Dockerfile

ARG UBUNTU_VERSION=22.04
ARG CUDA_VERSION=12.3.1
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}

FROM ${BASE_CUDA_DEV_CONTAINER} as build

ARG CUDA_DOCKER_ARCH=all

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

WORKDIR /build

RUN git clone https://github.com/ggerganov/llama.cpp.git

WORKDIR /build/llama.cpp

ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
ENV LLAMA_CUBLAS=1

RUN mkdir build && \
    cd build && \
    cmake .. -DLLAMA_CUBLAS=ON && \
    cmake --build . --config Release

WORKDIR /data
RUN wget https://huggingface.co/IlyaGusev/saiga2_7b_gguf/resolve/main/model-q8_0.gguf -O model.gguf

FROM ${BASE_CUDA_RUN_CONTAINER} as runtime


WORKDIR /app

# Copy the executable from the build stage
COPY --from=build /build/llama.cpp/build/bin/server /app
COPY --from=build /data/model.gguf /data/model.gguf
WORKDIR /app
EXPOSE 7860

CMD ./server -m /data/model.gguf -c 2048 --port 7860