0.1.19

address https://github.com/lucidrains/magvit2-pytorch/issues/9

0.1.18

0.1.18

0.1.17

0.1.17

0.1.16

0.1.16

0.1.15

add a final norm after all the encoder layers, just before quantizing

0.1.14

allow for multiple number of codebooks, so FSQ can be fairly evaluate…

…d against LFQ

0.1.12

will just use x-transformers for the downstream (non)autoregressive m…

…odel

0.1.11

fix mishandling of ema model device

0.1.10

force ema to be on same device as online model for now before update

0.1.9

fix a device issue with ema model