Y

I would like to ask why I can achieve the desired effect on the sample during training, but when the trained model is tested on the training data set, it produces bad results.