upsample layer error with configuratiion from sample (CPU)

[martende]

I have tried to use segnet_solver.prototxt config from this http://mi.eng.cam.ac.uk/projects/segnet/tutorial.html .

But it failes with error:

F0209 10:55:24.973695 6643 upsample_layer.cpp:127] upsample top index 0 out of range - height_: 23, width_:30, scale_h_: 0, scale_w_: 0 check scale settings match input pooling layer's downsample setup

*** Check failure stack trace: ***
    @     0x7f972e9275cd  google::LogMessage::Fail()
    @     0x7f972e929433  google::LogMessage::SendToLog()
    @     0x7f972e92715b  google::LogMessage::Flush()
    @     0x7f972e929e1e  google::LogMessageFatal::~LogMessageFatal()
    @     0x7f972ed47394  caffe::UpsampleLayer<>::Backward_cpu()
    @     0x7f972ec37402  caffe::Net<>::BackwardFromTo()
    @     0x7f972ec37601  caffe::Net<>::Backward()
    @     0x7f972ec3093d  caffe::Solver<>::Step()
    @     0x7f972ec31424  caffe::Solver<>::Solve()
    @           0x409279  train()
    @           0x406c78  main
    @     0x7f972ddba830  __libc_start_main
    @           0x407199  _start
    @              (nil)  (unknown)
Aborted (core dumped)

Seems that scale_h_ and scale_w_ are not initialized. Should they ? Does currrent upsample layer implementation support CPU ?

Here the config from sample: https://github.com/alexgkendall/SegNet-Tutorial/blob/master/Models/segnet_train.prototxt