Segmentation fault using python layer in caffe

Question

I'm trying to use python layer as described here . But I am getting this exception:

I1007 17:48:31.366592 30357 layer_factory.hpp:77] Creating layer loss
*** Aborted at 1475851711 (unix time) try "date -d @1475851711" if you are using GNU date ***
PC: @     0x7f32895f1156 (unknown)
*** SIGSEGV (@0x0) received by PID 30357 (TID 0x7f328b07fa40) from PID 0; stack trace: ***
@     0x7f328883ecb0 (unknown)
@     0x7f32895f1156 (unknown)
@     0x7f3289b43dfe (unknown)
@     0x7f32429d0d9c google::protobuf::MessageLite::ParseFromArray()
@     0x7f3242a1f652 google::protobuf::EncodedDescriptorDatabase::Add()
@     0x7f32429da012 google::protobuf::DescriptorPool::InternalAddGeneratedFile()
@     0x7f3242a2b33e google::protobuf::protobuf_AddDesc_google_2fprotobuf_2fdescriptor_2eproto()
@     0x7f3242a5aa75 google::protobuf::StaticDescriptorInitializer_google_2fprotobuf_2fdescriptor_2eproto::StaticDescriptorInitializer_google_2fprotobuf_2fdescriptor_2eproto()
@     0x7f3242a56beb __static_initialization_and_destruction_0()
@     0x7f3242a56c00 _GLOBAL__sub_I_descriptor.pb.cc
@     0x7f328aeca10a (unknown)
@     0x7f328aeca1f3 (unknown)
@     0x7f328aecec30 (unknown)
@     0x7f328aec9fc4 (unknown)
@     0x7f328aece37b (unknown)
@     0x7f327d91b02b (unknown)
@     0x7f328aec9fc4 (unknown)
@     0x7f327d91b62d (unknown)
@     0x7f327d91b0c1 (unknown)
@     0x7f3288f412ae (unknown)
@     0x7f3288f09dae (unknown)
@     0x7f3288f88729 (unknown)
@     0x7f3288ebccbf (unknown)
@     0x7f3288f81d66 (unknown)
@     0x7f3288e47a3f (unknown)
@     0x7f3288f12d43 (unknown)
@     0x7f3288f8b577 (unknown)
@     0x7f3288f6dc13 (unknown)
@     0x7f3288f7154d (unknown)
@     0x7f3288f71682 (unknown)
@     0x7f3288f71a2c (unknown)
@     0x7f3288f88016 (unknown)
Segmentation fault (core dumped)

I am using Ubuntu 14.04, GPU caffe installation. Python layer and prototxt are here . Could anybody suggest anything? I don't know what I am doing wrong.

Answer 1

From my experience, this is one of the most prevalent weaknesses in Caffe: this SegFault without an error message. I generally get this when I haven't connected the data layer properly. For instance, I haven't started the data server, or there's a severe mismatch in format.