Centos Keras(2.0.4)+Theano(0.9.0)+OpenBLAS CPU方式配置

1. centos安裝openblas：

yum install openblas-devel

2. 安裝python2.7.x, 並pip安裝keras.

3. 配置keras使用theano backend：
vim /root/.keras/keras.json
{
    "epsilon": 1e-07,
    "floatx": "float32",
    "image_data_format": "channels_last",
    "backend": "theano"
}

4. 配置theano使用openblas：

vim /root/.theanorc
[global]
floatX = float32
device = cpu
optimizer = fast_run

[blas]
ldflags = -lopenblas

5. 參考官方文檔:

http://deeplearning.net/software/theano/install_centos6.html

BLAS installation (with Level 3 functionality)
    Recommended: MKL, which is free through Conda with mkl-service package.
    Alternatively, we suggest to install OpenBLAS, with the development headers (-dev, -devel, depending on your Linux distribution).

    

http://deeplearning.net/software/theano/troubleshooting.html

    

http://deeplearning.net/software/theano/library/config.html

config.optimizer

    String value: 'fast_run', 'merge', 'fast_compile', 'None'
    Default: 'fast_run'
    When the mode is Mode, it sets the default optimizer used.