Hi,
I am trying to run Pytorch on my Provii and RX6300, the environment is:
OS: Ubuntu 20.04.6
Torch: 2.0.1 + ROCm-5.4.2
ROCm: 5.4.5
But when I used any operations related to GPU, like tensor.cuda(), the Provii will just stuck and RX6300 will return Segmentation Fault. That is, the pytorch with rocm did not work at all.
I have tried different OS (22.04) and ROCm(5.4.2), docker version torch, host version ... They all did not works... What can I try? I guess the problem may come from the driver?
Thanks in advance!