pytorch报错“RuntimeError: DataLoader worker (pid 83709) is killed by signal: Bus error. It is .......“

在docker容器里运行pytorch写的代码，报错：“RuntimeError: DataLoader worker (pid 83709) is killed by signal: Bus error. It is possible that dataloader’s workers are out of shared memory. Please try to raise your shared memory limit.”，具体如下图：

一，报错原因：

docker的共享内存shm不够

1, 在运行的代码前面加上下面代码（建议使用这种方法，操作比较简单）：

import sys
import torch
from torch.utils.data import dataloader
from torch.multiprocessing import reductions
from multiprocessing.reduction import ForkingPickler

default_collate_func = dataloader.default_collate


def default_collate_override(batch):
  dataloader._use_shared_memory = False
  return default_collate_func(batch)

setattr(dataloader, 'default_collate', default_collate_override)

for t in torch._storage_classes:
  if sys.version_info[0] == 2:
    if t in ForkingPickler.dispatch:
        del ForkingPickler.dispatch[t]
  else:
    if t in ForkingPickler._extra_reducers:
        del ForkingPickler._extra_reducers[t]

2，在运行docker容器的启动命令中加入以下参数:

--shm-size="64g"

具体改为多少依据宿主机的运行内存来定

3，修改对应docker容器的配置文件

首先停止docker服务：

$ systemctl stop docker

修改容器的配置文件：

$ su root
$ cd /var/lib/docker/containers/容器ID
$ ls
$ vim hostconfig.json

修改里面Shmsize的大小：

重启docker服务：

$ systemctl restart docker

Done！！！