I recently changed config in order to train CNN's faster. I got a gtx 1080. For most of my existing neural nets, I'm using the theano back-end through python3.
Since I've read that pascal architecture wasn't fully used through CUDA 7.5, I've installed CUDA 8 and the corresponding cudnn 6.X.
My problem is the following, pygpuarray seems not to be able to find "libnvrtc.so" :
Traceback (most recent call last):
File "/usr/local/lib/python3.5/dist-packages/theano/gpuarray/__init__.py", line 179, in <module>
use(config.device)
File "/usr/local/lib/python3.5/dist-packages/theano/gpuarray/__init__.py", line 166, in use
init_dev(device, preallocate=preallocate)
File "/usr/local/lib/python3.5/dist-packages/theano/gpuarray/__init__.py", line 65, in init_dev
sched=config.gpuarray.sched)
File "pygpu/gpuarray.pyx", line 634, in pygpu.gpuarray.init (pygpu/gpuarray.c:9428)
File "pygpu/gpuarray.pyx", line 584, in pygpu.gpuarray.pygpu_init (pygpu/gpuarray.c:9119)
File "pygpu/gpuarray.pyx", line 1057, in pygpu.gpuarray.GpuContext.__cinit__ (pygpu/gpuarray.c:13421)
pygpu.gpuarray.GpuArrayException: b'Could not load "libnvrtc.so": libnvrtc.so: cannot open shared object file: No such file or directory'
I'm pretty sure cuda is installed, here is the nvcc --version output:
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2016 NVIDIA Corporation
Built on Tue_Jan_10_13:22:03_CST_2017
Cuda compilation tools, release 8.0, V8.0.61
And I've found the file in: /usr/local/cuda-8.0/lib64
Strange thing is that "ldconfig -p | grep libnvcc" command return nothing...
I hope this post doesn't exist anywhere else, I've made some searches to prevent that but you know, the Internet is vast.
Ask me if you want anymore informations. Since i didn't managed to fix this problem, I obviously don't know which info is necessary to solve this problem so i made this small resume waiting your questions (if needed).