Skip to content

John istft - new vocoder #118

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

Open
wants to merge 15 commits into
base: master
Choose a base branch
from
Open

John istft - new vocoder #118

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
6330b6c
new istft vocoder
johnpaulbin Dec 4, 2022
2f7d72b
credit
johnpaulbin Dec 4, 2022
438620a
fix device
johnpaulbin Dec 4, 2022
f6ac618
Update istftnet.py
johnpaulbin Dec 5, 2022
52c682d
Update istftnet.py
johnpaulbin Dec 5, 2022
f933170
Update istftnet.py
johnpaulbin Dec 5, 2022
e1d83b7
Update istftnet.py
johnpaulbin Dec 5, 2022
930dc7a
pls work
johnpaulbin Dec 5, 2022
4bc5a65
Update istftnet.py
johnpaulbin Dec 5, 2022
c345aec
Update denoiser.py
johnpaulbin Dec 5, 2022
53107b0
Update denoiser.py
johnpaulbin Dec 5, 2022
c9ac307
Update denoiser.py
johnpaulbin Dec 5, 2022
4841567
Update denoiser.py
johnpaulbin Dec 5, 2022
c82068b
Update denoiser.py
johnpaulbin Dec 5, 2022
5ebe30f
Update denoiser.py
johnpaulbin Dec 5, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 16 additions & 5 deletions uberduck_ml_dev/utils/denoiser.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@
import sys
import torch
from ..models.common import STFT
from ..vocoders.istftnet import iSTFTNetGenerator, TorchSTFT


class Denoiser(torch.nn.Module):
Expand All @@ -46,11 +47,21 @@ def __init__(
raise Exception("Mode {} if not supported".format(mode))

with torch.no_grad():
bias_audio = (
hifigan.vocoder.forward(mel_input.to(hifigan.device))
.view(1, -1)
.float()
)
if isinstance(hifigan, iSTFTNetGenerator):
self.stft = TorchSTFT(filter_length=16, hop_length=4, win_length=16, device="cpu").to("cpu")
spec, phase = hifigan.vocoder(mel_input.to(hifigan.device))
y_g_hat = self.stft.inverse(spec.cpu(), phase.cpu())
bias_audio = (
Copy link
Contributor

Choose a reason for hiding this comment

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

try using rearrange from einops instead of view - its more verbose

y_g_hat
.view(1, -1)
.float()
)
else:
bias_audio = (
hifigan.vocoder.forward(mel_input.to(hifigan.device))
.view(1, -1)
Copy link
Contributor

Choose a reason for hiding this comment

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

same here

.float()
)
bias_spec, _ = self.stft.transform(bias_audio.cpu())

self.register_buffer("bias_spec", bias_spec[:, :, 0][:, :, None])
Expand Down
Loading