Dockerfile

FROM nvidia/cuda:12.8.0-cudnn-devel-ubuntu24.04

ENV DEBIAN_FRONTEND=noninteractive
RUN apt-get update && \
    apt-get upgrade -y && \
    apt-get install -y --no-install-recommends --fix-missing \
    git \
    git-lfs \
    wget \
    curl \
    cmake \
    build-essential \
    libssl-dev \
    zlib1g-dev \
    libbz2-dev \
    libreadline-dev \
    libsqlite3-dev \
    libncursesw5-dev \
    xz-utils \
    tk-dev \
    libxml2-dev \
    libxmlsec1-dev \
    libffi-dev \
    liblzma-dev \
    ffmpeg \
    libcurl4-openssl-dev \
    nvidia-driver-570

# Check if user with UID 1000 exists, if not create it
RUN id -u 1000 &>/dev/null || useradd -m -u 1000 user
USER 1000
ENV HOME=/app \
    PATH=/app/.local/bin:${PATH}
WORKDIR /app

RUN curl https://pyenv.run | PYENV_ROOT=/app/.pyenv bash
ENV PYENV_ROOT=/app/.pyenv \
    PATH=/app/.pyenv/shims:/app/.pyenv/bin:${PATH}
ARG PYTHON_VERSION=3.11
RUN pyenv install ${PYTHON_VERSION} && \
    pyenv global ${PYTHON_VERSION} && \
    pyenv rehash && \
    pip install --no-cache-dir -U pip setuptools wheel && \
    pip install "huggingface-hub" "hf-transfer" "gradio[oauth]>=4.28.0" "gradio_huggingfacehub_search==0.0.8" "APScheduler"

# Clone llama.cpp and install its requirements
RUN git clone https://github.com/ggerganov/llama.cpp /app/llama.cpp
RUN pip install -r /app/llama.cpp/requirements.txt

# Build llama.cpp
ARG RUN_LOCALLY=false
ENV GGML_CUDA=ON
RUN if [ "$RUN_LOCALLY" = "true" ]; then export GGML_CUDA=OFF; fi && \
    echo "GGML_CUDA=$GGML_CUDA" && \
    cd /app/llama.cpp && \
    cmake -B build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=${GGML_CUDA} && \
    cmake --build build --config Release -j2 --target llama-quantize llama-gguf-split llama-imatrix && \
    cp ./build/bin/llama-* . && \
    rm -rf build

COPY --chown=1000 . /app
COPY groups_merged.txt /app/llama.cpp/

ENV PYTHONPATH=/app \
    PYTHONUNBUFFERED=1 \
    HF_HUB_ENABLE_HF_TRANSFER=1 \
    GRADIO_ALLOW_FLAGGING=never \
    GRADIO_NUM_PORTS=1 \
    GRADIO_SERVER_NAME=0.0.0.0 \
    GRADIO_THEME=huggingface \
    TQDM_POSITION=-1 \
    TQDM_MININTERVAL=1 \
    SYSTEM=spaces \
    LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH} \
    PATH=/usr/local/nvidia/bin:${PATH}

CMD ["/bin/bash", "start.sh"]