[cyyever]

What does this PR do?

The argument is ignore.

[Rocketknight1]

Hmm, I'm not sure about this one! If you read the source for is_torch_xla_available, check_is_gpu overrides check_is_tpu, so there's no point setting both to True

[cyyever]

@Rocketknight1 Ok, let's just set check_is_gpu

[Rocketknight1]

I'm still uncertain, though - I don't know much about neuroncore, so I don't know if those devices type-check as GPUs or TPUs or what. That makes me a bit nervous about modifying this function!

[cyyever]

@Rocketknight1 Delay it?

[ArthurZucker]

Thanks for the PR, and sorry it didn't make it in! This branch is based on v4, and v5 is very different, so rather than rebasing we'd love to invite you to open a new PR against main using our latest tools and good practices:

• Transformers tenets
• mlinter and the modeling rules it enforces (make typing)
• Modular transformers
• Contributing guide

Closing this one in the meantime, thanks again!