Unpin transformers for v5.0.0

[jveitchmichaelis]

Checked if upstream fixes allow training DETR or if the CI is still breaking. Cause seems to be garbage weight initialisation by transformers, which caused extreme gradients that failed from-scratch training (nan boxes in output). This would then fail the low level tests in the CI, but shouldn't affect using pre-trained weights.

This seems to be fixed and released now, and the CI is pulling 5.0.0. This only affected a narrow use-case and should not matter to most users, so if someone hasn't updated, it shouldn't break other things.

I don't think it's necessary to restrict transformers versions in pyproject as everything else should work fine with 4.57, but we could include that (e.g. transformers<4.57, transformers>=5.0.0)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 87.16%. Comparing base (3146b96) to head (8be81c2).
⚠️ Report is 5 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1231      +/-   ##
==========================================
- Coverage   87.89%   87.16%   -0.73%     
==========================================
  Files          20       20              
  Lines        2776     2782       +6     
==========================================
- Hits         2440     2425      -15     
- Misses        336      357      +21     

Flag	Coverage Δ
unittests	87.16% <ø> (-0.73%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

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

🚀 New features to boost your workflow:

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