因為沒有安裝nccl，本機為centos7

https://developer.nvidia.com/nccl/nccl-legacy-downloads

需要登入一下

得到nccl-repo-rhel7-2.7.8-ga-cuda10.1-1-1.x86_64.rpm

安裝教程 ：https://docs.nvidia.com/deeplearning/nccl/install-guide/index.html#down

1）執行：

rpm -ivh nccl-repo-rhel7-2.7.8-ga-cuda10.1-1-1.x86_64.rpm

2）然後執行：

yum-config-manager --add-repo https://
 
developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/cuda-rhel7.repo

返回：

Loaded plugins: product-id
adding repo from: https://developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/cuda-rhel7.repo
grabbing file https://developer.download.nvidia.com/compute/cuda/repos/rhel7/x86_64/cuda-rhel7.repo to /etc/yum.repos.d/cuda-rhel7.repo
 

cuda-rhel7.repo                                                                                                                                        |  232 B  00:00:00
repo saved to /etc/yum.repos.d/cuda-rhel7.repo 

3）更新：

sudo yum update

4）安裝：

yum install libnccl-2.7.8-1+cuda10.1 libnccl-devel-2.7.8-1+cuda10.1
 
 libnccl-static-2.7.8-1+cuda10.1 

該錯就沒了