From a18e6233b5236ebcd91333e90f5f1d16cc5b9381 Mon Sep 17 00:00:00 2001 From: Artiprocher Date: Mon, 19 Jan 2026 17:35:08 +0800 Subject: [PATCH] updata wan-vace training scripts --- .../wanvideo/model_training/lora/Wan2.1-VACE-1.3B-Preview.sh | 5 +++-- examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B.sh | 5 +++-- examples/wanvideo/model_training/lora/Wan2.1-VACE-14B.sh | 5 +++-- 3 files changed, 9 insertions(+), 6 deletions(-) diff --git a/examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B-Preview.sh b/examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B-Preview.sh index 2bcb55b..1f25eef 100644 --- a/examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B-Preview.sh +++ b/examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B-Preview.sh @@ -6,7 +6,7 @@ accelerate launch examples/wanvideo/model_training/train.py \ --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 \ + --learning_rate 5e-5 \ --num_epochs 5 \ --remove_prefix_in_ckpt "pipe.vace." \ --output_path "./models/train/Wan2.1-VACE-1.3B-Preview_lora" \ @@ -14,4 +14,5 @@ accelerate launch examples/wanvideo/model_training/train.py \ --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 \ No newline at end of file + --use_gradient_checkpointing_offload +# The learning rate is kept consistent with the settings in the original paper \ No newline at end of file diff --git a/examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B.sh b/examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B.sh index b565078..c8b77cc 100644 --- a/examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B.sh +++ b/examples/wanvideo/model_training/lora/Wan2.1-VACE-1.3B.sh @@ -6,7 +6,7 @@ accelerate launch examples/wanvideo/model_training/train.py \ --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 \ + --learning_rate 5e-5 \ --num_epochs 5 \ --remove_prefix_in_ckpt "pipe.vace." \ --output_path "./models/train/Wan2.1-VACE-1.3B_lora" \ @@ -14,4 +14,5 @@ accelerate launch examples/wanvideo/model_training/train.py \ --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 \ No newline at end of file + --use_gradient_checkpointing_offload +# The learning rate is kept consistent with the settings in the original paper \ No newline at end of file diff --git a/examples/wanvideo/model_training/lora/Wan2.1-VACE-14B.sh b/examples/wanvideo/model_training/lora/Wan2.1-VACE-14B.sh index 633ea0e..28bd05c 100644 --- a/examples/wanvideo/model_training/lora/Wan2.1-VACE-14B.sh +++ b/examples/wanvideo/model_training/lora/Wan2.1-VACE-14B.sh @@ -7,7 +7,7 @@ accelerate launch examples/wanvideo/model_training/train.py \ --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 \ + --learning_rate 5e-5 \ --num_epochs 5 \ --remove_prefix_in_ckpt "pipe.vace." \ --output_path "./models/train/Wan2.1-VACE-14B_lora" \ @@ -15,4 +15,5 @@ accelerate launch examples/wanvideo/model_training/train.py \ --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 \ No newline at end of file + --use_gradient_checkpointing_offload +# The learning rate is kept consistent with the settings in the original paper \ No newline at end of file