Remove skipping logic now that set_epoch exists

[muellerzr]

What does this PR do?

I'm fairly confident that this logic provided is no longer needed now that Accelerate DataLoaders have a set_epoch capability and as a result we don't need to manually skip the batches like done here for random states.

Should fix huggingface/accelerate#2709 and fix #30500

I've verified the slow tests/trainer all pass with this change on single and multi GPU, so it seems safe.

(We have tests for resuming training at certain points and asserting that results are identical)

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

[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.

[amyeroberts]

I'm not super familiar with the dataloader class. Am I correct in understanding the set_epoch will effectively do the same as the logic here: restore the state to the last trained epoch?

[muellerzr]

Correct, it'll restore the random state to where it is at a certain point.

So doing dataloader.set_epoch(2) will give you the same iteration data as the third epoch in the series.

[amyeroberts]

Thanks for this tidy up and testing in different environments!