Handle loading non-existent checkpoints or corrupted checkpoints.

[zhengchenyu]

What does this PR do?

• 1 Handle loading non-existent checkpoints

Setting resume_from_checkpoint to true at the start of training will result in an error because the latest checkpoint cannot be found. Therefore, we should only set to false or null at the beginning. If you then interrupt training and want to resume from the latest checkpoint, we need to set resume_from_checkpoint to true. This adjustment is unnecessary. If the latest checkpoint cannot be found at the start of training, simply print the message; there is no need to raise an exception.

• 2 Handle loading corrupted checkpoints

I've noticed that an exception during the checkpoint process can interrupt the checkpoint, resulting in corrupted files. This can cause errors when loading the checkpoint. This PR add "latest" tag to the checkpoint to ensure the checkpoint is complete.

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.

[Rocketknight1]

cc @SunMarc