model training is too slow

[yuhongjiu]

Tesla v100: 50min per epoch, if 200 epochs, it will take a week. but paper write a halfday with a single NVIDIA TITAN RTX to train our model

[anse3832]

In my case, it doesn't take so long when using single V100 GPU. It's not the official code, so the training time would be different as the paper says. (But I'm wonder that you took so long time for training)