Skip to content

Replace ModuleSpec with Protocols for LayerNorm submodules#3090

Merged
chtruong814 merged 12 commits into
NVIDIA:mainfrom
nschank:norms
Feb 7, 2026
Merged

Replace ModuleSpec with Protocols for LayerNorm submodules#3090
chtruong814 merged 12 commits into
NVIDIA:mainfrom
nschank:norms

Conversation

@nschank

@nschank nschank commented Jan 27, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

What does this PR do ?

Creates a unified Protocol representing a LayerNorm and uses it throughout.

Associated design doc: Typed ModuleSpec.pdf

Contribution process

flowchart LR
    A[Pre-checks] --> B[PR Tests]
    subgraph Code Review/Approval
        C1[Expert Review] --> C2[Final Review]
    end
    B --> C1
    C2 --> D[Merge]
Loading

Pre-checks

  • I want this PR in a versioned release and have added the appropriate Milestone (e.g., Core 0.8)
  • I have added relevant unit tests
  • I have added relevant functional tests
  • I have added proper typing to my code Typing guidelines
  • I have added relevant documentation
  • I have run the autoformatter.sh on my PR

Code review

The following process is enforced via the CODEOWNERS file for changes into megatron/core. For changes outside of megatron/core, it is up to the PR author whether or not to tag the Final Reviewer team.

For MRs into `main` branch

Feel free to message or comment the @mcore-oncall to help accelerate your merge into main. The less complex your PR is, the faster it will be approved and merged!

(Step 1): Add PR label Expert Review

(Step 2): Collect the expert reviewers reviews

  1. Attach the Expert Review label when your PR is ready for review.
  2. GitHub auto-assigns expert reviewers based on your changes. They will get notified and pick up your PR soon.

⚠️ Only proceed to the next step once all reviewers have approved, merge-conflict are resolved and the CI is passing.
Final Review might get declined if these requirements are not fulfilled.

(Step 3): Final Review

  1. Add Final Review label
  2. GitHub auto-assigns final reviewers based on your changes. They will get notified and pick up your PR soon.

(Optional Step 4): Cherry-pick into release branch

If this PR also needs to be merged into core_r* release branches, after this PR has been merged, select Cherry-pick to open a new PR into the release branch.

For MRs into `dev` branch The proposed review process for `dev` branch is under active discussion.

MRs are mergable after one approval by either eharper@nvidia.com or zijiey@nvidia.com.

Merging your PR

Any member of core-adlr and core-nemo will be able to merge your PR.

@nschank nschank requested review from a team as code owners January 27, 2026 00:25
@copy-pr-bot

copy-pr-bot Bot commented Jan 27, 2026

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@ko3n1g ko3n1g requested a review from a team January 27, 2026 00:26
@maanug-nv maanug-nv added enhancement New feature or request Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review. complexity: medium labels Jan 27, 2026
@maanug-nv

maanug-nv commented Jan 27, 2026

Copy link
Copy Markdown
Contributor

/ok to test caa1bdc

@ko3n1g ko3n1g added this to the Core 0.16 milestone Jan 27, 2026
@yashaswikarnati

yashaswikarnati commented Jan 29, 2026

Copy link
Copy Markdown
Contributor

lgtm; just had a general q (its not the scope of this particular PR for sure),

looks like for lot of submodules the builder class can be identity or None or no defaults.. not sure if we should make this more coherent at some point.

@nschank

nschank commented Jan 29, 2026

Copy link
Copy Markdown
Contributor Author

Removed the misplaced assert that was causing the test failures!

@nschank

nschank commented Jan 30, 2026

Copy link
Copy Markdown
Contributor Author

lgtm; just had a general q (its not the scope of this particular PR for sure),

looks like for lot of submodules the builder class can be identity or None or no defaults.. not sure if we should make this more coherent at some point.

Oh yeah definitely, I've been noticing the same thing! "No default" is sensible in a lot of cases (for things that are effectively required), but IdentityOp vs None we should decide on

@yashaswikarnati

yashaswikarnati commented Jan 30, 2026

Copy link
Copy Markdown
Contributor

/ok to test f83cef5

@yaox12 yaox12 enabled auto-merge February 6, 2026 02:09
@yaox12

yaox12 commented Feb 6, 2026

Copy link
Copy Markdown
Member

/ok to test a9c3e1b

@yaox12 yaox12 added this pull request to the merge queue Feb 6, 2026
@github-merge-queue github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Feb 6, 2026
@chtruong814

chtruong814 commented Feb 6, 2026

Copy link
Copy Markdown
Contributor

/ok to test 1db71cb

@chtruong814 chtruong814 added the needs-follow-up Issue needs follow-up label Feb 6, 2026
@chtruong814

chtruong814 commented Feb 7, 2026

Copy link
Copy Markdown
Contributor

CI passed earlier here
https://github.com/NVIDIA/Megatron-LM/actions/runs/21736072386/job/62719070351

The CI retriggered after merging main again but that was unnecessary. Will go ahead and merge this since it passed PR CI already and has approvals.

@chtruong814 chtruong814 merged commit f68c7c1 into NVIDIA:main Feb 7, 2026
36 checks passed
@chtruong814 chtruong814 removed the needs-follow-up Issue needs follow-up label Feb 7, 2026
@nschank nschank deleted the norms branch February 13, 2026 19:51
daiyaanarfeen pushed a commit to daiyaanarfeen/Megatron-LM that referenced this pull request Feb 23, 2026
@nschank @yaox12
@chtruong814 @daiyaanarfeen
Replace ModuleSpec with Protocols for LayerNorm submodules (NVIDIA#3090)
a6f14e1 
Co-authored-by: Xin Yao <xiny@nvidia.com>
Co-authored-by: Charlie Truong <chtruong@nvidia.com>
BoxiangW pushed a commit to BoxiangW/Megatron-LM that referenced this pull request Mar 4, 2026
@nschank @yaox12
@chtruong814 @BoxiangW
Replace ModuleSpec with Protocols for LayerNorm submodules (NVIDIA#3090)
cf34c99 
Co-authored-by: Xin Yao <xiny@nvidia.com>
Co-authored-by: Charlie Truong <chtruong@nvidia.com>
@sbhavani sbhavani mentioned this pull request Mar 23, 2026
Open
yangbofun pushed a commit to xlm-research/Megatron-LM that referenced this pull request May 22, 2026
@nschank @yaox12 @chtruong814
Replace ModuleSpec with Protocols for LayerNorm submodules (NVIDIA#3090)
032fae4 
Co-authored-by: Xin Yao <xiny@nvidia.com>
Co-authored-by: Charlie Truong <chtruong@nvidia.com>
@sbhavani sbhavani mentioned this pull request May 26, 2026
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@jaredcasper jaredcasper jaredcasper approved these changes

@AAnoosheh AAnoosheh AAnoosheh approved these changes

@ericharper ericharper ericharper approved these changes

@duncanriach duncanriach duncanriach approved these changes

@yashaswikarnati yashaswikarnati yashaswikarnati approved these changes

@santhnm2 santhnm2 Awaiting requested review from santhnm2

Assignees

No one assigned

Labels

community-request complexity: medium enhancement New feature or request Expert Review [deprecated] Apply this label to indicate that your PR is ready for expert review.

Projects

None yet

Milestone

Core 0.16

Development

Successfully merging this pull request may close these issues.

10 participants

@nschank @maanug-nv @yashaswikarnati @yaox12 @chtruong814 @jaredcasper @AAnoosheh @ericharper @duncanriach @ko3n1g