本文主要包含2个部分，包括：
- 第一部分 nVidia CUDA Toolkit的安装（*.deb方法）
- 第二部分 Caffe的安装和测试
一、安装CUDA
https://developer.nvidia.com/cuda-downloads下载，
http://developer.download.nvidia.com/compute/cuda/8.0/secure/Prod2/local_installers/cuda-repo-ubuntu1604-8-0-local-ga2_8.0.61-1_amd64.deb?autho=1489134159_f583eee0b77d9f937741dcacfde61a22&file=cuda-repo-ubuntu1604-8-0-local-ga2_8.0.61-1_amd64.deb
选择Linux/x6_64/Ubuntu/16.04/deb(local)/cuda-repo-ubuntu1604-8-0-local-ga2_8.0.61-1_amd64.deb，较慢，建议多线程下载。
安装
CUDA8.0 Installation Instructions:
sudo dpkg -i cuda-repo-ubuntu1604-8-0-local-ga2_8.0.61-1_amd64.deb
sudo apt-get update
sudo apt-get install cuda
CUDA9.1 Installation Instructions:
sudo dpkg -i cuda-repo-ubuntu1604-9-1-local_9.1.85-1_amd64.deb
sudo apt-key add /var/cuda-repo-/7fa2af80.pub
sudo apt-get update
sudo apt-get install cuda`
验证
cd /usr/local/cuda-8.0/samples/
sudo make
cd bin/x86_64/linux/release
./deviceQuery
网上注册下载Nvidia Deep Learning Toolkit的cuDNN
https://developer.nvidia.com/rdp/cudnn-download
安装
tar -zxvf cudnn-8.0-linux-x64-v5.1.tgz
cd cuda
sudo cp include/* /usr/local/cuda/include/
sudo cp lib64/* /usr/local/cuda/lib64/
二、安装Caffe
1. 安装包
sudo apt-get update
sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-serial-dev protobuf-compiler
sudo apt-get install libgflags-dev libgoogle-glog-dev liblmdb-dev
sudo apt-get install --no-install-recommends libboost-all-dev
sudo apt-get install libatlas-base-dev
sudo apt-get install build-essential cmake git pkg-config
sudo apt-get install python-dev
sudo apt-get install python-pip
sudo apt-get install python-numpy python-scipy
2. 下载caffe
cd /cnn
sudo git clone https://github.com/BVLC/caffe.git
3. 修改设置
cd caffe
cp Makefile.config.example Makefile.config
vi Makefile.config
仅显示修改部分：
#有CUDA：
USE_CUDNN := 1
#仅用cpu：
CPU_ONLY := 1
#解决新版本下，HDF5的路径问题：
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/lib/x86_64-linux-gnu/hdf5/serial/include
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu/hdf5/serial
#查看opencv版本：$pkg-config --modversion opencv，若是2的版本，下面一句话不必。
OPENCV_VERSION =3
4. 编译caffe-master
make all -j16
make test -j16
make runtest -j16
make pycaffe -j16
make matcaffe -j16
5. 重装caffe
sudo make clean
make clean 以后，一定要重新安装caffe所需的第三方依赖库
重新执行1和4步
卸载CUDA8而装了CUDA10之后，caffe需重新编译，出现问题：nvcc fatal : Unsupported gpu architecture 'compute_20'
vi Makefile.config
删除CUDA_ARCH :中的以下两行
-gencode arch=compute_20,code=sm_20 \
-gencode arch=compute_20,code=sm_21 \
6. 使用MNIST数据集进行测试
数据预处理
sh data/mnist/get_mnist.sh
重建lmdb文件
sh examples/mnist/create_mnist.sh
训练mnist
sh examples/mnist/train_lenet.sh
