| # Please run `accelerate config` to configure GPU, DeepSpeed, etc. | |
| modelscope download --dataset DiffSynth-Studio/diffsynth_example_dataset --include "boogu_image/Boogu-Image-0.1-Edit/*" --local_dir ./data/diffsynth_example_dataset | |
| accelerate launch examples/boogu_image/model_training/train.py \ | |
| --dataset_base_path "./data/diffsynth_example_dataset/boogu_image/Boogu-Image-0.1-Edit" \ | |
| --dataset_metadata_path "./data/diffsynth_example_dataset/boogu_image/Boogu-Image-0.1-Edit/metadata.csv" \ | |
| --max_pixels 1048576 \ | |
| --dataset_repeat 50 \ | |
| --model_id_with_origin_paths "Boogu/Boogu-Image-0.1-Edit:transformer/*.safetensors,Boogu/Boogu-Image-0.1-Edit:mllm/*.safetensors,Boogu/Boogu-Image-0.1-Edit:vae/*.safetensors" \ | |
| --learning_rate 1e-5 \ | |
| --num_epochs 2 \ | |
| --remove_prefix_in_ckpt "pipe.dit." \ | |
| --output_path "./models/train/Boogu-Image-0.1-Edit_full" \ | |
| --trainable_models "dit" \ | |
| --use_gradient_checkpointing \ | |
| --dataset_num_workers 8 \ | |
| --find_unused_parameters \ | |
| --data_file_keys "image,edit_image" \ | |
| --extra_inputs "edit_image" | |