nodes/hao-ai-lab-FastVideo
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

prepare release

Browse Source
This commit is contained in:
root
2024-05-25 02:54:40 +00:00
parent 9b2951969f
commit de3f7d8a9f
28 changed files with 520 additions and 198 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/accelerate_configs/hostfile
+1 -2
View File
@@ -6,5 +6,4 @@ node120 slots=8
node121 slots=8
node104 slots=8
node106 slots=8
node99 slots=8
node122 slots=8
node99 slots=8
scripts/accelerate_configs/hostfile_2
+2 -6
View File
@@ -1,9 +1,5 @@
node117 slots=8 # your server name and GPU in total
node100 slots=8
node118 slots=8
node119 slots=8
node120 slots=8
node121 slots=8
node104 slots=8
node106 slots=8
node99 slots=8
node99 slots=8
node122 slots=8
scripts/accelerate_configs/hostfile_node106
-2
View File
@@ -1,2 +0,0 @@
node104 slots=8
node106 slots=8
scripts/accelerate_configs/multi_node_example.yaml
+2 -2
View File
@@ -8,8 +8,8 @@ machine_rank: 0
main_process_ip: 10.0.10.100
main_process_port: 29502
main_training_function: main
num_machines: 10
num_processes: 80
num_machines: 9
num_processes: 72
rdzv_backend: static
same_network: true
tpu_env: []
scripts/accelerate_configs/multi_node_example_2.yaml
+5 -5
View File
@@ -2,14 +2,14 @@ compute_environment: LOCAL_MACHINE
distributed_type: DEEPSPEED
deepspeed_config:
deepspeed_config_file: scripts/accelerate_configs/zero2.json
deepspeed_hostfile: scripts/accelerate_configs/hostfile
deepspeed_hostfile: scripts/accelerate_configs/hostfile_2
fsdp_config: {}
machine_rank: 0
main_process_ip: 10.0.10.100
main_process_port: 29502
main_process_ip: 10.0.10.104
main_process_port: 29503
main_training_function: main
num_machines: 10
num_processes: 80
num_machines: 5
num_processes: 40
rdzv_backend: static
same_network: true
tpu_env: []
scripts/causalvae/reconstruction.sh
+7 -8
View File
@@ -1,11 +1,10 @@
CUDA_VISIBLE_DEVICES=0 python examples/rec_imvi_vae.py \
--ae_path "/path/to/CausalVAEModel_4x8x8/" \
--video_path input.mp4 \
--rec_path rec.mp4 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--video_path /dxyl_data02/datasets/mixkit/Airplane/mixkit-sun-rays-over-forest-treetops-515.mp4 \
--rec_path rec_488_129f_512.mp4 \
--device cuda \
--sample_rate 1 \
--num_frames 65 \
--resolution 512 \
--crop_size 512 \
--ae CausalVAEModel_4x8x8 \
--enable_tiling
--num_frames 513 \
--resolution 256 \
--crop_size 256 \
--ae CausalVAEModel_4x8x8
scripts/text_condition/sample_video.sh
Regular → Executable
+11 -7
View File
@@ -1,14 +1,18 @@
CUDA_VISIBLE_DEVICES=1 python opensora/sample/sample_t2v.py \
--model_path 512_10node_bs2_lr2e-5_4img/checkpoint-94000/model \
CUDA_VISIBLE_DEVICES=$1 python opensora/sample/sample_t2v.py \
--model_path hw_65/model \
--version 65x512x512 \
--image_size 512 \
--num_frames 65 \
--height $6 \
--width $7 \
--cache_dir "./cache_dir" \
--text_encoder_name DeepFloyd/t5-v1_1-xxl \
--text_prompt examples/prompt_list_0.txt \
--text_prompt examples/$2 \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--save_img_path "./sample_videos_94k_7.5_150s" \
--save_img_path $3 \
--fps 24 \
--guidance_scale 7.5 \
--num_sampling_steps 150 \
--guidance_scale $4 \
--num_sampling_steps $5 \
--enable_tiling
scripts/text_condition/sample_video_257.sh
+14
View File
@@ -0,0 +1,14 @@
CUDA_VISIBLE_DEVICES=2 python opensora/sample/sample_t2v.py \
--model_path 257f_512_10node_bs1_lr2e-5_8img/checkpoint-5000/model \
--version 65x512x512 \
--image_size 512 \
--cache_dir "./cache_dir" \
--text_encoder_name DeepFloyd/t5-v1_1-xxl \
--text_prompt examples/prompt_list_0.txt \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--save_img_path "./sample_videos_257f_5k_5.5_100s" \
--fps 24 \
--guidance_scale 5.5 \
--num_sampling_steps 100 \
--enable_tiling
scripts/text_condition/sample_video_513.sh
Executable
+17
View File
@@ -0,0 +1,17 @@
CUDA_VISIBLE_DEVICES=4 python opensora/sample/sample_t2v.py \
--model_path 513f_512_9node_bs1_lr2e-5_4img/checkpoint-11500/model \
--version 65x512x512 \
--num_frames 513 \
--height 512 \
--width 512 \
--cache_dir "./cache_dir" \
--text_encoder_name DeepFloyd/t5-v1_1-xxl \
--text_prompt examples/prompt_list_0.txt \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--save_img_path "./testttt" \
--fps 24 \
--guidance_scale 5.0 \
--num_sampling_steps 150 \
--enable_tiling
scripts/text_condition/sample_video_65.sh
+16
View File
@@ -0,0 +1,16 @@
CUDA_VISIBLE_DEVICES=4 python opensora/sample/sample_t2v.py \
--model_path hw_65/model \
--version 65x512x512 \
--num_frames 65 \
--height 512 \
--width 512 \
--cache_dir "./cache_dir" \
--text_encoder_name DeepFloyd/t5-v1_1-xxl \
--text_prompt examples/prompt_list_0.txt \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--save_img_path "./testttt" \
--fps 24 \
--guidance_scale 10.0 \
--num_sampling_steps 50 \
--enable_tiling
scripts/text_condition/train_imageae.sh
-34
View File
@@ -1,34 +0,0 @@
export WANDB_KEY=""
export ENTITY=""
export PROJECT="t2v-f16s3-img4-128-imgvae188-bf16-gc-xformers"
accelerate launch \
--config_file scripts/accelerate_configs/deepspeed_zero2_config.yaml \
opensora/train/train_t2v.py \
--model LatteT2V-XL/122 \
--text_encoder_name DeepFloyd/t5-v1_1-xxl \
--dataset t2v \
--ae stabilityai/sd-vae-ft-mse \
--data_path /remote-home1/dataset/sharegpt4v_path_cap_.json \
--video_folder /remote-home1/dataset/data_split \
--sample_rate 1 \
--num_frames 17 \
--max_image_size 256 \
--gradient_checkpointing \
--attention_mode xformers \
--train_batch_size=4 \
--dataloader_num_workers 10 \
--gradient_accumulation_steps=1 \
--max_train_steps=1000000 \
--learning_rate=2e-05 \
--lr_scheduler="constant" \
--lr_warmup_steps=0 \
--mixed_precision="bf16" \
--report_to="wandb" \
--checkpointing_steps=500 \
--output_dir="t2v-f17-256-img4-imagevae488-bf16-ckpt-xformers-bs4-lr2e-5-t5" \
--allow_tf32 \
--pretrained t2v.pt \
--use_deepspeed \
--model_max_length 300 \
--use_image_num 4 \
--use_img_from_vid
scripts/text_condition/train_videoae_65x512x512_test.sh → scripts/text_condition/train_videoae_129x512x512.sh
+11 -11
View File
@@ -1,12 +1,12 @@
export WANDB_KEY=""
export WANDB_KEY="953e958793b218efb850fa194e85843e2c3bd88b"
# export WANDB_MODE="offline"
export ENTITY="linbin"
export PROJECT="512_10node_bs2_lr2e_4img_test1node"
export PROJECT="129f_512_10node_bs1_lr2e-5_4img"
export HF_DATASETS_OFFLINE=1
export TRANSFORMERS_OFFLINE=1
# export NCCL_IB_DISABLE=1
# export NCCL_DEBUG=INFO
# export NCCL_SOCKET_IFNAME=ibp25s0
export NCCL_DEBUG=INFO
export NCCL_SOCKET_IFNAME=ibp25s0
export NCCL_P2P_DISABLE=1
export PDSH_PATH_TYPE=ssh
accelerate launch \
@@ -18,29 +18,29 @@ accelerate launch \
--dataset t2v \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--video_data "scripts/train_data/video_data.txt" \
--video_data "scripts/train_data/video_data_513.txt" \
--image_data "scripts/train_data/image_data.txt" \
--sample_rate 1 \
--num_frames 65 \
--num_frames 129 \
--max_image_size 512 \
--gradient_checkpointing \
--attention_mode xformers \
--train_batch_size=2 \
--dataloader_num_workers 10 \
--dataloader_num_workers 8 \
--gradient_accumulation_steps=1 \
--max_train_steps=1000000 \
--learning_rate=2e-05 \
--lr_scheduler="constant" \
--lr_warmup_steps=0 \
--mixed_precision="bf16" \
--report_to="tensorboard" \
--report_to="wandb" \
--checkpointing_steps=500 \
--output_dir="512_10node_bs2_lr2e-5_4img_test1node" \
--output_dir="129f_512_10node_bs1_lr2e-5_4img" \
--allow_tf32 \
--use_deepspeed \
--model_max_length 300 \
--use_image_num 4 \
--enable_tiling \
--pretrained t2v.pt \
--pretrained 512_10node_bs2_lr2e-5_4img/checkpoint-94000/model/diffusion_pytorch_model.safetensors \
--enable_tracker \
--resume_from_checkpoint "latest"
--resume_from_checkpoint "latest"
scripts/text_condition/train_videoae_513x512x512_1node.sh → scripts/text_condition/train_videoae_257x512x512.sh
+7 -7
View File
@@ -1,11 +1,11 @@
export WANDB_KEY="953e958793b218efb850fa194e85843e2c3bd88b"
# export WANDB_MODE="offline"
export ENTITY="linbin"
export PROJECT="129f_512_10node_bs1_lr2e-5_2img"
export PROJECT="257f_512_10node_bs1_lr2e-5_4img"
export HF_DATASETS_OFFLINE=1
export TRANSFORMERS_OFFLINE=1
# export NCCL_IB_DISABLE=1
# export NCCL_DEBUG=INFO
export NCCL_DEBUG=INFO
export NCCL_SOCKET_IFNAME=ibp25s0
export NCCL_P2P_DISABLE=1
export PDSH_PATH_TYPE=ssh
@@ -21,7 +21,7 @@ accelerate launch \
--video_data "scripts/train_data/video_data_513.txt" \
--image_data "scripts/train_data/image_data.txt" \
--sample_rate 1 \
--num_frames 129 \
--num_frames 257 \
--max_image_size 512 \
--gradient_checkpointing \
--attention_mode xformers \
@@ -34,13 +34,13 @@ accelerate launch \
--lr_warmup_steps=0 \
--mixed_precision="bf16" \
--report_to="wandb" \
--checkpointing_steps=100 \
--output_dir="129f_512_10node_bs1_lr2e-5_2img" \
--checkpointing_steps=500 \
--output_dir="257f_512_10node_bs1_lr2e-5_4img" \
--allow_tf32 \
--use_deepspeed \
--model_max_length 300 \
--use_image_num 2 \
--use_image_num 4 \
--enable_tiling \
--pretrained 512_10node_bs2_lr2e-5_4img/checkpoint-94000/model/diffusion_pytorch_model.safetensors \
--enable_tracker \
--resume_from_checkpoint "latest"
--resume_from_checkpoint "latest"
scripts/text_condition/train_videoae_513x512x512.sh
+8 -8
View File
@@ -1,7 +1,7 @@
export WANDB_KEY=""
# export WANDB_MODE="offline"
export WANDB_MODE="offline"
export ENTITY="linbin"
export PROJECT="513f_512_10node_bs1_lr2e-5_2img"
export PROJECT="513f_512_9node_bs1_lr2e-5_4img"
export HF_DATASETS_OFFLINE=1
export TRANSFORMERS_OFFLINE=1
# export NCCL_IB_DISABLE=1
@@ -26,21 +26,21 @@ accelerate launch \
--gradient_checkpointing \
--attention_mode xformers \
--train_batch_size=1 \
--dataloader_num_workers 4 \
--dataloader_num_workers 8 \
--gradient_accumulation_steps=1 \
--max_train_steps=1000000 \
--learning_rate=2e-05 \
--lr_scheduler="constant" \
--lr_warmup_steps=0 \
--mixed_precision="bf16" \
--report_to="tensorboard" \
--checkpointing_steps=500 \
--output_dir="513f_512_10node_bs1_lr2e-5_2img" \
--report_to="wandb" \
--checkpointing_steps=250 \
--output_dir="513f_512_9node_bs1_lr2e-5_4img" \
--allow_tf32 \
--use_deepspeed \
--model_max_length 300 \
--use_image_num 2 \
--use_image_num 4 \
--enable_tiling \
--pretrained 512_10node_bs2_lr2e-5_4img/checkpoint-101500/model/diffusion_pytorch_model.safetensors \
--pretrained /dxyl_code02/dev/Open-Sora-Plan/hw_65/model/diffusion_pytorch_model.safetensors \
--enable_tracker \
--resume_from_checkpoint "latest"
scripts/text_condition/train_videoae_65x512x512.sh
+2 -2
View File
@@ -5,7 +5,7 @@ export PROJECT="512_10node_bs2_lr2e_4img_"
export HF_DATASETS_OFFLINE=1
export TRANSFORMERS_OFFLINE=1
# export NCCL_IB_DISABLE=1
# export NCCL_DEBUG=INFO
export NCCL_DEBUG=INFO
# export NCCL_SOCKET_IFNAME=ibp25s0
export NCCL_P2P_DISABLE=1
export PDSH_PATH_TYPE=ssh
@@ -34,7 +34,7 @@ accelerate launch \
--lr_warmup_steps=0 \
--mixed_precision="bf16" \
--report_to="wandb" \
--checkpointing_steps=200 \
--checkpointing_steps=500 \
--output_dir="512_10node_bs2_lr2e-5_4img" \
--allow_tf32 \
--use_deepspeed \
scripts/train_data/video_data_257.txt
-3
View File
@@ -1,3 +0,0 @@
/dxyl_data02/datasets/pixabay_v2,/dxyl_data02/anno_jsons/video_pixabay_513f_51483.json
/dxyl_data02/datasets/mixkit,/dxyl_data02/anno_jsons/video_mixkit_513f_1997.json
/dxyl_data02/datasets/pexels,/dxyl_data02/anno_jsons/video_pexel_513f_271782.json