einops>=0.3
ftfy
regex
torch>=1.6
torchvision
unfoldNd
vector-quantize-pytorch>=0.4.10
