Hi Kristofer,
Did you get a solution for the below issue? I too am facing similar
problem?
Regards,
K. J. Nitthilan
On Thursday, April 21, 2016 at 12:30:22 PM UTC+5:30, Kristofer Krus wrote:
>
> Hi,
>
> When I enter print(theano.sandbox.cuda.dnn.dnn_available.msg), I get the
> message
>
> Theano can not compile with cuDNN. We got this error:
> b''
>
>
> Regards
>
> Kristofer
>
> On Wednesday, April 20, 2016 at 8:00:00 PM UTC+2, nouiz wrote:
>>
>> Make sure to use Theano development version:
>>
>>
>> http://www.deeplearning.net/software/theano/install.html#bleeding-edge-install-instructions
>>
>> There is some fix for python 3.5 in there onwindows from memory.
>>
>> If that don't fix it, you can get more error information by checking:
>>
>> import theano.sandbox.cuda.dnn
>> print theano.sandbox.cuda.dnn.dnn_available.msg
>>
>>
>>
>> On Wed, Apr 20, 2016 at 9:52 AM, Kristofer Krus <[email protected]>
>> wrote:
>>
>>> Hi again,
>>>
>>> I have also installed Theano on another computer, and I'm getting the
>>> same error message on that computer, "CuDNN not available", even though
>>> CuDNN is installed. I have tried everything I can come to think of to try
>>> to make Theano find CuDNN, but I haven't succeeded. Is there some way to
>>> make Theano give a more helpful error message, to find out *why* it
>>> says that CuDNN isn't available?
>>>
>>> Regards
>>>
>>> Kristofer
>>>
>>>
>>> On Thursday, April 14, 2016 at 11:22:48 AM UTC+2, Kristofer Krus wrote:
>>>>
>>>> I managed to make Theano find CuDNN in the VS2013 x64 Cross Tools
>>>> Command Prompt (CTCP) and with Pyton 3.5. CTCP sets up some environment
>>>> variables and paths (some of which Theano needs in order to find CuDNN)
>>>> and
>>>> when I add those to the environment in cmd.exe I get Theano to find CuDNN
>>>> there too. I can also make it work in Git Bash by adding and exporting
>>>> those variables, by adding the paths to PATH, and by using Unix path lists
>>>> instead of Windows path lists (i.e. using colon as list separator instead
>>>> of semicolon, using "/c" instead of "C:", which is the folder Git Bash
>>>> maps
>>>> "C:" to, and by using forward slashes instead of backslashes).
>>>>
>>>> By trial and error, I could eliminate all paths and all variables but
>>>> INCLUDE and LIB, which needed to be defined as
>>>>
>>>> export INCLUDE="/c/Program Files (x86)/Microsoft Visual Studio
>>>> 12.0/VC/INCLUDE:/c/Program Files (x86)/Windows
>>>> Kits/8.1/include/shared:$INCLUDE"
>>>> export LIB="/c/Program Files (x86)/Microsoft Visual Studio
>>>> 12.0/VC/LIB/amd64:/c/Program Files (x86)/Windows
>>>> Kits/8.1/lib/winv6.3/um/x64:$LIB"
>>>>
>>>> in Git Bash, and as
>>>>
>>>> set INCLUDE=C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\
>>>> INCLUDE;C:\Program Files (x86)\Windows Kits\8.1\include\shared;%INCLUDE
>>>> %
>>>> set LIB=C:\Program Files (x86)\Microsoft Visual Studio 12.0\VC\LIB\
>>>> amd64;C:\Program Files (x86)\Windows Kits\8.1\lib\winv6.3\um\x64;%LIB%
>>>>
>>>> in cmd.exe.
>>>>
>>>> My .theanorc file looks like
>>>>
>>>> [global]
>>>> device = gpu
>>>> floatX = float32
>>>>
>>>> [nvcc]
>>>> flags = --use-local-env --cl-version=2013
>>>>
>>>> [lib]
>>>> cnmem=1
>>>>
>>>> Regards
>>>>
>>>> Kristofer
>>>>
>>>> On Tuesday, April 12, 2016 at 4:45:44 PM UTC+2, Doug wrote:
>>>>>
>>>>> Yea python 2.7.
>>>>>
>>>>> On my machine I have following settings for nvcc in my theanorc file,
>>>>> but that is for my setup.
>>>>>
>>>>> flags = --cl-version=2010
>>>>> compiler_bindir = C:\Program Files (x86)\Microsoft Visual Studio
>>>>> 10.0\VC\bin
>>>>>
>>>>> On Friday, April 8, 2016 at 7:17:39 AM UTC-4, Kristofer Krus wrote:
>>>>>>
>>>>>> > Theano can have a hard time detecting CuDnn on windows if you rely
>>>>>> on config flags in your theanorc file to get everything working.
>>>>>>
>>>>>> What config flags in .theanorc are you talking about? Do I need some
>>>>>> specific config flags for CuDNN?
>>>>>>
>>>>>> Regards
>>>>>>
>>>>>> Kristofer
>>>>>>
>>>>>> On Friday, April 8, 2016 at 12:03:50 PM UTC+2, Kristofer Krus wrote:
>>>>>>>
>>>>>>> Hi Doug, thanks for the answer.
>>>>>>>
>>>>>>> I'm using Python 3.5.1; I'm guessing the code you provided me with
>>>>>>> is for Python 2.7? I also guess Theano supports Python 3, or do I have
>>>>>>> to
>>>>>>> switch to Python 2.7? All the documentation I've seen so far has been
>>>>>>> using
>>>>>>> Python 2.7, but then again the documentation seems to be a couple of
>>>>>>> years
>>>>>>> old and I think I've seen somewhere that Python 3 should also be
>>>>>>> supported.
>>>>>>>
>>>>>>> My Python knowledge is limited, but I put parenthesis around the
>>>>>>> argument to print and that enabled the script to run. However, I'm
>>>>>>> getting
>>>>>>>
>>>>>>> Using gpu device 0: GeForce GTX 750 (CNMeM is enabled with initial
>>>>>>> size: 95.0% of memory, CuDNN not available)
>>>>>>> DEBUG: nvcc STDOUT mod.cu
>>>>>>> Creating library C:/Users/<user>/AppData/Local/Theano/
>>>>>>> compiledir_Windows-7-6.1.7601-SP1-
>>>>>>> Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-3.5.1-64/
>>>>>>> tmphps024jr/m91973e5c136ea49268a916ff971b7377.lib and object C:
>>>>>>> /Users/<user>/AppData/Local/Theano/compiledir_Windows-7-6.1.7601-SP1
>>>>>>> -Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-3.5.1-64/
>>>>>>> tmphps024jr/m91973e5c136ea49268a916ff971b7377.exp
>>>>>>>
>>>>>>> DEBUG: nvcc STDOUT mod.cu
>>>>>>> Creating library C:/Users/<user>/AppData/Local/Theano/
>>>>>>> compiledir_Windows-7-6.1.7601-SP1-
>>>>>>> Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-3.5.1-64/
>>>>>>> tmpmz83229a/m454ce67f097d7c8ae04965f0de2d545f.lib and object C:
>>>>>>> /Users/<user>/AppData/Local/Theano/compiledir_Windows-7-6.1.7601-SP1
>>>>>>> -Intel64_Family_6_Model_60_Stepping_3_GenuineIntel-3.5.1-64/
>>>>>>> tmpmz83229a/m454ce67f097d7c8ae04965f0de2d545f.exp
>>>>>>>
>>>>>>> Traceback (most recent call last):
>>>>>>> File "cudnn-test.py", line 12, in <module>
>>>>>>> f = theano.function([img],out)
>>>>>>> File "C:\Program1\Theano\theano\compile\function.py", line 322, in
>>>>>>> function
>>>>>>> output_keys=output_keys)
>>>>>>> File "C:\Program1\Theano\theano\compile\pfunc.py", line 480, in
>>>>>>> pfunc
>>>>>>> output_keys=output_keys)
>>>>>>> File "C:\Program1\Theano\theano\compile\function_module.py", line
>>>>>>> 1779, in orig_function
>>>>>>> defaults)
>>>>>>> File "C:\Program1\Theano\theano\compile\function_module.py", line
>>>>>>> 1643, in create
>>>>>>> input_storage=input_storage_lists, storage_map=storage_map)
>>>>>>> File "C:\Program1\Theano\theano\gof\link.py", line 690, in
>>>>>>> make_thunk
>>>>>>> storage_map=storage_map)[:3]
>>>>>>> File "C:\Program1\Theano\theano\gof\vm.py", line 1005, in make_all
>>>>>>> no_recycling))
>>>>>>> File "C:\Program1\Theano\theano\sandbox\cuda\__init__.py", line
>>>>>>> 256, in make_thunk
>>>>>>> compute_map, no_recycling)
>>>>>>> File "C:\Program1\Theano\theano\gof\op.py", line 978, in
>>>>>>> make_thunk
>>>>>>> no_recycling)
>>>>>>> File "C:\Program1\Theano\theano\gof\op.py", line 881, in
>>>>>>> make_c_thunk
>>>>>>> output_storage=node_output_storage)
>>>>>>> File "C:\Program1\Theano\theano\gof\cc.py", line 1200, in
>>>>>>> make_thunk
>>>>>>> keep_lock=keep_lock)
>>>>>>> File "C:\Program1\Theano\theano\gof\cc.py", line 1143, in
>>>>>>> __compile__
>>>>>>> keep_lock=keep_lock)
>>>>>>> File "C:\Program1\Theano\theano\gof\cc.py", line 1595, in
>>>>>>> cthunk_factory
>>>>>>> key=key, lnk=self, keep_lock=keep_lock)
>>>>>>> File "C:\Program1\Theano\theano\gof\cmodule.py", line 1165, in
>>>>>>> module_from_key
>>>>>>> self.module_hash_to_key_data[module_hash] = key_data
>>>>>>> File "C:\Program1\Python35\lib\contextlib.py", line 77, in
>>>>>>> __exit__
>>>>>>> self.gen.throw(type, value, traceback)
>>>>>>> File "C:\Program1\Theano\theano\gof\compilelock.py", line 40, in
>>>>>>> lock_ctx
>>>>>>> yield
>>>>>>> File "C:\Program1\Theano\theano\gof\cmodule.py", line 1142, in
>>>>>>> module_from_key
>>>>>>> module = lnk.compile_cmodule(location)
>>>>>>> File "C:\Program1\Theano\theano\gof\cc.py", line 1506, in
>>>>>>> compile_cmodule
>>>>>>> preargs=preargs)
>>>>>>> File "C:\Program1\Theano\theano\sandbox\cuda\nvcc_compiler.py",
>>>>>>> line 414, in compile_str
>>>>>>> return dlimport(lib_filename)
>>>>>>> File "C:\Program1\Theano\theano\gof\cmodule.py", line 299, in
>>>>>>> dlimport
>>>>>>> rval = __import__(module_name, {}, {}, [module_name])
>>>>>>> RuntimeError: ('The following error happened while compiling the
>>>>>>> node', GpuDnnPool{mode='max'}(GpuContiguous.0, TensorConstant{(2,)
>>>>>>> of 2}, TensorConstant{(2,) of 1}, TensorConstant{(2,) of 0}), '\n',
>>>>>>> 'could
>>>>>>> not create cuDNN handle: CUDNN_STATUS_INTERNAL_ERROR',
>>>>>>> "[GpuDnnPool{mode='max'}(<CudaNdarrayType(float32,
>>>>>>> 4D)>, TensorConstant{(2,) of 2}, TensorConstant{(2,) of 1},
>>>>>>> TensorConstant{(2,) of 0})]")
>>>>>>>
>>>>>>> Although I'm having some difficulties with interpreting the result,
>>>>>>> it doesn't look too good to me.
>>>>>>>
>>>>>>> By the way, which versions of of CuDNN are currently supported by
>>>>>>> Theano (Theano 0.8)? I stumbled upon a thread
>>>>>>> <https://groups.google.com/forum/#!topic/theano-users/JvfbAOXZicU>
>>>>>>> asking when CuDNN 5.0 will be supported (which by the w
>>>>>>> ...
>>>>>>
>>>>>> --
>>>
>>> ---
>>> You received this message because you are subscribed to the Google
>>> Groups "theano-users" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to [email protected].
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>
>>
--
---
You received this message because you are subscribed to the Google Groups
"theano-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
