[Bug] Reject negative training.max_norm - #4295
Merged
tianyu-l merged 1 commit intoAug 24, 2026
Merged
Conversation
taking-lying-flat
requested review from
fegin,
tianyu-l,
wconstab and
wwwjn
as code owners
August 23, 2026 08:34
meta-cla
Bot
added
the
CLA Signed
tianyu-l
approved these changes
Aug 24, 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
TrainingConfig.max_normduring configuration construction.0as a valid boundary value.Why
Trainerforwardstraining.max_normdirectly to gradient clipping. A negative value does not disable clipping: the clipping implementation computes a scale frommax_norm / total_norm, so a negative maximum produces a negative scale and can reverse every gradient direction before the optimizer step.Rejecting the value in
TrainingConfig.__post_init__surfaces the mistake during configuration parsing, before model initialization or training begins. The check uses< 0intentionally somax_norm=0remains supported and deterministically zeros gradients.