Theano

windows下

• Download Anaconda now!
• conda install mingw libpython
• pip install theano

Keras

Windows下

通過 conda install keras 或 pip install keras直接安裝。（會預設的給你安裝keras最新版本和所需要的theano）

因為windows版本的tensorflow剛剛才推出，所以目前支援性不太好。但是keras的backend 同時支援tensorflow和theano. 並且預設是tensorflow, 因此在win本上需要更改backend為theano才能執行。Linux中切換backend同理！

• 將C:\Anaconda2\Lib\site-packages\keras\backend\__init__.py的line 27修改：

# Default backend: TensorFlow.
#_BACKEND = 'tensorflow'
_BACKEND = 'theano'

then, python-> import keras

• 出現 tensorflow提示錯誤的話，需要修改下面的位置的內容: C:\Users\Administrator\.keras\keras.json

{  
"image_dim_ordering":"tf",  
"epsilon"
 
:1e-07,  
"floatx":"float32",  
"backend":"tensorflow"  
}  

{  
    "image_dim_ordering": "tf",   
    "epsilon": 1e-07,   
    "floatx": "float32",   
    "backend": "theano"  
}  

切換後端

如果你至少執行過一次Keras，你將在下面的目錄下找到Keras的配置檔案：

~/.keras/keras.json

如果該目錄下沒有該檔案，你可以手動建立一個

檔案的預設配置如下：

{
"image_dim_ordering
 
":"tf",
"epsilon":1e-07,
"floatx":"float32",
"backend":"tensorflow"
}

將backend欄位的值改寫為你需要使用的後端：theano或tensorflow，即可完成後端的切換

我們也可以通過定義環境變數KERAS_BACKEND來覆蓋上面配置檔案中定義的後端：

KERAS_BACKEND=tensorflow python -c "from keras import backend;"
Using TensorFlow backend.

keras.json 細節

{
    "image_dim_ordering": "tf",
    "epsilon": 1e-07,
    "floatx": "float32",
    "backend": "tensorflow"
}

你可以更改以上~/.keras/keras.json中的配置

• image_dim_ordering：字串，”tf”或”th”，該選項指定了Keras將要使用的維度順序，可通過keras.backend.image_dim_ordering()來獲取當前的維度順序。對2D資料來說，tf假定維度順序為(rows,cols,channels)而th假定維度順序為(channels, rows, cols)。對3D資料而言，tf假定(conv_dim1, conv_dim2, conv_dim3, channels)，th則是(channels, conv_dim1, conv_dim2, conv_dim3)
• epsilon：浮點數，防止除0錯誤的小數字
• floatx：字串，”float16”, “float32”, “float64”之一，為浮點數精度
• backend：字串，所使用的後端，為”tensorflow”或”theano”

win10 bash下配置環境

• bash Anaconda….sh
• source ~/.bashrc
• sudo /home/sun/anaconda2/bin/pip install theano
• sudo /home/sun/anaconda2/bin/pip install keras
• sudo /home/sun/anaconda2/bin/pip install –upgrade gensim

Cuda

安裝CUDA

檢視系統資訊

$ uname -m && cat /etc/*release

驗證NVIDIA顯示卡

$ lspci | grep -i nvidia

安裝gcc及g++

就是安裝C++開發環境，因為CUDA是基於C/C++開發的，當然現在也支援很多其他語言

$ sudo apt-get install gcc g++

安裝linux kernel header及開發包

$ sudo apt-get install linux-headers-$(uname -r)

為了避免出現標頭檔案相關的問題，推薦安裝上build-essential：

$ sudo apt-get install build-essential

合併：
$ sudo apt-get install gcc g++ linux-headers-$(uname -r) build-essential -y

官方文件

Installation Instructions:
`sudo dpkg -i cuda-repo-ubuntu1604-8-0-local-ga2_8.0.61-1_amd64.deb`
`sudo apt-get update`
`sudo apt-get install cuda`

3.6. Ubuntu
1. Perform the pre-installation actions.
2. Install repository meta-data

When using a proxy server with aptitude, ensure that wget is set up to use the
same proxy settings before installing the cuda-repo package.

$ sudo dpkg -i cuda-repo-<distro>_<version>_<architecture>.deb

1. Update the Apt repository cache

$ sudo apt-get update

1. Install CUDA

$ sudo apt-get install cuda

1. Perform the post-installation actions

驗證

重啟顯示卡
[email protected]:/etc/network# systemctl status nvidia-persistenced.service

驅動

apt install software-properties-common –> add-apt-repository ppa:graphics-drivers/ppa(可選)

apt install ubuntu-drivers-common –> ubuntu-drives list –> ubuntu-drivers autoinstall

nvidia-detector

[email protected]:/etc/network# nvidia-smi
Wed Mar 15 16:06:11 2017
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 378.13                 Driver Version: 378.13                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Tesla K20c          Off  | 0000:03:00.0     Off |                    0 |
| 40%   48C    P0    57W / 225W |      0MiB /  4742MiB |     95%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

cuda

安裝：
./cuda*.run

cd ~/NVIDIA_CUDA-8.0_Samples/ –> make all

cd ~/NVIDIA_CUDA-8.0_Samples/1_Utilities/deviceQuery
–> nvcc -I '../../common/inc/' deviceQuery.cpp -o deviceQuery
–> ./deviceQuery

nvcc -V

環境變數

vim /etc/bash.bashrc

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}}

vim ~/.bashrc