Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix RNN-T loss memory usage #11144

Merged
merged 4 commits into from
Nov 4, 2024
Merged

Fix RNN-T loss memory usage #11144

merged 4 commits into from
Nov 4, 2024

Conversation

artbataev
Copy link
Collaborator

@artbataev artbataev commented Nov 4, 2024

What does this PR do ?

Fixes memory usage for Numba-based implementation of RNN-T and Multi-blank Transducer losses.
The current implementation requires 3x memory compared to the size of logits (logits, gradient, extra memory = size of logits). This PR fixes memory usage to the minimal possible 2x (logits, gradient).

It looks like assigning tensors directly to ctx instead of saving them through save_for_backward breaks PyTorch logic, and it copies the gradient tensor (which results in extra memory usage).

TDT loss was not affected by this issue (I'm unsure why, but it requires contiguous tensor for labels-related logits).

Before (main):

  • Peak memory before loss: 2.09 GB
  • Peak memory after loss: 6.27 GB

After (this PR):

  • Peak memory before loss: 2.09 GB
  • Peak memory after loss: 4.18 GB

Code to check memory usage (size of tensors besides logits is negligible compared to logits):

import torch
from nemo.collections.asr.parts.numba.rnnt_loss import RNNTLossNumba
device = torch.device("cuda")

loss = RNNTLossNumba(blank=1023, reduction='none')
batch_size = 32
logits = torch.rand([batch_size, 188, 90 + 1, 1024], device=device, dtype=torch.float32, requires_grad=True)
encoder_lengths = torch.full([batch_size], fill_value=188, device=device, dtype=torch.long)
label_lengths = torch.full([batch_size], fill_value=90, device=device, dtype=torch.long)
labels = torch.randint(0, 1022, size=[batch_size, 90], dtype=torch.long, device=device)
print(f"Peak memory before loss: {torch.cuda.max_memory_allocated() / (2 ** 30):.2f} GB")

loss_value = loss(acts=logits, act_lens=encoder_lengths, labels=labels, label_lens=label_lengths)
loss_value.sum().backward()
print(f"Peak memory after loss: {torch.cuda.max_memory_allocated() / (2 ** 30):.2f} GB")

Collection: [ASR]

Changelog

  • In transducer-related functions that extend autograd, save tensors using ctx.save_for_backward(...) instead of directly assigning tensors according to PyTorch documentation.

Usage

  • You can potentially add a usage example below
# Add a code snippet demonstrating how to use this 

GitHub Actions CI

The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.

The GitHub Actions CI will run automatically when the "Run CICD" label is added to the PR.
To re-run CI remove and add the label again.
To run CI on an untrusted fork, a NeMo user with write access must first click "Approve and run".

Before your PR is "Ready for review"

Pre checks:

  • Make sure you read and followed Contributor guidelines
  • Did you write any new necessary tests?
  • Did you add or update any necessary documentation?
  • Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
    • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

  • New Feature
  • Bugfix
  • Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

  • Related to # (issue)

Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
@github-actions github-actions bot added the ASR label Nov 4, 2024
artbataev and others added 2 commits November 4, 2024 15:53
Signed-off-by: artbataev <artbataev@users.noreply.github.com>
Copy link
Collaborator

@hainan-xv hainan-xv left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Great discovery of the memory usage issue and a very clean fix! Approved and thanks.

Copy link
Contributor

github-actions bot commented Nov 4, 2024

[🤖]: Hi @artbataev 👋,

We wanted to let you know that a CICD pipeline for this PR just finished successfully

So it might be time to merge this PR or get some approvals

I'm just a bot so I'll leave it you what to do next.

//cc @pablo-garay @ko3n1g

Copy link
Collaborator

@titu1994 titu1994 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the fix !

@artbataev artbataev merged commit d19e9d3 into main Nov 4, 2024
162 of 163 checks passed
@artbataev artbataev deleted the fix_rnnt_memory_usage branch November 4, 2024 19:09
hainan-xv pushed a commit to hainan-xv/NeMo that referenced this pull request Nov 5, 2024
* Fix RNN-T memory usage

Signed-off-by: artbataev <artbataev@users.noreply.github.com>

---------

Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
Signed-off-by: Hainan Xu <hainanx@nvidia.com>
lilyw97 pushed a commit to lilyw97/NeMo that referenced this pull request Nov 13, 2024
* Fix RNN-T memory usage

Signed-off-by: artbataev <artbataev@users.noreply.github.com>

---------

Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
HuiyingLi pushed a commit to HuiyingLi/NeMo that referenced this pull request Nov 15, 2024
* Fix RNN-T memory usage

Signed-off-by: artbataev <artbataev@users.noreply.github.com>

---------

Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
yashaswikarnati pushed a commit that referenced this pull request Nov 21, 2024
* Fix RNN-T memory usage

Signed-off-by: artbataev <artbataev@users.noreply.github.com>

---------

Signed-off-by: Vladimir Bataev <vbataev@nvidia.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants