Docker: initial Dockerfile support

[QSmally]

Summary

This PR implements a basic Dockerfile in contrast to using Python virtual environment (as I personally had some dependency problems with it). It's basic, as in, it installs the CPU dependencies and leaves CUDA out of the mix. Maybe with build-args you can change some things but I don't have a Nvidia setup to test it.

How to test

$ cd training_extensions/
$ docker build -t trainer .
$ docker run -it --rm trainer
trainer% otx # ...

Checklist

• I have added unit tests to cover my changes.​
• I have added integration tests to cover my changes.​
• I have added e2e tests for validation.
• I have added the description of my changes into CHANGELOG in my target branch (e.g., CHANGELOG in develop).​
• I have updated the documentation in my target branch accordingly (e.g., documentation in develop).
• I have linked related issues.

License

• I submit my code changes under the same Apache License that covers the project.
  Feel free to contact the maintainers if that's a concern.
• I have updated the license header for each file (see an example below).

# Copyright (C) 2023 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

[ote-ci-bot]

Can one of the admins verify this patch?

[samet-akcay]

@ashwinvaidya17, can you review this PR please?

[ashwinvaidya17]

Thanks for the PR. I have one comment regarding the placement of this file. The rest looks good.

[goodsong81]

Thank you for the contribution @QSmally!
It looks generally OK but one only concern of mine is that the majority of use-cases of model training will run on GPUs. Current PR covers only Python3.9 + CPU.
@samet-akcay What do you think? Will we start from CPU and extend GPU support later?

[QSmally]

Thank you for the contribution @QSmally!
It looks generally OK but one only concern of mine is that the majority of use-cases of model training will run on GPUs. Current PR covers only Python3.9 + CPU.

Ty @goodsong81, @ashwinvaidya17, @yunchu,

I implemented build-arg support for UBUNTU_VER, PYTHON_VER and SOURCE. There's an issue where python3.10-dev and python3.10-distutils don't exist but I did get it working with Python 3.8 (along with Ubuntu 18.04). The default values if no args were given are 20.04, 3.9 and /whl/cpu as given in the README.

However, I lack the installation knowledge and machine to get the GPU dependencies, so I'd like assistance with that

[samet-akcay]

Thank you for the contribution @QSmally!
It looks generally OK but one only concern of mine is that the majority of use-cases of model training will run on GPUs. Current PR covers only Python3.9 + CPU.

Ty @goodsong81, @ashwinvaidya17, @yunchu,

I implemented build-arg support for UBUNTU_VER, PYTHON_VER and SOURCE. There's an issue where python3.10-dev and python3.10-distutils don't exist but I did get it working with Python 3.8 (along with Ubuntu 18.04). The default values if no args were given are 20.04, 3.9 and /whl/cpu as given in the README.

However, I lack the installation knowledge and machine to get the GPU dependencies, so I'd like assistance with that

Thanks for the contribution @QSmally. In this case, GPU support could be addressed by a follow-up PR by the team later.

[ashwinvaidya17]

Thanks

[yunchu]

LGTM. @QSmally, Thank you for your contributions!

[goodsong81]

@ashwinvaidya17 @yunchu Could you help pass the Dockerfile style check? I think the CI is regarding the warning as error.

[yunchu]

@goodsong81, we can update linter setting for dockerfile but I think it's better to resolve all warnings.
@QSmally, Can you apply some changes to resolve warning issues from linter? I recommend to use pre-commit tool to your development environment to get the direct feedback from the linters while generating the commits.

[QSmally]

I rarely encounter a Dockerfile with pinned apt versions, so I'm not sure how (and if) we should update that, @yunchu/@goodsong81

[yunchu]

@QSmally
You can add the comment below to disable that warning on right before the apt-get install command in your Dockerfile.

# hadolint ignore=DL3008 

[yunchu]

@QSmally I'm not sure but https://github.com/openvinotoolkit/training_extensions/blob/develop/.ci/Dockerfile this one could be the example for you.

[goodsong81]

Thank you for the contribution!