Closed xmfbit closed 7 years ago
Hi,
that's interesting, I do not get this error. Do you compile with USE_CUDNN
?
Best, Nikolaus
Yes, I compiled with USE_CUDNN=1
and I found that your release did't support CUDNN6.0 which I am using.
I have fixed the problem by adding #include "caffe/proto/caffe.pb.h"
in file /include/caffe/util/rng.hpp
. And I replaced include/caffe/util/cudnn.hpp
with the same file in the latest official caffe repo.
Another problem, in the file src/cafffe/layers/resample_layer.cu
, opencv2/gpu/gpu.hpp
is redundant, and it is incompatible with OpenCV 3.X, so I removed it.
@xmfbit I did same with you, but new problem occur:
tools/convert_imageset_and_disparity.cpp: In member function ‘void ImagePair::read_data()’:
tools/convert_imageset_and_disparity.cpp:156:31: error: ‘numeric_limits’ is not a member of ‘std’
value=std::numeric_limits
Sorry, but I have no idea about your problem. What's the version of your compiler and platform? You can try to add #include <limits>
. @wuyuzaizai You can refer to my modified FlowNet2, which can build successfully in my environment: Ubuntu 16.04, CUDA8.0 with CUDNN6.0
@xmfbit I'm adding the missing rng.hpp
include. Could you explain how the opencv2/gpu/gpu.hpp
include is redundant? I did not find any other include for this, and it does not seem to be included by any other opencv2 header.
@wuyuzaizai I think this is not connected to this issue. If the problem persists, please open a separate issue for this.
Nikolaus
The implementations has nothing to do with the API declarations in the opencv gpu file. I guess the author added this for debugging, because the last several lines in the file resample layer have been commented. I removed it and built the flownet successfully.
Ah, now I get it :) I'll remove it then. Thanks!
Nikolaus
@wuyuzaizai hello, I meet the same problem as you. Have you solved the problem ? Could you tell me, please?
When I try to compile your code, error occurs like this.
So I add
#include "caffe/proto/caffe.pb.h"
in the file$CAFFE_ROOT$/include/caffe/util/rng.hpp
and the error is fixed.My environment is Ubuntu 14.04.1LTS with CUDA8.0. GCC version is 4.8.4