CLIP score improvements

[SkafteNicki]

What does this PR do?

Fixes #2935
Fixes #2906
Fixes #2883

Add the possibility for more multimodel models that supports longer captions than the original clip model.

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--2978.org.readthedocs.build/en/2978/

[codecov]

Codecov Report

Attention: Patch coverage is 21.05263% with 30 lines in your changes missing coverage. Please review.

Project coverage is 69%. Comparing base (34fca16) to head (34bfe8a).

Additional details and impacted files

@@          Coverage Diff           @@
##           master   #2978   +/-   ##
======================================
- Coverage      69%     69%   -0%     
======================================
  Files         338     338           
  Lines       19436   19464   +28     
======================================
+ Hits        13480   13486    +6     
- Misses       5956    5978   +22     

🚀 New features to boost your workflow:

• ❄ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Borda]

seems like just the docs are failing with warnings?

[SkafteNicki]

@Borda no there are also failing gpu tests (the test for clip score only runs on gpu, else it would take too long on cpu)

[Borda]

no there are also failing gpu tests (the test for clip score only runs on gpu, else it would take too long on cpu)

GPU testing is fine :)

>           raise ValueError(f"pic should not have > 4 channels. Got {pic.shape[-1]} channels.")
E           ValueError: pic should not have > 4 channels. Got 64 channels.

Let's try to pass it as PIL image instead?
Or it has reversed order of dimensions width x height x channels?

[Borda]

@SkafteNicki seems to be hanging https://dev.azure.com/Lightning-AI/Metrics/_build/results?buildId=227210&view=logs&jobId=e9b4e556-a3dd-5eb9-47b3-21a54922f9ed&j=e9b4e556-a3dd-5eb9-47b3-21a54922f9ed&t=de22f8c7-8fa5-5553-2667-84eedbb59742

[SkafteNicki]

@Borda is there a way to get more info from azure? it is hard to debug without what is going wrong