theluyuan/DiffSynth-Studio
1
0
Fork 0
mirror of https://github.com/modelscope/DiffSynth-Studio.git synced 2026-03-24 18:28:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

wan-refactor

Browse Source
This commit is contained in:
Artiprocher
2025-06-13 13:04:35 +08:00
parent 7e6a3c7897
commit 8dd24169cc
47 changed files with 810 additions and 310 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/wanvideo/model_training/full/Wan2.1-1.3b-speedcontrol-v1.sh
View File

@@ -10,4 +10,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--remove_prefix_in_ckpt "pipe.motion_controller." \
--output_path "./models/train/Wan2.1-1.3b-speedcontrol-v1_full" \
--trainable_models "motion_controller" \
--input_contains_motion_bucket_id
--extra_inputs "motion_bucket_id"

3
examples/wanvideo/model_training/full/Wan2.1-FLF2V-14B-720P.sh
View File

@@ -10,5 +10,4 @@ accelerate launch --config_file examples/wanvideo/model_training/full/accelerate
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-FLF2V-14B-720P_full" \
--trainable_models "dit" \
--input_contains_input_image \
--input_contains_end_image
--extra_inputs "input_image,end_image"

2
examples/wanvideo/model_training/full/Wan2.1-Fun-1.3B-Control.sh
View File

@@ -11,4 +11,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-1.3B-Control_full" \
--trainable_models "dit" \
--input_contains_control_video
--extra_inputs "control_video"

3
examples/wanvideo/model_training/full/Wan2.1-Fun-1.3B-InP.sh
View File

@@ -10,5 +10,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-1.3B-InP_full" \
--trainable_models "dit" \
--input_contains_input_image \
--input_contains_end_image
--extra_inputs "input_image,end_image"

2
examples/wanvideo/model_training/full/Wan2.1-Fun-14B-Control.sh
View File

@@ -11,4 +11,4 @@ accelerate launch --config_file examples/wanvideo/model_training/full/accelerate
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-14B-Control_full" \
--trainable_models "dit" \
--input_contains_control_video
--extra_inputs "control_video"

3
examples/wanvideo/model_training/full/Wan2.1-Fun-14B-InP.sh
View File

@@ -10,5 +10,4 @@ accelerate launch --config_file examples/wanvideo/model_training/full/accelerate
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-14B-InP_full" \
--trainable_models "dit" \
--input_contains_input_image \
--input_contains_end_image
--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"

3
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-1.3B-Control.sh
View File

@@ -11,5 +11,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-1.3B-Control_full" \
--trainable_models "dit" \
--input_contains_control_video \
--input_contains_reference_image
--extra_inputs "control_video,reference_image"

3
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-1.3B-InP.sh
View File

@@ -10,5 +10,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-1.3B-InP_full" \
--trainable_models "dit" \
--input_contains_input_image \
--input_contains_end_image
--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"

3
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-14B-Control.sh
View File

@@ -11,5 +11,4 @@ accelerate launch --config_file examples/wanvideo/model_training/full/accelerate
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-14B-Control_full" \
--trainable_models "dit" \
--input_contains_control_video \
--input_contains_reference_image
--extra_inputs "control_video,reference_image"

3
examples/wanvideo/model_training/full/Wan2.1-Fun-V1.1-14B-InP.sh
View File

@@ -10,5 +10,4 @@ accelerate launch --config_file examples/wanvideo/model_training/full/accelerate
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-Fun-V1.1-14B-InP_full" \
--trainable_models "dit" \
--input_contains_input_image \
--input_contains_end_image
--extra_inputs "input_image,end_image"

2
examples/wanvideo/model_training/full/Wan2.1-I2V-14B-480P.sh
View File

@@ -10,4 +10,4 @@ accelerate launch --config_file examples/wanvideo/model_training/full/accelerate
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-480P_full" \
--trainable_models "dit" \
--input_contains_input_image
--extra_inputs "input_image"

2
examples/wanvideo/model_training/full/Wan2.1-I2V-14B-720P.sh
View File

@@ -10,4 +10,4 @@ accelerate launch --config_file examples/wanvideo/model_training/full/accelerate
--remove_prefix_in_ckpt "pipe.dit." \
--output_path "./models/train/Wan2.1-I2V-14B-720P_full" \
--trainable_models "dit" \
--input_contains_input_image
--extra_inputs "input_image"

3
examples/wanvideo/model_training/full/Wan2.1-VACE-1.3B-Preview.sh
View File

@@ -12,6 +12,5 @@ accelerate launch examples/wanvideo/model_training/train.py \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-1.3B-Preview_full" \
--trainable_models "vace" \
--input_contains_vace_video \
--input_contains_vace_reference_image \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

3
examples/wanvideo/model_training/full/Wan2.1-VACE-1.3B.sh
View File

@@ -12,6 +12,5 @@ accelerate launch examples/wanvideo/model_training/train.py \
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-1.3B_full" \
--trainable_models "vace" \
--input_contains_vace_video \
--input_contains_vace_reference_image \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

3
examples/wanvideo/model_training/full/Wan2.1-VACE-14B.sh
View File

@@ -12,6 +12,5 @@ accelerate launch --config_file examples/wanvideo/model_training/full/accelerate
--remove_prefix_in_ckpt "pipe.vace." \
--output_path "./models/train/Wan2.1-VACE-14B_full" \
--trainable_models "vace" \
--input_contains_vace_video \
--input_contains_vace_reference_image \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

2
examples/wanvideo/model_training/lora/Wan2.1-1.3b-speedcontrol-v1.sh
View File

@@ -12,4 +12,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_motion_bucket_id
--extra_inputs "motion_bucket_id"

3
examples/wanvideo/model_training/lora/Wan2.1-FLF2V-14B-720P.sh
View File

@@ -12,5 +12,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_input_image \
--input_contains_end_image
--extra_inputs "input_image,end_image"

2
examples/wanvideo/model_training/lora/Wan2.1-Fun-1.3B-Control.sh
View File

@@ -13,4 +13,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_control_video
--extra_inputs "control_video"

3
examples/wanvideo/model_training/lora/Wan2.1-Fun-1.3B-InP.sh
View File

@@ -12,5 +12,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_input_image \
--input_contains_end_image
--extra_inputs "input_image,end_image"

2
examples/wanvideo/model_training/lora/Wan2.1-Fun-14B-Control.sh
View File

@@ -13,4 +13,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_control_video
--extra_inputs "control_video"

3
examples/wanvideo/model_training/lora/Wan2.1-Fun-14B-InP.sh
View File

@@ -12,5 +12,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_input_image \
--input_contains_end_image
--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_full" \
--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"

3
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-1.3B-Control.sh
View File

@@ -13,5 +13,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_control_video \
--input_contains_reference_image
--extra_inputs "control_video,reference_image"

3
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-1.3B-InP.sh
View File

@@ -12,5 +12,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_input_image \
--input_contains_end_image
--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_full" \
--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"

3
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-14B-Control.sh
View File

@@ -13,5 +13,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_control_video \
--input_contains_reference_image
--extra_inputs "control_video,reference_image"

3
examples/wanvideo/model_training/lora/Wan2.1-Fun-V1.1-14B-InP.sh
View File

@@ -12,5 +12,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_input_image \
--input_contains_end_image
--extra_inputs "input_image,end_image"

2
examples/wanvideo/model_training/lora/Wan2.1-I2V-14B-480P.sh
View File

@@ -12,4 +12,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_input_image
--extra_inputs "input_image"

2
examples/wanvideo/model_training/lora/Wan2.1-I2V-14B-720P.sh
View File

@@ -12,4 +12,4 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "dit" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_input_image
--extra_inputs "input_image"

3
examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B-Preview.sh
View File

@@ -13,6 +13,5 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "vace" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_vace_video \
--input_contains_vace_reference_image \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

3
examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B.sh
View File

@@ -13,6 +13,5 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "vace" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_vace_video \
--input_contains_vace_reference_image \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

3
examples/wanvideo/model_training/lora/Wan2.1-VACE-14B.sh
View File

@@ -14,6 +14,5 @@ accelerate launch examples/wanvideo/model_training/train.py \
--lora_base_model "vace" \
--lora_target_modules "q,k,v,o,ffn.0,ffn.2" \
--lora_rank 32 \
--input_contains_vace_video \
--input_contains_vace_reference_image \
--extra_inputs "vace_video,vace_reference_image" \
--use_gradient_checkpointing_offload

39
examples/wanvideo/model_training/train.py
View File

@@ -13,14 +13,7 @@ class WanTrainingModule(DiffusionTrainingModule):
lora_base_model=None, lora_target_modules="q,k,v,o,ffn.0,ffn.2", lora_rank=32,
use_gradient_checkpointing=True,
use_gradient_checkpointing_offload=False,
# Extra inputs
input_contains_input_image=False,
input_contains_end_image=False,
input_contains_control_video=False,
input_contains_reference_image=False,
input_contains_vace_video=False,
input_contains_vace_reference_image=False,
input_contains_motion_bucket_id=False,
extra_inputs=None,
):
super().__init__()
# Load models
@@ -51,13 +44,7 @@ class WanTrainingModule(DiffusionTrainingModule):
# Store other configs
self.use_gradient_checkpointing = use_gradient_checkpointing
self.use_gradient_checkpointing_offload = use_gradient_checkpointing_offload
self.input_contains_input_image = input_contains_input_image
self.input_contains_end_image = input_contains_end_image
self.input_contains_control_video = input_contains_control_video
self.input_contains_reference_image = input_contains_reference_image
self.input_contains_vace_video = input_contains_vace_video
self.input_contains_vace_reference_image = input_contains_vace_reference_image
self.input_contains_motion_bucket_id = input_contains_motion_bucket_id
self.extra_inputs = extra_inputs.split(",") if extra_inputs is not None else []
def forward_preprocess(self, data):
@@ -85,13 +72,13 @@ class WanTrainingModule(DiffusionTrainingModule):
}
# Extra inputs
if self.input_contains_input_image: inputs_shared["input_image"] = data["video"][0]
if self.input_contains_end_image: inputs_shared["end_image"] = data["video"][-1]
if self.input_contains_control_video: inputs_shared["control_video"] = data["control_video"]
if self.input_contains_reference_image: inputs_shared["reference_image"] = data["reference_image"]
if self.input_contains_vace_video: inputs_shared["vace_video"] = data["vace_video"]
if self.input_contains_vace_reference_image: inputs_shared["vace_reference_image"] = data["vace_reference_image"]
if self.input_contains_motion_bucket_id: inputs_shared["motion_bucket_id"] = data["motion_bucket_id"]
for extra_input in self.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]
else:
inputs_shared[extra_input] = data[extra_input]
# Pipeline units will automatically process the input parameters.
for unit in self.pipe.units:
@@ -118,12 +105,6 @@ if __name__ == "__main__":
lora_target_modules=args.lora_target_modules,
lora_rank=args.lora_rank,
use_gradient_checkpointing_offload=args.use_gradient_checkpointing_offload,
input_contains_input_image=args.input_contains_input_image,
input_contains_end_image=args.input_contains_end_image,
input_contains_control_video=args.input_contains_control_video,
input_contains_reference_image=args.input_contains_reference_image,
input_contains_vace_video=args.input_contains_vace_video,
input_contains_vace_reference_image=args.input_contains_vace_reference_image,
input_contains_motion_bucket_id=args.input_contains_motion_bucket_id,
extra_inputs=args.extra_inputs,
)
launch_training_task(model, dataset, args=args)

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)

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)

31
examples/wanvideo/model_training/validate_lora/Wan2.1-Fun-V1.1-1.3B-Control-Camera.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-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"),
],
)
pipe.load_lora(pipe.dit, "models/train/Wan2.1-Fun-V1.1-1.3B-Control-Camera_lora/epoch-4.safetensors", alpha=1)
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)

31
examples/wanvideo/model_training/validate_lora/Wan2.1-Fun-V1.1-14B-Control-Camera.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-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"),
],
)
pipe.load_lora(pipe.dit, "models/train/Wan2.1-Fun-V1.1-1.3B-Control-Camera_lora/epoch-4.safetensors", alpha=1)
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)