Theano 在 Windows 上的配置

注意:不建议在 `windows` 进行 `theano` 的配置。

务必确认你的显卡支持 `CUDA`。

我个人的电脑搭载的是 Windows 10 x64 系统,显卡是 Nvidia GeForce GTX 850M

安装 theano

首先是用 anaconda 安装 theano

conda install mingw libpython
pip install theano

安装 VS 和 CUDA

按顺序安装这两个软件:

  • 安装 Visual Studio 2010/2012/2013
  • 安装 对应的 x64 或 x86 CUDA

Cuda 的版本与电脑的显卡兼容。

我安装的是 Visual Studio 2012 和 CUDA v7.0v。

配置环境变量

CUDA 会自动帮你添加一个 CUDA_PATH 环境变量(环境变量在 控制面板->系统与安全->系统->高级系统设置 中),表示你的 CUDA 安装位置,我的电脑上为:

  • CUDA_PATH
    • C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v7.0

我们配置两个相关变量:

  • CUDA_BIN_PATH
    • %CUDA_PATH%\bin
  • CUDA_LIB_PATH
    • %CUDA_PATH%\lib\Win32

接下来在 Path 环境变量的后面加上:

  • Minicoda 中关于 mingw 的项:

    • C:\Miniconda\MinGW\bin;
    • C:\Miniconda\MinGW\x86_64-w64-mingw32\lib;
  • VS 中的 cl 编译命令:

    • C:\Program Files (x86)\Microsoft Visual Studio 11.0\VC\bin;
    • C:\Program Files (x86)\Microsoft Visual Studio 11.0\Common7\IDE;

生成测试文件:

In [1]:
%%file test_theano.py
from theano import config
print 'using device:', config.device
Writing test_theano.py

我们可以通过临时设置环境变量 THEANO_FLAGS 来改变 theano 的运行模式,在 linux 下,临时环境变量直接用:

THEANO_FLAGS=xxx 

就可以完成,设置完成之后,该环境变量只在当前的命令窗口有效,你可以这样运行你的代码:

THEANO_FLAGS=xxx python <your script>.py

Windows 下,需要使用 set 命令来临时设置环境变量,所以运行方式为:

set THEANO_FLAGS=xxx && python <your script>.py 
In [2]:
import sys

if sys.platform == 'win32':
    !set THEANO_FLAGS=mode=FAST_RUN,device=cpu,floatX=float32 && python test_theano.py
else:
    !THEANO_FLAGS=mode=FAST_RUN,device=cpu,floatX=float32 python test_theano.py
using device: cpu

In [3]:
if sys.platform == 'win32':
    !set THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 && python test_theano.py
else:
    !THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python test_theano.py
Using gpu device 0: Tesla C2075 (CNMeM is disabled)
using device: gpu

测试 CPUGPU 的差异:

In [4]:
%%file test_theano.py

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))

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('Used the gpu')
Overwriting test_theano.py
In [5]:
if sys.platform == 'win32':
    !set THEANO_FLAGS=mode=FAST_RUN,device=cpu,floatX=float32 && python test_theano.py
else:
    !THEANO_FLAGS=mode=FAST_RUN,device=cpu,floatX=float32 python test_theano.py
Looping 1000 times took 3.498123 seconds

Result is [ 1.23178029  1.61879337  1.52278066 ...,  2.20771813  2.29967761

  1.62323284]

Used the cpu

In [6]:
if sys.platform == 'win32':
    !set THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 && python test_theano.py
else:
    !THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python test_theano.py
Using gpu device 0: Tesla C2075 (CNMeM is disabled)
Looping 1000 times took 0.847006 seconds
Result is [ 1.23178029  1.61879349  1.52278066 ...,  2.20771813  2.29967761
  1.62323296]
Used the gpu

可以看到 GPU 明显要比 CPU 快。

使用 GPU 模式的 T.exp(x) 可以获得更快的加速效果:

In [7]:
%%file test_theano.py

from theano import function, config, shared, sandbox
import theano.sandbox.cuda.basic_ops
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), 'float32'))
f = function([], sandbox.cuda.basic_ops.gpu_from_host(T.exp(x)))

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,))
print("Numpy result is %s" % (numpy.asarray(r),))
if numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()]):
    print('Used the cpu')
else:
    print('Used the gpu')
Overwriting test_theano.py
In [8]:
if sys.platform == 'win32':
    !set THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 && python test_theano.py
else:
    !THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python test_theano.py
Using gpu device 0: Tesla C2075 (CNMeM is disabled)
Looping 1000 times took 0.318359 seconds
Result is <CudaNdarray object at 0x7f7bb701fb70>
Numpy result is [ 1.23178029  1.61879349  1.52278066 ...,  2.20771813  2.29967761
  1.62323296]
Used the gpu
In [9]:
!rm test_theano.py

配置 .theanorc.txt

我们可以在个人文件夹下配置 .theanorc.txt 文件来省去每次都使用环境变量设置的麻烦:

例如我现在的 .theanorc.txt 配置为:

[global]
device = gpu
floatX = float32

[nvcc]
fastmath = True
flags = -LC:\Miniconda\libs
compiler_bindir=C:\Program Files (x86)\Microsoft Visual Studio 11.0\VC\bin

[gcc]
cxxflags = -LC:\Miniconda\MinGW

具体这些配置有什么作用之后可以查看官网上的教程。