[WIP] Make FSDPv1 only perform cat() during last microbatch backward() within FlattenParamsWrapper

[chrisxcai]

Proposal from @GD06 : https://fburl.com/gdoc/byomncbg

sudo rm -r /tmp/chriscai/xldumps/* & CUDA_LAUNCH_BLOCKING=1 PYTHONPATH=~/benchmark/fairscale_repos/fairscale/ CRYPTOGRAPHY_OPENSSL_NO_LEGACY=1 torchrun --master_port 1024 --nproc_per_node=8 train.py --dump_dir /tmp/chriscai/xldumps --model_parallel_size 2 --pipeline_parallel_size 1 --seq_len=1024 --gpu_check_level=-1 --steps=10 --log_all_steps=True --profile_freq=10 --dump_profile_traces=True --profile_with_stack=True --use_microbatching=True --model.n_layers=1 --reshard_after_forward=False --batch_size=128 --model.efficient_attn=flash --model.attn_bias_type=causal --model=small --mem_snapshot_stop_step 5

trace before change (SplitWithSizesBackward triggered every microbatch per FSDP module):
https://fburl.com/perfdoctor/qdt32ibh

trace with applied change (SplitWithSizesBackward triggered only in last microbatch per FSDP module):
https://fburl.com/perfdoctor/mpjsoarn

[GD06]

Could you please try to split this into three logger.info entries? I think the printing of new_param failed as the new_param we used to build partial function was gone...

[facebook-github-bot]

Hi @chrisxcai!

Thank you for your pull request.

We require contributors to sign our Contributor License Agreement, and yours needs attention.

You currently have a record in our system, but the CLA is no longer valid, and will need to be resubmitted.

Process

In order for us to review and merge your suggested changes, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

Once the CLA is signed, our tooling will perform checks and validations. Afterwards, the pull request will be tagged with CLA signed. The tagging process may take up to 1 hour after signing. Please give it that time before contacting us about it.

If you have received this in error or have any questions, please contact us at [email protected]. Thanks!