Revert DeepSpeed stuff

[muellerzr]

What does this PR do?

During the integration some deepspeed items that need further looks into before it can be part of the integration need to be addressed/looked at more carefully. This PR reverts the base deepspeed logic done in setup_devices and parallel_mode to restore the original deepspeed behavior

Fixes # (issue)

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.

@sgugger, cc @pacman100 so you're aware.

[sgugger]

This makes me sad, but if you really need to...

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint.

[jli]

I see this was included in transformers 4.29.0 (https://github.com/huggingface/transformers/releases/tag/v4.29.0). Could you share more about how this changes the Transformers + DeepSpeed integration? I don't quite understand the diff. Does this disable some deeper level of integration of DS with Transformers?

[muellerzr]

@jli this pr just reverted a small portion of Accelerate handling the deepspeed part when we weren't ready for that yet. CC @pacman100 if you could explain accelerates deepspeed integration vs the transformers one were replacing in terms of features? :)