8000 Add Encodec features for Codec toolkit by jctian98 · Pull Request #5758 · espnet/espnet · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

Add Encodec features for Codec toolkit #5758

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

Merged
merged 13 commits into from
May 8, 2024
Merged

Add Encodec features for Codec toolkit #5758

merged 13 commits into from
May 8, 2024

Conversation

jctian98
Copy link
Contributor

What?

Add several Encodec features into the codec toolkit: https://arxiv.org/abs/2210.13438

  • Multi-Scale STFT discriminator: done
  • Skip discriminator training with a probability: done
  • Loss Balancer: pending.
    @ftshijt

@mergify mergify bot added the ESPnet2 label Apr 23, 2024
@ftshijt
Copy link
Collaborator
ftshijt commented Apr 26, 2024

Since it is mixed with the current decoding and scoring PR, I will review it again after we first merge that one. Thanks for the in-time PR!

@ftshijt ftshijt added this to the v.202405 milestone Apr 26, 2024
Copy link
Collaborator
@ftshijt ftshijt left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for your contribution! In general, it looks great. Please check my minor comments~

Comment on lines 81 to 85
"fs": 24000,
"n_fft": 1024,
"hop_length": 256,
"win_length": None,
"window": "hann",
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You might consider to follow the new setup for multi-scale mel loss?

use_loss_balancer: bool = False,
balance_ema_decay: float = 0.99,
):
# (Jinchuan) re-apply everything except the discriminator config.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

One TODO here, to add the docstring

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For the note consistency, please use:

# NOTE(jincuan): ...

Also for TODO

)


def apply_parametrization_norm(module: nn.Module, norm: str = "none"):
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

For consistency, I think many of the modules are already in shared, could you please double check in case?

@ftshijt
Copy link
Collaborator
ftshijt commented May 8, 2024

Thanks for your contribution! Looks great to me.

@ftshijt ftshijt merged commit 970ad59 into espnet:codec May 8, 2024
1 check passed
@jctian98 jctian98 deleted the codec branch May 17, 2024 22:53
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants
0