1. Abstract

Caffe2 is a deep learning framework that provides an easy and straightforward way for you to experiment with deep learning and leverage community contributions of new models and algorithms. You can bring your creations to scale using the power of GPUs in the cloud or to the masses on mobile with Caffe2’s cross-platform libraries.

The original Caffe framework was useful for large-scale product use cases, especially with its unparalleled performance and well tested C++ codebase. Caffe has some design choices that are inherited from its original use case: conventional CNN applications. As new computation patterns have emerged, especially distributed computation, mobile, reduced precision computation, and more non-vision use cases, its design has shown some limitations.

Caffe2 improves Caffe 1.0 in a series of directions:

• first-class support for large-scale distributed training
• mobile deployment
• new hardware support (in addition to CPU and CUDA)
• flexibility for future directions such as quantized computation
• stress tested by the vast scale of Facebook applications

For more information, please check:
https://caffe2.ai/docs/caffe-migration.html

2. Install Dependencies

• Common install

sudo apt-get update
sudo apt-get install -y --no-install-recommends \
      build-essential \
      cmake \
      git \
      libgoogle-glog-dev \
      libgtest-dev \
      libiomp-dev \
      libleveldb-dev \
      liblmdb-dev \
      libopencv-dev \
      libopenmpi-dev \
      libsnappy-dev \
      libprotobuf-dev \
      openmpi-bin \
      openmpi-doc \
      protobuf-compiler \
      python-dev \
      python-pip                          
sudo pip install \
      future \
      numpy \
      protobuf

• Install libgflags

# for Ubuntu 14.04
sudo apt-get install -y --no-install-recommends libgflags2
# for Ubuntu 16.04
sudo apt-get install -y --no-install-recommends libgflags-dev

3. Clone & Build

• Process

git clone --recursive https://github.com/caffe2/caffe2.git && cd caffe2

# This will build Caffe2 in an isolated directory so that Caffe2 source is
# unaffected
mkdir build && cd build

# This configures the build and finds which libraries it will include in the
# Caffe2 installation. The output of this command is very helpful in debugging
cmake ..

# This actually builds and installs Caffe2 from makefiles generated from the
# above configuration step
sudo make install

• Comment, you may get into trouble when clone third_party of caffe2, please run the script manually

git clone https://github.com/ARM-software/ComputeLibrary.git
git clone https://github.com/Maratyszcza/FP16.git
git clone https://github.com/Maratyszcza/FXdiv.git
git clone https://github.com/Maratyszcza/NNPACK.git
git clone https://github.com/zdevito/aten
git clone https://github.com/google/benchmark.git
git clone https://github.com/Maratyszcza/cpuinfo.git
git clone https://github.com/NVlabs/cub.git
git clone https://github.com/RLovelett/eigen.git
git clone https://github.com/facebookincubator/gloo
git clone https://github.com/google/googletest.git
git clone https://github.com/Yangqing/ios-cmake.git
git clone https://github.com/nvidia/nccl.git
git clone https://github.com/NervanaSystems/nervanagpu.git
git clone https://github.com/onnx/onnx.git
git clone https://github.com/google/protobuf.git
git clone https://github.com/Maratyszcza/psimd.git
git clone https://github.com/Maratyszcza/pthreadpool.git
git clone https://github.com/pybind/pybind11.git
git clone https://github.com/PeachPy/enum34.git
git clone https://github.com/Maratyszcza/PeachPy.git
git clone https://github.com/benjaminp/six.git
git clone https://github.com/facebook/zstd.git

• Comment, something error about python environment. You may have another version of Python installed or need to support Python version 3 for other projects.

Try virtualenv or Anaconda.
The [Anaconda](https://www.continuum.io/downloads) platform provides a single script to install many of the necessary packages for Caffe2, including Python.
Using Anaconda is outside the scope of these instructions, but if you are interested, it may work well for you.

Another way, it is necessary to execute cmd 'which python' to confirm which python you are using.
If it is not you want, you should modify the user's .profile to change the 'python', then 'rm -rf build', then 'mkdir build && cd build' and 'cmake ..' again.

• Comment, you may experience errors related to confu or PeachPy when running make install.

Install dependencies of NNPACK: 
sudo pip install --upgrade git+https://github.com/Maratyszcza/PeachPy 
sudo pip install --upgrade git+https://github.com/Maratyszcza/confu

• Comment, something error like below

Scanning dependencies of target observer_test
[  4%] Building CXX object caffe2/CMakeFiles/observer_test.dir/core/observer_test.cc.o
[  4%] Linking CXX executable ../bin/observer_test
[  4%] Built target observer_test
Scanning dependencies of target mpi_test
[  4%] Building CXX object caffe2/CMakeFiles/mpi_test.dir/mpi/mpi_test.cc.o
[  4%] Linking CXX executable ../bin/mpi_test
/usr/bin/ld: CMakeFiles/mpi_test.dir/mpi/mpi_test.cc.o: undefined reference to symbol '_ZN3MPI8Datatype4FreeEv'
//usr/lib/libmpi_cxx.so.1: error adding symbols: DSO missing from command line
collect2: error: ld returned 1 exit status
caffe2/CMakeFiles/mpi_test.dir/build.make:100: recipe for target 'bin/mpi_test' failed
make[2]: *** [bin/mpi_test] Error 1
CMakeFiles/Makefile2:2612: recipe for target 'caffe2/CMakeFiles/mpi_test.dir/all' failed
make[1]: *** [caffe2/CMakeFiles/mpi_test.dir/all] Error 2
Makefile:138: recipe for target 'all' failed
make: *** [all] Error 2

Solutions
http://blog.sina.com.cn/s/blog_7328cc4d0102wdwv.html

4. Test

cd ~ && python -c 'from caffe2.python import core' 2>/dev/null && echo "Success" || echo "Failure"