Results 1 issues of Yibo Hu

Where did you implement the early stopping based on PPL on the validation set in the training script? Thanks