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

ARG MODULE="inference"
ARG SERVICE="worker"

ARG APP_RELATIVE_PATH="${MODULE}/${SERVICE}"

WORKDIR /worker
COPY ./oasst-shared /tmp/oasst-shared
RUN conda create -n worker python=3.10 -y
RUN /opt/miniconda/envs/worker/bin/pip install /tmp/oasst-shared

COPY ./${APP_RELATIVE_PATH}/requirements.txt .
RUN /opt/miniconda/envs/worker/bin/pip install -r requirements.txt

COPY ./${APP_RELATIVE_PATH}/*.py .
COPY ./${APP_RELATIVE_PATH}/worker_full_main.sh /entrypoint.sh

ENV MODEL_ID="distilgpt2"
ENV INFERENCE_SERVER_URL="http://localhost:80"

ENTRYPOINT ["/entrypoint.sh"]