Jel imao neko ovaj problem da mu se mean loss smanjuje, ali Test Accuracy je stalno oko 40%?
= Using device cpu
Loaded 60000 training images!
Loaded 10000 validation images!
Epoch: 0
Iter: 0, Mean Loss: 28.289
Iter: 100, Mean Loss: 5.196
Iter: 200, Mean Loss: 3.158
Iter: 300, Mean Loss: 2.354
Iter: 400, Mean Loss: 1.916
Iter: 500, Mean Loss: 1.635
Iter: 600, Mean Loss: 1.439
Iter: 700, Mean Loss: 1.295
Iter: 800, Mean Loss: 1.184
Iter: 900, Mean Loss: 1.096
Mean Loss in Epoch 0: 1.067
Computing mean representations for evaluation…
Evaluating on test set…
Epoch 0: Test Accuracy: 43.08%
Epoch time (sec): 170.6
Epoch: 1
Iter: 0, Mean Loss: 0.370
Iter: 100, Mean Loss: 0.354
Iter: 200, Mean Loss: 0.351
Iter: 300, Mean Loss: 0.345
Iter: 400, Mean Loss: 0.340
Iter: 500, Mean Loss: 0.335
Iter: 600, Mean Loss: 0.330
Iter: 700, Mean Loss: 0.325
Iter: 800, Mean Loss: 0.321
Iter: 900, Mean Loss: 0.317
Mean Loss in Epoch 1: 0.316
Evaluating on test set…
Epoch 1: Test Accuracy: 36.78%
Epoch time (sec): 162.4
Epoch: 2
Iter: 0, Mean Loss: 0.291
Iter: 100, Mean Loss: 0.279
Iter: 200, Mean Loss: 0.279
Iter: 300, Mean Loss: 0.276
Iter: 400, Mean Loss: 0.273
Iter: 500, Mean Loss: 0.271
Iter: 600, Mean Loss: 0.269
Iter: 700, Mean Loss: 0.267
Iter: 800, Mean Loss: 0.265
Iter: 900, Mean Loss: 0.263
Mean Loss in Epoch 2: 0.263
Computing mean representations for evaluation…
Evaluating on test set…
Epoch 2: Test Accuracy: 43.37%
Epoch time (sec): 157.8