Fix deepspeed for PPOv2Trainer

[qgallouedec]

What does this PR do?

Fixes #2025

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? 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.
• 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.

[qgallouedec]

@danaaubakirova can you check it solves the issue?
cc @RylanSchaeffer

[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.

[danaaubakirova]

@qgallouedec yes, It works for me! Thanks a lot!

[lewtun]

Thanks for the fix! I wasn't aware the self.deepspeed was needed, but I guess this reflects the fact we are relying heavily on the Trainer internals.