copy Whisper generate_with_fallback kwargs

[udeepam]

What does this PR do?

The Whisper longform audio decoding strategy (from the paper) is for each audio segment:

1. Beam search with beam_size=5 using log probabilities as the scoring function.
   • GenerationMode.BEAM_SEARCH
2. If average log probability over the generated tokens is lower than -1 or the generated text has a gzip compression rate higher than 2.4:
   • Fallback to sampling with temperature increased by 0.2 up to 1.0
   • GenerationMode.SAMPLE

However because the kwargs to generate_with_fallback() are not copied, if on the first audio segment we fallback the num_beams parameter in the kwargs get removed.

transformers/src/transformers/models/whisper/generation_whisper.py

Line 777 in 1681a6d

generation_config.num_beams = kwargs.pop("num_beams", 1) if not generation_config.do_sample else 1

Hence for the second audio segment we start with GenerationMode.SAMPLE instead of GenerationMode.BEAM_SEARCH.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

@sanchit-gandhi

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.

[ArthurZucker]

I think this was fixed by #29225

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.