[ModelRunner] Fix stop & bad word list pointer offset. #1486
Conversation
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
MartinMarciniszyn
approved these changes
May 22, 2024
Collaborator
MartinMarciniszyn
left a comment
MartinMarciniszyn
left a comment
There was a problem hiding this comment.
I'll move these changes into NVIDIA's internal repository.
Collaborator
|
This has already been picked up by @syuoni and will be released to Github next week. |
Merged
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Outdated suggestions cannot be applied.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.
In regression tests of the
ModelRunnerCppvsModelRunnerwe noticed that thestop_words_listfeature does not work properly for theModelRunnerandbatch_size> 1. Depending on the input we get incorrect results or Cuda Runtime Errors (misaligned address). Looking at the relevant code, it seems that the pointer offsets forstop_words_list_ptrs(and alsobad_words_list_ptrs) don't take into account the size of the tensor entries. This PR fixes this issue by multiplying the offset by the element size.