Dockerfile

ARG GRADIO_PORT
ARG MODEL_NAME
ARG QUANTIZATION
ARG MAX_CONCURRENT_REQUESTS
ARG MAX_INPUT_LENGTH
ARG MAX_TOTAL_TOKENS

FROM ghcr.io/huggingface/text-generation-inference:latest

ARG GRADIO_PORT
ARG MODEL_NAME
ARG QUANTIZATION
ARG MAX_CONCURRENT_REQUESTS
ARG MAX_INPUT_LENGTH
ARG MAX_TOTAL_TOKENS

ENV TZ=Europe/Paris \
    PORT=3000

RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
    gnupg \
    curl \
    gettext && \
    rm -rf /var/lib/apt/lists/*
COPY entrypoint.sh.template entrypoint.sh.template

RUN envsubst < "entrypoint.sh.template" > "entrypoint.sh"
RUN cat "entrypoint.sh"

RUN mkdir -p /data/db
RUN chown -R 1000:1000 /data
RUN useradd -m -u 1000 user

# Switch to the "user" user
USER user

ENV HOME=/home/user \
    PATH=/home/user/.local/bin:$PATH

COPY app app
RUN pip install -r app/requirements.txt

ENTRYPOINT ["/bin/bash"]
CMD ["entrypoint.sh"]