NVIDIA / caffe

Caffe: a fast open framework for deep learning.
http://caffe.berkeleyvision.org/
Other
672 stars 263 forks source link

branche 0.17,after InnerProduct layers,BatchNorm got some error message,need help #592

Closed 714142828 closed 3 years ago

714142828 commented 3 years ago

I0125 15:23:57.180774 19274 layer_factory.hpp:172] Creating layer 'loss1/fc' of type 'InnerProduct' I0125 15:23:57.180778 19274 layer_factory.hpp:184] Layer's types are Ftype:FLOAT Btype:FLOAT Fmath:FLOAT Bmath:FLOAT I0125 15:23:57.180795 19274 net.cpp:205] Created Layer loss1/fc (106) I0125 15:23:57.180801 19274 net.cpp:582] loss1/fc <- loss1/conv/bn/sc I0125 15:23:57.180810 19274 net.cpp:552] loss1/fc -> loss1/fc I0125 15:23:57.329124 19274 net.cpp:270] Setting up loss1/fc I0125 15:23:57.329149 19274 net.cpp:277] TEST Top shape for layer 106 'loss1/fc' 1 1024 (1024) I0125 15:23:57.329174 19274 net.cpp:138] Setting types for Layer loss1/fc/bn I0125 15:23:57.329181 19274 layer_factory.hpp:172] Creating layer 'loss1/fc/bn' of type 'BatchNorm' I0125 15:23:57.329188 19274 layer_factory.hpp:184] Layer's types are Ftype:FLOAT Btype:FLOAT Fmath:FLOAT Bmath:FLOAT I0125 15:23:57.329221 19274 net.cpp:205] Created Layer loss1/fc/bn (107) I0125 15:23:57.329226 19274 net.cpp:208] Test Layer for songbo loss1/fc/bn (107) I0125 15:23:57.329236 19274 net.cpp:582] loss1/fc/bn <- loss1/fc I0125 15:23:57.329257 19274 net.cpp:552] loss1/fc/bn -> loss1/fc/bn F0125 15:23:57.329299 19274 blob.hpp:251] Check failed: axis_index < num_axes() (2 vs. 2) axis 2 out of range for 2-D Blob with shape 1 1024 (1024) Check failure stack trace: @ 0x7f2a502710cd google::LogMessage::Fail() @ 0x7f2a50272f33 google::LogMessage::SendToLog() @ 0x7f2a50270c28 google::LogMessage::Flush() @ 0x7f2a50273999 google::LogMessageFatal::~LogMessageFatal() @ 0x55e1c33e3243 caffe::Blob::CanonicalAxisIndex() @ 0x55e1c33e2fba caffe::Blob::shape() @ 0x7f2a5186b51e caffe::BatchNormLayer<>::LayerSetUp() @ 0x7f2a51a54794 caffe::CuDNNBatchNormLayer<>::LayerSetUp() @ 0x7f2a516977c6 caffe::LayerBase::SetUp() @ 0x7f2a5168af46 caffe::Net::Init() @ 0x7f2a5168928b caffe::Net::Net() @ 0x55e1c33db167 test() @ 0x55e1c33df829 main @ 0x7f2a4e98cbf7 __libc_start_main @ 0x55e1c33d829a _start

drnikolaev commented 3 years ago

@714142828 looks like your model is not well formed. Could you attach complete log?

714142828 commented 3 years ago

@714142828 looks like your model is not well formed. Could you attach complete log?

after I modified src/caffe/layers/batch_norm_layer.cpp , and remake caffe,my mode can work well ,thnaks   | 115 | int H = 1;   | 116 | int W = 1;   | 117 | if(bottom[0]->num_axes()>2){   | 118 | H = bottom[0]->shape(2);   | 119 | W = bottom[0]->shape(3);   | 120 | }