You may need to install ‘nccl2‘ from NVIDIA official website

本文主要是介绍You may need to install ‘nccl2‘ from NVIDIA official website，希望对大家解决编程问题提供一定的参考价值，需要的开发者们随着小编来一起学习吧！

报错信息

在使用paddle进行多卡训练的时候报错，报错信息如下

W0111 17:25:32.685145 56257 dynamic_loader.cc:207] You may need to install ‘nccl2’ from NVIDIA official website: https://developer.nvidia.com/nccl/nccl-downloadbefore install PaddlePaddle.
Traceback (most recent call last):
File “tools/train.py”, line 114, in
main(config, device, logger, vdl_writer)
File “tools/train.py”, line 47, in main
dist.init_parallel_env()
File “/home/disk0/zw/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/distributed/parallel.py”, line 181, in init_parallel_env
parallel_helper._init_parallel_ctx()
File “/home/disk0/zw/anaconda3/envs/paddle/lib/python3.7/site-packages/paddle/fluid/dygraph/parallel_helper.py”, line 42, in _init_parallel_ctx
parallel_ctx__clz.init()
RuntimeError: (PreconditionNotMet) The third-party dynamic library (libnccl.so) that Paddle depends on is not configured correctly. (error code is libnccl.so: cannot open shared object file: No such file or directory)
Suggestions:

1. Check if the third-party dynamic library (e.g. CUDA, CUDNN) is installed correctly and its version is matched with paddlepaddle you installed.
2. Configure third-party dynamic library environment variables as follows:

• Linux: set LD_LIBRARY_PATH by export LD_LIBRARY_PATH=...
• Windows: set PATH by set PATH=XXX; (at /paddle/paddle/fluid/platform/dynload/dynamic_loader.cc:234) [Hint: If you need C++ stacktraces for debugging, please setFLAGS_call_stack_level=2`.]

分析原因

环境信息

• python：3.7
• cuda:10.0
• cudnn：7.6
• paddlepaddle-gpu：2.0.0rc1

通过上面的错误可以很容易定位到是因为没有找到libnccl.so导致的这个问题，所以导致这个错误有两种原因：

1. 没有安装nccl
2. 没有将libnccl.so添加到LD_LIBRARY_PATH环境变量中

解决办法

安装nccl

根据cuda的版本去选择对应版本的nccl，可以去nvidia的官网下载https://developer.nvidia.com/nccl/nccl-legacy-downloads
这里以cuda10为例

1.下载nccl-repo-ubuntu1604-2.6.4-ga-cuda10.0_1-1_amd64.deb
2.安装镜像库
sudo dpkg -i nccl-repo-ubuntu1604-2.6.4-ga-cuda10.0_1-1_amd64.deb
3.更新源镜像
sudo apt update
4.安装nccl
sudo apt install 
libnccl2=2.6.4-1+cuda10.0 libnccl-dev=2.6.4-1+cuda10.0

将nccl添加到环境变量中

nccl默认的安装目录是/usr/lib/x86_64-linux-gnu，修改~/.bashrc文件，添加如下内容到文件中

#设置cuda库的目录
export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64
#将nccl添加到LD_LIBRARY_PATH中
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/x86_64-linux-gnu

添加好之后保存文件，使用source ~/.bashrc让文件的配置生效，在通过echo $LD_LIBRARY_PATH查看环境变量设置是否成功，配置成功之后输出的信息如下

/usr/local/cuda-10.0/lib64:/usr/lib/x86_64-linux-gnu

参考:

1. https://forums.developer.nvidia.com/t/have-strange-problem-on-installing-nccl/60654
2. https://zhuanlan.zhihu.com/p/174710896
3. https://github.com/PaddlePaddle/PaddleDetection/issues/1444
4. https://developer.nvidia.com/nccl/nccl-legacy-downloads

这篇关于You may need to install ‘nccl2‘ from NVIDIA official website的文章就介绍到这儿，希望我们推荐的文章对编程师们有所帮助！

---