add Dockerfile (#1)

[nicolecastillo]

• This commit adds the Dockerfile that assembles the image that will be used for the Popper workflow file.

[ivotron]

this is looking great! let me know if anything of the feedback I left doesn't make sense and I can clarify. thanks!

[ivotron]

root is already the user for base images, so we can remove this line

[ivotron]

gcc is a dependency of cmake, so looks like the minimal requirements are cmake and g++, the rest we can remove

[ivotron]

instead of installing from repos, we can install with pip:

pip install xlearn==0.40a1 liblinear==2.11.2

[ivotron]

we can use 3.7 as that's the latest stable release

[ivotron]

we can also reduce the size of the image by making use of multi-stage builds, as it would allow us to not have to include the compiler and all the auxiliary files that get created when the libraries get installed.

For this we need to add a name to the first image:

FROM python:3.7-slim-buster AS base

RUN ...
# all the commands you have above that install the libraries, then:


FROM python:3.7-slim-buster AS lib
COPY --from base /usr/local/lib/python3.7/site-packages /usr/local/lib/python3.7/

the last two lines say something like: "create a new layer from the python:3.7-slim-buster image, and in that image copy from the base image the site-packages folder".

[nicolecastillo]

@ivotron When I remove the root line I'm not able to run the tests from the demo, do you know why this happens?

[ivotron]

not sure. can you paste the commands you're using to run the demo so I can reproduce the issue on my end?

[nicolecastillo]

Sure. So once I run the image this is what I do:

• cd xlearn/demo/classification/higgs/
• python run_higgs.py

[ivotron]

do you that inside a container? or as part of building the dockerfile? I'll ask on slack

[ivotron]

it looks like the library assumes the USER variable exists, otherwise fails: https://github.com/aksnzhy/xlearn/blob/43fd6819d5b514099db7b85b27ebd010e0ea230e/src/base/system.h#L48 . This is a bit unorthodox but that's what it does, so we need to define it as you had it.

Also, I hadn't realized that the package on pypi is quite old. So we will have to build from source as you were doing before, sorry! This is what it ended up working on my end:

FROM python:3.7-slim-buster as base

RUN apt update && \
    apt install -y cmake g++ && \
    rm -rf /var/lib/apt/lists/* /tmp/* /var/tmp/*

COPY . /xlearn

# build from source (installs to /usr/local/lib/python3.7/site-packages/)
RUN cd /xlearn && \
    ls -l && \
    ./build.sh && \
    rm -r /xlearn

# install other library (installs to same site-packages path)
RUN pip install --no-cache-dir liblinear==2.11.2

# create an image without build dependencies
FROM python:3.7-slim-buster AS lib
ENV USER=root
COPY --from=base /usr/local/lib/python3.7/site-packages/* /usr/local/lib/python3.7/

I built the above with:

cd xlearn/
docker build -t xlearn -f demo/docker/Dockerfile .

And then ran it with:

docker run --rm \
  --volume $PWD:/workspace \
  --workdir=/workspace/demo/classification/higgs/ \
  xlearn \
    python run_higgs.py

[nicolecastillo]

Awesome, thanks!