caffe 安装和使用问题列表

1. mkl download: https://software.intel.com/en-us/intel-system-studio
problem : error while loading shared libraries: libmkl_rt.so: cannot open shared object file: No such file or directory
answer: export LD_LIBRARY_PATH=$your_install_path/system_studio_2015.1.045/mkl/lib/intel64/
2.
the python install package, do not use the pip install requirements.txt, it does not work. install one by one by yourself.
    http://www.scipy.org/install.html
3. 
python use problem:
When I tried to use the python binding after compiling caffe with cmake "_caffe" was not found. This was fixed by copying the python lib from the build directory to the python module directory with the expected name. Further caffe's proto files are missing with the cmake build. This is fixed by the consecutive four commands taken from the plain make file.
Assuming you are in the main caffe directory and your cmake build is in "build", the following commands enable python support:
cp build/python/libpycaffe.so python/caffe/_caffe.so && \
mkdir -p python/caffe/proto && \
touch python/caffe/proto/__init__.py && \
protoc --proto_path=src/caffe/proto --python_out=python/caffe/proto src/caffe/proto/caffe.proto && \
protoc --proto_path=src/caffe/proto --python_out=python/caffe/proto src/caffe/proto/caffe_pretty_print.proto
4.  how to log the training and process.
    GLOG_logtostderr=1 ....... 2>&1 | tee log
5. how to log the training and testing table like:
  caffe.log.test ( #Iters Seconds TestAccuracy TestLoss)
    caffe.log.train(#Iters Seconds TrainingLoss LearningRate)
     see tools/extra/parse_log.sh, if you like, you can plot the picture like :