caffe2安装

１、查看cuda和ｃｕｄｎｎ版本

cuda 版本
cat /usr/local/cuda/version.txt

cudnn 版本
cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2

２、可以讲ｃｕｄａ和ｃｕｄｎｎ升级到８．０和７

３、安装caffe2，用ｃｏｎｄａ安装比较方便，如果环境不匹配可以多试几次

conda install -c caffe2 caffe2-cuda8.0-cudnn7

４、之后还要安装ＮＣＣＬ２

1. Install the repository.
   • For the local NCCL repository:

     sudo dpkg -i nccl-repo-<version>.deb

这里首先下载了nccl-repo-ubuntu1404-2.1.15-ga-cuda8.0_1-1_amd64.deb，之后通过

sudo apt-get install libnccl2 libnccl-dev   安装相应的库，之后运行如下验证代码：

# To check if Caffe2 build was successful
python -c 'from caffe2.python import core' 2>/dev/null && echo "Success" || echo "Failure"

# To check if Caffe2 GPU build was successful
# This must print a number > 0 in order to use Detectron
python -c 'from caffe2.python import workspace; print(workspace.NumCudaDevices())'

都ｏｋ说明ｃａｆｆｅ２安装完成。