feat: add support for tensor parallel training workflow with accelerate

[kmehant]

What does this PR do?

1. Add apply_tensor_parallel API to apply TP plan to Llama and Granite models Already merged in to HF/transformers.
2. Introduce tp_size user facing argument to be further consumed by accelerate (see huggingface/accelerate#3173)
3. Allows for e2e TP training.

Please review in conjunction with huggingface/accelerate#3173

Fixes #32470

Results

See significant improvement in both memory and throughput compared against single gpu training, and FSDP across different settings (checkpointing on/off) and context lengths.

Note: Please be aware that the effective TPS for FSDP would be multiplicative of the parallel factor (number of GPUs/devices engaged in distributed training) whereas that is not the case with TP. Therefore, when effective throughput is considered we can find FSDP is better than TP in terms of throughput. However, that may be compensated by increasing the batch size utilizing the memory gains etc.

Done on two models

1. ibm-granite/granite-8b-code-base-128k
2. codellama/CodeLlama-7b-hf

Tables below show the max cuda memory and throughput for various configurations showing the potential of TP contributed in this PR. There is gains in both memory and throughput.

Model	Method	# of GPUs	Context Length	Batch Size	Grad Checkpointing	Cuda Max Mem (GiB)	Tokens/Sec/GPU
ibm-granite/granite-8b-code-base-128k	Single GPU non-distributed	1	8192	1	FALSE	OOM	NA
ibm-granite/granite-8b-code-base-128k	FSDP	4	8192	1	FALSE	OOM	NA
ibm-granite/granite-8b-code-base-128k	TP (This PR)	4	8192	1	FALSE	52.4	7675.4

Model	Method	# of GPUs	Context Length	Batch Size	Grad Checkpointing	Cuda Max Mem (GiB)	Tokens/Sec/GPU
ibm-granite/granite-8b-code-base-128k	Single GPU non-distributed	1	8192	1	TRUE	OOM	NA
ibm-granite/granite-8b-code-base-128k	FSDP	4	8192	1	TRUE	29.975586	2256.896
ibm-granite/granite-8b-code-base-128k	TP (This PR)	4	8192	1	TRUE	26.5	5935.5

Model	Method	# of GPUs	Context Length	Batch Size	Grad Checkpointing	Cuda Max Mem (GiB)	Tokens/Sec/GPU
ibm-granite/granite-8b-code-base-128k	Single GPU non-distributed	1	16384	1	FALSE	OOM	NA
ibm-granite/granite-8b-code-base-128k	FSDP	4	16384	1	FALSE	OOM	NA
ibm-granite/granite-8b-code-base-128k	TP (This PR)	4	16384	1	FALSE	OOM	NA

Model	Method	# of GPUs	Context Length	Batch Size	Grad Checkpointing	Cuda Max Mem (GiB)	Tokens/Sec/GPU
ibm-granite/granite-8b-code-base-128k	Single GPU non-distributed	1	16384	1	TRUE	OOM	NA
ibm-granite/granite-8b-code-base-128k	FSDP	4	16384	1	TRUE	36.8	2084.864
ibm-granite/granite-8b-code-base-128k	TP (This PR)	4	16384	1	TRUE	33.5	5692.5

Model	Method	# of GPUs	Context Length	Batch Size	Grad Checkpointing	Cuda Max Mem (GiB)	Tokens/Sec/GPU
codellama/CodeLlama-7b-hf	Single GPU non-distributed	1	8192	1	FALSE	OOM	NA
codellama/CodeLlama-7b-hf	FSDP	4	8192	1	FALSE	70.7	3560
codellama/CodeLlama-7b-hf	TP (This PR)	4	8192	1	FALSE	42.8	9216

Model	Method	# of GPUs	Context Length	Batch Size	Grad Checkpointing	Cuda Max Mem (GiB)	Tokens/Sec/GPU
codellama/CodeLlama-7b-hf	Single GPU non-distributed	1	8192	1	TRUE	75.3	2849
codellama/CodeLlama-7b-hf	FSDP	4	8192	1	TRUE	26.4	5957
codellama/CodeLlama-7b-hf	TP (This PR)	4	8192	1	TRUE	21.4	7125

Model	Method	# of GPUs	Context Length	Batch Size	Grad Checkpointing	Cuda Max Mem (GiB)	Tokens/Sec/GPU
codellama/CodeLlama-7b-hf	Single GPU non-distributed	1	16384	1	FALSE	OOM	NA
codellama/CodeLlama-7b-hf	FSDP	4	16384	1	FALSE	OOM	NA
codellama/CodeLlama-7b-hf	TP (This PR)	4	16384	1	FALSE	OOM	NA

Model	Method	# of GPUs	Context Length	Batch Size	Grad Checkpointing	Cuda Max Mem (GiB)	Tokens/Sec/GPU
codellama/CodeLlama-7b-hf	Single GPU non-distributed	1	16384	1	TRUE	75.3	2599
codellama/CodeLlama-7b-hf	FSDP	4	16384	1	TRUE	30.1	2433
codellama/CodeLlama-7b-hf	TP (This PR)	4	16384	1	TRUE	26.6	6873

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.

• trainer: @muellerzr and @SunMarc
• @SeungyounShin (happy to include as co author - [WIP][Need help and discussion] : basic llama tensor parallel #32597)

I have cycles to bring in more improvements over this PR to bring in Pytorch TP support to HF. Looking forward. Thank you

HF projects:

• accelerate: different repo

[kwen2501]

Such timing! I have similar thought here. Shall we collaborate?

[kmehant]

@kwen2501 Absolutely, please let me know, how you want to take this forward. Thank you.

[kmehant]

@ArthurZucker @muellerzr since accelerate PR (huggingface/accelerate#3173) is merged. Requesting review and merge of this PR which would allow for complete e2e training workflow using tensor parallelism. Thank you.

[SunMarc]

SGTM ! Could you also update the docs (Efficient Training on Multiple GPUs doc / trainer doc ) ?

[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.