Implementation of HifiGAN. Due to computational constraints, the final model has been learning for <=25 hours.
Link to final model
import gdown
base_model_path = 'https://drive.google.com/uc?export=download&id=1Ymw4vR--v7uiWNcz2zstzZEoVYLrGJnp'
best_tuned_path = 'https://drive.google.com/uc?export=download&id=1oA-1fm27kx80KH5URJ0DV4s5cOj1MrdS'
gdown.download(<choose your path>)
test.py --ckpt-path ckpt-tuned.tar -t <test audio folder> -o <your existed output dir>
The Link to wandb report