Note: If you are using NLLLoss from pytorch make sure to use the log_softmax function from the functional class and not softmax

Hi Pandey, 
Can you tell me the reasoning behind the above statement in your README? 

Thanks. 