jellyfishovo
commented
2 hours ago the aforementioned proplem reslults in the error in hypergraph convolution process, and outputs tensors like (916,512) and (917,512). They are not allowed to calculate the "group loss" and report: IndexError: The shape of the mask [917] at index 0 does not match the shape of the indexed tensor [916, 512] at index 0
In your code, you mentioned group loss, which is the loss calculated after masking some of the hyperedges. However, during actual execution, I encountered an issue: after applying the hypergraph mask, the maximum indices in “hyperedge_index1” and “hyperedge_index2” are different, and the index values often differ by 1. Could you please explain where the problem might be? I have followed your code logic strictly while running it.
Thank you for you time and consideration~