1. centos安装openblas:
yum install openblas-devel
2. 安装python2.7.x, 并pip安装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. 参考官方文档:
http://deeplearning.net/software/theano/install_centos6.html
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.