in an exception getting thrown.

function ends.

error in the multi-gpu training code.

behave in the way you would expect.

with respect to parameters and one that updates the parameters with those
gradients.

- Made layer() recurse into repeat objects so that the index given to layer()
  does what you would expect.
- Added an operator<< for network objects that prints the network architecture.

thing to do for this kind of training.