Theano: CUDA 8.0 - GpuCorrMM encountered a CUBLAS error: the library was not initialized

[liorshk]

C:\Anaconda2\lib\site-packages\theano\compile\function_module.pyc in __call__(self, *args, **kwargs)
    857         t0_fn = time.time()
    858         try:
--> 859             outputs = self.fn()
    860         except Exception:
    861             if hasattr(self.fn, 'position_of_error'):

RuntimeError: GpuCorrMM encountered a CUBLAS error: the library was not initialized
This could be a known bug in CUDA, please see the GpuCorrMM() documentation.

Apply node that caused the error: GpuCorrMM{valid, (1, 1)}(GpuContiguous.0, GpuContiguous.0)
Toposort index: 73
Inputs types: [CudaNdarrayType(float32, 4D), CudaNdarrayType(float32, 4D)]
Inputs shapes: [(50, 1, 98, 98), (64, 1, 3, 3)]
Inputs strides: [(9604, 0, 98, 1), (9, 0, 3, 1)]
Inputs values: ['not shown', 'not shown']
Outputs clients: [[GpuElemwise{Add}[(0, 0)](GpuCorrMM{valid, (1, 1)}.0, GpuReshape{4}.0)]]

Same model works fine on CUDA 7.5.
Theano 0.8.2

[jetsnguns]

I had similar error "GpuCorrMM encountered a CUBLAS error: the library was not initialized". After installing cuDNN v5.1 for CUDA 8.0 everything worked.

[litingsjj]

I have a same problem, and i installed cnDNN v5.1...

I reinstall cuDNN v5.1.it's solved!!!!!!!!but have::serWarning: Your cuDNN version is more recent than the one Theano officially supports. If you see any problems, try updating Theano or downgrading cuDNN to version 5.
warnings.warn(warn)

[nouiz]

Make sure to use Theano dev version.

On Sun, Nov 20, 2016 at 6:38 AM, litingsjj [email protected] wrote:

I have a same problem, and i installed cnDNN v5.1...

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
https://github.com/Theano/Theano/issues/4671#issuecomment-261772993, or mute
the thread
https://github.com/notifications/unsubscribe-auth/AALC-_L4x18iiSeJVQ3KXQebhpU0XRyvks5rADEugaJpZM4I-pqm
.