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

prepare scripts

Browse Source
This commit is contained in:
LinB203
2024-05-26 04:17:03 +00:00
parent de3f7d8a9f
commit 8aebbc64c0
22 changed files with 53 additions and 567 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scripts/accelerate_configs/hostfile
+2 -1
View File
@@ -6,4 +6,5 @@ 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_2
-5
View File
@@ -1,5 +0,0 @@
node121 slots=8
node104 slots=8
node106 slots=8
node99 slots=8
node122 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: 9
num_processes: 72
num_machines: 10
num_processes: 80
rdzv_backend: static
same_network: true
tpu_env: []
scripts/accelerate_configs/multi_node_example_2.yaml
-18
View File
@@ -1,18 +0,0 @@
compute_environment: LOCAL_MACHINE
distributed_type: DEEPSPEED
deepspeed_config:
deepspeed_config_file: scripts/accelerate_configs/zero2.json
deepspeed_hostfile: scripts/accelerate_configs/hostfile_2
fsdp_config: {}
machine_rank: 0
main_process_ip: 10.0.10.104
main_process_port: 29503
main_training_function: main
num_machines: 5
num_processes: 40
rdzv_backend: static
same_network: true
tpu_env: []
tpu_use_cluster: false
tpu_use_sudo: false
use_cpu: false
scripts/accelerate_configs/multi_node_example_node106.yaml
-18
View File
@@ -1,18 +0,0 @@
compute_environment: LOCAL_MACHINE
distributed_type: DEEPSPEED
deepspeed_config:
deepspeed_config_file: scripts/accelerate_configs/zero2.json
deepspeed_hostfile: scripts/accelerate_configs/hostfile_node106
fsdp_config: {}
machine_rank: 0
main_process_ip: 10.0.10.106
main_process_port: 29501
main_training_function: main
num_machines: 2
num_processes: 16
rdzv_backend: static
same_network: true
tpu_env: []
tpu_use_cluster: false
tpu_use_sudo: false
use_cpu: false
scripts/text_condition/sample_image.sh
+13 -9
View File
@@ -1,12 +1,16 @@
CUDA_VISIBLE_DEVICES=0 python opensora/sample/sample_t2v.py \
--model_path LanguageBind/Open-Sora-Plan-v1.0.0 \
CUDA_VISIBLE_DEVICES=4 python opensora/sample/sample_t2v.py \
--model_path LanguageBind/Open-Sora-Plan-v1.1.0 \
--version 65x512x512 \
--num_frames 1 \
--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 \
--version 65x512x512 \
--save_img_path "./sample_images/prompt_list_0" \
--fps 24 \
--guidance_scale 7.5 \
--num_sampling_steps 250 \
--enable_tiling \
--force_images
--ae_path "/remote-home1/yeyang/CausalVAEModel_4x8x8" \
--save_img_path "./sample_image_512x512" \
--force_images \
--guidance_scale 5.0 \
--num_sampling_steps 50 \
--enable_tiling
scripts/text_condition/sample_video.sh
-18
View File
@@ -1,18 +0,0 @@
CUDA_VISIBLE_DEVICES=$1 python opensora/sample/sample_t2v.py \
--model_path hw_65/model \
--version 65x512x512 \
--num_frames 65 \
--height $6 \
--width $7 \
--cache_dir "./cache_dir" \
--text_encoder_name DeepFloyd/t5-v1_1-xxl \
--text_prompt examples/$2 \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--save_img_path $3 \
--fps 24 \
--guidance_scale $4 \
--num_sampling_steps $5 \
--enable_tiling
scripts/text_condition/sample_video_257.sh
-14
View File
@@ -1,14 +0,0 @@
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
+5 -5
View File
@@ -1,7 +1,7 @@
CUDA_VISIBLE_DEVICES=4 python opensora/sample/sample_t2v.py \
--model_path 513f_512_9node_bs1_lr2e-5_4img/checkpoint-11500/model \
--version 65x512x512 \
--model_path LanguageBind/Open-Sora-Plan-v1.1.0 \
--version 513x512x512 \
--num_frames 513 \
--height 512 \
--width 512 \
@@ -9,9 +9,9 @@ CUDA_VISIBLE_DEVICES=4 python opensora/sample/sample_t2v.py \
--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" \
--ae_path "/remote-home1/yeyang/CausalVAEModel_4x8x8" \
--save_img_path "./sample_video_513x512x512" \
--fps 24 \
--guidance_scale 5.0 \
--guidance_scale 7.5 \
--num_sampling_steps 150 \
--enable_tiling
scripts/text_condition/sample_video_65.sh
+5 -5
View File
@@ -1,5 +1,5 @@
CUDA_VISIBLE_DEVICES=4 python opensora/sample/sample_t2v.py \
--model_path hw_65/model \
--model_path LanguageBind/Open-Sora-Plan-v1.1.0 \
--version 65x512x512 \
--num_frames 65 \
--height 512 \
@@ -8,9 +8,9 @@ CUDA_VISIBLE_DEVICES=4 python opensora/sample/sample_t2v.py \
--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" \
--ae_path "/remote-home1/yeyang/CausalVAEModel_4x8x8" \
--save_img_path "./sample_video_65x512x512" \
--fps 24 \
--guidance_scale 10.0 \
--num_sampling_steps 50 \
--guidance_scale 7.5 \
--num_sampling_steps 150 \
--enable_tiling
scripts/text_condition/train_videoae_129x512x512.sh
-46
View File
@@ -1,46 +0,0 @@
export WANDB_KEY="953e958793b218efb850fa194e85843e2c3bd88b"
# export WANDB_MODE="offline"
export ENTITY="linbin"
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_P2P_DISABLE=1
export PDSH_PATH_TYPE=ssh
accelerate launch \
--config_file scripts/accelerate_configs/multi_node_example.yaml \
opensora/train/train_t2v.py \
--model LatteT2V-XL/122 \
--text_encoder_name DeepFloyd/t5-v1_1-xxl \
--cache_dir "./cache_dir" \
--dataset t2v \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--video_data "scripts/train_data/video_data_513.txt" \
--image_data "scripts/train_data/image_data.txt" \
--sample_rate 1 \
--num_frames 129 \
--max_image_size 512 \
--gradient_checkpointing \
--attention_mode xformers \
--train_batch_size=2 \
--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="wandb" \
--checkpointing_steps=500 \
--output_dir="129f_512_10node_bs1_lr2e-5_4img" \
--allow_tf32 \
--use_deepspeed \
--model_max_length 300 \
--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"
scripts/text_condition/train_videoae_257x512x512.sh → scripts/text_condition/train_videoae_221x512x512.sh
+7 -15
View File
@@ -1,27 +1,19 @@
export WANDB_KEY="953e958793b218efb850fa194e85843e2c3bd88b"
# export WANDB_MODE="offline"
export ENTITY="linbin"
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_SOCKET_IFNAME=ibp25s0
export NCCL_P2P_DISABLE=1
export PDSH_PATH_TYPE=ssh
export PROJECT="65x512x512_10node_bs1_lr2e-5_4img"
accelerate launch \
--config_file scripts/accelerate_configs/multi_node_example_2.yaml \
--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 \
--cache_dir "./cache_dir" \
--dataset t2v \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--ae_path "/remote-home1/yeyang/CausalVAEModel_4x8x8" \
--video_data "scripts/train_data/video_data_513.txt" \
--image_data "scripts/train_data/image_data.txt" \
--sample_rate 1 \
--num_frames 257 \
--num_frames 221 \
--max_image_size 512 \
--gradient_checkpointing \
--attention_mode xformers \
@@ -34,13 +26,13 @@ accelerate launch \
--lr_warmup_steps=0 \
--mixed_precision="bf16" \
--report_to="wandb" \
--checkpointing_steps=500 \
--output_dir="257f_512_10node_bs1_lr2e-5_4img" \
--checkpointing_steps=250 \
--output_dir="65x512x512_10node_bs1_lr2e-5_4img" \
--allow_tf32 \
--use_deepspeed \
--model_max_length 300 \
--use_image_num 4 \
--enable_tiling \
--pretrained 512_10node_bs2_lr2e-5_4img/checkpoint-94000/model/diffusion_pytorch_model.safetensors \
--pretrained 65x512x512_10node_bs2_lr2e-5_4img/diffusion_pytorch_model.safetensors \
--enable_tracker \
--resume_from_checkpoint "latest"
scripts/text_condition/train_videoae_513x512x512.sh
+5 -13
View File
@@ -1,23 +1,15 @@
export WANDB_KEY=""
export WANDB_MODE="offline"
export ENTITY="linbin"
export PROJECT="513f_512_9node_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_P2P_DISABLE=1
export PDSH_PATH_TYPE=ssh
export PROJECT="513x512x512_10node_bs1_lr2e-5_4img"
accelerate launch \
--config_file scripts/accelerate_configs/multi_node_example.yaml \
--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 \
--cache_dir "./cache_dir" \
--dataset t2v \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--ae_path "/remote-home1/yeyang/CausalVAEModel_4x8x8" \
--video_data "scripts/train_data/video_data_513.txt" \
--image_data "scripts/train_data/image_data.txt" \
--sample_rate 1 \
@@ -35,12 +27,12 @@ accelerate launch \
--mixed_precision="bf16" \
--report_to="wandb" \
--checkpointing_steps=250 \
--output_dir="513f_512_9node_bs1_lr2e-5_4img" \
--output_dir="513x512x512_10node_bs1_lr2e-5_4img" \
--allow_tf32 \
--use_deepspeed \
--model_max_length 300 \
--use_image_num 4 \
--enable_tiling \
--pretrained /dxyl_code02/dev/Open-Sora-Plan/hw_65/model/diffusion_pytorch_model.safetensors \
--pretrained 221x512x512_10node_bs1_lr2e-5_4img/diffusion_pytorch_model.safetensors \
--enable_tracker \
--resume_from_checkpoint "latest"
scripts/text_condition/train_videoae_65x512x512.sh
+5 -14
View File
@@ -1,23 +1,15 @@
export WANDB_KEY=""
# export WANDB_MODE="offline"
export ENTITY="linbin"
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_SOCKET_IFNAME=ibp25s0
export NCCL_P2P_DISABLE=1
export PDSH_PATH_TYPE=ssh
export PROJECT="65x512x512_10node_bs2_lr2e-5_4img"
accelerate launch \
--config_file scripts/accelerate_configs/multi_node_example.yaml \
--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 \
--cache_dir "./cache_dir" \
--dataset t2v \
--ae CausalVAEModel_4x8x8 \
--ae_path "/dxyl_data02/CausalVAEModel_4x8x8/" \
--ae_path "/remote-home1/yeyang/CausalVAEModel_4x8x8" \
--video_data "scripts/train_data/video_data.txt" \
--image_data "scripts/train_data/image_data.txt" \
--sample_rate 1 \
@@ -35,7 +27,7 @@ accelerate launch \
--mixed_precision="bf16" \
--report_to="wandb" \
--checkpointing_steps=500 \
--output_dir="512_10node_bs2_lr2e-5_4img" \
--output_dir="65x512x512_10node_bs2_lr2e-5_4img" \
--allow_tf32 \
--use_deepspeed \
--model_max_length 300 \
@@ -43,5 +35,4 @@ accelerate launch \
--enable_tiling \
--pretrained t2v.pt \
--enable_tracker \
--resume_from_checkpoint "latest" \
--multi_scale
--resume_from_checkpoint "latest"
scripts/train_data/video_data_old.txt
-1
View File
@@ -1 +0,0 @@
/dxyl_data02/datasets/mixkit,/dxyl_data02/anno_jsons/tmp_video_mixkit_65f_54735.json