Add QK layernorm support for dot-product attention in MambaModel#4067
Open
Phlip79 wants to merge 9 commits intoNVIDIA:mainfrom
Open
Add QK layernorm support for dot-product attention in MambaModel#4067Phlip79 wants to merge 9 commits intoNVIDIA:mainfrom
Phlip79 wants to merge 9 commits intoNVIDIA:mainfrom
Conversation
Convert static mamba_stack_spec and mamba_inference_stack_spec into config-driven functions (get_mamba_stack_spec, get_mamba_inference_stack_spec) that read qk_layernorm and qk_l2_norm from TransformerConfig, matching GPTModel's approach. Backward-compatible constants are preserved. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Member
Author
|
/claude review |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
Tests cover: default (no config), qk_layernorm=True, qk_l2_norm=True, inference spec, backward-compatible constant, and a full forward pass with qk_layernorm enabled. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Member
Author
|
/claude review |
Member
Author
|
/ok to test 367b8a8 |
Member
Author
|
/ok to test 1c46827 |
Member
Author
|
/ok to test b28a11f |
5 tasks
Contributor
|
I think this is exactly what we want to avoid; as far as I understand, we do not want to start to make the spec dynamic in code. :) |
Member
Author
|
/claude review |
Member
Author
|
/ok to test bdf2d12 |
Member
Author
|
/ok to test 488e448 |
Member
Author
|
/ok to test e9c20a4 |
TENorm uses __new__ to return a TE LayerNorm/RMSNorm instance, not a TENorm instance, so isinstance(x, TENorm) always fails. Check for not None instead. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Member
Author
|
/ok to test d16170f |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do ?
Converts static
mamba_stack_specandmamba_inference_stack_specinto config-driven functions (get_mamba_stack_spec,get_mamba_inference_stack_spec) that read qk_layernorm and qk_l2_norm from TransformerConfig, matching GPTModel's approach. Backward-compatible constants are preserved.Contribution process
Pre-checks
Code review
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!
All PRs start as draft. If you open a non-draft PR, it will be automatically converted to draft.
Step 1: Mark PR as "Ready for Review"
.github/CODEOWNERS.Final Review might get declined if these requirements are not fulfilled.
Step 2: Final Review
For PRs that change
megatron/core, once all expert reviewers have approved, theFinal Reviewlabel is applied automatically and final reviewers are assigned.For PRs outside
megatron/core, this step is skipped.Step 3: Approved
Once all required reviewers have approved, the
Approvedlabel is applied automatically.Merge
Any member of mcore-engineers will be able to merge your PR.
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.comorzijiey@nvidia.com.