0

I have successfully installed caffe. all tests ran fine. when i try to train an example(https://github.com/tenstep/DeepModel) using hdf5 data i get error. I read in some forum that it relates to glog.

*** Aborted at 1465829389 (unix time) try "date -d @1465829389" if     you are using GNU date ***
PC: @     0x7fb5bd59d754 (unknown)
*** SIGSEGV (@0xc0) received by PID 17515 (TID 0x7fb5bf7ae780) from     PID     192; stack trace: ***
    @     0x7fb5bd57fd40 (unknown)
    @     0x7fb5bd59d754 (unknown)
    @     0x7fb5bd5a6147 (unknown)
    @     0x7fb5befc9cd9 caffe::DeepHandModelLayer<>::LayerSetUp()
    @     0x7fb5bf049515 caffe::Net<>::Init()
    @     0x7fb5bf04a3b5 caffe::Net<>::Net()
    @     0x7fb5befd472a caffe::Solver<>::InitTrainNet()
    @     0x7fb5befd593c caffe::Solver<>::Init()
    @     0x7fb5befd5c6a caffe::Solver<>::Solver()
    @     0x7fb5bf02d3e3 caffe::Creator_SGDSolver<>()
    @           0x411666 caffe::SolverRegistry<>::CreateSolver()
    @           0x40ab20 train()
    @           0x40852c main
    @     0x7fb5bd56aec5 (unknown)
    @           0x408cfd (unknown)
    @                0x0 (unknown)
Segmentation fault
user27665
  • 673
  • 7
  • 27
  • the trace was produced by `glog` but the segmentatio fault seems to come from caffe itself. Can you post minimal example that describes your hdf5 data creation and the prototxt you are using? – Shai Jun 14 '16 at 06:32

2 Answers2

0

It seems like you are getting a segmentation fault setting up DeepHandModelLayer. This layer is not part of caffe, but rather a "private" addition to caffe.

  1. This layer reads several configuration files from folder configuration/. Make sure you have the required files in the right path.

  2. If you still experience difficulties with this layer, you should contact its authors, as it is a very specific topic.

Shai
  • 111,146
  • 38
  • 238
  • 371
0
::google::InitGoogleLogging(argv[0]);

That's all you need to suppress caffe output instead of messing the vanilla code. It took me a long while to figure this out. Click here to check the example from caffe.

Jeff Wen
  • 43
  • 1
  • 7