Dockerfile

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

WORKDIR /app


RUN apt update && apt install -y python3-pip
RUN pip install -U huggingface_hub

RUN python3 -c 'from huggingface_hub import hf_hub_download; \
    repo="rahul7star/Math-AI-Full"; \
    hf_hub_download(repo_id=repo, filename="rahul7star_Math-AI-Full-Q4_K_M.gguf", local_dir="/app")'


CMD ["--server", \
     "-m", "/app/rahul7star_Math-AI-Full-Q4_K_M.gguf", \
     "--host", "0.0.0.0", \
     "--port", "7860", \
     "-t", "2", \
     "--cache-type-k", "q8_0", \
     "--cache-type-v", "iq4_nl", \
     "-c", "128000", \
     "-n", "38912",\
     "--repeat-penalty", "1.2"]