Created by: aprilvkuo

为使您的问题得到快速解决，在建立Issues前，请您先通过如下方式搜索是否有相似问题:【搜索issue关键字】【使用labels筛选】【官方文档】

如果您没有查询到相似问题，为快速解决您的提问，建立issue时请提供如下细节信息：

• 标题：简洁、精准概括您的问题，例如“Insufficient Memory xxx" ”
• 版本、环境信息：    1）PaddlePaddle版本：1.3.0    2）CPU    4）系统环境：centos
• 训练信息    1）单机, 多卡
• 复现信息：如为报错，请给出复现环境、复现步骤
• 问题描述：请详细描述您的问题，同步贴出报错信息、日志、可复现的代码片段 训练时候出core

*** Aborted at 1558436524 (unix time) try "date -d @1558436524" if you are using GNU date ***
PC: @                0x0 (unknown)
*** SIGSEGV (@0x18) received by PID 26669 (TID 0x7f8848604700) from PID 24; stack trace: ***
    @     0x7f88b1236160 (unknown)
    @     0x7f8877ae1e54 paddle::framework::Tensor::mutable_data()
    @     0x7f88772df242 paddle::operators::SequenceExpandAsGradKernel<>::Compute()
    @     0x7f88772df503 _ZNSt17_Function_handlerIFvRKN6paddle9framework16ExecutionContextEEZNKS1_24OpKernelRegistrarFunctorINS0_8platform8CPUPlaceELb0ELm0EJNS0_9operators26SequenceExpandAsGradKernelINS7_16CPUDeviceContextEfEENSA_ISB_dEENSA_ISB_iEENSA_ISB_lEEEEclEPKcSI_iEUlS4_E_E9_M_invokeERKSt9_Any_dataS4_
    @     0x7f8877aa0c93 paddle::framework::OperatorWithKernel::RunImpl()
    @     0x7f8877a9f7bb paddle::framework::OperatorBase::Run()
    @     0x7f88779136a3 _ZNSt17_Function_handlerIFvvEZN6paddle9framework7details19ComputationOpHandle7RunImplEvEUlvE_E9_M_invokeERKSt9_Any_data
    @     0x7f887791333c paddle::framework::details::ComputationOpHandle::RunImpl()
    @     0x7f887790d4bc paddle::framework::details::OpHandleBase::Run()
    @     0x7f88778ac46d _ZZN6paddle9framework7details24ThreadedSSAGraphExecutor5RunOpERKSt10shared_ptrINS0_13BlockingQueueIPNS1_13VarHandleBaseEEEEPNS1_12OpHandleBaseEENKUlvE_clEv
    @     0x7f88772d2363 std::_Function_handler<>::_M_invoke()
    @     0x7f8877280f77 std::__future_base::_State_base::_M_do_set()
    @     0x7f88b1233973 __GI___pthread_once
    @     0x7f88778ab152 _ZNSt13__future_base11_Task_stateISt5_BindIFZN6paddle9framework7details24ThreadedSSAGraphExecutor5RunOpERKSt10shared_ptrINS3_13BlockingQueueIPNS4_13VarHandleBaseEEEEPNS4_12OpHandleBaseEEUlvE_vEESaIiEFvvEE6_M_runEv
    @     0x7f88772823a4 _ZZN10ThreadPoolC1EmENKUlvE_clEv
    @     0x7f889e1218a0 execute_native_thread_routine
    @     0x7f88b122e1c3 start_thread
    @     0x7f88b085612d __clone
    @                0x0 (unknown)
![Uploading image.png…]()

pre_emb的lod_level为0， emb_ner的lod_level为1， 使用sequence_expand_as对其进行扩充得到emb_expand， 并与emb_ner 进行contact。

emb_expand = fluid.layers.sequence_expand_as(x=pre_emb, y=emb_ner) emb = fluid.layers.concat(input=[emb_ner, emb_expand], axis=1)

之后经过sequence_conv_pool进行分类， 训练时候出core