Remove the FP32 Wrapper when evaluating

[lmxyy]

What does this PR do?

When using BF16 for training, the script raises the error of RuntimeError: Input type (float) and bias type (c10::BFloat16) should be the same due to accelerator's FP32 wrapper. This wrapper needs to be removed for evaluation.

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?
• Did you read our philosophy doc (important for complex PRs)?
• 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.
@sayakpaul

Integrations:

• deepspeed: HF Trainer/Accelerate: @SunMarc
• PEFT: @sayakpaul @BenjaminBossan

HF projects:

• accelerate: different repo
• datasets: different repo
• transformers: different repo
• safetensors: different repo

-->

[sayakpaul]

Thanks!

[linoytsaban]

Nice!

[SunMarc]

LGTM !