make error with cuda-7.5 and cundnn 5.1

[shomedas]

while doing "make all" , I get error:

py-faster-rcnn/caffe-fast-rcnn/include/caffe/util/cudnn.hpp(126): error: argument of type "int" is incompatible with parameter of type "cudnnNanPropagation_t"

Is cuda-7.5 and cudnn v-5.1 combination not supported? i am on ubuntu 14.04 platform

[CharelCHEN]

hello,i am meeting this problem as well.could you please tell how to solver it ?thanks!

[Microos]

Try this:
1).
cd caffe-fast-rcnn
git remote add caffe https://github.com/BVLC/caffe.git
git fetch caffe
git merge -X theirs caffe/master

2).
remove self_.attr("phase") = static_cast<int>(this->phase_); from include/caffe/layers/python_layer.hpp after merging.

3). (in caffe-fast-rcnn)
- mkdir build && cd build
- cmake ..
- make all -j8
- make pycaffe -j8

[baristahell]

Hey there.
I'm trying to train faster-rcnn on a custom dataset with CUDA 7.5, CuDNN 5.1 and had to do this ( @Microos 's post) to be able to do build caffe.

My issue now is that i get an error when i train the net, the solver can't set an attribute:

I0324 15:41:08.424839 5348 net.cpp:380] input-data -> gt_boxes
Traceback (most recent call last):
File "./tools/train_net.py", line 112, in
max_iters=args.max_iters)
File "/opt/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 158, in train_net
pretrained_model=pretrained_model)
File "/opt/py-faster-rcnn/tools/../lib/fast_rcnn/train.py", line 44, in init
self.solver = caffe.SGDSolver(solver_prototxt)
AttributeError: can't set attribute

Did you encounter this error after you patched the cudnnNanPropagation_t bug? I'm worried it's caused by this BVLC Caffe merge.

Thanks.

[vignesh1905]

@Microos Solution worked for me with Cuda8.0, cuDnnv5, ubuntu 16.04.