Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix reducesum onnx lit test to linalg lowering fails #3218

Open
wants to merge 4 commits into
base: main
Choose a base branch
from

Conversation

renxida
Copy link
Collaborator

@renxida renxida commented Apr 23, 2024

@renxida renxida marked this pull request as ready for review April 23, 2024 22:02
@renxida renxida requested review from rsuderman, saienduri and vivekkhandelwal1 and removed request for rsuderman April 23, 2024 22:02
Copy link
Collaborator

@vivekkhandelwal1 vivekkhandelwal1 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there any reason why we have removed all the detailed checks and replaced them just with the checks containing op names?

@renxida
Copy link
Collaborator Author

renxida commented Apr 24, 2024

Is there any reason why we have removed all the detailed checks and replaced them just with the checks containing op names?

I did so because i was feeling that the original checks were a little too detailed. For example, it made sure that there are two instances of int_0, and running CSE might cause the tests to fail.

I'm operating under the understanding that the lit tests are supposed to make sure the generated IR is valid & the generation doesn't crash.

But given what I learned from @zjgarvey in the standup this morning, we also rely on the tests as a source of understanding for how the lowerings work.

So I'd love your thoughts on how detailed the tests should be @vivekkhandelwal1 @zjgarvey

@vivekkhandelwal1
Copy link
Collaborator

Is there any reason why we have removed all the detailed checks and replaced them just with the checks containing op names?

I did so because i was feeling that the original checks were a little too detailed. For example, it made sure that there are two instances of int_0, and running CSE might cause the tests to fail.

I'm operating under the understanding that the lit tests are supposed to make sure the generated IR is valid & the generation doesn't crash.

But given what I learned from @zjgarvey in the standup this morning, we also rely on the tests as a source of understanding for how the lowerings work.

So I'd love your thoughts on how detailed the tests should be @vivekkhandelwal1 @zjgarvey

I think the checks which are already present are fine, if you want then you can just remove the constant's checks but that won't we much either. Also, what's the reason that we have removed a lit test?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

test_reduce_sum_negative_axes_keepdims_exampleonnx->torch->linalg fails
2 participants