FIX overwriting metadata when both verified and unverified reported values

[Wauplin]

Should fix #1185.

With this PR, 2 eval results are considered to describe the same object if and only if all attributes are the same except the metric value itself (see is_equal_except_value). Otherwise value is not overwritten.

cc @lewtun for review

[HuggingFaceDocBuilder]

The documentation is not available anymore as the PR was closed or merged.

[lewtun]

Thanks a lot for the quick fix on this bug @Wauplin 🔥 !

I've tested the PR against my own model can can confirm it works as expected: https://huggingface.co/autoevaluate/binary-classification/discussions/58/files

The code itself also LGTM!

[lewtun]

Small suggestion to help community developers understand what this is about:

Suggested change

	This is a test model card.
	This is a test model card containing one self-reported metric and one "verified" metric in the format produced by Hugging Face's [model evaluation service](https://huggingface.co/spaces/autoevaluate/model-evaluator).

[nateraw]

Seeing failing test here:

FAILED ../tests/test_repocard.py::RepocardMetadataUpdateTest::test_update_existing_result_without_overwrite - AssertionError: Regex pattern did not match.
 Regex: "You passed a new value for the existing metric 'name: Accuracy, type: accuracy'. Set `overwrite=True` to overwrite existing metrics."
 Input: "You passed a new value for the existing metric 'name: Accuracy, type:  accuracy'. Set `overwrite=True` to overwrite existing metrics."

[codecov]

Codecov Report

Base: 84.00% // Head: 84.03% // Increases project coverage by +0.02% 🎉

Coverage data is based on head (706c629) compared to base (711f688).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1186      +/-   ##
==========================================
+ Coverage   84.00%   84.03%   +0.02%     
==========================================
  Files          44       44              
  Lines        4321     4327       +6     
==========================================
+ Hits         3630     3636       +6     
  Misses        691      691              

Impacted Files	Coverage Δ
src/huggingface_hub/repocard.py	95.69% <100.00%> (-0.03%)	⬇️
src/huggingface_hub/repocard_data.py	98.49% <100.00%> (+0.08%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[julien-c]

pinging @coyotte508 and @allendorf to make sure this is also the behavior that makes sense considering the server-side implem of verified eval results

(i'm a bit fuzzy on this subject myself)

[lewtun]

pinging @coyotte508 and @allendorf to make sure this is also the behavior that makes sense considering the server-side implem of verified eval results

(i'm a bit fuzzy on this subject myself)

I think this will become clearer once I refactor AutoTrain to include the new model card API in https://github.com/huggingface/autotrain-backend/pull/823 :)