Expound error on recursively_apply

[muellerzr]

Expands the error a bit more in recursively_apply to be more straightforward. Example:

Traceback (most recent call last):
  File "/home/zach_mueller_huggingface_co/test_distrib.py", line 6, in <module>
    gathered_tensor, i = accelerator.gather((process_tensor, 1))
  File "/home/zach_mueller_huggingface_co/accelerate/src/accelerate/accelerator.py", line 1869, in gather
    return gather(tensor)
  File "/home/zach_mueller_huggingface_co/accelerate/src/accelerate/utils/operations.py", line 243, in gather
    return _gpu_gather(tensor)
  File "/home/zach_mueller_huggingface_co/accelerate/src/accelerate/utils/operations.py", line 223, in _gpu_gather
    return recursively_apply(_gpu_gather_one, tensor, error_on_other_type=True)
  File "/home/zach_mueller_huggingface_co/accelerate/src/accelerate/utils/operations.py", line 97, in recursively_apply
    return honor_type(
  File "/home/zach_mueller_huggingface_co/accelerate/src/accelerate/utils/operations.py", line 71, in honor_type
    return type(obj)(generator)
  File "/home/zach_mueller_huggingface_co/accelerate/src/accelerate/utils/operations.py", line 100, in <genexpr>
    recursively_apply(
  File "/home/zach_mueller_huggingface_co/accelerate/src/accelerate/utils/operations.py", line 118, in recursively_apply
    raise TypeError(
TypeError: Unsupported types (<class 'int'>) passed to `_gpu_gather_one`. Only nested list/tuple/dicts of objects that are valid for `is_torch_tensor` should be passed.

cc @stas00

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[sgugger]

LGTM, thanks!

[stas00]

Thank you for making the error more user-friendly, Zach.