Use configuration files“byol_resnet50_8xb32-accum16-coslr-100e_in1k.py”4 cards running in parallel, the machine restarts

[HUUA]

This is my software environment。
ubuntu 22.04.1
torch 1.14.0.dev20221027+cu117 pypi_0 pypi
torchaudio 0.14.0.dev20221027+cu117 pypi_0 pypi
torchvision 0.15.0.dev20221027+cu117
mmselfsup 0.10.0
cuda 11.7
nvidia driver 520.06

whenever i pass“bash tools/dist_train.sh configs/selfsup/byol/byol_resnet50_8xb32-accum16-coslr-100e_in1k.py 4”,1~2 minutes, the machine will restart。
Run 3 graphics cards, the machine no restart.

I first suspected a hardware failure related,But I pass things like stress-ng, gpu__burn，Let the machine's IO, graphics card, and CPU all reach the maximum load, but the machine will not restart，
So I want to ask for help, is it a related problem in the code