About the evaluation #8
Description
What's the difference between training and evaluation other than the training data set path? Why do gradient updates occur during evaluation ? Should I retrain the evaluation with a pre-trained model ? And Where does the evaluation index in the paper come from？