Refactor batch handling update

[yafshar]

What does this PR do?

This PR refactors the batch handling update, which can help to improve performance

• Added get_num_items_in_batches function to calculate the number of items in each batch outside the training loop.
• Updated the get_batch_samples function to fetch batch samples only in the loop.

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?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[github-actions]

Hi 👋, thank you for opening this pull request! The pull request is converted to draft by default. When it is ready for review, please click the Ready for review button (at the bottom of the PR page).

[yafshar]

@techkang, @SunMarc, @muellerzr would you please review this. These changes have been added to https://github.com/huggingface/optimum-habana/tree/transformers_4_49 and we get better performance in fine tuning for some cases.

[techkang]

Hello, could you explain on why this refactoring improves performance? Maybe the specific PR in optimum-habana？

[regisss]

Hello, could you explain on why this refactoring improves performance? Maybe the specific PR in optimum-habana？

I think it's due to calling .item() at each training step, thus moving tensors from device to host and syncing the code every time while this could be done once before training.