Training code for VQ-VAE is not reproducible

[zelaki]

Hello and thank you for great work.

The training code you provide in: https://github.com/FoundationVision/vaex is incomplete (e.g. UnlabeledImageFolders class in utils/data.py does not exist in the repository) and not reproducible (there are no instructions on how to run the script to reproduce the experiments in the paper).

Since this is an impactful work and the multi-scale quantized VQ-VAE is one of its major contributions it would be very helpful for the community to be able to reproduce it.

Do you plan to update the training repo ?

Thank you in advance!