Problems building caffe with cudnn enabled

[Moruk]

Hello all.

I have Cuda 10 and cuDnn v6 and when I try to build caffe with make all I get the following issue:

Can anyone please help me on how to solve this problem?

I have enabled cudnn in my makefile.config

PROTOC src/caffe/proto/caffe.proto
CXX .build_release/src/caffe/proto/caffe.pb.cc
CXX src/caffe/syncedmem.cpp
In file included from ./include/caffe/util/device_alternate.hpp:40:0,
                 from ./include/caffe/common.hpp:19,
                 from src/caffe/syncedmem.cpp:3:
./include/caffe/util/cudnn.hpp: In function ‘const char* cudnnGetErrorString(cudnnStatus_t)’:
./include/caffe/util/cudnn.hpp:18:10: warning: enumeration value ‘CUDNN_STATUS_RUNTIME_PREREQUISITE_MISSING’ not handled in switch [-Wswitch]
   switch (status) {
          ^
./include/caffe/util/cudnn.hpp: In function ‘void caffe::cudnn::setConvolutionDesc(cudnnConvolutionStruct**, cudnnTensorDescriptor_t, cudnnFilterDescriptor_t, int, int, int, int)’:
./include/caffe/util/cudnn.hpp:105:70: error: too few arguments to function ‘cudnnStatus_t cudnnSetConvolution2dDescriptor(cudnnConvolutionDescriptor_t, int, int, int, int, int, int, cudnnConvolutionMode_t, cudnnDataType_t)’
       pad_h, pad_w, stride_h, stride_w, 1, 1, CUDNN_CROSS_CORRELATION));
                                                                      ^
./include/caffe/util/cudnn.hpp:12:28: note: in definition of macro ‘CUDNN_CHECK’
     cudnnStatus_t status = condition; \
                            ^~~~~~~~~
In file included from ./include/caffe/util/cudnn.hpp:5:0,
                 from ./include/caffe/util/device_alternate.hpp:40,
                 from ./include/caffe/common.hpp:19,
                 from src/caffe/syncedmem.cpp:3:
/usr/include/cudnn.h:500:27: note: declared here
 cudnnStatus_t CUDNNWINAPI cudnnSetConvolution2dDescriptor( cudnnConvolutionDescriptor_t convDesc,
                           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from ./include/caffe/util/device_alternate.hpp:40:0,
                 from ./include/caffe/common.hpp:19,
                 from src/caffe/syncedmem.cpp:3:
./include/caffe/util/cudnn.hpp: In function ‘void caffe::cudnn::createPoolingDesc(cudnnPoolingStruct**, caffe::PoolingParameter_PoolMethod, cudnnPoolingMode_t*, int, int, int, int, int, int)’:
./include/caffe/util/cudnn.hpp:124:41: error: too few arguments to function ‘cudnnStatus_t cudnnSetPooling2dDescriptor(cudnnPoolingDescriptor_t, cudnnPoolingMode_t, cudnnNanPropagation_t, int, int, int, int, int, int)’
         pad_h, pad_w, stride_h, stride_w));
                                         ^
./include/caffe/util/cudnn.hpp:12:28: note: in definition of macro ‘CUDNN_CHECK’
     cudnnStatus_t status = condition; \
                            ^~~~~~~~~
In file included from ./include/caffe/util/cudnn.hpp:5:0,
                 from ./include/caffe/util/device_alternate.hpp:40,
                 from ./include/caffe/common.hpp:19,
                 from src/caffe/syncedmem.cpp:3:
/usr/include/cudnn.h:952:27: note: declared here
 cudnnStatus_t CUDNNWINAPI cudnnSetPooling2dDescriptor(
                           ^~~~~~~~~~~~~~~~~~~~~~~~~~~
Makefile:524: recipe for target '.build_release/src/caffe/syncedmem.o' failed
make: *** [.build_release/src/caffe/syncedmem.o] Error 1

[尚高星]

hello, Moruk:

have you already solved this issue? I have met the same problem when I build caffe with make all command. Could you teach me how to solve it?

在 2018年11月12日星期一 UTC+8上午7:30:17，Moruk写道：

[Tamas Nemes]

This looks much like a compatibility issue.

If you look here:

https://developer.nvidia.com/rdp/cudnn-download

You can see that cuDNN 6 is deprecated a long while ago, CUDA 10 is only compatible with cuDNN 7.6.5 which you can download there.

Here is the installation guide: https://docs.nvidia.com/deeplearning/sdk/cudnn-install/index.html

So try to upgrade your cuDNN version to match your CUDA version. After the installation, remember to start with a clean build by first doing make clean and then make all -j$(nproc). Also, remember to delete the old symlinks from cuda/lib64 because else Caffe gets confused which version to use. I was able to successfully compile Caffe with CUDA 10.2 and cuDNN 7.6.5.

If your problem still persists, please contact again!