[ET-VK] Fused RMSNorm operator to fix fp16 overflow#18786
Merged
Conversation
Pull Request resolved: #18772 Fused RMSNorm operator that performs squaring, mean, rsqrt, and weight scaling in a single shader dispatch. All accumulation is done in fp32 regardless of input dtype, preventing fp16 overflow when residual stream values exceed sqrt(65504) ≈ 256. The Python reference impl (`rms_norm_impl`) must preserve the input dtype — PyTorch type promotion would otherwise produce fp32 output from fp16 inputs, and the FusePatternsPass re-trace would propagate that incorrect dtype through the graph. Authored by Claude. ghstack-source-id: 364514329 @exported-using-ghexport Differential Revision: [D99841211](https://our.internmc.facebook.com/intern/diff/D99841211/)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/18786
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 34 PendingAs of commit ae17fd1 with merge base a231fbc ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
This PR needs a
|
abhinaykukkadapu
approved these changes
Apr 9, 2026
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Pull Request resolved: #18772
Fused RMSNorm operator that performs squaring, mean, rsqrt, and weight scaling in a single shader dispatch. All accumulation is done in fp32 regardless of input dtype, preventing fp16 overflow when residual stream values exceed sqrt(65504) ≈ 256.
The Python reference impl (
rms_norm_impl) must preserve the input dtype — PyTorch type promotion would otherwise produce fp32 output from fp16 inputs, and the FusePatternsPass re-trace would propagate that incorrect dtype through the graph.Authored by Claude.
ghstack-source-id: 364514329
@exported-using-ghexport
Differential Revision: D99841211
Summary
[PLEASE REMOVE] See CONTRIBUTING.md's Pull Requests for ExecuTorch PR guidelines.
[PLEASE REMOVE] If this PR closes an issue, please add a
Fixes #<issue-id>line.[PLEASE REMOVE] If this PR introduces a fix or feature that should be the upcoming release notes, please add a "Release notes: " label. For a list of available release notes labels, check out CONTRIBUTING.md's Pull Requests.
Test plan
[PLEASE REMOVE] How did you test this PR? Please write down any manual commands you used and note down tests that you have written if applicable.