Blob Check Failed

[Admiral JonB]

I've been having numerous problems with getting Caffe to work right. I've got a slightly modified AlexNet implementation from NVIDIA Digits (note, I'm not using digits right now, I'm trying to do things with just pure Caffe). When I try to train, I'm getting the following error/stack trace.

F1021 15:31:23.666375 30299 blob.cpp:115] Check failed: data_ 

*** Check failure stack trace: ***

    @     0x7f72e4994274  google::LogMessage::Fail()

    @     0x7f72e49941bb  google::LogMessage::SendToLog()

    @     0x7f72e4993b92  google::LogMessage::Flush()

    @     0x7f72e4996bef  google::LogMessageFatal::~LogMessageFatal()

    @     0x7f72e4764a6e  caffe::Blob<>::mutable_cpu_data()

    @     0x7f72e47811ca  caffe::GaussianFiller<>::Fill()

    @     0x7f72e47cdb3e  caffe::InnerProductLayer<>::LayerSetUp()

    @     0x7f72e48c7915  caffe::Net<>::Init()

    @     0x7f72e48c87b5  caffe::Net<>::Net()

    @     0x7f72e48ad8fa  caffe::Solver<>::InitTrainNet()

    @     0x7f72e48aeb0c  caffe::Solver<>::Init()

    @     0x7f72e48aee3a  caffe::Solver<>::Solver()

    @     0x7f72e48a1973  caffe::Creator_SGDSolver<>()

    @           0x411fd6  caffe::SolverRegistry<>::CreateSolver()

    @           0x40af52  train()

    @           0x40898c  main

    @     0x7f72e2e19f45  __libc_start_main

    @           0x409293  (unknown)

Aborted (core dumped)

I have no idea what this means, would someone be able to help me figure out what's going on?

I've currently got Caffe built (Ubuntu 14.04) from the github master, and I've attached the prototxt files I'm using. The training set of images I'm using are of numpy shape (3, 256, 256) each before placing them in the lmdb. If any more information is needed, let me know and I'll try to get back asap.

Thanks in advance.

[Jonathan R. Williford]

This usually means that there is an error reading in the data. For example, the path is wrong or the input layer is the wrong type.

Best of luck,

Jonathan