Hello World from OSS Silicon Valley

_ Prerequisite

• Docker installation (You can refer HowToUse/Docker/1.11)

_ Install&Setup

Step.1

Create docker image for caffe.

$ git clone https://github.com/BVLC/caffe.git
$ sudo docker build -t caffe:cpu standalone/cpu

Step.2

Test if the docker image can run as expected.

$ sudo docker run -ti caffe:cpu caffe --version

If there is no problem, you would see the message like the followings.

libdc1394 error: Failed to initialize libdc1394
caffe version 1.0.0-rc3

_ HowToUse

Step.1

Launch docker process.

$ sudo docker run -ti --name <container name> caffe:cpu bash

Step.2

Compile the executables.

# cd ${CAFFE_ROOT}
# cp Makefile.config.example Makefile.config
# vi Makefile.config

...
CPU_ONLY: 1
...
CUSTOM_CXX: g++
...

# make
# make test
# make runtest

Step.3

Prepare sample images.

$ cd ${CAFFE_ROOT}/data/mnist
$ ./get_mnist.sh
$ cd ${CAFFE_ROOT}
$ ./examples/mnist/create_mnist.sh

Step.4

Start the sample training.

$ vi examples/mnist/lenet_solver.prototxt

examples/mnist/lenet_solver.prototxt

$ ./examples/mnist/train_lenet.sh

_ Author

S.Yatsuzuka