MAC-OS: The nvidia driver version installed with this OS does not give good results for reduction

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi, I get above error after following the installation instructions on this page: http://deeplearning.net/software/theano/install_macos.html

Using gpu device 0: GeForce GTX 1080 Ti (CNMeM is disabled, cuDNN not available)
Traceback (most recent call last):
File “”, line 1, in
File “/Users/dirkwillemwonnink/miniconda2/lib/python2.7/site-packages/theano/init.py”, line 116, in
theano.sandbox.cuda.tests.test_driver.test_nvidia_driver1()
File “/Users/dirkwillemwonnink/miniconda2/lib/python2.7/site-packages/theano/sandbox/cuda/tests/test_driver.py”, line 41, in test_nvidia_driver1
raise Exception("The nvidia driver version installed with this OS "
Exception: The nvidia driver version installed with this OS does not give good results for reduction.Installing the nvidia driver available on the same download page as the cuda package will fix the problem: http://developer.nvidia.com/cuda-downloads

Although initially things seems to go ok: “Using gpu device 0: GeForce GTX 1080 Ti (CNMeM is disabled, cuDNN not available)”
I don’t really find an alternative video driver for my 1080Ti card / os Sierra 10.12.6. Current driver version : 378.05.05.25f01

Thanks for your suggestions.

2

How does the following indicate that “things seem to go OK”?

Doesn’t that indicate that Theano is actually unable to use the GPU it found, because it would need cuDNN to do that? Why does it show up as unavailable? Have you installed cuDNN on this machine?

I have no insights into Theano’s complaints about the CUDA driver. Maybe it requires specific CUDA driver versions, or a specific version of CUDA? Have you checked with Theano support (no idea how the vendor offers support, a mailing list or user forum perhaps)?

3

Thanks njuffa.
I am no expert but earlier Cuda gave already much GPu accelleration for Theano. The CuDNN option is, as far as I know, expanding that with specific optimized (dnn) functions. So first I think I need to have the gpu activated before CuDNN wil make sense.

I had both working on my Mac some weeks ago (with Cuda 8) but with a warning:
(Using gpu device 0: GeForce GTX 1080 Ti (CNMeM is disabled, cuDNN 5110))

“WARNING (theano.sandbox.cuda): The cuda backend is deprecated and will be removed in the next release (v0.10). Please switch to the gpuarray backend. You can get more information about how to switch at this URL:
Converting to the new gpu back end(gpuarray) · Theano/Theano Wiki · GitHub”

I guess the “gpuarray backend” might be causing the issue now.
(There is a Theano community - I also reached out there.)