support 1080P and cudnn v5? #237

zimenglan-sysu-512 · 2016-07-01T03:42:51Z

so far, faster-rcnn surpports 1080P and cudnn v5?

thanks.

zimenglan-sysu-512 · 2016-07-01T03:58:05Z

@rbgirshick can you do me a favor? thanks.

EvanWeiner · 2016-07-02T05:10:15Z

+1 echo Zimenglan.

Thank you so much for your hard work on py-faster-cnn. I would like to humbly request if you can please modify faster-rcnn to work with cudnn v5. The new GPUs like 1080 can't use anything less than v5. This would greatly be appreciated.

Thank you,
Evan

manipopopo · 2016-07-05T14:01:04Z

Maybe you can try to merge caffe master branch into caffe-fast-rcnn.

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

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

py-R-FCN use Microsoft-version Caffe which is compatible with cudnn 5.

zimenglan-sysu-512 · 2016-07-06T01:48:44Z

@manipopopo thanks. it's a real help, it works! now i can use 1080P and v5 of cudnn.
@EvanWeiner you can try. self_.attr("phase") = static_cast<int>(this->phase_) from include/caffe/layers/python_layer.hpp is the key.

GBJim · 2016-07-13T03:17:39Z

Hi @manipopopo
I followed your instruction, but encountered an error when I run /tools/demo.py
The error message is as follwoing:

WARNING: Logging before InitGoogleLogging() is written to STDERR
W0713 03:15:46.031435 22906 _caffe.cpp:122] DEPRECATION WARNING - deprecated use of Python interface
W0713 03:15:46.031481 22906 _caffe.cpp:123] Use this instead (with the named "weights" parameter):
W0713 03:15:46.031488 22906 _caffe.cpp:125] Net('/root/py-faster-rcnn/models/pascal_voc/VGG16/faster_rcnn_alt_opt/faster_rcnn_test.pt', 1, weights='/root/py-faster-rcnn/data/faster_rcnn_models/VGG16_faster_rcnn_final.caffemodel')
[libprotobuf ERROR google/protobuf/text_format.cc:245] Error parsing text-format caffe.NetParameter: 350:21: Message type "caffe.LayerParameter" has no field named "roi_pooling_param".
F0713 03:15:46.033658 22906 upgrade_proto.cpp:79] Check failed: ReadProtoFromTextFile(param_file, param) Failed to parse NetParameter file: /root/py-faster-rcnn/models/pascal_voc/VGG16/faster_rcnn_alt_opt/faster_rcnn_test.pt
*** Check failure stack trace: ***
Aborted (core dumped)

How do I solve this problem?

manipopopo · 2016-07-13T06:35:28Z

Make sure that optional ROIPoolingParameter roi_pooling_param = 8266711; is in caffe-fast-rcnn/src/caffe/proto/caffe.proto. Remember to rebuild caffe-fast-rcnn after merging.

GBJim · 2016-07-13T09:33:01Z

Hi @manipopopo
I have successfully compiled the merged Caffe.
But when I test the training speed om VGG16, I found the training speed is around 1.214s / iteration, which is not faster comparing cuDNN v3 and no-cuDNN implementation.
Is it normal?

manipopopo · 2016-07-13T12:14:34Z

I'm sorry I can't share you with the precise time one iteration takes. I don't have 1080 at hand. But I vaguely remembered it took about 13 seconds for 20 iterations when I trained ResNet-101 (GTX 1080, cuda 8, cudnn 5) using end-to-end strategy a month ago.

zimenglan-sysu-512 · 2016-07-14T05:41:53Z

@manipopopo i use 1080P to test the demo images, the prcise time as below:

Demo for data/demo/000456.jpg
Detection took 0.107s for 300 object proposals

Demo for data/demo/000542.jpg
Detection took 0.077s for 161 object proposals

Demo for data/demo/001150.jpg
Detection took 0.089s for 194 object proposals

Demo for data/demo/001763.jpg
Detection took 0.089s for 196 object proposals

Demo for data/demo/004545.jpg
Detection took 0.104s for 300 object proposals

GBJim · 2016-07-14T05:48:24Z

@zimenglan-sysu-512 That's really fast!
My detection speed is 0.260s in K80 machine (VGG16)

manipopopo · 2016-07-14T05:51:09Z

@zimenglan-sysu-512
I vaguely remembered it took about 190 ms for 300 proposals per image for ResNet-101 (GTX 1080, cuda 8, cudnn 5) during the test phase.

rbgirshick/py-faster-rcnn#237 (comment)

hengck23 · 2016-07-21T15:22:23Z

for those who are using windows caffe, here is a fork that support cudnnv5
https://github.com/TadaoYamaoka/caffe/tree/windows

For more information, refer to:
http://tadaoyamaoka.hatenablog.com/entry/2016/06/19/225038

duanLH · 2016-08-14T00:56:13Z

@manipopopo when Ｉ　run git merge caffe/master Ｉｔ　ｓｈｏｗｔhat：

root@duan:/home/duan/py-faster-rcnn/caffe-fast-rcnn# git merge caffe/mastererror: Your local changes to the following files would be overwritten by merge:
data/cifar10/get_cifar10.sh
data/ilsvrc12/get_ilsvrc_aux.sh
data/mnist/get_mnist.sh
examples/cifar10/create_cifar10.sh
examples/cifar10/train_full.sh
examples/cifar10/train_full_sigmoid.sh
examples/cifar10/train_full_sigmoid_bn.sh
examples/cifar10/train_quick.sh
examples/imagenet/create_imagenet.sh
examples/imagenet/resume_training.sh
examples/imagenet/train_caffenet.sh
examples/mnist/create_mnist.sh
examples/mnist/train_lenet.sh
examples/mnist/train_lenet_adam.sh
examples/mnist/train_lenet_consolidated.sh
examples/mnist/train_lenet_rmsprop.sh
examples/mnist/train_mnist_autoencoder.sh
examples/mnist/train_mnist_autoencoder_adadelta.sh
examples/mnist/train_mnist_autoencoder_adagrad.sh
examples/mnist/train_mnist_autoencoder_nesterov.sh
examples/siamese/create_mnist_siamese.sh
examples/siamese/train_mnist_siamese.sh
python/draw_net.py
scripts/cpp_lint.py
scripts/download_model_binary.py
scripts/travis/travis_build_and_test.sh
scripts/travis/travis_install.sh
scripts/travis/travis_setup_makefile_config.sh
tools/extra/parse_log.py
tools/extra/plot_training_log.py.example
Please, commit your changes or stash them before you can merge.
Aborting

what should i do？

manipopopo · 2016-08-14T03:22:12Z

As it said, you should commit your changes or stash them before you can merge. You can try

Use git stash to save your local changes before merging. Afterwards, you can use git stash pop to apply the saved changes on the working directory.
Use git commit -a to commit all the changes before merging.

longwoo · 2016-08-29T05:54:38Z

@manipopopo I am using ubuntu16.04+GTX1080+cuda8.0RC+cudnn v5 and tried your method. After make -j8, it gives me

PROTOC src/caffe/proto/caffe.proto
CXX src/caffe/common.cpp
CXX src/caffe/blob.cpp
CXX src/caffe/layer_factory.cpp
......
CXX src/caffe/syncedmem.cpp
CXX src/caffe/internal_thread.cpp
CXX src/caffe/net.cpp
NVCC src/caffe/solvers/adadelta_solver.cu
NVCC src/caffe/solvers/adam_solver.cu
NVCC src/caffe/solvers/rmsprop_solver.cu
......
NVCC src/caffe/layers/prelu_layer.cu
NVCC src/caffe/layers/threshold_layer.cu
NVCC src/caffe/layers/base_data_layer.cu
CXX/LD -o .build_release/tools/net_speed_benchmark.bin
CXX/LD -o .build_release/tools/test_net.bin
CXX/LD -o .build_release/tools/convert_imageset.bin
CXX/LD -o .build_release/tools/extract_features.bin
CXX/LD -o .build_release/tools/device_query.bin
CXX/LD -o .build_release/tools/train_net.bin
CXX/LD -o .build_release/tools/compute_image_mean.bin
CXX/LD -o .build_release/tools/caffe.bin
CXX/LD -o .build_release/tools/upgrade_solver_proto_text.bin
CXX/LD -o .build_release/tools/upgrade_net_proto_text.bin
CXX/LD -o .build_release/tools/upgrade_net_proto_binary.bin
CXX/LD -o .build_release/tools/finetune_net.bin
.build_release/lib/libcaffe.so: undefined reference to `cv::imencode(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imread(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, int)'
collect2: error: ld returned 1 exit status
.build_release/lib/libcaffe.so: undefined reference to Makefile:619: recipe for target '.build_release/tools/caffe.bin' failed
make: *** [.build_release/tools/caffe.bin] Error 1
make: *** Waiting for unfinished jobs....
`cv::imencode(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imread(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, int)'
collect2: error: ld returned 1 exit status
Makefile:619: recipe for target '.build_release/tools/extract_features.bin' failed
make: *** [.build_release/tools/extract_features.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `cv::imencode(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imread(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, int)'
collect2: error: ld returned 1 exit status
Makefile:619: recipe for target '.build_release/tools/convert_imageset.bin' failed
make: *** [.build_release/tools/convert_imageset.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `cv::imencode(std:.:__cxx11::basic_string<char, std::char_traits<char>, std::build_release/lib/libcaffe.so: undefined referenceallocator <char> > toconst `cv::imencode(std::__cxx11::basic_string<char, std&::char_traits<char>,, std::allocator<char> >  const&, cv::_InputArray constcv::_InputArray const&, std::vector<unsigned char, std::allocator&<unsigned char> >, std::vector<unsigned char, std::allocator&<unsigned char> ,> std::vector<int, std::allocator<int> &> const&)'
.build_release/lib/,libcaffe.so : undefined reference stdto `cv::imread(std:::__cxx11::basic_string<:char, std::vectorchar_traits<int, std::allocator<int> > const&)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imread(std::__cxx11::basic_string<char>, <std::charallocator, std::char_traits<char>, std::allocator<char> > const&, int)'
<char> > const&, int)'
collect2: error: ld returned 1 exit status
collect2: error: ld returned 1 exit status
Makefile:619: recipe for target '.build_release/tools/compute_image_mean.bin' failed
make: *** [.build_release/tools/compute_image_mean.bin] Error 1
Makefile:619: recipe for target '.build_release/tools/upgrade_solver_proto_text.bin' failed
make: *** [.build_release/tools/upgrade_solver_proto_text.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `cv::imencode(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imread(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, int)'
collect2: error: ld returned 1 exit status
Makefile:619: recipe for target '.build_release/tools/upgrade_net_proto_text.bin' failed
make: *** [.build_release/tools/upgrade_net_proto_text.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `cv::imencode(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, cv::_InputArray const&, std::vector<unsigned char, std::allocator<unsigned char> >&, std::vector<int, std::allocator<int> > const&)'
.build_release/lib/libcaffe.so: undefined reference to `cv::imread(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, int)'
collect2: error: ld returned 1 exit status
Makefile:619: recipe for target '.build_release/tools/upgrade_net_proto_binary.bin' failed
make: *** [.build_release/tools/upgrade_net_proto_binary.bin] Error 1

manipopopo · 2016-08-29T06:03:12Z

@longwoo
Have you uncommented `OPENCV_VERSION := 3ˋ from Makefile.config?

longwoo · 2016-08-29T06:31:39Z

@manipopopo Thanks. So we have to use opencv3 in this case? I am trying to install opencv3 and rebuild again. By the way, here is my Makefile.config.

USE_CUDNN := 1
CUDA_DIR := /usr/local/cuda
CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \
            -gencode arch=compute_20,code=sm_21 \
            -gencode arch=compute_30,code=sm_30 \
            -gencode arch=compute_35,code=sm_35 \
            -gencode arch=compute_50,code=sm_50 \
            -gencode arch=compute_50,code=compute_50

BLAS := atlas

ANACONDA_HOME := $(HOME)/anaconda2
PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
             $(ANACONDA_HOME)/include/python2.7 \
             $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include \

PYTHON_LIB := $(ANACONDA_HOME)/lib

WITH_PYTHON_LAYER := 1

INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib

BUILD_DIR := build
DISTRIBUTE_DIR := distribute

TEST_GPUID := 0

Q ?= @

manipopopo · 2016-08-29T07:15:58Z

I'm not sure but opencv 2 should be OK.

I didn't know what version your opencv is, so I guessed that you're using 3. The default Makefile is for opencv 2. If you want to build caffe with OpenCV 3, you'll need to uncomment ˋOPENCV_VERSION := 3ˋ. Otherwise you'll see the error messages you post.

austingg · 2016-08-29T07:31:54Z

@longwoo by the way, your CUDA_Arch should be set to compute_61 for gtx 1080 card

longwoo · 2016-08-29T08:25:02Z

@manipopopo Yes,opencv 2 is OK .Thanks.
@austingg compute_61 solves the problem. Thank you.

USE_CUDNN := 1

CUDA_DIR := /usr/local/cuda
CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \
    -gencode arch=compute_20,code=sm_21 \
    -gencode arch=compute_30,code=sm_30 \
    -gencode arch=compute_35,code=sm_35 \
    -gencode arch=compute_50,code=sm_50 \
    -gencode arch=compute_61,code=sm_61 \
    -gencode arch=compute_61,code=compute_61

BLAS := atlas

PYTHON_INCLUDE := /usr/include/python2.7 \
        /usr/lib/python2.7/dist-packages/numpy/core/include
PYTHON_LIB := /usr/lib
WITH_PYTHON_LAYER := 1

INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/include /usr/include/hdf5/serial/
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/lib /usr/lib /usr/lib/x86_64-linux-gnu/hdf5/serial/

BUILD_DIR := build
DISTRIBUTE_DIR := distribute

TEST_GPUID := 0

Q ?= @

austingg · 2016-09-19T02:06:03Z

@longwoo By the way, if you only develop on 1080 gpu, you can comment other compute capacity, which will make a faster compile speed.

vishaltiwari · 2017-03-21T11:17:29Z

Hi,

I am trying to build py-faster-rcnn on ubuntu 16.04, using tesla k20c, cudnn: v5. (cuda 8).

This is what I have done:
In Makefile.config:
uncommented the USE_CUDNN:=1 (as without this I am to build it, but only able to run the ./tools/demo.py --cpu , but the default (--gpu) gives the following error).


I0321 16:44:53.692109 26971 net.cpp:744] Ignoring source layer data
I0321 16:44:53.758460 26971 net.cpp:744] Ignoring source layer drop6
I0321 16:44:53.767998 26971 net.cpp:744] Ignoring source layer drop7
I0321 16:44:53.768015 26971 net.cpp:744] Ignoring source layer fc7_drop7_0_split
I0321 16:44:53.768318 26971 net.cpp:744] Ignoring source layer loss_cls
I0321 16:44:53.768327 26971 net.cpp:744] Ignoring source layer loss_bbox
I0321 16:44:53.769767 26971 net.cpp:744] Ignoring source layer silence_rpn_cls_score
I0321 16:44:53.769780 26971 net.cpp:744] Ignoring source layer silence_rpn_bbox_pred


Loaded network $FRCN_ROOT/data/faster_rcnn_models/VGG16_faster_rcnn_final.caffemodel
F0321 16:44:54.035586 26971 syncedmem.cpp:71] Check failed: error == cudaSuccess (2 vs. 0)  out of memory
*** Check failure stack trace: ***
Aborted (core dumped)

But I need performance as I am working on a video dataset, so I have to build it using "USE_CUDNN:=1"

Also, I have merged caffe-fater-rcnn with caffe/master build and removed the line: self_.attr("phase") = static_cast<int>(this->phase_) as mentioned in the previous posts. And tried to rebuild.

At the very end I get this following error:

CXX/LD -o .build_release/tools/test_net.bin
CXX/LD -o .build_release/tools/upgrade_solver_proto_text.bin
CXX/LD -o .build_release/tools/train_net.bin
CXX/LD -o .build_release/tools/upgrade_net_proto_text.bin
CXX/LD -o .build_release/tools/device_query.bin
CXX/LD -o .build_release/tools/net_speed_benchmark.bin
CXX/LD -o .build_release/tools/finetune_net.bin
CXX/LD -o .build_release/tools/convert_imageset.bin
CXX/LD -o .build_release/tools/compute_image_mean.bin
CXX/LD -o .build_release/tools/extract_features.bin
CXX/LD -o .build_release/tools/upgrade_net_proto_binary.bin
.build_release/lib/libcaffe.so: undefined reference to `cudnnSetActivationDescriptor'
.build_release/lib/libcaffe.so: undefined reference to `cudnnCreateActivationDescriptor'
collect2: error: ld returned 1 exit status
Makefile:625: recipe for target '.build_release/tools/upgrade_net_proto_text.bin' failed
make: *** [.build_release/tools/upgrade_net_proto_text.bin] Error 1
make: *** Waiting for unfinished jobs....
.build_release/lib/libcaffe.so: undefined reference to `cudnnSetActivationDescriptor'
.build_release/lib/libcaffe.so: undefined reference to `cudnnCreateActivationDescriptor'
collect2: error: ld returned 1 exit status
Makefile:625: recipe for target '.build_release/tools/upgrade_solver_proto_text.bin' failed
make: *** [.build_release/tools/upgrade_solver_proto_text.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `cudnnSetActivationDescriptor'
.build_release/lib/libcaffe.so: undefined reference to `cudnnCreateActivationDescriptor'
collect2: error: ld returned 1 exit status
Makefile:625: recipe for target '.build_release/tools/extract_features.bin' failed
make: *** [.build_release/tools/extract_features.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `cudnnSetActivationDescriptor'
.build_release/lib/libcaffe.so: undefined reference to `cudnnCreateActivationDescriptor'
collect2: error: ld returned 1 exit status
Makefile:625: recipe for target '.build_release/tools/compute_image_mean.bin' failed
make: *** [.build_release/tools/compute_image_mean.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `cudnnSetActivationDescriptor'
.build_release/lib/libcaffe.so: undefined reference to `cudnnCreateActivationDescriptor'
collect2: error: ld returned 1 exit status
Makefile:625: recipe for target '.build_release/tools/convert_imageset.bin' failed
make: *** [.build_release/tools/convert_imageset.bin] Error 1
.build_release/lib/libcaffe.so: undefined reference to `cudnnSetActivationDescriptor'
.build_release/lib/libcaffe.so: undefined reference to `cudnnCreateActivationDescriptor'
collect2: error: ld returned 1 exit status
Makefile:625: recipe for target '.build_release/tools/upgrade_net_proto_binary.bin' failed
make: *** [.build_release/tools/upgrade_net_proto_binary.bin] Error 1

I am not sure what library is cudnnSetActivationDescriptor' and cudnnCreateActivationDescriptor' that it's not able to find. I tried to look for this error but didn't find anything useful.

Does anyone know how this can be fixed?

Thanks,

hana9090 · 2017-04-01T06:47:17Z

@manipopopo
PLease after I did your step regarding caffe,
when I ((sudo make -j4 && make pycaffe))
I faced this error:

touch python/caffe/proto/init.py
touch: cannot touch 'python/caffe/proto/init.py': Permission denied
Makefile:647: recipe for target 'python/caffe/proto/init.py' failed
make: *** [python/caffe/proto/init.py] Error 1

manipopopo · 2017-04-01T09:47:02Z

@hana9090
I have no idea why it happens. 😿
Could you use

make -j4
make pycaffe

or

sudo make -j4
sudo make pycaffe

instead?

hana9090 · 2017-04-01T09:54:14Z

still have same error :\

manipopopo · 2017-04-01T10:18:59Z

@hana9090
I got the same error when I ran

sudo make
make pycaffe

After that I ran sudo make pycaffe, it works without error.

You may want to try something like

sudo make clean
make
make pycaffe

hana9090 · 2017-04-01T12:25:53Z

@manipopopo
I did what you write, but still have the same error of permission denied.
Although I change the permission of /proto/ directory from root to user!

hana9090 · 2017-04-01T20:59:59Z

@manipopopo
Thank you :)
it worked with me if I arrange them as following:

sudo make clean
make pycaffe
make -j8

- See rbgirshick/py-faster-rcnn#237

soulslicer · 2017-08-12T08:57:36Z

Made it work with pascal:

Clone this and compile then

https://github.com/soulslicer/caffe-fast-rcnn

Just change proposal_layer.py

     def setup(self, bottom, top):
         # parse the layer parameter string, which must be valid YAML
-        layer_params = yaml.load(self.param_str_)
+        layer_params = yaml.load(self.param_str)
 
         self._feat_stride = layer_params['feat_stride']
         anchor_scales = layer_params.get('scales', (8, 16, 32))
@@ -61,7 +61,7 @@ class ProposalLayer(caffe.Layer):
         assert bottom[0].data.shape[0] == 1, \
             'Only single item batches are supported'
 
-        cfg_key = str(self.phase) # either 'TRAIN' or 'TEST'
+        cfg_key = 'TEST' if self.phase == 1 else 'TRAIN'

d-li14 · 2017-09-18T03:18:50Z

@manipopopo hi, I followed your instructions but met the error below while make pytest

======================================================================
ERROR: test_backward (test_python_layer_with_param_str.TestLayerWithParam)

Traceback (most recent call last):
File "/home/lid/py-faster-rcnn/caffe-fast-rcnn/python/caffe/test/test_python_layer_with_param_str.py", line 46, in setUp
self.net = caffe.Net(net_file, caffe.TRAIN)
File "/home/lid/py-faster-rcnn/caffe-fast-rcnn/python/caffe/test/test_python_layer_with_param_str.py", line 14, in setup
self.value = float(self.param_str)
AttributeError: 'SimpleParamLayer' object has no attribute 'param_str'

======================================================================
ERROR: test_forward (test_python_layer_with_param_str.TestLayerWithParam)

Traceback (most recent call last):
File "/home/lid/py-faster-rcnn/caffe-fast-rcnn/python/caffe/test/test_python_layer_with_param_str.py", line 46, in setUp
self.net = caffe.Net(net_file, caffe.TRAIN)
File "/home/lid/py-faster-rcnn/caffe-fast-rcnn/python/caffe/test/test_python_layer_with_param_str.py", line 14, in setup
self.value = float(self.param_str)
AttributeError: 'SimpleParamLayer' object has no attribute 'param_str'

Ran 51 tests in 10.025s

FAILED (errors=2)
Makefile:536: recipe for target 'pytest' failed
make: *** [pytest] Error 1

Could you please help me?

manipopopo · 2017-09-18T13:45:18Z

You can ignore make pytest for the present. It doesn't affect py-faster-rcnn.

d-li14 · 2017-09-18T23:12:54Z

@manipopopo Thanks a lot, I run the demo.py afterwards and it passed the test, maybe make pytest is definately meaningless in some sense, but my motivation is just to find out what happened.

roxma · 2017-09-25T07:43:37Z

I think this issue should be kept open.

manipopopo's suggestion works, but still it should be fixed by py-faster-rcnn.

mammadjv · 2017-10-31T15:25:33Z

@manipopopo Thanks a lot mate! great !

twangnh · 2017-12-15T10:59:31Z

the method of merge and rebuild caffe works! ./tools/demo.py runs fluently, and the train with ./experiments/scripts/faster_rcnn_end2end.sh runs! but after I changed the train.prototxt, just replace the full connected layers with only one full connected layer and add two conv layer after the pretrained backbone network, and then train again, it prompts Error parsing text-format caffe.NetParameter: 1:3: Message type "caffe.NetParameter" has no field named "me"

vguptai · 2018-02-23T08:22:24Z

For folks who are starting to look into this, give this fork a try. Hopefully, it would help you skip a few steps.

https://github.com/vgupta-ai/py-faster-rcnn

hana9090 · 2018-04-10T12:16:38Z

@manipopopo
I follow your steps for my setup
UBUNTU 17.10
CUDA 8
CUDNN 5

But When I run the demo I have error
I0410 12:13:10.739405 81877 net.cpp:380] conv1_1 -> conv1_1
F0410 12:13:10.898735 81877 cudnn_conv_layer.cpp:53] Check failed: status == CUDNN_STATUS_SUCCESS (4 vs. 0) CUDNN_STATUS_INTERNAL_ERROR
*** Check failure stack trace: ***
Aborted (core dumped)

manipopopo · 2018-04-10T13:30:41Z

@hana9090
I tested it again and got no errors.
Are there any processes using the gpu?
Maybe you could search status == CUDNN_STATUS_SUCCESS (4 vs. 0) CUDNN_STATUS_INTERNAL_ERROR.

hana9090 · 2018-04-10T15:54:44Z

@manipopopo
I download new faster and build again your steps. I have same error :(
How to search about this error? or this status?
Please help

hana9090 · 2018-04-10T16:03:57Z

+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+

manipopopo · 2018-04-10T16:24:37Z

@hana9090
I can't reproduce the problem on my computer. (CUDA 9.1, cuDNN 7)
Maybe you could google status == CUDNN_STATUS_SUCCESS (4 vs. 0) CUDNN_STATUS_INTERNAL_ERROR.
It seems that the error is unrelated to py-faster-rcnn.

hana9090 · 2018-04-10T17:50:40Z

@manipopopo
Thank you it works :)
I just have to run the demo with sudo

William8Work · 2018-05-27T06:08:53Z

@manipopopo thanks for the steps. I got my through all the make problems.

However, when I run tools/demo.py then I got this error:

WARNING: Logging before InitGoogleLogging() is written to STDERR
W0526 23:23:06.432446 25676 _caffe.cpp:139] DEPRECATION WARNING - deprecated use of Python interface
W0526 23:23:06.432467 25676 _caffe.cpp:140] Use this instead (with the named "weights" parameter):
W0526 23:23:06.432471 25676 _caffe.cpp:142] Net('/home/wli/codes/py-faster-rcnn/models/pascal_voc/VGG16/faster_rcnn_alt_opt/faster_rcnn_test.pt', 1, weights='/home/wli/codes/py-faster-rcnn/data/faster_rcnn_models/VGG16_faster_rcnn_final.caffemodel')
[libprotobuf ERROR google/protobuf/text_format.cc:274] Error parsing text-format caffe.NetParameter: 350:21: Message type "caffe.LayerParameter" has no field named "roi_pooling_param".
F0526 23:23:06.433696 25676 upgrade_proto.cpp:90] Check failed: ReadProtoFromTextFile(param_file, param) Failed to parse NetParameter file: /home/wli/codes/py-faster-rcnn/models/pascal_voc/VGG16/faster_rcnn_alt_opt/faster_rcnn_test.pt
*** Check failure stack trace: ***
Aborted (core dumped)

Any thoughts on how to fix this? much appreciated!

zimenglan-sysu-512 closed this as completed Jul 6, 2016

GBJim mentioned this issue Jul 12, 2016

Installation Make Fail CUDA8.0 cuDNN5.0 #248

Open

ShadowLau mentioned this issue Jul 14, 2016

what's different between caffe in py-faster-rcnn and it's original version? #250

Open

twmht mentioned this issue Jul 18, 2016

can't find cublas_v2.h with cuDNN V5 #262

Closed

patricksnape added a commit to patricksnape/caffe-fast-rcnn that referenced this issue Jul 20, 2016

Comment out as suggested by github

60ed504

rbgirshick/py-faster-rcnn#237 (comment)

patricksnape added a commit to patricksnape/caffe-fast-rcnn that referenced this issue Jul 20, 2016

Add extra layerparameter to proto

7b1ee02

rbgirshick/py-faster-rcnn#237 (comment)

Jaspereclipse mentioned this issue Aug 26, 2016

A demo question about gpu #315

Closed

yxliang mentioned this issue Sep 20, 2016

cuDNN v5 support? rbgirshick/caffe-fast-rcnn#14

Open

pham93 mentioned this issue Oct 26, 2016

How to make with cuDNN? #383

Open

PangolinLE314 mentioned this issue May 2, 2017

mnc doesnt recognize any object daijifeng001/MNC#46

Open

dacox added a commit to kinsolresearch/caffe that referenced this issue May 10, 2017

Patched python_layer.hpp.

0f78ce0

- See rbgirshick/py-faster-rcnn#237

mbuckler added a commit to mbuckler/caffe-fast-rcnn-cuDNN-V5 that referenced this issue Jul 29, 2017

https://github.com/rbgirshick/py-faster-rcnn/issues/237

c148eee

nyanmn mentioned this issue Nov 17, 2017

Check failed: status == CUDNN_STATUS_SUCCESS (3 vs. 0) CUDNN_STATUS_BAD_PARAM #733

Open

cy94 mentioned this issue Jan 17, 2018

Loss converged but no detections? #535

Closed

fernandoFernandeSantos mentioned this issue Jan 18, 2018

Running py-faster-rcnn on Jetson TX1? ShaoqingRen/faster_rcnn#71

Open

fpoms mentioned this issue Feb 2, 2018

Object detection example fails (Caffe with support for Faster-RCNN) scanner-research/scanner#142

Closed

chuanzihe mentioned this issue Apr 9, 2018

What recommended CUDA, CUDNN, GCC versions? #803

Open

support 1080P and cudnn v5? #237

support 1080P and cudnn v5? #237

Comments

zimenglan-sysu-512 commented Jul 1, 2016

zimenglan-sysu-512 commented Jul 1, 2016

EvanWeiner commented Jul 2, 2016

manipopopo commented Jul 5, 2016 • edited Loading

zimenglan-sysu-512 commented Jul 6, 2016 • edited Loading

GBJim commented Jul 13, 2016

manipopopo commented Jul 13, 2016

GBJim commented Jul 13, 2016

manipopopo commented Jul 13, 2016

zimenglan-sysu-512 commented Jul 14, 2016 • edited Loading

GBJim commented Jul 14, 2016

manipopopo commented Jul 14, 2016

hengck23 commented Jul 21, 2016

duanLH commented Aug 14, 2016

manipopopo commented Aug 14, 2016

longwoo commented Aug 29, 2016 • edited Loading

manipopopo commented Aug 29, 2016 • edited Loading

longwoo commented Aug 29, 2016

manipopopo commented Aug 29, 2016 • edited Loading

austingg commented Aug 29, 2016

longwoo commented Aug 29, 2016

austingg commented Sep 19, 2016 • edited Loading

vishaltiwari commented Mar 21, 2017 • edited Loading

hana9090 commented Apr 1, 2017

manipopopo commented Apr 1, 2017

hana9090 commented Apr 1, 2017

manipopopo commented Apr 1, 2017

hana9090 commented Apr 1, 2017

hana9090 commented Apr 1, 2017

soulslicer commented Aug 12, 2017

d-li14 commented Sep 18, 2017 • edited Loading

====================================================================== ERROR: test_backward (test_python_layer_with_param_str.TestLayerWithParam)

====================================================================== ERROR: test_forward (test_python_layer_with_param_str.TestLayerWithParam)

manipopopo commented Sep 18, 2017 • edited Loading

d-li14 commented Sep 18, 2017

roxma commented Sep 25, 2017 • edited Loading

mammadjv commented Oct 31, 2017

twangnh commented Dec 15, 2017

vguptai commented Feb 23, 2018 • edited Loading

hana9090 commented Apr 10, 2018

manipopopo commented Apr 10, 2018

hana9090 commented Apr 10, 2018

hana9090 commented Apr 10, 2018

manipopopo commented Apr 10, 2018 • edited Loading

hana9090 commented Apr 10, 2018

William8Work commented May 27, 2018 • edited Loading

manipopopo commented Jul 5, 2016 •

edited

Loading

zimenglan-sysu-512 commented Jul 6, 2016 •

edited

Loading

zimenglan-sysu-512 commented Jul 14, 2016 •

edited

Loading

longwoo commented Aug 29, 2016 •

edited

Loading

manipopopo commented Aug 29, 2016 •

edited

Loading

manipopopo commented Aug 29, 2016 •

edited

Loading

austingg commented Sep 19, 2016 •

edited

Loading

vishaltiwari commented Mar 21, 2017 •

edited

Loading

d-li14 commented Sep 18, 2017 •

edited

Loading

======================================================================
ERROR: test_backward (test_python_layer_with_param_str.TestLayerWithParam)

======================================================================
ERROR: test_forward (test_python_layer_with_param_str.TestLayerWithParam)

manipopopo commented Sep 18, 2017 •

edited

Loading

roxma commented Sep 25, 2017 •

edited

Loading

vguptai commented Feb 23, 2018 •

edited

Loading

manipopopo commented Apr 10, 2018 •

edited

Loading

William8Work commented May 27, 2018 •

edited

Loading