Got a CUDA error when calling g.apply_edges() in a GAT-based model

error message:

CUDA Version: 9.0
PyTorch Version: 1.10
DGL Version: 0.4.1

Does anyone have clues about this problem?:face_with_monocle:

This usually happens when the index is out of the range, such as for a tensor t = [1, 2, 3], you write t[4] or t[-100]. Could you share more information about your code?

It is indeed due to the index out of the range.
I made a mistake when preprocessing the edge data.
Thanks your reply.