Add with torch.no_grad() to DistilBERT integration test forward pass

[jaketae]

What does this PR do?

This PR encapsulates forward passes in DistilBERT unit tests with with torch.no_grad(): as per #14642.

Also, a unit test seemed to terminate early due to a return in a loop; this was replaced with continue. (If this was not what was intended, I will rollback the modification.)

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?

@NielsRogge @LysandreJik @sgugger

[sgugger]

I'm not sure those checks were meant under the scope of the issue. I think it was only for the last integration test, but I'll let @LysandreJik confirm.

[LysandreJik]

The request was indeed meant only for the integration tests :)

[jaketae]

Thank you for the feedback and confirmation! I've reverted the change.

[LysandreJik]

Hello @jaketae, were your last changes committed? I still see a lot of non-integration tests with torch.no_grad

[jaketae]

@LysandreJik Apologies for the delay! I thought I had properly committed and pushed the changes but seems like that wasn't the case. Reverted most of the no_grad context and modified only the integration test. Thank you!

[sgugger]

Thanks a lot!