Open mbenami opened 5 years ago
Hi @mbenami, thanks for your questions.
I would recommend monitoring reward values for each iteration of training. It's calculated by calculate_reward() function by averaging highest three non-overfitting validation accuracies in the training history. It should increase overall by each iteration (while there are lots of ups and downs), and should saturate after a while (as in below example). You might want to stop iteration when it doesn't increase for 20-30 steps.
Yes, I thought about integrating TensorBoard, but didn't have time. This is something I want to do in the future if I find enough time.
HI @barisozmen thanks for sharing the code for deepaugment I would like to try this on my dataset. which value would you recommend to monitor on? have you considered to implement tensorboard/ tensorboardX in the code for easy validate of the process?
thanks!