Add Parallelism getter property to Accelerator class

[WoosungMyung]

This PR adds a @Property method for simply get parallelism rank in HF Accelerate.

Motivation:

• To make it easier to retrieve the process rank in Distributed Learning using Accelerate.

Changes:

• Added rank property in accelerator.py

This PR is from issue #3702 with S1ro1.
Let me know if there’s anything I should improve.

Thanks a lot for giving opportunity for this amazing PJ!

[S1ro1]

I think we want to adapt this a little, sorry for wrong assumption yesterday. I'd say the best thing UX wise is: return the rank if enabled, return 0 if parallelism config is enabled but no parallelism {x} is enabled (in this case all ranks are esentially 0) and raise a RuntimeError if neither of these 2 met. Do you think that is reasonable?

[WoosungMyung]

@S1ro1
Thanks for your fast feedback! I think it makes sense that parallelism configuration is set & parallelism is not enable, it should return rank 0 for all process. I'd appreciate it if you could take a look and let me know your thoughts.

[S1ro1]

Thank you, LGTM now. Congratulation on your first contribution! Just fixed style checks for you and will merge!

[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.

[WoosungMyung]

@S1ro1
Thank you so much for open-minded feedback throughout my first PR. I really appreciate the constructive discussions and your welcoming attitude.

[S1ro1]

@S1ro1
Thank you so much for open-minded feedback throughout my first PR. I really appreciate the constructive discussions and your welcoming attitude.

We thank you for the contribution and look forward to more 🤗 Pleasure to be of help!