Theano里配置GPU的新方法

来源：互联网发布：jenkins mac 安装包编辑：程序博客网时间：2024/06/07 06:31

这几天在研究pylearn2，里面有要用到GPU，之前一直很疑惑：明明显卡是支持CUDA的，CUDA也成功安装了，也可以检测得到GPU，为什么在python程序里面就是没法运用GPU

运行下面这个检测程序

# -*- coding: utf-8 -*-"""Created on Tue Sep 29 08:09:11 2015@author: Administrator"""import theanoimport theanofrom theano import function, config, sharedimport theano.tensor as Timport numpyimport timevlen = 10 * 30 * 768  # 10 x #cores x # threads per coreiters = 100rng = 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 theano.config.deviceprint("Result is %s" % (r,))if numpy.any([isinstance(x.op, T.Elemwise) for x in f.maker.fgraph.toposort()]):    print isinstance(x.op, T.Elemwise)    print('Used the cpu')else:    print('Used the gpu')

显示的也是CPU

因此，我只能怀疑是.theanorc.txt的配置出了问题，好在还有另外一种方式来配置GPU，那就是设置如下的环境变量

里面的具体内容为：

nvcc.compiler_bindir=E:\\Microsoft Visual Studio 12.0\\VC\\bin,base_compilerdir=path_to_a_directory_without_such_characters,floatX=float32,device=gpu,nvcc.fastmath=True,nvcc.flags=-LE:\\Anaconda\\libs,gcc.cxxflags = -IE:\\Anaconda\\MinGW,blas.ldflags=,flags = -arch=sm_30
nvcc fatal:Cannot find compiler 'cl.exe' in PATH

再在path里面加入cl.exe的路径，这样就配置完成了，在cmd里面运行一下

大功告成！

0 0