theluyuan/DiffSynth-Studio
1
0
Fork 0
mirror of https://github.com/modelscope/DiffSynth-Studio.git synced 2026-03-23 17:38:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

wan-series

Browse Source
This commit is contained in:
Artiprocher
2025-11-14 19:05:26 +08:00
parent 5be5c32fe4
commit e3356556ee
215 changed files with 5504 additions and 482 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
examples/wanvideo/model_training/full/LongCat-Video.sh Normal file
View File

@@ -0,0 +1,12 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "meituan-longcat/LongCat-Video:dit/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-T2V-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-14B:Wan2.1_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/LongCat-Video_full" \
--trainable_models "dit"

16
examples/wanvideo/model_training/full/Video-As-Prompt-Wan2.1-14B.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vap.csv \
--data_file_keys "video,vap_video" \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "ByteDance/Video-As-Prompt-Wan2.1-14B:transformer/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-720P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-720P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-720P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.vap." \
--output_path "./models/train/Video-As-Prompt-Wan2.1-14B_full" \
--trainable_models "vap" \
--extra_inputs "vap_video,input_image" \
--use_gradient_checkpointing_offload

13
examples/wanvideo/model_training/full/Wan2.1-1.3b-speedcontrol-v1.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_motion_bucket_id.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-T2V-1.3B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-T2V-1.3B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-1.3B:Wan2.1_VAE.pth,DiffSynth-Studio/Wan2.1-1.3b-speedcontrol-v1:model.safetensors" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.motion_controller." \
--output_path "./models/train/Wan2.1-1.3b-speedcontrol-v1_full" \
--trainable_models "motion_controller" \
--extra_inputs "motion_bucket_id"

13
examples/wanvideo/model_training/full/Wan2.1-FLF2V-14B-720P.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-FLF2V-14B-720P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-FLF2V-14B-720P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-FLF2V-14B-720P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-FLF2V-14B-720P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-FLF2V-14B-720P_full" \
--trainable_models "dit" \
--extra_inputs "input_image,end_image"

14
examples/wanvideo/model_training/full/Wan2.1-Fun-1.3B-Control.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_control.csv \
--data_file_keys "video,control_video" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-1.3B-Control:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-1.3B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-1.3B-Control:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-1.3B-Control:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-1.3B-Control_full" \
--trainable_models "dit" \
--extra_inputs "control_video"

13
examples/wanvideo/model_training/full/Wan2.1-Fun-1.3B-InP.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-1.3B-InP:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-1.3B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-1.3B-InP:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-1.3B-InP:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-1.3B-InP_full" \
--trainable_models "dit" \
--extra_inputs "input_image,end_image"

14
examples/wanvideo/model_training/full/Wan2.1-Fun-14B-Control.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_control.csv \
--data_file_keys "video,control_video" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-14B-Control:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-14B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-14B-Control:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-14B-Control:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-14B-Control_full" \
--trainable_models "dit" \
--extra_inputs "control_video"

13
examples/wanvideo/model_training/full/Wan2.1-Fun-14B-InP.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-14B-InP:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-14B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-14B-InP:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-14B-InP:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-14B-InP_full" \
--trainable_models "dit" \
--extra_inputs "input_image,end_image"

13
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-1.3B-Control-Camera.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_camera_control.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-1.3B-Control-Camera_full" \
--trainable_models "dit" \
--extra_inputs "input_image,camera_control_direction,camera_control_speed"

14
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-1.3B-Control.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_reference_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-1.3B-Control:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-1.3B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-1.3B-Control:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-1.3B-Control:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-1.3B-Control_full" \
--trainable_models "dit" \
--extra_inputs "control_video,reference_image"

13
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-1.3B-InP.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-1.3B-InP:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-1.3B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-1.3B-InP:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-1.3B-InP:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-1.3B-InP_full" \
--trainable_models "dit" \
--extra_inputs "input_image,end_image"

13
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-14B-Control-Camera.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_camera_control.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-14B-Control-Camera:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-14B-Control-Camera:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-14B-Control-Camera:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-14B-Control-Camera:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-14B-Control-Camera_full" \
--trainable_models "dit" \
--extra_inputs "input_image,camera_control_direction,camera_control_speed"

14
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-14B-Control.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_reference_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-14B-Control:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-14B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-14B-Control:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-14B-Control:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-14B-Control_full" \
--trainable_models "dit" \
--extra_inputs "control_video,reference_image"

13
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-14B-InP.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-14B-InP:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-14B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-14B-InP:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-14B-InP:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-14B-InP_full" \
--trainable_models "dit" \
--extra_inputs "input_image,end_image"

13
examples/wanvideo/model_training/full/Wan2.1-I2V-14B-480P.sh Normal file
View File

@@ -0,0 +1,13 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-480P_full" \
--trainable_models "dit" \
--extra_inputs "input_image"

15
examples/wanvideo/model_training/full/Wan2.1-I2V-14B-720P.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 720 \
--width 1280 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-720P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-720P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-720P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-720P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-720P_full" \
--trainable_models "dit" \
--extra_inputs "input_image" \
--use_gradient_checkpointing_offload

12
examples/wanvideo/model_training/full/Wan2.1-T2V-1.3B.sh Normal file
View File

@@ -0,0 +1,12 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-T2V-1.3B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-T2V-1.3B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-1.3B:Wan2.1_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-T2V-1.3B_full" \
--trainable_models "dit"

12
examples/wanvideo/model_training/full/Wan2.1-T2V-14B.sh Normal file
View File

@@ -0,0 +1,12 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-T2V-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-T2V-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-14B:Wan2.1_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-T2V-14B_full" \
--trainable_models "dit"

16
examples/wanvideo/model_training/full/Wan2.1-VACE-1.3B-Preview.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "iic/VACE-Wan2.1-1.3B-Preview:diffusion_pytorch_model*.safetensors,iic/VACE-Wan2.1-1.3B-Preview:models_t5_umt5-xxl-enc-bf16.pth,iic/VACE-Wan2.1-1.3B-Preview:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-1.3B-Preview_full" \
--trainable_models "vace" \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

16
examples/wanvideo/model_training/full/Wan2.1-VACE-1.3B.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-VACE-1.3B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-VACE-1.3B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-VACE-1.3B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-1.3B_full" \
--trainable_models "vace" \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

16
examples/wanvideo/model_training/full/Wan2.1-VACE-14B.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--num_frames 17 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-VACE-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-VACE-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-VACE-14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-14B_full" \
--trainable_models "vace" \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

16
examples/wanvideo/model_training/full/Wan2.2-Animate-14B.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_animate.csv \
--data_file_keys "video,animate_pose_video,animate_face_video" \
--height 480 \
--width 832 \
--num_frames 81 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-Animate-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-Animate-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-Animate-14B:Wan2.1_VAE.pth,Wan-AI/Wan2.2-Animate-14B:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.animate_adapter." \
--output_path "./models/train/Wan2.2-Animate-14B_full" \
--trainable_models "animate_adapter" \
--extra_inputs "input_image,animate_pose_video,animate_face_video" \
--use_gradient_checkpointing_offload

35
examples/wanvideo/model_training/full/Wan2.2-Fun-A14B-Control-Camera.sh Normal file
View File

@@ -0,0 +1,35 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_camera_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-Control-Camera:high_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-Control-Camera:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-Control-Camera:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-Control-Camera_high_niose_full" \
--trainable_models "dit" \
--extra_inputs "input_image,camera_control_direction,camera_control_speed" \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_camera_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-Control-Camera:low_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-Control-Camera:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-Control-Camera:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-Control-Camera_low_noise_full" \
--trainable_models "dit" \
--extra_inputs "input_image,camera_control_direction,camera_control_speed" \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900]

35
examples/wanvideo/model_training/full/Wan2.2-Fun-A14B-Control.sh Normal file
View File

@@ -0,0 +1,35 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_reference_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-Control:high_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-Control:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-Control_high_niose_full" \
--trainable_models "dit" \
--extra_inputs "control_video,reference_image" \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_reference_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-Control:low_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-Control:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-Control_low_noise_full" \
--trainable_models "dit" \
--extra_inputs "control_video,reference_image" \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900]

33
examples/wanvideo/model_training/full/Wan2.2-Fun-A14B-InP.sh Normal file
View File

@@ -0,0 +1,33 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-InP:high_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-InP:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-InP_high_niose_full" \
--trainable_models "dit" \
--extra_inputs "input_image,end_image" \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-InP:low_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-InP:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-InP_low_noise_full" \
--trainable_models "dit" \
--extra_inputs "input_image,end_image" \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900]

37
examples/wanvideo/model_training/full/Wan2.2-I2V-A14B.sh Normal file
View File

@@ -0,0 +1,37 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-I2V-A14B:high_noise_model/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-I2V-A14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-I2V-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-I2V-A14B_high_noise_full" \
--trainable_models "dit" \
--extra_inputs "input_image" \
--use_gradient_checkpointing_offload \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-I2V-A14B:low_noise_model/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-I2V-A14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-I2V-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-I2V-A14B_low_noise_full" \
--trainable_models "dit" \
--extra_inputs "input_image" \
--use_gradient_checkpointing_offload \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900)

17
examples/wanvideo/model_training/full/Wan2.2-S2V-14B.sh Normal file
View File

@@ -0,0 +1,17 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset/wans2v \
--dataset_metadata_path data/example_video_dataset/wans2v/metadata.csv \
--data_file_keys "video,input_audio,s2v_pose_video" \
--height 448 \
--width 832 \
--num_frames 81 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-S2V-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-S2V-14B:wav2vec2-large-xlsr-53-english/model.safetensors,Wan-AI/Wan2.2-S2V-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-S2V-14B:Wan2.1_VAE.pth" \
--audio_processor_config "Wan-AI/Wan2.2-S2V-14B:wav2vec2-large-xlsr-53-english/" \
--learning_rate 1e-5 \
--num_epochs 1 \
--trainable_models "dit" \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-S2V-14B_full" \
--extra_inputs "input_image,input_audio,s2v_pose_video" \
--use_gradient_checkpointing_offload

33
examples/wanvideo/model_training/full/Wan2.2-T2V-A14B.sh Normal file
View File

@@ -0,0 +1,33 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-T2V-A14B:high_noise_model/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-T2V-A14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-T2V-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-T2V-A14B_high_noise_full" \
--trainable_models "dit" \
--max_timestep_boundary 0.417 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [875, 1000]
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-T2V-A14B:low_noise_model/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-T2V-A14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-T2V-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-T2V-A14B_low_noise_full" \
--trainable_models "dit" \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.417
# boundary corresponds to timesteps [0, 875)

14
examples/wanvideo/model_training/full/Wan2.2-TI2V-5B.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-TI2V-5B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-TI2V-5B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-TI2V-5B:Wan2.2_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-TI2V-5B_full" \
--trainable_models "dit" \
--extra_inputs "input_image"

40
examples/wanvideo/model_training/full/Wan2.2-VACE-Fun-A14B.sh Normal file
View File

@@ -0,0 +1,40 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--num_frames 17 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-VACE-Fun-A14B:high_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-VACE-Fun-A14B:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-VACE-Fun-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.2-VACE-Fun-A14B_high_noise_full" \
--trainable_models "vace" \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--num_frames 17 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-VACE-Fun-A14B:low_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-VACE-Fun-A14B:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-VACE-Fun-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.2-VACE-Fun-A14B_low_noise_full" \
--trainable_models "vace" \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900]

22
examples/wanvideo/model_training/full/accelerate_config_14B.yaml Normal file
View File

@@ -0,0 +1,22 @@
compute_environment: LOCAL_MACHINE
debug: false
deepspeed_config:
gradient_accumulation_steps: 1
offload_optimizer_device: cpu
offload_param_device: cpu
zero3_init_flag: false
zero_stage: 2
distributed_type: DEEPSPEED
downcast_bf16: 'no'
enable_cpu_affinity: false
machine_rank: 0
main_training_function: main
mixed_precision: bf16
num_machines: 1
num_processes: 8
rdzv_backend: static
same_network: true
tpu_env: []
tpu_use_cluster: false
tpu_use_sudo: false
use_cpu: false

12
examples/wanvideo/model_training/full/krea-realtime-video.sh Normal file
View File

@@ -0,0 +1,12 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "krea/krea-realtime-video:krea-realtime-video-14b.safetensors,Wan-AI/Wan2.1-T2V-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-14B:Wan2.1_VAE.pth" \
--learning_rate 1e-5 \
--num_epochs 2 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/krea-realtime-video_full" \
--trainable_models "dit"

14
examples/wanvideo/model_training/lora/LongCat-Video.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "meituan-longcat/LongCat-Video:dit/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-T2V-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/LongCat-Video_lora" \
--lora_base_model "dit" \
--lora_target_modules "adaLN_modulation.1,attn.qkv,attn.proj,cross_attn.q_linear,cross_attn.kv_linear,cross_attn.proj,ffn.w1,ffn.w2,ffn.w3" \
--lora_rank 32

18
examples/wanvideo/model_training/lora/Video-As-Prompt-Wan2.1-14B.sh Normal file
View File

@@ -0,0 +1,18 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vap.csv \
--data_file_keys "video,vap_video" \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 10 \
--model_id_with_origin_paths "ByteDance/Video-As-Prompt-Wan2.1-14B:transformer/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-720P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-720P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-720P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Video-As-Prompt-Wan2.1-14B_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "vap_video,input_image" \
--use_gradient_checkpointing_offload

15
examples/wanvideo/model_training/lora/Wan2.1-1.3b-speedcontrol-v1.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_motion_bucket_id.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-T2V-1.3B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-T2V-1.3B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-1.3B:Wan2.1_VAE.pth,DiffSynth-Studio/Wan2.1-1.3b-speedcontrol-v1:model.safetensors" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-1.3b-speedcontrol-v1_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "motion_bucket_id"

15
examples/wanvideo/model_training/lora/Wan2.1-FLF2V-14B-720P.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-FLF2V-14B-720P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-FLF2V-14B-720P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-FLF2V-14B-720P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-FLF2V-14B-720P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-FLF2V-14B-720P_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,end_image"

16
examples/wanvideo/model_training/lora/Wan2.1-Fun-1.3B-Control.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_control.csv \
--data_file_keys "video,control_video" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-1.3B-Control:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-1.3B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-1.3B-Control:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-1.3B-Control:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-1.3B-Control_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "control_video"

15
examples/wanvideo/model_training/lora/Wan2.1-Fun-1.3B-InP.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-1.3B-InP:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-1.3B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-1.3B-InP:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-1.3B-InP:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-1.3B-InP_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,end_image"

16
examples/wanvideo/model_training/lora/Wan2.1-Fun-14B-Control.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_control.csv \
--data_file_keys "video,control_video" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-14B-Control:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-14B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-14B-Control:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-14B-Control:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-14B-Control_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "control_video"

15
examples/wanvideo/model_training/lora/Wan2.1-Fun-14B-InP.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-14B-InP:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-14B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-14B-InP:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-14B-InP:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-14B-InP_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,end_image"

15
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-1.3B-Control-Camera.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_camera_control.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-1.3B-Control-Camera_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,camera_control_direction,camera_control_speed"

16
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-1.3B-Control.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_reference_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-1.3B-Control:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-1.3B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-1.3B-Control:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-1.3B-Control:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-1.3B-Control_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "control_video,reference_image"

15
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-1.3B-InP.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-1.3B-InP:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-1.3B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-1.3B-InP:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-1.3B-InP:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-1.3B-InP_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,end_image"

15
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-14B-Control-Camera.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_camera_control.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-14B-Control-Camera:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-14B-Control-Camera:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-14B-Control-Camera:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-14B-Control-Camera:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-5 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-14B-Control-Camera_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,camera_control_direction,camera_control_speed"

16
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-14B-Control.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_reference_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-14B-Control:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-14B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-14B-Control:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-14B-Control:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-14B-Control_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "control_video,reference_image"

15
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-14B-InP.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.1-Fun-V1.1-14B-InP:diffusion_pytorch_model*.safetensors,PAI/Wan2.1-Fun-V1.1-14B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.1-Fun-V1.1-14B-InP:Wan2.1_VAE.pth,PAI/Wan2.1-Fun-V1.1-14B-InP:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-14B-InP_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,end_image"

15
examples/wanvideo/model_training/lora/Wan2.1-I2V-14B-480P.sh Normal file
View File

@@ -0,0 +1,15 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-480P_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image"

17
examples/wanvideo/model_training/lora/Wan2.1-I2V-14B-720P.sh Normal file
View File

@@ -0,0 +1,17 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 720 \
--width 1280 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-720P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-720P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-720P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-720P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-720P_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--use_gradient_checkpointing_offload

14
examples/wanvideo/model_training/lora/Wan2.1-T2V-1.3B.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-T2V-1.3B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-T2V-1.3B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-1.3B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-T2V-1.3B_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32

14
examples/wanvideo/model_training/lora/Wan2.1-T2V-14B.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-T2V-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-T2V-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-T2V-14B_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32

17
examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B-Preview.sh Normal file
View File

@@ -0,0 +1,17 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "iic/VACE-Wan2.1-1.3B-Preview:diffusion_pytorch_model*.safetensors,iic/VACE-Wan2.1-1.3B-Preview:models_t5_umt5-xxl-enc-bf16.pth,iic/VACE-Wan2.1-1.3B-Preview:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-1.3B-Preview_lora" \
--lora_base_model "vace" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

17
examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B.sh Normal file
View File

@@ -0,0 +1,17 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-VACE-1.3B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-VACE-1.3B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-VACE-1.3B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-1.3B_lora" \
--lora_base_model "vace" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

18
examples/wanvideo/model_training/lora/Wan2.1-VACE-14B.sh Normal file
View File

@@ -0,0 +1,18 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--num_frames 17 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-VACE-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-VACE-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-VACE-14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-14B_lora" \
--lora_base_model "vace" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

20
examples/wanvideo/model_training/lora/Wan2.2-Animate-14B.sh Normal file
View File

@@ -0,0 +1,20 @@
# 1*80G GPU cannot train Wan2.2-Animate-14B LoRA
# We tested on 8*80G GPUs
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_animate.csv \
--data_file_keys "video,animate_pose_video,animate_face_video" \
--height 480 \
--width 832 \
--num_frames 81 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-Animate-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-Animate-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-Animate-14B:Wan2.1_VAE.pth,Wan-AI/Wan2.2-Animate-14B:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Animate-14B_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,animate_pose_video,animate_face_video" \
--use_gradient_checkpointing_offload

39
examples/wanvideo/model_training/lora/Wan2.2-Fun-A14B-Control-Camera.sh Normal file
View File

@@ -0,0 +1,39 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_camera_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-Control-Camera:high_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-Control-Camera:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-Control-Camera:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-Control-Camera_high_niose_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,camera_control_direction,camera_control_speed" \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_camera_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-Control-Camera:low_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-Control-Camera:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-Control-Camera:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-Control-Camera_low_noise_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,camera_control_direction,camera_control_speed" \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900]

39
examples/wanvideo/model_training/lora/Wan2.2-Fun-A14B-Control.sh Normal file
View File

@@ -0,0 +1,39 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_reference_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-Control:high_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-Control:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-Control_high_niose_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "control_video,reference_image" \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_reference_control.csv \
--data_file_keys "video,control_video,reference_image" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-Control:low_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-Control:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-Control:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-Control_low_noise_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "control_video,reference_image" \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900]

37
examples/wanvideo/model_training/lora/Wan2.2-Fun-A14B-InP.sh Normal file
View File

@@ -0,0 +1,37 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-InP:high_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-InP:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-InP_high_niose_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,end_image" \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-Fun-A14B-InP:low_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-Fun-A14B-InP:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-Fun-A14B-InP:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-Fun-A14B-InP_low_noise_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,end_image" \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900]

39
examples/wanvideo/model_training/lora/Wan2.2-I2V-A14B.sh Normal file
View File

@@ -0,0 +1,39 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-I2V-A14B:high_noise_model/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-I2V-A14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-I2V-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-I2V-A14B_high_noise_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-I2V-A14B:low_noise_model/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-I2V-A14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-I2V-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-I2V-A14B_low_noise_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900)

19
examples/wanvideo/model_training/lora/Wan2.2-S2V-14B.sh Normal file
View File

@@ -0,0 +1,19 @@
accelerate launch --config_file examples/wanvideo/model_training/full/accelerate_config_14B.yaml examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset/wans2v \
--dataset_metadata_path data/example_video_dataset/wans2v/metadata.csv \
--data_file_keys "video,input_audio,s2v_pose_video" \
--height 448 \
--width 832 \
--num_frames 81 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-S2V-14B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-S2V-14B:wav2vec2-large-xlsr-53-english/model.safetensors,Wan-AI/Wan2.2-S2V-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-S2V-14B:Wan2.1_VAE.pth" \
--audio_processor_config "Wan-AI/Wan2.2-S2V-14B:wav2vec2-large-xlsr-53-english/" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-S2V-14B_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image,input_audio,s2v_pose_video" \
--use_gradient_checkpointing_offload

38
examples/wanvideo/model_training/lora/Wan2.2-T2V-A14B.sh Normal file
View File

@@ -0,0 +1,38 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-T2V-A14B:high_noise_model/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-T2V-A14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-T2V-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-T2V-A14B_high_noise_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--max_timestep_boundary 0.417 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [875, 1000]
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-T2V-A14B:low_noise_model/diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-T2V-A14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-T2V-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-T2V-A14B_low_noise_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.417
# boundary corresponds to timesteps [0, 875)

16
examples/wanvideo/model_training/lora/Wan2.2-TI2V-5B.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--num_frames 49 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.2-TI2V-5B:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.2-TI2V-5B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.2-TI2V-5B:Wan2.2_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.2-TI2V-5B_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image"

43
examples/wanvideo/model_training/lora/Wan2.2-VACE-Fun-A14B.sh Normal file
View File

@@ -0,0 +1,43 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--num_frames 17 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-VACE-Fun-A14B:high_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-VACE-Fun-A14B:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-VACE-Fun-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.2-VACE-Fun-A14B_high_noise_lora" \
--lora_base_model "vace" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload \
--max_timestep_boundary 0.358 \
--min_timestep_boundary 0
# boundary corresponds to timesteps [900, 1000]
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata_vace.csv \
--data_file_keys "video,vace_video,vace_reference_image" \
--height 480 \
--width 832 \
--num_frames 17 \
--dataset_repeat 100 \
--model_id_with_origin_paths "PAI/Wan2.2-VACE-Fun-A14B:low_noise_model/diffusion_pytorch_model*.safetensors,PAI/Wan2.2-VACE-Fun-A14B:models_t5_umt5-xxl-enc-bf16.pth,PAI/Wan2.2-VACE-Fun-A14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.2-VACE-Fun-A14B_low_noise_lora" \
--lora_base_model "vace" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload \
--max_timestep_boundary 1 \
--min_timestep_boundary 0.358
# boundary corresponds to timesteps [0, 900]

14
examples/wanvideo/model_training/lora/krea-realtime-video.sh Normal file
View File

@@ -0,0 +1,14 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "krea/krea-realtime-video:krea-realtime-video-14b.safetensors,Wan-AI/Wan2.1-T2V-14B:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-T2V-14B:Wan2.1_VAE.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/krea-realtime-video_lora" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32

16
examples/wanvideo/model_training/special/fp8_training/Wan2.1-I2V-14B-480P.sh Normal file
View File

@@ -0,0 +1,16 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-480P_lora_fp8" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--fp8_models "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"

28
examples/wanvideo/model_training/special/fp8_training/validate.py Normal file
View File

@@ -0,0 +1,28 @@
import torch
from PIL import Image
from diffsynth.utils.data import save_video, VideoData
from diffsynth.pipelines.wan_video import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="diffusion_pytorch_model*.safetensors"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="Wan2.1_VAE.pth"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"),
],
)
pipe.load_lora(pipe.dit, "models/train/Wan2.1-I2V-14B-480P_lora_fp8/epoch-4.safetensors", alpha=1)
input_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=input_image,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-I2V-14B-480P.mp4", fps=15, quality=5)

38
examples/wanvideo/model_training/special/low_vram_training/Wan2.1-I2V-14B-480P.sh Normal file
View File

@@ -0,0 +1,38 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 1 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-480P_lora_lowvram_cache" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--task "sft:data_process" \
--offload_models "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors" \
--fp8_models "Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--use_gradient_checkpointing_offload
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path "./models/train/Wan2.1-I2V-14B-480P_lora_split_cache" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-480P_lora_lowvram" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--task "sft:train" \
--offload_models "Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--fp8_models "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors" \
--use_gradient_checkpointing_offload

28
examples/wanvideo/model_training/special/low_vram_training/validate.py Normal file
View File

@@ -0,0 +1,28 @@
import torch
from PIL import Image
from diffsynth.utils.data import save_video, VideoData
from diffsynth.pipelines.wan_video import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="diffusion_pytorch_model*.safetensors"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="Wan2.1_VAE.pth"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"),
],
)
pipe.load_lora(pipe.dit, "models/train/Wan2.1-I2V-14B-480P_lora_lowvram/epoch-4.safetensors", alpha=1)
input_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=input_image,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-I2V-14B-480P.mp4", fps=15, quality=5)

34
examples/wanvideo/model_training/special/split_training/Wan2.1-I2V-14B-480P.sh Normal file
View File

@@ -0,0 +1,34 @@
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path data/example_video_dataset \
--dataset_metadata_path data/example_video_dataset/metadata.csv \
--height 480 \
--width 832 \
--dataset_repeat 1 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-480P_lora_split_cache" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--task "sft:data_process" \
--offload_models "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors"
accelerate launch examples/wanvideo/model_training/train.py \
--dataset_base_path "./models/train/Wan2.1-I2V-14B-480P_lora_split_cache" \
--height 480 \
--width 832 \
--dataset_repeat 100 \
--model_id_with_origin_paths "Wan-AI/Wan2.1-I2V-14B-480P:diffusion_pytorch_model*.safetensors,Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth" \
--learning_rate 1e-4 \
--num_epochs 5 \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-480P_lora_split" \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--extra_inputs "input_image" \
--task "sft:train" \
--offload_models "Wan-AI/Wan2.1-I2V-14B-480P:models_t5_umt5-xxl-enc-bf16.pth,Wan-AI/Wan2.1-I2V-14B-480P:Wan2.1_VAE.pth,Wan-AI/Wan2.1-I2V-14B-480P:models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"

28
examples/wanvideo/model_training/special/split_training/validate.py Normal file
View File

@@ -0,0 +1,28 @@
import torch
from PIL import Image
from diffsynth.utils.data import save_video, VideoData
from diffsynth.pipelines.wan_video import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="diffusion_pytorch_model*.safetensors"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="Wan2.1_VAE.pth"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth"),
],
)
pipe.load_lora(pipe.dit, "models/train/Wan2.1-I2V-14B-480P_lora_split/epoch-4.safetensors", alpha=1)
input_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=input_image,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-I2V-14B-480P.mp4", fps=15, quality=5)

184
examples/wanvideo/model_training/train.py Normal file
View File

@@ -0,0 +1,184 @@
import torch, os, argparse, accelerate, warnings
from diffsynth.core import UnifiedDataset
from diffsynth.core.data.operators import LoadVideo, LoadAudio, ImageCropAndResize, ToAbsolutePath
from diffsynth.pipelines.wan_video import WanVideoPipeline, ModelConfig
from diffsynth.diffusion import *
os.environ["TOKENIZERS_PARALLELISM"] = "false"
class WanTrainingModule(DiffusionTrainingModule):
def __init__(
self,
model_paths=None, model_id_with_origin_paths=None,
tokenizer_path=None, audio_processor_path=None,
trainable_models=None,
lora_base_model=None, lora_target_modules="", lora_rank=32, lora_checkpoint=None,
preset_lora_path=None, preset_lora_model=None,
use_gradient_checkpointing=True,
use_gradient_checkpointing_offload=False,
extra_inputs=None,
fp8_models=None,
offload_models=None,
device="cpu",
task="sft",
max_timestep_boundary=1.0,
min_timestep_boundary=0.0,
):
super().__init__()
# Warning
if not use_gradient_checkpointing:
warnings.warn("Gradient checkpointing is detected as disabled. To prevent out-of-memory errors, the training framework will forcibly enable gradient checkpointing.")
use_gradient_checkpointing = True
# Load models
model_configs = self.parse_model_configs(model_paths, model_id_with_origin_paths, fp8_models=fp8_models, offload_models=offload_models, device=device)
tokenizer_config = ModelConfig(model_id="Wan-AI/Wan2.1-T2V-1.3B", origin_file_pattern="google/umt5-xxl/") if tokenizer_path is None else ModelConfig(tokenizer_path)
audio_processor_config = ModelConfig(model_id="Wan-AI/Wan2.2-S2V-14B", origin_file_pattern="wav2vec2-large-xlsr-53-english/") if audio_processor_path is None else ModelConfig(audio_processor_path)
self.pipe = WanVideoPipeline.from_pretrained(torch_dtype=torch.bfloat16, device=device, model_configs=model_configs, tokenizer_config=tokenizer_config, audio_processor_config=audio_processor_config)
self.pipe = self.split_pipeline_units(task, self.pipe, trainable_models, lora_base_model)
# Training mode
self.switch_pipe_to_training_mode(
self.pipe, trainable_models,
lora_base_model, lora_target_modules, lora_rank, lora_checkpoint,
preset_lora_path, preset_lora_model,
task=task,
)
# Store other configs
self.use_gradient_checkpointing = use_gradient_checkpointing
self.use_gradient_checkpointing_offload = use_gradient_checkpointing_offload
self.extra_inputs = extra_inputs.split(",") if extra_inputs is not None else []
self.fp8_models = fp8_models
self.task = task
self.task_to_loss = {
"sft:data_process": lambda pipe, *args: args,
"direct_distill:data_process": lambda pipe, *args: args,
"sft": lambda pipe, inputs_shared, inputs_posi, inputs_nega: FlowMatchSFTLoss(pipe, **inputs_shared, **inputs_posi),
"sft:train": lambda pipe, inputs_shared, inputs_posi, inputs_nega: FlowMatchSFTLoss(pipe, **inputs_shared, **inputs_posi),
"direct_distill": lambda pipe, inputs_shared, inputs_posi, inputs_nega: DirectDistillLoss(pipe, **inputs_shared, **inputs_posi),
"direct_distill:train": lambda pipe, inputs_shared, inputs_posi, inputs_nega: DirectDistillLoss(pipe, **inputs_shared, **inputs_posi),
}
self.max_timestep_boundary = max_timestep_boundary
self.min_timestep_boundary = min_timestep_boundary
def parse_extra_inputs(self, data, extra_inputs, inputs_shared):
for extra_input in extra_inputs:
if extra_input == "input_image":
inputs_shared["input_image"] = data["video"][0]
elif extra_input == "end_image":
inputs_shared["end_image"] = data["video"][-1]
elif extra_input == "reference_image" or extra_input == "vace_reference_image":
inputs_shared[extra_input] = data[extra_input][0]
else:
inputs_shared[extra_input] = data[extra_input]
return inputs_shared
def get_pipeline_inputs(self, data):
inputs_posi = {"prompt": data["prompt"]}
inputs_nega = {}
inputs_shared = {
# Assume you are using this pipeline for inference,
# please fill in the input parameters.
"input_video": data["video"],
"height": data["video"][0].size[1],
"width": data["video"][0].size[0],
"num_frames": len(data["video"]),
# Please do not modify the following parameters
# unless you clearly know what this will cause.
"cfg_scale": 1,
"tiled": False,
"rand_device": self.pipe.device,
"use_gradient_checkpointing": self.use_gradient_checkpointing,
"use_gradient_checkpointing_offload": self.use_gradient_checkpointing_offload,
"cfg_merge": False,
"vace_scale": 1,
"max_timestep_boundary": self.max_timestep_boundary,
"min_timestep_boundary": self.min_timestep_boundary,
}
inputs_shared = self.parse_extra_inputs(data, self.extra_inputs, inputs_shared)
return inputs_shared, inputs_posi, inputs_nega
def forward(self, data, inputs=None):
if inputs is None: inputs = self.get_pipeline_inputs(data)
inputs = self.transfer_data_to_device(inputs, self.pipe.device, self.pipe.torch_dtype)
for unit in self.pipe.units:
inputs = self.pipe.unit_runner(unit, self.pipe, *inputs)
loss = self.task_to_loss[self.task](self.pipe, *inputs)
return loss
def wan_parser():
parser = argparse.ArgumentParser(description="Simple example of a training script.")
parser = add_general_config(parser)
parser = add_video_size_config(parser)
parser.add_argument("--tokenizer_path", type=str, default=None, help="Path to tokenizer.")
parser.add_argument("--audio_processor_path", type=str, default=None, help="Path to the audio processor. If provided, the processor will be used for Wan2.2-S2V model.")
parser.add_argument("--max_timestep_boundary", type=float, default=1.0, help="Max timestep boundary (for mixed models, e.g., Wan-AI/Wan2.2-I2V-A14B).")
parser.add_argument("--min_timestep_boundary", type=float, default=0.0, help="Min timestep boundary (for mixed models, e.g., Wan-AI/Wan2.2-I2V-A14B).")
return parser
if __name__ == "__main__":
parser = wan_parser()
args = parser.parse_args()
accelerator = accelerate.Accelerator(
gradient_accumulation_steps=args.gradient_accumulation_steps,
kwargs_handlers=[accelerate.DistributedDataParallelKwargs(find_unused_parameters=args.find_unused_parameters)],
)
dataset = UnifiedDataset(
base_path=args.dataset_base_path,
metadata_path=args.dataset_metadata_path,
repeat=args.dataset_repeat,
data_file_keys=args.data_file_keys.split(","),
main_data_operator=UnifiedDataset.default_video_operator(
base_path=args.dataset_base_path,
max_pixels=args.max_pixels,
height=args.height,
width=args.width,
height_division_factor=16,
width_division_factor=16,
num_frames=args.num_frames,
time_division_factor=4,
time_division_remainder=1,
),
special_operator_map={
"animate_face_video": ToAbsolutePath(args.dataset_base_path) >> LoadVideo(args.num_frames, 4, 1, frame_processor=ImageCropAndResize(512, 512, None, 16, 16)),
"input_audio": ToAbsolutePath(args.dataset_base_path) >> LoadAudio(sr=16000),
}
)
model = WanTrainingModule(
model_paths=args.model_paths,
model_id_with_origin_paths=args.model_id_with_origin_paths,
tokenizer_path=args.tokenizer_path,
audio_processor_path=args.audio_processor_path,
trainable_models=args.trainable_models,
lora_base_model=args.lora_base_model,
lora_target_modules=args.lora_target_modules,
lora_rank=args.lora_rank,
lora_checkpoint=args.lora_checkpoint,
preset_lora_path=args.preset_lora_path,
preset_lora_model=args.preset_lora_model,
use_gradient_checkpointing=args.use_gradient_checkpointing,
use_gradient_checkpointing_offload=args.use_gradient_checkpointing_offload,
extra_inputs=args.extra_inputs,
fp8_models=args.fp8_models,
offload_models=args.offload_models,
task=args.task,
device=accelerator.device,
max_timestep_boundary=args.max_timestep_boundary,
min_timestep_boundary=args.min_timestep_boundary,
)
model_logger = ModelLogger(
args.output_path,
remove_prefix_in_ckpt=args.remove_prefix_in_ckpt,
)
launcher_map = {
"sft:data_process": launch_data_process_task,
"direct_distill:data_process": launch_data_process_task,
"sft": launch_training_task,
"sft:train": launch_training_task,
"direct_distill": launch_training_task,
"direct_distill:train": launch_training_task,
}
launcher_map[args.task](accelerator, dataset, model, model_logger, args=args)

25
examples/wanvideo/model_training/validate_full/LongCat-Video.py Normal file
View File

@@ -0,0 +1,25 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="meituan-longcat/LongCat-Video", origin_file_pattern="dit/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-14B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-14B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/LongCat-Video_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
seed=1, tiled=True
)
save_video(video, "video_LongCat-Video.mp4", fps=15, quality=5)

43
examples/wanvideo/model_training/validate_full/Video-As-Prompt-Wan2.1-14B.py Normal file
View File

@@ -0,0 +1,43 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="ByteDance/Video-As-Prompt-Wan2.1-14B", origin_file_pattern="transformer/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-720P", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-720P", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-720P", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Video-As-Prompt-Wan2.1-14B_full/epoch-1.safetensors")
pipe.vap.load_state_dict(state_dict)
pipe.enable_vram_management()
ref_video_path = 'data/example_video_dataset/wanvap/vap_ref.mp4'
target_image_path = 'data/example_video_dataset/wanvap/input_image.jpg'
image = Image.open(target_image_path).convert("RGB")
ref_video = VideoData(ref_video_path, height=480, width=832)
ref_frames = [ref_video[i] for i in range(49)]
vap_prompt = "A man stands with his back to the camera on a dirt path overlooking sun-drenched, rolling green tea plantations. He wears a blue and green plaid shirt, dark pants, and white shoes. As he turns to face the camera and spreads his arms, a brief, magical burst of sparkling golden light particles envelops him. Through this shimmer, he seamlessly transforms into a Labubu toy character. His head morphs into the iconic large, furry-eared head of the toy, featuring a wide grin with pointed teeth and red cheek markings. The character retains the man's original plaid shirt and clothing, which now fit its stylized, cartoonish body. The camera remains static throughout the transformation, positioned low among the tea bushes, maintaining a consistent view of the subject and the expansive scenery."
prompt = "A young woman with curly hair, wearing a green hijab and a floral dress, plays a violin in front of a vintage green car on a tree-lined street. She executes a swift counter-clockwise turn to face the camera. During the turn, a brilliant shower of golden, sparkling particles erupts and momentarily obscures her figure. As the particles fade, she is revealed to have seamlessly transformed into a Labubu toy character. This new figure, now with the toy's signature large ears, big eyes, and toothy grin, maintains the original pose and continues playing the violin. The character's clothing—the green hijab, floral dress, and black overcoat—remains identical to the woman's. Throughout this transition, the camera stays static, and the street-side environment remains completely consistent."
negative_prompt = "Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards"
video = pipe(
prompt=prompt,
negative_prompt=negative_prompt,
input_image=image,
seed=42, tiled=True,
height=480, width=832,
num_frames=49,
vap_video=ref_frames,
vap_prompt=vap_prompt,
negative_vap_prompt=negative_prompt,
)
save_video(video, "video_Video-As-Prompt-Wan2.1-14B.mp4", fps=15, quality=5)

28
examples/wanvideo/model_training/validate_full/Wan2.1-1.3b-speedcontrol-v1.py Normal file
View File

@@ -0,0 +1,28 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-1.3B", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-1.3B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-1.3B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="DiffSynth-Studio/Wan2.1-1.3b-speedcontrol-v1", origin_file_pattern="model.safetensors", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-1.3b-speedcontrol-v1_full/epoch-1.safetensors")
pipe.motion_controller.load_state_dict(state_dict)
pipe.enable_vram_management()
# Text-to-video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
seed=1, tiled=True,
motion_bucket_id=50
)
save_video(video, "video_Wan2.1-1.3b-speedcontrol-v1.mp4", fps=15, quality=5)

33
examples/wanvideo/model_training/validate_full/Wan2.1-FLF2V-14B-720P.py Normal file
View File

@@ -0,0 +1,33 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-FLF2V-14B-720P", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-FLF2V-14B-720P", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-FLF2V-14B-720P", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-FLF2V-14B-720P", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-FLF2V-14B-720P_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0],
end_image=video[80],
seed=0, tiled=True,
sigma_shift=16,
)
save_video(video, "video_Wan2.1-FLF2V-14B-720P.mp4", fps=15, quality=5)

32
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-1.3B-Control.py Normal file
View File

@@ -0,0 +1,32 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-1.3B-Control", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-1.3B-Control", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-1.3B-Control", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-1.3B-Control", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-1.3B-Control_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(81)]
# Control video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
control_video=video,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-Fun-1.3B-Control.mp4", fps=15, quality=5)

31
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-1.3B-InP.py Normal file
View File

@@ -0,0 +1,31 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-1.3B-InP", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-1.3B-InP", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-1.3B-InP", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-1.3B-InP", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-1.3B-InP_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0], end_image=video[80],
seed=0, tiled=True
)
save_video(video, "video_Wan2.1-Fun-1.3B-InP.mp4", fps=15, quality=5)

32
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-14B-Control.py Normal file
View File

@@ -0,0 +1,32 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-14B-Control", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-14B-Control", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-14B-Control", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-14B-Control", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-14B-Control_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(81)]
# Control video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
control_video=video,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-Fun-14B-Control.mp4", fps=15, quality=5)

31
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-14B-InP.py Normal file
View File

@@ -0,0 +1,31 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-14B-InP", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-14B-InP", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-14B-InP", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-14B-InP", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-14B-InP_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0], end_image=video[80],
seed=0, tiled=True
)
save_video(video, "video_Wan2.1-Fun-14B-InP.mp4", fps=15, quality=5)

32
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-V1.1-1.3B-Control-Camera.py Normal file
View File

@@ -0,0 +1,32 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-Control-Camera", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-V1.1-1.3B-Control-Camera_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0],
camera_control_direction="Left", camera_control_speed=0.0,
seed=0, tiled=True
)
save_video(video, "video_Wan2.1-Fun-V1.1-1.3B-Control-Camera.mp4", fps=15, quality=5)

33
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-V1.1-1.3B-Control.py Normal file
View File

@@ -0,0 +1,33 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-Control", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-Control", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-Control", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-Control", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-V1.1-1.3B-Control_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(81)]
reference_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
# Control video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
control_video=video, reference_image=reference_image,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-Fun-V1.1-1.3B-Control.mp4", fps=15, quality=5)

31
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-V1.1-1.3B-InP.py Normal file
View File

@@ -0,0 +1,31 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-InP", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-InP", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-InP", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-1.3B-InP", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-V1.1-1.3B-InP_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0], end_image=video[80],
seed=0, tiled=True
)
save_video(video, "video_Wan2.1-Fun-V1.1-1.3B-InP.mp4", fps=15, quality=5)

32
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-V1.1-14B-Control-Camera.py Normal file
View File

@@ -0,0 +1,32 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-Control-Camera", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-Control-Camera", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-Control-Camera", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-Control-Camera", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-V1.1-14B-Control-Camera_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0],
camera_control_direction="Left", camera_control_speed=0.0,
seed=0, tiled=True
)
save_video(video, "video_Wan2.1-Fun-V1.1-14B-Control-Camera.mp4", fps=15, quality=5)

33
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-V1.1-14B-Control.py Normal file
View File

@@ -0,0 +1,33 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-Control", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-Control", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-Control", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-Control", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-V1.1-14B-Control_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(81)]
reference_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
# Control video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
control_video=video, reference_image=reference_image,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-Fun-V1.1-14B-Control.mp4", fps=15, quality=5)

31
examples/wanvideo/model_training/validate_full/Wan2.1-Fun-V1.1-14B-InP.py Normal file
View File

@@ -0,0 +1,31 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-InP", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-InP", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-InP", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.1-Fun-V1.1-14B-InP", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-Fun-V1.1-14B-InP_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0], end_image=video[80],
seed=0, tiled=True
)
save_video(video, "video_Wan2.1-Fun-V1.1-14B-InP.mp4", fps=15, quality=5)

30
examples/wanvideo/model_training/validate_full/Wan2.1-I2V-14B-480P.py Normal file
View File

@@ -0,0 +1,30 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-480P", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-I2V-14B-480P_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
input_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=input_image,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-I2V-14B-480P.mp4", fps=15, quality=5)

31
examples/wanvideo/model_training/validate_full/Wan2.1-I2V-14B-720P.py Normal file
View File

@@ -0,0 +1,31 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-720P", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-720P", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-720P", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-I2V-14B-720P", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-I2V-14B-720P_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
input_image = VideoData("data/example_video_dataset/video1.mp4", height=720, width=1280)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=input_image,
height=720, width=1280, num_frames=49,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-I2V-14B-720P.mp4", fps=15, quality=5)

25
examples/wanvideo/model_training/validate_full/Wan2.1-T2V-1.3B.py Normal file
View File

@@ -0,0 +1,25 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-1.3B", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-1.3B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-1.3B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-T2V-1.3B_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-T2V-1.3B.mp4", fps=15, quality=5)

25
examples/wanvideo/model_training/validate_full/Wan2.1-T2V-14B.py Normal file
View File

@@ -0,0 +1,25 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-14B", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-14B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-T2V-14B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-T2V-14B_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-T2V-14B.mp4", fps=15, quality=5)

30
examples/wanvideo/model_training/validate_full/Wan2.1-VACE-1.3B-Preview.py Normal file
View File

@@ -0,0 +1,30 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="iic/VACE-Wan2.1-1.3B-Preview", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="iic/VACE-Wan2.1-1.3B-Preview", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="iic/VACE-Wan2.1-1.3B-Preview", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-VACE-1.3B-Preview_full/epoch-1.safetensors")
pipe.vace.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(49)]
reference_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
vace_video=video, vace_reference_image=reference_image, num_frames=49,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-VACE-1.3B-Preview.mp4", fps=15, quality=5)

30
examples/wanvideo/model_training/validate_full/Wan2.1-VACE-1.3B.py Normal file
View File

@@ -0,0 +1,30 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-VACE-1.3B", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-VACE-1.3B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-VACE-1.3B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-VACE-1.3B_full/epoch-1.safetensors")
pipe.vace.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(49)]
reference_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
vace_video=video, vace_reference_image=reference_image, num_frames=49,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-VACE-1.3B.mp4", fps=15, quality=5)

30
examples/wanvideo/model_training/validate_full/Wan2.1-VACE-14B.py Normal file
View File

@@ -0,0 +1,30 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.1-VACE-14B", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-VACE-14B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.1-VACE-14B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.1-VACE-14B_full/epoch-1.safetensors")
pipe.vace.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(17)]
reference_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
vace_video=video, vace_reference_image=reference_image, num_frames=17,
seed=1, tiled=True
)
save_video(video, "video_Wan2.1-VACE-14B.mp4", fps=15, quality=5)

33
examples/wanvideo/model_training/validate_full/Wan2.2-Animate-14B.py Normal file
View File

@@ -0,0 +1,33 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.2-Animate-14B", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-Animate-14B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-Animate-14B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-Animate-14B", origin_file_pattern="models_clip_open-clip-xlm-roberta-large-vit-huge-14.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.2-Animate-14B_full/epoch-1.safetensors")
pipe.animate_adapter.load_state_dict(state_dict, strict=False)
pipe.enable_vram_management()
input_image = VideoData("data/example_video_dataset/animate/animate_output.mp4", height=480, width=832)[0]
animate_pose_video = VideoData("data/examples/wan/animate/animate_pose_video.mp4", height=480, width=832).raw_data()[:81-4]
animate_face_video = VideoData("data/examples/wan/animate/animate_face_video.mp4", height=512, width=512).raw_data()[:81-4]
video = pipe(
prompt="视频中的人在做动作",
seed=0, tiled=True,
input_image=input_image,
animate_pose_video=animate_pose_video,
animate_face_video=animate_face_video,
num_frames=81, height=480, width=832,
num_inference_steps=20, cfg_scale=1,
)
save_video(video, "video_Wan2.2-Animate-14B.mp4", fps=15, quality=5)

34
examples/wanvideo/model_training/validate_full/Wan2.2-Fun-A14B-Control-Camera.py Normal file
View File

@@ -0,0 +1,34 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-Control-Camera", origin_file_pattern="high_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-Control-Camera", origin_file_pattern="low_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-Control-Camera", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-Control-Camera", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.2-Fun-A14B-Control-Camera_high_noise_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
state_dict = load_state_dict("models/train/Wan2.2-Fun-A14B-Control-Camera_low_noise_full/epoch-1.safetensors")
pipe.dit2.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0],
camera_control_direction="Left", camera_control_speed=0.0,
seed=0, tiled=True
)
save_video(video, "video_Wan2.2-Fun-A14B-Control-Camera.mp4", fps=15, quality=5)

35
examples/wanvideo/model_training/validate_full/Wan2.2-Fun-A14B-Control.py Normal file
View File

@@ -0,0 +1,35 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-Control", origin_file_pattern="high_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-Control", origin_file_pattern="low_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-Control", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-Control", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.2-Fun-A14B-Control_high_noise_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
state_dict = load_state_dict("models/train/Wan2.2-Fun-A14B-Control_low_noise_full/epoch-1.safetensors")
pipe.dit2.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(81)]
reference_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
# Control video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
control_video=video, reference_image=reference_image,
seed=1, tiled=True
)
save_video(video, "video_Wan2.2-Fun-A14B-Control.mp4", fps=15, quality=5)

32
examples/wanvideo/model_training/validate_full/Wan2.2-Fun-A14B-InP.py Normal file
View File

@@ -0,0 +1,32 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-InP", origin_file_pattern="high_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-InP", origin_file_pattern="low_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-InP", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-Fun-A14B-InP", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.2-Fun-A14B-InP_high_noise_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
state_dict = load_state_dict("models/train/Wan2.2-Fun-A14B-InP_low_noise_full/epoch-1.safetensors")
pipe.dit2.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)
# First and last frame to video
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=video[0], end_image=video[80],
seed=0, tiled=True
)
save_video(video, "video_Wan2.2-Fun-A14B-InP.mp4", fps=15, quality=5)

33
examples/wanvideo/model_training/validate_full/Wan2.2-I2V-A14B.py Normal file
View File

@@ -0,0 +1,33 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.2-I2V-A14B", origin_file_pattern="high_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-I2V-A14B", origin_file_pattern="low_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-I2V-A14B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-I2V-A14B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.2-I2V-A14B_high_noise_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
state_dict = load_state_dict("models/train/Wan2.2-I2V-A14B_low_noise_full/epoch-1.safetensors")
pipe.dit2.load_state_dict(state_dict)
pipe.enable_vram_management()
input_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=input_image,
num_frames=49,
seed=1, tiled=True,
)
save_video(video, "video_Wan2.2-I2V-A14B.mp4", fps=15, quality=5)

53
examples/wanvideo/model_training/validate_full/Wan2.2-S2V-14B.py Normal file
View File

@@ -0,0 +1,53 @@
import torch
from PIL import Image
import librosa
from diffsynth import VideoData, save_video_with_audio, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.2-S2V-14B", origin_file_pattern="diffusion_pytorch_model*.safetensors"),
ModelConfig(model_id="Wan-AI/Wan2.2-S2V-14B", origin_file_pattern="wav2vec2-large-xlsr-53-english/model.safetensors"),
ModelConfig(model_id="Wan-AI/Wan2.2-S2V-14B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth"),
ModelConfig(model_id="Wan-AI/Wan2.2-S2V-14B", origin_file_pattern="Wan2.1_VAE.pth"),
],
audio_processor_config=ModelConfig(model_id="Wan-AI/Wan2.2-S2V-14B", origin_file_pattern="wav2vec2-large-xlsr-53-english/"),
)
state_dict = load_state_dict("models/train/Wan2.2-S2V-14B_full/epoch-0.safetensors")
pipe.dit.load_state_dict(state_dict, strict=False)
pipe.enable_vram_management()
num_frames = 81 # 4n+1
height = 448
width = 832
prompt = "a person is singing"
negative_prompt = "画面模糊，最差质量，画面模糊，细节模糊不清，情绪激动剧烈，手快速抖动，字幕，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走"
input_image = Image.open("data/example_video_dataset/wans2v/pose.png").convert("RGB").resize((width, height))
# s2v audio input, recommend 16kHz sampling rate
audio_path = 'data/example_video_dataset/wans2v/sing.MP3'
input_audio, sample_rate = librosa.load(audio_path, sr=16000)
# S2V pose video input
pose_video_path = 'data/example_video_dataset/wans2v/pose.mp4'
pose_video = VideoData(pose_video_path, height=height, width=width)
# Speech-to-video with pose
video = pipe(
prompt=prompt,
input_image=input_image,
negative_prompt=negative_prompt,
seed=0,
num_frames=num_frames,
height=height,
width=width,
audio_sample_rate=sample_rate,
input_audio=input_audio,
s2v_pose_video=pose_video,
num_inference_steps=40,
)
save_video_with_audio(video[1:], "video_pose_with_audio.mp4", audio_path, fps=16, quality=5)

28
examples/wanvideo/model_training/validate_full/Wan2.2-T2V-A14B.py Normal file
View File

@@ -0,0 +1,28 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.2-T2V-A14B", origin_file_pattern="high_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-T2V-A14B", origin_file_pattern="low_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-T2V-A14B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-T2V-A14B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.2-T2V-A14B_high_noise_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
state_dict = load_state_dict("models/train/Wan2.2-T2V-A14B_low_noise_full/epoch-1.safetensors")
pipe.dit2.load_state_dict(state_dict)
pipe.enable_vram_management()
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
seed=1, tiled=True
)
save_video(video, "video_Wan2.2-T2V-A14B.mp4", fps=15, quality=5)

30
examples/wanvideo/model_training/validate_full/Wan2.2-TI2V-5B.py Normal file
View File

@@ -0,0 +1,30 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
from modelscope import dataset_snapshot_download
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="Wan-AI/Wan2.2-TI2V-5B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-TI2V-5B", origin_file_pattern="diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="Wan-AI/Wan2.2-TI2V-5B", origin_file_pattern="Wan2.2_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.2-TI2V-5B_full/epoch-1.safetensors")
pipe.dit.load_state_dict(state_dict)
pipe.enable_vram_management()
input_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
input_image=input_image,
num_frames=49,
seed=1, tiled=True,
)
save_video(video, "video_Wan2.2-TI2V-5B.mp4", fps=15, quality=5)

33
examples/wanvideo/model_training/validate_full/Wan2.2-VACE-Fun-A14B.py Normal file
View File

@@ -0,0 +1,33 @@
import torch
from PIL import Image
from diffsynth import save_video, VideoData, load_state_dict
from diffsynth.pipelines.wan_video_new import WanVideoPipeline, ModelConfig
pipe = WanVideoPipeline.from_pretrained(
torch_dtype=torch.bfloat16,
device="cuda",
model_configs=[
ModelConfig(model_id="PAI/Wan2.2-VACE-Fun-A14B", origin_file_pattern="high_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-VACE-Fun-A14B", origin_file_pattern="low_noise_model/diffusion_pytorch_model*.safetensors", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-VACE-Fun-A14B", origin_file_pattern="models_t5_umt5-xxl-enc-bf16.pth", offload_device="cpu"),
ModelConfig(model_id="PAI/Wan2.2-VACE-Fun-A14B", origin_file_pattern="Wan2.1_VAE.pth", offload_device="cpu"),
],
)
state_dict = load_state_dict("models/train/Wan2.2-VACE-Fun-A14B_high_noise_full/epoch-1.safetensors")
pipe.vace.load_state_dict(state_dict)
state_dict = load_state_dict("models/train/Wan2.2-VACE-Fun-A14B_low_noise_full/epoch-1.safetensors")
pipe.vace2.load_state_dict(state_dict)
pipe.enable_vram_management()
video = VideoData("data/example_video_dataset/video1_softedge.mp4", height=480, width=832)
video = [video[i] for i in range(17)]
reference_image = VideoData("data/example_video_dataset/video1.mp4", height=480, width=832)[0]
video = pipe(
prompt="from sunset to night, a small town, light, house, river",
negative_prompt="色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
vace_video=video, vace_reference_image=reference_image, num_frames=17,
seed=1, tiled=True
)
save_video(video, "video_Wan2.2-VACE-A14B.mp4", fps=15, quality=5)

Some files were not shown because too many files have changed in this diff Show More