Fix cumsum dtype - #4511
Open
jloftin-nv wants to merge 2 commits into
Open
Conversation
github-actions
Bot
added
component: tests
cehongwang
reviewed
Aug 21, 2026
| if dtype is not None: | ||
| acc_dtype = dtype | ||
| elif not input_dtype.is_floating_point: | ||
| acc_dtype = torch.int64 |
Collaborator
There was a problem hiding this comment.
Why this is int64? Can it be int32 or other non-float dtype?
Contributor
Author
There was a problem hiding this comment.
You can see here it defaults to int64 in PyTorch
Collaborator
|
Can you also add a bf 16 test case? |
jloftin-nv
force-pushed
the
dev-jloftin-cumsum-dtype
branch
from
August 21, 2026 22:36
4fa9d8d to
b398035
Compare
Contributor
Author
Added |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
The cumsum converter drops the keyword-only dtype argument and always seeds the TRT loop accumulator as float32, so integer / typed cumsums disagree with PyTorch.
Forward dtype from the aten node and use a typed zero for the recurrence seed.
Type of change
Please delete options that are not relevant and/or add your own.
Checklist: