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

Why there is no positional embeds in the flow decoder transformer layers inputs? #381

Open
JohnHerry opened this issue Sep 11, 2024 · 2 comments
Open

Why there is no positional embeds in the flow decoder transformer layers inputs? #381

JohnHerry opened this issue Sep 11, 2024 · 2 comments

Comments

@JohnHerry
Copy link

the flow model in cosyvoice, its encoder Conformers contains position embeds while in its decoder transformers, I see no such addition. is that means no benifit here in flow-matching? sorry for it if I did not find the code.
@aluminumbox
Copy link
Collaborator

well decoder is from matha-tts, we haven't tested whether position embedding can help in decoder, maybe it can

@JohnHerry
Copy link
Author

Yes there is no position in macha, may be it assume the intput hidden had contain such info?

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
@JohnHerry @aluminumbox