forked from mrq/DL-Art-School
26 lines
973 B
Python
26 lines
973 B
Python
import torch
|
|
from scipy.io import wavfile
|
|
|
|
from models.audio.vocoders.waveglow.waveglow import WaveGlow
|
|
|
|
|
|
class Vocoder:
|
|
def __init__(self):
|
|
self.model = WaveGlow(n_mel_channels=80, n_flows=12, n_group=8, n_early_size=2, n_early_every=4, WN_config={'n_layers': 8, 'n_channels': 256, 'kernel_size': 3})
|
|
sd = torch.load('../experiments/waveglow_256channels_universal_v5.pth')
|
|
self.model.load_state_dict(sd)
|
|
self.model = self.model.cpu()
|
|
self.model.eval()
|
|
|
|
def transform_mel_to_audio(self, mel):
|
|
if len(mel.shape) == 2: # Assume it's missing the batch dimension and fix that.
|
|
mel = mel.unsqueeze(0)
|
|
with torch.no_grad():
|
|
return self.model.infer(mel)
|
|
|
|
|
|
if __name__ == '__main__':
|
|
vocoder = Vocoder()
|
|
m = torch.load('C:\\Users\\jbetk\\Documents\\tmp\\some_audio\\00008.mel').cpu()
|
|
wav = vocoder.transform_mel_to_audio(m)
|
|
wavfile.write(f'0.wav', 22050, wav[0].cpu().numpy()) |