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

[RFC] aten.to.device #1075

Closed
gpetters94 opened this issue Jul 18, 2022 · 2 comments
Closed

[RFC] aten.to.device #1075

gpetters94 opened this issue Jul 18, 2022 · 2 comments

Comments

@gpetters94
Copy link
Collaborator

I'm running into calls to aten.to.device in a model and I was wondering how we want to handle it. It looks like we should just be able to replace it with a conversion but I wanted to double check if we wanted to handle the logic of moving tensors between devices.

@silvasean
Copy link
Contributor

Once the program is converted to value semantics it becomes an identity operation.

@gpetters94
Copy link
Collaborator Author

Okay, I'll implement it as a decomposition to aten.to.dtype, since the dtype logic is the part where we'll actually do anything.

qedawkins pushed a commit to nod-ai/torch-mlir that referenced this issue Oct 3, 2022
* Add ScatterElementsOp and RoiAlignOp inferShapes, with dynamic shape support and separate verifier.

Signed-off-by: PengQi220 <[email protected]>
Co-authored-by: Alexandre Eichenberger <[email protected]>
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

No branches or pull requests

2 participants