search

Theano / libgpuarray

Library to manipulate tensors on the GPU.
Other
189 stars 96 forks source link

Device Kernel Image Invalid #426

Open thorlucas opened 7 years ago

thorlucas commented 7 years ago

While running the Theano gpu testing code, I get the following error using flag device=cuda.

ERROR (theano.gpuarray): Could not initialize pygpu, support disabled
[...]
pygpu.gpuarray.GpuArrayException: (b'device kernel image is invalid', 3)

I'm also getting

/Users/thor/anaconda3/lib/python3.6/site-packages/theano/gpuarray/dnn.py:135: UserWarning: Your cuDNN version is more recent than Theano. If you encounter problems, try updating Theano or downgrading cuDNN to version 5.1.

Despite having literally just deleted cuda and cudnn and reinstalling version 5.1.

nouiz commented 7 years ago

For cudnn, ir could mean you use an old version of Theano or that you have multiple cudnn installed and Theano find a more recent one.

Which test give you the exception?