I have a working Theano installation set up on my older PC (GTX1080, CUDA 9)

I have a new PC fitted with a 2080Ti card that I am porting my projects to. 

>From experimentation so far, it seems that I am forced to use CUDA 10.1 
with my 2080Ti (because CUDA < 10 installations fail to find the device). 
This should be no problem though, as from what I understand, Theano 1.0.3 
does support CUDA 10.1

I have followed the setup instructions here: 
http://deeplearning.net/software/theano/requirements.html as I did before.
I note that by using Theano 1.0.3 I am not able to use pygpu 0.6.9 as 
recommended. This is because the dependency solver demands that I use pygpu 
0.7.6

I am unable to get the environment to use the gpu:

[code]

from theano import function, config, shared, tensor
import numpy
import time

vlen = 10 * 30 * 768  # 10 x #cores x # threads per core
iters = 1000

rng = numpy.random.RandomState(22)
x = shared(numpy.asarray(rng.rand(vlen), config.floatX))
f = function([], tensor.exp(x))
print(f.maker.fgraph.toposort())
t0 = time.time()
for i in range(iters):
    r = f()
t1 = time.time()
print("Looping %d times took %f seconds" % (iters, t1 - t0))
print("Result is %s" % (r,))
if numpy.any([isinstance(x.op, tensor.Elemwise) and
              ('Gpu' not in type(x.op).__name__)
              for x in f.maker.fgraph.toposort()]):
    print('Used the cpu')
else:
    print('Used the gpu')

[/code]

Result:

[code]
C:\Users\Dave\Anaconda3\envs\TheanoPy3.5\lib\site-packages\theano\gpuarray\dnn.py:184:
 
UserWarning: Your cuDNN version is more recent than Theano. If you 
encounter problems, try updating Theano or downgrading cuDNN to a version 
>= v5 and <= v7.
  warnings.warn("Your cuDNN version is more recent than "
Using cuDNN version 7600 on context None
ERROR (theano.gpuarray): Could not initialize pygpu, support disabled
Traceback (most recent call last):
  File 
"C:\Users\Dave\Anaconda3\envs\TheanoPy3.5\lib\site-packages\theano\gpuarray\__init__.py",
 
line 227, in <module>
    use(config.device)
  File 
"C:\Users\Dave\Anaconda3\envs\TheanoPy3.5\lib\site-packages\theano\gpuarray\__init__.py",
 
line 214, in use
    init_dev(device, preallocate=preallocate)
  File 
"C:\Users\Dave\Anaconda3\envs\TheanoPy3.5\lib\site-packages\theano\gpuarray\__init__.py",
 
line 159, in init_dev
    pygpu.blas.gemm(0, tmp, tmp, 0, tmp, overwrite_c=True)
  File "pygpu\blas.pyx", line 149, in pygpu.blas.gemm
  File "pygpu\blas.pyx", line 47, in pygpu.blas.pygpu_blas_rgemm
pygpu.gpuarray.GpuArrayException: (b'Missing Blas library', 5)
[Elemwise{exp,no_inplace}(<TensorType(float32, vector)>)]
Looping 1000 times took 10.554782 seconds
Result is [ 1.23178029  1.61879337  1.52278066 ...,  2.20771813  2.29967761
  1.62323284]
Used the cpu

Process finished with exit code 0
[/code]

My Anaconda environment is detailed in the attached environment.yml file.

I have tried moving and renaming dlls as suggested here: 
https://github.com/Theano/Theano/issues/6681#issuecomment-482819279
But with no joy. Plus, I'd be a lot happier to find a less hacky fix than 
renaming dlls to pretend they are different versions!


This is my third day on this, so any help would be gratefully received.

 - Dave

-- 

--- 
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].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/theano-users/ac9cf5bc-3fcb-4d07-9143-d9ff3ff0b8cc%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Attachment: environment.yml
Description: Binary data

Reply via email to