fix: clamp VQ indices to prevent CUDA out-of-bounds errors#86
Open
xander1421 wants to merge 1 commit intoHeartMuLa:mainfrom
Open
fix: clamp VQ indices to prevent CUDA out-of-bounds errors#86xander1421 wants to merge 1 commit intoHeartMuLa:mainfrom
xander1421 wants to merge 1 commit intoHeartMuLa:mainfrom
Conversation
When using vector-quantize-pytorch >= 1.20 with PyTorch 2.10+ and Python 3.14, the codebook indices from HeartMuLa can occasionally exceed the valid range [0, codebook_size-1], causing CUDA device-side assertion failures in the einx.get_at indexing operation. This fix clamps the indices to the valid codebook range before passing them to ResidualVQ.get_output_from_indices(), preventing the CUDA errors while preserving the generated audio quality. Tested with: - Python 3.14 - PyTorch 2.10.0+cu128 - vector-quantize-pytorch 1.27.15 - CUDA 12.8 Co-Authored-By: Claude Opus 4.5 <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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
[0, codebook_size-1]before passing toResidualVQ.get_output_from_indices()vector-quantize-pytorchProblem
When using
vector-quantize-pytorch >= 1.20with PyTorch 2.10+ and Python 3.14, the codebook indices from HeartMuLa can occasionally exceed the valid codebook range, causing:The error occurs in
einx.get_at()when indexing into the codebook with out-of-bounds values.Solution
Added bounds checking to clamp indices before the VQ lookup:
Test Environment
Test plan
🤖 Generated with Claude Code