search

yulu0724 / SDC-IL

Semantic Drift Compensation for Class-Incremental Learning (CVPR2020)
115 stars 34 forks source link

About the print accuracy while training and LwM #16

Open eeeeeei opened 3 years ago

eeeeeei commented 3 years ago

Hi, thanks for your contribution. Here i have two questions: 1、while training in cifar100 with resnet32, i got the print accuracy 0 and LwF loss 0 all the time, the other printouts are normal, such as follows. However, the accuracy while testing seems corrcet, is this phenomenon normal? Is there anything wrong?

[Epoch 00001] Total Loss: 0.021 LwF Loss: 0.000 Accuracy: 0.000 Pos-Dist: 0.492 Neg-Dist: 0.621 [Epoch 00002] Total Loss: 0.026 LwF Loss: 0.000 Accuracy: 0.000 Pos-Dist: 0.488 Neg-Dist: 0.617 [Epoch 00003] Total Loss: 0.022 LwF Loss: 0.000 Accuracy: 0.000 Pos-Dist: 0.504 Neg-Dist: 0.632

2、can you share the code about LwM?

Hopes for your reply.