error training fcn-32s on gpu however run well on cpu

[zhjia...@gmail.com]

Hello:

I'm trying to train fully convolutional networks for semantic segmentation on gpu.

training fcn-32s using VGG16.caffemodel

However I obtain an error:Check failed: status == CUBLAS_STATUS_SUCCESS   CUBLAS_STATUS_MAPPING_ERROR

the code I use for init is below:

# init

caffe.set_mode_gpu()

caffe.set_device(0)

When I comment the code above,it can run on cpu.

How can I run it on gpu.

Thanks.

[Yb W]

The same problem.Did you fix it?

在 2015年11月17日星期二 UTC+8上午10:32:09，zhjia...@gmail.com写道：

[maina...@gmail.com]

Hi Yb
Did you find the solution?