add fscore, reorganize mod and change error in doc.

[tsanona]

Pull Request Template

Checklist

• Confirmed that run-checks all script has been executed.
• Made sure the book is up to date with changes in this PR.

Related Issues/PRs

#544

Changes

Add F_βScoreMetric
Reorganize metric\mod to mach existing order.

Testing

Added tests for new metric

[codecov]

Codecov Report

Attention: Patch coverage is 90.27778% with 7 lines in your changes missing coverage. Please review.

Project coverage is 83.14%. Comparing base (7a19b5f) to head (d2edd87).
Report is 8 commits behind head on main.

Files with missing lines	Patch %	Lines
crates/burn-train/src/metric/fbetascore.rs	90.27%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2648      +/-   ##
==========================================
+ Coverage   81.84%   83.14%   +1.29%     
==========================================
  Files         838      840       +2     
  Lines      107487   107985     +498     
==========================================
+ Hits        87975    89784    +1809     
+ Misses      19512    18201    -1311     

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

[laggui]

Thanks for contributing yet another metric 😄

Minor comment regarding the name, otherwise should be good to go!

[laggui]

Suggestion: rename to FBetaScore. It seems that many people refer to F-score as F-beta score (as it is parameterized by beta).

See sklearn for example: https://scikit-learn.org/stable/modules/generated/sklearn.metrics.fbeta_score.html

[laggui]

Also, can you fix the trailing spaces that misalign the right table border 😅

And the F in F_β score should be capitalized (and maybe add a space like I did for readability?)

[tsanona]

Yeah I was torn and chose the smallest name, but I also agree with this, will change!
Ah yes will format the table didn't realize it was misaligned.

[tsanona]

While I agree the F should be capitalized, it feels like it isn't consistent with the other metric names, should I also change the others or just capitalize the F?

[laggui]

I think it's consistent. The F-score is always capitalized (based on its origin), ROC is an acronym for Receiver Operating Characteristic. Other names are not capitalized because they're really just nouns (precision, recall, accuracy). I guess there could be a debate over top-k vs Top-k, but even wikipedia doesn't capitalize it 😅

[tsanona]

All good, it's been fulfilling to contribute! 🌟 I have now plans to adapt the existing metrics (accuracy+topkaccuracy, auroc and hamming) to the same format as the ones I've implemented. (this will be breaking code, which I'm not fully sure on how to handle)
I also want to add a few more losses and, or and and ops to Bool tensors and scatter and select* ops to all tensors (not just numeric).
If there's some priority/no necessity of any of these I'd be interested in discussing 🚀 Should I just create prs or is there a better way to start these discussions?

[laggui]

I have now plans to adapt the existing metrics (accuracy+topkaccuracy, auroc and hamming) to the same format as the ones I've implemented. (this will be breaking code, which I'm not fully sure on how to handle)

Not sure what do yo have in mind?

We try to limit breaking changes, but sometimes that's not possible. Depends on the objective I guess!

I also want to add a few more losses and, or and and ops to Bool tensors and scatter and select* ops to all tensors (not just numeric). If there's some priority/no necessity of any of these I'd be interested in discussing 🚀 Should I just create prs or is there a better way to start these discussions?

I think missing features are mostly captured in issues. If a feature request is not already filed, you can open an issue for the request and the discussion can start there 🙂

For other general dev discussions and thoughts, this usually happens on discord.