RuntimeError: Distributed package doesn‘t have NCCL built in

项目场景：

提示：这里简述项目相关背景：

Improved Diffusion 复现

问题描述

提示：这里描述项目中遇到的问题：

RuntimeError: Distributed package doesn’t have NCCL built in

  File "D:\APP\Anaconda3\envs\diffusion\lib\site-packages\torch\distributed\distributed_c10d.py", line 602, in init_process_group    default_pg = _new_process_group_helper(  File "D:\APP\Anaconda3\envs\diffusion\lib\site-packages\torch\distributed\distributed_c10d.py", line 727, in _new_process_group_helper    raise RuntimeError("Distributed package doesn't have NCCL " "built in")RuntimeError: Distributed package doesn't have NCCL built in

原因分析：

提示：这里填写问题的分析：

windows不支持NCCL backend

解决方案：

提示：这里填写该问题的具体解决方案：

找到问题代码所在位置
D:\APP\Anaconda3\envs\diffusion\lib\site-packages\torch\distributed\distributed_c10d.py

    dist.init_process_group(backend=backend, init_method="env://")

更换为

    dist.init_process_group(backend="gloo", init_method="env://")

即可

来源地址：https://blog.csdn.net/weixin_43940981/article/details/127423861