jik876 / hifi-gan

HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis
MIT License
1.98k stars 507 forks source link

how to use pretrained models? #119

Closed yasntrk closed 2 years ago

yasntrk commented 2 years ago

Hi i couldn't understand how can i use pretrained model in hifigan. For example in tacotron i was running script like that

python train.py --output_directory=outdir --log_directory=logdir -c pre-trainmodel.pt --warm_start

and it was training from pretrained model. What should i do with hifigan?

yasntrk commented 2 years ago

Ohh i didn't see this part.

To train V2 or V3 Generator, replace config_v1.json with config_v2.json or config_v3.json.
Checkpoints and copy of the configuration file are saved in cp_hifigan directory by default.
You can change the path by adding --checkpoint_path option.

My bad