参考：Keras中文文档：http://keras-cn.readthedocs.io/en/latest/getting_started/keras_linux/

1、

1、先安装anaconda.

下载python2.7 https://www.continuum.io/downloads

安装命令

bash Anaconda2-4.2.0-Linux-x86_64.sh

2. Ubuntu初始环境设置

 

安装开发包打开终端输入：

# 系统升级

>>> sudo apt update

>>> sudo apt upgrade

# 安装python基础开发包

>>> sudo apt install -y python-devpython-pip python-nose gcc g++ git gfortran vim

安装运算加速库打开终端输入：

>>> sudo apt install -ylibopenblas-dev liblapack-dev libatla

 

3. CUDA开发环境的搭建(CPU加速跳过)

如果您的仅仅采用cpu加速，可跳过此步骤 -下载CUDA8.0

下载地址：

https://developer.nvidia.com/cuda-downloads 

之后打开终端输入：

>>> sudo dpkg -i cuda-repo-ubuntu1604-8-0-local_8.0.44-1_amd64.deb

>>> sudo apt update

>>> sudo apt install cuda

自动配置成功就好。

·        将CUDA路径添加至环境变量在终端输入：

>>> sudo vim /etc/bash.bashrc

在bash.bashrc文件中添加：

export CUDA_HOME=/usr/local/cuda-8.0

export PATH=/usr/local/cuda-8.0/bin${PATH:+:${PATH}}

export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

之后source vim/etc/.bashrc即可同样，在终端输入：

>>> sudo vim ~/.bashrc

在.bashrc中添加如上相同内容（如果您使用的是zsh，在~/.zshrc添加即可）

·        测试在终端输入：

>>> nvcc -V (注意大写)

会得到相应的nvcc编译器相应的信息，那么CUDA配置成功了。记得重启系统

安装cuDNN 5.1, Linux目前就是cudnn-8.0-win-x64-v5.1-prod.zip。

tar ‐zxf cudnn‐8.0‐linux‐x64‐v5.1.tgz
cd cuda
sudo cp lib64/* /usr/local/cuda/lib64/
sudo cp include/* /usr/local/cuda/include/

Keras框架搭建

相关开发包安装

在终端中输入:

>>> sudo pip install -U --pre pipsetuptools wheel

>>> sudo pip install -U --prenumpy scipy matplotlib scikit-learn scikit-image

>>>  pip install -U --pre theano（注意不要带sudo）

>>> pip install -U --pre keras

安装完毕后，输入python，然后输入：

>>> import theano

>>> import keras

如果没有任何提示，则表明安装已经成功

Import keras 时会出现错误，需要配置keras的环境。

Keras环境设置

• 修改默认keras后端 在终端中输入:

>>> vim ~/.keras/keras.json

•   "image_dim_ordering": "th",
•     "epsilon": 1e-07,
•     "floatx": "float32",
•     "backend": "theano"

配置theano文件在终端中输入:

>>> vim ~/.theanorc

并写入以下：

[global]

openmp=False

device = gpu  

floatX = float32 

allow_input_downcast=True 

[lib]

cnmem = 0.8

[blas]

ldflags= -lopenblas

[nvcc]

fastmath = True 

 

之后可以验证keras是否安装成功,在命令行中输入Python命令进入Python变成命令行环境：

 

>>>import keras

 

 

速度测试

 

新建一个文件test.py，内容为：

 

from theano import function, config,shared, sandbox

import theano.tensor as T

import numpy

import time

 

vlen = 10 * 30 * 768  # 10 x #cores x # threads per core 

iters = 1000

 

rng = numpy.random.RandomState(22)

x = shared(numpy.asarray(rng.rand(vlen),config.floatX))

f = function([], T.exp(x))

print(f.maker.fgraph.toposort())

t0 = time.time()

for i in xrange(iters):

    r= f()

t1 = time.time()

print("Looping %d times took %f seconds"% (iters, t1 - t0))

print("Result is %s" % (r,))

if numpy.any([isinstance(x.op, T.Elemwise)for x in f.maker.fgraph.toposort()]):

   print('Used the cpu')

else:

print('Usedthe gpu')

结果：$python test.py
Using gpu device 0: GeForce GTX 1070 (CNMeM is enabled with initial size: 80.0% of memory, cuDNN 5105)
/home/aaa/anaconda2/lib/python2.7/site-packages/theano/sandbox/cuda/__init__.py:600: UserWarning: Your cuDNN version is more recent than the one Theano officially supports. If you see any problems, try updating Theano or downgrading cuDNN to version 5.
  warnings.warn(warn)
[GpuElemwise{exp,no_inplace}(<CudaNdarrayType(float32, vector)>), HostFromGpu(GpuElemwise{exp,no_inplace}.0)]
Looping 1000 times took 0.380512 seconds
Result is [ 1.23178029  1.61879349  1.52278066 ...,  2.20771813  2.29967761
  1.62323296]
Used the gpu