Fix CUDA illegal memory access bug in monotonic_rnnt#14
Merged
Conversation
SimBe195
requested review from
DanEnergetics,
JackTemaki,
Stefanwuu and
curufinwe
Stefanwuu
approved these changes
Feb 2, 2026
albertz
approved these changes
Feb 2, 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
3 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
The size of the gradient tensor in monotonic RNN-T loss computation is essentially
B * T * (S+1) * V. For larger vocabulary sizes and sequence lengths, this size can overflow the signed 32-bit integer limit. In the current implementation of the gradient CUDA kernel, the index for writing into the gradient tensor (grads[bts * *V + v] = ...) has a datatype ofint, so such an overflow leads to a negative index and thus an illegal memory access error. Changing the datatype toint64_tfixes the issue.