今天F8开发者大会上，Facebook正式发布Caffe2.经过一天的折腾，终于在ubuntu14.04上成功配置caffe2，现将经验分享如下：
1.ubuntu14.04系统下caffe2所需依赖包安装，很多依赖包在安装caffe时已经安装，这个我们不需要管，安装过得依赖包会自动跳过：
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
build-essential \
cmake \
git \
libgoogle-glog-dev \
libprotobuf-dev \
protobuf-compiler \
python-dev \
python-pip
sudo pip install numpy protobuf
2.GPU（可选择）
如果时caffe老用户，那么这一步自行跳过，安装命令如下：
2.1 CUDA安装
sudo apt-get update && sudo apt-get install wget -y --no-install-recommends
wget "http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_8.0.61-1_amd64.deb"
sudo dpkg -i cuda-repo-ubuntu1404_8.0.61-1_amd64.deb
sudo apt-get update
sudo apt-get install cuda
2.2 CUDNN安装
CUDNN_URL="http://developer.download.nvidia.com/compute/redist/cudnn/v5.1/cudnn-8.0-linux-x64-v5.1.tgz"
wget ${CUDNN_URL}
sudo tar -xzf cudnn-8.0-linux-x64-v5..tgz -C /usr/local
rm cudnn-8.0-linux-x64-v5..tgz && sudo ldconfig
3.可选择的依赖包（这个可以在后期需要时再安装）
sudo apt-get install -y --no-install-recommends \
libgtest-dev \
libiomp-dev \
libleveldb-dev \
liblmdb-dev \
libopencv-dev \
libopenmpi-dev \
libsnappy-dev \
openmpi-bin \
openmpi-doc \
python-pydot
sudo pip install \
flask \
graphviz \
hypothesis \
jupyter \
matplotlib \
pydot python-nvd3 \
pyyaml \
requests \
scikit-image \
scipy \
setuptools \
tornado
# for Ubuntu 14.04
sudo apt-get install -y --no-install-recommends libgflags2
4.clone&build
git clone --recursive https://github.com/caffe2/caffe2.git && cd caffe2
make && cd build && sudo make install
python -c 'from caffe2.python import core' >/dev/null && echo "Success" || echo "Failure"
之后，运行以下代码，验证GPU是否安装成功
python -m caffe2.python.operator_test.relu_op_test
在我运行之后出现如下报错：
Traceback (most recent call last):
File "/home/hzz/anaconda2/lib/python2.7/runpy.py", line , in _run_module_as_main
"__main__", fname, loader, pkg_name)
File "/home/hzz/anaconda2/lib/python2.7/runpy.py", line , in _run_code
exec code in run_globals
File "/home/hzz/caffe2/build/caffe2/python/operator_test/relu_op_test.py", line , in <module>
from hypothesis import given
ImportError: No module named hypothesis
针对提示错误，缺少hypothesis库，运行以下命令
pip install hypothesis
然后再次输入之前测试命令，运行成功，结果如下：
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([[-0.86028236, . , 0.61849821, -0.7184248 , 0.89307433],
[ 0.92490983, . , 0.90946507, -0.37061477, -0.67617983],
[ 0.07383067, 0.55851477, 0.90312296, -0.94337231, 0.95312852],
[-0.50963259, 0.83218467, 0.86731863, 0.63752574, 0.0466823 ],
[ 0.87043667, -. , 0.66159999, 0.10338549, . ]], dtype=float32), gc=, dc=[, device_type: ], engine=u'')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([ -2.72050470e-01, -1.68155816e-44, 1.22070059e-01], dtype=float32), gc=, dc=[, device_type: ], engine=u'')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([[-0.27205047],
[-. ],
[ 0.81497574],
[ . ]], dtype=float32), gc=, dc=[, device_type: ], engine=u'')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([ .], dtype=float32), gc=device_type: , dc=[, device_type: ], engine=u'CUDNN')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([ 0.80068815, 0.1997007 , . , -0.46049505], dtype=float32), gc=, dc=[, device_type: ], engine=u'CUDNN')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([ 0.73503637, -0.1312747 , -0.06368257], dtype=float32), gc=device_type: , dc=[, device_type: ], engine=u'CUDNN')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([ .], dtype=float32), gc=, dc=[, device_type: ], engine=u'CUDNN')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([-0.8335939 , -0.67032146], dtype=float32), gc=device_type: , dc=[, device_type: ], engine=u'CUDNN')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([ 0.05724426, . ], dtype=float32), gc=device_type: , dc=[, device_type: ], engine=u'')
Trying example: test_relu(self=<__main__.TestRelu testMethod=test_relu>, X=array([-0.66229254, 0.59582025, -0.85684592], dtype=float32), gc=device_type: , dc=[, device_type: ], engine=u'CUDNN')
.
----------------------------------------------------------------------
Ran test in .661s OK
5.环境变量设置
echo $PYTHONPATH
# export PYTHONPATH=/usr/local:$PYTHONPATH
# export PYTHONPATH=$PYTHONPATH:/home/hzz/caffe2/build
echo $LD_LIBRARY_PATH
# export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
然后就可以开心的跑caffe2上自带的例子了。