Aleksey

If I understand your question, difference between train/val batch size does not affect to train quality, because val process is only prediction (without weights changing), and val batch size affects...

I have created by myself on pytorch (https://github.com/Kwentar/Color2Embed_pytorch), may be, Authors will say if I have some mistakes, but it works similar

I have the same problem, can help to reproduce if needed (only in windows desktop client)