This project was presented at the ICASSP 2023 in Rhodos, Greece.
This repo rely on submodules, clone it with
git clone --recursive [email protected]:anonymous9992/neural_audio_energy.git
or manually fetch the submodules using
git submodule update --init
Install requirements
pip install -r requirements.txt
Example of training the small melgan variant on the ljspeech dataset
python train.py melgan --variant small --dataset ljspeech --gpu 0
Batch size is scaled automatically to fill the GPU memory.