Skip to content

feat: RGS for wanda++ #2537

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

Open
wants to merge 5 commits into
base: main
Choose a base branch
from
Open

Conversation

namgyu-youn
Copy link
Contributor

@namgyu-youn namgyu-youn commented Jul 13, 2025

This update introduces one of the main concepts of Wanda++ (arXiv), Regional Gradient Score (RGS).

Algorithm : RGS

RGS uses block-level gradients instead of magnitude (Wanda):

$$ RGS \ : \ S_ij = (α * G_ij + ||X_j||_2) * |W_ij| $$

  • G_ij : Regional gradient
  • ||X_j||_2 : Activation norm (same in Wanda)
  • |W_ij| : Weight magnitude (same in Wanda)
  • α : Regional gradient weight (default: 100 from paper)

Algorithm : Wanda++

  • Collect Calibration : Save input samples in block-level
  • Compute Regional Gradient Score (RGS)
  • Apply sparsity (pruning)

Copy link

pytorch-bot bot commented Jul 13, 2025

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/2537

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

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

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 Jul 13, 2025
@namgyu-youn
Copy link
Contributor Author

@pytorchbot label "sparsity"

- Since WandaPlusPlusSparsifier is still in prototype stage, we move it to the prototype directory.
@namgyu-youn
Copy link
Contributor Author

@pytorchbot label "topic: not user facing"

@pytorch-bot pytorch-bot bot added the topic: not user facing Use this tag if you don't want this PR to show up in release notes label Jul 21, 2025
@namgyu-youn namgyu-youn marked this pull request as draft July 22, 2025 06:56
- Since WandaPlusPlusSparsifier is still in prototype stage, we move it to the prototype directory.

update annotations and remove rebudnat error handler

update indexing logics for Wanda++ using Wanda
- Since both paper uses same index logics (extraction, sorting, and updating), Wanda++ can use the same logic as Wanda.
- For using it, `_apply_sparsity_pattern` is moduled to be used in both Wanda and Wanda++.

remove hardware-specific test case
- This test case is not fully implemented in `wanda_pp.py`
@namgyu-youn namgyu-youn marked this pull request as ready for review July 22, 2025 07:57
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. sparsity topic: not user facing Use this tag if you don't want this PR to show up in release notes
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants