training loss

[idejie]

I re-implement this model with Pytorch, but the loss is around 0.69 and does not drop in the training stage.

I want to know the loss when you training.

Thanks~