Artem Gribul

It seems to be that you haven't prepared files, containing mel-spectrograms, but the dataset trying to load it from disk. In this line https://github.com/mindslab-ai/univnet/blob/9bb2b54838bb6d7ce767131cc7b8b61198bc7558/datasets/dataloader.py#L100 Authors generate mel-spectrogram in runtime if...

Found another way to deal with torchscript. Just paste the code and call `patch_torchscript()` before exporting. ``` import re import typing as tp import torch ttp_regexp = re.compile(r"TensorType\[[^\]]*\]") torchtyping_replacer =...

Any updates here?