End of mini batch iteration about graphsage HOT 3 CLOSED

Yes this is to avoid index out of bound.
I have modified the code just now to allow the size of the last batch to be any positive integer less than self.batch_size, so that all examples are seen at every epoch

from graphsage.

Your modifications are cool and the size of last batch can be any positive integer now. Thanks for the efforts!

However, I still can not figure out why + 1 in original code. Suppose there are 100 training edges, with batch_size set to 1. Your original code would produce 101 batches, where the last batch is empty. Would you please give an instance when + 1 helps?

from graphsage.

I see. Yes the +1 in the code before didn't seem necessary. Thanks for pointing it out!

from graphsage.