I am working on a academic computer vision project and here are the train loss and validation loss. is it advisable to stop at 25 epochs or take this as is and use model from epoch 44, which has the best validation loss. Any advise is appreciated!
