-
Notifications
You must be signed in to change notification settings - Fork 4.2k
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
initial prototype for vits2 #2838
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks like WIP. Pls, tag me again when you need a review.
@@ -430,3 +430,133 @@ def forward(self, x, x_mask): | |||
x = self.norm_layers_2[i](x + y) | |||
x = x * x_mask | |||
return x | |||
|
|||
class ConditionalRelativePositionTransformer(nn.Module): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Better if you move this into layers/vits2
return int((kernel_size * dilation - dilation) / 2) | ||
|
||
|
||
class TextEncoder(nn.Module): |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I guess they use regular transformer not relative position one.
Ahh, yes. I forgot to turn it into draft. I'll let you know as soon as it's done. Thanks! |
class VITS(): |
@p0p4k If you don't mind, we need some testing like TTS/tests/tts_tests/test_vits.py Line 36 in cdc971f
|
@p0p4k all looks good to me. Can you add a recipe for people to get started with training VITS2 ? |
@erogol Ok, I will do some cleanup , remove unused stuff and add a recipe with test soon. Little busy with other stuff! Thanks! |
@p0p4k sure take your time... |
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. You might also look our discussion channels. |
To do some testing these are the notebooks at my other repo.
I still have some questions about the techniques used in the paper; they do not go into the specifics. Would be great if I can get some help.
@lexkoro thanks for good discussions.