Releases: lucidrains/naturalspeech2-pytorch
Releases · lucidrains/naturalspeech2-pytorch
0.0.38
add __init__.py
0.0.37
fix checkpointing code, thanks to @janjurca for raising the issue!
0.0.36
make number of blocks per resnet block customizable
0.0.35
add ability to use resnet blocks (with residuals) in duration / pitch… … predictor
0.0.34
specify number of convolutions per attention, constituting one block … …(depth)
0.0.33
allow for variable lengths
0.0.32
wire both up with tokenizer, but check later to see which one accepts… … the phonemized ids
0.0.31
phoneme encoder can now receive a list of strings and properly return… … the phoneme embeddings
0.0.30
phoneme encoder needs to account for padding, use an extra token plac… …ed at the very end of the embedding table for padding
0.0.28
get ready for phoneme encoder to accept list of strings