Code mixing metric

[arielge]

• Metric that utilizes generative LLM to estimate if a text is code-mixed
• Added language identification task + example card
• Format for Starling LM

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 47.36842% with 10 lines in your changes missing coverage. Please review.

Project coverage is 90.96%. Comparing base (4d3ea8e) to head (0ea4d57). Report is 6 commits behind head on main.

:exclamation: Current head 0ea4d57 differs from pull request most recent head 32aa794

Please upload reports for the commit 32aa794 to get more accurate results.

Files	Patch %	Lines
src/unitxt/metrics.py	47.36%	10 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## main #956 +/- ## ========================================== - Coverage 91.03% 90.96% -0.07% ========================================== Files 110 110 Lines 11841 11860 +19 ========================================== + Hits 10780 10789 +9 - Misses 1061 1071 +10 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[arielge]

Probably the easiest is to not run test_metric if a gpu is not available. Let me see if that does the trick

[arielge]

OK so this is still not enough as it looks like just loading the model crashes the test... Any ideas?