SIGBUS error while using multiprocessor training.

[shahzaibbaig123]

I am using mistral model and finetuning it. My server has 4 gpus. My training works fine when I use one or two gpus. But it gives me error when I try to increase the number of process in my training command. The command I am using is this:
accelerate launch --num_processes=3 -m axolotl.cli.train /data/axo_configs/mistral/extract/config.yml --deepspeed deepspeed_configs/zero1.json
Also, this is the command which i use for getting the image and make a podman container:
podman run -v /data:/data --device nvidia.com/gpu=all --security-opt=label=disable --rm -it axolotl
This is the error: