Implementation for multi-gpu usage

[zhangyu2ustc]

Hi,
First of all, Thanks a lot for sharing this package.
I notice that the current package only a single GPU during model training. Is it possible to distribute the training process to multi-gpu, which will accelerate the training process?

Thanks again.
Best,
Yu.

[mdeff]

Thanks for your interest. That's definitely possible but requires the code to be modified. This repo won't be updated as it's purpose is to reproduce the paper.