How to know when should stop training in your code based on Loss and IoU output ? #5
Description
Hi
I did train the model and I got training loss and validation for each epoch among training like bellow:
....
Epoch 150 |Iter 81200 |Loss 0.03061891
Epoch 150 |Iter 81300 |Loss 0.031306103
Epoch 150 |Iter 81400 |Loss 0.034883704
Epoch 150 |Iter 81500 |Loss 0.0326193
Epoch 150 |Iter 81600 |Loss 0.034357037
Time: 3870.109310865402
validate...
last iou valu:0.65
new_iou value:0.9206227973963147
Save the checkpoint...
Epoch 151 |Iter 81700 |Loss 0.03399411
Epoch 151 |Iter 81800 |Loss 0.030656744
Epoch 151 |Iter 81900 |Loss 0.033804014
Epoch 151 |Iter 82000 |Loss 0.034370854
Epoch 151 |Iter 82100 |Loss 0.033772305
Epoch 151 |Iter 82200 |Loss 0.03267585
Time: 315.97310042381287
validate...
last iou valu:0.9206227973963147
new_iou value:0.9007333493933005
Epoch 152 |Iter 82300 |Loss 0.029079173
Epoch 152 |Iter 82400 |Loss 0.031596527
Epoch 152 |Iter 82500 |Loss 0.03290856
Epoch 152 |Iter 82600 |Loss 0.03253841
Epoch 152 |Iter 82700 |Loss 0.029811475
Time: 315.04660058021545
validate...
last iou valu:0.9206227973963147
new_iou value:0.9206331145003502
...
How can I know when should I stop training? and how to know about overfitting?