Is 200 epochs enough to converge?

[eensdl]

I followed the instruction and upon 200 epoch the average loss is about 3.

I am curious if I have done something wrong?

Should the training converge within 200 epochs?

[eensdl]

Yes it should. Somehow Colab is giving me bad results. After changing from Colab to the local GPU the loss dropped to 2.6 in 110 epochs

[robbie2021]

Hi, i have met the same problem with you. But i get the Regression Loss value is inf, the Classification Loss is normal.

2021-04-21 11:38:41,798 - root - INFO - Epoch: 195, Step: 100, Average Loss: 1.7795, Average Regression Loss 0.7859, Average Classification Loss: 0.9937 ............................................................................2021-04-21 11:38:55,192 - root - INFO - lr rate :0.0001 2021-04-21 11:38:55,193 - root - INFO - lr rate :0.0001 2021-04-21 11:38:57,230 - root - INFO - Epoch: 195, Validation Loss: inf, Validation Regression Loss inf, Validation Classification Loss: 1.0416 2021-04-21 11:38:57,241 - root - INFO - Saved model ./models/train-version-slim/slim-Epoch-195-Loss-inf.pth ...................................................................................................... 2021-04-21 11:39:17,186 - root - INFO - Epoch: 196, Step: 100, Average Loss: inf, Average Regression Loss inf, Average Classification Loss: 0.9936 ............................................................................2021-04-21 11:39:30,308 - root - INFO - lr rate :0.0001 ...................................................................................................... 2021-04-21 11:39:49,252 - root - INFO - Epoch: 197, Step: 100, Average Loss: 1.7731, Average Regression Loss 0.7626, Average Classification Loss: 1.0105 ............................................................................2021-04-21 11:40:02,683 - root - INFO - lr rate :0.0001 ...................................................................................................... 2021-04-21 11:40:22,183 - root - INFO - Epoch: 198, Step: 100, Average Loss: 1.7419, Average Regression Loss 0.7519, Average Classification Loss: 0.9900 ............................................................................2021-04-21 11:40:35,509 - root - INFO - lr rate :0.0001 ...................................................................................................... 2021-04-21 11:40:54,622 - root - INFO - Epoch: 199, Step: 100, Average Loss: 1.7601, Average Regression Loss 0.7651, Average Classification Loss: 0.9950 ............................................................................2021-04-21 11:41:08,277 - root - INFO - lr rate :0.0001 2021-04-21 11:41:08,277 - root - INFO - lr rate :0.0001 2021-04-21 11:41:10,293 - root - INFO - Epoch: 199, Validation Loss: inf, Validation Regression Loss inf, Validation Classification Loss: 1.0122 2021-04-21 11:41:10,305 - root - INFO - Saved model ./models/train-version-slim/slim-Epoch-199-Loss-inf.pth

[eensdl]

Try downloading the project again.