Fix test failure on DeepSpeed

[muellerzr]

What does this PR do?

It looks like grad_norm might not always be a torch.Tensor, as that merged PR breaks all of our deepspeed trainer tests. In such cases, we can check for item before calling it. Follow-up to #29212, linked to #27326

Fixes failing DeepSpeed tests

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.

@ArthurZucker @pacman100

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[shubhanjan99]

Fixes the issue mentioned in #29207 , but I haven't checked whether #29212 fixed it or not for me.

[ArthurZucker]

if it is not a float, what is it that is returned? A list ?

[muellerzr]

A tensor, actually

[lucasjinreal]

When will update to pypi?

this one break training on deepspeed now.

[ArthurZucker]

This was released in https://github.com/huggingface/transformers/releases/tag/v4.39.0