Disable jitter noise during evaluation in SwitchTransformers

[DaizeDong]

What does this PR do?

The jitter noise was mistakenly added during evaluation in GPTSanJapanese and SwitchTransformers, which would bring uncertainty to the evaluation results. Now the bug is fixed, and the implementation is the same as the native code of switch transformers.

The former implementation is:

if self.jitter_noise > 0:
    # Multiply the token inputs by the uniform distribution - adding some noise
    hidden_states *= torch.empty_like(hidden_states).uniform_(1.0 - self.jitter_noise, 1.0 + self.jitter_noise)

The fixed implementation is:

if self.training and self.jitter_noise > 0:
    # Multiply the token inputs by the uniform distribution - adding some noise
    hidden_states *= torch.empty_like(hidden_states).uniform_(1.0 - self.jitter_noise, 1.0 + self.jitter_noise)

This PR also updates the outdated annotations in configuration_switch_transformers.py. Now the default values in the annotation area are the same as the values in the __init__ for SwitchTransformersConfig.

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.

[amyeroberts]

Thanks for fixing and for the attention to detail updating the docstring too!

[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.