🚨🚨🚨 The Great Deprecation 🚨🚨🚨

[muellerzr]

What does this PR do?

This PR applies all deprecations for 1.0.0:

• Users now must make a DataLoaderConfiguration to set items like even_batches, use_seedable_sampler, etc
• use_fp16 properties are now gone
• Accelerator.autocast no longer accepts a cache_enabled arg, users must make an AutocastKwargs instance and pass it in as a kwarg handler
• fsdp_backward_prefetch_policy is replaced with fsdp_backward_prefetch
• AcceleratedOptimizer.is_overflow should be replaced with AcceleratedOptimizer.step_was_skipped
• is_tpu_available should be replaced with is_torch_xla_available
• ACCELERATE_DISABLE_RICH is no longer a valid env variable, one should manually enable rich traceback via ACCELERATE_ENABLE_RICH
• utils.modeling.shard_checkpoint should be replaced with split_torch_state_dict_into_shards from the huggingface_hub library
• accelerate's tqdm wrapper no longer takes True/False as the first argument, one should manually pass in main_process_only as a kwarg instead

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.

@BenjaminBossan
@SunMarc

[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.

[BenjaminBossan]

+4 −247

🤩

Looks good, thanks for addressing these (sometimes long overdue) deprecations.

I also did a check on the code base to see if anything was missing and found these potential candidates, WDYT?

1. accelerate/src/accelerate/utils/modeling.py

   Line 562 in d5b7b70

   # TODO: at the next Transformers release (4.28.0) issue a deprecation warning here.

2. https://github.com/huggingface/accelerate/blob/main/src/accelerate/memory_utils.py

3. accelerate/src/accelerate/utils/dataclasses.py

   Line 545 in d5b7b70

   - **TPU** -- This field will be deprecated in v0.27.0. Use XLA instead.

[SunMarc]

Thanks for taking care of that @muellerzr ! LGTM !