Apple Silicon（M1 / M2）MacのGPUでStable DiffusionのLoRAを学習（2025/3/8更新）

conda create -n kohya_ss python==3.10.9
conda activate kohya_ss

python ./setup/setup_linux.py --platform-requirements-file=requirements_macos_arm64.txt

# metal用のpytorchをインストール
pip3 install torch torchvision

# metal用のtensorflowをインストール
conda install -c apple tensorflow-deps # 今はいらないかも…
pip3 install tensorflow-macos tensorflow-metal tensorboard

# diffusersのインストール
pip install diffusers

# LoRAの学習用コードをダウンロード
git clone https://github.com/bmaltais/kohya_ss.git
cd kohya_ss

# requirements.txt の tensorflow, tensorboard はコメントアウトする
pip install -r requirements.txt

.
├── train
│   └── 20_pochi dog
│       ├── AAA1.jpg
│       ├── AAA2.jpg
│       ├── AAA3.jpg
├── reg
│   └── 20_dog
│       ├── BBB1.png
│       ├── BBB2.png
│       ├── BBB3.png
│       ├── BBB4.png
│       ├── BBB5.png
└── outputs

{
  "LoRA_type": "Standard",
  "LyCORIS_preset": "full",
  "adaptive_noise_scale": 0,
  "additional_parameters": "",
  "async_upload": false,
  "block_alphas": "",
  "block_dims": "",
  "block_lr_zero_threshold": "",
  "bucket_no_upscale": true,
  "bucket_reso_steps": 64,
  "bypass_mode": false,
  "cache_latents": true,
  "cache_latents_to_disk": false,
  "caption_dropout_every_n_epochs": 0,
  "caption_dropout_rate": 0,
  "caption_extension": ".txt",
  "clip_skip": 2,
  "color_aug": false,
  "constrain": 0,
  "conv_alpha": 1,
  "conv_block_alphas": "",
  "conv_block_dims": "",
  "conv_dim": 1,
  "dataset_config": "",
  "debiased_estimation_loss": false,
  "decompose_both": false,
  "dim_from_weights": false,
  "dora_wd": false,
  "down_lr_weight": "",
  "dynamo_backend": "no",
  "dynamo_mode": "default",
  "dynamo_use_dynamic": false,
  "dynamo_use_fullgraph": false,
  "enable_bucket": true,
  "epoch": 8,
  "extra_accelerate_launch_args": "",
  "factor": -1,
  "flip_aug": false,
  "fp8_base": false,
  "full_bf16": false,
  "full_fp16": false,
  "gpu_ids": "",
  "gradient_accumulation_steps": 8,
  "gradient_checkpointing": false,
  "huber_c": 0.1,
  "huber_schedule": "snr",
  "huggingface_path_in_repo": "",
  "huggingface_repo_id": "",
  "huggingface_repo_type": "",
  "huggingface_repo_visibility": "",
  "huggingface_token": "",
  "ip_noise_gamma": 0,
  "ip_noise_gamma_random_strength": false,
  "keep_tokens": 0,
  "learning_rate": 0.0001,
  "log_tracker_config": "",
  "log_tracker_name": "",
  "log_with": "",
  "logging_dir": "",
  "loss_type": "l2",
  "lr_scheduler": "cosine",
  "lr_scheduler_args": "",
  "lr_scheduler_num_cycles": 1,
  "lr_scheduler_power": 1,
  "lr_warmup": 10,
  "main_process_port": 0,
  "masked_loss": false,
  "max_bucket_reso": 2048,
  "max_data_loader_n_workers": 0,
  "max_grad_norm": 1,
  "max_resolution": "512,512",
  "max_timestep": 1000,
  "max_token_length": 75,
  "max_train_epochs": 0,
  "max_train_steps": 0,
  "mem_eff_attn": false,
  "metadata_author": "",
  "metadata_description": "",
  "metadata_license": "",
  "metadata_tags": "",
  "metadata_title": "",
  "mid_lr_weight": "",
  "min_bucket_reso": 256,
  "min_snr_gamma": 0,
  "min_timestep": 0,
  "mixed_precision": "no",
  "model_list": "",
  "module_dropout": 0,
  "multi_gpu": false,
  "multires_noise_discount": 0.3,
  "multires_noise_iterations": 0,
  "network_alpha": 1,
  "network_dim": 8,
  "network_dropout": 0,
  "network_weights": "",
  "noise_offset": 0,
  "noise_offset_random_strength": false,
  "noise_offset_type": "Original",
  "num_cpu_threads_per_process": 2,
  "num_machines": 1,
  "num_processes": 1,
  "optimizer": "AdamW",
  "optimizer_args": "",
  "output_dir": "./outputs/20_dogs",
  "output_name": "last",
  "persistent_data_loader_workers": false,
  "pretrained_model_name_or_path": "runwayml/stable-diffusion-v1-5",
  "prior_loss_weight": 1,
  "random_crop": false,
  "rank_dropout": 0,
  "rank_dropout_scale": false,
  "reg_data_dir": "",
  "rescaled": false,
  "resume": "",
  "resume_from_huggingface": "",
  "sample_every_n_epochs": 0,
  "sample_every_n_steps": 10,
  "sample_prompts": "dogs",
  "sample_sampler": "euler_a",
  "save_as_bool": false,
  "save_every_n_epochs": 1,
  "save_every_n_steps": 20,
  "save_last_n_steps": 10,
  "save_last_n_steps_state": 0,
  "save_model_as": "safetensors",
  "save_precision": "float",
  "save_state": false,
  "save_state_on_train_end": false,
  "save_state_to_huggingface": false,
  "scale_v_pred_loss_like_noise_pred": false,
  "scale_weight_norms": 0,
  "sdxl": false,
  "sdxl_cache_text_encoder_outputs": false,
  "sdxl_no_half_vae": false,
  "seed": 0,
  "shuffle_caption": false,
  "stop_text_encoder_training": 0,
  "text_encoder_lr": 0.0001,
  "train_batch_size": 2,
  "train_data_dir": "./dataset/images",
  "train_norm": false,
  "train_on_input": true,
  "training_comment": "",
  "unet_lr": 0.0001,
  "unit": 1,
  "up_lr_weight": "",
  "use_cp": false,
  "use_scalar": false,
  "use_tucker": false,
  "v2": false,
  "v_parameterization": false,
  "v_pred_like_loss": 0,
  "vae": "",
  "vae_batch_size": 0,
  "wandb_api_key": "",
  "wandb_run_name": "",
  "weighted_captions": false,
  "xformers": "none"
}

accelerate launch --num_cpu_threads_per_process 4 train_network.py \
    --pretrained_model_name_or_path="your_base_model_path/model_name.safetensors" \
    --train_data_dir="kohya_ss/data/train" \
    --reg_data_dir="kohya_ss/data/reg" \
    --prior_loss_weight=1.0 --resolution 512 \
    --output_dir="kohya_ss/data/outputs" \
    --output_name=your_lora_name \
    --train_batch_size=2 --learning_rate=1e-4 --max_train_epochs 4 \
    --seed 42 --save_model_as=safetensors --save_every_n_epochs=1 \
    --max_data_loader_n_workers=1 \
    --network_module=networks.lora --network_dim=4 \
    --training_comment="activate by pochi dog" \
    --enable_bucket --bucket_no_upscale \
    --optimizer_type="AdamW" --mixed_precision="no" --save_precision="float"

--sample_every_n_epochs=1 \
--sample_sampler="k_euler_a" \
--sample_prompts="kohya_ss/data/prompt.txt"

# prompt 1
pochi, run --w 512 --h 512 --d 1 --l 7.5 --s 20

pip3 install -U torch torchvision torchaudio