Why is the validation loss lower than the training loss?

vision
2 
val_epoch_loss = 0

is inside the validation for loop , it should be before the for loop

1 Like