Batch_norm and Dropout kept on while testing #33
Description
I have tried with keeping them on and off during the testing by setting the flag "is_training" in test_enet.py. However, I get better performance while keeping them off and probably keeping them on gives false results.