DCT-NET.Pytorch icon indicating copy to clipboard operation
DCT-NET.Pytorch copied to clipboard
verified publisher icon LeslieZhoa

Training error

Open darkasevgen opened this issue 3 years ago • 1 comments

Hi @LeslieZhoa ! How can you compare the dictionary and the float? https://github.com/LeslieZhoa/DCT-NET.Pytorch/blob/b303f416c2caf05114fd38f10b0570df903b4462/trainer/ModelTrainer.py#L124

This method is called here: https://github.com/LeslieZhoa/DCT-NET.Pytorch/blob/b303f416c2caf05114fd38f10b0570df903b4462/trainer/ModelTrainer.py#L96

But the val_loss obtained in CCNTrainer.evalution is an empty dictionary: https://github.com/LeslieZhoa/DCT-NET.Pytorch/blob/b303f416c2caf05114fd38f10b0570df903b4462/trainer/CCNTrainer.py#L196-L206

I want to understand how to fix this and what metrics to count? And as I understand it, you trained the network without early_stopping. How many epochs/iterations did it take?

darkasevgen avatar Aug 05 '22 04:08 darkasevgen

Hi @darkasevgen ! I didn't use early stop. It is just a template. If you want to use it, just modify the evalution function and get_loss_from_val. As described by Readme，when you train ccn， almost 1000 steps, you can stop it. When you train ttn, you can stop by monitoring its condition. Maybe it is 3w steps.

LeslieZhoa avatar Aug 05 '22 11:08 LeslieZhoa