Blob usage in custom layer caffe c++

[Fateh Singh]

I was previously using new operator for allocating memory for intermediate results for the custom loss layer I am building.
The layer was working and I could run my network.
However, that is not the correct way of doing it...
caffe uses Blobs for memory operations

From one of the sample layers , I could get some information
So I did the following

In the loss_layers.hpp , where the definition of my new layer is :

Blob<Dtype> intermediate_sum_;
Blob<Dtype> log_label_;
etc...

In the custom_loss.cpp , the file where I am writing my initialisation,forward and backward pass :

The problem arises as soon as I use

 intermediate_sum_.mutable_cpu_data()

or

 intermediate_sum_.cpu_data()

...for instance

caffe_copy(outer_num_*inner_num_, label, intermediate_sum_.mutable_cpu_data());

All code before this line work and I am quite sure error starts at this line.

The stack trace:


F0624 08:38:50.733919 92222 blob.cpp:102] Check failed: data_
*** Check failure stack trace: ***
    @     0x7fce0a16c5cd  google::LogMessage::Fail()
    @     0x7fce0a16e433  google::LogMessage::SendToLog()
    @     0x7fce0a16c15b  google::LogMessage::Flush()
    @     0x7fce0a16ee1e  google::LogMessageFatal::~LogMessageFatal()
    @     0x7fce0a6003ab  caffe::Blob<>::mutable_cpu_data()
    @     0x7fce0a593c59  caffe::DepthLossLayer<>::Forward_cpu()
    @     0x7fce0a4d07ea  caffe::Net<>::ForwardFromTo()
    @     0x7fce0a4d0b27  caffe::Net<>::ForwardPrefilled()
    @     0x7fce0a5fa8f5  caffe::Solver<>::Step()
    @     0x7fce0a5fb3d4  caffe::Solver<>::Solve()
    @           0x407f09  train()
    @           0x405c58  main
    @     0x7fce095fe830  __libc_start_main
    @           0x406179  _start
    @              (nil)  (unknown)

I think I am using Blob in a wrong way , can someone give an example of how to use Blob , or where am I going wrong?

[Fateh Singh]

It was a runtime error when I run my network, make all works fine