Skip to content

Conversation

@Ankur-singh
Copy link
Contributor

Context

What is the purpose of this PR? Is it to

  • add a new feature
  • fix a bug
  • update tests and/or documentation
  • other (please add here)

Please link to any issues this PR addresses.
#2353

Changelog

What are the changes made in this PR?

  • Added disable_dropout function to torchtune/training/model_util.py
  • Added tests for disable_dropout function

Test plan

Please make sure to do each of the following if applicable to your PR. If you're unsure about any one of these just ask and we will happily help. We also have a contributing page for some guidance on contributing.

  • run pre-commit hooks and linters (make sure you've first installed via pre-commit install)
  • add unit tests for any new functionality
  • update docstrings for any new or updated methods or classes
  • run unit tests via pytest tests
  • run recipe tests via pytest tests -m integration_test
  • manually run any new or modified recipes with sufficient proof of correctness
  • include relevant commands and any other artifacts in this summary (pastes of loss curves, eval results, etc.)

UX

If your function changed a public API, please add a dummy example of what the user experience will look like when calling it.
Here is a docstring example
and a tutorial example

  • I did not change any public API
  • I have added an example to docs or docstrings

@pytorch-bot
Copy link

pytorch-bot bot commented Feb 11, 2025

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/2378

Note: Links to docs will display an error until the docs builds have been completed.

⏳ No Failures, 3 Pending

As of commit f89ed8f with merge base 8c9235e (image):
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Feb 11, 2025
Copy link
Member

@joecummings joecummings left a comment

Choose a reason for hiding this comment

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

High quality work!

I am uncomfortable with how many one-off files though we have for various utilities during training. Can you expose this through the training module __init__ and make the file "private" (prefix an underscore) so that I can follow up and consolidate some of these utilities?

model (torch.nn.Module): The model in which dropout layers should be disabled.
"""
for module in model.modules():
if isinstance(module, torch.nn.Dropout):
Copy link
Member

Choose a reason for hiding this comment

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

I would also check that the Dropout set to something other than zero. No need to reset Dropout to zero if it's already zero.

for module in model.modules():
if isinstance(module, torch.nn.Dropout):
warnings.warn(
f"Dropout found in {module}. This is likely to cause issues during training. Disabling."
Copy link
Member

Choose a reason for hiding this comment

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

Suggested change
f"Dropout found in {module}. This is likely to cause issues during training. Disabling."
f"Found Dropout with value {module.p} in module {module}. Setting to zero."

@Ankur-singh
Copy link
Contributor Author

@SalmanMohammadi and @joecummings I have made all the requested changes😁

Copy link
Member

@joecummings joecummings left a comment

Choose a reason for hiding this comment

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

🫡

@joecummings joecummings merged commit b9e5b1e into meta-pytorch:main Feb 11, 2025
17 checks passed
@Ankur-singh Ankur-singh deleted the disable-dropout branch March 4, 2025 07:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants