Created by: wangsouc

使用paddle v1版本的multi_binary_label_cross_entropy报错，

最后一层配置如下：

lbl = data_layer("label", 1)
output2 = fc_layer(name="output2", input=sim, size=1, act=SigmoidActivation())
if not is_predict:
        lbl = data_layer("label", 1)
        evalor = evaluators.auc_evaluator(output2,lbl)
        outputs([multi_binary_label_cross_entropy(input=output2, label=lbl)])`

lbl 使用的是 integer_value类型。

出错信息如下：

F0627 18:03:17.144136 20786 Vector.cpp:96] Check failed: id < width (1 vs. 1)
*** Check failure stack trace: ***
    @          0x1086a7d  google::LogMessage::Fail()
    @          0x108a52c  google::LogMessage::SendToLog()
    @          0x10865a3  google::LogMessage::Flush()
    @          0x108ba3e  google::LogMessageFatal::~LogMessageFatal()
    @           0xd71b3a  paddle::VectorT<>::toOneHotSparseMatrix()
    @           0xb40513  paddle::MultiBinaryLabelCrossEntropy::forwardImp()
    @           0xb3c6ee  paddle::CostLayer::forward()
    @           0xc34f96  paddle::NeuralNetwork::forward()
    @           0xc16f98  paddle::GradientMachine::forwardBackward()
    @           0xd3e34a  paddle::TrainerInternal::forwardBackwardBatch()
    @           0xd3c9a1  paddle::TrainerInternal::trainOneBatch()
    @           0xd380e0  paddle::Trainer::trainOneDataBatch()
    @           0xd385e9  paddle::Trainer::trainOnePass()
    @           0xd36d0c  paddle::Trainer::train()
    @           0xad3904  main
    @       0x318ae1ecdd  (unknown)
    @           0xad353d  (unknown)