New config_from_keras_model

[vloncar]

Description

This PR introduces the new config_from_keras_model function that fixes the shortcomings of the old one, namely the need to keep in sync the supported layers and their configurable attributes. It relies on the attribute system to automatically discover tunable attributes of each layer and include them in the config. Config is now more expressive, and includes backend-specific parameters.

Type of change

For a new feature or function, please create an issue first to discuss it with us before submitting a pull request.

Note: Please delete options that are not relevant.

• [x] New feature (non-breaking change which adds functionality)
• [x] Breaking change (fix or feature that would cause existing functionality to not work as expected)

Tests

Current tests should suffice

Checklist

• [x] I have read the guidelines for contributing.
• [x] I have commented my code, particularly in hard-to-understand areas.
• [x] I have made corresponding changes to the documentation.
• [x] My changes generate no new warnings.
• [x] I have installed and run pre-commit on the files I edited or added. -> As discussed, this will be done after review
• [x] I have added tests that prove my fix is effective or that my feature works.

[vloncar]

@jmitrevs @jmduarte This is now redesigned and ready for review

[jmduarte]

For the test that fail, i think it's related to this error from the keras model conversion step, which happens when we try to import hls4ml:

WARN: Unable to import optimizer(s) from qkeras.py: No module named 'pyparsing'
WARN: Unable to import optimizer(s) from quantization_templates.py: No module named 'pyparsing'
WARN: Unable to import optimizer(s) from quantization_templates.py: No module named 'pyparsing'
WARN: Unable to import optimizer(s) from quantization_templates.py: No module named 'pyparsing'

It's weird because qkeras uses this library but it doesn't have it as a dependency perhaps due to a mixup of names (See: https://github.com/google/qkeras/issues/109). I will try adding pip install pyparsing in the Jenkinsfile.

[jmitrevs]

I confirmed that this works in our use of extension API (after removing the unused config). Let me look once more, but I think this is a good change.

[jmitrevs]

Let see the tests finish, but if people don't object, I think we can merge this PR. In general, though, I would prefer more documentation on the attributes, what are the common attributes for different categories, etc.

[jmitrevs]

Assuming the CI looks good, I'll merge this later today, unless someone objects.

[vloncar]

Wait, we need to do the pre-commit stuff. I'm working on it, but there are A LOT of issues.

[jmitrevs]

My understanding from @jmduarte was that it would automatically format things, so you don't need to manually do it, though I am not 100% sure.

[jmduarte]

some are fixed automatically (black formatting, isort sorting imports, etc.), some are not (flake8 issues). @vloncar is fixing (most) of the remaining issues now, and i will take a look at the rest, then we can merge

[vloncar]

I fixed all of the issues from flake8, but I cheated a bit by ignoring the print statement warnings. pre-commit hook is changed to ignore this, but we should make a note of this and replace all print statements with proper logging (is on my TODO list, but way back). For some reason this still fails the black and isort stages. Javier understands this better.