Error for Momentum optimizer

[ioana-blue]

I'm getting an error when I use the momentum optimizer:
TypeError: __init__() missing 1 required positional argument: 'momentum'

Initially I thought was some TF API breakage, but I think the culprit is in this line:
https://github.com/CoderPat/structured-neural-summarization/blob/master/train_and_eval.py#L516

def optimizer(lr): return optimizer_class(lr) # **optimizer_params)

For momentum, the constructor expects more parameters, so I'm guessing the following fixes it:
def optimizer(lr): return optimizer_class(lr, **kwargs)
and the kwargs are picked from the above.

[ioana-blue]

I guess it's possible it is a TF API breakage. FWIW I'm using TF 1.12

[CoderPat]

I apologize for this. I ran most of my early experiments using the more cumbersome API on the OpenGNN, but decided to upload a more user friendly script to users, and some of these bugs were left unchecked.
Should be fixed in the submitted PR. Feel free to review it. Ill merge it tomorrow.