Dockerfile

## TO BUILD IMAGE:
# docker build -t ml .

## TO RUN IMAGE (ENVIRONMENTAL VARIABLES DECLARED ABOVE)
# docker run -itp 8080:8080 --memory=1g ml
## TO RUN IMAGE (OVERRIDE ENVIRONMENTAL VARIABLES DECLARED ABOVE)
# docker run -itp 8080:8080 --memory=1g --env MODEL_FILE="models/ml_classifier.pkl" ml

## TO SAVE THE IMAGE
# docker save ml -o ml.rar
## TO IMPORT THE IMAGE
# docker load -i ml.rar


FROM python:3.9-slim
# From pytorch/pytorch

#############################
# INSTALL PYTHON DEPENDENCIES
#############################

# install git for pip install git+https://
# RUN apt-get -o Acquire::Max-FutureTime=100000 update \
#  && apt-get install -y --no-install-recommends build-essential git

# create a virtual environment
# RUN python -m venv /opt/venv
# ENV PATH="/opt/venv/bin:$PATH"

# copy and install python requirements + ember from github
COPY docker-requirements.txt .
# RUN pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
RUN pip install torch torchvision torchaudio
RUN pip install --no-cache-dir -r docker-requirements.txt

#############################
# REBASE & DEPLOY CODE
#############################

# rebase to make a smaller image
# FROM python:3.9-slim

# # required libgomp1 for ember
# RUN apt-get -o Acquire::Max-FutureTime=100000 update \
#     && apt-get -y --no-install-recommends install \
#         libgomp1 \
#     && rm -rf /var/lib/apt/lists/*

# # copy python virtual env (all dependencies) from previous image
# COPY --from=0 /opt/venv /opt/venv

# change working directory
WORKDIR /opt/defender/

# copy defender code to /opt/defender/defender
COPY docker /opt/defender/defender

#############################
# SETUP ENVIRONMENT
#############################

# open port 8080
EXPOSE 8080

# add a defender user and switch user
# RUN groupadd -r defender && useradd --no-log-init -r -g defender defender
# USER defender

# update environmental variables
# ENV PATH="/opt/venv/bin:$PATH"
# ENV PYTHONPATH="/opt/defender"

# one may tune model file / threshold / name via environmental variables
# ENV MODEL_FILE models/ml_classifier.pkl

#############################
# RUN CODE
#############################
CMD python -m defender > /dev/null 2>&1