confusing about the input h5 file

[LiuChaoXD]

could you provide more details about the h5 file?

here is my way to construct the h5 file.

1. for each image, there is a h5 file which contains 'label':(1,0.0.0,1,0,0,1,.....)
2. based on the step 1, the h5 file list contains all the path to h5 files

However, when I ''sh ./examples/semihash/train_1.sh", the program occurs the following info:
F0911 15:56:31.543309 11268 semimultilabel_softmax_loss_layer.cpp:29] Check failed: bottom[0]->channels() == bottom[1]->channels() (38 vs. 1) *** Check failure stack trace: *** @ 0x7f494f6985cd google::LogMessage::Fail() @ 0x7f494f69a433 google::LogMessage::SendToLog() @ 0x7f494f69815b google::LogMessage::Flush() @ 0x7f494f69ae1e google::LogMessageFatal::~LogMessageFatal() @ 0x7f494fd52728 caffe::SemiMLabelSoftmaxLossLayer<>::LayerSetUp() @ 0x7f494fcbfb05 caffe::Net<>::Init() @ 0x7f494fcc1511 caffe::Net<>::Net() @ 0x7f494fdce61a caffe::Solver<>::InitTrainNet() @ 0x7f494fdcfac7 caffe::Solver<>::Init() @ 0x7f494fdcfe49 caffe::Solver<>::Solver() @ 0x7f494fc9ce53 caffe::Creator_SGDSolver<>() @ 0x409f28 train() @ 0x406ef8 main @ 0x7f494eb2a830 __libc_start_main @ 0x4075a9 _start @ (nil) (unknown) Aborted (core dumped)

[LiuChaoXD]

ok, I solved it.
One suggestion:
the author should introduce the 'label' data in h5 file more concretely.
For example, the 'label' data can be constructed by np.zeros((1,38)), and the multilabel information can be denoted by [0,1,0,1,1,1,1,0,0...]。

Is right？

(0,1,0,1,1,1,1,0,0...) is confusing a lot.