theano.sandbox.cuda error

[awesome-yz]

I have installed theano version 1.0.4, with keras 1.1.0. Have cuda version 10.1 and my .theanorc file consists the following configuration:

[global]
floatX = float32
device = cuda0
optimizer = fast_run

[nvcc]
fastmath = True

[blas]
ldflags = -L/usr/local/lib -lopenblas

[cuda]
root = /usr/local/cuda-10.1/bin

[dnn]
include_path = /usr/include/
library_path = /usr/lib/x84_64-linux-gnu

running the theano gpu.test() gave me this output
Can not use cuDNN on context None: cannot compile with cuDNN. We got this error:
b'In file included from /tmp/try_flags_dh61617g.c:3:0:\n/usr/include/stdio.h:27:10: fatal error: bits/libc-header-start.h: No such file or directory\n #include <bits/libc-header-start.h>\n ^~~~~~~~~~~~~~~~~~~~~~~~~~\ncompilation terminated.\n'
Mapped name None to device cuda0: GeForce GTX 1050 Ti (0000:01:00.0)
[GpuElemwise{exp,no_inplace}(<GpuArrayType(float32, vector)>), HostFromGpu(gpuarray)(GpuElemwise{exp,no_inplace}.0)]
Looping 1000 times took 1.511685 seconds
Result is [1.2317803 1.6187935 1.5227807 ... 2.2077181 2.2996776 1.623233 ]
Used the gpu

Did anyone face the same problem, and how can I resolve this? Also, is there an alternative that will enable me to run the training file?

[Villjoie]

I have the same problem

[sakshijain032]

Comment this line
theano.sandbox.cuda.use('gpu0')

[sanjayroka05]

I have got problem please help me:
AttributeError: module 'theano.sandbox' has no attribute 'cuda'

[Villjoie]

you check, if installed the cuda and cuDNN programs well

[dersugiritlioglu]

One can use like this:

THEANO_FLAGS='device=cuda0' python TrainingAnomalyDetector_public.py