如何使用Win7在Anaconda中修复史无前例的'import theano'错误

时间:2019-02-01 08:51:16

标签: python-3.x gpu theano

当导入theano时,以下出现     pygpu.gpuarray.GpuArrayException,b'无法加载“ cuGetErrorName”:\ xd5 \ xd2 \ xb2 \ xbb \ xb5 \ xbd \ xd6 \ xb8 \ xb6 \ xa8 \ xb5 \ xc4 \ xb3 \ xcc \ xd0 \ xf2 \ xa1 \ xa3 \ r \ n'     网络搜索中没有类似结果

尝试在kera和CNN中使用gpu,出现了大量错误。到目前为止,“导入pygpu”和cuda安装都可以,但是“ import theano”存在严重错误。在网络搜索中尝试了无数种解决方案,但到目前为止,没有人能奏效。     由于我的笔记本电脑版本较旧,因此使用的软件仅供选择,不能随意选择,     我的笔记本电脑系统是win7和32bit,安装了:     蟒蛇3     python 3.6     CUDA 5.0     pygpu,theano,numpy,mkl,mkl-service,scripy,nose等(根据协议要求打包的所有theano请求)(最新版本,通过conda安装)

already installed(many peckages have been installed,just list a few might be useful according to web search):
 anaconda 2018.12
 balas 1.0
 libgpuarray 0.7.6
 libpython 2.1
 m2w64-toolchain 5.3.0
 mkl 2019.1
 mkl-service 1.1.2
 nose 1.3.7
 numpy 1.15.4
 python 3.7.1(auto install while install anaconda)
 scipy 1.1.0
 theano 1.0.3

.theanorc.txt:
[global]
openmp=False
floatX = float32
device = cuda
fastmath = True
[gcc]
cxxflags = -IC:ProgramData\Anaconda3\MinGW
[nvcc]
flags=-LC:ProgramData\Anaconda3\libs
compiler_bindir=C\Program Files\Microsoft Visual Studio 8\VSTA\Bin
[blas]
ldflags=
[cuda]
root = C:Program Files\NVIDIA GPU Computing Toolkit\CUDA\v5.0

expect:
1.import theano
2.run CNN,keras or anyother deep learning with gpu
but the actual output,
'import pygpu' was ok
cuda 5.0 was installed
but when import theano, came out,
error (theano.gpuarray),could not initialize pygpu,support disabled
traceback,
.......
pygpu.gpuarray.gpuarrayexception: b'could not load "cuGetErrorName":\xd5\xd2\xb2\xbb\xb5\xbd\xd6\xb8\xb6\xa8\xb5\xc4\xb3\xcc\xd0\xf2\xa1\xa3\r\n'

0 个答案:

没有答案