fix

lang/ja.md

@@ -32,3 +32,91 @@ webからはatprotoを通じて実行されます。
 - https://github.com/comfyanonymous/comfyui
 - https://github.com/stability-ai/stablediffusion
 
+## LoRA
+
+apple siliconでLoRA(finetuning)するには`mlx_lm`を使用します。
+
+```sh
+# https://ai.google.dev/gemma/docs/core/lora_tuning
+$ conda create -n finetuning python=3.12
+$ conda activate finetuning
+$ pip install mlx-lm #apple silicon
+$ ollama run llama3.2:1b
+$ echo "{ \"model\": \"https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct\", \"data\": \"https://github.com/ml-explore/mlx-examples/tree/main/lora/data\" }"|jq .
+$ model=meta-llama/Llama-3.2-1B-Instruct
+$ data=ml-explore/mlx-examples/lora/data
+$ mlx_lm.lora --train --model $model --data $data --batch-size 3
+
+$ ls adapters
+$ vim Modelfile
+FROM llama3.2:1b
+ADAPTER ./adapters
+
+$ ollama create ai -f ./Modelfile
+```
+
+## unsloth
+
+windowsでgemma3をunslothしてcustom modelを作成します。
+
+```sh
+$ nvidia-smi
+$ nvcc --version
+CUDA Version: 12.8
+
+$ pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128
+```
+
+```sh
+# https://docs.unsloth.ai/get-started/installing-+-updating/windows-installation
+$ conda create -n finetuning python=3.12
+$ conda activate finetuning
+$ conda install -c nvidia cuda cudnn
+$ ollama pull gemma3:12b
+
+# pip install unsloth
+$ pip install "unsloth[windows] @ git+https://github.com/unslothai/unsloth.git"
+$ vim custom.py
+```
+
+```py
+from unsloth import FastLanguageModel
+
+# モデルとトークナイザーの読み込み
+model, tokenizer = FastLanguageModel.from_pretrained(
+model_name = "google/gemma-12b",
+max_seq_length = 2048,
+dtype = None,
+load_in_4bit = True,
+)
+
+# LoRA設定
+model = FastLanguageModel.get_peft_model(
+model,
+r = 16,
+target_modules = ["q_proj", "k_proj", "v_proj", "o_proj"],
+lora_alpha = 16,
+lora_dropout = 0,
+bias = "none",
+use_gradient_checkpointing = "unsloth",
+)
+
+model.save_pretrained("gemma3-custom")
+
+# Modelfile作成
+with open("Modelfile", "w") as f:
+f.write(f"""
+FROM gemma3:12b
+ADAPTER ./gemma3-custom
+PARAMETER temperature 0.7
+PARAMETER stop <|endoftext|>
+""")
+```
+
+AI modelの名前は`ai`とします。
+
+```sh
+$ ollama create ai -f .\Modelfile
+$ ollama run ai
+```
+