Download Dockerfile from matthoffner/llama-cpp-server: direct link, hf CLI and curl.
- Browser
- Download file 2.16 kB
-
https://huggingface.co/spaces/matthoffner/llama-cpp-server/resolve/main/Dockerfile
- Command line
-
hf download hf://spaces/matthoffner/llama-cpp-server/Dockerfile
-
curl -L -o Dockerfile https://huggingface.co/spaces/matthoffner/llama-cpp-server/resolve/main/Dockerfile
2.16 kB
| # Set arguments for versions | |
| ARG UBUNTU_VERSION=22.04 | |
| ARG CUDA_VERSION=11.7.1 | |
| ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION} | |
| ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION} | |
| # Build stage with CUDA development container | |
| FROM ${BASE_CUDA_DEV_CONTAINER} as build | |
| # Install build essentials and git | |
| RUN apt-get update && \ | |
| apt-get install -y build-essential git | |
| # Install Python3 and pip | |
| RUN apt-get install -y python3 python3-pip | |
| # Set work directory to /app | |
| WORKDIR /app | |
| # Copy your application code to the container | |
| COPY . . | |
| # Create a non-root user 'user' in the build stage as well | |
| RUN useradd -m -u 1000 user | |
| # Switch to the non-root user for any further commands | |
| USER user | |
| # Set nvcc architecture and enable cuBLAS | |
| ENV CUDA_DOCKER_ARCH=all \ | |
| LLAMA_CUBLAS=1 | |
| # Runtime stage with CUDA runtime container | |
| FROM ${BASE_CUDA_RUN_CONTAINER} as runtime | |
| # Re-create the non-root user 'user' in the runtime stage | |
| RUN useradd -m -u 1000 user && \ | |
| apt-get update && \ | |
| apt-get install -y libopenblas-dev ninja-build build-essential pkg-config curl | |
| # Switch to non-root user | |
| USER user | |
| # Set home and path for the user | |
| ENV HOME=/home/user \ | |
| PATH=/home/user/.local/bin:$PATH | |
| # Set work directory to user's home directory | |
| WORKDIR $HOME/app | |
| # Install Python3 and pip for the runtime container | |
| USER root | |
| RUN apt-get install -y python3 python3-pip | |
| # Switch back to the non-root user for installing Python packages | |
| USER user | |
| RUN pip install --no-cache-dir --upgrade pip setuptools wheel && \ | |
| pip install --verbose llama-cpp-python[server] | |
| # Download the model to the user's directory | |
| RUN mkdir $HOME/model && \ | |
| curl -L https://huggingface.co/matthoffner/Magicoder-S-DS-6.7B-GGUF/resolve/main/Magicoder-S-DS-6.7B_Q4_K_M.gguf -o $HOME/model/gguf-model.gguf | |
| COPY --chown=user ./main.py $HOME/app/ | |
| # Set environment variables for the host | |
| ENV HOST=0.0.0.0 \ | |
| PORT=7860 | |
| # Expose the server port | |
| EXPOSE ${PORT} | |
| RUN ls -la $HOME/model | |
| # Run the server start script | |
| CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860", "--log-level", "debug"] | |