cuDNN v5 support?

[ekcheng]

Attempting to build caffe-fast-rcnn with cuDNN v5 leads to these types of errors:

In file included from ./include/caffe/util/device_alternate.hpp:40:0,
                 from ./include/caffe/common.hpp:19,
                 from src/caffe/util/upgrade_proto.cpp:8:
./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:127:41: 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:15:28: note: in definition of macro ‘CUDNN_CHECK’
     cudnnStatus_t status = condition; \

Any plans to update caffe-fast-rcnn to support cuDNN v5 and the newer Pascal GPUs?

Seems like the main caffe just recently added support for cuDNN v5 per BVLC/caffe#3969

Thanks!

[tianxingyzxq]

I have the same problem，have you solved it yet？@ekcheng

[yxliang]

related to #237

[nils489]

Hi, I forked this repo and added cudnn 5 and ubuntu 16.04 support for the faster-rcnn branch:
caffe-fast-rcnn
I did not have the time yet to work through the other branches

[shijx12]

@nils489 Thanks it does work!

remember to checkout faster-rcnn branch!

[machanic]

@nils489 @shijx12 I download and compile your caffe-fast-rcnn , it seems to compile success. But when I run demo.py, I found that your src/caffe/proto/caffe.proto did not even have roi_pooling_param?
Did this repo correct???

[askerlee]

EDIT: Oh sorry for my mistake. Now I checked the "faster-rcnn" branch and it works! Thanks @nils489

Probably @nils489 forked the wrong branch...

[nils489]

@sharpstill @askerlee As @shijx12 already said, check out the faster-rcnn branch. If you try to compile the master branch, it will fail. As I said in my original post:

I did not have the time yet to work through the other branches

The caffe.proto in the faster-rcnn branch does contain roi_pooling_param

@sharpstill Probably @nils489 forked the wrong branch...

@askerlee Probably @sharpstill checked out the wrong branch...

[hana9090]

@nils489
I use this version for cudnn5 however it is not compatible for faster rcnn :(

when I try to run demo:

./tools/demo.py
[libprotobuf ERROR google/protobuf/text_format.cc:298] Error parsing text-format caffe.NetParameter: 350:21: Message type "caffe.LayerParameter" has no field named "roi_pooling_param".
WARNING: Logging before InitGoogleLogging() is written to STDERR
F0410 10:51:30.183878 46559 upgrade_proto.cpp:928] Check failed: ReadProtoFromTextFile(param_file, param) Failed to parse NetParameter file: /home/usename/code/fast2/py-faster-rcnn/models/pascal_voc/VGG16/faster_rcnn_alt_opt/faster_rcnn_test.pt
*** Check failure stack trace: ***
Aborted (core dumped)

[shilpastp]

@nils489
I am able to compile the caffe for py-faster-rcnn successfully. I am getting make runtest error:

E0416 11:20:37.078045 62161 io.cpp:90] Could not open or find file examples/images/fish-bike.jpg
F0416 11:20:37.078090 62161 image_data_layer.cpp:124] Check failed: cv_img.data Could not load examples/images/fish-bike.jpg
*** Check failure stack trace: ***
@ 0x7f5f396ef5cd google::LogMessage::Fail()
@ 0x7f5f396f1433 google::LogMessage::SendToLog()
@ 0x7f5f396ef15b google::LogMessage::Flush()
@ 0x7f5f396f1e1e google::LogMessageFatal::~LogMessageFatal()
@ 0x7f5f36e74a5e caffe::ImageDataLayer<>::load_batch()
@ 0x7f5f36f058bf caffe::BasePrefetchingDataLayer<>::InternalThreadEntry()
@ 0x7f5f37048115 caffe::InternalThread::entry()
@ 0x7f5f37c0f5d5 (unknown)
@ 0x7f5f3649f6ba start_thread
@ 0x7f5f361d541d clone
@ (nil) (unknown)
Makefile:526: recipe for target 'runtest' failed
make: *** [runtest] Aborted (core dumped)

Following is the working environment:
Ubuntu 16.04 LTS
CUDA Version 8.0.61
Driver Version: 384.111
GPU: GeForce GTX 1080
cuDNN v5

Am I missing some step or doing something wrong??

[nils489]

make runtest didn't work for me either. But the build was sufficient to train fast-rcnn and run inference on videos. Am 16. April 2018 08:51:42 MESZ schrieb shilpastp <[email protected]>:

…

@nils489 I am able to compile the caffe for py-faster-rcnn successfully. I am getting make runtest error: E0416 11:20:37.078045 62161 io.cpp:90] Could not open or find file examples/images/fish-bike.jpg F0416 11:20:37.078090 62161 image_data_layer.cpp:124] Check failed: cv_img.data Could not load examples/images/fish-bike.jpg *** Check failure stack trace: *** @ 0x7f5f396ef5cd google::LogMessage::Fail() @ 0x7f5f396f1433 google::LogMessage::SendToLog() @ 0x7f5f396ef15b google::LogMessage::Flush() @ 0x7f5f396f1e1e google::LogMessageFatal::~LogMessageFatal() @ 0x7f5f36e74a5e caffe::ImageDataLayer<>::load_batch() @ 0x7f5f36f058bf caffe::BasePrefetchingDataLayer<>::InternalThreadEntry() @ 0x7f5f37048115 caffe::InternalThread::entry() @ 0x7f5f37c0f5d5 (unknown) @ 0x7f5f3649f6ba start_thread @ 0x7f5f361d541d clone @ (nil) (unknown) Makefile:526: recipe for target 'runtest' failed make: *** [runtest] Aborted (core dumped) Following is the working environment: Ubuntu 16.04 LTS CUDA Version 8.0.61 Driver Version: 384.111 GPU: GeForce GTX 1080 cuDNN v5 Am I missing some step or doing something wrong?? -- You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub: #14 (comment)

-- Diese Nachricht wurde von meinem Android-Gerät mit K-9 Mail gesendet.

[yiakwy]

The errors are still there and the cuda has already been updated to the latest version. I tried to rebase the branch onto the master branch. Surprised, we still have the errors. Illustrated every year when I need to install caffe onto a new machine, always new errors occur!

[yiakwy]

Rebase the branch onto the lastest caffe and resolved conflicts solved the prolbem.

[yiakwy]

@ekcheng This can be closed now.