Comment by pk-protect-ai

I looked for pytorch native implementation for Xiaolin Wu's quantizer like 3 month ago, and found none. Would not it be much easier and more productive to integrate with pytorch? Some of the functionality which you have there is already provided by kornia, torch_kmeans. You'll end up with much less code to worry about.