search

facebookresearch / vocoder-benchmark

A repository for benchmarking neural vocoders by their quality and speed.
Other
199 stars 27 forks source link

Pretrained models #8

Closed githicham closed 2 years ago

githicham commented 2 years ago

Hi,

In your paper, you train and compare different models on various metrics. Are you planning on releasing the model weights you obtained as pretrained models?

Regards,

ebadawy commented 2 years ago

Not sure if we will share the pre-trained models (at least at the moment); However, we have shared the config files that used to train each model. Hopefully, this will make it easier to reproduce the results in the paper.

githicham commented 2 years ago

I started training a DiffWave model on the LJSpeech dataset with a single GPU (using your config file), just to estimate the time it would take. That's what pushed me to ask about pretrained models :-) I am closing the issue since the initial question has been answered.

ebadawy commented 2 years ago

I would suggest to keep the issue open in case someone would be interested to train and share their pre-trained models. Will try to do so later on once I get the chance.