educatedlucifer12/fable5

1FROM python:3.11-slim2 3ENV DEBIAN_FRONTEND=noninteractive \4    MODEL_REPO=yuxinlu1/gemma-4-12B-coder-fable5-composer2.5-v1-GGUF \5    MODEL_FILE=gemma4-coding-Q4_K_M.gguf \6    MODEL_DIR=/data/models/gemma4-coder \7    LLAMA_VERSION=b9592 \8    LLAMA_DIR=/opt/llama.cpp \9    LLAMA_SERVER_BIN=/opt/llama.cpp/llama-server \10    LD_LIBRARY_PATH=/opt/llama.cpp \11    LLAMA_HOST=0.0.0.0 \12    LLAMA_PORT=7860 \13    THREADS=4 \14    CTX_SIZE=2048 \15    BATCH_SIZE=default \16    UBATCH_SIZE=default \17    FLASH_ATTN=default \18    CACHE_TYPE_K=default \19    CACHE_TYPE_V=default \20    GPU_LAYERS=0 \21    TEMPERATURE=0.2 \22    TOP_P=0.95 \23    TOP_K=64 \24    REPEAT_PENALTY=1.08 \25    HF_XET_HIGH_PERFORMANCE=1 \26    PYTHONUNBUFFERED=127 28RUN apt-get update && apt-get install -y --no-install-recommends \29    ca-certificates \30    curl \31    libgomp1 \32    libstdc++6 \33    && rm -rf /var/lib/apt/lists/*34 35RUN mkdir -p "${LLAMA_DIR}" \36    && curl -fL "https://github.com/ggml-org/llama.cpp/releases/download/${LLAMA_VERSION}/llama-${LLAMA_VERSION}-bin-ubuntu-x64.tar.gz" \37        | tar -xz --strip-components=1 -C "${LLAMA_DIR}" \38    && chmod +x "${LLAMA_SERVER_BIN}"39 40RUN pip install --no-cache-dir \41    huggingface_hub42 43WORKDIR /app44COPY app.py /app/app.py45 46EXPOSE 786047CMD ["python", "app.py"]48