如何在OSX上使用带有GPU的theano?

时间:2016-04-26 15:51:20

标签: python-2.7 gpu theano theano-cuda

操作系统:OSX 10.11.4

GPU:GeForce GT 750M,CUDA7.5

语言:Python Anaconda 2.7

我已按照theano的说明在我的Mac上设置CUDA:

  1. 设置路径:导出路径= / Developer / NVIDIA / CUDA-7.5 / bin:$ PATH
  2. 设置LD_LIBRARY_PATH:export LD_LIBRARY_PATH = / Developer / NVIDIA / CUDA-7.5 / lib:$ LD_LIBRARY_PATH
  3. 创建.theanorc文件并在[cuda]部分添加CUDA_ROOT = / Developer / NVIDIA / CUDA-7.5
  4. 我从theano网站复制了代码并将其粘贴到test.py文件中:

    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 range(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')
    

    但是当我跑这条线时:

    THEANO_FLAGS='floatX=float32,device=gpu0,nvcc.fastmath=True' python test.py

    我收到了这个错误:

    nvcc fatal: The version ('70300') of the host compiler ('Apple clang') is not supported

    并且警告如下:

    WARNING (theano.sandbox.cuda): CUDA is installed, but device gpu0 is not available (error: cuda unavailable)

    是不是因为我的Xcode版本(版本7.3)?如何解决这个问题?

1 个答案:

答案 0 :(得分:2)

我修复了这个错误:“nvcc致命:版本('70300')..”为OSX 10.11.4安装最新版本的CUDA SDK 7.5.27