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

Feature request : Streaming conformer #13

Open
pfeatherstone opened this issue Dec 7, 2022 · 3 comments
Open

Feature request : Streaming conformer #13

pfeatherstone opened this issue Dec 7, 2022 · 3 comments

Comments

@pfeatherstone
Copy link

pfeatherstone commented Dec 7, 2022 •
edited
Loading

Can you support streaming conformer like how this paper proposes to do it ?
@pfeatherstone
Copy link
Author

If training with CTC loss function, do you have to segment the labels per block or do I not need to worry about that?

@pfeatherstone
Copy link
Author

Or is there a simpler way to train a Conformer model with CTC loss and do block-wise inference for online streaming and transcribing? If so, do we have to segment the labels per block ? (Which goes against the CTC design)

@pfeatherstone
Copy link
Author

This page from NVidia seems to suggest you just need to partition your data into blocks. Each block doesn't need any causal attention. Is that right? Then can you still train with CTC as usual ?

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
1 participant
@pfeatherstone