Task-wise accuracy numbers

[ZhimaoPeng]

Following the guidelines, I run the "run_base_CIFAR100.sh" for training and "run_inc_ncm_CIFAR100.sh" for evaluation, but only get the following results:

Use GPU: 1 for training => creating model 'resnet18' => no checkpoint found at './exp/CIFAR100/CIFAR100_base-data-2-fusion_batch-128_epoch-100_lr-1e-2_milestones-60-70_seed-1_s-30_m-4e-1/1/trial1_session0_best.pth' session: 8 number of images: 30200

Files already downloaded and verified
---- CIFAR100 Base Transform ---
~~~~~~~~ testing dataset ~~~~~~~~
---- CIFAR100 Testing Transform ---
length of the trainset: 30200
----------------------------- calculate and store average class-wise feature embedding -----------------------------
100%|████████████████████████████████████████████████████████████████████| 236/236 [00:02<00:00, 84.48it/s]
----------------------------- do interference -----------------------------
100%|████████████████████████████████████████████████████████████████████| 100/100 [00:01<00:00, 91.54it/s]
TEST, total average accuracy=0.0657
TEST, task-wise correct prediction: [427, 55, 22, 52, 19, 34, 12, 13, 23]
TEST, task-wise number of images: [6000, 500, 500, 500, 500, 500, 500, 500, 500]
TEST, task-wise accuracy: [0.07116666666666667, 0.11, 0.044, 0.104, 0.038, 0.068, 0.024, 0.026, 0.046]

As shown above, the testing results of each session is too low, how do I get normal results ? Thank you !

[CanPeng123]

Have you loaded the correct model? It shows no checkpoint found.