Segmentation fault in torch_scatter

[maicondallg]

I'm trying run the code and i'm having Segmentation fault when in torch_scatter. This code exemplifies the error:

import torch
from torch_scatter import scatter_max

src = torch.tensor([[2, 0, 1, 4, 3], [0, 2, 1, 3, 4]]).cuda()
index = torch.tensor([[4, 5, 4, 2, 3], [0, 0, 2, 2, 1]]).cuda()

out, argmax = scatter_max(src, index, dim=-1)

I'm running this in Python 3.8.17, Pytorch 2.0.1+rocm5.4.2, Rocm5.4.3 installed and the lasted Release.

I tried install conda install gcc_linux-64 gxx_linux-64 to solve, create new envs, tested python3.8, 3.9, 3.10. All cases i have Segmentation fault.

[maicondallg]

Apparently the error is in Torch

[Looong01]

😄