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

Query about ''offset'' #5

Open
RayYoh opened this issue Apr 15, 2024 · 1 comment
Open

Query about ''offset'' #5

RayYoh opened this issue Apr 15, 2024 · 1 comment

Comments

@RayYoh
Copy link

RayYoh commented Apr 15, 2024

Hi authors, thanks for your work. I am confused about the offset here. Since the residual is None, why need to use this offset instead of directly using the output hidden states.

@Aristo23333
Copy link
Collaborator

Aristo23333 commented Apr 16, 2024

Hi authors, thanks for your work. I am confused about the offset here. Since the residual is None, why need to use this offset instead of directly using the output hidden states.

Hi ! Thanks for your interest in our work. The offset is designed referring to the offset design in PCT. I regard this as the a similar approximation of Laplace function mentioned by PCT, which is inherently permutation-invariant and more suitable for point cloud learning compare to the original Mamba block module. I also provide a simple prove as folllowed:
0884066f2d91c1110e0f48904462486

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