Dockerfile

ARG PYTHON_IMAGE_TAG=3.7-stretch

#
# First stage: get dependencies, build the project, store binaries (whl)
#

FROM python:${PYTHON_IMAGE_TAG} as py-build
WORKDIR /build
# install and store the requirements (this stage will be re-used unless requirements.txt changes)
COPY requirements.txt .
RUN pip install -r requirements.txt && pip wheel -r requirements.txt -w deps
# add all other source files, force rebuild only from here
COPY . .
RUN python setup.py install && python setup.py bdist_wheel

#
# Second stage: create the smallest possible image for deployment
#

FROM python:${PYTHON_IMAGE_TAG}

LABEL maintainer="Quantum Analytica"

WORKDIR /app
# install all dependencies from wheel packages in the 'deps' folder
COPY --from=py-build /build/deps/ deps/
RUN [ -n "$(ls -A deps)" ] && pip install deps/*.whl && rm -rf deps || echo "no dependencies to install"
# install the application from a wheel package in the 'dist' folder
COPY --from=py-build /build/dist/ dist/
RUN pip install dist/*.whl && rm -rf dist
ENTRYPOINT ["python", "-OO", "-m", "language_modeling"]