[theano-users] cuDNN not available

[ephi5757 via theano-users]

When I try to run my python CNN related programs that depend on theano I
get the following statement:
"Using gpu device 0: Quadro P4000 (CNMeM is enabled with initial size:
80.0% of memory, cuDNN not available)".
Then the program crashes with the following error:
("We can't determine the cudnn version as it is not available".)
NOTE: I installed cudnn v6.0 and it is located in
C:\ProgramData\Miniconda2\pkgs\cudnn-6.0-0.

Previously adding 'cuda.disable_gcc_cudnn_check=True' into the
.theanorc.txt worked perfectly.
Now it does not appear to be effective.

How can I get cudnn to be recognized?
Are there any solutions to this problem?
I am so close to be back up and running my CNN programs on a new computer.
Thank you for your help.
Arnold

FYI, here is my .theanorc.txt file:
===========================.
[global]
device = gpu
REM device = cpu
floatX=float32
[cuda]
root=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\
cuda.disable_gcc_cudnn_check=True
[nvcc]
flags = -LC:\ProgramData\Miniconda2\libs
compiler_bindir=C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\bin
cxx=C:\ProgramData\Miniconda2\Library\mingw-w64\bin
optimizer_including=dnn
[lib]
cnmem=0.8
====================

[ephi5757 via theano-users]

Problem solved: "Using gpu device 0: Quadro P4000 (CNMeM is enabled with
initial size: 80.0% of memory, cuDNN 6021)"

see https://github.com/Theano/Theano/issues/3898.


I copied the corresponding files to their respective folders in my CUDA
folder. For example:

- cudnn64_6.dll to C:\Program Files\NVIDIA GPU Computing
Toolkit\CUDA\v8.0\bin
- cudnn.h to C:\Program Files\NVIDIA GPU Computing
Toolkit\CUDA\v8.0\include
- cudnn.lib to C:\Program Files\NVIDIA GPU Computing
Toolkit\CUDA\v8.0\lib\x64

After copying these files, I add these lines to my .theanorc.txt file:

[dnn]
enabled = True

==============================================
Also see https://github.com/Theano/Theano/issues/5972 .

I tried setting env vars as follows :

LIBRARY_PATHC=:\ProgramData\Miniconda2\pkgs\cudnn-6.0-0\Library\lib\x64
CPATHC=:\ProgramData\Miniconda2\pkgs\cudnn-6.0-0\Library\include

*===================================================.*
*Thanks*

Post by ephi5757 via theano-users
When I try to run my python CNN related programs that depend on theano I
80.0% of memory, cuDNN not available)".
("We can't determine the cudnn version as it is not available".)
NOTE: I installed cudnn v6.0 and it is located in
C:\ProgramData\Miniconda2\pkgs\cudnn-6.0-0.
Previously adding 'cuda.disable_gcc_cudnn_check=True' into the
.theanorc.txt worked perfectly.
Now it does not appear to be effective.
How can I get cudnn to be recognized?
Are there any solutions to this problem?
I am so close to be back up and running my CNN programs on a new computer.
Thank you for your help.
Arnold
===========================.
[global]
device = gpu
REM device = cpu
floatX=float32
[cuda]
root=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\
cuda.disable_gcc_cudnn_check=True
[nvcc]
flags = -LC:\ProgramData\Miniconda2\libs
compiler_bindir=C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\bin
cxx=C:\ProgramData\Miniconda2\Library\mingw-w64\bin
optimizer_including=dnn
[lib]
cnmem=0.8
====================