From 62c3d406d920e7a7e6778b67059299eb90011d41 Mon Sep 17 00:00:00 2001 From: feng0w0 Date: Mon, 5 Jan 2026 15:42:55 +0800 Subject: [PATCH] Docs:Supplement NPU training script samples and documentation instruction --- .../FLUX.1-Kontext-dev-NPU.sh} | 13 +++++-------- .../FLUX.1-dev-NPU.sh} | 15 ++++++--------- .../Qwen-Image-Edit-2509-LoRA-NPU.sh | 0 .../Qwen-Image-LoRA-NPU.sh | 0 .../Wan2.1-T2V-14B-NPU.sh | 0 .../Wan2.2-T2V-A14B-NPU.sh | 0 .../Wan2.2-VACE-Fun-A14B-NPU.sh | 0 .../Z-Image-Turbo-NPU.sh | 0 8 files changed, 11 insertions(+), 17 deletions(-) rename examples/flux/model_training/special/{npu_scripts/FLUX.1-Kontext-dev-LoRA-NPU.sh => npu_training/FLUX.1-Kontext-dev-NPU.sh} (63%) rename examples/flux/model_training/special/{npu_scripts/FLUX.1-dev-LoRA-NPU.sh => npu_training/FLUX.1-dev-NPU.sh} (57%) rename examples/qwen_image/model_training/special/{npu_scripts => npu_training}/Qwen-Image-Edit-2509-LoRA-NPU.sh (100%) rename examples/qwen_image/model_training/special/{npu_scripts => npu_training}/Qwen-Image-LoRA-NPU.sh (100%) rename examples/wanvideo/model_training/special/{npu_scripts => npu_training}/Wan2.1-T2V-14B-NPU.sh (100%) rename examples/wanvideo/model_training/special/{npu_scripts => npu_training}/Wan2.2-T2V-A14B-NPU.sh (100%) rename examples/wanvideo/model_training/special/{npu_scripts => npu_training}/Wan2.2-VACE-Fun-A14B-NPU.sh (100%) rename examples/z_image/model_training/special/{npu_scripts => npu_training}/Z-Image-Turbo-NPU.sh (100%) diff --git a/examples/flux/model_training/special/npu_scripts/FLUX.1-Kontext-dev-LoRA-NPU.sh b/examples/flux/model_training/special/npu_training/FLUX.1-Kontext-dev-NPU.sh similarity index 63% rename from examples/flux/model_training/special/npu_scripts/FLUX.1-Kontext-dev-LoRA-NPU.sh rename to examples/flux/model_training/special/npu_training/FLUX.1-Kontext-dev-NPU.sh index 51397e6..7ec976d 100644 --- a/examples/flux/model_training/special/npu_scripts/FLUX.1-Kontext-dev-LoRA-NPU.sh +++ b/examples/flux/model_training/special/npu_training/FLUX.1-Kontext-dev-NPU.sh @@ -1,20 +1,17 @@ export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True export CPU_AFFINITY_CONF=1 -accelerate launch examples/flux/model_training/train.py \ +accelerate launch --config_file examples/flux/model_training/full/accelerate_config_zero2offload.yaml examples/flux/model_training/train.py \ --dataset_base_path data/example_image_dataset \ --dataset_metadata_path data/example_image_dataset/metadata_kontext.csv \ --data_file_keys "image,kontext_images" \ --max_pixels 1048576 \ --dataset_repeat 400 \ --model_id_with_origin_paths "black-forest-labs/FLUX.1-Kontext-dev:flux1-kontext-dev.safetensors,black-forest-labs/FLUX.1-dev:text_encoder/model.safetensors,black-forest-labs/FLUX.1-dev:text_encoder_2/*.safetensors,black-forest-labs/FLUX.1-dev:ae.safetensors" \ - --learning_rate 1e-4 \ - --num_epochs 5 \ + --learning_rate 1e-5 \ + --num_epochs 1 \ --remove_prefix_in_ckpt "pipe.dit." \ - --output_path "./models/train/FLUX.1-Kontext-dev_lora" \ - --lora_base_model "dit" \ - --lora_target_modules "a_to_qkv,b_to_qkv,ff_a.0,ff_a.2,ff_b.0,ff_b.2,a_to_out,b_to_out,proj_out,norm.linear,norm1_a.linear,norm1_b.linear,to_qkv_mlp" \ - --lora_rank 32 \ - --align_to_opensource_format \ + --output_path "./models/train/FLUX.1-Kontext-dev_full" \ + --trainable_models "dit" \ --extra_inputs "kontext_images" \ --use_gradient_checkpointing diff --git a/examples/flux/model_training/special/npu_scripts/FLUX.1-dev-LoRA-NPU.sh b/examples/flux/model_training/special/npu_training/FLUX.1-dev-NPU.sh similarity index 57% rename from examples/flux/model_training/special/npu_scripts/FLUX.1-dev-LoRA-NPU.sh rename to examples/flux/model_training/special/npu_training/FLUX.1-dev-NPU.sh index c167503..8133594 100644 --- a/examples/flux/model_training/special/npu_scripts/FLUX.1-dev-LoRA-NPU.sh +++ b/examples/flux/model_training/special/npu_training/FLUX.1-dev-NPU.sh @@ -1,18 +1,15 @@ export PYTORCH_NPU_ALLOC_CONF=expandable_segments:True export CPU_AFFINITY_CONF=1 -accelerate launch examples/flux/model_training/train.py \ +accelerate launch --config_file examples/flux/model_training/full/accelerate_config_zero2offload.yaml examples/flux/model_training/train.py \ --dataset_base_path data/example_image_dataset \ --dataset_metadata_path data/example_image_dataset/metadata.csv \ --max_pixels 1048576 \ - --dataset_repeat 50 \ + --dataset_repeat 400 \ --model_id_with_origin_paths "black-forest-labs/FLUX.1-dev:flux1-dev.safetensors,black-forest-labs/FLUX.1-dev:text_encoder/model.safetensors,black-forest-labs/FLUX.1-dev:text_encoder_2/*.safetensors,black-forest-labs/FLUX.1-dev:ae.safetensors" \ - --learning_rate 1e-4 \ - --num_epochs 5 \ + --learning_rate 1e-5 \ + --num_epochs 1 \ --remove_prefix_in_ckpt "pipe.dit." \ - --output_path "./models/train/FLUX.1-dev_lora" \ - --lora_base_model "dit" \ - --lora_target_modules "a_to_qkv,b_to_qkv,ff_a.0,ff_a.2,ff_b.0,ff_b.2,a_to_out,b_to_out,proj_out,norm.linear,norm1_a.linear,norm1_b.linear,to_qkv_mlp" \ - --lora_rank 32 \ - --align_to_opensource_format \ + --output_path "./models/train/FLUX.1-dev_full" \ + --trainable_models "dit" \ --use_gradient_checkpointing diff --git a/examples/qwen_image/model_training/special/npu_scripts/Qwen-Image-Edit-2509-LoRA-NPU.sh b/examples/qwen_image/model_training/special/npu_training/Qwen-Image-Edit-2509-LoRA-NPU.sh similarity index 100% rename from examples/qwen_image/model_training/special/npu_scripts/Qwen-Image-Edit-2509-LoRA-NPU.sh rename to examples/qwen_image/model_training/special/npu_training/Qwen-Image-Edit-2509-LoRA-NPU.sh diff --git a/examples/qwen_image/model_training/special/npu_scripts/Qwen-Image-LoRA-NPU.sh b/examples/qwen_image/model_training/special/npu_training/Qwen-Image-LoRA-NPU.sh similarity index 100% rename from examples/qwen_image/model_training/special/npu_scripts/Qwen-Image-LoRA-NPU.sh rename to examples/qwen_image/model_training/special/npu_training/Qwen-Image-LoRA-NPU.sh diff --git a/examples/wanvideo/model_training/special/npu_scripts/Wan2.1-T2V-14B-NPU.sh b/examples/wanvideo/model_training/special/npu_training/Wan2.1-T2V-14B-NPU.sh similarity index 100% rename from examples/wanvideo/model_training/special/npu_scripts/Wan2.1-T2V-14B-NPU.sh rename to examples/wanvideo/model_training/special/npu_training/Wan2.1-T2V-14B-NPU.sh diff --git a/examples/wanvideo/model_training/special/npu_scripts/Wan2.2-T2V-A14B-NPU.sh b/examples/wanvideo/model_training/special/npu_training/Wan2.2-T2V-A14B-NPU.sh similarity index 100% rename from examples/wanvideo/model_training/special/npu_scripts/Wan2.2-T2V-A14B-NPU.sh rename to examples/wanvideo/model_training/special/npu_training/Wan2.2-T2V-A14B-NPU.sh diff --git a/examples/wanvideo/model_training/special/npu_scripts/Wan2.2-VACE-Fun-A14B-NPU.sh b/examples/wanvideo/model_training/special/npu_training/Wan2.2-VACE-Fun-A14B-NPU.sh similarity index 100% rename from examples/wanvideo/model_training/special/npu_scripts/Wan2.2-VACE-Fun-A14B-NPU.sh rename to examples/wanvideo/model_training/special/npu_training/Wan2.2-VACE-Fun-A14B-NPU.sh diff --git a/examples/z_image/model_training/special/npu_scripts/Z-Image-Turbo-NPU.sh b/examples/z_image/model_training/special/npu_training/Z-Image-Turbo-NPU.sh similarity index 100% rename from examples/z_image/model_training/special/npu_scripts/Z-Image-Turbo-NPU.sh rename to examples/z_image/model_training/special/npu_training/Z-Image-Turbo-NPU.sh