Fix missing diff[i] in legacy contrastive loss GPU backward#7100
Open
Chessing234 wants to merge 1 commit intoBVLC:masterfrom
Open
Fix missing diff[i] in legacy contrastive loss GPU backward#7100Chessing234 wants to merge 1 commit intoBVLC:masterfrom
Chessing234 wants to merge 1 commit intoBVLC:masterfrom
Conversation
The CPU backward multiplies beta by diff via caffe_cpu_axpby, giving gradient = -alpha * diff[i] for legacy dissimilar pairs. The GPU kernel sets bottom_diff[i] = -alpha directly, omitting the diff[i] factor. The non-legacy GPU path correctly includes * diff[i] in beta. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
The GPU backward kernel (
CLLBackward) forContrastiveLossLayerproduces wrong gradients whenlegacy_version=truefor dissimilar pairs.Root cause: The CPU backward computes the gradient via
caffe_cpu_axpby:The GPU kernel sets
bottom_diff[i] = beta = -alphadirectly, omitting the multiplication bydiff[i]:The non-legacy GPU path correctly absorbs
diff[i]intobeta:Fix: Change
beta = -alphatobeta = -alpha * diff[i]to match the CPU implementation.Test plan
legacy_version=trueon GPU🤖 Generated with Claude Code