From 03c281241c5c291488817c732985e19e91f0a17d Mon Sep 17 00:00:00 2001
From: smthem <2102561739@qq.com>
Date: Mon, 2 Jun 2025 20:31:57 +0800
Subject: [PATCH] init
---
Bagel/EVAL.md | 78 +
Bagel/LICENSE | 201 +
Bagel/README.md | 199 +
Bagel/TRAIN.md | 168 +
Bagel/app.py | 256 ++
Bagel/assets/arch.png | Bin 0 -> 168377 bytes
Bagel/assets/emerging_curves.png | Bin 0 -> 372807 bytes
Bagel/assets/teaser.webp | Bin 0 -> 1104072 bytes
Bagel/data/__init__.py | 2 +
Bagel/data/configs/example.yaml | 45 +
Bagel/data/data_utils.py | 177 +
Bagel/data/dataset_base.py | 620 +++
Bagel/data/dataset_info.py | 39 +
Bagel/data/distributed_iterable_dataset.py | 58 +
Bagel/data/interleave_datasets/__init__.py | 5 +
.../data/interleave_datasets/edit_dataset.py | 72 +
.../interleave_t2i_dataset.py | 212 +
Bagel/data/parquet_utils.py | 89 +
Bagel/data/t2i_dataset.py | 128 +
Bagel/data/transforms.py | 287 ++
Bagel/data/video_utils.py | 165 +
Bagel/data/vlm_dataset.py | 195 +
Bagel/eval/gen/gen_images_mp.py | 0
Bagel/inference.ipynb | 535 +++
Bagel/inferencer.py | 309 ++
Bagel/modeling/__init__.py | 4 +
Bagel/modeling/autoencoder.py | 360 ++
Bagel/modeling/bagel/__init__.py | 18 +
Bagel/modeling/bagel/bagel.py | 1030 +++++
Bagel/modeling/bagel/modeling_utils.py | 144 +
Bagel/modeling/bagel/qwen2_navit.py | 1157 ++++++
Bagel/modeling/bagel/siglip_navit.py | 402 ++
Bagel/modeling/qwen2/__init__.py | 68 +
Bagel/modeling/qwen2/configuration_qwen2.py | 179 +
Bagel/modeling/qwen2/modeling_qwen2.py | 929 +++++
Bagel/modeling/qwen2/tokenization_qwen2.py | 328 ++
.../modeling/qwen2/tokenization_qwen2_fast.py | 123 +
Bagel/modeling/siglip/__init__.py | 98 +
Bagel/modeling/siglip/configuration_siglip.py | 287 ++
Bagel/modeling/siglip/convert_siglip_to_hf.py | 401 ++
.../siglip/image_processing_siglip.py | 230 ++
Bagel/modeling/siglip/modeling_siglip.py | 1557 +++++++
Bagel/modeling/siglip/processing_siglip.py | 131 +
Bagel/modeling/siglip/tokenization_siglip.py | 364 ++
Bagel/requirements.txt | 18 +
Bagel/test_images/meme.jpg | Bin 0 -> 59919 bytes
Bagel/test_images/octupusy.jpg | Bin 0 -> 35825 bytes
Bagel/test_images/women.jpg | Bin 0 -> 86746 bytes
README.md | 34 +-
Storydiffusion_node.py | 120 +-
.../storydiffusion_workflow.json | 3645 ++++++++---------
images/bagel_img2img.png | Bin 0 -> 370836 bytes
images/bagel_txt2img.png | Bin 0 -> 432755 bytes
53 files changed, 13446 insertions(+), 2021 deletions(-)
create mode 100644 Bagel/EVAL.md
create mode 100644 Bagel/LICENSE
create mode 100644 Bagel/README.md
create mode 100644 Bagel/TRAIN.md
create mode 100644 Bagel/app.py
create mode 100644 Bagel/assets/arch.png
create mode 100644 Bagel/assets/emerging_curves.png
create mode 100644 Bagel/assets/teaser.webp
create mode 100644 Bagel/data/__init__.py
create mode 100644 Bagel/data/configs/example.yaml
create mode 100644 Bagel/data/data_utils.py
create mode 100644 Bagel/data/dataset_base.py
create mode 100644 Bagel/data/dataset_info.py
create mode 100644 Bagel/data/distributed_iterable_dataset.py
create mode 100644 Bagel/data/interleave_datasets/__init__.py
create mode 100644 Bagel/data/interleave_datasets/edit_dataset.py
create mode 100644 Bagel/data/interleave_datasets/interleave_t2i_dataset.py
create mode 100644 Bagel/data/parquet_utils.py
create mode 100644 Bagel/data/t2i_dataset.py
create mode 100644 Bagel/data/transforms.py
create mode 100644 Bagel/data/video_utils.py
create mode 100644 Bagel/data/vlm_dataset.py
create mode 100644 Bagel/eval/gen/gen_images_mp.py
create mode 100644 Bagel/inference.ipynb
create mode 100644 Bagel/inferencer.py
create mode 100644 Bagel/modeling/__init__.py
create mode 100644 Bagel/modeling/autoencoder.py
create mode 100644 Bagel/modeling/bagel/__init__.py
create mode 100644 Bagel/modeling/bagel/bagel.py
create mode 100644 Bagel/modeling/bagel/modeling_utils.py
create mode 100644 Bagel/modeling/bagel/qwen2_navit.py
create mode 100644 Bagel/modeling/bagel/siglip_navit.py
create mode 100644 Bagel/modeling/qwen2/__init__.py
create mode 100644 Bagel/modeling/qwen2/configuration_qwen2.py
create mode 100644 Bagel/modeling/qwen2/modeling_qwen2.py
create mode 100644 Bagel/modeling/qwen2/tokenization_qwen2.py
create mode 100644 Bagel/modeling/qwen2/tokenization_qwen2_fast.py
create mode 100644 Bagel/modeling/siglip/__init__.py
create mode 100644 Bagel/modeling/siglip/configuration_siglip.py
create mode 100644 Bagel/modeling/siglip/convert_siglip_to_hf.py
create mode 100644 Bagel/modeling/siglip/image_processing_siglip.py
create mode 100644 Bagel/modeling/siglip/modeling_siglip.py
create mode 100644 Bagel/modeling/siglip/processing_siglip.py
create mode 100644 Bagel/modeling/siglip/tokenization_siglip.py
create mode 100644 Bagel/requirements.txt
create mode 100644 Bagel/test_images/meme.jpg
create mode 100644 Bagel/test_images/octupusy.jpg
create mode 100644 Bagel/test_images/women.jpg
create mode 100644 images/bagel_img2img.png
create mode 100644 images/bagel_txt2img.png
diff --git a/Bagel/EVAL.md b/Bagel/EVAL.md
new file mode 100644
index 0000000..a2c268b
--- /dev/null
+++ b/Bagel/EVAL.md
@@ -0,0 +1,78 @@
+# VLM
+We follow [InternVL2](https://internvl.readthedocs.io/en/latest/internvl2.0/evaluation.html) to evaluate the performance on MME, MMBench, MMMU, MMVet, MathVista and MMVP.
+
+## Data prepration
+Please follow the [InternVL2](https://internvl.readthedocs.io/en/latest/get_started/eval_data_preparation.html) to prepare the corresponding data. And the link the data under `vlm`.
+
+The final directory structure is:
+```shell
+data
+├── MathVista
+├── mmbench
+├── mme
+├── MMMU
+├── mm-vet
+└── MMVP
+```
+
+## Evaluation
+
+Directly run `scripts/eval/run_eval_vlm.sh` to evaluate different benchmarks. The output will be saved in `$output_path`.
+- Set `$model_path` and `$output_path` for the path for checkpoint and log.
+- Increase `GPUS` if you want to run faster.
+- For MMBench, please use the official [evaluation server](https://mmbench.opencompass.org.cn/mmbench-submission).
+- For MMVet, please use the official [evaluation server](https://huggingface.co/spaces/whyu/MM-Vet_Evaluator).
+- For MathVista, please set `$openai_api_key` in `scripts/eval/run_eval_vlm.sh` and `your_api_url` in `eval/vlm/eval/mathvista/utilities.py`. The default GPT version is `gpt-4o-2024-11-20`.
+- For MMMU, we use CoT in the report, which improve the accuracy by about 2%. For evaluation of the oprn-ended answer, we use GPT-4o for judgement.
+
+
+# GenEval
+We modify the code in [GenEval](https://github.com/djghosh13/geneval/tree/main) for faster evaluation.
+
+## Setup
+Install the following dependencies:
+```shell
+pip install open-clip-torch
+pip install clip-benchmark
+pip install --upgrade setuptools
+
+sudo pip install -U openmim
+sudo mim install mmengine mmcv-full==1.7.2
+
+git clone https://github.com/open-mmlab/mmdetection.git
+cd mmdetection; git checkout 2.x
+pip install -v -e .
+```
+
+Download Detector:
+```shell
+cd ./eval/gen/geneval
+mkdir model
+
+bash ./evaluation/download_models.sh ./model
+```
+
+## Evaluation
+Directly run `scripts/eval/run_geneval.sh` to evaluate GenEVAL. The output will be saved in `$output_path`.
+- Set `$model_path` and `$output_path` for the path for checkpoint and log.
+- Set `metadata_file` to `./eval/gen/geneval/prompts/evaluation_metadata.jsonl` for original GenEval prompts.
+
+
+# WISE
+We modify the code in [WISE](https://github.com/PKU-YuanGroup/WISE/tree/main) for faster evaluation.
+
+
+## Evaluation
+Directly run `scripts/eval/run_wise.sh` to evaluate WISE. The output will be saved in `$output_path`.
+- Set `$model_path` and `$output_path` for the path for checkpoint and log.
+- Set `$openai_api_key` in `scripts/eval/run_wise.sh` and `your_api_url` in `eval/gen/wise/gpt_eval_mp.py`. The default GPT version is `gpt-4o-2024-11-20`.
+- Use `think` for thinking mode.
+
+
+
+# GEdit-Bench
+Please follow [GEdit-Bench](https://github.com/stepfun-ai/Step1X-Edit/blob/main/GEdit-Bench/EVAL.md) for evaluation.
+
+
+# IntelligentBench
+TBD
\ No newline at end of file
diff --git a/Bagel/LICENSE b/Bagel/LICENSE
new file mode 100644
index 0000000..f49a4e1
--- /dev/null
+++ b/Bagel/LICENSE
@@ -0,0 +1,201 @@
+ Apache License
+ Version 2.0, January 2004
+ http://www.apache.org/licenses/
+
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+ 1. Definitions.
+
+ "License" shall mean the terms and conditions for use, reproduction,
+ and distribution as defined by Sections 1 through 9 of this document.
+
+ "Licensor" shall mean the copyright owner or entity authorized by
+ the copyright owner that is granting the License.
+
+ "Legal Entity" shall mean the union of the acting entity and all
+ other entities that control, are controlled by, or are under common
+ control with that entity. For the purposes of this definition,
+ "control" means (i) the power, direct or indirect, to cause the
+ direction or management of such entity, whether by contract or
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
+ outstanding shares, or (iii) beneficial ownership of such entity.
+
+ "You" (or "Your") shall mean an individual or Legal Entity
+ exercising permissions granted by this License.
+
+ "Source" form shall mean the preferred form for making modifications,
+ including but not limited to software source code, documentation
+ source, and configuration files.
+
+ "Object" form shall mean any form resulting from mechanical
+ transformation or translation of a Source form, including but
+ not limited to compiled object code, generated documentation,
+ and conversions to other media types.
+
+ "Work" shall mean the work of authorship, whether in Source or
+ Object form, made available under the License, as indicated by a
+ copyright notice that is included in or attached to the work
+ (an example is provided in the Appendix below).
+
+ "Derivative Works" shall mean any work, whether in Source or Object
+ form, that is based on (or derived from) the Work and for which the
+ editorial revisions, annotations, elaborations, or other modifications
+ represent, as a whole, an original work of authorship. For the purposes
+ of this License, Derivative Works shall not include works that remain
+ separable from, or merely link (or bind by name) to the interfaces of,
+ the Work and Derivative Works thereof.
+
+ "Contribution" shall mean any work of authorship, including
+ the original version of the Work and any modifications or additions
+ to that Work or Derivative Works thereof, that is intentionally
+ submitted to Licensor for inclusion in the Work by the copyright owner
+ or by an individual or Legal Entity authorized to submit on behalf of
+ the copyright owner. For the purposes of this definition, "submitted"
+ means any form of electronic, verbal, or written communication sent
+ to the Licensor or its representatives, including but not limited to
+ communication on electronic mailing lists, source code control systems,
+ and issue tracking systems that are managed by, or on behalf of, the
+ Licensor for the purpose of discussing and improving the Work, but
+ excluding communication that is conspicuously marked or otherwise
+ designated in writing by the copyright owner as "Not a Contribution."
+
+ "Contributor" shall mean Licensor and any individual or Legal Entity
+ on behalf of whom a Contribution has been received by Licensor and
+ subsequently incorporated within the Work.
+
+ 2. Grant of Copyright License. Subject to the terms and conditions of
+ this License, each Contributor hereby grants to You a perpetual,
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+ copyright license to reproduce, prepare Derivative Works of,
+ publicly display, publicly perform, sublicense, and distribute the
+ Work and such Derivative Works in Source or Object form.
+
+ 3. Grant of Patent License. Subject to the terms and conditions of
+ this License, each Contributor hereby grants to You a perpetual,
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+ (except as stated in this section) patent license to make, have made,
+ use, offer to sell, sell, import, and otherwise transfer the Work,
+ where such license applies only to those patent claims licensable
+ by such Contributor that are necessarily infringed by their
+ Contribution(s) alone or by combination of their Contribution(s)
+ with the Work to which such Contribution(s) was submitted. If You
+ institute patent litigation against any entity (including a
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
+ or a Contribution incorporated within the Work constitutes direct
+ or contributory patent infringement, then any patent licenses
+ granted to You under this License for that Work shall terminate
+ as of the date such litigation is filed.
+
+ 4. Redistribution. You may reproduce and distribute copies of the
+ Work or Derivative Works thereof in any medium, with or without
+ modifications, and in Source or Object form, provided that You
+ meet the following conditions:
+
+ (a) You must give any other recipients of the Work or
+ Derivative Works a copy of this License; and
+
+ (b) You must cause any modified files to carry prominent notices
+ stating that You changed the files; and
+
+ (c) You must retain, in the Source form of any Derivative Works
+ that You distribute, all copyright, patent, trademark, and
+ attribution notices from the Source form of the Work,
+ excluding those notices that do not pertain to any part of
+ the Derivative Works; and
+
+ (d) If the Work includes a "NOTICE" text file as part of its
+ distribution, then any Derivative Works that You distribute must
+ include a readable copy of the attribution notices contained
+ within such NOTICE file, excluding those notices that do not
+ pertain to any part of the Derivative Works, in at least one
+ of the following places: within a NOTICE text file distributed
+ as part of the Derivative Works; within the Source form or
+ documentation, if provided along with the Derivative Works; or,
+ within a display generated by the Derivative Works, if and
+ wherever such third-party notices normally appear. The contents
+ of the NOTICE file are for informational purposes only and
+ do not modify the License. You may add Your own attribution
+ notices within Derivative Works that You distribute, alongside
+ or as an addendum to the NOTICE text from the Work, provided
+ that such additional attribution notices cannot be construed
+ as modifying the License.
+
+ You may add Your own copyright statement to Your modifications and
+ may provide additional or different license terms and conditions
+ for use, reproduction, or distribution of Your modifications, or
+ for any such Derivative Works as a whole, provided Your use,
+ reproduction, and distribution of the Work otherwise complies with
+ the conditions stated in this License.
+
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
+ any Contribution intentionally submitted for inclusion in the Work
+ by You to the Licensor shall be under the terms and conditions of
+ this License, without any additional terms or conditions.
+ Notwithstanding the above, nothing herein shall supersede or modify
+ the terms of any separate license agreement you may have executed
+ with Licensor regarding such Contributions.
+
+ 6. Trademarks. This License does not grant permission to use the trade
+ names, trademarks, service marks, or product names of the Licensor,
+ except as required for reasonable and customary use in describing the
+ origin of the Work and reproducing the content of the NOTICE file.
+
+ 7. Disclaimer of Warranty. Unless required by applicable law or
+ agreed to in writing, Licensor provides the Work (and each
+ Contributor provides its Contributions) on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+ implied, including, without limitation, any warranties or conditions
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+ PARTICULAR PURPOSE. You are solely responsible for determining the
+ appropriateness of using or redistributing the Work and assume any
+ risks associated with Your exercise of permissions under this License.
+
+ 8. Limitation of Liability. In no event and under no legal theory,
+ whether in tort (including negligence), contract, or otherwise,
+ unless required by applicable law (such as deliberate and grossly
+ negligent acts) or agreed to in writing, shall any Contributor be
+ liable to You for damages, including any direct, indirect, special,
+ incidental, or consequential damages of any character arising as a
+ result of this License or out of the use or inability to use the
+ Work (including but not limited to damages for loss of goodwill,
+ work stoppage, computer failure or malfunction, or any and all
+ other commercial damages or losses), even if such Contributor
+ has been advised of the possibility of such damages.
+
+ 9. Accepting Warranty or Additional Liability. While redistributing
+ the Work or Derivative Works thereof, You may choose to offer,
+ and charge a fee for, acceptance of support, warranty, indemnity,
+ or other liability obligations and/or rights consistent with this
+ License. However, in accepting such obligations, You may act only
+ on Your own behalf and on Your sole responsibility, not on behalf
+ of any other Contributor, and only if You agree to indemnify,
+ defend, and hold each Contributor harmless for any liability
+ incurred by, or claims asserted against, such Contributor by reason
+ of your accepting any such warranty or additional liability.
+
+ END OF TERMS AND CONDITIONS
+
+ APPENDIX: How to apply the Apache License to your work.
+
+ To apply the Apache License to your work, attach the following
+ boilerplate notice, with the fields enclosed by brackets "[]"
+ replaced with your own identifying information. (Don't include
+ the brackets!) The text should be enclosed in the appropriate
+ comment syntax for the file format. We also recommend that a
+ file or class name and description of purpose be included on the
+ same "printed page" as the copyright notice for easier
+ identification within third-party archives.
+
+ Copyright [yyyy] [name of copyright owner]
+
+ Licensed under the Apache License, Version 2.0 (the "License");
+ you may not use this file except in compliance with the License.
+ You may obtain a copy of the License at
+
+ http://www.apache.org/licenses/LICENSE-2.0
+
+ Unless required by applicable law or agreed to in writing, software
+ distributed under the License is distributed on an "AS IS" BASIS,
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ See the License for the specific language governing permissions and
+ limitations under the License.
\ No newline at end of file
diff --git a/Bagel/README.md b/Bagel/README.md
new file mode 100644
index 0000000..f6e79a4
--- /dev/null
+++ b/Bagel/README.md
@@ -0,0 +1,199 @@
+
+ 
+
+
+
+
+ 
+
+
+ 
+
+
+ 
+
+
+ 
+
+
+ 
+
+
+ 
+
+
+ 
+
+
+
+# Unified Model for Multimodal Understanding and Generation
+> [Chaorui Deng*](https://scholar.google.com/citations?hl=en&user=k0TWfBoAAAAJ), [Deyao Zhu*](https://tsutikgiau.github.io/), [Kunchang Li*](https://andy1621.github.io/), [Chenhui Gou*](https://www.linkedin.com/in/chenhui-gou-9201081a1/?originalSubdomain=au), [Feng Li*](https://fengli-ust.github.io/), [Zeyu Wang](https://zw615.github.io/), Shu Zhong, [Weihao Yu](https://whyu.me/), [Xiaonan Nie](https://codecaution.github.io/), [Ziang Song](https://www.linkedin.com/in/ziang-song-43b0ab8a/), Guang Shi :email: , [Haoqi Fan* :tophat: ](https://haoqifan.github.io/)
+>
+> contact: shiguang.sg@bytedance.com
+>
+> We present **BAGEL**, an open‑source multimodal foundation model with 7B active parameters (14B total) trained on large‑scale interleaved multimodal data. BAGEL outperforms the current top‑tier open‑source VLMs like Qwen2.5-VL and InternVL-2.5 on standard multimodal understanding leaderboards, and delivers text‑to‑image quality that is competitive with strong specialist generators such as SD3.
+Moreover, BAGEL demonstrates superior qualitative results in classical image‑editing scenarios than the leading open-source models. More importantly, it extends to free-form visual manipulation, multiview synthesis, and world navigation, capabilities that constitute "world-modeling" tasks beyond the scope of previous image-editing models.
+The figure below showcases BAGEL's qualitative performance.
+
+
+
+
+
+
+## 📢 News
+
+We sincerely thank all contributors from the open community for their valuable support.
+
+- **May 26, 2025:** Thanks to [@neverbiasu](https://github.com/neverbiasu) for contributing [ComfyUI](https://github.com/neverbiasu/ComfyUI-BAGEL).
+- **May 25, 2025:** Special thanks to [@LeanModels](https://github.com/LeanModels) for providing the [DF11-compressed version](https://huggingface.co/DFloat11/BAGEL-7B-MoT-DF11), and to [@Gapeleon](https://huggingface.co/Gapeleon) for the [INT8-compressed version](https://huggingface.co/Gapeleon/bytedance_BAGEL-7B-MoT-INT8). We also appreciate [@gluttony-10](https://github.com/gluttony-10) for contributions to the [Windows package](https://github.com/ByteDance-Seed/Bagel/issues/51).
+- **May 24, 2025:** Together with [@wangwei1237](https://github.com/wangwei1237), [@gluttony-10](https://github.com/gluttony-10), and [@KingNish24](https://github.com/KingNish24), we built a Gradio [app](app.py) and launched a [Hugging Face Space](https://huggingface.co/spaces/ByteDance-Seed/BAGEL).
+- **May 23, 2025:** We have provided a training guideline in [TRAIN](./TRAIN.md).
+- **May 20, 2025:** We released the official [website](https://bagel-ai.org/), [demo](https://demo.bagel-ai.org/), [model](https://huggingface.co/ByteDance-Seed/BAGEL-7B-MoT), and [report](https://arxiv.org/abs/2505.14683) for BAGEL.
+
+
+## 📮 Notice
+**Call for Bad Cases:** If you have encountered any cases where the model performs poorly, we would greatly appreciate it if you could share them in the [issue#11](https://github.com/ByteDance-Seed/Bagel/issues/11) or [Discord](https://discord.gg/Z836xxzy).
+
+**About Inference Hyperparameters:**
+- **`cfg_text_scale`:** Controls how strongly the model follows the text prompt. `1.0` disables text guidance. Typical range: `4.0–8.0`.
+- **`cfg_image_scale`:** Controls how much the model preserves input image details. `1.0` disables image guidance. Typical range: `1.0–2.0`.
+- **`cfg_interval`:** Fraction of denoising steps where CFG is applied. Later steps can skip CFG to reduce computation. Typical: `[0.4, 1.0]`.
+- **`timestep_shift`:** Shifts the distribution of denoising steps. Higher values allocate more steps at the start (affects layout); lower values allocate more at the end (improves details).
+- **`num_timesteps`:** Total denoising steps. Typical: `50`.
+- **`cfg_renorm_min`:** Minimum value for CFG-Renorm. `1.0` disables renorm. Typical: `0`.
+- **`cfg_renorm_type`:** CFG-Renorm method:
+ - `global`: Normalize over all tokens and channels (default for T2I).
+ - `channel`: Normalize across channels for each token.
+ - `text_channel`: Like `channel`, but only applies to text condition (good for editing, may cause blur).
+- **If edited images appear blurry, try `global` CFG-Renorm, decrease `cfg_renorm_min` or decrease `cfg_scale`.**
+
+
+## 🔥 Quick Start
+
+1️⃣ Set up environment
+```bash
+git clone https://github.com/bytedance-seed/BAGEL.git
+cd BAGEL
+conda create -n bagel python=3.10 -y
+conda activate bagel
+pip install -r requirements.txt
+```
+
+2️⃣ Download pretrained checkpoint
+```python
+from huggingface_hub import snapshot_download
+
+save_dir = "/path/to/save/BAGEL-7B-MoT"
+repo_id = "ByteDance-Seed/BAGEL-7B-MoT"
+cache_dir = save_dir + "/cache"
+
+snapshot_download(cache_dir=cache_dir,
+ local_dir=save_dir,
+ repo_id=repo_id,
+ local_dir_use_symlinks=False,
+ resume_download=True,
+ allow_patterns=["*.json", "*.safetensors", "*.bin", "*.py", "*.md", "*.txt"],
+)
+
+```
+
+3️⃣ Go to [`inference.ipynb`](inference.ipynb) to start playing with BAGEL!
+
+4️⃣ Use Gradio WebUI to start playing with BAGEL!
+```bash
+pip install gradio
+python app.py
+```
+
+## 🔥 Train & Eval
+
+### Train
+
+```bash
+bash scripts/train.sh
+```
+
+You can replace the variables in the script with your own before running.
+See [TRAIN](TRAIN.md) for more details.
+
+### Eval
+We provide the scripts for evaluating VLM, T2I and Editing benchmarks.
+Please See [EVAL](EVAL.md) for more details.
+
+
+## 📊 Benchmarks
+
+### 1. Visual Understanding
+
+| Model | MME ↑ | MMBench ↑ | MMMU ↑ | MM-Vet ↑ | MathVista ↑ |
+| ------------------- | ----------: | ----------: | -------: | -------: | ----------: |
+| Janus-Pro-7B | - | 79.2 | 41.0 | 50.0 | – |
+| Qwen2.5-VL-7B | 2347 | 83.5 | **58.6** | 67.1 | 68.2 |
+| **BAGEL** | **2388** | **85.0** | 55.3 | **67.2** | **73.1** |
+
+### 2. Text-to-Image Generation
+
+| Model | GenEval ↑ | WISE ↑|
+| ------------ | --------- | --------- |
+| Janus-Pro-7B | 0.80 | 0.35 |
+| SD3-Medium | 0.74 | - |
+| FLUX-1-dev | 0.82 | 0.50 |
+| **BAGEL** | - | **0.52** |
+| **BAGEL + CoT** | **0.88** | **0.70** |
+
+### 3. Image Editing
+
+| Model | GEdit-Bench-EN (SC) ↑ | GEdit-Bench-EN (PQ) ↑ | GEdit-Bench-EN (O) ↑ | IntelligentBench ↑ |
+| ------------- | --------------------- | --------------------- | ------------------- | ------------------ |
+| Step1X-Edit | 7.09 | 6.76 | **6.70** | 14.9 |
+| Gemini-2-exp. | 6.73 | 6.61 | 6.32 | **57.6** |
+| **BAGEL** | **7.36** | **6.83** | 6.52 | 44.0 |
+| **BAGEL+CoT** | – | – | – | 55.3 |
+
+
+## ✍️ Citation
+
+```bibtex
+@article{deng2025bagel,
+ title = {Emerging Properties in Unified Multimodal Pretraining},
+ author = {Deng, Chaorui and Zhu, Deyao and Li, Kunchang and Gou, Chenhui and Li, Feng and Wang, Zeyu and Zhong, Shu and Yu, Weihao and Nie, Xiaonan and Song, Ziang and Shi, Guang and Fan, Haoqi},
+ journal = {arXiv preprint arXiv:2505.14683},
+ year = {2025}
+}
+```
+
+
+## 📜 License
+BAGEL is licensed under the Apache 2.0.
diff --git a/Bagel/TRAIN.md b/Bagel/TRAIN.md
new file mode 100644
index 0000000..eb5e94e
--- /dev/null
+++ b/Bagel/TRAIN.md
@@ -0,0 +1,168 @@
+# Data prepration
+
+We provide data examples for **T2I**, **Editing**, and **VLM** tasks. The T2I dataset is generated using [FLUX.1‑dev](https://huggingface.co/black-forest-labs/FLUX.1-dev); the editing examples are randomly sampled from [SEED‑Data‑Edit‑Part3](https://huggingface.co/datasets/AILab-CVC/SEED-Data-Edit-Part2-3); and the VLM set is sourced from [LLaVA‑OneVision‑Data](https://huggingface.co/datasets/lmms-lab/LLaVA-OneVision-Data).
+
+We offer examples in both raw-image folder and parquet shard formats. For other data formats, you can use our dataset code as a template and extend it as needed.
+
+
+1. **Download the sample dataset**
+
+ ```bash
+ wget -O bagel_example.zip \
+ https://lf3-static.bytednsdoc.com/obj/eden-cn/nuhojubrps/bagel_example.zip
+ unzip bagel_example.zip -d /data
+ ```
+2. **Expected hierarchy**
+
+ ```text
+ bagel_example
+ ├── t2i/ # text-to-image (parquet)
+ ├── editing/ # image editing (parquet)
+ │ ├── seedxedit_multi/
+ │ └── parquet_info/
+ └── vlm/
+ ├── images/ # JPEG / PNG frames
+ └── llava_ov_si.jsonl # vision‑language SFT conversations
+ ```
+3. Edit every `your_data_path` placeholder in **`data/dataset_info.py`**.
+4. *(Optional)* Extend `DATASET_INFO` with your own parquet shards or JSONL files to mix extra data.
+
+---
+
+# Training
+
+The baseline training recipe looks like this (replace environment variables with real paths or values):
+
+```shell
+# Pre-training
+torchrun \
+ --nnodes=$num_nodes \
+ --node_rank=$node_rank \
+ --nproc_per_node=8 \
+ --master_addr=$master_addr \
+ --master_port=$master_port \
+ train/pretrain_unified_navit.py \
+ --dataset_config_file ./data/configs/example.yaml \
+ --llm_path $llm_path \
+ --vae_path $vae_path \
+ --vit_path $vit_path \
+ --layer_module Qwen2MoTDecoderLayer \
+ --use_flex True \
+ --resume_from $resume_from \
+ --results_dir $output_path \
+ --checkpoint_dir $ckpt_path \
+ --max_latent_size 64 # 32 for low-resolution pre-training
+
+# Fine-tuning
+torchrun \
+ --nnodes=$num_nodes \
+ --node_rank=$node_rank \
+ --nproc_per_node=8 \
+ --master_addr=$master_addr \
+ --master_port=$master_port \
+ train/pretrain_unified_navit.py \
+ --dataset_config_file ./data/configs/example.yaml \
+ --model_path $model_path \
+ --layer_module Qwen2MoTDecoderLayer \
+ --max_latent_size 64 \
+ --resume-from $model_path \
+ --finetune_from_hf True \
+ --auto_resume True \
+ --resume-model-only True \
+ --finetune-from-ema True \
+ --log_every 1 \
+ --lr 2e-5 \
+ --num_worker 1 \
+ --expected_num_tokens 10240 \
+ --max_num_tokens 11520 \
+ --max_num_tokens_per_sample 10240
+```
+
+- **When fine-tuning BAGEL, set `max_latent_size=64` to ensure the correct pretrained weights are loaded.** If this is not set, an out-of-bounds error may occur.
+- The total value of `num_used_data` should be greater than `NUM_GPUS × NUM_WORKERS`. (For toy data, use `num_worker=1`.)
+- For T2I-only fine-tuning, set `visual_und=False`. For VLM-only fine-tuning, set `visual_gen=False`.
+- For debugging purposes, use smaller values for `expected_num_tokens`, `max_num_tokens`, and `max_num_tokens_per_sample`.
+- When fine-tuning on toy data, the loss behaves as follows:
+ ```shell
+ [2025-05-25 17:01:37] (step=0000000) Train Loss mse: 0.4063, Train Loss ce: 0.5504, Train Steps/Sec: 0.01,
+ [2025-05-25 17:01:40] (step=0000001) Train Loss mse: 0.4121, Train Loss ce: 0.8152, Train Steps/Sec: 0.44,
+ [2025-05-25 17:01:42] (step=0000002) Train Loss mse: 0.3876, Train Loss ce: 1.3411, Train Steps/Sec: 0.40,
+ [2025-05-25 17:01:45] (step=0000003) Train Loss mse: 0.3825, Train Loss ce: 0.7360, Train Steps/Sec: 0.44,
+ ```
+
+
+You are encouraged to adjust any of these hyperparameters to fit your GPU budget and the scale of your dataset. If you encounter any issues, please open an issue for assistance. 🎉
+
+
+## Model config
+
+
+| Argument | Default | Description |
+| ---------------------------- | ------------------------------------------- | --------------------------------------------------------------- |
+| `llm_path` | `hf/Qwen2.5-0.5B-Instruct` | Language‑model backbone (HuggingFace repo or local folder). |
+| `vae_path` | `flux/vae/ae.safetensors` | Pre‑trained VAE checkpoint for latent diffusion. |
+| `vit_path` | `hf/siglip-so400m-14-980-flash-attn2-navit` | SigLIP ViT used for image understanding. |
+| `max_latent_size` | `32` | Maximum latent grid side; defines highest generable resolution. |
+| `latent_patch_size` | `2` | VAE pixels represented by one latent patch. |
+| `vit_max_num_patch_per_side` | `70` | Max ViT patches per image side after resizing. |
+| `text_cond_dropout_prob` | `0.1` | Probability to drop text conditioning while training. |
+| `vae_cond_dropout_prob` | `0.3` | Dropout on VAE latent inputs. |
+| `vit_cond_dropout_prob` | `0.3` | Dropout on visual features. |
+
+*(See `ModelArguments` for many more options.)*
+
+
+## Data config
+
+
+| Argument | Default | Description |
+| --------------------------- | --------------------------- | --------------------------------------------------------- |
+| `dataset_config_file` | `data/configs/example.yaml` | YAML that groups datasets and assigns sampling weights. |
+| `num_workers` | `4` | Background workers per rank for the PyTorch `DataLoader`. |
+| `prefetch_factor` | `2` | Batches pre‑fetched by each worker. |
+| `max_num_tokens_per_sample` | `16384` | Skip raw samples longer than this. |
+| `max_num_tokens` | `36864` | Hard cap for a packed batch (prevents OOM). |
+| `max_buffer_size` | `50` | Overflow buffer length for oversized samples. |
+| `data_seed` | `42` | Seed for reproducible shuffling and sampling. |
+
+
+## Training config
+
+| Argument | Default | Description |
+| -------------------------------------- | ---------------------- | ------------------------------------------------------ |
+| `total_steps` | `500_000` | Optimiser steps to run. |
+| `lr` | `1e-4` | Peak learning rate after warm‑up. |
+| `lr_scheduler` | `constant` | Learning‑rate schedule (`constant` or `cosine`). |
+| `warmup_steps` | `2000` | Linear warm‑up duration. |
+| `ema` | `0.9999` | Exponential moving‑average decay for model weights. |
+| `max_grad_norm` | `1.0` | Gradient‑clipping threshold. |
+| `save_every` | `2000` | Checkpoint frequency (steps). |
+| `visual_gen / visual_und` | `True` | Enable image generation / understanding branches. |
+| `freeze_llm / freeze_vit / freeze_vae` | `False / False / True` | Freeze selected modules to save VRAM or for ablations. |
+| `use_flex` | `True` (in example) | Enable FLEX packing for higher GPU utilisation. |
+| `sharding_strategy` | `HYBRID_SHARD` | FSDP sharding mode. |
+| `num_shard` | `8` | Parameter shards per rank in HYBRID mode. |
+
+**Distributed‑launch environment variables**
+
+| Var | Meaning |
+| ----------------------------- | --------------------------------- |
+| `num_nodes` / `node_rank` | Multi‑node orchestration indices. |
+| `nproc_per_node` | Number of GPUs per node. |
+| `master_addr` / `master_port` | NCCL rendezvous endpoint. |
+
+
+## Logging config
+
+
+| Argument | Default | Description |
+| ---------------- | --------------------- | ---------------------------------------------------- |
+| `results_dir` | `results` | Root directory for logs and metrics. |
+| `checkpoint_dir` | `results/checkpoints` | Checkpoints are saved here. |
+| `log_every` | `10` | Steps between console / W\&B logs. |
+| `wandb_project` | `bagel` | Weights & Biases project name. |
+| `wandb_name` | `run` | Run name inside the project. |
+| `wandb_offline` | `False` | Switch to offline mode (logs locally, sync later). |
+| `wandb_resume` | `allow` | Resumption policy if an existing run ID is detected. |
+
+> **Tip** Export `WANDB_API_KEY` before launching if you want online dashboards.
\ No newline at end of file
diff --git a/Bagel/app.py b/Bagel/app.py
new file mode 100644
index 0000000..f9c97e8
--- /dev/null
+++ b/Bagel/app.py
@@ -0,0 +1,256 @@
+
+import numpy as np
+import os
+import torch
+import random
+
+from accelerate import infer_auto_device_map, load_checkpoint_and_dispatch, init_empty_weights
+from accelerate.utils import BnbQuantizationConfig, load_and_quantize_model
+from PIL import Image
+
+from .data.data_utils import add_special_tokens, pil_img2rgb
+from .data.transforms import ImageTransform
+from .inferencer import InterleaveInferencer
+from .modeling.autoencoder import load_ae
+from .modeling.bagel.qwen2_navit import NaiveCache
+from .modeling.bagel import (
+ BagelConfig, Bagel, Qwen2Config, Qwen2ForCausalLM,
+ SiglipVisionConfig, SiglipVisionModel
+)
+from .modeling.qwen2 import Qwen2Tokenizer
+
+
+
+def load_bagel_model(model_path,quantize_mode,max_mem_per_gpu = "24GiB"):
+ print ("Loading model .... ")
+ llm_config = Qwen2Config.from_json_file(os.path.join(model_path, "llm_config.json"))
+ llm_config.qk_norm = True
+ llm_config.tie_word_embeddings = False
+ llm_config.layer_module = "Qwen2MoTDecoderLayer"
+ vit_config = SiglipVisionConfig.from_json_file(os.path.join(model_path, "vit_config.json"))
+ vit_config.rope = False
+ vit_config.num_hidden_layers -= 1
+ vae_model, vae_config = load_ae(local_path=os.path.join(model_path, "ae.safetensors"))
+ config = BagelConfig(
+ visual_gen=True,
+ visual_und=True,
+ llm_config=llm_config,
+ vit_config=vit_config,
+ vae_config=vae_config,
+ vit_max_num_patch_per_side=70,
+ connector_act='gelu_pytorch_tanh',
+ latent_patch_size=2,
+ max_latent_size=64,
+ )
+ with init_empty_weights():
+ language_model = Qwen2ForCausalLM(llm_config)
+ vit_model = SiglipVisionModel(vit_config)
+ model = Bagel(language_model, vit_model, config)
+ model.vit_model.vision_model.embeddings.convert_conv2d_to_linear(vit_config, meta=True)
+
+ tokenizer = Qwen2Tokenizer.from_pretrained(model_path)
+ tokenizer, new_token_ids, _ = add_special_tokens(tokenizer)
+
+ vae_transform = ImageTransform(1024, 512, 16)
+ vit_transform = ImageTransform(980, 224, 14)
+ # Model Loading and Multi GPU Infernece Preparing
+
+
+ #max_mem_per_gpu = "24GiB" # Modify it according to your GPU setting. On an A100, 80 GiB is sufficient to load on a single GPU.
+
+ device_map = infer_auto_device_map(
+ model,
+ max_memory={0: max_mem_per_gpu},
+ no_split_module_classes=["Bagel", "Qwen2MoTDecoderLayer", "SiglipVisionModel"],
+ )
+ #print(f"device_map is {device_map}")
+
+ same_device_modules = [
+ 'language_model.model.embed_tokens',
+ 'time_embedder',
+ 'latent_pos_embed',
+ 'vae2llm',
+ 'llm2vae',
+ 'connector',
+ 'vit_pos_embed'
+ ]
+
+ if torch.cuda.device_count() == 1:
+ first_device = device_map.get(same_device_modules[0], "cuda:0")
+ for k in same_device_modules:
+ if k in device_map:
+ device_map[k] = first_device
+ else:
+ device_map[k] = "cuda:0"
+ else:
+ first_device = device_map.get(same_device_modules[0])
+ for k in same_device_modules:
+ if k in device_map:
+ device_map[k] = first_device
+
+ if quantize_mode == "fp16":
+ model = load_checkpoint_and_dispatch(
+ model,
+ checkpoint=os.path.join(model_path, "ema.safetensors"),
+ device_map=device_map,
+ offload_buffers=True,
+ offload_folder="offload",
+ dtype=torch.bfloat16,
+ force_hooks=True,
+ ).eval()
+ elif quantize_mode == "nf4":
+ print("Quantizing to NF4")
+ bnb_quantization_config = BnbQuantizationConfig(load_in_4bit=True, bnb_4bit_compute_dtype=torch.bfloat16, bnb_4bit_use_double_quant=False, bnb_4bit_quant_type="nf4")
+ model = load_and_quantize_model(
+ model,
+ weights_location=os.path.join(model_path, "ema.safetensors"),
+ bnb_quantization_config=bnb_quantization_config,
+ device_map=device_map,
+ offload_folder="offload",
+ ).eval()
+ else:#args.mode == 3: # INT8
+ print("Quantizing to INT8/fp8")
+ bnb_quantization_config = BnbQuantizationConfig(load_in_8bit=True, torch_dtype=torch.float32)
+ model = load_and_quantize_model(
+ model,
+ weights_location=os.path.join(model_path, "ema.safetensors"),
+ bnb_quantization_config=bnb_quantization_config,
+ device_map=device_map,
+ offload_folder="offload",
+ ).eval()
+
+ model = model.eval()
+ print('Model loaded is done!')
+ inferencer = InterleaveInferencer(
+ model=model,
+ vae_model=vae_model,
+ tokenizer=tokenizer,
+ vae_transform=vae_transform,
+ vit_transform=vit_transform,
+ new_token_ids=new_token_ids
+ )
+ return inferencer
+
+
+
+def set_seed(seed):
+ """Set random seeds for reproducibility"""
+ if seed > 0:
+ random.seed(seed)
+ np.random.seed(seed)
+ torch.manual_seed(seed)
+ if torch.cuda.is_available():
+ torch.cuda.manual_seed(seed)
+ torch.cuda.manual_seed_all(seed)
+ torch.backends.cudnn.deterministic = True
+ torch.backends.cudnn.benchmark = False
+ return seed
+
+# Text to Image function with thinking option and hyperparameters
+def text_to_image(inferencer,prompt, show_thinking=False, cfg_text_scale=4.0, cfg_interval=0.4,
+ timestep_shift=3.0, num_timesteps=50,
+ cfg_renorm_min=0.0, cfg_renorm_type="global",
+ max_think_token_n=1024, do_sample=False, text_temperature=0.3,
+ seed=0, image_ratio="1:1"):
+ # Set seed for reproducibility
+ set_seed(seed)
+
+ if image_ratio == "1:1":
+ image_shapes = (1024, 1024)
+ elif image_ratio == "4:3":
+ image_shapes = (768, 1024)
+ elif image_ratio == "3:4":
+ image_shapes = (1024, 768)
+ elif image_ratio == "16:9":
+ image_shapes = (576, 1024)
+ elif image_ratio == "9:16":
+ image_shapes = (1024, 576)
+
+ # Set hyperparameters
+ inference_hyper = dict(
+ max_think_token_n=max_think_token_n if show_thinking else 1024,
+ do_sample=do_sample if show_thinking else False,
+ text_temperature=text_temperature if show_thinking else 0.3,
+ cfg_text_scale=cfg_text_scale,
+ cfg_interval=[cfg_interval, 1.0], # End fixed at 1.0
+ timestep_shift=timestep_shift,
+ num_timesteps=num_timesteps,
+ cfg_renorm_min=cfg_renorm_min,
+ cfg_renorm_type=cfg_renorm_type,
+ image_shapes=image_shapes,
+ )
+
+ # Call inferencer with or without think parameter based on user choice
+ result = inferencer(text=prompt, think=show_thinking, **inference_hyper)
+ return result["image"], result.get("text", "")
+
+
+# Image Understanding function with thinking option and hyperparameters
+def image_understanding(inferencer,image: Image.Image, prompt: str, show_thinking=False,
+ do_sample=False, text_temperature=0.3, max_new_tokens=512):
+ if image is None:
+ return "Please upload an image."
+
+ if isinstance(image, np.ndarray):
+ image = Image.fromarray(image)
+
+ image = pil_img2rgb(image)
+
+ # Set hyperparameters
+ inference_hyper = dict(
+ do_sample=do_sample,
+ text_temperature=text_temperature,
+ max_think_token_n=max_new_tokens, # Set max_length
+ )
+
+ # Use show_thinking parameter to control thinking process
+ result = inferencer(image=image, text=prompt, think=show_thinking,
+ understanding_output=True, **inference_hyper)
+ return result["text"]
+
+
+# Image Editing function with thinking option and hyperparameters
+def edit_image(inferencer,image: Image.Image, prompt: str, show_thinking=False, cfg_text_scale=4.0,
+ cfg_img_scale=2.0, cfg_interval=0.0,
+ timestep_shift=3.0, num_timesteps=50, cfg_renorm_min=0.0,
+ cfg_renorm_type="text_channel", max_think_token_n=1024,
+ do_sample=False, text_temperature=0.3, seed=0):
+ # Set seed for reproducibility
+ set_seed(seed)
+
+ if image is None:
+ return "Please upload an image.", ""
+
+ if isinstance(image, np.ndarray):
+ image = Image.fromarray(image)
+
+ image = pil_img2rgb(image)
+
+ # Set hyperparameters
+ inference_hyper = dict(
+ max_think_token_n=max_think_token_n if show_thinking else 1024,
+ do_sample=do_sample if show_thinking else False,
+ text_temperature=text_temperature if show_thinking else 0.3,
+ cfg_text_scale=cfg_text_scale,
+ cfg_img_scale=cfg_img_scale,
+ cfg_interval=[cfg_interval, 1.0], # End fixed at 1.0
+ timestep_shift=timestep_shift,
+ num_timesteps=num_timesteps,
+ cfg_renorm_min=cfg_renorm_min,
+ cfg_renorm_type=cfg_renorm_type,
+ )
+
+ # Include thinking parameter based on user choice
+ result = inferencer(image=image, text=prompt, think=show_thinking, **inference_hyper)
+ return result["image"], result.get("text", "")
+
+
+# Helper function to load example images
+def load_example_image(image_path):
+ try:
+ return Image.open(image_path)
+ except Exception as e:
+ print(f"Error loading example image: {e}")
+ return None
+
+
diff --git a/Bagel/assets/arch.png b/Bagel/assets/arch.png
new file mode 100644
index 0000000000000000000000000000000000000000..194e3a2cdde3101a4655ddd452a9343a6236b0aa
GIT binary patch
literal 168377
zcmeEtLCO
zzl=cdynG}uewAJp^=noH!5_P#kyBE7pCMsM3#DcB4olsVmzh)LvXCJa&@>!!j*_UH
zGA7wF040-Xm0S*R;+0*_c+I(H#Xk{(S;~FSsBG?9j$mtU@iw
zH0`yXot!)_`*IvxNsX3%MC0`3a8Oj@T~q4tOsa!tS(5bxH#;hAU>sLs7Cb004$iKZ
zwz*98q2$M`%rnBh=moU3;UT;qb#y-O_uihj%s$O#GT}vLTdKr)^}Kak7Jom}k%s-%
zS||)Q_+wz$Qg)o1b5Owbr|dG}jNuywr=7R;vTi>8FQ?W0DB>AoO}>+oh>ZF?q@vYl
zCKA5n-Q8kui0LDox${8|s%~iyoT<0oe%NmJ|LN}JsIU3u0t#(GB<&cEWeVB?471~5
znX14!lhi}TamF#6#Jh9u$HJ@3cKLm<;f2)15pH5&ue;~xrMD=w>)2nXBk=Nq*`1~4
z@j_7`hTyYO5v=z5uv>;eJAICgeMiSQsCEzpi+o=_*
zV73z(Hayf^I;_5s{x2vqlB5K|mY*4Q(WBqLS;F%C!dgQHMY->$cSZ$yornwB_{+N!8YM}B&*wI_V**x-jzZB7d;aqzTnTlTZt#kmYeILl>kA
z)n00%%}XgyO-^C%y>)+i5OORfJ;ieI{0`?8*UO@A{Tf0CG@j2)RH$;@G3%P%Zw!&h
zx-&Kun(@Rjn!2qw%1@Bq2Alle!jQBeu#_RIl_ZZ8Ac}rX_a+iUFE)veM4o!)t2=|(
zkh09zb{TO~XL8o?xZ2;%X2@LAmi?HfxXTyc!TcTd}d^7DdeWR;fA+&%t
z+B0CPQ9D}Gzl7z0awT|$?1?1uH6XrsBgthzkl7H4gC-@;AQ%)1inZ_e*tm1Eowi<2N5{Thtq7-Y`aS=Q$LNsZVHv^ZaFe)Nbpee`0Bb71w<#
zlr1#b4k-&SD`r(~kZI8Jki4v?FdElkmj-0$x2;Hfd}~efZNB5a6gwcBGqIcEHRKbd
zV5E@aJ+|#hw@N2WXXCxLHJ+tfoLM+qB(gi25iY(|H_!uVh#0jQ-UF{n@U#YvBC0j2
z+&=6I8AVMLSkJKS@9vwornf$8HCxkPGhV|KGw-nJkm}Ixh`4(l5FQvE*b>&
zKaE?7N69M#%oQ?DH8wL=HeQv{yA!wLf4;U8a9(qsxr@0oKi&p--}qVoj~=f9jvlMQ
za|2TSMuW~e?iz)f6{~sk%khGNl10I7oK*JHcNOV@t8bupk
z+v3&EmFT_Ll^0bB9z7lwDp7Eb1mOuI}+9-D7m155*q))n^E>@Qu8ytzeTK#_9#~%~n8{s#~Weu9E
z2B=n9hwN$;z2V97`F=Yxx}x~UWA!kkDdw6L#zEUR2r2F>UeuWuSZNx$PRPkNDKn{u
zq;`eCvEWrm4w7oW8^3_wvXe&4_fcTGV;Pcoi$_wuc&SNpA-MEb_r>Z<oS^yDy
z+;Ln&%*SuG(YFML#0I2nQT5RVBwl>RV%fBigfFU*A8q8xP4?QH2YV7^BKkJ_x@8Aq
zesZ?9>jt>&Y-%X}`B{@frQE~R&qOVM*Mn!dSUZ@gH_SI|PUSSAGJ##i+>|LY;ac|{%P`o&U={4`46D8Y??V>cYrMa%KW`I6X(#}Fq&XLZ--Ba4DZKy>+CA(o
z>UKu@pYAM)t!I;I9Rf=pOkLeZcEId$23HRK01^l4)ty>EP@0t9`{k3X??-^`i?bhR
zULrrC&4#8XwO-7&#oomQ+|707Dcx0-ZJnBhTDY2iT1^W!3l0^xH5qMfLGioUKRfjO
z6t4rH65v}|`tX%T;r!OFx$B3%o6^$v9l5{_`}|*@>LkqIw@k%FMJVa$
zjl8|%n>*7XQRq%oXF_w+X4C4RvQ>Rwaw<~Vz$dqv$q>1(aqhI1Q~)6>?}{D4OYe`A
zb3Y;c3)u@C`;o9c5kl|0w%LaB8_u6yyxDVNdI1Xib3bPbW@U?+CtqZ~gb{gmh-UbD
zHv4_Nespu$b;CF+p&AcR`_}BMCWZl@x9pHB>1Am=alqtVLliQB_L!v#LJNx{Bv~!?BgkLethWX
zbIo*I9q=VU?M&^-wExw(oJUvM(*mnYQYOXO%hoHwv(rb@iIW(JTOgcU{UY?NV6${$
zw#d41QYG##7VO=5ayx{*5fhctAH?$T>3rnRe655Syq%v^fd5SY!uM>vsL9TW60+QN
zNP|xqbRmBD*cb65-j4P<$jtxXmg@d?zpsw-hBMhjxyv&U`|j#9x4oJ#*k8C*k^Eb&mI(MpeI}Oj
zM4vfWGtV#IdiiSS5Z#6#d2jPPQim`m2e0RF^9;fZkAtctDK4E3yO3`4p0mo(K6gQw
zuF;mdAFY&?kysI73?$TN6i8@@&@;p>{*3Z}!XKWoAffyt9vKNK!WIejzw)Rco_~G5
zBksS-{PT(OBOD1G@rwv?d;dcI@7(D6zfk@=jQk6chV))T`r}8$Q^VZV($dk*#>w6N
zX$Fjlz;u?^b3;NRW&G=Y_EG)y38MZ*TTNYeU1cR0FTa4i!IfN16zoI!A=>Cz#-Cm4AS6Pit+R4?Dj-P{zgNs2Ni;j*?
z)YZaDSY77Be-%gk5@WD&cXt-%YtVTcRezeZsxAG&hEBOj&y(3Yij1?;V#C&@K;0s^Z92#Exm33wIKKe5=s)uRpOpV?@n1^n{%^^5
zf`Wf5`Y%cUUR2A?(pB0CfN0ZQ{D0rqf3^FsiT_nll=H8l|I1YTbDsYZiNU#6e+Fi1$ckE2FB3cp_HWU!P}ch%c6ZJ`v$(37|+kLVF}6Nu-Z5?=`)jK|8FS
zX$Bjr+EU$+x-pPnr@9rThRA#h{!x+1$fl212ki(r&|EXQp6ig9$T4U;fPQuoVi9Js
zR?K6GG9z-E5DNMHT>k6J*B*-^Q~rCl7yg(iu9EHcDOUx@+1bT@;RDxK`sHAJJE>RC
zi0J;aOFlzV46D+%-t7HTAq~LQ@BX0Vf@mdXl{6fDx_B*bO
z?D|+3_Sy1*C4+=!qhJG^Zg|vI$LFixYc+HXsGz_6PuNQKY2re;j%qr;T>+f$r&TPa
zVtLVcf!v_VkJ)zF0YEMCzrV;i59Ej#)sZ|Q5!QDl$j%@!=>i{YEZ1Z;yInB;9cY#f
zF3vf9J7Huns!Bt%K^=5|%k)OXhr?wYGJLe8UbH$hDKTc(W_2sR+;L=++?qc?rqGn#
zqBXHN7u8bVD8FBeWr}@GE|wed;-Ul$b5bOOCT#{V6B*I)wwGT}&k4C0u4T2D*ydVE2MH-nNl?537Wn;fwb^W9?@oZufEtnF*>l}&0zDFZCh(gp1#(cZf`ceC6nPl0V
z-*!pOk%~*^Zr3BWLYZF%`L*QVu0@ex_QM#o-E}>{Eu^RaM(z6L(nrVk%C33`
zerCu6EX?&($}$J!VMC8EDkPjyq|0oD3JwS;%@lS4A8Lg>zbr7#e;#}rh5gCjkF(U~
zy0sHh#HXWr#cWIt%eotdmm2y96E{6re1CD}=tMf{3wH`Ke7!rV72TZA^kQVF_F)`^
z?t3xyMvvDkf=WntCr7~^yW3+E*!c!+(Y@0LUF!?{;jtn3zVoE5G}xTxyCnHDtV&Z0
z@24{2UsJyG3FqPO7zrAt4kWAs9jipSS~r74?{=FR|Bo7TMr1$OaTUu9^P0U&A@zS
z@Gy3`pe<~-x6U_ug_gd*TphCNKluB1ixNJ61;>wmH?QYqX)$kEfX}B48$#wsuSTN7
z{K`DKXaYorE<}2F(Zd$(HY~S!5o%l+g($|58X>;bc-COXCsaEHoBl-n2^fH_
zqcj+sEZG(rv;XgId6R&s=Fp9m4hDh1HQ#J5WvdN`Y7+#rV7u^Z_#H{?j&6^B*!-9u
zj2K`MUFLrxR-T8Tq-KS~29LMhgZE?)L&@Pjl&iXP1
z@@3HR%S;cBt-8eB?X1P0OvfQvRq2rPxO=uNLWkT2;cM(wd3;VBty749Do
z{+xQb9K7_9h78~=Rm-KO#pTi1eoUR+FwJt^)2|@b8fsx-lL3`u62XWNcLW%Qt}-G4
zF(57(o$7u@B`PoVMarH?u-)AQu>p3pw)-^Q5IJL3cD3Gm+E%n^^VG8UqyIU?wf_@A
z`!#6ja&PJ5F-^i6)~eF+WZl1Mh1Y5sh&n3A^U1(oBK{>=5$}?xggq5hUkbC79e?V3RmVA@4*L=M-KS9|AHy#;f6D2XQ>6{x38cS)l
z9(;GZlS7G;C9At3LQ766m~!0RBLuRXP2ZF}ZKI)JGCBh>O_&l6SYFm$jN>yO)9P3&
z=JLrqfd-_xVTYC7iN14;U9qeFyL8>xuhKUB&QA^4h*|>1>*B=)wDlWo?XndVZDhk{
zEf0WaTazW1^D(VHF$2M#-jc|se`(BwjoWgSFH(nEaubvTNs^jUY&7_0hZUFQD60Bh
z#$vbQm0r&$PQRea|Jy|;kC@i)gO{!slh@klQd;8G8}g?itxlz(3ynv-{wK3(?8f0k
zShaj(Rp_~V(B(W~)m%Q-AE&mvGQ)a4@bT>^yi3{{S%e&92zqwJOg>tKS&z*B(#T`v7TH?4l)U7E|vxITAWFESOG9
z`(9}04Mkhi6gHvgufNkHG}KFE^V9TKlDQ2#OLf~TvUCc$B&tsbt4`Y&Vdu{an|j7%m2SJ7Hs^ANH!7&ty}Joi78)swxA-J|M&vQG{01B(
z)p_$7@M~{AaF+$By$Qd<9@pt$PpxhZ82sr9AO5XYkMP6=%7R
z9??j*M2Ame`fDLQ?BBd?A8W%T$}
z1j?GfLyyJ&Qkn94W`J7Y?T;7xqO9Vd7$Q`X4*b-2NQMFP3c^jp2~{I>5&{C-HyKA}
z#}%L^IlMQs#MYDE=j3jUfEdy{abSWeSe%-F;XD((YouzgSq_
z6yMV2daWYldRCCx<~{94X25UHD#WfoQu9@6c(##;uYDARM>@KpK>~>(zU$mtvn>bs
zKF;4}K&m=`HtO#ddsgIvrkx~2F8KmQdF|HG%!^t$%
z!Ud0xGW|=i;x4p^5Mg}01Amr6Apuo`E51_xS{<*Jjp5wPdgv<&nZ^aw^5pXt$;m{%tJfD~%TXLupC9YV_655j~O{
zIxIwrjO$8M68KTX*H|L$gk3D~Lmagb`ygJ*$KhHY*Sd>}aPJPNPZ*!$jHIx;Bm3=|
zYrD9BUFHMK@4nMOfGSB$B;NjJz|Hk)OxCvIQe?X!!GAnJcj}b@;giek3JRu6y)5?j
zY}42HwQQenfwXUh=@CY+&!dFf-PbRGdzCu@+w3PJ<0P2A`?
ziVn45+J7keZw_)~Fe!Ojig75jjBnQ0K=C6@IZg_5vNmHOvdqMTOD>lvTORWFhX#%j
zKZCi7pM3oMppgr+h9R9T2aLqj3UOJa7#cN&=U$8hxPw_qev`C=D)6L~XVPs7kF@8^
zG)p?HS_CN{Jsgst*v{P4U
zbuo7a+4A`xa0_hfKg{S!FJB!j(ob8rNth1W_^pjRf-{3}LSgbT+RCq0L8CD6#Hieg
znzYc?OHaqjhuD&kPE~K~#AWSk*<>=W|)GR3nzqdc(Ica-0q+10Jx^%*RPrNyQ7jm!f
zyy!%7=eNlp(wWNMM1qMqO1{r5xwVnCUljy|UuW2HcsdWr$6tcwe
zI{L-(fsp%M(oUvgFP%FMMFQtmaX5T!Lmgw;%88DRrkJzYq0K46klMAIEu({Y*QFK7Mh~yl~FYnjfTY|p`T*YSRVc4wJI@3
z+@{PCg**`V()HrU&Ox8(Td)qjFl9e-TZl(@U-O#|(OO>TZo%ISp|WnCs001Y&{e=%
zV<}w_7wVctc#govZHsdMwuOYpW=Gro#s({YPBP*!|2iGI+R*UuXM-@x+z-9yo{YAl
zoG5#qBVvn^>GJ!AfV-PfnJbiM2LoU68BmxQBPW=00h9$iSs1pfFH
zDWk?O4dTLUzU#!FQMKNw71lBWF&RI65+5Lh+<2LokbV}llged2zTbudlmpQ>l%pR?
z0v!$S!8^5wAPeJ3_aI2s)}1i`L$fz8O@wxgM5
zeog6%3?8*YJ(RI43CJM&yb^TLR>)^X!O|8bz5y#%h*;i@PbpRO{)e-P#bLg^-G?y1
zk(aO({xPO$bD?kbopGa}7kj97q%laVU}V789s1g<9EsYB0J=~_|Ml+jzNENfImJSX
z%{!>#t9t7u*YPt#Xwft2xeTEW_`CgUHV4YWDXddmhVw$UP<)
z_QWhoiWVL90Yd7#D_7l3wr05Z>z!Wa~
z`Z|t64z{55tM=``*sYRK3{5#3YVpR5w{Y>xO(Rj>_NDBHepFSeuR5QFe=6*`g;^;T
z8zt>8@aQ3PZohsCPgD`aN{>)`J5c@LZsM+H1!dC
z0x8pPb^Q8l|#C*p(xzskZQ>D7odo7vz8uUP6;W-|1
z*f2A!LOs6kA_zbD#9y-C2J!>5k}U=Q9AiiFl
z{cv_}QD;b`#mTZ?IgxRn6>Z{?`FjFVw%>_utykL-hkZ|89N*lSu8|EOZD6hHqEK!B
z+3wxBsPih2@Z`AK&&WH;wob%>yQ6cUURy#Lu@arwvpMDX<_bsYJo)F~bXBPc-BrgD
z(*Cn?S*kohENd&+=zRNUcW?O%F%x5s>aj;A+Z=VD@x1ABc3@x^ASQbJSgQ?}SM*&E
zi~;IW*D1k=aN_B16VV4a6Rgr*gvI8Xa46Zx(rEnQ!Q}}}3SXz*Nf6Q#bCDyx!YR+b
z@>%40I@ke;Iob;3p-oS6E>``ZjtKec#YVE3qYN
z&oRn-BN9DAj%CX|uMpo$X`%JINyp4=41=R4Q|!v8TcxI|7B%8;)78}rm89MRZ8+cF
zF>~yB&06YSoMKbAah&fz4Ybz4@LMUVeNqnGuQO!DC*x68m95WZJog&8OP1X!?dyDsT7R6%Qc%Mp
z7W8ryVt|aVX4*`GKlZbv(SlCv9b!;UNl0;Hj!NJX10-pY99)J-Qd$P6ZJ
zd%*0*mutG7H;_Qs`WLwuD=~>v!tjhaV_)%*!9ac$eLHXc)FU5qvUu&T>NE%XZ}hJn
zn6COJ2X!#(qxqk%2E%e0*p5+LFkqcRO=I^`ES=}Urhj$orLTa?*&ubhwGQtK2i>;7
zTdrIZ7z6lYO75hM#5cnByBuE^GxZJotCS3j#RY}Q!$A{r+7rnOSWG__0$F_c?e=-X
z^tjL**41LDlB9{1hf6u1Vr`NRt*{Fk>_z>pxjL)ILc|1I_~Uh=L9!iBRUwaUg>upf
z!9DQ%S`&af*!=~$nYJDg{`TsuW418_2J>}wwW!)u^Kn*jaECgEy_mu{u%en8QljIs#>uA`B`JH}!u$&O!7mf!BTAJ?G%6NliK~0*3Mn=Y!
z7}Gf5MvuAn>?Ye(z0i!X_gqHp0p$+dXQt^MJ)sNNDePrL*)J`N@oAV&qA$_%iC5xp
z547wfO8YNOl-k(M-_JoQuks1>>A5YNY
z9go@8G?RMcC;x(}X3#{+y~pTqHeOA};qr#GI|spM^L`6a(LEfMoQ7$I{))Aawn*~@
zujJWZ8@IR~uD!3;*6EZ1#!dkRY;TOcmwP5`jP*h#KhTJzv>>+AH%^@o0m*@qYQamv
z?G>HW^IQoh91YS8#b;--#~rs2%@2oZzV;IPA|>?@2eU&WeP&e{8;^uBr>rn(XJISt
z$Z}6OY}f)@*o8=+e(e3PkjE{~nC+h15j3PdX4ka4@Y;(?evHbY6v0oj=Z<1*Gqlt~
z-s&B!0?(yd?!K%Wx^6H_);Eu=0x9M&5(R6&yy4j13Cin(A)+54otil|N`{->L)Ly5bT^6_b-yIyDd=E`Z-8UA(BHJjp4E#ud}K2d
zJ06&4D?mZ8LIm`t3;0$J2;oKF9dcRgz(8x>33FWJ0(YW!7=ao8l3eHUvw#T}?4kSE
zAWDBHt#;Uu%eRNIMlFFp&XxM}dj3VPGfa|Y`<1cVaKBh8h|z~wHk?)djH&M6=NCx1
z(8tWm=I|haQy9_{fFIYmCSIKE;TqIF@wkEjvFwj1>kk_T`Q89KJ1l~!gnYPvALqU!
ze$82Iw4UknWvX-tU`r7?d(Zo)#MV^GUe`uFlBmQy0z1t4*`zikE-7p7dpGujv=cUd
ztsF0N87E6R*+%UBQ(c&AMa98LLHQdX2e?-2=h3R8r?R0`n%??qf6N)#}kbN
zn{E2&D1p>%EWOmkLP)>ylVUwCv@LYukySegX!h|JizoxF2KeN+$Fdzs6n%60LX%?l
z7`Z-Qai|{Ce5ayKm^}3YKG17>ghZ&AW@?usTSs+_ICyG$=kY{IwY;eYkSY;$=cxU=
z`~|omUNv#6>vTDcoyFa86)N-H&dzO{m?q|5Y;^@lG>wQV3_O#%Y8zR?
zYm2>IkEx4Naf|pHhHCdLm?d=KK3lCiIWCzY7rum{c!0%+@oD#oSq@vYM=4}?2~^F2
z8EMSX?b^R7R_8lxTwfN=5yss^Fo;Zcx&F|C<+5<$e||oeZV8wsSft}nY96NM_Sn_F
zRC}yDFf-QaBAZ;Jr_@Hz-+z4``S8A1M#^E${12CG(06#1bywNGwevWsLV^xHO;{ts
zTJkK>m{TMbC(m0wglpZl;mKU@28#5zWwGz|eYD~8c{qxXL1rjWq&gcNG_nbkyV&CV
z=GJLog=Y%)slm+%LBK)7gA{m6yeXffs0yr_b;p_pvuRU!yUwobn?`p$`mJW~;x1>r
zD$fL@fl=b1>_k)JV;hGJFA$%TdyP@1*Cpr*S+>R2tK`dE^()VEa+v~E-lVFIo9^xH
zn2|=LXwT`u)D~}MVx>neH;gJ;tiFwAA|Dww04U
zGvyWR9KWLyAr@p+`=^ojui70)nUqmHjzaVf`kM*`Zw5c>E}6GEF4Zd9;Id^5`-om|
z5f=+eb59yJDKwUH0am`5<`-1^o-x$}(j&ZzcHH>@Wmu42Eu;v5TCQ^=;|nvxZh^7R
z<(h%>7njv-qN+wxhp{NnM{69z0vlKjQsmqR`B^vP%3rwbmOA5c+Rqx&6zq6F+d54e
z=_r_h;?<_%iR1P0T4Ujht+sGikWo}``e&$q0a82|6#h+qb+KXV%*Kmd&1HUD=x=9>#c%uyeOUdwM8Bb$>9uzID>w*v^(ERsgs9lhu4*NN
zZ|bdp=_B=2o>zx!|`0AJOW1*(!fk>1MEcivk?)XUN+{K@BK
zUl!=*wBEh!zcm$jF_ijI?2*CPWZHW64)^#iVwBd|Wpz$Lv}(Ue^^LG!`{M7t7FYv!
zDfsPRrU;0gbu%;-1MGF6iI&)`L4KH{X?6?s(i27JfcpLh&!db{6p4;*2mTtw;%}V7
za@yh)*QG28IXNQG@5PCcGWLqxY@E_hgmlZcez(kT_}+gxUYpa+=v;gz-U#A~H*)JE
z5$moXp6?VR%e%7460B^(p<3VrwpB$E9Emg}-P6#L=tq#*ks(%pgME9|A-FX(I)hi{
zc2V;69vm7HR<^zhm`~Zzcdu%G9af134}T9$I%67+)AE=X&%XXSx2mXwKAJu;xIJ{>6NUn~h0=
zzNbC{uZ_UDMX7(zp!H(UkN911g0CzsZGA^qq{`U@4sj_s$Z`ucUY_ms5YO?3D##Aa
zIaD%*F0B5nL^Tb_;B$mRe~Qr>85c}M`_4=p>VF3rAJZD04mK)AQZF?~7TJ>@bJdc7
z!4az2-#sh!8G9~Pt{V|p(5nv|i<(!%BZbJ_vp`76$zeoTV~5~7oj{>Seop3cy}d4J;xi_weF-pTsupF)!>@o%oRU?~RKmOGi^Jg{$|v2ejKqT;&)?{t)7B-~C9uy_?I;
z>2!rq_b@GGro52xgF-((nJXHouhpb|do^P9UV}ZneS3OeZnz|Ly-V&my@wUHx1S00
zzJE36_AsC@>EcIYX4ZFeXkJe7!?v+v?Si~IA8fJ^Zi)1g@F#|1+knd@MN2o{
zf@3_~%0mNPaZT?o8Zq}3PNvDjuL+ul+MF6-?CaMDpLc1-)7EO;4u1>5*PTy4PQIFz
z7$BveAY>I7gI>>&>hDghFbuUhq^A`xi8>uu8zMY(dI`@F&Ss{r5ExwRe2zvUV~=R(
zfQM5>!Cnb|3$_z-U)Rjz@a)NZhKePGXGmXqUlJK)jztKW)*L9Fx!q|2n}Cm{9}s{k}==z)Ot_AM(FCDg{4rGX7+>KxA^c)
z%$0+avVqN7rD>i-eUFR;ziAy0YUQ#-5QLP1{xvwy$Ed!cBvczCZB6q+CUD-uw2od1
zzPo6G+A30{L(AFaylycR%wEx;u4?X$(rX9?3F3E?F>E1?^Fgt~V!4OMNcHKXrKwac
zWf>XIHqK;&&8|yXa8$UQxnyq>4s7Ex8~0P$RJcWeW7xhKauw5r;3ATP?3VGV#d;6yZ?$1%(
z(9K%(O`z{ZS4EeY-%j4|@=>4#O^sEH#r@E5li8``GO$QcZR{cVWwy`B)Hyq~6L;6!
z9`s(t2?2xeQeaK*!#gX#*}GjMgHVf5E;|a>(H_MSU?H~WurNu7KaZ>`JhMxg3CLh_
ztdm|o&#QB@9z({8YSoo_WNSZ3hE*wEdIIne)dzHsb&gh5%2{5eyzhQO(3U=?JUBDs
zg3On3Z~;v@!|z=EdhBO#R7F19WbXiu6Kw`a&9HfM&X2C~R18%c&Hchxqz*-@*)Pu&
z?f@2>9=rNC>qI|{Z?Ao#`pJY2Ebl%OacXqYa#Y`0fD0N^xCeHZAvUOAUi;QQ8n2uW
zT>bK5H#@nvcQk^$F|er|hW#3*e5BxJH5pLexAc7H4qqa~O?-_1oC?*1eb6=Mh+l9N
zb3>yU=K~-t`?&0qjoZNj&lhf(EncErnysFrrTc@{_d3A{8b>)&T9bVI?*_!NN{L1!
zU}SW|w*s1UoHC5?8lnnO0~Fk
z@SRaf@K`hA^wp;4x&x%rI;DVb+{UvWx@{+$29E7=g)MY(q_P!Sv+1;TOP>!%C70We
z`(8cO8Hre&Ua?2bHcqt~R&#P!<GMe)xmXo?WugT
z={l26VX4FXgYDIkGuJCnzbE8)i~BGUcjwLI{R>m|UYzXM6Xy$d4FnLiR{N^>lKSl|
zd}8n3Tfbvrk_lD5%g6zcJD)9)hGFelWr;AKb_KwTJ>b$Hr}q-IHu>^ouTrgo@}Vbf
zuLeC3R5t8!*4+D?uC5PaS8UW~J*lq@VdZz(2a~**iVZhtXn?B5DZBc2y}`_mFY3&(
z1FpQv5ZG(gN!Re%(aIY7s1EO9jqR2blrcn3a|egNx^SiGrPto7g4m^8QJMD9AwQf&
z8Xa=G&Na-2VUo1rMFT^>Go>I!;2%Zvfe2oNJ)GQFKy{r+C=Y|$0p%+l&|;9E^R3nL
zt=qHUF%Em$;#WYaZ_V~MHf?svTTP^s*D@dAU*7;7g=sovAiGeLuHyO
zV&??li@=|j9RZvIpo}rnpV0dPB7K0Wkj~9#jV0N8{0Da+&>q6z!uF4)Xy-T2_&F4s
zdX6t!_6k^p<#9lye{>I88`HGu-wa21_J$Ke4xF8@%p_5PpvHK_PIpNy&wO1DDWZN_
z@TY}(^;c_TRZ>?iPM|-Y@Ce9fWbW9mA)7ra;Q}01D!uU4@vnXnv>Wx3P0n)-6fK)B
zjN3nTBJ?K$WM7-eIz)KlI^jj(u=y~K&Sfck4JXrw
zTc#-%e9S0>;(2TnEkFJ%x$9RECc>BG5mw!BZx3HjXAS+O__?;D|H1EVdc8D%_3WPF
zL2OcGnpEQBmUp~3wZ)y$exY8iwCakKeG#%@+3xVsl+3&McMfhJUq~MS5HD)T)r$_=c(*|kwVGM9Y!17tee@ljPTv>IILC(~md^?x%{=%q
z{e?au7J^Js`7B|)GpNiQ0TX|6{QwTCpGiaV0KDVT^Fug|uTV|8zP$mqw$nbkE;}Bz
z0I;Gx<&WHtPX}mugDQ6duPPM-DO(P6qtT9^TH<*nxZ6Vm6m`bxA&MF{vJ-}(4*cwE
z-)l?EcXxqd3rZ0DdfKW0xJYDX%TSj6id$FO+Z()Md(R;YvU7J(9of+@@Z^(+-OTri
z{YG5;otmjB#0o09>r8zNdc4L)HtjyLLATO7UgY>oKP`G2Ep~W3Hq$0gVSPdqLLWaV
z8vfj9Gb4i~;{GF~kkPorpvwWFuRo8A_?-XB4NRT_aOci%>s=!hQ<7pgX?Gpj^+5MOVN
zCiTZ^bGlVdOk9G`wFV(1Vx(5Ib!!6m*|wC+7?sWFJGY1D*4XWo07cSPk6f%uHU?6C
zS{i{*t-#WB^Is)@nk!c8XU|#WO4i)RE~ybZ#BSG=db%TmuCE$_L0EH(Ce1cjoC6Uc
zN>}{6CG?K;9$?)@J)fVlE+X?cQfcst%rC?kkFIeM^u5n}8u!abw3qor`qlxh@2`%8
z9U&7w-v}!IV$lr5Qu=0BThX5nnj8EX>C2y=u-BiJvKJY@z4?-GK7Az_a7IR4pm%-}
zX<8bp%~dzL%2gVH)0LYM;WR(fQRg~M<3tO0I*}jm5NYt*Bp10_0-3!Sv0wwPTpIek
zo+YH7LN|C(+fSc#I)1g;&;jNq`7w4I;kV>glZADFPyV$Pf}gUOzxLB)0kV)ZrJuL)
z1>xcy0k-SV6Eq+=I|Om0x1r-Fb461JSQ6>1&}Mk%632~ahJ&8T?Y#lQ6DTc5HncP>
z*1@q9F!@JhQ=)g%N5rDH4CN`ei7!k#)MB8W53mUYb35xEuoztV*Kzc&p^Ir4fld)@0K;BXj(?qy?hDa0dy9&NfIS68zX+~fPX%WwHx2Bc{G+3qZE
z$6?dPO5oX*f{^bX5C~W_Ji~S?kb~jPwAfoJ01?#fDyp(7rwB1sLrcSeJ|VwMBe;0z
z{PwLkf{0B4J1y1daC>Zq5$VZvB~bELi~M!`G_x`Zlc$r
z`@{~jl7rN=C;3rHuu<%4hEU(=Yw6lml$-UoGD(GU{o%NV>H?piSB5OnePsrJ;_5chVRtm?f0W{cFc2c}9A|10+PRKM
zySxVs{&>n}&p5Cf(!HQw}KNr-G)*{)USS8wacGs
zwK>LHZWd;dy1~)6i&F@VfliesyYH>)3m*i-pOyLaS$IsX;M~&((c8cn|YyDMp0DVQf*L)s+`TG*c>vh5FS+YrwgI}
z1};YRZ}+iezgbehO%7nTnA;~VGJ?28{&p1a2*lqt*t6VorHkA7gN4i_%EizDwF57B
zj{KGfP5hg+qp!2%Gx@``b@o0-COx8C=SUO5^5Aa2iyF`^5U2y;Jojner3g;dja91E
zP76@X(*?F0$Rm3oQ1ZZ(>MOOD}Iv
zoZjQ_{gc$JvHRjOr+eqeu-iT0#y})VbV8?CY|bm{P|)d#Qae$~zBM<*2&Pkpf$fc_
ze?*7i@p^vK$l5DZH2fxIa}SEc&rR5Zb_pqri1-^~zi
z@8+EfCC=$)_N#(dKyg8z_fAS9VBK
z+QS&s;Mm|Grh7Zc5Zaqlp5i3TZ+-q;ugOU@uY?^~wbUU^>WCda?Y+Msl9^|bmq4*v
z8aEZ~xZG0*neQ|@vly64crrCJ@V)x>;L)Ag_s;W};+?*)2b4-sClUSRC%7o@@b1wM
z#kBg-5AV$$$E+swd^~&WI-Z6Zt))AYDp1eukVqpz<7G?ohUvYDNsuY&q90x|J?qaR
z#GwyX+i^eAY#h*WuqcICuUVMpBANo7VmHsH|Gr2$2uot6%Gpsrw-fAm9aKJ|g3C=C
zeV2{(YMoWr>uz6NdL0lpXo%(eE6UYej3Q`QZ4UYdP+DJlwvFH&5P~_+9)Xv6hF@q|
z&UCgkIlvfKUhj6$C**T|X&J?Ihl7AC!C(m&=*mYq8?k?IF};+33HP9zsDC=D?D~J~
zy=6dD>-q<3fFKBo?}^@^xf71sP5r{
z^6lAM@hi@rI;A_Dx6GA(oN8!N5fyH@Zb<*65v{XQD&{aF8pqsCfUso6K^`dq5JwC-
z@~SLIuL9IZEB7F7G&y!r#jg?1y-iN|24DH%RCv}RNV(QT9VRFDvVffx#yJSBvxLC=
zQ3B8Bq-*5hWO^zKBK}ccCqkkbx0+z@bPwL@$g@RT5hKQxppg!_zvy?*WClc3V1b^brWl{&9J-K`OQ8WATGmq20
z-6DGMg|UwIN`{0yXk@WI%=*`>8Irmjv|=G8Do!%Hdbdh!GOFtuRsWnkdLZ%kbY_90
zcPEE-Ar9*it2iH_JS7wNs))Rd2Q9m@Ft1*}>6~}KpH}rE18yz^$N)Ef*t;RCzlz9<
zp8DmWH2(xN!L&fz(}*ouRCC*Qcy?&OKZm;KQnt!LTz!#rf^1lJ{YI}tR$$j**fQ>oS%G)+w1)@Li7;*{$K@j`Lt-k1}x&R3*$UQ!+d?aU5-RLvLOsI%@ONZ
z&5S?)4d?2rZfIjw&S7qV0E;MNr<$TZxR3W#fK7SY$-+1VZLV@bMjNRNdH%QdBL=YhVEb~pD;)jgf(x=%#6W@HFE
zV`6FXnQ$8pD=X)}aAVEbxSWgdiv=r>Scp7lNI%IWkTC)P`Kz!OWCjNc3w@uC;BX(;
ziowvZ3Pp}|1+VIKlRaU9%O=GBPX`sSZk=I1(@eF1VqYD3)qB#W*!`%)O9CEAIUoR?
zaQ;RuSj0L_y8zKK;ba7EF6@L2%HDnMnHFOT
z-9_sl7W=ox>wGEAT&ND6Z?=*Iyhu2}l#O-gh;LL$x<|d9ISZ--^Z1*;%oE8fChI=#
z*3Q=hvAd$eZYY5ZfSwsgGF)Mov@{&qCF4-9soZ;?V!Jy4YMkzIGS`zep$gJUx%^WD
z5e~|OA9CP;#VXf#gZXO>9co5{ip^aZGFfKLGulZC9`rhqBk-pjs
znN_vGWl6V8+9%z?C3CSH%!OB9XCBr>3kg*Lie^mP=ddB>#)iZ
z=MHAto+p1?n(aRhWR0MiISgWhluoNw!#E(weFDG1f|9^|wSzA9S{x{r^a1D>B5CkPO`
zxvgaIw^S7w<h@X<{QS_A-&ejBw
z@pcPPGurlkHtzG#u+;V650HAb{V@3riv8}{c$%lj!C_-Ij`i(+pR7os;(}{#<;mQp
z!%h-&`O;Ne!$TwcC_T>@hMD;8Ww)@7&X$zd)eBAHTK+6ImLnH{kJG#MSmzbixEXk}
znmA*ceZ|w$LdTf9AQ2B_?#o8#^1dXwwX_`>vxX%P^6<8c9Fm2^sscISg{l>
znHr(Ph*)*-oa+-a9us@5OtB5uK+2k~0l;HO{Yf_-t5B-Te33~|t}IuFX${~_k$&-L
z@_elz?6$27*lBL3kcX=tLT58TW2NIbf7ZM^GrWP;7d_=J#zbwQY(*cgG+O%1ylS+S
zzPiveKkk_RLu@U15+F*K6
za%o0BgHBew|8U`JS-U&go9dvub{Jl|a>!k&IMj+@HQkh)1d}9ErSJ5#_
z2r}Q~3l@avX)rAeVDYwp!$&wTCq66%%Nxt+jl{3pKNArhC!S&Ciz`j16rh7<=!0}&{PCR{rZbi&7ME+A@CzZdg4?VdWVv(#xS)yx
z_ejmkN@uGfO=8Z2OPm}^*EAF?bhod4lU8R!#q8
zbrfe%M>*3SZg=s{H)9uDhU8sf@Lj2vk#iuOr^{KG*PCh<8ACQ)EGxjV;KETwi4RA?
zP}>_Z^>`rKnnKpGGBd!Xj!UFCPYANH1bUE4?LGJR%!Ao0=ian^8Nkcf?y!QzeadVO
z7SH~SR>beTcAgN~LD3ELM`aWwm*K9ti_!fu$n+CJTuxYsbi6Dl=W&)6^vD^0I9PP~
z??4tvj|8k=!KKU}R$E@*6;^g$m3F3vT~65fJ7f=n6jCmCo>0`*9Zf6K^JzWHLnmq>
zG+QLDi-|cU@=by`HoVhu@<_G5q3IKu)waV;s<3ZnImyLfY1uED5oPS3saCbYO84Az
ze_uuE{{w=+3`-i&rh@)2W{}r^Tac(i_KS(dqvOL>Vh#WH+dLLMZ$*V#!8@qAVEaLhGa7Udl#FtO+;rEnBTF!AA%;3~
zaIx$aiH3qo4{}{CFHo`KbXM1Eu>HxhOIF*0HIpH$N)D@{@(-I>RU;s@!PNTzOX4^^
zq)Hjt|IyX^k}%Mw%XD5j(0;3shcW4Uv5Tb3b`hTRSRHiW*&3yCV*wQyvNc%qLv}|<
zlNabqcpQ!TxE4q-(<(o=u=@N3a%oJNVmZO)@JtjTiprOF-Tpc73xIdj(&$wA8uW@g
z&Aegd2ZR=rgfqI8R^FNFQ{P|(9SX@zeo^?=T;&pF&n-pPATbSPJejZGHqm!)<|Oqj^Ot&N1`
zcUbG<(JmVKu%_7H^4F4g2mda1nSixTAz<DCa#dPlg>
zWFgiWb_KS>(Lzuc||Y{`25R&vL=MWbIXx*dGK-38U-I>>&UtuLmX
zSN*5Ng@BTRiW@F^x8v5A{4o7ThFo}MM6Ht%qE(jpE}q(zkllL5EQUg8GPnLqX83kU+Gk<{98Z1dxU|&VM
zDy`;my~Icf#rc|13>2;7SRf2Xc%$D2LvMC@Z6}XtQ(|yTTs_R-ETWwWf-Wo2E@=MTI*JZW
z-ebwTq8RO(w$}B7a@4`(WN6A0;Vp<5^Cvt~tr)-ja3#4(#O5b*)Y>ZQz0IY;5)^>j
zHtpc@VA-Ab5!?0SzWlHNH1LUSgaV;S?o6T8(Ld8c&XD#n+zQvlzD@d=plag-XFKW
zJzC*7#-xL{UcbzApy=WKvoa?QjY|Qa_dq_tP)tA^NIh-}X_Ufx#00cl^zzAo_<20i
z<~x3U!Mk((!%04Pm)5&QM+87
zb?1V`;47{`=r=}MmQ8)-sAVh80wva((c9&Fj}^qX!!A*j|=Z_?u;<(nZYb2v_Gp2b>>*x}gpm|C4p>{u&*
zmB;#96Y>$Dj5m~%`|NVyG4AObOgYa&{_3?leR6GQDj{x1Uwbp%l25aB?H+$#W6XEC
zcC_CBY9l1{;cjKC5u>#hpL;J|^8uQg>bGR`;;0ZXi8#PeWTnK~{r&u8yYvo~AOmH|
zM={IN47wCJ5Bt88!NP9=RGQ=?$$24M@yE_Ba`i@o`&Z)6W~hsbo!orDpCXN)muCp@
zInniJb-w}MEkDNTi_KWfQZJ{fO6u;jM^T0JmB;8SlLbJo_f^GSLqmZyxz4TU^&=<5
zZjd1wOeCE^R@r4cJsjdEwJPHUzi8IiKw_C=H}qbkC*saS!O?CUc53hMGl%+}x#YW2
zvuGEqNQd(`2yonP^-11nl#z=!YVqFor=s7a7$uaPQXMbAQs4j{Q(C8VZC@pEr2O4T
zV93_g?xME;=@@5ugqx8W!tEJg;^ta^(sHyNPAf*K9nEX5`<(8-AL?q`05Iyv85dYI
zmLIpb=UikzdR6mK+mQ44T8*EZpkGUtjZI5<>igFVX}PlGJ1)KpSKQHT!~Y1>F&%dj
zs8QI5MR_^F&MC5ksQBeuQFml8R0o4r}it=191!IB}04Ue^@|5hYF#rY}~bU<*P;`~#b+_a=u1t!ww+Cb7AHE4P{g
z&FfdtMWZ9$ooMzzw9bW6uH4d!jw&sr8I-Aj{<-bvv#)#M0?^*rqsa{wy)%OqG@#{f
z_p#qoF@L{6MZs5%O-F($wZ@qRbG0*jiL_|b26p&FUb8+`$79oK#DQl89`x4Xm$D?>
zZC#X#ul)ReChp-=*~_`vTwt*3l)JFRdIX)AgXNaJ`WethY2>T-#XV3P4AfPXGZtbce0uJH)INq2K3g+q!!RfAqkB?a
zliTPGLO2VRW}Q~v4Vb{I!;N*R;n2X&8d^rLS4_*@
z>!yY>T|v6oKd+LJ5SyM#Sb;t)Q%aogwJ^wY_PjpD<}XT+P&ez~Jy?QNED1DP_30eDCS
zucIj(9wqyAkjy_I;OWlH=Pm*EREwWhZ*F+J!ff}s%Z?(JYMwcc6Q1FcI}P^@Jx0q8
zdcC{j%JNkxH(i0s`wOCFFFP?v4H3q1bx1^Eq;66fs3AcIS!facK&^T@tCf`HyzsaY
z9M9ZdWPa+XS7_+J7&zrCw92|_^SQalc6s`P!aZ?k?WZlLOKt%=P|e_bMdI2x3LNtL
zA38`EO5oPRdyiARM$dL%+Z%}xi0qy$u)bZflIS#A@rvw@4tnNWVcJE=@)^E;t#RX#
za!PhIFfcgSdQR-i-MeqeG;&p0IN0S$=8xBKUg&dpT{#o?3-`9179$XDjX+mBY2Rp#Yp-S#0F(7yHHg!$%GnL
zvemqjuI7H%O#Z_$-36@61OAtzYhlme_r#p;xV_9-tXay|6o!_x{t0bcvEL(EO_C
zdJzts&z!^2LCbGIHT^;F$qQUYk8vQ*USh&nS|``#R#df>O=LG)Va5qwx(*C>*%uGF
zaT2ZfW8J(Q5|y}4Na>tb-NUC1^R<~gdLM(rKfp#&jFWCWl1;v><3Ta~-Q-O}-k%*I
zN*{BDMHGU!0E0*j-t_^owll5c`lvGUcTV||s%+P*J|V7d^5q~qXpr$_{Q(B*fZN$3
z{ob?1yoA_%uP`_#^s6JY=ug>TNYVs_EIXfCJwFxFn4_OH79q~teLv5hak)_y`P)c!
zww!ECUn+l`*@>RyTwmT@1(~RObx+u!2Ne8gi&GJGt(xID0Zt~a^*DX3HFT7ZBA(93
zNh++zs5>wTbANZC#irojc)Z7~`lKJSFziMVBlTu`?H*GFcTpJI`#JkU1bMV;i6@9G
z^c@u{6e#}18A|o=v&bcGT7BmM2ntgp(_IqZY`9h|G~4g`xVVUS;4cvP*GgJ0kYb*P
zO9*M$=_PsuoM}5Aer=##gRg{?&0pFoJ)?fYVVk;?6m+@<0^-7x#^%rH5)4<=PL9ho
zm~}Tx0&A6rposCYT2=t2+L7q=+$S8hV4+dLk(_7!u`Ai=8!mv8hq)PlJ~{m8m+)d_%V;hdNT(g1p;Gg5eHURkT&
zJ_=2NS+&zNJtYV-0^jiL%bbzIiZTt1IHt|!iI
zdx)3YdgrVDYF0R02a^6xzh2^1B_QdXOMCHdNU~Z$G#kfF^+tITrH0A{lWRN4S|z{a
znC;w9G<P@Q*&(A;rypX
zv!TEfl1~4IYh&li7Gd(X5u*N{Q*ExEaX*yr{L28#iz(Lham%@C9%&gq(7?Y3pWvt+
zlRI*m=e7=B?Y1yTNdRc!jBXMlY~=+E^Ps`o4RFVrJ{%mC5rXg_8*7q3M
z8&|#vchoxe#Dt8v%8>L;m`+Fb1sNaICHv?^SIwu}$yA?~Y#Fjuo!8CM`sB+k#(Sj_
z;mFzCinJcFFPIGo>nZx9>(#oa7HCx{L|+b<$1K^A@e!GWB{>-t61
zoQ%v_&~b-ye;(+>EG{3N((*^LOl#Q+I6U-*=4d&8Ml0pQLehF@AI?G-jq+{FWy9*v@P2QaTO?`B;*pbbohfw+Ows^V+y)8lU0
zoYV=G&i6+y6T36ic3f7k6taF`Qz06L3||86t01@*-M{OUlYDd
ziLaFm|EM=!4jzjX1+W1f+$GRIWnzUNOi=jTu~U6I=Yt>Tp_>ZmK(9D~?rOg;P!47g
zllX%T{4)`;`$Zy}@2~fi&*S3=I0ln2NE7ZFiepQe(0?1gBG|z$Z_9+feEj5D(^?S+
z5{q8Vi+sfPCu(Eb-&ox)u3Ru<-6UG!oN=piiEp+0lz7Y26e$_T31*zGKCfmmf5Pw1
z+@1xLXr;*Y#jvc{L%(;Y2r{H!u}SPkni*RV`GToh?QT|$z{Y3FgSt5KNq?pkyPK1>
z11c%NiZ#PQWIokpm$L4-H}cF#!lj{{1MmKf!1Jc*{Q@Z~ilZKMUlqkK&&v@c%9!Z>0Y&p|Kamu6ue@
zo@a_2+Idr(t+fu&hGf2pZMo;*>w04uYH-$SB6u>*XI99}zHg2)BZ2D)L09(DL);(g
zZ~8F0SflT|xNwX(`@`Bj816vk@h`W|b+)M%xh&4d*)Ocq>01I8HTG(}iZ66-30~Uq~0-
z=$aRJaLPV-p?G=6-ga%+Nn7;@XDX-E4TV<^#Sy+q+dcoc$SB}
z`qiqJfHNk&ZnAdc!JdA1#uUjn59n;ZT`R_(t@Qwv%aulDC-Z4US24{U8T{x=Kyz#j
zgO9YRsqdwqXimLzG1B3$@4n%7iBK0?YEzjZj({WMdmX0(j2Cgv7N7OSz=
z8$aQ7ke{1aP~u2g4y2bgJ$czu7!~xBxD)~%8ROi{Lk2Hu`AptIIU(j2b)}HAG_>*C
zwVp2ki70ZX>t9uwbp$HV{Q0u6VQ*pO)@8>_`ppSxL=r^nhnS$7VZzl->y|!Pm))(t
z?sl60eD3yCbT+VgYXEB17caBcvyydNYl#Y@mDh>8)c=gzQ*yR^d&yO8+OCKODcqXvwh^*%s@
z9XvyKE>{Le3^*6l0t27IO&d%k;~M|I{M)_m12=31?!itvgcdNqu37r>&t3pfmhnLY
zq`dqFA3!si0F=vtFd%#`N0Pr3l-+FGyQHk
z5YupCEDXx>&gNZ6-N~?(;RKgs@ZbDN%CAia)Hy3zm$z~?r;O1Tn43XFb&?wWDT0h}
z;O!dN6xUJaqkj2LGd~v~(a_sdsZR^}TVu7RkDM3;lo=jM5q4YMO!z$;{hlmARnC%F
zFEZ`a6#;)pI&1fq{C14lw#UQp!;}!;^(;NJ)P*FS_wXBAanB!yzPCBmbx9z2VN5^V`)*x97Uo|H#nM?52O
zi`=ICo4+pn^7r_VAGfvs-{E1f!9d2&@V~>uAx>f&^Ep??PbFMf;ZWZa-|TZJ
zpC<$Fq^0xqB4A>Y7N*CI#`*R{6@(*VYPylITjPIwxy`oM-`hVA+1r`3x>^Jd>)sqC
z>sHMyGz_5MM-{&J0LS|t66L+W9yZW>r;Ty#k@OS)A1D0rvi*#EzUr(?%Y%RXfJ8YTFo?=r9~@PGa8U+z#x2Mif
z^^_L-cU6;u{@Jsa>-wve^mNLtEUPWo$iSJ6&~48Ca_VU;r`9J7(0R~#&Ol0{@5p41
zLM(!;ZA|2E+T@e>J(NOEPtTRq7VPHysJQ{ngCQA6FWja{v)e-zIaqY$
zN}(a>e*b+o)_(VbVD!rT)LvM+az~cj!6-?`8)@+m$Da-v68!p(qR+%=`FuQ(#aAafL3JW?bh=61T+(yKqH+%83LKXtF$Cr>DE
z&Xm_TRYc6g>Ck%(3vDO=_gxtmSszRY-+Hg0FRn5YQ8X&rV`$*o9r+CM|8ZKX4@gTN
z1(O$&R$a8L;ER~sGocieG%SYL`Tti`|FMWoVh?Ntbgq8}y~+s64~qOUirs2h)ZP}2
zs(JA8^qXgqjmPi9r7S_WUhFmwz(0!1^&8L0;C~fVu-=^bSznZ!>um3RKrV2
zo?H0X_UFpydqmh_p_16@9mmB){~fD$EmBpGxnf%CQ_(hEDeJ;H@n_|SPBexe3}uad
zlWY06|A$2sw#8kvZ2{(XKAa{Wvbfq7{QLu3hU_azGifBQhdF;xTkq+bmMd-d#@i5?
zo*C#MHmv+#TPTeG!yp$cc|V-E$L49i#Pa&%^3N)pH!S$1WY-w?^GPCx{$TFC&5<8G
z#Gg2Wh0uWEhfUI*n;_frN5Ai+-#r&$Tl`l6unZ?gC_%dsJkiOmWunNlo&q
z$A8)>?ke6L7Vr}7`s;e>A`}>Dp=5_RzGvNOAJOao;Wz<5|KQ<+P6~tkJaOPTDoAv_
zm;L=o1KWGw;XalYj|F^iOgmP?Z+y~3fu4p_{VFYrV!!_6L6gnjrPw9n?N;$9Sh=Ok
zVM#$RBLk%X4WJ1t7O&Zw*UCl#*DVEBP1kI_%0|H|XP&yAmIk=1ux7`KY1
zGQ_ReRzcl;K%jcgMThnGb@Nv6fsYTrMezF4=C3YB5k^sU>;+>|tiSKW`?21Qxr~JC
z6wwy!z<83hBhUkW-veEGz$<}~kwb-2zVefoxAVhoo7qz8BuUZ#$9+)W`mEqpghstj
zIW_tjEqAzT_~^f9QNBZW=CwEh4-zwawnRyGCVtazs^Uf!raVCBCpc+Ar1B|jMuEEF
zePW*cwpC;!a~)9|qd&H6qprTZQPSKTwdLXI$%hZzRnjqhs3x-!Eo*Ak=iwur}82qK9)QQ`6SnSEtsxXuUfJ{UjN(+y}5rA#~_EA
z^)9vBq;M!ZOa*fM)>#LarTd##pOnPU!2lJFxEy-B3XC`wKPMqpUavzm38@Z;8|he<
zxYE^5x&iDrINrj;Pxg(T8AKt&P_<%L!w58x@1vfjG{x7Rt|z|?eBXt}nWZm~P>MG{4A|3Cx%nOLm
zJvKda@28atG{o6HQh)3VCqdm{c3S1>?x+@cR1tk)dCI
zFiaiZM=WpL%ypFZ#o*T{#VEV1l;(fLy7Kr+(|GA;QMi`GoC|6#wdG+OD9}X^F(;Sk
zr7W#?3FN*%(6QALX!SWv&_^g=YcMFiPdn|Yi*hl1q|3cANmS6I$}BoY5|^zUj}|sK
zb~%yZ==jldBB7Q2dKq3Ca^G-Tx!UaK%Hh;9bffit8QgYD?BQ+eMSN(pfAt9
zdMgcf6CE@rz_vH%*0)5qIB>bW>P@M2>sOBYlNMcgjU6dXC-jD{o
zvq!1%;$$@6-BGXv8CFrJk6V7R-GfPs1?Z5btq1o%)Q#U$`tdwZ)`n_J1-FzYp0UM2
zxXIzc?2-LqeyTa~_N^xggZ|tN>%1*>&}mA>V=LCU9pmnV6v1S^;H#_c{HHHTf=y_6
zkYKg9+@k0?S7vC1IfcaU((`6jHt5MehUI*H42o~(9`stC<5Vpb6)}!i5TdC_i+9WS
z6oM1++HJqaJg<5fPuN`HDLVL_sQES0rzl~?XjJXFtIUQOKXn$RF1YUjbN9+o%g=bx
zdk;din?Qq!G{(}2gVkYYK^+xdF774x7f0a~_U`XrOa`{X)eU
zJ$n85xtwhT0$*j&Z0nLhB!5Pqko!)d`MhApd4dBzRU{4}@OLY<9Uup^xpzs^_JGE4
zBqjUPHuz5FWdKBFE9*!b9d=K
z-YsxvHDFTRv}*az?_6Pl5V#I7B>|@+8joW4FLl_Fk6>K8LTLgf1Kp0{A1iL${0qb8
z%etSW2tFL6Gh@liEYJ{A?}Sc3mUW+pE6a5BZTXG!e<@SUdeFYW9_dLeGqhxqq>W8J
zSkeRCm|vD6#KZ#6#&yHHASLrq^%qad$*b){+HXrG^*&LoV<2)US~0_0J|xsBhi_dawYURLDL>kc3mC
ztXYF`zSONwC3C)HnFu=z!2uFf;n5yB%r!Sbc!;VSwv`M{psmf@+nT*iw}
zU;KT?(Q}H1KR3KtnR47-Xj4T46#mG=7y
zsHjLjZBfT~+F0zGzu|8#%AGcr`9<5JzCb1xCFB_r-eU6V-O*S`zHQdM(=*T784dfR
zZrv%1Q-i)d$yi(4sf1_O&PDv>Yl|g2kg2hB2(U7KzW+-pAYf$rHMpDnncC!B88m
zIh)!+nmhks4H*v&q!RS)TZk?m3srd9ScRKU+88ITl_gwuuzT%#j>2oO!cd9>x}W8Y
z9T1Q`TyT_yj09`2ue2`w&Hq`H_POWf#ip0v$e48eR{JS*jJlWmEgs&P!-gY>(>${-
zL-eKHNN&g+yp;J(D#jPFoJgIxB&?fafq0jaU>(O33}1DPu<7v>&v12Mz*#3`UR;3L9RJL`D?oNTNQ)k4W^jS#e
zqcm;%J9)4z+%0I}vIVAXe>p_WK-c}!gaGas-=!BgjW{q6D
zLXI4T^s?7e7Qs5Xv(wkR8_q|YWHlbkg2w#`B5~<*Rx0#jCcf${cJ4cekb_fkrhv6>
zans?zZaXKV)uC3@53QF@6knzR4V9uC_=s;fZk9Ih1gyrgJ?Ue2{4?04J7dG9Kk$^J
zM-m+uSIPcng+Mx<=B060ijS}Fg9y%%g5^tFcnjvp)~lqa=$tRw4hBHaQ^be*%$+uT
zsye3H%xOF%kMV)OM4#v8GRMaD$QlgXlPilB
zGTofWmAm7b+fEtiI-MBJz*?Mc{y4}Sf!*3k4oumxyOo{Kv|qU&p%H9ky*55h>V^mKRfOpUcC5Vi(p1lTPc`jw=$}_0
zfI_wTjvgP%$sQ-TH>h+Sr*Z_?EyK2Rmbr5-$nB}VD;@z?VemzpsKh-jO8&1Gqj~0N
z)MEpAYY%I|1B-8~2qp=A*x^AUYTrl5c{0mFgG9{pa%Bb276HR)
z_R$(;g72S&QZMNHt`6sFOH;b9z)dpeHxl=$0|FetYuEeX8zU
z{Mjn#AkdzWkGQ~^6AY)h^yEu&)^Ro7Wa!UF9F;f|?M#Yz@$VfZvZyg!lzX5gRu~BE
z2-!(S(sy4Pubm1f^JXvVmGHd3or9
z37(7|4r?@d^)KzcM6mw@ucagd!%Q{q+@zw7G3@DkqODZ9m7tsnuGBMO)?<7vtE!3tyR<P(RZ61xx>f!V5p;%EalgGZv6wD(fu)^Jk2k4iE9PS{GO!mYAZ>z7bdpT(P^2z
z3J0R;F*!^%lFc|hnx8+JE!jw;OE*Ul?|bsyg(39T7sa=-Ba_*Z%U>9p9o*PhHls37
zUI&fxBL^#PCU*%y(C|#W8a~NW)EZ4k&A8nrP0+s@L@4%5k;9y<1k()2Xg)_4U0F>l
zSoOSj?LSkqKeW~v05{w^Bx>L+PZg#oCq>aEC6OnU7s>bWszNX66;cE|zj$8S@Fnxv
zNT^^Xr4A&%39qU|FzTAAXc8U~na6uXkVkFvF;8EHMeg#zToOB=%vyp%Kh{A(kzxDJ
z7acwDqlfL~S+--LU2sMst6_~j=7gR;-tli{nBhW=zua8G6OX44XFPe&pe~ayjZHkg
zM)8d6NhJ~VjZcU!G=%fU%jwOWnr?cHN)&dDs-Q2y`tcwHWpJLtB&xg_yC{r5=3ZcFywU@j)D-`
zzumF28c@s!d!Z#ZGF~seGQGwogU(*PLdNCg&@_h})cV!XeMyQBn*+(v^DDQ|;vcp?
z44MUpBA>kr58!ZwLqNRN~o>q?JQ(L}z2$P($vi0lFJ
z`K6H#&r2*5KJ_Ab^h$_Ku`U*i6o-|FRvXbH(qtXna>Q3QUAscRME3O3>Z8x{rkfX-
z<$D&hta+sC(B9-EL`d`)*i0Wr%)#`G2wbyNpDFmLl52M@6&9G>U2#Rs@Uq{g5fwfD
z9YwkA8J{tX$H@KPab9V)`wA<&EN=Q?Gw^QinzN76!LPik)-iiW+I}=nop}C$|LEJf
zsj{PmU7IwndLFoDhUn|Jwwc#c){RvZzz9kJd|o)@Zc?O>EY+tJ0I^739)wu&8}+{-;EPcpt%4$dR1FM0vD{T
z924nW?c3E@f9)(W(zt`&0ANC^mLDwO3gWo^@N!eJMQ+-UvUjbva%k
zS5-u+C!NOvALQv+K~_(kKttpDh~dWDudzwTZj~Hr1zMskBz@~{EJVyn(D{ocj!ZdN
zp&cjqTH*NBwf;a6$IgJ>Df#*NK@RcKk2lABxaE4U`X(&ZBMWuvUi!)|4=~f*aEmr@
za;IH6bfuOb)-K3hwx0@NQPg5WEYS`NARRl*UoX%hT4e*%G}$MlRqA;fk|bZZeAThM
zgh*d$SB%tX2$(U5I$s|v9+#(_OiA;Wzxg!~=Dgr}Yc4@x9?#Cv+PqP{e7lh}$^`uG
z{ITa!yZ~`{`~X3bqPJxD785l7nMQ-t**sHHeb*?vS4nzD60__t`!>TK)_%s0wT=(Z
zk#?`uWu$NUnbd`XT!0M)gq-{EELhK9=~iAKL_n*@w&``YJExMTCrHTNy^FVoMULNFT&=e*2n{Q
z@>NSFS<_#Iip@PE;Q}LMXkAZB-avQM6k`2zVD}SCDNrhw`J7757M01$8D^$0KEI1o
zaLgtuEdyLP4$nQQBNTS``Pf>AO!j7~Y#Buq*nuwoZK43ogjGe$2oO!2VWeKEc-(e<
zCe!rEI*5X39gE3Z0p%gNUE9j9>G$yh*i+&25*_Ab)M<%OfBj>QUs)r6pf
z{Q8%p9PEF5J)ft1(fitvshCSn^Gdf8}2AU^Rw|5@8w7Cy&plDr?r)
zH^5S!SN$S)y9jZ!-kr2oY~-Cc-aS9tEf;)Zrg(E2zRqzrBy0!Hw(Okduzq^luTgJA
z2WAF)Jog>H;Y5OYNB}DG)FeWi3R^$h440U{qF#qHQLFNb74&+JbRR){9=gtBFK34AYt>euSm@#sG)mz*Jix#byRnvK;L
z4yhdXbqIx$_ef!;%7C*Pr1Akz4!T0d==x@%&Xy~Em1BUBs-9WIJb4wK*OT@!`x^2X
zq^^K?3xowT}QQ+!|R0v*60ERT`3|_hI#sISvmicrKl;bO}8cfpVjmh&}1E-~K
z0?-RrpN)rUxP6J{-NClhjilVdJrd^zC*OAOcWw#k)+fpb+opB5)l+wbD8s%;JUyjU
z4yQiaFL}CfOyg1OQdasHj=QYlSfn;n{#7{+X9=Rq!Gm8e2qZm!p*K|0L8fzMwZ4O@
ziu0gkQTMAgjtO-Tr&w}PMuN9dGmbbfzx;svETMxUd^0ifMh1f*)0*L#=1=a~1z&(*
zyh{znIq!$;j?+yrlx8e3IWFzc;K-pSaWfB(`$MWWB
z|3zz%Y&~i8v3^%BNIdk95?}Ih4|L!skV};T@aEHdeWe|;b3UbknMV<$&Av)udN@6
z=4$$WKhok!;3W*_PSF7$ZvP*~-a0JGZu=h=1wlnZ5Rnd%l1}MH>FyMPp<8ljL0aih
zx?$*)mXaP?8tEQt==j}(dXCRI-*dh1KXZ*TbMLj+`s}q=&mG={e4@QG7K%d*S>og6
z(Teo9Tgror`(bgNd&>5S&)i%(xB^zM-DHP9*L{6kZ28RRDd5?Wu)!N<^)1{49g0Vb
z{1!c7i=x^Ov|idWsqkQa;d7Z?n!Ik{RB4p38=kP%sSLrBJ(qP@nYPXtQ*EzlGEDDo
zQT}Rwbr*7Km?ZAP^G-s)J4Ud;L>~;Z_Kb8_ug0P(t!-mNUC6;m$*ILrv)Cye`amhX
zdC_I@LeKA^9z>=xNfpazC?KJ6&GGDqxE4*3vTlB&t_$B=4D*aOjT}C`!Xby~fz`^i
zzFe`>n)~_}XFrdh$^+q!B2Spc1Fci(hY&H$eZKc)NsX=^``=Ntnig#OIQCCHdA!Js
zv5y}C8K8LqpHv}1#r!Fb?K*Z)*rNu`FoDkweR=7RZ`^0w_d*`Ro7zN>|Sv>
zj~e!$zuPt+yu*@~3SaR*lA`>IruhZT4~=BorrckfEb@nXO>Q25Q)W3
zJP<7v)TxwGUWpM+`@~e8dRwzh<1-y<5$Sb>lK;DiWrv^ZY*6a$$u8sMQap+6$4Dy`
za0`wpdo*r|P~HSqW`*)(*!aiO19twHkmaSwV%Z>ApyvXMa1)ME4IPfn-us$rkAdhe
z8qhLH53JJMSkXLQUppi2BK2M>Jub~Fn~{j?JJ`z_XOfbBW93}d&}qS|k`sH8(+iYb
zFTcC{%o4)%fo`P1O1)`s3N@3tkaIB1yriqP!joZZId?{xI+6EwV0)E8b~_pV&`o{9w}7nY_=(mw~R
z4m%_g>HRMx!*B0MeaHxXNSERwkOA}Q;^%p}_27|U&t{J5K{a$>&L*b7wXnB^)R}#Y
z!)>vfIFZ>6DX){3_+=X5W7-(WofHMA8?0ZHVALx3d5P7YmX$eUN2f~nNOwnDXk!kQ
zVCPR0Yt;xnaB8OjrP%t+X9|yD`Biznv{IYik-{tUcI{{$S*c(qkm@@q>JMf0uxRpx
zvc8M10*)-JItEUpm?hm;1DD26QHUIh>^kQGm{-s(RU+OwYO$bExJT
zYTi$5-H5o`KUB{Z5Vk2sMUTs_I}{#<3Ial|(LEXDq|T{BF%>CDdr^BRkR2BoJ_;me
zGd4BZrYh6>mQ`8n@nME=ImdCTaW&6~j)(B=oAZctg~s;m#{dhwve%w6oib(
zwoRb`ci8DXp*ZGmmi@M-Bgkcmf1K9wbSRB+UM;^k&p
z1VOvU$4$p%WEG5;(gZjzo%S5=5O8JLyY^Dahkc_Vv2n=9w^6=i(H%Qn2WFQ}jo7wkr;KKA?8oX7qyxmV@
zg`4WkCV+U>rAt0hk=%N}u?2%5`T`C!9jTQKG#_)kRuYSpt*zU_ztxz8S}Wm%$Q156
zbJ$XKtMJMd^o(L3q<2|vX9niXF`3)
z`Z+9=X5QG%vBvQVq|{#Sl80&32oW{u<}=rvfNdhP>y!)7(PSudl
zbydNsOmOaNdFhqJsP#>yPh5ZLY7iR`s_lLwL8@a98&4y3ik4
zuxT3GKPzMrxAn2WJ=7&3pUbVqSa>rpqdi^JP<6zQe(!6>)y#_--D%>|v<#DQhV
zUa{ifkn3pD5B6eAEmxT-$EV8LTjw&-oQ+tdIXXpq5F$P>_mJM1E<~kT9A$%Vo~qW8{h)nnZn(6;bV+$5RY`v(uJv%>
z2?oo>{w*$Bkx=UMJbZOE{t3d0BHZHP(FJ(E43SWyKz?NzXfM}myMS01yW;C&gAs0p
z+=(H6)I7Y>9bis9et?_q
z3E!lZ@9CT|?H>Xrv;#NCPj7mO(0c_kd7^C^sk%JMsVBBFf)fPOYsU_6HA$K27UiFh
zjD3=h16?p#ESc!U0<)-|v#*POJxfQmlN^lFNH{c987JV7MNG3sqyo~l|8{5v*19RTsxq9mb)n`!NF-oJs*J*0$$~4!@eAb)^6UNIlE9G;-FHcSS&8S!A|y*PU_5#+Vtv3aN=B+v~8^u
zba|GMRbjfx;@W$1+sHU$ifd>1LpV;1_yWmS{cHB_ieBTez)C`+PG
z8zziFWo<;ZCO8dDwWubFgAQ!kMKX*C1L%2%yUgIBH4!35^f=386z6z|Xx6j-g+tHk
zRr|LzJbw1kbB&~NRLEGBG>9KH-pKSQ!!aGb%xwr2G85P&!$HuLIxEI@y|Xz75}7{v
z#%tW!>Ml3vueicgRkg+FCLK&t&bKZqakX&E%ZIG-wO9YqUC3EvKPmY9uJOQ^i4F2v
zK6o^XG>soxO?mLRa^trNV^ERTHP6tmxI7lXY$R|X7lC$z#AO^C-3Xn8%|$U}bbAjO
zu+0dN)Wb$MB2zhz67ZAYr-2ehwR{z~Z>nB)9zju6bOa8{k5+67_wB4`$%*u#-OHTr
zeV@sheB8_X)<(4%wUcV?fPC$?l@L5`S9a@
z?ev`EzVqW{Lzj!;H*a*tPM(9A&1|4NpjkfIh`G|edeTK+OwazYyuZW0F%Yc|nw!S?y?srm2RPpd?SBCQ%4c)s#f1!H3>d}y2Za5h4k{hX()d9hV)eXP%R
zoMzD7>BsTGy4~B%_5-T|m5h!o$n_%&0EmMncjym=!yx=xaHGL@MT&kKc-2J3VqQ{i
z=jeIuOblOmy81>8qjmCC%hXrcM0#f5#)L;l(d|6`Np?-2a=85ogD=sGI?(31
z>gD7ZD&(we2}c~Z>zE$rA`;c;QE~n!mHZT^&JKj2=9&IZH{8k*vK30DY3kBf
zJh<|oEu3va<@U>@X7<`AU^!aVb2XG$6u&BSA!JA?*R;-vZjYf?VLOYg&(GOPRE={9
zV*ekRrjT3M8478letA1rmr@%)I#9Ke_KkhPw)10NkA1GFF
z8O1sV0PPi?5fYAxZ%=?}V59nvp!zASMb;kM>-r%_$dEIOc`Uc3xS0iBo%(^7goqox
zmAVj93pLzInn`68BP^?t8T>)-cf`vLrx(asTR#HwjbDfhM%Po}2`6;MK#Vhq5xWnl
zX#uR<0u;-KTBkKAemHkR|&zQ-Zs)1!b{3H-DI+F
z+!^!CN-8C5)ZHmx-zXi^ji5sA5OAc~yB=upqaT0Uerfd8V>U+EqLQM0ZP~sfEAc|<
zL@_^_T&77L(j{2wddO38vLFx4LwurR!`V^s5M&5
z28J1D$!e2Yft&Ip@-VfCS>_Lh<7fwf_}DC%%K9@A7;4TQTX{u*U9llvf{v&}az?)0
z&b@fGu!?dx2a{h3Ax|txjCX12SW4mdJeWQ}L?9!eic+mUU
zlAiah2h(e4--fFTd>-mKI&vP{iL|^Y6!RL)XZ!^cQ3TX($KzaAf56!0qs6@58PQ|y
z^*Chqy<2r5Bd}`uF$a@EQg%a5(=jVJ#_tsML|JMtako0dB3c0SrSeF(;QZQsmk2E2
zk!;T$NIJ@&99VHG605Otch%BGu;txAO1+5DTs&=cIW=e$RPgr5pF_Ah-3?mi*
zgP~{r>il^^ESo9oz~Fg)!tfaN@KjYLgS)p$u;)`p6JRPl#1#@DB;uv=M*Q}r+-NtM
z4sIp`d8oOH2)D|i(B~hO^pM%Ql(DTTO8Hm1lViDDh11afgcFgpy{0QENk+XA&|c+a
zWuDrnbiB=&$5nT^x{vz#cnB&sM6Cg@pKLE}{xG>=i7FR>4pkgqgB>in4CJ;{Qfj)b
zmFQG&d%S0)EdM5inn)uJLm8-oX$F4X_dQ=Nq-vv|1i}cn38|?f9bV_xSli7e^p=$DIU4&)N>{IQz;*IgD0$(*S?T+uy36Nt5UbZw8
z6*3xvmqAkbVvcQNQyiK<)_hZ~J@;tdYv8u{JwqG$=|w75DG!1Wppx9S*88gm4OdY#
zboo?Sl4*$AuTL~?1W1Sg37dZTr*dbu{kD{ZYE|pQKygf-`6Mf#o6lmV4&SLeK+WY}
zWx(>Lj=!j>1NJz9a3Kw>c_BaD)ygR^SDLhS)YqKl<7QYS59PC^BXg|zLeB2987{g(I<+8;;W*G>_;$rQl-!tYc#%d?q~@dWJ;R;X
z`ByFxopX+FyCXRnP(sM!sgWUez6jYX@o11TPZCxf;VkY+NF>oG+d{iZ6?j=E&OB9{
zst_2eby!z?>(6(gGPudFa_A`!8j8bUj4Qqc>LZIZYEUUP8T2cjCyi02{lnEFmoN=Kh}p}FDfNUKcF7r-QJ9d
z9ZH6DzSZX2F^RS8XqxH{LpM>%#`x4=0w+)65ejvo=y3L!;A_nI%3dg3o~U{@)I!;~
zCe_}vlF{Hrl4WM(6+Kn8qZ*M|RlpdxonYq#C3yT;R&>(iHFm^ez^k9$_T9msvXfI&
znqEd+U9NuSYdCQu=?Pm%KvV=~IgGD2%+7r;XHc!haw0yA)Twd`$nY;&T@Heen@X?G
zM@ce55sq=|1W>V5cm5$D*gZ$4ogOz8dy$7eagUb83
z!rtZ)h87~z1KcA3G6*a7W5RSxef#HXifX?p~HzVK9!9~
z?oJ<$kl?q1hndKxBzIJ@kqmqro0pi&rtASw08~n6>mMa}eX>!>eo^VZ*N!YNOnS9O
zJYt9&hjz75S%VdOElWz*Q4ii-3gWvPanRKe2dDB!?Msrxr0ksn50;jXDrj@YHlzEF
zag#aMJI`v>f3TiyVuCe)FzL&42`93j0IeWG{-Z@-_G#Vy#;W69WN$k6HDq>HZVG7*
z=R|Ek?fYEg7IVdYRBTp9^z>Wqt18=J{Kl$A{n278Ku=!nFFWq7JapBt(5LuF$d@wM
z#5n3Tus&z8Hkd!&%K8f|&HP%#mkS>Y*eNJdg34OZv~-t)-jSu^K~aUIx+ZdS@v%3+
zv*?mGSy7Z5}}MN>q#FK^_vdhJ0*52&+EntLYmNe#+6nQ~QzIJYQEU
zbU1*hHcKN&?W^q>UD!IWXi@QaI)xJfR0VK8Cdy1A%w#r0HHQ^V6yIU4n2m6d!@CvBl(5txsD}5?Dj9;vUe1Zc|U3pml!&wbU@nvs#_9--ZrBKI6~E>tf7+Ge8hbGFyZa~iYW
zmm>=NKlY*!-P5Yo+4_JoXVKw0!j^vr&9msZFd@C7{&Wsa`&$iAL^{Nk#Hj?b#WH@p
zYP&nqAYiW*RsLOs#wG6{(ALGwce~y1}m5=0DR%
z3goQ%c)+`}XyyHmxEC%(0X)t+7xCFUzpDW2)d
z>c!tV8a79B=ujR+LSW-3dQ-z(R_VGqa@|!HaWt=W3p#C@2{)Tmpm`Fq`ResGdBn(&
zNF1QX_ZD(wn0Iv#YG-g+iy{r-XG=Ys8U+}d$H87ig?o>p>+U`2fZ_Mb*MphPcs)+U
z=Ui!L#Tr-1v|LzhXS+FyQ*d6ek&goy3Zsauz9dY8Mq8}*oUcwTMAAab0ZpMX>_I2EbbBimAnXamgLpkHtF
zW2WKMJHRQP>{#)exid0kRqjHP^e}}gVzfqA6&ovXxX_M%=_{8wZ2C>Qa}k^J8=!ES
zC*Z*x?>Eha5vrmXr?XDfW0K+S+N2=&SEk*O$z@CE8IPEjxU8z|XJHG{NGo(Ez+Ui*
zVcloX*?R;(d~|Q1TLL09mqO7P0@PAy^uzw70g7j#cEt|HU#{#_j%gjYBTpSti`662
zKYk<&IiCC&!K<(?6`Mi{2g>0&^I+nefMZ6ACS0)8fsX7U(L5)Jf`Z~k1xzD47IjX+
zZVN&f&95YG1NUlg`RNZut=+)WKpXx90TA(ND+S9=PVE!j#=@V6JM_hC&ciP@MMW52
zUKn@EC0ZS3@tt-JrngxMD=xfC6F!|f)~d{|Phd5htZ_N#Y`9)jsttEv#jr8Qpj^gb
zj0fWD4R+T^VLJ5+4tMXpkB~U8A1uLE{cd{I&f3TvjLGS6U~+bzc{){CZ@tvRAO`zp
z$!C$+h*RumwE$0*;ltQk8vokG+T861UreEvgMOal>QTKg>H)I%Vo-v&T7jvmLmzG-=b$&_n8cXlN$9cS!Wh<$5!9NP(ePqug#cL^pBD9S~5t71Wq51v07bxD$^O#M_A
z>I-GHulOYp762c=EIbh~tkPO;BXs3k%K5(imO+*Q8i|V*pu?y}etvgWnf<5L0n88)
z5dzHcOhL~XCQKgJ)1>IWPnYG_
z0wAinScu#xH6;SQKggan+ICdP1kpbf7t0H
zdx|Wg90U}s*rcU2aa~)tQUwNjjtq_jTvaA=f@#2Kmj=ooXfe|hYu%;&*SGr&c?6At
zc#7$T{Nd`P*V}^M*<7GjuXj|s8svDKXJ6@tzR<);I^@3zga{;!3vgqm=n&TdV9}6)
zRG0S9kmj&$vpq^$`L$|QWS@24J?}3-HoXt{GxN3H8N&3iB3M_h#3y$qq
zWl+;;NcueIyu&A}dz_@Rl2BA>uDG4vzQlu})A7Mu<#JEyca?dzRn5gV^WeczG1qrRYyvb
z4zr9$t&mp}_u7TQvXpFst^gZwn^UK+vNnakY1@3AkQh@SRIFB*`of$f=YwALxoU&8
zL(uG&k#;4kmC@!oN8>oB{y3SPfV`NdgE1X`a@ef-tWIN54E*A|;X(Z49(I`D=cjBy
zTypArA6A9b{a!?l{*~HoVI-f#6bPT|-lue~P36IY$G@_=KYrvKs;#~*>ogcFa}DyN
zF@6DkvKf;pVAFZ|IT5P4CS8U{pOSJn+{_f;V9(Cf^Q+EqlA2(@IiOs(36IUhI)f62
zH`vZp-Ad(X`*@6h9+pZU?xotqh!0bu_5fi~vT>_(D^f6*VWkct2r+w6NS`GYrBtMB
zQp2^mmaLrhxcJ)fN{EFMVwL6h#EU@+sdze%U6OQQU4Fvetcjf=#Y3RnZd8t?Xh-YP
zeAqsus4`bPZZcqCqMWr91!5YZpN)yyJEkMzfmAaHCgCi?i78Zoj@(txAe|AN(vEgF
zsNNohrXRC5$u;4SuLEr=9G%r%pazetyA4yD7P5q-{hogIOI?MUoew;E2PAtlq5}IK
zM~y7vY<51{;VgVPYN^zp{^i714Y10vDzv6=7kYckqh+G+Ijr#!V`!b;DOl4uK+VP=
zCg%NRQg)&MN$Ev;av;2M`b|KmGPqx_F27a+!^LehPjp?W1q*`=-5<^dssV5mh
zK^fgA69LgH6k&>?&CqRpT6Vz5l!A9L`W*(N&pYj__PFI8Bev}go+)+)V_?MAc2j$chcmHr9H>9{+ygmWUq{=A(+{?Jae18f
z5PUt@^)cZ^RDUmn$4L0Zq_0_zDVCkG_LtK1DmN7pKVs=+qAqaoIt-iixeU~M3N2=<
zFv?peKQ6YInk~N9P<8y7UsAt#+!HNI&cn=5VX-5c$zhu5@Zxyz-QanmJFH;N*@NV4
zb`MQU69tiI^I4*vhpo!{q%Zh#Wqh;v*CJ%fG1Y+w38Os
zGuev8vV#?^i;741gTD81d&E5lI(-(mW5IQwj;`!v%B`BNE7OB3dqpdNUZ4H`CdL`L
z9uupPis4jaWAoW8yevf*g^6%|=%~<97az69{lzmglUp_3^BtI%!o{xi~y3Sy7PBl!3;chtfT>
zbq5~il}ELNed5nuHFc}CupZ7BiosqAXJAzMX;L-IzKKaca!#RQMY=M)H!{labR-X
z`81U+FS_T3nZ~U6JfV|tg|;y@CdC=OM3YS#
zG4)lLw#$`Q*4q-3ERFc<)s%|m?}mNm;0fJJ?)Z(cru7N>S=O2jE1X1UpijC=J@`>!
zfhv{8Xy;;bw!9dc@sb5Z#X-a{dZ^JeR95dgmWEvC(Dk07z)A_MDaSp6l(z1>Bu1yG
zpQsqD<(1}rH33?a(Ax&c5OZ=0M7?3G;3Itx?u_$=JA$1sag4{K4e~MIXh}(5wOQxf
zHA7F>r)3NOh%S9xg=1#ba8=3Blu2Y=_
z<<21>WQc^41}bpjk*z=2ZPNF-TEY6JtRA|j$xjuwC?s_8g7MpzPr8j|sq@}@!?k|w
zNHOgaUmpvGK;Om0K&uRbbPnfdDf?=ox_5}e-?_BvAbl7I3hE?)*4Fc4FQXsNQv?HZ
zufH_q5lI|?AgJqmqO0rWh7KH4`yxQ2r|}bhEE~OA(PLO*fz+OZN8xcc{L{}Opr|Tb
z7S-%1hu+mWBo9$yP_L{*MSmHkET|!MUS%d~j$sDmcZ=pm(w5us5uWxd{}X+)+bYWo
z+^=M#7$58Mxc4o%Hk!iZwyl0ById`CP}I*9uv3!O@?%&53(}9Tn*x54tH=B&-RIh$
zDXaU^yq2P9TE|&Zyc#JcmT9N!->Dypq`Zh5Wt*nFgfQu`L8NFvP6$vHoXHuG+-{`A
z!usfYyPedX8bB2zVej%yp(fzI+uQ>xQTInZX{17u8$4wleggG2Hnu<+O(IQ)y-LJ*
z{)A$6^L6w?>$68O{V~!44!aBDiHDm^_6er>A(7ws#mo#dg4~-h9+S_DVlU9)pz;D$
zs-KKl8L<{Qtnn5(31rV$UWZkcI(-2ORY?O|tujd7jrIR*7kMw+5D+ulys>S0FT>@$
zGresCOOZr_xY4eJx70%YJ}&=DmxcWjHcp##zt#SPSgTY0#WE&^sqPpCPI0KY@IvHq
z_*Yd7W$LS=BypG`oU_{lvOs0Fsdy0g7|5-qh)E8PzzNx@N@9%g`D3wQz;D%E7N=Oc
zHcjqNt~3#zgy3(m1i(9$m=5M7i7iGl-P#$XlPN)}(uO2JR0B|YyB*vR{1gt?h7Q}x
z*(WEz-S8DLt>b)2*(ufW!g6Du&W{&BV~^D7PyeI5i9?hPo|;XZV^y^Yb_se^n-uh`
zF=m|toofFz@hPpUoaDi}T?c-NdYMte1ZlE@>S|{Jmb(y-hdTMLv(ftOirKblIj-&a`Hp{ief?9N)E2?9Al==tbf?=!(^E%Z*VFU_
z(+yza`OMu>lu%Cy?a)VY`}LMas%)IjVlp+YX2Lg{_GM8lJ9
zGx?_NL)j$IcwUq}oR>jA3n-fVzU*)E*%gLg*>JjstySqBu@ZvT{jSBwXG9Y@#Ff<~
z{n%3{y%2BHTw!Km<+u$|>`+g`Rw%>K_M>N}9lv)hdz)GwN`Y-l05@0$4}DNt;NKh>
z^=J{9Qxk!yyX#FMq7&PFXL3SYvu
zdx7@M#%D$wbC`?NEHLwFlM9XU_vkTle2n(kUM*_3tmF!Un%F{>k9v3woDcB0yw)|q
zC9Uq_K_k_DR7#74UNOhKHJkou{hVgqSNT&tz%Oo(n(2MY=b*FTwk~wjPWIa8j#TTN
z5S!v6Dsg*ABr*O3h|l}p=BTsLsgzdE*O;SMoH!YSyR$n?-|utzd_i<^4d4
zjE~nK7};>>dnv2=GEylkoIb5!;)l=sh`W8+Y@Lif{7o9%@vOm-6-9}SF6Zt%RmY)u
zFB#o=*E_kk%@f^k7vu5gj2b76KD7PuL#DpSoU1@6Ev;oE2YvO_soZ2R#c|`ic*yF@
zXxNDaPIn^1ov=0b*8;_)lth=ajz>Q3}ngyL2pV0z$Ys^zMR7h3BkPERAGtc%v^=P`4sv%7v*vdJ5uCHt)l~DDSmXldwd*C4
zdXsa%PLuEx>$yyeH=BOSW|>0OEV`TTtzky_7P(AN%`i}lJuF|DRf{1
zndIGjAqt#%D7;CUOP0#{C3-Jzvd_1&>kC39zO~Q(w2BY;t?Gk7$
z0$kino;y-F+wiCMJh6sC3Y^7HVkX(~==1}dDDO-?CE99f=yVi~ezrmj?qH8*d{H@C
z1q)&Q3Ixt+weu!Z+2L*lSvqozY>JA5*(LDc8YYjF*0Ss;$1jQB8_Q95@>O>F>a%%Z
z){L7;49d-Zd#Qt`r_CGD89^atwn^v+Uy!|zaWAv*C7WU_ks&7-+k496a6V|=gR-5w
z&5SLRwc*`iT2rMFA;h`Uy>pElVX!=7^!aEYxR)h$-rtp$$|2#wjE6&LH20~51F*es
z&g2bye!tzd1F9z^fcX;zqfx_AXpW}vOfE!2^+|a1N{aI^V5)ZGko>L%W&AVa!~ikI
zfOO)#fjrQ6Z0(1}-~;lpWQW$OzaNYc0b-5jRg?=cn=vw=TS*y5-QYuL1*eODPXFyr
zAB3%Wf{Ykt0n}(m89TUo{1{kySXD=@v{1+T8T;nzF4Dh8vK+~vX>AXF`@fcWW`RV~
z!}}_9cq1>U(`1?DSH9rY&TYQroD}`(z&SpXEoU6hbbjrK*RgVU5)~ToVwPADz48bJ
zb#(vaJR}8vhaU@Qw%Myp7HGvhJ(?x6{A=3NAFct@Q8KZFElN=R#^QgN;u43EKpHRe
zTPA=6coA0o<*5`;{W)$HuF;fUoul^fTg1SQ6GQIzndn1>orpjbf*T}AB;?ODqA{yX
z?QbH{^!H$kokc0YA0U?{vPTtH;l0aXbaS0F)c?C
zJ?Q7iTefg|F=hn(oXQQMdPu-nl~03%hcXv||HnWp0B2l2AB6^DvO^yk`x%L!{+bpP
zS=W3Tr4AY%;9cJeq6c!=B3bm0GaoHKP?-HT5yktki1dbDooCHPKB%a!2BBlG6nRdb
zN}<-RRi!<>DMb4x7d&2jcY>CaA<6O3j8U9!?z_Lp`^BQ~hism&O2C4vynspLt5!e#
zTc&?=@>S98mUPJiFliGv<9iZF_>`gi-^=`f<(zC?n0h0>Zj3<+bB3XIxf`Ua6kC_
zk%;gS@cAtu6>>k_Qj=A=XzS;@?bY?LC^zWYGYVhE3+MONvyI{d$@~ZGOyR$7=$jQj
z`XeuRb_|HrmKh%e6-aRkZ^E^!%~tA3K~w$OxGWogW&wZ?2U;iv>>O+(|03+q+aJ#r
z8ZBw_bs-e+wI2bj@3X3{s4=Z#aQ((E10$dpq`>46UHA2ic4?BxHp9O}6&WRT%9@p@
zHYzoeA*#K{<{2R7x(*D!Bj$3w3uZIL7MBpG>Q3^Gy6IQ?rv_{0xJ};D!k($`OmKE)
zQDg>1fjr9sx)`>3*_SROK#MT18lk(gzxi^rB$*)+FwYL`KJ;;R250t1tM;??R1ON(
zNH_GSvPo52yUg?$fAQsyNW8gF2)J^^4;s%#l5)r*zNq+)DX;b`Ch6T-?;b7KpJe`P
zkk>6F`Frd|`hhEQXVc1ohP`iFnf{E#WP_7u?X|z
zKic762OlBSMslpDPsOPRKIbT#_7C>!qOMKC2@`STUl>{_p=mLN*j5TpF*=?f>4w?W
zvKr^A?4_%4dwF(8(NmwEW3Yl#R57A{y_~w#*yL88ItmAQ9*Oi{FHGEM)NlB*+Y|S
zl0>IuylyVOV<%)~+t&vvWX
z`(QIQS{I?DMVhN&@1H%bIC^?Ta(TFUcl}fX)<2?ulQH;DUI-06iD$OL0Q1`9)kEd1
z^waJcNp>5{51iOT&rZ5<=;n5fzdCg4u%6_gIONHczDYJ|+xr{vrOi3YjoIrCtFL|J
zth9zsjwO4HZTgo$)ITH_UW!Gi8K1rFMKd29A*!d_1ablYg$1FZ#~L-355fDj
zry=7ZCbF4sqxy&4DChRjNz*panJc8${KueRwGyZstC)Ks^^bx8*x03Qurp-LzEaNg
z?AwJIPOc$@z|?KLCl(%^zUk<;C$LZ@_%kWPs?<$rD6V6KI7Bx~H(H7z6HD1#jQ
z72te7z5*K9xN|3%2n>_-+3$7JVBs0!J)WGJCcW`KS0`4bTsn3#
z|F!ll