Segmentation fault (core dumped)

Hello. I have paid for gpu server and I run it for Training a large dataset for kaggle competition. But I faced Segmentation fault when I want to use gpu for training. I use cuda12.4 and pytorch 12.4 with 2 a100 gpu in community cloud. and I really struck in it. The code was written in Python. What should I do? (I'm stopped it for now.) p.s I have used runpod/pytorch:2.4.0-py3.11-cuda12.4.1-devel-ubuntu22.04