Skip to content

Refine layerwise non-mutating calibration#1592

Draft
realAsma wants to merge 2 commits into
fridah/layerwise-configfrom
asma/layerwise_skip_in_meta
Draft

Refine layerwise non-mutating calibration#1592
realAsma wants to merge 2 commits into
fridah/layerwise-configfrom
asma/layerwise_skip_in_meta

Conversation

@realAsma
Copy link
Copy Markdown
Contributor

@realAsma realAsma commented Jun 1, 2026

Stacked on #1571.

Summary:

  • Add calib_mutates_weights gating for non-mutating layerwise calibration.
  • Skip layer weight checkpoint/writeback for quantizer-state-only calibration.
  • Keep FSDP2/Accelerate writeback conditional and improve layerwise progress reporting.

Testing:

  • Pre-commit hooks passed during commit.

realAsma added 2 commits June 1, 2026 21:41
Signed-off-by: realAsma <akuriparambi@nvidia.com>
Signed-off-by: realAsma <akuriparambi@nvidia.com>
@copy-pr-bot
Copy link
Copy Markdown

copy-pr-bot Bot commented Jun 1, 2026

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

@coderabbitai
Copy link
Copy Markdown
Contributor

coderabbitai Bot commented Jun 1, 2026

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 8f451a49-78bc-4bd3-b5b5-3f7cdf665250

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

  • 🔍 Trigger review
✨ Finishing Touches
🧪 Generate unit tests (beta)
  • Create PR with unit tests
  • Commit unit tests in branch asma/layerwise_skip_in_meta

Comment @coderabbitai help to get the list of available commands and usage tips.

Comment on lines +507 to +511
if not writeback:
with _fsdp2_unshard_context(fsdp_module):
yield
return

Copy link
Copy Markdown
Contributor Author

@realAsma realAsma Jun 1, 2026

Choose a reason for hiding this comment

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

@sugunav14 here is an easy perf improvement for layerwise FSDP2

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant