update train.py for bypass=False in VNet

[neeravkaushal]

input needs to be added to target and output for bypass=False, so that the eulerian loss for bypass and no bypass equals.