Releases: lucidrains/vector-quantize-pytorch
Releases · lucidrains/vector-quantize-pytorch
0.4.10
make sure codebook is learnable in the presence of orthogonal regular…
0.4.8
allow for making sure the quantization step is done in full precision
0.4.7
add ability to random sample codes for orthogonality loss, with ortho…
0.4.6
add ability to random sample codes for orthogonality loss, with ortho…
0.4.5
fix
0.4.4
add ability to only enforce orthogonality loss on activated codes wit…
0.4.3
allow for passing in image feature maps, taking care of transposes
0.4.2
get ready to deprecate and turn off commitment loss in next version
0.4.1
set requires_grad depending on whether training or not
0.4.0
release orthogonal regularization loss, from https://arxiv.org/abs/21…