Dockerfile

FROM ghcr.io/ggml-org/llama.cpp:full

WORKDIR /app

# パッケージのインストール 
RUN apt update && apt install -y python3-pip
RUN pip install -U huggingface_hub

# モデルをHugging Faceからダウンロード 
RUN python3 -c 'from huggingface_hub import hf_hub_download; \
    repo="mradermacher/Qwen3-Coder-30B-A3B-Instruct-480B-Distill-V2-Fp32-i1-GGUF"; \
    hf_hub_download(repo_id=repo, filename="Qwen3-Coder-30B-A3B-Instruct-480B-Distill-V2-Fp32.i1-IQ3_M.gguf", local_dir="/app")'

# サーバーの設定
CMD ["--server", \
     "-m", "/app/Qwen3-Coder-30B-A3B-Instruct-480B-Distill-V2-Fp32.i1-IQ3_M.gguf", \
     "--host", "0.0.0.0", \
     "--port", "7860", \
     "-t", "2", \
     "--cache-type-k", "q8_0", \
     "--cache-type-v", "iq4_nl", \
     "-c", "32000", \
     "-n", "8000"]