RuntimeError: index 99 is out of bounds for dimension 0 with size 99

Question

I'm working on PyTorch geometric I have a RuntimeError: index 99 is out of bounds for dimension 0 with size 99 issue. My dataset is a bipartite graph.

• Source shape is 32,3 [nodes,feature]
• target size is 100,3 [nodes,feature]
• edge_index is 2,745 [2, index]

Answer 1

The error message

RuntimeError: index 99 is out of bounds for dimension 0 with size 99

says you tried to access index 99 of the dimension 0 (maybe of target, which you say to be 100,3) when it only has a size of 99, so the only indexes you could have accessed would be 0-98.

It could be that in your dataset your __ len __ is just 1 value greater than your actual data, or if you are sure about the dimentions could be that you skip one sample in your dataset which then is not found.

Answer 2

I found that this issue popped up due to "add_self_loops" in the CONV layer being set to true. Try setting it to false and see if that works.