Fix issues pointed by other users in github

[CoderPat]

Fixes #1 ,#3, #4, #5 and #7, #12.

[ioana-blue]

Can't comment on the script that generates the graph, the rest look good to me. Thanks for your help!

[ioana-blue]

Note problem explained here:
#4 (comment)

[shellycsy]

Can't comment on the script that generates the graph, the rest look good to me. Thanks for your help!

Have you finished the experiment?

[ioana-blue]

at line 445/444, I get an error because I think the iterator now produces only features without labels as you removed the labels_file from the input function but you're still unpacking it from the iterator. Taking an educated guess, labels should disappear from line 444

And also probably remove labels from the call to model. I think it should work fine with None for labels since it's calling the model in inference mode.

[ioana-blue]

What I see is that TF is not happy with dealing with None tensors, so a workaround may be necessary to pass in None as labels. Passing an empty dictionary should work. This means changing the code in the OpenGNN to account for empty labels as opposed to None.