Repository files navigation

This is a reimplementaion of the unconditional waveform synthesizer inDIFFWAVE: A VERSATILE DIFFUSION MODEL FOR AUDIO SYNTHESIS.

• To continue training the model, runpython distributed_train.py -c config.json.

• To retrain the model, change the parameterckpt_iter in the correspondingjson file to-1 and use the above command.

• To generate audio, runpython inference.py -c config.json -n 16 to generate 16 utterances.

• Note, you may need to carefully adjust some parameters in thejson file, such asdata_path andbatch_size_per_gpu.

• model
• samples