CacheDataset with shared memory crashes during torchrun distributed training

steps to reproduce, run torchrun --nproc_per_node=2 main.py

import torch, os
import torch.distributed as dist
import torch.multiprocessing as mp
from monai.data import CacheDataset


rank = int(os.getenv("LOCAL_RANK"))
dist.init_process_group(backend="nccl", init_method="env://") 
torch.cuda.set_device(rank)

dataset = CacheDataset(data=[1,2,3], runtime_cache=True)

this errors doesn't accure if spawning multi-gpu processes manually (instead of torchrun).
this issue is because "broadcast_object_list" checks for authkey in each process, and torchrun has these keys different.