search

BVLC / caffe

Caffe: a fast open framework for deep learning.
http://caffe.berkeleyvision.org/
Other
34.15k stars 18.67k forks source link

Jetson TK1 cudnn issue #3370

Open geogeorgiev opened 9 years ago

geogeorgiev commented 9 years ago

I am trying to compile Caffe on Nvidia Jetson TK1 with GPU, CuDNN and Python layer. Cuda and CuDNN are properly installed using the Jetpack 2.0 provided by Nvidia. I get the following debug info:

In file included from src/caffe/layers/base_conv_layer.cpp:7:0: ./include/caffe/vision_layers.hpp:310:3: error: ‘cudnnConvolutionBwdFilterAlgo_t’ does not name a type cudnnConvolutionBwdFilterAlgo_t _bwd_filteralgo; ^ ./include/caffe/vision_layers.hpp:311:3: error: ‘cudnnConvolutionBwdDataAlgo_t’ does not name a type cudnnConvolutionBwdDataAlgo_t bwd_dataalgo; ^ ./include/caffe/vision_layers.hpp:477:3: error: ‘cudnnLRNDescriptor_t’ does not name a type cudnnLRNDescriptor_t normdesc; ^ ./include/caffe/vision_layers.hpp:504:3: error: ‘cudnnLRNDescriptor_t’ does not name a type cudnnLRNDescriptor_t normdesc; ^ make: _* [.build_debug/src/caffe/layers/base_conv_layer.o] Error 1

antoniodourado commented 8 years ago

I've got the same issue here. Any solution so far?

geogeorgiev commented 8 years ago

I've switched to Torch in the meantime :)