I am trying to run a convolutional neural network in lasagne using cuDNN.
Running my program I get the following error:
Traceback (most recent call last):
File "sarcoma.py", line 141, in <module>
main()
File "sarcoma.py", line 134, in main
train(args)
File "sarcoma.py", line 109, in train
cnn.train(data, args.epochs)
File "/usr/stud/pasa/Documents/Medina/medina/neural_network.py", line
338, in train
self.train_fn(*(inputs + targets))
File
"/usr/stud/pasa/.local/lib/python2.7/site-packages/theano/compile/function_module.py",
line 898, in __call__
storage_map=getattr(self.fn, 'storage_map', None))
File
"/usr/stud/pasa/.local/lib/python2.7/site-packages/theano/gof/link.py",
line 325, in raise_with_op
reraise(exc_type, exc_value, exc_trace)
File
"/usr/stud/pasa/.local/lib/python2.7/site-packages/theano/compile/function_module.py",
line 884, in __call__
self.fn() if output_subset is None else\
*RuntimeError: Cuda error 'the launch timed out and was terminated' while
copying %lli data element to device memory. str ptr=%p. dst ptr=%p*
Apply node that caused the error: GpuFromHost(input0)
Toposort index: 92
Inputs types: [TensorType(float32, 5D)]
Inputs shapes: [(1, 1, 25, 160, 160)]
Inputs strides: [(2560000, 2560000, 102400, 640, 4)]
Inputs values: ['not shown']
Outputs clients: [[GpuContiguous(GpuFromHost.0)]]
I am confused by the message (which looks like and incomplete printf...).
Searching online it seems due to the cuda kernels in theano running for too
long. I have indeed the X server running, but I cannot run the computer
without it, since the computer is managed by the university and I do not
have admin rights. Any help is much appreciated, as due to this error I
cannot proceed with my project.
--
---
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.
