Fix #17893, removed dead code

[clefourrier]

What does this PR do?

Fixes #17893

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?

@ydshieh

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[ydshieh]

LGTM, thank you, @clefourrier !

I think we can also remove the line _keys_to_ignore_on_load_missing = [r"position_ids"] (line 1395), but let's wait for @sgugger to confirm.

[LysandreJik]

Hey @clefourrier! The code quality error comes from a new release from black. Rebasing on main should solve the issue as you'll benefit from #17918.

[ydshieh]

Regarding the test

from Lysandre on Slack

There was a new release from black that has a slightly different behavior for the --preview flag that we use in the CI.

If you see failures in the CI for the code quality test, with a large number of file changes (>500), please mention to the author that they just need to rebase on/merge main in order to benefit from the fix.

[clefourrier]

@LysandreJik @ydshieh Should be good now! 😃
Ty both, I had missed it on the slack

[sgugger]

I'll defer to @patrickvonplaten who knows the model best. Changes look good to me.

One last thing to change is the line

_keys_to_ignore_on_load_missing = [r"position_ids"]

later on in the pretrained model, which should be now

_keys_to_ignore_on_load_unexpected = [r"position_ids"]

Just in case there are some longformer checkpoints in the wild with a position_ids key in their state dict.

[clefourrier]

@sgugger Done :)

[patrickvonplaten]

Thanks for the fix! Great catch :-)