0.0.19

add ability to skip prompt asking whether to clear previous results, …

0.0.18

need torch 1.12 for complex stft discriminator from audiolm

0.0.17

in contrastive learning, make sure always the full batch

0.0.16

batch size must be greater than 1 for contrastive learning

0.0.15

helpful assert

0.0.14

forget about weight decay, and make sure to clip wavs by data_max_length

0.0.12

curtail to shortest, as no self attention masking strategy in audio s…

0.0.11

get a mulan trainer out, remove wip

0.0.10

add patch dropout

0.0.9

adopt dual patchnorm