiAIT&<SC3(NK5JIwB}9k>R};y
znnp&xp3bCa(xYj?tM5XA$)KXmb9+n4ne?&mNQ3<Oy(NLCll#CNwa}!^$J3L~N{Zw>
zo}R!gsF5+}jfM17MtNRkUfNQzrXQy#`)5aEPNC#SBW%&>5T<6l*<CZtB^R(JQio@~
zxm|5X%$1%6=0-DMoI`6#8{Yi$QonI<O;|~JgynzDaQF}`D@+zxevq{oVEN2wec9?3
zb`7*r;<^6=i${i7{AB+fqRD55I7IqW2iNOK5NSu2d#_%LqTtl2#+%W01G0NPN*YPj
zXosy}+w&XAhTnj4^+Phc8qr$gPPEM$Ef&S`lkU)7Ye08-ZL*<Y4lXV(_TGHxnA>qS
zUYr?=jbv;s(#-#_p@rCm#<#A+a)UI6UB_O8AniSKwT-p122``N;k#bkz=*@lNc%T7
z+?ckv>NjFP>Alls(akotNe!7F50rny&&EPqHv}q@#<}$%T6Kfq4sBIfmrWX<8Z-i|
zWGUH(3EMaa)%#@N--OB9xP>tU`e5Grjld7rkw=GtA>wqlgQP`DWcZ^r>4a%KaSZ^J
zC#BMY1pWft(qRJ0WU!YHuC^(&Pown*3a?Ye=u`%qVNFzKdX-cvH!lWmXVr5rCCJ63
zxoT4H^jNViI!-%mCyujt5tfh}#FvoBdi0(}GZ@U<Ef`uv3?mD@!TgrrzUg}clZTh^
z+1?9_jn_8XIFl|>i8zJpS65aVUDl3RJGlcy0n_7tIEyF{R(ag$^`2<77yU)Ru+eg3
zKW^N@NoR*P?vmMUzu11OajEfC115dr!uj*<223OBH*PmE^b_M^99IxD;)Eb|$p~bT
zeGd4!Pp2Xj`3Ex3b=}PC(RdMt-iWfe1~!R#8AOgUzqQqkKsdZk`50&f+-R>QP6+P^
zW)XWCWAglvKw~~NOpZsKqrRbkYvM>Fa@F8k>7L(BHh2NlNcN<Ix$P;h9lNVE5e!f2
z6WSJykhyVU!q=KQBkzJ;CqPSJ%d`&7gB#XVvSQJIESRQxs9p2ZT#vUw<7CujL}J%r
zf3CY&$J!z>0wutbU!b8+ilM1gjKi&V2NtIjw|hP>w8POx*>P%Uwd2fjYx!Iy)T!VZ
zc12Wu)eW}&tBggAtKdSecHA!4AwJe&y@Q6(dgh@g%yS-#x;Q6Zoa8%R!a8pf$7epB
zI1B{{&8-EIn>_uQLE@PUtbqlvR|tHCmQouYQ|xSav)*>S3-1}uGfa3-nBh*_gZ&<c
z+!{N*j66NAt+cPd&Q4I$f@=$pn&cRxPzKWpoT937$KCQB;dawHxiw10cF9?T_p{A1
zJhIc&sX$<gz#|06k7RiQMFNik@am)+<B*4PE)xu7F|#)b&?(rZ4Ip!TGK7A%N}xr+
zBj6Jl5fC?t#@4z|e`bs&)}|&zS#||D7Muxm5s|T9pi)8E12y1Gr~!LwBI5`eXDiep
z7Y;l&BOq$KW6vEX_AGJ`vuo6Repox$^_U|T#vB0(o}63ItX&>^hZ$2%IE4Z&I9Rxt
z3JzEof}U$6$Z^^X0TcHQTrb-MHwqnO7Y6I_oY>_?CN3aD@{ZFA+&Fd|b`{HI!qxdI
zQqSOtuR{@7Cp0)+CfqIhn@DMRsH7?ziXqeAl528RF_AarI-XhiF_bB?^}x`x76w`j
z+lmp8L;Ig-hN9GE<$?0Rcu>#LcpCmyv{n=aBPeK>y|oXu2WG*-xRzK0Mug*M$(k~c
z6)8F6qpY3i^WTbr8uZkDqJ3B`)#Nh%HI#_a3dVno+h!@_zm$S0YY&uv*B)p!+K-7I
z+BYoBLF=X24)@VMQp!JVRz;s|t$k$tOYwM3?hQjP@Y+VotwoA&_@gxO%Vj(<xoU}&
z$OyL}*pkYrOaT{9B`bc1VmaEf&;`O0VLy-7kmiDcb4X`gOxyu++Qf^S!^p#7-!?>T
z9%n6d@G-f!a5;+r&@vASfEI~{;g5KlW)e}N-$EjRqk*N0Wg;$QAikrd6bs5F>ZXPb
zC~%*$IE;*lk~tAA$|pJWL?mf5zoWpdn1-`qA)Z5Vqg<*Xa+FWaJu*oLc`~_goW=};
z5=4-)7SdiZYqJl>2JV-+A~AB@n<Wv25yC~tJ$4kO>?{G&4?@0jwb$+n(Z?bx6H<)C
z7V82ui<zW4k2KdvKC#EC?g;`*1fC{v0f1}Zv6vAXu}=Zm`H|#2jQ!_OT_o&N)Q(~u
zM5sk&7VK7+%U};p>D4AqWEj?%$<F<==r!>3uL%Qm46fmqx&6zpAp~-*nPbnRHTyXN
zFAyMBXLS4`@WD$1anSOE03v|`I~nLvN(hIGNYMsk+Lb9O`AI_mS?TgWN$;3%R8>YJ
zkPJWYBU++g{D%O;xQQqGs;~L_nzDd!gRgn2Z+IFW-OqUlkFOO)v?S-5cZ-m96g2^&
zeQU3<FL_5rUC}Gu&F{!|ksM&umIk$OQ18NRw9AA1F>l(d-vwLQl|c#IGSQ;y&3W_a
zIpr;QNARwRD9kk2@AxPRbN$N8ybc9PVY4inHHh?Pb`h0s`FD6e_8f}o`5s%uHhC#P
zRGT~%M}+VU=0lx^2O_Z)#i6HyCw?1%LNuFlYP6*^_xtj9K;aM%Sa&KpDn(@-jd;+}
zQw`8a5sh_n9vpm6fom=~W|H4S+|*N(;$8{dR^F`aQ&2-2<*SrmM^Iy*#+*uOXf;KF
z4B@8f%m0i;;P-EzyHDDk=lqb24aO}*(MVy#OlVHC*HM>I7x@pu$`QyBQ9rgrwRZ^+
z9@w|2NqK}nvQcBUNi|ypG^*BG5#ocU=lfks*;8w67AFH_XQO_b8vimiE@CXt8LW)e
zQPeau)_s@SJZyT6LY)~C3t|_vKT?Od1(9NsA~nkzC_MxGY2>REo&310jYK<0;o+j4
zc0tc&JTXZO$S#RQdL6Vqk6=SzNgyG2<UM6y>Z?MMBAh}g(3-j|LV1FlA?3e|{%jfm
zN*~zg#mul@bIBh<J;{6aB!FFQIf1yaa5}yl-o!16IV_prZ{d&4A<&EvFyT(P&_e`%
zjy6w3Si_b{ul63foY6XM_NR1OZqa6kKe7u+RR}Wv9ufmG28nHg*i!8W@~aX^S|^f*
z=r{T~s0(O4#CZ_)6Op+_iW{(0{T74)%1m52NSiv)W+uj74t4UrLgkxUVs7SfF(Fgn
zpGu$@NKMcMg_RnJ&>%9ET6pq!Ot6N!tF6(^l(yWftu{6qw;HX+lb0H+cp(Enx76`F
z5xe6+^mbgjnkuv$5qTirhW$GB!<FC$VHU_k<Q=+(7=Y*3{RGAx;_|#QNv{wFf(W^i
zMBMQFn{BunhTjgOZp`xxQbBkWLPPN>jG468x1;TNumG+L=YSC+$=kziSs_#L`l_42
zD3rIjvKq7ZQJ;~t&{%RykOccpDyBuVLzMp8D7#9BO&s$GnLaph`jbYoG*|}piGv^Q
z?Z8P}=t?L33ZD2k0LUmv5G9Rs6eFec-h^q8Xt-Fx1riB%oLdfQoHfo_Vy9EYd?Ctk
z?}^qB_`5ot;tJ(J6@#AC8a+ijeOK$}p%8L+rNo5t9~cpcKj(*IiPDL&n&vX2)nx`0
zAp;I6VQE666aNSS1%?t;B!9>C2SO$)s)*oM@CZS%j^(~(f|SA6w4J*2Ysw7}*_i2Y
zai}al#r^<2;LMWH8`lY4%dLx<Bird-TBTzl>u_G`ou4>_jQfQAn~ahO{}<hw7PHZb
z?Tbl7GhXqASC@1N;s+CUkdVf|LLK@9eiZ;CkT9`xw-GRC$9K4#kSq?p3F`VGH881=
zZb5NL)QvMr8ygTpz!WA{`#Npt2s)NQWmTm}c1|9~IvMF7&uYTznSJ<1KVdPC#A+4e
zkys^#mqM`8WnEKQ=5#YRPxLAjt}m{*Mq$FCc`Pp<#{UWGvJL$<5Qr`=?}lk;?e*cA
zRFz(HVyQp7*L|{{T0}3yUmgh^kI)Q%g#-yl$bDSJ2q6eE2OVjBJ=HUYfdGVN7Yswf
zFo(28?63d<>gicZfiOf5X-`eHec3bSBrkWAE&|Mpu!-@`<4v`>1GN@P4PwlTxX!^v
z1~Dv+2s>(5ofxG|y*CYvq<EEG1wE=LAx$;4oCz2pXCP(BHJD&F3h3ZY@kXYfDxOdF
z!|2wrsoY0kU1EQRR`=;Vadesp)Myl6!e;{+q0v3NPS^jq%gmx5G0a<{b4Fr}{SE-H
zWrI5tvX{}AHW?dmh3Vc=SY}o>1gx|5W<JxNKcIHSL3uWjifS(R1k2hQ)aVT%Mr|E(
z-%Sovk40Yvh(J@kMU+CZ7DFK&T9l8-v%*g{<pv0z%x3FjP3z4}P|>*e0b(KfN(3W_
z<%>u_zQT|NxyYOH*EJ#&p*=}E!uj&#BQO*Ru4#~>D*Q+Vhem#+aK$`*HwP)1$H5uP
z?DH6p{V{>}0C?_}ivY4u3a`?QVi*NCeFuT+P8a$tMEGN}Am3VB&*F<@$^{Vv)^9s>
zpAGRjOT+2$hKrD0N@*iQ>_ED(Kc_kWLd=;}FfOkOJPQ|!%eUCyQj@<Va8RnSzed^H
zgkU;Z6F0)Mq^*P;IfT;jAs>+46b~DT5Lz8jC-r4mV?g1ax=-jO+iaX8FIS7tr&+1K
zX>8`QcK9j*DoM1ZQgd9I<TtI&0+0`j58okZgFKE+<jG!#pCiN-C|(H|oXu}xE$j^f
zB$K(-br~Y~2_ie0@%#Z*k?>__2%IB8PA%6dy4~UWi=NwgqiJLW{dZJjh%pcu%HGZe
zd?bL=)Br9VLagv;5JII0pBN9(QWk%kFikeSA%ZFfussVK7y%)<cXEP!!C;^0dw`Hh
z0%EL;kO}?dUoawTu?~cMN$)UXL!K9A4T71s53jLu(+%1l2r&x5&VKiMv^>(2aGH-y
zEay-&F%PaQ{uC{n^}}8<_2(+0p^O&EHGI$zustL*kA}>yOlcmYln^(6M5#X^@HYgA
zCfVN;_y+>tATTD|9%V@*vTp+5Ypanrtyqfkv7#zJQ8eX$5JW;COZiC2m;cjj7zS`$
zF%X<GRCpeEn^v_TMrfYn1{e)0iP6%<s@X15<bL8ipBAAf7Jd}_J~a@Lk2PvwiVsy5
z!%+BJ{8Zp|7GV`292_r>?1W(fu0et<tVArs$jlQl6?<l~+bH}kBER;z$+}Mr8R63w
zx1HG`B-*kFGDQQj;Bgr#`}AN4d^9s&)S~Bg@j=!2`w>RY1vja!SSX2uEEY}VYM(v0
z4xH}bYbzRaBA2;UMH(VrY(<?;6bcX0E+b5gkFO9wrH{mzkZIy05N^CIFeV}!;y@<O
sTAaGDe$0y*9xoBqi;r8X5*gSgGy%Qmm7;MB?>Zj(SC!fk>79D;zvwe&x&QzG

literal 0
HcmV?d00001

diff --git a/model/__pycache__/ccsr_stage2.cpython-310.pyc b/model/__pycache__/ccsr_stage2.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8e796a4bc6c4d8608f22e4e2a2c7471956d4b377
GIT binary patch
literal 10484
zcma)CX^bS-U9YRFtE<oH>ACjowfAZtGxpl?fj4VAUauipPlDGbamQAvrmMQAW~!@t
zU)Ai<s&*i=;6-SX00{(zkePr0heJXFB=`_MfdnES04X1y_<#sfv{EDxiU1qp<@bNp
z*NiuTYSq8ref;15eiasrriADA7r)@<J|aoqq0IE7A@dyGNRcJUksR5Vx^h$QDoqq1
zulQ<H#lPxn-ApqxEYq91sLM99!+%5k=elOo6m2v=-z_u?-D0!YEj3Hsa<eSTGk&Ez
z)0`1$-LH0Mo3q_ov(}w!&UNRT^WBB!f~d>-i`}K>l1Ll=q3+@4;qH;<k?zsvQBjuj
zk9ChXkBhYF*Sja0Cy>rN1^>S8{muJj=_Sc2I;C5ZQ*sx!)#gd$%T5LPifeA1Ldj{A
z%s5q)RNW=y&p65zsWE$pu+va@_GP=*^MVdnpAT-XHe{avICEU)Iv2fG+|YQo6$aO=
zz~Q>@1+L9_zVEWo>W0pi@A7Kwb=@d-dzRZ>a~%f*VTkILo*jF(|0=VCs2#Gd%NiNJ
z@anMbi`4Xrdv(hVs1-L}aifp;VQZ6{7sKm8WOsWQy7YV;Q-d%_>vFHHy79}Q<NBDP
zeA%YTi(b3E6`@H(<t0DtSn-Y6vLnmucDU}|@S-^43tkY#Hq8{fR-1)g3p;9iw9X?J
zJ-ZVIw*Qj76-AyMTtpA%t!>e)s4t><-wlSFy#vaTs1&<vVFO~BtVMX7epF<h!yC;5
zxKdMgq^9Cz9M#co$)GVA*w^{&3t<qm(Dz;EwP3if)k`fMeMV0a2?OuP@J1y}-jN2`
zSlX5Mq|ZMlOR<tDiL`+M9c4!y7>Ts2CB~jS$R+Y_1~od;CjPS*r8kR%e3DPhB$ue$
zDoV4vMq=*CvUK$*`sLD^J=A;>%%I`7t;$khJSs)!ucH<HUdNkgj<hASkD<M`Ywk%;
zN&_>_V~)bMj9DIhLm3n>QzmMPmJ(~n(nb-jv(&ny5E6Ejff<yi%hRzIrs}8Hdi4{D
ziT%i<{p^-#<yWM4GOL$x8hPgSTsyYHz`x1nfGU)!j@NCJc_t27Yn^9<(CRSTVT4Rx
z0_D2C)mpcM!1W_;5(TH!%eDAs&*d3M{1v>Qw-sA!M3@oR!6h!ZpwRUk$9L0WuJ*zn
z&$+=CC>-R)i{Vy0)k$sHF;`L6C~_@w{Wh0x@Z5-HBEeg+dm~O0pax!{ZCIFth?MAq
z>zmhY)`_^X9$`bYP_7b?*SS7qLfLyqVozXm?_FG354-M45XSCW7;dh#u-BEI%OWE3
zl`Bs_^7vEdp8il;{K5-YURe>ySZTE)W`RjM?z!dOO`ZkIVp3;`2=6q2CYy2vZ$Z2T
zq>sF7<O*_CDaecRF{LCI6jSaWz6VV&565a~tE)IKjT@E~*uc2OO>4+lNEa;Y>Xz-N
zHS8(ODK<0(o}{t`fJM9!iE~3L$)e-~EJ(HVM4ti}v0zTlDLFN#<udr`;zud^Wpf68
zs({El<+r7_?BpEtRt^jcVS7tzYN?oQX5I#OG<7g>g`2|-LohKC4)17!0`DkD-=Rg^
zA(DHQT>!jv=Pn*mzuG#2g3)spi4tCt8&My<Vrg42IoNHMm6JJQxv2b6=&Wt|vvQhe
zjbt`jLYYEk#8$r2RR>2P7PMq;TTYH_Dy$J_C_R`<%1#FPm83W*Cq<EdCMn@xjh?|6
zx)_)83VTV^tI<p2a`U!q>4^-OO%y;QQ2}$YnPlSpwgS2@BnD>8bmW?Znj)CHkYu*y
zw0u)#-MHjr#T*%`PZos`c?<Q6<5o)a)<>n!%C{kYAYi@`7j8>bcU#^lZp)iWv_rLV
zIj+1dy)CyD$8d7DjCdw(w=Hi257E~|o2p~pmc9ZVndpFpuVN*@q>h<qlPq8@F#zY1
z9N>JC2V6)B?3?l8#!{dF9zx0CqzHH<DFGf$%7Dj`d0H7##}m~l)}+DFxW0QLzAsUe
zquUD7_wPx_9i`kJVdx~vPQ|C=GZ24DqAf`1S0dfKgB5k;JsRsQwYe?tHWGDjRQ7<t
z?}Lfjkq5^Tb#UA%59;wlgky-<Z$reM7~B`n5N47SOTZFJe{OJp{BUv{vu6{ms3u{~
zd{S{{cCr}dWLh>jp41cln)C&k{fsz$1*!AsRXijOP6awrC8UZ;X>cl8K;3FG3wSwg
z5y(^?FVmT!^i#<q%|19CC{A^7I$oypLU}v6|F%Rmo9R3!cUNeXD2^Y&IQgWQoJy+B
z>`o@g$B)v;$k&p&<aBa4$vd?jC@>jRv~hNKB|e>;{G2q*Kd`$Z@N|3+n4=b&wDDka
z{7FfXtOt{0m<2U5=DfO;oJc9pDKAQ!D%SL!<oMvsc+44;d~=K~Ivv8)oHM^;1ljl;
z)<o*?jI*$#ZHu{*Gr-(<28?rHElJ&3d|DdR_pb>nDUGrGuPF}ifn|ls0?T*P7DFtb
z9<MK5-O`SZR!X$+e_-*@2#X)=zfCmx*a(M6KfZsxjs%f*Xtn>!r7-kQoT$GZZq*^X
zJ7HXp!@3u={4K|=$Lnq#%GC|X=&Fb9`ps~Q)mto#qKEB~y;g_ray+u3U=A)XFZW-+
z=a^ejI$o3-i}iS7Ez->Yuc3w5g~m58!*YW(hF!;=hal}g_M(TivN}|=weH$ZRL6*;
z%vk%^*X@Y5x8~L(H|~F?$HHqKwn+_{8x56z-A%_rTRQ+M;`-T+AFkQHf0MQ<tjh+C
zPYvomR<aUr!Gx`!h3b9Scdx-@tzX9&0(~%V-Ma4v9puqrXoy(dEkACN5*a;|CY>;g
zH>v`F@}xvslE7bpn;DouG8ycp{i|)r>_ce%j>2nHF+P>yW>^E2sa_?O%8m2B-Cc9+
z3o&vLX|Af&KQ&Qo%a-K@UTj&*=V1xieslqev`7C*G=sss)q<f##4xtd>&$Jr-Zj?|
zm^`?I5BHy0uD`nO;Y@l&CE^q=zj)<Jy~n(edGSpk3Yebs!&!vBu*#!)zyDC(TXvTL
zgL=!3+^BvXC!HSJq)TeIecF4oexd$I9VUJK+yf7Ib(lufZ_;jR=!Yi7I4<9>M=?R_
zk`l-=dlvY)L#HAX`8zVt_UzQ_(Rcxd-k7qu1~!R!2}F)Ex3$>|K{&id`3PtPT=&|s
z6~H@!S;St(m^?Qk(1_2Bl9Le^sBhq2pE}ZzTs62>8OQC#>pTx?Bzw}r+<eCIB72P{
zg5gPh0&da>sT(IIe6^8b<Xy1K1ZW9tmDZs-aKjo(S}Yon1=COuv}=5bXQD08I2m;*
zk?<Pq&yJ0Cw3mqyC;^uI2n~H)3{9nC9Bz7DSe#bm^<AF#g7HS#QEF&<QR=w0T&@!8
zRPYSDD5`$S_P5*@84DR#!G&Dy+C83u_~^iT2MwY1+&xd2XB`&ya8A51&3C+rbzUQm
zPklOZ7zz-YZTq1eKl;Ql@z^<5#{$^L34DT<QXL&rWNr0^j%%$4?-|ZBOn66_;g;vX
zeh+~EiPOu-)8pC|@A7Nx7$r@(w(zJ)jxh>lund6{R8{KQo316?ZdxZd$H|Bntv0-$
zEtcYuouW>80y6~eBS3y6%MmCLxF3L5rrj8eJe0GkVE8mOd4qsBh2lDpIXM|ZKXVDR
z33Ldo6BrW^H;BgCy+@yzV2SzEL{L^4I2N1<bP<uUpF|ZGls!}f)|49HGgBE$&^Wt7
z9kRjDW7B=2wwrw6AhD;BgNR+C=8L1+;jSkfu{7ZbQ1JBJf@XPX;vJ?;HRTivwBTUj
zVk$V`j1cr(BSDVSW(b(rH{p751KcQd5YG?S;aHK)^;BFyhU6@(<=atYS?oo^lyG&(
zb$JYLbQy}kJf^|vGT?5}&p=AUOC=RqS9F<vrd*XPih;Z#*YM8ECsC%z=3PCLw$RaH
z)K-ju9NPat(-oyAD|eN<`rTTV#?$btptYhX7(qe1^j~{dyKCf4jBAQDU_>~6rmQK8
zSdo&|-%HzxKL4%gs6kKd2im)pVpT5TS4D{!EwBH_q-~ng|4YdmvUXSbzIIou(tZr|
z(0<y)9JF4V?O-46J*D)+W>qrNt+n^`?-h<#<^Cx20<UeX+*+jgMh~TlUtYi)k*gL<
zv5asFf-Q-h$P{pKRI=i?DVC$H3SA&95%v>k4QVbIIE!@3#l#&Dr=)qgF^W7K^lcex
zb2w|EgOAAlr5Dl&04;OB0BDhD7(K+(G?R!DeHn=ujs})0mWjBKj`)s}P%J1Ls~Z|N
zpul}f<1jKJO2$;QC>LkZ6Op8i+_nO@ViwMZiFgjhjk1Y`$WbmacF80i=E>x~dI~cP
zOAtXyTS&Www9OtI8@OM_6^W7K-YAMFj1VqD?y<utWoHPGeh~7Nt9`F0L?4T&Oh_>j
zTWkxUQAj1#14wg?<P&?4>K-DnLf}yX=K#0{9*Y>U5qq5Sxv}Iti2bKfT_EgJ)Q(~u
zM5qO29_&_=OJEO8=~t&tWE9qzOV9no=r#27F9`#50xsc6-Tu{A5dyi?$g-!=n*A7o
zX9y6hGdg||_<%@H#X-yUeTW1K?4;lqC?$l$d8BBA5$(#5l-x9-|EP57U!>0%pRA~i
zMrh&P_Yf`76a6;8C~o4&uIg%TrmZX?+~8`C>gtY$H{)g<gvZ+j5iQ9&#;rmVZaJbR
zQZwh8yZJrIIV|c5PVrW5TW%J}0Y+_cSPKXB7Tm^WX_!Ca%sRDOV5??%SOT|9w5T`>
z&LVowI7`kU{8vR3W)|#sbR31beDP&ogMy^6SsKk6MtW1bh)Orzn>-ge7RB^jhb?29
zyyzpUO`eJ+LU=lJp-zK+kywi2&|AhEy#+uanhiP8Jt@ilmi#$TIK%_iok*5SQCUkP
z9<(xv2B;^9##&hi4!)znHJ2<S&g~*@>ZoyHw+L=4ZIt&YsG*JX70P!I)YzjjXW}Ya
z%}^jixM`Wy?_v>n{_WX2q}_Se4anGF+(Z<O6gJF+#w`0J>Qd^eQY*4@1ad^wk9~=1
ze}w?yf&D5qDUF&48#Q79)r15noW-?Pi1?u4xNc9RX4)Q$;vurrQNKcsf1Mf^FqUHt
zSH@~6Y8WZ&ev#VTYkG}BohcIwVi&YOR)@F+ks^{JRnr_QJste1=PDGPe7~fPMLS60
z-lCm$L9Y<)BrzbnBogU0(DovN4Fe^Hgxr>Ql|5;o3Q3A^3Z+16>Z%Ck32ugz|2q1!
zSpX<~V4oLK!@g~kKZJUc_v~S+ue2;*Tv%9L*AA}X7R4BqOz}5ulrJFAj1Vy4PPot`
z1U^ffCnBt3%cNKP_q~uFCvEnJbXslEW=9XP_n3^ojLeXXL1G&qwnY27{E7sU&Janz
z0o5_cl6pX`ibxNx4n%D3M3Qr$Vty9l0A)E`I!GHCx^{^5-7M<lJ%!3QwAk3l;c7yr
z&_9*XWdtb-v_)(#$x@W2BPV&hCf*#_L*3EZglKKGUtO!O*RR)G^@lIi*YJl7eBEN#
z?S|~81@YUp>3S;PvPATO{2TT=*b`U$tAt%35Ycz&8)5>U>$outJjCXCd75S+5CkD|
zB@VgnxYs<m8@lTSVK3r22FW123ZbR=491MT&Q`b;4Hv*O!a-oaOLM^E!;M+0op@%=
zj$s(en_O9o*l(g<kTq92F%z0gZVA$0zeiQHaCU&={{Uq#(xDUQoFo$n2TwoJOr~~B
z?$@RWV7$M5Cr)-q7B4w05wSWM2Qi|hQI=w*bnY8C=UBr93oepKz~l6?NaMtD+7dg3
zlLA8|xc@|J2>zW6XaedSxojd2ds1uk6zwuQ+8_tTki8|v28@8AF@ZRAZZwuCof@lQ
ztg>%lb*Ujm$be%?Senx8#79C<fw4r@=l9j`3!SK_BEr9dR|t#_miwjwYX<yH+o?&v
zs9eQ0=!E%jfv7A$!u}jR;M9_XKB*Hrmz(EPhql|hutvv1R^p=6e_-kmQZ5v7a5`Ed
z{9trznCwk-Vt*ng5zTnn8C_x0Wr*uf*+W7qf0a6Xj=*mKU<6@AW^W*H;6*pN9Ft5A
zoC$@ozo5|#Dx}*`To(1Bl+q@Kgb+0Osnxzl8#;uJB~V#KDUiLBgV9dL`up>muzu$6
zz0nU?&SUXf!FVKIN%18RuykqHP*yqJ&@B?Z3dQS+tFCdlaAYRSs|WFajJj+=&kY5m
zjZ3^y8rr)vI+KdhZ%i%qNB6ozR#c1VW%LlskbNyQ#5a*3Aqn}9D;XgQLFN#R+8_hV
zT2EO9f-sA;U;?BKq%~rQC0si=nKY$9C<<F#O&}~CeL-@vhrx`Fab5yDs7?<5(ymH3
zxKp(zN{Joi^SIc-l?JgajR=csN1bX{qTU-iMpB&ej)EQ)l#nQ$SxrR@&@+@XzfY+N
zyInvQw~JR(O;vVWvLwd$j}7GxLhKUz7Fyk*6UE_aB4ndcqzNAoq@+eS?iu<J0JoZH
z93+CFOO(z?kg-1m;MH_+Yf1t$8dI1ZaEj>$QrKu_It1*r&XDBUpHjQRusoeeMK${~
zDwq2y=W5WZS5q~RfeNtWgG6M~M*<@56m1g097As%)J}5skUTHkW<#!n?#YTa-`BMM
z+!QfQdVhdsau`X_%#R}hF$=pER3mT5Z)hpCfJR7F!WVPoLpW7}Z4K0-!mU&iC|bCc
z!ZUL+J6TegI6ZxpJ%{nwUlI5<0G_>WBcAM%?ko4g2sXks*FxO7+k<Kg5Dl3w$hF#?
zG~7ri_Cp5L->~R<8&Y(hhRa0jHX?XCl-5)1jxhdrH0R%oInxTp<u!q4;Zbq<I{Rm8
z@(%>|3m5inl)Xg=rsFkmK|D`NONf&LD4m@2J`qk@Ys`jZF9LPaK!$N9bk?r62Lumu
znKZW`3F8fHu#rtmh;28F=<!4oEpr>@Mn0{<hYQ+NX<<?t7s(vQfs7pJ%&-K9IsDXu
z40^pyU(u+m4W)y{;rXX$JKy?SdcOB8m_qRxOaM2wqeewo)~ty(Y!5&jTTIv8LqnaL
zJ)7ZDAf{kZHf0e8Ur8_`0}n(9^C5E7VqVt^Y}l5m)lR}mNPBjgQfCQ}H_bB?R_}7{
zdB^U8356N@T~suSVxkC<G~)4QZIqoJo9k&QzFPoAX#nRA9IkM(5Y?riqnHT+To(O2
zoi@1{I@qTI^L!q^s)*tAk591>WcCO8?qenr#uzJQCQ>Hk@-Sj8u?|F`#fF4D$O$V8
z@zom#*I2%0`<?@42m{C!xA!Gl9;s~j@rS0CbD)`+2RA5xgO-ijK}VhXa}@zvMvLSc
zzOnGxA0wGMS!4yK`kR7AsoeWZO8qr~e<VPB!Tz1V_X+$GfeGtO9Q<EK1wH^9JNSyJ
zDDNw(@&iRv{s(a_#MYGelw9dQjk>NQ4p`O^FVj^xL-;q$N`cV`jk8<_vqL*FTDn*@
z`x;W@5#!UN7NI8+t{wX$Y9N9!9yKt;w=>H@AY4Fx!gspMFi;Soj+VzJ%cuZ1GkzNB
z(rFap-SlIdx#@1>h`b0!^Rv@+9~d&k*EY77nna|M)1WqaFljuygjDm?a0q<IGg;KK
z<Mi;2*W{BHMs7PdsI6EiiKH|<PUOnZ>|X~?clZGqjX9M|J-q@A5iPgEZZ`~sQ_D+;
zT;qE%#B}MqH74Y$_%4O(FAI!`V2e1Ask0WRE(|f_e2T{lMD^l(n~FrXz5#7S|BFgN
TKZ5^MOX2GqrFuyEOwIpan>@0|

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm.cpython-310.pyc b/model/__pycache__/cldm.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..decb3eb7997eaa14a7f64268e86f2c0be09a09b2
GIT binary patch
literal 10352
zcma)CS&Ss-U9YRFtE<oH>ACjo^?Hx?F%#QKoZ!vI&aT&xtS7-^$8nLURMS;6(=&DS
zepS6kt2&2g6EB2u5){seFq0r8hX;^AAOVWR6E8d=#KZT7f`nE=f(H~4L%jU{U-dQP
zO`ux!@9#eT-~aw9Dwj<O&tHDyZSS9-m!uz2X8O^Pc^+@9$dcqruIx*JY{`LQp#XWs
zS1lF)s;>n(D>p3DEnU>*t^Dxc5dVe1v`o=P^NT^rDg|Y$98|1IP_?R}Jm=Sf8EZzQ
zb-y0WTC+jJY6NrETrh9V2MgAMsLT6{!IHHk(uRL5IBp#ePFN>`lh#R5R`5>+r>)Z>
zZTijNjCBU-qFeGG3C>z)W$7i!ExVPwl3Vc>c2(<9<g0EC`I={LpF_!cl+3twl+?W?
z<S)3&n$()TPuOWGJpZ!O?RLUVuD%f7S#8O@_-W>P%yX}H+DS{}`F0fEv_qHcekb%C
z#*2NAMRpLmJ>TQ?q!V~?;&p8=Sod5P17V2zTGvTBj{hoi!gwQMfyY`ozVPa>?TgfO
z&AZX_LTbg0SG@QWe$?LK=GEv{7&}21Lswo%5^4~ISzY0^RWErtay=h2RIfW!d9|~#
z(TmZfrSgg&ZQ9B0#CBr46KrzbyWNSCm@jm~IB{sE#IrY86xi6&Mu*mU;%dj)j6%nM
z$?3&$#|f{Z2kWf&Xjash(Y)`4!_D3Y<w#VD-L<g+u}sz`JWf9<GSB0U=K(y)l3mGC
z+?=br+FcnmCIkCApS===35$H+b6*RG`&zx$*3oD5l#wv-ehP0~!Q`9LAfHJ4@`3b?
zXJjc+QYDqPF`%pLsRJXG_O;YFkOzfS-p`>%N7}@H{;Kqa@}QU&Q!_23>aL2?{JxQz
z2eK^PIEj9RtmXhUuY(yh{B~7Y3XP|v_|h%3qTd^MQ_YonGW!(TYy0Mb^sF>6lOpCQ
z?aG+t@y{!R5@yQ9mS`!ljYQfmqjjEIZz_a@J!N2q)#>tVtc9ui>9yYYOlo33@@PN%
z6<Yb4^iFQ|8cw6gysqaYb`<(|*c?!WGS%$_tt!tY5o>Spd>GlA%yAhZlUG2wo^Q9e
zoG|qKn43hw8TE24xzqJ{&J};9PT1`w_Bs(}%yn>y2QDZD9oO}}teC6asLKmp*aL-w
zym&e4B~zW$wv%ubWvw#TV$a{;@@-xiu}my@EAeh8Spw9+YqSj;a}bddeQ<r}mcuq<
zu588F5G|Ce1mrEQ51CNzyd$v>V{`Amw6YZi-bxrH-g*@6thi3%tZcHV*Iilbh1bEI
z?JX}{X#)i-SFWtRBEXY@(+O7uT2|V=8!UJ4@I1H{^SZ}~_Ra%nvMJZ_mc&~^`ovF+
zLP@SGC3#UkrBvjSV#@vF4<YO2;aDwgbrnaZal^Jl2S~TMX%Be|>5^^V=sA8?!#;vJ
z#h#|XvsAVKu!J`ziEc<0S(Lnw3#pc#_;UawE-c6erJxqHLJmJ&{3u1ge8Ip^6%cv1
z`j)gIy9L+0TL8~O-rkigEt9lX?kzBgrGt-a+#GHgvWby=ct;aNcwa&KJ}u%tQQfQT
zGT^oQ5AcZk)%FP#jGl`~RPd7Mi2LZ3NV|g1!Ev*!n$8i|#nn$jZSBf$%2}SZ(%E<k
zWeSxMXZcE?4o*NaXzARpoSxWGSS!g<dN7w(-5l~OX?ajh%Od?;TEV{>KZh}NF)rm5
z_L8Vq<Cn(e<~`ZgQyDOyDu70+0u~Z8%_YTM1*Bg}4a}I^lp7Li%HZ!(n%kAL@*R~0
zNyW{JIdW8=E($sFMbs~jTPg7uKPkN_--85#jQK%Qx+hWHJ$bvlEAJ@r9@Qq*r1qBd
zmb{_3hFiF6Br{pNU3nLHi2p#esk`Pq={r!9sSa5BeXInS)G_mHng?v82H;#;0Gv;Y
zfD37f{V-YFUJ4b!V<<VEmH|(s6~L2e74TF#Pb)*}bgH`LhBP>tH22RWkECjPa#un6
z?16;bNy;4%h8{)Px#WCu0TNJ2wFL?NDx{wGu%b=*fX2E=ZSKkYtyDc2l|3f#`*^Bu
z%7as>IymiC2hHRO!ZD=mKS9c#89b8A5N6UdOTZFJzcx6VJei)x?D-TcYDk!~nAY5x
zy*x&FG%FjNPMfKIQ~IXNz9LRvLFy8Em5)h-bD@q@1*vjc8JtTOP`8@S0$$Hrgff*U
z%XDTa{am_8vk%UPid!F?PnPMtP`;6#y(dx4PBzb@`ztg`oFpH>IK{M_o=fZQ>|QP`
zCQs4G$T!lt^n7|eExL_8=r9>nw0&`ZB{`oy`nEL8KeoRj@O1hBn4=b&wEcK``dLYl
z?8nnnm<2U5=De|#p2;ZBtu9JCD%SL)^z`7uc+44;{BVpdIvv8)oIAf~g!$xYtcf(@
z1$SXj+ZA)A7l66(3>fFgTGFPw_)%%lJiI2Xq%y|xzh^jn2$mHl3oJj#S`4v#e!RYH
zbxV6XS}F0u|AED0BP_nZ|1QzwGb0=#{lUZQbtQ<jW2^mFu0@f5=1lXAsMmz--i(rF
z5;Z$v+wZwvGuiT*(63%dc2_gnXx@o>tl4H!96#xdOtvP}m)jvT3OZ+bdAa|_L&xmJ
z*?4hgFE*2j!ALXzzlIiK7fRo}4&x2d7^WS20fMyu%!?hYl{KNA?Jdu7<0eKNWyS_T
zbIXZod+S~^_LBaWx-7cc!8WNO^WvfYZ+h8SXzPSPMbf;u=|}61@86-V3Inr6<5PpC
zkCm(>Jy@~Li(QXB>3cU}wKi{I41qpaxL(uu!cFASVQ7okLC;Uxq(nv!rAa5u;*IM7
zpgbvs`U8If?&M$v$z-!v4zIQ)vk#&5I|^@5#rRZ)n_(?fW_p!WDmO0qPO$DemwU*?
zq`B%+|J+2eE!%b{>?F3md<n*o<Hwhg$a?gjMKjpUy*6wuB8IV%-eg|e>)iBQfytvw
z_*nnB<>sqf9h^y*s6?E?^%vLHnqAh3SSPsyL;=&2emINB7lwJ<?DwB&c9y+mz_8hN
zVlQsq!bxX`HtCX?Zoks`Li2L-15H@<&8HuGtkZ;bMExf1riOlEQjFvB{brmHq%IkO
zEVGXTKlkZWgd%@O=J~FZIX)UM!PXm77T3TgF|UBgQRcOGx)BJ6Hz*$ijeuL7jl>S&
zAi*qRFJnwz7!hd9XGY1%hzry=^lnWZX++){ysMn+b(1Y#1T~ULX=83{#_hz;I!y%I
zllp{~Nh4&QoS5*{R*sQ_!LAdaC9qXmhvvc)YbjZ=Xh0TBOFh!A^&y^%d!TVL>M|m+
z8ZbdO9js$xnHYf*V98I=(5J=FR4T^dW+!kqvAj;-<Hb%m-Y7ds4b4uR`EG5GtAsig
zJj1Svs?R!p&wG)vh;bEM$ko8<@*Kp+CJcDc5L(YY^n`idWl<OB#7onB$IDpfP2%{>
ztrLf#0HOH}KXQ_%K0Hi3^E7K>0qoNRK0`~XkB%v}d)=(J)#$>3hW`vJ-W8U(-Em>U
zhaop7elH_uk85k4>#wm>lr-Vp!l@=X#weP>as<v$RV8qCJX?6)v`%i0ld)B{H{byE
zScXS-jye?y%n*2l0J)N^K%hk6EC8=fdongTDHk)rKo&E5g8-d^Ro((JCnrPbXX^yo
z1Y80hfiVGbgJ^7{`_zXgSYjP&LX>4yfn&j$Ko=1i`(-K>ls!}f_LLg1W~MT>pmDZF
z9rEGOY14h8wma6sQDV=62Qj-w%@;?t!(C4}Vrjw=py27b1<hKOiG!Fi)s$Z-(1L@7
zm#N@@b3)K_jRZMPn;~Z6-0@g!8Q?~tgRJ6k9j+ZaT+hS>WJtlb+rAUWw#{C|a+z>-
zK8w@`@y6Go2+UI&oGt_I7X1vQG`v(&lXXRx>1WDyxuzJ%8*&5hy!<H26xn>B=du<$
zT8!F?5s*XspK7|IG-Tz0@<4yk$kTWlel@gK6a^zFXqWwKKhYi-MHAziVhtD(j-M%O
z$|6>z<n?#6cB0RJD>`b>Q~Rm*lUlhhSMaN&M2uF{e>`cMrS$(&iiWH`Q2tYUpw($V
z26|{;HZcdSmu5TKM|)SP{H$4(+;nU0UH#umC+l*56n%l$HdbzJQhcL_(!?)U@W$k-
zB~l_I;(~ZfDyK5VTwIl`_+1L;XsbdO2up<h5?Vu=3kJ?3opCX72gGR$f7}=aACCIA
zA!-p;%yjSxxxaKJ3j)wG4-0@ciH6ZbJWVr+K+*3Yk-*WwQpGY68qyKsQBn#A<r8&V
z!v++%Pgy8NMo`I^3K$iVJbEISv|ZR$;8x7S*)S2(q0mu2)et-?q{cp(q{BR!+&9i)
zhG7YUNLdSMzm&B(fMWyq%UF}xA{N&wi$IJJE<*0H<0xep2#|gd@|COoPFILN7Ezgy
zVkEX$7ob(jB-LX`bB*K^dz|W?Ah1H<DFROea1A^bvu7wzykr%|lJh9`KZ5EqVV?qb
z6zd>DEh+P0w}xB+duU3(K6N6au*O_=?jJ+1p`U+E7@!ky4NvCwufB>9$hB6UeH5+P
zFA#W+0I@ow;}?MsULJ~rw&(j02^8DOK%Y`VI9x)Cb`{gE3`r?W6Z+3fSN=u%lJR;?
zWi$fG@Iw#L5<T%B0gU1%uI#Cv=H)h&C4?J1%~d_!)$r!Lyo(6=MoC0V@~&~W1X)K>
z6Cm0%_lpOTdtB6&-16PRu56Xa0Y+_kSPKXBF5E_|GR&WFXWhnKu$5IEmcT6&Eo$z9
zyNI4M?vi^9|8)_CnFafu97kcUUwxT3pdcx7X2U3lZGjxjkIXKj(jD&(FT}1*VLi`f
z%h)C_`v`24r(%mJp3Xd|)38lEHF`)GP$Z{{H~tL(3ejxKX}$x>{zLg~P&mW`)}2bW
zN`YBhBObJKsRpR02*}!b7Y@Fwz%`d_1FaEsL5jjC#BDS}ale$5T{WrfS5a2mo;jdE
zhc?dFDZhz8#{u=66+LIEr*PeJtN)6H;rWk`-zSaF^Ik{>2;(LqYox?sI<y+>HPmGk
zM?QtHd<4Qo2$1bk?Og(d8}?OdQW@iv?ADmqVBf$|YJ1RFS+54Q-6pU@K%-t-J3=VY
za6PX}DQjk<!{TIkE?Ku<rN+NTjZ2uqwTBCC6*2~Yld2!GAZuo70mBp4C`_7hzu*$V
zG-Ey5B*rE|UN4$MZK+c*Q#EBhSEJD8dlhZ07eGf3*9){SdavM(N!&p2Nu&=NV2njX
zDh5gd@wzMTD+kg*6=E2%7)pUq)m0G(6runs;+yEtW&vO^;Fx$R^9nW`a;8vE%7Q%!
zVAa~TFD^Xnz;nWzxP>uBB~!vdICBdKR3n5;*cATv2&FF&0*Hd^v~5VX{v%g1Mxo9A
zj83a<+U)2dqmp<+@W6MG81evUaT`>bYTuJzk-!f*;s@vreUOK)fyzXX2q8le)@!D?
zK}*%ILXx1&z?Frxodep8MBmS&PCihmd|OM5?E)?~WV$$@5@;3DF?7K}R~-mAB4ws}
zQce-zN=>{4yaqT--P1Pcj!RqZ*Vmg{&0EcO^U2H2b^IX%cexyRLB#IZkl=wsmt4iR
zEdm(iBC&5{XI$}b5RQRH#PC}c!85$D=_Rlc5#i_6X<CL%5md^RB;va3-R!_)(Y;O>
zbz@#&5H7+o5sHq_V9cbm*^7GdZ~;6gd<?dSiP>+`?C>0M`Bu!#h1|N6z`&GuxUwFz
zZ&4$m<AnB;+k!$EseQ3>c7*tU4;3$swAv$NM&bPFN7@Z$l>$4XF+~aEEgm|06Qj`4
ze;RN6TL5HYB#61jd5X9a0c^w2NHpBk;6{qn1P(HfG!7pJFR^neBFGSoxJX57sE@rI
z9qSt9KqkYU)EYfSyWE~OC_ugB?@EaQQ(|b6AevnmjU`H_#%dX>?6<MH%p9YGgnLVP
znz9dwp@e7yhl#2$9;!bSfKgFJ9I%ErCgWuj%l(1@+6JT3b{f*RlpCP62^ZofQd#~0
zBlQW_n8f9zPH1j!Udr6!pnKV+gCSdUQR+W7bqE<d3K2OSa}nM&y0uJZp%eR_m_#(=
zRd;k3I+Vy0t4hf6uTzHsfnNu}2qays{B49EI`JJYCnT;zcZ$CLm>L*VNY|>kb?U|$
ztxe1kA#sXRt9_j|bPOFUptG7%BCDwY)1OTK_vSTWL(M;Yqwll7$8xrY@kq{+`b;5d
z>GrRsta7^QTO@iF>e~}{X5*ma$k>)wkK+Fnb?HGf4ke_6+rUv8Hb7@|CN-tsnp*15
z?{%My%{I}?=pl=NtaPC&{ssw>mQVz^*b!0@WDYvg2DvnsF$^Rk>;%CuBpvfeYs3yq
zkf?4hODT|!=ppT^sdgZ{`hw)<kJGh+aS3(^{tNi0+Wetf6Qw#a<|SMX;aY?kmPUkC
zwx>>wQlZ}4I!035>YjogHI$I*npw?643IOFGrvoz3A0gn2$z&MGKE!jJ+eH;*O)Ek
zKEm`8`wO(XPv?oF(?r-uqi7YrL&yk?F79*mJpnF9vsg+DW0~ljksxE=0pRs)aC=Gw
zGaAz(GX;*du=~i)GqWLJ>uruKitkgq(y%<6NJTZ5`+{XH4SMy45TsTP;@?S*QjbZW
z8Hlh|91x{Y<VIIWsg~qp^1Sf64Y>(|CnMZ^Pt*EyQ&cqReTY~{z7oL*V);cRAYWk?
zgIwfo`STi)iBPAcE@3#j@-YdL6mUa>6jkAsDmXOqN`+(R=JxWCk_8-`zRI4*c<fIJ
zyamAXw;Y6{J<@yCZWO~xxarxj=7KKNT8LoEbU~rLv6)3H$q4i#2CUz<>4F^MbDoCF
z#aj*nfGMT*46#G$!v2cp{A)31R>8QuF7PZIGA`d@e@9J-JC5L){SakeCIr*T8n|Dc
zCw(R4$PtuI4*8JmCh;*AAtWw=I%y!oJOc{%)dNB=IrLh5HYKAC9I>r$=d&7o*Z?Io
zE;S~lNnzXEE&|;!`tT`3X4okdgjak9$CT`4_BsKQxZLbI4AK4sQJ~DI{sC2yXl3UK
zTqHoEn&&7I9`M`?t`odIvi|;_YIHG1j4uhg?;}U@cHrMo^iT9hB$dTqC%jeARY&kd
z2j=Es;KJfNJw=EQ>vxDGNtO|}3z8&6l8?bCCyolCU}2wr9|gQ1OcVq;Zyz~EUcKq~
z9ejp>7(W=(>wcS-Mj8*!^s%Ys9BC%z!8OL8qa{3+Bi=Ig=PIJ4j26i?eCyz|10*xA
zhD@j|s6YmPCOH0tQh!F^ZwU}}v40@&j|9F<V4~LglqKE9z5;NHmi$2^;AO*JQr=Tk
z<)@0K{11Xl2yiLyDrV)!#u;5lFs`a2Xr-$qnf?v4RuCh!E^-|#1LeYK>0;IFZKTL8
z#V0^*LQgE5CiY!wAawTzH88|CEX!dioGX5YzR!TdN0>NX9&7wj0j@{<EZn5iC}gmS
zSc^3`-EABe7ZG9W;&k2nhK%sli_;M%Mi$a4(x4P2$b!m5W!BPg2z*a7S=6%Yb^}Jf
z0yn5$ERL8ti%XO1X<axxH%@f;@fD3TmCL-O5)BbAx1%74Le`)GDhL_lTPy@q>6<Yo
zB$D_bgzGO0EQu(GIE1P56(=oh7voZfyURG^>_eBDL<X_}IZpqJN=ZL~{{~+A)ztbi
J=}Qg&e*whrZKwbM

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm.cpython-37.pyc b/model/__pycache__/cldm.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..34849c6d7d4b4df9ce6c304895a621d4f4cd3d63
GIT binary patch
literal 10149
zcma)CX^b3KUazaGt9!b8dU|?J+vBq1s}pC&aW=<pyw2q~S#V~-c9Ko+hN7mcrl))A
zXuqoV(W=&>jF%WT<MoD>ICg;<i7S8*m-qk(i3s=vNPI$i5)unvFrQWe@nN+R{Qj@{
zvSVPYRsVYT@qhpO^()0<Uc&R;56`>beMyr3l`_LeL*ga8ks?cyBRR4s`Ep(Ml{yNL
zS3I?@;;VX^pQ&dCWqMr~b=i7$@HNCY=jZEr(MI#8{6f9p7wbj8R4@6{^=VO_@yh;8
zeMY2puj0?vXZ>ou>d)2Z{Q3I4zffNgby;uGU#c&OwBarL2kHm>gY|>{q52_Fmh)Ep
z!}Y@=o%c-tNc{-XQ%=D<>OWS0OqO1koT5{@EjcB3VMnbWLw?#RBVTs&?c*ppfsz@g
zf|826g#1ZIxgynO?-OQf3eUc2cRH=0$<>#GTkAEMPkoU&E_0m=twvnac(xG+*R8<e
zy4MO^oAIfh%R<W!ov!EdO5F0@D0VxR>u<P@gMlza<x0nnTef$V*+H}!GT&vj3}3i9
zX!{y9z2aW$x&gJ~#$`8p#S0tTJbxj)5k$7%!O*3b<Cq$RL0Xr4Yu%0C3?0|Q4AYlv
zs=UzJ-0VhZQd4=!3!7GaGq&u=YWYpByEj`=9Px!#5XCml6uZ_Y3w;Ya+HBD}4_;{5
z%`mXN*X?c;wd~*mda%|;mu5wM5zTvUFxc#UP>w{U*j)n~5X)o@!sGB6QQ167T&XTQ
zQeAN}j_PQ)Wzd%l%<Fvi)i8)z=y|U5RxsGp`o)Gm7^R4B4)2fR73A8K`q^08llP@>
z{G=?!N}?oE8-qE@uG%*eX-`XxeYu}Y<h=}Pbfoh)k0F)4AicNTpGu~Ze3DDl9kriN
zvU^67-<M_S+PQZm^v$Jp`>1;djPT>qW$8;&qB+ubj{P+1w7vYk^qhndop=gk7ItKe
z`u10pegPwAqTdj04fYqY)GnfZmfANJ+VZZ_&j%}$<>_b(6ZMm8yY>-Ql*b<AbPrfv
ziPj3FW!5hO^Ha?2xOQxXfp?3|0X-;Fot9sl=9xHTjV+!HLaWJahZ(5lB~YpBS&c0_
z2wX4Xd7{>oQn?o2>bN}Ph)<yvbh@#%K|~mF9US3;;tPJuaXdFI=4vPG@SGcTLCqi~
zUJSeOL?^Xj$6Q5Ot;n^=^)|VDljnx)5eYtu-J5Zm!0GccZNtJGM36)eT;IN7vu4DV
ztq2>Ug>sdEyv6ka!^y3WCH4$9_x+dFwnE=s3&Pml2*d3)$Byl_CJVcrwJY7=5;(E3
z<pygFpkR&o<drX6UK99OYj}>o+PTHEU`<TuEEAy}1CV4*&dX)I1@SA%i;5u^lqzzI
z%8Fc83UcqjBdB+EFj7rhU&l#l+_0>`2Cglhw+8HkbiuN&b!{)LVb5Vgv6Tt%ER`()
zEa8n507Dk}2iS~`o}OqGV8mwAA4r?BlXLR7b6_=y$lFp~ONC=S^MRyEbsdyc=J~+}
zAWj&u^v9Z@rTaA7ed6HzL_Am7=K(L?|8G2^e!X!J1@xT4Tf(d2jXpw;SlSWnDND&L
zTS?}Kd7_nAapWEOeL2mu%gJnX7G(;R5qo&5uk{atp)<+cj+`9aR@miOr*wZVnRYVB
z|3p&kPbWo@zMhouRio<|Ll@&xUSW4dy&ByemFMrsRyL6VjYI(i>jUP&zL~hNqp(ln
zVq##%OjE8(s40O-i%DijPRqAd#^Y%xE9S^heX=OT!4FWsIBKOtKZuq0<vS1>+qru(
ztH;GV64l<3+oc_OTe&9_4&ImJay;{a^ntvoIEItEZN!zd<&L}qT%gqj(Q4Mo-;urz
z-InNpg<r;cfKMIsSCcH@Tw(yuCpo}{WD0OGDPZ2)@lty^&;Sph>|lcVtwTu(a3z@r
zJe<ta3Xw7s)hSk`{-OBD-qHB6L`@FuC`cdMmykO|xqZUaag?2iPsXPp&Xh!3kkGG0
zT6Gs|YRdbB*&4OEBkw()sQbgRCj_2PCu&pfuOw>!uru8^@5whmiJv5VUy$y}@BJ|Z
z=aK$Vl$vpcFqIrx0@g%bq5oLCnjFSTvdPgztwI<=JeHlA-7H2tmX`GoCuX8=>-SVP
zFDOE}r?3yvtGF!nj|T;$N=OxxQvZ0efV$sLW&wXUZ4oF`9<R|kqV%tlMVh~VBG8;l
z|3th-Cy4TYP9D1>QO&j_<~g=^hDO1OKZS9ol45c^sW`K{nP5IXOCuv+P3Dpl$$@0b
zsqR7@$)K!uZSPEcGC6iv8swkYJ0oy*cpq4#7Mj#Pog98nQY7nivVvJqBTM&`YZssj
zj->SHOfO2?DpqwWIov-v8g&LM`)Gu(lLK7MIrF<lP>P?%sz|k+bQX5C9Wh&S61W@9
zfpH#MO=3EWpOyONgR8<?PGK#A*D&LMrFeV<rZu(<9BDWeaTWtipB$|)J<FwComxc~
z{~t^)4>9@a{yRjYpBds2>1Q5Xu_HmuEwA@3UkpR<$Px41uxmm-H^bPBL$eh$ysqP#
z@s?{sb-DqWPiDAj-U_?SY_Kqjp0tOSmkG_}w8$EQI$2#^?Y;ZRF}qPZUX+@BW<0k0
zXyzYkXd#NB&-0gHgF!;WLSrvM!1kVgt%bER6N=Z`a&0FvG2$>Y((UG!9ntnSTr+ax
z-d8&;yxzh#sUdTtfoeD1bS$*B1E3-{Pc^-8!}h#ev{hjz)@XccV0u`|THJ*(YM$!2
z>`BkP4&&3jfiVR7V5qvL=LSvW(P3a_Sbo=w8>Cq1p)cu)S-epN0CWelYe@os0d8k-
zRx+90r3Y79li6p`8U}imDn=(V*a#yXl<G}VncO(%+5U!OpYI~~A9$tS@v)j)wX9aq
ziY;sP94s8$i_RmFw(mWMW-xlY4H!^F1|uumWNyQ4U3VRU!-p5}`Q8hw=GCnhPNPHg
zA@G0cwJTT54r_(172g7ifa7sLoI>ac%Q`Z9z0+oE)m;S)%!VDgk$D4WoF3P>OKOjO
zx%HlT-aKQ%EH}?S{&>rT$wU3d?IwmkJub#Ud7c@?1gT3(8msI)uydczWN1QWJ9g^1
zXuJReZbVaD1A|1o1oB3i+t}`eAQ@hzd;~NCZnZXJD}aXsqlf(n#^kvnX-0fzm>iF|
zKz#%E#>9b!<Zi(|$~bN(-r`fBL$VVs%*|(<R%CC`L@+3+Pr&muLh6c%39r^N><!w?
z7YNW27&#7DJ6x%nk`{{wWY5&phuYPi<(a4p>L!ydB@%dEQRFf@)z=sc8Lz_5Y}#1Y
z<|?rSCBU37(^M;BDk>G@@_fs;n^<nE=klpmFxogH+X=N%>Wnp9t`aI$a1VP$RK0C`
zT{oRi1wV4tw>vxoLDGbk4w^#ixksKj&pIsZ;M{m&k^^}W>%309pZaOya1<aoyXl2?
z{M569#M5WlVcN+h0&mh%D#K%otZrxEj8;4Fh~W^!RCk2AZM7WO=>g7T?8q|m;J9|B
zb?Gg3h?04@tMHOZqA?0hunYl{s!G1S?OMW>rgifCC>imhwFxh$%Thctvdq~Off)iv
z2#}A+$Vp@c0!IOOdD306$P+o0iivlrNu7W=h2j>FIX)Rezwlz2Ln)WQCV>$_af4`X
zv-8xmV=S>YH4&6H4IB$j1-gjn*gC4Xp!9)4uqG4&pP9&5g67$eQHN|W@XU0NsO=VC
zc$oOpXh6hXrRIyn+QF{J9I`a#5K#2w+=6C#Y3v22Og7>C3AEr~;Yunv;EWLYTqBW=
z(`E>Q*tc93aRb~bv=N^gti!P)o9n6gfH==tR>QNS$g;qLwBYn8gn#oEvLxK2{{)a$
z^6<LK<Y$SWfs}@qO3I2Zr@y>hQOdG`ya8`4FE7d`<oV37R+p>8cGO!okSnR0qR1;+
zRg9sbbRIsOBBwuHeDkV-vG5bUbfqHFnzATs*=3~R4XS0iBA4J3R**xlsTK6BXyhkW
zFi%yP>^&TPFo%(jp^s*z@`4H<au@-D3pUbC4f4O}p)Zj(VJ#wxjip#d)B`b(L{4Oi
zdpIhY;yV->(AI?*AD9p9b+m^l7TlIaI%PQ`Wt;;kO>PXs0S~*K9n>PIm%5#j+*^7z
zb<1gy4{~dR_?RAIKAJ^@YQBR+3||2Y6^lf0Mn?!kNhlzbjn%e><ty-l($I*Eu#Pbi
zvdP_iKF*>q0y^#7j)EY|3HS<m1X(C}lT9>)Z*qyTM|R{OPqy;4OR!F*y+IkmIB6Sc
zuaLIdhkpV;$^fcF5QQybnFGzo$jh!3MX*H31|hZ>#RJ$00wmsrc;srY)e(}5U86D~
zgh=YJ2%uI-sk??W*GRmu$EofK0_3B!GX$Oj;2KyYVrMD;H05(6!T0caP<)_B=Rv_V
zNrWXpga|<_f>Gv`5{R!?nK+|i3}Q~u0y~FJ11J8Xuq4OeBA(QzU%!fY#l>2dJ&)Gx
zvjjdzfLN3fL5L^?By!5l4cGH5YrIfcX+J}a1Y5s=6m2P@%^EqFem~T&^i}BvLuJ&v
zfp_!}UC|Q}T@8aAuIg%TW>Z;`@YV67xU~EoU4%YBS+@&y!_k}!LLWIdzc;lnIrE|}
z>ln9lJ951s0whwsIH(n25|oz)`33THZ-ce!(}NQDpSR%^*2@k&K8!o#RGm3|D<Y&a
z3%Xhyg;XwGcoXr>Dl{6!vn<%tFb0r=y^tC<RJ!fn;<?DNDA?vY>`82o7d?cQ$;q%p
zY)xk_^k2{z;()|~fp;2jbQgdkE^QfHVcZ3ELpWf?iDao18nrayK`WDJfO>+EsFii#
zqdN*bYRNLt8u1gPa&cbp8jUcuSBQ&_8khE_QC4ox>{Gl#8|5pMZz5i?Pd#Tv&smC4
z2yZL1-o&EteDB5kq?&ow4aibpJdc1EX<O*{T9sWzT}oTz3J4QLAWFpL*hf_Ra|8${
z>|4~NG{PkroDna>5P@&gaG_e#UR7%A5%3960E=sl5b-_3aovtc&1|+<6c5fL?e_E3
z_!p>g0aG~EV8OLq%F5rQ>PIZ7J~Od^!O3bAv=n?k<cDXd_ej^7#Ka_2q^T10#ZYS%
zROk-qu(G_Yj5Go$=+hd3_C&AH2&a*Nz>`Q5RB>^$h~PqBi6L5d<UM6y>Z?KoBYZ+B
zkf*vXLV2mo)7Y<~Kbr-BTE`LbLTbuy+T;O2b&!%^Gzl*^EKgiBSiWlq*Kzk@3`@o$
zKx~sQAY_bSE@4u*%R_X&Oj{?)m7p#QP@27?ucpjF8~tB8tv6`1^pMp@JRs<v4r@U7
zpusjYNuvFVl4S16&=Yze5kHiwh~^lGS5f%>a|v#%67_447AP}tO(3;1K$a2fd*B>L
z-dCu+t;I$=hsy++Libbx9YPv~E*0o{p&(NEL`OUZk*Fk(H;2~%PpP}wCfz4#>%Gc`
zxn<rk8|IVe%?*5zfgx1%T|Z>EEJ$bHrVFL1h9#m5<jt^O!=AX}T_el_d5EFcN`gyx
zuIa`w-w<Kv)05N-=^^NnD{;tm$GzTyGoib!AnZgu#~?<8&mfc-pTU@MtJw{^(O>~Q
zBg}vHZA{F5oo0uVhl{Q$aqc{`VaKrj<ZZ5OL~M^53H>EBo7@m|!G4R1Y2}|L_uoe8
z>m!||9K?wa4Lj0Zd60jvIza)WO^%M7EEhWRFXD|z1dvG)BUBodC`?L6+}8SfoY~WH
z%<>XA0LPd`8b^;~H`p7APKHJT)qvJn+{0x0M(_sGFgFJMs4e=5R+(L`pL294ds~Wg
zFdC$Gp3Dz~p>xBLMCo88VM-Y5>^HHt)btXl$S}gugx)6>5`qjoC8|Drpnen~QBh3v
zeg$tt#z_;)ey<Fg23ylcs?s-=Yuc5Gy8_(yDXVALpP(_kR}zd5@#FXN=TcAA@0@q&
zILM~drQYKcs85+s2*b&+h;VSx4HWTZbYg!<;B7IJXv?Ra;icn10FTWjA+~>!I((hL
zF9Fo@DPd)AB23VVZgDv#=^Qu{#Pw%1x<Q3>e~BxfPLvXw5Them%BLol{LYBTV4WKZ
z-OS7@@FQT$&y!&^|LC1QU}O{4g-9(UK_Zp3WCEcouH2M$PWNkzM501FyW*m16b>BP
z)AIU5$mc86r3(c(5Qa9c;D%|a_}1`1B&AoISmyuT>;73RvO&Z`4;chxgp1|>IT9oo
zA@6YwBg7zR8w8~FGf5`p5r{yT0)j_K9A=T$hzFJ=q%vtrf$&2QX-`eGec90$Bqw`-
zg8#-j7|{6U@TJ=91GRaS3SK-X#Tu?Jh(~F}f>Yd8Cq^kz@3xMS6lZ!@!FXkqkcOIB
zPh|=alL{90F{Q>FMxh(r3SLX~(zN4}aWT3#tSK#WngqB{XNhCeL<mNsND^X*DRI%Q
zc!n<PaT}S&H6mEPg1|{`vEQLtD(T?Xgv@0$rm#fd<O-9HOt^eH1kAYR&}jH0YF8MP
zrxU5D=5jCPRSoL%S}OB1koR`{Fy-Xw4qt?jBA=+_AO@7+z>@APz)YBjG^N*sGYVHy
z>CH{h&$#c1Oh}j#sR%N80|^LI5^QiJ+wxa5A`qcDNi)JKbmV0TViRyvgXUA=a4I-1
zayW(8=45uW5Q;e*nZC|m#CYrzfF@o%-+Pf~Z`cSCyQJo(J7EM<;JRzUT=P57VgbS$
zlLfiPW;2Z=5^lYa0p~X@x{rp$oX3DTtSuX%xu2!9o}zUiPuSn0jQt%|433g<c|%}X
z_*PuL!TyPw{5^pO<p}!+l>HQ;w}gZQ-@gQ;l1HC^5TWA(9tmp_86&Y#M+S(K`ZBC9
zAaGCJC*+c+jth>2Y%*{{TW@F68vKg@8fR2$j7#HOJKvrHvSHKVUjwP#rchX1@ff^I
z(u3?v1W3B_e8*;py~l{er1tWkQWc3&cAUUT0wk7shN9Cx&%Er|{yRhC?q8@z7h^=Y
zvhRF~49S6k%R&)2(I3%F7JYq$F^aC>oyV^%_YO~x;G_DDh>>I$akwBxLZWaT2I3Qe
z*;6QDf`xLz3PFhR=0iuvr?1;y3pf3U&4Lx(&bMe$q}(7&mnRnTP&0AG!2Sl!;DkI+
z^Nd|x>djR|C>bq~YxpyP$NqrYyhVU)q%<OMfKozG{3WISn!rC2AbMi|M&REGe1pJ<
z<c2nq@CNRoYK4|;B4KEG5rsljNUJOuIy`}Cqf*v%xlk!j897F+Yp1vlmVnw|v{11|
zMjkQwllTWsgHRC(uZR5}HK1c>#F5+(e+gL)0^uF;BlNEasBi>iqt%fP9~R(Z!b@XC
z6vSm&0tSsTH`#3zfEKY=ermGrQ$vRM6N%jtMnf8pB5OX4*A^*7+*e;341vF=j2E@)
zI31sn(a#O47mFhyoJN2vRB`gbxp97j|1Hrt6S>sYDbNtnY9sXhFkn?0poHKo{>Fr`
zCH>us35g^AYQgn41(rleLL9>QG}9EFC9a%O+?^+y7JnZ>fD$s<0Hp7tUXTw$YSW)b
N@Kejnm9I+P{{SnKYY+ec

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder.cpython-310.pyc b/model/__pycache__/cldm_decoder.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..29ee333c16ff6d1dd5efe18de4bf2b5a54368ef6
GIT binary patch
literal 10502
zcma)CS&Ss-U9YRFtE<oHx%cd~_i7(Az7i9>S;zKz9h~(f@~-2!NK~rns_yBTI(omV
z-lM6ulguXGfF|Jz1aV{rB7~7ZARag*^1>?-;)RFr2`NZurAY9A2r<OV@BdYIPtSN0
zs8;>^yO00(zrTvgWmCfQk3Z~si!Vyj_bD^|Xvlm5Z>-3Y<VvpWOMz_3fnuQmdBs;P
z75}QQ1vx7>D$^}p)a9-G=-&|kg}}5-(MI!&LCGowWvd)itV&R|s-isS*Mb>qMx=GW
z9?V*^!JIW0%v<xpg0&DVT8pAC?=J<**0M+&{*mCQbu>6;9Se?I$3<DeKM|a?PKvbY
zH-b~vDWr>T$$uz#*m_u&UXk3gTe%~-6>o7{wN4{nb!*7iJah95O3tEW#;v2I?kyvK
z&Q-2T&Dnc|ou<O`uQ|P5H|%irh4A)TQ|84_GuLCDd#T$>ni|iyqVR?tx?K0Wq31AO
z9C$3UgUIds9<L|ez>5>FXM4fA=eigOL)5SKoTTgc*O(K=?T7^)Yv%akwNcxbsp(bk
zdfyAF6*pe@;urm>wZ+X#(akV+f*yvhypSZ+APTd(!W(N|@>=A2K4z$1aj5cAx83f?
zXwp=9#g966ax1Z&*zN`$u6wt-aT4>zZWt#H&6Id{n?-?*9ksi(&SRIlPA3W-{}rbn
z$6Y79gdVKB-lthnUq<tR7mhZ250oQODR$Sw2E;O1i|{!8sK|T*Z@d8DNtWzNmg44I
z)z$9EpfMTP*ZJ&AQJApE_dWNGaI~+r%Pk#!#!ndu1MjEs#uZH7k%sw1+LiaDZ#*qa
ziIOU*w21*-Wk($vskE!5#-2PZr1EYKH9FEJ{_~flFO-MHw3wP{Ayv0kl;(Gh)ZCM0
z>H2Z>D`Yi$sCg62py9Wz%2H@NDa99Vq80t#!kcQY)R)<(&|cd$_oQc~p_vphM`>He
zERTL(8I~|pF1AEViM11Hvy9exYTZ!?2|LQr46D=S*;tEH_0wy;{+ZOoe&q3fb}O{<
ztI~V9waZP}Dl)I<If)&G{%tl7RN;hFw;MF8JeNeQwZZdYWOtb3GD0S=fO0+GZf!VW
z==m`>iGnlg<yvyP=kc5?{z~1j*H7$qBFvcU;1Um9Pzt)P>w8%-S9?*97rd|!3I}=d
za@0?zI;kxu;VR0SWv<1Z-{$fyUKq1XEO;yNZY5a))WB=B4I6V1krI7yee0&fIx$x^
zVr+;O%2fjL2G>VSD0knJ*hjFr_dmJ15e43A7$x3%6m6}#PU5U~Sk&*WUhRihz@4oP
zFI;T_1*^n6FaF%?s{$XZE#D1n*P|M?(!0&`U|US=E)(&c1<+(uuHh|-w}kYu9~p&`
zTvtl+l6*p`$R)*;2S*=3*DIs3n%deL&P?NmZHEr9ZgbNfF&EM$+rHj+{H%sOi#f%n
zrocz3Y!P4yZ%iWHkSek$`4AgYEj{t)0LE-skPAvdEog-te!BQkihlWmfuAZM@^1BA
zsV%z&*Su2z(?a0hkt{6}wN~z3aEPUYk!#!>Z5YCdk$`wl6I6IlLHZso;vSLRHTH49
z%lGc%5%p`WV<;Fu=aH!3CD{=V&?}L)1*3!IW?40zC$@{LuR?Ea%U_nWJZq-2@iNL3
zDkIkNl|UUHgJ{sw`E5BpwxzIUlB4u+KCQYr<X6-3u$q=d`nj}%e>Hv%W9VXB$}8*@
zQLo0YOv=r>vaP2wU_Mm<jZ_6JBxagRirWfkzmyu7G1rmjB-E6_;H5OTEobFhDhrZ|
zn-_EBs6Jg1g5+(~FHKr0@!PLTUzYDe1VPCBaZ<V~QQcj6v%D>DDe(@~Ce@_&uJo?l
zR$Rj^+%b}wtlhS}4Lro(7H#UTc~|;v=*d(EEd4H40!-?dc{a@h&ZP$6d|CipNQ;1r
zX^DL|S=w9<6~H4XIhvLMkEIpB<7pM}M7lsLL+WIzy5%`(csyzBo=P4{)%5tbg7m|C
z5^~2Ww?`N{jj}Vz+2kBVppt5f68cq0Kkr~g9eIz&I!|rx%Dc@}-5ZxZBJlfYs&?ez
ziBuh)bgRQg@)+S5V)maPW={<tN@fT%>8WL438h~hKAb$Bp2Y0=6f2sOFlRBXxidR?
zjB+|F8=g!XseVKHWtn|RoW6q81@tN(k%nhN9jOXZ<+L(9lP;oeEu96tlC=nBDo<AE
z%uxEdbctpko(&bZK0KSO(0QS}oj!b5qMEI2p3}RlG)kN#Phgy4T29ZTb$50r7Z#Hz
zX=LQ*()sjkdNeJ%b30ICGN@?t{O)RUHa-1yX_S9tcU9o&<Q_0bEi`HK(e&iAk|NoU
zrYA59YGllLeK|dqQJz~}lD1T=>HF!);kn6}Gbs7)1Y2}EgsFLVVaEvb$x~Ppsl;>c
z;*Pd0=1R{2bCVe`&Y`uW4R`5d(y(!GO;|}~g5`h9aQFZ$D@+zxzMr)iVfpN2ec9@k
zcXYH;;>G_1i$}&-{BZwmqRFSnI7IrHgX?uAh_oYXgV!%dk$>t`<E^OQfb8x>Nh66G
z-LU2NU9XXBcnv66FC@dO5w#n)qdsf2SQN*PJ7b%z0sZB6$&P~3Sy@>by!F5_`*Aj2
zoEeObWZz<>ng3ry3$Y8WZ(f1*25Ai2j(q}xbnx`cU96Qgpq}jw&vD}hMjU4*7C>Xe
ziD`T5UL*FB!8<(`-RNSQ)R1}c$N*?~*;r`ngg`~oIN$N3b;tK_(^iFr*`)EQLBq#N
zR+B!=*v9#u#~$~+8!%fNH!+4lA52`Y;d@~Rd2|>VBX-dDlNKqF@k43S3A1?PIshn7
zN~L89`~|p`gBc`~(Ox;Y+NR8&LF@MvK1UUkQyFcBk-o|FDydX%T=1P>-E}S|$i<|&
z>eArMzG7Rk?QYmjY<uMbtRcsbFCvlk7(9z+Fq-=<7+XXP6Dz&Jyq4F!;kg2nhnMj2
z!E-B(Ya3mhNsp*RoWhltuU>8RST|zb<TelmOz-!@Swz0D%;Uyj@K~d};;jINjg}L8
zapNXVIy<!eE}8B2^W85rE;gQMz^rdP^~fXL2FxSsx8H7R=*RYpaa_LNh!cX;B_og(
zb`kiwN2ekb`Fk?Y_ngf0(Rc~Q-h{Ha1~!R#1w@WAueH^SKsbDk@-ff|xY2DVb_fp%
zW)ZuBF?nH3pfR5rC-+BOq`skdbLvPVa@XKq<y^0qZ15tek!(sEb6YcRH+I%(A{d|4
zC$vl&A#>%#gx8um_9|_L+ze5|UZZtrE?lvuk`;>vWWhAmL+zR$;kmdE8YiPJBNA&4
zHfYDeI@&A52$TRzzCh!j5JOX`7>AqP!0BLl-GRr8-Egu|M%EE(<IH($d0Zvbso)v*
zlBoKu<M+Ln8H*TK!G&B6oF30Xd~{&JgND$0{(&dV^Dc{eI4543<~v@-I&ToiXMUYH
z3<U_yxBbXTp8V)2@$^&dB<*C4z^7>`_3<&qcE6YPw&r^9py5EnjCX}8Zg*YS@L|Z!
zeaDxPx5u@s-79af<CHYv-omRUImRfQ!Eyu|R8<L_EzcINH?5PKlVog_?KV81KFjdP
zPE)5Mff)j)2#_zy3Is|79s=OCX;;Q3FXenD7~Z5NWHyUau*w@i=KjeL`k6yuoj{9#
zOJG7kmO(Vu?mhX@eJrsJYC@D{Re@u{nLrm28T$;C3d$a-0eeafSTj=@cqpK8_BwUQ
zha;~|_leqWTZ@N@J&PX1>}6`cG_D=(dY>bf_c;O-JUzFdS*x<|A!bZ9<roUI;9%iq
zDmdVr5cFIlL5|aA2%0#zJr-LAxKZdJt2kPRYsU`PGjRbKQn2lo@5HffvrAYm6Ryr1
zNIi`=z5+#Hp3vZQ8F07gXCS5FrIMPgE4oZSQ?AQ3#X#PW=kPAbr%|TJ=6yYvwb0RG
z+*XW$9NPa_(-mb-R_-hJ_4{*q8c)NohSrLrU<3v2vVZMI+I^#FVq8<K0VBfkGi6O#
z!ito<{(ja@^!aZ^M-6&vKh}O!E7#=;esz?H(Te(i?zhcS`hO`!L)Pvq|DoO2>a-sN
zJ+v>Hn1j|!vmNfEy{}Y$(yU5uy0!Mc{)5u-x;z+%U*NS(lv|4w-}s?4@e5IVOu{IU
z5*Z;E1YA-%l_}`rs$|7)Q#?mo6S_cHBJ7iB4QVbIIFEG3#l#&Dr=)qgF^)bQ_H6^y
z7I4-=2OpCM%P(b709xij0nj4RFn)-qX(pN@{uV$2M*~Y0%S3EQM~p{FDISzh)J+W=
zP~bjgu^1UqC1WaLR7mpZiD=ShVOxP)F$-tIL`;WbNBLAk^r(;;yJV7%@?>&fKZ6-Y
zC5R$rEu`I2)@Bcm4csr|s>H}~Z<a+QMhF)n_t+7XG72&X{UGEkR|nmm5PdA9ic<tg
zY_SNSS;{0;6KSrId}5DK-J=9n2s}aHNdT^a$71#r<%yT9!bEZ&#y*7{N`!ri+)=E9
z2(_dvfZgWg3fMza2KA{E8HY9Ivva?IUL!yMvM@mR!DT#|+rM@VA&|?>Jo_26W*;N)
zvjm9M86Cd}eDLx}9JD;&he)8{P6lpKN(hJNkfL41Wa$}_QkW+6pO&uti}a51W=&-@
z0*UVf5781m@gD+=<0h``sh;NL+R8G*4W8zzp6+URb6(y>h`e1A(UQDt+$llUQPc#8
z_RQVlp5z`CbtSiar?4$sC31jKTOQTILA?XF(W;E{$J|+W?he?>s*XzFmWdWMchOx!
z&lz{wJ%azbh{DW*{f<wfFjp?U#^<0QDRgGTD2Hu<9L$f*E~3&c?=~;Qu1#@0&t;Ee
zo4o8JvQ3_fEy8#@^Po<{1JPIm?dYxIjo${K5Y48X>Rl<#e@Fg0C>-Je>rN$GrO2$U
z5f9qAR0GsgL}cx}3kTm-;F?RekrZ|jJ9X8hv|9$ZRW_@86x7fr`5NUr2x{!nm@`Qo
zt!5~YA>6dw+P`8Ec>ewQdu9Cbycde-9XAn0BZUn!p*hR0p)R8?@*jkiBakD)e(XzB
z`(*-z2ewU3Dii#XjT*Bq)oc<#U<&-yiVz<(T+i!?)J(g};$(#EY}9wD@m*?M!dR|7
zS{a)|QPap+_lwl#LDOp#?#!525WArLi8{nhh!m3)shj3V>FMB4y-=g*<OdaPBHBR;
z4;JmT3wniUe+&u8E{R0?9B6w9!G@udKtgWIyULz4RD~o(IE7N6HFZsd@&q?S%6}dG
z*(?B*K1@Ac$_)FqL;euzN!~M>#HzJyUtC()f#-xbaFb$;OQ!gnHp&+fY(^NEa3@^o
zF#?~b%@YyUv1QV$gNI(qXq`6u6FRN6XtU#o*n2|8-$G_Y#vrjx5L>GKzWllblFkuH
ze;%r1m?!msS{0EVTpx(o+^MwSLdARq;s9j@Ts%mdIl6jC^xZt_<UNJTH?_prEZ}-V
zrtm+N(1ipk3baLRF3nSvrX#0Cye8fP*hAgX+JtCrZBSotY&32*T8+mqHrDZn41C>k
z-~|!8ZA1J94qZ_dTegTkkblE|6?@`}f1R)k1R??teM3yZ3mq?kfrt1!uTIk}1cD$$
zt|SrHUGGL0?uPDl!>AWq1qR6=yb7VE_zcEOx}AR1k4FpOIpH9%U!ytT@!{4i(@s3M
z?j$e_<t?tP$Lt%Z7i7(qPQrxdl3Ri_*l$o3Eu0;qcsih$>ClODPLm0QgQp*9CQ~~h
z59X!_V6wjhCr)+<83*L6#55Bb2MMC3ah_tObncrt=S0IL3oesLz~l7tNaMtD+7dg1
zlLAAexC2FN2>zWMXaedSxqK>*dQxlj6zy_5+OPn{kiR1(28@7_F@ZRAVLX;7of@lY
ztg&xmb(tYW$be%?Senx8#79C<fw4qYdZ7M5=tM;o5&l)YLSS^T+%FigX29RHojK`O
zl<T+(-Df^rA}T9Su<xJ;oLUlo`*lL+a`Qsw&<4GW>vSw+B`!&WN2U%T<3b?^r=ump
z4@S49$=*UI_H8kVXvVAV_!^TgLp*=V9uiXdYt-Qv2>c=dMj)YL<!>Qy(2Z|%IU$)G
zy89Hy-lGNv71E6;E{u9{Mrr$ogb*~vsnxzg8zM`r0xGL1C9-!4Fxtsj|6oBA*3ZI&
zH~JyVc_LnG7>~p&DZUf}mM-p^${NQlowY>tDip6LuDd4T!m*hwuN}t!3F^{^o*M~9
z2N!wcG_-eTd?qzz(41Q8Pw#c_DJ-%@^fG>kWyro38sfW1kdTD@$CZo_g&=c?Ms1jb
zWvyo{13{QaS}+092GSZa#4@fO++3DYAQXiyuBH%{uD&R_`J-S)*SH{o9aOh~e`!~x
zTimHy6Q#tC@&#P%;7Ws7mPUj{wWCh8t5EMv9V01jbw@#u8cIl%&a7o32Iv{dnct+;
zKD$w92RDq@Gfh>^?g1xvkWJ+tLhKUz3$(gNCki2~iI9y(ktTdZkdYeQy65O)0NiY5
zagZ2>E>SuoLB@Uyz^Z41+fx#l(U=yQ9dL?;%|<qwnGFFuturEdwoC0wqw;Jb71ity
zP|oFn;9N_CR=qBysg;8YaFWABWYUKMBJLD#5y4PQN(Z%*Ts<N$2)Egg8=!l#qRkI9
zZ7@GYO#8h*L^B19BxvT7NI=ZOt_9V|oAT$ij9Ne=q$=Tyx$+U5D#5k}YEj`<Dk&5#
z+)Ck@xw)M@DNLN6zQ&%%c<fIJd<DSD-*ga9_MrQ$YA=dmBi!(8#I1u~VpT&#L#7J~
zt#&61Hxi2dhynGtY`Wrx6kWh@II|505xfDV^$fdXjQ=Ig`BzjiI#I^ub%AH$QE~Yu
zBmQ82P2iw#VSj_Nw+X>?yaq0a7f5LdadHTy`zL)sgp<~qu;H&G1Jp@F8OE8=S-aXE
z5IoA|vfQF1j5n~sW<Dz+w%s)1XHreHENq&a#jFM&FKAPxrTx;ROy)QaWb8oaMkP4R
z(dQOq(Ccmbl15!?Djh5i&p$uk`QG2t^MmKX6pGJa0(glNH*3POW}ihHwgVuJEurh~
zk)h7bp2Kh{kRVuu0A+R$4__%ufCu83mk~K?v2M@}9oUwc)&55`B1wC8hQK)jB*S@*
z!s-FnUT~cNOh_7o{T+Z+78AvzyM6>JX<lTf$L4xgif<QsP%Z#Z9Xed$WWlYXpre=x
z0bCY;jZV9cS#+>Z4d(d*esvMU8JwJAA#nPK`W|2=6=RUF63j%(gj^o>A`)U9h(e1E
z?V^Aegq4N(>aD|Ttln_^E<S<4^#bII*ZU?dk5o4N_#;!xIn+$dgBz4TN6Y5iVMm?%
za}@zvMvLSczO(Sz?;)8vS!4xf`uixQgxvcRO8psuza>C?!M;b}9|*inV4rnvQI?c5
zdj|j?flVBI#Z;6J6jk}LqACA_xE5k-%KJ*8@<XGc>xctZb;Qeb70wX;4YO7fBQ(!*
z9n22x$Y|+e)$Es$B99oKAGHWQv2g9!?@$9GALda5LwrB85{AMBv`*m@krfyyh)~BX
z6O(0JfSVaVi*!*`n-TAdFpo7q-E9()7r|)j{B+%ihK%sVjnmCcB2vj&P@6oMES_CK
z%6fV<1it0jU(|~0_5%C*NUpT;;abuUy%9G(^ASc)yk$^Nv1}4hS+Jal*E)A_RXFR>
zCt)<hR4()TN;E{g(u#s03WbYnRS?I<7h;I;(wA&ZNLlff3fEr~_!FTQaXeEeFK9rR
pWyXaJs~3p}#Md`9i7bBux{3aml#+f7|C!Fhw>V1ui1f~!|G$ATp8fy;

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder.cpython-37.pyc b/model/__pycache__/cldm_decoder.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..2e0f1c5a6f1a38df7102e5148a1bec395e9204d7
GIT binary patch
literal 10306
zcma)C+ix7#d7s<P&Mx;tF7K2?jbz={u_;?l?4(iT*p?Nim6=qM<78@gGTa$1hn&mm
z%&a65vvCqCvD$`?>NW+6BnA1<0O<_{nigmuQXuF<QMCWSc?!@c0evXYq7O!Z_V=Bc
zU9M=i%`E2IbH9A&yPtWvUav}cKKSATZ~kRT`d7*f9}S7m;!PA;l3dA^11Xd(IaDkZ
zAg=_frQ%-=w6J8AiZb2OMP1n{7ypL%uY^^rD%xnlL|C(GVcn{S4XY70t)?h11(V^F
zH6_w|&<dxm>2SuH31_X@aL$?w=dF2BR}L1!MQc%{jbJG}W*rNUTgSr_)(KHo36{f?
z)=81B24;B5I)(IvTMJHyPgqaL(&r?%?l$gAZo`}3Rjo6~H{D6(C%x+CS(KbZ$&}kd
zNy}SA{=BQ)l-kpe2s3SkmtS{!JwNJl^`&TgwJr0BFEH0*o_p2rq-~9tJ8^W!j$Ez>
ze&jifPwaRsw!_%%2Oe*we&{8s*R#EF&2wE0gdtitdrs;*!7b)Q$$HE}kF`sD{#McU
zRcd<EyWRI9YQ>EkUh;AfceZ%-YJ4|JoUn(X8!x3PHHf3UuJXpJm%bjmUVs^z*Bz?7
z>aVZ&6Etb7yb;7*JH40MPGb9Em+Rg=KS>il??*}M&`hakud_I`v7>dL)_MG@?{wqH
z2|nlalf-wTtLVY}wLZ;?`Z}8Lcu}$0N1z;uO0l~RHXxSCI)umJGorFNlz5UQyOO22
zC0BK|`!eWD2Ih4>{Yo6AEDi$CeIqLNw0f<h7o*hiU%~rpcm=t3r9nBB_T>ZVyI+;1
zRLPV~+QeY4vZoG=Oxo8n<3Ju%GI_s*8Xf8Cy)#IauS#z(4JNXQteRCabypo!v+}-?
zRS#rYx_$Xg34JSh-2v*}1S9;qbVK^GlxeQCRbk&oowi>+kY125qMJ@&%-XJuQNQ$-
zGN@tXQt~6w)?hzPrOi6pm#KYMp)KzzgKD%qUY?IOKUP1!w%hMxMOExU&i8=THE69s
zT50tfFh9Y(p68@?90l8K7U)5l>iS{3$xCU>Ivczk#deoDE;CTe8=z7zusa)06nR0y
zt3<6irE)FZ?s>fAiocp4_4=v3Mnsr!9US3-;%lMrx`CG$bF~-uc*TqQpk|N~ugCp#
ztdrVtQm&$`UFTZj1?ybC$16kjNCY3H-n}$W;Pm+<ZNtVKM36)eT;ICuux`SYjRYH_
zg>sdEyutN?;bi~E68i);_rYryH{#H{7{#f#7ROr`T_<%ec3Ir-UA)<ku7eXh8(ws=
z0~A~&K6&}`H!cc%T<ipHXuBTOu$A66FM~xfvAakFcLqR`HMuHJ;;o5aLtan}xu(n@
zx1cP`lS)nAIrb4$yi$zR)>c<>Vj4GWJ92<?n^*0EjgYR{_U*nC<TdOCOenTA2A-#~
zd4L7Hi2`8ABL5h>5ed+ftN@JIt@*CBF1r=CdcOjegQ&bOSz0b6t<t-aCRsWtYLZur
z4M40gV(lMmf~FqPY>$YO9}yAVVxIxL_UOOyi2BveaTL&V0dE7ZiZ^*5JyK~`u&FF%
z(`-4LB_>LiQ^l2c<!{P)p54f%lS?R5sEpVoP(y8S9E@GcW_RW6_?E(Mq&lSsvsu$E
zA^)3MebCJ6B7G-o;9pJdU<_T1OL>L8C+gMYy-|7ffozvE8PLcSK!^aqD%iM`)^-*4
zv$UQWn6cEAXC%}#z^wJGv@7T3TPove(=CfRN>raM2(j=B)Gv%$DakKV<(u*Y2#>AG
zLz!7={eeWa59H0puDqo@lnDpll+(#{>Rstwd0lZ0w{qV|TY1Y}c^9}qtE-~bv|D{3
zeIGh5(*bKg#Cm{F9rMp*Wx(0Y0G!Jzfb-b|;6hfzy!X?^&80{KJchF48RoZ7WDUUO
ztO<BBo1+yXWoD{dpOFS9(o_4V(<d@DJF%-Eeda(y?gZry2vcWKb}l`iUVvCrGHqT$
zzXoa7J*=rK9}s5S)aHS_|74~f49h+y@cdM!cIClxrVdWJ&4Kw)zW1~AX~Oqa>7o4g
zPar@~4NjxfOk0Gh?9?K#ChBT~C(@PdBvw++PG{;2gd)V`q&u}&#)xO~vcbvB%=9h&
zp~~h2MJNvy_8s)9FG+*5Q4OgEQuVAcIGfF*?k}@xz(31dL<*Is7wH^P`VZLx%|AF7
zX>Mz9F1<)6i1MFjPdt#QW=j(DoY{YdM!|_ci*Y8hdUiHzxzl^4XfC}(BO^bP&1UDa
zW7&i|vj?>#gR(Z;`_H83vor5WMgC*^&j_5IJOCD{g(hu2m7RP+QY8DSY#FnlMwT8b
zx35AaoXY9ZZ7xV#DpqwNJ2^N%8g&XQdw+zl^98PE-MKv@YNXF$Rixt1yYqY6u9z)5
z58RFBz&J-%lbP<qr=)@T_^Pm$3s{TbHO%;5IUYX((;8a>jx?N#IEw<)=SM5d&vJ23
zr&h_e{|A#xLri|S|1Qz!$A@@C`V)_@*p(pWmR5IeT#Mu2)G6~Tao>b|?#8K^#-<;2
zg1+mS>4s-Qg?bU0QD(euZpVFQc37MwPdh^k%!Gz=eX@$6R#sM4cE0kFWA>ALyd*dO
z%=FOyqnUrHp@k@hUaww<Ed~h<OO1UN0(R%QSADFNnNYy?hUd76i4ljHk&ZVvoP@Tw
z=9!6??!41u@f{!Aq=w8(3Keg9`B-S{L_kGqUg!q#niB-uv{hj*wrPB7U<O#p#k3Ej
z)V$F1*wcY`2S%uQ7h?$Y!C>{wz>B)bqeEeC*kM0NJEU0Xp?}g7(|D5>0O$_p*P;ae
z0^BO$tYk938;`HFEwhiKH4OC`su-O}u@Oc(DA${$GP!X%aKbg$xsoFHA9$smvxjPK
z#kT#(Pi=eUGAtb@NUk7}x8HdI&0zfYJ20e(3`W+r$-Iu|-|<|5!=nrM%+9A*%v&2i
zPNPTkA@G0w)tfiX9`j@7r`tdgaD3Pgrw|9i!cNScou^EH#ajW4%#M?IiFp@ioFCU=
zm)tJ<4gYQPiusHQ^W41j<deP$(}(&Uwi_Gzsl#F%R1la+N|3tbq_M)T06UN9Oopap
zx##2_jK*s)<VG~bH84oR8z66#d7Z6Z43gn9luv+0zzu&rwIg^sFnZXpVoY8cl4ioE
zhRMSb=c#Yx-5op7nA|S7NhR0or5k(#bV&B1jk&EU*H4@^nh1s^^@*%1jgY%!V!~VP
z5_^p{^LYZa1V#=7)()4dt>nd`0ogNc^+>z+^SqSwLEU7s<wODxEJ-{@r}`>mF|%f1
zZ+0E5Ykh@Sf)ZfPmuRYGF%^}Hae37boi3K^?|6K|k4773WILfY$(^!}$5le53hrTW
z*-`Z+C+K@(KCXfvxf(h>UV<R$!deGSq4n%Xo;WYNEbiglcx{{mc^&J#L%g5+YvOPe
zAh^69#7_F`^F`vhOY9`=<T`=ZX(_GYu_boDm-n`2dhm?l7{g?Dh1qTUF6{Lv;?+Zk
zmXRmNwVVF+H`obER^hh7TPBIdC`7?Z1Wc-GgwB>{3zwSK$*ZGeV%6<+csqTT<B^eN
z&L#*<5jaJFd__jiBC8QN4ZtVI-4>fXk_)+*_!2dFlK`E9Ro?(I4^M{B&%R2)A+ScE
zLtsQumO(VP-h1}>LoBf_H6co~n!vH(RG^E9j;&Iup!7l^*kcO8ni|W%8v)I;FH(nc
zRCs85K-9Kv%^xNHJSvc|SE%{Iu(sIsA%`p;atJ7Td~QLrR^!kc%$aP=2^46-!NR3f
zaKI%Y^0`JL9jDC@3~{zSmRJV3Q)nY=qF9G(Cl1$h@d0sOvF%RaB#CW<2Wi3iQ3(I$
zCbG}rP5u)=TCT$Dnj}9<{0yWtyi_u&=yLw6$}MG5Hjp>qjaB6Z`J6mg8rJIa%&;Bx
zmJQ?@s-`INvNj{e&`>%DA5M|;pDzBZs)4ca6TNh$CDNL*AZz6%q~Q&!lX6RLz$a`W
zhh7uQ=-JZ9Pi$eH8D+fpaP(phBOgN_%}V7p6+YxJ5&{=&q?<YlVFM5S6KNCH5~A2t
zN@YYp5c|mFOs4pUtCA_cOA!KXRfzGzd|;nLdx&DeZDpi$mLpQeIgrxi#xN{!)a?vV
zTY<h7Zl@&gEWVPv=d{SjxwS)lOb;<1%_2fJKR_aduYiS$MIuO}Bb1?J6qqTe>ZXR}
zEAWEy5Q&WNjxiRxsoeWyT1H<4b~Y=!3W6=?;44%SY@r}dInxlvsbt1J*^xz_Y~|b6
zVVz3*MH#|5c^hfJmbW^9e*!<s0IEbVg^_1nXg)@sf4eS%CPFp{vBj3C#W?~b-i3JN
z>W<$Nl8c2@CWO#wN<{$eT29?<q`5}og*{1iA0t3MI(vq|vjAKJizMt4<%tKZ%1H1X
zJr9Zx)aX3skdQ=J0z`-q!~z&)PHBMnc3NX+G>k#a3R+;7(W!9auL(=?5M0BP`}C`~
z5U;q_F0)UfHTx8SUm-v&%7`FDl!Dg_Zti$NVB3cag_U-f8VR=kG*Yy$gf?qbVEX-1
zzs5V#rwx@+ZxXwshv<r)gy?D*<nUBa^GfT=qJ)1PKZ;MwKhQ<!1C({Y2GOLsC4@dI
zUUh%sKyv3qUD-A6S9WEqCITc-0!6I|lc2m&<mbuLy${y1nnelx&-?HStw|RiAI6<>
zXWUu*w?s%~8g#WV3aMPb`a0sAGtg)h*Ro+x!x%sg_CjvhQ0bPp%`1s(Q_#(G+0)n_
zuLlS*lapbK7@N*K=)Y(b8a2>{-X`8;2Y@0jn{uXS<@Z3{5Dr*zCfO>5NNtUH&@N>f
zpq?Q#YL{L3=&k~fTC$C_vX5AZtERR6I(V(I**u`wgf_}gQof7W!~u;tmA24oilPz1
zvns82u>d^({^BE2!@TT8BBaKv2yT&Pg+6ajvm2<(>4?mJVU7r-h*%tZhibo0fcDS6
z0l*t0ypeI4unns52~bpuYn>QDJHz$7o=8ot`z%QdB<G{vr^er+#x;!P+QrJ)42s%D
z&ZJ+XHXk*;Mlndiu0yh>cw<i6#1FWPD0s2jfY=`@pISu+Z!Re#p$;PWuu!KR&?|)c
zQ%FE&NhHi?a0j!1_`pC(Asu(+edRzJs6vV&0zxTJn7S&Wak<#i*zce}n+AYphCRk>
zxi!A-kP`#xPja5oB-Uid4#bUt9ePf52iF+Ju;h@7#TNNIqQ8jC65fQzJ4D`#v~40j
z2^y~ky|;7vm7Kn5qyI~%)edcz9#Rn_(tRJ9f^<Pzn@|Rs_D4!qdQW}}%3@F^g@AGu
zAsGbiM9}SWR&k+Jz6F*?Sp_kFX|qI8|5V>EqfS0hsC-jPjm-*f24srEQwd`JLM@;z
zLTXt#tz|}6&L;3y@m9bP>Ylbv=+#ztT5INpdDrZiPhT<D@P`bnn0n}iG26BwT0@6!
zg(f<-2oaDg!oG{$a3#1+I0f1e2!}2qvgehqm%{c!n4CAqsT8a)NRKON%yrkh<HHls
zJwJ+jiB(~c0m9%H8i-F}%+&Ap<9<>sfR}{r&K}YnaL#a1m1`beT60p^Z1NUY))Mv|
z)C)r9N;hRfE6E)}5bXD;iWdH18vh<jU!}t)Zjlh~uEM~{jWXyR4F^9ncJQMej!v3v
z4ieg5z?=L&02u=*f}BZ%0-SWNo7zB6OZz&`R$c_*<Lt^v<GgX!273+X1SZHJ)X}<(
zo0QVPh+ab)bW!x9w&*8XmG-nj#ns*NeJRC>qg|B@2LzBS!;wU3F_JI>j8#Tdm>W$3
z5ha14F^x@}Bg7I|iIB%g;vUwIVjC)oiOfkjBxF`}vFx`eVYPspX(Kbz?<%*on`2Ra
z+}9~9&#(hDhTlnI?g$%xP`#Y{m|^eA8XX2%gSxcy<QVF6o)Z#pJfI<5Q*>)jurH$%
zBfZIPiJ3%O-gJl8hIA$21!IPe5W?T44%-CY1!z}=9I?vx;NbhoHkVTp!I67NUhGF`
zXBkvTH;}lr=_NUt3CTGUpL}9$$#2lEa+9V3Yo6{@<`j4Uu)^oaq?!BZojztBk0j|N
z7C@4e^jij5DlW~GRSxpD7KlWJUiHLP(<qQPw1nl=Bgp5=)TIw~R>(jHmuSN@lyi4@
zAd<4v9$V)B-RsdMEV4tyLJx5Xna@HI{1ge2hY<6)KoPPJv<+FP4N5Sc^_)u}`N~KO
z%17EjS|cu4#Pxt%%2Ntto-nu74ARln=Owp%4E*OBmnHCk>Q?YC?W=TSI#ydnDRH2D
zSxPlrJrI}Dh&8vqr;fF2Q14A0BPnikPeG4Kl#uM4TFqq&P?HN5Myx-^V-$A5&ED-?
zAvN>+x6$2STk*-^5#SM>CZw(=!Y3L<fe`=8$%}5bOLXCm8^t^(k-%ydBu;XR{Q-d0
z$_KZ{L@uK-EwU`&pbB%0%&%%b1Wd4QLFVj-)UH;P=M$-@X8S1T@{V9xOM~*fEySl)
zg3Nc)qXbl?%Xkq3O16ktj-$5(50q{YYcLY#AX4cy;Z4F9RCZ>^$mp=|W7JZ?5Q17>
zLjp3D<QqJ}P5CV?rxQ>IX+=0%uDpcvB)G1DPE>e|N(R{pk5RZ?ZfUPfx)JB5ud)|0
z9{W>(kwIsb?>dO_dQf#%vlk~Y1nzh?V##4IwVDy)65|Dx&U!Zw1QLRSm;v$kY`R#6
z;3N*miETKDt=*%vp5u0i@4uqSe?k?-c``1q2`mfeh|71`-%*plCh)j8VSj_NuMv9b
z7E*%q-T+d`C9gh?(8E(6$!t<kBlaUo1>&TE4BHC`+*J3q1E9CamGa!YBy2N~|7JNa
zA$Hv~l8sCgEi0SV&567Q9}H+?rM1J-v`&^bj$`N>mx>Y`W${S^+3NZ_eWjwVww2;g
ze*R+jU;de%e|r(kpacxM-b<aNJt>T4_6GLAej7j>SV|Yng$>TDJ%=GgpCVv`AYp#*
z4F@Psfc6uF+7P|zFhBGo2c{$YIt}*61W2;8vjol)SOVZB3U!BEd&zY|Fy95L_<I1W
zE+$HfyLAK$X<lT<$L4xoif;vaP$U4CjvTA-o8T!?;7`niASz3Ki_W=)StMB8CW1gZ
z&Uy}}XiMHXImSHT@DKGJu@MPh5>SGTNO=fPrjY-L35%_e1mG26NFlm;@9659cbve-
z#|8)sLy~yC@6p0YJ3|>PjV<IzGjY4Y{sPU~Ge^B_>djRIM;R@UYxtfaVBbeF_nXMf
z%XRcIN(njlCzSeg0)I<@ID!2mfqx>95g0Mc(8PP5N(TV=v`R`$Bn(({MX*Y3%^NyA
zn5NO1)O5Mlnw&6dqILTM*TKxtdyEz;*2o?pMXoG9;^`1765*S%KcogkB+R7-hWILF
zC5nU(XPv@_7b`G15DQLLMn=f60Cyrm9;VW10OC*)k+EjSyNzPsBJ68j7_a-#kTJev
zas1rKA%&ZVkI5CuBgZ5+tmld$@P*9bqE=kD7uvTA$+3=4ztVo>6+H9tPp~HSG^nRo
zHffhUEKMY8oqv2)IN9Q(Dw<&|mpgAY8X{Tg#9<gm!bi0lh#2D=E5uUiTQDZ1sQCVa
z>#qy^i8zTkp2Ks_Q*_F>2g|W~g{WP8(}Jie^o#*W{|kCeJ`PQk>!X@7xzu_`3jPP*
Czlr1k

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_24.cpython-310.pyc b/model/__pycache__/cldm_decoder_24.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..2ec373f43ed49ebe4dcf9b8a830b8591efa02efc
GIT binary patch
literal 10505
zcma)CS&Ss-U9YRFtE<oHx%cd~_i7(A_S%UF-mGJLy$;EG5_#82TqG*hbk+3qOdY*n
zRqxSM+lglrZ$Oi91%fy-0};YVAP^565_#d3h&LX-C!`>ul_J3d0%C}l-~X$=X1obh
ztN#7n$N&4^Uq$7zDdAao|21#%B}w`LWu_kunNQ%26<Ly8$(4O6kS#e-EEFKG_^PGi
zU-h*hXXS=vx}}S{yp<pR8{)qZn3gHpXnrv$S*4(Cm4k{^3943Al;`|fFk{V#wC>k~
zS!*_!v*v<%Yd%=87J@}<QPk!ArC`}w7HPvj5*)RT2FI*p!Ex)jC@c6Uf|J%skv9EC
zaLPJ`bkQyOj|7iekIK@kl3R8wcO|#tE$*n+Y2>SJ4f&d9Zk<8NS(MDUb(GY-W#rGf
z$~CDud!Mk=RCxY1r`zp>ZLYo;-dStPy!aXBddzb#cUnnP<M~z;-n2uP>wYKn9L9@%
zk41J6xjo<G^`sMcapHAtFIe|n7Xx94`n9f;bR7RWbHaEdVu8n+Ilg#(*!C4_dd<7h
z^FnIHjn}>SB|mCybMta^D~z3>i=itoCJ8l&!mO_F#+sMB7P+2}8LC$us=VCU*yzP*
z(o}iHkJ@%}JF%VE?gVYFd$&7r67$7Q7$*+Rlz8?Aivk-v+UU?ak6rFK?I?8oSDjuQ
zcbxDtda%xVk7h-E8O{4%INa=gP>w{U*j)=75X)pO!sGO#BJ&Bn@dAJ+S+Xlxikov)
zSGy~N#$;e$=d&+IVZtKc_uMza;l9?cv~=_tJ!K>eyr0G!S1@^78srmcPu`cl`HU<j
zN~)yN76x>cU3Flj(w>$Y`|_ZW%6mD~=t!IR&tI0lR2~%5Vrr&^RNYZgn%^^0b6=LF
z8^_VFkk#y?=JQ|%4Zj^#mO|rcDZY3Mt?2hA-c)m?p3FXt_S&AgFFhv>%%q4pN;@)U
zdHjpYpoE!nu_ancY$K7j%4nUZ)@_B5u&WHrusU6yjkP#cKfTr)pG{5dM;`5GuR<%o
zCcT$iyV8`cBJ;YQlh{${-(mAW6;4QXJ3+I`b4kQnn>-&zcAGgaBV_UlDA)7t)}|AN
zo*#3QC^(~Dt|fQ69?!Yruha><y~JK8!i>2NF7dzxrJ&=wzLynqwHtML!3%q!aF7=-
zN4;dKliG3;uA;12=34Ce8(hB43nP|^1#czZ?IcTp8hDMiVPg&=Qlby8Z{KoQJLbw}
zj1AF3xk^CZ<ob{a<<5H&`v^Ao{-;+rqrh7Yqr_W}qU}}JNu1R-i+bJFYrXI)xU;qC
zg{v)~V3m01rJs9!Rp4W_<-39HdQ`*g3(u}}@9;c07jwJIM1N-iG})AEcuV3fA${ye
zMxi9vm6E(9pHM1tNipUA(T9-s%5bcvwzh_2)3{;Vp##L*+_Z=Mg>=cbZ}c2Lt6|S!
zPO+~k@KGvT1X#iwlT<gPiY!V##EDc(Py7Xd5hoVpf>KZmS|Nv@E`F4vU%p`ArwWL?
zTYXpBklli7-YtM{A#?9amX=9dEB7v##L~gbHEs?!4Ee-JM!cs9LcFgaeV-O_pD6D-
z`#9j0`w#Gl`nA?E6pWtpNL28W_=x-Hl}I~+*THqOteVaf=f%}eL3QoO@5ou6HPhL6
z8D$EU5qJ4Ypbm~fI%w(qj+`FbR#-F1QF<_+R^1%(t7&;qP0J$vd|JW38b6OQbTKaF
z751vASL0X5<>o!v)>9cUpDKVxssa`gGtDK%9R=iHN)61IYs+&IYRcg8QkvV5v+`|~
z1xdxti#c*spDqbm@)qis#;uh2txrks$oC+HAZPwKDczH(?w-6=-jTPJc$aFEYEpYw
zdRN|1T*EEgHIkXE-HyBiJjCA-ZR)OhPx@`B%2Wp|{VrAlOzN0<Hq8Uhr3T=9S^!)~
zi-3!1iG43w+FA}3z#}L*nw9~Nr4_*AX%+B9x<D&K>SU_A<vD3^JZbEmN*+no^!Sc~
z^rQO{a>psRPZ&ClvNOrq<Q$}+l4^?*`c+6p?_x!5d7s8QPi^kWd(BkcAC)~O@cVeG
zw&lTzR2`gjtAj@J1mPHx_8%c>PYoVPW(YIssbydZrC%F7nmn1F#O(PLE1HuqXECk0
zGrM_=ayly;oJ<?3epC8onSDi^zJk<6^eP{b24_MYsR~l%v@$r8E~0KNodvv_wFqS@
zPgdy6Q2K>*iDn<14HdUOIGe1{d7*qGee|A0HQU)dr}tKAlsHMA!Z^jWoSsSR?(A+Z
zEGAFW$jHy7^Xb|2Xj*jVcA>{)P|?=;z18GwdiopEF#p)zs=(99ePE7SXwuf>>B;9L
zMY11HPhb|*$e8oSa(XJGJh!?eZL3(*57Lu^bK@~*Q1ZPow&-*SQ}gb^t`X*w3s@6r
z#dGfBuC^oQO3wjv;~6l{p|zw9cj;r&pmA_bSV?7!<$uj^_z)~BOcq#vkhK_M`RsUo
z+3J>eb+l6A#s33~M@CruaQ_{m$!A75MEco->vbiFv?FW%*RMp8f9h1@&8XLa>~2R%
zBZ(TFu;uq$uaRte4d_@eB-^VIZ8YvgJ=SQkD2|_WMrK<BD$MPWDFwZ=va-^D^Pyw*
z;%vM)vl$!7#Au|M|6fB3u?xj-UWEY%X$<p@eFB2C|I8~Ltd%vOq3uo2apML?9A(Bv
zKx5O1X?yElBleR1+g%pj>|mSJka_XY255NMSZM2nKt<9x-}a+*$M^5hR)vw-r17ai
z!^cWilO8PD#`&(tp7gz&uv{CrFor-MtX!|*dtn=SbQszrcF^;a7AcX@Lut|pvv}h=
z04PsNrDX~H1-PAqB_xyGUOBkhrp%s2>-Q8sM-}5!8E%G=zRC0|sZ?%U^qpYcbuJ~y
z#iY6FQvb|Ev8~v4C+sA)y>bx-k>kggkjQ%UpF=a)&Ak@vEh2`oq26F#%j?|qT!G2M
zOZa&I`IW}?%?{3_OH?9G;p!{bt~I)>6R}Qm2Z#cuC;f01kuQw%xY6%F(devrD}Z66
z<-}gxxP_C>4sFsUGv9u`^QFe6##0Sg_KgdVJ=ST!LZW_?c2h$?F)7Az`F<l#2vV1f
zKvvi#;O9P_icsY5$voe6GUrF*CD?mo%HkT>B<2+mIm*1&b~ggy@HxuIKqKI0XCtvg
zI7u*z*j0?l3nK!J`OGLe8F7*NhTg5IBaO&ogNK!Ky>7C}i=aj_D{ahe&A6S|S*M9$
ze^Q^&GHHa&n-ddWZ|2yiXfxz#h!XZ1twVF+jWw05STrCDrl}rk*Zc_2#XZnC8Fd+v
zSaUE#+YZ*Tu|kYM39#ggH2w)OG?j{RxY-GuHkQ}vd%V~Q#~Wp29icYP{I{0JRYIK#
zo?$PGs?Rxo&wGWjh;bEM$ko8<@*KoR8%8{62(9NIdcr*KvZ#x5;-zW6<7KS#CUJb`
z+KIzZfYAJgA34d>9~~y1xxh}+PSyx~hL%zv9aC)gx>;{)t_vp`J~S+OS6JhA$AuXm
zhTNR^zKk3`u3hU~eS;mRqzMlfjy1_KMiC8`Bha9#O5kjJw(!1bo!lHJW2<a$zzOQH
z43F$Ibt)2=A#jQSxs$9wphVyi0A8E+W^8g)&S!$*^VEdQW^oEuc@xN-oD89#IRw@T
zv<SEa#sp*;L}MG>r$0Ku65FIEL|Il9I2N1<bP<uU&r+$N?4cU4r__KoGnIjp0vczp
zQ-^#wblh~GsO^rmc$nC;07A@Oq2^1Y+TpGz9I-s%2vG3!+=6DU%EU>`m}<&56llT0
z!qZf6z&Ro4xkiE<r_B&Iaqf65whVBi&_PylxDMBj9j<5M0y3mv+b!RTW7}qzv0Nrx
zoi~tr25)>7ioiUf!Ra#KZqd&`O2bPfHCb15nSQ2RmurfFydlrwU64<sOp(n8dM<0B
zqs6GL7y&u7|FNbk%ABk`P#)+H=JGV2hF=Y>6-B`a3fg7=+K;pcM$yE$rdR_;gyUz*
znzDoyDS7?<texod--?bJ^wfT={is&1%N6|UC=sI-_5YZ(%~JY*DMdrp9w`5=J<#g3
z9|JwKFPoTy)=RS;?xVf0RDRN|N^ZKf_P+kZ(($_7A4Oo`wT+cqixl7Jp)~OeQF~0n
zD3KBwQ5VEqQaP0=?&7Ls#qUr^M_UuRKv*K|lV}ZTE*LnEbjHQR9T2CadATtPKpggM
z1Jo99)<Op#ll#jrXMq4(=0O3_BGE8<h^J{Lnj`);KmtbtOBKsRct}T>M@cCpluy(x
z4I5D4K4sw;8G$8ZDri(l^5}_x(pF(dfm<;PXTwBThr&nsR6_u%kQ#erk`D7^a^E<E
z8HObYBxNn6y;9a@AC3*&FXNiT$Z>C$MKDGP7a{l95tK3tG6?-3<SSSEovsjlEToE4
z1W0VL2%uTYBvlh>u919Vk5S#@1Xc(<Mc`=wu7Ss5c7gK5OIBelIS*r>LJlRuJ_YY6
z)<J|?QWn5&b8-djp(*|P)QOD38uQt?Uqr8=pMOOdpc8NfPv-WoT}KGyN;A)X2Cdo0
z2>dJoVs%ExF9ILDJQN2l&-Wn`D87?{+msT*;d!KJS20<7hNKjx3H_&~EB`FLZG66_
zG8%!z_ko9KiJtfm0Y-5XSN2p-^Ku)?GQtg>=Bl3VYIt*A-bIvrqa>mwdDpmGf~=#c
z2@vg>d&PapJu2!-ZuxFuN484j0Hd}%tc8Pm7jC0f8Rn0<v+mqou$5IEmcT6&Eo$zf
zyM&%I?y`FX|8)_CnFaeDA4g%XUVe?wK|xaV%!W}8+X6Y5ADLZ5rQ6;eUWi?rLVKRe
zp2RkJ*++1jJQZ6+@^t1worVVjvIg4GTg4l{1wbL1EjiUYQkws+{0&ez!~@oyO14VD
zSz99>v~#HjsHX_Z+IbfazN^4Bmuw>`>>+&Ws!3_D3~sAzRre{Vp^fu3%C`~J*rzdP
zk~&(=P#{CNX}Pt3!6NYd+Y9&0_~Usm6ahSLB8o-|8)iatmR(0(MqT7T2rEY*M@0VE
zSE%+K0)z*)Lrp4U{E>|svkuj45kO!H{M3pNA2eLg>x$IOMu)}85ZT$N?^5G?)VPGP
zTzj}OHix37k+JTVsm;Tt*C^tdF|i<aLHlEMh|3TuCMi-k&7soM!Jm4eM$yR+D%x1I
zgA^Vv+G!W`3eo-q5|CXIiS#+p_7Z{(10{ik+>!T`eQBTyNs4d^r9f-ynh50yZibZq
z2KuvE04RN!dc2ew_8SiQL#QWt&u9{>*0OzZZ(#?X6W+vSiZLpg;&0k0Uqrkakzm4|
zaG^&Ce1SGkL|Dg`Nw4-Fc{!tX+U!s0wAP}{jviv~F&TdwnIRd2#I`_esrLKw>k>#h
zM<o44sE$FN)B|c&M0#+8AYyZ;(t-;W^HqoglofFIAZ_L7_94;t@~D&d6)NA-5@V}?
z8w#0j0H}oSBuG)9En;(No}x5uIW6Kf@fN@y>aMmyh}PEn_4USP<5r{9c=A$X9e>Ec
z*DVKL5V1Qp#BboxEmg5)i|7OSH|$rjC$9K62)jTag7DBc#00$1_7WI)2+{NEG|fUF
z2twpa5^>%2Zg${q=w2s`y0KMYkPO1B5L$}QV9cb`?nS+LxB#9L4g&i%ngbpmF3vLT
z#B=LT0>e<==E{1^zKME4)?8^POlU5-B}jw)2366**&&Lj1A2uHojB(-nLs#r`jKWb
zwG(oGZi)cL`#W&rWQUM(K)y;$Gm&wSAX*ydDOO77zJ+s6G~Bb`K8XZ8PA`u%P8_E#
zu`@U+Fhq)LP_%~N-_3z0pw5xYr}D5TwMI|TF1M=<3Q!FByHaAn2pAd@h(i}fV~NtK
zv6{vj`xaK08B&A{IHrW9Da}rNBm@;0OH`!?>JNlYR8$e+U&AW|MjOlhk^yT5{7u`L
zlYT|Hf!ojt^Wh#*S$T?m7d_zAlJJ|<37yN$i<v_kbT6&bv5=LxB=sMgI)sc1g&drY
zmIyx>-I^wQ6P?(1#3Z5_uezg~Ou7v5{3&}#Nae3nhhHG@ivSpbgo>5Fjle-CzQg5&
zWOC?ED2%;F4Gb!zOHtezb>ocECWeF%G{vdazCjxzORNGat0^V2cM34t$yonjK@--`
z!oxTEA<KCzUTYYS#49Pj6atp+?wZOP$3>mBMD!{YuP1K2#^J(|nJljz#{UWG(u1BG
z3PuNad80J6cYAauHKpI2TIx^lb^ihu*&=!wJ;X9(UkeTKJtRm-LjL1QMu<X?IYgs2
z$icGKGnRoM%p)zB0BHkhjTmAX*A8wjODPbF!WLIk2uoLAl-&GLFr#Z+l)w(ETfo1x
zr_x34RIQ0pVn_KRE_QIGK`cun!lK$$r`lDh_m+;46t}vophpcQBuZ!2G7$sx4CTyk
zQfk6(6xzWh<Bd#HRkN$W@ik;qxsMRL#Qp-U?$e1v2x}r_qfw*@-w|Y_Mi=in`W^t6
zn^_zrhM`N8&Pb54-vY4e+2Hn+1ZFg*MP>(_Vqvq9jb>&;z)ouqNuKRdyV9^cn@B}9
z`va77xi2`^(x6pu2x)5NpaPuaFcF#bt$>I-#oI(M6qC|H?Ic%^$P2=4Hsl8Ao~&r|
z15NACPZ86k_lIbvfRO~vd=d$WS=hCp8hK0pqLxt$XoOTHd@)x(f>R~9p@CXdxRpu@
zMGLo5cxG;HH%|%^r>C#67cd_CQvzQFu=2MY#FIVfKC9Y|V%P{bJsWZBpqp6L5Ydq7
zf<kMforN0-#eT$q`r9_$azlzPU^txFrh^DxpVE4U-4VwBlIHv?su-Rq<MO(|v+$_6
ze2Wo(u)ij7P`I$aLD^e`U^-p{7sLysw1hZ0gwn}L9}wZBwZ?4t>&O6g(m;lBCUn-G
zwhsgkbGa<HC<)^YY_OHjN{DT@jQH796D<o{=2kJQ!S@T=RB35anv}^L$AOF-=-jXb
zhdKP>f(&|ngFdEF*P2Qji^KCzFSNh^_w@Ym1u%u;GnfEg;>69Gu&mkV(1z^-h+|9W
zx_fA-bF=F(TnZ!z79l{HUB$y!8s5qKxLe_6M2=dl6Ldlcwq<6u{}GKy(w?0maE<`U
zaGs;Edcd_8T_*q&lEz?v2Vj-OMDg&tAAw4m7uo5txt^8c^Mx*y3&4d#hbx>cxK$K%
z6f+@!%i^!oY1c7}4)&?RJYT@CE@C+SlT$1NPXAEf1I(mi3=&p?nMj$C%fntmLaYN(
zXtAL^6!3zuvJhXreRz%4n~vYXR}i>ffL!sq-=gJ_%7!0*WNJBwnu&REgYxHS*_=D<
zs8fHgB0$S%kzB*47C!quBr_+AtiVivAElI#dw)WyKO^wB1c)!#_X+$1fwu@uSm!on
zNjbB(0q`Bz*uhs!MfpHcl^-jb@;``cA-1NxuM{f(X*6^lalopMc$u!k8N$C|)=FZ8
z=6SAz*`Xa7EnTdd{Ss2-5#tM_7NI8=t{wXwYCz<}JZfNwPiR)cP`H5BDSRce0s{pR
z>Ud>rvWyCFGvjBGE{bY1;$0EuvF4|{jU)0R7;T-OuKUoC5k9(cI+;mCDme>klLwQ<
zvnxnh&kTpa=RA`|t+;MCux||I$_Bn&OM0O<<faSHGIHWAgL;Z(lYq*C<wU&Jxr3|1
zSr5MoqZy`hncr8UA>x%*6a-NyTwJSyI5s{KLxh(;W@AFiimz0-{+hs_2)&5onL2qv
s1HvpbE@oK0L_8qAzNtxM`5Vwp^uMH(^kewXbQZqFQR+vex99x-1tl_`?*IS*

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_30.cpython-310.pyc b/model/__pycache__/cldm_decoder_30.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..43220f1328eaa30cc40427b9aa43d28d253e169c
GIT binary patch
literal 10505
zcma)CS&Ss-U9YRFtE<oHx%cd~_i7(Awl|3h-mGJLy$<nu5_#8gTqG*hbXE8COdY*n
zRqxSM+ev1FH=s$l0zn*^fe2wF5QqmZkr!Txc;n%FLJAUEDH1#&AclDP{lDt&=^1YV
z)vAAg_woP!_g7K5Y)W{vzq#ryz9>oGr_A)DA@fPRu_8;7E4i{S1+paviiHB?6<@Vf
z{Hwke<gDDNOt*AVm$&kxe?$Bi0@E@@8_h2UC94#at#VMYDnZq%it?Oa3udetk=Ffs
zFl)^QbJkoiZ_Nh_)<UpoEsDClzZ5K6%OY*~M}ni)(cqYMEI4i*7i9(iL~znNDbl9j
z2u@k2kS@9<|B>KP>rq*HMRLn-<*wvbyv1$RI*okQts!6Y%*`_>Ig64Rw~ms!w~YKb
zSGguNXYUhsnhMXq=Ja~qu*20C!aHkCnHN99T#tF~<!&o!YCPYH!kc#Ja^3HSp2K)?
z;IYUKBDe2*yq<IeFHXFk?FH+e>tY}bQNPx6lCI-lXHFQmBNlk9nd6JsM{QrGrq{e1
zeJ`X|+<4uKU-YBa7B??Px5C&7dKkL$LXuE}D9q{#Z>)LAYmw{un4x;rp~}nMcDo;=
zNmJz&KkC@Y?ZkFsyBl=4?%nRjNz50!VVpQLQ{vfe76mqT)b7$ck6rFMohWqtSDbzv
zcb)Juda&+#pJqjU8O;Y?INI!eP>w{U*j)=75X)pO!sGO#BJ)YS@dAJ+S+Xlxikov)
zSGy~N#$;e$=d&+GVZtKc_uMza(Z1HMv~=_tKV>8gyszSoE10|^4fBb#EAL6)xFAc3
zk}9dRi2+?@M;#iew5z4Yo;)n1@@@__I?^Wo^OvPBmWRc(n3`!JRku}?=68+M+>>SL
z#&Ps3WHo!Jc@xZ_;kT{IQfNFa#TRd(75(1An`*Aqm)WamukD(9(zDXgOp2JJv@K(n
z$G@NqOPDDaTcV}J+KIGTM(aGa?kI$W9c5^S)#>tVti`GN>9yYYY-(aZ@_0YH6<YZ<
z>Al?Am8NVJnb-52#EwG$4x0z6a6+ov4VqP+OCr|V;Q27JJIrwzA(K}?xt?#gHk>f@
z{Fs|W!5Q^(ExFV4c+M4nrEb{kC-yoKX3TYPi3cty1zp$my{wq4y{N|vUf2hPgS>b-
z>L*j3)RvQQ6=lsb*J96abNMzej9DfYyp?#jlPm#h;5FKYjX8)&i9Wc#b<1I$m@6AG
zHbe{MDgk+e>mw$VyYETtquAX0udZ%Hfwvk)iMJj_TdS^<IIA5N^?R$=`r%b@XKTX?
zS6e{AD)G*XKlA#kz{hIKcLUq?sD{}Wo?7YM;dyW_=608f{>}nuvMJZ_mc&~^`q&SR
zLP@SGC3#6cp;Y9OV#<S~4<YT9(O6AwZ4Jk!al^Jl2Z*=1X^;2|>5^^V=sSK^!=A;Q
zVqa6>V^p>Xu!J`zscuLWS(JQ;6RDP-`11f`PAtd;rJxqHLJmJ&{3u1ge8Ip^6%cv1
z`mWTL-GXc0Er4$!bMH!)mPuPH_b!;k(!t9$ZjLq#`NT*@yr&65yssdApB8bSDDOJ^
z1mKnX5AcZkwbn5djGyyJRPd7chzIDENZW$f!F98&n$8pF#nn$kb#2REk+VE&rnB)f
z$`mRi?(&sD9Ug;p(9-#BIX$+eux65@^l(0{x;f-m)AF#ImPPuxw1R&%ehy>kVqD5A
z>=jY3#;;7u&3m$~r!rtZRRE1t1uP_HnoEk?3dp~d8kjNHk>@1Tl)>YrG`B5h<y$HX
zl8T!bbL6N#T@te7ZPYJKS}F0{pO(HN--8r_ocW`qbWftXd-7&^Ti#OQ9jZ;LN$p+f
zUAe8ehFiF6Br{pNZFw7bh`%k`)LrwQ^jlDssSa5B9jpYH)G_mHng^Uq4Z!)d0JxAA
z0T<H}`);zdxg08hM^JJ!Edw4)D}cw-D&UE9fmVjp$y9aAbJFm5(%3zfJd&#E@ofd^
zNB1P;j#F-rFmxJaXOgqYIY>bz)fOf6tB{J`!HPQa9*uRL+T4?Oo2j}tE_+Pi_wiKi
z$iow<Iy~uChmGV3!Z9T6KSI);8a|TD5N6U-%fJ#!e}4FA@??4vv*%N+XimbM#kA(m
z?Bp@Z>8xycGHs;#P3f0p_GNMU3Q`x*t9(Qno(XlNDoB;n%J59Oh`P0O7Vv7;B9y5-
zS)nsS>F3iWntgaSRNVUTY_dY<h4Oa#=sk&Qwz7Fn@2=7)agscRaf)d<J(Jem*_~Wi
zOrEBZk)KQF)3fQ(wCK+5K#$3wqRsQWtI65)^w*_P{;}Ovfv1yuz#O&Eq|L|Elg~<u
zWIvvsz$~bdG3Slt^i)Q9Zgol8Qn9A*rzeN!CS%T^<hv7W(diJT=G}!IBg`kyU`?bI
z&$){`+P0W0JqOH9X23Xy){-{drH@O)#=$jVC6x)5|24zmL$ItcSz!4=)?$R^vy=5@
zt6SdD(MpLI{|_u48DsIo{kMrGFN|@B^hXY^*OegBj;sw{zY<0MsZ))&qJ9IiyAvgi
zBx-cSmfv^1MzY~Gpkuv|Y_CStZrq9btkGgo96#xd&9(+qnA;^&3VLT{Wo7WzL&xmL
z*?4hgGd7ZaqmgF*e+@0fE)>6c6$TunG0Z#mNeI%xg_pZnD{DYQ+Z&$a#tn=(&P<Ge
z#)cEq_SU^d>?MPDdMvux#WtxS^Wu>W(D1Uc(AEioillMA<45a`@86-V3L~>g<5Po%
zkCm(@eOR)M^F5C}>3cU}xi)TL41qpaxn9Hf!VdE2FtSJNpzkLwQX=Dr(xek+@y2xk
zP@a@Z%M$nta4QE(NG7|za&WaxnSBJU-&6P;RZLE0v>8VFCey2=Qn_)_cY<}-xs)Im
zljf>RgERYzZN;{`VK=eum5VTl96!E<MAl>QESkY??zdoX5iv{*^#=1=UiYTw3QQhe
z!Y2mLtu(H0ba5s<q7rcmS6{w%t<huMh;@@YKol^&-w$UI`NBAl8-u|UjqZxK0vI-0
zPVB{vTR7?L(Du7z=G)J9zu36cc&Y)*zVXask98Zckf`5&yQ!g{*e}L$`F<l#2vV1f
zKvvi#;O9P_icsY5$voe4GUrF*CD?lt%HkT>B<2+mIm*1&RxbkK@HxuIKqKHrx1HD_
zoFte<>?+3Og)xD~d}f^7A90cThTg5IBaO&ogNK!Ky<W1xi=aj_D{ahe&A8pzS*M9$
ze^Q^&GHHa&n-ddWZ|2yiX*1+$h!XZ1twVF+jWw05STrCDrl}rk*Ze5Y#eL8?8Fd+v
zSaUE#I}X;-ULi)H1X%I~8vleCno7ku-0TKU2g~aYJYMXElZ`U6j!+wC{#(oADxppV
z&#;$7)#n_)@4d`e#JCDB<Z9sbcn;#D10x<Zgx2#9Jz<`AS=7Th@zONk@iNwVlQ=$e
z?ZjayKxn@0M^5te$3}?@&#;rUlQjaLp{3Nv#}wQBUe?>1>%obJ4-HG+71p@jbz#Pb
zAvgDZUq%ie*RFN1zQK-D(u9W#$C~6AqlgB}5ol0VC2+PpTX^5JPHs+;u~oL)aDw_Q
z!y`LQor(lz2%I87?j$P^C=qxBfY+wI8JirH^O<0HlbVp(EKb2HZvdJ5Cqw9G4uN$7
zEdnlq2?1FK(OA3p^vCwG#5Sl2QI=H&js<4|T|{K;vs5Z5d!z>JDK%itOl9DtfX3PD
z)FB^^95>x3YP(}C9wzoIfDp5nsrk~lcC_n#j#%F32vG3!+=6DU%D$7BG1ZiBDA0m~
zg{P_DfOA67bBzQ!PMaZa;@t6AY#HE2p@XdAXdSK{J6zAi1!PFUwp+dv$F|KbW4TPY
zI&UC#0dIU2ioiUf!Ra#KZqd&`O2bPfHCb15nSQ2RmurfFydlrwU64<sOp(n8dM<0B
zqs6$b7y&u7|B<FE%ABk`P#)+H=JGV2hF=Y>6-B`a3fg7=+7Gn{M$yE$rdR_;gyUz*
znzDoyDS7?<texod--?bJ^wfT&{jgT9%N6|UC=sI-_5av!o2B&sQi_JGJy8B#d!W^6
zKL&beUotTVt(Rsy+(&y~sr<NEmE3e|?S1_RrQ>yZFpj{$Ynv#y7Ad~*Luuj{qV|}C
zQ6eQWqArNJq;e`#+{IPNir=P?j<zOrfv`l_r_dVGTrhAR>5PksJ0MO;^KxSxfH>^i
z2B<CItc4CfCJ&Zh$^rqj%!2};MWSK+5Kq%gG)Md`fCP>PmMWHs@Q{u$kCIYID4(dC
z8aAN7eagZyG6GA+RM4o9<k1rWrOm>&0=Hro&W4Gw4uy~MsfGYjAvJc%Bpv0+<i2qR
zGmJ_QNXlAByQQqn9vmCEU&b|wk>lPhi(rfpE<*0HBPeAQWDxp6$XBiox;-KKSV$G8
z2$0xf5kRw)NvbB&TqF6!9;3R)39Jx!ionwVTmz5A>>0`vFIk0&<UEXh3OSSr`xLyR
zSO*bmNm&59&B+z8ho%hbQztSGYs_coei6M!e*P6<fbN4Ucrv$t?K(mrSDJbDQ)taT
zPT;2r5UVpfei8WK<&ijOdA<*kK=GXn+@_Qe4$mP)yNb!uGbE)jP3S)<UHNC}9plZK
z%4h@<-v=I|C3@mN02s$jT-j4S&C9iwWrQ0%%~d_!)$r!Lyo)G#yCk9|dDpmGf~=#c
z2@vg>yTv`pJu2!-ZuxFuTeeE%0Hd}%s)d7k7jC0f8Rd_;v+mqou$5IEmB1|%Eo$zf
zyM&%I?y`FX|8)_CnFaeDpG0A<UVe?wK|xaV%!W}8+X6Y5ADLZ5rCZ(|UWi?rLVKRe
zp2RkJ*++1jJQZ6+@^t1worVVjvIg4GTg4l{4L~89O*z%OQkwsc{B=+`!~@oyO14VD
zSz99>v~#HjsHX_Z+IbfazN^4Bmuw>`>>_;Xs!3_L3~sAzR`)2Vp-u8N%6Aad*rPFL
zk~&(=P#{CNX}Pt3!6NYd+w=F!_~Usm6ahSLB8o-|8)iatmR(0(MqT7T2rEY*M@0VE
zm#OwE1PBjoo0?Q6_#+!NW?ib;B!Iva_^A~kK4`d}*AuCkc9+G;2-(@F?^5G?)VPGP
zTzj-KHix37k+JTVsLjKs*C^tdF|i<aLHiSRh|3TuCMi-k&5_d6!Jm4eM$yR+D%wP}
zgA^Vv+G!W`3eo-q5|CXIiS#+p_7Z{(LnVQP+?IEhJ!z;4Ns4d^r9f-ynh50yZibZq
zI{LF&04RN!dc2ew_HBp!A=Hz+XEcdbYuUcIx3B}x32)*u#Tb`N@i%RhFCyNINHF0}
zxX@z+K2Mt`BCKP}q*n)zyp+*8ZT81>T5Hi}#}Bdhgp9w1%!rIZVw)hgRQo;obqOS$
zBa;3CRL3w+>H)PXB0ac45V5&aX~Bhx`6|Q#$_lu9kT!F4`;h3ndDO{!3YBkaiLqJ0
z4TVfM08~PE5~L{57O}ZBPf?nVoEGt#cne?;bw_IxqP4X_eZ8^KxYcMip1jmp#~(8A
zb<2SlMC^_Y@f$dFOI2*yBKko74f_@Bi7WmM!Y&YqAUyO9F##`hyaWawLiD^kO|uXP
zf)KfqL|k{hn_ajYy4MY(UThT@B!loOgqGqn7&Ga1`cXd~Er92QgTQ{3=77hCi?d8S
z@!YzTz%Z1zxUwFzZ=ha~HCH+b6PinI3DRJ{PF1vUc8KEXfL^9UC(b!dCJ+vuex#X9
z?Swp-n<9Y8{tlct*&$>ckgpQcOk^A+h?d5Aij~s2Z{nO24fia#Pa*-2)5{}`6US*w
z><mr{43Xj*6s;lncXFT!sB`4<sXXdQt<h7o%k5~x0u)33u9O%s0!GFJ;?RZhSfX@l
ztfsNXzKPXkh7=(KjwxYjO0yFm2|)$M5>@Ge`U9a86;(v|*YFB~(ZO=RXuz5Qf75p6
zq+eEU;5Kxh`EZY@tUSfOgC1~dN%-y837yN$i<v_k^e(N_v5=LxBn=*$I)sc1g&drY
zmIyx>-I^wQ3!T`v#U!E`ue#%#Ou7v5{3&}#Nae3lho2+x3ji2_go>5Fjle-SzQg5&
zWOC^4Qy6=X8W>bam!h~c>cttQ?Hdw8&=jXu`vz@@EU^lxtfrL6-YLLnCu9AC1x;8#
z3lHDuhb-rbc&%YP60fBAQV3YOyK5?I92a%g649$ryq>u6nuH6-X0p6?82=}zOCNe}
zBp4ms<&D$O-ktH8)RaMUYN<cD*ZpU($QIGd_#u`d`&wv-?;=4$67nBcGC~xB%pn@J
zVGfqHp0NxBVIFD01V|f5Ys3)CxOQ-JSxSLW6t=jULRh-`qU7d}f*D=oq6BtO-2(ol
zU6n3!r)o`<5<AKlaj}Cd4Psdu5f;^sI@PX1y*G7?q`1``1wCpgAyGQBmWddkXC!BS
zgHrqKMxh;CGTz8ERW-W`oLoaTmHP;>OYG0l>OP$)gs>(;HX22m@Et)$YIO0QqwfK5
zxtYa5Vi>wa>5K#!`%M6=o(*nKNnl1}T4Z*>DHb*x*=S}q1njiVh~(KWwJVLvvx!tx
zv)@NKmj{A#Ee%@rhLEOK4l2M&4ik|{-wKGhQ@lk4Loq2G)J}5sh`b=&W<zd(?#YTa
zKhU(n{1h?m_x=#g6flyYnNJ}BF$=pER3mT7U(hmY0gaHVgfHgGM{ueH+Zw1vg<Gkl
zP_%F>g=gmGcJib!aeDe1dmiJlKOyi{04smXK|I-m?z5`BD29!2)3Xt`4tj}I4G|5Q
zE-19xoh;l)DE1=;)Ze!0mK#!Z0mI?UHXKCo29(w_?2a-17c}QzQpM;*8JE`uo`pxn
z<y(ySgZ&kOgTjUVHOk&51k>>vxFB92r6t74A(ZZ)^Z^l0T5H0FzlID@Ck<s7XF_M~
zYI{KND3{A}i;^(jzy_Q7tc2Kh(}+KkYNBOf)7&g(HTZr(n<_2smnLO0$8jKI2Rb(@
z!C{WRxFCaGZ_~##>RMCjU~zc<>G{t0{+^y6JP)Q&d<GN1OPsh_6P7jm9NMrQ0C8*y
zU3ZTRb#C??hD(72!6F1Gv#WUcN~1e@A9pLfjL1=ob%So`z_!e+_CKT%N!qhB1kMp4
z8P0PQRu8!Lg6jleLed!Q?*Oc_m?$1y_ajhA^CCMvHrKOKe7?|washbe(BTRv3vLw!
z9mPxt;IjB@blP>yqJw>EFwYn8tBV-U;N%nwfzv<K_W(1g7=whBU?x%~<npi=kr3-Z
z6k2R(7X`c^tSrP=Zy#P`^`_%@@f8HF7a&)>-ZyD^q_W}1ADLRtp=M$p+@Sm!S~lkn
zJL=S*s|e6CS|r!-sfEve7s<@YA}cV{-$yAW<lY}s>Q4#$Edk;S_B{gsK;Ufx`>b=D
zvZS2ZI{^3&Y~tW6rlNeHsLGENP5B?hwGdlV-d75h|1=u9jyPadN4!i|;SAy5Fl!|-
zLi0S=!R*kEjFv7|&3+Ln@`&*TQj5?N3)hbQHZ>sfVIDOw#3wW>VJKWc>lD5cS%HCq
z2z9(NF<Hh1xS8>@NEb!58S$<N^H}rK-6j!v5sbFZPuG2D$Os?ZINi)7B9)v4waJ6Y
z;@K6XtP7(d@Hx-^qE=kD7uYvOa;1%L*OGqdjkxK;Q;eK=%b=cO*(9K{U^x-5b?)G*
zaMq)*!f1x6T;}(cXoz^F6$L>Q3K!R^AdZcX#1P@7kJ*@zvf?WhuD>SmCqggcc&1KX
t(10+@jEfmoFA)!juWxD+S^fre6a6nKCH)xwGo6KRag_QI>76<Me*u)3o|gas

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_45_22_15.cpython-310.pyc b/model/__pycache__/cldm_decoder_45_22_15.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..db20315e6325148d4a4e7113e79c31d881f66c4d
GIT binary patch
literal 10517
zcma)CS&Ss-U9YRFtE<oHx%cd~_i7(A_S%jK-mGJLy$;EG61?jqE)tb$x~hA6rjFjP
zs`qHBZIapG4QLXsKoCb}AVL@k1mb~1A}_oW@y5gVgcKySQl#*JfEeQC_y4N9r)Rtg
zRIC2|-N*m?-(N-LvMJ&D!Qd<2@s}m(dz6`eG-N)GH&$dxawS*xr9igiK(SDOyyB~t
zihtGDf}E8bmFbo)>he~8^lym&LSR~^XruYXpk$STvQ-W$RwbxfRZ*VvYr%{)BhtEG
z4`!{|V9uHg=B@c)!CD9wtwm9n_m_fYYgwcX|44AuIvO0ajs?f9<D#tKp9oG`Cq>%y
z8^I~-6w*bv<UbNTYCS4TuSss%t=yH|inq9}TBnh(x;5l$p1FAjC1+7G<JM79_m+`A
z=PK8v=InjKPE+Ce*PULk8+N$*Qg~;rDf8l|nd>pnz1(djO^xSUQFzl1U9S7x&~q3s
z4m=jwLFD#*kJpoK;Khm8v%O&5b6pICA?nw9PSSP!>&yw`cEkdYHFJFN`l#)z)byHn
zqwj^(iW_fu@ymYH+T!Nr=vEjzK@US$UP=;b5QSM?;mtKKc|CGHA2U?1I#hYN+iv$`
zG-;~5;zu1jxt-WfY<GhW*S*`_IEndUH;fa9W=cG}&7#1@j@n&X=dsIOrxS&a|C-Z}
z<E|54Mi16q@6)WPFQfUu3rCy156Y3K6uWC-17ex1MR=TkRAfGmH(mhnBujQBOL23q
z>S}jo(3lMD>wNZ=C`?%7`=0w|INI0Rm6nb^<EM;-f%j8*;|eD4NW*+0?aF)7*PoH4
zL`juY+QfjavZD@-RNB>2V^1CyQh7Ir8Xaj9|M|<(7t6z9T1?HfkgD4%O7pu$YVOIh
zbmKVs6|$N=)O-%ipy9Wz%2H@NEyWjap%wk!!kcQY)R)<(&|cd$_oU~gp_vphM`>He
zERTOd8I~|pF1AEViM11Hvy9exYTZ!?2|LQr46D=S*;tEH_0wy;@tM@be&q3fb}O{<
zYtnnUwJS~8Dl)I<If)&G{v9?CRN;hFw;MF8JeNeQwZZdYWOtb3GD0S=fO0+GZf!VW
z==m`>iGnlg<yvy5=kc5?{z~1j*H7$qBFvcU;1Um9Pzt)P>w8%-S9?*97rd|!3I}=d
za@0?zI;kxu;VR0SWv<1Z-{$gdUKq1XEO;yNZYNm+)WB=B4I6V1krI7yee0IPIx$x^
zVr+;O%2fjL2G>VSD0knJ*mKz2`#-z75e43A7$x3%6m6}#PU5U~Sk&*WUh9Wf!JVxQ
zFI;T_1*^n6FaOLNs{$XZE#D1n*P|L{Kl_}0;e!42b1S_&JP-cG3fyI)!m|LHY|1sf
zCGnP!KK4VSP?GCPNnVmqC>6P+nDXH0LkN6jG*(kvTf^aL+_3G?0s3uj+9NJQx@6lo
z`i`H~u;($S*xeNP2$d}YEa8nwwi{AK79}6zN2;YK{ye~#9}99pDX0alki$<GKT6Rr
zUoh}f1w`JhzALq5x8Rz03*cTz-n){eWir>wy$e>cbZ~Tyo1+awN->fY?`eV@?<+{(
zr$yW+`n%3P26*NE13aRBt#u3q<L5jQ6}%)w;sJUk(zf7u@ZKz|rt`#qarKi>VB7MS
z<t)#d>1@1=GKI>B$9yGFhsPizv~+%3PLFLVteNB}J)BRgZVvg?v^=b)Ws!a%t>9mc
zU%(i;7?<)2drj1<@oSTE^PX($sSKD;6+k0Z0Sk$l=91#J0wyS>24>84<T(j7WpH{a
z&27tB`IgFpq~hkq9672_mxM%l8}&<*R!aQ#C#5gT_aKWPZT=)F-IJ*9p1fJ!mba96
zhia2*QhQf=S8gk=;TG;1$xPO6Tiym9;%|vIb=SNn{SFjnssomO4=VvCb<8}Q<^ktY
z18_bq04}6Oz{Rx0zLP9%E{6)>5tJNF%Yety3gGdy3V0%2pp_wYGF9F3oHRV1G<HuV
zkECjPd|N^K(LD*d<CNPY44p>VndEG84zf^5wM7a2Dx|1)u%eE<M`N9*HuvP+W~%Ou
z%N`T>eLPh=^6*5e4o|w(VIz5ha143-Pms5#hL0pOgqifzGO&cwUmQM~Jei)v?D-Tc
znv*bRF|D~XJ9&(9Ix8EVOdF|wQ~DK|eMy|Yg49LyDj$)CXF?sR3R2~?GCY$mqHZmn
z1-zQI2xTfyR_M%7`uTK;W*?pn6}LV-o2<}zp}d_wdQYO7t!$ptyQ?%xoFq?SoMKu|
z&!lyCb|)7Wlc#B9<mb})^lW-GExL0%&}K5IX!HE;YH~I`{WWQne{6SE;OXQZFh?yk
zY4h>)<nxju*^j3uFbisA%z0xuJ(W?OTV0a2RIKTH>B-@_$(S=J`OXAebUK8od3RyQ
z2=mDWtci5vId^eK+ZJ=B=YYA%3>fFoTGEEQ^igTpIJhRPq%y(s4>BA+1j`DO1(qLV
zEk;;AJ6T`0y5$`mt(182|G?srF&01Ef17CXnK2HLe)iycT?r!X$lBnID^cX1I@Ne9
z>Ng;}J5ka|qDD7t`F+=GBpY4>n$`=+0&7I=#+|6o8Z8#Z@srNjcxynBxm_}-pnX<W
zRt9f9bj*I7jTdK@V<Xu&9ckwO*U&=jLj9XpVa7ok!^mSFhaeq1^J*7sWew<Qd&6_w
zxPcMJnTaXT*l=Rn-n!R_y=3rCk3~1T*d{e(UOch{8eTRQ+BzXnku=VC{Ak_r{X4W(
zVQMyMd}`3}v69uK4|}$8zUQ$geeWjh*TyZ3A<zd~*K7D**g+m0MplU(^!=noN@V;{
znsmY}-nb3`%9B!QSpt6nZslMP$z;7(4z9K-vuDxzJ%!Iv#pF~*n_;AHGQCPFl^Yj*
zCs=o#O9^r@X|B37IJ2+VR&2W)b`#rPxd^k!@#9NKWIYDYqZzE{ehXF?5yQk>Z!oXr
zb#HpEz~td2d~EQ-O5^%Q7iZEVDiNn}_0?<F8a>vHSU0%?L;=(L{cskMFHH2fF&I42
z=&pDxfMKKM#9rLEg_F(>ZNE!q#Qj|Ni;YW-ry8*D8y6mXtlNN{ME&;LO%46Teld>A
z_Zx9Skh)|9vcfI_KlkZWgd%@W=J}qJxj-5(!Rnh(7T3TgF|UBgQRcO_dJzbR&rv=G
z8UZ)D?Zgh@D#0vbS1~3pj0rU6Gvnm`h>O%W^lnWZX+(Y-{H&bo^^y%<1T~UzX=83{
z#_h(=I!y#Cl=_60Nh4%FotW@?Gsiwjn<0Ngl(5%n9hwWDtf^$hq5)YjP4!T_=5st3
z_d(-i)MZ3s&A}M$I9Nw}g&2VnV9A$g{1akmDiz~!vl}=aEU!E8c(EH!Hp<94LT#LR
zaV?LlggO;G!(I_ppLP7c_bOu%<0`n2tAW$wIf#!AOnJ}{TF*c9gn8a&Q4i<DOVfPE
z%UI`4;`q$L6NjMyq4~BSImy!>86}>%z)sRm)(Cu>mQo)dQ*8HpS#N8u2Ui-NH0*g-
z*yMKCg)tw7+}!to8M%F2yVkw>COb|^6Min-Ym#G(;u<VRpg~oYz}fO_;e*pUxj9M3
zR@rXD73#ALkL)ycDiW9>aEbsqm8?LZMBotsUYqu5Y;sr5XM*8#)P&4taSB#>1IXMz
z8A3mE2&@xm5pW4i2*@&s#@fB7KeCS{wn0sZvaBj_EI1SBA|hj-p;AHFBQ;=8sR3(d
zDg##qG|t|j4*788zUe+u+Z}81FtKN0gqXcb&6mctqh0TF#PU8zfP$yz7Bp*B_Fct{
zsir(cffgJr{7nT1oD+hcYb3~V+6>VX=Z?o>%K$eD9b^?p>u~MZ;d&-6AVUha-SV9{
zwrzG9%VomVc@wE;@Wxl62+R{2oGt_I7X1vQG`v(&lXXRx>1WDyxuzJ%8}c091^G0}
z6xn>B=du<$T8!I@5s*XsA8ERx%*o0F<$?ZSE>GiW_|?!_Q51}zpk4N_{ZM;g6itk4
ziZx(FIDV$ADN9(9lGoqQ+KE2@t>~yhPwhwA4{PPRT*0r75;0m)|BwB)SxWyerD(|7
z1Lfbf2U?x>W1xrj3nu2E_0nvI`)Kbgl^-{&lACU=y{~`2bi6JP#&H;UZ4>3zBE>g;
zC{6rA)E<*CN~A<a>;;jRR8D1zzPKt`@!J&G(bj}65S9r01X@Fy3kJ?3opCX72gE69
zUT%!T5Qlx+0JR02wa~%G<iYYQSvY`}c~AhfNHmNe;%S<R=7_%ukigNvQpGY6B+?P=
zQBn#F<r8&N!v++%Pgzh#MtI4X3LO=aJbEImv{~3z;8x7S*)S2@p&(K|)euH1q{c3p
zq@z5U+&9i(hEWN^Nm&bNx0JQngJT2t%eW>na@?C`5t0$YMaVsN1f`6E3_?E$`O4Kn
zw<km&3#sB10TNp*0%(>pN!3J}Yb2l8V^sGzffWKz5qKJaYv8e%U7$Sil2w>U&coQJ
zkVA>EPa!;tbr7MJlm)QcoLm8WXv&~Ibt2=i#(Z|}7tw3v=U)*9=sviDCv*GPt|J6;
zrI}|xh1Tq&1b&(Tu{xvU7l9969*KjN=lc)|6y?dlZAuB@@B&h_tC%c3LsAOUg#MG#
zm4A`mF+NvQ8I3^V`@loAL{I!j0OPocD|@P^dAYW-jBtaexvHnT8s40jcM&UZmqfHA
z?;3YYkaZL_0ir!~x40*{M@3!9E#EC{%T|dTVAPgJwQx}H!fmuFqx>;<)}6Zxwz8_D
z61ZieMa^Awm(X*@U3QP)zb>LMvtYmDlPJv9%dhh}C`gK-*)YmsTObGXBeRRBbj!QL
z3$bfcfX{Q;lh`IN`v`H9r(%nEp3Xd|)9^r8)<8RYt9awL0VqVXDW`f@O7q{AzXl42
zc)+?-$yO;uYiq=Vb}rQb^%S95JMY55cNMtil5He~T?A2GH7V_u!EKez>K+9(v`M~3
z`3`~_do<=uQb(&93S<a3Ew}csSOlJbd+~l5e?0GnB8<mPMA1lL!%S$-vg@eJsEhmu
zVdV(qi1;7-64id00O5gcQ<KUBe`KS^tV=bU1Q3`4KeZyn2MyQrdLlK`?y@);Av+uO
zU21%f8kaDZYmZjO=1|l$GS>YCYV)w^HHv#?Oe~09(Edan;z~q{Ns81>bENci@TXp=
zQFQWyiZ&7LAccpEcG?BKLbN}D1Z0;)B7F|Dy@X)HP)Q&mx8+@BPa3L1k|LZ!DbSj_
zCPI0Fn<3@DhW>0807@UG9xr8vecK^_2=yfI8BJo<TDC84FYLf`!kf5KF~%iR{7oC>
zi-<ZS9!$6sF7z0IFVg0T2<zA~>D9p_uVl1NoBc7J)>^dL@k8uAA>(f%Ga_S<*d~ZA
z)&4+!Ljp<Xh@`&&)iKPIdO)p;NDuB2L~QO<T5zFaz5;Q8vI1@(q|F@NK_vQa9(D4b
zLgkxUVr&+0Pa)Gi0F}_q1Stx%MQkq3Q<SD7r$xLb-U8S|-O<{FXl-p!UvF$QZZ%qs
zCoeVD@rMk2-E!as5xZkU{00u)RTW#dh(3^i!+s5W;);KRunPnt6c2qvOu!2rFM)xF
zKs~Qc(<}snAVjVt5!YSsW*6>;?sdbc7h44e$soK6p{4i?#!R}Me$<ag3*b57Ah2Jj
zIpFc(`Yh8<Jh$#7Fbw4_uB^xG>!=rG&6Q5Vgyxc4f;8A~QWY(n9in(TpjYY8iE~bq
z350{EA8962J0TC|rU+oNzXK;ub_f{<<g3Iq6B!2yqNQ=3Vx@HMn>go0!)*(0l}Nzj
z^zumK#BtgZJA;z~L!`J2MQaHDog8Qa>KwUzDvx?nYxETDay#0v0L75MD<uYufRQnQ
zICNn=mMEPXt7)vUZ(wzqAw|f5V@g<>((J@XLQsLRL{)mA{y^wNMHLbLHM~M#bg<kn
z8n9---?W`M=~tB-xD(xHKHMfMD^Ia+qX(Q?5`O!2Lg#YxV&>2Wy-VwKEMz4vNrT6x
z4k6=0AqS_UCBhFzx2DP7LMQeuF^Oo#tM2$dlP*I%f65*bQu(XY;pYkb5&%XZp<?B4
zBXH1-?{GOGnH;+N6vp181_l+<)hKR`dT~Z+`-X%NG{vdazDXM*ORNGat0^V2cM34t
z$yonjK@--`!oxTEA<KCpUTYYS#49Pj6atoR@S4gR$90{xMD!{YuP5%kCgH-dnJljz
z#{UWG(ubZK2}TDudgC;-cV~PiHD%D8TIx^kb^ihu*&=!wKg2R*UkeTK9VAFdLjL1Q
zMu<X?IYgs2%)zqOGnRoM%p)zB0BHkhjTmAX*A8wjODPbF!WLIk2uoLAl-&GLFr#Z+
zl)w(ETfo1xtI~DuRIQ0pVn_KRE_QIGK`cun!lK$yr`lDh_oj}K6t}vgphpcQBuZ!2
zG7$sxjO5I3QEH#vD71sC#v7TYs%960lgr4aavvdfiTx#7-KP_U5Y|M<Mx#g*z9z^>
zjjrEw^i=?^II}oN3`3VFosl48zYSp3v%&2t3Cw6ri_8u<#lmJI8_mpyfSuMEkv!X_
zcBN5yHj#>I_J=6v@<4E|r9rFS5Yp7jK?OL;VInf=ivbaLinoYhC?=(Y+DWb+kr#y9
zY{(7JJz3G_2bwmRpCYFH-XEfw0!9)v^9dv%W?|QYYUEA%3tC1kpb=7)@WovD2u_t?
zTLZPIa4VG*iWY9A@XXxYPM#DdPETKBFJe6QX9T_iVC8Q)h$nl{eO9#>#jp`>dN$(L
zK`*hYA)+DE1%+0-lZ6`z#eT$q`r9_$bwi3SU^txFhJy&+fYN$~-7&`hist-lsu-Op
z<MO(|v+$_6e2Wo(u)iU2P`I$aMcLbgU^-p{7sLysw1hZ0gwp+!J|MzLYfaejH;@79
zq@fJsOz5m#Z4U?@<#JhWQ4+=**kCiCl@Qx*8u7EKCR!FY&COy~gRdC0snXJZX;LP0
z90xLXpmU=V9Ome|3o_{SHhof~t~Heo7Ki7bU+jGMAL#l1i(m@HXD|W0#EF|VVOg`!
zq7B;t5XY9#b@#|n=Vs4gxD-ebEJA=XyO4*kG`gAhaks+Dh#a+8H|T~AY|G4Q|6>}F
zq&+)B;2Z&x;XFrS^?++HxlRBkB#puT9>6M#iQ>_vKLV9BFS65Pb3H4?M+`kE7k~?g
z4p%r?aH}ZjC}u(cm&ISD)2?F{9qdzsdA@*OUBqw(C#P5loc^J{2bf947$mF&Gm$bO
zmxsNKgjffn&|*WoDBuNQWg)(L`|uj8Hyyu=Zy|8K0J-AzzCp_)l?^}s$kcKUH52pT
z2IVi%vN?CyQK$Y~MSzykBDsbSE`0X;NM=qJS%I1UK1wMe_x_Ype@@`<2oPVe?-KY&
z0&f%8XPw)WCFRWC0l?Q_69->073Bj(Req#s%Ksp)h1i<%zEY_Cr_s=L!~v^1;$^xD
zX9)j>Su2SVn&-I=W`}lUv~;m*_RC0-M~v@~T7;fhxOVJ!sR5A>^QeI#KBQR*L*W8i
zr|_-F3JerPsN<E1$ucg$&5WN#x+tp6h<8Pp$C{t+Hi^iKV6=69y6!_mM)>r`>1HMo
zspKrEO&&}Z&#oY4eRMPgKI++D)Qao&0{g~DuC(z5Thb4`5jTxLb7ADlTL$$PizhLa
zh0KYBt#b$0h7%usAx85|<udQDL_@?YttbehP&m3)1<`DLForlUeb~l?tQFs?aQ$_G
zMG=4zhctEmf)a#rW?am0dx<zfe1lVy$o@B=q3C}}De1@XpJ^_9k)za)Nbk(~{|l`p
BqNo4>

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_45_22_15.cpython-37.pyc b/model/__pycache__/cldm_decoder_45_22_15.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..28bf817a60d2ddd86664cf861a3392efb87f2db8
GIT binary patch
literal 10321
zcma)C+ix7#d7s<P&Mx;tF7K2?jbz={7s{4zajnR)Ez3?TGmRwM$<*#-xHDWXIhWO$
zSxH>X#;GaCY8yJL+Y~606y!q#q&F03TA+PMfuIjX(f$MHDL|hD6lj4KeFy@yzwgZK
zaz(psW-;HM`{g^|{md)XYFWbbZ~YIwW3NckzfflQXh?hkZ>-3Y<VvpWOMz_3fnuQm
zdBs;P75}QQ1qG{+m+6)+>WWq||2M>cDJWZI(MIzpf{IlMs#Z0qS+$^U)kS&1pA4p~
zDUsIwMlfwn2Q$`8Fl)^QbJkoiZ_SIkqQ4L<T8koW_)EbN>qv0aIvO0aj)}68zZ@L5
zj*E2JH-i(_38W|7ihnYA#Ck-QUYFdeTe~Z{HE(`LwN4>lcPEjb^vav3QE~<)Q*Hw#
z4Q~<ov#xSoYEC~O%rq5VeADT6J7JrvFNfQ!O_@)8iMby0+{>L-($sjd6@|C#(B-<{
z2|b7LiN41oJBZw#@9{>`3A{M*y0#ard9I6rFht{e*GW2#e}g$;ydJT@W6c7ezmd0n
zjhbHfZuY#8T5;o=7r)|1tu0=@9Ni9MC+K46+RI5o4WclsE4{VqC2vNq=VONYRfj4s
zch=W?F`6`0Uh|{2o!m)mC$>95o9o`4PMpMiz7xiYLo+3wz0RV*#*WrIw9ccKJ5D<a
z9shNw7snkZyo?^Kv(}?oQC~&#z8B`3eE`als1&<vVFO~BtVMVnJ|ildLy0F@vMX7N
zTX0oZyDNjfWME$B)2~Kh!Xn@I+_%DfPpem2dOk`O|0TS?f>)4hTN)G-X;0plzWp^>
zN|aPdrA-XxD!b~yNToe3HTLB}DV6sMsL_!w-#LX;@v`*Z(qJN;NXuy{Rd>`uIW6uP
zX?b6krJEPuk<ho4)$OD19WcU|rEAhxq*QaItrGht>a@M`zVy6=5#3}0V^(%#jQaMw
z%AkUg3-J#`TZ8>Lkv6MnU!?YJg|@t_49em1czHJ3{8;_?+HQV=6_v3EIoktPSEIE8
zX@%7*!2AUBx}KBRQRr{8S)d1Hs@n;gbzVp!*4p63FtXdsahZWyUIUeSzTMhz!qD?$
zUM6bID3xo;cGu$tSNv5vVYipqYea-G*TE4UD83SOT-W!qVy<?hE-!gu57Z1&;?<~^
zjCE35PQq1`HLF~UJ%63ccX(;Y9<kt~#JiJZ37kHkq;1%kg9wu7f$Lkh9oCMyvJqoL
zv{0@RkT<xVGo0M{p~RlW=6?3&^BYm%oe!hLTZ^Kt^RAOP=i4mmb<bb#g;&9etqm_c
z-vSEG6Q8{D#cSsUKF+s%H?UogYMA}Zv-X7x_EXQUbhmjCyowdLi$r;+03=zH%km`N
ziul#!1;vmn$_#P~%CbDEROJ4V&miZOe59tfx{9OIxMAC&1LWJhZ0G!hbj7xB_8dQ}
zVb5biv9~es9F@%jEZ~h507Dk}hxm?Yfu48;V8nOzkEC_kExF~pCGZ|(=3U9sGHGcQ
zK9V%a(m`62yqs?U@`VwP|4<Xe^?+u3KrH=$sObj#Jm8fF|BXk~ueOe&fSz-BYj{<>
z@h9k!NIQaGWhtFz%jqm}Q@orguDm0EAZK}YEuD@ppiH4M;ul{Hw82qucp;tLk<+7F
z3cHr*lpf5cb+>^0ucy^PJ*|rLt+a-JHNJ&0bTKaF751^HSL2UI<>h;_T})*_BUJz)
z3joXD=R#81QP@wCYHDD{LR+4ZP*VffR@1_coRx2>j3;%sDCQ_oeYzmz!_QH_Flwd5
zKTnho<a>}JTc!Ilvy$pPiE8i3o3$NzOSvx-4nB~R$z<vy=_7evaSgY0*GL*!%N=<K
zxInARqSdrpz9)Sb$}ZIbE8oX@fKMIs&!k1b+0+1>OG|+B=>*_HTEV<`lf})YPy;-I
zvZE>Hw~wVYz~!_Kcs!k>6(VJ(s#~3r2FH>Udnc1eQZ+rcqab~1UqbE}<@O0vr%`q$
zIh&k=d{a_wUP8Ybso7nusV(mlW}DRJp1k*Hs_qZV9us(eJXPEBU^!I>$KCqCyf5GR
zN%92Y`?7Rje(y(+q$dU^QEDa)!c=-<5m*y-mBAy)N_reCDW)e=bq3NA@^aFh+AU(l
zQ(4*IcxtBlmVRGlbAlq2`wIIOdR3RC!RfGqR1K+WS{s~B=TY|;=``S<W-UU6%9HbS
zjwt>6bb;m{oC!6zF*uW)rxQf^Pt!;4NmR2XiFr=#JxQbB#Gk@A6KOR)oi^O*-9k8*
zT%eJWpGjxaGwG3Z!kyWLev(01o6Wr^le6imkEJ~S*xr)@XUF$}MQWi*n~$f*pO+NL
zemq^qEU1yC`^wGB&<!UtdUWdx(w2%<olB1o&W=W%!pc4w;p=RUt66t$*9dFL(^wVh
zy0h;5uC^m)OV0v#qd73nq1B|OyYPZEFdtqO)^ZMO5xj;O|0~1eXJA@mOTdwaQxRv8
zWBTl9h1ppy?&{PkzViQIa%qUkPxs#;8hvJnN2H&9c*U*+F}JkZzjh^x{1Ye4uSPu+
z^0^%)W)hj5u;uq$&rCKv6S~w3$(}N!b#puFF|);@IDWzznqnqYmD?fn2>NAZWu^bs
z&m6NCXXC}09cU&8h9J%S3k@wqF%*3HDvUBnXqas53lOmVr(f$}t;~cbwl_S-jZKU=
z%#4)1x#7gLy*1B_y`=wsmqoWa*d{e(UYzTC)62#}TPFl667yW!kJcRD-=?h!!?8)@
zQv=h-O3o)eSf}Q>uE(D6y<4zG&D$75pbwU-XZl{)Mjjn<yTcB8e$pbvLJ$3uo|wiP
zHvm9)u)`K5@E71#0cRzX9bS8QrA?VVgVwOvXQ*OyBKbxb>7Yz+lFH=9Mc)b5T<21P
z+<)Me`lk=n+=^{?!cJn_D;Ht%IDUKyiL8D9c{GC++-t$2A~G16-6r!|UgwtQ3LG9@
zz~}ofu9!DAIyj9k(TBkQ)z_|HH@mD8u}-oL6amKv{csAAFHG&&?DroxJ1gD_U}(0S
z*o)2EIOFWN4!UFp+OKuqGcTD>ny}N&3y(h9F<}Eyzk_yTLqC2{jDzxhGfoInmy9%4
z*d<`+0iDUvrYv@y%$3o21s2_irnm+MiFpm=jWVyb)r~+he1`Hd&<MEESx@W`E)R?z
z_RAQPmxiPn^QmF-V8nUq8+y0L4m2YF3;t5U^}5Lhp8y?_;b>!SYs&4!&KgYw3zPbU
zR+&b~d^0iOjb?$pL7Vv^0a^khPXcR)Z`D+?V$p!?nWlQEUGq6!h<l)JGTAaBfh!ir
z9-~uzjj@PXGcY{c4%W54LM%ZEFz3rO)v}n1O2xRm+zFgEmfPuje4-PMHqOX)LT#LR
zXDyGbgi00M!{EE4>TSpGd15}Uf*-jWI9*<VAZf#F2Th^%>}Q@hFS;!1;@o&;oCA3k
z>%2w0pE+*ga1<c8xb8<z^3-#A;^_<QIPK&rfj4O>jp4DycCVZDwr09;k>Mf3c6WvS
zZFgK4?qSHw2c9h>myT=KJ6GRg$0%8b{|dL6BpRbw1uGCRsj3z@Tb?a^Yg#8SkCL%f
zwb$YH^jL;RMwU67ATUMX1Oajy8F`JYLf|9-pB(pJY;sM`Wn$uOYVr;NIt8n`0c0MW
z456QWjetX7jX;aQh@dQkXl}jx)N=<|Vr^<dlxEd|W5KCF7ZDv>rBXrZxk9kV6oNH1
zmVsLWnrFXC9g1P@s_8ya+qN}-nE12sK+ImH<_p8xeAfpYvUtEDpy=_r1<hKu1Gg|^
zvN3N^palmD-%`N=7lg>?8i{nAHbYdz+4fj$8Q@N#jjV}$9j+ZaT+hS@#CgfKTfP&=
zwhbPn1!qSg{G02@K8-j2PXKAT46kdF{4DV^kkar{$)uvo*{>`&lu6k@-hek&mKWqR
z@?2q9tIIRPcGO!okgKVhqR7kIj2J^h=^T7GMb3V@_%Evl#==kZ(v^ltYs!MG6_=2P
zH>gg^4Y>xNuz?(UO)R5lLnA-2fq7<>@!rGH^Er%c41F{!l~+{wki(D&T(FUDYAJ*b
zJoHbbO<0SGViPHm5e7jJB$ZQ{0wJzSruYtp3ba)r#&h$5y^i(}#e&<4NM|fZq>OVQ
zrOAz9eBiL#>7%v;eJ$KhN$xMcn)&Ip$cMSLMSM&TF(1t$VmIGIB7v`fg^EQYTB9ST
zp`;X{DJJTshUF{pg0fhNj5v=m7Q-ps`CL*&UqpB|OFIgpE@$8?lo55IXihQJ5YH*4
z#va*`d7f<Ln^$3-N_%-3;yPIyX|IyC+J}DvKgs~AL{x>5XPs+4MxKAODxxPsHVCoB
zmZ-%U0wmsrc;sro(-o471ym-4&`C;#0L@B9-A$yqM&gA%N_CGBARnDQN#H2}u7O2j
zc7gK516FAy_zs^3#Rn>M9&<=YA}j$SL<nL5j54RxKz#ki*clCD5VL|7*hO^8o%k!l
zk~{!c@MJ#y>J7vzt~876b7;+8An;2Bh(#F@gosk`YR=6q&-ZQnV4<+mZc`({)-NJO
z`-*9^MhT|hFZ8RuFTH4}jCzyU9X&)>^u$D0!yt#JdYV^QR~9Aw>-bTiTK=9cLLZ>4
zyA_Bg%`G7GQS!=r6Z?`oC+do>aksQ1TNM!?ffC4TMVJKTwLCvhp6*?+mQ~M7;D6qQ
zS7=ST@c1z9lsn_j;=dt6D$}5=g;7Z5>g6{P@SK50qX3r;dm6?7a<CUN!-h(?ylq~J
zU7MnBp39!V_ITAtteKn)TLjy5=0X33qZp}yHuTo<#(e;axNOR)o)$j_bwfB{#i?Yg
z6f3ng;z7HRYJhr*7^z)!;iJ0>JZi}{lF}YxA+DNK_Nw5u+Gc&9ViVdZKS}vEViWr`
z=2X%^t0{^`2+yjp+QtI#{QFA}NDcF%7mC;#FC(}`nicxIInA!2E~6td`-M3okRpO|
z?0u^J1_9bX`wakI8{v(N%b0CYO@{zQrMT9L5VSK~&+CfR)Ov@-Nsi=f)VtL9Luy>X
zSgxI~jLo2^X=F_Lb!zij(`yur6zn=AYYIGOv`zegTZx<(%QcAoq4KGebnxbqG7{<_
zf=>%|+5x>ns6UPbWR^t2d<K^>3y2R4lmyaoN8VHRrGYA>C?X(~0)?roA{v*8Esgy)
z`m<>OXlB@Bypmbt>kc_Fkp3j+8BJnMwrpQq9oT{Agtu^?VGK(SxL9nF&m#<s04?E7
zc)Ua8y+qq4;*+59D$slVldop<O&k4RI<2;7v-FUP7?JKL$mFC8(%OVFNVPvy(!$5`
zyHFN`A}Iuvs|d*;XeWYh7t@jprSc(I9%Us2|E0|Wh5r+MuZTK%U!n3%EipDrxFV1#
zAWtO-{tLB$wg{=E#iWuNZ8@F5TgF=gL#VsjI-yrv?KjrU4fD3yGM~6)uHg?ESTWVW
z3nI2{L$n4CT@OvPY!M<LSA=~VyWxs|lW+>OAtDZ4LS)ZNZ7+fCg?Kryk5egFUyvSG
zl8EcBcdG+WK=(Rf)QznYgA5P`ztBK@3S%anb}#D1`2u)B*zW8;%>m~OH&&VE;e|CP
zfz2jwab+!L-$K10WUjOmCbW{=5(L41m#S#tpQiEep!795Y~mIP;qEdFoXjYL-qCRI
zGh+uo+TrM=$>t!T{UyBd?*Wi8kRZqz*C@bA=ens4^rWz-<80+c5I)YXh&0X{XKk=I
za86)?6ha-Xi?~uL42<v%q(K*XKWdA9qE%s68<bq#E#8$9oH*K*$#6gbxilO}l;$G|
zBfwZ?M1`5rBoI*$7#h>q#5qDNft3h(j3n+s{V2AfqL|2>ghNbbMH|b0ZxU7uxS2LG
zBmItYQ@cJE<;TUHvhpO`M`QS%B<2pW;m^t!GaobPURtBWAZt*U`j3vGKI1td@x}uh
z!Zk&=<^=l+Ix*6l?1q?0wB>bocz;M&5}rS1=m;VFP3o{s;3I%$S;!Hqcn1!CC*J0A
zLLxYH56FxC0PQS;3h62mw>RB5BQqg6N8*!Dj4k;s+Er%K)L_lioywd74**v99GNt8
zpS{zE%;S+Hox}o2l9GN)Axp*WnX<}3-qr$<sL-pPxN{l>@`jeMym|=ve3`oRpw4m`
z=-@VOn1*t04-Z6A`pvOr{@=YGT)-k*L@e|Wmyr1^6v2;?AbAKej|&tb`#{@}h1#G1
z<5|zR1d^|aw4i*X4Wu>Vf<;^pxP>gGK;{W^TTLMyU4341i$}nJu5nQU52$Vl|I(gH
zSEpmOWt0*J$`_?X!_@<EDUDchtGnu0yBhW0)G?Cc)^`>3m_!N5&Z*T*rT{gWU}41i
zV?0J-7hLh(%oI{RyNDZI0ydQnIXnV9pwooZ)kOG2qbLyKe;Ikv6?cJd-f^{<#Ux@_
zt%Af!Zn577uo~Im_L#_JG^RzC1sqgij*<CQ&W3;q*3QYCeV^J@^73pV71eAH<y`Iy
zmbEk}&znMgS_Q~_Cpk<&WxACYF`#&hh~+4HOYlJH2C)JoVGbgdUK8FVd_ko@J4Qwa
zeIKHh5{3}e@&*!+sU+Xv32w^oY8jn?I!G(R(Q@S_oF~C`4RoTyV^mVePI!#M?Q#pd
zMbeEpH+_}8gz?y)0E`Set9aW%jMsyzv+CU_h9Pjvvk^-Ux`|Z}5tkS*D7DtxSs;)Q
z>_-fUzhl!4GXy7bI8JQCL2T^~rS%NALwx@wP5vXQ$j_5;c}-whI7eK*&Hk2}{1t(R
z#R>aslzpAhOSg~`oc9`#N-lZ%VT2x>@<?Wrf*P?OQ7RB84P@9}K;Wjjr|kp1d9IM<
z<|Sd9f&4d%SqZW0rV($XnrK<tEN@O^HTZHs8!N3GlqOZOv~e6m-?)&M;3)HN8^~7I
z*XeT=b+xJFhw{^x+W-7d^!)2fU<So!(Dh#8#LY=zG_$v`2liV4;=mHRV9srDUhX;!
zA^HRX8w3fn3vf6<`AxWwn+~gr=uL}tf==kbbY$P4!TyK<Np^Ocz*z!I0K7n<?tp7A
zyG{V+J4Y3N2VhmjL~(w}j$k3pi|qK=T+d4J(Lfi91mMD<V-<cAJS7VJiJ1^YW$|y)
zIX5ti1dH255J<;a&%qRJ$o=DE%mWVpRNoOBk?<t}CD@3Rhwx-_`Hz^e*a}GiUJ`~B
zqMLUPudaT}@jLkH0AXQB60iFmS{P|(D1)W3g&b-oZa3JUqgiw2u$N7}xr*Q@qXlve
zA2#^xyGUk!6PbCLjy^&uA?N;>Qh!F^ZwL@4uzw)%j|5TzBW4+zc+XMk007@uNwJB9
z0c$P~R;i77Lx%@bHyV?gE>{|p6Gla}Zl2>hm>GJH(L%)<**&DlmBrUQEkZ>sd^7e3
z)PRVDxzxZApQo&Zq443X6ZisS1tte#!STw-2pJaOPQ=f`R5}en94aC**6et<QS4iU
zeXVojb)OnC!Uru*Co^(L;b!4uaz(PpG06?<g?tEnDs!-?71!+s_RU;!tm7N7q!)TQ
z&kR2vVRh<nP=B#_(lJ?Znh4c8`|#RuzWFy+G|yNrbKxp9M7+|9f*=Zor)t#@H^wJd
zh^W%1VNA$U@f8Ty-xOFB(Gqb;2dAB-=&W%Ymf`ji(Y*Mw1+i0T8v~I37xapJ6#6LB
NN)=^tsqwz#{|_d7jqLyc

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_45_30_22.cpython-310.pyc b/model/__pycache__/cldm_decoder_45_30_22.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..e1cd31c82ac92b8711ac4a1e104b0ef5d013029c
GIT binary patch
literal 10517
zcma)CS&Ss-U9YRFtE<oHx%cd~_i7(A_S%UF-mGJLy$;EG61?jqE)tb$x~hA6rjFjP
zs`qHB?Ig2_H=s$l0zn*^fe2wF5Qqm3iM;Si#2XJEFGxW`D@B3_1jG<8zyDX=Jw4-1
zpj!3s?>_$D|Nbf}mrV)JKdyYmJN}X+eV;PZkA}?0@WzTPNv`C|z7)uo94HnFkXL-w
zQt_|)T9C7HqcYvnMP1&?kNyquUkFUg6m2xW7?iA1P`1iJ#i|5Vt18NKel3`>W<*-|
z>%puw8_ZdA!MrseELaP{qO~aM^8Qk=Y%Pnl;U5W(T1SIp*0JEYbzGDc{1d@R>!e7V
zej_+#okF_kmi$M8N3BO?=~c-syOq0=Tk#gRRqHhJRkwzG%`-R8pyVt{X52bT>fSQ)
z=UnBQ)SSIf*l8*}|C-b5b;Ax<UkvZ8HDzA>6mvc1xtF`Gq^a?ID++Jgq04o@8+s1o
z#ev5nJBZxA@9}!l4ZJw<dbSs=d#;OtFhu=Y&q=zDf1NpD+>Th_v1X1hULUo6g_>UT
zZuGs7T5;obFMi37T3g(_9Nh|IC+K15%8N-t4WclsE4;DhC9g%U=VONIRfj4sciZiL
zj3!N$SNy1BC$|&ZiS2IC;ktLb8z(Ve?1pjT&`gPEw^<a}*ipMn>pXV3>vW>f@n3cN
zaolyn%jm(n>wTIP^<^|4c;RTX_dz)lm11`-Y(OlNwFr;XkBZF4@Wu-Oo@B|cWGQaW
zRbB0_3>uSxeVxy~9EAyseBX252uJ%`yVBCpXZ)0rFz|j7Z(PCT9ch?Pq+NMW`o=S|
zlqji^N}CwaRd&>&kxIK-YV66wLMrd(P@^Ml;y-^``a*eFOpB?R7E*OvMQMK5NX<Q2
zmTnwJzd}~ChnmlV88rO1Rapv+r=|GfEwrNFn|M>rmHIOKB-(4c=AQJNG&GYU<|u8;
znC0=$E5j0I%EgvwDY14UZI;nGPpvx&Az?=unqhUiJR56qs(yN{H$I)3*pEEk&u)cQ
zeocBew|1o|TSeyeJSVZE(7(gxfhwGk>UM)>mFJR(wKjM@jO-3`Tt>*`6;Q6{+pP^J
z3_U;QCQ)!ky<AJ~^gNz(#b2o#_WFsvPJ|hA9bDpp3ra!Pb$u@@=4vnM@q!ojLE#`T
zUXJ?7R429NBwR&Vv&^;F^V?j$%?o3ei3M*Z-t8nyfEsv>wqauqB2uCcu5aCPSSRMn
zMvM*7Lb*yn-r)L(3FYp)68jJ~_ukL0ZbX5%8b*n?9z|QLu9G;c9TxR_tJnJBRd8o(
z!wXkiK*1{U&PzY@`l`UkYRh*6+x4i1+0TB+e&#9r!iAOI9i9jOVg>FpQQ=ttO*Z8k
z-jaArNFV#5Q7FlEr6e!OCzOg@QcQVp^dSVkG8(I?t*znkG;Y{-=m7mTH|-IZAziZV
z8-2&mYS?p_Q|xXEe3;4>0haK_B-;(CB8!p_@FUgI6Mqh1%#Q`RpcK@CR><L}iyx)v
zmoFIjsRAPJR^O4@vRiP?y9IDBB=23x(lVKA<=z3SSUNbm#?8@&A*C2eigz_Zj`tO$
z@6#gg6a8Ih9|gQ}{{bFRzt%d2g7I@6i3(m4BJlvd5@}m-Ja})GRnvLmzqtAdD6nn$
z%W{@y&2%<iMwvoo#ACh^sKaBB5n4LGEvLt}6xK{~lpfBfRX2zHYFZvv)3QiEpH}d%
z#?NC6U5rb4g}o~3)%ev(xp_~v^;8DTrwX8vs(^*WOmj(bTLBZ4QUf#QI`W)^nld=O
zl;*bOtb9vlK~iz^VvZcur%OVjyoLIuNh>9O>l4zK<$I7tkT!pml<rAXcTe6dZ_8Us
zyhF7~HL1NLy(703*KiAWjbtWkw=Hi25AnA}o4RY>lYR>dGt~i0zk`(klR9ReP4j?r
zsR1~j762F0BH&_LV&6@cHkU&M@CZtdre(lmX$A0jS_M3jF3`%5I+?0&c}^N0Pa3<Y
zl1EZCJ-)3V{pg;A+;Ph75r$5q>`ZbtIR{y&q}rl{eic&GJ6KUi-lMV3Q=5D8ZZlQ)
z#$}HQ{63zl9eH>nRfi|t>adYKK{$rI{Rha~Q^QA+8Ny6@Y8hBU=~su3CQqg(F?&A6
zismHDSxjs0%uXJooX*OIC(}l%-;{nyW?vGguOM|1y~;<V;h9iJs)AHGtqjkki>O;m
zX92HfEkc>flNCBMlzuK<qS=RML&dER&n7E$UMO#;kKU80W-FWL^zJH+5+})17^j$)
z(=%z^o!!ZW#pG!k8Tq+%K0TWrO^fc_4z!sJD%w21yPBL$Pk&t+<saK!6?i(i2h33m
zP1<}sJ^7rZNcQ9D3Cw~T8FSuPPETc&=T?`bEfs6}etL3vZZhT!O1?Y67M%`ZYTjMg
zF~WRu0c#@Nc+Oqi(YD20={aC-G6Tjrw3f8tE`3BAHV&={E2&Jd{I3}fAA)6t$pXs{
zvKAvOpPj5PTix=Gj#f&%_<vyW$QX+s?7vMk`OFxHNI!dUy{-h2c4Tev`jsg1Pn~MK
z8TA{G-JK|DBvGRqw*0>9HIfal0Zr?LWPvrJcH>UeXN?w%;`m8tY`is~$lNX&RM0*v
zD=ULHA3A0~&c=%~%dwH{n~pT||7&O=cA@^wt1#mrjbY@mk3o<Qo_VE<wXz2Ew7uau
zZrs3#<IKbqXlyt!ZExLc#9lIZyT_uNU2KyYGA|xk0u3)43vHbcs7M;;JASn8`2HQ*
zsxUR1G(I(G_*ltm(uY0UIN$TwlfHKo_G{x7#t`U(t?M;>FYF+X4kN3?4*Gu5A|*0@
zC`~$H7H?b!0Od)kv@C(Y0Jn0mhh(zeD+gEGl-aXr{jS32sA6&|qs=hVH<?}~mCB8a
zz7wpw&ZPvom^4>i8l2fzY%8|i4ZDeLuUv##<oNL=B(ffZ=g<t+bH4?vi-=)jt~Z$1
z^13%YS77q+5<WV3ex-4Jql+`?5tWEjxcbVqYmFZ3My#9M0iuBE{eC!$$QLGh+!zd=
zXmnS+6~M63a$+xT+`>s`hqm7(Gva=(`-R4(##0U0_l*mWJ=Sf&PNIJM?WTr)V!s&2
z<@=2|AxK>^0$E{~fS>zxDngOJEAxEM$y^|fmtgfxD2r=glbBaP<S6r6TfGQ`!{;a;
z1C4+i-F9M!aFt*dv8x!97sdn{^O<pSf5b)V8+x~<jx-{_4SrV6^?Jz$FM=A$xU@01
zHRE<;XPqX36-s?V%cK!9pH578y_sX5pv{oKAxhY5v<}UMPu5hjV$pysn5KHDUGqac
z7xzKqWYlFuV$H!A?KoIRdxaQ*5@5*}Y5WsnXet%saI+gY9W1Xq@OZHsPBzNOIznxn
zd2ua|tAsigJi}fVRiAPEzV`}a5#uVjkgI{y<2i_r4orE_5L(Ya^n`idWl<03#7onB
z$IDpfP2%{>!4rp}0HOJ|A34d>A08#1xxh}+PSyx~ik4CzA5(1ids%O5t_N2do;2)v
zSJ>os*M%`3hTPoufEl@cT)Wo2`UX2rNfUl9+-s6!jN%$BN1#DfmB88ZY~h2`I=MMX
z##Y&G!xieY43F$Ibt)2=A#jQSIhCwHphVyi0A8E+X>4*=&S!$*v($vlW^oEuc>~DY
zKN&(la|o;xXc2G;ObEy_h{oEzr$4-pCAL9Lh_b9Ia4a|z=prIxpQchl*&{VzPpJWG
zW-0?$1vJiHrw;jW<i6=XQQIAB@i4JxVT72yLd}=PwWD3{bHwsKM}UH-=N2?;RrX!Q
zjH#wPM1d9@Ec{Ic2b>dvo@*q?aoP;g6X%Y{V#@$G3LRt>N9%Cy*x`C6E+9h+w%zib
zIJRwe8OvqD)p-M{XYj^Xp$N<q8k{Zz?iT$Fq%^!#Qj>K>m+5E9b-AV($Q$w;-UazI
z$`sjrpy#p{I$Dg|iV=`Q`yXk#qRh$41LcAKU@lMNY53L9T2T~?prBp$ul-PaU=&S^
zYl<~sL^yt?tSL)ak&@Tn%i4)P|E=h#K~L>R+7D~xx?I7pjuJ6iQU8zqwpmL5FQsV6
z+5_d^wFg?A_G6%j_C*tO(0XaM!+o^(l**5rRmn}a*51>9P&!_h2je&lytavQYmwp`
zKa?hZA!?6F7$s67Bld#GODd-_MPFQ%toUsT>}YF37YIv)eH^VJ%>@JJk<Pf7xC7#p
zG%q*CVTi-NZGhSW&RXc;WAb46<t!XP%RDFmS|l3A5Aih3M03R70!ZL!V5wr62omWC
z_9!U@hVqHJsbK>O+@~xkBO|<IOofgLNgh2BR@y9VD{w1j;cS=)?obdZpK1sr6;flD
zOwv)FOzs<JFvF+>;iRmEv|GyB?7^{t`(<2{7&-3EvIxlt;UeT7JAzV1K?b28gnZ@d
zpxYCokA+lmiU5f%76CL%nWSnW%{7uw>@li)oWKf!rwBX^z%}q#%q~!#c*!bEB<Er5
zQ^=u2*ryO4#X5*kOUeS+ZBDL$Jv3!dpE{9oSYtjr_lxK?^7F3<19Tr;!IQcDYu6D1
zxzfzDpF(T)5duF=fLNW;@r%F*FOS4Q%kzDR1d8%x;5Ma%aCjam+Eq-Jo*^lPX+r-=
z>B>J#ZyTSjsf<P-@qOSSTB0ZZ1AuYd#FahO)4W_;Sw^_Q(_GclT@7!}%e#n`w@V^g
zl6Q@}CCEC8ngG$Bxm(<m+@qqd<d*Liwq>hC4lruVqgptqci}c#l~MkfJL}Hf1zTCw
zQ3>2K(W2%qx=ZLe<1V{L@Lv~Em|3vj@ktcs>gCt?926wQ&}<mxuq}{-`H|U0RJ!He
z;f2_>DZuBs>`82smwkk|$y2dKJWppH)M<DiENh@0y;Z#NTL2WI*_2bgE2a7G$X^GA
zLp)&Jsbs4ZqO~>RK|7affO?A1tetn^;JXT3bICT6!Y+cSu9}o~%iy-kW_6E(8rmdZ
zqkIQJjXfH3CaI&<3<WZTo0ePq7c2tLzrAq3j6a_DLJ`K}CZcGhuwf=NXW4bsWz<Ff
zgRpW0azy-(eTiznOn~sfwy8;Ff<LlRW7egbO#%o^fuC9t;)90kc|DPuX?IzijF6p;
z`VKX|M~zDu%e6-<V{<5K8X4<;k=i_LdX3_q850X)7qmZ7hqw}vVv-_t(;O*19sH>m
zY80J(zoJb<J4oT-qMdd@uMq7|AOYDWkw~8dZ7(6%FjNvq$ZdI7*^`E<kfaEwPztoB
zu8B~d;ATkqucJSk1%T3rsmDv1Vc&MhA3{CJdq$I3wU+IR+Y3AJobV>DRE%-S6o1o3
z`68muhzAqygbO`J;0v^QBEmYhOnP<j$jce6(`J86r?nPscKi@~PssRN$c)GsB(@1+
zOSRvVUzb4AIU?!LLv;-Eq#jVKBGQ9<1QDA%l@?s6n6E$_psau!2x&7%cMyrbn@63z
zr%?H(mKd7_+*8PO4?rb!GeL?1Z4sMG^Ax4&$Y~L;iMIguP<OO8AzE7-)Ylsuja!XY
z<H<{nb^IX%U$-21LB#IZ5Wj&#cU8rfEus(P->_f7p19)QAnXEx2*pF+5EJl1$4g+~
zAyCh&(=-c#APA8wNyK&6yV-@ip?lph>cv)pK{5!hLTD*IgE5nCryup>(E@l*I0)=l
zX%2XNxIW9Y6VI(X2@FGdi!19f`v&R-S#zb6Frm5RmLLuG>r_PxXNM@B4(Js+bmE-T
zWCG#f=|`H$)K18QxhVpe?C-#dlO00F0r@I1%|ymQf@o=+r&uYS`zFpg(Qw;>TO|_k
zIK4d5IB}e|#LnQPzz``eL(v+7e<uf;fI3GmpUR`2)EYfSyWEa8EI={j?@EaQBVc4q
zAP!v^k0nZ{#%daC?3-9!W=Ih-;FuDYrZhY8kq}g1EK!vns6P-oQBg&Ne+{n?7#%G4
z3kIwi@HcH|PWol#2JS@nnGd&#%F0vhJLmzYmW1DaozS`5yqGz(LGRK!9Sd2BOVZ%6
zsYA%PP{_gQXo>KH(XDB+H_?fGTTCLF@v1w%&!o!`&!4h~gjD_-b@+J#zW{&{NT^u(
z+Xx(V<2zhVNG6BwK83M&sewU-bTx{bqh6d*+P)zn1Wj>jwQtad$P%l7%4$l9?41IP
zb~4uAU(kg0v+(eZe!y~`h}Rm%Bk@X#FNJ`m8@#5n#&KO|EfKv6#p{WCuSvLYY$nTV
zhw*=cy7ZyvMuO47jovs7?cEulNlh6vr<VGYd)>c)MYf1u#t*R!+1Eltd>084l92zn
zk`ba1WDe1&4Rf%p^^9d82=hn_CP3OiS|f&7#<hc+%Tfx2qOir)6vEQg7bQ1;6wK%v
z7bUQR>K5=X?W%O0J5_6<l-N<eh>IOuX%NfOh_I-3)TwqA>b<FBB*m@nDCki`35n8~
zwM@hSJtH~u8<g5-Hwx|Gs_{mqsjAt9;N&v0soY11U1EQZR`=;dA%ry%ve78ggs%xQ
zQlsnl9DNmlE6ywq62s6XN@pa<*lz+@^=xo^N&+()(;~A2PO-4r$VM}>Az-I<MkLR6
zsa<JQo=v2pn*Bb?xjYb@YiZD`H-t2`a!>(Ia+rur`eH!Do#HJb7>Y^hpmvh0N8|<J
zHXCvSbWc{a`M#zN=BJ2hzxM}drht(I&3qgQh*{XRpc;8o{=Ak^3uuH?C44bgK7vyv
z*w#QTD%?sXg`$O9DLgYbx05G@iPO{9*b5ks{Rx4u09g534&uokbe~o2MKNrIo1Trh
zb<j(!YKUmabU~rj?quOcLa`q)p#HW^cioVp3m6V(w&5UxH=wkhVRwx2zo0q)k}5_g
z%DB8P@GLwkF5hCrAMCFP9273>uTl0EA()QWzy<LFDJ>yR4xx1aqz{O2(pnQX{552N
zI%z1wI1@T+SK9-EN4Z><Ta<+H1~%BtXC=h8n@0R>s)?3`O>?uD)!-`zZK|}iUz(K3
z9LIr-9q8Ps1cy2L?t%<@y-lCgsB2B7gT>+brx!Zk`#XAm@B)}Z@fl12FLB~#O<30K
zGibwh0K~B+blp8N)VbMn7%l}81d9-$%r4~ND~)dEecY|^G9pJU)(yI$1KTpQ+W(M7
zBx%pi5I9GGWH`@JSUup{i>?!Z2}xtHzXh<$Vxo9->5o7q&5P{x*j&#_@exB0$_3!U
zp~DqU7ThWdI*OSPz-95*=(OvYMF;!TV4g4FR~Ipy!O1BW0;hkV?*V2~F$M`M!Azt~
z$mL-#AtBa*D74tnE(&--SXqd#-afp>>P^S*;#&w@FF>w%y>HU;NM*y1KQgtPL(RlI
zxIy_dv~12DcGRgqR}r9Pv`DVugA1SiE|Qs(MOI*@zmHN%$h|+N)SnXg8v?`^?0W?M
zp1@lK_F3mPWl1@+w*l}q*u=qCOhtKLQI#Jln({x0YazC#yr&c@|7kRI9dW>_j(C}_
z!WqK9Vb)4wgywm!gV~`S87*C`n*AbD<PqaLq!ytk7OoxpZE8T|!#rwWh!1I2!ce$?
z)+u}|vH}AI5$bqlVzP`2a5Lj)kuHjAGvZwl=CS6dyG<hUA{cF*pRW7BkP$w;ak`mF
zL@GH8YLf?(#j`6&Ssxh<fscCj7q#NLy}-UPk}GX|!ItzxZ^TXG&s-R}@|HpU#o|d!
zWg&ASVe8z%wc*4^Ux?8>Q@PChE71_~N-GM2C=`yaRY5cxAB-W+OCPo|A#26ADqMd}
zU{M5M#34<czn}zRoEaB0++HG15Z~a`B(nbvXejz$QcC(U{AZdAU*stDBhuS*{{I5m
C)S^NF

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_45_30_22.cpython-37.pyc b/model/__pycache__/cldm_decoder_45_30_22.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..a9e17e7327e557d7a5a6832dcdcb620bb269657f
GIT binary patch
literal 10321
zcma)C+ix7#d7s<P&Mx;tF7K2?jbz={u_;+j?4(iT*p?Nim6=A8bD74S40ndhA?LC>
zGb@S2Y@C`(thS+}x=n#1NkKj|Kzc)grUlxE6bSlI6zxB7`Vyc|0{T#(MIV9y?e9A?
zyIj$3n_0}a=YILlcR%w=y<U~@{L9KWyc4fT(!Ww>_-IIc8gHV=lH^LR97v&T$)RGQ
z0C^=)EfxQ2poJx?RFvtKF6zowx%fB4e<iG1RnbNZCc>Ik3+q-rY*>x3X*ET8DVPkW
ztSOP!gH||gO@}kqOgL-JhI7_jIB(61x^l1(E?SEsZ3Ii<G3!`(+&Uhfuuh1wO0XQB
zv`&h2H88_d)+wYX+*)ute9C%CmOdl7b+_?AavR?Ku4<h@zUfXPKj~FB&!XfUN~YWv
zN?P6`^5<RUmeihpOqgjay!?jK>-kZat1m~}t8JN2e2%#u^W1BGCv9uI+=-*RcI0wB
z@FUM*d}7CAu^q;4Kk#@f^+PX7y`JrbYo6<3APmvE)pJtc32rkdO4ef*daPaI^S6t(
zuTj%m-krV|Q7dlT^paPCxU<Ep*W!Co;)Fd6-FP`osX-j&b(J?)z4VRP^#aV$yx~yg
zHGh4*pP)%w<&7Zj+Ufn&b`skUyIl9~`$?Mcc|S^0hh|DWd!5ChjUBD~w9ez#e5V^n
zPVgD0pCrB$T|*D%uk~qG)Ys8`$BT;1J_hATREphoumQ15)*(C&pAnVKp~RCc*_ABC
zExD?zJ&-|PGBB_6=~v?@WpNOA?we7ur`78ny%?pA{|eq;!7IqMD-Fu2v@ah>-~OsB
zrAnq`(k2FTl|6M}WYWHt83*#9lF9oe)aXc8@1H@cd`)_5X)uvZWYw&ask`c+nw9sB
zta>2J(w!?`l+d@5*Bzkli(rJ$OE;x2NtxzKTNU<A)M@+G1L;KxBf9AX#;oni81)NZ
zQ3f@PTuOc*+8XS~skB)~`!cofDzxQ2Wl)Wl$IJ84=Ev&C*LLT9tf-1T$oU?yx(2Nk
zNGq*g2j(Z3*Yljzj-z0k%>q3rQ(Zr7H+d<IS!aWnquA~;$7Kd;c>`4H1$Jk{i6SpZ
zc$KI%r&O+`+dYq$T=7@)qh3F?*N6xcu7e{yP<$=)T{rOZVy^b$9<O*&AJhy|;`O+n
zj&)KyPRdo3wd-6<ykMQn_jzT=9*N+i)VrVN37kHkq;1%kg9wu7f$LlM9M(;^vXNjz
zv{0@RkT<wqFr4iFP+}j$=6?41iyLw1U5w(?TZ`kZi>{M87rQL(_b%S*M>oKUoeeL#
z*Z~SI5}&;C*_#&yJ}!0wH?&=kYMA}ekJ-;(vM*m=>232ecoi#f7m4!D07$YXSLI2(
zHSuf63yLAvlo{j}lx2BRsmVLXK7^cCijmsd>MD*-<A!ZV4v=s2s$K9C(ly(@(|3Zr
zhP{Xh#oori3sg1_uz)vF01R2=pWr*91$vSdfDzv{-<8&7x8hbGRKR<XnGYmO%cZ4N
zdRNjUO9yF9@@laG$QMRD{zFX=*JGOPF|qVxqNdyIQ-IeW|2G~{zuGyD0(vgsZQxb$
zChwz1D(wn>m8EQ&EoZaDP04bqxbm+2j-2P&&1^ckj53AFh+hIV)CR}F;iYVLSI&-a
zDePvdQ+hC)HQf^Ozn;|x&8#lcce4im)#NV5(8aivSJ-=^UQON`l~*6ib~%#)jZ6WA
zEC8&6pG#?NS7AR%>zRQWOI>+JLQMl)ThB_na$dfrGM+ZwvY4Yp_1S`u4?jo!!l;#!
z{5)0Oksm^eY*ik~%u4GIC8~WWZ#H)2E#;9+ICw`+C)25SrFZ3Z#Wmc@10!waEqCQz
z-~z3#iB{8Y^`Z1#D7#DttbHHr0X}uiKa-UKXEOtEE~@~}XA^)6Sq<|(NEbJkA`S2u
z%8qB4-#(Ew0GG2S;K^)`R*00DscwBn8k|T^?VnDc%GB({u7dQL0|~hklsh0yokiKX
z^n7{&@=eLKc?ta*q-OW9rmlQIm~B&=hw}c@nR+lR`-s5vGnv|z2g{i{IO#SA<|Fz3
zPts=z-`Au^@>@TGBt11ajZ!mh5vH<Ji@=(ws|}t?SF)2>NjW>6sWXs{ke8G0)Lt1Q
zp2^DwCo?nCxAaFUn-dhFJW|-V(5t>A4bDb2q#8)ov&P_THjlc$$fg1RG;a|pRGwa>
zb42OiXA3m{;9R7+t--nUBAp=0f0{k@P@<YGNz8L*{}PRY6MqilOl0-!Y}Rt8_e#-R
zdYMK>ekPmE&Sl5433p}>`bh?5ZMOF>rRTFV?@2}eBm0*G&Q2Zxi_}7sHlN8(z9=b@
z{Y<uuSx_TOkCZ#tpc_u*^yoGhq%9Szx{#e5oF9!kg_XTO!q@o%SF`Tio)I<D=dmi%
zb?4prJ#AOamYoOgMsr}CBdf_wci|J#z<hF5Sjz>hMerJC{I493AA)I(EdfUwPDPwW
zf$8(373OETxTjOA<of@E$)zDCKiGelX!N5)JR<$ECs*uB5OYhbJ2$V#ad7IC`Q^B8
zLOyrn)J$X3k2*o$^~`j`Goeepi0mmdUN^VnJ~KNkPLgMxp($oURk=QyN6;@TD=Rx+
z{?IY|Nj_eZ+ks|!Xb95GztGS^6hpyRZ@?&ngoeq+J`Dl8^ZaW**2+w1Vtd1L+{DC)
z!^}w8n;T9-+gtO@#7lSH?y>l;k8M&z<|T!$H@$o;v~?n&A~i2`gLusef^FKWFdW-7
zJ~c1{tmI<ahjnUR=y~kfz`F};)Vzl=1o~jPdS>87UF6ZBusiIqAEX^pEcDPn>4|B)
zNecjU2Rm$00)GK+m2g%v+2M^RSK5}@N6{J<`wUf#PNdifBOR3MO;VZMxDq(wn(JIm
zk^2w4($3jKHMe5he&naNy>bO6j}s(Uk;vQcyohG7g8Lm<R73_Nv)g1|$Mf%cuE62Z
z1$=7flPl)!4Iih`Bl-~dzwz3wTV{{>G4s=Hpa?iV?1xi`17T_>=FZMDroZB?07hoV
zNxa0ohcnKP>#$31p#7TvmU-2@WWr82FF*aXZ^8zmeuwSGhJNO-7zY&uW|9)5E;(te
zu&cn%V>*+eO<C?axhtdb8Z5dIO>qqjlJEw|8)aT+s~3Z0_zdL}pb>DxUr+4_E)R?z
z_RAQPSB9jS@Tp<)aKw4)8+rG}4m2kJ3;t5c^?K<Bp8y?_;b>!SYs&Q#XN@L;g-LxP
zt4br}zL}WtcDux0r_Fqp04;%$CxNxYw`wbSv1mZ{Oj|wDuKfZpC4EpgnQS?cz!ggp
zkI|{V##qd(85o{j2kTm2A(o&7nDb?tYFSK0rD9xO^+TtN<@!4wpYWs6#u?d8s7-S3
ztmAQ&P^p4@7<_kBeZdL(o|uoT;76{8PLG!$NV+iFK~rcw`=KY!%Pxz1I5%D!=RjV^
zI`0zi=Z>2=90dq2uLrS{KKDYAc>XdwNjteg;0;<zYj|vl-S6eSt(hKNWO&H1-CbdS
z+rA6KJ&Jhs(6eRa(sAvUf8$Mdf|6DEuW*}5qA`k9uo3~2sv4oQ<=Miwrgie_D4AGw
zdmV01pXGRDWSO%G0#gJ|5g@0Lk=Mv-1Wp6+$#MV1CfDRbE+)P}O}<EgPQj{g0GWp;
zL+EE;CEyTPBhVo*A}Grsnp^KZ_rf8TSeKd*rCCkjSa2%PMMTF|sZ>yUp%Cmbg<wsM
zW#E>8=Gm`OhjLW7YI;D_wr$NHCH_1-kg!*&`NFWa*!3ZYEFN+QD0+NuL9<rl&@IfF
zY|I-JXu-k4w^VSzB_Z;;Mj{=j%@7rFwmp_u2DnpbBWt2qhifMe*K_d!abB_QPT(Yo
zZG#7C!TC`L|K=95&*M%06F^$7!t0tOKTG@!q%^!#GO6ft{;SF@Wl}bfH{gv`<puei
zJXadl>hjF69rczC<Ql4`DDtv4BgW8BItL$4k@KG}{;R5ivG5bUbfqQInzA5k<t3!y
z4XTrJOK!j?Y$1nU6U*q?(#TJ2VV)Ufy!UYQVh$r8Lm$mb<uw&P<S--x7i^@PItpO}
z5B(Eq6V?)<*i=eogh3Dl$>dC?K!~f7DZWdg0&P`@@xpvypFw+wV!>@?q;r-dQpP!u
z(&WZ4K5*3S3{YEvz7}q$B=0P~n)~Ur$S1kALwrmRF(1t$VmIGIB89Jjg^EQYTB9ST
zp=1=HDW~eDhUF{pg7R32j5v=m7Q?CB|9DzPUqpB|E4vD!F6ZDYR1tNdXihoP5YMS(
z#y;7RMV@TsJ2zmRO8Z3_;yQU7X}^}YI)HxyKgs~AL{x>5XI*GMMxKAWE}|zwHVCoB
zmZ-%!0wmsrc;xDi-xHFHg;XYl&}m9V0PR{%-EE|~M&gA%O?4k3Kt4LVMBq69u7O1o
zcA4_T16E}u_>P_j#RqD19&<=YA}j$SL<nL5j54P*Kzuu`u`?RRAZ7(Euq)_PIPuqo
zC3y(0<H>#c)!T?yTyK}z$I+U7g1|2kAQojr5F$#!>jgJ=ydbdc!-c|1yGM-#TYnNM
z+E+rGH7YRuexYCEZRwMS%BVMq-O)pIMNdL>H4Jiis;7CSb!Ab)zm6XTs^#zLBJ=^u
zdQgLC(%ce49~G~<KXD+rbE2;78V@SFvQ-lS5-5S9R)k4V-YD|(<mo;DYgx^r1pem(
zc!k!a3y%-uPPsGgEdE;}q%sY<S{Q{?Zd`i<0nZs|GzxIpu%}@RAP0LPH*BbM%iHFa
z#I-5<=DF-yY>(Fi#G1*;utl&<XCCxlG>VZLXhUxkZ?XeG5tmIl)3fq>pl%2UtT>Zw
zm13o~Mm%VjG7V7A5F@qAE_`%Xfk!RbMq1fNEW}mQ+I}6p*4S(wP;5dQ<tHiMMQq}L
z#+*u9Xf;LA2;o_kR=ZdLo_~MoF{xo*_978m<5dK=NV7tpx2M@n)a7(UX1_2;1X4s$
zj=fE_UnfBOXTJf!8za1tahb3Us__X>RElez7(qM3^}L=)O|APZNed+BqduU<-=M}d
zjOE(J%GeBw+D6W#U!yi3HoZo{NWrc{vZlaePTRx}xRoe)vD$#xA1a?(MF(##DI=i{
zBKV+CrybBMg!(f`KxRoK%x7>Jvw--(KuIATcjbNMKpLn*iXs9+DNvZYDxz_@*wWZ<
zqd%JlfM$k0#%sAXzV46{1L;q4p3x-MWXBG~)qx#)PIMRd8OE^Wkc-6@`8>kF2+$JV
zgvUEX-b=J?B0dQkuLixhbNbbszG<WXOQ+QiZI&KV5hK!lADM!5L0X$o2ATGUN>+ML
z{tA@EpiBw@<tjol2-=CD+m)>1LaBTMERV7Zg8$NHiNgP>zF$V2e4tSIrj{C;6<iU>
z6p*J91pkFvKwE^=vT|C>jINwb;H~1VfFaa9ZJp4ot?sne%nkFN*)gBJYOdiA8CWs(
z&<kU>Z9}w%4qXpTbZikKAXkKa8@u63aEEXTv>_r6T|#8fD_t*z?S*(bZ;n$bSYMDH
zSJIg4u6NgmC!l+N6!#LV!XN{L!7nrrpTd}_-|ff!q*wqi3EQ1LqB-E4;l?W0JiN5#
zq_ElKEv~F3>|3Z8gv^z0%7j*uJAxqC?@|>l{DU<99hAOChfUlfA>3Vsfs-3$&^sCq
zerD|8M>`yyG}#;^v_FS8`8@zK22un$lLiGi>0CFpfu5H3b)2od2*StNm667I<E#z#
zI?f49kU^-Ubs1MGrGXK>jx^|^=tphQPqZrSX@iQZyX6N`iW5h>Dj5z4AXkPXiPB;u
zVFVbfjHoa-ngk+B0z+dOn>a^^C9o19kCDVZtRKZTR1_1LlW<7LtmtCdZ%x8#0XNe|
zW~ARy?r68hqWrkHQ&ujq12l%;Nn-8@8~&_%CHFDI-qkfa46+7wY3J!N)aN`WB;I&H
zL%62s)}CNrLMKLglie0GiMG7y4(|`?O2P}q3>_hazeydo3A_u?t_nF~mG8sB_mgcd
zrzC<S_mI5U575pssF1EAaeLECaxxQ=b0j|b#MqMGq+R7EO#{|E-Koqe@Bm<i&yh(p
z_u)Hz!aN>H(n&0UBq`~)46;<*o++yw<ZUewi3+{yi94rJAa7_1%d1C_&zGr7AL^`-
zfevodhG{70?(je)Wv4y1%>TRB<I7lNhlqt9;u12Sg(CPd5+n~H=5c`{WFKf7vQQh8
zU_9$Nmq7BBkrtGXw1KonT(F4i0k@Q=6v#YbZmSujqpQzLZuuDa&o!<{-~rXG;9uHT
z>FRW>wu(~XK>3Q4YPfnJE~OD`ZhcQ3YuBLOn>t2P+~%Hw9+M~`**UeE%M_p{7c7ie
ze~iZ{?1C%4JGnw?<`;3JOTf0`lfxsxV>(SpT}^~fG>QTt{+E*%U2&J_<{ejyc}yaK
z)hbAw<QDsV0IQV`ZjXsvMq^rJS-?RR<`|h@)qDt;VBLbu+4rekttihYQc=zJQO@NZ
z!LpVH<#|VlPpbr(@1#cws7kl;A_kOf5wRRcZwVeK-5}OrB+Nmi(rdz-gfFP<%#M-K
zVc#dHrGg;@wY-i5WGcxwc!HbqSG1f?KpmtN;b^(?63&z0x&}H?;V~*1WG6gE;dZ&D
zy)x-WoSVMNUcz|nPXI;+omIZ)Aja!K)mhD6oWKyc>)D7UhrQHlMu<y{7gRdy-8>LT
z2o7Qf#NW5+h8cpBI2<Ro;UKnlpVE4c+abRHk|zHVRTSsRxV$E?ESw`Q-(!DEP5z3&
zlj4N^HOjt5=%rgo3C?>1NF|rN`XoXRPkAJ>NkNU+k0=$0lLj(uFCcJJ-PaC)-Xd4Z
zbMun0%|QN}<-CO0b<;>TGEKCsY*sfX@)~?OppBK*4olNIS=u;`p>JF&N^q3Lw+&>g
z>+AHnin`iXibMJ7OWl9|Cwl(vB`|{$Fz9+Ob&~d^Fq+w$*aQ150C8X`T`(6mIIs2`
zh7f&<fDM9#`2{!}pyDPxz)go$NA#w{{Lqgan2zl0G}s>zAj!_o5;#v_34oU<)E#o|
zW!DM8d>5$V?*Od2m?$YO*%2(Hd669-o9lTgJ{ssjkpNsia;(B{f~Q1*KQR-6s4V$S
zI_DN<kzjF~2m<Li>p7UBEqUkU81sO`KhSr?MkIVmKnXS?<sm$oLjEHrEVe=tfLDYe
zh3Mw}qpNG)bpjt>9Uv?WN#gasLklDA3}vu1wvZ#u#O((Ab2Mwu9QCrPH&+oHWwb!9
z;lqZ2eHY2xZz3}<*U`r)CFI;6Q|iwM{0#x(1ojUE{*gdNV8kp#6Ym8o9RT1PD=9IN
zFksCU!78;iZ|Lw~nnr6<)8$%ga>A&I*6j;i2Qx$OF<PitBYTJxxw810r$eYngm1?F
zfEo~yFqaw_;`5Z1C=x!LbqZf#tia?zEI3&i86m?0+=&Evm`bMsh(kq0#+n`PHi~_V
zu&;GtyzT=-#`vJc@pB`G6mA|qCRZen9FyFzK2Z#TPh}1lwc@(H(7sbhj&*zkmi8mB
z;F;maBdkgN4eBozPdX+KP7|S8=bu~~&bRo+isl*1<t|)}hDcUAaTvyt@Kmh^;>P&o
z3K3QMG>i#ZD!u~Y`Wpg^B3dF2>F~7k6rDA0!*bkSC7Ksswjg#2ZDRn^|AJnVk3%2j
OTB)W?F16m4g8u=YwT*`W

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_45_30_8.cpython-310.pyc b/model/__pycache__/cldm_decoder_45_30_8.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..2227cff2ee84f38fd222b5e38a6c767c453c7e2f
GIT binary patch
literal 10515
zcma)CS&Ss-U9YRFtE<oHx%cd~_i7(Az7i*RvySccIwb3f<z2^dk*HMDRo&Awb@YB!
zy+>1RC!P)7fF|Jz1aV{rB7~7ZARag*^1>?-Z#;ZYNI^m?MS=$e#1JpP|5x2TJ>yNF
zTJ`VmKK|eT{wgY$O$pEUuYAor_M#;HfHKpMhRmn&#)>RSuH?$T6v&nwC>9EkSA5k{
z@vr(?kh5~5GTqWeUEa!%{tfY82u#ZqZ8X0al&n%vw#q@pssvT5D#~+yEts)pL|XUj
z!K^hK%vp26yfq&zSPQ|TwJ7TH{!*}PEsM0_9|?|HM}uS5vEaCMT$C036TwOAq)3~7
zBRFN9Lb~Xd{D*>vt%qgl70E5Tl{=DK@fNpL>ooFJw}yPpGdItm<Sa^N+&W6?-ZJv%
zT;;0NoV`cbX(~Mbs?+Or!wy$p2yd@7WnTO|b3Nv{m%6Q_squU(3UAn<%XPmSdJf~o
zfyW{{h}^#K@p{q?yg2cCwim2>u8V;%MEz>dNxF`IjX7c5j#%KaW{xjj8?}9znqKv;
z_q~u>apQF_e$kIwTim=9-3((V=wax}3rRu^qA;r~ys_pbuSTxtV}|M#hbk|1+wFdg
zCQX%B{HS9mw-Vcl?QYQFx_7G^Cox~_hH>K1Oo?Z=SrpjVQM*g)Ja(z;bfVDlUvc_z
z+;zfB=)t<{eVP^ZWi%gn;b^n>KsgeXVs|ZUKrEBB2#?c`ip;0+#tQ(RWXY~%DQ?bH
zUG0tx8k2#2ozK1$g$av%-*evxNBdg4+|to!{FIR}@O}<&T*2fWX_!x>U3pLX=F_s2
zD5;W4n;6hlcGRJfO1oNW?8(DID(~h{qa$tNKYvO3a(P%xi>a9wQgvHJX@1v8%{^I`
zt{+FgLRPbfnm54=8h+cVEQQ9CQhebiTG8(<ys73&eVKg@?X_KVPkL4wnn@9Jl(uEe
z^5~b8VF@$kVoS7?SUZt6%V?da)*XeAu%iskusU6yjkP#cKfTuLucapTBaip9TcMR-
zmEO;-U2e)&k$FAON$e=}Z?k!z3MZtx-Jn_Jxg=t(4W17pyTcrp5i)rNl<WC+Yr_dc
z&yTrD6r52n*OJ>kkLO(RSL%kneqyf^Va8ksmw4cUQqXl>-^+@*+KYO;;DvoqILM2a
zqkb~gNo_d^S5ek1b1nA#HkWVl!kA@Z!CQ%UE6Ea|2416W*qDQel<0%&TQ?oniMg^7
zV?(r1t`d+pxISV+x%<AvK8DSG@Y?D|6nLv)lz8h=w6*FwiL=^aQNOo(wI5yqceXaX
zaJ2;#tP=0M__MFC3Vf`#d^fONk7}6x%*X7fpRhl%(!0&`U|%f2T_z$t3!urST*F%u
zZwcvRKQam>xvrGtCHaI>kxPmx4~{;7zE?(LHMO-hoSnuE+YTLI-{z)0Vlt#lwtc<t
z_*o5m7ITWtO@WV7*&@Ia-k3zYAys5i@)35VT6*Fy0*u+QAQzN^TF?qP{B-f76#eoA
z13y(j<lX9fQd@Qlu6d^b=7r$BBUxG|a;@BZ;1o*-L)W-D+AxF^BSG=LCaCeAg7iIF
z#62RvYwS~im+#%jBkI>$$51eS&LdI5OEM%LpjRSo3x)^l&9Z7bPwW?0KMVb}Ex#*g
zdDcv4<7Jd7R7Nc3D}g#Z1`(m9^V@QIY)fIyBuDAtd|GvL$gig5VKps_^mAzi|7!dk
z#?ZyMlvmg*qF#+(nUtG%Wm`{Wz<jCz8mS6cNX#^s6t@-dKq)maW3D65NvJ7<(MxG=
zTh7Y2R2C!^H!tSMQGL231j^f}Uz)U1;<rC5y(`~^D1xy0<D_&~qPn~CW_erQQsNz|
zO{z)lJ?TBUt+<9;xML(US-Wj{8+eGnBihtm^RD#U(3hzWSo&S81enw@^K6<2oJ$SB
z`LqDIkQM<K(-Qk$vb4DzDu72&ax^Ui9!o2L$I~j{iFAQhhSbSab<1<o@OaYLJ(WC^
zs_F4<1?h+PB;<}$ZjUf@8f9mav&lJ#LM7D}CG@M1p5DQVI`STkb)MSXm3Nz|x;HL+
zMBw+)RPD&a6RA2p=~jo0<T1iA#O*&q+@2afl*|xj(o@U85=wtz_;B)gdJ?neQ><uC
z!kopl=FaToG0N$zY<Mzlr1}l%mu2=<arz2U7tpJGL>it6b)+gtmD9@bOuC4=wR9Hn
zO4cHjsXSSsGehYY(<PdHcs5kr`tWSBLg$6@cKYyLiE6g8c~0-H(kO9~Jb`hFX*oTU
z*4^2iTv$w=q>+)IOXt(G>Cv?4&h0>%$)KXm^Si6b+4S@`q*4Bn-Bp37lY77%wa}!^
zN7IwfN{VDZnx4QcsF5+}_2u+bMtN>^N!n7erXQpyhvz0^&Y<Lb6Kv7x5T@qcg&iZz
zCr@Ebq#Dn;i#yu3m@7R8%uQy%IEU7fHr%C8O2fv%HDM){36}pg!{Gz4tT0(%`F_@7
zgypl7^<}GD-qF!Yi5LG5EFKwS@uU5>i6);O;}Gd*4zAagAkvPk4PL(-MgFN%jkltH
z1G2jlC5<F%bi<b4cfCfk;WeOWy^svBM$~THj{2<8Vo@AF?u>1>2K1QQC3_0WXJut&
z@YVyz?8n)7ab`F+l6}jOX8wN-EyOOgzj*~#9HcR9Joae_(!tX&cd=I1fSR^9Jjab2
z7;&7LSOSd=C#LPKdyUvj2JiG(bfb%HQbXp&BSWC!Wn-bO69N@U<9x@D)*au!O<NU~
zW|PLJ1`Qu8Sxx#dXB+2x9(&yPZoqtP+{73HeK2*shVO+P<k4Ydl-NPvPg<lz#t)@Q
zC(PoF>j0oUDV3Hb@E71#4(5<d#(U-9YMU~92Cd&$_#9PCPGz(iM*1extE5u7alv<j
zb=SF=AQzM7s!M}2`-*MFw!2|BvF()$u!<Z%zKBHDWAH4R!FcYsU~~~NOsw?=^IBf_
zhUW@Q9$vzy2G6ZDu5EO2COx7OaSB&nzIwIMW8H{#liNTPFumUoXA$|rLXR7R!DEf?
zinjt7Hd;>X#f_Ud>Fm(<yJR-p&vn1txY&520rS4`)FY2{8!(fo-+sHPp&#2X#&P+6
zBTfiXmyAGG*hS#y9-WF%<nPNo-*Yk#NaH0KeG|&!8rUS}6%aYfyw+AP0^#sE%Ev$>
z;6}Hd*daV6m__Ug#^i-DfyR7hoZKIAk@|+-&8Z`e$Zdn0m2<sbvcZd>MzSq!%x%rM
z-Pl>DiC~0MpU^UCgv_NA6JBfP*k@@o<Zg%(_A0GIbK#OTm8@7aAPc6c9%|S87|+Fh
z&^Q@&8If3Xuthr#*3n)eMxX>(@&y|IgczDi#W>vT22Kad>kd3#?1qz#GO~_P8)r^j
z%i}7cP6f}fmqgVU9KY|q%vi*@3NGYo;PiM7;-dpg9yEm3^A9{>o_AT)!#VNNG~e+u
z)_H?CKJ)LyVJJXozU@a&^5n-yiKm}pCut{Z1U^qosgI8-w)?%Tw>8&;Ck;m$=DaIR
za=YuomJdU2?mNJYygsg7?Ou6<9jBxTHy7SD$uUM@4VEL&psGsXY<aeD!D*e`oFrqb
zY`5VF^;w2TcA7dB3Cs{UMSy%tRv=I!@DKp6O}jKUc`N5L!SE(EA+uSWf>quCGWSo0
z(9awK>jYW^TmlmUvJ9fJcJIlL?_-H=P!pmos|p+o&IG!M$k=OCDkyuT2J9&{V9iWr
z;HiMd+3VCHAC9~?-6v|hZ7m)q_AH7JvzMv)(ztfC>wS(`-scEV@buh*X06J;r<gI-
zl!GYHf`f&-so;QfLeO)K1UXKdA$a24_E>Bg;6|Z?tm0@Lt{ppE&%^~}NWr#Sz7xl`
z%`RcNOt?C4AoVof_zDz(c|wEJWx(B{pMjKymr81~uIMuTOt~)C6a#rfp2NE!pGKJ?
zoA>ox)<Q>%aa%D0a%lf!O;?mTS-G#=*YD5eX*><T8d@ugf)Nz7%l@?=Y4?qyiE&M_
z28;;D&y+P~2`f_a`UhD%(dWMv9X05w{aE`^tz4HY_|;J&Ml0(7vEMdJ>Hnn^4OzRd
z{JVBvtJ8iA^w7RyVh&m_&33qt_JLCQNwX@s>DJl@`VULT>+)b6hJn{MQEn|#eB+1G
z#4kkcF$tqYN@Rpy5O_)DRHopItCAJJO>rG<P3Qt)iLlS0HKe&<;5^b97ZZ0toRa3{
z#yARb*tZQ(TfkWh9ehk4EWeaR18A8C1we~L!}uYdrkQAt_}c&p91Sd0EE6#z9kCuI
zrMOT&Q8zVgK!N*|#bjhemyD^%Q6b5rC!$K5g>40H#Vni+6R{nNA>~sIQKUj@?2<`3
z%9F``{S0Opl^~jwwUBm8S(`mLHgLa;s}du}y;&9!86jMR++#;j$|%Sn^n;METpe_K
zLiDkaDozm~vBe^QW+{_YO{BR-@`*h{b&nEQA@Br&Cjqzy9*fyilqX)Y3KPkB82c1*
zC=vE4f=96qBGi(y0Ct;`D_{>z8PumvWE|F*&(8e<dX4=2%fbNN2bb|=ZvWafgg`Dg
z^XwC7%|1!sX9y6hGdg||_~7M{IB0pk50OAYo($Zgln@TjAw|22$<i|<r7%tCKP_GP
zXXzc|&6>(+1QOo|9-<|B;y(l!$4y+>Q$5YgwUuRr8$8WbJ>Avt=DfU%P<gu~q9u9P
zxKo0xqo@fG?U}p9J;^;P>Pl|;PGMWNO5^~ewmhnZgL(&Uqg5H@kGZq%+#Rr$RUMVU
zEfXzj?xMSdo-^*Udj$V=5rvrr`yHP|VXj<ymCr#zQV7k4Q4ZSzIhY@rT|}i@-fdop
zU7O;2p35G`HhI}cgqu7STZHp;=0Tl?2cohD+R<CZ8@~-eA(~A&)w@!f|E~NEP&mW`
z)}2bWN)cLHBObJKsRpR0h|Jn~7Y@Fwz%`d_BPr}6hU%(GX}1h+t87;HD5#-L@-@nL
z5Y*VCF=vuGTFp=(L%3<VwSU1P@ci5J_saO=c`p=EJZ>V2MhY8dLUWc~LtRE)<Ua^2
zM<7Rp|JYZl_FV#m2ewU3Dii#XjT*Bq)oc<#U<&-yiVz<(T+i!?)J(g};$(#EY}EIt
z@m*?M!dR|7S{a)|QPap+_bb%qLDOp#_RN@A5WArLi8{oMh!m3)shj3V>FMB4y-=g*
z<cAe)BHBR;4;JmT3wniUe+&u8E{R0?9B6w9!G@udKtgWIyULz4RD~o(IE7N6HFZsd
z@&q?S%6|j>*(?B*K1@Ac$_)FqL;euzN!~M>#HzJyUtC_;f#-xbaHC?3OQ!gnHp&+f
zbVfLsa3@^oF#?~b%@YyUv1QV$gNI(qXq`6u6FRN6XtU#o*n2|8-$rIc#vrjx5L>GK
zzWllblFkuHe+jB%m?!msS{0EVTqB6s+^MwSLdARy;s9j@TtG;hIl6*K^xZt_<UNJT
zH?_prEZ~|#rfUEyp^FJp6ljasT$-mSO-D|Pcul+ou!p*%wF%MM+MvGP*l65rv>K0J
zY^>uC8Th*8zzZUF+lKfJ9J;D1wrml7ApeH_D)z(`|2kn82t*_v`i7W*7dl=70}pX}
zUY({{2n0chTuCCXyWWj1+zs99hEXrJ3Jj7#cojlR@fnPnbUXd1ACDHmbHYJjzeaPw
z<HPM)rk!|h-AP~=%3EAnkJ&d-FUXoJorDR^CAS1=u-~96S~xpI@pM2h)1edRoF)?p
z2TwoJOr~~19?VS<z+`_1PMquzG7iXBiD@P>4iZF5<2=Pm>D)JQ&WVQ07F;TkfXC_O
zk;aMRv?X>1Ck2K`aTki#5d1qi&;-;ua`{vq^`zG5Dca?Bv|#~?A%90o3>X0;V*+vL
z!gwrEIyF|)SYzM9>M}!$kO9Y(ur#IFiI0S!0%M7)^g#WA(20sFBK)g(g}~@wxnDM5
z&49mYJ9E;nDA#c%y3c&LOjK5$VBbX#IJG4F_UnYs<>rOVp$&Q$*XdZuN?ei#k4zmx
z#)U!-PDe|GAB=8Ilf8vb>^ou-(TrE!@pUF$hIsyzJtU;^*QvwL6Zk~{j6g!g%HKlZ
zpc~)jazZjWboVKYy-y7cDx{lHTpab{jMDZE2_a~TQ>%T0Hbj<K1yoj3N@VX8V6>C5
z{^5cqte=GkZ}cOU^F+MXFdm6lQhX@{EM4F=l{JprI%|pORVZFhTzgHzg<~^WUOSBc
z6V#;-JvS1J4leY@X=v}x_)KcbpgFbFpWf@<Q&?n+=w<v6%aDC7G{pCiAR!6)k1H7=
z3PI)&joL5=%UaJ^27)k;v|s|H4Wu<<h-F+mxVbE)Kqv}ZTumV?U42n<^GCsqu5m#E
zJE(2}|I)5Xx4BcbCQ6AN<qNpj!IcKFER6_@YDb-FSE1gUI!035>W+dQHI$GjomtC7
z4A3)@GrvixeRiYJ4sIH+XPT;--3LzYBAd!RgxDqa7ie{lP832|6CoRoB2D<1AR{%p
zeb3QH0l4AJ;vg{$U7~bGf{gtZfK|^1x2Gg9qcJTqJKz)xn~iKVGaCYST4zM^Y?s=V
zM&;Q=DyrEZpq$GC!MT<Ot$JNZQ!57*;3S8M$fOSjMBFLfB7&iqln!bqxq3ui5N@*}
zH$eAfMVlXL+F*W)nD%>rgk}mDNzlw^kbszlT??v_H{~yB8MS~$NL9iYbLAsARf25|
z)S|+zR8lBfxRt^)b8|a+QkXbBeT_Yj@z|dd_!@wfzv&>J>_PWg)m{|CM!4bGh+7A}
z#HxmfhD;X}TJ262ZX^`@5d-RP*>u$nDY}5+aAq40B6tHz>lt>(82?L}^RK94bfS#Q
z>jKZhqvG;SM*PA4n!rKf!u|$jZxe#)cnw?-FObp_;^YuY_fPtO2q&#IVZ&cX2B?#U
zGK@2!vv##TAb6C^Ww}L37;j*M&3sluY`ba1&!n1YS=cl;i&+gmV$h~aOZ%lsnapt<
z$k>6-jY@Esqt7nLpx4{<MUA@FR61B3o_~72^Zmc4=ZDXODHNZ<1n?3kZq|fl&Axy(
zYzIIbTSC{}BSW2=J%`~^AVIJQ0m|$?9=_7(V&2Ey3NIsa)MDMB8#=HpGpqfNXhf3s
z><oc(1W1PS9EH^buD#$o0ho|92Kzezt1KppM|b`RRMNc2PLIv?tQ6lc^q^b-o;q~6
z!pVYLML|a~69Tv_{yLp@9kb|QpBl{b1^ntFhBG)h#X{ipkMupjOe)48VI`P}lnJ>!
z>_sHRIuL~x8`?zyF9<6O@zq<0*I2#b_+5Mof$Igx6|eU#S{|uv`0+=kmUF0?m<Kl~
ze~y;Txx<b+_2()Aw2T(XHGFU3v)@B9bF#<^%=GtBN(s65CzSd#0)I<@_=0_(z&{Xp
zo4`Kn+@dTgXZ8*NJ_egO_=>40A1bQyV?|T`2XQUL)|3yFLghb=hOQ$HSk)0P(^WV_
z_&3a2NsQ1u&vh_6v?HUXi&e8<LW(?Me1_B_^u)roW4}WUh<unw4Gi%e%}N*w7tlI|
zPeoQ>pddmWuS`sqaRF{-{4CN%QEf)NE5bb1{B*ZTL|z1=t@G1$9~m;jmp4u~Gl@tg
zXF+Z9V6u321u5$jqapB3&;FuTT(=k4*GF=tjStw8e&~(3Y4nu~BTwEksJmD@NvSMi
zP84jNJGe9)_~`pEnr14OIe#S@B3@}lK@f$)&$TKDX5(`)gn8++HYP-^_*8}KuL>lJ
z_=`BBspA)fAZ#<^LWbIl#0cUOoSH=DzX1hB|4T|qKZgHIap8j;rG7+uXU_j$&{3iH

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_45_30_8.cpython-37.pyc b/model/__pycache__/cldm_decoder_45_30_8.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..fe092429de431bbad226d80f15b28cd9cf49450b
GIT binary patch
literal 10319
zcma)C+ix7#d7s<P&Mx;tF7K2?jbz={u_;+jY^PD=*p?Nim6=qMbD74S40ndhA?LC>
zGb@S2Y@CKlthS+}x=n#1NkKj|Kzc)grUlxE6bSlI6wN<yo&xkqKpzUU=tB^o{e5R<
zmn+(BGmH85+%Mnx?q^=D*Q*kqpWgU}cl=dJ`d7*f9}S65;Y}1-l3dA^11Xd(IaDkZ
zAg=_frQ%-=w6J8AiZb2OMP1n{7ypL%uY^^rD%xnlL|C(GVcn{S4XY70t)?h11(V^F
zH6_w|&<dxm>2SuH31_X@aL$?w=dF2BR}L1!MQc%{jbJG}W*rNUTgSr_)(KHo36{f?
z)=81B24;B5I)(IvTMJHyPgzgN(x)Z2?lvAsZo`}3Rjo6~H{D6(C%x+CS(KbZ$&}kd
zNy}SA{=BQ)lG@Xc2{Ublm)~@HJwNJl^_6IQwJr0B&ob9zo_o#jq-~9tJ8^W^j$Ez>
ze&jifPwaRsw!_%%2Oe*we&{8s*R#EF&2wE0gdtkDdQR#)!ENS5$$HE}kF`sD{&vy!
zb!vLcyVLg~YQ>G4Uh--XceZ%-T6`}`oUn(X8?U4(HHf3UuJYEZm%bUhUVs^zHyo<G
z=C7~!6Etb7yb;7*JH4OUPGb9Em+RhrKS>il??*}M&`hakud_I`v7>dL)_MGz?{wqH
z2|n%glf-wTYv{rJwLZ;?`Z}8Lcu}$0$DkaEO0l~RHXxSCI)umJGorFNlz5UQyOO22
zC0BK|2QuhO2Ih4>{aPHQEDi$CeJd*Vw0gax7o*hiU%~rpcm=t3r9nBB_T>ZV+h38T
zRLPV~+QeY4vZoG=Oxo8n<3Ju%GI_s*8Xf8C{WD0FuSst&4JNXQteRCabypo!v+}-?
zRS#rYx^v|V68cv1x&zdG0gUkL(oN}$Ql`1mR)u{Nb=rRQKzdQah;BN8F>AXrM*aMk
zl|c<7my#cdwg&rgDs9%$zD(`A3T=5$8C0X?@$!7M`LX)(wcU9iE2?4-a=r(wu0d-B
z(n_n>f%ysM^*krF<0#l>vp^5ZRM!vNO<qc4*4g0YD7L%IahZWy-T;+)f!*0~qR0yp
zUL|VHDV1yKcF*G_SNzrdsMk;JH6p@<>);3v6kiK{*A2Y9n5(_G$17gc2Q`C~cs=f?
zW1ZBFlX4Yh?K;;IFIeaDeO?)|M<Vzr_3o#60;kU>X&W}?Ac7=%;QH1*hjkOKY$Vta
zEtIPS<PEMD3@7_Pl-S3xxu1XT;zk^L7o#}!*5Y{UqU)s2#V(8cy^FW{(G757XTysw
zc7TG5#3!$Q=H^9#kBgna4Q<z>8fJg=WA^ix?3Y%0+q?``#RA+#BD^yIlB~&9c@l3;
z{2KCtV#qaR2Dt@gS)NpC^3Jgjq2`rhq_(!Yij&j0VcU@d+}pfr7wm*|&9?9KoglAa
zFJeNmwK4DlmCXYz;7t?&Ll*fb*p5hno@51J#CFYhrFGe@xYY+0upUI_1If~IVQH1#
zl{CrHL0OZ$T5JH~g%OMYP!qKEm}YxS9Q~Mx={EZ$;PuD<jYrh4c8;Tfo(p&zcvZZ~
z`{<EMyMkS1DVt`?*(@<rvYaZeyeofQ&hzYMHl18XnL=g6E`b_qgX3WEQZ~CQXUDe`
zb~Du}J($g!ZVCC{%<6+?Ru}2JSp)xSau;LhVqD5A>^)JhChv{Ps}E(noXLPjrT{_|
z09L`yrL?xIu%D#$%)pGLt~?{5rU9m{XQf>^FW*ubPn&L8%u%BHY(a>JU!Z<r)JjQy
zkt$!8A3}(1RUXOAO6w0Ls(mPLHg@GL<&jJ{__~}<rc>`q@5<|nYq*sMM%v0-?#jEs
z1zKGbt)|`TL+QKFb(s!W`##nKeCn8gCMyHZW(MF~Rso#PCIA<*8s>eFE^aPG8sIUM
z9nUboeIjcBE@w@^li3`t5GgZL-TI6)IFX*(Kb=06so9BL1?e*f5^^UfcR-jri?Va+
z`Sb$Bo04hs68ber%kE)KUHO18+om=T<^88K^<Y@`5rOAtGPNrYmNRv5(rpgRNAmrj
zq|Xw*uSt*Ow|@jddTMYQrDob9Ol7AQfi+QA8$6Y+WGAtba&|gXXCNFQE+^foy)s5T
zla~!nW@e^u>5o)4Cn!RBq_A(HSA9tuoQ-NoHIS-jjltP$9(8|}O#}XU-Xc<{JiSQg
zh|+(|7HIy#xkz(cgLCOcIzg2GEPLvqL^WHInCHy?B^m`M{v5`c$m-eItmRJcm7=-y
zGL4M<Og5XH%Z_Ce?#v$4lMKq*Z0}!6&u3@elZyOD_Ad#Xojd>*sf8wOK9ik%QBoxP
znQR%ephlJ+DR-_xHJr-n(QPhBTPjv{Av-xZKN@ukD|>&0uk!`2X5G0xBWk42V^yT;
z&b#w_+OC)_I}hBA=D;{dR+E|T!pEh7`Q)mwmJ3*m;5E$nUpXE>1k)N@0**ACia3h`
z)8|Jk%+GRhPp4MN_5TNxOG8Y4u>UU6=tqZmMEYY-uGp0z=9X4>ZeEY$;M6JeOL5<X
zeD21nnZ~9cb%MU@ndydSLX~<EnNw!GZf?hYW_DPdB+oiSOU#6(a(%Lnpk7v1R(8Jh
zp=0)we7qz#1I_f%4y2iXsiB1^hJLT!fK3Jo4U3I^3IcZL`PY4{m6=e)_J-%UiHQ-1
znUStHH=KmFx8|9Nm+rjNWAR-d+oXofOA1wQdihvr>qJ0BYF_9D@tP9^+q6|-H@0bf
zYG4Lf$;Gq}<J7#+^VqY2cNfN}c@JX<^uciT%)pDf$fHAHcGzJ*NIRrh=%Iho6VrH;
z769lDX4s+x{sP=8;jCmb!y8Yov@NrbqBRWm8LAkaNU;${Iw;qhq%yg2C2+zu*SVS^
z_aAtrowJ8(ZpF6!$WLv1<q9kwCrGX$k+<J@5zSx(_d77Chzv$nx5>PY=il{Qfy1K<
z_~gzfR?OQQK2D=Y^dazn<Mms&%pUV&=BL|05paCi52p|Z!qQI6ot<Y)f5lq?jLeRc
zc!_xrXPh6`VVB%K`wjnX^Qw8tgqdz$e)?(Ogb76b4%>|l{mfx84k`%DBqc~)a?)5~
zSAm_!bS6WSvfOiWPe$W47<40=;u;tv;SG>C%Dm22F9yl*8OkR>BjAR=p4t&S9vD6B
zS1~593`sNLQ^Vxpi1XAp^6rftXiV-G+@+H1_0kPK0XihR(Z<}?l<OzX8chTPllnwf
zl}5;2Gcn=qc8R?~oB0d@S^^_S0&9nB)mHLi(SYokwtA#p`vqP~`k-zy*>WO*Czd20
zqf>pIv6xviusgdB*0sJuEI|n{=PNYTvY3iW#kjochfWvE^>;iz;YXv5GqRmfo8-<}
z$KxuYQU&)gxbCR>yc6_2F&|gKk6aC%9xp+VbYZoFrqFu!Lr<KST^9FnZoD?mfxM1&
z-X-48{Wftp3J_dg4`L^M?u8=p{AG5Mc5;Kjo3xbH@YoW&-^+VjGd+07aFAiTyTbgo
zeHV6n6!GezW6Q{+<Jv9%##`(JC97~>;Wd*)V-%`jB?2Z@H9}|0vxRF->*Uo@GO_CR
zI=r4f%kjv_GG`M6rU;xOKt3ZQr;*hNoCe^N<L--1p2>w=Onjc2e1QO+f>qxDG7nFN
z(9ga?z#*_kphI9pP?kY7x88g1g+nZ{E;S)avzoxM;8dWCh>opNsi5>iA=qOI!I~P&
zz$*dGv(Hh7a#VO~dO*~+ZOtDg{yaL6u-B;h!mzg3^&y8W9&!jMdVFp{vsUBKE6kZ}
z%o!AD!NJ0{RB*s0A@aFKA|0pA5EOB?J(gGoxKn5&Yob_(YbOrZbMXOjUa{>?;3SD{
zg9mBB`B4b}<`%Ng<4yh(Kw7TC>zX7#OZ*I^G`v(YspxY4tI929QZ|q`;Eh$~1^Jvj
zR~pvp^31Ru^_C6f8mgu!^0GD~#?VkY2Omz6^PevMtEz#q@Dsgsr6tmuvLI{aC8XgE
zs*`d{ZonsOA%|WQ%jntC$WLrxo*8Al_i*%L4kI5!AI(bTH5ESOFd_mMY^0ky3Sk2e
z{S#>u))J!FR7zz;K@bDU<V>bGh^vw*zDtn;ZB>Zz!hB$#Mtg{2!EI%vbCx4g#yODE
z<i;>OaMbM#P+Nh%7H+2`?<~HSyXmyZC%Lsld`u59AI%~{H{U}dg|C2xibWz=qa&oD
zWE7w&r|PDL<ty-l@=%G4Fpn`7!l~SUDJ`Qf0z8|QT?IjxbMO_a2)a-(r<`dB=TtId
zpX|sYPqy-%8?a8L{h|zEoxF{-U&~t^z(0W>WdKzosKUszE;Jt_&%a$4!4n}HgxF$B
z)Z!cg67NDha&^b=3CYDmDicEJG^HYdb}gswHqu-p@xq>_x{nYbADvwy@Eicwz#<8|
zOnKq~t1=RNN6&-e12sC2IV2<zmH-hV1hD`{nNu1dzMa<C84Y6)vw{}b6?7_``0K)w
zJOtPA<UalCZNw|Cx6AA$v}PYC@GAs}MHvx<h*I!+!Oa~n2yFXsp|H~KQ6s_DpFoQC
zmC$C53QWIW>eqNj`h=k}>P=#I^blRqlMr1EgB+gfX<lhvS(Nav<4197`FpwueSoqa
z)F7HPw}j9~#jEa597yh*s4Kh1gUYUK)kJ^<N}#9}VG@)#iu^ozx(~owR<kI9|M>u3
zp*88k<HNX9?u<K&|CR`;OoOf#Mj@3O*WN_Da|RlX;#)TCX&3{@!CuG>8!FxMws|FS
zZ3@15E_)W+<MjZcW^yuY5o^<#2mKd~LZk-T(A&hD>;O>2WmC@dto$CR8^QrA&Lmr<
zP^qmE589<n1JpBwNbRx<AKg{pQA@UwR`wAKan-c8Uk9%>Hk$_&o6tu2Ny>K-n>e5`
zr_vT$O;I#Lcvhv=E*602-(P-AYM7V3NQBmS6~QgitkCD}X?7EJIUSMNFU%2v6cLkS
z?@;a62+;o7ZvpVe2ybLuCTxRhd;%1e;#wz0(9Uo@uP0Jd>pn};0?GNP52*1ssBsNr
zxpuKKHiM$Jku&L6sm+H?uTd;gu<MYlDejolHt_>4B??}wHX!zg%BNP*!JA9UNT`Dd
zJ}A^_2lNV|{tOb3SrQ5J8QjGzAU-fqQb@;Ld0#n@2C9&vh=5QE6sE3<Xk0F~H1@mb
z&!z#OnPHFdT5gT6JLJSb`jebzG>J9Yu>)~)V27R)-Nki=F)TUcVzEU&k0>zWvxGO{
z@eYyqGHsiPPlCp)LGSIHel4eO+UWn%X|+R}rH53+h;-jarXXFA)+Urerv0&!mEM!T
z3}rDWlR`kbijWL~b|UC@C9AklD&GLhqpX71zqDDR=zps3mr*AlC{(_wrN(9jHv}@p
z<EaF(f1wu879q8)oYpd<D`yjUt9UD52z5_eC-iEoJFPWy!@Os9%xABfYxqM3R!lwg
z!kBH_5Urs@w?h*hTZ9P66=C1TZnzTMA)Eqj2#7<M5ZUue*Gpl0AzaRz<5UXP7o^9P
zH0HYN-Sy!K=$;?Ny~L_8$N*vR3k}4lFlOp^`*A-h7Qjowc4v=h4mfAHu*x+LFReK#
zY&Lm|D{Be+7U~5dbETUyp_Sy0APDyRR7DH_AdP<qrLWUr6SqhRcUNKH<VG3vj)sGu
z89VsV4o4?VHU|mq&*Dw~0Dz2v6hY3UK><!W*G+Apr=@)zXDcs)@Nsr!q;cLjYlFRk
za{?1&5b9`M#*Ip8U_@^q4Z0}$QCsvAtx9{^pyKLo`GJ(;#L=!wh64h~mElODv=~Vk
z0mdpLD$I>0fryg8(3r+1&Jkh>tVGCTBykVxN3jhR#YE;L91=1sx>)wxldxLA&9spj
z>Gza7+O4rDKkn_6l}qdZjp28am^;FTKd)ZNeax_Tb&U>#tU+Dcd3p@>InN1+Hy+Rs
zt|_{;C)gL!iILu9x5Z4NEpNKR>qEMd@PaWzM+o6>Qip8<?*g=|LXKGF`*866WSh$=
ziQvdRBro;@w6hE<q?<@w-t>~3%!K3|iBCQ;w&b^HSGh^kfHhBdDsu`v09fI3WYWxi
z_)ecNk4KVp5(^+nO8PB>EESh$$|?tWTMI;@La%z_%4rnH8(PBh>JjAgW$Mz0IxA$L
zgUhsG8p^plJP=9QX^$=Q|L*npG8Wk(Vxfn)gv@852!4zN$wP>FT%ZWq2ik@#)CMIO
z&w9=!kbGsN1?3}cAgvJ>EaG~=E#)Z%GEbP>Y6j`(>hqFYJ_i1CjVls(Ky@qlm-bb<
zIUTF5qLesLz9OX>t{#X>X~dda-&4ohHK_Ndj*%3%xu>AVBuYqjPOauL1*pjd3nSJa
z<1q@m;D+x`u8^AfJ>2LHu&wyy@CfjjP7_jB6X6q$qCklM<>W;-+$Fkr$IW6MlSp8-
z3KA!|#r_b$YUP95V<MN)m=;+Ua8QLgM&?&F9|9&=w;*%&eQH-L%JYd-RI`1Qb9qOw
ztffJD-Vx%{DnaHu=}`iz(xtqJ0VP{REXUDXf(J@Bh&31qa}cTYn(!v!3o1LaV`OyL
z_X%pLU<g4iZy*7gO7ab!;HLa#EvFMu2Wdq(TCTi=^CY;gflgF-j7kRC36D{@U2bWw
zOu7;0rmwP>F&_IffRRCGmG3!-@p@2oR<jo;Fa+*;He$(PFSVKx;u7NpmCkxM4+Ii|
zgO~yF_ieghhTtR)$BAt?h^^hHw4UR3i0{9q$$vx@#d$I=uL&#*=ZMSq*xyr=zaj9X
zIAMQ_vab?)=@wFg^WFeb$tABoiO|DS9?5J{P$TvuN(JJifehOV2;5ZnwF98H$d&Tk
zyd-QhkpE^mFCli_G?I->6D=#7)y;{#1|JS+W2Lpj(zH&NHjZQH8<&a_9A)uo1KH~O
zI(@C8uC|rpP=5Au_ox3t&%eD4W>5kKUGJq%(w-DXGkXhrV7~((4lJb$=E4T&)t<u;
zqE8X9L69)N|AqrpT!aU>>9FdE-gKBB`jG?Ek$sH@`x62r+1XhF=LswU@Dhc(L$1Bz
zIw6?v0#*D2fK?Y0CB+>(f`v3Mvg2cOJuk&K13f4bfXhdYRrpQtlqm2gW<n5^CBIGQ
z+`=ppEN&A)ART8t2UD~q@0=WC9&q>v`i|I$gf9sw!A7J!geOzTf5e2vR!9QyiZG-Q
z-MoKvb<Mj@;NznMgoPnVyxw<cVWgd*43@?ga-^BK-C%!-X6>1yUN-gSDuSbo7RWVx
z*ATGpBANS5Wai~M`WU5zocmKs{RM%)BS4(M{)xap6UYdRm}O|<y+EY{0DNL4B_<LE
zthpjsrMBh`9Ue^6XiaLmTx(5E7&XzleSzy>X6QXe3l(c*50N5Q79aC;2o;I&&DbAN
z10oXUQUgPLow5=|!iTd?;RB2nm>h@&Co3Z(WLSVZksuFK=`;XwsEEi|v*X=Hv2PLf
zwJwa;ePGBK-?KP=Zsd@{&BMp!isX@Fk{i}b#Sr*X=5SFfuG<UkJB8#}$0uNEKk^Em
zDZU+HP3mq?cd>R-F?np7=+rv@<kE1w#TQmI%~&q?;A%8PveJpeFph+yYBdlx#s^mj
zsM3dFOo&qP5eV1c6i5`o5^+X{hn=VBsBsyVqxLG1y!fyMp;IUu1Cag~^qPDe>L^!A
MHDz+C^^O$$54#VIJOBUy

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_45_36_15.cpython-310.pyc b/model/__pycache__/cldm_decoder_45_36_15.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..7b73c6bf9ef3ffe57920751938920fc36a572c21
GIT binary patch
literal 10517
zcma)CS&Ss-U9YRFtE<oHx%cd~_i7(AzD@9E9oy@5NY<0!T_<sos8rKc)6+9`^nO*n
zM^kMlo(<lBCgBPMabyM}gpoiX9ylcO!YdJPJbX_`K|(7<f(Hb|5HG*~SAETR6R1}G
z`@4_-_rJf2%4Ji+Gx+RRz2h%S()TGd{b<O10&lFylH^LR>`Q@c$$?^_0C~k%EfxQ&
zuLU_PH!RaFUDV~R{P5op|AoM`OwmU3i$TdM1!b!oRIEx+wW^{#=huQ6YeuAXzaGq5
zv%#D-7tCAp!Gg6AELw}AF7Gb|%hs|;8~%~tsC6_rW*rNTTgOFN!9Nk4v`&h&={JH?
z)+wZmZpnWrc-VSamR^(GvRk<$xfO45N3~8PUv+E9*F1CU3`)+TWX7$dr0y*vf6i5|
zO3m4Ogq^0s^RGMIZYOMW^`-FkT2to5&oI|xo_nd&N}3wax1#Wd9lBiiJE7+=UhI1;
zvV+L&`5vz)oxqC|uWNh3y63tW2t(AbcAcc-_}7>d#v2g}Jl4$d#cRX1uTs;i-u0dr
zQY&t};l(feQEQu<m!g|t>;zp5U3n=<s6iBFb%i(AyyW%B^?b}wz2Z>irOw7iFGiE5
z$}4`<wv$_l?ZkE`Xmj1W)rpgsFLuH>acHK*vo}~2*x1oVht_%QQpagWq2s^i^y0YV
zgqP5Rb=G?{E9%Q=-uJ@cX77P=Br3)3TG)VCCTkHMrymuWPvDIg06fW(UCC11oU6Lp
z9T_wx1N%CkeI*JL7WuyCz8MbpwRX9sqtECmBVpkEG~T#^$=lK(pGbT1zVwZ!Whqfo
zC6%@?psVbv10$98wA9#_2ZdDL%b`X`+QfhUlJuqWpqLg@GcBa*j*8O!o{^gSvMgOc
zj(&x#W*;@52Qz5+?WnR88c#~`g_~$azqjzFnk)5W_Gz@&_RM|hd1+uKMa)s!kul4o
zUsMJq%#@2Q(Nba?iL_Nl>pZn?D};nyWnhNY>GEu>#i{z~wO;>hYGOa~Xg_-uTKQG!
z{oLB+rfe0N*Y%vljza%7n+K|JLaN&dnpK`lBG%gE`7pBE%yAhZlUG2wo^Q7{oiOzL
zn43hw8TE24x!v`6&J};9PT1`w_Bs(}%yn>y2QDZD9oO}}teC6asLKmp*aL-wym&e4
zB~zW$mXmN5Wz90zV$a{;@-1E%u}my@EAeh6Spw9+YqSj;a}bddeQ<sIro-AXS2km8
zh!)CK0`ex;hfFAU-j~?3*xUy{ySf<#-f9>n-g*>muewg+thQOy>#knyg;&6xtxYdn
zZ2<+V#5*tl%p0o$AFD0j4Q$t=8fHK9to`(J_LI-9bZ_%K_!ld1mx&6`0%)=+*YK9a
zTSEHS4~;@et}7*ZNj{-e<dR~_{i6>c@Ri|MO>J!rho^DFwnGQ#x4CH#xeV!&ZC~#>
zepbVt$DCq!Q{Xu&TLf6b8<T7|q>3y`KEjVwOHcd-fDu0y<bqOA3tAzEpDuosqF=sX
z;HL_Ryjy)w+K}CXYu+h<dm(x6NS2n#Tr2k;SjE!8(KT)kHw-DoNK(A7339xrAbpP(
zagXTl8v8il<$L$>i2AkGF%*oR^GH<ik`Rge=#@x2g5$w^v#grV6aU54PeFn0$nVNo
zo;B0ico}61l@X8mN}vvoK}Kll{EnO++g4aJ$x(VRpH|%*@~dfiP)*As{X$y7zZ$=Q
zF?2C5<rVgts8{3H#^vT++167TFrO-bMydi95;M&u#T^AqP)ZHVm}|>(5^Boe^irDJ
zk+bq`l?6$~&5Jp5RG%&hiSjn;m&UD>`0Y<g@5*-}iy&?OI4RwgsP3-3Ro;=em3WtG
zlWJ0XPkK+@P+Y?;+%b}wtlf^h13bjv7H#UTc~|;vD9ls`Ed4H40!-?dc{a@h&ZP$6
zd|CipNQ;1rX^DL|S=w3-6~H4XIhvLMkEIpB<7pM}M7lsLL+WIzy5%`(a6D=3ok|``
z)%5s|g7m}t5^~2Ww@(;4jj}Vz+2kB#p^|Eg68cq0QSV|!ZF!%@I!|rx%6rXJ-5-@b
zBJlfYs<!39iBuh&bgP3#@)+S5^7bDgZ%+*#N@fT%>8WL438lX<csO}HJ&D=#DONNm
zVa{S&b7yw*80B<UHaM9!QvHVX%QE|lIDG}F3+PooA`Q-jI#Ly+%4ubACS63`S~?4O
zC2JANRGzHRnW6Lx=@QL8I2$T%eQ-8eq4PrdM*8qwiE6g9c~0-G(kO9~Jb`hFX*oTU
z*4^3NTv$w=q>+)IOXt(G>Cv?4&h0{*$)KXG^Lwkw+4S_+rD6V&y;Xsyll#CNwa}!k
zN7IwfONwManx4QcsF5+}_2u+bMtN>^N!nJirthaG2j|9P&Y<MGV{Fms5T@qcg<T`e
zCr@Ebq#Mt<i@Vy6m@7R8%#CNjIEU7fHr%C;NrT40HDM){F_!-|!{Gz4tT0(%`F_@7
zh~=~6^<}GD-qq1ci5LG5EFKwQ@uU5Bh$f#N;SlL(4zAagAkvPk_20N0MgFN%jkls+
z1G2jvC5<F%bi$V3bG=5g={2Bfy^t)hMzqnm9rakF#iBTV+!+~f4Jb0VLk1PJ&&tY5
z|E&j(*^9IB;>>bvBoot-X8wN-EyOO=zj*~_9HcRfJoX6)(*D!0cCc2~fS$HDJ;#k3
z7;%&tn*xnZC#LPKdyUvj`tNjEbfbf9QbXp&Lrb9HWn-bO69N@U<9yqX)*au!O<NVF
zW|PLJ1`Qu8SxtJdXB+3c9(&yPZoqzR+{73HeXw=ChVO-K<k4YhmDoYgPg<lzMh~S)
zC(PoF>j0oUDV3Hb@E72A4)%~t)_djPYMU~92Cd&$_#9P?Pi43nM*1extE5u7alv<j
zb=SF=AQzM7s!RPd6UDY-+nunJ*!Ic=m_?2sUqmA7(SIJzU_JL*u)2sC#^!p1c`dJV
z!*c~D4=>^4{TEgm*ETyilP*z-IE5>(UcK7rvQET0$!#DCn4a{*Swy}t(c?zH|5&55
z;;jINjg}L8apNXVIy<yUm&}O!xz3jw7aLDBVBa^MdgPH#19lSio3xu6`msqdj?4EO
zaYB%~WCXIpE&@OI=v0Ive_!VLu9LYy8ZW`>8&ek7z$P)TfXGqiwYIwv2#3#6J_Z^A
zH#-}N9l}+DS;VejOkNlfXv}9u$;pU|)Hn2QP914PejEI(oa=RyO<n{wl5uHcZfnNv
z#LhZR1S^#KgqBGoWImmk@LDs+K1G`$e?ye8*J&M^3!kj1WW}NZSujoYP`l={JQw#s
z<7CujL}Jas7;QUP$HodV0wutbFVXlX#L!eK#^GisaN1a2r|<D%Cme5-k#&UHIP>CK
z9#;u<DtLyyBC0;;_&x7c#v;a5a3NO%r^|B?A8nZOpdqxLf8Yu8yvw34&WV?%`Hq*d
z&Ktz>nS&<|Ljgkb8-C;@Pd+zHJpB|qNjq61@EKZ4eRNE*-Row(t+_5-X?W7G=Uriw
z+Z`9id>C?b;sG;q`?z+sbLCBToRTK|T)5XH#~8&mSdKu0sw#oA?b*Tyr*(33oQ$oq
zy#ZIK$1*&!)6}U*V1~dc0_0S(0)Y~NhX8nO+NZI}T{)i#hR;(IGMmLISmjM1b8<3-
ze&!HZC(t6`5*QPZWe|;Rbf0`~f+e;|O^C9rDsU_~6X+r$W1po`LD@q!U{9$5Yi23~
zR|PcA-k=WoaOl42K2h6kYw<9#XJLexy-LlOMzzCTPdH+E!V#e0>A3~XT9t{bm@(Cq
zhbYj3gN47T;DB>N&~uFhIZm4)dg9#nSZo>KMxleO;&2_V9Xnjl#06wX!M0nz6UVmA
zE@8P$xH@kl^)%l23KW5PLW9#~z}=#sfs}@qN@}vM=ra9Gxh~ff19?N9!@D4#Mwud;
z_w`)XLPv{HTQLH1X#XQkSClzfxv$*U@6Y9FJPp4ZS}Tfz5frq`{<R-!_l=^7aZRxX
zj0nfilr?1uD^l|M2U$DO=f4#lHR!4RNc&-}T$d~O)lniwE9(C-X`7|=|5A#Etld}s
zUAwQ<X+H*fXkRul2d$T8JKRV6K&kw=S(V&$YwZL52c_e6xj%}-z-t>Tw-zbB(L-tC
z7ozr<gi#_TGGZ@?yrgm}Q}o4E$%@~hz>c;ibb+u$*eB5%(p)fb9_fsWi8~-pN%L}J
z6oxqL+XkpD;H-rXJ|_2<U&+D&w9JD7phco#^bk+eOf*OQEr0}$29_$8i6D`VV2_eg
zU?`ucTN*Z?z<tVsGBUzT##HF2kmS)5VWq9Yjsmx07S4u=;0^_m@~MU}QXw_=$Rr)+
z$>hF%1~Uvx5KhWkNPDHM%|09(xL?LqiIL;pEQ^qg5H3RQu_GvD6l4(kLC9CG_B&l6
z`dCO6rwEYPVi7>Ilu4>4(p)3?#2%r#M+vMDc!I!_09*r)#q24{6E9hXvE)3AeF{00
z2>TSmqgV$KYDrlDyUocJu!pAf>r*E(3Tw<~=Y9dbhJOBKVSrA+WjvYNzjh5Fkju?H
z`zf?$A0zP71c=oc9lr>C@bXX`v^?L3NT4WB25wPG2!|JtqFu#g=^2tzm?rd}l&<`<
z^p5fQn#yPd65j_Nq9uCbKLi-XO<dViJ<ZE)D9Z>pc$%wvx~t*Md3hJH@{N*+mgHUI
zP6@J(q9#DJXYLjECHJVPE4k%6g&o-{kpqm{@~{>T>K(X^R%Mt!=FYlvcfeLwbyxzo
zOth%Ei|!J7&bZ6&5&YLh6lNCecYGX$xpL`sJ_iL!F*F-SIcy8$V18tF5tVLxw|OCU
zZ3^&tE_)o?<YgZrZt_%Y5zo_^2Xz`A2+JC1M{gBx{5AlEXtv~3??`F>JMz~-;Sdj4
zcPiN`g=lS!c+k$J8lavcG;8NwIQXst*IcrVq_Bq|s;efYy)wA1vQ^!upoTWi*C^jc
zP-CCQoJs0vHA8_6;il!*{soJ`^KUQSE8~yny-<YlxQQqlDQuVt%~^I0bs2S${~)X!
zfgBP4V_%`#cL@+4*bX(RjPXY{YRo!Rvqb=bDezM(LVVD0J+CWLGaDTiCqrasqrOLt
z?^5Fu#&Yf9%Gex=nnuRDU#2z>nqH&0XU4>W*ahv6)gi7#q?n{g-86?vPX~YMg&IXC
zKdfkD(GF60uxO`U&?`jyV@N=DNhH$eK-)_QHVl*m5^_i0Q}(5SDkLewDU<@OscRyX
zC%73>{_E(^W&xn|Ve0WxX4r2y<PV{q<UONFtXj+V#qEV1cusf&S1QJ+WQxCOqkIuj
zXT*aEcfy4pA@D`oJP~0XTPD5Qf9REr)@idprqf!BHamKVy~kwyEo6pd3=-P{v8CGY
z%Wp^^=^T;t7oj=^c~TFkRT1gIJ%WhMok|NXRLoZ)4p3IW4TQ9nqdSO1-^-&;-dCu6
zOG}Kc0`4hfx(A>Vx|twFfwqXvrFn|dwB@vj*Th=@d#JnG1|eEo>(|#Cn~j@|R^#!D
zjdlDX17EisctON&+YrBjLw8lhmMx+W<lnGg#h$q0UnlGWfe6Jz-w+e<LfcDV;2}`Y
ztJ5?KfglKxD@nw4*SpbyyP<oXFzUuufk83|uR>@kK7%onPP-TN;^6{#PB;kc*JuuS
ze7HW#v=h&*I|&R!d7CTiG5ZGU1zB^YoiL%f<dz@}_8U}13ulKYo(||$I&|Wk(_{kS
z;OR%2$<$8B{kbUu81L`EiIW{d#sT>%G0jBAL4s&$oTpeRo%<HfIni+2f?FjL@Ho9Z
z(l~LPw#3fhq`(j<E<@28f`2y$nt(b-E}zQ7p41vWMZ4UtHYh+b<nKs{0V7~&Odt+j
z7>y-Lr^ad;YwVj?U1mrTGT@jJmZmg2@sSWzU@TFU9;iPMI#E$Ygnt#U5EyMN_e%z>
z8Spo4XHNPR<vQ*}C(MW2L}ld(_8s(qQ%k~cQYUmSH!oxkZP2~APRBx4;*!*VWa<zy
zE);TbI$9$9V03Gm>@9R+-xiaIX1wZ-?lb8!#Pg@@At9B&Mjd{hz%K$|1QIG%{uTlU
zo%lAF6OzfHJE1W4J~c3?kgi5?bJUG9N}CuGLeLbaR{JJxh%B)RsH~=x$lfWyXeVR+
z!v#%PKMN1u=tnH)v3RXvJQA;@_)-X1y1{EIYaG{g))LXHP`sYF_Zo)_M`p6Tb{PLB
zs7nueZYUTX+~|$c(BAFQnbeeib84wSx!1j?u*eqC%jhANA^Tcri0>jnLK5;HS298r
zg3KWrwLuP+wVtsI1YsU&!30PfNNdCp%eZ!Mb6HA(P!zVfnnGB*`l96KkAfLp<AMZs
zP~8Ilr9G9dbEj%eloC727jUtID-B{<8W9%Nt~%AOLcO<ijHI~LT?IXAC?QcgvzCb%
zpl2v&ev?uYcB9Y^t{SgrnyQ*z2#zl!o60?e*d_KCXmyWH6hc@NAsdY%P57E1BQ?5y
z&(T)_xZ=#>ATbPGqI5=rjQtjXRnG>urz9|=F)cDX;1mm+jchbC8v=G(dr0zZkJ^=n
z<=I3ks@Wf)oXdT|xt0d4dR<6UD+d+eB!`K}q%Q_U+$r8Bf}xm{4r(X4dPH6jZnGgb
zK=))tn;&Xge}0OXCcQsGGX;z!Xy%hhK+M9f1=Yw~@)xy?T0kSDD&dQ{@)4XW!3_=6
zqQb3IQYc!umBKS~bGvy`m^eLsjlGER*q;*kDu9*0=^&o$LHAkJZWO~txZ&A|TL<05
zs)mS$OcxYd8|^IINGSFr2GrlO>8=}6bOFQR%r+fF@cNY2GwhBq{+Bf8Us1*IL>ZUY
z1)haR#pRof_=Ej5frG+@{SC_ACIr**8n_@{Af+Y5$sv?ZPWpfdC#^MR!(T@RsFMaV
zj5DFL_OyK<c$mv&xkX7BZ(xJ1d{#niyJf`Bq?%}1*fO_@Sq;8o(56aDlhUM2<~R;y
z<Ur?!B{<CCcNb*P>l^e*jk?xU+E^T(e|oX~y}zgD2QPvt6raHa@De9()`VrvK8H4J
z7eE|aLf73xL!Fylhv8BnL9hq`%IrcOzS8h!-pAbvFC%i)Vx6E9I<PG>tNo8?M3VOG
z41se5NQUzqh1CPDz2rIpn2<CE`#S)uEGCMFm;MM;(!9t{kInV06dy5kp<Do-I&`?g
z$%0!&K}RtY0=O*x8l83>v*=)-8qD(r{OTfx(?2=ILg4g|^gX~#D#jpTC76kn3AsG%
zWhBHp5QP>S+Cu>^2rCQm)mw+xSiRx+9efLc>jlUaulr3}9;s~j@kgeXbEuh^2RA5x
zj+V{2!;U)j=PClUj26i?d~o5j-$OEUvd9X|^!HIp3Ay(tl=?FQe@lS)f_;y`KM;7E
zz=U;fQI?c5dj|ktgN+@0#Z;6J6;=6>qACA_xE5k-$_GlJ@}EXS*AWM->WG)=Dx4ww
z8)mH}MrfYrI+z{Wk<rq{s@X3gMIJG}LuwIvV&U4c-=PLXKFp&AhWL<XB@BfNXr02h
zA}cUZ5TTA&#wN?C05>y!7U`m>HY45@VIFIKy4yG+FM`q5`RTfk3>o3m8>f?*M5L0l
zpf-6hSv<Rfl=ZRU5csHPvZxi;?FRPsp<LO(7i>u{^oHCt`pkuqD{mRpUo4))R2DKP
z61L7ATpLb&_=OnFGnLD{zY+}*ue72dh(h7$S`|dI@xd74y!2rk6S7u(tHSly1r|jB
yMjX=A`3p)A#+h*;!|g@l1n~_{O(Of>fQF*~C8eYv!+)l^@I{VNKO(&|=l?HX7@`&c

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_45_36_15.cpython-37.pyc b/model/__pycache__/cldm_decoder_45_36_15.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..345f91ff83be90057f8575752df508ced8a923df
GIT binary patch
literal 10321
zcma)C+ix7#d7s<P&Mx;tF7K2?jbz={7s{4nJ82X-wq?a>WhRkiCz-~b40ndhA?LC>
zGb@S2Y@CKlthS+}x=n#1NkKj|Kzc)grUlxE6bSlI6zxB7o&xkqKpzUU=tB^o{e5R<
zmn+(BGmH85+%Mnx?q^=D*Q*kq!RNo}op@D}{*^MrM?>P%coRjIBv*3fKni6`4iyUp
z$SZ+rsrXj|Ei753qD;4RQCGIg#lIo`D`C~DiZ)s>5!S3)Shwn7!)k<0t0~G$!DKjP
zO^LJ~w8Cj?I-Id)!dYuJoU`V_d23$Om4k(F(OMK~BUlQLS;xZT*75L!bwZR?g5~g}
zbyB3Off=5%P9Z(v)`HXF6V?;5^cl&myN&yj+wkUhRqG7$O?MLcNw2zj7A5CUGUc{V
z(()FOKkq6xrS|kA!c1G?<u{#P&yTuXeI?pnZOeS(bIkRa=U(+YX<OsvP8{8_BbV!e
zA9)Vr6FVM@?J#!xfyY~^A9_jZ^=vO(^IR7LVTjhvo|F1caEm!nvL3V0W9<^3zg4t-
zotobCZuh;2T5;osm%JLpoh@Fy8sCi)C+uP9#w%$`4dN)TtGu=9rEkWr7hs0wb%!di
z`s?fc1Wnp1Zv=7IPVc3*lh}UP<+^vzPtt_X`%#iQG*jx?>nsj!>}cJmbsoR!JKZ>P
zg3mbpB=Mc-Dta(~txvO}zK-TQUQ}%M5hzEZQtYmS4Txp34&ib5jHql5C7xu-u4E~0
z$yHtLz6|=3fq9)zzZOR+i-W*(--?PotzPTs#VB?BSMdHCUO}#1X;4n3efdE8&R1nA
zRWc=$HZhp1?5P7IllHaDIFJXGOx`b{Mn}4O?+jAqtI|75gNbY+t7er<-BkzGth{ez
z)dN|UZeRYQgua!$?f`XP1S5Pwx*>f@$~0Hns<3aNPTQ{@NH0kk(M=~XW^Gr-sBeEo
z8PqUxDfyvjYp|cB(q<j)%hbNB(3ba<K{Z+)FV9DtAFCf<+wBjqqAK<v=X=2F8njj*
zt+aX#n4e%?&vQ~cj)HAA3-q8&b^Wm2<fSxboef@&V!O*6ml>$#4N$2U*qse0io77<
zRif6MQn{9H_dH&5#b3>jdi~U1BO*+=4vz3Z@wL!*-N4I>x!Q|+yy8WDP%}u0*W-RV
z)=BL+DOXX}u5&H%f^{z6<CP(MB!Z7p?_QcGaQb|bwqauqB1obKu5aCSSU2IyMuH8|
zLb*yn-r#z{aI*g+iM@c${rn3TH{#H{7{#f#7ROr`T_<%ec3Ir-UA)<ku7eXh8(ws=
z0~A~&K6&-CH!cc%T<ipHXuBTOF#BUK*w4RcKlj2)Z=09Ft5|`%NR)R5K$11NDo^6A
ziC;rrPz<@I%pkX*EX$KhP2M^75#+p5jMUavS8;S2H*7m{fP9-*?Sh|>uG#kOz7ym%
z>?KSn_BIAyq_TN{1-yv@V8|l>7~c^s(37kHjQFnkp0qBz6}Ni70^Wnnyf0Z=E-kIn
zdy*zuI!J4hSBnimzA)nPA8CTP9?@)%h@~G9HQi#L0=)L<zwwCr)y{Dg&~pKA1FwoV
z`2amqX;<*8EM?PdIh!SJN|sZ_m3QTD$a$XK$flD^C{w76_$5$7ZEzeMUdm>7<?Q&D
z!fvEGr3bTF(=8$Y8(DqO%<3Y2Cu`teP3~X}U5rb4g}pE8)#UwAdG&#8mopjA$P_@x
z0>CQxxs=v+753A#o*9_2)Rku>)HJ}g^{lij=jB@}<7v|^i#bYEpDhUa@C(#0j9Mwl
zFH+?j@&ibbt;$20S!w-&M70m(&Bm_0r96}g2j7s>$#m*H={<Q}aSgX}-$+||%UyXF
zxIn9`qSdrpeIR`g$}ZCZYd^qxfKMIs&tzr5*~|c(%PN5L*#zK1R>Qpa)5XoDNCP~E
zvf~-%w@+jZz~!t7cru%#6(VJ3s#~9t1}D-}`=`?<GBrD~s~~;mKtk>W<qil_XHj-8
zJ)d5Hd{Z)QUP8YHso6cOsVg54X4}-}fxQ1@rXCE-o)UO|I#avyU^!C<C*9`2d??@h
zY5EM|`>OO%e&@%Kq^Aa_QEH|w!c=x@5m*y-wZRkVN_G+}DQBlMbq3NA@^aFh+ACwk
zGkMwIWM*dimi|y>bAlq2hYI^PdexVt!P%&WR0FAc))<`4=27>T*)-sv<t-wG%F~N<
zjwt<yY=Pz<oQpKKH8_`Eq!UE>&$1^TNK~^WiFwZKKTD(F#Gk`B6Inewo3-5Oy;3xn
zUZRnapUGykbJ?+M!kyWJev(01o9+E))AQMx_oX8L)c&&qXD1JUMQWi*n@?vaUy>Bb
zemYymEU1yChsy1%&<&?@dUTr$(w2%<UC2%j&W}c&!pc4v;p=>Xt66t$&xji7^H>$>
zy7TV*p0+Dy%gzILqd73nk=10TyYLBVU_QPotmOjMB6tlm{#TC2kHECXmVhG-ry|ax
z!1Vdi3iGpE+|#L5a_#@Y<kAq6AMU?PH2Se29+7_G@fEug#N5*A&W&qv9Gp63emU-&
zkk8#XHPhJiqfXFwJu}_#Oz2WCB74e=*Ujy?&&&>sljIp^Xo{IoRjyCw5%kN-%F51{
zKXS}|l8=|<cA%La8iF+QFEz9f#Zd6o>oCe7p<%MIPeZ`&Jpa0nwK5Z$*xv9QH!(5d
zFf&s2=7y8d_SQTz@zR}ldn~@=W1G~Fc}bz`O)nn{ZJh|HNX-k~AYOBVV4Joo497N&
zPYuifE4i5VVV#;6dLDZw@b17GHSb~!fj(HSo*8&i7kP9j><&BZ2Wf{C3qABtdSV)H
z(gFb8!46xLz+Zq{C7hK^c6j6Qm9}N}F|>xoK0_6w6Dc;rNC)M5lT;=*E(cDy<~mnW
z<o*M%v~%`Q&8^tBANi?muUv-7;{?eSB=Yt<FQFN%;C=@d6_LTn>^7O#@%%fUD{y#p
z0iW9W<cfJ~!^dg#h&}}VufKltrrBeD%=~m4C<2ZT`{5MgK$zNzxwG@M>92SzfRWj8
z5-&0D;*9g-I_#1gXus~iV_q?zHDRZlm!5pmH(>)&zr%K8LqC03jDrdSGf4?jmz*?K
z*cD*s5uM4<rY!fI+?COI4Hn&qrnm+MNq7U~jWVyZ)r&zge1`G~&<MEUucvkdmj^}<
z`&Ep|D?`#u_|!0YIO06@jl8>K2O5+A1%IjJdcAamPk;`|aI`VEHRbw=vqlra!lXWt
zRizPf-%Lz+t6gGm&}KeMfR@0>lfc^HTeX$ESTrDermY@n*M5<gl0K-LOtzdz;EE-Q
z$LLgFXDnvc3=Ge%gLSR15KB-3%=rpUwJfHhQZX*C`k~Xsa{V2TPx#Sj<BV)4)F!!i
z*73MXs8qo{48A+6-gbh%C+6cS_>rrj)8i!wk}k}4&=gwFe&mVsvdiKg&W+c`Igr<}
z&O5~Wx#K1dM*)J%>p|?K&%IbAp1;IS(oU`uc$1dW8Xj9>_j`G7Yo-Sm86GlhcURco
zw(r7lk0M?@^lTZqbX>dXUw@08pkx*PE8J$1XpCYNtVF=1sz&H+dA9JaX`Q?}N+wp_
zUWeP$XE`1jS>|klz!ZT~1juP*<TbJyfztqda@>Eh$u+r<i;1_X$rlOGDOmLlAoK8K
z2>tA<1RMfu1Udvp1Z5dSbL+k5UOdDS>rxY<G^+_53r+>Pi0IfVl?qBP6oNgb5Ui=O
z4BQgXJo`L#C`W~>rUyiA+t&P1;?Kha344v2FAQsoT_1AD;vt8CqQ~bJG;1{u-NKy7
z#=Jp+791>mO9cm95+a{#B+_x(3{eqh+hd7kfIEdYvL=dkxOU=jJr^Gk=M~%T1WuCJ
zHh7R0oF9eoZ*C&{Jl^C#0i@+Bysk;|v&7FpO2bPflZr0qzpC6)CS?P81KwCwUXahp
zbERReF3$|xQE%BmuAyp*A}?z*VhjzXbMWC5IsfV6zp5G-3qR3IS6U*iDGRbzUP2n)
zpgJkH<OY1g7INq{v5cNAjr_zG=9y8(dk;r1<}mUx^wF$TUQ^*i4nrbv!A82NqYyUm
z&_9tjVJ#twO{G*u7z9C(OwMEqgt#i1;=2?o&{l;QFU$w_8MKEe7Ti`wI%hc|Wt;;k
zO>PY114rG?0JRn9YvFcE^3LLGxt~spe4JZ5#K-gy^U*9KcJqBCQuqp3s8}STH9BG%
zN=6Zya;k1>SiS-;D36uMi1Qd@F`UZ1kEdnyMTBRwva2BKat^*i6;T(8=9DuH@tjI#
z?2{c?<jGdPeI3@Rv|p4Vu9LTs_G@{o1NbNKqYR)*L{%7h)`jL{<oUPjB6=cZgAiM6
ziCUZ^K;m79N3QPpJt4VRNM%9@ou*U-(5~gw-A0;gBwpB)RQD7C^3mC|1fB!n8dxM@
zmncs>U{ywf@923@e4s|>F^7aC!V(}tgdi5cD04~!#JAHLJELI?Vph-syNphS6Ms!u
zl84|Lp4_Kjy@hziwRV|(9Ie?W2>c2GVo^o}A)*w#UT|~A3j*6dTqvxxyVOXq^(T>{
zeI>M6qXN_Km-;o{l|E^xjCzyU9X&)>^dv-A!yt#JdYV^SR~9Aw>-bTiTK>K+LLZ>4
z`!$Fr%`GAHQSqw#69<wzC+f<salf)FTQw0Nff6WcMVJKTjUqo!p6-3Hmenju;D6qS
zS7=ST@c1z9lsn_j;=d(AD$}5=g;7Z5`qeiP@SK50qX3r;dm6?7a<CV2!-h(?ylq}d
zT$`eAp39!W_IN!&teKn)TLjy5=0X2OqZp}yHuN^}COZHWaoLnJJuANt>V|N@iZjVp
zDOPH0#DjJz(*X4hF;cti!bf)%c+`?@q?LWdLR>Yi?bpF;jm_o(#U`{-ev<NC#3l}C
z%&D}6R#Oy>5S~?OwTlJd`S+I}ks9V@FA}jeUPW+=G%NIZdz#%qT~0@2_6u`FAVmb_
z*t=Bwbpo`1_L~5_F~S=emkHaT8lM0~rMT9K5wtU0&+CcQ)Vj}-v_Nt`>V0bbO=?`j
zSgu{HjLo2^ZRAY)HEQ!w(`yur6zn=AYYIH(v`zegTZw`fs||?#q4KF!bnxbqG7{<_
zf)5LI+5x>ns6UMaWR^t2d<K^>3y2R4loZl&SKe0+q=71=C?X(~0)?roA{v*AEsgyS
z`m<>OXlB@Byp~(z>kc_Fkp3j+8BJnMcI-f09oV7gM0aqXVGK(SxmawG&m#<s04?E7
zc)Ua8y-eFC;*+59YS4Q-r(etIn>PBtbXx7uX6YdnF(Ta$kSRzPq_qiUkZFIUWTp4z
zuRvK0%A^obt|BCZpq&W1UCt^ll*%{3@+hky_%CgiDEy!5`(@P02MU#MYN@eV!4-i_
z0eLDx@L#9}v_(iQE2p*0=*rmy-YVV-7((6C)(O4Z>P~CT+%WH&9rKwh<{JKxffZ8^
zy)b6mHbiUa(Dl$n#}*+1az)s8up6!fw+W{}8zSP+B}DeT()Ci<UWk|T<~Wst^#$p1
zC5^f6dUt$y0=nl%aWAnd3^G6%{6YiqDU6x=-G1CpiUshJu-(~1ngh-mZme?6!%J&U
z3Y$&d;>udWzKwc8$Xw~BOlT#!BM5^19#zr8KTPA_Md|Bw*u*Uo!rfIEIJr><y`$ma
zXT}bGw8PO!lg&Xw`*V1c-v=OLAVrWfX;6TZ&UI58=xJ$R$Jxq@Abgx%8EKq1&e~va
z;GDn&8H74omvN<18W_<VNP{kle$*EIM61%CHmJC|TfQ%)IB~SAlHq^=a%DJ@C@n@3
zMu4%(hzfI~Ng$#mFf^vIiF1Tl0xJ>n7)jj2`cZ5{MKO^%35SHtiY}J@&LpfBa5HUW
zM*3alwsvzY%8!dXW#w6RfX47UNz5H#!=G0#=RRiGyRt@yLDrxy?L0Y#`kd#4#2XK2
z2-g(d+7s+c=)_2GvRh&%(Uv#e;r$_9NqE7Sp(BLwx2VH5f%gE~RUt>L@;x~CezMKw
zltggk9+DUPA=+6671C8CZf|-?PG&-Kj>IRQ7+dmNw5!~tX~3GNJC!*F9ssQHIWlSH
zK6<B*na3kZI*A34BqjZpL6(Z!Gi8;7ysZTyQK45oapyD&<P9xhdG!eL`7(9sL!A{e
z(7|omFb(D09Uh3J?6k+0`G5C%bP0>>5V6ohTteovPy|0gg5)8@JT6d#>;r8>7HWeM
zjAuRP5=g!>(t`4lHjvhc3l?!b;Fj`~0+}bwZ8d{*boF`3Egu8_xyEG)JfON2{7d^P
zU7e2AR#8eEC|{OR4Ob7ur8HvAt?#L0?Hbg3Q^!b(+uT#oV-h7KJEvB2nF7@0f`t+5
zkMS6VU2w&BJ6A}}{33323D{PAa(D!IM5hU<tBLT5Mo}Qd|8nx8EAA5AyyI#yk4Yr3
zS_O%d++u$KV72nW?J<$dXiSSN3pl9493%6qnhyaJtXq&d`vJA973KLvDyrE&%DKEF
zSk}^@JZ}r}X_X-Jo%ARHRq0k<#DJ15B9`OmEx`k&8^junggJ;*dQEtf@CB8f*)cLY
z?E4tCR4{~~mN$@qOeOgSPjFNIik8y}sDrd394%K~!g&&0*FYyKJVqsh?1aZC+%C7Y
zS0>$vbJJJZ%NURSDZt2}v&wfJ#CScZI;+`>6Bq(_JR7m(u$NlR2yuz=f=Xw-n+F03
z!9mP`_<J_pFhg(>hvURH9K_b{QCiP&JH+>2(d0j-isC#Om)8WAg>%H^yX^0%$zKzA
zT%54KLD|;`y>tsH!Fg{0spOJZA4llnDUW0}DX0<q5v2lg(m;mo1q5!Y``Q7}TjWZ4
zZe9|$8OVRLoR<*0ZW_r(riqr7&FbbvUV|?Ow6W6KVQE??OB=^A^o>hJ368S(wt;MQ
zeVsm6QCHhaaVS4~x%)5wOwYf)3}#RQ23_x^PSTzfMl*X0dtko}APy|03+BQG=hdFW
z5TZ{JutAV8zW|2=RNRCIxaqL!h~9LVANr94(~*6h2K!?IB-z<n0_O=V0q_!qx<jtL
z;yNLi?*di)J%Cjg6D7qZJA#EYFS6rfb3HG`M*}@55`arbj#c<g@RTU<CuTwrl_kGL
z=iI_95-e^LK_DGxJqJ^?CGVUZV;*q$hx(4#h=eZ*D8WXgJcK7x$bZCy#a2iH@QN^`
z5Z%0Ybal-;PT=FK1B8VkNxa^7X<?+Dp$wMB7ILJSxZPlXfoAQQqh2=k<|=}tj26f>
zeAp1M?;)A{O=RZfI{Fx;gq-^mO8q&3za>DN!2XfIKM}|XjF@F;;=M?v0|0zuB_$>j
z2CTUvSf#e+4ILg#(`Zd<x?F2bP8c=Ox_yD`U}oq&Mhg{dWDk%cR~BFMbO;rR@XgpC
zQUf9q=28Pge4errMZ$-(PT>oT6_^}|1t%*bBV<^BJCPs{Q|UASaj1yMShM5ZMzL=Z
z_O&jI*L`Tn7$3Aaes1KD!p+0S<cj2xW0D)zCyF8Psm$S`R$R9i+P4eIv5s%R(thL>
zJTv@wgf*$ZLH)(zNyp^DX(CkX{NroG`4-<;(L7_h+=Z*r5Xnj>4#PMSo~qSA+!&u+
zA)-p3hA|;a#aAF)e^X#lL`%dW9iDccqO-<rSdQB(MDyay7Q{}WZ45yAU(jpvap<F5
NE7g?ArPjMr@IOp>javW!

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_decoder_lr.cpython-310.pyc b/model/__pycache__/cldm_decoder_lr.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..0ec2f4944ad6736cd61f0f973c118dc17764e8e9
GIT binary patch
literal 10508
zcma)CS&Ss-U9YRFtE<oHx%cd~_i7(AzG4%+S;zKz9g_7V@~)G(NK~rns_E&OI(omV
z-lM6u6VE2z0F!Ve5X6xgh!92sfq3AM$P2GTyz%foAq5Go6bT*>5JSBD{$KSq<4vGi
z_3!UK{@?%pDk_&v3C}Nov*#UoNs@j*ndwJE=Hqx{MV2I2a%Eo%WJ?Yd3kAq4zG|uX
zSA8wWS-D}EZt0>fZ{>&ohWIZ8re%sYnqLe`Rw*c3<)C6!f~r*&<vG6=%vdubt^4(0
z)|w6Gthr#`nhzGNg<#QI6m@xjDOk3aMcVL>1V^o-!7=MtaNIgB$_oC8;G}g@q)opO
zoU%?KU35$SL&3w=!?N_M<d)sa9m%bDi#w`y8u_YQL%!yjTW3&m79}%o9VK;d8ToUr
za#d>1-XrWZ6`p_1>2^C|o2xH|x7V67FMgW29`oEwomSG+c)k^dH|)^my59*shw)<H
zW04(1ZqN64J?R8qoOoT^3)Vf?#XuOMezof)9ml`MoG{*qSm3c{jxSytwta<~UiGf`
zypURP<8?27$&Xsw+`JUs3}YwgV(7|?NkR>xFsm!PvF0VOMXu*#hUyiEDlc_5HhM9d
zG*w>lqqd#gN^B>#J3*W4-mOlY#C)+6#)(5SC7!*(qQJ(EHafJ<W0yKkI|?2DRi_un
z9Vfhm9;~z8qghd3M)STG4mW!blp|3ocGtoN#4=fn@HqXb$b1}cya3=ymh4KF;^th{
z)$YikF&Ws``RvP4n6SwAJ@<`pxUaR#EggMEPZ<dV@2Bv_6-?fi2KhwVllP@>JS|I!
zk}9dRg#le<R~;Csw5O%UzC0+T@?H)#I?^Wo^OvMAmIuYOn3`!JRd-aB=J$-$+?QqP
z`f>CtWHtM!`5c%*!*55GrO<d%iZ9$mEBd{OH`QFJC$mqXy|!oWOV3FIGbv(@(vFN-
z9{qwcC}E~tY>Ac<+eoCXGFs=Ubz31M>?#8@tWK9_V=YeAPp|d*XHpaUkw^R4tI*1?
zO7G>?E;nVX$h@xSBz6?~x7j>Ug%eWUPSC9KToSR?CeMeF-DZx<2${SB%JqD^wdsVR
z=f~V63eKpPYsu}d$8)auD|NzdFR|B&Fk`NROFVEvDd@Pa?`6eY?M7W*@WLJ_9OT8z
zQ7@V5q_&)dt0-%hxfXl=2A6N~!iZ&J!CQ%UE6Ea|2416W*qDQel<0%&+czE7j=8cK
zV?(r1t`d+pxjtk<x$~aHK7!4?|Ebl@DDYOpDDl>#XnWOl5@)r|qF#6PYA?J3?rd#(
z;c5#gSS8+h>E~Ww75G?f`EFpl9@Q}0XDi*?JP*#r-0m{b-&p`nHsu=Ll6Xr<AN!F}
zD9LrDBrnM)l!{zZOu2vb0i?Y$9IL6Vt>M@-ZrFC{0P!|A?IC|5U9#=#J;%>#*mIau
z>}v`<OJ$1yOL$|F>V{O2MahRak!tCQKMyeC#DZK<3Ti<s<nYtQk5cr@7YzJV0g-pB
z?@Al8TX4-g1@J9o?j6a}GHGk&-UX9bI(WIp&EbY2pBTxA_cTF>_Y|b>(IV~<<y~VR
z1H63iJ|0oO);flQ(Q_V&3SJT)aUZ=BX-DunxNeqJ(|O{&xcW(`t{wRuIm@$VIvX#e
zOrbL3E?)`M!7)e&EuG(y(_`BTYbH5L59ZUVn?rszEf1<`S)`v&EBIIA=P`yZ#-+T%
zUKRCf{OY*eyer#!Dg)+I1<*)Uz(QiCxum$Gfc#6Tff;jcc}_x289ZJ}b31ZYzOAw#
zsknJDM~>>#B_T`RLjBUXl@h=8N$DN=E~F6T%pWJEyAsvimAA?}^0pH1Qf*RAYVS(#
z${UJnxP?1LGLyC2k#~TH_&cIa-8Jt@zYSHH>VT!+#Y%ul9W&3SdBC~U0Gv+?fD366
za4{{h?<GrH%b@~z1SLn)GT^ba0(d;F0-i`0Xk|#9OjWl$Ck>7#jlEOJL#dh`-%*f$
zcwa*9IOX;UL#I)8COMm&gA`O!ZBatM3aRK_tf(#T(^%)J&0Tq~nX3DvvPT4dA5GP^
zJUEf6gOhG`&`2I597EFnBP8vq!9&RmVJ1Db3@oAatAmG=$J3LTJ)dGla}wq(rZsnF
zH;++HXJvzvX(QEdNWUVpFN@Pxkh*|g<s;JIOsFGOL8_cq24~Vm)UBnnfLF2>p-kn;
z3Y{5BKc6nq?1QtR;?@UelNCBIly9UD-<7ClJDcb9-YSg}C&?2Sr<j)0GilwO-OYu?
z<VhMC`MGpHJ)0g)i|*Vm^q34P+B(0tnw(8fe_a~pAK6<KcsjWc%ux$X+IloS`JALk
z_M_<u%z_#jb6#IgPi2(nR+pr06>IuIdU9}XJmw5azBk4eoep7Y-d)%=!hG@+)<jzI
zoV&QI?TES3bHLnq28?rPEosAD`lvK$99$DtQW<0UUo#v&0Lu!K1(xq;ErwV=J6>P5
zy5(IRt(182|G?sr5f(q(e}`!D=@AZ*e&*nMT?r!X$XfsP%TeT?I@NeH>NOy{+fmX;
zqDCid`90TbB%59XI@Sxx_G&~MjoVR=HCil+<Hwzm+17vxb30^8LGP@rtn}Y};F!HQ
z8!ygm#zrzR8foVL*U&=jLh+kdV8B5d!@Of3hal}g{YnRGWesR(d((5=xPcKznXwVj
z*mPpr-n!R_y`=wkmqj-^*d{e(UOcn`8eTRQ+BzXnku=V?{b=3s{oAxvVPrOGd}`3}
zv69uK2TQhbzU#5aeeVV=*TzkZA<zda*K7D**hU^5hW3aZ^!%hnN@Vm<nsmY}-nb3`
z%9B!QSpt6nZs%YL$z-=z4z9K-vuDuyJ%!Iv#rRZ)n_;AHGQCPFl^Yj)Cs=o#iwSZu
zX|B4|KQmEmE4JMUJBe+tT!2C3`0+&~vL5~C&<u8SuLXOHh+%A~H<;J*IyXF5VDj)1
zKGuJJrEzVugEQ$8m55Wg^2*h#jV|j%tdraZqJZg1Kb%G63*$U)^!txBIxF4^VAyCm
zu@^UP;-s@fn{>&{x1aBPv2n5SL<5$6<Ecj;=`>&=QNKyMsi7a66yvyjzY!+{sY^y6
zE9@fhbB|6%DDwAYp6@!D^P}++?7cB%aSdz|^9qO@WnOE$8-Z~69OYx65pc7!k=P-e
zB$!3)3dZDx5rM{hW|W+axJZ3N@8;BzM&z-<!^*i{H`(MxP$QX@Hs-cw+)nJQ(?qa8
zsZVH`G(zUhi3zVYbL^9}8S*ql344v!p}FwJno3qI8juCkR1dXleuU@Z9%!75x{OGy
zIhdhs2kY2aAx5ACSn@?0|AZKtO2s(b>;z65%j@(#UhIV9jWV*1P#b6dTg&4rp-u(Q
zu$M*EXC1%iy~0?;xC$=hYT$Hv4&tK?BOWw_*7FZMVV-wc)WtdR(lp=kGS+#6I6iaj
z#9=5vXnw<woaD)8hl!`3Vkc=QYXm+`OR0~JDYkpvthY7Sg%b@Q8kW2(tZ}>J!i*0?
zZccn(Mh+j>u6C}x!H!eXgog{qn&cRxhz83MXi!xpaJD^Lc;B>6ZjO_&Rkk<a1oc>k
zM|PSz6$#7`I7NWmNmd|GBJdCZuT6V1HaRNiGr{mVYC>kSI0dV`31m)AhS1L(0_y}?
z1Y81R0<sLEv5oGN&rYz!HmM0wmQ@9g1!n?XL}ctUR4OQYs0QpQHDJw5W#FWM#@XxC
zAs-GMH{B;{yKOBVCiX0V5VKdP`O>I%xa$c=EKfKB6g)k*pjoRjaS}78n(_?=T5z!N
zG!-0hP6&Fgks!xuGsI1t+a8N81KcQdkX0P6!?k0F>zTNK3@O-l%Xi|~w%H{tmkC$r
z4WypN8()DUFi&W3x(v8m^fQps@KQ-l))if*pDEYnnqnYt$a8oX<kKiqWb?kB%UbAY
zF={JDKo0GHtm%p}CoA`r`}+O4JdLN}S3_$>Q80pncG<u7BkjIXG%>Cz)_@V=_?fb%
zEMY}TUVlGpC;I%iqN4^qwI6Fgs+H?<1;08<#ArqRKPGLnl>T2z(U7(K%D-#(wL0y`
zKo9LpCgz~^(rkzOXzwePpERqIn{KVWum7-gye{`g5g2%FW98N&#W#8=P5eUC9+NOi
zq(ny41u>UYPGyR_xGGukI~3B<)`Tt)mI(U<T0@!(2F@d$aWQcR#3^ZBZj1sDhke@s
zwFR8D(80&#{_@LNAb^&6Pyn<@G>jhNX_|@Vh`$Yxz|p``#WE2d(h=rSQVI#>6Lm|&
z1{AnYSvW>UV9A&Y8WoZ}dLp2-RoGGBR?NcLFcH?F@KHY15I`!V#vYlZ!#tVX*Uw;v
zVF?0BSqo{el(pH1V*~fgxGFJn+?!<)j1j^`$USxhrHq0MLO%%k%GG|SD?}d)sp1p?
z5?d?+XqGZb)kK<WB%jzLRQD)>6#`EXcoKkX;IWuJMS0>St1y<Fhp|r~hZ13*f_D__
zAVMuE3t+c7xdQgklzx5cL`Gqa`Rv>;px4mPzbp*U3Al_WbNkn>Ap~-{nP)$P*6gDM
zewF~SI-}zkfe&6Dii4Kt`w$5f-^su&N(tfcJW{l)m@GX*QVP?A{?pQxf0o`hK37v2
zjX>i2z(ce|PyB}fqqvDHd#a~-xea9*;Ra80RZn*{yg4uLB1*nd648>pYuqV8)=|_1
zi1y6A;=be_6?G-Ie5bG@TP1RUQCl9?!a=<Qx6!H$^T*s-ckT|@%Bl`a;FgINHFwcn
zLeCj@**${)x`@Keg8hz<qcB%4y~gLDASrrg!zhPsfgH?_%r2tRZSOWO#I8-DJ<ny2
zW1GC}Be+eTiY+2}I`g1T!vg_X1MTRo;*H+|pb*WLoa!AZ&3{+^Iw&0C0qagBTczNv
ztq~8}xl{wxQv_x0ybA~4Rp6RSwviO}5I%L)q_kHCw^g>P`xMmB#`zlM+X!mx)0i_!
z9j#_4kRjZ(+}gij5qSRXg?nZE@w^v`03J6HMI(g`God-luAwfYF7h9Ql_QWNB7f}5
zRQnDA!UNl(CY3S%$VQD>hibM6ATR}fYDI_-8m{McMQUcF!{TI!>}=F`sqtNET*6qc
zJzN=^Ls8SnSocfR=0VeI6!FZMSP;9Q{joa4Wr!4$6seo$Q0eL5PrXp1=;Q|#Z7kYB
z3J(_Tv<rHLXnza|$S#RQ`W$F`3BiVel0ZW4$a~7ZG*E>kML300pfz<(gz^M8L&|?0
z{n;!4ls-&7Udjyn4Tt<8)RVktG>KJf*}k~9umjHtZ{RY;7?n)%H*J(JBHoNhFyT(P
z&?5xCK$|BbtYgchSNjjWoY6XM_9t{&Ytd#$53%={jK7V{kc>fMTOhVn`+fO!2_&5(
zlKuiz#~@GY0ktY3J-9&-vAI)e!G((X3d8}*3b=cawsLg)km!4P)XDn_m2YW@u~ong
zg-ka9R6=(Wq$to9vAHx)QJS`#7V(;R3t$g*SKA;&Yis@bdSkP3v(aiiezCERKV;zR
zmIE({*lio)H*n~ds@Sqc^nv^v_G{P^SN!XQT_6xac<38q0$yl)2@E`h=y`RTW+4y+
zA#x>&xbAv4I&e31uM<Yy*eWnc2H{l*EyZUrX3}Z*qFy{)0M7{rf&Dto0gn$CXPI{5
zxpgOjVJL5NWj$u!K)oPquCx;-G?&~Gq``iJs%YWt5XI90y+VghoO7B?ARIjXNHdw*
z3AsNvMF8Xd9XN5aL&!KFUnQoQ$T&z4EsgUOE2VSa!Z{}z?pbi3L;@bCmq!{Wj?<Rd
z8JrXtBE>Z*T0`*f=0For=g8$#dDxR$qo-(>+tmgID2DtUDKTIK42=oIp$nt2MCsI6
zO=FFH3#-cvDMAJuQ^L}eW+y%pf(nc!s?r1X2SO(*s)+Ef;uQj;jpcsPfHedDrtQp0
zzp7lvZRmvgaF3|0Ji)$;9&l<&_)Y4B&gJHX%%Kgs7uV@n$Vyz2`j1Q<LdJzc4o*i)
zgddD<O_RNePV75k648uT-O)`ZU50r6lszP*^4F-tFA(@e0E|FF#me78;Gh%V=5j(Z
zIdmr!#@?d_1{Kn!DDI59aYktqLqZ6e;?!#2pbe2FRsogOloHuH1sLsQtbeee3F~L!
z!5jUM<vbRzHH=5%l@wnJ0ZVsxO=XSaqRv_(dKHS-6E|MtaN)>Ime&sB{{(gELC*~Z
zql3G=Q5xF2Jvx(`(r->J^{4l`_Y@Y{B6=A;#4==G3k~r-BuGd?{^LqUh(eG#M58vy
z!LrsfmVqG5BQ2N!X#;7E7-AXM4sI?>DG-Xn7FSaUOIKf%-2726qibA{zz(Wgz`wMo
z(nao6t%*`%NBIISc5tOZEK4K8qS{rb+Eu9cmX47Wx4NsKM-3$;N@vzG5d-uL<;-tV
zYQk<5+QB8`^-NP$v#Y@IHDpt{hY-8O{sOJ;(TPF`Ya(Q$QKSjq5oDxB7w<Xx9srk{
zSsWyWp-Ys`NRYAL0<h}Y;P#XRW;CWnW(S;NVY88qW@bacPHPWIp6yY)(y%<6NJTaK
z1C(>QFF4oIpjEF6X=>%50-WS95t;O@fQUQA+e9!FlhQ%$Bv+5f3&L$S<Ob-TtZ4HC
zP3zB35!0mihiImNkp#_r0ttv&*tMV<c}xC+mQf36gj6MbF;_l<Qzf{efm&3!l}ZXl
z3%62uW^QgbPYM&Kr?0UWFdq9;0$%~J@;4pClRfA@tJ;lX*a$Z~8*%HPn^@Hl(U9qa
zLTjU)g&PUQe#C(KTQ=QtLy9h7IGow0g9u)q(t3v75yt<L=KL$F7@jEO^18sY@Tj<a
zlM#Qgzb0@{xUj!L*;|BQI$i@8#0#Xfgg7~b(#c655aFb?#%%aaWPmzpAj3ElI%`kc
z2ZD#WT$Wpugz*M8*ve-m#I{>T{7kBemW3^ItC-c``vq;Pv@|JA%4Ck?Kt>L9ZdiiD
z9DZ>@2ED#PAJeF7O{I;+;rXW*+TZ_sdVcr<m_qRxOaL!&;$}@)*6g!r!*&70u_bif
zJv7w0*>xB$1rh{{5TMMi;^8Ze?&NW|!pn#pwOA+Ugbr-W%xeE58j++uJ44_c0g~Z7
zM`87VYcIM^045}j!Tt`wDvOEY;dMU(l{7E1(_?czE5+vvT__iTrw$#iaI)Z5QP5G$
zga9s!zecBB$1FP7rv~$U0l&J4;q*^Vu@E@@Lwyf0lZr7&SP5n#WkN0wdkG1#4n(2F
zhW1dv3&P4meD&7hHCAspeg|Jc;Ccaa#p`~HmPaZZe*BTC<s51z=D`ihpQB}S?y#dy
z{ke(&Eu%$p4WC;0?Dvq&oGh{eGyQ#(QbO+i38nswz~2%ezF^-c@DBvuA~0c{Ta+c`
z%-#mTcVJ@&UojQs14UJStZ2&rAg+bjn)1F<sQjnV&~?NCt2*Lkx(a6q|AtvBi4mIT
zxejKBc4V}4v1;~9NRda3FOXV<o>;hc?02XEkq`5zfgwJjSqVeo0$Qi=mB<PV6hx@w
zm9fb(D!|Q*pGCSTs?CUZMVQB$pYAq}$ctdKb$+_;LqkUR=*H<}CK0LRET~N$Ocu|s
zAZ0x}90H&7Ocu4`y4}FOK9nmP_;xMnh2D^xp7{vAabe`gTLyI&D<>J1#mk9)t#bz#
zhT|T77)Fy!<ucc=L_@?YttbehP<Xjk1)*$wCx##|eb2^(pcNmgaQ!ubLJ@rtCp2~V
tf((RZW?aY+dy&{ce1KDv$oMy)p6Gu`De1@XpQ$Z;j-%9%NN>;i{|mK=pydDn

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr.cpython-310.pyc b/model/__pycache__/cldm_lr.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..5cb81091fadf1db0f53ba3e87bc2b60d5bb067ce
GIT binary patch
literal 10270
zcma)CX^bS-U9YRFtE<nM>ACi-wfAZtGrnRIyxI6zuS2q)MBa5A7l}$WT{Sa3Q%CQs
z>OGq3IPq-aB`}6?Mv6jaA_OuZ5Fa=sP$WJfBtG-ueGL*?2?;)M#b7VL|Eunvp7AD7
zt@`)7kN^AMui|Rel<@rfFJJWk{sl?;5oIPH4Vf44CW<Ucp5)1a6w0<7DmDs`R|3^m
z@vjD2Sg;GjGTqiiUC}NM{|)h93QgM-ZM2{qR_sbxwX0#xu7!2GF3Jl*Bb>6QL|PA~
z!x?)fY}(Cm)}9UL?748>o)>k+U?E(z7e(3#j)X_;qv0|8Sa{q%F3L*5iSVR-Ql!nm
z3QyUmkS=?b;Gytg`(asnRr0D{?T+Nt{P}IwK8<|cYark7&CN3?Ig64hZyF`j{vz_{
zJmsp?nz={VX(_z;n%nJmqIIsm7~NiJ$-Mk|=K0L`E_K>zOXJ0M9NlmtkLy7v@?FNu
zeV@fn7<;|I=hJB?^pn)@I)1q7dmaYD5Ytz?ZrX8!Ys`(3wU~uIYZdtXwPD*=sOeSz
zde4uj6*pe@lb3?Hy~WK-@y#f4!!Cxdy_lxdAdd37(i<y&`daMy0cNOQajEiBXKk&Q
zph-*RwIE)1(p#zHCQc_@=emEZlcWis??g%J(oCuEtg$$Bu%opOt@GHWj=LU5Zt$wx
zOOlQoT|y7mS?$rRsIQ`V-;aiyy$8yXs1&<vV*_HDtW9{Fd{ks!z?;kg_>wJqlC5|J
zPxZ7rGH6T&_H{n<avY^B4g%kMBO2~&<#Jm`pV3oA!od4Eyh#m{uS<hsD(%X9(yu%%
zOR176nY4)kJ!MB77@4%IWyYR7C}r|)0W~_(CjN_;q_0#5<*b~USt(PuRg@NYjm+GW
zW$F5H^eg2xd#HI6%%I`7t;$klJSim?ZlV?a-ol${p45}s=g?l;HTR_Fq=A{1F-K)v
z#w?G1Ss7F?Qz5ZMONp(e(q<K{i`05uAtdZ512d{mmgi&5Pt;GY_4*ew6Z?@z``NA0
z%CAcA7gjE}WV_7#uJ5K!90j-8EKr3LQoT;ts`Ek`v-SorMzOQbT#pelc@3272Tps#
zjUqotxJeY8Q!m%j+g+a*Jn>iQMBQHMtP)`+TnCr<;DSon@w~v#i@Dm3yS(H_Jy1Bv
zi&x`bI?+jOyD3*u)~a$X@q;xk-{Pea%OrxgQvX((CqNCnLECUJ2N5aJ2iLc5x@<k+
z%0_|>(L%XOK;Gc`kO}3^`x5&EHuu5H%Nud%FGq3eug3A#vgfAm@;Zxq-Q}yj=nA;A
zz2Qg8ZJ=P8c;}^`dwp5pW4Rr8p%bvB?rmNK$6`+J2+`eH08KXK2HuKzD@Y&vu~DkX
z(@I5NkWVNzxuTeI|L6lqdTBUTOIumNacSIeoX7>j9d0^9zCya<IM;h_kk_#1FsInh
z1bCLp<^dM*CM3-bsV0k(k8vQ?(vy4%V8nqXxulfTl2$6<r;8t@=vOQm_^ARS@73Rx
z)?}~bnRiOySIF8slC9-Z)-JpY=CF0}aD$t}4MQ$5k`3=`f&lL+NZ+GH+#`y+#y$mj
z`QCjzqJE`)3<aa-JQ6j$BsP*hdZp5~;Bj!<46A3e#BoXevrt*v@;h>#XRT}|Swxva
zWyDp15~_n^kP2EhyDewOwiMP%3zQzrW_7QC{Bl+u)U&EcKcCg`uO`o93|)*%d4;_y
z>eb}cak+U{cJxdJEM^Lzk*R>C)XWNLd0PS5S26=L7S`pagqkXNyOI^Q<-B}LWno(L
zieioe)n^MrhI|$E3*%Nw^3~5u@5px{eIQ@{Ag$b$sP3-3S>2Ymlw^l$(|X!?S9(`o
zQ#`{f-7(Usyxq3E4Ll^@7j33J^RDz=sK`tQtb7kE0VZ|KJd+gxo0$PPo0S0PvNGU&
zR$)I(7d96o1@H(;j%HQBV_6OGcvc5Ik<HP{kUE*EUbQI=j;Ge{sq~>t&5my?NI$$M
zA$Oc|dxW9WC_9s$P0vC4DVa7ep<j(u^A1+DF7MG;=c&zIdAF6Rd!w>P1b!dQ)OC4q
zB2xz^z52jPA0r$?!u|s!?5V*+=@elmJGBTbq4b-BhttQilbF4jVMR>|bC$D)H?>p5
zD5vwX!O6_Z^c&Kz%Ixdn^cAEopjY*XG&mFKNY#+4X0^eYY#wzh*$m*7yhS8adAdYr
zhSD!(3pD%SY@~S8gR|)pofpd2vWM?VRI`=Ob9#4~MoH5235-+Ds@a)r+MC%aMCJ5J
z8X5U!Hk+Nzj%H=AxdZ(rgNioK?=Gijv(xWM!~7$=%K}d)_kcNSp-G#MW+$JM6v=rs
zJAqkHBV*3%i`l81^1S+jw54KAKgv!H&W*>MLdg%u*rL-ROwD?8J4RGYpTe3*8=mv#
zceHIWS9T7V8_$4o4y`4#yoFCn1MA?Lu#(ys%m0$&@Bvs>m@Ke-KW{O_^4anF^3^Tw
z=xC)R^Zy4HkBqSR@&4OHlTVLui1afD*Xv0TX-8K2uV0Si;M6JWt+;1FcCW{&mBv;l
zY6m^fx6%#Yf^PLAvbn5y&AJ`;nAK)+l05E?%ry(D%j=K{1%0!$wA6p=fn)ZPe7q#L
z6|HpNSfrW%UqcJA3x#i9f#C*e46}~C072S+`jrmW$}DJRXTx{B#KMT9%vkwb8*W0|
zTlKBPPy26oS$v~|ZBj$#CqvzD`T1CA>qbCDYMoyX;#D^YZqrtUaoM8rseu(>CCh0K
z7OZu?>$ArL{{}2p>n6q!=!13ZTY(>~BaaS4JH!clLE0uIGI}UYI$;KHG7SLAlQO73
z@E71#0Tz%<_ImB$YFjdU2Cd&$c#|r|r!w3OBYl(WRZ^+kxDdGEs^?zpA(xQmnwI)!
z_7&Tb<8-1<>NramU<kQEauJEVNB=oAgT35q!_FdN7#rvo^V@#shVKbX9$vzy`p+*}
z*ETvhlP*z-IE5>(T)k>_Stn+l^fnL$Oz-!@S;T=b$`h;Kf6VGE`AdM2)pip<v2Nm|
z^F!P3lACQm-}#Dl(R#vyMQ=Uz$RiyKmJ#*aZ#OaYWBbK8t{||IlpuA<31o>~1b*((
zsR%{>zRZhVH+On8UV)uArYx?3O%h%Mk)zCSZ*^l34sTLE0U7}}I%}yD!8w9i#I9gW
zUK$Z-!ly>b{SoJ>Z{**cIMSFrHF#D9&+n!iybNk2bJD@w_LSF2+*O(gb|>|TY?DUF
zy*M%9wN`<BmNr8khA3gL(K<8_URX=Xi$w#nU|Q;-cCAnFLec|`lTnuwiQR<xx$a^e
zYfHoklmJV<NaLRnLsO|3hnt<yUB~h|eV>;*(RibbtRvJWx!>0Ixk{*0!87b-QT0VP
z==rZO7BjAb3%MG)U0#6rScmZr8ba&Y2c9r5dMxhZoOorD?|2pKyg?kFyL93(6d<&?
z7Q}A)<g>%X(@(LJw38JApQoivkB%vEdfmLY-R#17hVKju-V;{1)A3-wM-eyoeO^Y6
z9@nmRuDrpHQ__TI3x}HI7^6rAD-f`#susFiz9YPDS|>Nh$;7TYYjA#gEXN}|O`XaF
zrU;xOKyD-}5vUM&2!J;xy%>ial=Hb@c$1ot*(^@Mu5JLC`zJ%_XD)$N0&M~wfiVHu
z2GQ7B_sM7XvBWm02~n0^2aW}20$oI8><d&XD0`>|oC!5xPfcVTLF4Ro>QIb^4x1hj
zwcWPo4-<PHI7rwl)O=x7JKXg?M=b7h1SoiNZb7qlZQnV}nQFo(6llT0!oyT>zy%@b
zxkiE<r_B&Aac}!9u?=ve&_Q;2xDL-rT(0Nh0y3oJIPJhq631beuv{iwoi~soIi6gB
zA}~*AaJme*Tl6!K((qDAL)H~trk^QK%MHaq-jJJk=j78UQ)KhLUdUVMXfbLlMnDej
zKh$(ZY0Aod<-UHuS)}na{2FMjC<;bU&@TVieyrU$$|lA&#Tqao96wXmlm)CvDe52O
z?L?pdR&>;$r}m-t<3@E_uHiS05;0m?|BwB)c}o8;rEJLBedXV@``R?^$3PG5YbNHP
z_0nvI`)D61wVyPrQkZP5eW3qy<@mJRA4Oi^wT+cqn-t&Zp)~OeQF}tdD3wwf(H6v7
zGC7ke-r}ib#cxw6M_UoPKv*K|GiVKIE*Q9obk4=Z9T2CadATtPJRJ6IL)0RqnCswU
za)0sVJODt;JSYI#BpOB!@ifgub0psZNa1K;sbZN33+V{&C>e!<imAG(VFL=>r#uWJ
zBcNnV1dB>(5j_z|+AM7=a4TluY?uh=P}r!LX$TyZGGmua(qW!V?(1hT!>|Maq`ZZ+
zTglt(!Lfn+Wn7im0v6Y*iXe;-E<)}x3NkPXG6?-3<SSSEovsjlETW221W0VL7@$?j
zB~=S)u919Vk5JvC1eORqLEuRMu7SrA_7vrx2Cz$G$$1$26mqB%_9<vbu?`~CiZTaw
zYsxjSho<zWCr)G()|k!D{Q`Oo{rt<q0Nn?d@#Jp*$~A;QF1L#8XV98`lEBXrAXaB|
z{37tdt3z?n_JaT-f#NzjxJ4-;9G*vtHk6QrFeIfkN$5W<UHeDrZR5>`%4h@<-v=I|
zC3=$I0~p0kJlR)$%`dDeiwHOPny31@r{OL5MGw*OwTg(A6g}fk1+tE!CP1`r?w0o?
z@2IG&c-1?lZP~7n1B}}0uoe#L9k`8lZJ0mi&3Mf_U@N;mEP-1lS~R?QZvj20yhZN_
z{-;G0W(Mqcd>n<ja_KeRgo31~nFFI7wgqx9KXSW>O1J#myp(tjh4g%nJ&tYiYJi|N
zc`A;G;OWeVI*kqlVhyyTw~jaYDgcFOHswt3NLlfB<@Z405D!>)COIkvWgU%p&?#gZ
zpq?QZ>l8gW_?`mSTyl)Gw2QE*r>2$ND!8q-S>L0ehBnSOD8G)N#vYA1l}@A86a_Mb
zn^su)CoBTbzkd2&6@R?wM`Ua;ZX$|C3L9oZYldAzT~1x(KL{&FAV);}*w?A{9Rh?0
zwoOfHWBiegny?PlY!X0V3jEZL5g#->-|te&o?7d$Bpo6<AN5^oe3u$mFqY>GSH_ws
zY8g4}evR5ZXnKtzojDT=Vi&YOR)-exCrOcM(;O;29sH@68Wf%UsHTlYJ4oTdqMdd@
zuMq8zApzMXkw|ZXwigg=7$_+u<hH!4>`4PvNK%ATC<R(mS41dJa5JR*4*Ihh04RN6
zpI36je$6F+2=yfI8BJn0+D;(uE1b}GqZ_zHF-9d5{7oC>^N2Ge0!+9QF7yb2FVN<R
z2&b`S(yRT4Ue0NqHv1Dgt+Z*gqo<4nsR}{Hzk$S%j6q_XAht~Vf&97zk}eQQL-gx|
zBGd)69^yQR`iaP#mEi&`Q@;UWfHDI&4$@`;XfslMw}?7<PoeTnEj2bvxR;PA?oTC9
z45TLLe!|RjL}(D1%1pc^yarf9-O<+QVoF=-Pp?`V)=jHzJ$}(z#UC>8bE}~r#_YBO
z(HpvSGgWRoBJx1K4f{6s!<FDVVHU_k;2pY#7=V}7{S?L>!t%U6Nv{wFf(W^i#$5OO
z8y&bBy5EW7Zeo`hq=N7$goff%7&Gmx_u^hMTmUZ!=Yahh%>i!@mt|#f<h-!zrZ5WS
zEv~F4>^;=yBrP<S+!iFkev^u6(d-bV|2E1l(_s_GJVd4s4xE0ZkxUJiL385Z$9p?)
z(qw<oNt2I~{1yNi1u3GWNs(ftbl#gV4N?vFD!4x)!H#n)B8{`gIZNzJhL|rz8LmCi
z8UlZ(K&N<>a-fP~Pil>xqFrG}8<e0Big%>cfbt(25r{vRMq`Q6iLqM73L`qq4Jbke
z98$v4ghnU+5dsPfC91Lm^#?*GDyoR!ui_PgVjau<iUCpvU(<G)(zliCAhLa?!@Z%h
z^aT3>dcc__p|@WrbS*b8<c@6Ey-1UZ5Xpkne`Mkia_$rIZ!$_E{9klynd~i$&wfWt
zBAW5KH@dl`OAtSpu!Dp&{zdBW3k1FefDuTT*u`527<7``Tuw<AN8SW={XR7?sE{r}
zaYxima!T7bAcTM^PptNfv>~#<YM`=)QXxC11Y@0y^pEB=VfD;Cc%vV)n8#wZf$>PJ
zlETX%Sm~~=rL1sV%-IV>uR`Ja;)ZJ+CLEc^^2%ZSpP(*1=(nLjba8h#N<(XJjLxK?
z^ji~4{pr2#kxkepdKo<>B*?B7n&GdJAmIqPkE<9V1VQGYBW+N~3OU0-07A11h9O~C
zL|P+uScCxe3VBL_Fhmb&SIx9N+0*AGuXvO$0*ni=iSb{;Kh+iw)S4*Oi7_wWItLdS
z#IQ6X?5G`eVw4*7-qbOY;?;K)^k|@jG}Y8fE?|J1p_CyhFu`mT(7~1B^;|#IJ)i7{
z@ug!+xre~I#Qqem?$LQd{AwalqfvYbUk&7hM%U~Gy8XvhW*+@WVBQj)GZJI$I{@}{
zKDaX>dl`*sld%C;SXgRgnVI<zu+G+7rCfV{pW0Q1<@rP^s=3@3ENg2}qt}HPwF{8@
zZhDw{O!_E51e%h7D20Mhx<Wd%A|H|Ggr98476_iqX7eLW>(5S5(SGj(#6t3w2u2Xg
zXOMt=g&_-akvHWpYeXhOdy;m9^X17$U?>t?(;!7v_>l?@jr>UAig|^dBBW#q2dA&F
zPh&jxM+CkBU>9$?2q61Vcy_%TCol?b_znWqVK=qw5yBsn1*P`ddLCaS>;^Fd)^9m<
zoel9hhv9Hs8!ke2eM;*&Vu#X&{Ta>q=TtE~OvdF^foI`Darq`A#$bOz;Gk4te~GfU
z3Bh!-25yAsNLvXxatNjShkQVGllT~m5E~hwP8!It#(=_Ib&t?XF1VIFoRN(Nj@Z;U
zi+K$`Lx2(*mm2$}X=&5kECb!J@$lh6ZpSGUHdg`$S8^LOu`d!JiObEd%MhPW5x2=r
zWAcO;iB@)oz&QdWFnNJu*C8*w=(*vWBV+Gxs74oKB)IqQevBN+KY>R=@jB5Tu}_wK
zi||%MR~_LF9hjSgJqrWx<OCr;sNVsSBw2=Wa*`xOlBd94LP8uBg14kW7%>zt2?GUT
z$y<kxk=JjyK?ffVASw^W^t-=KOM@hk;Qk(&Sk9qlVjf&P{0UmNnuq;c>d#fgH5o0E
zYxpuDVBbSB_f5!v%GKXdN(sU7hm`tb0)It-sEhqAfxja_F4)A%g#q;}mAwscf|mRY
z67Y)^Q&B!rROLfOQ~n3xAcUlp50p~vKMhOQ5k9Ny2s7y_d<gs-W}_rVXr1ReSO&_4
z(bC1L*)JhQjwC+VX%l)9;SsU#QUf9g=2HVhd~vcAMZ$BkPvN7ACFpkqc9W&C#vc{n
zLL<lnL=+5WMFIwZJv-TL9LN?iT>JcF-N%NE@r{bx$t?;JQh6-6LMd_-8Bv+NI2;0B
zzw9q+$@97)BiFz-s9r3Nm^qImlLKj=J2*E^boi+hjWdzUouUd2ku0_2FpMMN|JgMJ
zdGX~G!k_eo7ZVame8R!?*94YCTtXbe#Q6$QL+6XjsvLJ0iPFWVDGiCtV*_%W{uh*r
TehmLjy!30R(?_JYo5BAAF5OKt

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_psample2.cpython-310.pyc b/model/__pycache__/cldm_lr_psample2.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c854963239d0e3287c41cd014323f21a6dff711c
GIT binary patch
literal 10342
zcma)CS&$s(UGHmpdU~$i*<0sG)-|%dYs*PYu%g7VBnL-crz|;+o20sWr+a5-w~x`+
zJ(70lv6HM4E5Ifk;VO>o0tJ|YBE<to736_8igzBq*HA^NiYj=ZfEZ%&`+q&xu9d({
z^}oOS_<#TV>!@5dCH%ho_3PfX7bWTYl$rc#$b1S<tjLn&O0Minfo#cvVxa(e#aAsA
zZ`Id=oRu4t>6R|)@>YKEHpIISn3gHpXnrv$S*4(Cm4k{^3943Al;`}ZVA`4%Y2B{{
zGuBKnYt068)?6@e%?AtCf~d>;i@}n$B+`a|Bsgjv4USpIg5%b4QC9Fz1ShSNB5nHh
z;FNU=>7rZmPX|v}Psq~Cl3R8w_a(RDE$pb)8RV<(6!KG^xp@{P=TI{3)=*OOmXJU1
zDmSIZ%tOLXL*e;Xold76wz&FYxV_qtdGR&oddzdLwwp;q<N0P3-mycM>wY`*9L9@1
zk41J6xn1AmwWJ+*apHAsFIe+j7Xx94+RcuWv>pEzbHaE%Vu8mRIlgdf(DoWNz3JWV
zdLgyq#_L}EX+LUiar0_)H;kR2gP|)gCJ8l&!mO_F#;TXR8o8d28LBrNs=V4>U+>0f
z(olKDk6LzeFR`82ZU-%{d-vLL67z+27$*+Rlz8?!ivk-vT5r=jk6mp$ttfQ-mz{1L
zx1I1Rda(9dmu5wM8O?iMIN0n%P>w{U*j*DF5X)pu!sFymMdnj@;&}j1vSe4X6gTIp
zu6ADrjmf~i&SzeV!h}V>@40V;gMF=DZ|dkX{FRX~@O&0eT*2fmsh>}zJ$YaH)^oCy
zD5;W4n;6hlcGbR-N_$#r?92T^D(~e`qa$tNoxdu5soXE7#nem-sk)=0G{0x0=DsXT
zw~wP=A*<O(&6{8b4gVcgmO|rMDZYFct?2(2o>X(CuFO7*_S&AgFFh~y&7_DqN;@)U
zdGd=&zl51`u_ancY(0@S%V?da)-8pQu&eaVusT_ujkPdQKe^W1pG!^bM;`8HuR<%o
zDZQ6lz21<mBJ(<)lh{${Z?id|3MZtx?VwTRxg=uE4W17pyTu%r5i)rNl<WC+bHfQk
z&yTrD6r52n*OKjy$8)ZDmD*vao7iham@(JEB_6n-6trE}_p)NHcA^e1cwrY54)WsV
zsGCf5Qkzb~Rg^W#T#G$_oy+%lVaPJE;H|{Fmt+Z01D~R8*qDQel<0%&TX!ASin+28
zV?(r1t`d+pxISP)x&5BRK8nq~e`94M3cQstO1!lw+FEg)#93*vsM}e&*$r=iJDVF`
zxY7g)R)}{#{Zp^62z;zGeK)XuW_L2~yR_Wd=6SF#CU=*J0M7wvvMEpDDT$|q^syfp
zg_2xTO7fz7LaE3l#guzTA4A*AgRvUg>MG7n<A!aA4lr+X(;hGw(k0uz-F5t|hCPot
z#l|MU$Ea)pU=dGDV%?A`vMBiwD^e}};x7OUS+O7&l!98&3OW3B@uw91^92KcRY2t3
z>bugq>=s<}egTXNp?hDlv`pMuxp%=OmJVi~;^tt(5KxSS#Cw{c#D@yf4`~q(iS%x<
zPXJzj_y|8yzuG*8g5mE15*0ioKjI#GCDM*ycCg(HtEO|rdU5p?=&l|4t8$iSjdUhn
zLYYEk#9qD<sQqIQ4_Z36Bd5o<6xK*`l<v=^RX2zHN?Pt$)3QjvkXG<k;}<Z7F2<$2
z!d@2jYW(u3+<YM0dMX3vQw7jSRlq`Grn#iJqk#TPseu`DEqPW#O&LsHN^?7MR=%aO
zAgQ={F-MN-(?uan-bVf6sFf1G{fhKe`2oZb1kE2Or3Vt#J&-rcJMxwi?^112O{U(J
z-j&xC*KiB>jbu7&w<GTW5Ak<Io0@ArkbV=oGSvY~zm1gulR9ReN%MfSsR1~b769kd
zBH%(=V&6*^H<v;M@CZtdre(lmX$A0jS_M3j&eO_}I+?0&c~<HlPwIQ8lGCZ09^X-r
zeqvuj?l|T42}5U4b~ZVeoQD`xQf)y(zY6K-U96}j@6%WpsLca;uaT<z!?KG4zfY!W
zOYWaY)&5Dh+OH>15so2h{}H10RR44`O_)hfEdfg?{d)h2<mvPzX3wWs(X50yi|Ld*
zy_?4<XR@;X$+Vv8ccfpC*;mBrD@a{NuksP8e>T*SsvuQPEB&+S0_s-N8NeG^i%_QW
zWSPzkrC&%FY4-lPP;qPhbICHD7s}VuCmu*tvz5(rW^aW?iIe0Rj8ja@>DjdA&g|yG
zV)87FjQngmm!3<HrbTyl7m7><6>VPFTS?BPXTB*7@)!431fEXr19Q|ulQy4BPd+ax
zlKo_Q0<)k-#+<j8(o-4bxz$B!OU0VLpPuZWAB{PUlJAYMMW;iUnseuOjWC~F!kS1e
zo_7~^wH+~6dLEb?&46(ZttG9yiyxQz^@D4|N-85P|7(WB$6#4uvcU4Ati=G!=SJ(x
zR=2dPqm>db{2y36GQ{GC`|l7<J~zZ6(jPgvURQ!hJF?n){dyGnr%u)1in?{k?pBo4
zlc?Sfn|{~z>dA&zhm!R|GQR53dVM?UvU-z6as0G1wA<>?VQ!mjDJY)h<>lU6j~%lc
zXXC}0(O6H$RwK>){~B6|U1)yu1}r#8W7v1>QxK%R=dQJ}R#u0Kwl_S-jq4b3m>F3C
z^$jPc?X7wB*h_lvbXat!jcrmx=EVaepzdX3p{)}F6-oU<%a7I^-`}RK3M;cg<5PpW
zkCm(>U6``<3muO=?R$4%y4LSv41qo<TCeVVVGDV57#JjW(DjohDUsoi(xek+@WeF$
zP@a@ZOA`1Ca4QE>NG8L*a&WZ`nSBJU-&6Q3Rg6w$uo*`BCey2=Qn_*2cY-z7xso6k
zljf>Py|ZJ*wrtz&u$|cU@?}^=jvrq^BJ0t69?f7lcbhP{h!{qedYySquYJdJ1tt$K
z;S;?Vmg~1R+BlOAQHeN(8`o~$tan&DV(nxbhytd^{cskMFRb&p-s?S8Z!dexfMLDq
z#9my#i<8a{ZQLcZ-+sFNrTUfnGj*8u^-C8ow(BsFsNcBV#L!QTi*a1OUyl=l)FmU3
zWp)Mlc}S-s6#08H&v%^6`_Xs_2H%LXxCS<fc?Cp{GOxMSi9k4fmhv&s2)NN+PwWt0
z63ilY17q^SkU(QTJxq>AT%f+8cX#4QBXZf`V&z<~lWg!JsFCbS8*^LJZaa3?Xd)P(
z)F-q|8X<G%#DupRIra)|hFlF%!d|6yXfE8bhLRPF24ulB)I;qWALY5Y3mPY*E+Z0a
z7ItXM!8+ELi4iCPmVA-MKOu&uQZWuU+kw-<^4dL*7u(@zql~O0)W(?u*Yvnbs8hi+
z>?Kk4dB^X1*BFZ!SHXo`4V(_oL434e#e;^>dhW3&%=0dbIyfg@n&dlP#yam1$7jBs
zI1B{{&9D2BlRW#eLE^bf>?G}EmB4GXl-lr^V!Jyqp{>~tyl6PlFy&ogj@xY)c6=Ce
zbL{*w^7y!Rvwh<YcASzXTwHk8B*z$qG+2&6ovJE<v*p>s{ibztbCis&vb_#3sLL`u
zvNP1FNMM@4DFWnAvI2n;fztqdYSNvt$y2$I35GYR37O5}6s+<FkU2gXLO*i|tPyAu
za0!eEh#N#>>z!vmHpUX$pe96FRuwoFoC$Ohk+IKFsi5qE8n7qSfHgglu?3B@*QrB3
z9C&WJPt>+;EgUBHEP@cTYt(#kSUcGDm?M_P903ZRoLkVWRT+DU8B<L-hXO4)Sh$)B
z4mc+SJ=aK(<Fpw9C(gFVV#@$G3LRt>2kUU{*x`C6E+9h+w%zodIJRwe70YG9)p-M{
z=kUZgpa{$p8k{Zz?iT$Gq%=HKG9~MZF4NzXYx0z0AaBUCc;@9ZC{twfk)F$1=x8x)
zD@H&L?SH80iZUxJkCaFHquD%-r{O<^){3HF1O@G~xAp_=kx?`;t|``l5#jimvZgFz
zMM_?OKWiuY{I{Z`20gVOYCo7N*W?QRHI#_aiu!+y+h!^Kzm%dOYmbzF*B)s#+K+)A
z+LuktLF=X24)@XCS1LbhRwXytT6<srm(uZ?+#7~q;I)mETay&u@JDIl7ozr<gi#_T
zGQuthx}<U{Q{cr_$%@~hn2xq8bb+u$*k{lh(p)fb9_fsWi8~-pN%L}J7=bwK+XkpD
z;H-rXJ|_2;Udkc?w9JD7ph==(_#>XCnP`sq*8vhZ8d$1WCgMXn;yg-9F`;~-Zfe+o
z0{1D4$H<5*852>XLXt;MM3goQI||&288{mz;yM&R%BLD4NQKncBa?KHCzJd3S<Emf
zK_n?_A?=m2Hv4dF;C>l5B}R^Wqb#B^LbwRI$Bv+sQIJ9C2O(d%+G}@&=wl&OoFYJC
zi$ws9QYNVyNOO(k6T3)tPZC%r@C<=x0k{Spi`ga06E9hXk>os#eF{002>TShqgV$K
zYDt*~yUofKu!pAfY7-|i3~S6~=YAQz27dl^VStXob^J27fAtnZAlDmt_7iB$K2G2#
z2@tC@I(`xO;N^ihXnMX6kw5{S3~W<M2!|JtqFu#g=^2tzm?ZQcm#+Mi^p5f7l*(uX
z65j`YL`(FG{{UbZH*sZ8^)xTHt}G$k;AyVv>8^$+=jC05$=6FFT9S8-`z6Raikbk?
zp1D`tm)xVGuH=^Q7j|T;L=G@&%Y#}tsQ2MET9rZmm^<Un-UnM*)j<i|GSOnnU2qrC
zbJ|^UkKkPsQJ5L9-|<lt=El`m`79J9h0kml<*+S~gZYu!MO3=wZSz9x+7#RKT=q1!
z$;&>X+vKU(B9x~y59%~L5Ro;|j-D!>_-z0R(QL}8-j>q*@5|o=g+n}G-Kk`&6rHs-
z;z2u?YJhr*sH~lL;o!RpTyx1blENP1r>>fm_R8S4%4T(+f*RT=KSlW#f*SiY=5$g+
zt7!^k2sbUa`p;Mde*gB#hh@BY-V4dtVBAC$jTAP_gvJcJg}RKo$bS%4jzEqG{jsl7
z?G6FL1G`U6DkJ=njT*By)oc>bs9I}Ahz}aB=XEG$O|Q3EoD7hijrxEZe}fvAFqUf%
zR>o#g)G#vE{R*{t-1HiSJToR1#4c!mqz-W#BE=*{YNk0*dOG-1FHBK%@`H*t673*`
z$BTB_1wBHvKZOKjmqa3c7PP&HU_)O?AR%|;J!N0&t3r|@oI)wknz|}Nd4iiE<-dvk
zYz6>IAK2%m%&=c~$R9#I$$LhVSW`{g7Z(?H;5p$P+@=`Ak_rB%jq(KqoDm8p+zA(Y
zh`>+M=7|Vv*fQzW-szVzTBptah)%0b+U)RGM1oX>AmeW%F(6}**d~ZA)qYogT>?qx
zh@>I<^?n}e0$LAo9z^{_WUij#CM;FI24R3Q16L2yW)5gG5`8a^I(c8A@=Yx<HVe3*
zkm&+|N}w1>P0&S!nd*qpATpJjcnWw7u!g#;t<w#aw%V($)i>&Q>&^PpSL$ncAp<|R
z9C$&*wrz;sz@aOuV$&9p2l8#$FJV7i@oy7mfjmUup=*c%c%kJbFzyha=haDig)k6A
z$dx4Gy6fF(!`0Beb{KVHtH2-?ghwGX6raYJNxRjJy76EEJSUt3_RBN}ygl5S6~&SB
z+?tcXD3rIjvKF&%p*|yNp|RwqAPM%XR7{IzhbWzN_%%9g;+WH9`ryFnPa4V8U>VF#
z9Q<f+2Tq#o4?5}B@WeC|83hTVq;Z~Nq;%e!Fbxt77c96)BEgPx%Oj1m#yLyuY>JpK
zL>cZs(Ha7OH%F&<lX9SnK~HLpo}yiDSL+v`5c2n>#DMZ27!im+7lvbr(uuJe#wz<2
ztS&R42pMol2}=_io%lxxC@_?$>K~{-5He9wMFf8nj}R0sEcZ(WNEv)h+nJSqQMnBw
z8#5g)5S8U;*zchSoLLfj<2s>hxp_HrWP{F?H98iu4i}}~#fd}6xKGHx$ta2Nf6=XB
zvbWHQeMd|pn(?YTyuzeQ5YL~mgM>8xS?Z7y_&ETKK*Gez-$THl9dC0vAz2){6V&xb
z)WD!Zx(&r;Q76tQZEQdY0aKh<?HjZqvcM{!vMHrRc1{7tIvMF7%xl8xnScC7KV&hF
z#Of5rBe6;fFNI*GOS^`$%5g(yEfT#7h3ko{u2Gn9XdcU}hw*=cx^$u61_IH+CEhR%
zt=$@)$&}J-Of2=s_j-5<i)<3T41WbA$gUQe;ct*2;Rv~ps~8~!LFS+%t)EMC8N)yT
zLbD5oAz_$DS|fH?f&g`MSxSL0L=S0CO|^a5)fXf;f0Ql)jLWcz@h;#^wfO_JCQ5Z;
z%*(jW!9@l!ER6^|YFC{Yr9!<ob&RCA)m;TWrcgqfYI-#jFhI^g%KRFo#>_?m9o#M6
z&h%5&^~io0-99#yhX|}o?9b8aA)P10uO<RD8pW6J`9MZ!bl;w%&j4_znZ-_Gn72gd
zjKmoG4FIc_4Q@}!UPfbDWNg3{7M2=WW@a`7tg}|5kZI37YF8SRXA`NY=5kN4tffJX
z-WFoi%0ccs$zkd->FWRyXo|OpQYZ+eE2Kk9@)3Dn_{oM`2f>rsY<{3=y}1c08uvax
zEF@ovU<9#z1_{Vl7_uN2c~kzPMr0ziCuv7GU#@%vh9bdr4N_EvAF1Hb$d44Rn48<p
zLrNBKaQZ6yB*tTZO5p1NR{pMo0I~;#XH`2<45Q$VXCqJ@bP}r?BK$E~P-w2VviKrl
z*N+&me$S?RZHUi#42R>|a1gTl8A|IJVh7TN{RPeWmsBx0OvdFkfoI`DarrL$TWa!G
z1P)3S_SYzTn-EMVYv4wBp0t&aBZp8rKI8+ko8n<35ki6wsFV6KtTCW)Pu(Z<l5IB5
zk(aB*7t^dXx2bRDvv&A80V+wfrBY*DniMw8%_5Kwix1x;XoEbCP2|a5ho2+F6)0Zu
z8Jx}gSPT0+0g}nw>^KY&`~;Dm%y|Aosv_ab&Js9JfJ7?KQFJ@txffk0c(b8r1pRkZ
zql+;R8Oq-FC44N<fkX$mbO^D+qk#`Y5k4^<qNObU24T8}p>z>c(ShxG*uV$~$-R>k
z<O>G-P~QWDOcIb}gdk)>KlvBz(@2PQAmn?2(t7~BAj}#BGw&T<WA%>XxADyZV)<Za
zuk&qM9_dLq%||AdbEuh^2iFzfMa#zQVXv6_a~080MvLScK5X#W?;x3ZG-P&VO7keC
zgt++=O8prD65xU++22#@9|*idU`)2c>>`cGz6^k`u14OpVk*i9imLoj(Uku|5D9@S
z<$a}4`A?&+>%ehUM{r74;d$U~m{UbDLgND0!DvuPjFv7|&3+y!azF9CPm|CS3qOke
z7Bvu&504rc;)9jtFckilbqe2PEW;{5I5=J&*$Kk}T!Z*oSczDMk(noAD%RX&w^8_8
zM1HLclXV{&GQuY=PCK(hNVH`U<PxRGy<`;t>-1m<d^|H=)UxY#g3-q#j64g=poU_V
zBnz@YG|{Sc{@@~Tw1ZEsXuyeF=2Mkuh<Ldf1wj-F_tB~#MvO185Iv<Y#h4Ii;tLS2
zzbeor!W!Z<CJtI0IUO|ax-ukQA(9tgwoFN6T^mpX^j=g-`Z2s`@z8%tsU4Btnf3n{
DLWpA4

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_psample3.cpython-310.pyc b/model/__pycache__/cldm_lr_psample3.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..477d7336a160c4e93c9d33caa299b46fe834de07
GIT binary patch
literal 10342
zcma)CX^bS-U9YRFtE<oH>ACjoU3-sS+cWVxi4(k8$MNo(B<peTu5sLnRjTQ#>FJp|
zdS6xV(W-VJv*1N&90GxWLdZ-&fWsjn0TL4U6G$NO0g&Ls;}Z%3tpo}m5D-JW{Qj@{
zn(-!3t@`)7kN^AMufk%{l<=&4`33j#bCUFJ%1l2RG9SkqDY7Ivk|TSPFE?dhX`%pm
z#Z#Lq{#8%&GtJDfOmFI<F5Aow{|)h<^UbCy+Gt+hFEk5&v03y>&5~bkmPL8StN1g`
z8Ijh#sz2MD^=r+VKi8b|=bQ8XLUTdXWxYjzsktQ5hIhz6+&t_bX&&*9Hjj$3oOjGW
z-aIbSrdRh*G*2L%cM9G~|B>b+vh;%F6rIvt$tk%D+iLSs<jYP4`HE|9o<hlKl*~9)
zlvLd%<j*+DHK{RspRm(Vc=koR+wBBxu09vsS#8KX{}OXt<~o-;t+=7_Y%2_IT7koL
zuM@a7<N3bJLdy@Gp6Bvv-0|HgcDt7AuepwcfiOh%TGx&{ws)P`L9`w+-(`&qU${PO
zdzqSEb8qzAfLd|mWjFeS7q+&zc`3XVM7H0>(52_%m>PsZT9<oe)s0^a9oNGQ<*PPT
zUh1r`_aZcDsJ!HbZ7aSVTXtl1{5IF!+np$m_(CU$Vw+})U2C0%zJ(pFcW9kQE_Lj7
z7}(wmb}x!Lc5n$jSZA$Av!cF;=6yF9ZuUMXN1{^fu7wSVWwI9Gar#k_`8eKa9>A5F
zvLiJWC*!D&c2@?C$-utOXP*y)n1!C_I<Ex7eXU+;>F6_hibxoEKaDpkVe+;#$i~vH
zyeEC>DOrk@L`kGg4Cp93>cB{(T`e*8<UuZxcQdHbkv8$4y(GO}9ORRHVkWsn-BwYW
z-8B+(PnM+{N6{~r*6g9?RWO5w-?l1Cf$^jiUATo-^m`3&qB&AeW}ilTZP(nBo|Xn?
zoW~r6Z5gvX_IYJcz)YE_DOyTwJ(f0$Xq~0jZH17qqYTWTJYAlSwJ=pbz1AC_NlffV
z9_?qhL@U20y`5RTg44(|x9i%m6$aiNHV0IpOm#Ybqs%jL$XXjb8-!My*$yLQ@)9W5
z^{m#09R#iyag!)GrCzSZce*amIO4C+3A(-5S|h@YxDGCH!371s<2at17IU>5c6rVX
zdZ2KS7cYjrc&d}yvSY5ItWo4z<a+B|zRhzZmWc#!#qRAmO@JDBg|=Z~4kA*b53X<B
zvROOg%0`3@(L%XOK;Gc`kO}3^+Y<XQHuvt;m5tDMSAsBh*TQgX#j#_1rOm=#cja0y
zxC-uUZMeZo3n*A2-uc8QUtSUTSZR5VZ+Xn>rrh_`a`z6;f^{*uvqS`V8bFgxxq`PK
z-U8A`zH8(Pa#bnFi}Eq0Bo`D@?jL>#Z7&bUYG|vgI5&+OmKE5*yv0px$Y4kpEbB(k
z_R<>m5zHwzHU*xhvIT%eyb+0YLn_Il<bAA2we&=v1sJhnPR=PgHK*k=`03(DDf(q|
z27ao5$UEgXrFGfKIp*CQ7#BkKuGG|0aofzi2`*{sVCD)phZ}}~Vk9Kq)&wQqSCGC>
zi?~mucb#1XymJ2m9#OyAI)Z}Ha~6pbUXmYCAH8B}TQED=ZkCmkIbyx2{3+<JZTX9G
znrDq<Hd;cNLS@8Wp5m*6BM=W-GPf-!N46B!h%=NP%q3+fgZxTT9F&uyNI#R5@UKSC
zU<_T1OL>L8AnMiVg>kugPqy?#2FxZ3ppmG6x!6oHaei9?{TC7gGiKUyO+rl(OkPMb
z+j3gIr7}M*Iax7BhU$|=Axz#t{o=Tl620*$>5K9`h#?4?Z^eas64l+4H;dcymJ;ny
zZCs8kZ%S{<>xyGIxw}R@leXKIw}FS~8=_6sG4DxVgRV?;z{1zD5@1rt%(F=ru$CBr
zb4d<xKFI?vBn9?Pytug(D1e7hayTgh9!W}oN0Tz(v1FcBhSc#yb&54<a5S#(o`_E-
zYI1a2LHdzB3Av+`+anA;in3Gj>G%x9pps|{68e=$NAF-oZF!HzI!kTt$-9k2-5Zsi
z6Zm~BQQPw1SfUP&JLN$=ew=U&QTulgwI>EA;~BzCa$*TsLg`Nq9*Lhwj$`(0f)&*y
z%$ZLr&dg30qdc0H4UQ-EM87HhjLd#qoW6q81@tN&k_M*&9jOvh#iTSil`NocHJJsx
znzjgJDvy`x%uxE-WRYecoDLMHIyfCK(|MtMJ$dAwL^WIKJdf_K&?r$He*oj;lVWlz
zsXDVenIIoONh2d)OXiZ($>AjL)OMiAWKhxO+1-`+bn@t1(lCE+cSYdo_#QAvEi`HK
zvE=yEk|J4;CC4xeYGllLV<|b2Ql3*@l(tl?>D$Ti!I|-xGbnjyj4e7H!ql8Izhea1
z_&nA`YVnM-u%m5@xso%$+;|3zb6_n=-C6voG^p=i6IN0hWBEHN4j+PLg~<ZT57HJx
zET0~)FJ0Zzj*eDJwD5mm@z4m1@9)1&H2KsBhe&^L|9Tw>BJI#>|K%%T=$$xGe=Y3Q
zA-mgQT#v(gCun&+$F0X3ZXHV24aoSahwJq_VUN{YER3Qj?2+A8hYoW(WJ^KuEH5wj
zUwi17y(k?oN{z;PJh2*S=Kt5wLhM5Gn^$4MK^nupW6wd5_Mf`k!CF}zD%#p`Z6~T@
z#8GB!1=Kg}h_<)p)+0CWf1%65n;mSE8ZtK;8Ub}T9Sd#k0H}!TXWL%5W_#Wp+N!WJ
z8#F#OsC!t+O5B4fTR+=%*%O|76Q*nZ7RC_hgQ9ioo*T50M~9(7V);EUZjllhJ(MP$
zFpD>;0)X<QL|T%-Uw~T~m_jlc?xp>!ZOH6HX#KXrYg92lmEmSs1C^;>C6&sJ3!d$-
zIrhaExrj7ZRqCIbD7IzG>I9wGvX(EvBC@^cA`)qj{?lj%!@1Xj!9~O{w$$s)ZMmJB
zt|Ksca0wslKeJrFzR|&%bcssDDO|mL?OMIdIw9-CcYr8hdeRSP5qiQpkLvyY<Mqz6
zy9^lATXy6|^;<aU^w1_<Qv2<5o!9FZ>mR7Yw6C8(cdk>1iA4P-?WTr)d{T_#^1ON!
z6QnLFfh@CUfuH+yDngOJE%R*GPQ4$E7hv#>DT`}hlZcl<<S27nTip<Z!)ug}fJVTL
z&U$PG@RDE_u@^BW&y5H);xnV<WW)vP8@RWojx;2f4K7y3al7#b&x0DtuCy>WpK&^o
zy+#wk0Hr<wH)({_of8vYZ)6yG8SE+nS^`_8b!ZOUv4)Zsiw0!DG}Hs_8Xw}Bs0SJ+
zqb?;9UV|Olwy}=&Wnu(MfF(acLmwAIQ>hq-n;qY7V|ksv%k!OJyis<P8k(Ibb>Lbq
zR|$10c!pgPRX=KbJ@+zWA>%5zkgLAk<r#>NHmrEi5L(YY^n`iVVPO~N#0%4W$BS6!
zP2%{}w-bk<0HN7+FSO$)KRir4b)MC+0QM4r&(Kn;qhpG!-q3{RwJyA9IMFcW9bt}J
z9S3%N0Q^s!Uq&7u*RFN0zQT@C(u9i(&zj^IqmTy65I8|qCEwn1E#ZFCI=MMcM!aaP
z!wc%M6p!o_b;=W%A#joa`I9V1pg`ae0A87PXDsqm&ZdIlbJXPX1jH#6H-OB^$q@RP
zOJJQqo4^KvF#&OdXl%Xv<cB9%VjeXSlvM_f1!n?XL}ctoP{jph57mG*r3U!SRK^lC
z&aP31Y%uiPbdRX*4qrG(>}docVppj7;;44G>j_6JO*jG+JUzFdSzelWi78V}Ifnu*
zI9Rxv3Jy3U1U=VCkmIx&0w?wzmqpwFHwqoZ^TTyGR%CNM6&H{pIm>E!b`)6_yG)o8
zt`4~_PvMQOLJ^q9G&o%b+%5VUNNISfq$2BzF4NDHt8zs#kT>KS-g)^^lqs_LK+mKt
zbhH??6(b;r_TSTVMXAZk1LcAKpq8caH2f-PttbjcP|z;@*S@PgF!CnGHN_e*A{;+c
z)|5r8NXhE&rtL(Z|5kL=pr`gd?Yos?RW9LIMTr<Kum8uSZJN^mOUWCu_CWb}?SWRM
z{TS$>{g{b4XuUMs!9Lo%O6mK}s$`~HYwzm+QaD<b`=byHytc7&Ymwp`J(MPXc^+>>
zu39X`GQuthx+HQUQ{cr>$%@~mn2xq8bb+u$*eB5%(p)fb7U`6Wi8~-pN%L}J6oEMC
z+cMPVaMnTxACdb@&!>?9TIPNM&?3<=dWffKCJ`<AMI>T48d$1WCgMXn;yg-1F`;a%
zZfe+o0{1D6$H<5*8B<ZCT%1KuM3gpj+X~!@SvVUe;yM&R$|f2jNV&w=C6jcRCzJce
zDa<e|K_n?{A?+5@HhXYv;C>m`Bu0*VqbQ;=LbwRI#}1>EogqN_LC9CG_B&l6`dCC|
zLW+^tVm*LHA(d3;kmef6C-xZCJx*YSz>@^d18@yI7BONY_Cd<$#**_O_CJE^0%4z`
zcNFU&LM<rsV7HoF0()pmzdCgyqp-$YdhQ=Xuc4oRMHrwHa0O55_OD(?2;@p5%RY+M
z?1u<ELx5PF(eaDG2Sj=*4qC3~K_pOsCj~!EDIpv#AVnLDXjg`$<faMz2c=8@D1E_r
zwW2Z_p@nzfL$pLs^y>hlxQQdXs;jw~b!7?R23K=bS9dhL88_=7Ouk+a(UPoV+$}WW
zmLqB+HFK`Ho8Ob1!=kR>6z}G?<z|5#VAK|ewQx}H!fkAphWR7TtW&!SwrZA#C2-3`
zi;A=0ETZR(v*aAYe^o?bX2E_($5EK8mtN#GC`byQSuo0BTObGXBejdDbj!WNbCF|F
zY|nMrGPcQ!9-`aisaPVEr!yDoG}srBr6>-)WxUZF02HFxloP!pCD~t--vWh0JYe04
zWT_OLwKU>EE0bt|dV;8|m383YI|^KL$ui>HF5;(-8W(np;I`6cd5?k`+BjdKd>cWH
zJsNW+uA<cp1u}%2mRbF0ECSEJJ$s+DJI}fS85@k7h@z3ghMCZqWv`+xrLHQqA}dEA
zM}+>^&r|I$5Fk9TuTqoJsEM#qBNk9iNPxmwTx*4h4;qf^c13Dty~Cn-i0pLKuTkS)
zrp5(~<yga&u^NgRM#{Qhp*9bjUZap_%EW@$1?`X3A#OvYh@?o>G>1x02Y>3h3PmU1
zD`{iV4pMlyXs2D!D?~d<49G5tM0yRhy@+7LK#3tCx8+@BPa3E~k|LZ!DbSj_DnfaJ
zn<3@DiT-RB07@U&=Y`a;U$@C0LOscQ_5{^eT9zj+E-c@*gPXWbF-9d*{7oC>3kWzP
z6im1iF7yb2&(h|J2&>pK>DB(p=hNe)&3>Owt1a5>=*c5NszQ)4oxzZdL1LRAwnY1;
z{IUd+&JamM^y`Bx)CII2;yj4@iO5_%!A)4A{se>p$_!jRNShg;&4~5gEb8Pvg~~Uz
z*x1bBf<mSX04jlEAT>c36=tF%LW9UuV&cu=HNYC`j<!xWRN894x>nz)->SFjPh708
z;SU-3xkcagLw3i4==E*7qRO`{5qTirhW!Ti!xiraVHU_k1RlDE7=Y*6ZVclN@p)dJ
zrdJ39L4;h1L#{jS%??}*-R%TnH{v-4sUSQGp`rK;#*91dUf7F<3*Z^y9I)S{IpFQ#
z)-0c@_{^Fe!zh%uxUv?pUqyXN(n4d&EkP3Ox2TvF%??od@1X259X4^yNiuzK;PfMn
zWNNSsYEuV4-rK&DCi{a<`X#&(u{aq8F`}eVmSUuI-kUHDVhtB8xJV+wj&sW*jkCr%
zOYBsFm@h;b?my8Q0)Hn%r+AHWpo(EnYK@+vU1moc<e(6;ccs{X@*f%zh(G5>V~NtK
zu^PrI`xaK08c>7`IHZK7DUDA2BLox}N>qJmU;Vz2iHa&B_-lBDplD;cuNxp`@HK6x
zCVg4C0V10)9WD@+<qxnwLJv5z<d{$Dgs$c0h18MtyBF8!SjakDl=|nU4k6_}A^)bM
zB*Oniw}#1HM<@0NViM7emz~iSCS8KK-jp3Ar196O!&?M?1pp%m<1u?10fSC-hs!a^
z;=q}ru6Jm3g9_<36qiNaD5bQC0U-oTermO^(1s47V+mANQ3_<|<Y26mk^bJiCaj+M
zhi~+K7V}uFRxlokRZ@5f1S?(IHI!9OH*||cuR`Iv;;L&LCLEc^^6EkSAEPck=(nLj
zv~h_yN<(Y6M`uz|`i-fj{@`Bs$$DxLy^Nk55@c5k&G5%akZ^?D$5o6Ff*^Cykv7OA
znUrB50HN6h!;mn{BCQcSEJ1)enKY$97@~)?t0vl>?C1-UlRZoq0mcQ`#Q4wQpK7!F
zYE6{t#F!Uwor8-EVptjxcGQkKHA;zkZ|WFHamqUidQ?zCnrdb>6)-@~P|Ex^r6$aF
z5n0?V-bnRR*>TB!7~eiNl=}#*OYASv>OP$(j!qMS8ja#h_<SHGG`erk&}RU+(@bM0
z5zJenb4Fr}{Vo8nrh{8kvX{}A!q|W-EG#v$%*=EMSZD1<F4dlYNbL&4@^m5<)m-kU
zEUQ6{-biJC26Eqy4^odwUk8XlQ?x~tats6N3hB^-d`O-bezGCgLGWZYo9}5_e{PD3
zCcXC&i-}Q)U<9%JFcOfjFl0e4@}~TGjmSi3PtuNXz8v`w3`K(L8l<QSKT^S=ksm2s
zF(<Q=g_O+U;Ph4Yag4|QjKD7e@a!!c0c4jHUb!1aFbZzE76R3N7y2wf_+z>t*II9<
z@kPR}7cyY|wng{a5TEljTqfGE5whE%w4Ne%C|%fJ(VTxR=1eOXm)8WIg$u>yTkIdG
z$=?vzFICvzqU;SqFrBP{8{v7<Rzi*(K<VU=_sMRGhmA!@f(%e64P;njK;f>sN9ZNn
zY?31{SBuUiX=!Ft-^`}%@O1)Il4wh%#-udPZJL{TARiVVzDLl8c^sR_lf4c<M~Ev>
zyy7u9n_tCRSd#$BWNvnCh6sL)$WCfJ{|Qx*@MWh7oFza`EzeMN+vl0*9NT}jp{E4>
zPgJ9eF`}lC{#)nqu>ej}12}&GvBIN)4?__?F&?6&Ecz+JG}-jJ2&(A7_B?E01cc=N
z@hS3!pnPB7eS}OB5M!l;Oz0>7f)QDZbs*#`j99^jIbqfyn0fo)8p}6ruY+$65X%QU
zyWOwQ@<>m@X+AWyoCD3oJh-m-3$$$14tmAZpR0(5GFl|p@L_|;ejmxyqam{^RhmaA
zCB)62QtHnM{5=7pN%pS<zC++=2~5aVm|dh1+0OystE;g$t(c1Po}wz>Q#9p&5JW;C
zOL<qxmHyMH>pE~;))Ac2Rd^owH_S?2jL<mCbub!K5~HPyRkLp(MeZlQ_h}J&BH>4|
z-=hXZB&<UX4DrFrau5iAi=V(Z8OyK=5Dt!($9BS~0M{U18df5fVPxi|U%1RocN>Sl
zMdX*CovwS|kRd*Cu{)_9LZU5=AX79T4IY<}YMvYpfsbb<i&}P^u0Q^Ggpp^#4QeP>
zNwOdfL=&y@Gy4~TqaA*7MFURdQlF|oLqyB1(D%bYxR1Pq7%{%ULiCir6k|f5i7!C7
z{-Qva2y2Mbm^x^2<ihkZE~H4jNF*=5Y^g|OT^mpX^uMSS^dtDM;iX?isUDKPQ1kv5
DF_?bJ

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_qsample.cpython-310.pyc b/model/__pycache__/cldm_lr_qsample.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..1220882aff793eaa3776c382aeb7dd31ecfddef2
GIT binary patch
literal 10347
zcma)CX^bS-U9YRFtE<oH>ACjowY^8L?U{I;#0lPP;&{D=WIYa^HI6&6N;O?IJv~!L
z@2l!PTGb9@Hh2*ln<EfV2$=~8a5y9+Kp=rXfdmpC00}-kKA|AcN}%uo0WrkO@BgZ=
z8E*pBs(-)x_`m=CDl8UF3C|zC^`?9I8A<vUWu_kunNQ%26j_oS$&o$Dmz%P$G*N)O
z;;Bs)|Ej0?nPz5KrZ;s_mu+T;|AzR_`DW7;Z8R_M7n%jX*ev>`X2~x%%c4BvRs5Oe
zj7aNV)t_z7`n6`wpKH$f^UZmGp}8RHvfiS<)Las2!#m<1Z65WHHIMnno5w|2&O70s
zY@QTp)2sWZnx~M?I|c6{|KaAtvh=*<6rIu?$tk%DJ8JV0<jYP4`HE|9okqzSl*~9)
zlvLd%<j*?Fn$(!RN7!j7Jo|#(?RJ7TSDy)PuQp_!e~~#ZbDhhbR@~5dwiO09tia*A
z*9lyk@qFK9q2-58&vSV-?)YvLyIsrm*ImcKKp3LB*0tk~?OkJb5N(9acUdFD7p@K4
zK1WU0-0M9zpjO;?$&Ehgg{^IFUJh>tk?nUebm^HmrUqe<*5zJab>kO8$MrBn`KnEo
zmpdC9y$DSjDld6q+lp_+mK|9gzs+^`Rws%hzR(Gx*ru6c*V<sAZ(&Cp9a`tH%N@HN
z2DbOS-HW1*9b85a)>-e-tf()ddEX6&o4p6hk*E~AYheRonXE;4oPJbfK7luy2XLjP
z>_|<;$vCQ`-H}0KGO(}n*=NHbW})Y~&db4YU#nMII{J*BA`%APPvebBn7l0wvaz%$
z?@MnzDNC`ED2cR%0Uc#m9T<tUrzOU|Jjf;TUIsNf(kA}1m!;Q=gM5-t%p{knJ1R=E
zdq!gJ%d&L+IQr$%ntjx~0%p+g+fijHFrJX2i#O4Vey`$9G)L;m?9*tk?V0=1Q_{eU
z^O&QsBV(3FKd%f5m?;xAMN5fo#L`v~t+Uj+tq>A+m4O+Qr_0l^7N+W_*LwXkiHZHl
zqy6laXyt3t+nLoXIE_4WyRIEuVc^|nb3hf!RHx%N$~+T?thLFrL1?v^?Jz<nFM)Dh
z&uVSjLEw53H;IB%>g8H|yX*3dBmN4VpxcYBbt24&>);X>Tu|^kj^nv$F;}}`m*?D|
z2MPyy@nYDEr#h)EJLW3N8bz)}uD8METRb;nnMm+f?B0sg1gL>mXd4#hAR;CD;QIDW
zo3$gZY)04+EtIPS<V~&*nNaS$EwK+{bML;ivKjjBN)X2GdKhl6ICgBWv{~5euB`Qf
ztKiPorW>rZfP$50pIv)NfX98i6RZfdth7AGw>)NDPnmGJdz)v$z*vH_L}YjdK$A_m
zg0~>v0@BC6W8?~QRVm1e@(HCR7Zg+OAAJCYFAv9RXsfF@J&hZd71+SP#Z7C-W=Iz-
z>w3@j(i-*=%qg}v1)id^1%O4o5eaugD#@bceGEyp^hBQp7%^l{&M7%Hr{yyE>EcHz
z`ekzleyV`TJLNZ|4cW;#=A9ha7ozu$)YMX;+swQHUTNxJ=?XW88-|!-Br4w41U=qU
zkiJKYxJLwdja>q~a_>GKQNP+chJw*^4v7+8k|I$by<%xcusoP=mX(t^V!){UDQK`A
z`HOOzXN_bwT0)sZWyEBj;;VyW5E5E4w<9OVwiVWhGn5|8C1odr{7O<Bl#`-JKb@5D
zuSQQ}3|)*%d4)YM>ecA^ak+U{w)8{>%q9w;k*I*V*i15Uen$Zp6cPh7X4-O1LQN5@
zUPv-Ka$3HvGCwXkSusb3>XStwQeH>>;<%L(z5Xfbi}GCvBZ!-C#)Z2Q)!mi1iaYYQ
z675oLT#hSmNN>m+ieosrJ4QT{w%d_+fQRVoqD|E??@C{V#!Pg;!q>18U{c4-vq=`P
zmKcC@Ne*y6$pbDV1@=z7xV020fJabrG${feOG<#plQQ6mWS&-r)X7A3iZy9)Jg)DZ
ziXTeU<oJ$)^uzlSa>psRPZ)XxWvAma@mUB%CD9fn^ed63-o=XA@;;4qj@sOn_Zo@1
zKPo#f@cU?@w&lTzL>-)T%7c3R7~vQK_wOKZPYoW5X9zRNsU=_ur9VD+IDR}iiP^IW
zR#cNPXFjPoGrL)g@<>`XIGNNF{f6{YGW#)c`U+AP(W`hw8k`Pvq)JE?lhWXHvVgkP
zWESvh+9HstJYJ?VL+NLeMVftZCQzK};7q(s=Y{f(<l(y#)oiEpJhHb!qeOB10gRJR
ziplAu>dfwDf_(f0jf{LPnM=+jN0Yo$+l4xlK}B2V_EzFE$s=z{!~FTZ6@jOd`@kHv
z(4?(Llao(Liex>SoWLxokum4>rQ}pfc}{sz+E%fqZzU%OXUAjCpyZt~w&-*SQ*+Mz
zt`TJ83s@5=$Ft7DuC^oQO3ngv;~6l{p|vD+XYr%bpnh;oSV?J&<!`4rd;pdeCJQXz
zPg@MJd}h49bahL+I$9~w!vBHABO@%nzyA)=<dY*DBK^UG>vbfEv?Htim#&1Ncj{FA
z)v#BG>~4o~Jr3)gpyl-(w;pf0b*Nf5AQP+}Zq#pwJyvhAFp3_xN7h>%n#}2tMFsV<
zyu93h^?_sdqIA3{H682m#CD{a|6fB3u?zigUWFY8X$&imJp)17fAYBw*2?No)YhhJ
zJ5e1Yjxu9gpuTBGw7qq=9=UP<3tbl8=wO@Fkh#&&6sWuDSZHepKt)_X*Y?77+w*SI
zR)wwEpz*0e-NQ;&;vNjz`nj&l9{1cEFktI9F@``N^sHO=+@Os-It<Mc%kOz{i<HRd
zp)~1)S-epd0F);s(vk%J0^H8PAd<;^FCAQMLuMaB>$eqNql)pV3^&6Xs7&=LsZ?%U
z^lX3Ku`k8QMWnf^QvdWsu`OFxC+NhMwR{nFk?lp7kVt#<pF%U3&%GARE+U4py<TT-
z%kA879f8TiOZZs->E-&h%?{3_OH?9G;p%g1YxOScgsc<a2BLuJNk5!L=m{G=s`vYk
z)jP}XGGI_|*^wL7Z{noWLz{F-t+>y1UaMcKf1nNnzkcET`A!{%67`$3n;QDDNimMg
z^XgGdkh-J<vdlgX{M@5c5sLh6nP<Cp>H}%K0JCpQSzH5~M7#tdN15B&?uH;7UZZ>j
zGy-mRHexG)uLQG*y?`-!ZbYCFpBW`5BQ8+iz`Z$jq#-$OaI`Xx+l@DQ9@I$IrG>fq
zjMItib(#ofDD?@rNh73AotW@iBg4qoU{?vy64)xOLv!GiHI%eiG$0G6p&n}2_z=%T
zJ<vEAbt#eX8m!T_jdg4+6C+RpEcr<q`lJ||O2s(b?D%#Y%j@)Ap6>+Xjk4p^(CkF1
z8`pBVN~lx8Gwiac`Vrgfxz8~cGOmIPx$4_po`Lvi!<GjPq4nGYPnc&N7Itw?yfDpo
zyohz)AdXKxJaHHb5SrcaLOXuq!^6ar7g!w&U@sE*3@xQPI;P0#b<^Iw)`c$(R~iPr
zBaCvZ<G`8^fd7dL%*gNK+FIx8%j^UtO*p#nuSt$E3T&_pfm2jf^6hQc5>7a+lbhpY
z#EaGje4!po@yJe7r#yif0uK=&uae~m6bL*Fz$?>EjYa;-xl}NGj+%U)fH;NXCXhKf
z8A3mE32YE(6WAm$CLnGQjcs(F`0xZv%%diPvdX}*;7p*4h>ZO(s<@!+p&GEJ)BvBE
z%2<NN*&21o21EZ%_lVkV^M%92p2iU(c7>WRj%tUyo^Zs{gd;$~({l@&<)w+Qm@?Ir
zizv{7gN4JX;D9qi&~uFhIZm4)d}7~rS;P%+qtHP-KU{}nMK;${aRC{Uv#ge9N0DW*
z=Ll26)gjmANxacjC<60@2B*t_yG1_(DGe``RAgPzW%`+NRjw!o@`hZ)J1;+iGDSA;
z>zTBLjuxY~Vg%&S{=1s4C^cERuiV$~*RnL8hF=A(6-B`a3fiUr+IO`3M&87@rdR_;
zgyUz*nzD!$DOvs9w4Lbl--?bJ^whqqeWy~a$|d}&C=sLO_5YZ(O;h@RDS1QI?koSU
z-PfwL9|JwKA2l%tt(Rsy+(&y?DSfY5mCST&?Opv}3dgH*e-wy;*EUveEmC}=htk9^
z&*F{9Rg0xqM(_opmqbov3coliS@Am*+0j;oE)bRo`ysT3G#3n<MLOkT;tq&Y(!AUl
z#UT#+whXm7oVC!w$K?LfvuQklmU&PBv`93J9^z@5NyLnP0f`un29_$8i71hdXpfRm
zWGEY}TN*Z?z<o-iGBV;z##HPm7iZBEaiy)?jsmx07S4u==nh4ZvWbQ`QZ6y}$Rr)+
z$>hF%8Z!(_5Kl^5NPC5}%|09(xL?Ma#K>`P6h%x%2p1vu*in?Svjj*#2>Hs@ey1x$
zAB(6=NHG#ytOw92q>}19(p)3?#2%%(#|W$tc!IzM0Iq?@B1UY)K1li8SaKf5{zp(<
zAna2Nk76A}s0C#n>{gRYU=K~{SEo*76xNta&;4WQHT3hZ2m^EiuHZ@C{?%&;fm~^1
z*+<cu{Q!Zd2@tC@I(`xOfJjfpLCf_#hy)7rq~IqhC4|F8q-cW??aGjp+%%zozjWyz
zr7sw-R8&SIwD2Byh?eMyehpw0H*sWFbu~A$p)4WX;A)QQ>W+pt<7OQM%Qp%lT9S2)
zJB23Pazss}X3jPD^81o=RMZum;+@=%+$@j-jN0O`77pqixQ)%yFn`RMb!vCOR?YIT
z1a6sVQE?WWMf99;mYgH_uZk$lEZFb(I0|$1@(a8M1xZ0P3r0C?3*=yaq;?UNZo9X6
zE^;i2@VO3K#x{A;LyVg|6-xy4bml^x1_$D@6vd&pj5m56fI>7|a-w&nB>SuKo1k!r
z2dq1hER|xkmPR~iWfBcgPY|26vJM=4M}ccDSw@`OLlo6f<HBAM+*aBu?^94i8|N#O
zZzHI&Ph-x+RkWI+K!$MBGOPcLMd10jkKZHh&a-Yn#s=diqG+VBVJ0+Y*(<0^sjEt@
z$jTAO5s^Rkb5#5D1PBl8E7YVkY9egZhy_#=5}<Gv*IFUsgNEa}U6GpE=&&drB0C-R
ztJL_HsBr;fIo5DxtcIe7k+SZWsm+6?*C_CrGO-|bLHlEMh&vG}A}LZe&7soM!Jm4r
zLea_hO4?YogA^Vt+G!W`3eipy1F}mZkzNCBFCy45P+~~P9eGdLmj<ekqzI=_3bdxK
zicp^5W=Q#Opg)@hfYJx{c_B6IH*E5UP*3unJx=wNmgR}-3(I%y;0Eqgj8VxHf73?!
z0>aJ+2ovsv3q3;M$7%CKgjH;r^lJa1XVc@P&3=zgt1a5>=*c5NszQ)4oxzZdL1J4V
zwnY1e{E`Hc&JamM^y`Bx)CII2;yj4@iO5_%!F^bwehI<=Wd<%Fq^%6lX2kkl7IpHz
zLgiapY;5IlO(D}Y0F^*7keZ<D3Nz6Wp+RISG4bZ`8ek1|SKFX_Ds8o2U9WG}Z`ND&
z$1l~_@rMlj+@kOLA-ios^!hekROMTih&+&Q!+st6;fi;iFbm`%4i8;J48U`3H->SC
zC_OJv(<_95AVRLhA=e%EMhC8j?skH(8}S^2R1hA8&`^8^W5%6!FYHCb1@Me;4%lzd
z9PsvVca~38d}iH_VHC>STv?CUub@69X`!*?mLLiCn^a7TW``*Kw^8;S9X4^yLuC5k
z!0AUC$<$yO)TR!8yte}<P4)+!^ow{SVsSDGVnj)!EX7FaytiN)#2T(yaGgYg9p{!s
z8fT4jme}b8F<*!>+=8Mt1paP@PH~NLpo(EnYK@+vU1nDs<e(6;ccj>W@*f%zh(G5>
zV~NtKu^PrI`zBVG8c>7`IHZK7DUDA2BLox}N>sgdp#DI}L`4-5{2E>%DB4)=YX(Re
zd`;V_Nk6Sz2a!#f4%djv@(0)-p$D8<a?B@nLf3NhV(Q5H-An6qEMy%nO8xUwhmdlg
zkbl!r65;=%Tf=0pp%eRkF^Oo#%g*Q`lP*DAZ^{l5()erC;Y|X+41f`Y@tD1ZfI%m^
z&E=S6ao|i**E=-2L4|ZDiYudTlv3KnfDi&EKegJIX+uZQu>>lsC<U@}axm7(NPll$
z6IRdsgE#s<i+L<oD;ST&Dk;1Kf|aiB8p<lCd%8uUSD|oSap5%%6OPPddG#><Pf(X0
z^xIG%+PKOarJ=Rkqcf=}{l?T%zkjcLWIeTrUPez239_q&X82<yNH{|7<0?i7L6AA<
zNE>95Ov*42fY9uMVMrKek=BSEmLNc#Oqx<44ADc{Qxk1pcJu|w$sVPP0OKNTV*Kau
zPqo<twI)h+V$6%U&cQ_nF)WP;J8D;*8l^<Nw{(o8IOSahJt`<6O*ONc3K$?~C}n<&
zQWIvoge-0tuc!K{?6_n<jPD>D$~^?uCH9wSb&t*yN2iHEjYjb$d_<5E8r`~Q=wkrf
zY^KqV2<9!(IU_N~eg}Y8)4{DN*~@55VQjz^7M2=WW@b7Bth06_muk;Hpmv2}c{-7b
zYA*Lvmers}ucxv<1G#U<hpES;Zv;f3DcUAVIe`Ing>+~^J|fQxKiQD$Ab2vH&G$5|
zKQ~21liml2#l$E?FoIZq5DCav7_uN2c}xDhMr0ziCuv7GUyghPh9bcY4N_EvAF1Hb
z$d44Rn3LJfLQ3XvaQZ6y1jb{3M&K6#c=o1^0J2L8uiOnI7zH<63xR6C3w;(K{4rgS
zYi+dC_#$D~3mLF}%c9$Dh|hT%E)#9q2-)pYT2B!>lrHSAXwJVDbEXxH%j*Kq!iD1U
zP4*Ae<ZlQZlq&3RQT93^m`>Kfjqp5aD<MY?p>%S{2V^(J!^R>cK?bOk1~RNMpm0y!
zC-jnSHp!8ft3~INv^2A&Z)MYV_(lOLNwlR>V^SLDw#=<OkPnLwUnOY6JdRD|$zF$_
zBg7RbUhx>5&97iBtVw`mGB>+6Lj*rYWG6MA|AeYY__8wu&JiG|mS-rs?eoktj_tqF
z&{Kl`C#uoK7*W$m|E&x7WB{kB0bDqQSmDvYhoK0c7!T1>7X1WanrwPq1XXlkdmc70
z0zz{C<P`ZrP`<D40YWATh_O;aCiIhk!HBHIIuP;|Myz1NoG@z;%)E7YjpZA**TI(u
zh~<Nw-R_rZd88-dG#{B-&Y@;v9$Z)a1zI+0hrMFz&s9W287-1)_`JbmzlUV%(U94d
zD$Qe*65{4hDfQ<B{+<BQB>Pta-zM-g1SVuF%r4T1>}LV+?bX<uR!l{CPf?ZcDw^^?
z2qGbnrM#=;O8;rpbsab^>j+NiDm)MT8)hXhMrfSlIv5QqiP6%<s@d0(BKH$t{j>-@
zk?^C~?@|LI64s#xhWKn{IS7Qm#ZTdrjAd8_2nR>YV>@9~fNKyh4J#4LFf#Mf?_B1l
zyN$!&BJ#`6P1n6|$Pgd8*qzi4A<>pbkSQ9F29HZfHAkz$Cp42qEjv!vAALK*$g$uC
zO(hmdq9Bb!6RGmE2iJg;9e#I3<4xsKm#RQRM9Zzv_rpMVkGzBsF+RgW@RU9mV?v&Z
z4?wv7g20xDYKX&_I%jd>!tyaLrZ~Jr6fZt(sYqm88_)vuzo-=SWB9M(rC&v<9+AFK
G^ZpmN2!S~O

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_qsample.cpython-37.pyc b/model/__pycache__/cldm_lr_qsample.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9b2a428344c2d9a990c3142c01fa9912ba0c7629
GIT binary patch
literal 10144
zcma)C+ix7#d7s<P&MtSC%jKPtsFAGO+A=B2j-87l$F?jxjm)%?Y$wyWli|*Ax#V0{
zXJ#dFF&if)6RQ_0ikkvOleQ#U6g8TMUIMfaZ6AW5f5dqz&?mhWXwiotK>Pd7>`k=W
zW*76#xnI8X-Os#QE*B;IPQ9?_U3*26{*5w&pN7Pj@WhHNNv`C|z7)uo94HnFkXL-w
zQt?)OEy!89ewl9RqAqXc`)@<M3qjE;iZ+@*5tOV_P`1iJ#i|68)}$!U`PE>`ni6T<
zuLaZAbTDJh1hdv`FlWsL^VYnm%liw#qO~Z}hQAaXvyKJFt>eK7>x3vP_{+gb>!e5*
zeKR;^okDuTE%~Q|r>v)B=?%#(yOll3t$6dhs&xkWNw<o8)hlkDMaemoOu03b)VxLH
z&%4SksXqOfFjH4}{w=57ZiP**z7p=N)nz{MRpxrkbFZ};NnPXlMikz)LznA*EA$-3
zC%PVs>>zSGzQ=1xEAZmPYujG1?zt`o!VtAvZ6|3t{%z)j@kYb~kJWR0{&v6Z>(um?
zcc<fp)QTH7z4%o>YHai3wdh_LJ3$*mS6)dHY7m84UE%FDFL^6+Js&en-f*b$T5DsY
z6QfC8<rP0_+R6RIc4E5~G`a5GZ^cQ>=UZW%I5bn@*&8ehZ0u;GMe96%t>rYM(DC1J
zI&s``!fWWkTI(H}74>B_?|NZ>vyVYJ5|v_i4QxOxlQjsBgWr(K=1}5Emh4KF;^th{
z)%Il2mkiA7eEPL0OjzXmp8Iy#-_zRlhTb2gjCTRgSMUgOZA!g-A|1$w()Yh9ONo*y
zskDW`TxDPF8L4!jrN*J$E2Q#44mCQ`#rtQF%3qV-Tk1`u6KOFmr0TBPE2jAaBP|}v
zvUKOlyAt{qvbsamy$eS84e6%zH7V6xX}iF_hdS+`cqqLjVMI5Xz?h|78Kb`Q4W(DY
z$hr70L|cRXb0Te((LPV@n+k1tU+ERY<?-@twE406@wMIg2rDXL4|28#tgb?91=4bB
z*Ma#7=CwU1v7^x6VY5IF%2c-%)F*i^iCANk=flWuGRI{GYIy}z>iKqK(+NY*k9m=(
zHKSCnB|B}8=UnkBwZe8MvDb+RW3GcEJWzZoXt}QMWyM@=M{Qp4!Vahzq{PcnCmHLc
zHk^d3D65yb7JL2%m+$k!fIVWtM~Qbo$r3nyUZri=n1cwC=z;6o_Z-%Yxw08!L$pw?
z5|B5!-e)+u^|8c0gU$WoE32DP;H`#H;;l!~_Nwb7&T5lIo%ZUjPIv>H*x2;K)do<o
zN__I_mv62Le5^KnH?VzX-^n;_rM<)R;8M)(E)vb10gz-(F3MFrCGoGw3yL9^lo{j}
zlx4Z9l;rNQPa))${z!FgZ4HN}al^Jl2WYo>(e85+(k0uz({cQ)hP{Ld#jeJ{i&Qob
zuz)9401R2=pWry60{X=(07H(O{7~AE-GW=(D}d`DE%zi#%jBe$`%uy(O9xR^d9l9%
zNESxi{jny<>M_msn3(x7(a~-8MZoKi{~JG1zt%X80{UIRQ^BL+i9bS*MA{XcDog1!
zTTW+*kK*M-aphh413AmHo9T3X8D$EU5vTZSp!JS}w{z+2uACm<R@lu%r*v;Nopf`^
z|5jS=O{Qg$zMEF?R^z)ELl@&xUSSVKy&69pmKPt$c0QE>jZ^`I6aXxOlXFRFS7AR(
z%Bg`Fb4__hLQMtyT26Dja#p^rGM-Gjc`-+h>eB@w8GecSg<&ft{$-+kAU}Zo*e*Ph
znU$0uNL2ek-m2`%+sY%EaPWbgRFkO>r4Qu|#Wmc*o{`kDmb>yUaDi6WM5}4H_(1v*
z)Lg0qmi`#)0X}uiKa=JGXHx@kE-e7grxSn+X$kZ0C5u~2p$2#iWye#@Z=XmjfXnG5
z;K_83R*00Ds&08k>YYeV9h^>{O4anlu7dQLLkYPPlshC$okiKX<a}}gl1)jqc?tb0
zq+s{4rlx#In5|Qr2lBzwsd_jldq&{-VyZUf-g2t;PP&sl^O1c2XUVgK?`zT{`MsY)
zhMwx3MyZ+92vg~)MPN<Tm3mJlE9ptBB%hv6)fvb|NXn`^wV%g`XR@;1$<$2sZT*qT
z<^)A3j}-QM=v7{ldS}BDQWd1iX{C2Iok!hYr_+Ex$y$U8l_#rojwt<kx<K>y&V`y=
z>zzwh=>$>!ujx|{B&yk##5`vXF3~7B@#iqkL|RVIrZsnZKNrp=muY0=XVTg9TzV{>
zaA)?RnPgDbR{h{oay~utQ0nKOIk+TncJdHdq!yaAbum5plB7uX#dI07phlJ+DR-_x
zE1b&c(Vbk7wpFa^LVB`yemLqBR`$^lU+4R{nsw*)jj)nDk5!SDJMYf#YrA5$^gM7k
zoCD(=SxsuX3!jsE=98<!S}tHMg4ZzPe`R?56ijPu2{_VlD&j2qm_9#TVRn{_`#QCX
zum3-oTpD2Vll^yzMqe1<5$Vr7xnfs>m|I%w-n<?~{;5;u*Q1UJ`P_^WGl|Sr*zh~9
zXC|AT2`%b{WJ#IPhPe}UnAu=a96##}j4%^Q%59Nh1kJLtveNzfr;gc)v+?500yLA6
z2}m>lN<#}#4Aow|0douz8pay?5(I4b`PW-mD>I>k?M=^dV-q6|GD9_QZaOh-Z{0Iv
zFX_JDX3^aiwn+_{7x%Ti>1AV~trG$jiFu*vN9&I7@6c9-$ylfHse$QZC96pXcBy%x
z?XhQl?=I|6^B%?!=z}KpOy3Ke$fHBw;;@5`pEO9Z(2w4vC#LblH2}~ZEU-lh`~|q3
z!&%8>fmfbfX<cS7pfzmu8LAkbNPi=YbWo-@No8{5ithyLu5&d(?mzHI-LoS#w_@9^
zu$9>M$`u$pjvrq|B5U7$3C&>tb{epyhzy1Xx5>PQ*ShPu0*6Nz@Wt*IR?OR*Eu2Q1
z=tJQD#_P9knQhjJSS#59ih$!$Kb%733nM!=yWNXsYsFgu49$iUd$D;BXPh0^s7q#=
z{buVu^Qw8tgoSQie){Q_3G0XYjoOV3eQ{KbgYtbdP6$$$j5Jo*Rbb~aoyowO%(tD)
ziP3low%m}WxCRD^c?IN+GOw}SjzBVehVn7c2)NnWNbC^K4vZf5YZ#Lk2BaDDsX=ly
z;ym>Yy?bK^8j<G(Pbuem?PQZrfDXxIv@y3e<+fsHohE{9Nqs`ANF!uknV9f)J;&ap
z&3u^vErF2_fwjY{sw-KsXh8N%T|LsS{vyxC9Z)x!Y#EWj35#Qo(W$=9Sj4Oun4C=q
z>)Kc$mY@Wf^A(zESxiNxVq9Kq1x^#oZFN09(F%tfXJk8}HqQLAhR0Pxr3&t0uZpU7
z9KYj<`M3&x<Z9rwc@Bc434<Lph1Rp5dg46qvZ#%7<E3#9<Ylb$F7bZmvWde{fZ+Uw
zA34c$FZL79UuGw1CpQSZMN6p-jxDx3?X0&o(}r^l-xyZAD=crj<-%kSLtY&Dw2YiN
zuH9<gc$=M|WD%Y#9A=VejG`1QN5G`2O5kjJw(zQHoxC_q##Y(hfWy;a86Fu~=4^t%
z6oFF&$X#UQFR~JW(*V3W?z!0Hlw8Qf#5>gFT>^9pR(TW19GwiIpM8^nLtvdigTRoW
zEQ4rnqy5~ABP_8dH6co~CV^wYsX!MI9b2PPLFs*kV2>#TYicZG3z}!YP95@L-$~Pb
zqP87t{wVQhfq|I4M$H!nwf$X>IAn3eA)x5-xdqKym61c3G1-_ODA0m~g;%NIfOA6R
zbB#ngPMaYf;_P@VwhVBm&_>ooe;uwJJ6zAi2gG^7wi~_^$F>a~qy=Y3A^e+L$UcuJ
z{!ajDxd^YTN`9938%Sw*sHCdsa`rFEHKi&W$Q$s+it>VdPM*sRYIS*L(2jb`267cu
zQxthwn-ODZD4l~3r^wk~7w@8KU@ZJaFI}mLw5BY`T7C&>c!O$HuE`bngf-;QYhoEa
zYa029HOw=kjQ1Xl-k-zB#?XheQh7;*4><^izy%xXriMb;z>nTU+Jv>3C^nH2837Q4
zKT<iBDFouGWQy-nkU(1#V!UrYus6^iqF8WS9_ftbh?H>-q%^rPhzuNcJAKp^ps$77
zDaqZ%*D?>C7WpK%Hi(buN6bgFh^Wnvkx1YxV4-4>h|}nZW+*AeX7Y);rD6FByr3*f
zA|t|Mj74t>_dlED(HF6ut-`K?c*{BX3Pr?QD2|g)HAHd>sc}GdWIs=~@|_#7PNjo>
z86rAa8|k2wwK{}<0zb+Cszf}6k!Ri4e2hH*dRfFxglrIEi!D)$a|B4d3-QR+ZmTUM
z7YnFN2%*!I3IXb+jJoScbB)9cdz$K=AwWJlyF}nQ0Iq>WVs@GG!~<4gDEN+^2gL_U
zbRKg^NFpo&B18yc0gN)IR6u;)+SnNlVi2=}7T6VZ>O1k*g(W!x*YV4I`nB7LS6r{>
z*=Nz3eU8BA2@s1iA_x(s;N?CyH$30B?a@MErQM@Of~~)R6zwaf%^C%me!tSc^1k#1
zLuJ&P#P0A%bVa|I=xPw;@KjIpavRE`gtv}Ag{b8p>mu|4%GxVIG-+-Qp^t)BJeW9?
z+&NK~ca6Qmu56V=fCNgQUn{~SD6jPM^W^F7fwip3ehK`~J$Qvy)rH50ai`oFcNXuO
z2&qhit`>$Nl^fUILa1{F8jV6*HtcB_1IWQ%$P61Q-S&2PA$DzwyLm2q7Te=xA5ms<
zGHelM)0qeT7Y?JN2HMawi6`y?P{d_RPW3eZ5Y!FffEA~btx}ZK)`$n~T&e-;DWaox
z-i43uD)6W!+eiurh=sUnQaUJu*D70+hZLL8hWRSxn}|&u(wI|84Xvgq8X-KZ+*%V0
z!0+E*eoShZ=e>}u0>+C7Zjok%KCe%+o2bj^h|GRrjtHcPa2$J|YQIf@_Rqcpz$-(%
zk#QNbO{!@Tpr{nr8WDnahU<B4N?B7IEfyzzBxj@UQRDAY;}XVl?f%Nx42tST#-!h(
zHlH@VMqx<7t^=~BkYh&M#1FWS=<{N+0<k|(KBa;V-ds|KLLEf#Nuf?VphpPxi%39b
zNhHi?a1FD7_&`rdARTw*1LaWasX~e(0zxTJn7SsSahcfC*zce}n+AYp2IhGwv&J_Z
za$+F;NzOBx#Hu!IUtAj4f#-yGag$*TN=95Pw#erZ07ht*@FqOo0rFm^Z4>cH(0C>2
zz3%DPGWw>C{x6-@8njvZk>N(HEJ*hu5`EGIX>CCnq}rb-Y3`v6QLpz9v_q+ifQ*5V
z69wZxpW^N*Rlf^{N11`3zqFMDvW!GO$fHg^RH%GQON^}ot_EZZ#8U~>04W8!A}FSM
zQc4j*N{e_3cnmOuy02}}oshQHt*x7z<~_4vK6}+%#|s%)Fy+7tBDP~gtOgEU3QaU@
z5gs5ngwam8;@=^B0%eGULyr*2^Fq^0V0R%(&L_ty6l^cZjw?yTb=SMwf)}8BtuSiG
zR)K;4g|RQR51+!ANvqk3I&psiJSXgS_I;WIjv4N$Cd5JW+`5y%UX!=EvL3VFMSZ4o
zgf@~Jf)Ln8R7{KhByIl?rLWT=6PHNJaUAi$7=ymiaNsjz2R_{0@Px_MpcDQoo|pzB
zBOpO=Gp<lzlTLI?>*-1EK*y=diy(NMS{`YfG)~!IZ$glP_)`dTw9exaCD${;H<5;c
z(C<fW(NDC>?Q6Y)tGoF<DJekBOIt-U7!Wuv3`P>A{gH(7H`dq>u(r%t5{Sq#!qAwe
zCXNx}2rMM3NYak#hcOKm#YEmD6k;+bnppOGRgf^am^Lya{Xn^+-5QJT<El<sxy1ey
zjp1*Scss&=zbIbGe9NGHb)61_tUz7rK0Su|jMs#u8xLd%mlWMV4_`wk_9Ftf#Z00t
zpL7Q|hIAF-`C|r-5We4|4m$)s1gIAUSy}n}aO_+04wn-Wy`eisT7QmqmO+Jd`H1_P
zcASx!keEZU$tT8^{Ps|M!k#u1x=Wc;VEM!Po+DFc?$dYrgjqb4pj9k@Bq-^%6f#uY
zmnmx;<ZUewi3)w{iCd;&0B>Lc%WFrF&zGr72WqS@`yAY#4bo7`&B1|4O1D0?%>TRB
z<I7lNgNTKG1w3Rn3kC3xNRS+ac*n(wkb0nP5Rlf(rMZkpAn~BQ1&@&Q%OkB34=hSZ
z<+79lS%)6dftqTEva8QaZvGgB>y0ZgdGRjbO||(aYKtfpym&=QG+a0kkJ5-Gx4f^8
zjZ&fBTRKKk+{t|f<5f{Y3TbLBlPEw;CRW(*Q)<Lv6l}qj-knShO}ZYL2*XRiy3!&y
zM}Wt4mXNoa2%2aV`$5z%BQCnq&e2Uft`f6IL=0P15I6}M_6Gn~EgRe(6Sj=Tw8)--
zTPln&GP;V{5HP%&^+Kj3$+omg{qk%g71dns3SPA|Xv;f7Zdy4AdM7zbIYqjF7eSwR
zo2cYC29)4(lBz7hDwu;TrN@N(2<K1f&W_Q~sPB+WNSG3-2r_vS2?$dXY;gUy<Zoz1
zAVN`+B7|q<%1aW&Cg6rfN)s-kg5x3=QTSPIZa)v9Siq6#YwTr=$36xaT5wkWo`ayS
z2mNMEwxbxD|E^~vcpS77YcfP2V!WWx*l1>vKf<jaG2r~ZO_#}#nB=A4ur?h8)9zDR
z&(PYJC+u%%^1r2u{!ub6uL~>-Z-~qH*gsH{za#LZ9AQ62*|!M26(l5h>lGlCeDLCv
z2pt{pP*_t;YA7~RWPmuSC&SnR0uR(fLN1wJqZ~P~T6{arN^@KKRz7Qo4+5ZnL|ZB~
zMx{w%tGG1*)We{|2LM_>k3$oAGR5Jv;1CoFbt^uD$M_wrg}qIH<S#F_9fo*%g6K+S
z7ylVmk)UN~37jWDo+-~!#5v%(S6nA}x2|V|{7+P)i!l(L$=>#5+|9Qky#X#CL8)+3
z;FeI7O^k<#C5wNXa9qPs5=Bt(*Ma9b$mpuvJvm0WV60E{9g;9fJdBf(Fri+!4Sk`D
z0IaZ1iSc+rSS|=9-aop!$-9o<!sR|<uwY=X{X<$9sYFQNrLlz^X(p}`*iX=`K6BKi
zrQTdcB$CkrxrXlteD+63W=;y(RheQuMkyg){*qFEMd0rV5EZhY6ZjVbL`<V~59})8
z3inWjPno3HM8bdx)sGCRwRuB_yD(|gs+umBYSjs2f>G=G1+IgUpllc|RIHIbK#KfE
zd^FP_RK&s&Vt+^th!mJh4Gi%W$VwOrhsipH4;@yZ=Mi{~SB9E@P=E^vKZ^+wV=%JY
zMEJv+9q%>_G>aIkbz!{j6GKM$Zp3M2mIaBgELK~h6#0jYcy_ode33F*)Qao2gTWUW
zj64L(psB<nNrYz+Um{5B{F7_IIrcwUqVdLZnO{_*A>x%r6a-Ny+&`;=U@X31LYR`i
zv|>VDiBBuI{+7U&2tkO$7@caCqBF%6Q-;H<MBCzX5d<V5(G5U)FX$!tIAlJ3EP}sM
KU8=n=`TqmLC``uy

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_qsample2.cpython-310.pyc b/model/__pycache__/cldm_lr_qsample2.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..6aac493a4c247b51b13ce95333df04eadfe730bd
GIT binary patch
literal 10342
zcma)CX^bS-U9YRFtE<oHx%cc{@6~I2#$G2j!JCbb-F0x*lgPV{<04k6rmLo>XX@yE
zRlP@3Z6}!x-hjps6d+O*G7}+?0fG3yA%P<C4I#ltKD@6%LMtJ`2L#v<FTek*?w+3U
zCeW?=*SnAZ``@pka@myd`-i`O$Gh^PB>k8&lRpia7x2W2EJ?29%DxoHmK-P+3XoTP
z)l%_ReJ#jYxnY@Z>7p)g<%e%WybFP8nWBy67lV>j3d&YFs92StYE?yf&aVYi)|5!=
zem$7Brh^%4CYZHmgE?z1n78IdUEW^^7Oh2*HvB`uVe4>k#5xijwT_Cif`2SHZXFkC
z({BVPtP@BV-I9MYc+7fCmR^zEvRk<)xfO4IOSMiRUv+E9*F1CMG)m5(WXi3hr0y*u
zf7Vs5OU>yAgq^0s^RGGGZYOMW^~Lb+N>k>=FEZC-o_o2|N}3wax1#Wt9lBiiJE7+=
zUhI1;vV+L&`5vz)oxqC|uWNh3s^_{G2t(Abcb%l;_&1mn#%mD^Jl4$d`5VKwSE%WA
z?`F>nsTDU~_u`lQsI|$>%hByHc7iU3uDqBe)F2A8y22YPUh-PxdOl{TUUR7Oa%XL=
z7o$m2<rP0_+sU28c4E5|w7Krx>BLFQ=R0AXI5bn@*=sBcZ0u;QL+d<px#P5>(D7e!
zdU4!w!prEvI;%aJ74>B_?|b2JvkyQy5|v_iEo?w6leGwslRp)i7x2V$0G?#Yu4E~0
z&Q)FQo(vk3fqk7%zZ``Li+ta6-w22MTDjWN(P#84BVpkA0-m^n$=lK(pGZ6MuJmiq
z$Wo%DN-Aw&Kv&sT2SzIGXsNL)4+^QglS7S;w262Avh;R&P)v)dnHExYOGRmZ$4Jdx
zS(a`dMZZE;vx}NH!3-MyTdFLD#?w-K@itn~|1CVJ=1M)8eF5#Y9dlQDP8yg=5p$Hb
zWX$sTSCv5tGv#7Sw3OIdB5jn>I!~?J3L#-z8JJ;pvOF7WexiPItvA1vn%Iv#+Rsjf
zR(@UjAh&Y0DO*M6bv-AsqtL(0W`Qc4km`1VW|ilXh_%*vK8)-(b6iHq<P}h^=i9Ay
zCk#D5<|a{aM!j51?sh$%bH%IF3A??-UM0eexehMzzy+nC<GQ|=6?3&4b$P)Hd!TTT
z7cWP>WTKPWauTkhtXbw-?D=b4zQYS6mWc&#CElGROMn`9jkaN94kA*b53X<Cc33;+
z%6g0q(L%XOKwjtikO}3^2NL@<HuurB<@G4=mcuCVR-<Th*>w_Uxy_<pclmlRyaw)U
zt$X2e3n*A7-g)U4USAgYSZ?`lVEfFznQ`BRrS4sx2kT;TcaaG241gw^at%*OJSC)$
zd}tI(a$PCO3-U3gB9|0X?jL>xZ7&VSYHBMhI5&+OwjDaayv<E}$Y4m9Z2M-<@v|EC
z9Oe`on*h&J**w4ko|weCAys5i@(EU?TKdIb0T{7jK`tl-wV)Mp`0L_NDf;IN2L7sm
z$h+0|q&3+sxaPeA7#BkKo@8m6xV3WcflDkM%v|H<aKjK#jD*AonxMo73epc~5f6y;
zZm`b+UVZQoKT*HZI)Z}H?;H{pJS0EjK6)k6mSA?U-88GFv&4FF^;PJuE%_UAmS@d$
zI$lJXLS@8Wz7nW|BM=W-I=dyOM>ZAKOmdVS%%)X0hx~F{9#qq^NI#!e@K)pJF@`S2
zrM$vk5%p^P%DCLTFWY)51Lji&&`4FlLSm-5q`0Mk{!6KW8FOuUMnX*)OkPTJTXI&u
zsj?ucxOp*0j_T6|Axz#u{ld7F62J4R^bPqw#1I6{A0(yw64l+8H_BV`rV?*cZBk8Y
z?@8~;Yl>^Qg?mOam9^WFw}6NE2ck{gHSbH`gRV?<z|!}z5@1rt%+qNea3(bXXVU`U
zTv`O2PfP4a$->5Br~n>9$>FpNcqFX=9!;x&$I>}k8B)hn)h*9RgQH1f=R|TcRnwze
z3eu15O2{3h+%9406v|E~XOgoJgG#E+OXyc29lebewdGwJ>m0SYFYh!{b$3*DUf}of
zRBg+HW2rhg?p6nl<O#wtMD0I9)SeieOr{7k>4`;P38lY0cr1A`J&xJ)DONNiVa{S&
zbEmfR80Az}HaMO(QvH_n>oWVgIDG}Fi|AE8Bn?i7I#Ly+%4ubAI-N(|N;(aAEo%|V
zRGuu+nW6M6=>p9@I1?&veQ+jOqVq!eTKd?1iE1{pc~0#t(<pJ0JcV(JX*oTe*4^pt
zTv$w=rje1KNoUhD>EX2K&TK=G$)KW*b34n)ne^2A(lCF1XIbFs_%1L<Ei`H4@$~p}
zk|NoUr^hf0YGllLb1^-UQJz~}kTzAU>Bs5u!P)VcQz-e-7+Z8YgsE9~Zrcd+$px&5
z)Z$rpep}lTbERj2x$z7b=fGOhhP&_?Y0%ieCak0~#`3>rID7<_6($QTKg?PTv3zE{
zzHD`i+d5h)@%;aR#X}=3ezN}-(d08D93uUx{p)olh_pj1{nxKXk$>Vu<E^OIfb4EZ
zNh66Gov`KiT(6O=dkrXAFC^ou5v?`uMm^SOu_%t8bVhbt13Jv@kSztpv$V9-f9sKB
z_Tp^3I5Qd>$)43nGylJa7Gf8g-@FD34$>I*9eV+SwExVN4%W&VP|^0f=eThLBaSj-
zE1<FN#I(ItuMvAm|J^Q&ZgsFtYRJ5JXaqF8Y%H{OLZBjPoNN2hs^j~2X{*A@Y|{AD
zpy6XB%SjKWY~x(lV^8|tEtsy2+ZaQj4~o`n_+HpX9vy}Ti5>L(q(w?(^rJNCglRl+
z9RQRkrP87V{sP?0!4#6oaIfrNZBu5SLhBC{K0_7bQyFfCk-o|FDydX%T=bn_)paf<
z$i<|&>Qevoo?=_F?M~Q9Y<uY<EF#B`FCmfj=s$;MFr0fW7+gdQV@tikyq4Fw<+%cr
z2bb{K{_{(X8|xjMNtdWZoWiv$*RMCatP`<Lau<jKruX{cEFxc6=W(Oof1=S@@|FO@
zM$3u4xN#dNogLa<m&|_q#m?J}OO2-*Fzp)`&Y$lzU?Ne!y>=5rKe1Pg<MRDRoDifg
z8G$UZOTf<qIu)VFKahF8>tx=K#!E2x#+1c1uu04-Aaay>t<7!(!r?QNkAX(O_0C#i
zhwzeM7O`s>lNUw=8uO`9a&N?W>Kl5uCyq2Cmklmf&h@&<Ixm76$*#09w>9N<VrP{m
zf&of>Ld&EPGIvf)c%zwPuhM47)et4@HCl(}!X0ZWS+Qt97EDt;(60Guo{M{+aWd*M
zBC%#*hqfK8V{M5Tff8WJ7is)sVrVKA<8ZSRIBhJi)Ax9>6OK2^$T~u8oH=kUkE?__
z6+FXU7FAz%{GNA(v50XMT*%eH>GB-JM;lf=Xb7!mA9=z&@3N?ibK<2*zT;)A^A>S@
z=G%$GP=L_<njbmI)6WhQ&s<=~X(uZLzDP@{kB%v}d)=(JHPeL`4JR6=yerIcyW_%+
z4?}M5IlqiNKCWHwTzi8ZrKAZL7oIiAF-9Q`mLt%hs!HH&dbV)CX`S30Cu6H@ufYrI
zu?&yw6m=>Rm?Ch30Qr-wK%hk6Bml2Xx-&L;D(5o6@Fq1Ovss*iRbB@&_fCe;&m01)
z1X=`K0%HQQ45G2M?$giiVTr9%6QV4u3LFd01iFaG*q5kOQ1(y_*b{2NnwrSig2vhF
z)FB@ZJvZGaYP)OAA0+lHf)KMS)O=x7JKXgiM=b7f1SoiNZb7qFWzS2@m}<f~6llT0
z!qrr8z&Ro4xkiE<r_B&JaqfC7whVBi&_PylxDMBj9j<5M0y3mv+b!RTW7}qzv0Nrx
zoi~tr22XqqioiUk!Ra#KZqeUBO2b1XHCb15nf|6+murfFydlrvnUhbUOp(opdM<0B
zqs6GL7y&u7|B0q6%8aZ$R37RNXYw?jhJOvM6-B`a3fg6F?L+OMQ8Y2GDb|1y;rN-d
zrYvAZN?!jcYbW~rx1ys4J++@`AJ)osxq^QkC1SLq{vUg7vy}c{O3{$DhswWe54Ae&
z$3PG5YbNHP_0nty`)D62m7g}NlACO;eWd?O>1bW<k3ul;+Q!PQMT&3qqcrggQF~0n
zD3KBwVHX5lQaP0=@Zzdu#cxqeM_UoPKv*K|^JooeE*LnEbjHQR9T2CadATu)KpgaK
z1Jo99)<Op#k^74;XORF}=6(UtBGEAV5l_=hG)MehfCP>PmMWHs_>hh`kCIYMD4(bs
z8aAN7eahl7G9pXHMAWE|<k1rmrH#Ut0=Hrs&W4G&4#ki1sfGwrAvJc$Bpv3-<i2?t
zGYm@*Ny=JCJEg46E*u-UU&eKbk>lPhi)f4xE<*0HLnvhwWDxp6$XBlRJ6$3ASV$Eo
z2$0xf5kRw)NvbB&TqF6!&Qsmv1eORqMc`=wu7Ss5c7gK5OIBelIS*o=LJlRuK1J^+
z)<J|?Qs%&JGjav&p(*|P#EFc;8nfBCUqr8=pMO;tpnKpdewo|9aswfdtIa(7IkaY<
zA@K7Ah}9V#zX*Kr@=zSKJl}^%pa4$>?odhyhv$)^UBzVS8In?%B=nz^uKbhquJL9~
zWi$eb?|na_CHlp`2QZ48xU#2unwMKs77=dnG*|U>SHqL@@-D*UYb6ma$-BnA5@a1k
zO@L_6+$ru#?qN|^a?AG$Te4Lm2N<>GVJ#fgdvF`A$}oS#opxvLfvv3Sumo<IXi;<L
z-39cVau?l0c-KV~W*Y2wbR31bcKJ0v0|iOpGaE)ZYzyRIeq?qLm2P@>c_DUfitTwW
zdlK8^WgpRP@>FaQ%F~$#bsFxA$Qo!zPZdx64giH{Hsn<ANNN6e<@Z715D!>)D%mPU
zXKjsm(9WeApq?TsYv)}!_^txiT(XU%u!H!it0twLGPtd>QQf7WhBnUEDBng<W0%I9
zO6q7eMS%?ArsY=t8H>R0-#+)Cj2F*)AsHKtn~0*3!iJg9oMtyrmr)n_55me3$PuAG
z_I0ZL1_8nY+oC3wG5*L#jai3kHV7av1%7Hphz}aB=XEJ%O|5lUoD7kjjrtxnzE6!y
z7|XSXD`PV#Y8n~qevR5ZYI==Ao*5GhVi&YOR)@F^kz$e}b<-RwJste17itup{J5fx
zMLS60(W0GpL5~paPapx=C6P#<0c|fJ*f3BMNXRXDN7<DIs*t1zr%(#Crml!kp5SIk
z`EB%P(*RKVz&<Z!hW(mD{t)U(-ZPrSs<muiTwK_J=Y+R#n_`SgCit5+%I6VqMkttY
zCtT<e0zXHaCnBt4%cNKPCtuEJoi_VZI<2&5v!h=T2~rh;jK71#kc>fM8z8n+`=R{0
z1d`4XNkjDOgFMs)v>xI-i28}hTqDIzSgL*#!T@Ckt{$X~9MEPY`c58o@~%SV8(Ly)
z6mUTy(**#PKrxV-po<DK)e)gVWGXfB6z~{e4Ru>vqZ=x1rC(odtT%2qT8$?!HCFLL
z27Yci@PdflwIO-~hpwoKEn7q$$hTqN#(ucs-z3Zed5FM6*AN5nLfcDV+#x>CtCRE!
zVIYW*D@nw4*Spn$tD$?HFzUuufk7$=k3wiDK7}!pPP-TN;^6{#PB;hbH)sxcd$=_#
ziX-Q_RVRT_C~tCQHD>RlJ|k(NvE-H@3HF;*Op9g*DE+rlc7+a`IOZgoJ~(jtlSVQ%
zSOzl_2S481zLO^VgHD=!l=!y*$S6n<C5`hGBc=1+fN79uxM0CW5(##kTOMhgHO^UL
zr&GjyA<A(7iPjMK+c`SL>y!gk40}>*^c3xK+uEQ2g^<4|B?grL(1<|%xiA__lunG*
zG*%eVVP-%PGT@LBmL@bh@sAKtU?@?Q?yKJyGEq@Q1b-cm5EN}J_iY2D48ErA%t+r>
zZi2}6m<|_+%F<Kphv)%kmW1A3ozS)1yqGz%LHE)s9Sd293sV35#35wdC*<E`ltlQy
z=+-pZTj<1oM@%A`@v1wz!lX+O&!4b^gf#vZb@(L$zXE^}NSIjpI|vwb;=5c<NEV0g
z1a<vBH87}<ZbNZd)QvMr+cO}9fGJL__6^z)Szr}VSxqUCol}6ZPDc92bDFSv<{rJ#
zPgu-jv0B4;Bvwh`r4X!iY1dR%IBw{y1)^7>a6NI=H4YPw%wu`wApVb0mmc)nP#`+E
z#2cldwcDdJsVV*D#8Q8DuLop3wTNCuzXB3uR}0PX*GQ0Xgxtqfj1Yn#bI_4C$fdcA
zVITmZ*#*OpFw7&Z5j!kGfV#OXr9c>>hqR-n+OF*C^OBoCOcw#hMcBl67x1Rq{JvTf
zr8+U@MO^3LB7+!~MuZ);txk+mq23!hMpE4Bwt^ltl#r&HTFC?qkTaArBn2jzjRHEj
zTfCX+r>g6b{V=|LY$^{BSeMwJqSXUBPl#Vl1Zp&jFX8injL_)5Jx8Ab;7&7(e#9_u
ziOv~`G4@>mtDX&RPsm<IV_IZvz!esj8d+v$HUzA*cC(Oa&mT~`(y%<6NJTZ5`+{XH
z4QlkJ5TjNOa^Fb~QjbYr2Z%sZyh)ToK`31z9a@qP$#cR_Hsl5fp3G+RV@>PNPEgTa
z?|sBV@|6fi5X<M0fP94^3v!V+<gaQ(CPI6Xc7*fg%7<Vm5?s?DMOFBb3J#6@Na2dP
zx$QipWB~`KudvTyJoZNfz6oIEZ#xJedr){*wHw7S3T}Be0@Xn`v8o}$ACm=z)>=D@
zFA{eBhym+&Y`WKm_?*LVIIeXEA-g`M^$f8?>B9bu=KOQ27#=3$@~XhIaG|(-n-OEM
zzaX$*s<6LA**k<_I#~lZ!gHjpgd915(!E38C%Y*gHWncy<AFM9Aj2913U}09LND27
zdpYuQwfKCRmF70|jeOP)Unf8%iMCW~?3E^k4RfOi<ip~__XyfBk7E;gve)6~2yq39
zS9}I%a|>%>UnW2@nVVgQA%dSEvXdFl<SQ`}zU(xCvjj+_@*G9C1D<=)b%HmWdPdNH
zLp8b>1CgQZZC}900$oUSfC~o@D?A$bFcjev;~`qg;%^bA$)?vuP(=r}=U@XPASCyX
zPmnJd>=S+W5i&_YjFk~Gp`ZK<_7W0e9SHf7-eJUsydcaP1T*g(Tx0c?<9G1Q0b==J
zXRrGmS{~_1IL(J9mUEz)m<QJte}b0HnS)+2_2(+0p^O&EHGJ6Mv+pCBc{F5pWlHlf
zrG&WoLrVQIfxjX^G|B#!z~2!d$8BQe!t8pM%H9RQS65?iS}_&nV?|YdqG-zhAc%xO
zmhzEOsQjnV&~@Orsv|h1tMEMVHq2U4jL<yCbub!K5~HPyRkL43iri0p@6#gm#KMna
z-=hXZB+R1*hWKD*DGY_bWu3q`8B4GV5Dt!)#&*J}0M{UX7FHsbVPxisn2I$!*=-#D
z7Li};++^J+hK%rui_^*M5E5-!1i3^haxYm0z&be`0w2%pEo#YiyTSP55k{VcWl%%0
zN|FUxAev~^I=g=nINISSS2W;6F7v5MG(^1Aih>{th5Kk#5F^GHScsm|mtsr^H1P!p
z*IyIp5@8K-8WRUCBo!Sr?z%E0ULukgU$)dFvaSuN0eUYeCH)BAGkEA<Q|gDLcW3<n
E1&~)_s{jB1

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_qsample2.cpython-37.pyc b/model/__pycache__/cldm_lr_qsample2.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8ddd32e99132744b105460242729d4c07b616995
GIT binary patch
literal 10146
zcma)C+m9R9d7m4HLoUhXa_=i?HIj9sEw5xbv6EPlV_TA)R^GIdY$wyWlUdFzm%DSJ
z&I~0jF^n7UCRQ)2C~gWAP1^3FMNy-9=p{h=(Doq+`qusd=cz!S^j4rnAA$hw?>obr
z)@mCP^Ub+mzVqGByk0IBCH(&7pTFr{drgx5jWUCuhQycf#EL9QuH?$T6v&nwC>9Ek
zSA5k{@m75;$XU65nQrN#E^p=gZ$rEbLD4FTHkv;Xl&n%vw#q@pssxkPq$tn%)nLk+
z5^3G91=H4aFk{UGv({`dXUzrk*1V|8`wPLMwJ6etzZ4v?js!=oqroxjm?$gw%fWH$
zxJVa$GdN+LKzhP0`6q*CtY>8DP0200m0iiLc=J1|bqe`Ow~Bn#D{h`f$r+SPxiys3
zyhY^Cy2>r7KK+<5Q&)KYZKvICg-x!$7H+TBWj^s0=6cL?ueKUVUE}#i6yCK%m+O8j
z^c=<~x*m({AaXmt$7@L|@Z!X4+g`Bdxh@965Vc!vCuuqUZRUjWdc*>c)pLCQcE9a4
zYI@7N)A2%T#f_U@{JI}Cws`SsbT5pZppBs`uO$gJh{CL{@Xo52ydAloj~OO!I8=GH
zwZ7ho(WI{OiXS!Y<bGm1vE2%qT=(v`;w0wttuRg;nkn(@bruCScC_B2bsoLia+*=-
z_-{I$IBq%NRrFx3wGPdS`ZAh#y|BO8$DkaEO0l~JHXxSC8idEeZ%AcxDDfmqb|p)3
zbFS)YyE5oY2Ih4>{YDffEb@KNeJAYiY4v(T?~hW(yMX7H@d$ElO1*p{?aK$!cfTP^
ziIOU*w28r7Wl!xHskE=9#(~@`r1E|aH9FG8`=^k~UzOfp>P@5*X)!IN>W<ngruls%
zEgs0Sbm#JW68aXhx&zd`2S)f+>8A8mDb-wQtH8d4I&HsrAiXSML^qkhn57*VqrUrf
zrB}kpx%kgSTZ8>$B5jt@K2PnN3T=5$=@rA}@$zi6`LX)(wcYs`D=K0Sa<&Jou0m@C
z(sHZUf%ysMwLK@XqtM@Gvp^5ZRJRq>CwVT3SYw0d!^mzj$7Kd;c?DGJ`F3N&2}941
zd6B3!qg1XX+ij2MT=6Qk!geRI*N6yXu7e{yP<$z9xvuYJ#awMiZC>!g4yYNV#LH19
z8SA7roP?_=tCzVJd;U6?@AJZdJz~K}iFZHA5;%QcrES=lg9wu7f$LlM9M+7vvJqoL
zv{0@RkT<yAXE?d_iNrpO&He1gg^eiiE`(9ytwqt+1=mTO3r!Yv+81th!W-bk#)cPO
zXaEHlh)-Vs^34kZ9~T<F8`wUx?_`{IX{Ei*^I%g<?k*D1odS?#O)kn+JSFk3$P0=g
zmy{Xg7L;YVs+8pJk*84dN`Iufwz`T_)3{;Vp#!|zylD4X3F(q;-|0AhR>NM#gkoD`
z;3X=X2Ux%pD*%Qp@=vfFkpcbU6@VejO@1V;%WlCf?iRpy5SP1>rDcNB%6%khlBI*H
zs=V0W07MHT_Wnc@boH2KdraK?m<Z`M`y$}=$N!C=s9$XyMFIWJ;i=$J@x&jaM<VSA
zR+Xi6nk}cZ#7OaSqPX&o{GpuX+0Ar1zJxM`%7|5bHPCuT!Q8oYc1KQ+ZYk_$qEosz
zn@+kp<bORa_a@V_NZ(B>c&qVUjG>EhDX*}HqF#+34$F%VWILbAfJUkSLJR;F!OFR$
zw4<<}B<0k=jJc*fBcY}OhApSL9XTuCQW;Ms-MpA1NA>A~5Dh;^{lc)768}6=K9nCo
zfNT{W$;?X14<xF6Aa7Q7<SpfqOgQ*ZPO8b&N76^~y5bscVb@4%S<4-H2e?41tD@Dk
zTYMn>0D3Of0ZV^~^#Gqb=ATLPfU~IqIF}Xx=hF$mg|viucaz1<rBDMrg0iD2=C_Zf
z6~N_m67YCBM=M0iOjWl$BlV6YC-zS!&!lR4Y)3)*)PaQDG0GhfrcR^mOma3k2hpaa
z+Ps8*71FSKSW{CzAk5aO%>#M=*;G9klszZ#d_Gm1a&I|Rd&k|$p7}_=|C8i-!uM6_
zk^KITAw*B~PNLLIYJ{ou#3HaJ>Po$5l9lv0R+3Lors@m?BSdA@o!ZM|#8X*W?|5pa
z`j-AkWpjcelt&8t9rP+MNxjoy38@NF<+Rc}ozA1~FVktjA7w2<h02o)bdD(fX}Uo3
z_s)cxTkD-kF3<_0{9n^&9!ONPC5d@X?O&u(aN;juoQbrYo=$7-^j<EUOD@sK$j_v+
z>6!FMI^oXjK{d&stj+rV#pG;y>Y>!nKevBT;OzJTut+U5Y4dz~{AEdz?DOd|W<iZC
zJyPymg<3d~(W5)LAZ@8w)w%R|@9c2YDXi?{A->M`aW(7C?HOSuc@e83HFwsX-_v%)
zZ0T9xZa4?VIkcM8bQeA^^~@(%g|(c+S_H3Q#{bIj_!LZQYza8ha4O;~`j|dDTw!*W
zi+eh?im(4am|PlQ^3(lyh(<p%z$4P1eR9RF1TnX?+P!%_iu@BN%&$cq6Y{wkC1w(t
zt+3&DT+d84JQHfv3(1r+qjhsT>M*mxqBwru8CYQ^w3OQ-%LuAvWo4!NwWp5RiL>$I
z%mg%(kqt;Q|3X6xQ4HN)ya9U*5*pST`w|3f_r+^1td*Hi!uE#exUq>52brOsH#eM^
zwzuY)v6pl|XtU^U3)`fI%!~VK-t@Au(AEioio`tE^rJP$_qS=Q!e*?~_|(Akv62f(
z2ZpJ6uI;hseeW&|QS%<g5a@#{^-SLjo5-U>-{i1^j-ND0vCxm+q$j5F#5DlW9ZaxA
z3H$}PmBU%dWP(?oTxnfqpFwLF>oZg_JdyrJ80nx)Z<5O7#%13L)?DXGg4}=Lk-Dcx
zYHr20TVX4)?Ul>0b{s#xf<)H7`!brr0PZwkOc5CjEpC%}4X<_Aa|I3$FW`&aS60m1
z8!enho9IK}|Hidjx6C$cMXZ%<14Y2`s2@%t@`aTho89htv$f)_0ETA6iM`mohcnKO
zYt$vP&3>)*zInyGXu?D{FFpHg%Y^wu{YLG^hCV+k#zFbM87Bm(OGX+i><X~+n9gKi
zPUhQA=EZ2d1Y>SUQ(Oas#JmFXMw!>xYDXX$K12B!XawA7ttWN}ZwE#X`z4IY3j@-O
z`P3je8gZWbhTgrg1C7Y_f~%Bsy>_y}CqRc}GuoKjnsQsQvqlraxTHRzRiqIzw@gfU
zyPjij(Pq9(fR@0>iNM<7R@IfPSTrDermh}pSAU7;;tr^rOty?j;DyDp$LLhAF%~gv
z1~zBY!MfI0h$Sci=6sE&S{74LsTh|RTY=NWa$8-GPqf0}#u?d8sEsqntl@E$P^p4@
z*z2O|UB~ZuVm_{dAGsPhZJvW5X~JR$O`-MdQ%{`dT^6-*ZoD+kfxL`$-X-48d^T}7
z3J{!M_ai5H;iZ1!#Y^ls?c@f5w`nQ0!Lh}5r=9h-X4>$M;T*$kcZKO~w_MolVaSUk
zr<Rd7$F*Cn8}G1Vlq|w^g~v=1jZv6_<p`KmRSBFe&lYYqt&<mr$=E8}>+pCwEW;xs
z%bZORm?Ch30QrlI97a|ma1wx5$6XhjypnU7n0S|(yhnge!76V6nWK{-^s{dea0sjs
zXb>0@lw}ajt+!uzX@n)#q$WgZ)+BH&I2Gt3qGPL6Dk#0L5bQC9U`>r>Y(ewvSExfi
z?0adtPt>+;%^xQIEHV(YH>ml-ptis35r-^}I0O_uKDVG*t1|KkGbS5z1O-}fuy89C
z9B@vEe6Ep5$7wSJM4WAp#g+l?6xzs|=&!@IV~6XR_<%Ss*mlEr;@GyqgS6o6D1?7=
z3)vU(#QzB(Ef?W+Rmsm1e*-BE50z9EUC#bRxu#TQ19=19SW#Y(&&YGRL9H&&4BAm|
z*+8zMYKkH+YcpaD4W)DN;S@Rh>*8Hh4UC1q=%p()k=B$2S<5dW4R27b$~CzHpRk4;
zdQB{&XH6qNv4(kOl=0q!(fe~4*%<n8Rw^&4@F54$5V&AN-PBMB8~D+iNSm-06U8P{
zA|nEV_(v+IGQ~h#l}zy+iV|q6LX7v#2lgh~Llg^c%Ojn!9Fa24fs`gU2BCq&Zl{mh
z0`#?TJ0-ch_(tZU(;}bb)&}u0{fPN!77@1jArc9E1uRr75`h{W;S43E;7mSIH#IC@
zfftm8No0h0jIr=d;r{26Jo+NIvsu_t5O6sIU!jPA3k7oWsfJKaAvN~Nj_l{jR=#rs
z)~U4LFGENtYa{KKvQ`K1PvA!xK$QrnF!HSXnvap^UoVTmiI5FKY_TP3afSeicOf3R
z+HJLk<YEDp2_bZnQXxRSlu>sbX|9oYVb4<Ca|FmoXBP>)0Khe{NX#x#o_N423<cle
z^Pu=ZiOypV2}y({K!gZEEPzqwlnRKiTN^v0K@4J6&;q-RPJJi-y09ci;5vSpPrrH_
z@rvv9Jo_A4v(FRw0s&%CMg$?E6ujK$=7#6{wmn)Xth9U7NU-%QNYTDx+N@E4>Guo$
zD<4R&7%HRQBzA{CqAU8vL|20#ho^d)ms?jBCA@X~DMl^-P#2*OP}Xh<qDgae2z?a1
z;{L>e<j#q@yld<hc4Vt00whoZ{aO(wL3yR0pC?au7p!GX_DkS@?!qgysxCY}j63Dd
zxU+cIL`Y>CbhR)Hsoc2wHe#JK&}bCfvSCld7(fp8LT1=d>6W+63$bfc;LUT{^Vl9Q
z`v^0WlVOWEo6bDwzi=2HHPD8hNjz~EfFdrNa;m5KhoEi<2dp@iY?Z>KwnjW?=TZ$&
zPZ1uq^DcaJSAj<@*+x>>M=ZoulhS?}yjIzqJfPTwHq2Kk-$ZQUfX19kYG^e@(FoyL
z<yM<m0Dk}e>SI#FJnw~M6);{zaEml6^m%=n-9%kRM`ZR3b3`CT#N*fpRQnqQX#eb6
z0K7898yS}|+n|~j0g6g-tq~z;XSkl%rj#|c-ePgmM{+jmE;ar(H7;Q+*Y2;3&7i1m
zWK8-^YV)+|HHt$Db{&v4#T+x*CVs$$M4uOn6^Q+T@+lQ`@aB>-6zU*?PYZS00X;&f
zpGN{ROCn)DgL{|-#0Ppx0_nIT?<)sVPZd%W5fDm&!qim}jmyNA#(oR^*)#w&GceCf
znKi!dkP`#xPja5oBv!Ry`{LHX4m>Bki>nM{P%`3Tu|+<Q2ry!^gg4>w4v_aMZJUTs
zg2pRB?{!bUk<mA8^ndBJ+MvzSj|?|rWkI?Rk?4~yNNW?yAl3dzNplZnh<d$;pdCt8
z1Y`_^oG2Rqg%p=hsrqd&Jjx72{iV$ukYyzLejau5fkNe*T4HP#a5o@RB%Vs321qH;
z9YHbGlTwNhQd-1Qz+-?J)IDvTE`_w!Zf(unFz=ZS^Z6_08eYi2f++`H5V36=Vl{B+
zR%oJOi|_!sA&hpy75@(56DUI<9D0OEo)?;40=o-gay~gup<sJKc3ep!uDjmd7Q6u6
zYlTrewh9dVFN}SmefSi{Oj^xO)QS5G;5lKpv+vRzaLjN~H6ad~=hmDA_L{uKm9?1t
zHtI8-Beap+5QM-!rea$3r)m59D7{99Ok5%<M{&dhV+{I6!-3C?9r$o}!xJW3gHHG>
zcw!oijDQ5e&A38=O*+v{t*0lseI2JNFM{B4YI&q_(l}*<y#+xA;!h#W(K?S?lw8jU
z-$EJ&LcbriML*Fhx2N?AuI}b{rKA8gFKrgdU_juwFc?Xc_D2%R-&kee!`d=qNgyJ}
z2t#9<nm9&?Be0ODB1t={AI3CP6cc%qP>9K#Xkyv#S3$zyV%o@z^gZQ{c55uUkGnc$
z<s$oIG={%P;_VRo{j7L7^DTq+l{GpHvI2Ff`|KF%GhP#tZak17TvBucJ$x0N*bfNY
z7Bh*qe9|3U8PZjR=Z_gULim1%I&2g82%ugRWM$>=!?AD0+gwga^oH&jY5ghMSq2r-
z?ISL1+Hpo^LShcZCZ8Bv@;gKE347X5=q_bWf#nbDdyY(*xu@^+3A1=8L918*Nl?;j
zDP*X)FjH1J$lF>V5*7N?6W2_`0N%g?mRAoUpD$CF4%Ap*_Bpsf8>FF>n}Y+9lx}@&
zng4gM$Ct3k1`!MW3V6tD77F0+ksvt;@s5iVA@xApARw)mOLG~IK;l7p3mzfqmq%J7
z9$1u+%4I19vJO3@eKpk%WLKY;-24#=*Bh5%^5R{<n`-k<)D}@Hc=587Xt;179;FdW
zZh21~8>K?MH+77pxRZMd#;c-)6w=gcCQ*QxOsufqrPPSSDA<BKy*rs2nshxf5r(&Z
zb)`jajsTD8EFo_-5j4>#_JgQjMqG5KoujLE+$Cm_h#0o2AaD{i?DqhyS~j>nCTtmv
zX^}kvw^SHmWONm?Az*kl>xE27l5J^~`sLX~Dyq5M6})O`(3W?E+_Z8K^iFb^a*A{V
zFM>Yt7E#Gj3@E|nBvo00RWJuxN{<Qm5ze2|ogJf}QQsk%kT4}u5oGcf5)h^&*x>qY
z%3s%rK!l<sMF`Kzm6s%lO~7@HlqOt61;<4$qVTia++H3+v4A7fSJ|r=k9`6#wBW4#
zJqJNu5BklTY)3IP|6R{U@Hl8E)?|o4#CSoWvEIxge}r2<V!-)*n{JaKG0988VQn}F
zrroEso}slbPuO44<bO>S{i9@DUK3ar-Vm4XvA?4xe?#C&Il_L7vTqW4D@aK2)+<0N
z`QXJT5jr~Hp|GZy)KF}s$N+ItPlmAt1n#Q`gj_PcMmcg|wfJ_LmF70}&3x7lUj#t^
zh_+N}j7pQjW^r=@sE0v^F95WD9)~9KWQxOS!67IV>Q;OPkMUbr3wwtE$zNV<I}Gvk
z1kshuF8&j$B0<Ye6F5tNJX4;dh;zVmuena}UR}=!`5&l87h@nglfCUrxSVf8dIMZK
zgi_(8z%8LDn-~ufOBVko;kbsOB#NNmuLI9>kkM7Sdwh&=!C0T_J0xL}co-)mVM4ud
z8~Q>Q0a#(3665iLuv`#IynlFglXo4zh1-3^V8Os%`}?#oQi+hjOJfT;)J$9@upgmW
zede%BOTD>@NF<{Lat$94`0Njm%$yXmt1`uSgi=Dh{5hrmg23MrASz@(CGbxKh?qv}
z9@tgF74D%5-!e(DiG%?Ysvj9rYx9N<cVW_~RW)5M)v6Q51f$mVb6f`_LD?``s8}O=
zfE4+S_-dv>sECCl#D1R|5GgR18W`d;kd-hL4wH2PUplNn&m-^}uM9Q+pa2&TeijoV
z#$aT(iSUOtJKk*=XcjS6>)d$Vr-qF1;fT}9EDI7}S**51De?~)@vL>SKLkEW87*qX
zb=$%4lMF^Kf@M%cu}YHRS;&_t(mMO(B5;uXkCte_v0UaDm1u}~r4a=|6bk>(svsJR
zZ<r9Lq;IX55Ln{d3a-B`&?RCJ;xtCbnx*JaamSP)@d^>Q_+A7NNr-d<klqV=Nj?gp
QPhX4RuT+<6A4vZH03cLOZU6uP

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_qsample_0_8.cpython-310.pyc b/model/__pycache__/cldm_lr_qsample_0_8.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..48c4a2deb1d077451b5c6fb3a6764f88e124f27a
GIT binary patch
literal 10348
zcma)C>5nAmS+A?BtE<oH>ACjowfE|^J>w&>9lTk`@p>JS^(6AH<G4svs_Clf>6tou
z->TlDskRf(25&$U2nrA>3Ym!z$bdk6;E+I(_<|6B03Y73K|(7b!3PAyU@yPtt?r(l
z@g~r%dg{H8=Y8(CqH@`k@O%F6-u6ECk|h0vGSfc|na|*f6<Ly8$(4O6kS#e-EEFKG
z_^PGit@>J!vvQ*{-O@!}-pY^OhIkhO(=tUH%`XNes}z*2a!|1<LDi~?@|<4_W~>>J
z*8O@gYt04?s}amubHThdA1qi4qAu?*220kGNE`l<;HY&pIA$FSj$6k?S;0RMoU~4g
zwCOj4Q`RY@i*Ct(BzV+%RF+<q+_GD_E4dYKVOzCMBVToE$k#k`^9)MPqGZOcqonRF
zA%D(Qu1T%g`-GjA!t<{=y<Ru$aP_6|&T32M#V<0~W1f4t+fG^<&$pxSrX9Lm_q(Cz
zFkT#ZEV6^h?fV|DC*8n{6R&4`!J6l~7zjhuul1ay>-g816UOTi3q01!@rCQ7wy#jr
zYu=5%7g8&3yza%H^`rI{H!nxG!q^FV7`pOOl2C&v%<2knta{07k?Z-Gp?cM!%FEsL
z^?r;dEtOaNsADI$6WfXHZqVVnce@)WF<<D0apKTSiD$2~D6p}k^)9XR*yXO%i9*MJ
z)#=A^*9kA92kWl&X;##i(R|>Aqs`t2<w#VD-L<g+u}sz`JWl^qWIlr@o(J$GOLiqo
zadWQfYIkMOm<;UeeD>uiOjzXmp8G~P+SlrpwvImIUl|Dl&lm8-6-?fdhWSL=mG`7y
ze@2!PB~?;s69c-+jyg0_X;({)J$YD2<=q@=bfitZ^OvQsm50T&n3`!JRku}?=68+M
z+>>SL#&Ps3WHo!Jc@xZ_;lHiQQfNFS#TRd(75(4BlWMNim)RH4UfVVIr01lenG`Wc
zX<NoDk9}1cmM~K;wnR&bttZlE8Ljiwx}y*hc9fwRR;SCeu@<K4r`LMpOR0(d$m9L&
zR%qqdr1x{HS6Z@FWM0p65<3e0J8TZ9!U?HvH)vIPE{RxsgXhD@?l8w?giKxm<$Au|
z-f+Uu^J8uj1!vUDwd79E<2hHnO5L#6PwX`!%$V!o5)WKZ3c9ZAds#78dr^-Uys!@n
z2YK;w)K8{5sck3WD#}`AuEn0e&gI*@FlL!p@K)m8PO=23f!AmoHs&BACHmm{)-8v1
zVy<k&*bpt0s|4f?u8){d?!GUvPhxW)d~szX3cQstO1!lw+FEg)#98UEsNY+;)(@|O
zJKGyxxY7m+R)}{#`-`uy2z;!xeK)XuX5YxT&wkQ=ez|vt=fS<0-(4aaJPV-7rd-2Q
z5>E-~V?Q+tCAqGY<VE>}QjtrFDG!c5guItWW3{x^RUDnh4ciVKAm8SuJ>oH>OSXNZ
z@Az2_dk%Aoy-k5<scZpY5l>8#-H<A>DESySQZ4=BuK<j>u^<<ef?CiDIsA3;rxgA3
z1p|LoK;+%(d(yh>7F_dg0lW*jdsniwOyXL(_rNHY4t}n2bF^W|C`NMPeN7PKeFf?J
zw21peeb?Ei0k7PDfS;&eZ68Cy_;((O3LX+6@c_LNX<P6+IB%9!(>db4xcYghuWk8V
zIm@$FIvX#cOrbL3FkcDO;W0=EEuGtz(_>o-Yb7~K59iXVn?rsjEf1?{S)^Y`D|oB%
z3m8Ke<5FH>uZns#esxlA-ji)Tl>zgq0%)WvU?DNnTvFUtzyhVzz>K+$+>lUH2A`MG
z+_s#RZ>cOuDsEoPk)!%_QOJ|GQNK88rNnQ4UV2x)2T253^M^_4o<w!`<jwN7yrslD
zRGU<j+I!M_^19+0ZsD$x%w+Ah<!#_0{-J17cg=g!_n|IR9kBERtOS_UG4pJi2W+GU
z;9Ob&oKK5@3u%e{I9c3W3KhU3C^?#z0gt5>z~gBZ@I*RKD?{pJs=DQdG(4U(cTXjc
zq-uJ6TS5BKJqfwvl-nZ=okrQ2<ZN;dl2A#t1quBsq^5VUqK>>rW1Xiq_vGDHs_u=;
zE(rWSmZ}|jcp_DYC*A6>nLJK7hP3@hNZV7xN0J%BOnPbwSVHNq3?EINNKaz+e2Nt{
zB+OY%YwpZW9;2Mj%7!P?W~$$menV#85T~yobrHSFN2KAIP)DkQR5`5-&!h{eTTN#H
zuVyVmnaY!8Iy01hC0(T1hi5~@tq;#8%XD5SUr!&sCsEB-HqYtZ6&fW@k|!}vF)gQO
z(z-jllM9Q<Q#3O2jdU(On;uPzZes_!Oa>Kgp5I+b&ZeioEsgRQc2@+RPVND7)IyUs
zA4^X@Cn=KsSb74pphm`=H<r><8RfavMQKaLntqa=9G;tuIfIfPPq0O&LztR#=XZ=S
zpFE8<k!C#SF6?OAVy^TYFgKY2;~ZK`+H@B`B@LSg*MyZ+CRqO042KWFvchD6<p)`d
z5th$R)|ahrX-7vZC0_VHuy|yQ#gF&jCYpR^j6<Y9ad5q^1d(=Rb@2L?DDqF8YQ7cq
zn~>d|C}}29vm3VkzUwuU4X+73>xE>0HKX<Bov6>6Z5GAx6VBLlYeJ2=T{5SjdzP1%
z2X8%e%zm7W7iV^3GubyBY3Bde&_e7&`I}c^#6cRv#ABa<ARRpON*8NoO=xL*!*krY
zi4n({i6PM3aAMltn%9iIWbjUpMK`<HCN*SUJhB6tUN#omIw4SzG|zYZXwC8cJG50{
zXtrp4YS8qtl9i+nYqoj5=dmYz?<TC*<}HjN&<9=XHGMDaAde0so5T+Ke$pl-GX7DT
zbiyp2xDEiylTv9(0)GK+<zNlTWV=@muC^t!PoVYt3U5%w<WxqRVWe*|y-F&T8y9^i
zSaY3A334%MuDUchv#;2eZMz$G6Wd<C2&2gH<4Z_nJqFLA8Eof%8#Wga!^Bu`GOz7*
zZ+fo4<l!ZJdho(>^ZG^?XVN1o5vOqVm220UJ=Tp_H@O2u0n_{aa2Am-4D`4;7(CwW
zE_=&>VYBVTUfjHelg<upze{Gq{ZjX9%}dQEo3QSiPhYsuZNf^Te*5jFhJJj%7{}%N
z%{U=QT`~e$W|x4U`*bQok-snVe9y@oAdQz`^Gzs=YhaU@S3u+_^V(az2!z8Ml#hW%
zz>V&DVux^)U>32f7?T&q1RC?1adLme1?n4mx2BFXBCiczR?hW$$p$Zi8p*V@F}F42
zc4KFaCV~x0eL~Bm5i*ZXOnAMOW1pwZkhdX9*lV;7&4owSQnF&vfGn7ndZ=CNlROvq
zLE~i9Wkg~%V2XAetYdwd7=aRC$(LyS6Jls872|NT8#o;-uRHK~u^Ub{%E&rGZJhaV
zZI7#jIu$&_UKUkfcKp8g3S$xDD!7oVfz#tTh>s2odC(AA&pq^ndERAF59h>7(|pIv
zSm#aR_{_Z%hoJzW`E@^XlBb>>C7yYjour+t68Iu5r9M8U*zWhT-d3XrM;d-Kta(>h
z<aXDEDIbR1-1mPOIelEa*1h@$J5EUxUM`$#l4Fdb8Z1YkNmZ4=+45}Rfzvv<IZ4J=
z*<Obu)MptU*=g!jBrrqZ6ajK6S%E-_z#{;>Hto^a<gA>}1jC!ugv@4f3RZao$lN~}
zLO*i|tPyAva0yHZ$TEn=)_YGqyN@NdK~0FVtSWFUI1}h1B4b~oQbE}xHDFJv0c&O|
zV+$H*uTzJ7IC9=}pQ!DQwQ!i&vmipuUZLiT<J!@#_c>x|pCdrQ({l@&wJQ6LV#ZWc
z{-Hn%4i?^~f&<P8LC-Z3<T!1H*okw;W3gp`8-)(CilcS7cI<FH6Bm#n1>0`>P8{1d
zyNu;B;p)7B)H8VEt55{y2@Ot{0e6f322vUxDyhl3qRaF*<+@x`4CD>DfoEPmjWR_x
zALzNPg^m{EwqgY2(EdYBSCodVJWw9!4;pzIPs6{4){3HF1O@G~xAs%*fl)Lut|``l
z5#jimvZgFzMM_@(AZsW3{I{Z`20gV8wV&3?b-99n9VKG4qW&NIZL^gAUrN!CwFk<-
zYY((K?Z-e5?dvAyp!L#hhx=$BD3zZztCE{;t$m>XOX+xB9*m<f@Y*KItxbw={G&AS
z3sHMa!YGjv8Ic#nT~axfDfZ&3WW{e&SVvnGx<FVW>~m-hX)YKzk95Yx#2pZ)q<Oh9
z4niFEZ3EO6aMnTxACm`5FK593TIN9k&?eC^{t-{pOf*OQJ%9v`29_$8i4c*FP>+&Q
zSSX*Un;JHtz<tU>GBSco##G>_kmS)5L8Z;YwgR_e7S4u=&<=%=@~MU(QXw^V$s`@+
z$>hFq1~ZIG5KPKiNV}!1%^n;ZxL?LKiIL;pDvN-O5H3RQu_GvD6l4(kLC9CG4!S)d
z`dCO6rwEYPVi7>Alu4=<(p)3?#4b?XV+587JW1dw0Iq?@V)iuUiI=RxL~<U+K7|}g
zgnbI&QLKXqwWQ30-5PQQ?4c=x`qYVx!y0qhxnD%Dk)MA>7@+&$3VxZ}zj_@ZkSnb`
zdmgRXrwIH40b+GV$1egOygU*IZO`{15-7%#f!mZ4!r=v^Xjd^=dWNJFrV0J$r7Qm=
zy<@ytQyGmw;`_jlXo-ID?*okECa&zMp62D&l_i85Jk3=--PQ2qyu6D@`Fcr2OY*L9
zw**;7Q4=8AGk1%7l6zFtmE7{(!nSOc$N@%ec~lDr^)B2-t1`+Tb7$SgU9go^9hJZ>
z6D?})g1d;GGwzam1n;_t!pwsGj!&X6S1-TD8&HrGL9=0$!?r*U=0|21QR$X<hZkbk
zrtqHUvL~=jUiJ~-CQro{(L9}bP^aO6psay*^i=W0Zv#+>W>Ze}u9W6~PyRM29O41%
zP9<BV0IjVN58Am@1JqLlX6?KS2j5lTnoG8k6m}6pb=9P_TL!mPHmiFS)X*mR8s$3(
zYV6UNGf5q-W+;#$+_c>4KVuR2{o9N8%Xsm;7m~5TxQQqlDQuVttyy**bs2S${~)X!
zfgBP2W8a|KcL@+4*fuq(Oz=lGYRtM+vq=DfDezM}LVVD0J+DV8Yi7O6;$(#EY}EIt
z@jYr>!dR|7S{ZAgsAXiV`*mvbu<135dS*;4h+WYBL>=NnM2bm@)J=1w^mOp2UZ_!Y
z@}r725$zy_hl_UF1wBHvKaK=smqa4H0oq<fuwkeqkdWK*uCgZ$RUt_cPN5WNO<fhC
zJi*P7@;m6yW&xn|fqh=e4EuG5{2|nnyk|6tRcqV6xVx|e&k1khLd6)DOz}5ulrJFW
zjA$_7PPoux1ina{CnBt4%cNHakG!1GI&Jo6bXslGX2-uG5~L~w8Gjdv5gCKTHbHEu
z_9OXq2_&5(l7{Hlhk2+AXg$Pv5cLz0xn_#XuvGmfgaOJ7+&)N~IiSr*^xZt_<UNJT
zH?_prEZ~kpraJ&Cfnp#vL3b5qsv|;!$W&_LDc~`{8tRU=PM1{L>Y%>X+-Tlvwwq5}
zYOdjh4E)@3-~|!8V?*=?4&71}+qQ^2kZ;4jgZ*&Dzd@J<@(_fFt|12Cg^riNxI>7Z
zSEuO}!axuqSCWY9u6MHwS3~!@VbqJQ0)tc#9)-|Qd<J7C-A+I1$D;-CoNx}<Z_*s_
z_Hc1l6i3c;Yfb{AP~PIoTFkzU`i!K7#**8DB-n3JF)f-MqV(TE*(-F|#4(SM>4O8O
zKWQXWgJsZ|I{3-n4xBXEA9T{>qr|@rKt@4=C~2Ih7%83iCQO4w!yOCml1Q-Q-111{
ztZ~i~JCh>j3sHtEP_%}?-^tM_UZWhSV$_pbqo-(>+tG#vD1`i7DKVh@M@9tV&xP?=
zqI7Dkma)o+4l@IakO7C3ur#I7iGPHE0z-+a^g#WAkco;aBKT`~grMkPxnDCt%HV6-
zPDA>Rasx!R&vdv$RF<D)KSB>Uvn2HP>x8c5=EcmB4SJW>=vc@)T$BbErVb(FJ|X|6
zqa?!rMYop8-a;q#yJ8a2j91<9Ehb%pc>a_fB&6}TsKYN4_*DRmK*Gez-$uZo8{gq_
zLb5n?r>N@>sDVL+bRmlSqF$U)+P(oH1Wa*iwQtad$O5Z?%4$l9?3@CObu!XFn%9KY
zGym|7e#~N?h}9a#Be6;fFNI*Gd%Kph%5h0&EfT#7h3kpiu1T11Y#z(2hw*=cy7Zyn
zMgq~nJ>EDCt=$=)Nlh8Frk48id)+7NsZI1U{uPiQyIN?5zd?e8Bji4=VuTO`nS+kB
zVJ^*O3<CiO%`O;*gkc_Ojo4uc0@TfADFwn1J)~VV)%IjpUy$7VQMw2)F2W|pyMQ;<
z<`2}GDAkEEFXB1}7a7E`G$QP%9d&Ay3iaO9F_Pj|cNFxfp@cNm%xWfJfSi$(At^A$
zY!uMJ_2P|8KUH0i?1#w(WJ|e^z`Df#46W|dc|!bZB2c4Id<kC=WQ0an?m7Ar0N0vX
z^dp9OOLWdijIr+lSoLghdrI~)8q*?U1Fo>J)W|Y3vms!eby|f?d;XBxl}6>+L@KJe
zJP<5vX;7m#gc!AQko!(@n0ie5KtKeV;w_>S3PR}$>ClpVM4lIZvLQD?@MJcdA8FcP
zZi<TbdmkVclCMNCf>=I>1mr6WS&)moDSuTXG7;L7v?H7^S3Uwmk>I)pDXPMcRB&kI
zM+#TW&F$nNB?~w>eU-h4@z@^|_$GjrzvUo+>_OpK)m{|CD7fj_2vi5X#Hxk}e@qt?
z+UuPxzDU^hBL=MBw&`jc;&UFu;kY&&gzN^C)-%M8qzn6Vn)5HHVsw~{%WDG9!iD1U
zEk=yN{*u5!slxsWWp5LL>0}Mu2+xzY5_04aO7{=>fb6Dt*hGYoj0ftZp$uybDBM-|
z2)$&R?dQnL)#3|jR+`(?H}hFLe4qf8B-&D`v0s`LHqFf<kPnLwpCo9bJdRD|$zF$_
zBg7RbUhx^6&26lOeVG8sWN!8xh6sLw$WCTFldr@`__8wu&JiGy%5xOm4tVY**9qQi
z=@~))E!F5^3`B;qxBWD}7wAEv13Y~QvBIN)4?__?F&?6&EdCZ@nrwPq1XXlkdmc70
z0z&fO<P`aW!9Lda03nkE#8??26Z*-&V4p=otOFrm(mRaUkQanmgJ9<E!)vVGbo?$p
zJ3uTS?CkZvOUolY38(qU)N&3r6Z7D@;!n}C)i~@GQ-7`^8p>#qT*J2wKKlWZnMXrr
zSEe+NQc8%MKcdv15cq2XM3e0A3H$>Aa@?j?F3hfHsq7s9e0VkSrWI3BK2lWWLq${m
z2SFqRvXl>$Lghb=rmh3WRUN@8U4`d?w_(<bVuaRtu7lB_k{B&rteX8AQsjQ(lb<%B
zCl-Db`#v=wB4HjiFvK@2%V8+|E$b9M%UFh0fN*fUJh2nT1-J(Bv#=7e3?nm7#8j-g
z>28zow}|{&=cnsFHe`gaT%2xZhmdH?BFH65k$cH10M?n&5cqy(e^JY>+Y6?@A7SKN
zSO&Ef3ng)o#iEH^t#b$0fzus*ctvAQ<ubRbL_@^O?I;MMP<W751z}=*hlKzteJ{p@
zOcNi0aQ!ubF%j7i2QqcmLRisR<GL%u<0Ybc@o`H{A_LoiCZPACQqqs%-M~Ztno>U^
Iz0>gj7d{7L{{R30

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_qsample_0_8.cpython-37.pyc b/model/__pycache__/cldm_lr_qsample_0_8.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..520b098fca9b7d017708554667d5fbf14e8774ae
GIT binary patch
literal 10152
zcma)C+m9R9d7m4HLoUhXa_=i?HIj9sEw5xbv7K0vV_TM;R$jM~Y$wyWlUdFzm%DSJ
z&I~0jF^m)MCRQ)2C~gWAP1^3FMNy-9=p{h=(Doq+@<*Jf0)5h3ffjuT0<^#H3~ySi
zZAi>F=YILlcR%x5xm=X+d+F!j^xk?^lKzb{gP(@Pm+{1kEJ?29%DxoHmK-P+3XoTP
z)l%_ReJ#jYxqg{$>7p)g<@;|#ybD3mDvCCmKM|CyQc$+aLB*;Blh&jt&-vA0%9;{s
z-LD1H)^spq%>=X7Y%pid1@qRtsLT5c!J@S&(uThj9I=iBN3Em5G3%HpEBMR7aqGB9
z7kx80VVyvF!Y%nHgQu*gWa&-GExVOn$*p+vJF0aG`AN5meAO#%o<_+TluWrbl+?UM
z<j=awO{qTph%i%Ec>Zmt-EM_VuD%*>uhwNg@eXr6=DAl}jij#ed?O0)*rCgHzZH58
z;}czvMRpLm9pB@%q!oB^;<arrSo2&L17V2T&9;-Y9RC(`!gxJmfye4OK7Xs<_6=%!
z)4SdALTbg08(#dHA2qgk@oIE8jGds3p)0Q@2{nkqtgi6xs+YVSxt@<1Ca*hGd9}5^
z-igtquJVc>HSOeHVmq<j3YuK^?zQ41=JTyEP8^yk@$7XL1vYlH-lBCLz1nh`QRw(@
zI-NLfIpJ0GV6C+d&5HUmns>dhzu8Bi9EnP?y9PEOmdP4~$H8w%WpgO;BujQBOL23q
z>T0_(=t~CXbw2%i6ecY4eb0S2?C)vyT0`%TQpUT0=hyKFa&1bzd?M}32h#VxAxnvp
zDyg)I!CYle?HQ@GucgL;+$*H=ehxJ{(#3nHkjh_`K3M8aq!Vc|Eu`v>+AF5{eIqR%
z$g*_%^7|6{7P7hn)V&Wz_)Y1C^i?U<TxqMozKc3-zjz?MC}Bi5nZTH(9T}s(_jRRL
z!pOP!FGX8}{YxTkmeD>>?VAd1c~9vT!{zbvY_$2Y`th~h{sb#3Vh?h*2du6_YX#DB
ztJi?}3FfsuC$Xc@-)6Hw56V=x71SqrE{RxUgXhD@ZZgMZ25NZ)RO<P5W5WqU&yRVL
zs5PTht|i-TkLO(RDz(CPC$ZOv2xG2;BRo)iDQLN_?`6eYZAWch@WKwL8KlI^Q70Me
zq&A#{t0=3NxfXl=I+yS9!hk(u!AFUAFUb-(eO{$)*qDO|lIVf!TX!ASjJdKAV?(r1
zt`d+pxZY<tx%H{UK99});++c{QQ%z&qr_W_qOA+AlQ<WeEb6o`-0Xzc!HJCxFTBtI
z3N8?zy!MqF7X&^oG<-L(eP-XzIL*FjzqHcc=6P@`=64r~?oI(nvL+YhDxQ+~SL6l7
zkW0!8atq3`TvbYP_sA27d8I#6U0Yqn!D-yE?a%@0ZC<qd+=O(=wr_VFKdWIcVnVUA
zG4KMF%>yjpi4_1t7Wv1xjwpeC@e06@>n1;z)@8Tg7IzEaJV?x4$<i`eY2`kaG|AFI
zSXEx^ZvfJT5r==O3DSB*vpph)enhl%i+u_3+N1x*Pt>nAj-r5m=kQeUsCeQ}&?Av{
z1h>jkI?a~TS>mO5IZ<4BNB&68^6W-B9bZD3LS@7)z8Yw~qu}pcI=dsMN4FGqBhe|{
zn@uO(9P+=NmV1+FS)}iz6};8>4#v>MxRh7e15vNW4~FH%`?8%+Wk4fU03itgi{R#5
zQrc12Pm^+LV8&cio{>;f0ne7x+>V@;Z>fwYlWtzjk)!%_K}d&RqJCl6N{N4&C?Co9
zAw#wb4`pU0<@*xV-j_EkJMxzDP$nFFBq!Bm>SO6+d0lZ0x3Fs@wXEfiyaQaI)m71I
z+AZFfeh5{U>VTy`!FqsC9rMqmdBEA!0Gvw;fb;1D;6hr$yt~Qb=2EBu9zogB6!Y81
z(hA^mIth3@oud^ZWu~fIo{@UTk`wzUlc!QOJ+`ADed<6$?il3`2ves~b|yKSoP%^z
zQf*#BzX~bYJ*=rI9}s5i)aJgt|8%Mz49cDncs`%1O}V$6s=edxWY2sk-}`CuEaCgA
z^iclbCy=8jdM8n8CN;uTdSVe+6LqEDQ^`tt94pDECsTC>vJuj<>Q3$DG2*GLtam&$
zQ+-Q+sIoah5z0e_{XTk?m!#h5u!K|vsd8HBolfUb_c!S@;E%Hwp+e=!1v*ER{w!Uf
z`Fm$V&8_v$Bp2ueQU0&#Q}-pR*^<OOr}i(>C^+%wFwR6;PEV&bcX}@u&Lx*<WaMYk
z+4M|$B%N?)_Mn|)P}XLB|6+1BJ@r89=bzcXC~$WC09d3JnzVU7J^rGkNcQ=38MB~9
zmL4j%uR=4N$mr3XT#&X@tm<5Pymxju>J(P?$q-*>`?#8Q=k|=Sl01)9k)}KA&hKeE
zVz%@wa5tO-;~ZK|YPt(wkb36htHN5&VJ(8!Fynt^czgn;HMRsCX*d;e7JW>g9j-7t
z%f&sNTE*A?A51O{F!|a3J4B<O8{iS?&p*CmSAv*ZTJ7Gr7DfJv6Xw^VjtTkPj1n`6
z%vRX&JFaIY8=eVG>V;%YnbEqr9d(%5U{M@D>kJGr6N<`hk#Pj=va+(${n``9?8Mo4
zab^XY$;b?(nSZ6Bg(!x4FJ6a91_=#=jeQvcw)^}WEv%K9(8KnI=eV(n5eJ!}sy8>B
zn6|g(nX#92KWww;P7B+lhRlomn%?xXvC!5Dfr`XD*Yu+`$M?5stHNxo)A-cD^s$l)
zNe8y6d9LlTXMOJuY*F(r#t`U(HuX&33!BKJL*MGKgN~mxNU_k5-lQj{@x(O%&>gI>
zMG5=`xRt|M$z+9B9$#r)W}icA*y}S?F+7p}Mi}X!OmC9P<i=&+3D#WaN`l;f;E}qg
zM`~`xwp(E<vF(-1FnAn4zJf&7zWXAY!3OR$U{4Vl42^D+c@3|1$8!Y^4=><L-IrI)
zTN^E$Mw{qE;Q#s?H*cD4){0mw*#?S$<554HLgWiWJ2t!B^JZ(sTLBEsh7)_Sc^7A#
z9oMK!W}f|4>jU$OdC`QGZeDu&>6Qr#i29A%jSYQ%RE&f2eKSr7QkRT0R@fC_=MkOB
zz@p5zoy?KZcnS90kfyi>28nqE<c%_~vDJ=1GJJ;eG0+IO(OOUJ5DpKF9`<V(lNSc0
z8S|+@ax~&R^$oqdV+R_M_XTe$=X&jAgHM1C$!xSSw>9OqVrPvef_+JSLaRt4WS*Ir
z@K!y?-lENXg#ayqkspDz!?UU@S+Qt9_Do$p)UN&l&&3^3H<@f1k-!m)V~^3PzQI_;
ztQnY{O$X~*Um=#D1eo(xnrc~0MWtd~UTg(U6U%LNJwDM2hZ|>PJE1nte6xnfRYIi-
z?qRQqs`nhf<B9pW3V!5j;Iw%Tf}{zf9W;g3vrjy6o_AT)#<}s*I0y1F)_I3`KXco}
z;V3|Ge%+6p<hd96iRUk|<Fu3O1m32l)CR{E+nsjS+nQ;^L56<}%iR^$x7~7Kwud1v
zj{I6i4jtESwywX+j#07*?-fonNi;^03YH^aQdK2zwme&S*0fGu942F{Y_G%V>97os
zj4X3DL12o&2?FFYGV&Q&iNHw!ULE&dY;sJ_Wn$tzYVtk-It8n|0c4I&hS1NxLBJuf
zMxa4pNKlqRG`HS<?u8MSSd*F%rCF1}vEWpoi-?Y`QmLTyzCy6a6oNH1mazrRv)`Z&
z`LOS(={`~0wl#m4__N?Z%wDJF3xnGJu16fQIN}gc^!VI@X06J|Da@E`%oh}B!NJ0_
zRB*sKA@aFKA|0pA5EF5>Jr-LAxKn5&Yofmn*Nz>oXW|3mykOf6--%<}1`pDLv!f9H
z%}r#V#}of2fV5nM*HtAyOZ*L_G(1#NRdhM~7v-8#l?~(#cw<F*K|UkT<p#C7JTquV
zy=4QrimEA!ysXWLF*KCU!G}}i?5~S=Q8h3Y{-T$z)I?fS7Gy2IgfzTCwJO)-3VgyE
za_BX&jGi@({KOjOnNh}j4@U3LVPs?I!&#}kq{4?B1VrG14RupPA#C7BZz65NT1*t1
zNQsOf2tpvKoXQjiaaA(KcPLPxtqL*THy_xWXb({=xGj%##&Sf;I0sUi+!#a$4!fN`
zY75ZU!tIpg?&9m2mrjd(oLd{j$MhrSqgh1c=0`{*@D;F7u}H*fbVM|il;Sh_MBUV|
zd<9-m7AcVt<uS%0IE8yJC3*Bkd}p(;qafyT2EIZOF&B#E<WmjNoI+~slO5U5ldXLF
zI;>M^zh8!^PS!@+FJ-L`;Ge*cGJq-(Q(@#;_cb3Q&%a(4u@fO1gxF$B)Zz>Q67NDh
za<$uP3(3U-DicEJB&9-tdMTssI?`Mt@xq>_x@QQGkIpU<cn*MTV3C+zqCD|{RTv7s
z!{<TqffAj^91@ZUOMnOwf>;2f%qbNRU$-`PMuQl{te^#U8J+r0{54@oj=(kiGM|3+
z7UC7x>Us7OTC*<@_#y#fQAPwIq7=N`=jMjz`?ft=D6F)*)JU-P%Sh3_V%n@xfa&)u
z{VN|zFB>YO-XwO1KcXx8#Y9(wAcv=VnwMKw7A3rO{3%Q=|40|14^Y-_38G1Ja|nGD
zyyE`Ef#lAKy1Z-b7ItK-BmyK*0{vPMCP8_npPwgBcNeT>P4-LRf9}F7w5l#VK8!o%
z&bYI9*F;EV8g#WV45?hd`ZmIyGtg)h-m+m&!x%sg_CjXZQ0bPp%?q(>Q|!%i*|XRl
zFZ+lzlapbKP@B#?=)Z6nAvMs3o=H4$7l0xzn{ukB`3InG2nVb<m28zFrM5;qXy;N5
zP)`vdwev1~bXS2#E!jp=*hehHRg=<w8N626oIIe|gf`4qDc?kF;(*4SN@{2|MbQZ1
zS>;xnSO9+i{>md#!#wYWWEC)8L~x5VEA)ANn%zKMMn`1!3v)yuMTF$ohgAD50<?ej
zZ2(>w;*E^Um~Bu^ivUHXxYmdev@=}KYg5XaT5qv9=_5HCb(b1{hZ>hKmTUJ{#%54d
zH!>#uCbfCe^csaC1-lN&n!=75Z4*D>PNL6?#R|m!K>3skI(TzQ847g}!Dod!?SLL3
z)XyUUnI(}hpTSkk0^$QbC4qF@k@uAYsiz7liU<g$Kw;{th{k1NOJl!_{%jfmni-hq
zrOX;%cgTr>^d~vbXcDX1uzhiHU<aNP-ob5#F(?^vvDhM?M-Uj{S;Cv}cn8RPg|<z^
zCqd(tp!d2bU(e{9Hu}GGT5Zr~=|_eev9ciD2T1fu7o@cbWsqurs-(FGGDN-JL(mST
zDgrVFLQWKr|6+>!r&Rq87#?K?0{_xx4#+YReLs&n`9PuaO)W7t3%DMTDHu;BPy?hC
z=$fFI>Paa@2q`V%Dc~`{4C<b?PWM9EYPYs#ZkTt?hWYFja}6(KV8N6FFNoN-4Y3+H
zbTKs1utj))+z>`P;fjBo@ClS577jf^B+m;?FM-{KNI9Pzr%<rHAUm!k5!YSsP77Xu
z?zO_G9a{wk{ujo+&^~+$V<xR;C+fug1@N4(+u8SM4mf7GtC|o8&2wu`0((u~;>udg
zeh>AT&Jo&3ZU{nPpHML^`m?nC1C+i&hfG`|DMxX{17i&OM#F*6j2-xJcf%7VTZ2yc
z9Xv4&Mn*t_;AUK*z$TsOrq<Jw+`f)el@~$qIJG>|IBA@+!QO%(1M#O2=4hSAMM|z`
zgl{1Y1EJrK+M=ImmD|&L1y^_TyHZksnwK_<WH2CbTo{ZbO8X-T<!`LA?_+J5u_O?Y
zV}zkGO-&pl#1U9XRFR||)emDDDvF7`Nhri*PBgLX52_$xa4~IUM*6;TTe~?H-N$vE
zvT~9A85+aiB=L5L{eDrrocWeP`^p*}23di+)O~sk^%<`TNjDzI5H2aYfgZk!PV9#S
zZi$&hTR!OyZVl-w!t=)r93gzaOC7ced<;-83bL~D_u$yK;%zP`Bzi-4jI{m&?JR=|
z=>igWHtjegGa)gDVv|pdE&1J{_=G)eD0G)Hr@-=u^*u+X%-oZA`j}Zfl%Q2CfFvmC
zwG=W`+?gq>9OP{+5Qz$X>WQ1CVE}Jn0n4k0kk6N?O9yJKFZ&$ap$*bd%FV%nNJ_Up
zw#@&#*P}~VWP^x>eg!;aHVXytPe_m)gm}lriI94rZ4i*w%cZ%DM<DT_yakVt^vffy
z5f3a%NaeDW0$GP1(!QE%2ePZrOK$!Mh3k#WFnRGV;7zsp$7+ix6})&^N;F(J5RcM`
zCAYk%j*U{G-kUl`QryWs1>;pwLJDbWHIpbnOeR*?A5dz<VH9k^wchPa4NbZpnFzy+
zzq-;QH%EX+be52}nh2U`6#GHcFC#9x*3QvwJFXM6NJI=<RS-A{8uo_(RxKOc9uu~V
z#<a+ufLkhzFfzJ|*$^<in)O1aB+0h4O8xR|A{EtK?h0PDG-%7)LT*|)2zn<uOgTlm
zgcm`dc#EjyC<c_^a+0bn!77-8ETzYU`v~Vx>CTSP&#3Q^Oh}j#sR%N83ke8Q5^Qk&
zHs!BtL?A*@k|Km><;qJE#3tanMoJSdqJrZh7g6|GZf-9Rp;*9?>8tD&jK@9&7+P>v
z{;q?dt_S^QO}3*Ln*WYxBX}IN6KgU=AY#0r&{%I~kw3z%A2Hzko=q3ZkeKA9;IK9v
z1k>(OTF=nhmnZCRY4X3LivCeDF0TnJ3vY<aciBHulfNhMxEx_WLD@G6y%i)Rc<U7)
zm3;8x;|Luc@K9J&Oll}LQe=QQsVBqO0s{Bd141sDUZWg2uv&a8%}R5d`er_BhmQiF
ze?(g<HAbaLVY9e70o22w!$$yGKaWEbc{0V}wBQgF3Uw<!gU9%7tcATxfaEVPwjG9e
zdV=UmW*7fCRgs`&rwN=TK%ObjQN%gmxmR5$c)zY^g#6D`ql+;Royp$zCEU-qA-w@E
z9YU#aQs9<Qlue9>h$V}Ehj3iOP!dH@@YjLoImqa$+&w-<xL~Z$^c|8gNj!{`kuafN
zxD9=wivX;!PKoh&L0B#bCEh!{y2(3^-@*kyVz6Lful)mB7^y@^;H9yJ9BL-664;N?
ztUhztrKR3nMI@5Z0=b572z>U(NM=q7*;SchJVGfUUjB+we@);Y2oM#rpAq;u0V1Z+
zx(9ZZaD{uQ!skp<Y$9R6gz86z)Y`nE!(Es(YE?~_OSS5RF~O*H{T$c9NKiJ67An@r
z?juEhBR-sI5GrEf2(dq+21E+Xr3Qxh5@aO|g~Mc>z{d_N(DMkq#w$b3KPbQjgrCKP
zh%p%1Z6f?(&5m~)2AV~T)jBs`_n9Fhd^_T_GRuO*R~D-+QHuOSMm%es?hk>lQbvnf
zaou(>{#6DeFTpaXtyn0D@hs{~L}{IUd>uH;{-;Yc=2$NCjY>2`ywZq*APR*GXjKr9
z#g|M7Rniw%Ovo(pc?H+s78ny@2yq~zlg(0euDE8(@OXvjTYNHtpd_Ta0Z8u!y(Axn
R+@}vl@K>r!wGSo#e*k4HPr?8I

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cldm_lr_qsample_lr.cpython-310.pyc b/model/__pycache__/cldm_lr_qsample_lr.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..a4369cce036fbe473a039607f07cd60288e9b689
GIT binary patch
literal 10346
zcma)C>5nAmS+A?BtE<oH>ACjowfE|^J>w&>3Er%)^*SW$N#tF}agnH0(^b>cGj;U7
zRlP@3Z6}^hya7!hC_tnrWF|r&0|N1ZLjpzO8{!Y(!}~Qz(Mm}00Rb_@%kO!syQgQo
z33RKTdhg?TpZl$-Ts9^AmVeUqUVBNBeoUF^pN7nfcw$ACBv*1}UkYSP4ipOo$Sb~T
zsd%ft7UZnls7$wXQJ1&!qqia6g}}5-(MI!&LCGowWvd)itV&R|s-isS*Mb>qMx=GW
z9?V*^LBnbUbJkoiZ_Nh_)`F<Z`-{PnwItGpe<V0+9Sx3I$AaV5aZy(APXs5elOk>U
z&ES-E3hAO-@*fEvwH}qF*Ce;>R_;n}#aq}`t<%U?-5T;W&)htNlCvn8aqB3ldrQcl
zbCqjSYxX{2r={@x>rSuN4Le+YDZI1Vl6mp-%=MV(UhcM&md5k#D7<NhF4z5T=sAoR
z2Of*;AaeV@$LmQq@Z!Yl*<P^bxh@965cO+4C+Rx=b>@Wedc*>cwQ_vn`l#)z)byHn
zqwj^(iW_fu@ymYH-s0xv=vEjzK@US$UP=;b5QSM?;muVqc|CGHA2U?1I#hYNyT0C!
z(WIsFiXV0C<aS~^vE2<iT=#Bw<0R$_-7roZnkn(@bruCScC_B5bsoFibvjY#_^&zr
zIPN;(W%OX(wLZ;?`ZAgiyl}MH`=A_&O0l~(HXxSC+JwjHpNh<jc;a~gPqJiJvJ^Mx
zs;+id293$UzRqV~iNb_MzVEqjhNFG0UTN#-Gyav4Fz|d1Ph7#|9ch?Pq+NMW`psu#
zDN#}-l{PV;tL&&lBb9cw)Yy}Ug;d_np+-mA#5;dk`bv3NOpB?R7E*OvMQMK5NX<Q2
zmTnwJzd}~Chnly*3>yC1sw{=ZQ&N2K7FyB&Z9J*wN`0Aq4(+vFb5D9s8k$KFbCkAa
z%<|Zmm0<}p<zh>;l-PPAZI;nGPpvx&Az?=unqhUiJR56as(yN{H@=XX*pEEk&u)cQ
zeoguyw|b=|TSeyeJSVZE(7(gxfGV7j>UM)xmFJR(wKsS^jO-3`Tt>*`6;Q6{+wBb}
z3_U;QCQ)!ky<AJ~^gNz(#jDf}d;P>-Bf^Zi4leP)1*M?ty1thcbF~-sc)<(%pm2~E
zFGu}ks*~Dw60V}GRpwgk`RiQ1%?o3ei3M*Z-t8nyfEsv>wqauqB2uCcu5aCPSSRMn
zMvM*7Lb*yn-r)L(3FYnw68j`J_u=ZwMih7}VU&1lQM9$<I*GH=VNt)ga;+a;1$VYL
zyl|xr6s!>My!<O~tO$Ipw0$?QeP-XtxDUzY-W{F?^J02;iAeA)fF_%A4NpltC8Ur2
z)F_nXx>AxC<r7LpE-9uwIQkIkULK9r(pFb-avC>mJ9L11o16BC#gH!9_Km*dXEp3O
z%qg}u1)im{1%O37F$s1<s>q_`W6Vgk^ozd)FlNSrTu=&XK`Z3&*TtVw^v@Rz{8a&w
zcdPG9>#|#L&ASD#F2wF#$<i``YvtYtpIAEBxyH@Wh9RODiHQ$1L5uelr0>%r?i2A{
zXP*YVa{mE-qJFh~3<cxgc_b=$NP@%z^h%^{!R}zZSyoNwi235`XQ92e<@e+)&synh
zyo54^%80>yB~XXQARx4KZd*=|Z7Hmk<S0FyORH`U`IWRhtfpm=em<?>t;Ww|3|)*%
zd4;_u>ecwQNx6AXw)Ip7%%=*Vk*a`&#7uKZaa#cglu`pT<~njiLQNTLUP^P@a#p^j
zvLLCrc`-+h>eEFbPToQN;-r-lzw=q?J^3C45k$=&C#8E5)!maf%iHpn67NuLQcY^_
zOYh6;ifg!qyGAmTwcD1rfrt19qD|d3?@8Z-woG-v()Y0vU{c4-vuPf%ks5$=X#sFP
zEdnm2CHA9aadRnD0FR*LXj%q5mR10dr&Yic={&6rsgtSdmK)OWc+%WGl{}KF>G5p^
z=|}e@<c?Erk1%u^WoMGJ$vFr@CDj%r^sA7T-oc7G@*a(Kp4!}#cU!5tH!iy%@cUS*
zcI4rSR2`mltHWmUIN=z=_MaeZPYoYQW(YIssU=_urQaAnnmm!7#O(PLD{4rXvzXT0
znVmdFIh~aaPo~XOzbXBe%)Ta0UqR|3dX<kz!!x0dR0XMWS{a^67f`pF&H`S|T7)u{
zC(CqZDE(5pNV5;mhKgGso=ukNyimTLK6+20nyqY})4MA)N}ME5Vw_@HPS2!ucXlTi
z7L%uFWaJy^TzWP=nik#04pf;8D%w21yONwuPk&1q<uB~62t1wK1LmlOCT%{Jo_tPH
zB>S=S1ZF{vj5%*CrKd8=bE}KemWnm~I6XN$HyLvVB|n;Ai%y3yHRsOn7-2qn8fzlO
zc+Oqe(YD20={aC-G6Tjrw3f8#E`CZHHV>`|E2&Jd{I3}fAA)6t$pXs{vKAvOpPj5P
zTiw!*j#f&%@PA<O$QX+s@4rnn`OFxHNPptsdR+-3?a1ojjVn>)pE}iiJL)$fyE{?R
zOrmBtZ2NuJYbG0B6Kd8A$^2?Y>&-h+pEcVoisL7ovE|l;7IV8~O+oc6FE0<?e(0F}
zI2$j{%*JN2Z#UA+|F5Bi*oE#lufm3dG=_!8UW6bWJo9Q7Yh_I+X?w$S+_;Gm$C-&8
z(A;oh+TNPijJ;&=ZjVJbyVxc*WL`Wn1Dako7TP)?P?0pxcl>D0@%=lrRbgkgXnbnW
z^s$naqz_}ZdA{edCw%WFjMwHZj3LkmRqHi<FYF+X4kMGq4*Gu5CM7cdQJQqZES|Uy
z0Lqh6X-NWq0dD1B49R4=R}QYWC9_YU^#=-XP{rg_Mw?-zZ!*0~DwP`-eJ5CRol6OF
zF=?*4G&r-b*p_X(8+H@hUcLyM$noP#NMt<*&!HJi=YAU|7ZJn6R&O$|?R9T@uE6Bs
zC474D{Bra9Mi*z&BPtQ6aP`$|*P1=njaWCi14IGS`~7ehkuU7?xH%X+-s~=W%Yb3C
z?ZjT(yoHm_4sE|nX2Jbx_bbgy%_o~M?we0vxX^9FNTPoG?WTr)e7_jS<@?PzAxK>^
z0$FC4fS>zxDngNeAoG0B$^0LUmtgWuD2r=glbBaP<S6snTfGQ`!yA;3fkwcM?s{T}
z@RMK`v8x!97sdn{^O<pSf5Zjq8+x~<jx-{t4Ng|h^?Jz$FM=A$va~U`HRE<;XN@L;
z2}*rJ%cK!9hfYj*y_I91rOlADAxhZmv<}UML)KEVV$pysn3j5|UF(xP7xzKqWYlFu
zVl`lib{wo@eVG`65@5-fX#5jmXet%saI+gY9W1Xq@OZHsPBzNOIznxnxo~ZdtAsig
zJi}fQRbO=czV|9)5#uVjkgI{y<2i_r4(xc)5L(YY^n`idWl<03#7onB$IDpfP2%{>
zyAy|@0HOJHKXQ_%o*gBgd77Q1ovaf0JT0X@KBn02_p;trqX$14ZZwQ}R~Y1W*M%h?
zhTPnDe;N6FT)Wo2`X)P0NfS;kd~1?pjDi|0N1#bnmB88ZY~g^@I=MMX##Y&0hac2u
z86Meb>Qp2!L*Ntv@+et>K#9O30K7Ks(AeawoX-TqThxTiW^oEuc>~DYKN&(la|o;v
zXcKS=ObEy_h{o1?Pd&SjCAL9Lh_b9Ia4a|z=prIxU!YP!*&{VzPpJWGW-4O~8fR}%
zhkQ8l-E^O*?T)o@nAo!zLd;&J=8NOn(XRJ7VricvK*7^<3!1em`+j1^R8#JuKno5Q
z&ZdF`&Iv)!H4@}FZHCZ^bH`(`Wq=!n4zh})b+~rya6J<jkRb)zZu?Fg+cvw5<uc*w
zyouB^c;c&21m+11PL~09i~a^u8XhXC$-1J;^f%?YTvH6>4Y`45UOtU7MK&MkxvYhb
z7UQ;J1mw{EXPT}k4Ow}hJkTFB@-&`?e+{h_MZpLP+GTI;r`iLfXkuJbtN|m!@iS#j
zS;UHzy#8U<PW1V2MMn*KYCqF{S}WJ(3jTGJh|!Arf9$u-Qu==>MMKseDF3cK(CV}w
z13k2_nwW#uOS2vBqkX7Ue%`D~Zo0Mhq5dzW<8^s34#L1|n<%$7DZcTK(!?)B?J)_X
zL`q}?UJ!Ok<y5B7i>s0qzfDmcZB^(3VTrKMpf#kqVBkE`85a|GK%A21<;FM$aoD#F
zP+P!R3mtq+9xT0*#R6!V2L(WzM8o(;JWVsv9PxJn5;z)Is#qo>L^>ipN=i|oe4=h@
z*nk4}DT~O+h%FgYaic<#M^D6*HVfMd+=^K^8zv$<6hX?T8e&L=)Yv7Hbd)EP`^Fi}
zFe*VTDQh9^ma;Z`aBSdy8P_C6j(e*t;xR(F2)W0Opp;RNLFfk|U%5Ky_JrtTAyu3r
zKw^tU0IgCcsai;LjpP%%Ky{B1SSIi!fu{ht1|Eyq)08J(vI-N)c^LZ?awrk@DSk(>
z4kFZ&G7olZ$Q7`MrVQ#+Co&Fe%w^|(5xqu!{uN<>?t?4%Wp4lKb%a2!wDRni(3*XU
zz%LUZR%dklBJjb>BXQ97d><l#LOdC`O(`K9o=1vy6_cfBNJ?Rv(0@_7^3T${##=R&
z(Fi2I5B!Lh=okMXz&LK=%AV?JUT$4kLb$=xT-DQE4NuO?y9ku8mqfHA?;3YYkaZL_
z0ir!~x40*{M@3!9E#EC{%T|dTVAPgJwQx}H!fmuFqx>;<)@|GcTUpgn3EVQ#qUJ8R
zi|9GyF1bhWu8SzlEZFb(Bnor&^6R_-1xW!k8%8;73*=yaWOfmiZh3cjA$Dzw?s+bI
z0^8(eAMtJSRBRE<)0qc#8Xkzr8fZsP6;J#Q0EK8a<y7xVY5ou7Z-K%g9<c6IvQ>)D
z+8Xhool7-9Jw;sB&bx5%T?MYWWE)9g7ZFrfO-j3Ea9d@wx<^3`ZIZ81zJs8~9*sGZ
z)X{2&0vW<h%dP$di@@*SUbtVzi|4(Nj19(3MA1lL!%S$+vg@eJsEhmuVdV(qh~OXl
z8r8l>fbhV!sYzvmKeAC{)}@+F0tifjpV|@PgNEyQJxW<K>s=NnBV=czzE6$sQR5QE
za_!N|SOY~ZBV*mKQk#cOuTju5V`4$<g7zos5ceTcOj4w7nj@vBgFp2`jiQquRkVp{
z2Pr&Uw9_u=5u*KZBp|ya66p=l_9B7}LnVQP+?IEhJ!z;4Ns4d^r9f-ystDx?ZibZK
zL4P(20HqJ?^HOHmuRG)qp`PSDqe-k<+xErPg&lZKcoX+2#<*mPziFd<0U>7ug9&%S
zg&rgD1=>6jVI5m0y*ha0m5kPDvp=WPYMVAY{uPlRRUyduJ4lSk7$mj{VoS9j%5O*@
z=^T+XM87`FLtQ}YA<l!SpNPygQ{086>enF*P-furLE6j#ZAPN+=20i_DOA3xCB|j}
zR}?Z`0Z<7P1E~qRsxVU>5gJ6MQWH-Bj{(+DceHi7qtaFf^|j_k^H#IneBx4b4KHNi
z=avI6h}az)qBn5plB(FYMdX2e8}@DNhb#UK!Yq)77(8?hF#s=gyadJ_BJ{jEO|K9J
zf(W^iL|k{hn_ajXy4MY(UThT@q=N7$goffX7&Ga1`cXd~Er92QbHIL^=76_{d$XcA
za-Lgr5*UT@7FX6{_AS(BBrP<S+!iFkeus)_(d-bV{~pR-rNbtUd4x<K960?+Bbgd3
zgT~asPxf};q{;rElO`V}{#^hv3KB#~<2=Pk>AW{#8YCL7Sa6j@f*t3UM;d31bC%ed
z6fs|jGTeZoH3a@nj!y9!<v<mqp41vWMZ4UNHY`9P<nKy}0p&k3A`pKrjK>nCQ)9J^
zRYr7}8Bl}_IHZK7DUDA2BLox}N>rr>>JNlWR8$ecU&A8=MF-3MiUCpvU(<FP(zlfx
zAhLa?!xf^k{3QD!dcc__p|@WrbS*b8W{zyoyR=5fLe}A;G`KKz2pRVY`8ORU5&kc_
zwM_OlI<emulZa-#>W(il=@P{Ar|cjhjlW49evQDd17HLaCRYA70tVgq4wn;>#i2Vz
zU4KLk3@W7iP+S-F;*8Sv4G1A%ic_n7lQu*aSOrv8Q%Yp#6kx2Ak^a%VCaj+Mhi~*_
z7V|``)-WE4RZ@5<1S?(JwUkwkJ34ET=v63OPh56Q!h~b<SYADh{}a@u5B)Y0hz_pt
z#%XBn&iG7f%Ahs1)L-1|K3Pv~qL=ZnfCSmqLNojw5+ocU_i+^?gdoTqbfgV)X)a?J
z2ta6d!7wBY^GIvN4oeWAZZ1nH5QgX>?W(D^C%gKB<mQjkMSyV;HZk4>ys0*Spw>jG
zPK<dG*EzVzAcmz8VMpz#Q=?R<_oj}K6t}vgphpcQq^V|BGXVqSjHC=nfhlIAfDUdK
zZ)EzZ>Uv~9Ozt0B%6$aZCH5C+b)U`?;#U)a8ja#h_<$fIG`exm(T4!I)y$$FG0a<{
zb4Fr}eHXy0XM@{QvX{}A78x6Gg@vU?mYJCi0qd;ODrDO82h^@KD$gcTQO)InU|CCp
z8oeRJsFj1<cap=@W778lBG43X5v5QNN>@mSmgFPyyzr9^xe0<Nv)TMe(*|=>RJ7mw
z0I`sKC4v#e@);x`Ut!3CT;xsp%Nmi1(4M3n;e5IB5g3XD*EL8{6@H|GLnA*@xMFT@
zCl4uEz`^OO>;;U+{*=Ji0j&Hj2LWUc3eT$cq8LWOP0vQ4I_M=<HAMJhx}eZr?_}{s
z!mb}NVEwjDH`@@O^B4}twc#LSH=wkhA$BBP*k96|e?=9e!(?1u6L=Ob6qj!?Vhr}z
z1P)3S_BSYdhY(CBYv4wBp0t&aBZpABf5-=9H^svyB7|f-P$vy#SYtrpuDVC)CEIL2
zM_#TLUr4jk+@`*n&)VVp1gIp@mP(EN(xk9yZWe)jSbX>*K^x_9Y$8wgI{X|Vu0ZjM
z&){rsV=e591V|=xv*$2G@DoIKGUJ(iB}T%Rogr|J0Etwdqv&?Pb1%70@K#ID2>S1-
zMi*lsGL*gTr}4Q!4-y^V=|hMW9u0gLitvf?5G`f#Hwn{Z)9WIrq66FWuz?W}k_RWJ
z$QKOuvAzcgnIs^_$_SazPyPja840ltgnUWwFk(Yq5M~X6nYRzGv3k?-yZGt=v3#(z
z*ZU4FkMty*<|9+fIn+$dgX@YvN6S{@uvbj|xr%5gqeXHJpEmgH`$%RU4Vhh;(mYBj
zA#VPJQh!F^ZwU}hvVS1(j|9kZn_9UryPl=8cLDI-)x?`tOhx%fQI(%5n({vgA|a5a
ze5e#E|7kRJ9XPJ)2u|rLJP*7LvsM%%w9a!Kj0Tm&Xz60r>^G1i_Y+_Iv<W@2@T1uG
zr~wfP^QeI#K3Q1~L*Z{(r|?zAGOPlGgX86ioiHxIHHe>um55~+nRz0nV$Dr=n}okb
z<kvbsUH7peBYfoIbTd1IL|Ya?E>ViyOI87}PLGDb=QI0@T6W!DF!lKeBj3U@sHs>d
zNrNmDP1I_gJGcxS?&z~C8geR^c~vDEB3^DsK@f$)fwU@!65}f@#82sKF(yQs_y&aQ
zuM32Uz=k-FsiPLMijEq$T^SlL5z&ipTWS*7*9H^;y%&{|ehlvh9{Sgm`Vr~fhX230
C(_}0F

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cond_fn.cpython-310.pyc b/model/__pycache__/cond_fn.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..284b18e542485bf54c21722542ee1c8c3c8bf073
GIT binary patch
literal 2949
zcmai0-ESN<5VyS_xA&2>O+(cJ1-BH)NmVYbK(rA>g?<DcNUJ7KhtOto_9dI_?QU&f
zL!;v<mC74`076P0`!DS)PyG{GpfKaTyWF)@IPzrd9ozHqo0-dIW@-e+HS()m0PYW5
zOb!<oH(~0#Km-vqAwAlrJ=SIv*DP_`4*Yi#x97EeN>+$)h4+LA?|`){u<}I(Rux$}
zaN2=jYor-Ghr39V6+!k$sw9g<5!}m7oTX8++N4FT+fO^?vV)S7qXCPXF!e7$BxzGY
z+DuUC2qs<Oz;vZ2JoxVm(rt)}2%flYUsPpfClEf;t}AMy4!c#+u)8ys&f*zaL;f7;
z2I7P`2{Lurn35`Q@Wm-H_ryJ*?HO?zESfKZm3}Ouv?B{occMfV6%%R`DO32!WI62T
z5q$QQ%w+`IN}9e(J3iPk8pt!4w_xf8AOkueAsH~Bbl?C5Q-L}I7pOb%fO>+00xWt3
z2WxydzDr{hM@hUdc@i0!8XoteEy=frN-mBi$IRDiOD~UCc#pM%FLWV5rn$^lR3u`&
zYPc_Cni(m0r1^&4;Bm^czT$b5OVwQD>)ZIW+LDSJV7I0YU*=N};bSP&Nisc@@5X=y
zuB7fKhVR8m!Z#&1nH9_WChr@bb?vGfD-AjsG_RKlnH<|@?=7Ri7vj`3FY_`M<!ad~
zY2u#LM&@|C4H(Z9Fj;==EWQec!ef!KFY2h5Cvm#<rV+WyI#TQM9?iQd>v41A%F@!3
z)&0g2C;)bD456{X^>)@z1ll(CRg{+BvwUOP5(^)`ppAlHpnFARB3|xxLFZ1^1Gk`F
z;6`|Sm_vL)ZEd1WB@mj^mzH20`Qw3w=5eru)E$gCUcXU_ZgA_c#o9Q6_aN&32{ImQ
zz8PtNNS5*+R&L+_abeq-Tram;Td~>hZ!Ut$t-H}<xhd1uicF=7O#DdRj=SByhH7wM
z#Tj5}{L=b%l<tCMpbfY&R(vm0yBEfvSDU`V3{n*&0T9SNgrHGX+`&Tk%r9ImlkRJR
z69Pn?0J)bpTiaPrwh-c*S+?8i0JK^FZH+~$wRU~!vujtce_n3iy0!M8Ww8hARD|7h
zF@IbHVHiUih9PRAv2OWid|Lk1=>Dlu!B`y0DU|78{*MOo<9dMVm-Rr9HS&x?rL-rS
zOwGb4%Uj(`auW86Iu_Tkbi&85q473&e181+gg{X(1;d|~ib}4eK!%zQwt&$<-h*i?
z)HlEk$PO7eyF`76^=XG%jvcZia_Ah9pV@$pd8k}FP`3^}u!0@2d9VU4QFCw&gbv+g
zLs$4hny9l~cu)`k6c|(RCWL)2QljaW=FX$|J4ntT!JthT4$g-voP;4(QH`#frwyk;
z=cvo3Twh9c9FQ0d<S9&RBN{x`(mv31URPlHZFLTgT!m?!J^kF!hSlaQ!Xu4(G7O7a
z7y@|u3G$6FeAthY;TO-=-ok;}RXC9N>J(gCcqUWe9e-V>I#cR2>?zb)y^91RY*m<W
ziB%8*p)hvnPkjkUKp2i!%wc~sm;J>&_9ye%OIE49ICK7z!)h0)J<$Q-sewqvq1!<I
zzY9MlU1~dVqYeG@FgPN%A6Er-+p6u#)vhn<qG5Y#?J03dyAEw_rl^0vc4q|Ew<fKs
zgBoviQY-`L7kw6pAqV7;BJu#I&^eDh{P{9MAA|1PhwQcvRx=>~j22bEP!ee!hR49N
z0@v|U^r9csXz<Ve{Dh9j**pg56u1yiI4!uo1!lkw;aAlWrP$`_Sx{22BS;$I0L2t?
z4E&}UIEUU5`HlR}4nWfQJK#ky1ky-&=SUq4Y5J?{&oTH-=Q*NReG23`+P`114dDh?
zxZkW6vr0bf$Ix6uY|CNc<k?=~_GFX-%KE*+%M~Kdjg9Q73vi|gqQ1#cxNwoak%NVc
zPfQ@rf(;Yl4or=yU31Plb8H@J&y=@DS=~_YfslG1$$2CnAo&mpM}o*Nb*alV+Jt05
zMbt&y#&?RDVYE6D(|$q~tb5Ap^AT`gz|<Vb+rdBk!f$*+6-tA%eDgSmrds9|x^w6e
qn}myLsxX=by=xQ29-4T2oOng>Wtrn&VGL|RI!OZBptH{Wm45+HX|-zr

literal 0
HcmV?d00001

diff --git a/model/__pycache__/cond_fn.cpython-37.pyc b/model/__pycache__/cond_fn.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..0609d46e879b7e05701fed9aac191383708bbf87
GIT binary patch
literal 2913
zcmZ`*OK;mo5a#kFS(e>Aj5O(UgSLPL1eViFQKJdcyxah(6C^$qDhis`u4LL2$?PtZ
zCQwffkRQ<$u#Y|W-|*T~{z4B$zgg;GC(Q~wTrOvKUf;~<=F(Ej!T9N~^Kuodzv*Q<
zd@OEY>boGq5iE5^tj9)f&t>%Pre4p((@XtP&<mNf=?Gs0FC7say1fS8LeaomLpBb*
zUQ@V_olf%&?s7VA)y$ttmFBUinhy$-<XN0<by(FJlv%%i?Gxv8G_bgVsTGKHdQ3Py
zS9;RNFOWX8V6Q{r3h$-g3uR-!NmKZ*T@i>7dkfP&@V97v33|QuMN=%`>!NJWzSFtZ
zYfmhS_Dlbe^_IjE^j@x-n`I*6tS>uZ73hAP%Bo=^ZDM7rkQRA?{UYwmDpazNF}9U7
zrONstbf!U17xNaTUIiJmu@gCC7nF@XP*>27H}*mOaR3?!2JvqEFAlc&WZcUVlf-HA
zLh>{=GBZ3G#Y4%56BXBHoU_IEYN$76Z*-5ffv*k_1k>5zn<^FwT{YPkGRut=Jl1?i
z@9-q!d8v327gBZB`1T%&Rzs<{!E$)&@D+Ya5D6ouewynE{~$rUxstj}4S$xTDc_ac
z<d!WjO;H-25A53@Q5rfKGH;X%na*vq_tp{Mt4U@$S9l$V`fbB1X_ArDMiz9t4HnN8
z7I`su7T<!Q5EdINQOBbqO|s#;MieUVORejBG#{vZ#Lcy<>+9=Q_iO7=0Cw(7V6nsX
zUS6hxY@7Ki&g$=Zv2)rI3mv|yjY2TUy&^U-uXhK~xu1{V7SaoDgz!n8@KrV3C7Ws>
zbWUGd199q)M;4azU<s-F6mhzKr)J&Z)?sU{Sp**->i-Keo@l-sYd|E=_|KcSAO5ns
zXH20ty4_)7_R8HgsNB69KbN~Q>u$<Ss@Nn?<?UoJC^a(vp-OUKX(nlVFU}5N8MJ{L
z6UCq9>frM1^HwKRltGGWfog&v?+AiMEzk}XGUTSOWjc6^anA6kqn5$<&#mrWK9XI6
z_->vbbjjJ>p~}mm`?$>R!_ECYXx#_mx}@;Udyl#ne%Q)HG|1M9=T$R`5@cc&kutgs
z;#>&X^!MWIv|_AH`8uv5F0}~Kr%FJBp0lbzM$}b7ILLd3nrM&U<PtvB$H@X~hWY@j
zYLUulRJ-CUY-oEkd$BT;e}bXoR|_?3^y_-nD3lbmV5U}zYCa8m2AH-~QT-n~`_9-q
zaMX1wRQt@X+#~ncIr5I3-`p{qtx>}EQNWG@7~&qgD=>tNnHm(@*x4vnp^zr-?^OY6
z1MorNg;#-qXR#6;zqa`tcGY<zD@5M19zj4cPNRqls<z#Xt$1zLX6M{fA*i`Jo|@lh
z<KW>O4SF(6YcknCQP>K^G3)>N+0$yM4{_jY%-Z><Bu~uQq9zhWSv-<aRJEcAcq&s`
zx1;Du8K;vkfvw1uhl*BtC<p2cuB`%-D>y3LmYL3#I*UC;8mo&$DB@Oy6W*~35<nCM
zl3c6Fe@)+I)8F!i74Nf_%PdD9A7_JdjE24s^1goi${8@*bKBH&kDA90x+ry8JJEf;
zy8EJwwq5~Jy+uH}Q7!)X_|6oX@8tA6Rck<!-Zbb*N|DW)Y&OoJbHoT@03>?laX^1c
z>YQ=WiAyvp>pF<Ap`#n-wPP*^|2h_cp)}Syisor$&wfkNY)&*oB-9DT=Q;geK+!px
z7EnK6F?NseKjoM)YWeIU)EDkC@<w=|ltrHLg+KO=f@9}*=MVP~oF?3dXJH~TNd%Xi
zMKUqm+I~YJ?|5$rvFb~ZH{_WY4ci>9aYg%`g=$&JCuM?08&QjnDzC_&RsKlE8L(B3
zs-RGWGe0qMq&~-)su`CiC+5m0`c4v7K8ZMiGYc_-{~b(CrO@$O-bMG^+YXzibz5Bm
zkNSwn$3#9M@+pzeh!EauT^3~)?;=G=5yfen<Wx(OaP=kS{ah8)$}fn?YIGT^Z!k4o
zcM|mHf{-ct3ipKOd4XDA=MOn>;scv(Ygwi!eigfG6T}{xOPfQbYJOMe@^$jMb-@|d
JWNmNx>VKJKsT=?R

literal 0
HcmV?d00001

diff --git a/model/__pycache__/mixins.cpython-310.pyc b/model/__pycache__/mixins.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..a0f5286a4ad82a6b6b63b775f55ee032acd01176
GIT binary patch
literal 640
zcmZut&2AGh5cc10wxN}v;(~aBT(StYQYux@sz?X{IS@{Jv9{~Uc3bZ{wKs+I+TM8%
z-h-#`l@qVPiJ3%&s$!)1JU{j~-`MT!3>n&leS;r({-or-NlcE=!x+IZ!!2uh!CO&?
zRu+;|O;}Z^x6DYRZkbUxqDYLGu~G61->{L4nY#jSU1_2`?p9H~X{vr&QS-iSVRD2X
zCJ4X^Zdf5e8V<?`NFW6@7te6aIv$*7l+9D4OmeGkc(G+t96E@P-?e28A6#97|JZzO
zx|p1neRUSoi!Y_GLrgvNOW#!j=iC|X$<g1V2Zdl5lJV(x(ys#8xlgg`2Lz;a4ZmOK
zXRd|3b3L3ocac}lK)wJUTvytBcJN|v^73GRQ$9JFeac(BVDpw_h4Eq)ht}1)A;E#D
z18Gv?`uW2h#{ce3N5hzD-IXn99fw*Y=h9NVt@Y)yw4076Do=ZP+9%4;`<V33(@%Vc
zE;#SEsrz@EP4tLr2?BYbr86ePn6GQ2G^YJ((RB3#djI6^+5h-rcDNDwivD;_`jlmS
GSN;LwosGx<

literal 0
HcmV?d00001

diff --git a/model/__pycache__/mixins.cpython-37.pyc b/model/__pycache__/mixins.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..a15f3356a67bb3dd74cc39e53d204988c650508d
GIT binary patch
literal 625
zcmZuty>8S%5Z*sK&XGt&q=|TfYi=7#gu*EjiU?OEDUh3$<DK)F*xqB;1}<&#7`y;A
zufS7mOT{ZtG2_#L6fx3#{&vUv&3rRIJe)9$_urqv5B&b5!!9I}Q_L_!u*~v?wLIso
z$VDr2$>~lsDpzlql~z44tL{afSaHRs$uE4vrZQ&k6ZpmzHp<uCCaSZ#>=z{!cab7F
z#SAYHfaTn>Tz~`x38bKw;)IQC??Xo1@~_egD<9N7&-bmOX7TBHTU2o2stWwO`b*tm
z=j5j7%T-KoKNr3VG4;?7zAFWGp&)l0^DSmj2$rD~UwosfRe)ycX)?b@u)l-fbG>pc
z=+5<U<J_&bMPKO3yJ4-bhVG*7s&WOLE}aE^Rxg+H^Gn_0q0ueP6=v&AoHVX7HBB5y
zJ&YrXyXTKh%>Mnim`-A5OjopEOq>{lnnOeUz?hFi(QGT8q&!*kWGTwf`<V33(;}{+
z3(orks{S>!Wl!iXL7*JBkqN=an`s);ezQhnkHr1a?$Q6$V)k-7;wxJ6D0$2?`3H8h
BkGTK<

literal 0
HcmV?d00001

diff --git a/model/__pycache__/p_sampler.cpython-310.pyc b/model/__pycache__/p_sampler.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..1cf61b254e0863a6d9c6971626ca12de6e8a6d02
GIT binary patch
literal 16149
zcmb_@d5k32d0$;!U0r=lPfzd6?Ce#$T$0<%<?QTAJhY-IN?e(uM2=;V7Oj?;VsF2i
z*`DodUe)Z*j$b1*he8k=Ok~J%4oUEgkHl*uP8`7SeE_!MAa>3_5~F|sK^!1a{z(wZ
zQA~;4{JyXHp4laBqzC==t9N|w)qCIlzUk!iu7=M`KlNPj>$f!R&*@?OW8&c%e8V>Y
zgeLT+*3zqb%cvSHvud`is?|zWQ#!R7O}lFIZ@TK>*KB55nQDgHtft$_R<m4AHFK?e
zHP2<cS!m5vXIjN-5zlGiys1@ZMMk*zEj_1+tjN8oiQKMHofF1OS|$H2tV64$V&@0C
zQKQqYHRIIFTis?5TUQ(PD7Ic`gi-7~TWdCJ*P6j<MUSm$Q?&3~YSnH8{<R>g`C<Kf
zAhwzTYMtk{8lu*&2Nffpzg4>#G=s>OLA}!sBe_+lIpR|_QNwIE1HXY<>g=~WvQ=v~
zc52k2ZqTCop8|f<_&$qo_=2u!k*4%W-!<eXl)h&OeP5SPMkeYkq3`O#__A?S+tqtU
zloF=Wc8#mrtMffGvhZvvO_{squC9z5hP)Wr<8_9nYcFY=N29bzDedc;u)m3(Z|KP`
zVlGEy_VufTTUT0SDa^K{^?)-e&ucx*(6iUIUOLK9D`x)8myDj-bCjh_mAawHnlRhC
zuokplMrD+P8j1N_mBD;j<$PWHrbTlqN0>Ksd0Uu0x9zBmFc&qdThxweJy+y<S<I4)
zQp(|3W_o#*+jEs2WmR@pU(oh-T~k@*KB4upn}1uSRSu=JvQ-AZ_U7j!Tcvk(I{(@a
zTAJp1_&h6vS`>ISuZbh~I%{6kXa!*ubVD!dcw1rMt;tTytF^t3Y^*mxDxPSpt!;%k
zF|R8-^&ku{Fs5}JZQF~k2VQ%tbuEyS$ZiGRty()Gq!B16!5VCCfy8b<8a8$U;dMKb
z=pbZlv=X2iPTHSpnH#^qBfVg|)&j*|@EY7x1`mhchgMd`)?aH2#t{ZgfL^_`)s8~1
zCIjz_msT!ZdhEjGl~)-bW=dAR){xkROOt5OA5`2BSRz(`EkOHdmJ5R^I+zUu86QnC
zHRQ4JA%m5z@qW^D2S;39N#6DOWFr!EUIXiA)a`Zr)UaVF+-gMEy)!~IT4#7UgLm1+
zig_G5@w)gZX<M&eef~wSTkGPKfDxxGF5rlT*J(Fz17P+2J)`-W&CabrB&)g7t;t%8
zPS_t3%9W`@AJIobM~!4XgQMh?5a}KK^beV~Su}29$GxrY#Jv4w`N>gU@dCU#-eE+&
z3F(CwKscj~f*6WwjrL#{4$jLp^-il*do&E_4I}K_5LKfs8g+;$>@@8iy;MR=_jwbK
zwzj|8W$)vBYTcmT0Q=NyA*hx(i;ryejCTv~5O}NiL&xVn;;RV;O%??Lr4<g2_<qPn
zr{CXP8AMyMeT8-`U?DJlgafPx_^~e}5<FFlXmIsg5Z9U{urpt;wcDM@t0z(hJX?=I
zVLK;y7sG^dAzK>lC|IW*V?L|X*ujUi0)!QWG&DgOU|mz%t9DN(jj*f#gs};Yc1ME}
zGJys9$5>v8^=+9)R|fBV^5XSQE4YYt1lKy98y7__s$E={ovrT0m$uq3VAr6?+865`
z5nQ|q+4hO&KX<W(hvvnuAF_~<%iY_t`xJ>tP)Yac^RyOPSXtY7Anewl%7)yq{8Y13
zuQkJ`Q5f`pN58Ohw7(U7!X7Xf<lIiKzZ+P8$NO(U%b!Xvb{pY9SH>pxCU%+ULt=PI
zDl#^kK|4;#T6;Y}k06TO&)1q;!6&8c$T$^tn~f;W1R$J%h&r}w-EPnp7y(%*Dup<Y
zrqSA~g-Hgh=0y-TWFQiTi}gm;AzAH%y5m%^9m!gp;U)PEu^ne1Dtso*YJLJyY|*G{
zc1_mm=s>bGPS>|2jjCozAGDKkvF?u!PR`OB&P?quy>|EvfIH{vWxarJSvT$)hO1k5
z%Vti`=@$NqsC9G$?bcn#wDqDvwK=_{yQrtHJO?Co6JYQmGIQ{K7T@qqfSqEb?LqDD
z>l=pBH%y_wr7QiYwrA~YJ@XZ<J+H%j&|#cVuf0@cZ>0BOnjF^q3tHqTb0f2><JZ{L
zVFY}iG6l>Qm?={&;fgY$*o}Qs)@JRi;0NaCPtuOpBkv}iaoE?)-la#clCbLDh8B^v
z+uj<>zH7Cpew}nMi?1M@kfZ3BNbRYy6P<=6O@bi`+9B|c@Da~KmqQ>Xv*LjXq8?2k
zxh6X`0g)H^;dSU6Le;EE=5=}|gr9-1Zvj2&{zM`WR$l-8q}XVb$94pp78nSQScg6Y
z;^7PQS2PlHqqpHzd0*t&e4^3|-lYrP^78VleX-2am6v+G7A9-$@8ummwncEWL00eb
zY9$+|qK>Rzk6n_MKA%Wzpo*kLoLXykY7rPPjBa;>*l2g-G*^A-vRvd2ypq@sw!6^8
z6)QHeuW@Gh4)M&?R>r0b!f9?P`~5?JksWpX<UOwvJr@CtjAz*TLq^GP^wM3|kWZqr
zj_<^02~W@98^#cEfmYRpRyBlPH8V!W6i*3Dq=X$<B8{&jv^7}B@HF7Rh-_e^CbwpZ
zyePbBRnucNGopwZCzBK3CuW1phAT=8=YlN#9@T_Hzra=&M}RHY?=?Tx*9r1TU-ViK
zi@;ZSpI%4B5_-*uaV*Ce<6ySq;skmW#mTXGP6e}pMbB$sPvHe6aT==}s~fI1C(eim
zFn+E-esW~;hG!Y`gF*Sen2O>d@h~tQ2{L>0``Y{BJA73SVx8sT>Wbo=sJsc^qPp<^
z73+EN2xDDLux4_Z3-~@No)*iXfQ$I9;Cl(DWn(Ru`TaixYWXm}j}6b8-e9oa$BCAB
z#}~yD;>kBtSo6^X*a_=L{@R%5#TD^U#=InQ5Ud}I^DmL^7B3}Ae&-ii+&oWi1>{l0
zUMWjcRza|(sUI0x1J9Qnpl%1&GlZe{{MD<UeR0`)`8sU1db1XWWLlH{TyF=0F_T~(
z;&>KDsq8dgxcXuuow&#KDC&k+E?%t3?Z(aJj$FUk6>AqSUs}1myz-I9pSZk?NA4Kg
zSV&Nladxc%Gv5#4N5hF{vn6&%xMOFq*4VjOlRgU{I}wC+(szRomC2%E-FFfN1&!td
zZqTox3_ug^tmaz+w%8qgwUUxgp`Y{s;?lsnC%KpCa``m1ehdJ@%nF0%TGD@5JuSN&
zA401U`F;&!LQ=%G?dUd?_0IgnK`jqD#+l)4s}tdaECvrW`P&BG3&x^GZvxG%ZI~Ny
z-vLw5=T^@Uc)LZ?!qDE($fmH~GI|ztGI}R2{jSkVMXpFi*(j$#n!7q&cIa>&#%RwL
z23-HV0-M40pM&e~z|D75>ZT^2Q}$j#zyVk3W7=Zkpl5pN%@0R2%0@s!zhQ)*{j~Om
zj=76)<n4`Fk;d#coOaDGQSE3>Jk3!nTFO|njWxTdODi{7GwP2-G#|a=5eVW81x5C$
z|2)-mEd(^x3YY5_P&dD^sPr{j$=8yV>>Z_^`*iM07<V+mN2^E8G1MHNs5vp}b1dm|
za=b=hojmfC!cNhAi!hr97>?mwe2;MAODpsQL`L$PtaKk%5c-^^-QktZ@9R_J&kXQ#
z5BEU(1Hu8tW_NytS%;jx=Rs%i#h~s-49T8t5Csy>TD{(p>`uX|d2L8zuO*yH2!@zG
zF4w~=!zXGQcK6PaLvqDq@xo$*!Ya$|u=nTK?V1|C3io^11MW;NVRY~my#3hiKf#Fa
zI|~2nW9(|d|FLS@jd17e&|+Y}o+Q(__zGb7xfz)QgTQ`C2@(ja-@1#BB<HGfRubPN
zy!#0%F^_+WO3xEm0U+DG1<@zigqHJEmFh~oZEV9xg!3F*<on1IG=hXl-^5`{v2XW0
z{^2-9VZ<;_VIZua$I!1@s5QH=v;6)9)w#)Mzbk{A2P)fS^IWD?Sm7o%X^x;jIWj)M
zmDpUr)o5>p)6HhHvtB(8;NP>$QO1C%W}`bj`q-$cb!ob>1Qh|YYILLz5!|52XMA#0
zM{^G9{d%(j4^717QF|+>wc&pSYnKkKy?iGtNodPY^8O9wFz;=3X84?JM|EMUVz|=<
ziUN&XLHR?3t^_^S-^oEod72vXjBZc|vzCu-TB7GUSUk``O$>bdWdI{rfLOQRbF6}1
zwB~gSzF83-SqZ<!>qV4peF-LNQD4MgNiUcsz>+?%&s)}AE48HCcNbA|EO>Z#9m9Cf
zxw~_0QbkP6d{5mVzaPty&jEZO!k0fl-OmvqqL-f~@Ph<CM_`iNB}FVGy)UfzT8Cri
zq5M2O=jnOKr_~ZYPLgYN;b2`t{D(CR89=@Q^vT#mc%XWkklvRMWR(ygMBDN2#|zaH
zLrjP9L-o{T^Ff}Nrw$_;AJBh%X0rZJ-dLRMGsYkHYPl@Qq?o48S*8(QY5ykUpHeRY
zuyg7VY8vkur?>kOS&!mWyVF3tW7?cb0TG*1pGE<Og}$Z942KF3BIv=G;xGa3B*Ix%
zyfCg{m{6d>*u~98S;PH?wvigvm}PANrdP$N{i{Jhfb`AKUS(&z1?RdSrKdp4wJi#%
zdz<IB{YY*5E0r<zb5M2l-^n}2j_mTw?DIo|J>ROWBi?Xs_1RCoQb{x>Y+eouQZNmk
z_-L+?j-<s!2XMXIzh3~h3LDp>SmPkZ@4XhihDfBmi8<kOb9_()+e58Li0K36ypDL~
znBpc2JE$S<F<`FA)$ln@t_YFh{uLN$Y6{V=-A>}!bf$w>u8Ds1|F!-W!4w!S=TJ3{
z{hpi%=wC*3f|hXGYw(7S7Pnfl<s+by*hMH$wEbq|M!*`4J#_iQXp&z5s91@tCs8U}
z1lk0KvCbr9=4|%kmj4@#VF|z~xsawMqkxDQVrBT7HmL_gViPN$AZ5CS8<6tbdw2kW
zYRcHqK)4PQZemxPRM#h5sv&P7sWFIfVNp{pU5RlmENXgA4WDggZQyDY)?X_5wKNlR
z$WJMI>KXG>Kg|CbnqSul;<7EJ>@^dXo>BWXnEb=@nPiZm(3zlG=8zwW9=+59mqZu5
zgjJZz0EVbOILpvk!&t%10o^4e=RKe%x)2#~0Z4sB1~lKj2!|a<_B3Hjw1Akd?O1N&
zhIYuCV^J^SqDj<;P~>14#EcjV^2VTh@ZW&nhPyG?o)5-=!(oT=%kWrylII6l=S8Zd
zm!oYRm{SucIQgo>K10*$0LstzZLz-|+QT%{fhk_-tS3hVlkbC0lUVci829_&Ir$oa
zR|)(LfII0iIj~8!Bn+yB(QDH#j`~kah@+ch%9j=BKm#M8KzsWPkQK`%qr}3ZXc_Rq
z6LZUfx#hg)n1y%U)Xu|`O7fmK?<r2Cq@{-dkbjq{OaV#RAh1DToxpVh5rHW=cVc`-
z(#{iaav5(TZ(`WaBllkCfl;vuA8lg6q}B)WFQoOc{4)>8KT_P2@^4TB#k~l{EkBOI
zkl>>i{UR#lmy%kx;fF^lx2b9BeHY0BB#Gj0Czmc9XvTV9!K;QYMilMAq3t|0`O^0q
zDk&};Teu<>ai+euj;xLv@<8mZHdWPav<^Nr!EhsX@Ssvc+d`y9AqNA3dKTiGE<%uJ
zq0D`=XSWxpYSLrRDfDm<Cb5vG;7~{aDgr@|4Fp{T<!O9``@#+aHXD7BIq{OF?Bus%
zu1d11V*MsIlw8zm7FS!%+dfCYs>Pv5>L-^}7YAF1?eclwc6_K>|7OKaSc&!_QK|H5
z@<#|LfH=n}{mxd@-HNJi@<f|c?YAUb1;X*8=@ZKjW=$+aIwtb6i%v-|n&jx@jG}H?
z)_ayQ?Z8ZoBp(?T5FBsJ>#w!z<fIQiq~J67vaNX&a}Xi+9H>{=oVctq_L03J4D4?P
zmsGT)P{9Z=rT0@k7rTs{7MuctNN<_FES^)yAwdnoVCI+1Uapt#<&im(+Cc6I0$i{6
z@*;&GSD`n9I!D+zt50K81{YQYv#8DO<&{GhVX$)xD%+cBuW;Wq`ff#qy_tRO?;fRG
zmA#@sKa_@J(4Vl(sw~FNyrKzQ{#sxwU28j+w6Jpvz3J}@__k9~iSTmnOiC5hjLPhq
zy`n0DE_M0K!#VfW%-)<TCiyWN<sPuBS>(iIRZe8oY*|A&r?MO6y(9a;ySc22hxgfB
z##(a}pWj>9r}g(sYm`;xuCnY-Vk@Sc<mEf2WTS0ZTQOEw5ygcTN7%%a>(^c;VTaFu
z{#a#^?~IZgUaJuWE%`bc<ZS|51dam42;+8<yhGQ#h+-#Pun2?L=PVX-Pu2rJ03TB%
zt6JhxFy0v3h$&%!e3Wpq{h1Bgl=0=SBWEctNL;Mp`Zwf9zj|CCqf1Pn@aZZ#={>jF
ziH4?DAd)L+=UT0P<5o=~BE%?DJAv%@bZ-{HOpjCbCc-V1(ijUe_mMV}KT6;Y0z+ox
zeW|7*WRBgEe-lk{n)t-uUWxmR$$IZ$8uVk-Y*BO7#brJc*0_u-?LisC4lryYMfz;E
zLkvrD#&*kyQjiiXniirHxx0mwgJ&C23G1$Hy1I*R4!6z6-!)!$Q#swCI$X9nM$W>G
zG@==>z0Jh~Y(6pH1Y46p8nX3&!9#M(4Dm=dGVE)Ci@8CBmxSgwD7tFlKG~rZF|f40
zVS^nUm1cHW2J@#`oSH0jVJ+{c!N?gTn;l1+(-#a>pJG9PbT;(NvS1+Qc9lCO7%~tI
zxhT7rQ^e+oe~Z)&b1%QIZ6M;kt_yu#N0gh^fHDdApV2?3{T&Sqo>2viE0TyPYbwv8
ze@4xKh0?1D_ZY|(;CqeauK1&Pd$~*CO#(O-O|Ak|W;q_2gc#X4qNo+$IV6?@tJ<HU
zy4MMe7>I~nb_gs3#MaG5aEpVK96_lR?@f>+S|rkp3yD4$m{0O23FlV{5P3?bMDo$%
zT)Wd|;&0W$8?n93(VS`Wa|RhFze$uwOIyk#Z{a=5wB9XPj>$BryYE@4aawd!=M3ZS
zJScP8s~e!3Agl>}@gC+#E=Q2z=tKs)Z4MonFmMr(iL5>3MQ+%AqDxN@oh?izItN7O
z;;IB0jW}=@MCbO2?kNx!X%kok6GWHo<(TL|JgkipqRV5ZJQH1>v=3x5h)GeOj7YII
z%0llHB4ZDPMfw3mXM^a9AUa%}^ofr2pF2f#MM(T2#+4Lb52zU?x>+>~qI3F0r{m_n
zfeaSTu8@C^j_)v<n@i|qjoMjeX4bYaie2^<s^ow-Tdn~;(hb<XsS)YIV}vOGK0Q(p
z?|w3Ik_4+WQ75YP6K{MNv{eo_vcVw#0pYtBP0D{rtzRcFWohOHghTlLE72lxk~>>M
zT=uU-k4uwQ=s4QMfV03B0i`)W=}`JboK1>Jl2D4XiP&O}9Tv)s)_ObxHwX7O$RtL@
zsMU?)nRc*_+uBhDf*3Eyg?gunm^%asUwhW+&SUVofI<OZwwYdoi7${(CJcN{d@Xzt
zo`}F1^tT3~y5<U7q~EldCPB@@c?*^ZN9e#IWMOUCBm}=?k!92)`vJKTi1<Abr8gjq
z0W&HuGN9^G+fYug6s2J?l|aRFXv=bsT)#&-%JXH|3|fywMR36kxN|<5RY#CDJTJ1q
zvH;s>fh@AG=&zoH7+>rmhk0X871nfhROM05i5#vsmryFJB{i=W!P9ws3*hTxz?D-)
zHLG+nvxIOC!ChQo_KvILu=2i!_c+lz**n$q)bYI|qNq+FS9w;Q1T3jjT=oFxzKQH;
z*hD1(OYO@>0@Vq1^2>%grMxd2w0FJJ<SC4|3-F&hNs#*Slb1Yk|1)@FYj0ki*@qJ0
z@`8d&QH3(D62ax=Zy3D?4(S64Fwy5Mq|c%{jkkXQlIbke+JkCtawX#{O=(c94?cmg
z(^Hx%-@l*8C;c}6mYT!0WSMtC+gMU1bqsSH!3jNt6&_O$QH_{q>Gd!uV+Lmnb-Qsw
z<XMfMM6S$8jh_$;EKyHk*Um+rI!D+xPD7%iBo-m-&Op{ZAdX@e@=&lvZbf-a!B*76
zssiddtIm;vMd?BHFqPl|l(5=|$i`LW<ahH`ro$8{?K~*ORVk*NCD2IuFbV-}e07XA
z-tXM$93NJv_p!nSO&miW`y+p4TF8pyH;k|7^3S4k>JfDYm+_CZ7hsm0;MzZrD#KdX
z@Zu!ZhX3=^nmDDh$u&KKgXbqt^M3TlRPe~*RPcV|22}ACee-MT{E#lrL(6!pv3@<)
zZ^@s5*pWX=;2D7W_^kn%d-SE7qmcSR`oMU?|1Tgxfjqe6{}|}M52!|&^N?FNOnFRM
zdocc}j_9yI3R-t!R0?`@eAMGAg;88bqbQX@SvcP^{%0VC-x-*WHhQG-M{|?6j^8zl
zdT&0xc}3ZHFL>f?kCAk(n8`&LxhJd!D;fEVRQ@Fb!xyinuQ#AC!>F3Rl@wyT-ocGK
zGS03yV&hh0S|1htu*vz&@Hg6`b1OV+V22igbrDW<L0^O$jZ9Bi*^Yi3K`Xe^ADz@Z
zN%%BzS6Yp2x^x(2>P}<7qw{TQ#gH%3!p{S|Q%LY7xA}p*W5gDcd+L<*2km>uNL)7V
zPg^0s3Ggj)(|3-4Hu*OR1bv(MF9=H3rxannm^t^3?(LWr55JW=M%AhNiAPw?*8RHd
zgkdd;+GA!2TO~vLSH3`#U@LI{j{mM{wBvODpKs(oD&IM_g_}X#X;287m7xb^=+Tw`
z6|Zuh{Yz=0qpTAOi?0zWm=e3g_(_<&7aKa1XDr(C689imHe{<OW~VO1oqwmjo+JoZ
zYq-be8~wysO4B5CJsMkNEXkjy;tcYYecD6W;47GTVdM-!y?5G)^NLhs_G+6*cS0r=
zuELRzPtEXZ*{sJwG;aE1#yuNg_jCuysvg-{i4*V@T2JYqdn>cl*#)RvNL+#;5zC*W
zCU!7>o=X3Sz&|GN3k29kdxJ`M2=L7!(NU~lOPqosUm$P2wO)mLfcqx@+Sb|{<yP}Q
zIK(bp6ZlDXcXrImiyZ_7kkH+cd>fCn2V|BY*ED3GxmqAX3;kiAe?s%x;5lTC$X~`o
z)dt_;q3W=)Mf2ymG%&GXLXts7kI2uaViM%4Vnul1wX&BJTjX$aZW$*wut`~UfwBK8
z$zpt}=)$9GPILC$A&hy1n41Ei=Vx-oq5)?b$=q;#5RcBm=`QFc6R80D5+Z14b<bEa
za%i#N%cTgG(&x<caK|a(n;i8eYh<vVGmX2)O>*4PuAk98;9Y^6o<n>)2O;vL{uD*J
zjh)l)D?E}reKt?)L?nX`$t0o|HZ1=KI2EcM4hdzBL+Mfu8vJ)Q9m!iaDU%$gBl+7&
zE_jX|CrzH^%r29?Z=;uo$(TkM&e=l<Zv!q2oFEF-!Oej|ky8cyliLirNC<urRv4mv
z)5(M(nSxoA>zhtCOeYG|PWK`6YP=8Sp3SHnj7j8^!BENKxrppD3+5`a)YKg2L)aDu
zEg1^hMoHyiOwOq?ltc;cn)}wl^%iKg9Huj9>=K)VFm&i$xNY)HB82QX<oFQG+aoZ4
z2P>LCY&Gf#Rx`e$@?k65+~;>cqVh0F9GE0om?R4be&!IG{7=!mT0kD3ftm#bn_ZDd
z?SGH(G*Mf?&EwlJFlLlHG%(P=Qe2h)3`d{n!wQvtg}^^2@EZhvmB3ItAaebqR59t!
z$={^9-y-na1W1s}-vK~Wki@jSLEv`@kf@fwM}Q;?JEQXVsl;yRA5!Tt0z{LNMerX}
z>0c1|BLaUwfMl`UBR~RLeuKcjBruYFi;3)GiT59=>4S<rUe}+{h(9IpuL=Ac0{@l(
zhx!zi{yhQSjMG&5cL3GQJ!OWS$c!H+F;PX=Tp@@5?;zceU^;{WiHIC>h6W;G$wrh8
zt;jCY?RRrXDy9DoLGF=3Vr0<byj!+OG*~I;dw}NrUBpw$hY14M1y^X9A0_Z{fQpq|
zN38N`RgT3-Wo_3Wy8YPo=}H3sC5v*-_vs%V`z>kT7oEEAC#ghN2^V>!V~6|*9%DPn
z!jO1bO?C;G1UL+Qi%J~SvZ%zG?R6@V5h*F|6{p)B4$mg%x<~c@fxw>;7<v6%nx(pr
z0DJ-8@HL%sXt8Imp<CZGkY#0<#=C}Pe9uU^-z{ZJ&l=BJ-?LC}q256~qEg>Ajty!o
z)Hqc0;QecORBq4*1>iTnTeOXLthD==#sanEFazPU&|;v)NE?4ybd3LNW!&%B&tQFr
z^eLhb4itTI-?1wnpw9}uDg9aB9)DXKb6Wj54d8P!#&^*3|JYgMFF<VHw(_~R?JXnI
zANv=BvG&w@9gIb<@7Ne?V62rZew%l@Qbr*Dc9+Z<K4d;$K2W-ClXL}_1e#JvE$!Y;
zxSQCydb%lE%L#df%fmdHK?PEOnw-rU=Q&8Z%-Vge&D0TRsX1Kc%mlV@5`I5Nvpq?G
k2;ndaiJhmC8J-4#B==dzHQWkAgXW+~k;s>F9xomFf3BaQssI20

literal 0
HcmV?d00001

diff --git a/model/__pycache__/q_sampler.cpython-310.pyc b/model/__pycache__/q_sampler.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8677eff46713087754fbbd59d48ae0d6d76d04dd
GIT binary patch
literal 21269
zcmeHvd30RYdEeXT%{~|ofC0eek&-Bm2m&OvGqNm;v}M{7tXQg-vDByG+y`*LS%CW<
zNDzLLC<gUMbE;N7Zrr9`0;F4*txej~bjvyE>Aoc0m%Ox1nzVN1rfF=-RxDd$f8Txc
z-YftpN^z3Yo(#@>bKiY;`QE+X_uX%M=w!1g37^-#^f~9HpOU2Cpo_teii^kab-x6_
zBqledmRyxvN>yp8Rkfv6wU%DhWxA&{jjADj6IBzxYBSkNR+GJZscK5xOIOqQ)tZ@B
zwwe|9^k%M=ujWPBXck&y)iF^{G>ff$)qSl}wS;SvC9g@<ah76f{7yU}u?)*zlUR0B
zsZKKG8L5(c6*x*2JupAA;x#($S~Jj}U0Z27fp)G@_X6$7hU*39*;=z%yU=tNDsrHC
ztE`1z{d}vo<e*S&)h;>q1;?w|ZvCRe)|ymnKC#wdwRYX9D8a;)+GVHdcs6(Hown=o
zwK`1}9ICMzroHUg4b)O&yWQcfTC=fHqXu<_u<Lzt_)Xw@7GL*CS&}@-mpyq?;g9?B
zmcrz1nLp&IsMDCdDKq6IWlGwVyNahX)t5GvbJFt@UDeZYt@)C#ZmOHIuUu02yk`v7
zDUvKbBdtz(38wqf=Ot!*5iLI>vjj7*Va_cq>$ZH3#+H4_(|pV}Ep;(wSe}r&n4xPd
zN!^5(q`R2;!_O&QwQKsCulo8WiPxCgmYFswb(4P5H&MeeU&>EnzO-+CUizX&bNVJz
zFUkBWQ@g3Q=_i?5mZ+{Q%}Cu8%XHJ2CFAM7DQ3xcvwmhP<r`kwPjAYT(zYy1e%enx
zD0S1Tf6Pz#8I%&f;V1EHtbWWh{KTdVLde#>SCgcah0j^;)I7(kSxxM{)mgN>M$2(M
zXT`O=j<x1G)*|n;tXkXZ@WxUDG-I*G;^LZ%9kW(=r|!7!tQcttTidq0i;mS^Yh7^o
z@L*RQ>q@Qd(V!kEC>+&UT?4IMJ>fPs9A>R_IMIPCMk7oxR6J>KsyT7<108NT>$Mh0
zch+i%mfX49weCH6asdBYn~5=8M-ZS@@2s^w*Q#;H`qZhDv#0K#J$>?dF&<_LF<)$O
ztiq|`VbC5_++aWv*k5pPzdy@tG(`uqp`#d&rWomR|6rF0vqiC<G~Lb~r%#5@dOlo;
za5$?0{Ke4iCH&N|U?^N^co(fBj5S(E1Uk{PtOH}w4~+y~Vk>D`&z(E}l(kY@!7gD$
z>@u6h7Bj2UZe9fd_PsTu`I^nn6^Dh`oLQ;yT8nnrjtS+=$foz{Bc!87h)=YY0uv&=
zouA$&(=xNhWvsZhwlXwt4=p>~$}^UOC&x4NsW&7&cNT=xUnqznuhwWstFUukQB&`<
zTD23dLr>^o<>H|lZPutmL}8_A?dYLGTDrxPShTb~Z0D@^iCtT9>J6|@z2<^yiL=C(
zEgZ3~;29ii;a0<meed(t5QBzDfj|kvXp3((Sbz6>i_0BvjknLxiaA0EjBa5x>JEPF
z35jr=szx-p@G8W$#0hLn)NAc_$Fu68lmXAyJy6)j0kMj1NV$+Ljkf13(TWK^t5e_T
zL#G0S6@)ZSf)c=z<V(*RU71dVP5DnLt2o-OOE`p7jDqt=nL8QC>zqyy0p56g{$i))
z%mWYSLZ@?SzTRQZJod+>6L9{Shfm)3(CLTo53A3fedeR{Ei`D(uiCDVFnn(1YLI$_
z<ReIA<<S#_2w_!{Htuj&YB<MYR+xLF*{Ro>?xQF~?cb1RH>P^)(4*>zyCCE?GQFh$
z@*V3d7*>2LnIP40qk}R~u`oeOFuY6L&huy$s7<FG=)Bfma?rx@g4D-q%{Ax!oOgJj
zyDQCx7bG1}jYAY27`2rZr_Im<f{;~mK^C|AxL0$-+*Qpo$8B(jh1?d%jjBn4+Q!la
zy0h-_T96bd*$uWHBq1kkL6_C+&_IDky{hR&UaO-43DO`@U*puPnx?~`9rg=kyT5C^
zNKcp_S>G{Kx{m>*_NC;KoWr*yD>oG-C2KcJYDUh;8vY8XHDv|&wVS4D$OVOJGjdT*
zp`O0Hf<fvkK=dIRGx0r(uX_z(qu@zfIP<sVWyP15RVKeI`|^~urEN=H_2W`|LWb@j
zLo=aPTe@c~C$^!N?AH2|l4ttra&l9~ud*pa19*?GGUzPOQAX~#Cw&#Cy0T3twOacu
zm_ab}`)S4No^_e_*zFz5)~OTcNK&m_#Sy}5SFJ@M_%77E`b9c`h3s<NAt8!}p~xQT
zJJcv9X_5+_({?d#e>`GX9ORIO;jFk21X1^f2D!jHH3oU-+3rOgHZ)YT#s#C(H4XS-
zNct9rr-MHf2Q;d{|5hTb-^&t8gwQL{5gf6E!w>_<575J?FXQ@8Be1eQC1$gUN@uN8
zv)0_)-19w|EJUk7^<vEpaqX?;bvZDYbGbn(@7zKq9q3+%*DnSs5|y^tkw8J!aZpmA
zFE%?h4-DvfS67@sX|DtcQFZTWJ|!9mOajAMU%@e4(E=6g8YJUq2=XIK8K~TGN4ceB
z_cj3<w%4)4=X`?bxd5OfEya+JDn-SVi#JmWe;Ac@e1|><arGF!t^$T|q^iuMs>0-|
znpBc1dq2~d&J0Im34Bc^EkZqJscW!sSlTgAlUdYQmgTN#)x<zeo)u7ICNt~<Y|Ke6
zr&v+IeNLKwr(nI%FGee|ag3JfwVD`cYdYDmEm}>oGRDijO{*z3jaGR!ILeF|F`8{Z
zJAf7ic5q;xL(Z6^(e)zOlUYuY9R`*IbuqRXc7)x5{`-3UhkJ%kc&8ZoE~j+MkqYc6
zyBi~oJISqyZRu^t+a0UBfM+SjuE375$~71k)yePbsQ0krV$||&jyj7Kyd8Xt>;#(=
zBTt1RCo{=;d{43`*eTG~X?*X+m&Usf-}^xi2A=i-vBD369v{N@;h3Um&S*u>5VeXu
z-Os*<J#tOQs!Z>g<vn7g!MPsAeD8Li_p-;tJfJw*(}Ab?-9JT{y^lREM&G~Z=<g2i
z33is96C)pB85}?#46@IVj>Db_m7$Fv7h?H5*<O&_o-p0TsUwbR7#!4&G||XCKvv>P
z2POs%4eR{5b5B1tXFYon3T?evb6v7ONY7YmJ4}pB2UtADS!nLO(|q#WQ=t?WEiQWA
zihE{$zQ)%Zm*+ZsX?}$*&YwPY^7P!vhaY(G^c*fl!@vL{PK^iY#RhC4+l8|ME)SvJ
z1*!hnff?Z%n3rqZ7DCzx4~#nL1<{Afr1=RQfh47mlivnhreDQZ09DxhlC3c)tk>m*
ziq7AIcKk3vP>f6<Iv+xvhrgHZJ_Z1xuenZhF>D`KkIHz{hH!6qw$0GTC56Q>Oxb`^
zvoSHWQFBqlAQ{iLFyv!NqjRM3+)(gb(5NMP5@;~evbqc(2cQnUN9!sKPgf>w1zHX?
z7nNx*D_sq`3tESMk4>ekdnu-SX)oi0!8T?1HK6Or(Ac{MQ{W%T`VdF(kL-hg#Dp)z
z^!3XUf5JDmatvMyKQSYfLoY_Mn^?Wi%lihrHu5FKefmSvXJpJ>fVaa~9%Bj2Zoso4
z*+r`D&xxyjbQgC@z}W!KDbywWRERU`$32>lo^hO<X5=~AruGw5FKTg+Q7ub}`bpGH
zESG(Gkudq85R<JbYPn7OK8=1;;dq2SYGzQgf2ig_zs*e8=HOrr1D>MiAs;J6^Oa#<
zL?g^#Uu=uU#FsF%7zAJV8)CW*Z4qq_)9MJ!Cbs2~{zsy5MT^vq`*$!C6r0}o8NoVa
z5uXR0!C-}2)MrRx{yQFwC#zPkcet=Qq5ZuO)7T3kr_uoNNbloI?wR<CZpF>5V`R0R
zv4nULVuO57bE&xXM}=)V(tQCwins-AzD&sI;3>F7f>iGc#!k<m`tA1$dlW{ZR$Fhl
z8+XQ<m2iHLWEvD-4iw~6l4rpn(DQUg0%76RoA~f?)K%`}N&JMoOR^y3+4EHT5P?$w
zWaG3T`j{{^_#{>7D;!T77*M9+VF)y`5jn{beuzM*3JQ(7ryo9WU!aq()eUrXgtqtq
z+EsJ4<_h#tyEj2~-|)4)!kx=ID%+$R-b+|$?kW~(A3=M#XK+ANVsY(OqrK*i-d3BP
zrRshFd(S3QVg$H;8!Mx|XZp8Vr$%oUaYTTu8Xay!1UJY%931S|(VS7eU2isE$+Do-
zzu$6dZP=;K;;CJ0PhU^-bF{=C5bGDqVX?N=e0&`)M|E<fB3|hnxyt$`GyfP3SHwBi
zTgj*)KSDRMVz{Ud>c7~wQHh=v9`)!1>Y(F|X91K<4r1MS%hYmmL7R{@*sujyu0{MR
zFBVWX<Y}n#1-XpBqMTEUfJJ#ip3t<Lnm#QXH_IrQ8Z6<PrlP!M-rSfOJ|c!@-t%nW
zWJ2%{0^Ani^QWmf+1G;T`S()kqXfu>GeYj19+m$QfNIjKbwpsn<sYZ(EM3QZS}oG$
zFu7JIch-f(zuS$N0r)3qc+q4R9;hCsL2t<i{8Kal{Ma4)R=iL>5RbGQKU5D5-`>d+
z6V#wj;}QJ_M~3Tn<&E-in*sjV>&|I@0gp9GonwMV1g5=k&(BgT2VmpSF4Q#GGDxiV
zB(m%Udb`trUuaaF(lL-wr$}%@v5?m!o)j(>xKg@Mri42Mfg5*B9K2AjpqP+DM%hI8
zrz9Z=BrWT4jarf>p?X!6+Al_q6FP5VeN{LuT5v-2LK)<cy0AuZ2W$1%y6ySv_Q}e?
z@w4;j>OGTnOc?xg&kElNPB0N6bFE_wXFvGyN_b*I=M|n)R`Y7`SoG)WpOJ)|w_~hl
zdqEhCRuRf|-ys_XLaZ0O7vQetm*ES7Z$daxnX?|BiZn2Na5yi)sXK6RlZNfo;2(;Z
zYZx1`Ps13&o!$!=(n(Ez?8U30)1US<dgL1Gg-BlLEfG|K7@b{@ajf_7L>T^AcyS1Y
zt5$<Tjy-n^J;_4F;Ip8TAO)8vYunAnB}bfS!jr*2iCer1P|-qJe~BuZ1X=`Qhj8dx
zG$a4+H*nJ>Z=+dEL7Eno96X2cDdKNbrH;A;Dli`+<plaiq<muw7a&l{SC%CZt|<sN
z)T>ph>k%&1khKujC`7o>sHv9XhN2c4HC^lRwc%^a2r@(a)5Bj)QZWbrpl^&^V}5Ff
z`9Dna%Mw9EBz50dRH5l9wV#8^AMejFgTz8-h-w9geA+v4stYdhX04D_Xlv*b4|V%2
zLwgNn1>qG6xP<#1Q4>XtA}#=_57X#ggd+?`VQE5_XklQAZ3?-Gcy`R21KvU6qT!);
zp~wggVn*}@d85-F{1@?Cyc*%+ShvZ5akpLhCEgaB<oOQPd5S9O;b>VSb!un_haWX=
z6RTD`hWuDh7khWZ?shWmnBvLKQn*D>`EGZbP@6Xfxc>$49Ip}hqXcY#>xlt5utK#Y
z463>QW78`3+mA|!snvnQR~(?x30%O@a)VU7LaC?}g|H}S3T*ID-7=wWnQxhD?#-0G
zarf{cx#yUBiW9y{sB-|wKT_aBad(kQ4FZb<mI$~6M&#Uq!5Qgv9(t1d@g#f=-8PQz
zjpvSDfeIUK2w^(bBl+h(gbN}6)E)AV4(?(37uDe4F5}?lkE1grc>h74Lj`|6tQ9(Z
zyp{Y4-5Pn`GHHM$QS9|FI3*nNz_*G=bt##rU__g?adh~h_v*^W&@IpqM`l5?zPN<c
zmm0EWjI}mZ)eYQ@J~TnR5F5DgbtbLBQ{y911%i4E;+z6TNJjD1ZMAE(%Of?3fomNt
zOt?ujBy*VL62K7wr^hm!E{u{{9u{q(2Vpb=ZINE`jN}{PZ$(|;9|i`2d>IQ$Zm4P+
zA;9KUTX?{#g;*!(W#ClH(b8eLY_V<|HjY~RawQeA60Jjcq+X>GJs<xxK#&nb+MP9T
zWzDOm!Yf*wYVS_CR3IGN8{M&NG^@}&7HOL(&RZ~xazQ0SC!-W(P1D}elu-j_s3+OT
z(12igV_ti)T_@*O^q~WujCP@GUd9|mh+PxMD|Ai-^^|R-_%H?Qn?x{=Ruo4t988IA
zy_>=+Bk>2j04LJRYB!B*9f>xmfg4PHPVHv8*=`o;HTp6VY~bK}v72Q&oLsqX9(5)&
zuvZ^KuOtF{1k<R^Y-N3u0*26YbAGy;Z=V!x6KK2U<+k$M(x057gr2PeLpzk>K4?#)
zjQMHw&3{~Ch>&YwD_LrrrzE#?1+D4tllZoEuSnyH1S8$g`FTINsdfu~0dy(zpN{9;
z_VZi&{6d&Sv|Q?9bbkzqLuo(5lKxmpLOJ87mrGmY+ZcCs&M)lVW_1p@X2|EgHMvdr
z_ezVDgO*wlvODyxsC<~0ud7@rZE<ZySy-qVO`PIzo2Y#0!m}jj@cFg(R>~r_7DmFg
zhUc{S_u~d%C*Tp71_<EBT|tHug~Q>+7F-aNp%B|5)rZWJCC7Ha$K=ThilXEU-V6-*
zl+c0C(U?O2OgnALkF%GM*cIeB0+0xkyW-NW?q^7QV?#sO6tWLn?_212v8v^;FjQ||
zsMRlBsd0FO#E|-i!#g&`7d@!yfnIOI-BKwIu%O^Rq1wJkSp5Kjm>I>oRCN#ObyxWJ
z;Z~3!KC#zN20g|U=iV{u^n-L;qua<nbehQO6J><{qcXbfVAxQK^w_LR3`^FVkt)GU
zL5E;L)!?1T+|21Ft_^r4w3~*Sl2iC*5S8Emrt)G+&&Vd#At-4o84Zzpcr&1TtK}VR
zJ~ZDDTNf~I%+|k-i!d4w@klx{^lOFyX5`@|q4@>!u4;(8o0LZfmX?<du!HF*1Ut-u
z`4d8%szT^OTi#B9k(0>C+mCx@PcTruE(8H`{?IZl1OqX5%1;dlh9pEo#!GKyd}4F>
zznOkX-O6rD%kX$F$xL37;pHYAP$mKYqw+_kKOup^lYS2U3M3**lAjf#Kkw(ExhEDv
z?op6aA>tZgO#Kl&J^u`WO#;{ziGKv(`k3%XhAu{-9Ff<GKStnT0^+Fs9F;~Gh=`rH
z3Csfo+U16GMK~#i2jzNUZ-VU8B9Ue_7oG=^`o#Y<jrmyuM4ntwBH3s`rrl`^;&0X5
zOM$U2yg8%fXRh=zbAFyEkJxfLi~NqaG*x>ur<tmtK{@r7rVr9$N<XG3Hzz=uqgGus
z+}?!sWz6yaJ7M{VcSHU9GqfQ;OyKhbMwF{`v{sM`UwJpHZ^|1}!#dZ%h{&O^TebB%
zF?95yM5ma+iL8(iOXpS*%!o6bv?#EjDNgE$QK4fA4Xb~G(}@j5Q!uKG6FlXqTVQ>#
zWX3~?3s*3#eZ7Pr46YMQL3k-E*gfTEw{ed9IZDP^2>I(NpixQk5tfO$1t*5^32_VU
zgya5!`#AY^EBU=SW()TRL^!d{D4Pfl5nfWE=kPC5n~@F5#LEe#eM|v=&?#tUM|mQp
zPE)RwvN65KZlHUpJZfM`hf+~+;8BW9kPeGsB9|^QzQo2g+Cq{9rZ^WGt)(CjOA8?+
zWPHPE)mrg_eA`(<z_?!lKgzQ~uHI?F?GITdBCmCNr08=J1-ce$+aHIb&5-xU6ns^D
zHGJVA@xYw)w+Okum|_M?T+@V#1ma-k%TR`dlTh2zNy%F_pp+TUX<eu<-C}nilsYKr
z{6R0V3^D>t`gxWFk(AnspX!#p1eCTCNM#)N(xOGC*J8rUig;WecPG69h%Aph@v=AO
zPa?a!%+eSIndQqxzpyByE`u{Ek8336Bc?iy(msD0?Q&&_<w0^Y=$Z42{ytx3g=zSY
z2yP-~+THK(hg$hV!1_S<VE0hh^7n6**qDC+8RSL(AmBd#kSJS#r7t399ZFb<K{b3y
z388<$Klqa3AM&l26k4h7VLI~$@3R6<nu7$Xow%~XE7AT4ot#_a{*i5*TB1DRn<$|@
z)v)m|DBU}DX_Fjjb0-*U(m#wRzXK}foj6DD^2diU8N^hVuxfWb2tU^&l0R|lb}SdR
zTm3$Ne5>pu3W`;bmZ$uZKZ7|YupdXU>R9=$8PsTk;S<aI*(6r~Zs2vmKkDBNhA*N2
zK~i)5iST!1rzdHj{Yg;Wj9;d4DBhhGtBI8*TA&uku;)aX)CR&OY(v@zS`5?3A%BVt
zhgan1Eod8+?maQp+_MYC9P=xnnA@eA&;23Q+#c1uBCr1N-gE;xvOpc6Bvuwi-*L!~
zG0??f+&RK>Lf^cD<p%W4BWz0On|ETzXT7`pS)#(_qtG`|Vi{<HcS93A#-_nNIcS0f
zaTnzQP4I+&+&=+&t@yK~38Hk5f1FA%Rd-M+o}l9&?a3@;qju*&M=3u6m=QESv73B>
zWclm>jl7-IASrO%KfDb*CnYw6cT?uyu_Vc|{g;$i6#iS@tUvD`LGH+Wd(zLc1ETg@
z-if$2gW7|rRbBl<5<BE)!>khcvgd|(^;XhL$dLKnWyst0i)6|x%IZ(}bFoyJgF;{}
z1oFi|zQX?!jwt?@3A_)WJ{aU8<%Pb_;Ctgo0U!^c>*Jjy<kOJDA$&K9-YB9PU!A}i
zcS+@tPFt{5sE(vquNM@ap<X(vKm)xV@OAVOb<~S8Z+s2=Tf%!lI)3x8bPTje;E(1e
z=MsJwHEO;3=;|5YfQ11?2g;ok8R>+&u%?9#aXrcZ3RV170`Y@a6BiqZ$hclLaV0DS
zM!kas5jRjSHUi~JV^mcs*ltt2vjFqB%{o`yvrt!Vlrd@<KA)UihSvuP+%O7Fc|Uxt
zP*xrrR#rl9e`sB|8tW7zv+H%2_p|Uvr?@VQ41ntjUm)y{1H6$72Mse&9DZF1G$f?e
zDWTV^HN`@pJv9wGWl=&_NNTZxgm;mQ{VF+{-q`<i_%;;67EQeDMDOI#%K##@c<c??
zdPA*P_^n)5s%G5|eHztt-LCVF>()H4-B&I{3+~U-lpg_j6$=AJ67SD7+CifCdL93Z
zH0jJ5!rBPPQT#-xQ5MuFi}F>@Sd}^99N{XB!)ORi#K6fD7^Gt7pBp}#@NCe_1gy<Z
zixy;vxgyFI2#<t|gq2QvDa>xs7LkM@(osT-kOE+Vwz!4_yClJ|UmkBI*tD9wAp(fO
zWZ#sBX4q+mt{1#eAbc`SB<mp$5Q!Xk<%Vv<CnJnfY)Rv?9r|{*F_TcOIdPLvtobWc
zI!Hixend9So~G_Rwfl1fK0%-j5EOTMZFq+6rU5D^L%lsFcK%~@Ym2~j0{0U5aRNU<
z;3o<E6ag|=LhsB^Q|WIKAW9143!&8;a}wTOfyNB`7XelK!rI~@y<{O?a|%)v2eZT1
zJko+F;kJT@HZbA!!kZm<*+cPM{vwT=BjR)I=m!5ARB3=UkqgCtPF#RJ=`9dbsPQyu
zR+M6`JyiASvZw0zT(>|2CV)UNwB7$a4VNE4$VPUI$b3PH5fBVC@-vB)N|DdQzd#*Q
zm@;}mfvrsyD{y2qM<qmgF9tsihH!rtqun!>DHIeqlJLR@OdP~wGH^KM<Y^?sOv874
zr)()Fl?>{Qw=z1xV&a&34;)binG7JrY0$$&nZ7x>s7@#ox9B-bJ&}Mz&aG*!?{qk(
zDmVA5aAOH93i1)z0!Am{+{z#-lz}XJNPdLkL&|8#VXsFFGS~sieWm4bVSiCl3hnyd
z=282dn@7ls{(s3lf`VSc;Wk4GdTc5QBMFAytxToyKM+%CV!%|I+{08FZnK-IRQ^3S
zm99bMor+DR=|NKo>go*QS5QZ%5u6hSQfa_In&}xx;&*ogi432xf2bDsO`{P5X#aP`
z0NR}vh7F)Uq=g|z0I9~YiYu@F;4P>@S|;<D44}y&1Lz<OphG<a$Qm+$4zaQ@fDQ`-
z=t$21x?{iqx)TP_U2JN=0J;ON@ZAicqp<;W_mBZJM1?SbUf9zBnud`d89+mn@-7UZ
zW8WnM=vTc;Z0ht4ps#xO#I=zDq`2Sw12=%a00StJ6`=vNn=No_BZ&VD4sK4TH~(t{
z{yG8j@r8f_wL($)Dh_%6w{W#__BPc1@UYxl^XGq?#{D}4{w{%^Cm@WRU!W2x_5AM<
z`1=I@0fB!=;2#nA#{~WffzJ@wBoJ%o{3*Kprv&~Pfqzbblym+s04m2vlyLrwbc=Lw
z{!0W%9TZ_j{x7LSsyF{v1pYOFe?#CDfqzTj0RmG5NZ;oFj=;Ys&>`?22>eF^|A~Nz
z=KM01)(QM>bz~mUC_9wmuT$-aW_*L{-my~5e}%e<U{I_U^Z!hB|AoMRCGe{RevQC?
zBk=14UL^1v1W4QE|D6D-y8M3-_%ea75cr=2UM28V0-q&7IxPQR1db5+O#=U$z;6*C
zg_Zvw0>4dQm`H<}U3IsR5PuoO`?eZtMtvI{6&7$`N2OPqbIR|7hFZC8cB>OLM4uTH
zQuv<5*G=L1aVj7}L$6K4eL)UmWlI$fcnx_3nt1sc84$wxe$dmGDFTy07!}bLINr@}
z4z4bGy$G)CZr&4l6L6{GRUbItHDoX7aOP4rje_8+-g8MS$f-bBmBOXMC2hbZjY!ds
zHi?lo`!Ijb&%#NcC%k<8la=K>UL(?=XX<$AsD$~De}k}q(jCWmp_?A=H^iTkaM!0;
z>gAmnQ!t<a_ifQbUWJbY&mku<JrlK<z?@RAjT*HPPU8IVvw9_eOk|fFgNt|)VTK6j
z@-FyJ_>;hQ5a)5>`Xih}jMFbuye!w9TK!3XDh!uw;dF#s&xBhq1GnBZ!rED6qx_~f
z31__yC*8C^jr^iC%c1sluRKzlhnr4zRh&`<KOMX2(Bt~pLMVTQz`X|qG9ZM7T>S;A
zi-j=WntPEdwg?Ct`qNbU0RUt&y|}Y>Z}B@NJRff(boUgnePJp@t&qk--u?x`Ay2>}
zAY2bWLZuNQITLygevEEy5qL-b0&zA7sVKzYi152g<B5goi@{-07AVO1NbYH0r+Wg7
zd}h~FaP4y#F_SN-<Z?sG6%1!GrZX_6BlCIMKrROvvdED(-pt_D!^m_lz;MP}D(Fe~
zGH+@I1Qz;hy7?VIwR{ti_mSA|!2I<z=8isP6k<{GWe^QmZZA=ihUrHUm5j*Xz*c}Q
z3<QNKBLhKPh>`?Sv-+YW7mE^NuN*{4UWgKk-jOJQCltbg%2{7WDKj8SB5^a=hQ!Ta
zo1$RL3`7auF@z{d<2n>2BxDK@7WtShOG1=n{V~**AWBTc{_r|IM9C<|IS7h8GD1j{
z<adY?dM43g_=zAe3PNBU1*4C5Ct}PdcEida9}pOX*|;BJ7UBhg(ao;@h*+BmgvI-U
zViE$71jVmFP)y<-PlzXf5`rQn1jUyjD28hxDBQ0JF_DVJ1lsd&VBaghj~x3Y%s0xh
z<x!5kT|MTVGizVvc%2X#Q7?)E+!5xfo%$tW(odjmCil*n6t9sDv8efG$rxnN31U#Z
z-7)a8_uNhtDhmG|LiP~?j{{V+Fw1m7=)9GK0p3t<*Wj?UgOp8~ruYwWC}(V&{@GXW
zP6F>Qcj~qsz9II08cPUMMx|u}vjBk+zFfnhO!QP-EI`l4dznf?vDK(FPe5qABsfHF
zMv!QCMAjV_`z*w>klSCQ7OxTb9RlAX&|iE}DpK7U01}CA2?&@H<XTFRweKi+mrhZY
zHx*6!j-sc&T}&6xDo<$N(NM3U-b6iekG`$UL^T>}Oscu-)-|G6CTfEM##g>wFqAj6
zMCx10B;Cni1{zPp9R+ukgz~L|sr-(XOucSA2K;tuQ$QQ+HQHoeH!APGO*EI0R9@Fo
znQs~vjtF{NM%KPQyec7X2Ikc;uY&eiyKkCl<s0bnnvqp*C^_ZpT0ZkN>HCel`=b!{
z{k45Zqa3H*5xBfIh>L-B(t7Kp02i}|i_ybH%VusE=Ye6^_Zv}PV`PoZ-g6ju4h4O+
zOyM;_A(b)`ey*-is6x=GAW=b+IuQ|^vOQ|n@zV3k)zF#~;S$4Xv^((QD;+mbpFaO2
zH_^6wxXD^`A>X)j@q1rU1zuHainou0tVpVx6Mp-}HYc(V(sbLM6K~Rp&_!skRS1WN
p2@px|M(ct3NI1izq@szWq5YEZj^jXwj~rg|3{(PBHy<dD|6g(*MY8|^

literal 0
HcmV?d00001

diff --git a/model/__pycache__/q_sampler.cpython-37.pyc b/model/__pycache__/q_sampler.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..f62e3cbf0c09d6fa9e1b864802ddc91ead24e631
GIT binary patch
literal 32803
zcmeHwdvqM<b>GbF`@L8!fW?F4kfJCq1p**VJ<KSwWKotCONcBfma~?;TJ8+M1sA)E
zZw4fY%%0|$@*h4uc9<q@oSJfhHg%J<O`Eh$(zI#QG)*5#TKDZv-PBD|$Lh&B_3?3g
zQk|Z<zk9!#*#{ORVOw&P5IB2x=6lcgxZnNVd%xNDW->_yf1g|Zg!zf2s{Ah9_<tc>
zJcf_;Sr9`p)TXkoR@8N^qOFH2q4jVjydJ4UREpD@(MpuRW0g3*L(Rl`qLT39l9eRK
zr79_Whnwm3OeMo{k!E&1SIO~pw3%NYtBmn=tXWteuZ*u3D@EMLjl^q8Wx_}rDSS^p
zr5I@=^O|C0u4|PkLwi9fXWu{@m2$+5e{aWbv|F{N8+mcF(=^@ig+|?W!_PD<+l@a_
zYc^{ao91#^b;I_Cv5xP^)9bZW6BmW`+GVqP(X?w-tA5EeHk%Y0e`>Q~)LM13thtj{
zYgf#sX;+0=Z?`O4Y}TnV_ejmCq1r2EwSiDdTy3?*dac>mu2F)zMy<nN7M~bCPvB#H
zNL3VDaa3Equ8G$ibw@MQT~&1K5W>QSdR;ZNk89J)b+xN)J!nUakfU7JE-0^@?1t<x
z?!%7agsz9KtB!VA6X)z`f7lDk#<`aiJ7z>2<ui&AeI4mOrW!FL{u=7sL6`2T7pQF2
zQS7jT+Gdn4N|e`=N*CpKqpM0cw)K;Cg5pv0&6l-qs2g{}PRNN|R>ZFvp_XQZr<88O
zNjPzYd>J(+odjx3Iq}aZuZO8tCvJo;tK#cMsGDr%orDo8DHK*xW|eNzNOx1HC%rXq
zN1Qm<qjxh-dMD{b?Ua+cu1+bts;W3CC;5=lP2Es8CSOvVn3Kj`%!xV)d`CBa22UGx
zV%JqnhfM8*VMR&m_<KT_HQUr{dJ{81Z?EWfW8JiDvt#LYTi>)yeMPj_^;%1Bi^ggL
zGe<WXD=V88#!c^tcHOkBc`j)c1KiT>OQzo1T)$|F!NRVZ`qf&?rb2DZBU!4sv56UX
z?X=a{HVwVg7Bm|yE{$4(qJm1jsuno%eQlweTeWpe;Ca2lDTR5TrN3)wsjvODmceCM
zCQn1X-rj84mR=L4{*g0F^JgBIKfCk_mxr2U%U2o#y>Mo*7^KHsZWw4HT7S_*e6N;y
zzlt_$Lq;x-su;@hKz|m$Wh>mDRNdYjXP4x&J}o;!7N<ARelEJTimw_v3>Q}$_9gw8
zVKml{adZ4<*+Pps9};o9cqpk`FI;%~S-n&1V3bfIM%kFh5F2{C)w~9R)_Z+M^);LA
ztEM4abG}m(wRIZdYA{jG4-I;6e#m*$knQn@l3PMkZ||Qs$kfe7;|h9Q-|P(3>ouz?
zhw{8`;>qz0z1b_Lo;8n&)9WZOBfHjU`Mt2WUJj|Z*Vk*OEt8(mM$ZLBHCjfa4n~EZ
zroN+xl5^=6Poh)TdaYg1ALmi)nDqvhPrYVg*3!!2AzMDCU&S++`tq%c<8kk;tAPb7
z8-)o<E%b-@R)zIO-|Jjq+MA+vo_fq=CNMmNeyMxbqsJs<Ih8QY!R0r=xD`QUd$L|@
zwc55`ms|#Gwr*pFZ6D%ZwB*bMZ)vn_bCr6G*IAwN(jV;$U{uNr_*g}dRmD+WNq1G+
z7p|*6tZjr}P<~3mHWWe$*nzZ#C0E@Nw2v@(^U1|a?R9ezZ7?sk+m{!Onq6C5746N=
z;tQLtXV7EV%36!{wqY(_0AG9W)6Xxi<EFW|QMH)Ch=tCzH<OPLd&CUsJbIeiLJf;5
z+xJ+V8aA_FB`iGBY}ac|>rq_z>EBf6x2L^s@FumtxWMDK(_T-Y{cU{}#qwV{?Is(R
zzgxN?^oyJ1^={E(7dfPILrt^gMntW(Y9fVcyU7pLnw#d6LbQb&u{zC$?Iui28j~ip
z8?ALZX3Ibh@Ij-Tbu)<SwY`!ZTwIlmVOkAg8gj9@YNHY-K3&Dkb|dDNEoyFpn^bKW
zTW$huqRP{<k{Kw-4O6a4YDLuQNI-nljny{=<*KA;FKEepu3GI4oG8!}=7#!r41v~p
z>Y?$ZnpBHw9-pGBsadV4rtu%fe*qzJgcY>78rAX~mPKe5p=ni|MDBWs%D^8@t~fqV
z;A1%;*K}Li!Ir<Pu4#_C7BbXNs*XCX?1XofZs;YYwWKQaFH`MA?C4r-7h=h%)ZeeP
zv=1v=hwZo%T1#A4@vU7~Ap~@skO9G^8`^jg8+8a9xwcE&a;WwrnEAYNpQQe++xit6
zQOnzd^)sh05L@b8!+s%Z*Yp+Uau;iM{SxiA%&1IjfMX(|<j+HS2NDGnm{@>qwk(v}
zE00zuc5SdeSu1XM`0MsSAs0ovW`McbRqGOV7AmS)6TGVFo(lW`*!nt(r(IsIeJYjP
ze=B~~%cZl#z(NEvV*amU=R?8419&a!F|^*(aI5r>aBWqZlk@tSd3|AF;T4YoGau!q
zUaDELtzKV(!9)%7N`s`_h2?U}jo58bzvL!~FIIU(Tn#~G%pf<i(rnjkEH}%()-heJ
z)p27S^scj_#0j`1Zq(fBV9zaw-4OcKO$5*2=7ze`4GGg4UK_=#HwX}_?RHf@=TkI~
ziTf!DP17;`Pw~PK4<V#Z!Tukuq{r~FK8Gb>DizgGDw?5ILdJWIu!(;Fk;u)k5jA43
z0mPV50vR_oh8W>WEVw725eB4r!c43s`I90#tq?V`C?RL$Q9?S1r*s*a#u!6<!5Bvg
z*?r@S#suPX1MyV$B=SxfCB)@}xG}UGaVuC{X2DFEefM6Arf)`#8Dkc?#yMACilCI@
zK6&SE#*IV9VdR}K6FZZ;{EYo|1SRaPJ7pHV792q-x^eV1NdA?nZ?T+X#ywn4={Ds=
zM?6ulMaOR@jC+j}T+;NnT+)5UNiJzdmSkKoquzhUjLJTs`GwKJNO|K#eZwPX+<!A;
z%p0d)i=cOAy%L_}emzJj3pcaIqOruK%z34}@0%*+jB%DrIpme{R8Y#DJk7gqrj2vP
z16;ykuY~s>xP%9dhq#0zUI|Zgf8I&`58q4~?>5eJZr#iMJqH+#??U?@F}|DgANBHo
z;6SxMYP^SYAM<iQbD-Rh8IN=BdyF`?$7kKl3nXtDFG#^``=@!+c$#1!b`_g-3A`t=
z*Z}N9=}0K{^%VqIJ8hsHY-#$_7cM;a?1KK{CCJwGX3eq)3X#mS+A<9;nYP!Uj3)ph
zM7#OSg=gg+!znJ=cE>uuxL6ZgjVlXnvAWnXRu<2mSvtG0^zH{AI=g^dPUuF_5VI!S
z)Jg-etZG5~(R7o&La_}*Snzb?erw(Mm71vXCK8n!TAkz-`lD}>L0N7&j0*=cOBHmD
zzBNY!4M7Q^RKdv6b?S0CA|63Hp@X;uAB)l6C8aO%-4y#M2wD=h%;t(r9|RBYKk+Iy
zo`zkm8pvaj5T!(;5V4}__T<1YE%*tSA)qIE6}|^K0?khfH-R7)$D={QSLk6NSSf3v
zHE3QyBM`^JUCqD)mLNWn@C8vSWQ0Gdb;A&^kXq_uu4~<hoirkL%1%4rCD&ExS|EI>
zKoH%ip+ReN*3LNK!O+^AfYv4s4NV+6o2XcGqB~gwS`#NWtCXZ>CDDy-tU-U0bE42D
zsh2hDC+z&kRMbBP9ZYnsV8qaZD0D4Kb(}(b_2RBbv51*KyQ64#5@9hXDcg<kNt<e<
z=bWTnmULRXlzxiBITX7ug&Ii?FClDdZQ4;+(2^UfwItiJGec>2X#&h5=Zq|oo&X_p
z2stzma=4dfR;D@9A7Y@ToKttuUsPjBDMFD+b&y^PWgew6@u4;zHNe~Dw`}V!L|UXd
zM*YJro7z=}^55f^%PEq3;*ayHPi_A!uOPzcPh-LXLPK`ztxHx&n6@c^vFr7=U=$1y
z_0?dey($+h6%ds4xLCE$2X_<|B-c+8mOQUBmtpQf+OUOWkotK>%R||hp*0CoV4|nx
zddK2|lEzJXcfemB#{24f7$HNl4A-_A*7os0{9|<tF{pknW<rWjCPW6y1X&^mafRhK
zzJb3UX1rG<5G@z-V|2v}{sLV+Nn{a(pw~LMpuspuOj1y!Bk;s-6jDF*9&VVRt2jhC
zh_gtUHT$?_nXI1k;3Y!=Pq2OPoEss%u;oTj5X9Vb$XChMnjHwzRj<Oz_~3oDBg`v%
z0$U{1o}pHStqpYCIMM20PWKQ8qQk4}jn<|$935)5S1WTM)qRspa|uvWH#)<)XM0iW
zXNDsS*gY`S8f{SpZ*Gt(+dbS1qdNWYYQ5P2;5OW1FMi#uwE(Bhl`|tl&z4i-aq4nn
z>uxq+(z0)J!M*GbcWNlW?`bzn3cVh(7tc{~1#EF%Px=YPQHscLaef$Z3lG~cyU(!p
z-`|?r$Y?zVqND-7)ij`XHk?-{H3`7uS^+yj9{*8ohObI$NiBv5hSUFKcqSZH_0U{o
zdv<WI7^vm$?f?#YcXy!u<B!+wAOfa5yc>x3VND%qGZ2LNm&N-*54IJEr;(+%4T$fd
z+Yb<VhREOsAfBbLWg_1TQc2jgHroO%@j=|X8M+U)uu6e$2f21-YHyg_14cyz`+#_!
zs^cUhwt~u0D)g4yf%rZu07j>F^;TO#<xo)4sBNKgWH5T~)-Xv4daT~x2HazV;Ul+*
z(qNjtZDPNefvrNkj7Jo&5cw#PDiLD(;#DHV<wcFiMIwQ6T<TcInC0@*c~;$@L#l5N
zb3ZbIx%bxus1e)pL?qRABdvA=QU(KqFd9LB^c}&6kq%8!2>j~hxG1m&7Mj;Ca1Lu<
zVdS-r*&&GK1QsH|K+wuwhq1UwMq6bq8ia(3$`qj@t@c~0r*9<*FaVOQ>s*KK(ld9G
z*81Wm*(UXklUr5W*{Uv;`vk(h!ld_5`bpLkF1*MpRmhZVOSkls%TK)jrLvS^0Y=%n
zLY4_AI(v2Xq*iL2y{D`fJwq)@D>Lft8Et%Hss5_{Doi@!3RK8YX|kckFt-9Zl?tXm
zGCjcX(<e$3^fhZxEBkADur)kTgDrx|&2udv*_muzD^R9GOY5bTk6bf6GwQ2e7eP1=
znln<cMt=`hgyLWHioK>c$T`8&($bSGmt*1-W{I1GUBqZrn~lpR%af9Uksef#HA%b(
zQVvUTYmEXf6KN6|S|D-X67;_iX{AAwLLAnIJcOL=9<b{tM%h%X8$$bK*#px$hPYr_
ze-<~Gtcs(pDVW508$JOxAuWRN2=gW}g^*`~4AE#31NBm9#194PrTb`bA9ccOa7_U=
zi^%V=5<(r~Q%-c~9`#c?)c>O`4F(SKO9(p=C%O{irSb~^<-te|E|q}m4a{F&D?ebL
zKGVf|vFCNUXlRh=FDUBH7Y~g(Km&}iWS*5H?=K~?G5hNQlYVfv?l*;4x5LUH$n@(d
zm^>kval;2DSfzdD1zJyoMUR-9eluuUA}=O7GVQYx{goH=hU^^u4wqt(ZRE-chNVgy
z%AQ5^ECtfTQMdZAW?%#dA2mo5K%6~Ae#pZ_-(0byw#GeGJkws4Lj>q`r`1TL9UJBY
z6)ZdP5h5QZ^8Fy-9*@&Ca|8l8ymZ}c@7bv<d+CRH#q>s>2*BI8F9+nY(^v$zl_D{Y
zNf;)9oI(;M6}9dA2Di$6%L!C~cR7f^R_<~T?A-w?@O$OXHuE6cT@J!(oqhP<J{*Jq
zm$*d4B(efBnuCZA#RqI=m~9;DU+yFz53r60Zo@k6-=B5tDGOprkFABE*Sv+a{V>|h
zVo7L^SVFS(pjhIEK(;Og7KbzR)QlWi@cIzlz69b6UO`iYMNvb7SBc;u?e^6z>1V>$
zcC@)d9TG(jB}%+>`@~=e?3b+^6=PHlNqTM=PGN?dsIRQTo3#c<*yv`9f@FVi;YgOm
zClJ%NaN|S_<#OE0J8(dS7<CN77Ws?7#ncJyhPu(#(ojgO?>>SQahMsxaL$aAIvoNW
zOa^N(TNvb`+TpZ7EGR9Cw24;OiIVFm1hCMuhwi`n9#_4B4kYDnB?X_J=Cvvtjw<<J
z$$Ngam6G3q=(H;L-gXsYarH_$DfhXcKg1@A6IX~_1##0{WVO9%cQ)-xQr=PjRJ<73
zTW$#6Ps8Ju@oQzte}x8%ym^#(UM+-3DVW#N@XiTq!^*#beC4<tEp8YO8I;QaFVJA4
z8P%0mueEZ_*aU@Ys4f~eMvN`gPII*zw^c%p@Z`~U;nr(t==+4D>`-q*q=xAxwj1dt
zF#>S*#R$PX^T|**h5HB`YY_q?Q|RSTH{H#2GdI*Na;A-}!JQTcq_1@|Mg&HrY&VCn
zxDmxzei}Iw@a!R)LTGv?<HX5R2#7W7q`JA*Gn_Vtw62}q$?YmXbdsESck%|(;VQ_3
z^i;~2lS1CyONya<P<dYYu+mD}1uBDGh$Bwc$vKJZp>E#EV~!?nsNy-l23$Yr<aWlL
zymS;^D|S(uGX}@tl#@0R&R9{wb=pa-6?Z0f;osCMIr-6PHeP*B*?1LxOKEa}*qPd;
zHhWhq<ZGT>W^O5M6`>xx;Wb_k4artzZFw2SQApk}L4?HW#TUJK{-yVnOYBf1o!!C>
z+gukvfN>Qzkqsi#ATDg99r!Sl#~X~pye`}l;B1whc}cslYF14wZn6xy1->%-Bi$%$
zG{_+4smxPUX3A`lAAEIH)C@N(;B^E~IE!zql{o|M?Z!Y6Rr1P{smGVwb^wn|Lz+wC
z7i;y)S8D>+6)rNeZHji4oQiCKY&TMG!c<W%^exI@+6#wT)g=-vPwtmW#D*LFRk2MC
zj?vnwZY{aqQe`pqKFaiQiVjmWP8pa@c<=Kyy!-rXWZScZC13GYT9=maV<<;SM%5y8
z{8@IZD}>3#E(we8q!v|^A#%w}t0fH-*J*r5wRjkAdB_3#Kq$Ir!4K53XTi%S?>#j4
z0i$StspruovU%}R|7~1Iw>+>@LcGAw2K<_Qdnz&6n~b61dKV{OdMtBwEs7-*cVe_;
z#LHj^F=oFZX3{{vyD=>61pM+JMqJ!u78D*~<^a!sq)aihpv9hal6}k~0d|qLQ#)yg
zmObpvM&xp6C$p=pWpUC#T~%RGrWUA%O54F_wNEKn><K4}Jb7XlMa9YRHj#63SX%LA
zxvDhZjX*lskw+vJ*5GcsXQjziNqXY4%If0s7@Gv8m5*f;))_vE6zni2@zM{|?a;cT
zH6qrDoCa~jR~qJ3wiB|pww&L89uJPDo0S`j50b<WQJJ42LbF=%1SXw~n{KsRyk^#G
z)@3)k#hT;cSsU+oKD)n66ZRPLDH$?9(d3I`5+?E8o4iTwgf_WHf8!TXA34W<2SxsI
zo?{Mt=lbWF_%nE5@g1ID;zy_vB_hv*-1+npKT3J_n^WS?QtY22G9-AW{7C@L`1=Q#
zO4EaKXkRJi#E>u}d7D9%4=xEQ%MeqezZ^aRgCh)qV?o5Ij)NINt|6%t>?h9Bi_fSa
zld?qGlWItcfqc%4N`+LDv_m^!ePBsxgh9z?z@26WaywX*vJ(Ld%j`m$ak4CT?%^zw
z2SvpYXXr7778+X4?au^8J@(Kj@uFuV6AOJKhbbL&FksFY)RR)bGCQ!;qfpeU>B#oX
zK4XKd^^v_zQpVfc+uq^}Eq}*)T>Lm5dSqYc`uiHo;>p`|_o4l}n-&$S!5^SZ7<}C<
zXUO{*;ud*jyD51oPYOi%NCBeB%twAW(m5o~l1XEI)y+XQ2d5yMEriiyy<@w%mbnT~
zz+M0}@Gr_mMgoMJt+$(S-+&t)yVl9|mf`j&$|An$ngt3QDA4x*apXk)uY}=<rQXDv
zF+#6}X~ke|kqa7Rb{tIve%Of+!dQ!91w~#CcXM4x(%k}%^#M7G6+5;DM>9}t^F{=#
zaU3Ww(H*yAKnvqolSRY<F@UDM6ccuaovd<*owW0)BZq^DQ}&oM2}iXlxR-^xC8RGn
zxfK;*aR@(I%5Nm#E;o&<ac3Iok|i8FgrnLFa%P+{r{JhYdIn|(qSxUJ*PR8jIp@q?
zP;RKNoI=~@x`(=lyGNYaouZL(=5TB=>l^~jIfwcB2x$Iw9F+uK8HdX_ge^&(Ip@&F
zHRrH%<l`FkSeM+$2I7r0gquS|DIMP(k$0T_DEH)qb95JC5?@a`F<j+}a7$ZN;X1B$
z34Mom2Ghic(%b`%Hs$De{$t>4_dskr?i2@G(%;gEg1$Tc5R4>`D9*&K(-}B!N$EEJ
z5NN2lQ*tJ_9~8LejXN`_rGRm{7gsY7{LuHzVNakvGtRxv32@kP<eDY4<xI%$p>dw1
zQFbPsDL=myW_b=!+<k*H;D)yKyglcX2$SB>Ubz6l^<=jk%!~4fIdPwJ5_96t=EZ~F
zyqLc`FYat!+z4&tZ#grRwL_Q*m@CFO>lRLdQ{*uR4r5*)G2k4JZyiT2dJyJ_QDg{n
z6f5d}`<Qb-&9k+80AX-t!~rgj16<r|%o$05i#*4vG`=3Vn0HP&^WZcmocjq};Oai-
z6kS0ZvS&8l&^CS%?Z4OKO;K>4)0j<k?3j4VBsj?g_{Doi@e<}BC;Dc}?YIN2<dcrR
zi`Gsl#uN@dFC3U-q>U0BVk6>L?Xt7r9EB6`LTdr~jcE@3f_*9oEg^J<Landb_ZzcL
zTDk|r<av5<Y;VP-m{%-}<`uWg=ktymk&W}{<I@56I1QQW5WKZr^^&Vz6=WI~e}Tw=
z`IsaY`j9Ex`e_jCO#qQN@D53ibX)S^I~`c35GOQ=J?nA^TA=WbCc}tVdAT4*4&;iE
zWZ9Ri=|qr=!zdTIP-B+}L-REcN9q$!1XqUvqN7L^!#~wedQN<=gek|y55A;qT)}fj
zS)v@!4?Tm{vm%FaC<NudLIFQXBzSl?cBuhVK3-7}yXsv;>uvb@TCR4f;c8bK!y<gX
zYBl*lEHpDMqkYx-L?E!h6EP1BhNJ`vlj<>b`@04O7HJxmYMu4Q7P(gR4tEZh4LnT(
zLs?^AELP5JXJvWPi`W$K2oB?HL7h&H*|jG30MeJ~F`ptrQU^SHlPe9JqvTV6Zx90A
zo_kK7v4?rPi6iuM(wokhvNOubH&uN*RMzobMg{SD)zhP;>eaevTUO1sTYH4bJ&FOe
zBY=5QF`#mr-awbZI~Xr-XtdDr^4$*Nb!ye@CLFZj<3WaamZx;cQ#v^%pVxiu!aOVP
zL<s3cok|L`Qy1y2-DIGuwqzf$){V}x8ZEKJDG2Rb?7!i%w#<TKa=W!Ey_mx*aIR%H
zV~Go6T>i=?&Qv6b%6vH-img(wiw3)<xKlla6J*AAOR5ENo|W~5O`Olb(GNK2;=ngW
zLtn_64-8%7N>%C|3EPN2P2>;}Rzn`9t9=zHq#}_Jme4D!P29p>6{1K}Y>LRUAmv4g
zpMo_Yb|`9>$QdGkk;q>n@)Jb<G7*N7tS0*$-Tn*_nqsbcQR-5HwS@!I5T2n`fx}hx
z;^xW<9qQ#%^=^_JDXQ{Vaf)ZFl)T+IRIoTnjKjs6Z83m#votHMsvjZ#8dVm>Vuho-
z`0IRw1xu%n<1E!vr5U~o1b8X=)2%1@dzJ+>1BiedC^bTaP25~xgQ{?ImM+X;f(CAw
zl%9NOh8?pR-{Tnd`l|nc+}bJybZ`i#Im|Q0M=bhl#5Q(NvywlSPebjQhBgT*P$)MG
zuwsx#X$Dq}8Ff-UuIbtm!b%~0B8AXNbv~3t4&;HFG>#bL#CJ9{3BA+gK6w^64^A;7
zUTUsRXmjLV&2{G0ld6tVmPpBp+S6cT537%=Np1V+Z5bGje5cXRbc%^O!RqnXZ(?ro
zc0_s;`lA&~42-y-_31l-O=Qy0Al8n8O*mx(!GS>&gc6890c;|{BjMy2HWdMzk_46j
zn~D%kCJ>jBqKTJcvQM0z3Sd(SLh}@0Q;A^{fun+x2W*;R*py?@hG5eSu1Jh81>*56
zME4@L!C8pzMU;{SfEt(J2>{TCP;(H89(3kd03uPD0aRCa=5{6kp$I%>89e120#8{-
z?}Jc-z~qo~7|`g5qk9nQC~2wtAQpos37se%iyV3%gp%-R(m57Ds40iwQN9mC0Ukl{
z8cdTIN<+$~l5-S&aSx0}#{r@49fC#!EoHIjUQaBV9Bgs_6H&TA3{p435DzAHrX8GM
z0w7Y>AlN#ysE0t*30%!WP342A`*<WyIQP*=_~6Nz+ykO$)TaQHN`O!$Kfec}{J4`C
zdBrKE$h6Dllf8h!b>A%Upw)CAv?_N`1#{<=vv=N{bjp3uYWD7Y>7Tj7^T0Xid<oIp
zn{F{*2wI(>8RS8$`=NpvyJKiI4`?+{vyY(_u0}vBitC3~rvb(Qj{vRa18Bt{3(#uM
z90j!cBHDk#<6n4%M0WwRDe1(R{}8mAWN0;l#{|&I=PbA58WLJTkrF_w#RGGaj0dg0
z3~06J90Rmk1hmR{(CYsJTKS<Rgc7uhT1AFdnE+ZX4DtfzSFHHi%e|O~ERN<Ox62ps
zl*q<M(Z>q`Pg(Gw)hLAVCmLRfpTfQ^2u6y(N`$01mfnUWBo-*oK%ke3_R&D<eYcfE
zNWf+x<lF<_BBm05gIfAGiNH)l0BjggiJzs&p98ryk`jN5k`R^>e_NLHcPQh})AemF
ztNQ^c@e5SOFB18?M80Y0Bz}pq6LJ!Nj|e{{+gtuV-IDkvewoNWAo351kS1IF3XyX}
z*m_1NN&F)sze<Fm%s-~9UnBBQh_Ipe*XfGvAKx<25&x7*8A3Ko650G$iu$&|8}ZMm
zsQZWn07v|Dy8RbKeuKzw68SA6|B}eRBH|GFJQ1Gp|C+7{w1|I0<lhqccSQa@k^eyC
zKN9&A5w?_kfv%1c`A<Y#BLA5Pn-;%FSO0|w$;pxwaOEBz0W|s+g-$DsK%FvnE6kC;
zC4v}29l$4V5$>2j$a=3+Pte~0V))oy#Bdie98?1Q4kCs@1bG)5c;>i2Dp)eKarhmM
z4SxL{i48tK5*vJZ1U86R|L;4F4c>nj8$5%J_;#k317U+N-Ub^yamTX4U0{F%xVN{d
z+yw@=w0peMfWho7fWdFj+9IRRw-FV*Rkn!o9k26pS}+dWAy#xC;}#nr2wBqKG(Jgu
zp1{ZYI&=Y0Ad_Dc-rh<_naS9&6JoEdF!?<4+gEWOgG~#M*by9<4#OiL1hrchW`%ec
z-cxY)#;FS^=5jWl;(!`CkJqi|@LKsi8z9EGRJ>lyQ?F|{;Sq)E91h#;p%jIOQei|I
z@?z>wGmiSRP6j539JR|y<30!DL>S7!$Qpb&P#;bU!IumE7EpYT!2~hE{+S9)7o;GB
zdJ%r02^b@iM)H$;TTk9%d7KW(!%TsaX4r@TB{w}8#|=K$R>dB7#`t8*GJJ?8p$zw1
zI5nzu4qgb>-@>(G&??!c)F)%zEM80hK5V%(5))1q%E&nUJks#<D8YR`1BIaqzl}*!
z(?hvff^sosWI6N;HY@Krl+xxX6n+~eBk!aF<zflSy)pc50bAmwHb`YX$zDr*Piq9l
zO;63UkMamF&SO2n2&Yw?!r~}4yI#fAlgBg;to4~@`}l2m`MwIBp7MrOJFNZb>39A-
zHR}+OA0skKA2Z#jkJ+I_2U>@$i(&-~Yg~r(ESppwcTG>@a+JbIVd5tCwZ&#VPUxu$
zuY|F#myt!urE!Fe45N7%M5O`NGr5Lgh(!o_bl|)RjIeBOh2a#Zt?^H1P~iL@D)yO9
zx93JODzDPI&s}CwZ@^s&un1!o-{meoZ@e9MnFL=`oIDI{9`|7bTW*Nwd^F%W&x7Yo
z!JpOVIq=om%Ml&Cj_`Io2jyhD)8IQVgYQhk8+R}7nF)B$m%w{wnD@-|@t*$;-s6Y*
zyhpRLL%e6^F7J_Q<$Zb2r?G-Y@}AkZnfDB8Y+VgU-+qVMHl(2)8)QG#a}WFqaUgQw
zIXDM`>UDqz6>tb?fC~-1CTUOIcrD-#HS$1wndKvLC!)U@d;$*k2dc-x;4wHtK1yMo
zk_kM`hvMQw%10rr6q1iJkdI>BOrLy|4df$Q(^<$z(m{j#)HNJ^g9jia9=gsr5nRFJ
zJCKij8LB@G$x!`i3d{-8kdN?^RLDoEfE$qHl!w%l3pha$^+?}#c(dYkOdQftf@PxN
z7H1(7<shfzASL$8MD#QiH!=7uTOY+)+dRue#Gibbh*~u@qD2z}G7+_EoV{eJRkBTx
ziBL|aTLOO~4^Zl#BF^3P<f3Uvb5bt)C3^z?>tRSh)8I{btvbs^p9jbCLrVyia?w}p
zDab|1KrWho2lJ-VTO$|sM52A!(u`zFJ6JXcTXhGQh6Xv3_&us}6iZ5YY~k#=bOfx3
zbT>-&mcr0&M4`f<C^R)v6gqnj2yF*d!8y-Af>S-ye{=SZ-4O>6+ki}!^<}Ej9O)g7
z*uar|l&0MgN*m-zK1yRb$}T`o@^IWNP8nTSQ+zrASZ)BDf?pLq40k8;6wa6o$y2~+
z9<mHDTp&+<xKEy%9KdotdFqROl2Z@Mea$Y7kf)ON^Z=H7C(BbaZz0dhVI7JiL>SQR
z%e_3Iik7YTeIj2Wa;ILyZ#@e;_rMFp!k$B`ZX-?YVPS`FCrgcHW9J^I`=StaT_6e_
zgHjD!FOJgw$q|KE?3lnIj!Ae-7UZF|A#v_@!qF5&F~!LN(s`_kAsyaY&@cXNT_RS+
zBGb#LvxI|)1nEd*njRrCLCMp{f=Fb7ID|urdqpM)SAM%8JY57cn!pQK(p`H0Jt>!b
zAu0pq(rh3ueI4S`EDKSyEJTsGq=_Whlpjh$RGvcd8qtz5))S((-n6G77UIYqKh6Ju
z%1YJ*brs&#FGV)0ZNjTf6Bkk21lKmvSKAh9^Fw>J-59J5UX6HnKVAX@b)P-EGe<tK
zyP7dxRCW&Gg)4{gIfD4{U2R8)w;=rbP%nO403thmD@KtIVM>b4VI@Q{_OZ2lc8-Jj
zmf&wZ@w$f6=j?mot%yU5I0QCFUbps1W0DSe%3n%=kKz*gb`}EU)E^1O+V+KS;>X-p
z8ms3P@fxE=%f!)+v+sJ>ft|9p3m5geeo4QopMF%o$QSiDz5y=Mpr7E_4li_CtJ`r4
z$7ASij^*-OvjKbt70Net+>-apSK{D(GWh8e{K8bD)$Uks0>8e8-wWg4iK}-u@t$4!
z4Kcg|iw@B&pkOz$i5K@j>}qYCvgy?9OKy0r-Drt8#U<#)QSx>FVR}k_QFQsx+b1ED
zNL&eREm>kvR>lv_p*dA*Zxye?XQ>%KN#}<RWBR-&K%aY{N@=#1UZ+in+}!R=j!px%
zC;d|#>`&FNSj5>7jb47=59|x_WjOMj26CQ+3UUUgH}I=<+V<g*@^;X;V81G%eZhY9
zmAlYg!t+}}_aQX|%o#p(Cv(OGj!*mh(Bz1HDCbNJ?L#A=JG}yIAJ{&HJwU?tp9gHm
z(QUx?sXo~Lo7frr&>n36y&>2>br-e|s}y^%eP2sP>Fr}mfe~XQSdU6x^Nnv0QXeu}
z6i0Cmn6lmtP9fjbgP(f}FN>SFUuLJ{xBc$8i=T&T)o_lm>L#o7enR|OIIh#xD*fcM
z7ZbyeN44wKD*p_PcrP`dAx@pH8bnTkxKa7~bb;O0b6SSf?j5}pbj7a9(9g=(QAr<W
zVwwDg8aLKz^PAbE0gOQtPbJ<_f18qhiO7E=^1DR-JCQFFd4tFw68S$wdP-uxB41wd
zFv!dJSa@f)5|4%>3YdI0ntvkyR6)&`XiLDRaBT29pP>6R-T;i_fd#svdrjUy6vam4
zhu{}f)PjEN5Gn<!%KfQ=uoPY|f&Wu#;ZG^k(X_l<pu4%DyJV2^k!bJw(P%=3pHK_u
z{qU1&;gP=Z@9SN2f0RpjrQWp;9iQbf<Kg1*!Fo^xynW@!ivLMhi$r2X{(#8WK-|!C
zPd~%bN9AbKSYMFy$65&9aPJ4;rTk6))gw2<ZxUO8>5ktGNyG1^DB4=!H-*cSwPewg
qKY8&GUD4o;8dW#`h^*mJ460(4K<Er6{eCP{?B!`SABjI0pZNbK_UTUm

literal 0
HcmV?d00001

diff --git a/model/__pycache__/spaced_sampler.cpython-310.pyc b/model/__pycache__/spaced_sampler.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..523e4d05d04a5067fe9b6dde7540c3fc38ed69be
GIT binary patch
literal 16422
zcmds8d2k%rS?_y#PK`!mNtS%Hy_d&pOS1QNHk(b>WaG`QvaC%4$}YX_c|G#TbLn+U
zmX$Xo6*(0GYzP&?KOrEc0u`qyt^nbv;;e#ypt!4`DX2gS!u|yXImjlv=J&mx={am~
zj)Fg6RQ+}LJHGe&z3+Xu+WCA&!S7+^NBs{s6y*!_GW^r=@;I*GrvRA3)P~YjSJbAq
zqBZpuy=kl%O>@OmsZMKHE0(;cR&3n$M!K0^N%!h9D;ZgrUCH8ZG;+=SN?z8PjY4x`
zWkNn%jbd|hWl}z;8l~pc%2czwQpUT@($|!geJsPWxKBT+upG-@Q&@gSTbW_nd8JZ#
z8FN%BW@Nvk6V}@;uMwFSHaZPIGS1ekVPrgA55mZP!fQ0Viw%FdqDDrz&YHNJ=bGNC
zkB3szyX?Cc{m^rR>Ls6TG^o^ma-+_?R@JX)(exE>({K2p%l&G*6@+}FN`pm*J?3H9
zP2a7flp4FOHg9^3`nE?6sv6C%_bcF@!u15M;AvG+LPe;dx})*82z6Iu>YmCU4t10n
zOx;nL_5p3bvZHpj&}6z$cC@p~3)5XaH1KW+Md&;Fjw-av8ea^p;WAB8mGjE_{xHQ%
zp?q9n)+f;N{VGc_`x?gF#j@_HX9>3|l+X|uZBFR|XZ$>^bTLBLT2;EKFio`>`R&hZ
zUA=1yL+HZ1tZ<L%EtMHFN;fUi!bS<lco~t#cv)e8T=|4SV+xz;msNh1>D^4r7HOu>
zDpWSB%q!gt%XPCDB^R2)mZMB`^CGvK5muNL*&TI8*;7?TWJTs7rJG&<kVuIfo>IaR
zY22;#cZQZo?WiDxy!YLPqGTNWp5VS0`i|!`u=h^8=7ja8AB27<aKg5;5%^Axx0{aF
za@xGUS_jQItX`{a1lTdB!`oFq2u?^$tJvC>6JGM2)<*N9&qtA6@trGPD<q^LC@9A2
zuWx|XuAU6)+dgwTZBBF$NNh9{ph^bq4YeRE-`eJmzvVSSx+k2vY{~t50_VP^rJ?zI
zEhcdUz9c}W+TLh|f#Y%CdH1QM6Q>?HaeC<mi4P;iGq2S-R^ikr8ng!$*BPdW*<bWg
zKN#gie~30lLq~~^h8XMez;KuT%xZEyX}FtvoL-8*^<2CVF*>J?`AgKTRopx*7#^<F
z!%NOl#_G+ZaytEQ*}{xvKQxl_l3Pj3diLzOXPl1L!7c$KcA1^P7Bi>aYFq`t?0aiQ
z<24%XD?W>7bEe~QuSq-XCWLZkY|{ty5z|pUo=<-(<xGh5ZvOQ)nU<N=H?iW*MrUN)
z-n86!E6+GSz8v2$px&7Df)gN|!9sxzg<ieYUxk~;l_k}7v+11-eEPx=E0>_Ew^+Rj
z7KN3jwWE)UY3X&o#G$3_&33_gv)r|gU#;W#RJ{OHOD9Wi+45263ckU2mR}E@-1ot`
z8abfxq(GoF!~Pb(9<ssi_ZFA?;RbJ=p%wEb6BysZKGvJgV~<J1cp`&naQS7hYlRco
zp00YWRy%a6F_*!at%jhm?Spa^gP3x`Tk5UQU!@h3=d4P7`#+KjU{+w#kOXCbRYfQ-
zSY4GQ!jAf5+B!trbp=952NuX5ZDA==xA+8_O7O~)i<jC>e-ZQWFSgs47praNFJgZJ
zl7NfnA6dHp;nR;i5Eno3#QEnIn`qEj3_4!bXKo<b3}5J6jWUlClLVb~9y>|1p{bRX
z?K@=4!FgDCw9&45jo>jn^xMCpp4i^sTZ|rQ_m3Dj-FB|G9{7T`^F^SQze+C3)PugX
zjC8C{l#!=Cpd-(VXcg%VzZIF>Ypwcd;fGP?onB+Z|1QqkJTik$qaH?SAEe_GSx1)F
z>G&;%9^i(oQi$@X8q7T^j2^Kl&-|dyu@3UcMQS~=iCeo^zsU5rLheOrIVHEwwxTqc
zg)50O%8ww54C)nSYuu}%0ddnPRo&pSSC%9~tJhC;2V2Ld=nE&t)^`$x!Q%j#sf=1y
z3%JUvc0<!Ls&S*N=hU2P;ID{MTh&l++^}^^EoxMnQ%h<F<#h2SAgSsA{T~sUjq3?q
z!8L&GVyNsw@b9T>no!qtroN;Kb-%J}>?vLST}o?Og#w{MJ)u^+W@xRY_Mn{Hs`d9N
zp)K^a^p1+VwxdD~c%#r66c;EeV>Q8&&>`BjJrdWt_feb(d7{6IR=gTIo3zJ4Pc}QJ
zPM#%p)wv2e!o90bO>)1BURb?EQdlxDKN#VpXc+VDvA!dXhFHAkgnlak-T^*3wUFju
ziSej-kpxi<N040PZI6NJg>G;O5{I}<!{hR_(>o!2JJ@{_=t=s=Jb|#v`M(|?8}xFd
zCLxszbi|2Rg+v76<O6y$8Zf%S*T`8p@0O#vM5QO3Qzx8-g@qS-j9KzmIn_&E5YM%@
zme<wDV*X~G^xlQ#N;WdXHm_caGQ=-kxg(K=qSK(H$gDNmUWfx2gjYL$q_sLxN*3LB
zn$O7wawd`GZ*?GxD@LSaU88jJ4bjBdQbs!WgYnZ+c6*xuH9Ks(@pnE+^i1qROFNpS
z-mR51TP@wlX#8;$R&kB|XgZJM3cdsu=PN5JQ&u#luIOnkt+Pj%!AxfP220_xnNov}
z%re(t=CG`9p(Ix`Se_NG87rxDj-6o>zP*-aMF}VU4Bhv`V4*wume>^fW_zv5Lv3w8
z7q>;LeQX-|@@%+%Mz-zanPvOYqQK^cFwFZCzCrIb95Z&nFS3Ief2b^(M~)q0htYqs
z*MGEUJa<Q8KH`^N7gLcPWp@D6lyC2r_mppn@7A+Af_avb*%jHH?5=CD5?1#8k7B)>
z-6OG1$5_+3^f6o&_6BwwJ9#gz6Sz)dA1useL4N-tc6|xgsbnYV3;Odtjhc8zX4!r0
z{%a=YjD4b&7{dOYV=c1>*n<-5?Efg%huFgs>wcC4%Y8J;pO^Cgd@TOAKO-5@IWjrG
zsX}S7N<JH_364|~NCE3x#bg|H+R&`QCY^I<&p!Lif^*>#^tfun3j$KRNwTiCd?qmy
z%T91S0fm*f8&98oCgxkR#icOp1ZNf(J-$`nTxj#v#SW`2o<6m7dSU622Om1UfLGZt
zvKsZ4?{P@mI!pjJfQ=1HUurc`W`G-#yl25j_NK>O$(F2`MOH}?_J8zDx{j2JhwvaE
z>s-J!x@$rM)M4r>uEC(EU00VYCO?aI{3t+F>TCYQ5o0OO-%7Q_mB8kVz;D#z_DS(L
zSF>HPvwG;d41EF;(UxVa7R31W^vFgn^czO$WVGcG*FqMZ6B)1JyP!xa^d*ql%9_3g
z(;m=-q&K=6!`ID{e1=?ytk#+FlGZgKt<gF*_jk0e8D^LnX2YBSvG1rb=ONuyD5_nH
zX>beV1r9FU0#k4c*l+>Z!ra8y3v0K)V4RE8yfPab@9A!8{r+%5Sa4RTm$l&8r<C`r
z7`q4~-&&hwDU5ExpjX@yl@7+l+Z5HJri?jTm~#eYDUpfijPiXUjYr?Ok9<YsFxjK_
z(^M`?!MdrGWn}pb%BI(5g<7MTd@P>H?tW^yNBcg9e*0s5G<%fHqvXIy$-zOJ`MAxY
z;Sz><%ASV>R*J@(g+krOFpqt49m0u=X6P_5q4*xpbPxIw+8m+Pku#g#Q^)!r?c<d#
zGB?%V!E7A&?Dj9oqeF(^InWvGG$;at11YUuKZMQTc-3l~OH&K_&Wj0+y%?WVLXcp3
zGhYqPByUueG<S}Xv2w<d>_xH-a=9#IlGe{jlWeT}GEDQN1x%e>e9&>G;0cH_y*H@$
zJrm;J-z3d2*hYr8RS&lBOmqop^Al$pW?w#3!c3Zz?+u!b$%r8=zkCBfK8>=<oji?O
ze0HCrCwaz~>FFH=9tI$-z6sXHq;kz?smSbbd~IYweS~2h8Dtgl!_>ne5Gy}YhwN!Y
z58fY{<YEjW6CI%+y^+uqyhaBanA;m5ni_p~JKW#Asjx-b&>8A(1nXF&DT3~3&*-2m
z#NxWmdTS#Xuhtvw)#w0#d)p@aB?fpr>z(o5^Mk79sqxAZL<GpH-sUb?aGkuO(V;;Z
zjoB}EtBpErHx`u#^-bSv!7}q}r*2t#x{~E@rzL(yu3y53<F%bg-s9znX2uGVl`fEn
zXkaArcN4l2<XCSd`wjVBRFRkH`ejgf<+hD;^t^N=_a#sp9fJ!1TCM=LZoOt31+{2Q
zs|KvLB5bu1?%Ml`c(&9zD6d6z7JntRpqBtk>a;p-7&i=aPPJ~#;>k8(``xfL?KS(x
z_WY=b7#TTAZS~HN8b)TTU5DLD0<Zt0XA_qcenj;U1L}st)6(RFsn><*lja`0%fX~%
zlMsavBV;gYJMdAL6?m$ZH8UyE%gPJ{WkvJ;q;J>}cSy7gX;d{~B=!8QWH4XcAm^yF
zer(GP#g@BN8DbVUGn(EvIme_yvT#A#Zs0f4jUPD2mY;aryDBk%gN7pAJIo6`SX6_t
z1{{qh7v6;HLeEnSY!#^r4ZKEuZ>{rU_#%8td=s`RY$xgIVg6Ra!w50`z)Dx)c^YD~
zq!0NX?AJca8l8>Yr_mX~=husB5ML!{TkUG>nWR1Kf3nAVe%u#(O9aN9Oy`zN7wbJb
z5YS(Ml}S^$>eMMN(lZC?OO`7ZzYkOrW#E!wEw@p>47Q2_X=!nfo-P7ZjQC^*R1gx_
zAdq;iV%LYg-t#~GGAe@-fL+Rf_Dfm;o(=ds@Hegi^}9qmW<Em7DfCZBc^5ArP(^5K
z3JBMhgd1x?I+gVZmr6*ti%T>jT<AYkO2G|T3jK%P&E(w@#u_4Fkkw{<HxwOX@V5zT
z>>cA%JB<H!8edfiB5Yy`tES5X@;NAI$^MKUkc176P^~;6?+H(y>f)4yC!F}G(ALl|
zLG`U48QN=z7Wjb4D;)24pPI<E-9H5&^-&tV9pOkdU)lxW{7oRH0F31F2)ZR_bI5x^
zr)U)QEhw@-4LXeI3-U&%+nm4tc}rF!UL5CJJb+2JTb`F>TU?S(Hyxd4sE|I6mbEYZ
zMs{%YQ<FA{AiN3XJA0DzJ41Ub8GF+ZPq$a&ErNjlR)>joFKg)Z*Ky|f8i7>;mjEiM
zAwIB0rNj)P!r-%M6$kCdImG^TceH7;On(AA0U2P$a!D&mW>GXWSWK~Sv>_br*KED;
zYR25YXH-Dm7H6bGOC8}UBjS-1K*>||n|KNd!%?2nFM%+eg)rpj&>38Dz$<+e@LF6d
zwU}gi_<N{oj91Q*0!XLS-HIL3(g}z8{s$Td6vZ!E{e{@Rd-Q{E*Oe2?jSK{kSd^~T
zRuS{|5H7biT2xfEP}~2}010#5#)~kSvH@FFAVLl1HwhL=PGUr%g}$eEt=8;VNowfb
zL<<{+qJb!`O-2#eF3i3)n4OHG*pJAz&@F(?LR-X@&nv=;?-hNSKZ_Ye>LwPHl=LVI
zw}0cR8~Xr?iJa|)-lN(6(qXw=xo+Do#Fx8S$;9MI>kx~nE<ODKf%g*PoJ8riH^R;a
zOpy3Zu2-)nUMdid8;<W-zCWsz=C1*r4Db}~l3LVB)yru`)i8|L3~gM68|g__0<;R4
z9@rGO)~b?u-~Z9f$!wAO+a}a4BE+r@kp(>tj(lwov3;gtebaE@(~3f@!NgAOncWOl
z84-W%0?g=_^lldKCL-)8fg?bFUhn3*`EDL@b#o1ob(r|?>*kpWGr!QCK$*=f?A24~
zm4-8)U>2ph-Mp|VNC4fjAhO+w){<<SLfeh7usgA*{OEp)<nI<4+TkhbgZ6}FQe@G0
z;#~?uXw1N|Qk9l{N(tIm(3<|f4_C_!ON3Wu7MP+SCPaEi?-oT7bgA;6PsZF66T4HQ
z7-tWxmAk+$CXq3a6*-m`lVt_ZIgwo}@9x_J-t`4hymg!P1<W<KJH4mu&g{|rdrvh=
zKFKUgei&PqIv+jD*L5xxp`^5;Eic251<44*QRk}{FA&GX@AGe}%u0WH9Q`utq2J^`
zhzc$U(81&L01+Hn9i%o;2n=z&_!P`S$#i9mpH%);-}P~h$(D~w^2r~rj4W8y=)fN!
zoKmOD`YlR~aaWPi5fwOs4hW3}^3{zFFyy<i5fm<k>EhN?%k3~xUwjsa=j@AK_3{;u
z!zPv}&268zU5Y7$P<<n_+JM1XDGeP#dHSTf_#v9r4-rTXV=vASBCqBO{~#?wiq46<
zwbVbDQkE=Iryr+kgQ}wpqMb<Rk<SSC^`Fu0rUM&uk=`-ehr<*+1w2-!3_FbkOHnsq
zpXF{8OdIbO>@wqqrDxO(t{ft92VT|QmoamyO=So~*jmm&v<~(h^jUrOrei)b-pH}0
zU`BGR{|PT~1P<(x^iJrk3_**&kBgY*Pm+JhK&;KC1R5M^b<M(YutiE9hXow|lw_y6
zWV+Bm_fj~>X{6#DK%L!V3{-APMu21=w9HDzK!-acGDD0Z4c3qgv%5J##~dC&W?t5J
z^Lxq~e15AcQ&&~^`DhM!CI<gW^*QCk3J!Q$6wt3oETXK4ykz|oVgi2E)N*`!w7zFI
zj-wH0<3CK`M+jh76n+8V`lR&8aEA({!a;Vs^t#5zh!nM-qO!3AM8wWF2|Nf88Jl(g
zigeTP7x8|*czc2z&?1p$REXt4--W?<QOJLq0Fft`lt^|@lxwwHlK7ima5=KJq%&un
z{OnH88u}ihyan`|%Oj@ynxPx777SaLG^l1?Gt6OH%$UbC?Zz}HbKFwuqq{v}{R_tU
z|4vvUL;RWl3~k7#34DgYm~xfvuNAn$d%u&_x7F?aqdM0R##mc1Km$I-JHcV&Y=QVO
z&K59bT0!i}vn2<rH&>$xD8F#mBB>BAhh`^?CR+X~f-{2xSiu*fLbct1iw*G*8Re+v
z@QMB32|SR?Qz{5ALZ4+wEnyli9hZR%_H76<puZY8NG-!Gmbzv@C{oNoXZA}FyVAvD
z?3$!btXXv8pEpPi=#sERkQ6SXhr-kvI1gZ2<XIXDMX99;yITrVP!CGr6H}<m$`-j^
zi*lHkzV8Xt?hA_;aRRA`)8V8*rl*)@Szwugpr0Z2<0I+|hrqLDyGUnTn-YbZD)x&!
zo^vdRXvQ3#%3@ATi`iM^h2kn8o-hwwIZ+goLS++k@COjwLByhaKpcQd^D%tK!S12%
z;jSYN?CxVlaS&;Xlj0CyNgS5X4&c-$kSGZSpv0h-d_aq#I4BN%Kof_B^8t<4u6u;k
z;^BHW0WNolAhnZkPW&d@A0@86J1vgxfn&<&83Fz&3S~qNz(>nJsdewTMVr)En>)e7
zXT=eG{T<-2cY=rCC8kDaGCWgLffTvxA^0sGRYdvq+p&DyZvBH|3Q>r%Tm@xqPL#wv
z#@L4)x*Ic`7k5($o0f{>J)n#U>@8%=+Ci3=GV2h6<^!2^kj+S$br`F5EOf*%!nSq<
zG7C>^7SiM>q{$s@KUN_RX;PH6cpj1_6>*QKfV%D!$4Hvs=`L{(Jwe?qVYYXZLMO`c
zef<SVhbAcPI4C6}Oi4L&ppo*eC<L_e(IMLSX8TV2c#k-;hZ)W&Y#u4C_kP>7kYxuh
zYadbh--O4+z2YchZ}+xlAcGIe(l3OSq!bD(J4B_yzdfa}!y+3;<=|a9KDwK)M}LwE
z?!7e?e6xOis`!Yy{xNYpp^M{Cg`MR{y%ece_%DI&@LwkIHh?PG()}OlHguiG6?_r^
z+y|NhQkfBtBy%NBX{JPEDiQiL_||2e&kJK0nhuo_9rk*`OdRQDLQ@&;^`J1(OO{bD
zid+f<`&&l(u!;Kw^dt){Quw2>$uz~iW>D+($JWmX3*QB^Qz{?CO)EP8UfiTEE;W!!
zn*RnB{3e0q!=u!tI((!^J5OEdJy_K?0&GE~U8+agmHN0OD!M^KraD5oZn5^2;5SGd
z%>wHzEXIO53!4!sUr;G+^#HscupQqpDtKc5TdbQm>s#ct9;EDzW52G-;7LW}v=dUp
zex(rOi-UJQzpf$i(GRMW{S$hP>soA|$o8}ph^1v}btJ~gq^*~26m1`PHcp*}`>=t`
zXiDIvq(SN8KlY01Z0i*V_X=>^RkvpcWvgzLw}Ze7!`6_(_p>xHx$va*@#g?uhSwBv
zH)Pz`TUgmRA)o&u3SXJuKu8SUadKu#spvo|Iuz<V<5Z4ItBUJ{hZ71z#?Pb+A(MCw
zg7^dPlVeCXH-ChlGo--<(&ZgVQzk&HuiaXW1A|5l@hch0i&e=K4MI`C$f#`~K_N{r
z?mU4!Mwgb9*QHB8ni&`ZkoN6XY`7r>QCf5jBs(CF0)a4O>QXf<I;qlPU+SCgkY(u5
zn737hN!jJs3A{)^D!?DdQ&hUywBw{gaLOg9oQT!ogr@mtsY)u>zd}#HO5oQBe2#!r
zk)`4G>-74&1Wp4)>cv<gPL2sOTA|`VHAeixy|_`UQO;$YhnXR_vm0l9W+mChi4~Cz
zgBXdPZ63$s4APpAgBI`)(GUe9x4=!>{5BO@ID&{7^WTv#IEIuTYD?XV#?Q;AMB|F}
zGkT4(vZm+Y!Hl4{$+=1mKQhSJl$j+mtjT|l&}V?L|0!}ZE){f;r_vZd=u@{ir=Ov-
zPELa36S-nhgCU6|PpJLyZ{}cd7SxiCK%hDY@6nyAqb+GU)L5_OOoFA<G5t79QAz|Q
zBX!Og7!SvE?ZyF}j8@dEM^y)Smtdmi;4jUAr#!4aN}f|~`^Y!tCDjb(*Z+|#g5ITa
z`TaPR5clvvQ5q?jGbuFMt}azZ1KuW>sSx&ABtww8cqla2Ecl|x_oQ`mP!%nhlqr~^
z@Bxug8A_iL6g3!(P!|nRU?!A}35hiaFJr7SY4A50P=~TTmC1t2M2e&a<8-)<Io4(p
zVxSblZzMfiC!ttQb<3F3gvdg%EQ<`Di%2>#pqQF#P)Olpnu3^tzY@x=)}6)}C8;><
zgUXqLf@!nNOJr2%x-+5#pVN#KAJ5CVK*fM+U_&*?LN%Dhf4}755BsTbpO}THgHqet
zhq64%J{>}V?UiLwRzUdZvk>y|Z}o*dD?$aH!TOXaAqN6%y*t$#1!@lxUj`%1^hWwi
zj5H&$&LkuKJx1!6^+)>G(UHW=?kv;br_-Q@@86vhGi-8CW2LgPI}bnD0bB=BU)s}l
z4?&$ejD4Fy8`@{OLv5C-atxvT)pV)}owXyoN1?n=L9s4>LId`>@D8YPSTU$@bF^;Z
z-E5zj>n9tn|CX4A2%cW%zXx)Q#lb0h`h5a_K;VxF{2>97O8k!it{ajs47rdbWBOC7
z|1$!APJrYF{|kWa3uCGD6f<62MJjn`@%%>XY4BZSl(phi`m>O$Z#nneVy(V<|044L
z8v#EFH}PWNgH@lt@05EH{|G@TE8C@uPSv^OTyai5=3JBy)t7HzxH@H>Z5s=n*6Ow$
zK!{P?s&bE%O#CltQY4u8UlAY<DxGxvuj%Pw0)I<@Qb&da&fnAPKM)|%!y^KJL*VZS
z`~m@oz~>1hJU(_Q!XJlhKqTnaTkTE|rSVZnx0Z?b)y@Wz2Pg#_(E^fl521f#ZXkW-
z;Ye#k5O%!qQe>>P>n%<i0K^>();Q_Kpl{>vL}_I{mL=bo1d*Tqk2K$ZBJj@y$Z^5H
zNI-fF-;1XeeQPO@Qi1;~RlH0<N(BuL!gRuiSMoF+m*(ptAq46D{nv30jfRvD38&1a
zzqRCGvx4sj(p<hMwJ-SY!GDP>=a8h)=fZ|rq+|`%0PjXhhnjgcXO=ZM0`Z<xr@@;k
zwPWy}(~hBL4*ZxR0yjqjM&>ZWduK40g!euV$pH@Ve}?xaeD`kf%Mu&|eZD(2#CM4s
zPl6K{L}`fc?o0S?89X`RyWQOSisXS)Bb;`+$NzpJEQ3oEr=1qlh$Y3mb`M-}w3N8*
z&yDdww67GuJ<h|wO5}GtMpK#oUp(YLPa}VxKytn-Gyi+%?p8;RTOh~r<7CC8e3{PC
z<Cxy-pCd|+8k{2g)v`s}i(%T|0Mw^%z};ECm97Fy<}LK?q>=Puu5q+*S@QYHT!dWP
zmIv+9jWRAp4Dm0%c+R;l{kLYXCWTDgcGY!df**gHa1{w~0`~z#R-ALh@ihv+LO@y-
z+w>$uHDv8d|8neorjyD)1rViLZ5cD+a-$?m{t8vvv}X8k>EVT&po%F1Zw7cCS0K>B
zR=`>^nreJQLk5_pYp-gC_6^O<e7%&_o;1E;pwvLAjZ%1vzpl;qOAM6QRC3pFiENYW
zm*4@ow67N}?G+=H`I<IEH96p;-Uezk)MzQ~Yeif8FC(4#s`WVLa*H-aw86fiP425!
z<y&ZToW6?wj4us;nT0Wp-k9K}|7EAOucGBwtgQBB@Zc{Q`P_e68>or<eues4z4_9Y
z(0pyoh5A~(zD73pW$USLw@>lQawjUY2%lZ;kRc`sK<>TVaf*?1ivW=cf0clg*3uq)
zg<j#4!vD2Bd+zC2QXXlr=0Z$|!9tQ_)-OO>TtnuiMtK=*Sdf-Vt;LC;qAXPh3o>g*
zT5Iu{U!q|iCqQH~bPyAIitI<@5gx;2lwc0vfP#z($SCqG;GfJOv5=T(+7FiY{Wt%H
B;pqSX

literal 0
HcmV?d00001

diff --git a/model/__pycache__/spaced_sampler.cpython-37.pyc b/model/__pycache__/spaced_sampler.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..796de0de9969d0874ca73de9c1a21688c3288a19
GIT binary patch
literal 16581
zcmdseS&Up)dfwgY)?Qt#F80RllA@?Cs@2`(Xe61IC5^<HF>R@>8POxTM)K9|I=7qE
ztgX+zO*Y%NU}KtNz(!^)$4<NqHSYEj#N$11Y<Q3u-X07G4w5Is_rVApAc^v17<tk|
z?C<|i?cJnCHjFG)sB_Of%YU|i|4+S~&u27zp8bor{jWAO?Z47X|6?Na5`MwgkqAxb
z4XvqH^`=oZnr78(T2-r=s3vr(Ga7c)=4Y~+!qaS|o9Sx0SC^?~xGr1G;%PN<&3rY_
zb%{oyIaQtFyxk}^PgGBEKG`TWr>oP=a<z>2lt|y#swYK8WbvH&tR`|Ie_s>%eWN-n
zj5oAO;RDQ3t0dyo_jJN~yX7_F#Lb;f!;h`&^;#HPuhxSwPQBta8s2)tU#sY`6>f_r
zo{1YxZ_`Jn)bwuq?z$g(Zcw}Bi=76Qrarq<7hbF8SB!Y(u6M_8_@OKPTDuj5a;HXv
z#pgWXVc0vqTSqB1c3W-P^cwX&j~dhrnjJm`{F3;+f?x0jUDHA>(nEdUkoP0~z!3VO
zE<2%#GE3<Dx-h<M%xn94*Vug~Ob9d5_KoY>JBwX2wD4|4T4e5<`+8*DHssTxJzDmL
zw*B;*T9_1xNc);5?Dx^`OS(vk)P0P3fGs`LuM=!N(n2f3XbV~w5LJFg>jHk)-qgCu
z-G3dXsU9Ogc-!cj-Be^nW|X+C$v+Wh%MjMA)=fw0D20;$h7mJS8Y5<-)Yr84EgCgS
z3G=ot|CcbknN~4M3v*7RvN`RP*3F1qH;Zv{yGvmrO7S>OHy`B=GLao-qwKyus~zgP
z7G<N%vsyR%p1wWvrWPfm9Nv<V9i{QKw|^b0wWH*|4s^(SpR+VA<KXj(^u5q`Jf{H+
zaM~MASa144=yw7qY&$!F?`+6+)A3qPTh=%0z#K=^H#T+xoSV~;?V2A1OAKif2i$VP
zTfWoUX|DTn9N1mox$Ct;0vZC36jXnE2N-wnsi3~+3#ZeTgbe{hqnQBIV9?%B%Ut=H
zwsibmuL%TRa_Zbt`cDMTCstQS=I^xx!w7svL#Ni>X@!B~N#FV6mDQyypIo}S`VPax
zNNVOAb%|ZLG7bjqfy;G)DPs2PKI(^~EcJ(IV>EPRcr?UBmrss%>CbF~_mhS@+T-e~
zTI&t95el4B$NU*|YZDI-8-~o?dU(sZAVj@+fv3}7%PwZj{m_W##Yaiodj0y1*PM>m
z!6^YEPFXDBh=tQ`HSQt7?0b7g<24%XyS`Afxz_Qd*Q6762Z(ZQ;?Rfqq421#=F>k)
zo(ZAe(NFJ?X`4m;4tCtx>5Prro0hAN@|xpg<ygZIdll*hOF*3AMu8cHUcJ@dg`?wg
zNv+*%dQSyDtuVyS4M5dfqFw`|!cNoP(V`SCJz^yeZEbJ1%g(3x)H;5x4)Upa0kD=R
zi;ryWf^!#Z@SU|sf#Y)@($yG&s!0JsX@>nHeiX3b>Gw8Q`r(djU85cInF&lDVIS%d
zdh9U?1t+!$2iHCT<JKf4dowk!)oO=MO>r5}Y%K(a?VaIW3>4;qx71sqzezjBbXKFj
z^dY$bMy0)hUr<J}sYTj5xvoxfVPF4oW7~Q|`(+Jc$OHsPAY*wo)_3I;nsV~tcdguN
zH~kgN!C!B;Z?6b1^j0=yd#AJV#!l;1>=;B@Yo*o}{>pXmwU=*vex-?*#!ArfYQAs-
z<}z}*bMM2<^TZ>8CY=|aq6yLLY;Er`ZaE}}<>wpin%4+kK&IdRLw#v)zPA@W%<fYd
z7~Nj3w;fnS+t~zI{#0^trXKV~W^7_};tUggKol>FXce0czZEB>*V^>a!Vlxj7re%f
z|7j`PGEM}YMm>zvJ}}28td4E3)A3sZJ-`e_r4Z*)HJp33Fiv6Byzqm%#6B>^#d^J(
zB4+IZz2k(x8%i%u^OW4W*p1WR6)xjuH9rO@wy0M%yCJ<A8W1y$leHa5y{cIf2rbnw
z*4^Rh$!S{Q)WrUtKw)r=c4#`IXY{gO#ILLydci2`IeaaAODIX9tYoBg+bD8b0i^|$
z=5$#>@0v+4@F5IK;r9xDL4@R<6KV&L`-l3L5$Ri|(7&Qb`n+~v9co?kO|7-6YxGsE
z4icfgl{|!Qa$M{0(^|$`+V0sf6`5P<eH~9@Uxy~piA(|QrE6~AgtRsx&5c8n&8GJg
zKz}CRPt*R^Lgx;hXwVbH&XuRG6L0F=gV2!PJ!gYi-MSanZjs1kZsiAKY!eL?lb+~1
z)@X#Edrs)L0>B-@BNBz^2LDr|;)RdD7LEZ~mu*jg!-Z~e3*v=XO2d;(+VoBUzZ<;0
z3GgK96$2osJpV^Atzj>R)dtoi&=L5*3GoNOg9Y?vG~{Z-)$pvGFY;(E;pCEYWyx7y
zUVf*?g_)W1RBw5Knrm+_!DVdW->H+TyS!G(#)+^kYq#PIF-MnABsNfV85k5NHX3a&
z1i1y_y^bFntxlZeqEB3v^W1=E65IZ62ZFa^#U}POP7l@)PfcuPY)U_vq>ZxMI|OLf
zVcS*f{4C)ysnS~7FdU%&WhM;y97<|b-2V`pd<nl`4Xn)9s=CmshR~~~xGXFmUsxlF
z2bQo!@;)pX-zH1rfx&i$RZWT~{nS>Pe;G>WHA+PeJ@TS}9@#-XwackCVhZ&|apJyJ
z&5zYn-xBJkMHzL4LERK)g1QY5i(m9J{>Xc8#FGzfF(YQt>jd{2X)%COdPv{72PrWx
z7SMOvPaTvGc@3ju3?LjGJL8vnGdKk(r^T83Q0J>B|B-OcigOHSM!^wFzRlm9Z>!Ph
zWUvp|lirTdEO?K^g$FtDn7DX9fqBgK`agx<-%8)dA7sQO@dWps>-Am!7=0H-h5OEn
z6xi5GoPUEQ=o?Ca?tO!a`v&Rkp#G2zKjvLZ&^jz=NrUMdmt<XZ+E7iwdYl{AufP7<
zvUBqm6rozf3j)$xNj7b^e8G^34Gv(u0*z0$8?Rn}P4NP5aVrcv!L^kYPwv+5EVt$6
zN=Iz0T)nb-b$RvMo_Y4_GG4i1Y&YsH-;;56qYfR~4PdG^;>-{z$Th)QLY(T)Hcs8~
zq|3x^E5=hJxk4X$lCHq=<t#E0WQ&V*i=IYgAT?nzYAy&4Tc@v867oCHPM${+m->n`
z@fal~<qK5%A`;BR3jD@~YCkBR<P#|u45A*ou0WrFgq>#F5QVnBH#2rj%l(FH(5BME
z0bfFoftjAg%RpeJur#P+8Z8C_P}?%MU?w9?Kx|rFLtue(5OX9XA@WRNeZ}Zn5QAu~
z4Cj5Ln+P)^5oW_&1kBynVHiV5>QMK(wlH8~Uk&pSXbUFxC79SLnA0g3*tXn??1O@U
zi5n$PX>-a9PIr^rTQGK~A{(Z^e%lCsF)V&b$M`2;Fxy)tk;Dva7{;1AO{K$e@m8i<
z)SSe;ZOl7^vSgG|^G5khNF&lZXNVh;1LBa{&r&&;f=g1V$Z+`_%4WCbBYguic~1{k
z)m#o1sO=#kz$x@xP!MSaC^?OiGh-!Zhiy)&Hs?l51ZK)Toe29yBhG1Mc%5hrav}ic
zd4h=_&GEbdCs9u|*F&ftXmf$~hi5i>s895NtPhu4WRBEdWU9~Z{T5RY8B{lbaIkkE
zV~2Ff)}tT#5(ZJN)|PCtK$UoRfNAe4f+YY0NS~6M!L`8~RSlXui=_BpbC}C8_aN8B
za%Rx_^K9BobYFvsJ7@t!&nbEbalxXFGrc$H$vs{CZ(d|`1{#3n?bd_6ivuyuwlguP
zQ7+~~+sdRRiC}a|2;vHBAN(CYL*4#ynMSoD<WEtK3I5ZR`wS(kNJvF)f(r`PSLG}f
zB{~u-j%{dXFj8ZSObkg(MV_OCb&JBMg1~2AO3w^nm93|MJo9v%ASYfBCjbcQ$R`PG
z!E1D&@VLDJ<LU8tw<G;K)ycvZDIZs<yA^C><EANf$9u+SxDXreHtVgOV6xh5v^V3^
zNZf}unP&*_Db_oay-y9RnpY+(OOPHwwR&5+;LUaNJH}^+Wi)2L+^se0FpNZ89@aN~
zuLZNm-?(yY>D5YBzD!$w9Z9t?VA8y|)v3X|+K%e%L_vS2s|E7D3@tGEIRaOLi0kcS
zzoEQ96?q2NFN1!`$2Q6C^X#hai>5X@1}`Deaxjwg9898uRn%t;WhkYL5+p$pU)xyV
z+?+nAmrXL4=sROASeEXXrxSap#%0CWD3C<S-5v$$VVr2S>p(&pc1#c^P{4xlyO4oA
z&~K+?i47?jP+f>YHlpCL4K9QxxHid>1Vki^t9>|r%jAI7w(LQPS=MGrXd2$1_su3^
z5Cc7fO_nB1ke=(3jL-EQ@-sTyi@R<Z?YgU#5q@!$*Yp<UEV8=4e3LCBupIV_2hQT!
zE8qEMMX@;OD(opCF9YnA;aEeaMw1JVz`EIUx&m5-m7<{ortkQ4-VNV{dq&=YB?!xe
zohHKH9k4M1Ods+fz=<=$X-PBkJy?f*t~EXzKB(~-!F|>X6A)`9ztsk8Vi;S!wra_q
z=($AS?QIcw_h355@?Grr_&@-Evj_H`Q>Q>f&)A}stW}coNnlBwfqO%=+(!Mj&+JC7
zp_I)T`31_oiKJpFYFwuRvT`J;I+KKv+U_}Y{tYUF91^XRf)k+#8Ympmh5C&Fo502<
z=C5GCgx3LDC-E{s>#ySl$f`xgmIfp)hVbu)B#24FLs<`rsf3~gs>C29hEhbOiGC@R
zB6_z6?{;KuA<P8Pols9pGcktzYGhBmV|;3d@qe;qz)wLj1}jQL_J+x%^1DXQ!$8|S
zPAUWbH-^7VD?boEb)^e>36~s2G}N0ugG2w;iHA-dG6jxRa<!_H?~@Ywi~IBdq#wuD
zhoKOw{%k9N<u?JCq8iNJ5ziV>>4=wr=xH4EF}T^E29YKD0@2awA(Yssyuog$&2hd(
zQXF(UmU0Hi;*wZ8LPW1oAuW!!wJ-I?PH=pwgEj*Rcm(7ZdLr{52KG4NdSr-K+nef$
zAeq0_VU!}4oa6)F53-XkC2v#m4ia#WmnqL2ffO$$-MBDZJ8k8#{UooL-*(5F4)*HL
z0W6q7qLqnp%)q}0!D=cYS~m8c7#EWdg#*4aoR<l`^%3HdoHxu@%p-h-gyT40>6bt_
z&J8R|Qq)E`XO&*Rh@{Us2UC$-R5kFSD=+LEsenYT?ymCpveyf9{#!H-Y$TdZo3VFk
zeDM$KR<UJ?h9OoRTZqtzI9=P=L}=MVRNCHYQIXmY299PFiXZUVUA#mIq1{de#R$=0
z@WBh<mgKiZyf-or&92>AohV6;yeH5i1vAn@q%=k55;!nS!Y!Dkg5t#;ZVQzI&}_6#
zx2(vfI4(G-xkj6X&)>Wl>vyn$q`b#j_|_ZuT;&)j4g_#7$Q{r1HxQfV^6u@qkZbOp
zN=6~-U_T_T8fdafNs9pI7^K_Y2|GJsHKX2m%X&3xw_?+We4047e1B9H-Txcl0vbn4
z74?!yN?*~)!Aou#lgiy#zY3HzbW#)JBZKh(?E@AHj>+9<4QH#)j9m=qLA~hQSmkR$
zLnB=6rb3-mKX}xQL&VsHfxS;h+5zo0WFk!S<YA(l!3iMpjuV3U{S~vD#d`u#Zj``@
zV7_g3bKQJ5|DL`}k?q75V%u;K{B1Wc5^xd}x>G1i2^(km)99Iohn><aN^=MKC`BH7
zsGNl;+ns8?%59Tq8;6C1sYC6@7Ac~BP!wo~+@KHI6O<EC7JaAQ)P(jq?ep4OS}PNl
z2nNUV6Hy_WiqiXLw-^<Hqv`i_`Fei@<e!PA4yL1`irR0LyMPv*K$Jck<wQC<QPz;p
zMcJ+L!O26!8Cr8u@%T2|@4l{Wzl-=m?qKFnJD5GBIrnlKbnzpz#tc*0qUMm@FpXEt
z3FcTa*4C<a1ELcqCR(rG?BV%;d9gCbj%^imvg)DVly?XtZc|dHWC2MG-&Y403Mle~
z@R_2EIq0J<N7l(a*z{c=#7&leT;iNRS{d80!O=ldh)t<~S-(Y>Qru1HiMSx){YOM9
z;8!<3EpXRFjDc_|I;L7rueHO0e&Y*e4yV?=+U>iZg!RlI6MMdFyX4#tp(V$OS_7tN
zr8Gj6OpmNo-lbU)?-@b(%XqCOLfmP&D|e8OlSDi2?rPj4RhBc0)aeJQ+M;S)`SBaL
zO2c_X+4^~OJ3_*Wuk<MGB#1~^a~WGN!-gwxc&lVls4D}H-i%@E8IwX|IepGRihK@F
z+eldmkD&+bKhr)!@MEJKA$Z9vAMhl#2QOhcwLYmA@j^ZgCiTBUMuo?~PDy2j0xS?;
z==;lv$v$8u7$L9}U2FlF>svNRCKV-#WaQgm2uWtYCNpWMuZKwxb{dyto<m)#$1JEk
z!OQ{IaL_W#%z}tL6J<u2MH=iP7iJG~5s^Loi6U{^JjfqvTLoPE(KmH?3~2_s&`BJ8
z-T0~oVoyf}^eGa%C~Hxk#l%!J1=32bDXKF1ZekVa`gW`wgf%Wrk5HPT${W;Hm;5-v
zE+6H^VI77X0qo;dmd;gr{b!U+&>fjb(x+q@No?Jz`*+z7B)=E$mEyzkc>FZuf)Xr!
zpNRZXg83tq5LQb@U^1%WT&vw;nrV8$?bzOB@6aUHraC=S>w5`dFQA{6C-(v&UkXT=
z!E=bb8RL>MbHo7agE$_?{u&_vf5x#W!z_G&XYwmpu>3!aU-GLo334-jO!SgJMtvTJ
zQ}R8k{c%br6qsxu3BVbD@MECT{J0)E0_DGB6nvn&SwLhT*DMfrg?yoKaGX_Hbwwe;
zcNvt6GsXP3-rS6*U>3l&jq8l?%``h<Jk|0y5ziVHz)^5h2?!DuaiP|3zzK<95(kKs
zhQ!VX)Gpzd$B)@&!y=UKeR0c0eK#zG6Z8iOVF>fS1<im0Efh$BW`f(0(2@=kq(^Pp
z#Ma)nNcZWI#*iZYSZ;*LEwERl=?J<JG^rAF*;Ka_CZQ9RzznBRm*p0@UW;;==TN~E
zYEOnmj4_3~kTc<l=p@1|GYFEH-P!ky-PfV@%#z;qp7GAhV9s;hdBB>E3LAPfALWtH
zi4?*>3&@qDg=i+4o6|%ZzYKyfrvNV(6{8c8F0u>onNYfqKu`BHw7O@a(}1hLvxF6$
z>7MPL>pIctgOef`ox%Odd~_CRAv(u-2kF%NxU&get%R^O1dU3gGtt>E8_~JQ`LaR#
z*ga2H!)U$8g0Y{al-lu|qu#hZSryj7OmyK8B7*a?k&RrTjF8KwE{YF~?qkQaNlvu6
z2$3}xoyQ(L2BC5hBJ1&JdVD6MGfij^nvXvV56|;jRDN_jfg7IG4mUpw&;_Rt=A$z2
zgtoO1m7-G^=LF8{5@vZSx<n<S#QO6Sz>_JQ^l8+b5eO&ac~+!H#Pk_4&0_i-c5X3r
zqD6wXbsl0GIZ=iHy#N9F7=qtv2+$(e>2$t50a}Tkh$_Idi_szpP~;wuo}e5w{t{+-
ziS*m3te)GyLpO4PlL=tmWnc_)j6pNNl`=5m<>Ro0apcknj(pVl(|JA-oj=4pXEia4
zJGM`LY@Eo7IfU~}`OR=KdNR6z5ayGuWvCwWT>78F%Aj-(r3+LV{EzUmI2C18a1$Q6
z%j2i|C_H3rc=C8`_^5sZ>@c^l0y{1bu;Vh6m@|kq$NH^UzbpSGn2`J_N?t-z11I7S
zsdV(?J9ytf20RHW67E(btV@=gx?@duqp2h^XTYm&o3L@L11NY@M!4AP1#5b&R{{#!
zXs<?;Krb$%UK9Y{eLAvW*DWCod?`vGcNQw5jaEr~X?(Kc@!YVe$Mz4sscqlEI&HSo
z$E=^%FncyrSj(!UB{^&5&rvd1d^LHi4o`CsR+D#CCbnyBL@I;WxK)phyY)!{Rdj;}
z-@Sz0(Gu;u!S9ijngLV|ag-wTbCr_HlhH5ed*41TnUv>J87<BFE_wckR|qFzLtDHK
zz!gJ&Crz77kG+C|r(&SKL_E{N&4?P^PYS&Tno=v(zCe9-Tv@8)q8?u>`+(H<z0<F&
zo7`}`HgHj!?i14`Mh=rKeyBTpX2rp?0@$gV+cOZeHMb_)LEwd9>!USZ%wwRxjA_5L
zkPi$baRUDnsJG%|?{5V1U(nP}?I6sE5Cr*&S&ll8qYi~suQ`3)`AIfQrHS^^BS;p9
zCfO^K88~7BwGkXrpc|8-C9hE94C%4~hZkbD1p|at+pSF%gS9pg8ss3U(t46K$XOm_
zV+Yql(v+&sQ@DTW($>p5hfm_!p-lyaq1{rJFRo6qN!-8{6Wm@wj1d=%sTw9REB83R
z`W<(~wC2$6LyZDb`;}To3coT1ewZd$I%*qBQafapk~JiiC1o89&|m&ERk5w{GnD&T
zO1?(P14^2du=)Ew<$i$@!me0fSN6exqHrU(xmks!fGD85zO%7G7e>`3(hPa`U3FzO
z%Q&iTwZtj76>v+nEjiYSpfY0lxKbCe?O81lZU*k4%^{6zgIW=%mp_k}stt<8OEtwh
z9F3pnTwimmDm{;0ap{y)KIJ&8Vn%r3xw63%TV%5Hl^zZovvyj|0Ag<`YBDYrbZ|we
zG079Ak0I)B5L;+qp4!xfVh%=o4z>^sc$nrT_~6JMn$ht!7m!{w9Ag!wb0&U?lDVia
zn;B*3qh482myFXe`N_O5>WjLA8?UQm?gLT|Eay4>1wCWzo&QK?1NY;{I`o&Yf?9^w
zfzK)YD9_q!2r7mK0UsaTn5DQA-i!m2byExeJQ)4ZBeJ-pLVD@*VS)p><lQs6IrgvS
zAphxV5Hz`NK4gy_v_Y8Q&`YhTz}o8+L(Rd*t28nL-aHH17Bo24$ZTk2q_G+>^+($z
zCfb~UbSdp_gz(j|=dcR>b-G)|tfrzY^w)BfLB0rG*Mbh5!1bY%@as)Oe!*WX(wOrM
zMk%pQauS+t5_)V3dM!CAa^2ae1RozR(m*DC6Y_}WQ;u@bI#SR&vd}u_@K2u{T?&#f
zhbN;s$WiFe&?3?z&!umL(3X3p)V4sSK_{FQ@GbVW2vO|L?|vB0VXsPbbqezFe|D#P
zqd~*M^`zm5vx5<T8zas#?AehK{|qDcONS%=_4tU<?7=)Nk<_8loedWbPDQig#GxTd
zW$oZJe2{1GJB#|#p>c2y+M9!uoJAWtd3tJMnssxPLg(HvsU|$Xb>ZMK*bLLq<je0H
zfW8o3gtmwsJh&7s&_0Gwh?CJm|8f|#-7jHZ=OE{2@P{>)i!4}>C!TavF3Mj-vHT?@
z5Jumi{I94yH6Im_@?TN8b*#_pSE=o<QS$4Qe58!pyZI0J3&KWy^XV1*XQC0vr{MNq
z34F+jtDm^yuH(-mbUkUWwC>cLTh3kQsTZ7e&eT5mJ4~ofR}J^9<xXpJFBRbbcodeb
zEV5XXzd<u5sVIMw5?(gDkA90@S1I|olzfsBw((io{T{u3lM<49@^>itZAyNZlAoZ2
zCDgy8+_->IUSYUYaSIfeo$IZ3Cy3Kn8LqDLUlwYe9o!Y6>+FaEk)&M#kT|h}yGhT*
zMjN8I<At|kYpY#vNzzpyaA9?;8%-81P<}T`E2oqo`&gWe*Z=Qn(!Wp1A5ij#l>8AT
z#Hu;|QZ;v10~Sg0kE!fWC?R)fY{18x9tZ*a=K<4kX}&IQ`{0ie{nttHfQEDr77nsI
z{_d)S6ARt~JS|URftY&{VnK<7o(O=e41q>`$$M*o!Q;;kBg~#^+1IO>Kl(1B-2y}c
z{v2_{(>pf$fR)c;0vY_6m9Ilsfr<P-!^(Hx8nE$K!G23{Y4q9n^avX#=6(W<y%3c~
z*!al-8!v;!57>A&M+Pag(&;hAJu_gVt6>?8pBVSd2pjzjm}|du$VR_EoEc@@(F_<@
z@&97RRhj)KWE%e*h$fHW7Z~(qs`{s(<KswJ{soOT;c~PUf!rp#d<ON8pi2t>52?}|
zrOCbWajF+2&ovs7-MF~vZLP_l#QPcxl1IKR#JI)X@}N4qamJ-EGyZ{%e9m?0Z?wId
zB<^^(Yp%<;MJ3U)vW+~-5wDMJb%RY}b(-9u#H6H7$!$u=^5<BBiZl>U;J`ziY_&P8
zCi(Q3@Bb-P{v{<bCBu#69Q}Vwej5_9U4kzmN!eCH18XYS#m|=X;v8K;wDj@?^(?08
znZs=g#HCA=qjy8SKWl>t_DgVCQ7<`1O3vFE_>=Hm)JwPQl*(V!OD_)dJfF%O;*9q?
zWN`5|@DA|0>4Jfz8wb2u@;_3J1tUws|3I(*8A)uue&baokIpwlb6K56uspc8*Dt_T
z!v<gVt>!trvdqQ^0_L=>ah9rsWxoHUtTGm#6&mI_N@(+r+toPryc*#JFlsHBLqf&?
S{ke@(NPwJPOr)Mko&0;S2d8uZ

literal 0
HcmV?d00001

diff --git a/model/__pycache__/swinir.cpython-310.pyc b/model/__pycache__/swinir.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..f8aa99500edcda5a91d1f26ac7ec92268e5263ee
GIT binary patch
literal 31251
zcmeHw378w#eII5pI2Vh>g3I0Iso^1!pvdK+n-V3F)J2=3Ws(YQI0_76W_Pi~0?V0M
zk=$UDd?IZ!X_YodTgOMc>oonsBz1bZiTgFp(WZ&hHqFsC$+(9TC-uiP{c<Rmm0Ira
z_kS~kW3iHKCuzQZ2EKWidH1~cfB*YEw1$VXF?_~9|EPWR@5W-E<3sl+frlq?xqn@Y
z#jKd6HDYD{>J4pPFYEL1a(vz>8=6W>loOICIiD)0Bre`a%%{ugAU!jm(qmd|K3g6N
zVsi7t5|c0IgP4(W0d*LS(fMMzC~=9#hWWAb*!;%w#`*E`xWpwJ6Z4bhNx4roHqCD?
zZ=T;$-h$_}m1%69-&WqH#h#5>S!?J<%o?&&bK9@S%Xe8hYxqX2Y+8A11iw3wqhO69
z$7qmaCt`}$2E=R#Vs;^B%-V>UjX}(A#Ee@Lh?xjt?nca{wFxnsf|z>{v)S5$m@Prf
z9>i?5wjpL)5VIFC+pW70a~EPt=-)oXnAQ%&><D7+Ma)iX7h-nVg}MFqw(CaufTf*_
zm3H6ZJe?`U{rK_bHQ#uuUi19;Iom5~exkMD)#q_Dbng7qFI3+3;`tX|Jdd2|r<~Tp
z3srB{&v>oM5pQv!Vf(|Lv)HUvnk{F(+NhvV#<8o8Q*F-JCEXvbRU5U%M%A+`3vSb?
zEO>t6^b4n6IClpC5kv06srl-ReY!O>V>{2+uhyG2_N)721P|^Z+}N?QX2r_7rCFLC
zpEE3d+PE2;)~vW?+{gfa5>^s3le`%-V&xRxGv&vhZ!Dbod{|>G8Q$TS_pCmS%RPg@
zi(QYcXsZCSc6=e-)|TUKeIbWm-P2d%ZLO`Z>g`y?XdBmK_)SQx-i}Lb5>KgiVpT&*
z+KQcv-PX>OGJe8cT(BKKU8&TYb+1zK^=8vgo~}1BOn!X2cG!<sIrh`FBYuh_+*<T}
z!@^A|=^L)ynD&S2&C0Z0^%fo5^+#swmSuNE<WQiKoIz!kM)jKQIGgcCw=p4=O07|K
z-Ad(l?DeM(&bH?5gUyy_Uu?B59ki-m_27)tT3k4IZn248&D3V?=D}LavJXC0pPoK(
z>g>U(gY$T79CWYX4V<ZkYt9DL%-Y>Nf~=OtH!oK~TgqX)r^1)5>Bta%c+a6eiOYQu
zK|8h*Yiq08GKOqLi|+L34nrQKhH-#hXB0(ChBJ;IXMzE{;!GmAtvg%r<7-!)os5%U
z7h_V>E$2$rvH0NT5M;IRYiLVD>~oN#l;&JDtMfMIDqE@G-4`3Y&s8dy7po1G;*7CY
z<JqTAzbK>O+{5SLV5n<$qfx1J^}$4{!?v9f1e<WV1io}^QY&O6u4klDk`|2ASv<63
zbFr^Cmi6WMvay_y(E`}Dwapm6wh{a$R+GydQ?3r3D}t+GHMtUB#pqfo<kSG3#&WWq
z#L7r7eK>34bKIG6kIEgNt{ye_p4e|bx8J<5-#l50;xYAB>q^C~FWKhadebW%HC2?!
zsI25WYXd@?U6rXEE~;!U&R2q`eW<>xw!Vj;Nr&jc8Afm?&VZnEhfDPiNy?os;ZriO
zv}UUdwr{*#x3Bmqtg`t<&-SymR@1A`EVdTiQp&j-`Te1;27T>>uRZ5$7kurcp9(tF
zHB|yGg`mJf)$!_Hz17^0YTPt}d}gb*TQfB0Q9RT_<b->{Cd_dXmwOR`$7uti+7xSJ
zvc@n~TjbgjyRI!8oG45k02`ASP9-J|lbC2HR^ph%U<$1iCegqos{C=}ArK^F5<fFw
z5`&le3dWrv&^1=!R9}$US5wU_{5%&XK4;fwW<9etZPtm)BapceX1GwdyxC6rU|1Oh
zcnS)L_XQLjPJ(lQ!4!jo2ucY*4sdi1GvWw?`xxBMq@;6{_YW|5h{3}Qh<F4l`sX|!
zytre(Y&)*a`IQ;pJ`~LOF~rnx^?X7Ia8Dq*L!QcUD=&nJ1SwNCtPv|^rEh2;Q5icq
ztqW^mlv4pqkrslL!HgFC5jEP!J<kRlx0==O2lSi46kM#jwq?#QHoW?QS-WbPBoJmb
z%xmtwaNzlK$4jQdT9`|Wykmm^*I%~Hg_bM8Y+kHa-6@%eQ=SR>w9s-q*Sy$L)8x+9
zr#-Y%ZCYlt)jSYBtMQpqi@g_ZSoQgEw$CiiU$h;W?&iXxXVzw`&8FQ5+9Tuy<E(Ir
za1rmr6utzMa(>n}XB9ZH@hc(Vc!&>MXAS%irEmn?wGDTMLo%p0Wtq#DUas)n&AlLg
zjr}GW9Sm>-O`E5t&GXKpz2B@_cy+S@^4zRmY}hhbrq?nrFWSzv{pKYb!OPXgqCI6k
zWlvWz73erhMu30$QpE+|14KF9XjMJ4ZJud0QE^vm?*fQ9b;~v_RRelvsxnwc>~w`A
zx^JI(;Ly|qW_=pd>_zqC^_mqj2tflqH3uAo>0ruRoQ2L{T#=FO>h+<iLs6v*PHXP%
zR|=K`b334_^Jkm_7DA1PF#M3@@i~sm{dr*aE;ftIk`5#otHgl>$y#ac_~zJhx}8{<
zXs0c01CS^ft>j9|ORr=s<hA0PVq@U(Cm~1R$q1i-V8Mt9`;|rBAt2v4^MZrOM9;{L
zXQt<H4YxDx><r2vZV1o$_K;<)>dU!y&Vo$x0X+<-QBQXq(-=1yTOMu?uZ)1!qF<}=
zmBOmTgyp1;_;fr0njiLuJ%0#;@mQ=q+>_7f%7=cgn=hyjZ_hW3-v5!H&XrLMOz3hR
z1g2;u+xb;pi+Rbp4W$283ZrCL=~p?{V;I$q?WBy@IPN8fu(A`9l9H53Zxh<dS{W<*
zs;X%-_==%dIdfac=9Y42X4F)F_Qw@=j)=EI-0e?<2&fVwAh5!geO2%|7*vBx^7G#~
z_K9bIed)WOKt$a2tP*~(21;9<BI-~)*3-NrI`=ct>IW0RwWj#7%nMZqjGG63JK?pQ
z+N_^gvYnReZ(Z93Hd^tx!u=#lL0&;sUbS|~PtV)#><pGCxJF=3py3WtLEjeV{Sn8W
zsk>;Va&d8b8cTU2I~50damd^GnP`dn1{ao}iI$g>VTRPX*0eWYz3Qi2byFG<&ee%t
z(RqgX(jms;ox&>4IV3nQN@|EH{B*6=axAwp8{J*;lPb>72Wi2(u=9!@R8)fV1S>Qd
z>}0Tu0Xh6KE|-)vt0y&n$>kgId$i=23PwR2({kE=t*GtAcfXd!)3}~Pj+~y?i`vqp
zqFBd6($tX&hzG(CS-gm8ngigpHE_D|E<P7)6O2zBPz#8gi?_l3&H+QLCRSw4flUIR
zyP6<Z%hiq$D*~auDHf0G@#WNVdO6chp;Tflww%IUdL?Nk!Bpvm*m4%}Oe1aqyl;xF
zWbl-gn)oZLYi$hj#+QfMLqQ)5to$LY{GoOhaXDZDZGv^{R(er$zO${*#aJWoN*49U
z+c~g&In;``Y8$JGwvI9e-m_B@`M|M<ISlp_*eXlh)g=EP+~Y2<V$pJ@5W|~cR91C0
zV&L}UB3q3_a#SI>qYfam^n(%l?<MD?m0-x~Lcl}GQ!x|5a+9v~^YmwW<OS9+Uv)2m
z2R(G~K)pE)mIRM)QU@fC*?kwT1V1IS<p|k~df3JNb?^)Mz&me18W2o#rZlP;L5fU%
z3eqy=16fIG1wYI9jEB|k0u|ZLJMpZ*Usy-s6bbNtX0hpl-rK+)7nn!Z+<D${;gyEf
zVinoK;CuP{iHfhCaUNk7X<fVQYnOcOW#1q!@W-P?7vVr6ASeSa^zor0NiL%vcOQaW
za!T8(W%Ytq)bG)ZFBNqN+IkjCu80Mf)wB9$JY^w}EKT$;xIm!wla2^xH4a<&;XNhL
zlek{nia=IZ>?5S95L)mH3NfdDB=#}NipH7O4ItxVujhNjwPUa6_ILQOW2JZ>`ns*D
zaT2r=;TAt@U&Rzu9DBim(lGWK*xT4vZE16V|2x@wgkz!)a^38@yMVxgeui=sI!wku
z2Z@urhhh?s*J*1ij<2>EdHCFy+Lt3p4eIQ!55o~Zqa+i`BPCtEG|@E%fa(Wl8iAhx
z^S9vkPwpa;R&j|A!q6x6?f8!An7)nuuP^0GM!-=CTcuPVS!WMh*~_4Wz)w{wR;vbu
z1F^Q!M1X`kj>|{3=-kVoGYuvmcQJSz!D(F5pLER7^26C(`5lG0q|}ho^AU9~V%+y5
z=#h3Jg#+Xqd~IVcfh#$k6)6a6RY*Z8rBE%W1F0ut<?SqFp%JTa!zd3~qgL@oteg{B
zXc*(X!5=$!rQSTx^$K8~x1AG>R_#*tZvzBOrhxB`l1xea>ZKDwdH_zh)mWr-VD9Y^
zQBDa?MDk*z#pI4~5wRp+hCqUM=vS*KFc4Y0Q@4m<J=&LZ6w)M@iJ>viiQLyIFyC0H
zC=oyjA&_v+h7tkNOr=CLDP=&ku_3Zq7L<RC=u{Fi9|Odtzw**j=Pllbq|`0Sbqhri
z8#hSOiZ~mSEkGio%(Y%k%u-o^?tD;9=e!oAWY??LAP+83Vqd?aNWBXJ+?;w|l#@LI
z;PDz{eNJ`L6g6;{kcI+OSVkn3N_(X%Uazt<qSmi3_0-o9dr{$Os+yTiYQ00vr69-8
z;|8C?lc7XNRdM*Ch)6yd>liv@(4se$2%##gK;4B9JRU+wf<Vae>qO3pOPaHnw5yWr
z#8whQ&5{NhK>9<HFXle>4iJay+AEqC3-8*)ZH@I*GG4Z=d6XFwc#e}xBu8b%Uj-Ld
z$x$x6WH|5fhTB8ewPxJQuZ*lxGA%?kjCw_opyU{pdwGwtR31ED#!9@({*R&Nq@|&b
zW$<hp@dO@md87#mb9n?Y<B-%wD5;4=l~M#|$?RJ9^NX$rj?6rCpbKe0$0bY)sgLs>
z2JdAc><T#*p&HPYlfpaO7~I8x3iCinA@@?6=o7O1LO&K%OcdMlvz<=)nF!DL@%ehw
zPY|RI3+-^o);n_y1SrN>Hhl`F3usiY`I%@%s6t1a$C#D60#wylfY=h5j@U=%-AE}F
zx}>Bw^UI6N5RheZzKKC+9K<jngYgOz`0jqT1NG%RyC2w7fccznWG<6I#H|QT{V7C=
z{V5w)l8_1uS`re^Ua(<0m@!yE-q>tV4k{uBl9FD~m$nTeU>#9NN5$bMhZ`=6lepX@
z7DZe8dYvkNkLr8ap4!j}PeV*tHauMk|3k6suuKt{hRE|6(KGHWq)o?{VWWE6c_B1G
zKuSCnduaj+m?VCSB03pLO@SqmQFqe@lou&W1DBM#eEjk;F9AV@OHzlRl7$dL=25P(
z7}P0@LvE-O_=`7y+JwSW8p}$*6!)-5u4@ol)U%G~B%hfFTmTjWrNk~}Fjc@axbS)J
zPbd)2I;ZgDyaNHm_(r{EtA!y@K~=6&4wIqrhX_x$Rhh0g8p5Vky;4p9%THbvBHK|H
zm(%t6nToLZ>Dh`@?D}aKLfC}s$7@Z`*Jgd~N;$_&p^-yDk^oT$l0JaOKZi>g`!r<a
zoL0bPypfBmUn5RS+9cR{{f(S41_>LM27PJQKp5&>9yOxU@Iy$7gz6K7Bpr|h(@`K?
zLs-T<YnbDX*^NSWf?{?%1?vJ)E1(VN>VOAMPIzktV+g^qf!ljEzMN@->xObI2{jtT
z=wSAtlH(IEaM~1<DP*VH`X${tgFC3)5N3fGF~d!$sZ<OUc2LBvz}h-DysD{Gz(`kZ
z(z78CN;;HDv9H$+z-QRgZ;_tLGvlIOYUpJKZxp>@&?^n_99t35cp}Dhd5WzR@H<K|
znO~IxC`>Ap40d0Dkst{Bc2aPm3wv#S-W3_Nq&XJTib@jw7Thb)6G$YG`i9eLG{k<S
zU{ny9AWFeM1QTEPT7J^9UnU?QV;!$BxWV8$f&gBX9pl>=+iT$PwbLSt`e}C-y8YUs
zSI$UR0rpqRnX4h3mUF`5D2lR1Sh)ukzmChLppz!3#<Kv~Y<xSwG8Z4mT~XVrj{$tM
z@y&=UYI*rx+By(q1A)|;M|EqYQ3CBrHbwz_J!w=xD-P+DS|9_z;0NXskWzCBQM6F@
z^$q)#=Aw&<KjvIUibH8}X?TAZSv|G{5jG-X$3Ra-1Gg@1aac{?x~7=58>>vL;TQ&p
z>FP>;Y3Z(k1JNbfqFnSz;4X<vhT#@&AlY&y#uOcpRR}r&_=PnQxL0Tbo*<~hvM3?f
zxWdINT!o|!$iJ%OT`GVfDXa2GQ8KRzp&XMM*<NoOR2791uYi5hk>Qbn64EQMyNLb@
z{VgRFk(P)KST>m6H*(Cfmm^1p&;Z^mLKfbt?bM5uh&K-$J0%w9`($9}WmZof!`X+x
z&qREw5{kK~L^5`+@QzdV7LxH*WcguSE-R&EY!pf3^KYbgq$l$G3WmcRHC#QP$R79<
zqPrzzO2pHeNW;LMkcO$MOF-UDSxGw$Y(@K^QO;1&j#$XeMyyQwH{|EwbW@vs-gagn
zqAmT7DUW9GY|5JslzaP`xIR(#Y)8h0K2I{6fwHeldiruKdmYy5&r6`_v9NP&xklo+
z#NIksY#<#VzeHV-gfNnb!6QiyiZR0Sr-ao`oKePPLiRJxlL$-cE^fF3=>l6DkxmaM
zYY|dbtLYnjWAN9a_UYvPAhHeOBcCL!XOWl{efnl_ic1^2CwDMsID`@rUU(37V2#E?
zoG_<BT@HT$D8QjeCT6&#yBi_kK(KiQZ)@q}Q09U5?!emz_wIhA!}mZCDbZ>2W8lD$
zn>3O>32~T($OXV)WTjvRi{n@S0L*E@uf$G6IZi$z$dg_-4?L-YJYo4TZ>OG2lqXCJ
ze>-`yUL07Bq8ZJ0hLr)281$K|hQe2%@CL^MKA99f*`@FvTG}4TzEE$|&t)IoT%yRO
ztv##2givl?aaz!e&{*9uZY`y+D%HzgU)#4dQZlQK4O6*i!gZiI74WNv2n1A6!wW?~
zqvN9>Y(TtL^_-6|(}$U9NCDB^jObuZ@mBa-jH7DV*RLMJbr{zXT=(_kQxfcF31eGv
zlQnYSAg8o~zFXg|@jE7x36#b9(pV4Fy(wYlsuPOD70^$ZKV}GZSQ_oeHOH2c`>9Pj
z2A{0X!dQ=degiwLqy91Iv3t6<4yk0PqK-X<hhRxi6)a0!kvBN-!L`Psi;d1OQV)aZ
zSrPdI;X^(}=7<LrM))}HZU;-dGa*9N29U8T-+)Ya7ACO0U;>tkJ?|l8>vnE^vh^W0
zynz8{@okf-??L8=*tTN&4Xv0?r}7g};4)<md7^;Hk9!_LH+izS$mubLFG&^dlhYX?
zS0qd%SBfyD;15f3MPFEwE9DE@Ay)-Id!p*rYeJT)2Qf+JaT9JUP;QZ)_o!?eR33@`
z5)Kk3tgkcnKtcvMTB}r5OLeV#Nk1g^QN2^40yog_c9A;Ky#?NF9XFPX)H!##ulzab
zc~7?>!3AC|ecIlL9}Sd+UsK-FkETb<`P5shT>s*fxA5puHUTW^s`pY)$G#k0?pMez
zMP5g1)UUL<!k*A{tMl;9?&84)`5-;j$*9&LT$Si(WyA9d{2*^7zDt2jKc<LSF1mIF
z&cd}z3oW>BDH{PG^}?);q7Kap$N~4SFbB;Gh;*y)M4PwgTh6tvnxCmQT&#V*fU^YD
z6ljf~sZk*qet18OOZY$L>hOQJK<g{|Evi~WM<Qqzl;hCN*iG%Gelvd4xS6<_yqUV0
zzL~k1ogSi_?2R<k#!w|c1)fmcT{3jafzkfq_V6v^mUaueEa4rq<Q6TEnTp)OB-$gv
zl6FyuAp&=40G>hCm+oHHycjAmZl!KPzih?8ifLfQMlI-&Z=s^iv0I5-nOo^w`mLC!
z-HLnqb=}!{3p8s}tchndufLl64*jxD>eOLCApeW_C6x6UG}oo=dUGA<2ay>yQFigO
zb+_IG)@|0{MB9Kv3p5(~)FEH1-yTvrWNBnG>wkrTG$-}58l`*{h4Xv(_y7YTjk|E?
zXCGsgwbuN?<2Om5CJ~v8>6zr88rry~Q%(j)g5NJdPTtd@84(XW^%yWO;No%Igf)fc
zR(Xw~E=D~X4nH9_6hGe!pUOcfs|+TmQJtSqPCrS7-~DW7;q*A;$UFZKU#&CFZ#~4$
zKW2%4$iQUq0}MDULgL#T5BjfXjZmMwGoH+3|0JCogYI}~!+_D~(eRP=N#Z(*%MIBd
z(brPzsyPMhMIaXsR)`$&WnqKb04}+w#*8p-f<Hzc;gliehmQ6C2$f3L;YuB?UU{KV
z(h(;0+mOwVUu;|&#D#u{Z%C65B^_utLnnn|*PFd5*BQQ@(?lg$Z4OmC&fh_>lm=r-
zQXM%Qbf#!M{_^AKj^nZL#;y*m!&?%)aDZiSAI43nhQR=*6@R&cO(Im6T&5lhHXID*
zW>L@RVM-eVdRq^ahH*#6RCq@MM(8LY5&sks{X@uv@8b}XtD|BhFBiDVLRbHVr*!^)
zZ+VlC0%_leyIEF3#6%57b_^B0k=~lmf<0tv4Oh>H&GF-&M(|ebA!It(^`~L4#|9j3
zl_Ha<wn(J{wlLl0*KC|Ucm726Qy5t&9_kjhQmkbN#fES0WzN#qI6=5jea(zdmVh7H
z3kMm`zs3o||NQ??PLR`1^Ba{fM@=QT$sUtFd1-;_(IA%y-F@ls)~UnU726O*Yp3=Z
z;Q8f11Oz4$CcwOCgC^P7FcILVPMHC$>7CC4M~bSzPaI}#wIU+I5ATV0Nle^7MBstJ
zYr`hHqRak5@~hZgh|qxb6m**QENJkOp?(Wit{j{T@|N~0IoV`;6t+^`Iq0QeC#7bN
z;yRs2;k?j%0I3P2T1d@CIg*>9`BS;P)ZCE5*RSd--9jpAVII_DWmgmH<_YRhZF#x&
zNV@=S8`xR6fwTQ#*yQ@-MgrIHQof5*{c6~#AHf(p`x#IK@DulO$YuCH|Kac9=hIJ|
zzpeZErac31m6vUpXj|AEmV!E<QH2!^gfMnH;fp<&0Oe;pV1w8uWEuF9`5Ezr;`R$i
zc+7(0F+a(MMLQEQnnFSSd<ZNR*+r??%;ZV*?TF<Y?4zGOXTz5iZV%N4>=JhM(zUbp
z=@<R1fa=q9z4C`tU!?6o-Tz`!B@L^j#ik^}t|+?ZPA|TiG5uG3s~_iZ!YRCGjVP`)
zV&>)5f113+lgOPnlJE#O;PnuG$;*fL=|4&jkHb>3Tc3o<a7;^nQBS0intnq!_?vrU
zX_y2);`4hrTdo&s$WP*OcOw9g%8R;Otd?jOE}WcKDY3yq+u>Y)7Wtij!{Fx_{1k$a
z)w-JzWLf)Jt$PsnJA5-8H@mTaRr(h2radwqI2T#Y(ltsv!WW`rbLgJp2gwbFig~4s
zM+{W*Dx_;_WMng<GOQudc~W{vg6kb>>tK@Pdnjv_aJn-0@~V8xP(>-cWJgaaQzyZ$
zu=MZpEy<I^%M<M6R)q9YS<0$6rn_pX-?y~OuJ$3;BTIlo_7>JNu9r`+y@>ampd^t0
zVtNmFPo~yz^?W$I{JeGqK@aaa+@*$u)?e&n*n(o^Zgg+;&O1MYg4L^N)s$W3)HcW(
zGi4>wzj4sswUEMu8b(!?C}PU&xVjxVa(_otsFozNgoRW)Uh9(GRK%xKva4K(Wx`S&
zC8H=U)dtGWj=4@9_jTG9G$7l<)@UjlvZ>tUUK2&>OtnvIP>qi~*s1cX)k8&UA?KfC
z1_UPyroQujJf7*<a!*&Va2>f^jvW;a9GPIUg2n_!JN)8FZJlK(qzXGaXwKr(j1F^G
z7YP-=KoCzUA$vc#=0G$d4p&k%h*z+qOysmDL1Yc(^B1Xf;MN3bEi`hIzF;Acf-88Z
zTHop5%A!oa?RNPW_(RkR2l38NF#k_7pn2zZN}=qig06Iu80Qxmm<&4WmFp~_WBpM1
zqITGuVlzy_hC}R|hKCt71p;^i-yM94a?ywP9B@He`)~sr9}pKjZMaFCURFduN`&1Y
zI5yG;(CU8xC4}GtkG_ld7a54vYB%q8MuDIiVulGcDV7kp!QmTOLR2T;60uI9icsJV
zP$@JTGIbjjpQJn<LR1Ru|Ibha{gVNfuF2T{hp21P$p{T@WB;;9q+AV`1NN}cHH^wq
zljwtG=RcyvncL}s2AsEx6^Wo;Cst(h;%%Qz7Yj*2T<hXP=jV|bY}BXmQ(YQ>-Ck6z
z!PoGc+oe$c%fN36Cf3JqP0ZL|9)2SZbAExal35pbDO~eQ$kc^|ewmLZgU&)BL?7Zp
z8N<z5{F4p+76LA{gL~r0AAXX!A;Le2%OzFs+1y7A#Pbf6rV?xtkUKZ<0V`pXu6Q4S
z6@uoE__#~;Z79#Wi;QKZf@#4e318e{^JqEu_2(|AJz4T*o!29tI>y&y7jxQ9;s9k5
z4_{~yo<e~b)}MMG5(UKo1{g?25Bh28hj3$kjWX^88znv_#~hf5qb6m;(0@Xr^Pd@X
ziko~C1V$uqH|yxaAj5@(Acr*=B%BME<-0km$O|c5#_>5utjx0{3_tzig0dVft%i5O
zK>`DHals~cbsESyXzFcq8ru_E&Xjphohg7r1SFGt)I9PHO*v4Yi<2SAq`y(f4V_dc
z1#m}Mbb<i(*v{Kl9s40d*QCG_+aok2@O~T+zOYLkWQG=f3dwLaeiCj#5RIoOfI}^g
zc<j#vhe4yyOOtSzgHwr@fTIXqP3bfQ=OGYA?4jl{K$4c?j!r&b9&7GGd5YSny2_#K
zI^|Fc%QazU;t4EpwMl;r=B8IPk2?;a4wxHi$G;s=CE{|X+tpb}FR7uDdjCD{mXd1B
zmkyeBR~$})!y9bt&Bq+}q90~Wk0J1}V-aUK#T%^|?0jBWfW1#S&i!j9=3b~@wHxPV
zv9B3BqqASfgASf{oR*_-%CE2%B8Sk+?^zZ~hg0cm^S)LYgiPK=pry&ffH`0i!bcHK
z&gnRR3p%_$Mni~;Pzjdw9=5iZL5ac34A}Vk``OsPXW0}NHYw~_%3<7^wZS#Wk2qs@
zuMI0Ds{=X@=(FVrof}vmvP$IG9DgM2>ARZtInQ3G?gLKfiwn=+*h?#_g>#f_$A*<n
zZGR2(at^&&z<=E{tBotwYp&U9Hm-@+ktmeRd(FcOoDNq_M|2#|i#!Abz7!3fyrPbE
zFwdQhtgAS+M(u*fS`m|K@9y6JXX{q_a;C=B8h$uUlyYU|p2v--#BfFdICc%Im`CC0
z;`Sx-08WE|G!4cTu4usuRq5jJ#g*nIJ-D7@U4h?^))k0|tfokP1QPk*(3mo4`n)5Z
zbGQPW&oC%5_$&hf_JFz&*rgzLkpOw|;H}!Ms^q^j$z-sT!7c_YIzwR0;f97jLo)&W
zfcD7{1_Bt0fPRg_0mZv@A>#Mf=j233CaM-XPE7KiL>M4PqB4+|{a!?L5f$koX;Ld(
zFIm`SL6tW4cTU-vMp<8K{WZy|)?b*PMe0jzh;&m}-%KGbBV0p7&xZt*ANL2auzHqe
z;G;AHhsBQrIv6UYCe2S&2#Dv>FjNG?G(homKHOz_CLhA=^cHK(PU0lKq&04*c>W&F
z-HQcVR5w|hZwzB=>K4?IL40=FkQ4Wea@HQ2%gqg2lXl)tZsPU_p1TJbvOKaTh3D@v
zrC{wu+9=8(L`}P_-SCzw_Ql?9-GkT-eX)D2y@(w{*^OwcWbMP#xOK18x4)~_q;<fW
zLfV9N(4MrD(h{U&>yW(>Wj9SHt;5z4q;Iw!weGX-$K94eIq{}R>jCRQq;EyJZFAev
z>O<DUc)rUXS}|8+D76DAU2Wiv9<h$fo9-0UxC;!#V}6nzUZ>7he;LbKjC-ffM(!fV
z%@?kn=b0SPax66HVRF=bzse8YK2MG^&)P1y2esGmMYtx+tT7P2e?MKxfc&t#vEQ3)
zAg=3C?u33jqBw!8${LG8sBkeKyg%x<2=$$<xBYd9;MlM7IUJS#$~8Vw6>Ahes!53s
z_w3qDmwS=J{vS0Tc`zz3wsIVKBrZ$gt3M^aZkPnSiW%u2%E0j_s)EQaIBTh=mNOBj
z24^dET9uq$<dQgrWY1Hz1$>){PuB*<koUXLIbp2pEgp(`@YNVXJEPK>)!wbmgoVGT
z<DXFXS901@*C<5lHI5z4pxoIVo-T#z*Ry~|gBN+i^%XoGO&V(*w1YXSLYl==nC+|L
zt=-)^Yhj?85&b_iXp`|kC%2Bu66bnE9%sE%cTqO(cO7=tSt!s4lOK~6-P!ZrHPYgR
z7QM6NlaoycH^uj<Hp0!(Rh*gTyh+_XM}+nICPmeM6~1#(uifdRCJ1<sICkXVeFyJ9
z_^^1|Dg)}dv-{1%@{A3+E&74Yh<HYKEtrB_c(i+UN^>xV2AdtC7Kl^DcPhl&@-Sr{
zp5mb`M;>^fGfj(uK<Ox(t7F3;hLb7O?g>=g^Qg|On=<!w^SpZw_B>j36!H+s<5NK{
z?xlro&%vH0(r3#0ip;&{9^t^zkk}db$jje(5?qN=M+k>Sm0-!ZLmt-o&EDV|AL*hP
zI5N&C8|9~-ct)8lk%G_fJ>kckAHj_upIeymjakUjescD51&7S}$%gtJmZRt}L9+Dx
zoZ|N@mR+k}Q(Nh_N@Xgb0)z^aI4<|2kaGI$&g0Gql;RUSSZxe0hVA(BR$SY#^LKk2
z`d@la%5IT+Eo1pE&sa&c?}Dy4386Sn*YH&`=BZ`|DQP};;@U-|WZKy1Udh6S9jgqr
zClKb^;|Pb_<IYQH6^>HwJKqt5_=^qnxC-s!D(^?zTXDdXY9S#<kQJNR{&>_=#KwAV
z(C3DF_GBaM&aujP8+$e@6B0AYU7eLp9^J9Hmj$`zCY%qn^G({TlVY#lg19YnTbE7w
z-L|}=y#r;odE0x+{JowscOmb%^C|SnRPWw4adZ>Z0w}q|+lka2WJTqOwEWyI>}<}+
z5ox={Gb)BMcU!RMd-t&at8x09F7HHbd*m2_y<Q15jEPF(rR81TKI`6k%o=VR-=$gk
z_6{otz1R*bLFN-NBQRm?+7!cG0frBCH`-3&8FMhV3BMcK8H8i)EW(ZLA%x@Y9KwnA
zPWAwPth+t{R`I$PK3Rzg9BIcGhR8YD&LiAp@qTlA6ycV(hHz_}J=oUX_+8q!Ys<UQ
z-rkja+q*?M@~b%VAT~EC^RvIb(b|4V|C^d~TILZm5za40>_GEDq~9g!uOOYUFv{7o
zOz~MDq$Rxk+?2J02Ux>rLDHET<RRqc-Xu>5OY~1TfL>*--O}0*cD1%W?2)vz)4Kar
z*h7&2pmzw+y!|@w4@=A2+uOpth&$5!R;lqGsqrVg`;gmcZ|j|>>x9q4O;2*}K~CM-
z_qN*Rn{IDMeM2qxqmGPaN^7qL{pn7Bpt)Ic?3GsFjj^VN2L{AFI3Vt!Ag<?ayXOJD
zc{o^^c;oIdf$iuK^x`NXp*@LwkG64eVCAtk`c^sC#+s`<?ma<qM%KlXqVU=u2<D#y
zwh}d0DL(VzD<A#%u222fCuV;4;isQ<zVMkRW=_2Sw|-&gXMXJog?fLCc~3Ani2&yW
zY70G5GbLtfch2Dw^|iQSCuOHtY&E8wsg<%8XAjmrdmd8q_apP5Q(>JSM8a=VZe_J&
zF|`^0)Y(X2ALOL0jw*PXRf_ceL%e&2!2=Awi2?lw)L|XsUd_%bAL|eDj!vS=d$?mY
z{CA8av2cErfoKLlg}d|4zd-2cx~V31q5n%p7P})A?R-7gO*-#T9LE`D4f`rjTx{B&
z8kyrt1M(Y46*Y#RiKxDuwBRStL*{^1kQ#Vs6zytL9lnqP$tUlDBh*6LFK_LS;7B#4
z#!}}P${A}aLMdN<nl<HLpuX$HFso=}<V5F}aa@A~ukColZYu4IY)){UpJcH(&w&mc
zddQafxlkj?Gt<?G{|w*dXBqqw1F^bNovwUEp2MAAtOnd8PpYg4<0$7fs#6`)j3guM
zr(tP#7wA>=0xS3g%O->Vh@SP6j26ApN0>H3E~O$IV&!7Xy5{^_#*Rs>@;R$SqeX3{
zpAH}WjAX+3?Vc!0c)la5T3k3B-5rVU?(<WUI*a2k8#r)o(N93X+qi~v$>r>M#W}dn
z9ro_G8T>B>zr!Ghz|S=n>I-fKCulnLTIud}?bbY+01nvhx<DNq??O4HobMGs{Cux0
zA#PR<?J5p1N8os_yVK4ev0GnYAdvrOynCI&fbN&&Ku{63onL1jD!j^RXXYYa2A~<u
zI1UwEaNrcCjy=LKaR=HUz1??k&9k}u_Aw~M3~2At2(e{1I1E$tdDvoUAie<Srm;lF
znLnxNILZUt#otK7<*D<OOpep(M&BCG#(7{qoTk21(5VO4znC`k|6?TdKT9U{*Nv3^
z1-a7tpBlLKX7oQXvictzL;4>XIsFffVf{5Dum6EDV*Gx76z2f)EC#sp@JP<U>FUpN
zSsXxv*d&e;8ArU1expr1h4EwPJyUql2){*bES}TpT);1m@*h#^nF!Z8RC4y5#Ydz_
zTuV&sehUH*Ha<_A<2IIiJeXSW7Y(rS30Krf*@<}NjIkSeakQ^I!ziQ7Afybug1XYE
z>+$}4(KCG@yULZYpZf&*A+$l^;org|^u7Opv7+VO$U7>X{iIqE&T&2%TqCmHxFyD`
zICOXl?Rlc!-N}Ly?Bd-h!1d8MUYyEeo_sovK}>%shhYcEys;#Q*O63oFTNg9g$g2r
z;%iAfSlTT14V18NAdWQEmczDsoAyVVtxC5X@HD#dnvRS~eQ7+HwZ1ItHtgWPPNd0q
zMhy*Wa<Z>J$|zTO=<F%9?c9Yr)L|ER;q`I~f$f2x6q@995|{gR2<V-%s{ZOgomj)u
zC}dB|I1XC|f1!=jD8N{9_IQe7`TrX<?%@zFTnVTsW$z0&Vua^c@JvvcwQ3%Wod(^v
z6lv0)@#70s3(9S(o)sJLKX~`;4E_d!at6XF{Clj*Z0Gh$|Ev^byG%tSY|$7#Pw1dw
zvH;lN7uxlSV0!xV_KXYHRJ5*U@j%PDI5+VZ9Qrg)h62Y06Q|neKBqeqkX~R#pWwM5
zG3=8MQ+S?JN^OJ=b^t+^iT1_Excd0NCghLQNJXX+oP=-~Z7ZTF7+RpxAZYS{D&+_C
zbI7u0W013n>3vXJ2#o5{zm=Tl_;5{P*Aq~V{38cW)XL(5Ut%~?{iS+FimQfeSs;kM
zHEN`I@b;!y=s(mwPHM*!xKS9CFkZUAj{)lmpAup!exh%DV)VFP#h2cuGP*deAUI!u
z2lEc8)Mc%?B6aYzh4;Q?9a0KqJ}6}jl=<{qmg%&M<AYbg>CcVK73OeS5wI-&Y2Z~c
zQ_GtatrTkf5OOi+hV^nb??&rID~-HA_Evd=^s)8oLOMr-?ZH!v<Ha#*LjDh+4J#JV
z#?drnG!2ZVoGpcNFqF&Kauy7AnvHsf)Q-6fLz)NYienNZctP%Y5(Quv=5vAF-vvB4
zzsKPB5%{Tq;l=L8obaX<<)#3(75{6a80kMi#?rU~_qz$Ye<CHlYYTRz>UQxWUu3oT
zcPebtXMq2viaA*Z83rQ^h8Ub;K7zKNoyH#|@w5=Qf8(Egpe0;r;Q(&i`BP@eG4n73
zo`cxq#_|x;A4X74T@ATka<jdw%DD$Q=ndf>#SMM#@}&DX8JLSoeDH)Xz~Yh^6R|sS
znfT(UvG^DWS3%#eFXaQoG|2OXt1QAtH}KHsp+{Yd7-q#ckM>ABJ;b#lYm9$E!zYh8
z8~P7mq%77Z|HTBDI!wK65)<{-gZ!v1s!9ILMm=G^Vb;Z%b@@jCg9}^#lZ+Kse-U@^
z>4!96lNnWLi&GVjG&6Law{e?~18|qPJO4F~GKoDs2RCEfaNZR>8Q??>xAAcF=)}WQ
z-{I8{?((AD=<!rR4RVBypey=Ddn9FeZDEt-42mlGib&ad5(-)TtBu)cK`DFDhP59a
z#)R7j{CwdVBaj7kEjDdSYh7lRsE%IqKdlHX3X2fUaZ^f=gC70_x6!PvyQ9rzNoi!A
zqU&~G%|8Q4O?;#NKAh*Sb0)x?1o5)dwn$p|6S9POO8ZH#>dZh)`7nkMJlZY{8P4s;
zj&v3def|$3);YwWcggr^6qTGjN!TUdR2Yi5-7w8hT;zcG8TqNyt*bpVNwy_{OJ?#S
zZUof%KMw?XxUlKMz`&b1ZUMDfIHGHX2k1b>WEt3qi@n}ZW$_M+f*#3aYe2Sf2p@>C
zoNI<t&g7)}S3!2we>PQ1-LSECUZtz%UjCcg`yesyhf7Fjiun&K$6=O(QXgWXSA`5~
zo_@E0&U5cJ_nvv~!pZ&S*=KPqP6RoeCw`Dkl>D^oz3&iA2QP)iI?)FXGalPERS^|N
zV;36wPVt`~)cFE%f05))iKys6fKkruUk=Pr5w1uAhAsAbv1!hlSInAu;Bm8wA3T(^
zIEox{Bf;i8G$5%~vw}Z%z<;lGiJ9c!@OKIgoT^Z9?k7C`8e&3r@c~BgR3=}az3l6a
z%MRfV8|qPIKj7it^3WP}`E2BS34M7~HJ;rJj6#XAKwJ_E4n3LI@b4(BfR7jnASZid
z=CCV9Nl|Jycibtmd*6jD&UZ5y#M*DbOXDAH>i9D_kSTNx#-C4yD+5PV$w6%3>jz}(
zSrX*Kq9yTB_KHhk(2{^u*bVryNOUZcxGWO+3j@hP&R>><;%-(F_>%=J4cH@6IH3x1
zADn%#V6cA;e`zAA<S%ek4bF;0s_+_WLGi4|BzCQ(Lus)zbjIPqIeqS>1cFRx*|W<T
zj7vqdA|<UnRHO0rB7`2>DW{pT&~)+Fev*fHC({Qs&LeC<W`O*JfQ>W6J3<k%i@Xs5
zkaLvTA7vmQl#qtY23v6+V{{)L8|j%fd*MepX=FO64oHJUQN%BTo2E<}35I=zdH{>K
zFq%Sg7Nk$4O*c50*t$r=1tvu%Mh*d0VuP{tozdbw*=kPX|JSgQQvOJiVwPAMy;-by
zAzg2K#q*CSrTkv0YU;REoxcDyMl$i%m9Kd-E*Vh^(-BeB!YQ&Sb1?_PK(AtW*ukbf
z7#m&e>w{<*Xo&we%JofxXo!C?TrK0VrGue>7Y;)u>Zv<OTg*c@8jfh;;8e-M_Y?A-
zRA`2dd>u_)ppo!1QLTL)61}3DfUcH5%^nhw87b<sl5ZHs(v~;rX^)V_PDM~bmFIwk
zttcA!m;tnohZJLxlInrh!T%LU7{l)tBD{`+Udi>~c-3BHEo&2mJX#I1I50qW(7aM!
z@@-^=@%UB9^uf84vYdKXQvV2L;rpt@boal6<1(%*9Gr(QYA`OHTp09uw05Snd!YXt
zmz=1J)UjRP%Et`|{N(d;K)a}ll-@i(2PIRoJ_8hEJ7lSy?OfO<gFoh@NFO3waQTVH
zuG(uSgohC9kQ;Z7vBcvHPB7r~I8XELIRub-S^~HZ7qs&Z#^1}k4gxXxxQoGWAXr2`
zm$ZSddHOcBmy>G1a95}LMcTSw<2Ut9`seX~B>z+2qW_k@RsT(WTmEygZyvihnZ+(p
z2pZ$@@znT8e(%;Hb2wWx<Au@W@I9KeUz&1$6|IA>yS4!GsJbmLD`RwuX=%mGG@D9|
z$2Tr-2;-(YQkHWsUw0RWVO%BSj~ws^be+RY{Dj0f|Af_!E?kpb74fDearX6vhQDJ?
z^b`x>5A<wzigKYV^H44VMa@qMT63oY8vuDjb^bX?VmWEooBS^j0&IC$bw=sfT}2H<
z`JSBT|KGX}83N~>e1#F-J<8x1gU1<MW$--=zK_8t8T<@`Pc!&w2EW4Kvk3f=({vE#
zD+O@KsqTUSfd;~WpMDH)fd4rqK&d}xO#@G)Njf0q6oogTcDaC<Vm!SeJyIMk)?(jW
U%oGn7pDCU$-do&R9G{&0ziM#0kN^Mx

literal 0
HcmV?d00001

diff --git a/model/callbacks.py b/model/callbacks.py
new file mode 100644
index 0000000..0574fd0
--- /dev/null
+++ b/model/callbacks.py
@@ -0,0 +1,74 @@
+from typing import Dict, Any
+import os
+
+import numpy as np
+import pytorch_lightning as pl
+from pytorch_lightning.callbacks import ModelCheckpoint
+from pytorch_lightning.utilities.types import STEP_OUTPUT
+import torch
+import torchvision
+from PIL import Image
+from pytorch_lightning.callbacks import Callback
+from pytorch_lightning.utilities.distributed import rank_zero_only
+
+from .mixins import ImageLoggerMixin
+
+
+__all__ = [
+    "ModelCheckpoint",
+    "ImageLogger"
+]
+
+class ImageLogger(Callback):
+    """
+    Log images during training or validating.
+    
+    TODO: Support validating.
+    """
+    
+    def __init__(
+        self,
+        log_every_n_steps: int=2000,
+        max_images_each_step: int=4,
+        log_images_kwargs: Dict[str, Any]=None
+    ) -> "ImageLogger":
+        super().__init__()
+        self.log_every_n_steps = log_every_n_steps
+        self.max_images_each_step = max_images_each_step
+        self.log_images_kwargs = log_images_kwargs or dict()
+
+    def on_fit_start(self, trainer: pl.Trainer, pl_module: pl.LightningModule) -> None:
+        assert isinstance(pl_module, ImageLoggerMixin)
+
+    @rank_zero_only
+    def on_train_batch_end(
+        self, trainer: pl.Trainer, pl_module: pl.LightningModule, outputs: STEP_OUTPUT,
+        batch: Any, batch_idx: int, dataloader_idx: int
+    ) -> None:
+        if pl_module.global_step % self.log_every_n_steps == 0:
+            is_train = pl_module.training
+            if is_train:
+                pl_module.freeze()
+            
+            with torch.no_grad():
+                # returned images should be: nchw, rgb, [0, 1]
+                images: Dict[str, torch.Tensor] = pl_module.log_images(batch, **self.log_images_kwargs)
+            
+            # save images
+            save_dir = os.path.join(pl_module.logger.save_dir, "image_log", "train")
+            os.makedirs(save_dir, exist_ok=True)
+            for image_key in images:
+                image = images[image_key].detach().cpu()
+                N = min(self.max_images_each_step, len(image))
+                grid = torchvision.utils.make_grid(image[:N], nrow=4)
+                # chw -> hwc (hw if gray)
+                grid = grid.transpose(0, 1).transpose(1, 2).squeeze(-1).numpy()
+                grid = (grid * 255).clip(0, 255).astype(np.uint8)
+                filename = "{}_step-{:06}_e-{:06}_b-{:06}.png".format(
+                    image_key, pl_module.global_step, pl_module.current_epoch, batch_idx
+                )
+                path = os.path.join(save_dir, filename)
+                Image.fromarray(grid).save(path)
+            
+            if is_train:
+                pl_module.unfreeze()
diff --git a/model/ccsr_stage1.py b/model/ccsr_stage1.py
new file mode 100644
index 0000000..a19e97e
--- /dev/null
+++ b/model/ccsr_stage1.py
@@ -0,0 +1,406 @@
+from typing import Mapping, Any
+import copy
+from collections import OrderedDict
+
+import einops
+import torch
+import torch as th
+import torch.nn as nn
+
+from ldm.modules.diffusionmodules.util import (
+    conv_nd,
+    linear,
+    zero_module,
+    timestep_embedding,
+)
+from ldm.modules.attention import SpatialTransformer
+from ldm.modules.diffusionmodules.openaimodel import TimestepEmbedSequential, ResBlock, Downsample, AttentionBlock, UNetModel
+from ldm.models.diffusion.ddpm_ccsr_stage1 import LatentDiffusion
+from ldm.util import log_txt_as_img, exists, instantiate_from_config
+from ldm.modules.distributions.distributions import DiagonalGaussianDistribution
+from utils.common import frozen_module
+from .spaced_sampler import SpacedSampler
+
+class ControlledUnetModel(UNetModel):
+    def forward(self, x, timesteps=None, context=None, control=None, only_mid_control=False, **kwargs):
+        hs = []
+        with torch.no_grad():
+            t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
+            emb = self.time_embed(t_emb)
+            h = x.type(self.dtype)
+            for module in self.input_blocks:
+                h = module(h, emb, context)
+                hs.append(h)
+            h = self.middle_block(h, emb, context)
+
+        if control is not None:
+            h += control.pop()
+
+        for i, module in enumerate(self.output_blocks):
+            if only_mid_control or control is None:
+                h = torch.cat([h, hs.pop()], dim=1)
+            else:
+                h = torch.cat([h, hs.pop() + control.pop()], dim=1)
+            h = module(h, emb, context)
+
+        h = h.type(x.dtype)
+        return self.out(h)
+
+
+class ControlNet(nn.Module):
+    def __init__(
+        self,
+        image_size,
+        in_channels,
+        model_channels,
+        hint_channels,
+        num_res_blocks,
+        attention_resolutions,
+        dropout=0,
+        channel_mult=(1, 2, 4, 8),
+        conv_resample=True,
+        dims=2,
+        use_checkpoint=False,
+        use_fp16=False,
+        num_heads=-1,
+        num_head_channels=-1,
+        num_heads_upsample=-1,
+        use_scale_shift_norm=False,
+        resblock_updown=False,
+        use_new_attention_order=False,
+        use_spatial_transformer=False,  # custom transformer support
+        transformer_depth=1,  # custom transformer support
+        context_dim=None,  # custom transformer support
+        n_embed=None,  # custom support for prediction of discrete ids into codebook of first stage vq model
+        legacy=True,
+        disable_self_attentions=None,
+        num_attention_blocks=None,
+        disable_middle_self_attn=False,
+        use_linear_in_transformer=False,
+    ):
+        super().__init__()
+        if use_spatial_transformer:
+            assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...'
+
+        if context_dim is not None:
+            assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...'
+            from omegaconf.listconfig import ListConfig
+            if type(context_dim) == ListConfig:
+                context_dim = list(context_dim)
+
+        if num_heads_upsample == -1:
+            num_heads_upsample = num_heads
+
+        if num_heads == -1:
+            assert num_head_channels != -1, 'Either num_heads or num_head_channels has to be set'
+
+        if num_head_channels == -1:
+            assert num_heads != -1, 'Either num_heads or num_head_channels has to be set'
+
+        self.dims = dims
+        self.image_size = image_size
+        self.in_channels = in_channels
+        self.model_channels = model_channels
+        if isinstance(num_res_blocks, int):
+            self.num_res_blocks = len(channel_mult) * [num_res_blocks]
+        else:
+            if len(num_res_blocks) != len(channel_mult):
+                raise ValueError("provide num_res_blocks either as an int (globally constant) or "
+                                 "as a list/tuple (per-level) with the same length as channel_mult")
+            self.num_res_blocks = num_res_blocks
+        if disable_self_attentions is not None:
+            # should be a list of booleans, indicating whether to disable self-attention in TransformerBlocks or not
+            assert len(disable_self_attentions) == len(channel_mult)
+        if num_attention_blocks is not None:
+            assert len(num_attention_blocks) == len(self.num_res_blocks)
+            assert all(map(lambda i: self.num_res_blocks[i] >= num_attention_blocks[i], range(len(num_attention_blocks))))
+            print(f"Constructor of UNetModel received num_attention_blocks={num_attention_blocks}. "
+                  f"This option has LESS priority than attention_resolutions {attention_resolutions}, "
+                  f"i.e., in cases where num_attention_blocks[i] > 0 but 2**i not in attention_resolutions, "
+                  f"attention will still not be set.")
+
+        self.attention_resolutions = attention_resolutions
+        self.dropout = dropout
+        self.channel_mult = channel_mult
+        self.conv_resample = conv_resample
+        self.use_checkpoint = use_checkpoint
+        self.dtype = th.float16 if use_fp16 else th.float32
+        self.num_heads = num_heads
+        self.num_head_channels = num_head_channels
+        self.num_heads_upsample = num_heads_upsample
+        self.predict_codebook_ids = n_embed is not None
+
+        time_embed_dim = model_channels * 4
+        self.time_embed = nn.Sequential(
+            linear(model_channels, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+        self.input_blocks = nn.ModuleList(
+            [
+                TimestepEmbedSequential(
+                    conv_nd(dims, in_channels + hint_channels, model_channels, 3, padding=1)
+                )
+            ]
+        )
+        self.zero_convs = nn.ModuleList([self.make_zero_conv(model_channels)])
+
+        self._feature_size = model_channels
+        input_block_chans = [model_channels]
+        ch = model_channels
+        ds = 1
+        for level, mult in enumerate(channel_mult):
+            for nr in range(self.num_res_blocks[level]):
+                layers = [
+                    ResBlock(
+                        ch,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=mult * model_channels,
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = mult * model_channels
+                if ds in attention_resolutions:
+                    if num_head_channels == -1:
+                        dim_head = ch // num_heads
+                    else:
+                        num_heads = ch // num_head_channels
+                        dim_head = num_head_channels
+                    if legacy:
+                        # num_heads = 1
+                        dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+                    if exists(disable_self_attentions):
+                        disabled_sa = disable_self_attentions[level]
+                    else:
+                        disabled_sa = False
+
+                    if not exists(num_attention_blocks) or nr < num_attention_blocks[level]:
+                        layers.append(
+                            AttentionBlock(
+                                ch,
+                                use_checkpoint=use_checkpoint,
+                                num_heads=num_heads,
+                                num_head_channels=dim_head,
+                                use_new_attention_order=use_new_attention_order,
+                            ) if not use_spatial_transformer else SpatialTransformer(
+                                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                                disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer,
+                                use_checkpoint=use_checkpoint
+                            )
+                        )
+                self.input_blocks.append(TimestepEmbedSequential(*layers))
+                self.zero_convs.append(self.make_zero_conv(ch))
+                self._feature_size += ch
+                input_block_chans.append(ch)
+            if level != len(channel_mult) - 1:
+                out_ch = ch
+                self.input_blocks.append(
+                    TimestepEmbedSequential(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            down=True,
+                        )
+                        if resblock_updown
+                        else Downsample(
+                            ch, conv_resample, dims=dims, out_channels=out_ch
+                        )
+                    )
+                )
+                ch = out_ch
+                input_block_chans.append(ch)
+                self.zero_convs.append(self.make_zero_conv(ch))
+                ds *= 2
+                self._feature_size += ch
+
+        if num_head_channels == -1:
+            dim_head = ch // num_heads
+        else:
+            num_heads = ch // num_head_channels
+            dim_head = num_head_channels
+        if legacy:
+            # num_heads = 1
+            dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+        self.middle_block = TimestepEmbedSequential(
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+            AttentionBlock(
+                ch,
+                use_checkpoint=use_checkpoint,
+                num_heads=num_heads,
+                num_head_channels=dim_head,
+                use_new_attention_order=use_new_attention_order,
+            ) if not use_spatial_transformer else SpatialTransformer(  # always uses a self-attn
+                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                disable_self_attn=disable_middle_self_attn, use_linear=use_linear_in_transformer,
+                use_checkpoint=use_checkpoint
+            ),
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+        )
+        self.middle_block_out = self.make_zero_conv(ch)
+        self._feature_size += ch
+
+    def make_zero_conv(self, channels):
+        return TimestepEmbedSequential(zero_module(conv_nd(self.dims, channels, channels, 1, padding=0)))
+
+    def forward(self, x, hint, timesteps, context, **kwargs):
+        t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
+        emb = self.time_embed(t_emb)
+        x = torch.cat((x, hint), dim=1)
+        outs = []
+
+        h = x.type(self.dtype)
+        for module, zero_conv in zip(self.input_blocks, self.zero_convs):
+            h = module(h, emb, context)
+            outs.append(zero_conv(h, emb, context))
+
+        h = self.middle_block(h, emb, context)
+        outs.append(self.middle_block_out(h, emb, context))
+
+        return outs
+
+
+class ControlLDM(LatentDiffusion):
+
+    def __init__(
+        self,
+        control_stage_config: Mapping[str, Any],
+        control_key: str,
+        sd_locked: bool,
+        only_mid_control: bool,
+        learning_rate: float,
+        *args,
+        **kwargs
+    ) -> "ControlLDM":
+        super().__init__(*args, **kwargs)
+        # instantiate control module
+        self.control_model: ControlNet = instantiate_from_config(control_stage_config)
+        self.control_key = control_key
+        self.sd_locked = sd_locked
+        self.only_mid_control = only_mid_control
+        self.learning_rate = learning_rate
+        self.control_scales = [1.0] * 13
+        
+        # instantiate preprocess module (SwinIR)
+        # self.preprocess_model = instantiate_from_config(preprocess_config)
+        # frozen_module(self.preprocess_model)
+        
+        # instantiate condition encoder, since our condition encoder has the same 
+        # structure with AE encoder, we just make a copy of AE encoder. please
+        # note that AE encoder's parameters has not been initialized here.
+        self.cond_encoder = nn.Sequential(OrderedDict([
+            ("encoder", copy.deepcopy(self.first_stage_model.encoder)), # cond_encoder.encoder
+            ("quant_conv", copy.deepcopy(self.first_stage_model.quant_conv)) # cond_encoder.quant_conv
+        ]))
+        frozen_module(self.cond_encoder)
+
+    def apply_condition_encoder(self, control):
+        c_latent_meanvar = self.cond_encoder(control * 2 - 1)
+        c_latent = DiagonalGaussianDistribution(c_latent_meanvar).mode() # only use mode
+        c_latent = c_latent * self.scale_factor
+        return c_latent
+    
+    @torch.no_grad()
+    def get_input(self, batch, k, bs=None, *args, **kwargs):
+        x, c = super().get_input(batch, self.first_stage_key, *args, **kwargs)
+        control = batch[self.control_key]
+        if bs is not None:
+            control = control[:bs]
+        control = control.to(self.device)
+        control = einops.rearrange(control, 'b h w c -> b c h w')
+        control = control.to(memory_format=torch.contiguous_format).float()
+        lq = control
+        # apply preprocess model
+        # control = self.preprocess_model(control)
+        # apply condition encoder
+        c_latent = self.apply_condition_encoder(control)
+        return x, dict(c_crossattn=[c], c_latent=[c_latent], lq=[lq], c_concat=[control])
+
+    def apply_model(self, x_noisy, t, cond, *args, **kwargs):
+        assert isinstance(cond, dict)
+        diffusion_model = self.model.diffusion_model
+
+        cond_txt = torch.cat(cond['c_crossattn'], 1)
+
+        if cond['c_latent'] is None:
+            eps = diffusion_model(x=x_noisy, timesteps=t, context=cond_txt, control=None, only_mid_control=self.only_mid_control)
+        else:
+            control = self.control_model(
+                x=x_noisy, hint=torch.cat(cond['c_latent'], 1),
+                timesteps=t, context=cond_txt
+            )
+            control = [c * scale for c, scale in zip(control, self.control_scales)]
+            eps = diffusion_model(x=x_noisy, timesteps=t, context=cond_txt, control=control, only_mid_control=self.only_mid_control)
+
+        return eps
+
+    @torch.no_grad()
+    def get_unconditional_conditioning(self, N):
+        return self.get_learned_conditioning([""] * N)
+
+    @torch.no_grad()
+    def log_images(self, batch, sample_steps=50):
+        log = dict()
+        z, c = self.get_input(batch, self.first_stage_key)
+        c_lq = c["lq"][0]
+        c_latent = c["c_latent"][0]
+        c_cat, c = c["c_concat"][0], c["c_crossattn"][0]
+
+        log["hq"] = (self.decode_first_stage(z) + 1) / 2
+        log["control"] = c_cat
+        log["decoded_control"] = (self.decode_first_stage(c_latent) + 1) / 2
+        log["lq"] = c_lq
+        log["text"] = (log_txt_as_img((512, 512), batch[self.cond_stage_key], size=16) + 1) / 2
+        log["samples"] = self.sample_log(
+            cond_img=c_cat,            
+            steps=sample_steps
+        )
+        return log
+
+    @torch.no_grad()
+    def sample_log(self, cond_img, steps):
+        sampler = SpacedSampler(self)
+        b, c, h, w = cond_img.shape
+        shape = (b, self.channels, h // 8, w // 8)
+        samples = sampler.sample(
+            steps, shape, cond_img, positive_prompt="", negative_prompt="",
+            cfg_scale=1.0, color_fix_type="wavelet"
+        )
+        return samples
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params = list(self.control_model.parameters())
+        params_2 = list(self.control_model.parameters())
+        if not self.sd_locked:
+            params += list(self.model.diffusion_model.output_blocks.parameters())
+            params += list(self.model.diffusion_model.out.parameters())
+        opt = torch.optim.AdamW(params, lr=lr)
+        opt2 = torch.optim.AdamW(params_2, lr=lr)
+        return opt, opt2
+
+    def validation_step(self, batch, batch_idx):
+        # TODO: 
+        pass
\ No newline at end of file
diff --git a/model/ccsr_stage2.py b/model/ccsr_stage2.py
new file mode 100644
index 0000000..9beaf31
--- /dev/null
+++ b/model/ccsr_stage2.py
@@ -0,0 +1,407 @@
+from typing import Mapping, Any
+import copy
+from collections import OrderedDict
+
+import einops
+import torch
+import torch as th
+import torch.nn as nn
+
+from ldm.modules.diffusionmodules.util import (
+    conv_nd,
+    linear,
+    zero_module,
+    timestep_embedding,
+)
+from ldm.modules.attention import SpatialTransformer
+from ldm.modules.diffusionmodules.openaimodel import TimestepEmbedSequential, ResBlock, Downsample, AttentionBlock, UNetModel
+from ldm.models.diffusion.ddpm_ccsr_stage2 import LatentDiffusion
+from ldm.util import log_txt_as_img, exists, instantiate_from_config
+from ldm.modules.distributions.distributions import DiagonalGaussianDistribution
+from utils.common import frozen_module
+from .spaced_sampler import SpacedSampler
+
+class ControlledUnetModel(UNetModel):
+    def forward(self, x, timesteps=None, context=None, control=None, only_mid_control=False, **kwargs):
+        hs = []
+        with torch.no_grad():
+            t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
+            emb = self.time_embed(t_emb)
+            h = x.type(self.dtype)
+            for module in self.input_blocks:
+                h = module(h, emb, context)
+                hs.append(h)
+            h = self.middle_block(h, emb, context)
+
+        if control is not None:
+            h += control.pop()
+
+        for i, module in enumerate(self.output_blocks):
+            if only_mid_control or control is None:
+                h = torch.cat([h, hs.pop()], dim=1)
+            else:
+                h = torch.cat([h, hs.pop() + control.pop()], dim=1)
+            h = module(h, emb, context)
+
+        h = h.type(x.dtype)
+        return self.out(h)
+
+
+class ControlNet(nn.Module):
+    def __init__(
+        self,
+        image_size,
+        in_channels,
+        model_channels,
+        hint_channels,
+        num_res_blocks,
+        attention_resolutions,
+        dropout=0,
+        channel_mult=(1, 2, 4, 8),
+        conv_resample=True,
+        dims=2,
+        use_checkpoint=False,
+        use_fp16=False,
+        num_heads=-1,
+        num_head_channels=-1,
+        num_heads_upsample=-1,
+        use_scale_shift_norm=False,
+        resblock_updown=False,
+        use_new_attention_order=False,
+        use_spatial_transformer=False,  # custom transformer support
+        transformer_depth=1,  # custom transformer support
+        context_dim=None,  # custom transformer support
+        n_embed=None,  # custom support for prediction of discrete ids into codebook of first stage vq model
+        legacy=True,
+        disable_self_attentions=None,
+        num_attention_blocks=None,
+        disable_middle_self_attn=False,
+        use_linear_in_transformer=False,
+    ):
+        super().__init__()
+        if use_spatial_transformer:
+            assert context_dim is not None, 'Fool!! You forgot to include the dimension of your cross-attention conditioning...'
+
+        if context_dim is not None:
+            assert use_spatial_transformer, 'Fool!! You forgot to use the spatial transformer for your cross-attention conditioning...'
+            from omegaconf.listconfig import ListConfig
+            if type(context_dim) == ListConfig:
+                context_dim = list(context_dim)
+
+        if num_heads_upsample == -1:
+            num_heads_upsample = num_heads
+
+        if num_heads == -1:
+            assert num_head_channels != -1, 'Either num_heads or num_head_channels has to be set'
+
+        if num_head_channels == -1:
+            assert num_heads != -1, 'Either num_heads or num_head_channels has to be set'
+
+        self.dims = dims
+        self.image_size = image_size
+        self.in_channels = in_channels
+        self.model_channels = model_channels
+        if isinstance(num_res_blocks, int):
+            self.num_res_blocks = len(channel_mult) * [num_res_blocks]
+        else:
+            if len(num_res_blocks) != len(channel_mult):
+                raise ValueError("provide num_res_blocks either as an int (globally constant) or "
+                                 "as a list/tuple (per-level) with the same length as channel_mult")
+            self.num_res_blocks = num_res_blocks
+        if disable_self_attentions is not None:
+            # should be a list of booleans, indicating whether to disable self-attention in TransformerBlocks or not
+            assert len(disable_self_attentions) == len(channel_mult)
+        if num_attention_blocks is not None:
+            assert len(num_attention_blocks) == len(self.num_res_blocks)
+            assert all(map(lambda i: self.num_res_blocks[i] >= num_attention_blocks[i], range(len(num_attention_blocks))))
+            print(f"Constructor of UNetModel received num_attention_blocks={num_attention_blocks}. "
+                  f"This option has LESS priority than attention_resolutions {attention_resolutions}, "
+                  f"i.e., in cases where num_attention_blocks[i] > 0 but 2**i not in attention_resolutions, "
+                  f"attention will still not be set.")
+
+        self.attention_resolutions = attention_resolutions
+        self.dropout = dropout
+        self.channel_mult = channel_mult
+        self.conv_resample = conv_resample
+        self.use_checkpoint = use_checkpoint
+        self.dtype = th.float16 if use_fp16 else th.float32
+        self.num_heads = num_heads
+        self.num_head_channels = num_head_channels
+        self.num_heads_upsample = num_heads_upsample
+        self.predict_codebook_ids = n_embed is not None
+
+        time_embed_dim = model_channels * 4
+        self.time_embed = nn.Sequential(
+            linear(model_channels, time_embed_dim),
+            nn.SiLU(),
+            linear(time_embed_dim, time_embed_dim),
+        )
+
+        self.input_blocks = nn.ModuleList(
+            [
+                TimestepEmbedSequential(
+                    conv_nd(dims, in_channels + hint_channels, model_channels, 3, padding=1)
+                )
+            ]
+        )
+        self.zero_convs = nn.ModuleList([self.make_zero_conv(model_channels)])
+
+        self._feature_size = model_channels
+        input_block_chans = [model_channels]
+        ch = model_channels
+        ds = 1
+        for level, mult in enumerate(channel_mult):
+            for nr in range(self.num_res_blocks[level]):
+                layers = [
+                    ResBlock(
+                        ch,
+                        time_embed_dim,
+                        dropout,
+                        out_channels=mult * model_channels,
+                        dims=dims,
+                        use_checkpoint=use_checkpoint,
+                        use_scale_shift_norm=use_scale_shift_norm,
+                    )
+                ]
+                ch = mult * model_channels
+                if ds in attention_resolutions:
+                    if num_head_channels == -1:
+                        dim_head = ch // num_heads
+                    else:
+                        num_heads = ch // num_head_channels
+                        dim_head = num_head_channels
+                    if legacy:
+                        # num_heads = 1
+                        dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+                    if exists(disable_self_attentions):
+                        disabled_sa = disable_self_attentions[level]
+                    else:
+                        disabled_sa = False
+
+                    if not exists(num_attention_blocks) or nr < num_attention_blocks[level]:
+                        layers.append(
+                            AttentionBlock(
+                                ch,
+                                use_checkpoint=use_checkpoint,
+                                num_heads=num_heads,
+                                num_head_channels=dim_head,
+                                use_new_attention_order=use_new_attention_order,
+                            ) if not use_spatial_transformer else SpatialTransformer(
+                                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                                disable_self_attn=disabled_sa, use_linear=use_linear_in_transformer,
+                                use_checkpoint=use_checkpoint
+                            )
+                        )
+                self.input_blocks.append(TimestepEmbedSequential(*layers))
+                self.zero_convs.append(self.make_zero_conv(ch))
+                self._feature_size += ch
+                input_block_chans.append(ch)
+            if level != len(channel_mult) - 1:
+                out_ch = ch
+                self.input_blocks.append(
+                    TimestepEmbedSequential(
+                        ResBlock(
+                            ch,
+                            time_embed_dim,
+                            dropout,
+                            out_channels=out_ch,
+                            dims=dims,
+                            use_checkpoint=use_checkpoint,
+                            use_scale_shift_norm=use_scale_shift_norm,
+                            down=True,
+                        )
+                        if resblock_updown
+                        else Downsample(
+                            ch, conv_resample, dims=dims, out_channels=out_ch
+                        )
+                    )
+                )
+                ch = out_ch
+                input_block_chans.append(ch)
+                self.zero_convs.append(self.make_zero_conv(ch))
+                ds *= 2
+                self._feature_size += ch
+
+        if num_head_channels == -1:
+            dim_head = ch // num_heads
+        else:
+            num_heads = ch // num_head_channels
+            dim_head = num_head_channels
+        if legacy:
+            # num_heads = 1
+            dim_head = ch // num_heads if use_spatial_transformer else num_head_channels
+        self.middle_block = TimestepEmbedSequential(
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+            AttentionBlock(
+                ch,
+                use_checkpoint=use_checkpoint,
+                num_heads=num_heads,
+                num_head_channels=dim_head,
+                use_new_attention_order=use_new_attention_order,
+            ) if not use_spatial_transformer else SpatialTransformer(  # always uses a self-attn
+                ch, num_heads, dim_head, depth=transformer_depth, context_dim=context_dim,
+                disable_self_attn=disable_middle_self_attn, use_linear=use_linear_in_transformer,
+                use_checkpoint=use_checkpoint
+            ),
+            ResBlock(
+                ch,
+                time_embed_dim,
+                dropout,
+                dims=dims,
+                use_checkpoint=use_checkpoint,
+                use_scale_shift_norm=use_scale_shift_norm,
+            ),
+        )
+        self.middle_block_out = self.make_zero_conv(ch)
+        self._feature_size += ch
+
+    def make_zero_conv(self, channels):
+        return TimestepEmbedSequential(zero_module(conv_nd(self.dims, channels, channels, 1, padding=0)))
+
+    def forward(self, x, hint, timesteps, context, **kwargs):
+        t_emb = timestep_embedding(timesteps, self.model_channels, repeat_only=False)
+        emb = self.time_embed(t_emb)
+        x = torch.cat((x, hint), dim=1)
+        outs = []
+
+        h = x.type(self.dtype)
+        for module, zero_conv in zip(self.input_blocks, self.zero_convs):
+            h = module(h, emb, context)
+            outs.append(zero_conv(h, emb, context))
+
+        h = self.middle_block(h, emb, context)
+        outs.append(self.middle_block_out(h, emb, context))
+
+        return outs
+
+
+class ControlLDM(LatentDiffusion):
+
+    def __init__(
+        self,
+        control_stage_config: Mapping[str, Any],
+        control_key: str,
+        sd_locked: bool,
+        only_mid_control: bool,
+        learning_rate: float,
+        *args,
+        **kwargs
+    ) -> "ControlLDM":
+        super().__init__(*args, **kwargs)
+        # instantiate control module
+        self.control_model: ControlNet = instantiate_from_config(control_stage_config)
+        self.control_key = control_key
+        self.sd_locked = sd_locked
+        self.only_mid_control = only_mid_control
+        self.learning_rate = learning_rate
+        self.control_scales = [1.0] * 13
+        
+        # instantiate preprocess module (SwinIR)
+        # self.preprocess_model = instantiate_from_config(preprocess_config)
+        # frozen_module(self.preprocess_model)
+        
+        # instantiate condition encoder, since our condition encoder has the same 
+        # structure with AE encoder, we just make a copy of AE encoder. please
+        # note that AE encoder's parameters has not been initialized here.
+        self.cond_encoder = nn.Sequential(OrderedDict([
+            ("encoder", copy.deepcopy(self.first_stage_model.encoder)), # cond_encoder.encoder
+            ("quant_conv", copy.deepcopy(self.first_stage_model.quant_conv)) # cond_encoder.quant_conv
+        ]))
+        frozen_module(self.cond_encoder)
+
+    def apply_condition_encoder(self, control):
+        c_latent_meanvar = self.cond_encoder(control * 2 - 1)
+        c_latent = DiagonalGaussianDistribution(c_latent_meanvar).mode() # only use mode
+        c_latent = c_latent * self.scale_factor
+        return c_latent
+    
+    @torch.no_grad()
+    def get_input(self, batch, k, bs=None, *args, **kwargs):
+        gt, x, c = super().get_input(batch, self.first_stage_key, *args, **kwargs)
+        control = batch[self.control_key]
+        if bs is not None:
+            control = control[:bs]
+        control = control.to(self.device)
+        control = einops.rearrange(control, 'b h w c -> b c h w')
+        control = control.to(memory_format=torch.contiguous_format).float()
+        lq = control
+        # apply preprocess model
+        # control = self.preprocess_model(control)
+        # apply condition encoder
+        c_latent = self.apply_condition_encoder(control)
+        return gt, x, dict(c_crossattn=[c], c_latent=[c_latent], lq=[lq], c_concat=[control])
+
+    def apply_model(self, x_noisy, t, cond, *args, **kwargs):
+        assert isinstance(cond, dict)
+        diffusion_model = self.model.diffusion_model
+
+        cond_txt = torch.cat(cond['c_crossattn'], 1)
+
+        if cond['c_latent'] is None:
+            eps = diffusion_model(x=x_noisy, timesteps=t, context=cond_txt, control=None, only_mid_control=self.only_mid_control)
+        else:
+            control = self.control_model(
+                x=x_noisy, hint=torch.cat(cond['c_latent'], 1),
+                timesteps=t, context=cond_txt
+            )
+            control = [c * scale for c, scale in zip(control, self.control_scales)]
+            eps = diffusion_model(x=x_noisy, timesteps=t, context=cond_txt, control=control, only_mid_control=self.only_mid_control)
+
+        return eps
+
+    @torch.no_grad()
+    def get_unconditional_conditioning(self, N):
+        return self.get_learned_conditioning([""] * N)
+
+    @torch.no_grad()
+    def log_images(self, batch, sample_steps=50):
+        log = dict()
+        gt, z, c = self.get_input(batch, self.first_stage_key)
+        c_lq = c["lq"][0]
+        c_latent = c["c_latent"][0]
+        c_cat, c = c["c_concat"][0], c["c_crossattn"][0]
+
+        log["hq"] = (self.decode_first_stage(z) + 1) / 2
+        log["control"] = c_cat
+        log["decoded_control"] = (self.decode_first_stage(c_latent) + 1) / 2
+        log["lq"] = c_lq
+        log["text"] = (log_txt_as_img((512, 512), batch[self.cond_stage_key], size=16) + 1) / 2
+        log["samples"] = self.sample_log(
+            cond_img=c_cat,            
+            steps=sample_steps
+        )
+        return log
+
+    @torch.no_grad()
+    def sample_log(self, cond_img, steps):
+        sampler = SpacedSampler(self)
+        b, c, h, w = cond_img.shape
+        shape = (b, self.channels, h // 8, w // 8)
+        samples = sampler.sample(
+            steps, shape, cond_img, positive_prompt="", negative_prompt="",
+            cfg_scale=1.0, color_fix_type="wavelet"
+        )
+        return samples
+
+    def configure_optimizers(self):
+        lr = self.learning_rate
+        params_decode = list(self.first_stage_model.post_quant_conv.parameters()) + list(self.first_stage_model.decoder.parameters())
+        # discriminator
+        params_discriminator = list(self.decoder_loss.discriminator.parameters())
+        if not self.sd_locked:
+            params += list(self.model.diffusion_model.output_blocks.parameters())
+            params += list(self.model.diffusion_model.out.parameters())
+        opt = torch.optim.Adam(params_decode, lr=lr, betas=(0.5, 0.9))
+        opt2 = torch.optim.Adam(params_discriminator, lr=lr, betas=(0.5, 0.9))
+        return opt, opt2
+
+    def validation_step(self, batch, batch_idx):
+        # TODO: 
+        pass
\ No newline at end of file
diff --git a/model/cond_fn.py b/model/cond_fn.py
new file mode 100644
index 0000000..f9283f3
--- /dev/null
+++ b/model/cond_fn.py
@@ -0,0 +1,78 @@
+from typing import overload, Optional
+import torch
+from torch.nn import functional as F
+
+
+class Guidance:
+
+    def __init__(
+        self,
+        scale: float,
+        t_start: int,
+        t_stop: int,
+        space: str,
+        repeat: int
+    ) -> "Guidance":
+        """
+        Initialize latent image guidance.
+        
+        Args:
+            scale (float): Gradient scale (denoted as `s` in our paper). The larger the gradient scale, 
+                the closer the final result will be to the output of the first stage model.
+            t_start (int), t_stop (int): The timestep to start or stop guidance. Note that the sampling 
+                process starts from t=1000 to t=0, the `t_start` should be larger than `t_stop`.
+            space (str): The data space for computing loss function (rgb or latent).
+            repeat (int): Repeat gradient descent for `repeat` times.
+
+        Our latent image guidance is based on [GDP](https://github.com/Fayeben/GenerativeDiffusionPrior).
+        Thanks for their work!
+        """
+        self.scale = scale
+        self.t_start = t_start
+        self.t_stop = t_stop
+        self.target = None
+        self.space = space
+        self.repeat = repeat
+    
+    def load_target(self, target: torch.Tensor) -> torch.Tensor:
+        self.target = target
+
+    def __call__(self, target_x0: torch.Tensor, pred_x0: torch.Tensor, t: int) -> Optional[torch.Tensor]:
+        if self.t_stop < t and t < self.t_start:
+            # print("sampling with classifier guidance")
+            # avoid propagating gradient out of this scope
+            pred_x0 = pred_x0.detach().clone()
+            target_x0 = target_x0.detach().clone()
+            return self.scale * self._forward(target_x0, pred_x0)
+        else:
+            return None
+    
+    @overload
+    def _forward(self, target_x0: torch.Tensor, pred_x0: torch.Tensor) -> torch.Tensor:
+        ...
+
+
+class MSEGuidance(Guidance):
+    
+    def __init__(
+        self,
+        scale: float,
+        t_start: int,
+        t_stop: int,
+        space: str,
+        repeat: int
+    ) -> "MSEGuidance":
+        super().__init__(
+            scale, t_start, t_stop, space, repeat
+        )
+    
+    @torch.enable_grad()
+    def _forward(self, target_x0: torch.Tensor, pred_x0: torch.Tensor) -> torch.Tensor:
+        # inputs: [-1, 1], nchw, rgb
+        pred_x0.requires_grad_(True)
+        
+        # This is what we actually use.
+        loss = (pred_x0 - target_x0).pow(2).mean((1, 2, 3)).sum()
+        
+        print(f"loss = {loss.item()}")
+        return -torch.autograd.grad(loss, pred_x0)[0]
diff --git a/model/mixins.py b/model/mixins.py
new file mode 100644
index 0000000..a86d3c2
--- /dev/null
+++ b/model/mixins.py
@@ -0,0 +1,9 @@
+from typing import overload, Any, Dict
+import torch
+
+
+class ImageLoggerMixin:
+
+    @overload
+    def log_images(self, batch: Any, **kwargs: Dict[str, Any]) -> Dict[str, torch.Tensor]:
+        ...
diff --git a/model/q_sampler.py b/model/q_sampler.py
new file mode 100644
index 0000000..8f0c9dc
--- /dev/null
+++ b/model/q_sampler.py
@@ -0,0 +1,1043 @@
+from typing import Optional, Tuple, Dict, List, Callable
+
+import torch
+import numpy as np
+from tqdm import tqdm
+import einops
+import os
+from PIL import Image
+
+from ldm.modules.diffusionmodules.util import make_beta_schedule
+from model.cond_fn import Guidance
+from utils.image import (
+    wavelet_reconstruction, adaptive_instance_normalization
+)
+
+# https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/respace.py
+def space_timesteps(num_timesteps, section_counts):
+    """
+    Create a list of timesteps to use from an original diffusion process,
+    given the number of timesteps we want to take from equally-sized portions
+    of the original process.
+    For example, if there's 300 timesteps and the section counts are [10,15,20]
+    then the first 100 timesteps are strided to be 10 timesteps, the second 100
+    are strided to be 15 timesteps, and the final 100 are strided to be 20.
+    If the stride is a string starting with "ddim", then the fixed striding
+    from the DDIM paper is used, and only one section is allowed.
+    :param num_timesteps: the number of diffusion steps in the original
+                          process to divide up.
+    :param section_counts: either a list of numbers, or a string containing
+                           comma-separated numbers, indicating the step count
+                           per section. As a special case, use "ddimN" where N
+                           is a number of steps to use the striding from the
+                           DDIM paper.
+    :return: a set of diffusion steps from the original process to use.
+    """
+    if isinstance(section_counts, str):
+        if section_counts.startswith("ddim"):
+            desired_count = int(section_counts[len("ddim") :])
+            for i in range(1, num_timesteps):
+                if len(range(0, num_timesteps, i)) == desired_count:
+                    return set(range(0, num_timesteps, i))
+            raise ValueError(
+                f"cannot create exactly {num_timesteps} steps with an integer stride"
+            )
+        section_counts = [int(x) for x in section_counts.split(",")]
+    size_per = num_timesteps // len(section_counts)
+    extra = num_timesteps % len(section_counts)
+    start_idx = 0
+    all_steps = []
+    for i, section_count in enumerate(section_counts):
+        size = size_per + (1 if i < extra else 0)
+        if size < section_count:
+            raise ValueError(
+                f"cannot divide section of {size} steps into {section_count}"
+            )
+        if section_count <= 1:
+            frac_stride = 1
+        else:
+            frac_stride = (size - 1) / (section_count - 1)
+        cur_idx = 0.0
+        taken_steps = []
+        for _ in range(section_count):
+            taken_steps.append(start_idx + round(cur_idx))
+            cur_idx += frac_stride
+        all_steps += taken_steps
+        start_idx += size
+    return set(all_steps)
+
+
+# https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/gaussian_diffusion.py
+def _extract_into_tensor(arr, timesteps, broadcast_shape):
+    """
+    Extract values from a 1-D numpy array for a batch of indices.
+    :param arr: the 1-D numpy array.
+    :param timesteps: a tensor of indices into the array to extract.
+    :param broadcast_shape: a larger shape of K dimensions with the batch
+                            dimension equal to the length of timesteps.
+    :return: a tensor of shape [batch_size, 1, ...] where the shape has K dims.
+    """
+    try:
+        # float64 as default. float64 is not supported by mps device.
+        res = torch.from_numpy(arr).to(device=timesteps.device)[timesteps].float()
+    except:
+        # to be compatible with mps
+        res = torch.from_numpy(arr.astype(np.float32)).to(device=timesteps.device)[timesteps].float()
+        
+    while len(res.shape) < len(broadcast_shape):
+        res = res[..., None]
+    return res.expand(broadcast_shape)
+
+
+class SpacedSampler:
+    """
+    Implementation for spaced sampling schedule proposed in IDDPM. This class is designed
+    for sampling ControlLDM.
+    
+    https://arxiv.org/pdf/2102.09672.pdf
+    """
+    
+    def __init__(
+        self,
+        model: "ControlLDM",
+        schedule: str="linear",
+        var_type: str="fixed_small"
+    ) -> "SpacedSampler":
+        self.model = model
+        self.original_num_steps = model.num_timesteps
+        self.schedule = schedule
+        self.var_type = var_type
+
+    def make_schedule(self, num_steps: int) -> None:
+        """
+        Initialize sampling parameters according to `num_steps`.
+        
+        Args:
+            num_steps (int): Sampling steps.
+
+        Returns:
+            None
+        """
+        # NOTE: this schedule, which generates betas linearly in log space, is a little different
+        # from guided diffusion.
+        original_betas = make_beta_schedule(
+            self.schedule, self.original_num_steps, linear_start=self.model.linear_start,
+            linear_end=self.model.linear_end
+        )
+        original_alphas = 1.0 - original_betas
+        original_alphas_cumprod = np.cumprod(original_alphas, axis=0)
+        
+        # calcualte betas for spaced sampling
+        # https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/respace.py
+        used_timesteps = space_timesteps(self.original_num_steps, str(num_steps))
+        print(f"timesteps used in spaced sampler: \n\t{sorted(list(used_timesteps))}")
+        
+        betas = []
+        last_alpha_cumprod = 1.0
+        for i, alpha_cumprod in enumerate(original_alphas_cumprod):
+            if i in used_timesteps:
+                # marginal distribution is the same as q(x_{S_t}|x_0)
+                betas.append(1 - alpha_cumprod / last_alpha_cumprod)
+                last_alpha_cumprod = alpha_cumprod
+        assert len(betas) == num_steps
+        betas = np.array(betas, dtype=np.float64)
+        self.betas = betas
+        
+        self.timesteps = np.array(sorted(list(used_timesteps)), dtype=np.int32) # e.g. [0, 10, 20, ...]
+        alphas = 1.0 - betas
+        self.alphas_cumprod = np.cumprod(alphas, axis=0)
+        self.alphas_cumprod_prev = np.append(1.0, self.alphas_cumprod[:-1])
+        self.alphas_cumprod_next = np.append(self.alphas_cumprod[1:], 0.0)
+        assert self.alphas_cumprod_prev.shape == (num_steps, )
+        
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.sqrt_alphas_cumprod = np.sqrt(self.alphas_cumprod)
+        self.sqrt_one_minus_alphas_cumprod = np.sqrt(1.0 - self.alphas_cumprod)
+        self.log_one_minus_alphas_cumprod = np.log(1.0 - self.alphas_cumprod)
+        self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod)
+        self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1)
+
+        # calculations for posterior q(x_{t-1} | x_t, x_0)
+        self.posterior_variance = (
+            betas * (1.0 - self.alphas_cumprod_prev) / (1.0 - self.alphas_cumprod)
+        )
+        # log calculation clipped because the posterior variance is 0 at the
+        # beginning of the diffusion chain.
+        self.posterior_log_variance_clipped = np.log(
+            np.append(self.posterior_variance[1], self.posterior_variance[1:])
+        )
+        self.posterior_mean_coef1 = (
+            betas * np.sqrt(self.alphas_cumprod_prev) / (1.0 - self.alphas_cumprod)
+        )
+        self.posterior_mean_coef2 = (
+            (1.0 - self.alphas_cumprod_prev)
+            * np.sqrt(alphas)
+            / (1.0 - self.alphas_cumprod)
+        )
+    
+    def make_tao_schedule(self, num_steps: int) -> None:
+        """
+        Initialize sampling parameters according to `num_steps`.
+        
+        Args:
+            num_steps (int): Sampling steps.
+
+        Returns:
+            None
+        """
+        # NOTE: this schedule, which generates betas linearly in log space, is a little different
+        # from guided diffusion.
+        original_betas = make_beta_schedule(
+            self.schedule, self.original_num_steps, linear_start=self.model.linear_start,
+            linear_end=self.model.linear_end
+        )
+        original_alphas = 1.0 - original_betas
+        original_alphas_cumprod = np.cumprod(original_alphas, axis=0)
+        
+        # calcualte betas for spaced sampling
+        # https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/respace.py
+        used_timesteps = space_timesteps(self.original_num_steps, str(num_steps))
+        print(f"timesteps used in spaced sampler: \n\t{sorted(list(used_timesteps))}")
+        
+        betas = []
+        last_alpha_cumprod = 1.0
+        for i, alpha_cumprod in enumerate(original_alphas_cumprod):
+            if i in used_timesteps:
+                # marginal distribution is the same as q(x_{S_t}|x_0)
+                betas.append(1 - alpha_cumprod / last_alpha_cumprod)
+                last_alpha_cumprod = alpha_cumprod
+        assert len(betas) == num_steps
+        betas = np.array(betas, dtype=np.float64)
+        self.tao_betas = betas
+        
+        self.tao_timesteps = np.array(sorted(list(used_timesteps)), dtype=np.int32) # e.g. [0, 10, 20, ...]
+        alphas = 1.0 - betas
+        self.tao_alphas_cumprod = np.cumprod(alphas, axis=0)
+        self.tao_alphas_cumprod_prev = np.append(1.0, self.tao_alphas_cumprod[:-1])
+        self.tao_alphas_cumprod_next = np.append(self.tao_alphas_cumprod[1:], 0.0)
+        assert self.tao_alphas_cumprod_prev.shape == (num_steps, )
+        
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.tao_sqrt_alphas_cumprod = np.sqrt(self.tao_alphas_cumprod)
+        self.tao_sqrt_one_minus_alphas_cumprod = np.sqrt(1.0 - self.tao_alphas_cumprod)
+        self.tao_log_one_minus_alphas_cumprod = np.log(1.0 - self.tao_alphas_cumprod)
+        self.tao_sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.tao_alphas_cumprod)
+        self.tao_sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.tao_alphas_cumprod - 1)
+
+        # calculations for posterior q(x_{t-1} | x_t, x_0)
+        self.tao_posterior_variance = (
+            betas * (1.0 - self.tao_alphas_cumprod_prev) / (1.0 - self.tao_alphas_cumprod)
+        )
+        # log calculation clipped because the posterior variance is 0 at the
+        # beginning of the diffusion chain.
+        self.tao_posterior_log_variance_clipped = np.log(
+            np.append(self.tao_posterior_variance[1], self.tao_posterior_variance[1:])
+        )
+        self.tao_posterior_mean_coef1 = (
+            betas * np.sqrt(self.tao_alphas_cumprod_prev) / (1.0 - self.tao_alphas_cumprod)
+        )
+        self.tao_posterior_mean_coef2 = (
+            (1.0 - self.tao_alphas_cumprod_prev)
+            * np.sqrt(alphas)
+            / (1.0 - self.tao_alphas_cumprod)
+        )
+
+    def q_sample(
+        self,
+        x_start: torch.Tensor,
+        t: torch.Tensor,
+        noise: Optional[torch.Tensor]=None
+    ) -> torch.Tensor:
+        """
+        Implement the marginal distribution q(x_t|x_0).
+
+        Args:
+            x_start (torch.Tensor): Images (NCHW) sampled from data distribution.
+            t (torch.Tensor): Timestep (N) for diffusion process. `t` serves as an index
+                to get parameters for each timestep.
+            noise (torch.Tensor, optional): Specify the noise (NCHW) added to `x_start`.
+
+        Returns:
+            x_t (torch.Tensor): The noisy images.
+        """
+        if noise is None:
+            noise = torch.randn_like(x_start)
+        assert noise.shape == x_start.shape
+        return (
+            _extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start
+            + _extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape)
+            * noise
+        )
+
+    def q_posterior_mean_variance(
+        self,
+        x_start: torch.Tensor,
+        x_t: torch.Tensor,
+        t: torch.Tensor
+    ) -> Tuple[torch.Tensor]:
+        """
+        Implement the posterior distribution q(x_{t-1}|x_t, x_0).
+        
+        Args:
+            x_start (torch.Tensor): The predicted images (NCHW) in timestep `t`.
+            x_t (torch.Tensor): The sampled intermediate variables (NCHW) of timestep `t`.
+            t (torch.Tensor): Timestep (N) of `x_t`. `t` serves as an index to get 
+                parameters for each timestep.
+        
+        Returns:
+            posterior_mean (torch.Tensor): Mean of the posterior distribution.
+            posterior_variance (torch.Tensor): Variance of the posterior distribution.
+            posterior_log_variance_clipped (torch.Tensor): Log variance of the posterior distribution.
+        """
+        assert x_start.shape == x_t.shape
+        posterior_mean = (
+            _extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start
+            + _extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t
+        )
+        posterior_variance = _extract_into_tensor(self.posterior_variance, t, x_t.shape)
+        posterior_log_variance_clipped = _extract_into_tensor(
+            self.posterior_log_variance_clipped, t, x_t.shape
+        )
+        assert (
+            posterior_mean.shape[0]
+            == posterior_variance.shape[0]
+            == posterior_log_variance_clipped.shape[0]
+            == x_start.shape[0]
+        )
+        return posterior_mean, posterior_variance, posterior_log_variance_clipped
+    
+    def q_posterior_tao_mean_variance(
+        self,
+        x_start: torch.Tensor,
+        x_t: torch.Tensor,
+        t: torch.Tensor
+    ) -> Tuple[torch.Tensor]:
+        """
+        Implement the posterior distribution q(x_{t-1}|x_t, x_0).
+        
+        Args:
+            x_start (torch.Tensor): The predicted images (NCHW) in timestep `t`.
+            x_t (torch.Tensor): The sampled intermediate variables (NCHW) of timestep `t`.
+            t (torch.Tensor): Timestep (N) of `x_t`. `t` serves as an index to get 
+                parameters for each timestep.
+        
+        Returns:
+            posterior_mean (torch.Tensor): Mean of the posterior distribution.
+            posterior_variance (torch.Tensor): Variance of the posterior distribution.
+            posterior_log_variance_clipped (torch.Tensor): Log variance of the posterior distribution.
+        """
+        assert x_start.shape == x_t.shape
+        posterior_mean = (
+            _extract_into_tensor(self.tao_posterior_mean_coef1, t, x_t.shape) * x_start
+            + _extract_into_tensor(self.tao_posterior_mean_coef2, t, x_t.shape) * x_t
+        )
+        posterior_variance = _extract_into_tensor(self.tao_posterior_variance, t, x_t.shape)
+        posterior_log_variance_clipped = _extract_into_tensor(
+            self.tao_posterior_log_variance_clipped, t, x_t.shape
+        )
+        assert (
+            posterior_mean.shape[0]
+            == posterior_variance.shape[0]
+            == posterior_log_variance_clipped.shape[0]
+            == x_start.shape[0]
+        )
+        return posterior_mean, posterior_variance, posterior_log_variance_clipped
+
+    def _predict_xstart_from_eps(
+        self,
+        x_t: torch.Tensor,
+        t: torch.Tensor,
+        eps: torch.Tensor
+    ) -> torch.Tensor:
+        assert x_t.shape == eps.shape
+        return (
+            _extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t
+            - _extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * eps
+        )
+    
+    def predict_noise(
+        self,
+        x: torch.Tensor,
+        t: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]]
+    ) -> torch.Tensor:
+        if uncond is None or cfg_scale == 1.:
+            model_output = self.model.apply_model(x, t, cond)
+        else:
+            # apply classifier-free guidance
+            model_cond = self.model.apply_model(x, t, cond)
+            model_uncond = self.model.apply_model(x, t, uncond)
+            model_output = model_uncond + cfg_scale * (model_cond - model_uncond)
+        
+        if self.model.parameterization == "v":
+            e_t = self.model.predict_eps_from_z_and_v(x, t, model_output)
+        else:
+            e_t = model_output
+
+        return e_t
+    
+    def apply_cond_fn(
+        self,
+        x: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        t: torch.Tensor,
+        index: torch.Tensor,
+        cond_fn: Guidance,
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]]
+    ) -> torch.Tensor:
+        device = x.device
+        t_now = int(t[0].item()) + 1
+        # ----------------- predict noise and x0 ----------------- #
+        e_t = self.predict_noise(
+            x, t, cond, cfg_scale, uncond
+        )
+        pred_x0: torch.Tensor = self._predict_xstart_from_eps(x_t=x, t=index, eps=e_t)
+        model_mean, _, _ = self.q_posterior_mean_variance(
+            x_start=pred_x0, x_t=x, t=index
+        )
+        
+        # apply classifier guidance for multiple times
+        for _ in range(cond_fn.repeat):
+            # ----------------- compute gradient for x0 in latent space ----------------- #
+            target, pred = None, None
+            if cond_fn.space == "latent":
+                target = self.model.get_first_stage_encoding(
+                    self.model.encode_first_stage(cond_fn.target.to(device))
+                )
+                pred = pred_x0
+            elif cond_fn.space == "rgb":
+                # We need to backward gradient to x0 in latent space, so it's required
+                # to trace the computation graph while decoding the latent.
+                with torch.enable_grad():
+                    pred_x0.requires_grad_(True)
+                    target = cond_fn.target.to(device)
+                    pred = self.model.decode_first_stage_with_grad(pred_x0)
+            else:
+                raise NotImplementedError(cond_fn.space)
+            delta_pred = cond_fn(target, pred, t_now)
+            
+            # ----------------- apply classifier guidance ----------------- #
+            if delta_pred is not None:
+                if cond_fn.space == "rgb":
+                    # compute gradient for pred_x0
+                    pred.backward(delta_pred)
+                    delta_pred_x0 = pred_x0.grad
+                    # update prex_x0
+                    pred_x0 += delta_pred_x0
+                    # our classifier guidance is equivalent to multiply delta_pred_x0
+                    # by a constant and then add it to model_mean, We set the constant
+                    # to 0.5
+                    model_mean += 0.5 * delta_pred_x0
+                    pred_x0.grad.zero_()
+                else:
+                    delta_pred_x0 = delta_pred
+                    pred_x0 += delta_pred_x0
+                    model_mean += 0.5 * delta_pred_x0
+            else:
+                # means stop guidance
+                break
+        
+        return model_mean.detach().clone(), pred_x0.detach().clone()
+    
+    @torch.no_grad()
+    def p_sample(
+        self,
+        x: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        t: torch.Tensor,
+        index: torch.Tensor,
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]],
+        cond_fn: Optional[Guidance]
+    ) -> torch.Tensor:
+        # variance of posterior distribution q(x_{t-1}|x_t, x_0)
+        model_variance = {
+            "fixed_large": np.append(self.posterior_variance[1], self.betas[1:]),
+            "fixed_small": self.posterior_variance
+        }[self.var_type]
+        model_variance = _extract_into_tensor(model_variance, index, x.shape)
+        
+        # mean of posterior distribution q(x_{t-1}|x_t, x_0)
+        if cond_fn is not None:
+            # apply classifier guidance
+            model_mean, pred_x0 = self.apply_cond_fn(
+                x, cond, t, index, cond_fn,
+                cfg_scale, uncond
+            )
+        else:
+            e_t = self.predict_noise(
+                x, t, cond, cfg_scale, uncond
+            )
+            pred_x0 = self._predict_xstart_from_eps(x_t=x, t=index, eps=e_t)
+            model_mean, _, _ = self.q_posterior_mean_variance(
+                x_start=pred_x0, x_t=x, t=index
+            )
+        
+        # sample x_t from q(x_{t-1}|x_t, x_0)
+        noise = torch.randn_like(x)
+        nonzero_mask = (
+            (index != 0).float().view(-1, *([1] * (len(x.shape) - 1)))
+        )
+        x_prev = model_mean + nonzero_mask * torch.sqrt(model_variance) * noise
+        return x_prev
+    
+    @torch.no_grad()
+    def p_sample_x0(
+        self,
+        x: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        t: torch.Tensor,
+        index: torch.Tensor,
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]],
+        cond_fn: Optional[Guidance]
+    ) -> torch.Tensor:
+        # variance of posterior distribution q(x_{t-1}|x_t, x_0)
+        model_variance = {
+            "fixed_large": np.append(self.posterior_variance[1], self.betas[1:]),
+            "fixed_small": self.posterior_variance
+        }[self.var_type]
+        model_variance = _extract_into_tensor(model_variance, index, x.shape)
+        
+        # mean of posterior distribution q(x_{t-1}|x_t, x_0)
+        if cond_fn is not None:
+            # apply classifier guidance
+            model_mean, pred_x0 = self.apply_cond_fn(
+                x, cond, t, index, cond_fn,
+                cfg_scale, uncond
+            )
+        else:
+            e_t = self.predict_noise(
+                x, t, cond, cfg_scale, uncond
+            )
+            pred_x0 = self._predict_xstart_from_eps(x_t=x, t=index, eps=e_t)
+            model_mean, _, _ = self.q_posterior_mean_variance(
+                x_start=pred_x0, x_t=x, t=index
+            )
+        
+        # sample x_t from q(x_{t-1}|x_t, x_0)
+        noise = torch.randn_like(x)
+        nonzero_mask = (
+            (index != 0).float().view(-1, *([1] * (len(x.shape) - 1)))
+        )
+        x_prev = model_mean + nonzero_mask * torch.sqrt(model_variance) * noise
+        return x_prev, pred_x0
+        
+    @torch.no_grad()
+    def p_sample_tao(
+        self,
+        x: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        t: torch.Tensor,
+        index: torch.Tensor,
+        t_max: float,
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]],
+        cond_fn: Optional[Guidance]
+    ) -> torch.Tensor:
+        
+        if cond_fn is not None:
+            # apply classifier guidance
+            model_mean, pred_x0 = self.apply_cond_fn(
+                x, cond, t, index, cond_fn,
+                cfg_scale, uncond
+            )
+        else:
+            e_t = self.predict_noise(
+                x, t, cond, cfg_scale, uncond
+            )
+            pred_x0 = self._predict_xstart_from_eps(x_t=x, t=index, eps=e_t)
+        
+        # sample x_t from q(x_{t-1}|x_t, x_0)
+        noise = torch.randn_like(x)
+        tao_index = torch.tensor(torch.round(index * t_max),dtype=torch.int64)
+        x_prev = self.q_sample(pred_x0, tao_index)
+       
+        return x_prev
+    
+
+    @torch.no_grad()
+    def sample_with_mixdiff_ccsr(
+        self,
+        tile_size: int,
+        tile_stride: int,
+        steps: int,
+        t_max: float,
+        t_min: float,
+        shape: Tuple[int],
+        cond_img: torch.Tensor,
+        positive_prompt: str,
+        negative_prompt: str,
+        x_T: Optional[torch.Tensor]=None,
+        cfg_scale: float=1.,
+        cond_fn: Optional[Guidance]=None,
+        color_fix_type: str="none"
+    ) -> torch.Tensor:
+        def _sliding_windows(h: int, w: int, tile_size: int, tile_stride: int) -> Tuple[int, int, int, int]:
+            hi_list = list(range(0, h - tile_size + 1, tile_stride))
+            if (h - tile_size) % tile_stride != 0:
+                hi_list.append(h - tile_size)
+            
+            wi_list = list(range(0, w - tile_size + 1, tile_stride))
+            if (w - tile_size) % tile_stride != 0:
+                wi_list.append(w - tile_size)
+            
+            coords = []
+            for hi in hi_list:
+                for wi in wi_list:
+                    coords.append((hi, hi + tile_size, wi, wi + tile_size))
+            return coords
+        
+        # make sampling parameters (e.g. sigmas)
+        self.make_schedule(num_steps=steps)
+        
+        device = next(self.model.parameters()).device
+        b, _, h, w = shape
+        if x_T is None:
+            img = torch.randn(shape, dtype=torch.float32, device=device)
+        else:
+            img = x_T
+        # create buffers for accumulating predicted noise of different diffusion process
+        noise_buffer = torch.zeros_like(img)
+        count = torch.zeros(shape, dtype=torch.long, device=device)
+        # timesteps iterator
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(self.timesteps)
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+        
+        # q_sample for the start
+        ts = torch.full((b,), time_range[0], device=device, dtype=torch.long)
+        index = torch.full_like(ts, fill_value=total_steps - 1)
+
+        # predict noise for each tile
+        tiles_iterator = tqdm(_sliding_windows(h, w, tile_size // 8, tile_stride // 8))
+        for hi, hi_end, wi, wi_end in tiles_iterator:
+            tiles_iterator.set_description(f"Process tile with location ({hi} {hi_end}) ({wi} {wi_end})")
+            # noisy latent of this diffusion process (tile) at this step
+            tile_img = img[:, :, hi:hi_end, wi:wi_end]
+            # prepare condition for this tile
+            tile_cond_img = cond_img[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8]
+            tile_cond = {
+                "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+            }
+            tile_uncond = {
+                "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+            }
+            # TODO: tile_cond_fn
+            
+            # predict noise for this tile
+            tile_noise = self.predict_noise(tile_img, ts, tile_cond, cfg_scale, tile_uncond)
+            
+            # accumulate noise
+            noise_buffer[:, :, hi:hi_end, wi:wi_end] += tile_noise
+            count[:, :, hi:hi_end, wi:wi_end] += 1
+        
+        # average on noise (score)
+        noise_buffer.div_(count)
+        pred_x0 = self._predict_xstart_from_eps(x_t=img, t=index, eps=noise_buffer)
+        tao_index = torch.tensor(torch.round(index * t_max),dtype=torch.int64)
+        img = self.q_sample(pred_x0, tao_index)
+        
+        noise_buffer.zero_()
+        count.zero_()
+        
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(time_range)
+        time_range = time_range[-int(round(total_steps*t_max)):]
+        total_steps_use = len(time_range)
+        time_range = time_range[:-int(round(total_steps*t_min))]
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+
+        # sampling loop
+        for i, step in enumerate(iterator):
+
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+            index = torch.full_like(ts, fill_value=total_steps_use - i - 1)
+            
+            # predict noise for each tile
+            tiles_iterator = tqdm(_sliding_windows(h, w, tile_size // 8, tile_stride // 8))
+            for hi, hi_end, wi, wi_end in tiles_iterator:
+                tiles_iterator.set_description(f"Process tile with location ({hi} {hi_end}) ({wi} {wi_end})")
+                # noisy latent of this diffusion process (tile) at this step
+                tile_img = img[:, :, hi:hi_end, wi:wi_end]
+                # prepare condition for this tile
+                tile_cond_img = cond_img[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8]
+                tile_cond = {
+                    "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                    "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+                }
+                tile_uncond = {
+                    "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                    "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+                }
+                # TODO: tile_cond_fn
+                
+                # predict noise for this tile
+                tile_noise = self.predict_noise(tile_img, ts, tile_cond, cfg_scale, tile_uncond)
+                
+                # accumulate noise
+                noise_buffer[:, :, hi:hi_end, wi:wi_end] += tile_noise
+                count[:, :, hi:hi_end, wi:wi_end] += 1
+            
+            # average on noise (score)
+            noise_buffer.div_(count)
+            # sample previous latent
+            pred_x0 = self._predict_xstart_from_eps(x_t=img, t=index, eps=noise_buffer)
+            mean, _, _ = self.q_posterior_mean_variance(
+                x_start=pred_x0, x_t=img, t=index
+            )
+            variance = {
+                "fixed_large": np.append(self.posterior_variance[1], self.betas[1:]),
+                "fixed_small": self.posterior_variance
+            }[self.var_type]
+            variance = _extract_into_tensor(variance, index, noise_buffer.shape)
+            
+            nonzero_mask = (
+                (index != 0).float().view(-1, *([1] * (len(noise_buffer.shape) - 1)))
+            )
+            img = mean + nonzero_mask * torch.sqrt(variance) * torch.randn_like(mean)
+            
+            noise_buffer.zero_()
+            count.zero_()
+        
+        img = pred_x0
+        # decode samples of each diffusion process
+        img_buffer = torch.zeros_like(cond_img)
+        count = torch.zeros_like(cond_img, dtype=torch.long)
+        for hi, hi_end, wi, wi_end in _sliding_windows(h, w, tile_size // 8, tile_stride // 8):
+            tile_img = img[:, :, hi:hi_end, wi:wi_end]
+            tile_img_pixel = (self.model.decode_first_stage(tile_img) + 1) / 2
+            tile_cond_img = cond_img[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8]
+            # apply color correction (borrowed from StableSR)
+            if color_fix_type == "adain":
+                tile_img_pixel = adaptive_instance_normalization(tile_img_pixel, tile_cond_img)
+            elif color_fix_type == "wavelet":
+                tile_img_pixel = wavelet_reconstruction(tile_img_pixel, tile_cond_img)
+            else:
+                assert color_fix_type == "none", f"unexpected color fix type: {color_fix_type}"
+            img_buffer[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8] += tile_img_pixel
+            count[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8] += 1
+        img_buffer.div_(count)
+        
+        return img_buffer
+    
+    @torch.no_grad()
+    def sample_with_mixdiff_control(
+        self,
+        control_imgs: torch.Tensor,
+        tile_size: int,
+        tile_stride: int,
+        steps: int,
+        tao_steps: int,
+        shape: Tuple[int],
+        cond_img: torch.Tensor,
+        positive_prompt: str,
+        negative_prompt: str,
+        x_T: Optional[torch.Tensor]=None,
+        cfg_scale: float=1.,
+        cond_fn: Optional[Guidance]=None,
+        color_fix_type: str="none"
+    ) -> torch.Tensor:
+        def _sliding_windows(h: int, w: int, tile_size: int, tile_stride: int) -> Tuple[int, int, int, int]:
+            hi_list = list(range(0, h - tile_size + 1, tile_stride))
+            if (h - tile_size) % tile_stride != 0:
+                hi_list.append(h - tile_size)
+            
+            wi_list = list(range(0, w - tile_size + 1, tile_stride))
+            if (w - tile_size) % tile_stride != 0:
+                wi_list.append(w - tile_size)
+            
+            coords = []
+            for hi in hi_list:
+                for wi in wi_list:
+                    coords.append((hi, hi + tile_size, wi, wi + tile_size))
+            return coords
+        
+        # make sampling parameters (e.g. sigmas)
+        self.make_schedule(num_steps=steps)
+        
+        device = next(self.model.parameters()).device
+        b, _, h, w = shape
+        if x_T is None:
+            img = torch.randn(shape, dtype=torch.float32, device=device)
+        else:
+            img = x_T
+        # create buffers for accumulating predicted noise of different diffusion process
+        noise_buffer = torch.zeros_like(img)
+        count = torch.zeros(shape, dtype=torch.long, device=device)
+        # timesteps iterator
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(self.timesteps)
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+        
+        # q_sample for the start
+        ts = torch.full((b,), time_range[0], device=device, dtype=torch.long)
+        index = torch.full_like(ts, fill_value=total_steps - 1)
+
+        # start point: LR        
+        img = self.q_sample(control_imgs, index)
+
+        # predict noise for each tile
+        tiles_iterator = tqdm(_sliding_windows(h, w, tile_size // 8, tile_stride // 8))
+        for hi, hi_end, wi, wi_end in tiles_iterator:
+            tiles_iterator.set_description(f"Process tile with location ({hi} {hi_end}) ({wi} {wi_end})")
+            # noisy latent of this diffusion process (tile) at this step
+            tile_img = img[:, :, hi:hi_end, wi:wi_end]
+            # prepare condition for this tile
+            tile_cond_img = cond_img[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8]
+            tile_cond = {
+                "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+            }
+            tile_uncond = {
+                "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+            }
+            # TODO: tile_cond_fn
+            
+            # predict noise for this tile
+            tile_noise = self.predict_noise(tile_img, ts, tile_cond, cfg_scale, tile_uncond)
+            
+            # accumulate noise
+            noise_buffer[:, :, hi:hi_end, wi:wi_end] += tile_noise
+            count[:, :, hi:hi_end, wi:wi_end] += 1
+        
+        # average on noise (score)
+        noise_buffer.div_(count)
+        # sample previous latent
+        pred_x0 = self._predict_xstart_from_eps(x_t=img, t=index, eps=noise_buffer)
+        tao_index = index - index//(tao_steps-1)
+        img = self.q_sample(pred_x0, tao_index)
+        
+        noise_buffer.zero_()
+        count.zero_()
+        
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(time_range)
+        time_range = time_range[total_steps//(tao_steps-1):]
+        total_steps_use = len(time_range)
+        # time_range = time_range[:-total_steps//(tao_steps-1)]
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+
+        # sampling loop
+        for i, step in enumerate(iterator):
+
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+            index = torch.full_like(ts, fill_value=total_steps_use - i - 1)
+            
+            # predict noise for each tile
+            tiles_iterator = tqdm(_sliding_windows(h, w, tile_size // 8, tile_stride // 8))
+            for hi, hi_end, wi, wi_end in tiles_iterator:
+                tiles_iterator.set_description(f"Process tile with location ({hi} {hi_end}) ({wi} {wi_end})")
+                # noisy latent of this diffusion process (tile) at this step
+                tile_img = img[:, :, hi:hi_end, wi:wi_end]
+                # prepare condition for this tile
+                tile_cond_img = cond_img[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8]
+                tile_cond = {
+                    "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                    "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+                }
+                tile_uncond = {
+                    "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                    "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+                }
+                # TODO: tile_cond_fn
+                
+                # predict noise for this tile
+                tile_noise = self.predict_noise(tile_img, ts, tile_cond, cfg_scale, tile_uncond)
+                
+                # accumulate noise
+                noise_buffer[:, :, hi:hi_end, wi:wi_end] += tile_noise
+                count[:, :, hi:hi_end, wi:wi_end] += 1
+            
+            # average on noise (score)
+            noise_buffer.div_(count)
+            # sample previous latent
+            pred_x0 = self._predict_xstart_from_eps(x_t=img, t=index, eps=noise_buffer)
+            mean, _, _ = self.q_posterior_mean_variance(
+                x_start=pred_x0, x_t=img, t=index
+            )
+            variance = {
+                "fixed_large": np.append(self.posterior_variance[1], self.betas[1:]),
+                "fixed_small": self.posterior_variance
+            }[self.var_type]
+            variance = _extract_into_tensor(variance, index, noise_buffer.shape)
+            
+            nonzero_mask = (
+                (index != 0).float().view(-1, *([1] * (len(noise_buffer.shape) - 1)))
+            )
+            img = mean + nonzero_mask * torch.sqrt(variance) * torch.randn_like(mean)
+            
+            noise_buffer.zero_()
+            count.zero_()
+        
+        img = pred_x0
+        # decode samples of each diffusion process
+        img_buffer = torch.zeros_like(cond_img)
+        count = torch.zeros_like(cond_img, dtype=torch.long)
+        for hi, hi_end, wi, wi_end in _sliding_windows(h, w, tile_size // 8, tile_stride // 8):
+            tile_img = img[:, :, hi:hi_end, wi:wi_end]
+            tile_img_pixel = (self.model.decode_first_stage(tile_img) + 1) / 2
+            tile_cond_img = cond_img[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8]
+            # apply color correction (borrowed from StableSR)
+            if color_fix_type == "adain":
+                tile_img_pixel = adaptive_instance_normalization(tile_img_pixel, tile_cond_img)
+            elif color_fix_type == "wavelet":
+                tile_img_pixel = wavelet_reconstruction(tile_img_pixel, tile_cond_img)
+            else:
+                assert color_fix_type == "none", f"unexpected color fix type: {color_fix_type}"
+            img_buffer[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8] += tile_img_pixel
+            count[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8] += 1
+        img_buffer.div_(count)
+        
+        return img_buffer
+    
+    @torch.no_grad()
+    def sample_ccsr(
+        self,
+        steps: int,
+        t_max: float,
+        t_min: float,
+        shape: Tuple[int],
+        cond_img: torch.Tensor,
+        positive_prompt: str,
+        negative_prompt: str,
+        x_T: Optional[torch.Tensor]=None,
+        cfg_scale: float=1.,
+        cond_fn: Optional[Guidance]=None,
+        color_fix_type: str="none"
+    ) -> torch.Tensor:
+        self.make_schedule(num_steps=steps)
+        # self.make_tao_schedule(num_steps=tao_steps)
+        
+        device = next(self.model.parameters()).device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+        
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(self.timesteps)
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+        
+        cond = {
+            "c_latent": [self.model.apply_condition_encoder(cond_img)],
+            "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+        }
+        uncond = {
+            "c_latent": [self.model.apply_condition_encoder(cond_img)],
+            "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+        }
+
+        # q_sample for the start
+        ts = torch.full((b,), time_range[0], device=device, dtype=torch.long)
+        index = torch.full_like(ts, fill_value=total_steps - 1)
+        img = self.p_sample_tao(
+                img, cond, ts, index=index,t_max=t_max,
+                cfg_scale=cfg_scale, uncond=uncond,
+                cond_fn=cond_fn
+            )
+        
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(time_range)
+        time_range = time_range[-int(round(total_steps*t_max)):]
+        total_steps_use = len(time_range)
+        time_range = time_range[:-int(round(total_steps*t_min))]
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+
+        for i, step in enumerate(iterator):
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+            index = torch.full_like(ts, fill_value=total_steps_use - i - 1)
+            img, x0 = self.p_sample_x0(
+                img, cond, ts, index=index,
+                cfg_scale=cfg_scale, uncond=uncond,
+                cond_fn=cond_fn
+            )
+        
+        img = x0
+        img_pixel = (self.model.decode_first_stage(img) + 1) / 2
+        # apply color correction (borrowed from StableSR)
+        if color_fix_type == "adain":
+            img_pixel = adaptive_instance_normalization(img_pixel, cond_img)
+        elif color_fix_type == "wavelet":
+            img_pixel = wavelet_reconstruction(img_pixel, cond_img)
+        else:
+            assert color_fix_type == "none", f"unexpected color fix type: {color_fix_type}"
+        return img_pixel
+    
+    @torch.no_grad()
+    def sample_ccsr_stage1(
+        self,
+        steps: int,
+        t_max: float,
+        shape: Tuple[int],
+        cond_img: torch.Tensor,
+        positive_prompt: str,
+        negative_prompt: str,
+        x_T: Optional[torch.Tensor]=None,
+        cfg_scale: float=1.,
+        cond_fn: Optional[Guidance]=None,
+        color_fix_type: str="none"
+    ) -> torch.Tensor:
+        self.make_schedule(num_steps=steps)
+        # self.make_tao_schedule(num_steps=tao_steps)
+        
+        device = next(self.model.parameters()).device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+        
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(self.timesteps)
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+        
+        cond = {
+            "c_latent": [self.model.apply_condition_encoder(cond_img)],
+            "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+        }
+        uncond = {
+            "c_latent": [self.model.apply_condition_encoder(cond_img)],
+            "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+        }
+
+        # q_sample for the start
+        ts = torch.full((b,), time_range[0], device=device, dtype=torch.long)
+        index = torch.full_like(ts, fill_value=total_steps - 1)
+        img = self.p_sample_tao(
+                img, cond, ts, index=index,t_max=t_max,
+                cfg_scale=cfg_scale, uncond=uncond,
+                cond_fn=cond_fn
+            )
+        
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(time_range)
+        time_range = time_range[-int(round(total_steps*t_max)):]
+        total_steps = len(time_range)
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+
+        for i, step in enumerate(iterator):
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+            index = torch.full_like(ts, fill_value=total_steps - i - 1)
+            img = self.p_sample(
+                img, cond, ts, index=index,
+                cfg_scale=cfg_scale, uncond=uncond,
+                cond_fn=cond_fn
+            )
+        
+        img_pixel = (self.model.decode_first_stage(img) + 1) / 2
+        # apply color correction (borrowed from StableSR)
+        if color_fix_type == "adain":
+            img_pixel = adaptive_instance_normalization(img_pixel, cond_img)
+        elif color_fix_type == "wavelet":
+            img_pixel = wavelet_reconstruction(img_pixel, cond_img)
+        else:
+            assert color_fix_type == "none", f"unexpected color fix type: {color_fix_type}"
+        return img_pixel
\ No newline at end of file
diff --git a/model/spaced_sampler.py b/model/spaced_sampler.py
new file mode 100644
index 0000000..fca59b9
--- /dev/null
+++ b/model/spaced_sampler.py
@@ -0,0 +1,670 @@
+from typing import Optional, Tuple, Dict, List, Callable
+
+import torch
+import numpy as np
+from tqdm import tqdm
+import einops
+import os
+from PIL import Image
+
+from ldm.modules.diffusionmodules.util import make_beta_schedule
+from model.cond_fn import Guidance
+from utils.image import (
+    wavelet_reconstruction, adaptive_instance_normalization
+)
+
+# https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/respace.py
+def space_timesteps(num_timesteps, section_counts):
+    """
+    Create a list of timesteps to use from an original diffusion process,
+    given the number of timesteps we want to take from equally-sized portions
+    of the original process.
+    For example, if there's 300 timesteps and the section counts are [10,15,20]
+    then the first 100 timesteps are strided to be 10 timesteps, the second 100
+    are strided to be 15 timesteps, and the final 100 are strided to be 20.
+    If the stride is a string starting with "ddim", then the fixed striding
+    from the DDIM paper is used, and only one section is allowed.
+    :param num_timesteps: the number of diffusion steps in the original
+                          process to divide up.
+    :param section_counts: either a list of numbers, or a string containing
+                           comma-separated numbers, indicating the step count
+                           per section. As a special case, use "ddimN" where N
+                           is a number of steps to use the striding from the
+                           DDIM paper.
+    :return: a set of diffusion steps from the original process to use.
+    """
+    if isinstance(section_counts, str):
+        if section_counts.startswith("ddim"):
+            desired_count = int(section_counts[len("ddim") :])
+            for i in range(1, num_timesteps):
+                if len(range(0, num_timesteps, i)) == desired_count:
+                    return set(range(0, num_timesteps, i))
+            raise ValueError(
+                f"cannot create exactly {num_timesteps} steps with an integer stride"
+            )
+        section_counts = [int(x) for x in section_counts.split(",")]
+    size_per = num_timesteps // len(section_counts)
+    extra = num_timesteps % len(section_counts)
+    start_idx = 0
+    all_steps = []
+    for i, section_count in enumerate(section_counts):
+        size = size_per + (1 if i < extra else 0)
+        if size < section_count:
+            raise ValueError(
+                f"cannot divide section of {size} steps into {section_count}"
+            )
+        if section_count <= 1:
+            frac_stride = 1
+        else:
+            frac_stride = (size - 1) / (section_count - 1)
+        cur_idx = 0.0
+        taken_steps = []
+        for _ in range(section_count):
+            taken_steps.append(start_idx + round(cur_idx))
+            cur_idx += frac_stride
+        all_steps += taken_steps
+        start_idx += size
+    return set(all_steps)
+
+
+# https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/gaussian_diffusion.py
+def _extract_into_tensor(arr, timesteps, broadcast_shape):
+    """
+    Extract values from a 1-D numpy array for a batch of indices.
+    :param arr: the 1-D numpy array.
+    :param timesteps: a tensor of indices into the array to extract.
+    :param broadcast_shape: a larger shape of K dimensions with the batch
+                            dimension equal to the length of timesteps.
+    :return: a tensor of shape [batch_size, 1, ...] where the shape has K dims.
+    """
+    try:
+        # float64 as default. float64 is not supported by mps device.
+        res = torch.from_numpy(arr).to(device=timesteps.device)[timesteps].float()
+    except:
+        # to be compatible with mps
+        res = torch.from_numpy(arr.astype(np.float32)).to(device=timesteps.device)[timesteps].float()
+        
+    while len(res.shape) < len(broadcast_shape):
+        res = res[..., None]
+    return res.expand(broadcast_shape)
+
+
+class SpacedSampler:
+    """
+    Implementation for spaced sampling schedule proposed in IDDPM. This class is designed
+    for sampling ControlLDM.
+    
+    https://arxiv.org/pdf/2102.09672.pdf
+    """
+    
+    def __init__(
+        self,
+        model: "ControlLDM",
+        schedule: str="linear",
+        var_type: str="fixed_small"
+    ) -> "SpacedSampler":
+        self.model = model
+        self.original_num_steps = model.num_timesteps
+        self.schedule = schedule
+        self.var_type = var_type
+
+    def make_schedule(self, num_steps: int) -> None:
+        """
+        Initialize sampling parameters according to `num_steps`.
+        
+        Args:
+            num_steps (int): Sampling steps.
+
+        Returns:
+            None
+        """
+        # NOTE: this schedule, which generates betas linearly in log space, is a little different
+        # from guided diffusion.
+        original_betas = make_beta_schedule(
+            self.schedule, self.original_num_steps, linear_start=self.model.linear_start,
+            linear_end=self.model.linear_end
+        )
+        original_alphas = 1.0 - original_betas
+        original_alphas_cumprod = np.cumprod(original_alphas, axis=0)
+        
+        # calcualte betas for spaced sampling
+        # https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/respace.py
+        used_timesteps = space_timesteps(self.original_num_steps, str(num_steps))
+        print(f"timesteps used in spaced sampler: \n\t{sorted(list(used_timesteps))}")
+        
+        betas = []
+        last_alpha_cumprod = 1.0
+        for i, alpha_cumprod in enumerate(original_alphas_cumprod):
+            if i in used_timesteps:
+                # marginal distribution is the same as q(x_{S_t}|x_0)
+                betas.append(1 - alpha_cumprod / last_alpha_cumprod)
+                last_alpha_cumprod = alpha_cumprod
+        assert len(betas) == num_steps
+        betas = np.array(betas, dtype=np.float64)
+        self.betas = betas
+        
+        self.timesteps = np.array(sorted(list(used_timesteps)), dtype=np.int32) # e.g. [0, 10, 20, ...]
+        alphas = 1.0 - betas
+        self.alphas_cumprod = np.cumprod(alphas, axis=0)
+        self.alphas_cumprod_prev = np.append(1.0, self.alphas_cumprod[:-1])
+        self.alphas_cumprod_next = np.append(self.alphas_cumprod[1:], 0.0)
+        assert self.alphas_cumprod_prev.shape == (num_steps, )
+        
+        # calculations for diffusion q(x_t | x_{t-1}) and others
+        self.sqrt_alphas_cumprod = np.sqrt(self.alphas_cumprod)
+        self.sqrt_one_minus_alphas_cumprod = np.sqrt(1.0 - self.alphas_cumprod)
+        self.log_one_minus_alphas_cumprod = np.log(1.0 - self.alphas_cumprod)
+        self.sqrt_recip_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod)
+        self.sqrt_recipm1_alphas_cumprod = np.sqrt(1.0 / self.alphas_cumprod - 1)
+
+        # calculations for posterior q(x_{t-1} | x_t, x_0)
+        self.posterior_variance = (
+            betas * (1.0 - self.alphas_cumprod_prev) / (1.0 - self.alphas_cumprod)
+        )
+        # log calculation clipped because the posterior variance is 0 at the
+        # beginning of the diffusion chain.
+        self.posterior_log_variance_clipped = np.log(
+            np.append(self.posterior_variance[1], self.posterior_variance[1:])
+        )
+        self.posterior_mean_coef1 = (
+            betas * np.sqrt(self.alphas_cumprod_prev) / (1.0 - self.alphas_cumprod)
+        )
+        self.posterior_mean_coef2 = (
+            (1.0 - self.alphas_cumprod_prev)
+            * np.sqrt(alphas)
+            / (1.0 - self.alphas_cumprod)
+        )
+
+    def q_sample(
+        self,
+        x_start: torch.Tensor,
+        t: torch.Tensor,
+        noise: Optional[torch.Tensor]=None
+    ) -> torch.Tensor:
+        """
+        Implement the marginal distribution q(x_t|x_0).
+
+        Args:
+            x_start (torch.Tensor): Images (NCHW) sampled from data distribution.
+            t (torch.Tensor): Timestep (N) for diffusion process. `t` serves as an index
+                to get parameters for each timestep.
+            noise (torch.Tensor, optional): Specify the noise (NCHW) added to `x_start`.
+
+        Returns:
+            x_t (torch.Tensor): The noisy images.
+        """
+        if noise is None:
+            noise = torch.randn_like(x_start)
+        assert noise.shape == x_start.shape
+        return (
+            _extract_into_tensor(self.sqrt_alphas_cumprod, t, x_start.shape) * x_start
+            + _extract_into_tensor(self.sqrt_one_minus_alphas_cumprod, t, x_start.shape)
+            * noise
+        )
+
+    def q_posterior_mean_variance(
+        self,
+        x_start: torch.Tensor,
+        x_t: torch.Tensor,
+        t: torch.Tensor
+    ) -> Tuple[torch.Tensor]:
+        """
+        Implement the posterior distribution q(x_{t-1}|x_t, x_0).
+        
+        Args:
+            x_start (torch.Tensor): The predicted images (NCHW) in timestep `t`.
+            x_t (torch.Tensor): The sampled intermediate variables (NCHW) of timestep `t`.
+            t (torch.Tensor): Timestep (N) of `x_t`. `t` serves as an index to get 
+                parameters for each timestep.
+        
+        Returns:
+            posterior_mean (torch.Tensor): Mean of the posterior distribution.
+            posterior_variance (torch.Tensor): Variance of the posterior distribution.
+            posterior_log_variance_clipped (torch.Tensor): Log variance of the posterior distribution.
+        """
+        assert x_start.shape == x_t.shape
+        posterior_mean = (
+            _extract_into_tensor(self.posterior_mean_coef1, t, x_t.shape) * x_start
+            + _extract_into_tensor(self.posterior_mean_coef2, t, x_t.shape) * x_t
+        )
+        posterior_variance = _extract_into_tensor(self.posterior_variance, t, x_t.shape)
+        posterior_log_variance_clipped = _extract_into_tensor(
+            self.posterior_log_variance_clipped, t, x_t.shape
+        )
+        assert (
+            posterior_mean.shape[0]
+            == posterior_variance.shape[0]
+            == posterior_log_variance_clipped.shape[0]
+            == x_start.shape[0]
+        )
+        return posterior_mean, posterior_variance, posterior_log_variance_clipped
+
+    def _predict_xstart_from_eps(
+        self,
+        x_t: torch.Tensor,
+        t: torch.Tensor,
+        eps: torch.Tensor
+    ) -> torch.Tensor:
+        assert x_t.shape == eps.shape
+        return (
+            _extract_into_tensor(self.sqrt_recip_alphas_cumprod, t, x_t.shape) * x_t
+            - _extract_into_tensor(self.sqrt_recipm1_alphas_cumprod, t, x_t.shape) * eps
+        )
+    
+    def predict_noise(
+        self,
+        x: torch.Tensor,
+        t: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]]
+    ) -> torch.Tensor:
+        if uncond is None or cfg_scale == 1.:
+            model_output = self.model.apply_model(x, t, cond)
+        else:
+            # apply classifier-free guidance
+            model_cond = self.model.apply_model(x, t, cond)
+            model_uncond = self.model.apply_model(x, t, uncond)
+            model_output = model_uncond + cfg_scale * (model_cond - model_uncond)
+        
+        if self.model.parameterization == "v":
+            e_t = self.model.predict_eps_from_z_and_v(x, t, model_output)
+        else:
+            e_t = model_output
+
+        return e_t
+    
+    def apply_cond_fn(
+        self,
+        x: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        t: torch.Tensor,
+        index: torch.Tensor,
+        cond_fn: Guidance,
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]]
+    ) -> torch.Tensor:
+        device = x.device
+        t_now = int(t[0].item()) + 1
+        # ----------------- predict noise and x0 ----------------- #
+        e_t = self.predict_noise(
+            x, t, cond, cfg_scale, uncond
+        )
+        pred_x0: torch.Tensor = self._predict_xstart_from_eps(x_t=x, t=index, eps=e_t)
+        model_mean, _, _ = self.q_posterior_mean_variance(
+            x_start=pred_x0, x_t=x, t=index
+        )
+        
+        # apply classifier guidance for multiple times
+        for _ in range(cond_fn.repeat):
+            # ----------------- compute gradient for x0 in latent space ----------------- #
+            target, pred = None, None
+            if cond_fn.space == "latent":
+                target = self.model.get_first_stage_encoding(
+                    self.model.encode_first_stage(cond_fn.target.to(device))
+                )
+                pred = pred_x0
+            elif cond_fn.space == "rgb":
+                # We need to backward gradient to x0 in latent space, so it's required
+                # to trace the computation graph while decoding the latent.
+                with torch.enable_grad():
+                    pred_x0.requires_grad_(True)
+                    target = cond_fn.target.to(device)
+                    pred = self.model.decode_first_stage_with_grad(pred_x0)
+            else:
+                raise NotImplementedError(cond_fn.space)
+            delta_pred = cond_fn(target, pred, t_now)
+            
+            # ----------------- apply classifier guidance ----------------- #
+            if delta_pred is not None:
+                if cond_fn.space == "rgb":
+                    # compute gradient for pred_x0
+                    pred.backward(delta_pred)
+                    delta_pred_x0 = pred_x0.grad
+                    # update prex_x0
+                    pred_x0 += delta_pred_x0
+                    # our classifier guidance is equivalent to multiply delta_pred_x0
+                    # by a constant and then add it to model_mean, We set the constant
+                    # to 0.5
+                    model_mean += 0.5 * delta_pred_x0
+                    pred_x0.grad.zero_()
+                else:
+                    delta_pred_x0 = delta_pred
+                    pred_x0 += delta_pred_x0
+                    model_mean += 0.5 * delta_pred_x0
+            else:
+                # means stop guidance
+                break
+        
+        return model_mean.detach().clone(), pred_x0.detach().clone()
+    
+    @torch.no_grad()
+    def p_sample(
+        self,
+        x: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        t: torch.Tensor,
+        index: torch.Tensor,
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]],
+        cond_fn: Optional[Guidance]
+    ) -> torch.Tensor:
+        # variance of posterior distribution q(x_{t-1}|x_t, x_0)
+        model_variance = {
+            "fixed_large": np.append(self.posterior_variance[1], self.betas[1:]),
+            "fixed_small": self.posterior_variance
+        }[self.var_type]
+        model_variance = _extract_into_tensor(model_variance, index, x.shape)
+        
+        # mean of posterior distribution q(x_{t-1}|x_t, x_0)
+        if cond_fn is not None:
+            # apply classifier guidance
+            model_mean, pred_x0 = self.apply_cond_fn(
+                x, cond, t, index, cond_fn,
+                cfg_scale, uncond
+            )
+        else:
+            e_t = self.predict_noise(
+                x, t, cond, cfg_scale, uncond
+            )
+            pred_x0 = self._predict_xstart_from_eps(x_t=x, t=index, eps=e_t)
+            model_mean, _, _ = self.q_posterior_mean_variance(
+                x_start=pred_x0, x_t=x, t=index
+            )
+        
+        # sample x_t from q(x_{t-1}|x_t, x_0)
+        noise = torch.randn_like(x)
+        nonzero_mask = (
+            (index != 0).float().view(-1, *([1] * (len(x.shape) - 1)))
+        )
+        x_prev = model_mean + nonzero_mask * torch.sqrt(model_variance) * noise
+        return x_prev
+    
+    @torch.no_grad()
+    def p_sample_x0(
+        self,
+        x: torch.Tensor,
+        cond: Dict[str, torch.Tensor],
+        t: torch.Tensor,
+        index: torch.Tensor,
+        cfg_scale: float,
+        uncond: Optional[Dict[str, torch.Tensor]],
+        cond_fn: Optional[Guidance]
+    ) -> torch.Tensor:
+        # variance of posterior distribution q(x_{t-1}|x_t, x_0)
+        model_variance = {
+            "fixed_large": np.append(self.posterior_variance[1], self.betas[1:]),
+            "fixed_small": self.posterior_variance
+        }[self.var_type]
+        model_variance = _extract_into_tensor(model_variance, index, x.shape)
+        
+        # mean of posterior distribution q(x_{t-1}|x_t, x_0)
+        if cond_fn is not None:
+            # apply classifier guidance
+            model_mean, pred_x0 = self.apply_cond_fn(
+                x, cond, t, index, cond_fn,
+                cfg_scale, uncond
+            )
+        else:
+            e_t = self.predict_noise(
+                x, t, cond, cfg_scale, uncond
+            )
+            pred_x0 = self._predict_xstart_from_eps(x_t=x, t=index, eps=e_t)
+            model_mean, _, _ = self.q_posterior_mean_variance(
+                x_start=pred_x0, x_t=x, t=index
+            )
+        
+        # sample x_t from q(x_{t-1}|x_t, x_0)
+        noise = torch.randn_like(x)
+        nonzero_mask = (
+            (index != 0).float().view(-1, *([1] * (len(x.shape) - 1)))
+        )
+        x_prev = model_mean + nonzero_mask * torch.sqrt(model_variance) * noise
+        return x_prev, pred_x0
+    
+    
+    @torch.no_grad()
+    def sample_with_mixdiff(
+        self,
+        tile_size: int,
+        tile_stride: int,
+        steps: int,
+        shape: Tuple[int],
+        cond_img: torch.Tensor,
+        positive_prompt: str,
+        negative_prompt: str,
+        x_T: Optional[torch.Tensor]=None,
+        cfg_scale: float=1.,
+        cond_fn: Optional[Guidance]=None,
+        color_fix_type: str="none"
+    ) -> torch.Tensor:
+        def _sliding_windows(h: int, w: int, tile_size: int, tile_stride: int) -> Tuple[int, int, int, int]:
+            hi_list = list(range(0, h - tile_size + 1, tile_stride))
+            if (h - tile_size) % tile_stride != 0:
+                hi_list.append(h - tile_size)
+            
+            wi_list = list(range(0, w - tile_size + 1, tile_stride))
+            if (w - tile_size) % tile_stride != 0:
+                wi_list.append(w - tile_size)
+            
+            coords = []
+            for hi in hi_list:
+                for wi in wi_list:
+                    coords.append((hi, hi + tile_size, wi, wi + tile_size))
+            return coords
+        
+        # make sampling parameters (e.g. sigmas)
+        self.make_schedule(num_steps=steps)
+        
+        device = next(self.model.parameters()).device
+        b, _, h, w = shape
+        if x_T is None:
+            img = torch.randn(shape, dtype=torch.float32, device=device)
+        else:
+            img = x_T
+        # create buffers for accumulating predicted noise of different diffusion process
+        noise_buffer = torch.zeros_like(img)
+        count = torch.zeros(shape, dtype=torch.long, device=device)
+        # timesteps iterator
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(self.timesteps)
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+        
+        # sampling loop
+        for i, step in enumerate(iterator):
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+            index = torch.full_like(ts, fill_value=total_steps - i - 1)
+            
+            # predict noise for each tile
+            tiles_iterator = tqdm(_sliding_windows(h, w, tile_size // 8, tile_stride // 8))
+            for hi, hi_end, wi, wi_end in tiles_iterator:
+                tiles_iterator.set_description(f"Process tile with location ({hi} {hi_end}) ({wi} {wi_end})")
+                # noisy latent of this diffusion process (tile) at this step
+                tile_img = img[:, :, hi:hi_end, wi:wi_end]
+                # prepare condition for this tile
+                tile_cond_img = cond_img[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8]
+                tile_cond = {
+                    "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                    "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+                }
+                tile_uncond = {
+                    "c_latent": [self.model.apply_condition_encoder(tile_cond_img)],
+                    "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+                }
+                # TODO: tile_cond_fn
+                
+                # predict noise for this tile
+                tile_noise = self.predict_noise(tile_img, ts, tile_cond, cfg_scale, tile_uncond)
+                
+                # accumulate noise
+                noise_buffer[:, :, hi:hi_end, wi:wi_end] += tile_noise
+                count[:, :, hi:hi_end, wi:wi_end] += 1
+            
+            # average on noise (score)
+            noise_buffer.div_(count)
+            # sample previous latent
+            pred_x0 = self._predict_xstart_from_eps(x_t=img, t=index, eps=noise_buffer)
+            mean, _, _ = self.q_posterior_mean_variance(
+                x_start=pred_x0, x_t=img, t=index
+            )
+            variance = {
+                "fixed_large": np.append(self.posterior_variance[1], self.betas[1:]),
+                "fixed_small": self.posterior_variance
+            }[self.var_type]
+            variance = _extract_into_tensor(variance, index, noise_buffer.shape)
+            
+            nonzero_mask = (
+                (index != 0).float().view(-1, *([1] * (len(noise_buffer.shape) - 1)))
+            )
+            img = mean + nonzero_mask * torch.sqrt(variance) * torch.randn_like(mean)
+            
+            noise_buffer.zero_()
+            count.zero_()
+        
+        # decode samples of each diffusion process
+        img_buffer = torch.zeros_like(cond_img)
+        count = torch.zeros_like(cond_img, dtype=torch.long)
+        for hi, hi_end, wi, wi_end in _sliding_windows(h, w, tile_size // 8, tile_stride // 8):
+            tile_img = img[:, :, hi:hi_end, wi:wi_end]
+            tile_img_pixel = (self.model.decode_first_stage(tile_img) + 1) / 2
+            tile_cond_img = cond_img[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8]
+            # apply color correction (borrowed from StableSR)
+            if color_fix_type == "adain":
+                tile_img_pixel = adaptive_instance_normalization(tile_img_pixel, tile_cond_img)
+            elif color_fix_type == "wavelet":
+                tile_img_pixel = wavelet_reconstruction(tile_img_pixel, tile_cond_img)
+            else:
+                assert color_fix_type == "none", f"unexpected color fix type: {color_fix_type}"
+            img_buffer[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8] += tile_img_pixel
+            count[:, :, hi * 8:hi_end * 8, wi * 8: wi_end * 8] += 1
+        img_buffer.div_(count)
+        
+        return img_buffer
+    
+    @torch.no_grad()
+    def sample_savex0(
+        self,
+        steps: int,
+        shape: Tuple[int],
+        cond_img: torch.Tensor,
+        positive_prompt: str,
+        negative_prompt: str,
+        x_T: Optional[torch.Tensor]=None,
+        cfg_scale: float=1.,
+        cond_fn: Optional[Guidance]=None,
+        color_fix_type: str="none"
+    ) -> torch.Tensor:
+        self.make_schedule(num_steps=steps)
+        
+        device = next(self.model.parameters()).device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+        
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(self.timesteps)
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+        
+        cond = {
+            "c_latent": [self.model.apply_condition_encoder(cond_img)],
+            "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+        }
+        uncond = {
+            "c_latent": [self.model.apply_condition_encoder(cond_img)],
+            "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+        }
+
+        x0s = []
+        for i, step in enumerate(iterator):
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+            index = torch.full_like(ts, fill_value=total_steps - i - 1)
+            img, x0 = self.p_sample_x0(
+                img, cond, ts, index=index,
+                cfg_scale=cfg_scale, uncond=uncond,
+                cond_fn=cond_fn
+            )
+            x0s.append(x0)
+        
+        img_pixel = (self.model.decode_first_stage(img) + 1) / 2
+        # apply color correction (borrowed from StableSR)
+        if color_fix_type == "adain":
+            img_pixel = adaptive_instance_normalization(img_pixel, cond_img)
+        elif color_fix_type == "wavelet":
+            img_pixel = wavelet_reconstruction(img_pixel, cond_img)
+        else:
+            assert color_fix_type == "none", f"unexpected color fix type: {color_fix_type}"
+        save_path_ori = '/home/notebook/data/group/SunLingchen/code/DiffBIR/fig4/realsr-out/seed_231_baseline'
+        for i in range(len(x0s)):
+            x0 = (self.model.decode_first_stage(x0s[i]) + 1) / 2
+            # apply color correction (borrowed from StableSR)
+            if color_fix_type == "adain":
+                x0 = adaptive_instance_normalization(x0, cond_img)
+            elif color_fix_type == "wavelet":
+                x0 = wavelet_reconstruction(x0, cond_img)
+            else:
+                assert color_fix_type == "none", f"unexpected color fix type: {color_fix_type}"
+            
+            x0 = x0.clamp(0, 1)
+            x0 = (einops.rearrange(x0, "b c h w -> b h w c") * 255).cpu().numpy().clip(0, 255).astype(np.uint8)
+            x0 = x0[0]
+            x0 = x0[:512, :512, :]
+            save_path = os.path.join(save_path_ori, f"{i}.png")
+            Image.fromarray(x0).save(save_path)
+            print(f"save to {save_path}")
+            
+        return img_pixel
+
+    @torch.no_grad()
+    def sample(
+        self,
+        steps: int,
+        shape: Tuple[int],
+        cond_img: torch.Tensor,
+        positive_prompt: str,
+        negative_prompt: str,
+        x_T: Optional[torch.Tensor]=None,
+        cfg_scale: float=1.,
+        cond_fn: Optional[Guidance]=None,
+        color_fix_type: str="none"
+    ) -> torch.Tensor:
+        self.make_schedule(num_steps=steps)
+        
+        device = next(self.model.parameters()).device
+        b = shape[0]
+        if x_T is None:
+            img = torch.randn(shape, device=device)
+        else:
+            img = x_T
+        
+        time_range = np.flip(self.timesteps) # [1000, 950, 900, ...]
+        total_steps = len(self.timesteps)
+        iterator = tqdm(time_range, desc="Spaced Sampler", total=total_steps)
+        
+        cond = {
+            "c_latent": [self.model.apply_condition_encoder(cond_img)],
+            "c_crossattn": [self.model.get_learned_conditioning([positive_prompt] * b)]
+        }
+        uncond = {
+            "c_latent": [self.model.apply_condition_encoder(cond_img)],
+            "c_crossattn": [self.model.get_learned_conditioning([negative_prompt] * b)]
+        }
+        for i, step in enumerate(iterator):
+            ts = torch.full((b,), step, device=device, dtype=torch.long)
+            index = torch.full_like(ts, fill_value=total_steps - i - 1)
+            img = self.p_sample(
+                img, cond, ts, index=index,
+                cfg_scale=cfg_scale, uncond=uncond,
+                cond_fn=cond_fn
+            )
+        
+        img_pixel = (self.model.decode_first_stage(img) + 1) / 2
+        # apply color correction (borrowed from StableSR)
+        if color_fix_type == "adain":
+            img_pixel = adaptive_instance_normalization(img_pixel, cond_img)
+        elif color_fix_type == "wavelet":
+            img_pixel = wavelet_reconstruction(img_pixel, cond_img)
+        else:
+            assert color_fix_type == "none", f"unexpected color fix type: {color_fix_type}"
+        return img_pixel
\ No newline at end of file
diff --git a/preset/test_datasets/19.jpg b/preset/test_datasets/19.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..69a5675382446d03ae72494a853681538315bbe1
GIT binary patch
literal 13552
zcmbVy1ymeMyY1iv_uvE%?w&z|yE_2}ch}&-9fE}5?hb<!+#LpY*Wd(yoOAyF-uv!b
zZ>{@Y&8q&sn(F#y@2Z-v?%Msj@VWuOkdcs<06;-O0Y1Nd0I!<_P2wJwW&nV+G(7+j
z001BWu%KW7P;cH}nGg!@Z)^PK6aCA#dh<X2>m2AepA8BcK>TK%-ZJ}}kMm|@-m+yV
z?B8v3-~7?HENqAbz@Yt4D`H}AZ2CdT!rsN+*}~r8gP1D&2Te;m6MJ`O06Qxi2R|z>
zKReq8Ha31vc79fFz#FRUe~`VtCFtybq7S`AhWRi4TM+;cfcY0bcmNygzc~=T5&Aa=
ziZ}nia)5&VCmQ|R_t5|HL-7BK1pr9B<-hvk-}d#t^y3?UslO^-PXQ=@^Z6ns4}g9>
z1EBp~0+{&!)%bN8@C|^1h=}+O5#`-G6ij3!WK3Li6clt^d~9r7Y;1f?l)uxz%s++y
zeuP3rMn*+L#Xv*Dz(GSp!};rghVxGo%>PM%*IxikL_jtm8wQF30F4O+g9-Jz3V;P5
zz4k&0y`hAAn-zaejDONwPea4P!6P8PLqdl7@0CyhXqbOjVgR6Ep<tk4q2UnV5#eA^
zINmBTVPMIz*x*Ezjj$;k1K8nl;&W<5DXDN(x{NtEodT!C5U3MW&o?hX&Rlr8wLvtW
z6V*(_yYpPQ-%qP=nO@d?>zQ%gzLKE*o*!JF)H}O#UC_`sw|k>u<`$A%*w_!*yXBG8
zG<Oe8DFP47@84m*joBMo*uOC$z{7F<4S}2u3-%3xBLy}bdq6zSUkJKXrp`I2j5nPE
z6U0<;LC%~P)bM`+K%n93o_+(7_ufQYo!iB9>pvj;-x1LMCxq7}01C|CzljMD0_=|O
zhue_6<NrfGmuRr5o3H3Q^Hg9?k;Yn+Hhq3<4+!M)FP-^jCzxYn$BN)7N3#$@iIj3#
zRQ~W|`j`OkhSq|s3}4_C0O$97Vd-Pb*_n?N@bg*31&WYE68zMw5StXKz~yn<-0++c
z;Gc>1Gcvxk6Gm{kng=E+NwuuxjC4{~G4@%fBr$A6_xyna6jlNy=2zbaM_#pZISG5<
zG3iH#^t(2VBsT8X#-X#m-g~a~AHUuAd@pIl?fab%+6Yl$YfgPbB<m0AScZ2rc+l6k
zpJpBUKR-yU=l3ck>BLl6Ti{}@2z0ty7f$Qlk^Z8;YBi8q_*FDBK*~H@mQX|D%IM6A
zx3HP5Y7quUHUq~~1Si_t%Nca~xpdjufzc6`icr)SOVu_*m%5rpv|UjbRy_9;gwdK_
z_|N4JMn~ZLccj1Nqkqbda{UT*(^vt_IiY&Oiv%{dxT{!)jI0)KB%z$g7n?-=7Cq+P
zQ&-J0T7n!&pX(K=9o_CN2|JqcJ}?m~UZywjzlg<P*s2_@Y%qEtSe#p@13Pxd;%L=z
zU4iQCE-0j-Aa}LRg<zoc@(NA+X=SY<G3{p}fW3{O)hpn2a_{1}qG+}`XRhw*liwwS
zSkaDgN(S6#g?^IHl6%eT>9rMu=<`c6TR3H{q>T-brD+`(yQ2NvoLH(YX^7^7P*)6&
zY18$w2@w0D3bl-Zx+6pJPG!L9oC|&VC47X1wxkYk-oVDO(~+^F#rko5Q0+>vPKzEj
zCwSc1O>mhj<l=$*#cG5F_C4?DrA)%7w#$)NreEmEfD>$T#As#}+CBMM=h}y4)!Hkv
zj`;v83jmC@x&;yfMK&UWTG`Dh6KZ)+luUCr4xaML;a+CCq~2yVi9&%rTXaQ0o(k+o
z(4<v4S1#uqM$NvId~!R!DFcJmSFBJvcwRT@R%w3`$~dH$csVIb|6DUyiExbvMd_iV
z#V=<4A#unjHA@ZAgm&(~y*nbYHl<+WpaqkrIa(^E270O-2^T^2l{Jr$&X@b^wH8^y
zY$#bG18rZmjL8Y3OkgXxru=E2xvKE`$sj361MlN4tL!?4!PWJt-(`AiEZiC87t3!K
znDZKEf(8^+gN}q>%Zjhl&wjcWB#&jMnpQx-ekv1>`!kv?v~$NCzlJ)jx)4M;UGv?_
zUeI6OKlLq}fmGvLNpHD_ZKqBg;#7vpra?Nt;EV|;Vd4GB%N6jqL12>@4Ym|a2#`X3
zrp1ln$d#p})0FAj2k{m!{`boIxPkZ0&m8S3{+|~Oyc0j!3|y0aw>wi5rwp=Yh`6Bk
zWb$q%@_9;JQZR%Rqq%3jA6;v*Zid0do~e%glI4P5iazfVd%40$4znF{Ywi2s#pdry
z@*k}lm9}f4NS~>n<-{$O;LK$E=Km?z)!BHCc1tgAW;Dri)3rr>1*~%AVZ6<~sg2W7
zX1E4)`TXOpQXM8rXJQj4oY{bxF@9D|9GLf**2w9fd7nhXy<;2Y7$iaj^#s2=?gg$`
zuY$CQ&Kv=x(<+6+;x6i(9m{aGSYMOR1Xh)B{f+WOgG@n%ng`q@-n}h-C`9G9G4?a7
zB?xLWdU~{egVm!Ku~2n0*Mgy05MAdQgMK7rv9WIfzq9h{1rAPi=X^wlW2g74NjDps
zUYNa=JNw!jI%0B!iS$U+t33lrzC7Q!k$<ai;UxZ!xXI~i-V=)HPi4J#P9pZCf8$N?
zQ0Sv*@d1w?T$i=xO=7&~{8Di<9$@XnQFA+qLR=bKqS4|Te|o3>3NQe@2XnQ43>Z}$
zQ>pP!fe14(g+fzd_$KPjT&IO1w(7BVYfv73-OqM;@~xR=t09%8Z*?wT&SA3FJFFPE
z%bZ(Z4b`BYH#cHy(2v7kiKgOfh)#O+jdaU1GBi?F5Lk@WB?%i)kJ#WcHRI9za#puA
zD~Y)tZ9=mh3dR3JFREQ<@_Fu0&3R@>8-{w?#%g0dUr~=@mWvmeChATQ+Minn2J-%t
zmX0z0ujq&_W;TA2isTB0h&6e1Ve<toJmB?@k2<$_ZBh=i^bICNBI^U7jdJ#We<)Tp
zCs-dG6(h}qSOqhbSbT}0n&_p0bGg>dZaDvwl~v+Z9d*6(9P`$2o3D>~n7wr&V;vkf
zp?IUDWhvi2r(0*}`W2%cb)NIEnMPF5hy=DwQ~_QA<GQwFo5clZSL7W(gs$HU2?8hq
z6o8YtADdLFuhUn}D$t@~ghgmUI|W8&tC*!!S_9bGA9Cd0$+vy47pEQau4o_UKSuq7
zo}1f!%L^>B4w#Hla>p!9t+xjR(FVhA>O=ykDQ;c?TUX_sM8`naodtOx(YgrPCiE9)
zvR44CeZeb0-=NCDp_rROziOoomWa+<kET^i?AR|T&7Ez1(ZOJQlD`k=oe$bax2~{1
z)DmKUcm~hCxHo5=TR=~Qs2ADlOSJi96nfhD!RdD!P&Ne{lw};G4MD(|mx3EP8fMEs
zb;lEP{?ID(*M&JGtXBu15<?XYX?pSx2qIQF&`6H#1vpYrCKio4{PZe>PVXpnS6g22
zjpAB|DI=~4{*92yr|aGj^852Efam)((T@I^{e240OWH4wh0B>7LR@Y>OWBbk1?yYe
ztccdg?@|k=!kflwH(n`Su(TK-E0|w^*^ko6@gK^gTDkT16cVH~YHTkP1UzxYr~@GO
z%n5P;-dBLY)ptjPu<4*~AA2(ma|oLRwOt2fO4o<`om*?hFU^jMXSpFTFsrecDJQ#i
z$~p*4wmRsIe(>X2l6e^czI5$ajF8N%@e=?!PvXpdp9zmGslL^p%kdIXLD;qhnEG}4
zhBWB<%j?XQ8E~3bDe5-)qH80&UPMDAE+6V!9&9g#viu@mZ5sYDQ*}s9G?p|yngwAr
zl<L6HP+vQE`_r5?N<+^j*_!>OiaueDw`)=T!MYOL3S7I{T~j7ojTFpFbHeYpk<`&A
z`JyIR#0}G8NagFi4w|@MEA4)0zk#&*ulQzniRX!UyOykzjTgP-xHcHf14Q78ltiiE
zcO)o)0J44s1@~E@GHv;1-T5xh&AGOPi3~o{qGyB}^rJm=^@qF+TWv+VFQQ|VzNYMG
zz@B9V2@L_TV8r^<xOmp++vptWfO`fggIQm;Mf2EV6PhTSaCergPwL0NJ?dX*H#g!5
zIb0x|?P}df6X_QddjxA1vvwrhJ~kdH#TH3OTR|Q1$gLHC41<_HK__J>wTk6-SFspz
zb;gd-t$ThBScIAy99^4&C!T5U#1#|V)Ac?z3!HHAz`l-?&p!&h`!NTwq-NrN%&w4A
zIwpm1Xg2-QV)Sdn1<%+mkiPxVpb>suW|3$NPJ#{gRiicY7o1nG)4j8>pF&=?#%q7=
z-#gHziqzim?Icp&DdEwE&0uD<Gr=sZv(fGqKbsz*{yHQ;PjxKW-<TBoJ$_3$m87oK
zQBk=B@u%%an6kf2!M&O1;3?5yb7Y`e+ia79O5)I{85cSvEelJeusXi_!Z+)(mi0P&
z<pELFF+xMj5l}RxL^7Dn-jI9;t#&xjNim^F3f6l$HB5a0pS=PsDa5U^6W?<!>>-OS
zc%EqL&+u1Dpv(4+mp`O}mu(daR-#*p=V$$R4j<E+-LCcgC6{a~16E*;+WgF_rofMG
zmET<Pwq593OrcvH7w(!`Aa>sAr+V9!{dIS0*c143q)c9ZcydlrCMJpp@p<zbH(&Ki
zgUhQqu`#xY&i3s(=KV|!Y;(FLYBsBO^za8C&9J-J4(pS-5gNwsymi*??@K%7k69{C
z(H-p#Zxe_GcUJTUTBtr-DRCOjKbfw3eK;l}IlXo`SIID89Q`z}8#6-oJ8OcNjLY`N
zhs%PfZ(CISjg$<~Vr5?h3L-`7h`mFns|Fq-In^$hI11xDk+V%t-)<UXv?UCQE;6kD
zpjV2#0{A!wV=<bt^2WRr5jDVyT#&k>QzlT=h<8bWfAd(J92BkdrqgY!?><{~x5A_=
zdFRj8dtFtJHl#I%<GRm+#Pt`<(bI-_!NIc`wwl#!yDbR#Gc8qLAm*SI$GXx*GD(Wg
zTpi!YEX}=FfRO-ksZM!%pR$+cdX6AHX&d97NmNMaJV@R=6z0RzQXzuV{U3(+HeW?L
z-K<vtG$09Zwc6UDqjLkb$UIw+6zSI_ou|Tv81EH^l1}L}R;+61P1@JFt0Ip(mBOF)
z4Gk(irm>rOBZ*JR;Jd^r@kQ>&2%?VpENhSLL%wlA2wPU>&D8aYo(|h;q-t>TV#Ojr
z_^V$))m6v%`q(Q#{%L!9D48`|vG^6hkO;-eR1iMHX6eAkL|`u@ogJ?#d5g^0{0i99
zs0X0h4gadl8Y{`>%6nvb1$2HmaVFI{MPwuEBZi{!*i!#m`(btc@}S$yw1q*7<TPwr
z_*{Ku%&otJlzz8DfalJB`r_Mro~w{xQ7Xh2-F%Dj6eul_%i8n|Nj{^YyawJU5zC-U
zUG?aE?5-WpQhl9RXM2l1u@^l)QXOMH6hFV0I0MwsvWJXxyWYXR)w7bZn!KU4dB15L
zh#gDd;#KgzKKs*@w^i35Z!=z8zYYqj;#8fao$=`fS(3iI={+s;=JnYUzR;L)qoY?;
zsHSAnh{rcX!Dy!0%IZ}EF@(!79ef{EB9sw>{Z{~ap4f|1kBwYYf((bO@4K~yBW-Gi
zsi~t35M{BN1<X|TGflK?XXWoz?cBn8$08?X5n|}(^$LC0)N7$4<4!mEJlk>0Ea}k0
zcW+&6{tO!iW7x%q7&FemzwN}oCk1{0{$7sp<}tipeEzAPnKN@ZR7ne3z9z#a8;V`g
zYwgwJKI~-<jLEheu<m^y&1iYzcSSWu7jZJ>RM9_nc(yi=SN#h3Rxn`mHfpgCQc=f1
z`9Qek&Pqt@MX=XOBCKGi^*0O;%ALCPVA+B>694nX6<V<1UOx50G&Rfv9lF60GX{yS
zqK9EtzdB=7dQ=L+7f?Cn!5Ke#kZ+{Y$~;$JCB|MVIp-^&o=RHaC#d%Ab`-KxI>76(
z6&Cs1!cDgIl)sLkv!&mMp>taI6@V8mJ`VCH>BOBjbFr@{`4s~Hj2Du1%UB+hWGbmd
zSyz;VKS1oCV3V=r+}cQf+9KqEZYiA~PMf)j8s!Wu`eqyB2$$VH4L?bhg$Q%VSCY^R
z%9&+s80F@%>K0-#)$=CSIU_Vybi6M;Te7l0W(kzMvLRae6jM>rdAr3;D`~HR1{9zY
zpec-ASVYC;TW>woAp?;-X1)Th=5)iTVDhMpD6<J~kETp8Oyn+eY|!PaAhIFt*9z{M
z)@2@ywY9n+sX$^BzDlyTImUR-#svr%0X^WBkGoy04{vud>cm_8I#6o>0evUV!|8;r
zB^hpmu&T`QL2hkw!_Rfhh0i(K5zkXVg;G2gQWGMiL<N9ZTE_RAzZWs+iyL%EEPn5^
zcy7u@g;teUI~=>uEmSVlj`mV8bD`UElFTl|Dd5~8XSUBw^OK>mPz#WPDw*X<%gTu*
zyu5|eEyxhd!Ls_KPg-_U`?zA+ZCM&EmU+tPcOQDqq{V7%o}v6jR_~or7Dk{5wBU6k
zh=YAqNac{AiK%~Z&=6;rtx@H+i2#kL;d!3krgk{11B}<&>xVn{Y;E@vFj*0v#z53r
zN7m&uUS7<n7fMBCsw1QPHlVWmIi$ix_NuLsg!3^EqF-$L9q&{ThSvL3nCP&kijY8n
z9CrRjCk2k>BegBkRpNOMKizu<)dC*_+XgU8F*JnNu82rUPe9v~bPw`-v_k1%A$s&k
zYi@xi(U2Ra3-#{g1v0nNWDnb$BJJK~_Ej=1nVHyOa!{{eRvcclO7nF-8C|>dsK*JB
znb0e!^U3gORxjvF6W*ZU!DQGA_%a!>y9|2rB@a|${VmNRbV>3zk9%#Q-<OxhayAH5
zC&{{HKUK)Wq{oY{V3PgwEjdh#DqHvs+L$06)Z3-5<B`BbQw<9i&2RpDS@xV#__{@k
z<g_*7E%^Rz%cYwVMz;)sJfvk{FCK1v(EifT=ZIyzJjH%$oIk$1;DXu0pH*y*vQkt?
zLZ~S<z`agk_-Bz>EDnOEts>s^ohAsSW}^$~G<o)7Jm^fV(zAR-F5tdBV!V$|>nw4#
zJ<aCmrmmE*hXe8YGRUK0?on_O@?simHg;4Y4rZGZ2$hw31>{C+I1`0yW23Ub<=qXf
zA>1zf2#L&OveC4BP`%cFw%_%mGw6$Dmcs6Kxz`vyb}v~SWcCc7=A_)nssbTbUtH>5
zAA|DIP7zjLtQwM{>!=??R@W}B_U&&IK6H}g(oWMZ`0>R&*s63Jngl<15zXu&!AHl$
z3Pk!9B1=ym;V#uKeHI7|<fm2U;j35LRuUZff?{d;K<oNbJ!`>J4$yIw=I8QW-~vUX
z;3tl}j3YwB(Z~n1c8D|Wx^`U1KG)KxuTu&%rnJf0JM+%ZDIw(T!IwJRZ9d#2NMD^K
zK%(dg^EnT_dcLLCl=mH}SiCb1dLvy7_8gpFZX}V3fB7<1nvjB=!(IWUTFljFY>OH@
zTcESg*F!=NazD5>h2~IXjgib$9xHLd&HF=ApLhw7UI7YyZZj^<dA>-!;ee!XlP(XW
z4Aw_DAy!LicTG=3;Fpq-98g#MiW`GKz7qX_&l$#J4CED{oc;>PGJh83im9n&w)_)3
z;XHlE5%2e5^A&)+mk==NQ}sHzaeCx=r*&-K>Jv*n;q6bAxC+$Q6-@bEMn2bh_v1w9
zn%qnpNu-apo2(gg7o(?}gk!e0+|jl2hS-(akt*v{rY17^u^cx5_R=IyP$<1-;ff!%
z_mJd8)*LUZc7}zlu-dybLCBHS={)&jrpv@1HORJtR;=EgbgzWxV;hEH>~L|4fbVU>
z!bT9BQ011N)Z{B**XO-o`W*2+%WB>$K-w0a=8?tKjymYjVlOLK_<66a-sJi*9cZol
zO?Y?I`KJ)1>=i&!%i#Sb&xT155B>4cr}GB!r|sDgZJ@xD5EnZwGYMB^pPnGzHNq{I
ztJp(mmCz4$7LUjE!%|9~LeOGMh}Ks!bNfGy-f445%{NV_$)u+avF(+gZ5!Hp`UgDE
zf227N(|Zj(<i0&Ye6482ITR=cwGFugSH3-#IL4iCFPbkZ8^w2>Z;DsNSDC_kde~#B
zx~){#j7=>aQ;s3Ez-v7+LM6atgqy>*p}qo4t6!X1mLJHj-E-{EStG-}KaS+l4B$n1
z6WG_=IWBCRMw$QSlaInDK{C2RCn2<NTAhIj@z<^zl_}M2zIdGaQ!yHaVpFbL%dZRO
zek6Wm%LMXF-!osYj7>a~M7tl@tsX4C0^n~FEW!6FY8TEPb~CM;^$ZIymYA0##w|LQ
ziyaO%7h`e_Iag(2yH;DqQDtzr!XYWZ9*LTuLtXPro!T0H@3nKUO38&oiv&o#>5gAs
z<&WjJ)o^?LL}&aMzd`-hhF?OE$S^`=-86<+|3{o(M%|t6)12QAogf~zU)0?a(qwUx
z%4EN#&EX-szkO^hK%5o+@aW{PfCRN#mN;S7Qs0@zeQzxo$!Um8@I2cUY^d*Xy*6eK
zIYaWk>YNJ8>*AZHAO=7}KtM!*M|x8hkdcs((eN?QP*KrHaPctlDaoj*D99+tX&87|
zXy`cU$SGJw*f{wF1O)}DnLkU5eUjn<3JUxc1PTcm84VSU7z2Y?;3LIHf&b_9kFGEX
z82m4~LN9>q|4<a(1cm<)4PJWz4Hy7uD3~|l00s^Y7WVCIpSOAdEG8TlIhzPPwz3fd
zg(J?}#(fSVrD#nT?%U2i6^F4?;5#I|_xJ>a)HJjo>F7DRxOsT__{BbpOGrvd%c!cU
zYiMd|>wrv5&CD$<t(;w4-P}Dqy@I|4hlGZOM<gVEPfAWn{gIZNmtRm=R9sS8TUX!E
z2ySX_>F(+6>mL{#8lIk+orBCTEG}(sZSU;v?H?Q-U0hyW-`w8)xqo<rO$PNQF8w?G
zBbEF&UT<o|KVk&JU%a58-Txwv2@6Ng29G78j9}!5O~D?3h$H%@iFCcA<WM=sHFlap
z!lUBcq`vrz+TYCnXT$>kznJ|8vH#@tzX=%I3&{jse7wcI-R$lTR^GCfZUiRO>Yh*5
zf*S;6lzc#I?lY%Sd1jDs?gZ(sPw(@(*L4>1p}FkTlomSMizRyVsUB@GW6}-hds;Ty
zu?e}`D*4#mu$pmTS#o{mz-JY^+u5FrQR(MIL^R~QGu|^)2ShAs^kG(h@A-lwLX~x`
zHANcj1dbZb#6R<C&d{g7*8KQ)nagdd)_Trs6o?0!g<qJV$}?n!R|~Lon#lWQK=@pC
zmb!n22HKSp91|${?kZcDHSd%s=xCm&W5$Z`yKr*~d$qs?>h#*Ro-~cue_-7<wiERe
zeGjp{;HUEf6ogmh6+pBN<f8l$`>ah7%>+1`O)gtWn$TIN!dN1%j7*o!6|fLf19Yw*
z^(vq=1-fYIyc{rO_WVhGr#(gw*J@`VyF^7oAxfm&RGfWwU9~<!<ilM7rJZ$Z-5hnG
zNYf7Cf$bUIE)0y`OyQPQfzt{afUg|rt#`>`4|V^TCDU2F5-Cd?=iu1qP$wXDa^f`V
zBwZ|9oPSdPE+gM~yf5{JYR>IgkxzB#SiXO+FiIO|s0EL7R;_k}97p4N=R{9G9yoRa
z=01eAqBQCn@M5|iqVSn1fxVjWl1ei2nV5CrP?TN#_ITBVzG;_~&(LAbmD5^6n*%nr
zztvg{2Q(cY>L^=-nWGVHD43-mG+cuuOWDa&MeqsS{C4Q<(~vFlFPkEHQM{+ZDu|p0
zG{*R7=`3@Z4+SZ_eWav0-iMtJ`!J+ld`jj{E6V^r^~Txv`)&eDllCefzqUFclm%!@
zCZ;DS>dh(Ta91gXkDaXcrb*NC#fRS(N11x+B&Ykg1*n;Gt(zo=Xo8sLzn12^S0;ay
z5(dJl*3I1V#y@H#vO?#YaU?BtH~p3dM<_uRECH=4FeW>`8xnfxq16YYS1_QMd$Gx2
z%!_PxBI#S#zEd!+R}GE3yCp45(`06Ia+3f^e29BDK)V!>d2_<u0-sVOY6Py2g*EyS
zvbC;a&7gCx9^6Z`4gGH7LqBV0SlO-n!m)?si1(3ap`ukM$lZ~s+!6X@07}v;!3X{Y
za?iK~JuT%(U?AqpwRrGC7ApBo>QBO9=;(9Gf?BEG5<8vJ_kzlBoQY7HIa>{)W;vDW
z;Zl@S!J8hdH2v`;Z|Z_++j#U1$>J*08DrS`Z_bk%?y+C_TN=N!kbk)@a~&MGU~$ye
z$iZnJ50$rawEUPiK=mbsQzF08aSXW4*d<)D>v8>P17?<#xvHA78%NWZE88<{GC=pD
z&#xn(S{o?IZC~P4`w(9m9bZH}!+W146wQwCjoTzixct6~QokE>zPVIInsR?{uZcX^
zUSn8cy475!Wt2M~_Icx)SVRA;+0~mRg?thHJo`{9(viMKqlhy&M0!V4{L?48)*H^Y
z%;+@n?x@7R&vwQn$G`%&ZO42-CY%0y4z(#)lLQGtokX1_@%pHK>G$}DX%&T4i(4o?
z$rToRC%naW{an;j@4wkbh!GKXnrB<{gcEgHQszfC>t<6lP@W4m{lbgExn)Pn7*eAb
z4CTiR$gVB9Pem`OX#EZMC*h-FJS@uqwJE`~b;2ws(74w#1)_<W^TMs7c}1)yP#OIA
zj!B0Lu$5w~F_n+ix7$2K!Z6u}zIkwF>Z7m$A49_yHoU_v`2~`<#A11JS}c*Y7_d2k
zCR?e{TNAS;)bVFRFtA-^X$hH#x575XMwxk%fqiMn1rfD9-LQu;MW7|nTm28qz-qG4
z_XolQCAZ1nMUn|TN5C*f&%KOd^px|$s_%MWb)nP}W`EP$0URdOj%kjHZB+8X!|9yX
z=sZLIk^`$*Jt<VJ1nM;QC|;og)ubPCW~oB%v0gLrFx^9^YK+DfYB?uf50FS!HAXKe
zl{>zD@4Wse^@4%CWA;-yNtWM~52WiEj_+rlbKQnEG)taT8kh=}zhu7xH1@bDtARqw
z1_A;ZJ1A*8IibpBbR6%xrzuTVX<!`1xjz3G#L>*9pjRj0bSogSFcUNRCc|%33LlUs
zKc*4DSAwXG15bccu4ZmK+I3e3Qk<FluBZ34lpkg-FAzdj-|rZonUq&1SFZB0e-eSB
zP$QxhXWs`CX0;?8PGmPIVm1Ku-4tD6E+|T)A|0~DI>y{jA}-EHr!O^@w|4E3r;jx<
zLdFu8`c!&0714JNygV|%sEgO7VKuflM*m|BUbb=D3>(4jQ`ARSVQ=}IJXs<?%8?1D
z(G-qoN{{3-o$p|E4U4Z9!&?fb+_kT?TuB{j7D%r?mk3*0TJ=el#0l$`ijOF<8%3~j
zzK5Je=SQQ<aT}X`_l(*M8!*T!&<SC^ga_i7<tNozSQ>0Q#&QRHPRbRa7m?|VSaO>x
zD(gf!PS<h?t;)8!IA-<N7v9v=l8{VE9LFDO_APl@Cl3PQoaJ3OAmvLd3~|UB2@!^4
zYcTMt@fvVu$PIERJL>sAKd`4qbSoR17Ovzu-H=r<&Xb=)A;6xCQ`yzBkXUmcj<U0x
z2My-(w_=-&cH-r({xMeq8>NfyvCMf^VPp390LQ!~O1Jd652ZJ{d#<bAuW|d+FV$B%
zYpw`~)vHJ9GDJQG2SVPbX&i*H%i}l`LlPHy5E^(14L4?DNf<=1K<#|uY($heTMIde
zKkz9-u5;b7WF|5T%Tn^k<PwrMjeYWh(v+3L9onvOKf#XX=kAkOCo5YfnZ+vn5h;gj
zYMC;%@$4B76<Q;a+<hvWIa3RpITHVRm}gac@X+4yL@W&(P;(>@+p9T+$CJEN@Qp{Q
z@xTs4n&5iEzB^rePDVV^QuAbSVwFlwVe^tcot>St#=K;Oo}kF$rvuRHbkv2{X>c}j
zSM|(_-HDz-HYp8dKmAIDZLgpeS|@I3INfysH^obQtgiTRm<RocLBfnYMliR|PT#Oe
zT6mam#V;OL*{q0%5@U{v9ru%4(Gjap=HX1ElVFmzZrY*3L9&%38}R74po@&znh>vS
ziW@QKD7>Z}w#UI}t;8A&b&;gR)?}Y7THH5RQ(#<RL*t!yj|6E2pFG8Zzw~7o042du
zPa*dF6`(YWBQ4;Q#Fmx%w%ckDYYT{P0-#)1gnd8^ic%lCj8=`x*m3T_IWM(v7Rb4R
zVMcCwS`Rq+SXm7hjZ6w4;M5f2Fy*7poxC@cQB*?JVz>+;K~;xIYsS21(^vlF_8E;Z
zU6*q&wq@1>c(8!6IaHY<cCJuRj_bZ`4z)k}GwK-*Wcj}Fmj9dT$}bvmLkDl~$N9>G
zmiCw;J1X#xFY}p_e^jN8^JK6fiaX{@U)d)?@?ah?HrNSK<f-)FkLBhRffh|mSk@Lc
z%u}BoL&?UhKY*Q%x3Wl)io2%E$`AQ0k)`lU!8pGp>%zdr;ZqrCbFIR&Kwl`hA<U#9
zDcoD^<FD0Kku!8<eHkzp$!gpC$*{^iEI5I<thUe^)mg%!E*ILxKC5WDm_@Pc4;yo`
z5CL`pwGewYhf!G&0r2Qu_(zcwI*AK6vuV94h9v_|@n68BqodVSXT6;ic;ukOeg`Xw
ztf(NhX%iwREBfdAQcWIFt%wKWh!N{puka$ENJ*`D3fH@2Zx0iMJ@uvaU{lXj&O_?c
z%m)3X5&^_1a3XI7K6@`5Y+^qH5=6l|XNEwP1AKnoV6Dsp(x#A=k`?&6<0NTcV74-h
z4ij}l8u-K6Yr5mz_H=1|YphOFe?PMiW=Ww@Mz9SnS7?EBc&fL?hN;?$6aY=S(rY=b
z>%hz-MOhb>!fYHrM@SeBxA&_!<jU-V!I^I%?>Qb45gV+vTcWx)dHKn3_97iAk^pCi
zYr39~fZMXX=};J0B3+0W8?{&O5*f}^`K?-ylO6pm-TcWK=+C(pQh3WSc_YuzXZI)O
zjP2J@ydqhWE8cv~c)+&9uos5)=o55;6=7W$?Qa|rT7*vAyrRZ}!u2-9sA6>T*kJ@n
z&TNF;*?=jZkhv#mM6FpyXT+Z0CUxMN2+r_IY+F03h3>JNv*VuagnSPUU0>O{Xp)ZY
z(q6~Oq+UQwZdrMDR3Ksr%wFbR$&aOs`0t)I<||$r!f8@3D(nm;rC=S=)jcVPsuizU
zK3yj;tL%=65^4F65q0Zn0(;SlichL54)>2w5PIINvQ}r!{IG!O;_(8{KJmU$X*-8G
zu%*x$(tb{*-I~&+tp;pN9#<J`A*an<g{92NxV`A|P%;-;Mxr{Y1J0GQPhL#Jj5EVe
zH%)7l;M5C+SuSv<5F=`QOx~xo?)UhGtsG9PKn5P1)&BfjIx9T9hos4-X3<9pT<HyE
z)z;v>h}O?;`mv1ts3fR-&H2aL>7p|Mr>cvVS8E9%A`qFE#ADu*yxpRf<s>r%e3kAu
zBJJaoQYhG5UZnn6zMwhOwn%@CY(-StyLnS`%j)O=n(gP$x7`dk46PS)=*Ab2zamWb
z0m5|NclZa)>E?Y6`Ge8tDx(LczK-u>w-o1t!(!5{dxDF_e)ZHmx*3a?Udcnn1kmQ3
z)hxdyB%Ifp>8YufXDi+2A|LWXW+1}8rbg<Vr`lT%#Jh^h3hf}K#QsnL&j$#VeZ%or
ztae%g7)_v}M6+duGoHTWp^mkK$PT7nXdufgU^>5%>7|C9?~I`=BkWkGuSGHL#-Umz
zKVwANsi>qR;Q%Jiv%L5hXmak4@$v`;Lc7JTv$?{myhrfdFdSaj?44tuF8KVW`gwOa
z2k2J(3s+`B9b7}k*%f*x$pv4u#aFp$!8AWCfp2yuE*g9*!J*O~A5^Z0+h4@is1#SS
zlq10rbCcwd)#9`oVl8ioJfW&Z%07!45;N87Ho^f1>k88@sGug75v3ss7r`b;8G-;y
z7_!{=-t7{f3vTUw;D%vr@o$a#pMwp^tOKOR=RK3NgoO!U%C9DjLCwGA^@-1ZgA9=W
z*kf=y!+*Jz*HI?B8>$jaTRQ6Na<#acEP2T-4Att4JX@(DjUKZq)ph2fiMbp7Ol+v8
zq8F<-PuxM-p&D`-+pJ^K^86Tu@Fx1ZT0NIb6oHCb+hbNy5(hE8(e@%X_tLbpzQ&BN
zl{<?Y?=7QSJYZ-fM;G{6D0Ucw8pN(T<+9FVy+Rdg-pftP$msUnEw~Cs0LN@16Ba%z
z<5;l9SBc}|mhm2vj8Ze;+2vh!_ea)p;j0_?s~Ydb>eN%ohCzevLbBD^I>)pg#Xk~H
zIBwYCU+2e}^2C%aWe}_YQF*xqd~rv(w}!<v^|w-kK77QdY?qSgr;EuD{GKDsM_Iwc
z1Yb{v6`~q{gB$8+f9o9-a0#XWhiLAfu9baRF5kq7aggFHR&yw+9GzlMPArx+awk6D
z_9BHku&$eGll3VJ+PAF0Li;rORCJhjbt@UDQjYjxKHHL`Vf-h|8Ne2>z0__kVYP3E
z<s{z+M9p#84*T<fq@;guEUyu}d#Dm9X=OEd`aKefUifWIHR(DMXaB7d($B)~DGKuo
zp+qq7r<sCykr5$B4H968w!zacg6qqv^}dNvbL4L9`Ya@*b1O1h%(eVO54^+5?$2Yr
z<)}Q+bY1d1pQ*}a4Natrtvv#-#=X2NmR_MCD7M1U@m$_Y=g2TSW@DQe{y3{bOkrMB
zMU$%_fmWp=u&8FQ`pkx#&w`j$sM`w?z$zmssp1nY0s6rY=vgb^koX=IPU$f|uY$_K
zyOGHtvqzf@tRkHjJl3#8pqA<)5$AX(@yjzezp<gI?Yo#gw$#8(_Cw~ZhEoY*tMo!*
zcJs{II0Ro^DjI?64m3L_8iPp5wW@b0?V^~fdtP3afo=A+?iJwUT49*SH11jCovI|I
z29?%=81Du~XkeG6B-l#A2@0f>32u<k#33z<BK%&ti+@CyOY2m~#rVt}ldm4ERndRA
zlKgV=b^C3VJx-`%)NE5z^~NcJ3vAvf=_+Kvo^)UkWr*hXY%5EpiT)DHW=h7=yj5Yf
zM(o4BVxT+D%!J1N%jw5q)<ro^Qnb&{hN;vn)_yC7sQh>o*Fl`mPZyB<Hq{;bvrvkp
z3Mx@kzEj9%^$PF798+e?5(+d?80lw52es?sp<P|7L^%l4(=&MialIV+rvCRLYw;6H
z1v~c1_4*6j;9;yX*5z4TvTcpdO`4iFNr*}D3BK*OY`P?ysE4D@M#u}fe#FN$^}>vp
zz^lF!oJd(azjpGMz-HOU<FflvFnfLUG7w=>y$GHyF_6BMLV6}hePm&1HgTo=8}5FE
z-5*Vf#<v$|n9g|Mv}B=+V2L3_5~`mx)z42}0lk-|H(>iiy@e7k33*)9w^#LE7Hw|c
zolYjJ*+5&B;8s=jOSHQnzEDPtdi*)o3{BypTG-7Ga|gA5(i=t}-oGjyPSZQZf3`NO
ze78_X@^m;AehZ|x*45_vc^yKHbUIRfknPGPOx%b-HUDMfSQ$%iRC8ZI^?;9SAfn{N
zl8^j~r@6w~2rbKcXL;cA6@X1z=_8MuJJRfN0Ec^NSC-fYCO@JMsZPw+4TSOYt0_~h
zHz|XObz+F9e@|wi_KoB06C7?G6*w8ruvw*<$bSX#39+hSH<l|t@MmMe&+ozE{Tz@w
zvtW)sJ?#WD6kV{4mvD$OhhJH2rgxT1^Q@ZTK2uzIL5=P(_XXkv%Jv@cB}qrCe+8z`
zicJOC1j~C)Tg`R@>7HPVTEk9-t0se~^Ap){<-tR5nl~_M`p1bKLh?nXW9C-bNa}+V
z5S4v>ByEee2f<@dNax_&k_g_`HQrfYxgEMHsjsR+tjz_Eod|vfQFwywmN0jM1byH#
zk^6AQ6lrMF!@M3}saWc`4hPAi{}L~~qh|Rw_QC|V6B6+(Pn>PDVLQ=YiH&slqEJUZ
z(&Cu57T!MPCnjqr5K!&h>e^k*LZD=6<;&5|@)fjlXVFdh!U7t{m{-{PcJx@&8J2jD
zDpgeM4|($(Ke5;Xorry6x5A7t$Lh;_d?S4_|LRQ2;j5PS@XbR2*R}MA2*|nqLgNKj
zE88!_VLwbQ<Ge8m*})bsY1~i3&Q+^^ZI0T}2@Ie<a?M^II4|t9es+{u!}T*`{w!=t
zhJZAabhnaAd<V;#-f5ZlmGZ6vxvGqZjnuCIukDz{%ATH@u#-LKUzPwe|5v~<gjl)(
z8g0e6rOd*2$ds7&H&pUN+W8fIXKUTj$JW;5qyPgnaa<`qPa7v*PR)r&Zh_RalrWpf
zdVc48hZr6-<k$oHipExAv5`#I%n!cDJd_94-?2XaPW!?$1Klw+C8Ia2u(9pj__8n!
z4IeeG_X>FTi!9T!*1Av*P8h=lszsBK3}|`OixLzu-cGA4UF&IEGZ!r$D_7MZboysA
z92SC?(V_j$Nrpvlu7m(4qxR$)b+A&}P)Ac%)OBH&N6`3k@vd|fU+~5T!wFT5M!W|z
z=|(kP5gYzs;Q>OQ7}7mQ_;sz?)V9*ogce^gZ>cD11O^~Qd7L(YCNgM@?}W(ZJFji+
z{CKQYDfy^JHr<ZI8;m<bC7)<ok{5FjLLfo9LRbo!kz{>d=2nE0aMq#Q3Ujj!l5LER
zy&JMBaIG*HN%vHSS>i<2*32OI%Ak@UTl^SJgPc0CG{6*9O6+S6`GJ>ZZiSF&iv{}A
zcqa6t^vjZS)%w$6k1B1c`9&pxk5L&<j^$yFnWUKwv@ouUi#DH$#w)<nS%D|#&vMj^
z{4b;W;)gT`)tLO1;znPSD5fR{c8sj{6X1~m4YD^**(HC|R^it&JJmom2?5mx6TZ4R
zOzw(_5gLBa2v4D~uBYSK0_7p6+O7I#O|A5mBZZwyaWlh*x#!QucY+Q2IE=$&g`O4X
zhaa_?fp+{1bQwG!2Z3d}iZeRods2d+N!3cz01A<J%IiUz*O26A2!lnMG`fC!`$Trp
zDf5bUq?1LKVjb3M8SXvh{ZF9+YpI<Cd6XKR{twZ;-~7tsT+FUXyGuuSaIbimGMB)u
zwRFsCRF|bCXBMT{<E);!xj6OIbrW<Vxe*VIG4WkB8cfO3J6dtrS<?|8Wlq8^qBqK{
z0pa{;MYD4n)XP9P9E~D=LXxT4D7A+m73AoFyZDehV#1HbF565~nkEOB{VJm9d5#a)
zCERiVmAuT1y62C%P1Z9}q`~5Jn1e_O1lmNCQ7ZW-L8w<uIGVu3sd#F0V-+md;mnNE
zIy$j_aU3;CX{X4FP=-6s)_~vQxCIe7JWUl0K4S9T7WJ)z`<I`(`yB&u!(e!iV?O4V
zqKzLwel{l49}|Asl0V;4SAIee2rUNonR=>>f2|K+91t#*&W*S>PKafvt%ybj5;I1)
z5(NhemWN*n&5*#^cGTI!RV~yX1XmP4F6>DwK1#OZXvXr_l~&MZ4?7*RLp8Dj?($>y
lKvzRo`_-r7xD)h_5mWDS5@cc!RYR^hmBl&B&8A)#{}<+yj12$)

literal 0
HcmV?d00001

diff --git a/preset/test_datasets/49.jpg b/preset/test_datasets/49.jpg
new file mode 100644
index 0000000000000000000000000000000000000000..de46b72825be0f49126adc1ba2ae35d5a0a2a6d6
GIT binary patch
literal 2790
zcmbW!c{J4h9tZH>7z|U{Vlc$mmk^RBThU`J%h+X<bu42WWUG{xCyj~FB!ej-yFoK%
z5Q!L%HpWsRWn^qA>kyt?J<mP&o^$`Y_r5>p`~BzpIiK@+o$t>3ozH-Xot3Q>00IF3
zXt#i!pMdd6XG?QeM`t5MaG-WrbWkuNUQb6Cp?*F#G%^@rYiEtHbF#EX#2e~sXd!Ut
z{lY`zk`Sf{q%L%46gUa+^YZcY@$&QY@k1c|0#FesR8SBqE+o8HL{eNzN>UsKgC9_l
zhs!9*!e9t3gp#V7hNgzJytb~kx~_`4hWf7|AP58k6@c!CLiek~VQ}^TIXmrur~u#u
zB!WST0FNjLEDG9r0m$zn`9OaK_-8;o;N5*80)o&zyA3oEfCmHy^YDWC_;`7DyD#m&
z1H7VqVhVcZ{QFUU5JjB0eri^cfRaW1a|!4EMdic(F;@klFi9zCxQeQpx`w8~5yPX%
zV@4-`v$V3dv9)t?b#wRd^g0s|7=#H92@Q*lJ0G8rn3R0&`i<XjrrpA4=iDXU%e|jR
zdQ@Ce`nasTqO#!$mDWgSG&Og0zUb;^_WaSy8W<cJe)s;v$OL=x%M@pNW_E6Ad1ZBt
zyT0*b^Op+*fd6Lg-hadX!zH@w;^E~5^Fn^PKs*V%2Q13Vr=Z6#W{!gR;r1))rwWK$
zWEIsv7gRdzyeQ!x(+`Cy8%(Gy{i6LP`|n^^|1a6UVE=ZF0DHlp-OmGy0>^=+!Is?m
zQvH@cCrOogRZt{FB@dl&##+|-O00vDX}R1K#^sD0oFzo3c!hSDOEd1MvQKRDUm#!m
zBmp6khmI7HhUa9`6GAw*I*Tc8zA^iZiBfa<)|HtzQ{)@bHc>v!7r%iZGkmxd!^Mil
zwO3JKYUAM*nYt;*b?@<3i-E3qGfh^1sRp)IF;(%>(EW_b@uqVkb1pu!xrUNX{J$R)
z|2_7m%fjYxHt|>pHJJ6Zk!(oXY__w`yG}W=Sa#FLU{4jNk(312{ah@fa1NcI+H~r>
z>OliTZW-<lBx|fd*pInZ`__9S^L)_6p}No=0Moz_tq$xcXx^G85K?HQNMS7TdM{b1
ziWpcLWhz;F!#l!ZB+)A>M_c5FN5VNf&XGJ~e$dk^85DZi4p2#rio?FFF&cfjhkNV2
z!Jk)`ayBrGx6-r8<oh!Xwr$Thgsy?jZlm$%yhyh1m-L*uIT!K0oVGl$Po3VFrUG(I
z2pcwOi|<fj&|7iQ4|F4*O`y9|OXp#_UD1lKtBVn^1E!YY_9oc`u@v&%O2298nN?jU
z+(+5V>2MBrYMK$8)eZ7z@Pbk;<23^5@sb*cpsoZqel>bcEc4(@_`|76D{-8!>%5#L
zTG!XPI&r}h{Ty|WDdB+Nb<NjfmEN|k!1O=z>Z4v>Ftsp78ARtogw;-`vgm8VpfLBw
zhOtqpD8K8Gxjkxe0fQC5<QS&<qRHaiM&^q~u0A?Hq+S1M`ude~-?mmM$!&|Aydcah
zFI`X_o%DQ>IZtrEbJHNQ7exrdTx@~gPYYD>j1e%w6Dbsl!V=;)w=}J6*=*y_WMV6u
zvd^*f?3>|F&+4J9NYsj|5#chK*q3;`qgkh_$TC4?0PTYuTPj{yzUga?9(Kw!Zk|aN
zFFdX75avAa><8DR<pt@Mh(NgSa(}L>t=Zyn$9d0$;4SwH7_kU<1!b4^F@6WLHsYL|
z%AyZ#%wD3Vv#aqm&%*L_U)E2FDH1m|_VKwt0F3KiiMwXft-=RVXf;UhQ@Ll)lkPS(
zV*?tD*rBirZpB^Aa%ju*fDI4h%8hiFcS;O=M?bmi-Ciho<IA9tTv~&}W6sCGbVds9
zY|g`3uA_`}w-Vvf&}aW5P1jEb?rp&^VWFj%sq}^w<PHG&^y&SmjP_ZUD-9wredq89
zo#rib=|RX<*Y`;#t2H5hR1x%P%=PvKnm>lGV+%ih==-CA;?``L-xymKQdfmM0|@~K
zjNP^zf)XMeouvc)>|AP4p$&68K+j`(mNi!0)@o%vzCa|IgYSqCV_yg+eUH{+*bJqH
zZ?W+d4EYTcT046b?k#h=BS*0^FSIQ_8UBVTQ=SzkO3IVMJedq2(?a*DrW0O&1iR(i
z4k23Uz*Fm@8FMMc)?Vg|%cDlhNBfW%|Cb2AtJTJ$14R-mgAR~0v)js>H(osppk|wg
zDqT~?_N~5M7H8+LqiT%osH~sW3F8WfmBo6~lnXSL2PZk<I5(;=-{{Oy!Sb+mNlAFy
zr|}XODK;UiS2pHF+%e?a$Te(N!qSX<P&1>`kyG`<cjZcK>d#A>g9-KX!qZ*4g(^Q9
zgp$nK)(X{I{$S$lsKs)gU;Q>}IU~qd$%f?ehJpu%7Vo3bbRdQHvU-zzJN(H^&PnBd
zz0e_R(rdZ+?>b57JBd=vx#W*=qvhC$*Bsqrh+y!69RS`rs1@N$J=L=phfd24`NUwD
z#*(*JB{=L#_cAwwbv_T0A%<|&BX@~2G43=;yz}i`&k~xMMop-rONvUuSZ3YwACW!s
zMcY-B#qO5U4?QU#f^o_b%M?%}VQqm_vvkO(agbU<W7XKx3Y-PKqkoiy2bgt|*8F4!
z*2$0cI5AX5t$TfD5~q$G5GNlzemZ-Cc5ll<Ap_P+kPRNE-ir8q4WDfwt=2S$HH$=P
z**l!|&KM4aeHX9u?p0L3I6T*MoP~ear^7QQ@TF7XYe-e%eWq0!_Z1uOEWxy^E4P_7
zPv@`Y7&gJhG!w^2FwN-Gs94K~iyPm4)UI#);2!FipnmI*(wWAfmj+C4#G1~4`H5eV
zgSccFGU8qBkp<gtwYI<m;^D&LZEVwqKZ4dz+rx&8cx~Uo^*DjrH>X!6rus9mvDQPi
zabL2LvPm(%4M!1KeSIZ4g9G__2MUL`ZWX@uFd2wqjlXF>rTbDAFn&;qs$0^2ify}K
zm5u5<c%^K<8++7qNNuQgggqI?_3ZO=61oDda1cEcsXmaCT-eo?E_)9~cX!Yc3vM4d
zoHd%k9E3k$TXU`QCpu*t`&Fyrs~Ui$v^edI&dqmuUAQAHFgnR5e@Lq0Pfg-jrWRqj
z#h6^pRAzD4-A|#gkT`6JnT!$b#?XL+lBWaE9s4uppr*Y^>6ga0F?9=DwI0W6qM&oL
zzH}$!8~T6~Iq%zXQKt<&yRTan*h_XT_`G$&eP3y@hf(xKn)8{i<1TO4Mq66!%@-z~
zVMISaVJ>G}nyC`zYwenUZDQxN0|--^9yQ)}6DVN^YXuo^tG1uuJ2D6_e1}i%0fs{h
zUT3WqAK7nfFBa>L;n<tAAFUuCzO*67oDaKy>e}#=?XOV6t5~P!ZuNC^xizCP!ui_7
zHyuQsw0Ucd%Idg%(Zb0TgGmT~!%vpYyxSLse`q<>O$SYiN-k4wX(nm=k+@TS*Mst#
zW*`+5r^WRf?0Tw~C4=PLG~O$?xs@VcV`*1kRy>p!ndBu*SU_%~&gNbYRV!?^xtf`L
OFn@k{OZmahhkpPZm>o_4

literal 0
HcmV?d00001

diff --git a/requirements.txt b/requirements.txt
new file mode 100644
index 0000000..1871195
--- /dev/null
+++ b/requirements.txt
@@ -0,0 +1,19 @@
+torch==2.0.1
+torchvision
+xformers
+pytorch_lightning
+einops
+open-clip-torch
+omegaconf
+torchmetrics
+triton==2.0.0
+opencv-python-headless
+scipy
+matplotlib
+lpips
+gradio
+chardet
+transformers
+facexlib
+pyiqa
+pydantic==1.10.11
\ No newline at end of file
diff --git a/scripts/make_file_list.py b/scripts/make_file_list.py
new file mode 100644
index 0000000..df4ebb8
--- /dev/null
+++ b/scripts/make_file_list.py
@@ -0,0 +1,36 @@
+import sys
+sys.path.append(".")
+import os
+from argparse import ArgumentParser
+
+from utils.file import list_image_files
+
+
+parser = ArgumentParser()
+# parser.add_argument("--img_folder", type=str, required=True)
+parser.add_argument("--img_folder", nargs="+", required=True)
+parser.add_argument("--val_size", type=int, required=True)
+parser.add_argument("--save_folder", type=str, required=True)
+parser.add_argument("--follow_links", action="store_true")
+args = parser.parse_args()
+
+files = list_image_files(
+    args.img_folder, exts=(".jpg", ".png", ".jpeg"), follow_links=args.follow_links,
+    log_progress=True, log_every_n_files=10000
+)
+
+print(f"find {len(files)} images in {args.img_folder}")
+assert args.val_size < len(files)
+
+val_files = files[:args.val_size]
+train_files = files[args.val_size:]
+
+os.makedirs(args.save_folder, exist_ok=True)
+
+with open(os.path.join(args.save_folder, "train.list"), "w") as fp:
+    for file_path in train_files:
+        fp.write(f"{file_path}\n")
+
+with open(os.path.join(args.save_folder, "val.list"), "w") as fp:
+    for file_path in val_files:
+        fp.write(f"{file_path}\n")
diff --git a/scripts/make_stage2_init_weight.py b/scripts/make_stage2_init_weight.py
new file mode 100644
index 0000000..d8f57a6
--- /dev/null
+++ b/scripts/make_stage2_init_weight.py
@@ -0,0 +1,80 @@
+import sys
+sys.path.append(".")
+from argparse import ArgumentParser
+from typing import Dict
+
+import torch
+from omegaconf import OmegaConf
+
+from utils.common import instantiate_from_config
+
+
+def load_weight(weight_path: str) -> Dict[str, torch.Tensor]:
+    weight = torch.load(weight_path)
+    if "state_dict" in weight:
+        weight = weight["state_dict"]
+
+    pure_weight = {}
+    for key, val in weight.items():
+        if key.startswith("module."):
+            key = key[len("module."):]
+        pure_weight[key] = val
+
+    return pure_weight
+
+parser = ArgumentParser()
+parser.add_argument("--cldm_config", type=str, default = 'configs/model/ccsr_stage1.yaml')
+parser.add_argument("--sd_weight", type=str, default='****/v2-1_512-ema-pruned.ckpt')
+parser.add_argument("--swinir_weight", type=str, required=False)
+parser.add_argument("--output", type=str, default='weights/init_weight_ccsr.ckpt')
+args = parser.parse_args()
+
+model = instantiate_from_config(OmegaConf.load(args.cldm_config))
+
+sd_weights = load_weight(args.sd_weight)
+# swinir_weights = load_weight(args.swinir_weight)
+scratch_weights = model.state_dict()
+
+init_weights = {}
+for weight_name in scratch_weights.keys():
+    # find target pretrained weights for this weight
+    if weight_name.startswith("control_"):
+        suffix = weight_name[len("control_"):]
+        target_name = f"model.diffusion_{suffix}"
+        target_model_weights = sd_weights
+    # elif weight_name.startswith("preprocess_model."):
+    #     suffix = weight_name[len("preprocess_model."):]
+    #     target_name = suffix
+    #     target_model_weights = swinir_weights
+    elif weight_name.startswith("cond_encoder."):
+        suffix = weight_name[len("cond_encoder."):]
+        target_name = F"first_stage_model.{suffix}"
+        target_model_weights = sd_weights
+    else:
+        target_name = weight_name
+        target_model_weights = sd_weights
+    
+    # if target weight exist in pretrained model
+    print(f"copy weights: {target_name} -> {weight_name}")
+    if target_name in target_model_weights:
+        # get pretrained weight
+        target_weight = target_model_weights[target_name]
+        target_shape = target_weight.shape
+        model_shape = scratch_weights[weight_name].shape
+        # if pretrained weight has the same shape with model weight, we make a copy
+        if model_shape == target_shape:
+            init_weights[weight_name] = target_weight.clone()
+        # else we copy pretrained weight with additional channels initialized to zero
+        else:
+            newly_added_channels = model_shape[1] - target_shape[1]
+            oc, _, h, w = target_shape
+            zero_weight = torch.zeros((oc, newly_added_channels, h, w)).type_as(target_weight)
+            init_weights[weight_name] = torch.cat((target_weight.clone(), zero_weight), dim=1)
+            print(f"add zero weight to {target_name} in pretrained weights, newly added channels = {newly_added_channels}")
+    else:
+        init_weights[weight_name] = scratch_weights[weight_name].clone()
+        print(f"These weights are newly added: {weight_name}")
+
+model.load_state_dict(init_weights, strict=True)
+torch.save(model.state_dict(), args.output)
+print("Done.")
diff --git a/scripts/test.sh b/scripts/test.sh
new file mode 100644
index 0000000..e69de29
diff --git a/sh/1_pip.sh b/sh/1_pip.sh
new file mode 100644
index 0000000..11de99b
--- /dev/null
+++ b/sh/1_pip.sh
@@ -0,0 +1,3 @@
+cd CCSR
+pip install -r requirements.txt
+pip install -e git+https://github.com/CompVis/taming-transformers.git@master#egg=taming-transformers
\ No newline at end of file
diff --git a/sh/2_inference.sh b/sh/2_inference.sh
new file mode 100644
index 0000000..18e735a
--- /dev/null
+++ b/sh/2_inference.sh
@@ -0,0 +1,14 @@
+
+
+python inference_ccsr.py \
+--input preset/test_datasets \
+--config configs/model/ccsr_stage2.yaml \
+--ckpt weights/real-world_ccsr.ckpt \
+--steps 45 \
+--sr_scale 4 \
+--t_max 0.6667 \
+--t_min 0.3333 \
+--color_fix_type adain \
+--output experiments/test \
+--device cuda \
+--repeat_times 1 
diff --git a/sh/3_create_weight.sh b/sh/3_create_weight.sh
new file mode 100644
index 0000000..3ac895d
--- /dev/null
+++ b/sh/3_create_weight.sh
@@ -0,0 +1,11 @@
+
+
+
+
+
+
+python scripts/make_stage2_init_weight.py \
+--cldm_config configs/model/ccsr_stage1.yaml \
+--sd_weight ****/v2-1_512-ema-pruned.ckpt \
+--output weights/init_weight_ccsr.ckpt
+
diff --git a/sh/4_generate_data_file.sh b/sh/4_generate_data_file.sh
new file mode 100644
index 0000000..fe8fd24
--- /dev/null
+++ b/sh/4_generate_data_file.sh
@@ -0,0 +1,10 @@
+
+
+
+
+python scripts/make_file_list.py \
+--img_folder ****\
+**** \
+--val_size 10 \
+--save_folder 'preset/train_datasets' \
+--follow_links
\ No newline at end of file
diff --git a/sh/5_train_stage1.sh b/sh/5_train_stage1.sh
new file mode 100644
index 0000000..b5abd7a
--- /dev/null
+++ b/sh/5_train_stage1.sh
@@ -0,0 +1,3 @@
+
+
+python train.py --config configs/train_ccsr_stage1.yaml
\ No newline at end of file
diff --git a/sh/6_train_stage2.sh b/sh/6_train_stage2.sh
new file mode 100644
index 0000000..8638f3d
--- /dev/null
+++ b/sh/6_train_stage2.sh
@@ -0,0 +1,3 @@
+
+
+python train.py --config configs/train_ccsr_stage2.yaml
\ No newline at end of file
diff --git a/train.py b/train.py
new file mode 100644
index 0000000..f2327b8
--- /dev/null
+++ b/train.py
@@ -0,0 +1,32 @@
+from argparse import ArgumentParser
+
+import pytorch_lightning as pl
+from omegaconf import OmegaConf
+import torch
+
+from utils.common import instantiate_from_config, load_state_dict
+
+
+def main() -> None:
+    parser = ArgumentParser()
+    parser.add_argument("--config", type=str, default='configs/train_ccsr_stage2.yaml')
+    args = parser.parse_args()
+    
+    config = OmegaConf.load(args.config)
+    pl.seed_everything(config.lightning.seed, workers=True)
+    
+    data_module = instantiate_from_config(config.data)
+    model = instantiate_from_config(OmegaConf.load(config.model.config))
+    # TODO: resume states saved in checkpoint.
+    if config.model.get("resume"):
+        load_state_dict(model, torch.load(config.model.resume, map_location="cpu"), strict=True)
+    
+    callbacks = []
+    for callback_config in config.lightning.callbacks:
+        callbacks.append(instantiate_from_config(callback_config))
+    trainer = pl.Trainer(callbacks=callbacks, **config.lightning.trainer)
+    trainer.fit(model, datamodule=data_module)
+
+
+if __name__ == "__main__":
+    main()
\ No newline at end of file
diff --git a/utils/__pycache__/common.cpython-310.pyc b/utils/__pycache__/common.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..fc7865035ead68cf92c1c2b1e348b86369d726cf
GIT binary patch
literal 2069
zcmZ`)O>Y}T7~Yxvu;a!}NYjs$B3AtXOp#qGkXk7!pwe7WgG6o?mAly)C!6etJ3DSk
zw7xXq*j_jyq*&ro{tWlIa_V2`f%3ewX_Ka6teu^C_nmj1=lycgZnqqa{nyXs-+|-&
zL!0Ac2Agm3lz(9oj$o;iu|CUqpEKI?)a|=?yQ!D?eIK$5!k2*vWJ5G$a~O!GnA!LH
zGomG#3?v`&CC7g8#3?bm@AX@vE&Z)ikYOHr;<T89bXLsQQo$knKrBGkhV3&}x_Hdi
zS-J^YO>s_~$4#dNUvoN3FM-$TaMQRSjYdhnVchTXoz)IAJkM85P-;cOK*_X-1ikfG
z<uRvIhr`F4p1XL;WlWWG00#&@<v;TYpSTrMi<%wqici_Z!^*9^DPxYtdJ9r=%QIfN
zdj!Z^HLUxR!3;1oMp3*SZOD#~o6Av}Xfu;!qeAI4SvT!^ImikzPNhOv9Mjm4I?`Ho
zd_|aQVNh!rH%?2_sy721e&MJED82fow^?Memls;D7sYljE`;ojq$-O%N_%U!Z+vy@
z=IyU*@!h*?KlR2sNlWartjN2g9W#$x2gUl<V5o|00G~Brwxo=iXC7<ew+J6#Yl67Q
zhsj1v1m6r=(*f(i8(0WdIggoYYUIrxd&)HSf-O0FOn3>omkU?Bd(M>q$(}Pf&0jTv
zjucSrUc2|pNXA-<a9i$#4{PbeP#0m6mpaOI5^32r!6;HuR$BfY&oqCKJNJ|-RE^Sw
zCF6-C)*YtK0+H3Lr!&}^^T&+Uq-$Pt_*haPnykq}ramNPuaxQVJ;ub^GWM{j9I#-!
z>pwo0>WNCU3=t)ncD0I<JbaSq%}{S9WjGw?u}+Gd_+-(x49l^Sq1Eci$5CpJwL%P0
z$TCDk!+0~wH)I&)J6WM*w`R=LCER+F)oGEG(RwPyz*;2YC1twjs*gyW*uBGpKDu};
zk590m)So&?ZA5Iqv2&-+_s#=<!Vpt7@E)BK#FVeDAf-J*Y%Lq#UT9i2DpDi$Ok2rE
z<3!1Fu%RN+;r4dJMkhhHk#oB;Kdb_BxJ|q!3)n1z;JsE?ATz#osA(Ywv>Z@q>*9`e
zObp%(DjEuiW%Lw_icv9FY|1A#4-fd%vO+GRLmha+MNzxpi?W_lh8{`1%2n4Zue>fi
z;qQ|rpS-chCaiMHTa~A7iQv62&Q?${bgJev+d1K_nN@WSNEBtvt43YE-FM<Z3<&ia
zo{}O<Rq~*P_QNJ<FBSXU`2&sSH~xrP3{#&%#rpAuYk2pDaZxMAM>NLrq<dS-<SUf=
z=*|~pokD)e*}X#T5S9MMncd~Ih@-T;(>>CE4<b3&LcENHh1jkkRt%+1&SSsAeQ$ws
z3zV9%o61~$hEpftDZ=z69*gU;vjTFIhn<F6!pnGRf^J0KGafji`vJ1jB~_I1QD<3c
z=1x#S>%58swa>{||Kp^j^pfcSHi$GV49LV`()F)G;m8;NqjeJcO_(M}*nrKli~Jnp
zE`QBA|98Q|e7IWRiZnZHr89>vw?ltYilU$>>S`7f%BN8ArlJ%uE(jYBeLOV*S~7Z}
z?P+x@3hvkc6}FtO(0Q_b3>p0&DBI$i%XOyRp|mUKG0mf^&GHNG#pbzSG5EakA0<8T
AG5`Po

literal 0
HcmV?d00001

diff --git a/utils/__pycache__/common.cpython-37.pyc b/utils/__pycache__/common.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..09613837d0d70000377060169d3d45ca649c83e6
GIT binary patch
literal 2060
zcmZ`)O>Y!O5S{M%*!AMIjWGmBq>KV2Ocd=VcMBy1tbE}}K_b#BXc<rUuE(C8S-N`x
zTjL8RM{>z6A(Ho)Uz2-ZbK);>%BvnbfTQ%PtGjD@YU)+Js=e9mc07!)s&C|9k>~wQ
zhm(P`_c^Bi4MKQ=72bpm*n|%_qcbo3fsfTMf=M_Cp<9!YY{|9=U-%*tt!Lq2K{Dw0
zh+p!cZ@=JTL3Ex4gN_V$&d?NRpj#AO=oa0XizR55#R@cCao*i~ZKEZ)IQt?HtKuB2
zFA09vOD?>Ge_q0E>rOhI=H=M>UzB^B3A4N`Hxl1Q+9*;-N)}ZrXf-ocW}M!`2#@EN
zSoJzY?HvKG;B)>HKi~&`&D3?nj(E-I>>$9-uY);bp22<xS_;B*Ui*jSPOxcN?<Ip7
zc+i@r*={<PNr<oNX^|VdkWZ$SGDW^+yUlhusl=?1iUMw1V`)-jR1zxkt5HVH4eMux
z9(0<+5HJ7fsk6BB>Slktn#g`x8M#$eyM2+Gv_DqWY}&s&E58HyY+IK7tP-++D<6$+
ze*Hs#X7WNqHJMapZ@On!@Zn*#wKE*4YBEH~TCl7gJ=h8hSO-6K4mw0S%BpgdkKL7w
zpwgg~Ah<Aof{kFcx7$)H23$F0Pnp44uuI+{8&Qb4@Qr`y&G|2^30q`R1t<2e-F`Ba
znUNyim3#4nM*AQ(Rh*aFq@~GIBYQTQrYfE2B(Uvo<=$<jD&_pMK8)Bv<e5pBT7x4l
zNR3#<(Z2JBod#}$XBiJQLGV}`KXn;9pbfMvM*qc1cpcUNo7w{gw!iiLBdH#%+{hSN
zl10y`G%w@Fx!I1*cCO>mtjtVam4s%J?#futl#Jc2zI>Dx?zK*gAu6e3WHipU({e21
zwA`CiO7<KM0DBi-Jq>J0<T~9dq!>Dj=W#_lmiy{G>=UjC;@eQ@xsU0fe2fjTe~xTY
zx`zD81)Tl4cgwrqIbcXB8wL+A{)h$uz<hH9%nrzL_W>Ka@7j(_jWkke+g0-6ELTzw
z$0`-aE?71_?Ryuz+n42WF%UZ4hM-hLY>BB4ut%;#7gyT>5sVyClAHR@8p37-g$5N5
zCC0&eicQU^sB1Ro2abtHeC~i9qR=mSM}hEB+&(~~yyxKP8`_}u)xA2<_e3DVXJjd)
zfE=;|R{MHd2WpCbclwF96V*(#MEkkxmk8Iwrn(AG6ftMN)s%7X-})iX$?LnAn$ndc
z75CTBg4ls;GBx|%`-A_=9|n$VCk7fq_?GuRwLU7whDgUuo{I0Q&N>AP^*+5p#p!qv
zdHYwWC8FF<qS;$7sw^$^jo$0~Z@EbI;asGII2SLUxL7lkI)#q?jF28X%Kt@5E!i(*
znfU5MysJJUah(J;u|`mRss!RQ%wr!n!1il6CM|Uliw%k#?TI331K`z;49aM#G8>}y
zCfY8af`xW^1ur!5Cu4*6b4^^6#Sk_q0I)D56UPi3T*ZagarvKHr(N&EGzCUe-|^TY
zzsxT90SovlZ=|9d2?M5$WZ6b$kN&X~hf)y%)FK4Rs8ZRsBK}$*SZ#o|Uf2lj8I93(
vIoF(Qbf@|2AiB5;apvD}%0<>Sz4m%TxEsViy#%qs+I-PpYhR4kqSe+vuEX(^

literal 0
HcmV?d00001

diff --git a/utils/__pycache__/degradation.cpython-310.pyc b/utils/__pycache__/degradation.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..17e8adf69b0937b7f6e54c667a57b21242de1a8a
GIT binary patch
literal 20296
zcmeHPdypK*S>KtReea!4(&;VB<ClDE-&>YrDY9%wvK0AocBFHR&o-yM?YX--?Mpqg
z(w*3YBywy8B9j;46;e6~50OeB#Unsb1V{pZgcSc$Q&43pMS}jIfKpKff*^k1*E74b
zFG(>DG2oT9x4Wmir>Fb-e&3_Nxk_JOI)cxu*I&0jP>w{tMTPJufx@HsI=`YtA|fKR
zVq`}BYBL&_^_eLD8Z!p}#%5yp)r;{`VkV(c+h{RaO3kD=Zxqv|%uELPn25g;naPTT
zNaD9oq(mCO{URf>_+2NaM88<~iZ(NV+<;h*+<N3Th>gf?NNK4)lm^8nls5XM&0-5m
zgMMkN*oM+3v0dyCJ1lK!b0$LRye@X)pEjQmw~1Y7vBiomZ56lUpLtQoyq*zvh&xfg
zO*|{^5xbGwF5V~Z6+_7F5XVJM+=tvwaYEcLhLO8XOp8%5hTJakoR|=k$lZ>*J>mi6
zOw{ca4<dJmc)xf^>_hHO@x0hC-i_Q{;)lcm@i20Cix<Q}@d$GFh#B#ycnrDS==&aV
z2)TPvcUU}*+z{%Hh@;5m#EasXcmlcm#7p8S@icPx3++TC_wp6m!Ks|)C7h~Nu=7P^
zH<oHe*FKY%cHXsS%N1GT>H)bhH|thr7i50LDddY*AxfS7PYl0A0f+E)1QIuLNvnej
zHe!btbfMQZ$}b?lrmaTlD!zIhCHzL|H-oD|KDg$$H&8a}Dvj2SrP!LjreprmbKaqR
zu}~{wrlz}SnX=*{D@?~;DCNyk-j(*UIpm(Ln8GeuWyh|R^F_ZtH=0hH_&g>Toc)}m
z5_-;}%Y3ntcXO@PE1lK2r<-q7o5%CQ&X>)Sm3L~=!VO*j9>*=WCd;AQlto@*A@`fd
zyUci$XLCizGD<L0Bp3Jea@C7*7cW|<I9}AT%U)DeTu+<M#k?3#Tm7tfF`kx}6tijt
zW|B2OcR3;lFyOVrV~drNHCC>;)?B4>YOGKZ)>ze&4z17FiGAY_JT$p)uc|(D=*01{
znrj!GF=1i%Mc&0Sj#gK^K3-;j=*jgM*oh<Al^NGIYjOj!1)6^FX`zuN)M8gbBN<T1
zc2LQd2+@nmt?8?ZKD4^FroW`G>P$1B6VL(L4FW6XK{NGev%IELjVcRmExHPdsq6LV
zd=!*o<ac8#N3F6Y16nCr&Z4;>Z82;Xuv3<_%L``SHP2bH(v>t$S+Z;uXC3>TWe(Y8
zA`jlEj?G;LrT811+i#w=-9;u&XE9$z&(JW#W=ml**C4}X(BStzM6>_Oo0@{^FKMH`
zqjv3|9=XC47?ZTKUOX@J<ps-2(&EwDdGSTZ%@<C235gm!ABZs%dMHoNFMIkjzAG#E
zU19Q53DJv}qJe-ZwzvaDhlq1mW<cAe>011HTGOwmb-4?b1!9KaV<3~lm%(Nd88@<|
z0Zbx*7KRW(#6xfqt80K3fDZ@;zk}2wS|@^t=11{hP4sy7@DZ~z&-*kNfS-9ZUvnJd
zXkj~mk~wezD8C16l=40&+-T<}ht1F~({qPgFZlE+%()eF|9q`n*nhIIKqrSA)ImVi
zPF7D2oA^JO!;fh@{2*!{0BPr}UO418QM3j?0=?GCBMS7<P51<=5Xq;BH@?CYl8gH^
zQM3UX3trSNpYfvBauwiID@oE8a)y@;V}xfg&Gexqu*L!_0j9mUnnXKk7*#Ppxd)>#
zsAMyDYKA6<Q0|}wqF1Jc2as_AB0?jP!>LB%XF)#?2yux<05}9L0Fo%6rAZCXv`_=d
zl0!|30ErVbVq0!T8niR?dCtMw*=5r?Wt~Ot)(I)V?0_sB+7cSYms)Bk=oUYC;wgyv
zV4A@!@vNv-lVsT?2eBtPLz1$<^umMzAtvTgyN+t_LE`E1jLcDqARq(~1{zNC<|c7r
zdp4`3i4N{ZVY)z=6nqE}sLXBWkf|F&t3_nSg}gYgy##r%8efgqWAZjd2ttQ$5(S}v
z5Ge2K8i{h;MGZ-g^U>9WYt$2U1LTveC($ofPp)Y@B2;3j@-HBphE&Oi=)-m@uB=q;
zf?1JfzT7M}NI_A(&zQ%`%(2a8B}n99vr;aum?v4dpCqoXfviSV#mY$xE6kHD>-nO+
z0KUcbZuq2b)*`!zhUmkiH*}Z=#!&T}(|FXdZ3K#EsS<iZm)viDA92n+Y|ZDP!0tDv
zDrKwjG|hz_g2ID2^bp-Gc*t<r0}H1Z1AX<3MU<j$Qi56aVW^UKp_og_ZB(13q>qw*
zN|@|Zjfr{Mth}4*@1>-T3pO->XEvap=g^GhWpaH#IKtGjdRn8L9L1Fa%_I0^kfATL
z#_Pz`V{NPvTTQGc>hYVfMzV!9zz$1>Vu;lgSR*bBqzQb9QBw6-Jq5`dQ<ArC_|i9B
zPg9F}n%M{?uu&}UK)iQlBSk4i7Dd66H)Im;?#YwSLhX?I&1a#Tlq~3R=HrfQL-Pf-
znMc5VPlxOoQ{AO)z{4T&?wmZSXygAPOU|{JD@q})2J1F#^5uuki9mr+wBM%vyq(ya
zDWF6DeltugZwFuG(p@>(#3f1ksk{@37pqpzT9R3th1umyH~#hvT21D1IZS;=C}9rY
zO}TDtzN^XRA;NF4{Au)a$gE0k3|Ky^-KUtI)Zh-LCmrnj%=9;Ad0}iJmAgCBf1;J?
zt7OIJYa>Pbl+_|JTLk5VIc5$`+`s(8L$tiPT!7>AmbA(RYrnbZx>aZY*cfy+*BNEw
z#9A&alG!p^sFcP|*DSz8rR<DNj_-MJa^INkj6jA;Ya~AsOo16gjD)rRcQC;>D*iuV
z0_(0f!2~ZY;W?|y3VdR63=G!<202c}Hc`I5$>Tn`eGIJ>TW{)MYqDrs)LfF|B-@#-
zw?g|Qr^#kmx2JR`{R9-+^V)~tG!e$O2&uK(BBvtGX4J>1-nWxOS78f5B^Gfg%F}h7
zd%<$zUK8k*;L)O7ANA`WMtw3gTBsLBOZfK)^+9d2Y}Dhc31O@zp<GAmaVXVE+&6VG
zwwi*Xo~~)~Jbr1`hJT;axKAdy&ypU#T2Cz*B6~3kU`f@}gAt68VrxIeS_MkbD*EzQ
zu{_72Bv(q%q7|URPuN(0NV!|n(VbgV(%}JEo@EC~o3f4zE5QX;+SyMD5DF^6TW?-_
zZ{8NJ&RVr-LBoD=eDuL#bMMOy1(;3kU=}xM7tGA$=Lj_%fkw4BTD6DC5T#6W>~h71
zG0z>;1X`4Rb|+Vhb^#jM8OvAROzzX_(ZHPFwRhN@7@gev@@ShLe~W^in<>qlc=y_!
zPqNRBT_9$u=D1LCAY06e5N5ennzQ66tMJzbpO+Q5|I{k^n>c!;wc6jp(QBJj<6x=O
zlaF53Sk6yf+n_G<6RKszHP0Y7Ov7h_#R6G+iCWp7hlibYcDQfJgQjc-rBehhR)0o8
z(VvOe-#P^^6WR%$4E*}!VwqhIGAmv(=))R_7vnDNAhk`+Z??fJ4ISi5SXPJBputUB
z;pI(gTk&r*zJ_jS8}wv!tF{xhY>2!U&02I4GJeR6C-AyD^;UovjFz_$ypmnPD+TaM
ztR?|o02y*Lu0=%ZR7Bc-YlbayFRrG`ehZv;FpN<h10KO0pbE21UyQG20DT!i-e>Vk
zkQehI@1uacYzXo~SL+!-ULPPYThG+9fV@l{o(-5^8HPNR0D1lS40NjFH-Nm5TG@0;
zDFy2tlml$u0Od&ERIDCC#qm)f)<gdbpiu4w&nb`GgUB6`@1pWUNWOzFK5slC_u*$s
z?nf%$jl>5l`7l)+q=bwN`3NO<A-UYA?4C!d;xS6zgJdS%TqsG5uB@NKlzW_#Ba}=~
z!Y0sB%C+f8_jJHdh;AED^f&0~kVANIOD7cdM+X5%gROw{6dLygBzX*1!9tOnu#m5{
z`GV{nt_g9?7>+l5?O<%O$rMe32yxEDu=zl<Nl0+Ly;O+ad?TL#z1*B<NP}<kjC_)c
zZREbSgWP=@TMrR=a|+iRB3GABWB4nZkmU4*J^cCBBUk8rB&YLyuaNjHkgq)WkU>FZ
zHYolI8AufzlF+%U(6+H^O~!@X>f*G%j?;HI_L2hXJ>(d>@9HTL!}Xz->u?5!t|xGP
zALL-rKkko^XpE3ViznJz&=}1YDYSSI-uPgQgh=x&nk_PDvDDTgG)69)ztEMnOZIZ(
zpetbK)}R0-AZLqk6E-P0Z*sW|rz4#lsy3#(Qnj2`z1FE!s}<>*FWw7-<KAKOUOo<>
z44v!HZwTF}=<oh4H*aY0<acXaQbRQP$Dbk0(7EOe3l)h|*=nUM=xE!YAT`O)*+tu3
zF`;^zR=%*<a*xo1H0u4~!$;qN8?Z%XZr=NU)Bx!UFmHF!4Bn35(A23+gEt$3awf)G
zdt_!^h$x5>X4Z!*xY0}|L@84<eL;yY=3*?ky+mQLVizn&k~S_$0y0SC5L@N%Dv@!c
zK;+R?tY>JE7nA}t!-c9U@Z1~yiT=C#HV1u>V{V&Ow{{*UfO|jYG_W&?BmXQ8{f*m_
zf9L16C;whYZTMgO-;6EEe-77B9<}|+YdWkB^p7RK>*sWNg4Q^$OzWKP#f5#w7M7R5
z0d&4-i;fT`IzWu#>wFr?1)L+ddsYY>UyHn^Avnwsu~(88v<o<EhMFI}06z>46yq=o
z5NEI>Qjem)zLi5-nz?!X0IpyZL?N)^4p6+HoigMx`c<u08-Y%BV;+VCr3h`oJvee9
zk$)bmHhCDvYPn#}+eI9C^z?!Xr+Pyl=-!9+j8BYBj);+hEh&&n7FR}`l@bn8q+J*(
zR?d!8^NuqTzD+AO7HV!~9_PS*#FXim=cNTtYq_vOr{D-0LKKB#4pnVBS}5h0n~^G=
z&|)rDq<yYZb`h#Y-V($LxbPkn*@}mP;hXWW;hTry)x1D(nje;9q7=X#Ll&+G%dv4d
z+BpWD&#I6gLbRsufCYa`r~!mWggojSi9lQ;5SEolDVTuC1skaEh)}u_5|juiTtmKa
z3PGVGF;`zQ2EWIiJ0h_0qEmdNu0k+kk|yQpOA{&<!*EVZ*VC&uh+SLuw3W*sdbJ~p
zsaf;j$eZMWV!kvd@&`XbJV+;!*~o4!c|EJ`){N`M^>ev^B4$AaL<Bz5jvgqIuhTh*
zj2hvJf!@Fy;vO$MFDoSqT_GpCd;wK#EO=3D3fjC(6?7ar6IXkt+z?4B&|WFs=p|~!
z)7;Wy7O&E*H>25L+92G}jql!P^uzOzC0~Rl3%GQJcI!wBlhaITnyl>zWa_%SlbD*=
zb_u%(D;sS4VYonF$4+tJP!#9=VA-Wu9WGGlB>GYuF{zwWEZN$>wI8!=JU@o%9u1<;
z8V8xI^>#h4w@VA=&{VBdT`@W84=%l9<!a3}aj?07zyNi0H?%lxo<;Z?f_{Lm3l`jd
z@SnXrY%(+MnQZb0OX(()PLapX<S<Ti$PZWKcoj9w<leoxkOyvJ+@E8|Nn6_#)pc+g
zJs3FuMsxM2+E6iw`&y>0j8)QIm4*67<xz5@<(F;e3So3Ep$IPF_fq06Mq;MS3b&U*
zFcM0$i2ZcDxU{MkVmb}T+|mMvN+R|sP2sEva>a7UZkFg4C3F!8yWq!ysm7lp@}mQN
zBXc(qTy*Dk5M31&1)6d2NgzXIBDbrHNGKQu!6gPlOJc~ywnX56*X3>XsDJhz7l4OL
zQMi~8n6efX6l@Z&$4?NU4&ks(hYJa$hww&K8<mr&3Hmfz<4pZNOz|=JIp3i+aK_E4
z&BE?5$z8clIdUH(K#6wLht@*TMhvd3)XD;+SE37LN*pAhE~<BAg=&b9ylAE7wi|jI
z@<Ox`(wDJIR|&_mnIzC_yJp0U|A6;2YW(K_;n-PRYC$l<v<8BG8-;r0Qe-XilCeq=
zPO;Uv3)FIT7xA&@b!U5>BVsQ>4Os)O(Z4kVv9*YA3fjosIQW4IvYIHb<M>)aIn_>g
zW0ydl=*idX2|4Y?mJ;fYWPK4%0yC5N!V3J1{<=cUxQdWMEH1`^y%@(mlg*wzdZaM^
zX;_o6Gq)nb7amh6QeY5uoRj$5om~mq9C~(Tx*`jUq$f~-clWq%2GnVUfLA22^n?n<
zodb3bn*^c(g7yYbh&%-EGS--NvoTT#NyCUa+8Z((U2lk%J*Rdvz$C&|TFfkRkpD43
zIyrP->kRAwNpy}lhXeTW(l&HR3T`Y|UW^}&qBvd8IAdF95k^kSs*W?4Xw)B}q^|lm
z>7Pj6)8;t7!P6E!?X1_2#o`wQ6j*v`23W4?in7~#pzLhb{U!nJPTXQ3(~on8euV!G
z=-?0iK5am|GivCA#3iqx!7Tx=-U7S;-MVp8;Ei?#UZWTA$`wq8msv?LNktz{XrA(e
zNW549%2QRk$hV*g0|-HitWiQxBF|9L<}=#V2_;yqf52+VWorK8NG|JY2Bhb-U<ipM
z`m$Ief3JU%0X`3ojmg|lwQCSPIPKUZfG=f;9+)R|gbPp5c?42lj!mhX&=XD{0b~+5
zg?z?@pqFEemtICW3s}kUQY?De-eer*CvSVoT_nvbrP&_zVL<!6N`Od&5eHFZHTiM6
zpa0;bsrMucd4Xp()nlt@9Q7=`8?;fEec1m=y7VzhZegQ>qqNTb>Z92+^aJ?2K>w$4
zhu+&pn@Sr5Fa~@>RgL)StA3FJTBE#?*iMF0-dI2>Z)vw4>Sc*jI8a)PV1whK4L1A@
zK1UDCbevEmY|!Q~kTj7&{wTF#GN(YpcB;MO#`_5h-bRZ;cQL@<#`S)@0)UOZz8zbO
z&HZ^?xbcR{PoW8@kXrII<wqz-+b4gDk~Rw4>CdQhdoY<hFd6yN)aw^0VR9nXGx!jx
zH)QeWP=Jy~fe6?PYz>+05trXkVFe9K3%MF7brBbF#+M*V*R=BpLr3sr6oHdb1Wv*+
zum)EGHY<7wr-w_KOIg|uvK-w$L?uJXLrfLD-Ga@98mY@EwCcwxBL00?S1D@N&4UO8
zJfI?hbHf<qFw{g?kjebl0TT^tqzdI1@BOF~`G%t35l^*%*H~(E975$3>{GTo!s`84
zOJS~9nHwwR?eds&N?oQGhwz>e^lQ8^bk|dR-Zk>n<itMZuV{57yyYfONMb|e-#}24
z=|P{SEnxal!shdM!3)7*&FB$o<1bY+MBxa`;hwKpd|z!OlI9^XL!dq-$<UDVNOA)R
zv_CWHB@1WVLzQAhdi{r<dFq+tv+p^2eDdh=$6jayI4@eX%3d5n3Gm%`M!~X+o>8nU
zOiG)Eq=P^&;nR!a^deW4t+Qm7;lPm80@zlx^sEa*mB8943C6Oljr}$^kli0vX$<3F
zJcm)BzhB!9c|V9(f|8K*cO#XTag*LCa2i)({>*^JAA|YRjlAC>y#Kaw-bQ$|S3q6b
znKqgVgyyf(rKY@OY2Q=3Sx<1>LJ?0AMX<2Ub`_SN$M7w}vJVx6&4ky^2S5xLA{};8
z(*_#+9&Df`Hvwlako=<l5zUQVf)iN4mv<=zeAh$_M!T^i@(~SAVh!k<cqQqk3AGR7
zEfhqNBvAv~MwKpTmooH5iL#V@t1E>uvlp~_iV(k^f-Qw}HgXPA%*Q22071d^WK(7M
z?vXBckGvMa-P0X+&$Qe<((~?-w!4$0$-+MSLOt8C&oaJ!mTj>(|Bx-trVXjpv%cKP
z=HCd*9lR}*cVV{S?5u@HI}N&?(xPEXb@PpNT2MD|kA-AKPwNiuI0TH5zgXQ@y@Ka^
z;^ts>Dk-AU#%m}OWC{jX!V;##lpG(K@PpEB@gC!RfXBiIZa*$e%bFFkLw}%FD7{OL
zAiG&OLg2Zar5J3yKkF!co2;Uiw}Zjo{E`YOSyw60WC5Oqmp9RSz~uJ2K;f+T8v)R7
zLg_rBr$~SLxPBV%gJ0IBE=T1W`YO%u7m<SjtdEX+nvK~cp<*Z_l%){+m(vI6_zQvI
z)q_8RDO{y#Oh(R)beBue3B&I1pl8Q}Br(bfe6N~F_(8I(n7(op(_4s2#S5m*_By6c
zcqM;{5>h5%#e9kKzeY)`+?2nKx_U5QGr}UjV-|F@Ps=I4K*<*=`3*{b6G=`}F095%
z@Q|-!NclUs+%Z=0MDllOgfCO_dzAb>C0!=>8jbV^xR~o#Cc-K;Wb5F^DE9$MK1j)3
zlo*sWG<CSy_%-7+6pLAxoi_59(ZY)>k8Z9{O`c^6OOg*!{}D>gQ}PNWA4W2hBK?C;
z?$tV;p)0Ra(ojcm{7Pzz{4GkDD7N+7{W(0AL)19{#l&|-z%bdaZ-$bxS=#{hWI)?)
zG)o3lmwwbiZTTZ~yQ$jpZ9rs)+G64WD+HIUS>rVAO*p>Z0m`=zP<B_Lh)UWG*I%c)
zRwnZXJ5(fq6TMQSw50y7TGCgD1fxiTPYfCQ9>Q0Cl#+``I=wig*fqU4BrW~J|Bq3h
zCLgv<^0sz+zeT6}B4NZQXn;RO(wj9dP?Jv~!Mvzt4-NBA=n~Jzq1=uAXF<$W$0~lA
zCdD@Y*Dy-&#rhe#$Il{x`;Kw!=ct0*s_oto-qAm%3vH`IhjuOAt(Minc7B!S@)_#+
zbtJd8K-99Ge!kk}<~u@;kM{e1qp=2mhMreu=;o+ue28L5RuR=W6jGvm9&~bwgDIS9
z!~0;I$63Gvy>=hH6vI2#tfk^CA&S4|0Z$@&DE|lIr195Wpg}{=g~W${B9>q9LGX!Z
zkD4T!5yeEFeH?|l_I$p8_quwjgm2-Q>FF9`L=Yv_2yJMej^AL^REijm$?<BL>DNLd
z8|V6l@<&SGT&P7zPutg9QEPpm#+d=j05d;N@4I4Ny-tL0e(r^(R&Kp>#^_RYL7fjI
zLoCMP`m<=|4IX>?`Qyi?AD?~V*~gF0zVErmo;o)D0%=u<^T6xs>?EL~u&9W`eiEx*
z!e5Y?tZFvPFPyTlo{6IhN?AK&Y$UG4WM)H~|JNWuhkU*LS(3Dy;S7*}fg-e|KgZ7%
zfuw~<3FZQ5bM8i=tCJh#31=O`3eRij`A-Vv>1g*jGbY{|!bE6!fACb;5_$e2+YT?o
zQX%ElFTuyvK2v|?TPP50k2fK2OtP(H+S0vY*9`3&XA%#N+ip2F;H{@zP#SB_<eyT3
z>O>fl2samHBAn^-m%^7?-=OPlXuLfHh;6j@k3@S0`fTp&*W_Q)xcbyIRA^!zsE<T@
zbo4>L(?^c@HR`IW<vfl@bomo#6I}UZkmoC;s?#X+t45(*&{ox|k}f~lxS}!nV=muB
zRc=@^ry3N6jdyz}fF!gwA4C)K&e^CXJja0y<takUb@)$o%nb3N-x(YX;xFB}SQ3Zc
zPjW_sh#nm5`3gC$hx$tgyC^qA$^Dd!QbIbie1MYoQbIz4;|^6s&<Pa1xI=I2OAggh
z2kms&tqu?9_==Bm<fkde-hB4Kvxn*PRL5eK6dI)ye2sD}{77aoH1JM^wwX7a75wA@
z{C$mD(K<*Q=qw^3bxHlBgNCe@OeP1C8<GR*4e6cf+x2^NE&cEM1k!8zK29H`z{9K`
ooS6RC{)6dgdRKa5dUM)LZ%L=p*>p0!F1-W!ag3nK>-#tU7lU03f&c&j

literal 0
HcmV?d00001

diff --git a/utils/__pycache__/degradation.cpython-37.pyc b/utils/__pycache__/degradation.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..d9b09efa746d76730f22a7783e1aca807c9c05cb
GIT binary patch
literal 22425
zcmeHPTWlQHd7hb_y>NF)>q6^pdF=SYm3b3oIkw_RmM>DAD6}HWaZF#x;hy1g$i4Z@
ztVA-yHm#gkX`|HfLobO7h;0faKvAIRLmvB3^rc1nQ1ofg0!CA$&_mGzMPKsLe*b@F
zXJ?m7Qg(eqqb29;xzCw%{{K7w|DS*M>A}IghR-+tVz2Y9x~Bb{675eGk*9EZKh!l%
zXhN@QGwNHP(K&6*r1;yMG5I?^lg77E&D63pS)J;ps=3<0%mByDYQ9#ODIlH}nYXl=
zqR5IIz6ZsC$m4rR6hsl<cZq2+B<^}kpV=i|5WB=~#CD5|VvpF1*dFmUF)Z#zY_GT^
z_KE$74T~4W0dWwqyTu`KSR8irrF{jB!UaPd!Jod6759jvsIlKkEgcZ|;?KHaU@Vu#
zed2!P9~3W%2gO5(9THy`4~r4R4vUvXNgPA$i1>zhM2sSKk9b9liwVSziWzZSOd)o!
zcvYMbk0NG?*ThNj7-IK{*Tv)F3B>LfZ-`Uk3y6JA%!(()7ZH0v%!$+D3}O!oTRbJs
zBKDA&7f*|45PMjZ#k1l$#72Y==fru$O2QG(i!UQ~Oe}~C;;V=~BJ>xvQsow{*mOw`
zvtH9FyLJ`Py|q@=cdyvewS8x{-jFrU-X$yZvwmZ?BJEYLY*(FfiW=jS#zias3@%R~
z@U@$IfK}g%6|ES;2y}{95Z}<(Q&ft}2#~^eioOddZO7YXQGF9>Gf-hFFqhIB#)g5h
zr`Ez}>}t7H#V9R*(XnL1M^spzTdCPr&Gx0cVvYD$8<ucuPTg}Gb-Nnnm&WsX3!k&H
z;+^6crO<K~O%|#R+b?xzuWru9J$*aQwl3Miwd+>RvAvdba6>=3$0f&a$$H{8bzw_P
z;wkG=pAnDqXl@ypMhWK8N}158H^VeH2~*{U7p6S79;QUY5B1BXbeQI0tFP5C&BF?F
zVph$-Os*BheyYh`=<t{4Cl(tuXQJNlo%u#%c|zE}JyDU3R&(OTR{erouap;^`b4=Q
zoQY@Mg@vb|zckVE-KsYs94tPnV<yL&tKlHebkzIQZuISC5NL&?`lK%RAXcU)pidWx
zEFlAn3KA)RNDhHW_G?5gDz;&)EAr3-eZzRoST~qtKqepq)EfpgEP!N!R42V*P>xCq
zeIvCFf(eWuwU7d#nD#>$#(Agi$Tq3e9B<L8NLP$nWh|5<@z9p-TWgMN^rehtN7kL{
ztmm#d)`(jt>fm+Sw6-fClxTfRr>v{4zsR)dE!s`AOms79btR@sF%_<W1V4Beo%X9A
zY6z;mq=l-`ifP(^p1H*Yn3lAvVaAqrz2by9nmn4jFth0Sc6m9>O61`AK#PUMLxqOD
z5*jPGR#)-8%G9R<qLpxfIs%-y;yxrj-c3L{yY!>_urBXGUYWRnJ|?aKT#PVNi1^x)
z4g}GFCyX0{NQN*X9q7OlAP&d`-@{ZR6%ZMu>~TC!2Qn@`d(LVs@aoLB(at(=w>*#7
zShybWV;*!5THgXIYIa2X+pX8ssFhe-dfH^|N<@{ynqRd}Ewt+8Q*-gO%#Fr0LFm*=
zOkWwb@IO_;m*sl=AZqr3V{6V<;1g9*6b2LmmA2|*3gyvEL}aO8D58Y-|B~sVl!+*z
z>H<kCValyv2~*BW6NuERNm{^?85WbM5SmOWgQ)39&jAAgo5PG6L@z;@O|dX_0=+~q
zQq=F)<wHn!(E(8?)4`*N_<#+e(;ma2PWw&8xCYShi9P@^1Sf!w6u_iI2`_X}0&O18
zbf^JDP3(sy*@-M@Rp|4Qhk0}BmbdI&MeO#8s14F>lJICbsF!G)Qma6>`1A><EEd|s
zY>yJpid;2F-mK(atVzj~BpEQRFi~LF5(~(^i*i4QAT;C^`7otuLrOq`iHcLaxH%MP
z#Y1{g&+D>;_~kP1Er<~iP@L<{BNCWGU(U&WK6Zy|`fEV{^~`!ENXt`-2!sJaBn2V?
z5l}o5EVA{Yj~v=3uBFzqz8Pc#6Qq+1a%h(havS<#jZ(aa{5MD@-b3sJZMa^;myM=d
zwi?p1>z!mw28!AvhJ3!xoZ6{Y!#+D|HR{z>YmRsFIpXFPNNQYVtj?iZVa@Tz&V_CT
z{EG5^xTJ2@wNnxO(1&}E1(q0FDEq^SH>%qn;>5Ew2{j=!p0Ykhe6yZ)7HmkWr>yBl
z-HD&3GqEF(ZZL)|HtH@MWEk861E(4LB3X>LB}LgJ@v@M^5G833D-Fm4lv|`=kb)r!
znB)iIfra|4yr1$Pq@afb_Qaqw+oqqFP|c%i+U}e_q?@{B5LXQ8auj)G8V7v}(Zy`>
zeI$Z(4_l<yv+LO)vjbb?y4V7&uw*J`SRVjeWQ2(@i;I|KAV>!T*lp8lw++l_$IS<M
zsuARwg^&UZrEL$}d0!S%R8r;br~TCNKEx|KH+K;NhCF3mgcMS9AhB7`dA<t~7nEk5
z1Lu7;p|n`)E<GJSo7mF5xj97_{~y_LzDqArqGz)`Z=)6$f6+SL78(@Y_eeaSCDvvN
z*d%Fx7zU2znNq$lCm#jd$a^WU5QOPw<EkT>wUu;J=*QooNi)h^E{{>$M<`$pe}H2B
z*!*aR%@f2Qv;3QA=aB-X?d`C9Q9njZKY`pXrYE`UV`lp8Szeg?NzCrg^zU{veUsF6
zyERsImz}QNvTKt(ZcSJt#~)dFV}z!*RBB`Rf+L-J**Rq``hL?pH8BBM&G*Jx4RKb=
zi=?28mm9T-H(L(ip;7lHrY283Huc1W>y2Uml+Ktv)*b>gh8Rmq`gbrPQY3zbFhR+z
zJK#Z>moSetWdnvUIR=K?2^8cwRid5$wlQIM2NNO^eFx1I4BWj529BdF7Kv%^Cpmy&
z;1DELGN<f=2E1%|!_bjZ*Yw|nIYpQUH4=djYRj58iu^R?V~1z`s8b4^3F5QJK&bvn
zU~oP(EN(TAR$1;X#rdfye+~J$L~n^!=ne7$-y_imxw(2S$gF3Dxt@dYtpynf;5poP
z;6{3V0789!dEgyg<`AQCo6(*A6?ZDM@3aJqV5c+~SW1iHjTE40Ajl7E=!In#$S5Sr
zBLz8R(Ecw>&?Sh}jT&Tb1-~#M#<P!o@%AKo?>0qx*a=o<*?7_;+M{5#UBDtfn=k=e
z?M!gvhnMv`w}ln7)vP*@$6uWse{9q``FboavqIh;#WwZYBeVE95*5cFgDs9X-BD6m
zDbne?-f*Gva|1PiF4cgo%+0D>h75PbiNrmN+jQGKFy_yn9JP**Pn~>yyhrN4P0@g>
zDLI}r)V2G+!uC8ilvuTv=R+vMZe%rtu<EVayd%e1;9nblSvFuARI?N<;`q7l>}Uzc
z*X~x8+f${Ue7vMH0!**%QN{d(Y8p`%nq;ggWy7Jfp65u~v!0W(AtOr*CJ$s9W?OZ4
z0fu;%_+jZKYn#$Iltj`t#YQ7+C!>*tL)}9P3yGEC!NL4bhMC!wc4RfowcD`#5~jIH
zueNefjc+#QomhhTb4;^GGLUxnuzo<_r{@sT_W{GyO=C|gkJJ(5LZUf?TCqe!Y7;5{
z3}u%-3Mhlh^ASQBDMSR9-B30FD9f(r0A&CiD&Nq=z_KR)A*#(FMFz?Be0`6~!?6fM
z8s#zk5%>YNFxvc$%z6PpSYQBr3NZq}v;x3)bngftu$X|rM0rpE1P%fMi$NhM0s;#`
z0n%G0C@=&f1qd9n3y{h#Z3BS`p|oSpQX<+N)B}uegL)*fD%MY+<m5OI?D0F$NC#f=
z{x;rL$Ifhzk&-O_rK}{=u(G|7=~telfV3g`7zJ#A8B_||=PBcH3Z6g^<~wU5Pf;$b
zX<wk&lN5ZB0v6a<S36C~9=ptgn~*NSA~EXyGg^8h2(<luP%ky44;y=oVPM~(HujxG
z-8<OFg5M{FecBG#7YW~yL~;l25^y(x)2v9$o0#gLS%<w6kav94dbCp|A#wZ)%Yq*P
z;E~?96X40GsM3c-x&xb#E@Ft?5XN=}HzNi&Lq3gR+`1b<$*7#*&(F`?qVqeX5#say
znOh|6mFY}}KBUOT2fcrfgz^(0PXM7GWEGB-S~>2+eI$VL2Lp6YKo*0<l&`-R42U#J
zFLaf{WX+|&hSDrbt2p&(x6MSYvuOPW%5y0HUQao-r}EP%A3*t^LRxFL&x$<Xoyw1*
zynyn*=_!xevk|;#U+;@PHFqUGdTk?4w{MCPzGka1Vt1I#wm4mfNu18IRUOM;Z8~1J
z4&}9)&4%=?S09FQ`|zmsFdtM=giagjJF%TnGFtOlt{!WPWEGB!R2Ll<k46Y1$$V#o
zWvDWa*KE`UxmZL4q$>8jTXp?a3u=w!*yY8pdn6tt&X2lJx?r@8)LrUX=iZ;FvfCE}
zKkH&2I9kIozSp>1GCFuMljfy8H*;5lD2Nhfb|*8q(o7*iDbq89?GzW5(vl>#Fk4=1
zxMjy<O;3`@!_ZKIOJB6AM8<J~$fKVeH&dWVt|;MJC7LS1V~^Vt{r9!H4@c0(+%~Ij
z?L9}Z_7ujX9oYqa)PVCzCp3_5|K&NqouY4*L%J=sl3zj}VMZBjN=BFw?iE)!VHVC0
zcGVSK{&IAHox<h)E`sauVd%BiL1{DI*520Pg=mWOTe<7{b@(|zSxjArZ4U?98K{w|
zVePOMq|n|tzz$TM*n;sSO3;fLq*(O>=v~*BP5CZ;tJ>>ymX~H2H5Z_c1_<Fug?q5a
zOxAuGb2IfU%pdi#wcu89?6##PUU<!I?eHfbKQVcHVroo`m0d|*d$PJZ=B?Inm?z!x
zShaC=tZ94RSn@XA2v~0UjRp81MDBT3lwObyto`-!DtTbQJq_MFo;A{R>3FMVuXOyP
zHUo&Y*pTj8qwd2mknC^pQSo7qs<K@Ol1^k8f>o&Fh|#nKydWbdFbk<Rs4+01L^z&{
z!}-m9kbi)3FeBXhVAiPE0|vW?Mh8A*#`1ik9o`}ll2L*w32SQ{)+pQyYcJ7OGKY?8
z$SZ0X<-W0G4u8Owhcy@wQqycTS6)GBNrMWFrQ^zIWqNN)KQx*yctBqX^;O~pwOER)
zS+&|tZE{am?b^JsPrpk{NM2r=eokL2wMk?aWI<FB@#^@KRWiVPrx8)@Zc!WfCB>>?
zSX_{e8aYyto++nM%$i}C!cqZ`zD5~zNI#QNE6G4HlcNN!AnSbq!?>R0f$ra8MBkxt
z52HIRKa7+4BBUQWq|3o?<$x}~j{IA+TIXWkr}1_>uRu&9AqO692CUiI5}a9t@oiY!
zzmHX7KeiNX*YKKM0;j^y$-H4KW#9+RvGr6P#-L>D?pDml8G8c5Jl}SJj8E8EChzM+
z;?^qG$aJgLT(#Km6qe)X>&=#L;l#ZHZxwZ>JhC`yU4;iFJkS8F6$jR3m}y@hwU|Rs
zOm&!sH|P%MPLpBSVkcs2ByyN#2P)Og)X9^jgrWENWHiQ2N5DNpROMhjdN9Djc4Li(
z8cP?%l3l}Q`!5r?5>?-CJVn-8d&Tu`5pcsELF*W1qgHCCOVbWhXZbuVz%vf1S$H3N
zVMaPl2VTV{w%b~Tops>|mnUaxxL9L4WIwCJezo7i4Z9LKCaTImCbA>K(h8>jfPTn0
zqDvd8G6~o8$>O3oTTulf<QoNFClj7dX~fd|H5~C6^6?-gAE&JbR!2Yr+b!WIwUH8Z
zV3Mhipw5M;90Ui?+Dv^IZRQg7xn_`I7ms5Y*jX6S@6Z_dVMiK^DT|5i*8gCg+8%VY
znh~6qt1kS%;h$X>Ai6iGV3vY10?-xZdvcC)h=jsaqviKnulCqN^pMcch=kb1pil%}
z<zSF9;d@J$q;bWFMPwEu)-R9<w42(7_L{j)j&|vF$RUC58$O(}uNmH>0ZAOF0XJ;m
z82xRSaOj4cUAvCFfHOr%BI{YGOytx}sHW-}ej27~xN@uVAS-|3r<bzoj-+T2E&?NS
zxS;C2XMA5FWJWpaslK2RW^m73r{xwc2GIY%Kxu<2c>r$ZFu%iJ7>uFL3A6T*zQk;e
zTwJ}}kmW@Z7wDye{_hengkHvL2Mr1Id{Oy-&jT$-ErQK9W^N6c*y!Lf&gfzVR@_qp
zJChhWzBNe3O}E9zEr+%fM3N+;gqar+!(N{jB!ENwb*@0@k7*n(H)O{R^De|m^06&D
zVVWP09EttVyy7}n@%jQyEpOTKRqA<>0+)KuCun~T8KFMUt~Q~*80uHUAxs#*oT4By
z%rmTVPG4+2v<0@#HvPXN-Z+9=XoX$a&G+HT8;BVucK2PW`|%BNpQNuH0oUjPT!3w0
z?f|%{K7ea(3AlG?lq8O=X*4E3wV%&p&Ds!4>nH07!gLwpWK+^U*@dukRD)nvIu!5@
zy+E;^!-2au!z-xh-?u^7E9m8?h8ok$h{tG4n>Jm%agzlvdeZ@j*!JBZ{Y>R<Xn}!4
z#$y6VT*nb3^f^e7q|aTWHy+Z{>ITH2m(Rfgn_!&f6m@oDBMFh1c9gK7Y$z;%g##>>
zuo})g%^bDsdFG=e=`?DyTX2yHoj^$_%sv^gj9f!E(L<|c!7C=8HUmv>Lm)I>HD^o%
zN%B~sqE!lRgFM=&<eRBvkY<TDVJHken{a^M`;&!|=a%qt!%b|YP!hCAaWHO>hkJ?%
z1?0k;8WRpSH71#UE!4YG%XoEWLjwV25(H!_0)jdeqp)J0tl}G!Xn;{YQDK>8g?K9>
zE!ml?Es5#A?THD@bg)BDraq4Iv!X74jb@0Lg+5taG2<j8-j5l8x1%2#A5WBQQS3w0
z&yi?$Hq#F#=Dux6=+|-Ef5U|jI4z}OEX>6M-H(w88Gnzmvy&1Ec+q@*w~jRePz8ju
zwnm`&8I(aF+t9D!bs9K>r{D;lf+IMLOdBvufzDDl^L}cnaI;91M5YFR5N}vOOMs6k
zy_W~NgbrfJ_fTsHUJdyBwxL4g+_eDu6|k9$EW8-f5aytl0L*gsUjt8LCAJBj$%E@R
zZ}PEzxXC}Xf*0^w^XwApmEC36A7ia>qAM|9ZOl*9Y_~q)EvsUB-4QQHLM_94j-R`5
z;`3t{rj9?MY+v0*sgGP!PKbSw8@G}6NMof>M}@SaHy*5)@n$NVT|2&QR41CMjsw*>
zSW>opP3+@Tt@}YPvF@~d1tEL^zCmBR@Y>LosW4Z*;y=@<He@*T%nKJ@xHS9p`Abvh
zFP(j*2bsfE)v1RWI7-197@B3rt%hc`QJIo6=q<8CM^qyjqh3dTn5{ckNvDSmiTA-Y
zTP0YXf<%fLlDy%crapRhz<u%77j>%y&q-p)Vbs@#Oo)t!AUqD^mBt)Ys|R%XJ>>6*
zS=hcmRxN!7oOBjidOs$*L*D;a%TYh_CJ_t*7_EK}-L?hJ-=d<9Ffm6-rn-It=Wf~x
zx~T3e@itK0BuU&rUorZVe;uJe?FQakAcX7MCUw4}s>A+s+p79M`dJ)JV6V7gd{_6=
zH*sVkU|zmCfE`2^Y1ov^!`gRs9BJsl_v~9aKTrF@vv_9@o^v_mKm}K+>-x<Cy?@tb
z92h{K#p`-7Kzm9s0F@ta9MDk)`MBN09wX>=BvSnge0QzS-L<zh+&#bP?uD+qYg^u3
z>$yAaWyK&1H|5uYVr)n#M23W7(li3Uy?~7)9dm~s6r<g(XuqG_-SB>)?Zd)=Gu<v8
z-E4c<a!-o2{(gqr&C32Zw%f#xw51^ewqiKV(g}yUuX<TIG850kLZbF66~-^_9Ve?r
zyCk`fZL)eyjvbF2b8qt=lYFquTO4fCoVYAoPQqmJsTxe^XExwyJPTVL%<8lB<_dma
z!0SNAJn<|@f<NGG_KYAD@yLi$ff4NWWAr--WShNCPW|}F6i88tZ~;D}B&Gepc+-UZ
zq)(Sp@-4Jg66v=QgRRCGpA2;uqsc*#QMQu!bmo`&C+S!guhTS7f0N*ahLO?M#`^Dj
zkVTW{GwqgL4?(^jWF_(}1pE*O)-iN-f<f<VD<QHyWOcrjQi%B-G?8y3fI9L!6#s1s
zy7y%HF7krTSXrRx8UZ<&eB{%p#`6xkhVVjSgc=r|9Px<0(-~3MU_O!Qc)N4+(vVWt
zmS|us163(jqks${%Is!REY?_HdFG#e$U{(?_1T^&ucJnoQO6slLG{GEf6Y<T1_h5$
z(4^o^3M7J=0g|WKgGH?!?`$3w$Kn;%mgnyWluPp$9^8`J=IOzR8?+&^(y@gOLRQf@
zWRS&fpS}k|*Df=e%0Lv`MHX&4b^p}GuwV2O!z>&jC*Ys;5u9g#2%GaA;Q!13e+SnI
z(&an_ghCJri4wOk5_}+$aCnnQ2tcNnFO_|1Xp>a<d$@5d6_WZz*E6`hkJ;=<LU70I
zNYX{*#KZQq4vus~dw5T;ji^ih{WE&fJbKnT_QpRzYDXsj6zxMazAdzY-=`v;kB&;P
zr8dy^*xW<`KZ}=tvw87{G&!UM(5Gwp$xcP~<$lJ(Ptv?7*D;X3)<xo(wso;HQ(ICF
zc_(z4s=Fu$Z1Ptf;)#umKf*me%fv(D+X6Q_Q*s-`b2wpK?46gt-G+YN!~JiOX(e_|
zA|fKqk}>20BFd&n?kw;dfSVDld*(G9*jDK0c~Upi_&pZZ(Qx3K!tZ?H5CbigeGzX)
z;b-2UKSIlctjoRym1w*C^2PHO38nBWrQ=t4sQB)JUB-LhTPi~zp)1RqE%@@mZ#8y9
z>m5#1Vcb$G2t1hZD!LWrLKm6j{8;TF^=Cd&BcY)@v9*fQ{Xp@-E=voRy+FT@fpKl+
zsI#MELg(CQ%^Y89R^t7uluG~>3y;5ng${?G|LV(^p1=Iu?3XV-cYgM3FP**c{N-0j
zr+{A`e%^*naFi4dCEMVijV5KLsH)BKk5sVeIFmsZRFWR~KHN(fhXKtd2;iS$?kx_n
z+_z7ce}qxtsOb;!b&K#UCQ?GYc%yp&iN5D+AkRP}H|}ftHU6zZn>=nVF^u4yVGM$%
z_tU4suJ_DdWv%&j=;)-jMJd>jdq*10dKU!(1n|oV+mmcJnRYdgHTsb!$OUY2vh5Rb
z+b&xlJ>`PR2#pfwokSFph!&bKX&Ezv(Nsk8#|D-6NF9d~@Yq9l|3P#|%91t+iu)6M
zmyBsRQun!ZW+*5(bk<8JvGjfU+_@-6m8x8h<FwY0S5c>3^38U9hmvoh9`#S(s(*^L
zdr`S6WyrWs)Q{ubf1K(m(V3y3<Nh%^EsaX%eUn}Y!u!GGEtr^bb+C?e7UW{0=!BW#
z<Rk2R`~zV#Bkb#O1^yuTjXfU|?a@2V-gw)46aE~L!b!8NerxF-iakU@iGooI_#MFG
z6gxw~Sqh$}fW1GzgjksI=r@BTdt|bw9Q!!&F{QjhRo5t>BUe6Elyq{YthHo3RYE!m
z%#vj`lAYv_D4<=C!G_mAK|D-Gtia};#4mics?KSe1&`>C4tmvJ5pr2Em)n)wLs$Mt
z{@(n9nw}rUbpqF<ayBor3wrUfp;Z28es6wX-pcRK59EvaT>h^7VZ<lVQl)q2_x>-7
CU*^UD

literal 0
HcmV?d00001

diff --git a/utils/__pycache__/file.cpython-310.pyc b/utils/__pycache__/file.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..efa2be63551750c690c0cf5d313d2e6fc7d82249
GIT binary patch
literal 2708
zcmZWrTW=f36`q-$z3?KMwiP)_9CVYWZP>bG9VA5~q(uT725J~WA+s<_q#%}@A-UGx
zJhQYcVF?6~(KolR1%d$5W8Q*1<j2fwp8ON)G<DAzij?BaV$R&pob#Q_Y`(EkCs6+J
z@}KhS79ro^;9_aR;1j6&8z6!RnvemuETJhIu+$ki6vs}&Q+MFnK2N;VANXl72w?2W
zn(!o(wQVxt!WY3)KB(7fwK{x_T2s6!HkPZRB^}%)F3E;yL%l4U;tJH&>6W-EHlMnK
z4e^@g3A^!|IF|F4xCWeC;=0%tZ#<=gOM>1fowuHY++?@I48Ira(zyM3k;o1;!CWOp
zq_l*kO_3jDNgjzIOb*L@IEfQ!{HZL5B37n_!(jr)hecG*jHhHdSJ{{%>$BoQ8$i7U
zRri5R$%2-o0>qBlzqS{wVz<fPHx^FioX}IU;1yTbDtf{yeo85MRDS>{a`=t`&k7{J
zN%YSTU<JRwLB+Wa+ci{m9K-WMX2#W}iVNdK1&l<;RX*HnT$ENDXA)<ICvheXN2$Le
z>M|_9_+xLDr?Qvjr5xq?es7!$*(;>dc@`zT`yc)8gFn3g(TB_FTet4-_2y-q=pHid
z7Kf%~VOj%%D(lyQaN1-64JfB|%06d5)L(M9{XKWpRapGWIpW=i#T00fO^;mY%juox
z?Oze{{3R6Q$LZ8U&E=C)o8}}>lKcQ5&-P)I<kMlH@~M(qo69(okEJ>sX4X#H1ZnhS
zsN*9!_TaKr*?>+Bs<p_s&;g5h6#*;s7k)_|(#o%xpwG}QcRzC$!J@XP3$|!fb(r~h
z(S$wh#HpI6RDg$_@KdnCW1{vdSMYsTy<fVOcT6i6`@&r`D9PTgyc5r!^Q!<(GPG^w
zsJ*kje2c(Y2ID|@%egOz@GJ5K`H~-_A8b_(^#C~0M;niL<$=wEUGUb%5)li?o#*Iy
z3Q7M6teuJQs1e3#G?hAxv#?V$EZ2q~M9IFw_%(qplDGtDjSH|!Wt>E2s)<YdVGe%c
z7Anrl&W5^%oY9_Y8_63;-b8X82|_w>t*uM|k+BR1Iy80b5LuLh<JD!PlEa}c^TO2D
z^Va#6<O*;xZ5vKj=%p2Z4YT@PAUp`@>)=O7H=*S~H=)ly$3A-rEmOZi9!u`%+&>9L
z2iC3Oku4e<0*#!~&)9-4m>oM67zf4<gd(k+iccH>o9=eF`Wc*A0_R%b)Y~ut;dKKV
z(}1KQvt=ew$lY*V%G5Xj@2g5)#SF5B0MJsUwo;*=wrLw;McssMf6Rfd3dR&}usi<^
zor(%NC#qGFV@Mn&wQk7;I|XfE1o8_<@Mnw?a4B+37p~poRqjL35uDzE^Hlw^au%Ll
z@y~eB%0m~0a6q0x<)2c0x2y@^0&eM5K04=@E;?peuSn$!3Lz6b<KWajpjC_h!U(ws
zi8Dkk!%42va8{N@2zm5Icn~K^xJs>I3W+8G2c3kO977~W>abg{hfwz9WCu^_onCJm
zm$Uh(JI>SI%}A&?%0|6OG?u@Ol6ab>fP8P1<fC310qV*w4bPWsu7WhR<ol_b>K%In
z-U1?p??J}w?1X>ayBC&uIFf7Wr0A|UT0t=zt+w}PGDI)KGdg-K!xAP|dP%X=tQ+2+
z<?}=!*C?FKL3KRS4gWj|ck@i%2<IBu9G1Bn&qB}<Z#cgcCA=*s(L5=4?EdqmRq6}h
zgM^)c_zz|hBv4kIS$^;%al+N^?m4@@n#*%`Yk$0`-UV%}B{7s=5gnCYuVJ3S2>|uN
zMX2Z6rual&fFF?G3Huk2#N)4AyA*QQG{!SI-Y+0CmAa2_wS0n)Z@z!jx_$Oi$Julz
zeCNo02T?Gz4zGA)7U>y06mk+jF~JIlPD5cZod>nTB!h3pz61+sg`sDHQKTg*ZTx7Y
z(QwAImumG3JdMwLCom4^Yier;QNPDEn*{tbi$W^JC^rE@Wf?Zc5vU_<9(XlDTsePD
zCn`_z_v0?G)R=dJOCD{}77JL*zN=fbNw*+paH>8<X<W4l&7C{yHE7fpPC2{%-A>!8
zYP|BW0FV>}C{Yd&(**8GJnGuN9;^1CX1%csvRe%;QpN#O*=O5ubU5RH8EnF}xyoLC
zK@a??P&TCp9|JJ+ME(h1CXLVpkmQ=IO}8BOE~CqjzwjW5dyM^$IqV0<*?*bKzGohL
N;raFdT;Fc_{|91c&F%mI

literal 0
HcmV?d00001

diff --git a/utils/__pycache__/file.cpython-37.pyc b/utils/__pycache__/file.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..4d38007aab4060f64b31a7d6a0db594841e4ecb3
GIT binary patch
literal 2625
zcmZ8jOOG4J5$>KhA4{#I_1abhc`P`H={RI8AeV(<D2gD!!D1mx8*D&0AcpKFIpRF1
zo}u=E8DCacAr~J4*amhl0Sx4jgU|jUea%UKA(!~8rk1<W45nwgySlpStE#W(!Nx|D
z;rgulLjJqM*ng;VaW(L9A5ZZQG=d49umQcTgs0ZPO6`Hosc$Dv>JD7<?Id365BxM3
z1f0o+v}EIbCfx5W;R*kxGiZu;R%6np8QH?!#=VkjfA5P8(Rt|(Hn0{FbWdz(xCRZI
z;yN^R#J0F8wqEkVHNl^<?z^vH9oy?#+S!eZQoH-hMIyUg2TPSKB2`FCZHfFaOY%qz
zF*q#q;UrF^_NTHOidaFQLj%JE3WkfQoM}(Va;dU0C*`<YJcP%e@DvZwOj*rKRsoe$
z>+he`*3sR{ddU8EtF|lqjGwdGsT{Rk@iVJ(&N*k#w-49>(E5r!VSi;6S#IrVuDZ5%
z@`cQ_Ta+qZXfIlzC%Ufkv5R)eu0q?BIMYrNXHq+4-;YdP$Mk>i_h)%3`&nMfQJ&BH
zA}XW)ROQP>|LHQ@jkD=^CbRxH7qb5_o=hHm`J}%r<D}@5-rnL^cMK|P?*sA=iNF|l
ztTu1*fIIv>u5M!FNBhL-JzGrCF0$!~i*Gr7^m_9*jJ^IJuCA~BIGqxr+L1?Pq1%%@
zN%BMBn$6Km^67A)@~M(Vp|4X<zL4s8n3-@CI!L3VVG*B5MP`gWG6UBJo;CJAqXX={
zqn|1Uy#H1ENB=C@GhX?XCHO0f!M9)Abx=3zrm*VP(Kl5SV=wA9q^vW$YM*m~j5>49
z5rw&@?p3aE=B}ERZsncw%B64N)~&~^^3FW7%C7>fvnY<0t?s>%bha5*TIdJDTaA6k
zgkQ1m*bmMr#r)m@t6J*LWA^;c0ejv=&qLgUJ!EHV1&u-G(Q8UGMa}*RBA$uxq!q?#
zG?hgdXJNOYt-R3AVU*0(CZu#wERwhc#@YqGQfWJpnQGIN^DIZMxC<3$Wp_ipOPb%K
z=6z~zQS&QmZc{_h4qOv79UyM2<mW@(G#QaasVoLf8L8xWSd{rfH`e1O0awipXwjRd
zc#P4jaDIZ(;tm=XI7p;hNTeoGX$x-)?H0dnMij}R0#}+TxmVG@T6s1^Oy!X*MH^Mc
z&iU6?&1=i_?F!LB-$vc>%C4M=4W#*A*HIr}%?dTwpr$^g4PEtHymSi;A+uG{t1gW=
zMJZEl1KYPfyh(Ji_8(BGvM^SeMCF^f)$gHVf9#;8i_(}TjTEtH#RZ=S>UPOa!4M@k
z8R-b?9G;*D+l4KhR~Bc;Cw9tf*T{I4`wafoUS(r_L+w^}?U@<>jSgOUl%MsgA5{K1
zFTO1s0$PxhUgc9}{@_wpmd%P)zTl{r;FW`n_JLF*dTcSa3x*8I!*G(TG@O;?A_QT6
z8Xm?;5?*j<n1W6cWacEy<QTOaspDR=8RB{(Cp)yN*y;DDaXDL#dgDCpe-;T9N7<-9
ziN^AGQ4&wH6sY${Nj~bQ5fE2qYIr$iSPBHS()&j>Eq2Tfv<plMKL%ZPcf!9s*$vA)
z9LcqHQuNk>#!%v;3-SG#3@OEEO%c72VTpkYzob}s)(fA`@?|1OYZOkFaGlol!Y?P`
zUY^NM!({<A$7Qa@vk*Sg4wsvfg%9N<S|;U=k-wa}AiwZquyq3S4`vb;C}YknKm0#)
z!VBr%rMi8!l$Yw(xp>vRE8dt$QVqXlIyt?bBZ8p;fO=&k@^c+iItj1959D{k{VPb)
z^0&5K8M&uhIDf|T1t_U3_UTZTM|8&K^OMfQH%BrKsh#ldiF=n&(7XvQXw9NxhND7G
z;v*eg;LvR;DyGYVRv)2Dr^Z|&0)?Wo)4?bzBss19XjD+(v}ZOg)URnZo#fp>+wfO6
z)(N8iKy!uw{5KH=hf;;>fS@u8TjL1s2*Uwylf#Y6b9AEel>RgR43z~DF1r?Rk9Vw&
z74VKZQMY-UZxdhK$1J#L3$I6y)H`^oZ5p!o_V>D*##QZ=#|wa@P@qiNXmlIelX%oK
zzc<G1LBnKY54IZ*4N}@hsLau=9ZEQDBMgRb4NsZPulPZK6v}XU@Hv24Ch`j+;et+&
WfN$b=VAJ0Ei1F3E&2QYe<^KXhS;D{o

literal 0
HcmV?d00001

diff --git a/utils/__pycache__/metrics.cpython-310.pyc b/utils/__pycache__/metrics.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..55f8b41e3a5a06376a50bf996bd5ecc7baabafea
GIT binary patch
literal 2652
zcmcgu&2Jk;6yMno+i^bnNn6?i11cyhTyIklhX5tu(h^izQ4)#PRkhvhjO|U<J7#8_
z;>f*_^hWsy5K_vGGylyV;glO^PP{kkIEfU*1u@pVc{B6o&71e}TZ`G*ss-)gZx{Hd
zQ<n7yPL4MRCQH!OT_DV2B(++_NLqy1%t`H@({hLfyqmf`ujK(>WhGYLBdyYXi&fav
zp2eni?N*tQHOrqq08Li&fB^MLJ|jCDw}<gYEW^Hr`D|N?9iD|f!3HV67rQ7uZXS#!
z=nd#92QsoYtr6L@wrq9bYcLMyKY=N@2BY({w@bb!BQR~tkspsN<}vSuUC@y|vZt*V
zK#Q^MzHU7SZT~7mMvh$8a2G!ErTyFjo{V`|-M6)~>AoVUYr+#)^G2|81srn0>VQLj
zA*>3gT{f+q{SuoQxoq~bb<KK3E?VHGx#)}yFOAZ8kVcx*m9^$7onPV67F|m^S(M(?
z;$~AMiqq9dC&Gu#;kuA<mu_>Z5|R1!YL!A;<!vr`7V|~g)w-`1gMerCXUSI5=PZfp
zLUsb21*kk!6Xh8y4<#xu6n!o`YSE}fwMnl-=hu0rg!C8baxdy|MV}?QOJ$UGIDLGp
zMi-vcsOm<2PUo{4{UBJV(Qb`C1IthYHqdhWKTs1((GNF-WL(nuB-3C~1D4={w4Z#(
zQ$>@E@(7&68Q3P;G|J#^u(?jxyBwT27K<dOnb07<Ez)f`4(D{2(|8;R@avKBnkyX+
z!?+t|8Bgi_h7c*-ybjA$Wb_d|+=2TZH8vWuPVe$|G)VO#y%(j5Pa>oTGCStmb}Ax}
zF)^21!GYq^o$JQWg-;blV_DKR;RW#>yGC<TmsnZNb*QD%FceujJb*;9d`IFS%RmmW
zs$t(@;C^7C-B=nke}2|k{NdK`KMudV<Ck;KL{Fl8ZflWy8Z#^}8v{T3Fn4=A%5pCi
zorPPzmph;@cQK~<^s$h%=1zJlcY2EdYRQYB_OC|J6+Iq+(>eHWD`1h1f{qk}ey}#k
z9wb>O?(!^%1>?b8c<dL;tHD4gsS0{r%OqCy{xCmx7#|+?RfbRkU}mXHK;Ao3v2TzY
z#I<J#{D5AyFA<4bG4vA+4=q7g4}owC5YB4ZjBp2$uI)0Hd3$cl<=%ej6_h{sHp{GR
zXoXjpja(b>j#c;EUDBFj)8?cFCui6!oUAejOFQ=-tSqmr<xa+RUXmQ6-;6OlhYfjj
zdIP$;4n$j<h%)3QAWR4d14w7@5`bF73cKkOc?G0=7h=&*`bu7gS?&Rd@>C*(`gZOr
zp0<scaeP5uL%B*ACRw7x@O_vm6rUq=WM|G;SwGw~fnAX2@WfOmWG_mS9ex6!@hF0{
zG5ahq1waGDI$%1!-HzUcw@u2wA$t7*pj&a%WO>4zgq$&8EYNwAhV=>hKnc@!BBAU|
zP1E^i<Ew9KboKrhHTw7_pw7Y*zfRZsJWkp}`gCIaQ<}7)!odVOxe_nIR%lAp7|s9(
z3Pd|9|3V7!>lAYok3|OXff(~Y+B13D1dW#4RGEh-oYf4L6%RI+A!4T~-9XSA2b59x
z0{6y`rCJ<1@Z`HE?pih3dknkKD<%iS?I<0<H{mqz7y+tc0qBNkth?aojY!AcqiZ0i
z-+~l^;TjIE#5#~~BEcLqo2O(lAVC_20Z1sUqO)g?^0?T_ON?(PG5_a+JIvgDD7jaJ
zKo&#hlBRDPOXM`D1>TtRHd=x>c2rJge+iaMCCw0*>{N>U*2160V#}f)55v3~j{oHV
zKNE)E4x)6t<0viV8>n>#NM6oF*pU&-OGU=#E`DhwMFBMCm;uHVENX)QJ}?UZ;8m>i
zcFBE*Tsm8<lY4%}=$2PuH}_036fj?kMXx6^gBk`U&56^l$Sa=||I+T@7e}F~6{}((
Jw(6dze*@9@j(Y$A

literal 0
HcmV?d00001

diff --git a/utils/common.py b/utils/common.py
new file mode 100644
index 0000000..e8c57c9
--- /dev/null
+++ b/utils/common.py
@@ -0,0 +1,51 @@
+from typing import Mapping, Any
+import importlib
+
+from torch import nn
+
+
+def get_obj_from_str(string: str, reload: bool=False) -> object:
+    module, cls = string.rsplit(".", 1)
+    if reload:
+        module_imp = importlib.import_module(module)
+        importlib.reload(module_imp)
+    return getattr(importlib.import_module(module, package=None), cls)
+
+
+def instantiate_from_config(config: Mapping[str, Any]) -> object:
+    if not "target" in config:
+        raise KeyError("Expected key `target` to instantiate.")
+    return get_obj_from_str(config["target"])(**config.get("params", dict()))
+
+
+def disabled_train(self: nn.Module) -> nn.Module:
+    """Overwrite model.train with this function to make sure train/eval mode
+    does not change anymore."""
+    return self
+
+
+def frozen_module(module: nn.Module) -> None:
+    module.eval()
+    module.train = disabled_train
+    for p in module.parameters():
+        p.requires_grad = False
+
+
+def load_state_dict(model: nn.Module, state_dict: Mapping[str, Any], strict: bool=False) -> None:
+    state_dict = state_dict.get("state_dict", state_dict)
+    
+    is_model_key_starts_with_module = list(model.state_dict().keys())[0].startswith("module.")
+    is_state_dict_key_starts_with_module = list(state_dict.keys())[0].startswith("module.")
+    
+    if (
+        is_model_key_starts_with_module and
+        (not is_state_dict_key_starts_with_module)
+    ):
+        state_dict = {f"module.{key}": value for key, value in state_dict.items()}
+    if (
+        (not is_model_key_starts_with_module) and
+        is_state_dict_key_starts_with_module
+    ):
+        state_dict = {key[len("module."):]: value for key, value in state_dict.items()}
+    
+    model.load_state_dict(state_dict, strict=False)
diff --git a/utils/degradation.py b/utils/degradation.py
new file mode 100644
index 0000000..aa75976
--- /dev/null
+++ b/utils/degradation.py
@@ -0,0 +1,765 @@
+# https://github.com/XPixelGroup/BasicSR/blob/master/basicsr/data/degradations.py
+import cv2
+import math
+import numpy as np
+import random
+import torch
+from scipy import special
+from scipy.stats import multivariate_normal
+from torchvision.transforms.functional_tensor import rgb_to_grayscale
+
+# -------------------------------------------------------------------- #
+# --------------------------- blur kernels --------------------------- #
+# -------------------------------------------------------------------- #
+
+
+# --------------------------- util functions --------------------------- #
+def sigma_matrix2(sig_x, sig_y, theta):
+    """Calculate the rotated sigma matrix (two dimensional matrix).
+
+    Args:
+        sig_x (float):
+        sig_y (float):
+        theta (float): Radian measurement.
+
+    Returns:
+        ndarray: Rotated sigma matrix.
+    """
+    d_matrix = np.array([[sig_x**2, 0], [0, sig_y**2]])
+    u_matrix = np.array([[np.cos(theta), -np.sin(theta)], [np.sin(theta), np.cos(theta)]])
+    return np.dot(u_matrix, np.dot(d_matrix, u_matrix.T))
+
+
+def mesh_grid(kernel_size):
+    """Generate the mesh grid, centering at zero.
+
+    Args:
+        kernel_size (int):
+
+    Returns:
+        xy (ndarray): with the shape (kernel_size, kernel_size, 2)
+        xx (ndarray): with the shape (kernel_size, kernel_size)
+        yy (ndarray): with the shape (kernel_size, kernel_size)
+    """
+    ax = np.arange(-kernel_size // 2 + 1., kernel_size // 2 + 1.)
+    xx, yy = np.meshgrid(ax, ax)
+    xy = np.hstack((xx.reshape((kernel_size * kernel_size, 1)), yy.reshape(kernel_size * kernel_size,
+                                                                           1))).reshape(kernel_size, kernel_size, 2)
+    return xy, xx, yy
+
+
+def pdf2(sigma_matrix, grid):
+    """Calculate PDF of the bivariate Gaussian distribution.
+
+    Args:
+        sigma_matrix (ndarray): with the shape (2, 2)
+        grid (ndarray): generated by :func:`mesh_grid`,
+            with the shape (K, K, 2), K is the kernel size.
+
+    Returns:
+        kernel (ndarrray): un-normalized kernel.
+    """
+    inverse_sigma = np.linalg.inv(sigma_matrix)
+    kernel = np.exp(-0.5 * np.sum(np.dot(grid, inverse_sigma) * grid, 2))
+    return kernel
+
+
+def cdf2(d_matrix, grid):
+    """Calculate the CDF of the standard bivariate Gaussian distribution.
+        Used in skewed Gaussian distribution.
+
+    Args:
+        d_matrix (ndarrasy): skew matrix.
+        grid (ndarray): generated by :func:`mesh_grid`,
+            with the shape (K, K, 2), K is the kernel size.
+
+    Returns:
+        cdf (ndarray): skewed cdf.
+    """
+    rv = multivariate_normal([0, 0], [[1, 0], [0, 1]])
+    grid = np.dot(grid, d_matrix)
+    cdf = rv.cdf(grid)
+    return cdf
+
+
+def bivariate_Gaussian(kernel_size, sig_x, sig_y, theta, grid=None, isotropic=True):
+    """Generate a bivariate isotropic or anisotropic Gaussian kernel.
+
+    In the isotropic mode, only `sig_x` is used. `sig_y` and `theta` is ignored.
+
+    Args:
+        kernel_size (int):
+        sig_x (float):
+        sig_y (float):
+        theta (float): Radian measurement.
+        grid (ndarray, optional): generated by :func:`mesh_grid`,
+            with the shape (K, K, 2), K is the kernel size. Default: None
+        isotropic (bool):
+
+    Returns:
+        kernel (ndarray): normalized kernel.
+    """
+    if grid is None:
+        grid, _, _ = mesh_grid(kernel_size)
+    if isotropic:
+        sigma_matrix = np.array([[sig_x**2, 0], [0, sig_x**2]])
+    else:
+        sigma_matrix = sigma_matrix2(sig_x, sig_y, theta)
+    kernel = pdf2(sigma_matrix, grid)
+    kernel = kernel / np.sum(kernel)
+    return kernel
+
+
+def bivariate_generalized_Gaussian(kernel_size, sig_x, sig_y, theta, beta, grid=None, isotropic=True):
+    """Generate a bivariate generalized Gaussian kernel.
+
+    ``Paper: Parameter Estimation For Multivariate Generalized Gaussian Distributions``
+
+    In the isotropic mode, only `sig_x` is used. `sig_y` and `theta` is ignored.
+
+    Args:
+        kernel_size (int):
+        sig_x (float):
+        sig_y (float):
+        theta (float): Radian measurement.
+        beta (float): shape parameter, beta = 1 is the normal distribution.
+        grid (ndarray, optional): generated by :func:`mesh_grid`,
+            with the shape (K, K, 2), K is the kernel size. Default: None
+
+    Returns:
+        kernel (ndarray): normalized kernel.
+    """
+    if grid is None:
+        grid, _, _ = mesh_grid(kernel_size)
+    if isotropic:
+        sigma_matrix = np.array([[sig_x**2, 0], [0, sig_x**2]])
+    else:
+        sigma_matrix = sigma_matrix2(sig_x, sig_y, theta)
+    inverse_sigma = np.linalg.inv(sigma_matrix)
+    kernel = np.exp(-0.5 * np.power(np.sum(np.dot(grid, inverse_sigma) * grid, 2), beta))
+    kernel = kernel / np.sum(kernel)
+    return kernel
+
+
+def bivariate_plateau(kernel_size, sig_x, sig_y, theta, beta, grid=None, isotropic=True):
+    """Generate a plateau-like anisotropic kernel.
+
+    1 / (1+x^(beta))
+
+    Reference: https://stats.stackexchange.com/questions/203629/is-there-a-plateau-shaped-distribution
+
+    In the isotropic mode, only `sig_x` is used. `sig_y` and `theta` is ignored.
+
+    Args:
+        kernel_size (int):
+        sig_x (float):
+        sig_y (float):
+        theta (float): Radian measurement.
+        beta (float): shape parameter, beta = 1 is the normal distribution.
+        grid (ndarray, optional): generated by :func:`mesh_grid`,
+            with the shape (K, K, 2), K is the kernel size. Default: None
+
+    Returns:
+        kernel (ndarray): normalized kernel.
+    """
+    if grid is None:
+        grid, _, _ = mesh_grid(kernel_size)
+    if isotropic:
+        sigma_matrix = np.array([[sig_x**2, 0], [0, sig_x**2]])
+    else:
+        sigma_matrix = sigma_matrix2(sig_x, sig_y, theta)
+    inverse_sigma = np.linalg.inv(sigma_matrix)
+    kernel = np.reciprocal(np.power(np.sum(np.dot(grid, inverse_sigma) * grid, 2), beta) + 1)
+    kernel = kernel / np.sum(kernel)
+    return kernel
+
+
+def random_bivariate_Gaussian(kernel_size,
+                              sigma_x_range,
+                              sigma_y_range,
+                              rotation_range,
+                              noise_range=None,
+                              isotropic=True):
+    """Randomly generate bivariate isotropic or anisotropic Gaussian kernels.
+
+    In the isotropic mode, only `sigma_x_range` is used. `sigma_y_range` and `rotation_range` is ignored.
+
+    Args:
+        kernel_size (int):
+        sigma_x_range (tuple): [0.6, 5]
+        sigma_y_range (tuple): [0.6, 5]
+        rotation range (tuple): [-math.pi, math.pi]
+        noise_range(tuple, optional): multiplicative kernel noise,
+            [0.75, 1.25]. Default: None
+
+    Returns:
+        kernel (ndarray):
+    """
+    assert kernel_size % 2 == 1, 'Kernel size must be an odd number.'
+    assert sigma_x_range[0] < sigma_x_range[1], 'Wrong sigma_x_range.'
+    sigma_x = np.random.uniform(sigma_x_range[0], sigma_x_range[1])
+    if isotropic is False:
+        assert sigma_y_range[0] < sigma_y_range[1], 'Wrong sigma_y_range.'
+        assert rotation_range[0] < rotation_range[1], 'Wrong rotation_range.'
+        sigma_y = np.random.uniform(sigma_y_range[0], sigma_y_range[1])
+        rotation = np.random.uniform(rotation_range[0], rotation_range[1])
+    else:
+        sigma_y = sigma_x
+        rotation = 0
+
+    kernel = bivariate_Gaussian(kernel_size, sigma_x, sigma_y, rotation, isotropic=isotropic)
+
+    # add multiplicative noise
+    if noise_range is not None:
+        assert noise_range[0] < noise_range[1], 'Wrong noise range.'
+        noise = np.random.uniform(noise_range[0], noise_range[1], size=kernel.shape)
+        kernel = kernel * noise
+    kernel = kernel / np.sum(kernel)
+    return kernel
+
+
+def random_bivariate_generalized_Gaussian(kernel_size,
+                                          sigma_x_range,
+                                          sigma_y_range,
+                                          rotation_range,
+                                          beta_range,
+                                          noise_range=None,
+                                          isotropic=True):
+    """Randomly generate bivariate generalized Gaussian kernels.
+
+    In the isotropic mode, only `sigma_x_range` is used. `sigma_y_range` and `rotation_range` is ignored.
+
+    Args:
+        kernel_size (int):
+        sigma_x_range (tuple): [0.6, 5]
+        sigma_y_range (tuple): [0.6, 5]
+        rotation range (tuple): [-math.pi, math.pi]
+        beta_range (tuple): [0.5, 8]
+        noise_range(tuple, optional): multiplicative kernel noise,
+            [0.75, 1.25]. Default: None
+
+    Returns:
+        kernel (ndarray):
+    """
+    assert kernel_size % 2 == 1, 'Kernel size must be an odd number.'
+    assert sigma_x_range[0] < sigma_x_range[1], 'Wrong sigma_x_range.'
+    sigma_x = np.random.uniform(sigma_x_range[0], sigma_x_range[1])
+    if isotropic is False:
+        assert sigma_y_range[0] < sigma_y_range[1], 'Wrong sigma_y_range.'
+        assert rotation_range[0] < rotation_range[1], 'Wrong rotation_range.'
+        sigma_y = np.random.uniform(sigma_y_range[0], sigma_y_range[1])
+        rotation = np.random.uniform(rotation_range[0], rotation_range[1])
+    else:
+        sigma_y = sigma_x
+        rotation = 0
+
+    # assume beta_range[0] < 1 < beta_range[1]
+    if np.random.uniform() < 0.5:
+        beta = np.random.uniform(beta_range[0], 1)
+    else:
+        beta = np.random.uniform(1, beta_range[1])
+
+    kernel = bivariate_generalized_Gaussian(kernel_size, sigma_x, sigma_y, rotation, beta, isotropic=isotropic)
+
+    # add multiplicative noise
+    if noise_range is not None:
+        assert noise_range[0] < noise_range[1], 'Wrong noise range.'
+        noise = np.random.uniform(noise_range[0], noise_range[1], size=kernel.shape)
+        kernel = kernel * noise
+    kernel = kernel / np.sum(kernel)
+    return kernel
+
+
+def random_bivariate_plateau(kernel_size,
+                             sigma_x_range,
+                             sigma_y_range,
+                             rotation_range,
+                             beta_range,
+                             noise_range=None,
+                             isotropic=True):
+    """Randomly generate bivariate plateau kernels.
+
+    In the isotropic mode, only `sigma_x_range` is used. `sigma_y_range` and `rotation_range` is ignored.
+
+    Args:
+        kernel_size (int):
+        sigma_x_range (tuple): [0.6, 5]
+        sigma_y_range (tuple): [0.6, 5]
+        rotation range (tuple): [-math.pi/2, math.pi/2]
+        beta_range (tuple): [1, 4]
+        noise_range(tuple, optional): multiplicative kernel noise,
+            [0.75, 1.25]. Default: None
+
+    Returns:
+        kernel (ndarray):
+    """
+    assert kernel_size % 2 == 1, 'Kernel size must be an odd number.'
+    assert sigma_x_range[0] < sigma_x_range[1], 'Wrong sigma_x_range.'
+    sigma_x = np.random.uniform(sigma_x_range[0], sigma_x_range[1])
+    if isotropic is False:
+        assert sigma_y_range[0] < sigma_y_range[1], 'Wrong sigma_y_range.'
+        assert rotation_range[0] < rotation_range[1], 'Wrong rotation_range.'
+        sigma_y = np.random.uniform(sigma_y_range[0], sigma_y_range[1])
+        rotation = np.random.uniform(rotation_range[0], rotation_range[1])
+    else:
+        sigma_y = sigma_x
+        rotation = 0
+
+    # TODO: this may be not proper
+    if np.random.uniform() < 0.5:
+        beta = np.random.uniform(beta_range[0], 1)
+    else:
+        beta = np.random.uniform(1, beta_range[1])
+
+    kernel = bivariate_plateau(kernel_size, sigma_x, sigma_y, rotation, beta, isotropic=isotropic)
+    # add multiplicative noise
+    if noise_range is not None:
+        assert noise_range[0] < noise_range[1], 'Wrong noise range.'
+        noise = np.random.uniform(noise_range[0], noise_range[1], size=kernel.shape)
+        kernel = kernel * noise
+    kernel = kernel / np.sum(kernel)
+
+    return kernel
+
+
+def random_mixed_kernels(kernel_list,
+                         kernel_prob,
+                         kernel_size=21,
+                         sigma_x_range=(0.6, 5),
+                         sigma_y_range=(0.6, 5),
+                         rotation_range=(-math.pi, math.pi),
+                         betag_range=(0.5, 8),
+                         betap_range=(0.5, 8),
+                         noise_range=None):
+    """Randomly generate mixed kernels.
+
+    Args:
+        kernel_list (tuple): a list name of kernel types,
+            support ['iso', 'aniso', 'skew', 'generalized', 'plateau_iso',
+            'plateau_aniso']
+        kernel_prob (tuple): corresponding kernel probability for each
+            kernel type
+        kernel_size (int):
+        sigma_x_range (tuple): [0.6, 5]
+        sigma_y_range (tuple): [0.6, 5]
+        rotation range (tuple): [-math.pi, math.pi]
+        beta_range (tuple): [0.5, 8]
+        noise_range(tuple, optional): multiplicative kernel noise,
+            [0.75, 1.25]. Default: None
+
+    Returns:
+        kernel (ndarray):
+    """
+    kernel_type = random.choices(kernel_list, kernel_prob)[0]
+    if kernel_type == 'iso':
+        kernel = random_bivariate_Gaussian(
+            kernel_size, sigma_x_range, sigma_y_range, rotation_range, noise_range=noise_range, isotropic=True)
+    elif kernel_type == 'aniso':
+        kernel = random_bivariate_Gaussian(
+            kernel_size, sigma_x_range, sigma_y_range, rotation_range, noise_range=noise_range, isotropic=False)
+    elif kernel_type == 'generalized_iso':
+        kernel = random_bivariate_generalized_Gaussian(
+            kernel_size,
+            sigma_x_range,
+            sigma_y_range,
+            rotation_range,
+            betag_range,
+            noise_range=noise_range,
+            isotropic=True)
+    elif kernel_type == 'generalized_aniso':
+        kernel = random_bivariate_generalized_Gaussian(
+            kernel_size,
+            sigma_x_range,
+            sigma_y_range,
+            rotation_range,
+            betag_range,
+            noise_range=noise_range,
+            isotropic=False)
+    elif kernel_type == 'plateau_iso':
+        kernel = random_bivariate_plateau(
+            kernel_size, sigma_x_range, sigma_y_range, rotation_range, betap_range, noise_range=None, isotropic=True)
+    elif kernel_type == 'plateau_aniso':
+        kernel = random_bivariate_plateau(
+            kernel_size, sigma_x_range, sigma_y_range, rotation_range, betap_range, noise_range=None, isotropic=False)
+    return kernel
+
+
+np.seterr(divide='ignore', invalid='ignore')
+
+
+def circular_lowpass_kernel(cutoff, kernel_size, pad_to=0):
+    """2D sinc filter
+
+    Reference: https://dsp.stackexchange.com/questions/58301/2-d-circularly-symmetric-low-pass-filter
+
+    Args:
+        cutoff (float): cutoff frequency in radians (pi is max)
+        kernel_size (int): horizontal and vertical size, must be odd.
+        pad_to (int): pad kernel size to desired size, must be odd or zero.
+    """
+    assert kernel_size % 2 == 1, 'Kernel size must be an odd number.'
+    kernel = np.fromfunction(
+        lambda x, y: cutoff * special.j1(cutoff * np.sqrt(
+            (x - (kernel_size - 1) / 2)**2 + (y - (kernel_size - 1) / 2)**2)) / (2 * np.pi * np.sqrt(
+                (x - (kernel_size - 1) / 2)**2 + (y - (kernel_size - 1) / 2)**2)), [kernel_size, kernel_size])
+    kernel[(kernel_size - 1) // 2, (kernel_size - 1) // 2] = cutoff**2 / (4 * np.pi)
+    kernel = kernel / np.sum(kernel)
+    if pad_to > kernel_size:
+        pad_size = (pad_to - kernel_size) // 2
+        kernel = np.pad(kernel, ((pad_size, pad_size), (pad_size, pad_size)))
+    return kernel
+
+
+# ------------------------------------------------------------- #
+# --------------------------- noise --------------------------- #
+# ------------------------------------------------------------- #
+
+# ----------------------- Gaussian Noise ----------------------- #
+
+
+def generate_gaussian_noise(img, sigma=10, gray_noise=False):
+    """Generate Gaussian noise.
+
+    Args:
+        img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32.
+        sigma (float): Noise scale (measured in range 255). Default: 10.
+
+    Returns:
+        (Numpy array): Returned noisy image, shape (h, w, c), range[0, 1],
+            float32.
+    """
+    if gray_noise:
+        noise = np.float32(np.random.randn(*(img.shape[0:2]))) * sigma / 255.
+        noise = np.expand_dims(noise, axis=2).repeat(3, axis=2)
+    else:
+        noise = np.float32(np.random.randn(*(img.shape))) * sigma / 255.
+    return noise
+
+
+def add_gaussian_noise(img, sigma=10, clip=True, rounds=False, gray_noise=False):
+    """Add Gaussian noise.
+
+    Args:
+        img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32.
+        sigma (float): Noise scale (measured in range 255). Default: 10.
+
+    Returns:
+        (Numpy array): Returned noisy image, shape (h, w, c), range[0, 1],
+            float32.
+    """
+    noise = generate_gaussian_noise(img, sigma, gray_noise)
+    out = img + noise
+    if clip and rounds:
+        out = np.clip((out * 255.0).round(), 0, 255) / 255.
+    elif clip:
+        out = np.clip(out, 0, 1)
+    elif rounds:
+        out = (out * 255.0).round() / 255.
+    return out
+
+
+def generate_gaussian_noise_pt(img, sigma=10, gray_noise=0):
+    """Add Gaussian noise (PyTorch version).
+
+    Args:
+        img (Tensor): Shape (b, c, h, w), range[0, 1], float32.
+        scale (float | Tensor): Noise scale. Default: 1.0.
+
+    Returns:
+        (Tensor): Returned noisy image, shape (b, c, h, w), range[0, 1],
+            float32.
+    """
+    b, _, h, w = img.size()
+    if not isinstance(sigma, (float, int)):
+        sigma = sigma.view(img.size(0), 1, 1, 1)
+    if isinstance(gray_noise, (float, int)):
+        cal_gray_noise = gray_noise > 0
+    else:
+        gray_noise = gray_noise.view(b, 1, 1, 1)
+        cal_gray_noise = torch.sum(gray_noise) > 0
+
+    if cal_gray_noise:
+        noise_gray = torch.randn(*img.size()[2:4], dtype=img.dtype, device=img.device) * sigma / 255.
+        noise_gray = noise_gray.view(b, 1, h, w)
+
+    # always calculate color noise
+    noise = torch.randn(*img.size(), dtype=img.dtype, device=img.device) * sigma / 255.
+
+    if cal_gray_noise:
+        noise = noise * (1 - gray_noise) + noise_gray * gray_noise
+    return noise
+
+
+def add_gaussian_noise_pt(img, sigma=10, gray_noise=0, clip=True, rounds=False):
+    """Add Gaussian noise (PyTorch version).
+
+    Args:
+        img (Tensor): Shape (b, c, h, w), range[0, 1], float32.
+        scale (float | Tensor): Noise scale. Default: 1.0.
+
+    Returns:
+        (Tensor): Returned noisy image, shape (b, c, h, w), range[0, 1],
+            float32.
+    """
+    noise = generate_gaussian_noise_pt(img, sigma, gray_noise)
+    out = img + noise
+    if clip and rounds:
+        out = torch.clamp((out * 255.0).round(), 0, 255) / 255.
+    elif clip:
+        out = torch.clamp(out, 0, 1)
+    elif rounds:
+        out = (out * 255.0).round() / 255.
+    return out
+
+
+# ----------------------- Random Gaussian Noise ----------------------- #
+def random_generate_gaussian_noise(img, sigma_range=(0, 10), gray_prob=0):
+    sigma = np.random.uniform(sigma_range[0], sigma_range[1])
+    if np.random.uniform() < gray_prob:
+        gray_noise = True
+    else:
+        gray_noise = False
+    return generate_gaussian_noise(img, sigma, gray_noise)
+
+
+def random_add_gaussian_noise(img, sigma_range=(0, 1.0), gray_prob=0, clip=True, rounds=False):
+    noise = random_generate_gaussian_noise(img, sigma_range, gray_prob)
+    out = img + noise
+    if clip and rounds:
+        out = np.clip((out * 255.0).round(), 0, 255) / 255.
+    elif clip:
+        out = np.clip(out, 0, 1)
+    elif rounds:
+        out = (out * 255.0).round() / 255.
+    return out
+
+
+def random_generate_gaussian_noise_pt(img, sigma_range=(0, 10), gray_prob=0):
+    sigma = torch.rand(
+        img.size(0), dtype=img.dtype, device=img.device) * (sigma_range[1] - sigma_range[0]) + sigma_range[0]
+    gray_noise = torch.rand(img.size(0), dtype=img.dtype, device=img.device)
+    gray_noise = (gray_noise < gray_prob).float()
+    return generate_gaussian_noise_pt(img, sigma, gray_noise)
+
+
+def random_add_gaussian_noise_pt(img, sigma_range=(0, 1.0), gray_prob=0, clip=True, rounds=False):
+    noise = random_generate_gaussian_noise_pt(img, sigma_range, gray_prob)
+    out = img + noise
+    if clip and rounds:
+        out = torch.clamp((out * 255.0).round(), 0, 255) / 255.
+    elif clip:
+        out = torch.clamp(out, 0, 1)
+    elif rounds:
+        out = (out * 255.0).round() / 255.
+    return out
+
+
+# ----------------------- Poisson (Shot) Noise ----------------------- #
+
+
+def generate_poisson_noise(img, scale=1.0, gray_noise=False):
+    """Generate poisson noise.
+
+    Reference: https://github.com/scikit-image/scikit-image/blob/main/skimage/util/noise.py#L37-L219
+
+    Args:
+        img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32.
+        scale (float): Noise scale. Default: 1.0.
+        gray_noise (bool): Whether generate gray noise. Default: False.
+
+    Returns:
+        (Numpy array): Returned noisy image, shape (h, w, c), range[0, 1],
+            float32.
+    """
+    if gray_noise:
+        img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+    # round and clip image for counting vals correctly
+    img = np.clip((img * 255.0).round(), 0, 255) / 255.
+    vals = len(np.unique(img))
+    vals = 2**np.ceil(np.log2(vals))
+    out = np.float32(np.random.poisson(img * vals) / float(vals))
+    noise = out - img
+    if gray_noise:
+        noise = np.repeat(noise[:, :, np.newaxis], 3, axis=2)
+    return noise * scale
+
+
+def add_poisson_noise(img, scale=1.0, clip=True, rounds=False, gray_noise=False):
+    """Add poisson noise.
+
+    Args:
+        img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32.
+        scale (float): Noise scale. Default: 1.0.
+        gray_noise (bool): Whether generate gray noise. Default: False.
+
+    Returns:
+        (Numpy array): Returned noisy image, shape (h, w, c), range[0, 1],
+            float32.
+    """
+    noise = generate_poisson_noise(img, scale, gray_noise)
+    out = img + noise
+    if clip and rounds:
+        out = np.clip((out * 255.0).round(), 0, 255) / 255.
+    elif clip:
+        out = np.clip(out, 0, 1)
+    elif rounds:
+        out = (out * 255.0).round() / 255.
+    return out
+
+
+def generate_poisson_noise_pt(img, scale=1.0, gray_noise=0):
+    """Generate a batch of poisson noise (PyTorch version)
+
+    Args:
+        img (Tensor): Input image, shape (b, c, h, w), range [0, 1], float32.
+        scale (float | Tensor): Noise scale. Number or Tensor with shape (b).
+            Default: 1.0.
+        gray_noise (float | Tensor): 0-1 number or Tensor with shape (b).
+            0 for False, 1 for True. Default: 0.
+
+    Returns:
+        (Tensor): Returned noisy image, shape (b, c, h, w), range[0, 1],
+            float32.
+    """
+    b, _, h, w = img.size()
+    if isinstance(gray_noise, (float, int)):
+        cal_gray_noise = gray_noise > 0
+    else:
+        gray_noise = gray_noise.view(b, 1, 1, 1)
+        cal_gray_noise = torch.sum(gray_noise) > 0
+    if cal_gray_noise:
+        img_gray = rgb_to_grayscale(img, num_output_channels=1)
+        # round and clip image for counting vals correctly
+        img_gray = torch.clamp((img_gray * 255.0).round(), 0, 255) / 255.
+        # use for-loop to get the unique values for each sample
+        vals_list = [len(torch.unique(img_gray[i, :, :, :])) for i in range(b)]
+        vals_list = [2**np.ceil(np.log2(vals)) for vals in vals_list]
+        vals = img_gray.new_tensor(vals_list).view(b, 1, 1, 1)
+        out = torch.poisson(img_gray * vals) / vals
+        noise_gray = out - img_gray
+        noise_gray = noise_gray.expand(b, 3, h, w)
+
+    # always calculate color noise
+    # round and clip image for counting vals correctly
+    img = torch.clamp((img * 255.0).round(), 0, 255) / 255.
+    # use for-loop to get the unique values for each sample
+    vals_list = [len(torch.unique(img[i, :, :, :])) for i in range(b)]
+    vals_list = [2**np.ceil(np.log2(vals)) for vals in vals_list]
+    vals = img.new_tensor(vals_list).view(b, 1, 1, 1)
+    out = torch.poisson(img * vals) / vals
+    noise = out - img
+    if cal_gray_noise:
+        noise = noise * (1 - gray_noise) + noise_gray * gray_noise
+    if not isinstance(scale, (float, int)):
+        scale = scale.view(b, 1, 1, 1)
+    return noise * scale
+
+
+def add_poisson_noise_pt(img, scale=1.0, clip=True, rounds=False, gray_noise=0):
+    """Add poisson noise to a batch of images (PyTorch version).
+
+    Args:
+        img (Tensor): Input image, shape (b, c, h, w), range [0, 1], float32.
+        scale (float | Tensor): Noise scale. Number or Tensor with shape (b).
+            Default: 1.0.
+        gray_noise (float | Tensor): 0-1 number or Tensor with shape (b).
+            0 for False, 1 for True. Default: 0.
+
+    Returns:
+        (Tensor): Returned noisy image, shape (b, c, h, w), range[0, 1],
+            float32.
+    """
+    noise = generate_poisson_noise_pt(img, scale, gray_noise)
+    out = img + noise
+    if clip and rounds:
+        out = torch.clamp((out * 255.0).round(), 0, 255) / 255.
+    elif clip:
+        out = torch.clamp(out, 0, 1)
+    elif rounds:
+        out = (out * 255.0).round() / 255.
+    return out
+
+
+# ----------------------- Random Poisson (Shot) Noise ----------------------- #
+
+
+def random_generate_poisson_noise(img, scale_range=(0, 1.0), gray_prob=0):
+    scale = np.random.uniform(scale_range[0], scale_range[1])
+    if np.random.uniform() < gray_prob:
+        gray_noise = True
+    else:
+        gray_noise = False
+    return generate_poisson_noise(img, scale, gray_noise)
+
+
+def random_add_poisson_noise(img, scale_range=(0, 1.0), gray_prob=0, clip=True, rounds=False):
+    noise = random_generate_poisson_noise(img, scale_range, gray_prob)
+    out = img + noise
+    if clip and rounds:
+        out = np.clip((out * 255.0).round(), 0, 255) / 255.
+    elif clip:
+        out = np.clip(out, 0, 1)
+    elif rounds:
+        out = (out * 255.0).round() / 255.
+    return out
+
+
+def random_generate_poisson_noise_pt(img, scale_range=(0, 1.0), gray_prob=0):
+    scale = torch.rand(
+        img.size(0), dtype=img.dtype, device=img.device) * (scale_range[1] - scale_range[0]) + scale_range[0]
+    gray_noise = torch.rand(img.size(0), dtype=img.dtype, device=img.device)
+    gray_noise = (gray_noise < gray_prob).float()
+    return generate_poisson_noise_pt(img, scale, gray_noise)
+
+
+def random_add_poisson_noise_pt(img, scale_range=(0, 1.0), gray_prob=0, clip=True, rounds=False):
+    noise = random_generate_poisson_noise_pt(img, scale_range, gray_prob)
+    out = img + noise
+    if clip and rounds:
+        out = torch.clamp((out * 255.0).round(), 0, 255) / 255.
+    elif clip:
+        out = torch.clamp(out, 0, 1)
+    elif rounds:
+        out = (out * 255.0).round() / 255.
+    return out
+
+
+# ------------------------------------------------------------------------ #
+# --------------------------- JPEG compression --------------------------- #
+# ------------------------------------------------------------------------ #
+
+
+def add_jpg_compression(img, quality=90):
+    """Add JPG compression artifacts.
+
+    Args:
+        img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32.
+        quality (float): JPG compression quality. 0 for lowest quality, 100 for
+            best quality. Default: 90.
+
+    Returns:
+        (Numpy array): Returned image after JPG, shape (h, w, c), range[0, 1],
+            float32.
+    """
+    img = np.clip(img, 0, 1)
+    encode_param = [int(cv2.IMWRITE_JPEG_QUALITY), quality]
+    _, encimg = cv2.imencode('.jpg', img * 255., encode_param)
+    img = np.float32(cv2.imdecode(encimg, 1)) / 255.
+    return img
+
+
+def random_add_jpg_compression(img, quality_range=(90, 100)):
+    """Randomly add JPG compression artifacts.
+
+    Args:
+        img (Numpy array): Input image, shape (h, w, c), range [0, 1], float32.
+        quality_range (tuple[float] | list[float]): JPG compression quality
+            range. 0 for lowest quality, 100 for best quality.
+            Default: (90, 100).
+
+    Returns:
+        (Numpy array): Returned image after JPG, shape (h, w, c), range[0, 1],
+            float32.
+    """
+    quality = np.random.uniform(quality_range[0], quality_range[1])
+    return add_jpg_compression(img, int(quality))
diff --git a/utils/face_restoration_helper.py b/utils/face_restoration_helper.py
new file mode 100644
index 0000000..6f09a3c
--- /dev/null
+++ b/utils/face_restoration_helper.py
@@ -0,0 +1,517 @@
+import cv2
+import numpy as np
+import os
+import torch
+from torchvision.transforms.functional import normalize
+
+from facexlib.detection import init_detection_model
+from facexlib.parsing import init_parsing_model
+from facexlib.utils.misc import img2tensor, imwrite
+
+from .file import load_file_from_url
+
+def get_largest_face(det_faces, h, w):
+
+    def get_location(val, length):
+        if val < 0:
+            return 0
+        elif val > length:
+            return length
+        else:
+            return val
+
+    face_areas = []
+    for det_face in det_faces:
+        left = get_location(det_face[0], w)
+        right = get_location(det_face[2], w)
+        top = get_location(det_face[1], h)
+        bottom = get_location(det_face[3], h)
+        face_area = (right - left) * (bottom - top)
+        face_areas.append(face_area)
+    largest_idx = face_areas.index(max(face_areas))
+    return det_faces[largest_idx], largest_idx
+
+
+def get_center_face(det_faces, h=0, w=0, center=None):
+    if center is not None:
+        center = np.array(center)
+    else:
+        center = np.array([w / 2, h / 2])
+    center_dist = []
+    for det_face in det_faces:
+        face_center = np.array([(det_face[0] + det_face[2]) / 2, (det_face[1] + det_face[3]) / 2])
+        dist = np.linalg.norm(face_center - center)
+        center_dist.append(dist)
+    center_idx = center_dist.index(min(center_dist))
+    return det_faces[center_idx], center_idx
+
+
+class FaceRestoreHelper(object):
+    """Helper for the face restoration pipeline (base class)."""
+
+    def __init__(self,
+                 upscale_factor,
+                 face_size=512,
+                 crop_ratio=(1, 1),
+                 det_model='retinaface_resnet50',
+                 save_ext='png',
+                 template_3points=False,
+                 pad_blur=False,
+                 use_parse=False,
+                 device=None):
+        self.template_3points = template_3points  # improve robustness
+        self.upscale_factor = int(upscale_factor)
+        # the cropped face ratio based on the square face
+        self.crop_ratio = crop_ratio  # (h, w)
+        assert (self.crop_ratio[0] >= 1 and self.crop_ratio[1] >= 1), 'crop ration only supports >=1'
+        self.face_size = (int(face_size * self.crop_ratio[1]), int(face_size * self.crop_ratio[0]))
+        self.det_model = det_model
+
+        if self.det_model == 'dlib':
+            # standard 5 landmarks for FFHQ faces with 1024 x 1024
+            self.face_template = np.array([[686.77227723, 488.62376238], [586.77227723, 493.59405941],
+                                        [337.91089109, 488.38613861], [437.95049505, 493.51485149],
+                                        [513.58415842, 678.5049505]])
+            self.face_template = self.face_template / (1024 // face_size)
+        elif self.template_3points:
+            self.face_template = np.array([[192, 240], [319, 240], [257, 371]])
+        else:
+            # standard 5 landmarks for FFHQ faces with 512 x 512 
+            # facexlib
+            self.face_template = np.array([[192.98138, 239.94708], [318.90277, 240.1936], [256.63416, 314.01935],
+                                           [201.26117, 371.41043], [313.08905, 371.15118]])
+
+            # dlib: left_eye: 36:41  right_eye: 42:47  nose: 30,32,33,34  left mouth corner: 48  right mouth corner: 54
+            # self.face_template = np.array([[193.65928, 242.98541], [318.32558, 243.06108], [255.67984, 328.82894],
+            #                                 [198.22603, 372.82502], [313.91018, 372.75659]])
+
+        self.face_template = self.face_template * (face_size / 512.0)
+        if self.crop_ratio[0] > 1:
+            self.face_template[:, 1] += face_size * (self.crop_ratio[0] - 1) / 2
+        if self.crop_ratio[1] > 1:
+            self.face_template[:, 0] += face_size * (self.crop_ratio[1] - 1) / 2
+        self.save_ext = save_ext
+        self.pad_blur = pad_blur
+        if self.pad_blur is True:
+            self.template_3points = False
+
+        self.all_landmarks_5 = []
+        self.det_faces = []
+        self.affine_matrices = []
+        self.inverse_affine_matrices = []
+        self.cropped_faces = []
+        self.restored_faces = []
+        self.pad_input_imgs = []
+
+        if device is None:
+            self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+            # self.device = get_device()
+        else:
+            self.device = device
+
+        # init face detection model
+        self.face_detector = init_detection_model(det_model, half=False, device=self.device)
+
+        # init face parsing model
+        self.use_parse = use_parse
+        self.face_parse = init_parsing_model(model_name='parsenet', device=self.device)
+
+    def set_upscale_factor(self, upscale_factor):
+        self.upscale_factor = upscale_factor
+
+    def read_image(self, img):
+        """img can be image path or cv2 loaded image."""
+        # self.input_img is Numpy array, (h, w, c), BGR, uint8, [0, 255]
+        if isinstance(img, str):
+            img = cv2.imread(img)
+
+        if np.max(img) > 256:  # 16-bit image
+            img = img / 65535 * 255
+        if len(img.shape) == 2:  # gray image
+            img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
+        elif img.shape[2] == 4:  # BGRA image with alpha channel
+            img = img[:, :, 0:3]
+
+        self.input_img = img
+        # self.is_gray = is_gray(img, threshold=10)
+        # if self.is_gray:
+        #     print('Grayscale input: True')
+
+        if min(self.input_img.shape[:2])<512:
+            f = 512.0/min(self.input_img.shape[:2])
+            self.input_img = cv2.resize(self.input_img, (0,0), fx=f, fy=f, interpolation=cv2.INTER_LINEAR)
+
+    def init_dlib(self, detection_path, landmark5_path):
+        """Initialize the dlib detectors and predictors."""
+        try:
+            import dlib
+        except ImportError:
+            print('Please install dlib by running:' 'conda install -c conda-forge dlib')
+        detection_path = load_file_from_url(url=detection_path, model_dir='weights/dlib', progress=True, file_name=None)
+        landmark5_path = load_file_from_url(url=landmark5_path, model_dir='weights/dlib', progress=True, file_name=None)
+        face_detector = dlib.cnn_face_detection_model_v1(detection_path)
+        shape_predictor_5 = dlib.shape_predictor(landmark5_path)
+        return face_detector, shape_predictor_5
+
+    def get_face_landmarks_5_dlib(self,
+                                only_keep_largest=False,
+                                scale=1):
+        det_faces = self.face_detector(self.input_img, scale)
+
+        if len(det_faces) == 0:
+            print('No face detected. Try to increase upsample_num_times.')
+            return 0
+        else:
+            if only_keep_largest:
+                print('Detect several faces and only keep the largest.')
+                face_areas = []
+                for i in range(len(det_faces)):
+                    face_area = (det_faces[i].rect.right() - det_faces[i].rect.left()) * (
+                        det_faces[i].rect.bottom() - det_faces[i].rect.top())
+                    face_areas.append(face_area)
+                largest_idx = face_areas.index(max(face_areas))
+                self.det_faces = [det_faces[largest_idx]]
+            else:
+                self.det_faces = det_faces
+
+        if len(self.det_faces) == 0:
+            return 0
+
+        for face in self.det_faces:
+            shape = self.shape_predictor_5(self.input_img, face.rect)
+            landmark = np.array([[part.x, part.y] for part in shape.parts()])
+            self.all_landmarks_5.append(landmark)
+
+        return len(self.all_landmarks_5)
+
+
+    def get_face_landmarks_5(self,
+                             only_keep_largest=False,
+                             only_center_face=False,
+                             resize=None,
+                             blur_ratio=0.01,
+                             eye_dist_threshold=None):
+        if self.det_model == 'dlib':
+            return self.get_face_landmarks_5_dlib(only_keep_largest)
+
+        if resize is None:
+            scale = 1
+            input_img = self.input_img
+        else:
+            h, w = self.input_img.shape[0:2]
+            scale = resize / min(h, w)
+            scale = max(1, scale) # always scale up
+            h, w = int(h * scale), int(w * scale)
+            interp = cv2.INTER_AREA if scale < 1 else cv2.INTER_LINEAR
+            input_img = cv2.resize(self.input_img, (w, h), interpolation=interp)
+
+        with torch.no_grad():
+            bboxes = self.face_detector.detect_faces(input_img)
+
+        if bboxes is None or bboxes.shape[0] == 0:
+            return 0
+        else:
+            bboxes = bboxes / scale
+
+        for bbox in bboxes:
+            # remove faces with too small eye distance: side faces or too small faces
+            eye_dist = np.linalg.norm([bbox[6] - bbox[8], bbox[7] - bbox[9]])
+            if eye_dist_threshold is not None and (eye_dist < eye_dist_threshold):
+                continue
+
+            if self.template_3points:
+                landmark = np.array([[bbox[i], bbox[i + 1]] for i in range(5, 11, 2)])
+            else:
+                landmark = np.array([[bbox[i], bbox[i + 1]] for i in range(5, 15, 2)])
+            self.all_landmarks_5.append(landmark)
+            self.det_faces.append(bbox[0:5])
+            
+        if len(self.det_faces) == 0:
+            return 0
+        if only_keep_largest:
+            h, w, _ = self.input_img.shape
+            self.det_faces, largest_idx = get_largest_face(self.det_faces, h, w)
+            self.all_landmarks_5 = [self.all_landmarks_5[largest_idx]]
+        elif only_center_face:
+            h, w, _ = self.input_img.shape
+            self.det_faces, center_idx = get_center_face(self.det_faces, h, w)
+            self.all_landmarks_5 = [self.all_landmarks_5[center_idx]]
+
+        # pad blurry images
+        if self.pad_blur:
+            self.pad_input_imgs = []
+            for landmarks in self.all_landmarks_5:
+                # get landmarks
+                eye_left = landmarks[0, :]
+                eye_right = landmarks[1, :]
+                eye_avg = (eye_left + eye_right) * 0.5
+                mouth_avg = (landmarks[3, :] + landmarks[4, :]) * 0.5
+                eye_to_eye = eye_right - eye_left
+                eye_to_mouth = mouth_avg - eye_avg
+
+                # Get the oriented crop rectangle
+                # x: half width of the oriented crop rectangle
+                x = eye_to_eye - np.flipud(eye_to_mouth) * [-1, 1]
+                #  - np.flipud(eye_to_mouth) * [-1, 1]: rotate 90 clockwise
+                # norm with the hypotenuse: get the direction
+                x /= np.hypot(*x)  # get the hypotenuse of a right triangle
+                rect_scale = 1.5
+                x *= max(np.hypot(*eye_to_eye) * 2.0 * rect_scale, np.hypot(*eye_to_mouth) * 1.8 * rect_scale)
+                # y: half height of the oriented crop rectangle
+                y = np.flipud(x) * [-1, 1]
+
+                # c: center
+                c = eye_avg + eye_to_mouth * 0.1
+                # quad: (left_top, left_bottom, right_bottom, right_top)
+                quad = np.stack([c - x - y, c - x + y, c + x + y, c + x - y])
+                # qsize: side length of the square
+                qsize = np.hypot(*x) * 2
+                border = max(int(np.rint(qsize * 0.1)), 3)
+
+                # get pad
+                # pad: (width_left, height_top, width_right, height_bottom)
+                pad = (int(np.floor(min(quad[:, 0]))), int(np.floor(min(quad[:, 1]))), int(np.ceil(max(quad[:, 0]))),
+                       int(np.ceil(max(quad[:, 1]))))
+                pad = [
+                    max(-pad[0] + border, 1),
+                    max(-pad[1] + border, 1),
+                    max(pad[2] - self.input_img.shape[0] + border, 1),
+                    max(pad[3] - self.input_img.shape[1] + border, 1)
+                ]
+
+                if max(pad) > 1:
+                    # pad image
+                    pad_img = np.pad(self.input_img, ((pad[1], pad[3]), (pad[0], pad[2]), (0, 0)), 'reflect')
+                    # modify landmark coords
+                    landmarks[:, 0] += pad[0]
+                    landmarks[:, 1] += pad[1]
+                    # blur pad images
+                    h, w, _ = pad_img.shape
+                    y, x, _ = np.ogrid[:h, :w, :1]
+                    mask = np.maximum(1.0 - np.minimum(np.float32(x) / pad[0],
+                                                       np.float32(w - 1 - x) / pad[2]),
+                                      1.0 - np.minimum(np.float32(y) / pad[1],
+                                                       np.float32(h - 1 - y) / pad[3]))
+                    blur = int(qsize * blur_ratio)
+                    if blur % 2 == 0:
+                        blur += 1
+                    blur_img = cv2.boxFilter(pad_img, 0, ksize=(blur, blur))
+                    # blur_img = cv2.GaussianBlur(pad_img, (blur, blur), 0)
+
+                    pad_img = pad_img.astype('float32')
+                    pad_img += (blur_img - pad_img) * np.clip(mask * 3.0 + 1.0, 0.0, 1.0)
+                    pad_img += (np.median(pad_img, axis=(0, 1)) - pad_img) * np.clip(mask, 0.0, 1.0)
+                    pad_img = np.clip(pad_img, 0, 255)  # float32, [0, 255]
+                    self.pad_input_imgs.append(pad_img)
+                else:
+                    self.pad_input_imgs.append(np.copy(self.input_img))
+
+        return len(self.all_landmarks_5)
+
+    def align_warp_face(self, save_cropped_path=None, border_mode='constant'):
+        """Align and warp faces with face template.
+        """
+        if self.pad_blur:
+            assert len(self.pad_input_imgs) == len(
+                self.all_landmarks_5), f'Mismatched samples: {len(self.pad_input_imgs)} and {len(self.all_landmarks_5)}'
+        for idx, landmark in enumerate(self.all_landmarks_5):
+            # use 5 landmarks to get affine matrix
+            # use cv2.LMEDS method for the equivalence to skimage transform
+            # ref: https://blog.csdn.net/yichxi/article/details/115827338
+            affine_matrix = cv2.estimateAffinePartial2D(landmark, self.face_template, method=cv2.LMEDS)[0]
+            self.affine_matrices.append(affine_matrix)
+            # warp and crop faces
+            if border_mode == 'constant':
+                border_mode = cv2.BORDER_CONSTANT
+            elif border_mode == 'reflect101':
+                border_mode = cv2.BORDER_REFLECT101
+            elif border_mode == 'reflect':
+                border_mode = cv2.BORDER_REFLECT
+            if self.pad_blur:
+                input_img = self.pad_input_imgs[idx]
+            else:
+                input_img = self.input_img
+            cropped_face = cv2.warpAffine(
+                input_img, affine_matrix, self.face_size, borderMode=border_mode, borderValue=(135, 133, 132))  # gray
+            self.cropped_faces.append(cropped_face)
+            # save the cropped face
+            if save_cropped_path is not None:
+                path = os.path.splitext(save_cropped_path)[0]
+                save_path = f'{path}_{idx:02d}.{self.save_ext}'
+                imwrite(cropped_face, save_path)
+
+    def get_inverse_affine(self, save_inverse_affine_path=None):
+        """Get inverse affine matrix."""
+        for idx, affine_matrix in enumerate(self.affine_matrices):
+            inverse_affine = cv2.invertAffineTransform(affine_matrix)
+            inverse_affine *= self.upscale_factor
+            self.inverse_affine_matrices.append(inverse_affine)
+            # save inverse affine matrices
+            if save_inverse_affine_path is not None:
+                path, _ = os.path.splitext(save_inverse_affine_path)
+                save_path = f'{path}_{idx:02d}.pth'
+                torch.save(inverse_affine, save_path)
+
+
+    def add_restored_face(self, restored_face, input_face=None):
+        # if self.is_gray:
+        #     restored_face = bgr2gray(restored_face) # convert img into grayscale
+        #     if input_face is not None:
+        #         restored_face = adain_npy(restored_face, input_face) # transfer the color
+        self.restored_faces.append(restored_face)
+
+
+    def paste_faces_to_input_image(self, save_path=None, upsample_img=None, draw_box=False, face_upsampler=None):
+        h, w, _ = self.input_img.shape
+        h_up, w_up = int(h * self.upscale_factor), int(w * self.upscale_factor)
+
+        if upsample_img is None:
+            # simply resize the background
+            # upsample_img = cv2.resize(self.input_img, (w_up, h_up), interpolation=cv2.INTER_LANCZOS4)
+            upsample_img = cv2.resize(self.input_img, (w_up, h_up), interpolation=cv2.INTER_LINEAR)
+        else:
+            upsample_img = cv2.resize(upsample_img, (w_up, h_up), interpolation=cv2.INTER_LANCZOS4)
+
+        assert len(self.restored_faces) == len(
+            self.inverse_affine_matrices), ('length of restored_faces and affine_matrices are different.')
+        
+        inv_mask_borders = []
+        for restored_face, inverse_affine in zip(self.restored_faces, self.inverse_affine_matrices):
+            if face_upsampler is not None:
+                restored_face = face_upsampler.enhance(restored_face, outscale=self.upscale_factor)[0]
+                inverse_affine /= self.upscale_factor
+                inverse_affine[:, 2] *= self.upscale_factor
+                face_size = (self.face_size[0]*self.upscale_factor, self.face_size[1]*self.upscale_factor)
+            else:
+                # Add an offset to inverse affine matrix, for more precise back alignment
+                if self.upscale_factor > 1:
+                    extra_offset = 0.5 * self.upscale_factor
+                else:
+                    extra_offset = 0
+                inverse_affine[:, 2] += extra_offset
+                face_size = self.face_size
+            inv_restored = cv2.warpAffine(restored_face, inverse_affine, (w_up, h_up))
+
+            # if draw_box or not self.use_parse:  # use square parse maps
+            #     mask = np.ones(face_size, dtype=np.float32)
+            #     inv_mask = cv2.warpAffine(mask, inverse_affine, (w_up, h_up))
+            #     # remove the black borders
+            #     inv_mask_erosion = cv2.erode(
+            #         inv_mask, np.ones((int(2 * self.upscale_factor), int(2 * self.upscale_factor)), np.uint8))
+            #     pasted_face = inv_mask_erosion[:, :, None] * inv_restored
+            #     total_face_area = np.sum(inv_mask_erosion)  # // 3
+            #     # add border
+            #     if draw_box:
+            #         h, w = face_size
+            #         mask_border = np.ones((h, w, 3), dtype=np.float32)
+            #         border = int(1400/np.sqrt(total_face_area))
+            #         mask_border[border:h-border, border:w-border,:] = 0
+            #         inv_mask_border = cv2.warpAffine(mask_border, inverse_affine, (w_up, h_up))
+            #         inv_mask_borders.append(inv_mask_border)
+            #     if not self.use_parse:
+            #         # compute the fusion edge based on the area of face
+            #         w_edge = int(total_face_area**0.5) // 20
+            #         erosion_radius = w_edge * 2
+            #         inv_mask_center = cv2.erode(inv_mask_erosion, np.ones((erosion_radius, erosion_radius), np.uint8))
+            #         blur_size = w_edge * 2
+            #         inv_soft_mask = cv2.GaussianBlur(inv_mask_center, (blur_size + 1, blur_size + 1), 0)
+            #         if len(upsample_img.shape) == 2:  # upsample_img is gray image
+            #             upsample_img = upsample_img[:, :, None]
+            #         inv_soft_mask = inv_soft_mask[:, :, None]
+
+            # always use square mask
+            mask = np.ones(face_size, dtype=np.float32)
+            inv_mask = cv2.warpAffine(mask, inverse_affine, (w_up, h_up))
+            # remove the black borders
+            inv_mask_erosion = cv2.erode(
+                inv_mask, np.ones((int(2 * self.upscale_factor), int(2 * self.upscale_factor)), np.uint8))
+            pasted_face = inv_mask_erosion[:, :, None] * inv_restored
+            total_face_area = np.sum(inv_mask_erosion)  # // 3
+            # add border
+            if draw_box:
+                h, w = face_size
+                mask_border = np.ones((h, w, 3), dtype=np.float32)
+                border = int(1400/np.sqrt(total_face_area))
+                mask_border[border:h-border, border:w-border,:] = 0
+                inv_mask_border = cv2.warpAffine(mask_border, inverse_affine, (w_up, h_up))
+                inv_mask_borders.append(inv_mask_border)
+            # compute the fusion edge based on the area of face
+            w_edge = int(total_face_area**0.5) // 20
+            erosion_radius = w_edge * 2
+            inv_mask_center = cv2.erode(inv_mask_erosion, np.ones((erosion_radius, erosion_radius), np.uint8))
+            blur_size = w_edge * 2
+            inv_soft_mask = cv2.GaussianBlur(inv_mask_center, (blur_size + 1, blur_size + 1), 0)
+            if len(upsample_img.shape) == 2:  # upsample_img is gray image
+                upsample_img = upsample_img[:, :, None]
+            inv_soft_mask = inv_soft_mask[:, :, None]
+
+            # parse mask
+            if self.use_parse:
+                # inference
+                face_input = cv2.resize(restored_face, (512, 512), interpolation=cv2.INTER_LINEAR)
+                face_input = img2tensor(face_input.astype('float32') / 255., bgr2rgb=True, float32=True)
+                normalize(face_input, (0.5, 0.5, 0.5), (0.5, 0.5, 0.5), inplace=True)
+                face_input = torch.unsqueeze(face_input, 0).to(self.device)
+                with torch.no_grad():
+                    out = self.face_parse(face_input)[0]
+                out = out.argmax(dim=1).squeeze().cpu().numpy()
+
+                parse_mask = np.zeros(out.shape)
+                MASK_COLORMAP = [0, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 0, 255, 0, 0, 0]
+                for idx, color in enumerate(MASK_COLORMAP):
+                    parse_mask[out == idx] = color
+                #  blur the mask
+                parse_mask = cv2.GaussianBlur(parse_mask, (101, 101), 11)
+                parse_mask = cv2.GaussianBlur(parse_mask, (101, 101), 11)
+                # remove the black borders
+                thres = 10
+                parse_mask[:thres, :] = 0
+                parse_mask[-thres:, :] = 0
+                parse_mask[:, :thres] = 0
+                parse_mask[:, -thres:] = 0
+                parse_mask = parse_mask / 255.
+
+                parse_mask = cv2.resize(parse_mask, face_size)
+                parse_mask = cv2.warpAffine(parse_mask, inverse_affine, (w_up, h_up), flags=3)
+                inv_soft_parse_mask = parse_mask[:, :, None]
+                # pasted_face = inv_restored
+                fuse_mask = (inv_soft_parse_mask<inv_soft_mask).astype('int')
+                inv_soft_mask = inv_soft_parse_mask*fuse_mask + inv_soft_mask*(1-fuse_mask)
+
+            if len(upsample_img.shape) == 3 and upsample_img.shape[2] == 4:  # alpha channel
+                alpha = upsample_img[:, :, 3:]
+                upsample_img = inv_soft_mask * pasted_face + (1 - inv_soft_mask) * upsample_img[:, :, 0:3]
+                upsample_img = np.concatenate((upsample_img, alpha), axis=2)
+            else:
+                upsample_img = inv_soft_mask * pasted_face + (1 - inv_soft_mask) * upsample_img
+
+        if np.max(upsample_img) > 256:  # 16-bit image
+            upsample_img = upsample_img.astype(np.uint16)
+        else:
+            upsample_img = upsample_img.astype(np.uint8)
+
+        # draw bounding box
+        if draw_box:
+            # upsample_input_img = cv2.resize(input_img, (w_up, h_up))
+            img_color = np.ones([*upsample_img.shape], dtype=np.float32)
+            img_color[:,:,0] = 0
+            img_color[:,:,1] = 255
+            img_color[:,:,2] = 0
+            for inv_mask_border in inv_mask_borders:
+                upsample_img = inv_mask_border * img_color + (1 - inv_mask_border) * upsample_img
+                # upsample_input_img = inv_mask_border * img_color + (1 - inv_mask_border) * upsample_input_img
+
+        if save_path is not None:
+            path = os.path.splitext(save_path)[0]
+            save_path = f'{path}.{self.save_ext}'
+            imwrite(upsample_img, save_path)
+        return upsample_img
+
+    def clean_all(self):
+        self.all_landmarks_5 = []
+        self.restored_faces = []
+        self.affine_matrices = []
+        self.cropped_faces = []
+        self.inverse_affine_matrices = []
+        self.det_faces = []
+        self.pad_input_imgs = []
\ No newline at end of file
diff --git a/utils/file.py b/utils/file.py
new file mode 100644
index 0000000..9cf5d11
--- /dev/null
+++ b/utils/file.py
@@ -0,0 +1,103 @@
+import os
+from typing import List, Tuple
+
+from urllib.parse import urlparse
+from torch.hub import download_url_to_file, get_dir
+
+
+def load_file_list(file_list_path: str) -> List[str]:
+    files = []
+    # each line in file list contains a path of an image
+    with open(file_list_path, "r") as fin:
+        for line in fin:
+            path = line.strip()
+            if path:
+                files.append(path)
+    return files
+
+
+# def list_image_files(
+#     img_dir: str,
+#     exts: Tuple[str]=(".jpg", ".png", ".jpeg"),
+#     follow_links: bool=False,
+#     log_progress: bool=False,
+#     log_every_n_files: int=10000,
+#     max_size: int=-1
+# ) -> List[str]:
+#     files = []
+#     for dir_path, _, file_names in os.walk(img_dir, followlinks=follow_links):
+#         early_stop = False
+#         for file_name in file_names:
+#             if os.path.splitext(file_name)[1].lower() in exts:
+#                 if max_size >= 0 and len(files) >= max_size:
+#                     early_stop = True
+#                     break
+#                 files.append(os.path.join(dir_path, file_name))
+#                 if log_progress and len(files) % log_every_n_files == 0:
+#                     print(f"find {len(files)} images in {img_dir}")
+#         if early_stop:
+#             break
+#     return files
+
+def list_image_files(
+    img_dir: List[str],
+    exts: Tuple[str]=(".jpg", ".png", ".jpeg"),
+    follow_links: bool=False,
+    log_progress: bool=False,
+    log_every_n_files: int=10000,
+    max_size: int=-1
+) -> List[str]:
+    files = []
+    for img_d in img_dir:
+        for dir_path, _, file_names in os.walk(img_d, followlinks=follow_links):
+            early_stop = False
+            for file_name in file_names:
+                if os.path.splitext(file_name)[1].lower() in exts:
+                    if max_size >= 0 and len(files) >= max_size:
+                        early_stop = True
+                        break
+                    files.append(os.path.join(dir_path, file_name))
+                    if log_progress and len(files) % log_every_n_files == 0:
+                        print(f"find {len(files)} images in {img_dir}")
+            if early_stop:
+                break
+    return files
+
+
+def get_file_name_parts(file_path: str) -> Tuple[str, str, str]:
+    parent_path, file_name = os.path.split(file_path)
+    stem, ext = os.path.splitext(file_name)
+    return parent_path, stem, ext
+
+
+# https://github.com/XPixelGroup/BasicSR/blob/master/basicsr/utils/download_util.py/
+def load_file_from_url(url, model_dir=None, progress=True, file_name=None):
+    """Load file form http url, will download models if necessary.
+
+    Ref:https://github.com/1adrianb/face-alignment/blob/master/face_alignment/utils.py
+
+    Args:
+        url (str): URL to be downloaded.
+        model_dir (str): The path to save the downloaded model. Should be a full path. If None, use pytorch hub_dir.
+            Default: None.
+        progress (bool): Whether to show the download progress. Default: True.
+        file_name (str): The downloaded file name. If None, use the file name in the url. Default: None.
+
+    Returns:
+        str: The path to the downloaded file.
+    """
+    if model_dir is None:  # use the pytorch hub_dir
+        hub_dir = get_dir()
+        model_dir = os.path.join(hub_dir, 'checkpoints')
+
+    os.makedirs(model_dir, exist_ok=True)
+
+    parts = urlparse(url)
+    filename = os.path.basename(parts.path)
+    if file_name is not None:
+        filename = file_name
+    cached_file = os.path.abspath(os.path.join(model_dir, filename))
+    if not os.path.exists(cached_file):
+        print(f'Downloading: "{url}" to {cached_file}\n')
+        download_url_to_file(url, cached_file, hash_prefix=None, progress=progress)
+    return cached_file
diff --git a/utils/image/__init__.py b/utils/image/__init__.py
new file mode 100644
index 0000000..3bbe064
--- /dev/null
+++ b/utils/image/__init__.py
@@ -0,0 +1,26 @@
+from .diffjpeg import DiffJPEG
+from .usm_sharp import USMSharp
+from .common import (
+    random_crop_arr, center_crop_arr, augment,
+    filter2D, rgb2ycbcr_pt, auto_resize, pad
+)
+from .align_color import (
+    wavelet_reconstruction, adaptive_instance_normalization
+)
+
+__all__ = [
+    "DiffJPEG",
+    
+    "USMSharp",
+
+    "random_crop_arr",
+    "center_crop_arr",
+    "augment",
+    "filter2D",
+    "rgb2ycbcr_pt",
+    "auto_resize",
+    "pad",
+    
+    "wavelet_reconstruction",
+    "adaptive_instance_normalization"
+]
diff --git a/utils/image/__pycache__/__init__.cpython-310.pyc b/utils/image/__pycache__/__init__.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..ba710e5d9158be9b211a2d294dfb0e05db775cad
GIT binary patch
literal 591
zcma)(L5tKd6vvZkr|nETJA3l##Y-<%1_c*Hgk4z>6tM`BOUNXdZZt0`X}Yre0X+I4
zy!n-K_2gIZqRDtzj|TGN<)6p@<-OR=W{u+d^82e?EfD&V4*#yGp%-yaA0h-2SR=p$
z;}kFS0y2^v<4otUBugkr0VOG+A{DI23f5!|HK~s|xo%)XHgG~tFdC3pe#2xdiXOG)
zPaN%A9I6X-IJ|oM@-oHm``7&yb7NbEtz$-5pxoJ!GUvjUOXH=x6G}D@5VcSpl#VyM
zi_o}Xzq{dk?&#=4&1P?@lasoYAsd-!7vbJ#b}hB^QRdc6-p$-AYr=gd*yz=@q$;Y+
za7m4Iz_hw$DWR=h8jQ;l@`PnVkx(X739E!OURNhH5pUX4Qi<5($59SP53-skpi}De
zXcr1@0j&8CTT;gd4Tf^7t-GrcN=lj5lzvBUn+hlVy6mnj$j(?V_tt*uxD~P+rJJl_
zy6ev#J$|}7d-7L4KkwglvsZfR6tF>dG|oYJN*|6lp?U^Z%vwHAUpA%o8r0wi_%v_w
EKhx2f$^ZZW

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/__init__.cpython-37.pyc b/utils/image/__pycache__/__init__.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..68ad810bb1a1e8f2276d0a638746e66f52fa0da3
GIT binary patch
literal 624
zcmaixON$gS5P<tJ{hIEc9uYhVUcB_hhCPeOvWx7ZA`XJ&5;CbwPu4t|bb4j?AMM$H
zldC8Hf+v#!S#MUzM^&Y$N>!b=?K<-E^~Y0IB~kP<7=NYq%3FWu8xM#e7Lklm90xrS
ziA+&?(o?aL8Omgia#^54mZ+2!s^l82WsT}n%t|z}MXlVRjkhz%p@0%9SVIjBw6KA0
ztWgKsUDQAM75jpH?8*y%I9z|ad>8aDyAQh~F;ic-&JYD!Vruk+i7~FDOj%|g9R*oN
z>3yy|aN%FhE?i^A{n;JeQ-de#YO+|34V(ErbLm8&Pu#Qb<dz9$y_;$^TeDEhwQ?^A
zkje5}hPn3=MHyDwNFw+>2?53rE;4DrDj*BU1B!q$pb7}1Hesr!@%Y$pQ3<}on~9C5
z9H?4miRaMeDV@uymeLpae_B-g?~DplEwp*8;R=k25Ews1rVE7weqRqqE!m*7WqYl^
z4L~dzMx&Q$*e&WkS0g<#HBb#~2rD?h`aCR_i+SLZjBLPu5Zq$?YPxgfYpG!o>`mC&
U{M@67oA~8wn<YWwe~!2Q09y>75C8xG

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/align_color.cpython-310.pyc b/utils/image/__pycache__/align_color.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..975b7fd406f049107c89db9404a0c8b1b9d3f4f4
GIT binary patch
literal 3716
zcmcIn&2JmW72nw}E-C8MvMk$mv;9C|wvlMNF<PrZTvtX52Zj*Tx^&T&SaOEsO3Ph(
zW><~`GC*n<=*9gTq~3h%rPrc=La*6t3Iw{f$fYPyq`x;y%CsCKmv)Jroq6-->%HIn
z-eWpHUo-GL`|=OslV!vBj2g3#jmA&#s=uJ%MqqHp&6ouYUn@2TR$wub*)bc~fvxLK
z><rwXqU&z#4XQzPPz!2kSA@kqVRda@<+W#aFgGKa<8?^t{F{71*j<w^@}*}^Fwd8T
z%~v3?`31fTjRk%Yk~Ogi$tAuH$r68yWEgXWU&WZ^nK9qs*PyY&%|}MV53e!9sC~!Z
z_#Z!*()YL1IF)`k`bcnJb!0Tm{H{y~{=>-sW#nJ)XW39~HJiOC>*wvwPC94~_wzV9
zjBeeydAq6j#uL%bqYagX?N}(;Y{zN4IS5rIWK+pO%JWz>JG2Pa*c=`<{vQW9g_n(;
z1}mHggRm!1ad$<cQi)ovn<t$tN|P{dn1#2SmOVvnH~sLzL#<!9StxrVD_oW4vLiYs
z%{O^Gd>y>oc)xgvA~T*EV{>GVjHm1vfI4Ag3oUlmvNLmJjqIo9*vTyDx+8bw9NVOM
z>~ND?C+1T&u8b;=j2HHfgmaA^Uj0le|G3+5C9PUi^F%$)g*Xz$TOkjJS#%&;Q387>
z9nnfsISAwED5L{Qmj+a(>TV|v2gBA2PgWq8bdI8$rL9a`rkF3!eAX<Fe%6#i<#9G?
z1|G&mWZIxU>CmkghFrz8|NfxaPY0rzq?u@^=|23;MRO>mBC9ta-MR7Zk8a($U5dB2
zAN``4XHl%0dX6Uc-%DB=BEUznh*L(Bb_<wUzye=T;hGC<fz>S6T4S~;*C2YMqhuAm
zk~m^wM>&UjvGyc95V6QwQgl+-Q05vL+Hi87hR{NCkqSCkSzJV|Se_2m8>Nk1BZW&;
zTt@NvKl`8k{?GsV`}>966=C+;&d4_eA@vN!$T&7m0HTbI00qwWO||fW!BPJ$B;4Zm
z@65}_ugtMEwxO%x#GMms4?Wy{&PME#F?I-4C)3^vubkL`+O$pDPN_W^Uupn$BezSu
z;vIdu9mbtJ4m05oM40$t0uIQy-vuLl4c0bm;12(u?5QnXqZXapzdpHmTmHNE{Ss=W
zE@_CN@~?N}G|U=X{ykp-yRm-|#<{?H91N7De%2Qd$%ybx?eHLt_}LmX3y%&WE;ibS
z8%IK>oBn;#4T04y|EAcuT`qP?0pRHEUCb3F!yKH3z2#E)s-NaD_uFD-dpoCe-%r29
z1HQ0$G&rR@*>DS6MYwlXERv!E-1oyFTG$*q9?Pt-$!ZNxUWImH;c)U>RAcRebpnr;
zY(aTagJFZJmeHv5@S5+y6)oy%LGc|-qX>##WtA<lD{Pg0n|VxrkEX8!x-e|kqREPy
z$_`_p>#t#sjG!_1Au2PN167sS0{%c^s5^ul;0d&L!q$zETh^``L{lTXY^@q2tE^oz
zfFBkv!A$oioCG_UoaDd4NfVs#y<BD5wNiA26y)29OB6Y|k9(=RRWyj=Q2P3oX+TDC
zoDZTzGn~>m2r;CeDsrtJ!$nU5u0Zj0%CAD_tkq^vN)UN%PcISs(z@ryylK&fr9n`-
z$`v0Caj5{2@52L<+^TzniPy;ALA6|i@Nwo^-7z=aGIcFZ%__Kd_BjB_I_CZ{UPYwm
zd39E2>-cGT;b^8UT$m}!)hjrmAalk|Lo|BGi~h(^n!2RA1*S6gz)T}!k6w#g&n*N!
zu+e3Mz6fp(cj*TXvNC6n&;*vEAM%|F_n>b;UV+@=&WQ2qa~s@d14p%^n<beHhw-65
zp;^Bj=W<F1<rSVKlZN#<wu<LZcPEv<-jz4dhWT&eGYO-AyQ^D>`wq_zad}-X4x)}|
zm^(YC7WAZ#qNqwSjH3<^g!^wJXMsZD_GFq5RpD^}lEQ1lE1WEqoj%e99YW<Tn&AgD
zgNqsHHnN`;)-dFS3)dao;tfY`l1>vvFt;ycB4SWNtO~bm6mFVl;HPumEb7x}LuPn@
z(Tc*6TXC7o?EkDfp1EYMvlY{3@*X-)$&0VJm0Wsu{eOx^W*p<PBUdqzoqz!D8^69V
zwm2JE7mTr;S#XTbSjfpEXWjTA<v7mRcwB|V(E-Ce`s6Dd#Suye!&F6Nqc?kOw;w5g
znu_^p0#{1`N>1o93qMYuAnz2&bdt`Y|F)L(qh9}|u)LF!qHe>Mckn4{#Q&-lNBg31
zWSI1Xwgy5zS-o&$foQLSDoxa)*$WnxzVNV~mYg^I#i`+6Io5A5o#x}3QwI;%aJgLO
zneyk5zh-s{CYp7>i%JIG)eMVR{;`Et$-Va2`Pe4ImP~pSew|_8@fC$e1X!}xgc$rM
zAjcW*3<uOj`Mjn`L_3NNrQsS(?-sT>qoQDWN-$>z3YMm#seyvUlAd0wER=$&!Z{1d
zb(}iJtBI)Yn5=G^WGj+m6IaR)L9%h7sI*!<?X+5@KP@EWlFTRRZWc&7aoaYNL=uwa
zdnkf6Ejx%5z_yve=L*TnK<Qk$Oyejo)ESwk;FABWD9JQ5v<>uDX11*}J)Puv@0EX8
h-lzDlXsfp2xx}S)gy9uyWpQb7WpQO;>EhbLe*sGX%-jF~

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/align_color.cpython-37.pyc b/utils/image/__pycache__/align_color.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..59bd49599b4ade5390e2131fdb1d921c73aa5d44
GIT binary patch
literal 3843
zcmd5<&5s*N74PaVf6SLBGf5_!&29@&6vNJV6R{u$l#oqWDN-1%*pUDWu-dM&-822+
zR8>zhBYUMKLE@7CU=K4#Zb-nLKY=TKT_iL&xUKdAzgKNLGecH;0SQ}OuJ`J7eZ2R3
zzj}UuX{m1E^Rqv!iBDE7>oXE&9~Ye;;!%G@!>zW(8MkBBX83htJ8{|$Bb^(w#BIB#
z@5Nr?w=1Ua$JM0Pt|j$$oms*W4zGM_^D3{saN7%Wxy~0LFY-5q%a=Z7e3`Gj@Y+kl
zh1`JsD!+vBWymXh4e~m_47tH?VEiiVH~2N!ug>k?;MXyJiQ6Ap&9}lE%(CiV4>teH
zA50kp53)FuK|gvXc%XVR8tI@fvm|&N1wW61TSKi!YP;1MM0%Kaw|ZI98tvzCbQs-v
z_x5`&#W$adZXRta9d=`(WUCuz-BuDREo4i{B;$E3T0L3>Yix}UoBxZ0oFcZ?PLmbh
zqa+*%bo^bBs!XC+@8@YxM_C%iO}nV>X5~mx-_1UJ^tiOQ3txwFAavoYESEjevuQ?r
zs(3s+5AdkRXted*n%HA|Y&~bk0MQAXIFQ&}a<x5n#_n@_;%NuA{@5RT$1YhOd)(&E
ziT#{SD&q=5>+VQ|)*9e3pK0eG_nN+>Rf}4lswcS+N1}K;<l#t12ci?D@OIi0oivk4
z7)M7T9Z&l7LS<&|_u?=ab(&RKfnL%%iki+k+W4hdD$hF?%VW<)DO4WonOMbp5jnP4
zoDLZ2ca~hkv|s$VHOvyxN;56GS+?Kep$=OEndPI_M|t`<N(a56NC5&aS`VXs|NTcl
zY2`YKRm)7%!Vv~($6y2iDOPaMXexIAnq{n^sIhFHEwg2(?)dDAEiXg!e-bBI!(NIj
zPs0Nd3*C{Tmmx+nH@Gu_kqfYpi)hY9L@tq*&L&r=X;1?Ov*aae$h&2;j$W}ko8P#K
zTyq2J*SY2@^nd^7@U!3k>GQwcMGm-q5$bckBCr@tzd&bf9a|?rwPs^rle2wWy{T=E
z{;#3o4tIZJU$cH;Pn?MhTZiz?y%T58<u3PMvN5}AO+297JDH7Ec;&<e_GdC#`=#~t
z{n7$?F!uXIbJe52J_zGp9*0^4i3rmmOhHMS2YnFKSAc)34k`=o%YoW9J(B3m!L2C|
zYzN<d7?jvAZOKB6RB)>wXQ6Ix2loSoWQv1>FwO<m;~+$x1$ro;k`W~w$>Bj3@%b7w
z3y%&WE;hS|n@2)sTfsxo50O0E!ELemUb)yQ-9kq<cQIF#j&e{e_LfT#s9~1JJm`uu
z+uJ!M@;Ut%J_wz|qvVuGqFE_i6@dtxSfs`BeWiq?y#JAunJg+u%wagfC^iejCsG$K
zc@c2ChPMj`hm~(rk97;yYgcIt9T-o0aCEz7boACm=q>dSnhuS0VE7KEDPdQsvUS#A
zZ=n4elTF&DT`U6qaJBBxbVXh1!&sR9IRuhR3<5ty=L`bDq$SdTMsO_p9^nY+0_&cz
z4QuR|z3UdS=-4f#b!+UDy{i_`#KH!gc7IBl@Onv^!8yvDqK5C~N}E7S(HBxsq^AK=
z6!Ia`+6=2GiQ-TOCjSgPqd3l!C^ZDE3@SprGRzbO+PuTXKmxqL^lZx4VfC!X&R~`R
z^B;S9h2K}6z3|SLE!uPp1Wcd`UpyNj#Q>0RA_DST)DS@0(}+>D+qH5HLdQI;W?*5q
zWg1$Uxm9oz=nL?XRGxhwk1};ywaOOp-C#x!z<A~5;=@r<uI4DA8*t832N0-q<80v2
zEy`{DPUV3E?y~m4Rby+99x&ER2lukWy=xZz;{NqH7{{=ZqrE*s6nKlieV58P*jvz7
zpm(@GX1xB=1>3p6(!$a0l3qrm_%N7~aL|o&IU|TNsn62O^d|rPy=mvqv(8`c(Sf->
zM6UW=97H|Qw0Cw+9h^_Tg{G)UF^Z!eAVbWJ3JL~=Z|+l7R5=hxY2U0Cp3Y=%h|<Me
z>9R#Lyh}6qn1ND{eNZ@~kQY8;b#R9_J-JCXTWHz~`$DE71_Q*Z@Jpfavs{CoE`+aG
zoLxM0=0|v2(e-2@`4LHb-LBdVwqajl@-DQe6t*TaOU*+`VVo!Y2N1M%jPyqFWIwxu
zf(c3g(E8=g2|_w{)~t!E9R$J@Q54j%w_$yk$}VqWJz4*$1)XOu4g2Vm*Mx@S{!K=i
zipWu49;4l1q=H#>7-T7eD+Opdr4$xHoIOP;EKq)>y~E(9(G8=)@RhR6M0uSK?8@)p
zr>N5%pgM81FA7hF=|C8#pq?Yw7k(^IBB*waChE}a1&hjnRI#4XT=f3snfK2f>sOer
z<Y2Fks;J89NU5qVA3}f5?R1$Kp8W<o8cb_A6?gf^4y2N8-HG?HOHM7h^C<i(r@j*?
zx*Bo8lBcFr5IhAPo?*){p&|NvVMr0T9o+<_(*O|KDf}`gpZ4mEQs&NSH)g7tbJ{B<
zA-ytLE)_F_3*M7guu6=lq#kdP*)}~!k4<dQCYZcOKWjy$)8SdK(<x)=;QxUn_eo05
z0#z{5Z7WSBAz6M8P5X+`9YhL1+tP@)La~x4Q@odTA62Q6>M=vQhVcp~O0|K7@qyXO
mnQfb@&lG&sAC&)}?ovuA+Nx`p+_}gaPGiM<mm8NG%l`)P!|QSY

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/common.cpython-310.pyc b/utils/image/__pycache__/common.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..c4ce7bb543fe9bd9e158e20bbd578c60ea7e2a62
GIT binary patch
literal 7506
zcmb7JTWlQHd7d-3on0;|il!-&lI5}NxZIQ?DJF6xRgq;;Vywib4T-kwwbjjX&+LWV
zomromB~4<sbs`y(fjVfArzA+h^-H1GB*;@9^i=ezK>Jc)(1)N6qVOp#{LrG8sNa8P
z_Cm69&{@ns=YILmfBDY8-tcfn!E^WV_nf+|DE~s4!N)}AGG6}|AY9?BuH@q{%QNKF
zy4uk48jH*Hydm3|c~gE<`4qld-D;%s8MM*4aaYL?@sr%*>ANhS<<IggA4Y1JkMJX?
z8{sw|<wsF+gpcv3P%_HL`P29w<r9v!uJL2e7@wq6+2m7v`mUaT3N^<1I6p3HkV4G~
z{tRlK=Io}Ddu|VFS5|T?GOjm@6$gn-x#gCEn&%elC6#KE$H13f4X^(YBv3+SoqZ1*
z581e~VTe=6tAWN<uHBV&SCu>L4h!^kgX^EF?L$5MbjZ4+99KS}KInf>DKl<h42#C^
zjbSEZq;^r>VQn>3*VEjbQbIO~HB2jQO_rnvO7u_-HT>8y<zqF}@sp{tI;CLq*2*4j
zu*5*R4;9EA;`{j|$ak=++o?A8$dzg3ZG|ZxsKmMls?brr@&Ktwom+_5ha!anN&fTN
zeAR0>^R5>-Yo50;U-Gy!-*kjeY@NS(e&MU<7SErJi<g#e-k5I%wYooFqccEhqv5%8
z&Fv_CvEsPShfQ&5jMhUFsvk`#JEuyH8#tm+5?-@V6k_gT-76LA{-wEIhX*XOgyR<*
z&AM}FhcwVvXv7mBa);?QodmnDjN{CbU4lms*a`+Rgx7Mpu+W%tKBjWw%?_bRWI>MM
zP1ZN92;%4-kz6Ly8Z|f4{o0NbnOG2gji7~93F396U3uf3D0Suf(pz{V)on%wHdNfk
zG}U6$$r(|)S*sT$XQGV!6yzWpnnkAMz;^OYPGB743R+oo<N%*&rT#R?n69%4W<DN!
zJYwi<8mWvv#>UtQCSFAKqaXc{p1qeJA%o{%UoH{Lx)0ffo+|tYGJ&!S0SUF^B$#@V
zQ)SKw)LjjNXK;vKM{S#d4zV_%Hfm@}wK<2p8Kk(zb*VN+BJvi*+mLE=pojl#*c&BL
z8}yfI10|7a(;p*~j4`yEm1>ic>q)H-^PZw$U94^O%2`)|WO+JFg%*CX3Z+A+EGbh@
zg#?|vRE0xQ*3~0Hyb@$CW1%5A>g$R;DJesLOy$EMJEug^<zC~9<XeoO;h`zJf-OdA
ztfNFobEN1BDHl>ldt$pll@mmyw2nl2qZm{pz2wyD5Lda1C{16y6-SUo5K}}*io|gu
zB*uaiNjyX3B#~!{*dV!#c#hJiK=Q*-g?LY8Q7j=_{BR&UPPKlcSchs9c&#DDJU>p2
z4wa4`ED_JsEH4oG5)rx00Yx~#P+F63gN(2dHp<4J1P9e&l1-92oJOP7<q{oX_n|Z~
zTAuGeBXM8#QV@Ej13SE-d;-ht9T~{O3VX)}^4dN1o_0?!$DYFaMCs+p30j}9fQ3q+
z-ckL9P^qxCl8APSM4OZYo<s}r<<^EOmO_Q6r<Ar5YCO|oqKw{}?EOZlp|^IA!JiCq
z>aEG%8hUF<?=|$I-WV^`%IZxe)P)yfNB8ynS~)HuUy&^~u-#W$l?E*D1-t5r+K%T2
z#kyUt*P6cl#tmC|K{0UbnT1*VJl-?sKo-BYU~{J;9LLYiWimFNTaMlG9edLeK@Emf
z_O~Hmc7QYTn;z=Qo`^?-7q)9;!D-eaIpsMy#w+!@9aJ6E!C&}x&6T-&%@5?*R0YfL
z*u{9gNzb^esQ4FTiWXCARD64eI=<&}X?Jhip`9di7wqft1c7I-^+t2}oPE6<&+j%{
zL2ocQppU(D2bi^MfxT73V$oY}4?Al(*s5n2ZJIUS4t1;&AMF`<%R1(Lbw7XS?5j?>
z*s6o`tD@!fdr6)Xm%1Hq6azX2;?$G+l<VHs!A)PVmu2Y}Zq%;P=Jstj;ne?UqqNzp
zH5{lfPZukx&e~fLmE6FJS3Gxcmrl?UunpK)3+)pg-H@g79FiKZ+?S><XZpO76L;CT
z477k5<YwZ52BhK%@^HaU<R2%PNFK?p<fjw~Ay?$^eJp~{$HDow=hnCFYH`y+%CVhT
zrf2Q-7KAqL*cF>(S(Vt{{OaX>R-WuM8K9@|`WHa%D<UmDE&)-cqcJG27)wgjq=bp%
zp|T6$M&Pug!B~(|>npMMqA{fld!R*HX>&2sNf@JaPa^?~HK+=z4JoIDh=TJAk+De)
z`%oyd3W>_C((=ef=`gVUcV{Z8?<2Nuk0<Z)FXLTK?%yTM9)2II302FQp-kaMaUOR0
zBJp5hKMxMHX9wCdvE1X39&>2l_~2Xxb4Wu=yfX%PxAz#2WT2SCTEQ(a(AyFOF$b74
z+}&z{YHy>`Cl5v3IVpJ*(`lr4PCk)OvhzM>kpm`wL?ZDyjaVVfZnlADL)<^UuYLf#
z-Ul2xRix2KEP~{i_!^~qs36K*B<TkPZ(iC!&F3`uxUzHn(0cmq`a-nw#C3^3=}kj(
zb=Pp7(oB#X_tzjUQkLGYQ?)J#7y<kGH5ZERmYhhZHAY(9ae<F@GUDQ^)W+&+ePl?j
z%n!$!m6#4WMb3z~Ea#+N;w+Juh!8G{Qpr}Sfj<K>Ixz~2FshEGGg%WLIm;|&sTtNw
zGhKy{tKupe@4GV+pJ}^Gc>QUpS;%&wrt<bC@xuFhTWjlWqiu#NS0^C8@aQuETZiIO
z$@Kq0%Q{__KZO#zI&Pch9rj%)zC<4>Zhefq-TP`g)dpn6T<H#zW8Ov!9XCOJ9CHw)
z?dUK9Y*z>D#=S9Akq7QrH!v^bX`cC1i`$1rXm)hq0`x~qIl^Fn;D@CDu8EW!ucHsl
zbztmd4MPXV&t-R7Z*H%8qEsbpAe{s3IjqrKxmL$zZ}DnutzS9l!0nk;I1^96ZO*LC
z+ND{W97?Wdgf|@FB7(8M)yAy75jPR%a5%B@5|ocp!YSj*7?2h>aBbZ=dP{h2#ZHEy
zyVu_Dq3<g(WM)ySS>(LtR?xw$;HKF0b18|DBeooAsEjgi7wawOnh>6d^v#;Hh5MT4
zZZ2}V5jLG-kV{K#jLdkFh;2k{En=mJRU@_)sm&_ht-cD!f@*ac^Jh}LKr*RONavR!
zo#W~V-ccafadnp2OuU2YeX>bTgl-~IHXkApGVvucnVeO~*g<Y8tfF<;E;$jz8&C(7
z2q}Q{ag(Pyz_lG!=8@w@N9WctAW8QujZ%T^eMU4R=@97tv8XQl8FXRSKyMlWBPMgv
z74i?v73wq>q;L4R{K6!U6#ok<wgk=sTiQ3iai!~}fok4eT3Zs`{LBGC>q67(0a_$z
z)0e>F`s!P!Z`fB>=U!SkW7C->l1T`aY!NUcKh2$T&8-pUvIF?_O&BVIUpF1cZg|20
z01k?XZUClAnXCpu)4wo3@3?bYwT)WS;kDwNCo1!lo0kiHUd~&H=PmT-O_0GCsX&5Q
zw`xJvmJtMd#+|jlk|@Qj+?Bk`+3yiZzdcL%r5Magei1w^4uDd7yC7jp7joUII>6R~
z@T)9OaLc=PsakZQWCIB93jtPN7vsGJeMBV}E&)7+KjOvDpZP<}k5S&E(Ra@O{BQsC
z&C8GUw?|I@^ys%QKl-cR{q^5`>)+2^mdYx%d=K{+MY^m6U@ctEY2sZJR(=`$U{e3h
zm4}az!Sg;<<dLX+cI_9R|Ic6Ccv$)G2mkgDuZvp`QSlxs1Ud1Fob{ze6z<Wrd+#Hv
zDNqewMfsI?IcFl~L#8b<18J)wHSi)6yl%AWFd&X-v|!?1M{Bqza?E)%W_@ICIkieP
z$fvwkfGAKU(${K5|DY+ziptvJc4@673eDh0=<m}?^x+dQ2B%b;&A=q+OwfY%MREu>
zP+)?1;{O&T0G=fGb4L}~`%uCT6uyJdq%1eqO<YRHzp23Cz)Wl^;<@{X<4mC*rl0*-
z`H3n6g*u$rY4nCskctgwgWx7EnkfG|dSnP$@u81#y_`~ZEgIi7QJ>vV{UtfN%7@YR
zHMAWWr`n@%hDT6x3nefzd{mZ_(>#hkx2X^2L%fKOz$9e?V|}Q@aIvOPD#ePeAR7$d
z$CddGqbo3s#rR52&1==IQeG=nw{q&rT*TbG?jqBd74be;B4s77tyRRF^c9f?X76YV
zvuE}mqk2a_eP;Gda@2*(JI2E7;@PuL1VR$>4Fa?&<p}^SVzmcwDK)xdCI>F#HiVA;
zj@Cu!Xa~Aj=miM-sJf5NawFmrwf;6pWX7%%D8Jc?47b&2Zp(m8luB-L;tVy@OCF-H
zdWmiUX3-C}n+RIAYHn~2$X^D6E~6QcSp%PZy5Temp5Ttir#FlBf^3l=PHaM46ANq|
z&12t{9bER$Q0%_|l2x-nT9#(93<MS_b&_RO!dezUauyzKnu#qm+~>(icnN7qPj(dv
zKz;fH0-~9S7$P(9U*ySD2(AVJN3fBEvy)<kunx6`fTUL9<Ibsj6z|-79z==)bYEVW
zBw43Y(jqWJLe_btq0y~i)%h@p#+Tn%SuMPJ?Z(Yluf1EidHwY_U%U2BPRkF)M^>TW
zb)*Mqnm0?Eq~8ZM+rm%<YVh|M*e6|`KwxTui8hkbRK!>VNP;fOhZxh-?$^*h{<8rA
zB*};kgy;~pMVyxYdf4s$^#JF8;I9V*aa**IAkYLWWjd!FM*D;*=+fg;5KzX|GZa^+
z;4nqjI|Gq*8Jxs5AcvOX%R~r_K;W5(Ss-Pd0@nY8*C!UMADvL>Vg%8N1L*_e-u<5d
z(CdjY%4aZ!JS3bT*;vWAe8xvGD=0V|;ms74{y7nO6Brtp>mO0-kAaXZTDYR&n)X;X
zt;aB{-=Q8)gY4rzah_f&mITod=O&?K2}D=v51Wty^sBIJN4e@|x*|%EbNW*$R40`-
z!S-&e7kaHw)@!Aql@Vy`h?66#u9tCf2n|pK@F3<YE(-wgIddgOxg@t>d{2fW3r?88
zOpw#?XG4(}wrV`6ie<FJKL|-(0h--H6weSg$_I{(j#_*RBxlB4B}Rx<B5x6So5(F9
zlDm?~v!y=2#w)`FW9&uzbT-aZX8t}NPZfVe-RKII(^pqlmUC=5Hz99>a{fqr^ZILX
z__;|9<w{8ZVq0fPk2~kOf((TCCXpA3P)Jwa)g&mEPEi^asZt1ZxIuiPA5c45<-p2&
s%a5#!xE-|W&Lv``Z-ZEhMb=8n2kwY^=IP9obv1LsnzBw#n<t0=7w@F<ZU6uP

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/common.cpython-37.pyc b/utils/image/__pycache__/common.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..8d2be63abfbfbc0acd3aa9d37ab0a2bc641b0422
GIT binary patch
literal 7517
zcmbtZU2GiJb)GwaJG)#`luXN{q$pE6KirgBQVQ(YiXhASi<sK9ATgG`vAZ7bo!uRB
zcV=^EmK2HEAc*Wp2I8VYUfQ$`xP2+|l7}Y9Ly!kQ1VJ8x0_{tIK~V%~k(5Azw#Y*O
zCr$F5JF`Dr|0G6cF?VL}z31M0&OPVc@0`6nJzY?6z4)Dn+?K5<|3M#vOUKJ4+~K!T
zaD}s$Qc0hz!tkxO)V5a9SX!r74B5x5nDUvc<nYv5Ry$uQppVXt*ObZ>KgBJce~nd&
z{6oCRr|~w;XZR7c&G6HFmLEmU5q^w6gqm4C#~;S?sH<&i?lCI)@xRshBYgffz4DN2
zY|fz;C0{`834Ri_4|8@~DSh|=YgX1uEHPHwPQ%4Zq2Bdtkr((*tEN&bE+e}&Jj3sz
zxcfw;#L6c73$PorIc3WbPvBdPG_G>(HQ9DmxySCYNZ&NL{-)YfzM$SwVm6>Vsk}kG
zKyhEGGj4#&q7jEwW=a+1_EFzwJvCN0^W0p(81tC*gwoSwO>V44kJVVik3FKiuEsik
z@~y5eC}7rFJ0ON?3<X{EAvwY`JdPq(UYY5EAFiBGUQifTEQO>B+TyiuDJ3;g%PR@H
zD@Y)+c>Bh3GibZZeh|4EL9n&Voyb{kh@jh9zTNeo_54Py>H5pHfV<0Ay?Xuf>aFE&
z<h8<Oj|7f}c02IPo!z(d7aFeb-tCBs$7n&cI$ha&wC4JeE2=dSbgGUJ<qNH#=Cs0#
z<zaVhnkC%OX?I%gA?9e=pQN!)pa65ndL(}KK$*kVWDJsEm{b6)+2<uynCL8WC`c<p
z%;l!+T2!A$@d)m0d4gnlKt-vLXl>6=^w8UL6BFy9rxA6rC_%g=+U4h8Omdf3uY3-7
zqWYc006)$yCTThyw`3&wj@PP6u_OigsLDY!bW2PrfZgnw?m&r&B1RUpvgF8wpr|FB
zL!sz%OlP(>Bg+CihMzcv`k#N}ujx8?`sdV3QU3So8cB>UvUb{IsUH80mq^)%V#L}>
zQbj%cQstKssrwpK&EQbBzS=Wk96D6lfN`j?DUD+T-)5BK8rP+97@3+|P-#ONM;XRp
z4(KxD0EILTn1VEp5mh0jil+99(l~N*4Y|!}J~Rich4rjqJsX&btk1`}*upP0lzg0@
zFcdKjo~7C!Qm7ALEmCL3wP|ECa=)Y|{V8S{D!Ood9<<+I(FJYHp(T9~9T!ue9$(Wy
zfuuPP^?PNsDyiS4nM7|pQ8UqNuGfNY%Y`O+dg5-lVg?lA2o<wb9HoMEUXUe=ho~S;
z6c1BDo~Bd~WJO{gMP(Xxld@eGP7R;V-LcO(Y7N^?3wBiHUR&BkWsW+%qjq+(Ml8@Q
zWd7nL6~yR(>C|c_r2Z%9Al-$+X7HOe4%^N=n`h!N^jW`Nqp|3sC+&h%j-TP>ff~%f
zLX<w-^p^4l9CvVf>>ExxD35(>_tpE_eZ8I@9o8Gl@SDy;>kSsMSc%koYWQlbG+0l`
zY$8WCL3Y$rA}z*~TU)C5O04kwg3?oBjTg4^;wPZiB=t9A4b<9whT~?66SXF(HBf6A
z^$$Ts)EF<;>gsJJ*2TZa;OK$=K&z)U@R+;+uUT~3VELJDqYV#z-fjlM+Y9{2Y1#Fb
z*9q<CZ`mS<oXE8oSC;IjaG!k=#ko(d*xYRh*9}YMLczv$$F;kmYj3+E^5AbJy$$8E
zBkWGt3D8y#L^>J{X4@lAPqR97PM75v&$L>0)O68?Qzf)LUw*Z`Fp^_a6HW-%cGC4`
zlqsobgy-cOEyim%LVJ-2KkswtpkJ|LJ9{miw^!2%qQKr5j^^^Ry;@J__dDHaIG7wT
z1TTXDmTWJwcRVZ>)RH}L)^@>aU^_O=nzBQLO_HO%h-0;dd2bw+Z`r=;)}3w(l3y2H
zcSI#+PEr~W-f|+^29nf!(bQYP&Lq?4?d!7keHpbqV(t*TnWR1lqr~i04;vcF-Nj0p
zOZE;_r8KtUwZI?e(v7+TegTYi(La;XEm<q~A!`ZhBW)T;W+W>)@j#5%fhsVAWF{SG
zOe@|)AI{sE{$mF-%_F^)@{}qe<ceH=NJIb!*tpOR{MN4Bbhcf*xwf0?^pd^Vh0>;k
z1GUMP)qr=w_0y+^M*u+&azNx`!}BN}!0F24nUJZ{*BFdfd?c$;vl>Qdbsy-C&}v_U
zvmh9xucgNxjVaH*l9p(-?Q@AvikRew)(Bvv!A#I>NF7ZKi1w+K#Mq{eBeaxQ)y&-1
zX?b)?C;%Y@MACb2A!{F@x&g(zbNH7S0UWLz<`6Cqzt`2QRO_0dEFh?O3O@M)@&D>!
z{wMmgiT+G{btoM@mCm8@A+-kPkWQB5W{k=0;J3Ik;far6wc?{xd<+Go20;cpB~2_+
zU5ScMP(d4*g1G4c`&H`s3hGk`Cna`ztUST?PQRz@B=JyyDUi-ogpT0`Vc}vAI5<Xt
z@-_7f`t58Dw2%Ql_&B{YK@erMz_$w*$PxhBI~TXmBEgHP?45kaT1NdwO11XKuIWs=
zrt+aRiDmj6EvBRkI>iB@L(hjB@M~_O)2b7#<@$ixI=Ohc2Z$Ro*H|&6%~qyUyUQF)
ziWiJj2y)hmos$?MmoZP~7g7i1_$<Ruoz2e{3(N%GE;38C*zg@+FacFOj=sYnlR^s$
zJFegkPs0dfwhtqgp`Su%;$^+3_4J<6Gh>yj$D!aj?{7wI6J|{{C*Gy3>l9JH2^04%
z1gGXb_GP#QiA{3cdL4n`%W5yz0}{qu`970lejhz_gpv9j<{+fp*Wn`Az7G73pfgtS
z4OFseU|z=ayzr)$_K%I&>_eM~9zEp<<Mcom8U4PAH#uHkADe4p>}(BFljC!X@W)>F
z=I(kRYE428^5}rKhrcS<y%s{ob5~Pe{><d@ZZEFms0joP)y0h^yS8N0aZ?()=q*?H
zNOK(SwY6k#rCr1lHb|Pd)a)cD+&Ti!i0r$KFn90h9TE5qI~#^VxP#xoE3BnpnkBi8
z!#M(D5STJj4of+So)dOG(a@L_UT|7n_nHuaNc3&b-9gwF_}k|=1(qGxiAs4v_L-RJ
zBnjI}*ha!?32P>7Cs8|1+&ih=j=YOC5X=`z#RwHExdJq`$UdQ3>YO^G&f)76$|WXN
z(K^&k8iRCQ>gJE}5;O5xa;ls)$Q?p&8myu9**+aHNLHW?7!q0l?c*lT^?`c(s{F<m
zH~Koa9s$nuMLG&;QG*ch6ffcl?8>Jb+Qf*AGC=%@tTiwSjf9c2IVmdrJ#)r7%?T}>
zJt?2iqNC2QU}{$&U7%0<)@Lpc&T-(Lm#%DF5rgl=31u6A*VPC;GCUee=&`!~xih!y
z%j@MQR?gbAZ<%%yh9!RlRLRe9x9<5Kp)xzd$-fO}MF8!#>)P!=xPZu!gDeSvtklb9
z6m`P$%ge4`-to4)j>|o#9Eip;eJ#s{J|^d_rt?-u^JduL{mdZ4v^!qZv}H2FUi6pj
zPiBU(B)O82p8a`(@vkfq!f~P{DK7%B&KPX9cdHWE3_#hPrVI2f2;s{54B5P7*P4zG
zLmR_*?+d|3_L%Y(jWCy7xP<tWNl6#KeD?P(KgEHK*%zPs?vMZc=B0-I!pxa}Jo@UT
z#y9`xAAa;_zkKqNG+1fr2Z(eOY3>;WTe;NuLG<-`{b!fo%4%M|)cE^r-}}Y?fA!W|
zjbDH3|NiAUap$dc>?h8>bwF|P!Iw~^y?+{9Dw)U*k@rf>NV=m$je^94DBIl@e1t37
zU3jsJpeY&B+AC%%*~Hv&y+$*t<brO5d{HCOH#{etwDzKCY@FMzZPY}y6MX~2gtQ1{
zT4$$V?~CkFwum!H5HuU+eJGBl{1>qs{uhb}aEnfud#ad!0Q>60ru#@;%6endL{z#M
z<#49J6Kt8{@&n|27SImw&i+REJ5{C=b(~WbP+O@l;5bi^{6wUQ`aMt-2ww52*AZVX
zDEoOD-!G!QxTS_)k)x}88h!r?eP`yV^--L6M^N)^)WDDMSy@{|?NQKtpJ*^2vP66a
zUS}#YHj90R$eL2Elry%X>1gHwV&;2{Vqo}zxwVp7(V9E8idJjxl+?9y!u*Qv<71=-
z;<FHmRD+7P(GX>M6eJo1y{D}#ojv$1n)mcGXP3@q+g!c0XRIupd;IY~4EZc5?}5e=
z=Fw^4QE-`UG`V95ef|kA0~`h;s}7ThBTzj8W2GZvg$SQRk(lW@2$0|DCWhZ_cXnl-
zCdp+%Tr%@a7*bRo`0_B4I2LK3ptFmTf&$YCqumY?on6n5o<txZ^Fik@I$+pKcvSLj
zw_ObccSR+??X;@0M`b!Q{<I~<Sxge7K)5*0$agSKNMKq~i#V)iHH!gEL3w9!&#OfV
ztt<elB91P>(QkNgks=eYk_XpSya4;tpAt~hOx*yia9q*Br9z1{C_9prq^SKIBS7_z
z-cXj@TDtWm^?(v`2alnUY5{C7qbFJ1p<1#ih@wjG!~!h48?C!{qh#*-^K0wX8`o~#
zzH#lP>h0C%Za#bM#gbN;N;j}U+3rM-(6wULw#m*XE!@ITRqF817&s)mRoHRLP;KH3
zOZhk+DX;)+pm_O^ScY~^N}c{Kfn1Ul!<Hg0BHN4nF8xWdKlqaZC;h~q6bACWK$hrZ
zLX7AO;b@0pK0ybWgr}4<0*^UL!c#Jt((wI>G`uQ4Ol!7L4C0U+y|aXglzXZ6w4?uq
zdyqRtC<J{;bM9kWKlncZ@QPy3cg;{z;4F7%Twawzh;&v|bvcro!~g=J7ImcAlc}-U
z{uLEjU|WDvS_si>9bUCfMA;5dCSFjqG!#FHo^Gj?1jdjaC!J&&AXn+XotOdQt8j9U
za@C(3Xeg!J=^v+9omYAU(g(e6411w&*h@n%BhogJkVi6JuOm!|4U|acLCaNy27v1&
zb1lWQq_&XzP=yl<M=e2?2(f-1$h5lS@u(?2g?{+MA#1C`th>nO88Wpy+QBsEdx6F@
zQ(csjxP}UGor=#;af1qy@!0+n&41z!i9kX8^fCN&2IH2asp3PZ7N=2^^!4?%>m_!*
zbX-P#a`r^Kx%zCHfPR@e%G@&<U+VTOdF+;bUyyeYpQhpz71E3;CX?V-o{Z9)NP9xc
z!;j>?3X03+O2*bLIgsjMT|gAjZMhdoj3F%&dI85p+F`MxZarLBC|tEp6&9@1mU(>o
EcT#ci#sB~S

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/diffjpeg.cpython-310.pyc b/utils/image/__pycache__/diffjpeg.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..2a2a50d4ad0191c5dfc522e7696584728273cd05
GIT binary patch
literal 14822
zcmcIrdypJQd7szr&hD*Fr_<>z+v~U1$U0e)K{k#h`-F`h5XKS#A2#c~ow>a|?d_g>
z<|N&+2L#DBBoPUOXF^~P@<0R#1cy99gv8;MP)-F18K@$G0_uu?Ql+X?igIv%zi%G9
zdk;%MF<0|V_t!nW)7{^zf8Ezyw=SpQdg8gaxeK>z+V`31y;yn}x0g0F&C$j+M|X_6
zKCa8t7&ql<ja%|e&7{ZEI_sGA)J$gF)-~6h&Q>+Y`j8Hq5}Fp85o!y~3e5=}0JZT}
z&dHxMrU%CJZo0G1wK{`teo<f4kvFH;JL}>a<3sLHXM;1isPnfv{H^uU&XCXzLWe=E
zMSXg}9dZn3<Ksq1CyhI_r1c?beaIPcHl5YRhn>yN7CblJt2tYpZD%!So12~<Ib)7*
z0%yCk1DqWZXEQiEoh!h(BI0ZT=St@)aIT6tTfr$fSA%nP#MuVUHO{r*TpMw=gL9p8
zJvi4#oE_lY;M@q#jS*)jIJ=x8IK_x_1voc3F97ES5$8&9b~`tNb92PG3Y<O8C^(}L
zrvT1g=N52oi8xn-v(MQN&i;sV4LAp!Tfw<C;#}(($F$-LTOY{1yy?^?Ypzq6RLz;f
zl<&`aw~vliYyQ;S#GXoXW^|_BoN3hj*65Mi<mCMiyyV_o?w+P9)Mm<6*DExde!;1o
zta-JGx?7l-FC5IpZ8+ZS9?vgVPPnHlQ{_h0mF6Csb3MP-Y<Q#lZ{5HDw$bTXw|cWv
z@o%0jJ5H@pCHuhsYlY^PD8WaIdXSr(YgG7Jxn8uFhQM7KAfZ$%lc}vFyGZtv{0aza
z=J8Ohv@}Svkz_jwR76|4gXBdZMO#U8>yVb_RGO&_%Worj3CTf_xaI~jHj`WdqHLyK
zP4WQAogn!2VLX&nx}K>K79Am(A*qtQ90YCh=a#MoQPQ`=%)XLjisUhphe6QhEqJJF
z$T-RryREKbcAn%ki3@_7FU3RMKt_pVf@GG&0SVu!Aa!03QjR}A>lTeAc0(N?;m;H;
zwTbB+B=Uu$A(xCLDdK=F-5qje$YccB(YS?MndTS^L&%baY()o&FQ!wt$8dW)L42*F
zx3xvwacV1!(MD~H2C+3-Hxww0u5ufGtxUmF&AA4prx43Yu7GQal2J5*l;2d9sURh_
zAWiz|^O~ZB{N&!zspgD33MqCcn#~iVm8Rp4&brE@q>UcCZSR4D`)<26WRHy<dvJ8l
zuhqR#5&lsq&LqWm&+L4VW2PiOz|rwwWVM~9I)ulHp7N`tr@N78qhBX9m3NG`wnJNB
zKD7hAbWEhI5p8P^ub6Hv7RJi;%3QtdyM@Ve1sy9?no7CeY%?4Hzgc)}u3WGA^EnxV
zyHwS?T~Zv8sA$(@y;=5)w-;U+RxM1T9d-qc&CH@T5305&Y~VrHpHq#VHssyg3uE1y
z(U<HA2YnxUMcLN0w|x3=n9yp6mpCF(?xDjn`k0AK^g)sj33Vsf5b;ZXvlMpoYOp<O
zvYoYc+sr>_>xzR|!QFRpba`Fe1>$OOaN4-x=;P*-+9X__W1h9)<SZuzr;vJ5v$XLv
zWGWr3d+^@7_dPT=F{bXvK(ww#2OB3{<$DDbgyyb*!Ic`)(Y|?Q@2CL$^^30q@iiaP
zd&cM(i<++a=CoDCOp6mUgqolx_$kMNdrU9rex_sl*-j4i;9lv(+FEHK;^x~iXPrYG
zNFObl)wkaD=ntN`=HCugCx%{g+f!fq-l6K3Pk!!$+ebfp2!)*kU*C1z`<}i~{p1hc
z_2ZfS-?;E2d=;+$M4`GDiRzc%{<Y74>8rto>UaL?>1V(5%u^Q>r><y56*VnNJbGB&
z$AZBi<;_7k6s=)kmr6DGg;FUn8nZ#FtW<eE$cS+4-xuh|!EOZE2g<5E<NA1s7B@)y
zZo_LTG&EJ4^kLTFbV`l68E8;?rtGWQ>GMWld2W4Dy$bbVo3f=+rC#>DQmHE*>Q_<r
zLfjs$<)FS*w{+_{LpPq=Y;4wBoBL*RPyC$~&QW}Eo;Y7)xIJ3Dw%%d51Nm^QMGLc`
zPaAE_z%=qs>KwFj+S;Nm7;U4IYU|VKMV+(IdRwAe{fb6q6DzUnxbzshHc|E~Q-#xo
z{kW&xT6M}t;#AG?r$R;JD=U<Z%&>)S1N|?v#pELtEvdSTMEn-B7(~md&3MI3kiqQD
z%waZUGP2=hI89#$TOC8?z&crTPjP<Av+j9Q(WeC|=}C}=VVJA?%XKva<8i9292!)Q
zV>)DS)$@95Q~xwXwa#ZKISmK}BqvuY%`~04I@9@5iAo=pWJ)EcS;3@psuev}ur88>
zid(P8(!k$TuOXq?5W%DjU!)_mHA~N%LmA;#!05Xu0KC)@5UBr#u_?%6`PkfqSDu-z
z!<V#*g^*Y9y39heQJ{?|Fr;a?b#J99;Fz;@-WM4H^qT^z?|B{StL{gcs*|vj=XDwE
zus6~#=@N_R=`T`xhxOL>72mnKfYY63aF5~kz5voM;2nJt!t5K<X4`O5P70B<89gO8
zX>HRwmSZmJ5dXB3aqM%}g4tIeY9a5fUEgY(qEr^u3QA?Rt-id~)_dEC(mH<|Z>Agr
zCeR9>lP1(H+q#%Y`daE1%5;yqt)eYSBde<0{H{y1k-SdjtW2y`eRN-}-=q8bUr|5P
zE1^zPoz*xAwfkt%2(_G<dx~Z-aKcp($C6iTxk1{4C3W0jLj_@x4>Mh<&pB=h5q2@H
z9>E(ys#0$@+(5r4$dyl4OS8>peV-F#=Nca3a~DlyqC_zjO63IK%7`5(?F)>`1a7s=
z5{Q~IYV{a)kL331d-Wmw?a^D?R!VJT4-})-o`$;0OwFK*^eVF{`ceV6@1m;lij{fr
z?s~Iw;uzG}_o>O(6q0o0$!iVx@>yDB_Yx}XF~nY`!t~6h_+jIy#LEbkMVlA+5mebg
zKcSk;NPqg(xeES8#=wJmPH%1L|JK#DnVOtvGv@*>NZi%q#y0I?W5IxKkz;Pt7NAuK
zo~<?l%4uUe0?KK_u}BwF3#oRhV?xvE=u%D^oOC<gu@)I_BPe1mscAvmGOZ^QvHdx<
z?J^x(;u@!JIyL&j15g|K$y3R%hjK3(K;ydf9IjCO4<ZW@eH0JHRgQWCiHLZv(!}bZ
zI@g?o=j##TbZ8I)9qvb5a;P(dp=`s&(epXMP?+~sxq;Z)bDbU;6~j!QbON^orNeek
z->DBH5@~H-HlLT2bIJ@Q+Us#zmWsJ3=w8d^Jbh&B_`%Z$Ti0Pd<W&^nEJUqdt>G4q
z^SklW2&xdu%SiUcs}v^w9$WN&2+}X%j8`39qEq9XfoLc_o$+nQJVSSoI%gcw9@+h{
zMX_737E(Uo8YkU0Ml>h0MQdZ#f;V%>**ma?i%QL?G}T`^0#Syj!Zoy0lhz&DV_Q*w
zz{!H<A)xvWt*Sd2+-PqLT3bl_>pFw}`pys|;?9OeStXW+V=ms?*oIIq8qUB7%90u*
zqObX*=dEysE6dz2Vb~BBP^J(wA=;UB5t7ovh62d3`Vnqi|9xnQBN*kLIaJ&f>i?5q
zh9S$#kP1>QS2aCF{~cs&h`W3kS`T5ELex|7u_7fd5}1{y7Z|fOMKl26?_<hW#1RyY
zsl;mPI7?CzU8@Hr-DDhxK&?`)yRm??L=n{p^!Y%a3-pr+%<HpL<<-m(RmCGLG6)-e
zrG8W|=$rK$^w#=Ns8%;b9D^io-ir)exQ;a{BW0|3l9(9|{~Tsq;$;}5=@enTHbE!r
zCr>1b{mV5galqXl0VF@Z^d@|MIK8h$3a;@zWH2;VuOpGs6sz(mvt<&o<U$f7z&Q_f
zSK?kvZw*ExcuCdgt1aBUE-tIct>EsvIPknK-UPDdvNl7jepc^U*xEt%A*}LY3S02&
z1p;_n^)oM~L6-gx#o_E~m_dxK?254!1XnRH!CSaAp7Y$$&@fbuFQ)nJyf0n}(r;+E
z#44FSJWt7@f~p3pz@m#c?I0a`Q#<~_ATwX`3Hh8iLYw{})>*~Lz90OHl+-~zO*cE7
zj6x#VqScjnU{O^ts0^NCxGzqD_>fs&_l-&8jG;Ebv0^n3Nk`B;LbOJ2?8JHymJ)V}
zD?|jXZBy2e2zODEmK;FpMc71f%zWSCn{BO=p#?(<D@hn7esCD-Nt8#+RJ4N}kktlI
z)J6qn0zr#<7pr|31a?AVJ@@o%x#5J~kArc(H)2bN;7OoQ00iT(ttP;Wc<Z7V0((Ty
z8P{O(mcxp=wV^v?ktAb{A|JPA%+^6u{KBq4r-xID+zRf#D;{_CL9o%&>{ueEwKc?a
ztZ(5Zves`wVk!B>MOAT7`h@Y*WH7W-9OWQW3FT1zAz7;k;NOt{M<A`?q-U2Ydg_n)
z?w^1F8vY2==SgJhpJM8xB!5aG=1FwmW6Z8%o-UC8Ns!it?svqN=~FCRb2iy!>Gq$q
z8dED~kzp*ai}VG(X34dtQ<Hjvxn;+>hzyNWMo>2!zQ7LhF^njnn{&@xy&g_}4-i=>
z*Q-s19f_G=j(K7qD2?7xkbd)oXu|SSXEZ;3Mq5N^T*?T?1}7VFa{d5);5izyyuU6<
z56VK{!0v?zOmTf^(>{aW2*rpVKFA$&vEm|(T^F1});EDszeNH-aRtcjlfe&i6+$Fp
z%0;Jw+<cd@mK=VIJ<EX%B0e34ZVv&P=^gdw5y~uHOW_`48*c)E)e?`-$m(G;jQ+yd
z%`x6@%3=b*0;0phLI!mWL}Hn3jN0sW#xYN7swmde2Iv6PbHQ$)*0j}5@<`zgM1dpi
zw37l1kap5S(TYF-cIzXF1Mk;;F2*HhU%9ycnGrXqR`I*d#Dj%RvJe>nYUfX_T(lg-
zqUB%+g7agRp2L6vLVGVBie6s5o(!WWt*eid-(LgX@2|lPtP}3Mr^Jl>Eo=Q9NHN=8
z$*lG+L<F!rmi7(bleJJ_i6X6H7ru+;ykU?LI}fQC)NM${ppi2+8?BMJ%hbv}y~g5v
zR!#C_swDJ~sz!1<$%{$ELOM)MlXQPrOr!)&RDplRoh*2U)594>?h5llwd55$iGhcX
zKwqM0W?MI1=teLc`o5)(j`!IP?~7YOma8};kZ$Y|t>m9Z#jrCUW@@dq{Ie{4BS>qc
z=hL1?r5KVVAqK~!FT}v2CJrwUFqmziUP*w#)j0zULDPt*Nr1sz9AE&LZew{Bz8CtB
z%yR!Bz3KNK=hMrCDloCQP@iDe{*0tMiAm6lJ&ibK7}Tv)?*FcV{2*r7*gFcng0r@l
z;Yj~4?rzLr4qJ-c3hus(Q^+e3hq0~*w03sWeGqLJlTr%et5+h_KY<%M`e^3=nG6wQ
z@wIEs{3f&#&3w;ST;|NHFY-sk!>^d3)ja%{Sbddg*v60e3JO~rdq3jx2A&<=zy`j`
z)|p#-z>lC5D0!w9Ar3FNkDFtji_QOE0C@NwT3xP<45#5_1^t!9zlp^^!e?3h1BmDV
zi0BX=$U)lmMPS68fd~Z119n*F0FIa*bkcAyhQEG#$lu^(09)8jHi0b!v;dG`c80O-
z3XO40z!v5RVT+B(<s#UErRk_N1T~UUz!ni`(M}6&@uK+4+Pz;Cf(`>m)o*f$?;;r|
z`7M&SknAI2u&)l1yp@Dv9wQ3>LWW4=%a|fUpx#Ns@L?H1px#ZMgo(e;)Z0iH)TrMk
z`5ltCle~lEcS#-uiO~ajhYW!bLa+2bW`B=_wki}sF#`OgJrJG<K_~!02vLOXdTS_>
zl0+FUwe<$eronC@j9LjHIPtbyKa7wH7fk2)RWh!xW$F(=u#lIYh);SQxzeW+Q;(9A
zNmfy&7qL4w3Ul^$ctlvU^>IgHg>w5}l`)aw;56NY_tghjgSkuUQeJ?<m%FM;i24d9
z+!&HB2tF<0%q7g%41ccy5^pGdjSLZ(wJg_t{LWcaYi;WNgzjuHAdWANN9?aWilC!&
zAq%V-$TRR{V$CV&f(4`*ARegIwjx}ao=LV(1!>N-GaXZqX4W?sxucV5eerq1xd@++
zBe8D8CA+6L$=)gO;@!))QyBupR|x*>!gv;wUt*cO9I_Mss@p-wW&d`q-Ph#ZP=_(H
zk%kLeeHXLTYwQm^%CrdKBTW4|xwbf7;DO;rgnBcWK}tR?M6%Da>f<C&kgOuKFT+nt
zcyHx$`gRHNTO)nbdP&)(#JmK}s5XCtxl79KP`ID-s@Y_-bVrnS0VQ!GwNNk5!$HXS
ztQn4}71GPN@cE^;%r6z`jW(6KAF;Xwr_dy$t})!+VGt}tA(v+<$B3cM=m06PQ;DbP
zYt!&49V||;^N6Dx@-)L|H9QDF2-9vooy_~nMM%&7;e3p%g)5WfB?@5g#$jb5uGn9t
zXQQx^Y&l=%rpoH*CzdGtliv`-*=L)n`5dcbxV?KoXqmZN)U(foy&~=l^QaO=1~yy#
zuqAoyEkU<+ct`9<VUcK(OVIad==Ve`-h_5zg*b(@`VA7Q2Do=H{RR>$gL;N!nb@;>
zGZy<{U{yl^u3@crEf>mo<+A4vYzLO>-kp0QOrM{D(?=$bMR0w77qez~D8j4&^XFG#
z8TM*CmMzZ0Z(a654(d7lAaNc8cE%=-zICn7SvwS_^glxR7ENUClA5?q4kYcvSqni>
zu1`)a2|wEH1USn>OmcctWDmcQTn18{a}&4XoXg>&Fk*Qrveg%V_Z8>f)A%2r8)EQM
zedow)v}5USUZ`IEiT8hYfAGT#alwtZ*dN<D{_LR;#@xrIOiBL;NnDS9UVQ_P;$R4c
zd>!fKq0Le9UP1Bz$ty|3k&6$fSjtvNcwpt%m|Z0<d1E{mgZgzi5X2)veWTR16YP6@
zvHu(gu@ifd2(1A$z(91!gFw)Ob7t?BMkg))4W}3uxnp`c$AK<!wm|eVQP9hBNUNY5
zoMT?0(n(~*8J9Cbp|odu_Bmu!U&Zonj3hp)Ox`N8{shODreZ60JX)Lj$M%x$hSUCf
zw4hF~dFECKIosx?{sIIj;ouRs7m;X9c5y<-6@Vi=viQqJ3MnGIy}Ln{FI>(*NJ@eh
z;+gg{lE;ZYc>*PrClEq;3OX3ZZ4of|O*9#516N`}zEW}$s6$r6L4LlM#TYZNYB)MA
zE<4D~gps78*$C317nb2grde+)Z%wHBeVorhyhacx{=*5bE&aousEvqKg;8OAtwVyw
zPju&7NWdo`#&dLlf{gn#0R?XZ3}`zBV89!pPkKAkwgucXu~y6a@WKwZM$TZnrHw<8
zOc};PZegG?NbT#d&F$i_6kf<`XW@^sNTu4@cCJ0JD7672=D-7>C{SWXpu~(rK+;^u
zqa@SDe<Cd8xt?@v&>Ztv1=^hf+<E438;b^N=ed~;9nZGDov7zZ-yh4Oi?jVb7-H{n
zr1*$~poK|O)YjTeb!ugx3ht?u+;CqfI>-AnMg-4iFg}O5Qx`}+PjVN@dr4X(uP6B|
zNO7I|d#3-LL^dZl>FPg70+Mf&e20YAJ1}QzI1CD0(ks<#C$N<@Q$8K!WOrN`Xm1r~
zJ8p4Egq{k~&M&Y<i4OjOsW+2Ih%mv_DyHf~6z%OO%n#%t+QU4hVdM}qjOg3(yw0%U
z9F~cBY{5RoHvh4|VevM1l1F`g2V1Ay^Q<9I%&c%LVDw$FIrDA;yLXes!Bz+kB_pyt
zo5i7IZjbaHO11+#{-=x9kB?8)yl`tI*$yd$2QmvN@ajBJTAv3V6CI2X|DmIDa$kr@
z#|_~y<Lqpc$C|?c5l0a!G@gxWbl$sPsFUX8oc%5wlCMLU@0-K{lR0-cw|u=qWzO^Q
zza#xsS8fsSF4QIq_aF?5-{ZEXgM*{kJI3++J+gTx3URzICZ)lRZ;#T{Z+P3f0`9Y5
zG_Z?2odV{AgDBB<43MHEofU-lFY$UY7lt4Ig>-0e4>84-LkRmbNCkGR$}*)m-=Fvy
z#59!EV6eZ#vdgntH1ZV+Ngi(iD8*iM+Q=K&g4qyHNMcnh>|}iP8PsRNZR4yBpkg!G
zjUMy?4|CDd7{>NF0W<6-pV6`Rl3TD?*2Xc)GdiCzgqyIeNiAa0?wC<1kE36Xg`;B`
z7-byc+>8kJ1Y%$ZN5q)JnT`ziU?ToK;7bmCo2P-e&jn)+BR<IkqBx_21EK`4ZJ2%&
z$3+e74`pe-!OMnb89)Kg2c_FK;WpGsi($p0L5`o20!$E^e%#<OP?n3)eRpCh`{k2%
z^K_wm$ojs<?3^D)y}Jt&m8ny^3#vM?yYSk*y9>8Gk~G5Cu#dohlUVt`5aGX7zzVGR
zmxzLg^MW&OcaKM++ohl4JQ`1AVxprhd`=}rNVe{M{0oYObIxv%I$5sIxnf!&g&$#c
zF`fe1jWPRnl8=(ifdnSbY^v`u`+bnuO3E7PZgM})N_s8Orvm-dcxsAgJjYX~_*iDP
z$Fm9&49s@*lWbq7kQUIImFB6$VPhB;($@0^_lEKS7DkLKaX<*Gtf543BRg5tkFUL=
zp^|-pZO2w}Ff)`><QA_Bw~h!+1}XU;03{HoI1D?q35`a$4lqzjMkEA4yQ_o`vp`k}
z$E6*2X4aoqPcdyx*6MZn4;=YNksytKb;D8f?lx<8i&b^l%%@4H{1R>nAg&l?Nqi+?
zm4s7b5Tpw|4o6%M{X$^h5guv36TQ~F`#~(-PBW_FKw4(F32Xe+GOcF~%XoGuHG}~_
kfJLlr+Uem9!&`^94R0CVF?=QdZbW(;(p!<<vU%_S10Jsi00000

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/diffjpeg.cpython-37.pyc b/utils/image/__pycache__/diffjpeg.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..42d5a94006eb23d27b2657e8f21f861d48bddf24
GIT binary patch
literal 16804
zcmeHOeQX@Zb>FYM+dCd9lA^w4<@_xwifB`gEh~;ii?VDfwyao6EHQS^$D8HxsN>yv
zcTXZO$2g7TIBBCaNt>i;-Q?I5Xws%Z0jEGwAXN|`XxajWi>5`f4UnQm0<iffK!Ktt
zf~LRU%<kSDNqtyWiu{p1?7P{wGqbxh?`z(BbGW}hqv1I7`bKBzpr-vhFQQ`*g$bPA
zc|+5LHl+z&7!`d=m;c6;DgUi0Oa3S3lT%5Zb<9d)J~d_Qnq$tTg(VUq378VLNQ+F#
z7Jc|-MZXvj8^oX(0wybl!x|gKh}a}Co5dEf6)m=j?P3QoJ8^xR*d%s|w~3s%9hj83
zLkx*K#a-fVagVqcnBBM<MT?x+Bi=6d0`m@Wuh=KX#C_s^V8=^#$;Mk{#Qw|1T;EjI
zNw)hPt3BXkm-S^G*XG;?@jz5#YS0;M4~g+*o!`<<i_Lh8lz3;!!vBMK|A$aY;eQi)
zw!}NdLt>nzlC`YQ^*MvW5C>i{6r3?0(c-5K%BKyAhsD85+SIUkmv{vK8=ug`yTzlI
zH1VjDoEy1lPHh6_keC2wBE)P4<}q;?n8P7v3ow)72rx%N%vN9?7e|3P8e+Bq^MrU3
zm?uNbc3_T)<G>sbF*|^Hk9Z20r$Wq5V4fByfH@IjZUg2SaT1u5A!Zja?-kDi^K6KD
z8!)HDX<$x=m>e+g6YmG+{UPRdV4f3afH@Ol?f~Yjm;z=h#M~*2Q`+bUnqSO3T@&S*
zvLkXcZf!m{>-%-@;MiEH?9Vo)_Z4gNWAl~Te6{R1$BvX|W{#hH{E1BFXwA))=L;pr
z%T;TBPL$7=z4CO$$xScj4rHP}gje6^`Gw*+XQ4P-sFobr-AfI}^UJlWH@5$s`}aRQ
zHdl8_?+``*9rc0`<!XuO2ln47VP=Kga$-~uGBb^8k<AK~6>`o>9|5^*Wh=pMg8c;V
zCdd*D5NssaP5=hfRvsaEh@hX~VS>jA4iF3xY$mvk;5mYm1dkF76WmR3gkYYaMDR4h
zodgpE?<JTec!}V7g8K<h5G)cb5I6)+666V{3F-ubpa_5uPY8dp?u;5MHm+Uyg=4(E
ziQq7S1yIxhc<?`ga}uYw6TsKndP`fz8I`s|ifq)jXqbPac~6eB>bOqTFBdq2Zmm(}
zG~}WI&*X5dnCNQM2oiqHEzSlBSqpEcFML&VIi~+}d~CKh?~H*QoatKa+?Xi%g|QOq
z)yGaXs!yTE;;d60E7pV~3F+{$XU7_Tx#EpULK*`R&X5%L)fa;dFY@vkJNO+CkeX(?
zoP?ea<x9yAy$6LBKI@$34hy5D?a-FAYF3!QStHt3T%nZM!B0*$mvfVaO0iKX_)c!7
zP{bGIiZ$1Dyn0Q2xnIk@)F@QS{$fUc{9|s(J19#WityI%nM$qTj~>juS5?i;pdWq!
zI-9SfHxI<RPj&FD<2T%DS0A$N!Q5o0X80!i)M$_6dq`QD_OaI|R6#3GtZ*E{tFsew
z*f9mE@P*{H0%U<;D8%ReT3&tT?Ld1ZBQ0&~86&H^q@N-Y(ZLbr;q3-+G$<o&$`JaL
z`KmSp^(4$oHWU(+4g_lARn5|-k{r{Z|Jf%FkDr~Ko^+362%5L!)2ink*Y|R`p+rRv
z!zydYPmJf0-;~11CwnshzUG5iFB)xQ8Px2Xb5a2cvt;tQB8mSb)m;e~+)aoyludF;
z_fu`#Pq#B7LlsyusRUYDzAwaQTey~(enC3Oo+y3%u@`^%#vT7~s5Cuz=HYLA_umec
z{`&l1erfyIHxHHC4}5R;U0?XlmC~R7@UuUj-~W$S!eRec)cN$^ee1h_8(b-U|Ifek
z)9=6WjVq(N+nFzXKv?y~iP3=|;Wg@x>yk(VJD)E@j^^`$QLP7wg6kF*gOntN{o{dt
z8t7_}K3Q-J^Nx>3lubd>cdA~^MMtyc86RR0LONe<%!5mk^9A26FMum8&#BC~&!9f&
zA)U_`D+SNX=R3sUQp7!o)06gyKA`W?v-(!Oxw&WF_C;@4lpKfpoRuU_et`D?Kud2+
zLq%XVmMx6BK4-Kr&U12hE`#IdtS#D-(K6bJmOhtU)+Nt<GG?D%W+FpL{@;CCekb^8
zy5JXQa|^lsIA@)5Y1T*KLRt8;ikaAC4MP#Bdgye}+n7DZFX2*5x)TIa?(-5uXNmH>
zH<}7kn6vo?=0HwCT1|nd`ErO|O8dY%Uv@5VRtgZ#rb{IeB;=O_Nict-;;-h%6a>VD
zf-A_J9>-L(2RQ|sdZ!?)g|`5`R6z=$mrOoCUlWZA%h`OMj2+%d<@2Ie#8h*F-6?>e
zKb-%fQ>jEWKnV>prDAgVd`XZbTKOWjVHw$!#1(;y4^BRhH!BQ=CTBsVg;R}buP|S)
zK=e09a|)O9I>=nDnxn+X(Kx6&6|Yw?a8CFJJUy!3Ucn%lSU-WWr!sb5B;X+6GdUU2
z*YeBNC-Wv1u+k=abNiaNTuc55)TeNoWV>$z^pbyDUk0iB#+=zQAes`eM$PbFUd>qB
zw6-N6iXghA(7J3bnLYI>(%A0m^{tjES<a&P1<RQ&tLNHk>D_&VWu4DP%Y-l>^epv1
zV}cpe&Hs$~;A+gMD87ps*WsnOlXaPKarbrlh?|pji_=%DJ~ke4|JZnM6ZcIuQ5;WZ
zcb_Gg05~yfC>g+uqoZcfcg}G^h<UHvbb_P@=_;IHs0fS5hv+U=8p6rLN*ztQr_mxv
z6f3o=6X-{SOyPVfU$50F<042msvc}y2VJDX!e~O$q<e;~Qc_Cf#{;7{jniEvCnOV-
zQ9atQT4qGQ55GbDn%mY=Y%DF@51_L9Ji!YDzd{fT1(yb9=Rpy3NKx=2ZXL(j5e25?
zrXq0hK_=w!W<|lk;YzJ|?i7T8?^6!kk&DZrRZy<NE~r!HJJ%8bTk-H!0)X1xlw!a*
zAr0eEOUZ%?>dJ2q^mEq~2h`La&1Vf7vAL!9UDrwjN`hD#OydFyPAbD!jBVQU#*zWv
zB*xsPErBaxcUvtOqjSb~n4fcoun3nEONmyZZ9-Depn^dIOtO`1Tg$Z6VO+76tZ9K;
zQ_WXnvcE}gy+O4ZE8f{X7seiZ0BlDsd?9|i;v3096wNC?!=c!R%PEjJS4-~u3Ai#+
zERreKu&^jKY7OYsE+$ASffDFYZc^bZo*PiS0R>9EDuO|E?Yo64taHy19U4|bof_{P
zPG#5F!+J*FsSm@hSv8lpfKNzO<cSWSSZ;&^S{*mbC&?p|rw=R~Xx@e8m{)XR|G|Rx
z%2g+Kn%|Dshp~qRg`ADpb|x{>p664&{{ZOa5M?_{8`-#QEa?kXKRK83ZDC%dLP}gV
zj%Y8OdEO%7E?G+npH7%$%NWr_YKzvw$_Z^VxUzR(0U6#k!@G&zyCYx*5IdApD=}j|
zqP;Z!99p;gL>e#)BG-3lC0(R&qQ@=hZz<{bw+H+U?Lk`R?V)A4NX-vNShU;N0%I&2
zqHhHE;u<5N{SJ2La(ER0!_%-)SV@^WEwZ`;V|ju%r1<BEk^6D{<j@MoF}!-?(C8*5
zMH)aV33?QzAklQ(n&&>ra=HxL(1*D9U@E$>%!)pi#Po0kX0hf4M!oDl!HSmmlIy#a
zuI_sXXivH{LV|>}``xExxkg(P#$>TjaUwS19rs<ltuF?8Bhb%d$xx}!79!!@YuS7V
zH9a~!D9Q%)oPHALX1%#V5!PB^#u11u`}=SKGeXBonYKJuV6jyXqvJB>Tw3%HYI6y?
z>soXkS<R2e!r|vDnxZ8^uN4wMeC5M<eL1-pn^P!5LH{aIa=72ek{nGbn<%A1gI(%l
z_bl(r!R(AE#mcKV)oR(cVW(&H=0G@>w?LSXxSD(br$E}@X_vSn&iLRI@$mKlTy-&?
zq7J#FcP+>5Abl2I2?*XMtdJaCcJxy4S_%>AsyG{&-Rt={A}7%9$O!~fFv7)-Aui|}
zo})xkv|!{v-<zkm3!qmdajBR$edxHn6<P2J#cv#MrM2zoQ9)`k@6#&<L+vDQe+*!K
z`hN{~<Y1EilX`PF9)XzjCFd8TLG3Z^XVKaRRr|Vc%orC9_YSCU_#eOvFzrX^PtvOg
zu%?AC0+NfXT$t7^lO{Qha@<MEE4ZVVAvdM^XM2k+TUt9s8HW;A!@n?mqikX+4=G4T
z?H~g;ZWS)uY7vqQmYMr9@b1R}Aj+gs>@3s^RiV_poQ-bNHmUqffj$jKA?Iy9zFJ0a
zN>bSa`iPOy?*MOQ^yW}!#v%sO<&;4p-n4N`qWUZ!DETabb#B5Xu81=}B9gcXKAVjB
ztb7$do9tQPN@nR5mJzZ4fAZR=@$E{taSnr2@kZI~FHpUQ&Ef(0C7Z4GR&im=V-1_R
zzsBMJbppv>pJC|}1eXb<np(|U?k9QcHwZpOu#UpI0{ps+)tRGP$5^DG*ZIIM#v-nV
zvE<T<m1tjdD=^2JOU&r8PdPnB=ZI2cDd|<VAt&u18#&0OQ&}8sR4R(gx?JqJLZwu5
z5v`iP^@@z|PvPLvWb0LAuud&MaZ&S=7qw+9*YYWev4Kg4n2g^?o9Z%EUe@oA%LCz>
z6^cyG@&={<eiP47JV|>o$eeQEMPO8>B0Y0*O&%DP`{klT8Vk4@cRA=mrbwrtRQOW7
z1)0SdaedwS&*&m{3=d=Z1Y2iQ_t0O3x6^1xH9E=mmjIyXq!~$b{drR@D%6rd7_XU2
z$%Ut|&V|nm%Z`bq6lxnV@Kf8k7_eI@VV;9yMXFpI{xl0q7P~rtYpWGsql6Y%UhH{J
zYb8YjKCYxlN{D`>f6H!uCAJKEB`ujzE-2S}j^4DhS(J-@r<-W75JgpPFF36I*|jU`
z16WZXP%$39%*xL&XfQ@Uf<JeTfNNPySf{Kj#$Qz<;Zmsu);VX<b3ez1pC|ZjwoZ4J
z+3U3htsv?Izg5-u<U&2LBsZ-iP=0{!q~}$$M?ez;_}TiPF<@kk=1BC(WX*l6<Vt#7
zg)Ob|gM7~$@=K8){+@3TNtof3OI(kHA$Rh4^Qyb=>=7_nXe4jzlye=!m{Sz5wdwQV
z(|CGoy-_}F!q6K9r*!Cu`aRTDA4?Mo{7xfwef#yN>^62qjdZ=-RhW;XlEOL3hEy3n
z_UkSclcDVjYt*bh5ZkVFFVS|D?&R2Z<yB<6B5u>d>K;}u-NnDK)@o|C8s!+TQlnT1
z%5}pT-l7H#Osuu9twtF_SH4%+r-RZywe{x4?ipD{?r*_9C58MqPDSL5!&okHMV#?L
zlH+j`L_XOe?B>o+xhL+T7I36*9ydehUqU-Y=QL%mN#~n5x*a<2dcR-NIpxh-b8!uw
zuVXH*p>y|(oRVvri`&+x^X?hBb#zV&8R7$_FjsU=-1>CB8+W?R#n@0>Klvb44!x&u
zKgOg3&p|@cO*&D3geR{x^aP!sGLj~pm?oT<SmaA5rtT-(aH+QG=FC8F>dT0?wfjP6
zW)=?AegVho+<-{Jo-yEp-7q)k4~Z09u*lepU9i#t3zw+bW&#VuuQ1_)HKmhvBd#)`
z3zm12;a%mRjqkz*8#-WHN$G;E!v!0@U90=8lmoTzL<t{@<0psQ3H&L?=mnH+*s<!K
zXFH0g$Z6UpLaO(pEImO$*W`LWQ1`>U@iM_H1PcUD5-bwXJ?4Ipphd7mKsR3GAeC)m
zFR8MjF0v9W(e;(=K2l)hB+VfQVo*9tn}Z<%#Ukih`47;a)TyxS*Lq1sl>5-@DNX#M
zcuK<=e?!eU2I&+B?4um8q?&}c<T8DT<7=Zg&CclCxs2Eljcg=J-h!@dqu)I}G9>L1
zcS~eQRtAiocH9!fl((_qL}AICj8V#!!!>l}FkYosbVLr$FhyY*6(nz5Q$Zf%7fNp`
z7Gd2l)gex)KG7LRIyM2x2|}mi5fMnlBN1g}1Tq1cU?Kj9zzm?(vchO1&Al+83Yb(Y
z)i!1Pk@d}G8h)wfUtMDz6-I2MB~WLnBQtjP$b20{5%*rrDMC%B-1)mPx}))vh?k8O
zI>xV?WN=#M3|%ebCvK}^K^WIi#vNd3g5V(lNW14)rVTBFmTrw$TUw5Y@2JqUTL&^o
z$X8SGO_%jO0-s==_0v<F$_#F584DiH?F>wg^i1h3;0jXJ<9tx8SxuL?Tf!BCD&ydN
z^*E7dM_6_cK46khu2NWp%9Ufi{#xm}eDMc(+w1d1f!`SNMVJWDd9Ag)kwB*$kx881
z1OW1HK=GHzA#m4U)Dh4^W)1#LUz>y8Zo}h%Oc|sd%70V+tE9ur40olP--#(|tzV%l
z-(@k1tSDzg+=!&zo#;}Wjw<$6=}K%V#>r7P7{p&cQU<TX{`eV^@?!-@S;BFmBJhp^
zP`xoTp)02Wc?ry6ViGMz1&UAFIhMRe_yZC{hwXvf2l#zVV!3QQq3j9CYgP0cu^Cm6
zvKz>7z%s_EIEpOhewknuGttbbasDMUfsHcI+z4*&UQM2nBVykpNINN1yhrzikzc+B
z68uag_(A$k@XA!#eP$7XVZKz9q|@taQLnn{eWp-d@O`AN>&v8UI7XUVul6=?C4fQZ
zc!7`VVsGMZ348a;eB5y)AW8@8-Eq>c#D@twi1IOqS0?^T62{xegUm?vW8%Eg%w4%r
z<*J#xHgb0}*UjARGF&-t@+iIQ2hUvoeD%K$NvbRT*2o#{RPvv$ly3j}Yv0@-{OC&Q
z-uvyZ?VS4QA?b_<35=sz*!sn9jSeUm>>mQT+Ip1+=`r3oPS6SCOVg9NpfaGZvTwsM
zrOL!r>6Vgo-W_%Pa!fV)fU!qr;P%WxN03QThh`kzSdiU^3ni|=rUn)W%)14%UN*bq
z^X;Tm?MMq;W(rZVdHw9$>m=(*0TN4)8`8z=cvML}JXBOKG(Jg9G6`34WvVM^F9(-n
zsdTf{(oDR&q451jIntDFn$0BJ-ch~<dY82FK|Uj9%#=%94`a&b19`Inkbxskx1Y5+
z(+PnHN4jm*K7*fU5QGFMS^O{xy#}stUIYo|WgrCqNk1j8k%lk-5de|@2!6<a75IRA
zG>mwB1sy7OqbnfD7V{!@$x1I+kX`IXwOPw5W3Q03aDvplGS^)S#2~2*5;+ODs8wpN
zcm3GM7dVB{l8|Ct{hC{PM?U6DNxQ`q9>}8~Cw0IiPVf5w(2TI3g^p-PYT-k$sx=wy
za1r0Ig@O3SNwBQmO0{emvoPTkNc*rv1QG=<BKe?|P$ffK$}II&cap(-Ycs)0-Gw%`
zT50IzG)jq9x|M15Ez8=7k8DPKL_<;n!fJR@O2$Z1g0U8JDT_O)7WU*VW$C#PHeiO=
z@HW`(KAc%z;WU;F)Xp*i41y%x{Fhh;thIn5a&)jyuq!yyokScZ7RoS4+!aMadA>Be
zb|C=lg|%3fWDxGZd6Rb0HKfuF#bG8e$E(6M?iYX??RS5Nx4udsBfEdZ(jOCijo?oR
zzD~duc3{qzvDX7p7%yKbpF^_6d|@HT$P_b)Kz}D$`6C2#1dI$#SsW(`FR0kg@AB3>
z!D|F{f-e)CAy~(3`F+y+cHGqZ7}2p|OzkviP-XPZ#&+xqx(msZt7t!xmSH-^Es5@s
z_dgShPb6jxr(NQTe6{F^#pfPix*dW7QDW>z!UFbE8vBtLIqu$%WCwP%SF!oy)3aqy
zMUCS~aZc^f$l-=p;SQ6^B3y>yHaN8pvUq-6#p$AsuydhauW?6{vayl(SETT*mcmV5
zy&MsDCwJBD#+JAWIN{g~@;MvMUgk}CxnjffvE{P2>e@W3y}9yC?kKF;s2vk@1a^`j
z0SlY)_Q?bi$xWwYvz`JwI=zFq=oQ>8-QDWLW~e58MB=f90a>46q#Dl5xSU=Usm>@f
z>2>s{<o2U1NtY3yA;WWl9SH(CpV-S6E2>a$2Lrude2uKGzafQW(KWNWEInx?v1_e4
z6wOI2m5cmV9z6n?v5XdSTM-#E<IJv|h}6ktUv>AwJxD~OV;8wu7j*=sN0uz!ZDH%g
zMV<c;q<bMq^>$(zP8DH>iy|a!3kwO~DTsA!+}Mo8z&X=h5J>Fi6|yK(3{HpmPXRBl
zfVY`}%S-^M1E_%+ZVN$HEVhL(@6v|)Fok{DK=3?G1pr)H33fPgn3J1q*?_E*l!6({
zFS+T25Y8S6ldl-u{K5NDXn!u2w7n!sxx0|-Y`i*Ftv7sS^zY407iTZ*&AFxNy}1wE
zw>NkH3vnlGhQKMeN3PuiskT!>V652+ne&k0Kksyc-_jCa`Cnv_a~}q#N>hMSNg|SY
zXg+thVegX@B+eHq4F{I{Cvkr?vpV=k<3xFjKf#j2PE738aKFLJKLdzFs$6eAMl5>@
z^m3rj2Kt4m#4PuAOeHSxZ?)h}yP#sk7S_)|{~pOShu*ZL1wzdjBRYRJh*kZz1G5Xh
z=H_5by0K6_{Z^KTWW{gu>Cy^KTE;-eC2sUCwF6`k<sc#V3%W;9R@%o_f}>hhYZ`+$
zIAY3Q|0GLt#UmrCa>a83xFF%o*ZoEJGb~#(<w`~F$d#K&gCw?*V}n{JRi%@rA~RBC
zo{AI@(z0`D@k#HkH07iPCoMNAG~_3AX|H$<71)ocjc||B0+W6nE40`!(59t7h~GUJ
nbnchQ=tuNH(@xq$!&`^94R0CVF}w@Edr{to@>Z0$4Bz*E`yG_j

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/usm_sharp.cpython-310.pyc b/utils/image/__pycache__/usm_sharp.cpython-310.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..906e0bb05e0850062d44a567c2cc80967ea74e87
GIT binary patch
literal 1235
zcmYjQO>f&q5Zy0Qq%6C3sy1%XYk}$#6-I(2K+ptm&=d&bL)2b)+0xvVM2qAqyJUhG
z<jDqd>u(^Bx#S1!sXz~f{z5LvEM+GiV4r4pXJ<xl9@Xyd5P@BO`bU1C2>s_aH;uvO
z2`uvnKp=rL6m5TTMEX6As33x7EcYT0qi0Ak;awu(NqXj=4K8RD2s}d*e*+oN$p!;X
zunp#ER!V*JG;zl_&6RX!egTju!U935@Ew|CK?S=Eq2N#qL~w~C_6{)=d60g`27k?7
z%$BiUZ;s%hc6Ac?OJB_89l&yD=5K%!mAFP1q$M4~sKx?+j|5o}eE>eaj~ZOVJ*OpW
zJ@80J8e018pbXlurZsJc9qLGj+Z_Se&>FQP!D`wO3=~06-pPhQoxT<AJnOHdE@U>@
zwVtWgQft?g4LDDWwB+3SIu>bVEKOb?MF$I6K8q`3(ztluzq4Mcbh5NNzh^2g|CEJM
z+LB^zsZga2OC1+xt&Ft8sxYrADbFRh-)Ol=4RFm*tNC2&3A4<|Y_8qpp^tQt$J!C|
z9@~)fB#VvV{5|^k$8@Q3IW1HvPnB9t6D8ynnl`G4v+3;NuXi6DJ-oL)K0cnkoK|I;
znQ5BG3puS!&W)SKo%LHCLiQb4W)ENwk8pr}z@P9wK0mk~+ns9`{)Zs2e;;pLhsm`b
z+?h+{!8(RzmH;Kptwv|~f;0>~c8D6UBrUDI4(2&VHC?gBul<&FsPQ4%zlt8XiPqj2
z-9}9SFT6T%GVW*-)Pp)0BSGuIA!K(EFiUo_`2_X8&T`^wr=*@aJp#|t_|#Z$?nYx-
zlBMfOsK4zm(tKh4H!@u;-ISJEnx)EwWv5xCZ2;R;RB>im9-CDZ8Z|F@?|MCh`nIdn
z!=Xzzm@EAz)?y4Bvk&0-yhQ?FbcFZt7@zNd;q)qWLbRi;?;iXgoc|-h#CPH=;#`9G
zhMeb0RGITfoWFvWwm1F!^v@7)cdrDJ$SmXhMrYr-l&-yA(r!-12MEv}xfS+jz=qR+
z^@=J7O*koRdGiyM=c?%G1T`1df6=$mzq(Q4D+9r!ZKJ=rkK8zw1V8~0Ko03=PLn1B

literal 0
HcmV?d00001

diff --git a/utils/image/__pycache__/usm_sharp.cpython-37.pyc b/utils/image/__pycache__/usm_sharp.cpython-37.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..9f34252182fb5678c65b1f7fd0de904c2ba1e88a
GIT binary patch
literal 1220
zcmYjQO>g8x5bbuiJ@#Z0cG<8SBu<C}b6HEda9IIbR;<vnIS`PLFOB1F&p6|c+3q$g
zBKwp``#0<(7cTspJ|S^H^B1^4m6IXVsNYn(tE<ZIm7gCT&I#C$zdusHF+%>t&E7aT
zU&1oa03?yLB+2bRO}uH(5+*&#N?!SiPsvLnx%6*{^cA}dE@#&)2_-!z%U}l?$k~np
zE~yLiqO?k%zR2+L-mn~+`4m8rgh~QsqIV>xl1Y9OL9w}<$?%3I{2k#W@genr3x7C&
zeZEfhc6SOFHP*=BRVKrnGg!DcKLJ=`X-BTT!5axB9hLMkk>1ABAA`?6BR%clTrkTA
zA3SgLdS-)RX2T)sSjUF>NJej@!+`|sSx1J2<Q*G53e-S9{@KohI#U}OJr``0u2s1_
zbiQe~N^5M&g+df{VTEvkPG!*=$FiSKlZPv1U#6`wMOwd_?m6E!I$OJgugfO2->cd*
z+IjWXG1*ubS)JBq+Zg5MZEfDPN?j@8-qUJT7~oi3w0W-dk~?lxnQK%V`Y2Xas*xB_
zBN8GjQ)7hqmHhoGUN@DB>&B{!rrE?YwQ0Q4O}mZH+xpw0US(@l$5|s){G!P7=ihu6
zx3(xvTvX{w#jUA?K{1|gf6);XIDlo200=#z3mVb@@FRNlaIf0aJrRFl5FvkjvBM^_
z{WxF}AJ$i}%o@OgY#q6z*Iv)T<0qu|tv9gFA3@Gz(y<NigDx2ONcsSxgS+S%3flR{
z<O%6RxZ!spPVdP2u$y(^F_EmBoj`Vs04e#|?jNXkFXb}OI7kz5dI6qe>4kBAj><Tm
zmBn@$=||H+QLS9?Ru!u?inP|stea9geo?mCg|IDTo0g7Oso5lvX>uzj*PqzakFm}K
zPcYpqZ}i(#%VXG>2LJ@Q^xg+P7w|ixR}cQ<^DcDhX{2rH9{dyZKLl6?ID2(kDVV*v
z5LF}F68(h`Z=j{ylSw=M3B<d@yZOjeSqgFA*^=rHFex@T%`_@yP`1!Jil!&P2I}wp
yx~*Uya13sF3$muFntGxWhPigZ>#2?D)_sZpF%UkxZS-^erc8)c;N?Sa!Ttr-_$JE$

literal 0
HcmV?d00001

diff --git a/utils/image/align_color.py b/utils/image/align_color.py
new file mode 100644
index 0000000..36da591
--- /dev/null
+++ b/utils/image/align_color.py
@@ -0,0 +1,119 @@
+'''
+# --------------------------------------------------------------------------------
+#   Color fixed script from Li Yi (https://github.com/pkuliyi2015/sd-webui-stablesr/blob/master/srmodule/colorfix.py)
+# --------------------------------------------------------------------------------
+'''
+
+import torch
+from PIL import Image
+from torch import Tensor
+from torch.nn import functional as F
+from torchvision.transforms import ToTensor, ToPILImage
+
+
+def adain_color_fix(target: Image, source: Image):
+    # Convert images to tensors
+    to_tensor = ToTensor()
+    target_tensor = to_tensor(target).unsqueeze(0)
+    source_tensor = to_tensor(source).unsqueeze(0)
+
+    # Apply adaptive instance normalization
+    result_tensor = adaptive_instance_normalization(target_tensor, source_tensor)
+
+    # Convert tensor back to image
+    to_image = ToPILImage()
+    result_image = to_image(result_tensor.squeeze(0).clamp_(0.0, 1.0))
+
+    return result_image
+
+def wavelet_color_fix(target: Image, source: Image):
+    # Convert images to tensors
+    to_tensor = ToTensor()
+    target_tensor = to_tensor(target).unsqueeze(0)
+    source_tensor = to_tensor(source).unsqueeze(0)
+
+    # Apply wavelet reconstruction
+    result_tensor = wavelet_reconstruction(target_tensor, source_tensor)
+
+    # Convert tensor back to image
+    to_image = ToPILImage()
+    result_image = to_image(result_tensor.squeeze(0).clamp_(0.0, 1.0))
+
+    return result_image
+
+def calc_mean_std(feat: Tensor, eps=1e-5):
+    """Calculate mean and std for adaptive_instance_normalization.
+    Args:
+        feat (Tensor): 4D tensor.
+        eps (float): A small value added to the variance to avoid
+            divide-by-zero. Default: 1e-5.
+    """
+    size = feat.size()
+    assert len(size) == 4, 'The input feature should be 4D tensor.'
+    b, c = size[:2]
+    feat_var = feat.reshape(b, c, -1).var(dim=2) + eps
+    feat_std = feat_var.sqrt().reshape(b, c, 1, 1)
+    feat_mean = feat.reshape(b, c, -1).mean(dim=2).reshape(b, c, 1, 1)
+    return feat_mean, feat_std
+
+def adaptive_instance_normalization(content_feat:Tensor, style_feat:Tensor):
+    """Adaptive instance normalization.
+    Adjust the reference features to have the similar color and illuminations
+    as those in the degradate features.
+    Args:
+        content_feat (Tensor): The reference feature.
+        style_feat (Tensor): The degradate features.
+    """
+    size = content_feat.size()
+    style_mean, style_std = calc_mean_std(style_feat)
+    content_mean, content_std = calc_mean_std(content_feat)
+    normalized_feat = (content_feat - content_mean.expand(size)) / content_std.expand(size)
+    return normalized_feat * style_std.expand(size) + style_mean.expand(size)
+
+def wavelet_blur(image: Tensor, radius: int):
+    """
+    Apply wavelet blur to the input tensor.
+    """
+    # input shape: (1, 3, H, W)
+    # convolution kernel
+    kernel_vals = [
+        [0.0625, 0.125, 0.0625],
+        [0.125, 0.25, 0.125],
+        [0.0625, 0.125, 0.0625],
+    ]
+    kernel = torch.tensor(kernel_vals, dtype=image.dtype, device=image.device)
+    # add channel dimensions to the kernel to make it a 4D tensor
+    kernel = kernel[None, None]
+    # repeat the kernel across all input channels
+    kernel = kernel.repeat(3, 1, 1, 1)
+    image = F.pad(image, (radius, radius, radius, radius), mode='replicate')
+    # apply convolution
+    output = F.conv2d(image, kernel, groups=3, dilation=radius)
+    return output
+
+def wavelet_decomposition(image: Tensor, levels=5):
+    """
+    Apply wavelet decomposition to the input tensor.
+    This function only returns the low frequency & the high frequency.
+    """
+    high_freq = torch.zeros_like(image)
+    for i in range(levels):
+        radius = 2 ** i
+        low_freq = wavelet_blur(image, radius)
+        high_freq += (image - low_freq)
+        image = low_freq
+
+    return high_freq, low_freq
+
+def wavelet_reconstruction(content_feat:Tensor, style_feat:Tensor):
+    """
+    Apply wavelet decomposition, so that the content will have the same color as the style.
+    """
+    # calculate the wavelet decomposition of the content feature
+    content_high_freq, content_low_freq = wavelet_decomposition(content_feat)
+    del content_low_freq
+    # calculate the wavelet decomposition of the style feature
+    style_high_freq, style_low_freq = wavelet_decomposition(style_feat)
+    del style_high_freq
+    # reconstruct the content feature with the style's high frequency
+    return content_high_freq + style_low_freq
\ No newline at end of file
diff --git a/utils/image/common.py b/utils/image/common.py
new file mode 100644
index 0000000..2e8ca3c
--- /dev/null
+++ b/utils/image/common.py
@@ -0,0 +1,235 @@
+import random
+import math
+
+from PIL import Image
+import numpy as np
+import cv2
+import torch
+from torch.nn import functional as F
+
+
+# https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/image_datasets.py
+def center_crop_arr(pil_image, image_size):
+    # We are not on a new enough PIL to support the `reducing_gap`
+    # argument, which uses BOX downsampling at powers of two first.
+    # Thus, we do it by hand to improve downsample quality.
+    while min(*pil_image.size) >= 2 * image_size:
+        pil_image = pil_image.resize(
+            tuple(x // 2 for x in pil_image.size), resample=Image.BOX
+        )
+
+    scale = image_size / min(*pil_image.size)
+    pil_image = pil_image.resize(
+        tuple(round(x * scale) for x in pil_image.size), resample=Image.BICUBIC
+    )
+
+    arr = np.array(pil_image)
+    crop_y = (arr.shape[0] - image_size) // 2
+    crop_x = (arr.shape[1] - image_size) // 2
+    return arr[crop_y : crop_y + image_size, crop_x : crop_x + image_size]
+
+
+# https://github.com/openai/guided-diffusion/blob/main/guided_diffusion/image_datasets.py
+def random_crop_arr(pil_image, image_size, min_crop_frac=0.8, max_crop_frac=1.0):
+    min_smaller_dim_size = math.ceil(image_size / max_crop_frac)
+    max_smaller_dim_size = math.ceil(image_size / min_crop_frac)
+    smaller_dim_size = random.randrange(min_smaller_dim_size, max_smaller_dim_size + 1)
+
+    # We are not on a new enough PIL to support the `reducing_gap`
+    # argument, which uses BOX downsampling at powers of two first.
+    # Thus, we do it by hand to improve downsample quality.
+    while min(*pil_image.size) >= 2 * smaller_dim_size:
+        pil_image = pil_image.resize(
+            tuple(x // 2 for x in pil_image.size), resample=Image.BOX
+        )
+
+    scale = smaller_dim_size / min(*pil_image.size)
+    pil_image = pil_image.resize(
+        tuple(round(x * scale) for x in pil_image.size), resample=Image.BICUBIC
+    )
+
+    arr = np.array(pil_image)
+    crop_y = random.randrange(arr.shape[0] - image_size + 1)
+    crop_x = random.randrange(arr.shape[1] - image_size + 1)
+    return arr[crop_y : crop_y + image_size, crop_x : crop_x + image_size]
+
+
+# https://github.com/XPixelGroup/BasicSR/blob/master/basicsr/data/transforms.py
+def augment(imgs, hflip=True, rotation=True, flows=None, return_status=False):
+    """Augment: horizontal flips OR rotate (0, 90, 180, 270 degrees).
+
+    We use vertical flip and transpose for rotation implementation.
+    All the images in the list use the same augmentation.
+
+    Args:
+        imgs (list[ndarray] | ndarray): Images to be augmented. If the input
+            is an ndarray, it will be transformed to a list.
+        hflip (bool): Horizontal flip. Default: True.
+        rotation (bool): Ratotation. Default: True.
+        flows (list[ndarray]: Flows to be augmented. If the input is an
+            ndarray, it will be transformed to a list.
+            Dimension is (h, w, 2). Default: None.
+        return_status (bool): Return the status of flip and rotation.
+            Default: False.
+
+    Returns:
+        list[ndarray] | ndarray: Augmented images and flows. If returned
+            results only have one element, just return ndarray.
+
+    """
+    hflip = hflip and random.random() < 0.5
+    vflip = rotation and random.random() < 0.5
+    rot90 = rotation and random.random() < 0.5
+
+    def _augment(img):
+        if hflip:  # horizontal
+            cv2.flip(img, 1, img)
+        if vflip:  # vertical
+            cv2.flip(img, 0, img)
+        if rot90:
+            img = img.transpose(1, 0, 2)
+        return img
+
+    def _augment_flow(flow):
+        if hflip:  # horizontal
+            cv2.flip(flow, 1, flow)
+            flow[:, :, 0] *= -1
+        if vflip:  # vertical
+            cv2.flip(flow, 0, flow)
+            flow[:, :, 1] *= -1
+        if rot90:
+            flow = flow.transpose(1, 0, 2)
+            flow = flow[:, :, [1, 0]]
+        return flow
+
+    if not isinstance(imgs, list):
+        imgs = [imgs]
+    imgs = [_augment(img) for img in imgs]
+    if len(imgs) == 1:
+        imgs = imgs[0]
+
+    if flows is not None:
+        if not isinstance(flows, list):
+            flows = [flows]
+        flows = [_augment_flow(flow) for flow in flows]
+        if len(flows) == 1:
+            flows = flows[0]
+        return imgs, flows
+    else:
+        if return_status:
+            return imgs, (hflip, vflip, rot90)
+        else:
+            return imgs
+
+
+# https://github.com/XPixelGroup/BasicSR/blob/master/basicsr/utils/img_process_util.py
+def filter2D(img, kernel):
+    """PyTorch version of cv2.filter2D
+
+    Args:
+        img (Tensor): (b, c, h, w)
+        kernel (Tensor): (b, k, k)
+    """
+    k = kernel.size(-1)
+    b, c, h, w = img.size()
+    if k % 2 == 1:
+        img = F.pad(img, (k // 2, k // 2, k // 2, k // 2), mode='reflect')
+    else:
+        raise ValueError('Wrong kernel size')
+
+    ph, pw = img.size()[-2:]
+
+    if kernel.size(0) == 1:
+        # apply the same kernel to all batch images
+        img = img.view(b * c, 1, ph, pw)
+        kernel = kernel.view(1, 1, k, k)
+        return F.conv2d(img, kernel, padding=0).view(b, c, h, w)
+    else:
+        img = img.view(1, b * c, ph, pw)
+        kernel = kernel.view(b, 1, k, k).repeat(1, c, 1, 1).view(b * c, 1, k, k)
+        return F.conv2d(img, kernel, groups=b * c).view(b, c, h, w)
+
+
+# https://github.com/XPixelGroup/BasicSR/blob/033cd6896d898fdd3dcda32e3102a792efa1b8f4/basicsr/utils/color_util.py#L186
+def rgb2ycbcr_pt(img, y_only=False):
+    """Convert RGB images to YCbCr images (PyTorch version).
+
+    It implements the ITU-R BT.601 conversion for standard-definition television. See more details in
+    https://en.wikipedia.org/wiki/YCbCr#ITU-R_BT.601_conversion.
+
+    Args:
+        img (Tensor): Images with shape (n, 3, h, w), the range [0, 1], float, RGB format.
+         y_only (bool): Whether to only return Y channel. Default: False.
+
+    Returns:
+        (Tensor): converted images with the shape (n, 3/1, h, w), the range [0, 1], float.
+    """
+    if y_only:
+        weight = torch.tensor([[65.481], [128.553], [24.966]]).to(img)
+        out_img = torch.matmul(img.permute(0, 2, 3, 1), weight).permute(0, 3, 1, 2) + 16.0
+    else:
+        weight = torch.tensor([[65.481, -37.797, 112.0], [128.553, -74.203, -93.786], [24.966, 112.0, -18.214]]).to(img)
+        bias = torch.tensor([16, 128, 128]).view(1, 3, 1, 1).to(img)
+        out_img = torch.matmul(img.permute(0, 2, 3, 1), weight).permute(0, 3, 1, 2) + bias
+
+    out_img = out_img / 255.
+    return out_img
+
+
+def to_pil_image(inputs, mem_order, val_range, channel_order):
+    # convert inputs to numpy array
+    if isinstance(inputs, torch.Tensor):
+        inputs = inputs.cpu().numpy()
+    assert isinstance(inputs, np.ndarray)
+    
+    # make sure that inputs is a 4-dimension array
+    if mem_order in ["hwc", "chw"]:
+        inputs = inputs[None, ...]
+        mem_order = f"n{mem_order}"
+    # to NHWC
+    if mem_order == "nchw":
+        inputs = inputs.transpose(0, 2, 3, 1)
+    # to RGB
+    if channel_order == "bgr":
+        inputs = inputs[..., ::-1].copy()
+    else:
+        assert channel_order == "rgb"
+    
+    if val_range == "0,1":
+        inputs = inputs * 255
+    elif val_range == "-1,1":
+        inputs = (inputs + 1) * 127.5
+    else:
+        assert val_range == "0,255"
+    
+    inputs = inputs.clip(0, 255).astype(np.uint8)
+    return [inputs[i] for i in range(len(inputs))]
+
+
+def put_text(pil_img_arr, text):
+    cv_img = pil_img_arr[..., ::-1].copy()
+    cv2.putText(cv_img, text, (10, 35), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
+    return cv_img[..., ::-1].copy()
+
+
+def auto_resize(img: Image.Image, size: int) -> Image.Image:
+    short_edge = min(img.size)
+    if short_edge < size:
+        r = size / short_edge
+        img = img.resize(
+            tuple(math.ceil(x * r) for x in img.size), Image.BICUBIC
+        )
+    else:
+        # make a deep copy of this image for safety
+        img = img.copy()
+    return img
+
+
+def pad(img: np.ndarray, scale: int) -> np.ndarray:
+    h, w = img.shape[:2]
+    ph = 0 if h % scale == 0 else math.ceil(h / scale) * scale - h
+    pw = 0 if w % scale == 0 else math.ceil(w / scale) * scale - w
+    return np.pad(
+        img, pad_width=((0, ph), (0, pw), (0, 0)), mode="constant",
+        constant_values=0
+    )
diff --git a/utils/image/diffjpeg.py b/utils/image/diffjpeg.py
new file mode 100644
index 0000000..3392821
--- /dev/null
+++ b/utils/image/diffjpeg.py
@@ -0,0 +1,492 @@
+# https://github.com/XPixelGroup/BasicSR/blob/master/basicsr/utils/diffjpeg.py
+"""
+Modified from https://github.com/mlomnitz/DiffJPEG
+
+For images not divisible by 8
+https://dsp.stackexchange.com/questions/35339/jpeg-dct-padding/35343#35343
+"""
+import itertools
+import numpy as np
+import torch
+import torch.nn as nn
+from torch.nn import functional as F
+
+# ------------------------ utils ------------------------#
+y_table = np.array(
+    [[16, 11, 10, 16, 24, 40, 51, 61], [12, 12, 14, 19, 26, 58, 60, 55], [14, 13, 16, 24, 40, 57, 69, 56],
+     [14, 17, 22, 29, 51, 87, 80, 62], [18, 22, 37, 56, 68, 109, 103, 77], [24, 35, 55, 64, 81, 104, 113, 92],
+     [49, 64, 78, 87, 103, 121, 120, 101], [72, 92, 95, 98, 112, 100, 103, 99]],
+    dtype=np.float32).T
+y_table = nn.Parameter(torch.from_numpy(y_table))
+c_table = np.empty((8, 8), dtype=np.float32)
+c_table.fill(99)
+c_table[:4, :4] = np.array([[17, 18, 24, 47], [18, 21, 26, 66], [24, 26, 56, 99], [47, 66, 99, 99]]).T
+c_table = nn.Parameter(torch.from_numpy(c_table))
+
+
+def diff_round(x):
+    """ Differentiable rounding function
+    """
+    return torch.round(x) + (x - torch.round(x))**3
+
+
+def quality_to_factor(quality):
+    """ Calculate factor corresponding to quality
+
+    Args:
+        quality(float): Quality for jpeg compression.
+
+    Returns:
+        float: Compression factor.
+    """
+    if quality < 50:
+        quality = 5000. / quality
+    else:
+        quality = 200. - quality * 2
+    return quality / 100.
+
+
+# ------------------------ compression ------------------------#
+class RGB2YCbCrJpeg(nn.Module):
+    """ Converts RGB image to YCbCr
+    """
+
+    def __init__(self):
+        super(RGB2YCbCrJpeg, self).__init__()
+        matrix = np.array([[0.299, 0.587, 0.114], [-0.168736, -0.331264, 0.5], [0.5, -0.418688, -0.081312]],
+                          dtype=np.float32).T
+        self.shift = nn.Parameter(torch.tensor([0., 128., 128.]))
+        self.matrix = nn.Parameter(torch.from_numpy(matrix))
+
+    def forward(self, image):
+        """
+        Args:
+            image(Tensor): batch x 3 x height x width
+
+        Returns:
+            Tensor: batch x height x width x 3
+        """
+        image = image.permute(0, 2, 3, 1)
+        result = torch.tensordot(image, self.matrix, dims=1) + self.shift
+        return result.view(image.shape)
+
+
+class ChromaSubsampling(nn.Module):
+    """ Chroma subsampling on CbCr channels
+    """
+
+    def __init__(self):
+        super(ChromaSubsampling, self).__init__()
+
+    def forward(self, image):
+        """
+        Args:
+            image(tensor): batch x height x width x 3
+
+        Returns:
+            y(tensor): batch x height x width
+            cb(tensor): batch x height/2 x width/2
+            cr(tensor): batch x height/2 x width/2
+        """
+        image_2 = image.permute(0, 3, 1, 2).clone()
+        cb = F.avg_pool2d(image_2[:, 1, :, :].unsqueeze(1), kernel_size=2, stride=(2, 2), count_include_pad=False)
+        cr = F.avg_pool2d(image_2[:, 2, :, :].unsqueeze(1), kernel_size=2, stride=(2, 2), count_include_pad=False)
+        cb = cb.permute(0, 2, 3, 1)
+        cr = cr.permute(0, 2, 3, 1)
+        return image[:, :, :, 0], cb.squeeze(3), cr.squeeze(3)
+
+
+class BlockSplitting(nn.Module):
+    """ Splitting image into patches
+    """
+
+    def __init__(self):
+        super(BlockSplitting, self).__init__()
+        self.k = 8
+
+    def forward(self, image):
+        """
+        Args:
+            image(tensor): batch x height x width
+
+        Returns:
+            Tensor:  batch x h*w/64 x h x w
+        """
+        height, _ = image.shape[1:3]
+        batch_size = image.shape[0]
+        image_reshaped = image.view(batch_size, height // self.k, self.k, -1, self.k)
+        image_transposed = image_reshaped.permute(0, 1, 3, 2, 4)
+        return image_transposed.contiguous().view(batch_size, -1, self.k, self.k)
+
+
+class DCT8x8(nn.Module):
+    """ Discrete Cosine Transformation
+    """
+
+    def __init__(self):
+        super(DCT8x8, self).__init__()
+        tensor = np.zeros((8, 8, 8, 8), dtype=np.float32)
+        for x, y, u, v in itertools.product(range(8), repeat=4):
+            tensor[x, y, u, v] = np.cos((2 * x + 1) * u * np.pi / 16) * np.cos((2 * y + 1) * v * np.pi / 16)
+        alpha = np.array([1. / np.sqrt(2)] + [1] * 7)
+        self.tensor = nn.Parameter(torch.from_numpy(tensor).float())
+        self.scale = nn.Parameter(torch.from_numpy(np.outer(alpha, alpha) * 0.25).float())
+
+    def forward(self, image):
+        """
+        Args:
+            image(tensor): batch x height x width
+
+        Returns:
+            Tensor: batch x height x width
+        """
+        image = image - 128
+        result = self.scale * torch.tensordot(image, self.tensor, dims=2)
+        result.view(image.shape)
+        return result
+
+
+class YQuantize(nn.Module):
+    """ JPEG Quantization for Y channel
+
+    Args:
+        rounding(function): rounding function to use
+    """
+
+    def __init__(self, rounding):
+        super(YQuantize, self).__init__()
+        self.rounding = rounding
+        self.y_table = y_table
+
+    def forward(self, image, factor=1):
+        """
+        Args:
+            image(tensor): batch x height x width
+
+        Returns:
+            Tensor: batch x height x width
+        """
+        if isinstance(factor, (int, float)):
+            image = image.float() / (self.y_table * factor)
+        else:
+            b = factor.size(0)
+            table = self.y_table.expand(b, 1, 8, 8) * factor.view(b, 1, 1, 1)
+            image = image.float() / table
+        image = self.rounding(image)
+        return image
+
+
+class CQuantize(nn.Module):
+    """ JPEG Quantization for CbCr channels
+
+    Args:
+        rounding(function): rounding function to use
+    """
+
+    def __init__(self, rounding):
+        super(CQuantize, self).__init__()
+        self.rounding = rounding
+        self.c_table = c_table
+
+    def forward(self, image, factor=1):
+        """
+        Args:
+            image(tensor): batch x height x width
+
+        Returns:
+            Tensor: batch x height x width
+        """
+        if isinstance(factor, (int, float)):
+            image = image.float() / (self.c_table * factor)
+        else:
+            b = factor.size(0)
+            table = self.c_table.expand(b, 1, 8, 8) * factor.view(b, 1, 1, 1)
+            image = image.float() / table
+        image = self.rounding(image)
+        return image
+
+
+class CompressJpeg(nn.Module):
+    """Full JPEG compression algorithm
+
+    Args:
+        rounding(function): rounding function to use
+    """
+
+    def __init__(self, rounding=torch.round):
+        super(CompressJpeg, self).__init__()
+        self.l1 = nn.Sequential(RGB2YCbCrJpeg(), ChromaSubsampling())
+        self.l2 = nn.Sequential(BlockSplitting(), DCT8x8())
+        self.c_quantize = CQuantize(rounding=rounding)
+        self.y_quantize = YQuantize(rounding=rounding)
+
+    def forward(self, image, factor=1):
+        """
+        Args:
+            image(tensor): batch x 3 x height x width
+
+        Returns:
+            dict(tensor): Compressed tensor with batch x h*w/64 x 8 x 8.
+        """
+        y, cb, cr = self.l1(image * 255)
+        components = {'y': y, 'cb': cb, 'cr': cr}
+        for k in components.keys():
+            comp = self.l2(components[k])
+            if k in ('cb', 'cr'):
+                comp = self.c_quantize(comp, factor=factor)
+            else:
+                comp = self.y_quantize(comp, factor=factor)
+
+            components[k] = comp
+
+        return components['y'], components['cb'], components['cr']
+
+
+# ------------------------ decompression ------------------------#
+
+
+class YDequantize(nn.Module):
+    """Dequantize Y channel
+    """
+
+    def __init__(self):
+        super(YDequantize, self).__init__()
+        self.y_table = y_table
+
+    def forward(self, image, factor=1):
+        """
+        Args:
+            image(tensor): batch x height x width
+
+        Returns:
+            Tensor: batch x height x width
+        """
+        if isinstance(factor, (int, float)):
+            out = image * (self.y_table * factor)
+        else:
+            b = factor.size(0)
+            table = self.y_table.expand(b, 1, 8, 8) * factor.view(b, 1, 1, 1)
+            out = image * table
+        return out
+
+
+class CDequantize(nn.Module):
+    """Dequantize CbCr channel
+    """
+
+    def __init__(self):
+        super(CDequantize, self).__init__()
+        self.c_table = c_table
+
+    def forward(self, image, factor=1):
+        """
+        Args:
+            image(tensor): batch x height x width
+
+        Returns:
+            Tensor: batch x height x width
+        """
+        if isinstance(factor, (int, float)):
+            out = image * (self.c_table * factor)
+        else:
+            b = factor.size(0)
+            table = self.c_table.expand(b, 1, 8, 8) * factor.view(b, 1, 1, 1)
+            out = image * table
+        return out
+
+
+class iDCT8x8(nn.Module):
+    """Inverse discrete Cosine Transformation
+    """
+
+    def __init__(self):
+        super(iDCT8x8, self).__init__()
+        alpha = np.array([1. / np.sqrt(2)] + [1] * 7)
+        self.alpha = nn.Parameter(torch.from_numpy(np.outer(alpha, alpha)).float())
+        tensor = np.zeros((8, 8, 8, 8), dtype=np.float32)
+        for x, y, u, v in itertools.product(range(8), repeat=4):
+            tensor[x, y, u, v] = np.cos((2 * u + 1) * x * np.pi / 16) * np.cos((2 * v + 1) * y * np.pi / 16)
+        self.tensor = nn.Parameter(torch.from_numpy(tensor).float())
+
+    def forward(self, image):
+        """
+        Args:
+            image(tensor): batch x height x width
+
+        Returns:
+            Tensor: batch x height x width
+        """
+        image = image * self.alpha
+        result = 0.25 * torch.tensordot(image, self.tensor, dims=2) + 128
+        result.view(image.shape)
+        return result
+
+
+class BlockMerging(nn.Module):
+    """Merge patches into image
+    """
+
+    def __init__(self):
+        super(BlockMerging, self).__init__()
+
+    def forward(self, patches, height, width):
+        """
+        Args:
+            patches(tensor) batch x height*width/64, height x width
+            height(int)
+            width(int)
+
+        Returns:
+            Tensor: batch x height x width
+        """
+        k = 8
+        batch_size = patches.shape[0]
+        image_reshaped = patches.view(batch_size, height // k, width // k, k, k)
+        image_transposed = image_reshaped.permute(0, 1, 3, 2, 4)
+        return image_transposed.contiguous().view(batch_size, height, width)
+
+
+class ChromaUpsampling(nn.Module):
+    """Upsample chroma layers
+    """
+
+    def __init__(self):
+        super(ChromaUpsampling, self).__init__()
+
+    def forward(self, y, cb, cr):
+        """
+        Args:
+            y(tensor): y channel image
+            cb(tensor): cb channel
+            cr(tensor): cr channel
+
+        Returns:
+            Tensor: batch x height x width x 3
+        """
+
+        def repeat(x, k=2):
+            height, width = x.shape[1:3]
+            x = x.unsqueeze(-1)
+            x = x.repeat(1, 1, k, k)
+            x = x.view(-1, height * k, width * k)
+            return x
+
+        cb = repeat(cb)
+        cr = repeat(cr)
+        return torch.cat([y.unsqueeze(3), cb.unsqueeze(3), cr.unsqueeze(3)], dim=3)
+
+
+class YCbCr2RGBJpeg(nn.Module):
+    """Converts YCbCr image to RGB JPEG
+    """
+
+    def __init__(self):
+        super(YCbCr2RGBJpeg, self).__init__()
+
+        matrix = np.array([[1., 0., 1.402], [1, -0.344136, -0.714136], [1, 1.772, 0]], dtype=np.float32).T
+        self.shift = nn.Parameter(torch.tensor([0, -128., -128.]))
+        self.matrix = nn.Parameter(torch.from_numpy(matrix))
+
+    def forward(self, image):
+        """
+        Args:
+            image(tensor): batch x height x width x 3
+
+        Returns:
+            Tensor: batch x 3 x height x width
+        """
+        result = torch.tensordot(image + self.shift, self.matrix, dims=1)
+        return result.view(image.shape).permute(0, 3, 1, 2)
+
+
+class DeCompressJpeg(nn.Module):
+    """Full JPEG decompression algorithm
+
+    Args:
+        rounding(function): rounding function to use
+    """
+
+    def __init__(self, rounding=torch.round):
+        super(DeCompressJpeg, self).__init__()
+        self.c_dequantize = CDequantize()
+        self.y_dequantize = YDequantize()
+        self.idct = iDCT8x8()
+        self.merging = BlockMerging()
+        self.chroma = ChromaUpsampling()
+        self.colors = YCbCr2RGBJpeg()
+
+    def forward(self, y, cb, cr, imgh, imgw, factor=1):
+        """
+        Args:
+            compressed(dict(tensor)): batch x h*w/64 x 8 x 8
+            imgh(int)
+            imgw(int)
+            factor(float)
+
+        Returns:
+            Tensor: batch x 3 x height x width
+        """
+        components = {'y': y, 'cb': cb, 'cr': cr}
+        for k in components.keys():
+            if k in ('cb', 'cr'):
+                comp = self.c_dequantize(components[k], factor=factor)
+                height, width = int(imgh / 2), int(imgw / 2)
+            else:
+                comp = self.y_dequantize(components[k], factor=factor)
+                height, width = imgh, imgw
+            comp = self.idct(comp)
+            components[k] = self.merging(comp, height, width)
+            #
+        image = self.chroma(components['y'], components['cb'], components['cr'])
+        image = self.colors(image)
+
+        image = torch.min(255 * torch.ones_like(image), torch.max(torch.zeros_like(image), image))
+        return image / 255
+
+
+# ------------------------ main DiffJPEG ------------------------ #
+
+
+class DiffJPEG(nn.Module):
+    """This JPEG algorithm result is slightly different from cv2.
+    DiffJPEG supports batch processing.
+
+    Args:
+        differentiable(bool): If True, uses custom differentiable rounding function, if False, uses standard torch.round
+    """
+
+    def __init__(self, differentiable=True):
+        super(DiffJPEG, self).__init__()
+        if differentiable:
+            rounding = diff_round
+        else:
+            rounding = torch.round
+
+        self.compress = CompressJpeg(rounding=rounding)
+        self.decompress = DeCompressJpeg(rounding=rounding)
+
+    def forward(self, x, quality):
+        """
+        Args:
+            x (Tensor): Input image, bchw, rgb, [0, 1]
+            quality(float): Quality factor for jpeg compression scheme.
+        """
+        factor = quality
+        if isinstance(factor, (int, float)):
+            factor = quality_to_factor(factor)
+        else:
+            for i in range(factor.size(0)):
+                factor[i] = quality_to_factor(factor[i])
+        h, w = x.size()[-2:]
+        h_pad, w_pad = 0, 0
+        # why should use 16
+        if h % 16 != 0:
+            h_pad = 16 - h % 16
+        if w % 16 != 0:
+            w_pad = 16 - w % 16
+        x = F.pad(x, (0, w_pad, 0, h_pad), mode='constant', value=0)
+
+        y, cb, cr = self.compress(x, factor=factor)
+        recovered = self.decompress(y, cb, cr, (h + h_pad), (w + w_pad), factor=factor)
+        recovered = recovered[:, :, 0:h, 0:w]
+        return recovered
diff --git a/utils/image/usm_sharp.py b/utils/image/usm_sharp.py
new file mode 100644
index 0000000..7b83532
--- /dev/null
+++ b/utils/image/usm_sharp.py
@@ -0,0 +1,29 @@
+# https://github.com/XPixelGroup/BasicSR/blob/master/basicsr/utils/img_process_util.py
+import cv2
+import numpy as np
+import torch
+
+from .common import filter2D
+
+
+class USMSharp(torch.nn.Module):
+
+    def __init__(self, radius=50, sigma=0):
+        super(USMSharp, self).__init__()
+        if radius % 2 == 0:
+            radius += 1
+        self.radius = radius
+        kernel = cv2.getGaussianKernel(radius, sigma)
+        kernel = torch.FloatTensor(np.dot(kernel, kernel.transpose())).unsqueeze_(0)
+        self.register_buffer('kernel', kernel)
+
+    def forward(self, img, weight=0.5, threshold=10):
+        blur = filter2D(img, self.kernel)
+        residual = img - blur
+
+        mask = torch.abs(residual) * 255 > threshold
+        mask = mask.float()
+        soft_mask = filter2D(mask, self.kernel)
+        sharp = img + weight * residual
+        sharp = torch.clip(sharp, 0, 1)
+        return soft_mask * sharp + (1 - soft_mask) * img
diff --git a/utils/metrics.py b/utils/metrics.py
new file mode 100644
index 0000000..e1c3e49
--- /dev/null
+++ b/utils/metrics.py
@@ -0,0 +1,66 @@
+import torch
+import lpips
+
+from .image import rgb2ycbcr_pt
+from .common import frozen_module
+
+
+# https://github.com/XPixelGroup/BasicSR/blob/033cd6896d898fdd3dcda32e3102a792efa1b8f4/basicsr/metrics/psnr_ssim.py#L52
+def calculate_psnr_pt(img, img2, crop_border, test_y_channel=False):
+    """Calculate PSNR (Peak Signal-to-Noise Ratio) (PyTorch version).
+
+    Reference: https://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio
+
+    Args:
+        img (Tensor): Images with range [0, 1], shape (n, 3/1, h, w).
+        img2 (Tensor): Images with range [0, 1], shape (n, 3/1, h, w).
+        crop_border (int): Cropped pixels in each edge of an image. These pixels are not involved in the calculation.
+        test_y_channel (bool): Test on Y channel of YCbCr. Default: False.
+
+    Returns:
+        float: PSNR result.
+    """
+
+    assert img.shape == img2.shape, (f'Image shapes are different: {img.shape}, {img2.shape}.')
+
+    if crop_border != 0:
+        img = img[:, :, crop_border:-crop_border, crop_border:-crop_border]
+        img2 = img2[:, :, crop_border:-crop_border, crop_border:-crop_border]
+
+    if test_y_channel:
+        img = rgb2ycbcr_pt(img, y_only=True)
+        img2 = rgb2ycbcr_pt(img2, y_only=True)
+
+    img = img.to(torch.float64)
+    img2 = img2.to(torch.float64)
+
+    mse = torch.mean((img - img2)**2, dim=[1, 2, 3])
+    return 10. * torch.log10(1. / (mse + 1e-8))
+
+
+class LPIPS:
+    
+    def __init__(self, net: str) -> None:
+        self.model = lpips.LPIPS(net=net)
+        frozen_module(self.model)
+    
+    @torch.no_grad()
+    def __call__(self, img1: torch.Tensor, img2: torch.Tensor, normalize: bool) -> torch.Tensor:
+        """
+        Compute LPIPS.
+        
+        Args:
+            img1 (torch.Tensor): The first image (NCHW, RGB, [-1, 1]). Specify `normalize` if input 
+                image is range in [0, 1].
+            img2 (torch.Tensor): The second image (NCHW, RGB, [-1, 1]). Specify `normalize` if input 
+                image is range in [0, 1].
+            normalize (bool): If specified, the input images will be normalized from [0, 1] to [-1, 1].
+            
+        Returns:
+            lpips_values (torch.Tensor): The lpips scores of this batch.
+        """
+        return self.model(img1, img2, normalize=normalize)
+    
+    def to(self, device: str) -> "LPIPS":
+        self.model.to(device)
+        return self
diff --git a/utils/realesrgan/realesrganer.py b/utils/realesrgan/realesrganer.py
new file mode 100644
index 0000000..ce10fa9
--- /dev/null
+++ b/utils/realesrgan/realesrganer.py
@@ -0,0 +1,339 @@
+import cv2
+import math
+import numpy as np
+import os
+import queue
+import threading
+import torch
+from torch.nn import functional as F
+
+from utils.file import load_file_from_url
+from utils.realesrgan.rrdbnet import RRDBNet
+
+# ROOT_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+
+class RealESRGANer():
+    """A helper class for upsampling images with RealESRGAN.
+
+    Args:
+        scale (int): Upsampling scale factor used in the networks. It is usually 2 or 4.
+        model_path (str): The path to the pretrained model. It can be urls (will first download it automatically).
+        model (nn.Module): The defined network. Default: None.
+        tile (int): As too large images result in the out of GPU memory issue, so this tile option will first crop
+            input images into tiles, and then process each of them. Finally, they will be merged into one image.
+            0 denotes for do not use tile. Default: 0.
+        tile_pad (int): The pad size for each tile, to remove border artifacts. Default: 10.
+        pre_pad (int): Pad the input images to avoid border artifacts. Default: 10.
+        half (float): Whether to use half precision during inference. Default: False.
+    """
+
+    def __init__(self,
+                 scale,
+                 model_path,
+                 model=None,
+                 tile=0,
+                 tile_pad=10,
+                 pre_pad=10,
+                 half=False,
+                 device=None):
+        self.scale = scale
+        self.tile_size = tile
+        self.tile_pad = tile_pad
+        self.pre_pad = pre_pad
+        self.mod_scale = None
+        self.half = half
+
+        # initialize model
+        # if gpu_id:
+        #     self.device = torch.device(
+        #         f'cuda:{gpu_id}' if torch.cuda.is_available() else 'cpu') if device is None else device
+        # else:
+        #     self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') if device is None else device
+
+        self.device = device
+        
+        # if the model_path starts with https, it will first download models to the folder: realesrgan/weights
+        if model_path.startswith('https://'):
+            model_path = load_file_from_url(
+                url=model_path, model_dir=os.path.join('weights/realesrgan'), progress=True, file_name=None)
+        loadnet = torch.load(model_path, map_location=torch.device('cpu'))
+        # prefer to use params_ema
+        if 'params_ema' in loadnet:
+            keyname = 'params_ema'
+        else:
+            keyname = 'params'
+        model.load_state_dict(loadnet[keyname], strict=True)
+        model.eval()
+        self.model = model.to(self.device)
+        if self.half:
+            self.model = self.model.half()
+
+    def pre_process(self, img):
+        """Pre-process, such as pre-pad and mod pad, so that the images can be divisible
+        """
+        img = torch.from_numpy(np.transpose(img, (2, 0, 1))).float()
+        self.img = img.unsqueeze(0).to(self.device)
+        if self.half:
+            self.img = self.img.half()
+
+        # pre_pad
+        if self.pre_pad != 0:
+            self.img = F.pad(self.img, (0, self.pre_pad, 0, self.pre_pad), 'reflect')
+        # mod pad for divisible borders
+        if self.scale == 2:
+            self.mod_scale = 2
+        elif self.scale == 1:
+            self.mod_scale = 4
+        if self.mod_scale is not None:
+            self.mod_pad_h, self.mod_pad_w = 0, 0
+            _, _, h, w = self.img.size()
+            if (h % self.mod_scale != 0):
+                self.mod_pad_h = (self.mod_scale - h % self.mod_scale)
+            if (w % self.mod_scale != 0):
+                self.mod_pad_w = (self.mod_scale - w % self.mod_scale)
+            self.img = F.pad(self.img, (0, self.mod_pad_w, 0, self.mod_pad_h), 'reflect')
+
+    def process(self):
+        # model inference
+        self.output = self.model(self.img)
+
+    def tile_process(self):
+        """It will first crop input images to tiles, and then process each tile.
+        Finally, all the processed tiles are merged into one images.
+
+        Modified from: https://github.com/ata4/esrgan-launcher
+        """
+        batch, channel, height, width = self.img.shape
+        output_height = height * self.scale
+        output_width = width * self.scale
+        output_shape = (batch, channel, output_height, output_width)
+
+        # start with black image
+        self.output = self.img.new_zeros(output_shape)
+        tiles_x = math.ceil(width / self.tile_size)
+        tiles_y = math.ceil(height / self.tile_size)
+
+        # loop over all tiles
+        for y in range(tiles_y):
+            for x in range(tiles_x):
+                # extract tile from input image
+                ofs_x = x * self.tile_size
+                ofs_y = y * self.tile_size
+                # input tile area on total image
+                input_start_x = ofs_x
+                input_end_x = min(ofs_x + self.tile_size, width)
+                input_start_y = ofs_y
+                input_end_y = min(ofs_y + self.tile_size, height)
+
+                # input tile area on total image with padding
+                input_start_x_pad = max(input_start_x - self.tile_pad, 0)
+                input_end_x_pad = min(input_end_x + self.tile_pad, width)
+                input_start_y_pad = max(input_start_y - self.tile_pad, 0)
+                input_end_y_pad = min(input_end_y + self.tile_pad, height)
+
+                # input tile dimensions
+                input_tile_width = input_end_x - input_start_x
+                input_tile_height = input_end_y - input_start_y
+                tile_idx = y * tiles_x + x + 1
+                input_tile = self.img[:, :, input_start_y_pad:input_end_y_pad, input_start_x_pad:input_end_x_pad]
+
+                # upscale tile
+                try:
+                    with torch.no_grad():
+                        output_tile = self.model(input_tile)
+                except RuntimeError as error:
+                    print('Error', error)
+                # print(f'\tTile {tile_idx}/{tiles_x * tiles_y}')
+
+                # output tile area on total image
+                output_start_x = input_start_x * self.scale
+                output_end_x = input_end_x * self.scale
+                output_start_y = input_start_y * self.scale
+                output_end_y = input_end_y * self.scale
+
+                # output tile area without padding
+                output_start_x_tile = (input_start_x - input_start_x_pad) * self.scale
+                output_end_x_tile = output_start_x_tile + input_tile_width * self.scale
+                output_start_y_tile = (input_start_y - input_start_y_pad) * self.scale
+                output_end_y_tile = output_start_y_tile + input_tile_height * self.scale
+
+                # put tile into output image
+                self.output[:, :, output_start_y:output_end_y,
+                            output_start_x:output_end_x] = output_tile[:, :, output_start_y_tile:output_end_y_tile,
+                                                                       output_start_x_tile:output_end_x_tile]
+
+    def post_process(self):
+        # remove extra pad
+        if self.mod_scale is not None:
+            _, _, h, w = self.output.size()
+            self.output = self.output[:, :, 0:h - self.mod_pad_h * self.scale, 0:w - self.mod_pad_w * self.scale]
+        # remove prepad
+        if self.pre_pad != 0:
+            _, _, h, w = self.output.size()
+            self.output = self.output[:, :, 0:h - self.pre_pad * self.scale, 0:w - self.pre_pad * self.scale]
+        return self.output
+
+    @torch.no_grad()
+    def enhance(self, img, outscale=None, alpha_upsampler='realesrgan'):
+        h_input, w_input = img.shape[0:2]
+        # img: numpy
+        img = img.astype(np.float32)
+        if np.max(img) > 256:  # 16-bit image
+            max_range = 65535
+            print('\tInput is a 16-bit image')
+        else:
+            max_range = 255
+        img = img / max_range
+        if len(img.shape) == 2:  # gray image
+            img_mode = 'L'
+            img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
+        elif img.shape[2] == 4:  # RGBA image with alpha channel
+            img_mode = 'RGBA'
+            alpha = img[:, :, 3]
+            img = img[:, :, 0:3]
+            img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+            if alpha_upsampler == 'realesrgan':
+                alpha = cv2.cvtColor(alpha, cv2.COLOR_GRAY2RGB)
+        else:
+            img_mode = 'RGB'
+            img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+
+        # ------------------- process image (without the alpha channel) ------------------- #
+        try:
+            with torch.no_grad():
+                self.pre_process(img)
+                if self.tile_size > 0:
+                    self.tile_process()
+                else:
+                    self.process()
+                output_img_t = self.post_process()
+                output_img = output_img_t.data.squeeze().float().cpu().clamp_(0, 1).numpy()
+                output_img = np.transpose(output_img[[2, 1, 0], :, :], (1, 2, 0))
+                if img_mode == 'L':
+                    output_img = cv2.cvtColor(output_img, cv2.COLOR_BGR2GRAY)
+            del output_img_t
+            torch.cuda.empty_cache()        
+        except RuntimeError as error:
+            print(f"Failed inference for RealESRGAN: {error}")      
+
+        # ------------------- process the alpha channel if necessary ------------------- #
+        if img_mode == 'RGBA':
+            if alpha_upsampler == 'realesrgan':
+                self.pre_process(alpha)
+                if self.tile_size > 0:
+                    self.tile_process()
+                else:
+                    self.process()
+                output_alpha = self.post_process()
+                output_alpha = output_alpha.data.squeeze().float().cpu().clamp_(0, 1).numpy()
+                output_alpha = np.transpose(output_alpha[[2, 1, 0], :, :], (1, 2, 0))
+                output_alpha = cv2.cvtColor(output_alpha, cv2.COLOR_BGR2GRAY)
+            else:  # use the cv2 resize for alpha channel
+                h, w = alpha.shape[0:2]
+                output_alpha = cv2.resize(alpha, (w * self.scale, h * self.scale), interpolation=cv2.INTER_LINEAR)
+
+            # merge the alpha channel
+            output_img = cv2.cvtColor(output_img, cv2.COLOR_BGR2BGRA)
+            output_img[:, :, 3] = output_alpha
+
+        # ------------------------------ return ------------------------------ #
+        if max_range == 65535:  # 16-bit image
+            output = (output_img * 65535.0).round().astype(np.uint16)
+        else:
+            output = (output_img * 255.0).round().astype(np.uint8)
+
+        if outscale is not None and outscale != float(self.scale):
+            output = cv2.resize(
+                output, (
+                    int(w_input * outscale),
+                    int(h_input * outscale),
+                ), interpolation=cv2.INTER_LANCZOS4)
+
+        return output, img_mode
+
+
+class PrefetchReader(threading.Thread):
+    """Prefetch images.
+
+    Args:
+        img_list (list[str]): A image list of image paths to be read.
+        num_prefetch_queue (int): Number of prefetch queue.
+    """
+
+    def __init__(self, img_list, num_prefetch_queue):
+        super().__init__()
+        self.que = queue.Queue(num_prefetch_queue)
+        self.img_list = img_list
+
+    def run(self):
+        for img_path in self.img_list:
+            img = cv2.imread(img_path, cv2.IMREAD_UNCHANGED)
+            self.que.put(img)
+
+        self.que.put(None)
+
+    def __next__(self):
+        next_item = self.que.get()
+        if next_item is None:
+            raise StopIteration
+        return next_item
+
+    def __iter__(self):
+        return self
+
+
+class IOConsumer(threading.Thread):
+
+    def __init__(self, opt, que, qid):
+        super().__init__()
+        self._queue = que
+        self.qid = qid
+        self.opt = opt
+
+    def run(self):
+        while True:
+            msg = self._queue.get()
+            if isinstance(msg, str) and msg == 'quit':
+                break
+
+            output = msg['output']
+            save_path = msg['save_path']
+            cv2.imwrite(save_path, output)
+        print(f'IO worker {self.qid} is done.')
+
+def set_realesrgan(bg_tile, device, scale=2):
+    '''
+        scale: options: 2, 4. Default: 2. RealESRGAN official models only support x2 and x4 upsampling.
+    '''
+    assert isinstance(scale, int), 'Expected param scale to be an integer!'
+
+    use_half = False
+    if 'cuda' in str(device): # set False in CPU/MPS mode
+        no_half_gpu_list = ['1650', '1660'] # set False for GPUs that don't support f16
+        if not True in [gpu in torch.cuda.get_device_name(0) for gpu in no_half_gpu_list]:
+            use_half = True
+
+    model_url = {
+        2: 'https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.1/RealESRGAN_x2plus.pth',
+        4: 'https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth'
+    }
+
+    model = RRDBNet(
+        num_in_ch=3,
+        num_out_ch=3,
+        num_feat=64,
+        num_block=23,
+        num_grow_ch=32,
+        scale=scale,
+    )
+    upsampler = RealESRGANer(
+        scale=scale,
+        model_path=model_url[scale],
+        model=model,
+        tile=bg_tile,
+        tile_pad=40,
+        pre_pad=0,
+        device=device,
+        half=use_half
+    )
+    return upsampler
\ No newline at end of file
diff --git a/utils/realesrgan/rrdbnet.py b/utils/realesrgan/rrdbnet.py
new file mode 100644
index 0000000..11b23a3
--- /dev/null
+++ b/utils/realesrgan/rrdbnet.py
@@ -0,0 +1,183 @@
+import torch
+from torch import nn as nn
+from torch.nn import functional as F
+from torch.nn import init as init
+from torch.nn.modules.batchnorm import _BatchNorm
+
+
+def default_init_weights(module_list, scale=1, bias_fill=0, **kwargs):
+    """Initialize network weights.
+
+    Args:
+        module_list (list[nn.Module] | nn.Module): Modules to be initialized.
+        scale (float): Scale initialized weights, especially for residual
+            blocks. Default: 1.
+        bias_fill (float): The value to fill bias. Default: 0
+        kwargs (dict): Other arguments for initialization function.
+    """
+    if not isinstance(module_list, list):
+        module_list = [module_list]
+    for module in module_list:
+        for m in module.modules():
+            if isinstance(m, nn.Conv2d):
+                init.kaiming_normal_(m.weight, **kwargs)
+                m.weight.data *= scale
+                if m.bias is not None:
+                    m.bias.data.fill_(bias_fill)
+            elif isinstance(m, nn.Linear):
+                init.kaiming_normal_(m.weight, **kwargs)
+                m.weight.data *= scale
+                if m.bias is not None:
+                    m.bias.data.fill_(bias_fill)
+            elif isinstance(m, _BatchNorm):
+                init.constant_(m.weight, 1)
+                if m.bias is not None:
+                    m.bias.data.fill_(bias_fill)
+
+
+def make_layer(basic_block, num_basic_block, **kwarg):
+    """Make layers by stacking the same blocks.
+
+    Args:
+        basic_block (nn.module): nn.module class for basic block.
+        num_basic_block (int): number of blocks.
+
+    Returns:
+        nn.Sequential: Stacked blocks in nn.Sequential.
+    """
+    layers = []
+    for _ in range(num_basic_block):
+        layers.append(basic_block(**kwarg))
+    return nn.Sequential(*layers)
+
+
+# TODO: may write a cpp file
+def pixel_unshuffle(x, scale):
+    """ Pixel unshuffle.
+
+    Args:
+        x (Tensor): Input feature with shape (b, c, hh, hw).
+        scale (int): Downsample ratio.
+
+    Returns:
+        Tensor: the pixel unshuffled feature.
+    """
+    b, c, hh, hw = x.size()
+    out_channel = c * (scale**2)
+    assert hh % scale == 0 and hw % scale == 0
+    h = hh // scale
+    w = hw // scale
+    x_view = x.view(b, c, h, scale, w, scale)
+    return x_view.permute(0, 1, 3, 5, 2, 4).reshape(b, out_channel, h, w)
+
+
+class ResidualDenseBlock(nn.Module):
+    """Residual Dense Block.
+
+    Used in RRDB block in ESRGAN.
+
+    Args:
+        num_feat (int): Channel number of intermediate features.
+        num_grow_ch (int): Channels for each growth.
+    """
+
+    def __init__(self, num_feat=64, num_grow_ch=32):
+        super(ResidualDenseBlock, self).__init__()
+        self.conv1 = nn.Conv2d(num_feat, num_grow_ch, 3, 1, 1)
+        self.conv2 = nn.Conv2d(num_feat + num_grow_ch, num_grow_ch, 3, 1, 1)
+        self.conv3 = nn.Conv2d(num_feat + 2 * num_grow_ch, num_grow_ch, 3, 1, 1)
+        self.conv4 = nn.Conv2d(num_feat + 3 * num_grow_ch, num_grow_ch, 3, 1, 1)
+        self.conv5 = nn.Conv2d(num_feat + 4 * num_grow_ch, num_feat, 3, 1, 1)
+
+        self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True)
+
+        # initialization
+        default_init_weights([self.conv1, self.conv2, self.conv3, self.conv4, self.conv5], 0.1)
+
+    def forward(self, x):
+        x1 = self.lrelu(self.conv1(x))
+        x2 = self.lrelu(self.conv2(torch.cat((x, x1), 1)))
+        x3 = self.lrelu(self.conv3(torch.cat((x, x1, x2), 1)))
+        x4 = self.lrelu(self.conv4(torch.cat((x, x1, x2, x3), 1)))
+        x5 = self.conv5(torch.cat((x, x1, x2, x3, x4), 1))
+        # Empirically, we use 0.2 to scale the residual for better performance
+        return x5 * 0.2 + x
+
+
+class RRDB(nn.Module):
+    """Residual in Residual Dense Block.
+
+    Used in RRDB-Net in ESRGAN.
+
+    Args:
+        num_feat (int): Channel number of intermediate features.
+        num_grow_ch (int): Channels for each growth.
+    """
+
+    def __init__(self, num_feat, num_grow_ch=32):
+        super(RRDB, self).__init__()
+        self.rdb1 = ResidualDenseBlock(num_feat, num_grow_ch)
+        self.rdb2 = ResidualDenseBlock(num_feat, num_grow_ch)
+        self.rdb3 = ResidualDenseBlock(num_feat, num_grow_ch)
+
+    def forward(self, x):
+        out = self.rdb1(x)
+        out = self.rdb2(out)
+        out = self.rdb3(out)
+        # Empirically, we use 0.2 to scale the residual for better performance
+        return out * 0.2 + x
+
+
+class RRDBNet(nn.Module):
+    """Networks consisting of Residual in Residual Dense Block, which is used
+    in ESRGAN.
+
+    ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks.
+
+    We extend ESRGAN for scale x2 and scale x1.
+    Note: This is one option for scale 1, scale 2 in RRDBNet.
+    We first employ the pixel-unshuffle (an inverse operation of pixelshuffle to reduce the spatial size
+    and enlarge the channel size before feeding inputs into the main ESRGAN architecture.
+
+    Args:
+        num_in_ch (int): Channel number of inputs.
+        num_out_ch (int): Channel number of outputs.
+        num_feat (int): Channel number of intermediate features.
+            Default: 64
+        num_block (int): Block number in the trunk network. Defaults: 23
+        num_grow_ch (int): Channels for each growth. Default: 32.
+    """
+
+    def __init__(self, num_in_ch, num_out_ch, scale=4, num_feat=64, num_block=23, num_grow_ch=32):
+        super(RRDBNet, self).__init__()
+        self.scale = scale
+        if scale == 2:
+            num_in_ch = num_in_ch * 4
+        elif scale == 1:
+            num_in_ch = num_in_ch * 16
+        self.conv_first = nn.Conv2d(num_in_ch, num_feat, 3, 1, 1)
+        self.body = make_layer(RRDB, num_block, num_feat=num_feat, num_grow_ch=num_grow_ch)
+        self.conv_body = nn.Conv2d(num_feat, num_feat, 3, 1, 1)
+        # upsample
+        self.conv_up1 = nn.Conv2d(num_feat, num_feat, 3, 1, 1)
+        self.conv_up2 = nn.Conv2d(num_feat, num_feat, 3, 1, 1)
+        self.conv_hr = nn.Conv2d(num_feat, num_feat, 3, 1, 1)
+        self.conv_last = nn.Conv2d(num_feat, num_out_ch, 3, 1, 1)
+
+        self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True)
+
+    def forward(self, x):
+        if self.scale == 2:
+            feat = pixel_unshuffle(x, scale=2)
+        elif self.scale == 1:
+            feat = pixel_unshuffle(x, scale=4)
+        else:
+            feat = x
+        feat = self.conv_first(feat)
+        body_feat = self.conv_body(self.body(feat))
+        feat = feat + body_feat
+        # upsample
+        feat = self.lrelu(self.conv_up1(F.interpolate(feat, scale_factor=2, mode='nearest')))
+        feat = self.lrelu(self.conv_up2(F.interpolate(feat, scale_factor=2, mode='nearest')))
+        out = self.conv_last(self.lrelu(self.conv_hr(feat)))
+        return out
\ No newline at end of file