ci: cache reference metrics & clean audio tests

[Borda]

What does this PR do?

Resolves #949

Before submitting

• Was this discussed/agreed via a Github issue? (no need for typos and docs improvements)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure to update the docs?
• Did you write any new necessary tests?

PR review

Anyone in the community is free to review the PR once the tests have passed.
If we didn't discuss your PR in Github issues there's a high chance it will not be merged.

Did you have fun?

Make sure you had fun coding 🙃

---

📚 Documentation preview 📚: https://torchmetrics--2335.org.readthedocs.build/en/2335/

[Borda]

it seems that hashing is not supported by its nature for list or np.ndarray, see:

• Numpy array default hasher python-cachier/cachier#43
• workaround for TypeError: unhashable type: 'list' python-cachier/cachier#136

[codecov]

Codecov Report

Merging #2335 (5d217bd) into master (efb3a25) will decrease coverage by 1%.
The diff coverage is n/a.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #2335    +/-   ##
=======================================
- Coverage      69%     68%    -1%     
=======================================
  Files         307     307            
  Lines       17352   17352            
=======================================
- Hits        11961   11821   -140     
- Misses       5391    5531   +140     

[shaypal5]

it seems that hashing is not supported by its nature for list or np.ndarray, see:

• Numpy array default hasher python-cachier/cachier#43
• workaround for TypeError: unhashable type: 'list' python-cachier/cachier#136

@Borda, please note you can easily work around this using custom hashing logic, which cachier supports:
https://github.com/python-cachier/cachier?tab=readme-ov-file#working-with-unhashable-arguments

[Borda]

note you can easily work around this using custom hashing logic, which cachier supports

@shaypal5, yes, I saw it but this I believe is quite often the case so it would be beneficial to cover it out of the box (I mean list not numpy.andarray)

[shaypal5]

note you can easily work around this using custom hashing logic, which cachier supports

@shaypal5, yes, I saw it but this I believe is quite often the case so it would be beneficial to cover it out of the box (I mean list not numpy.andarray)

Cool. As long as you are aware this shouldn't block your use case, I'd love to get contributions by you into cachier!

[Borda]

An idea bout increasing the testing range, we could move importing the dependency package to the def _reference_XXX(...) function with eventual skip if not imported... that would mean once the reference computed on with feasible configuration (mostly means package and python version compatibility issue) it will be used as part of a cache for remaining configurations.
This would fully cover the suggested improvement mentioned in #949 with minimal engineering overhead 🚀

cc: @SkafteNicki @justusschock @lantiga