Dockerfile

ARG UBUNTU_VERSION=22.04
# This needs to generally match the container host's environment.
ARG CUDA_VERSION=11.7.1
# Target the CUDA build image
ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
# Target the CUDA runtime image
ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}

FROM ${BASE_CUDA_DEV_CONTAINER} as build

# Unless otherwise specified, we make a fat build.
ARG CUDA_DOCKER_ARCH=all

RUN apt-get update && \
    apt-get install -y build-essential git

# Install Python3 and pip
RUN apt-get install -y python3 python3-pip

WORKDIR /app

COPY . .

# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
# Enable cuBLAS
ENV LLAMA_CUBLAS=1

FROM ${BASE_CUDA_RUN_CONTAINER} as runtime

# Install build and runtime dependencies
RUN apt-get update && \
    apt-get install -y \
    libopenblas-dev \
    ninja-build \
    build-essential \
    pkg-config \
    curl

# Install Python3 and pip for the runtime container as well
RUN apt-get install -y python3 python3-pip

RUN pip3 install -U pip setuptools wheel && \
  pip3 install --verbose llama-cpp-python[server]

# Download model
RUN mkdir 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 model/gguf-model.gguf

COPY ./start_server.sh ./
COPY ./main.py ./

# Make the server start script executable
RUN chmod +x ./start_server.sh

# Set environment variable for the host
ENV HOST=0.0.0.0
ENV PORT=7860

# Expose a port for the server
EXPOSE ${PORT}

# Run the server start script
CMD ["/bin/sh", "./start_server.sh"]