search

rishikksh20 / hifigan-denoiser

HiFi-GAN: High Fidelity Denoising and Dereverberation Based on Speech Deep Features in Adversarial Networks
Apache License 2.0
208 stars 45 forks source link

Pretrained checkpoint #1

Open Gorodecki opened 3 years ago

Gorodecki commented 3 years ago

Hi. Thank's for your research! Could you please guide me where download the weights (pretrained checkpoint)?

rishikksh20 commented 3 years ago

@Gorodecki I don't provide any pretain model, you can train this model from scratch

Gorodecki commented 3 years ago

What data did you use for training? Config used a similar one in the repo?

rishikksh20 commented 3 years ago

I used LJSpeech, I add noise to LJSpeech and then tested on this repo. But I didn't get satisfactory results from it.

usamazf commented 2 years ago

I used LJSpeech, I add noise to LJSpeech and then tested on this repo. But I didn't get satisfactory results from it.

Hi there, I am trying to do the same using the augment branch but I get error related to tensor size after about 350 iterations worth of training. (Probably some specific sample causing the issue?) Any ideas?

image