Created by: jeromeJT
@lcy-seso @reyoung 在直接使用dense_vector_sequece接一个fc，再接lstm这样的浅层网络可以正常运行
尝试使用paddlebook中语义角色标注例子中的栈式lstm实现深层网络失败，麻烦帮看一下是哪里出问题了，谢谢~
配置：train.py.txt
错误信息：
I0620 14:29:36.920446 25470 GradientMachine.cpp:85] Initing parameters..
I0620 14:29:36.941530 25470 GradientMachine.cpp:92] Init parameters done.
F0620 14:29:37.149533 25490 Matrix.cpp:3275] Check failed: label.getSize() == numSamples (50 vs. 3050)
* Check failure stack trace: *
@ 0x7f3fc48a762d google::LogMessage::Fail()
@ 0x7f3fc48ab0dc google::LogMessage::SendToLog()
@ 0x7f3fc48a7153 google::LogMessage::Flush()
@ 0x7f3fc48ac5ee google::LogMessageFatal::~LogMessageFatal()
F0620 14:29:37.155344 25491 Matrix.cpp:3275] Check failed: label.getSize() == numSamples (50 vs. 3050)
* Check failure stack trace: *
@ 0x7f3fc47fff09 paddle::CpuMatrix::oneHotCrossEntropy()
@ 0x7f3fc48a762d google::LogMessage::Fail()
@ 0x7f3fc4675065 paddle::CostLayer::forward()
@ 0x7f3fc48ab0dc google::LogMessage::SendToLog()
@ 0x7f3fc471b4b0 paddle::NeuralNetwork::forward()
@ 0x7f3fc48a7153 google::LogMessage::Flush()
@ 0x7f3fc48ac5ee google::LogMessageFatal::~LogMessageFatal()
@ 0x7f3fc46fc38c paddle::TrainerThread::forward()
@ 0x7f3fc47fff09 paddle::CpuMatrix::oneHotCrossEntropy()
@ 0x7f3fc46fe54c paddle::TrainerThread::computeThread()
@ 0x7f3fc40a58a0 execute_native_thread_routine
@ 0x7f3fc974b1c3 start_thread
@ 0x7f3fc4675065 paddle::CostLayer::forward()
@ 0x7f3fc8d7312d __clone
@ (nil) (unknown)
Aborted (core dumped)
