axial_positional_embedding
einops>=0.3
torch>=1.6
torchvision
vector-quantize-pytorch
