This release adds cropping to the skip connections in the waveform models to enable the use of different max lengths than multples of 16384.
0.1.7
This release changes model outputs from `.logits` to `.audio`
0.1.6
Remove dependencies from pypi build
0.1.5
Patch release. torchaudio not found via pip during pypi publish.
0.1.4
This release introduces a new model `UNet1DModel`. There is currently a pretrained checkpoint on the huggingface hub. It is still training, but beats the waveunet checkpoints across the board.