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.

Sign up for GitHub

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

Jump to bottom

Project 2 Issue : Getting Sparse Attention Wts #50

Open
prateek27 opened this issue Nov 12, 2018 · 3 comments
Open

Project 2 Issue : Getting Sparse Attention Wts #50

prateek27 opened this issue Nov 12, 2018 · 3 comments

Comments

@prateek27
Copy link

prateek27 commented Nov 12, 2018 •
edited
Loading

I am implementing attention in the following manner, but get sparse attention wts. Only 1-2 elements in the activation vector are non-zero. Anyone facing similar issue or can help me figure out the problem in my approach? Thanks!
PS - I am reducing the dimensionality of global vector g to match with L (384 or 256).

attention
@raghavsi @vineetm
@raghavsi
Copy link
Collaborator

What are the nonlinearity on g/l?

@prateek27
Copy link
Author

prateek27 commented Nov 13, 2018 via email

@raghavsi
Copy link
Collaborator

The layer before B when you convert Conv4 to FC, what is the activation there?
B you have softmax right?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@prateek27 @raghavsi