0.0.29

complete variable lengthed sequence training

0.0.28

fix readme

0.0.27

release level-by-level quantizer decoding from @lucasnewman

0.0.26

oops, actually release acoustic prompting from @lucasnewman

0.0.25

release acoustic prompting from @lucasnewman

0.0.24

appreciation to the few who contribute

0.0.23

make sure generate can still work unconditionally, fix readme

0.0.22

fix key padding mask for non-flash attention, readying for variable l…

0.0.21

fix masking of upper quantizers during training, thanks to @chenht2010

0.0.20

end-to-end text-to-speech with audiolm, spear-tts, and soundstorm