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

ViT Support #1377

Open
cnut1648 opened this issue Dec 16, 2024 · 0 comments
Open

ViT Support #1377

cnut1648 opened this issue Dec 16, 2024 · 0 comments

Comments

@cnut1648
Copy link

cnut1648 commented Dec 16, 2024

Hello, am I right that currently TE does not support ViT? Specifically, I am looking for ViT that uses 2D rope, e.g. Pixtral's ViT.

I also checked Nemo codebase, but it seems that they are also building their own impl for ViT instead of relying on TE.

Nemo defines ParallelVisionTransformer. This class init a series of ParallelVisionTransformerLayer layers, each a subclass of ParallelVisionTransformerLayer_, which in turn a subclass of ParallelTransformerLayer_. This class defines all operations in Nemo's impl.

Thanks.

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

1 participant