Random Unexpected segmentation fault encountered in pytorch DataLoader when I use gdb to run the code. bt only gives “Nostack”. This happens when I transit from ubuntu 18.04 to 22.04.4. The following is the info it gives when “run”.
The following is the env information: enter image description here enter image description here
Is there any conflict between ubuntu 22.04 and the latest pytorch? The reason why I need to use 22.04 is that it contains the Network driver of the ASUS z790-e motherboard.
I tried adjusting batch_size and num_workers and found that when numwokers is reduced, the crash time can be delayed. At the same time, sometimes, this segmentation fault will directly cause some CPU cores to softlock and Ubuntu22.04 to be completely unresponsive.