feat: added method to load pretrained models from huggingface

[Marsmaennchen221]

Summary of Changes

feat: added NeuralNetworkClassifier.load_pretrained_model and NeuralNetworkRegressor.load_pretrained_model to load pretrained models from huggingface. Currently supports only Image models.
feat: added ModelImageSize, ConstantImageSize and VariableImageSize
feat: added support for NeuralNetworkRegressor with images of variable size. If you use a VariableImageSize any image which height and/or width are a multiple of the VariableImageSize are being supported by the model
feat: added NeuralNetworkClassifier.input_size and NeuralNetworkRegressor.input_size
feat: changed Column.get_distinct_values to keep order of values in column

[github-actions]

🦙 MegaLinter status: ✅ SUCCESS

Descriptor	Linter	Files	Fixed	Errors	Elapsed time
✅ PYTHON	black	20	0	0	1.56s
✅ PYTHON	mypy	20		0	3.23s
✅ PYTHON	ruff	20	0	0	0.32s
✅ REPOSITORY	git_diff	yes		no	0.32s

See detailed report in MegaLinter reports
Set VALIDATE_ALL_CODEBASE: true in mega-linter.yml to validate all sources, not only the diff

MegaLinter is graciously provided by

[Marsmaennchen221]

@lars-reimann do you have an idea how to test the pretrained models? Currently, those methods are not being tested. Maybe the same approach as in Safe-DS/Datasets#164 could work, but I would appreciate it if we could do this at a later date.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.76%. Comparing base (4a17f76) to head (e3b5df0).
Report is 72 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #790      +/-   ##
==========================================
+ Coverage   97.74%   97.76%   +0.01%     
==========================================
  Files         109      111       +2     
  Lines        5641     5691      +50     
==========================================
+ Hits         5514     5564      +50     
  Misses        127      127              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[lars-reimann]

do you have an idea how to test the pretrained models?

Not right now, no.

Currently, those methods are not being tested. Maybe the same approach as in Safe-DS/Datasets#164 could work

We've already used all our space for caches in this project (10GB), so that won't help us.

but I would appreciate it if we could do this at a later date.

Sure, no problem. You can add a comment # pragma: no cover to the loading methods.

[Gerhardsa0]

@lars-reimann do you have an idea how to test the pretrained models? Currently, those methods are not being tested. Maybe the same approach as in Safe-DS/Datasets#164 could work, but I would appreciate it if we could do this at a later date.

For testing these models, maybe it is enough for now that we check that the structure is correct.

As example we could load an model into safe-ds and then check if its loaded correctly:
the layers, the size of the layers, layer type etc

[lars-reimann]

Looks great, thanks.

[lars-reimann]

🎉 This PR is included in version 0.26.0 🎉

The release is available on:

• v0.26.0
• GitHub release

Your semantic-release bot 📦🚀