Releases: lucidrains/vector-quantize-pytorch
Releases Β· lucidrains/vector-quantize-pytorch
1.2.0
add a new technique where one can do cross entropy loss on the distanβ¦
1.1.6
fix orthogonal regularization, addressing https://github.com/lucidraiβ¦
1.1.5
allow for turning off norm in random projection quantizer
1.1.4
random projection quantizer needs to have input normalized to mean 0 β¦
1.1.2
add ability to easily get codes for multi-headed vq
1.1.1
should never require gradients
1.1.0
add random projection quantizer, used by Google USM
1.0.7
cleanup
1.0.6
add the structured quantized dropout from encodec paper
1.0.5
add the structured quantized dropout from encodec paper