test net accuracy is always zero

[New Caffe User]

Hi All,

I am training Alexnet on my dataset with 30,000 test and 12,000 validation images. I see that the training accuracy and the training loss trend is good as expected but the test loss is very high and the test accuracy is always 0. I tried with 100,000 iterations and the test accuracy is 0 till the last iteration.

I tried increasing the regularization parameter to reduce over fitting but this did not change my results. Please let me know if there is anyway to debug this. Am I missing something very basic here?

Thanks!