[bug] fsdp training save checkpoint twice

[cyr0930]

What does this PR do?

FSDP training save checkpoint twice as pytorch_model_fsdp.bin and model.safetensors.
And saving model during calling _save_optimizer_and_scheduler is not intuitive.

(Maybe I have to rename final checkpoint or save it as safetensors format.)

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?

@zach-huggingface @SunMarc

[github-actions]

Hi 👋, thank you for opening this pull request! The pull request is converted to draft by default. The CI will be paused while the PR is in draft mode. When it is ready for review, please click the Ready for review button (at the bottom of the PR page). This will assign reviewers and trigger CI.

[SunMarc]

There a two kinds of checkpoints:

• one that can we reloaded with from_pretrained
• one that we use for resuming the training

So, this is normal that the model is saved twice. However, it might make sense to only save the checkpoint with from_pretrained compatibility in the last checkpoint by default.

cc @S1ro1