Dockerfile

FROM python:3.9

ENV TQDM_DISABLE=0
ENV TQDM_FORCE=1

WORKDIR /code
RUN mkdir -p /code && chmod -R 777 /code
RUN mkdir -p /.cache && chmod -R 777 /.cache
RUN mkdir -p /.cache/huggingface && chmod -R 777 /.cache/huggingface

COPY ./requirements.txt /code/requirements.txt

RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt

COPY . .

RUN python3 -c "from transformers import AutoModelForCausalLM; AutoModelForCausalLM.from_pretrained('ibm-granite/granite-34b-code-instruct-8k', cache_dir='/.cache/huggingface')"

RUN uvicorn main:app --host 0.0.0.0 --port 7860

# CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]