Dockerfile

FROM python:latest

ENV PYTHONUNBUFFERED 1

EXPOSE 7860

WORKDIR /app

RUN export PATH="/usr/local/cuda/bin:$PATH"
RUN wget -qO- "https://cmake.org/files/v3.17/cmake-3.17.0-Linux-x86_64.tar.gz" | tar --strip-components=1 -xz -C /usr/local
RUN CMAKE_ARGS="-DLLAMA_CUBLAS=on" FORCE_CMAKE=1 pip install llama-cpp-python --no-cache-dir
COPY requirements.txt ./
RUN pip install --upgrade pip && \
    pip install -r requirements.txt

COPY . .
RUN ls -al

CMD ["python", "app.py", "--host", "0.0.0.0", "--port", "7860"]