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

Changing the temporal dimension #23

Open
MLDeS opened this issue Sep 15, 2023 · 0 comments
Open

Changing the temporal dimension #23

MLDeS opened this issue Sep 15, 2023 · 0 comments

Comments

@MLDeS
Copy link

MLDeS commented Sep 15, 2023

I want to change the temporal dimension to ~6 and use only the first few Hiera blocks. I want to do it from a hub video model. Accordingly, I changed the Patch embed (Conv3D part) to have a less temporal stride instead of 4. After that, I only want 3 Hiera blocks. Following the changed Patch embed with Hiera.blocks[0:3] throws error. The inference way of return_immediates=True also says requiring a mask argument, which I assume is because a mask ratio is needed when requires_grad=True?

What would be the best way to do the above? Calling each Hiera block separately after a custom patch embed like above?

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