demo/quick_start/api_train.sh报错

[huangpingchun]

在参数--trainer_count=1 时候报错，无论是use_gpu 设定1还是0 都存在这个情况，当trainer_count >=2 后 是正常的，无报错，使用train.sh是正常的
错误详情：

I0526 09:56:27.058514 22934 Util.cpp:166] commandline: --use_gpu=0 --trainer_count=1
[INFO 2017-05-26 09:56:35,802 networks.py:1482] The input order is [word, label]
[INFO 2017-05-26 09:56:35,802 networks.py:1488] The output order is [cost_0]
I0526 09:56:35.804144 22934 Trainer.cpp:165] trainer mode: Normal
I0526 09:56:35.804183 22934 GradientMachine.cpp:85] Initing parameters..
I0526 09:56:35.806304 22934 GradientMachine.cpp:92] Init parameters done.
F0526 09:56:35.899184 22934 MemoryHandle.cpp:49] Check failed: size != 0 allocate 0 bytes
*** Check failure stack trace: ***
@ 0x7fcca0ad410d google::LogMessage::Fail()
@ 0x7fcca0ad6458 google::LogMessage::SendToLog()
@ 0x7fcca0ad3c1b google::LogMessage::Flush()
@ 0x7fcca0ad732e google::LogMessageFatal::~LogMessageFatal()
@ 0x7fcca093de9b paddle::CpuMemoryHandle::CpuMemoryHandle()
@ 0x7fcca091be13 paddle::CpuMatrix::CpuMatrix()
@ 0x7fcca091bf7e paddle::Matrix::create()
@ 0x7fcca092c2f8 paddle::Matrix::resizeOrCreate()
@ 0x7fcca0881eae paddle::Layer::resetSpecifyOutput()
@ 0x7fcca088207f paddle::Layer::reserveOutput()
@ 0x7fcca085e897 paddle::FullyConnectedLayer::forward()
@ 0x7fcca077ea03 paddle::NeuralNetwork::forward()
@ 0x7fcca0c31ef0 paddle::TrainerInternal::forwardBackwardBatch()
@ 0x7fcca0c3304b paddle::TrainerInternal::trainOneBatch()
@ 0x7fcca0c2954d paddle::Trainer::trainOneDataBatch()
@ 0x7fcca0ab7664 Trainer::trainOneDataBatch()
@ 0x7fcca06d9032 _wrap_Trainer_trainOneDataBatch
@ 0x4c468a PyEval_EvalFrameEx
@ 0x4c9d8f PyEval_EvalFrameEx
@ 0x4c9d8f PyEval_EvalFrameEx
@ 0x4c2765 PyEval_EvalCodeEx
@ 0x4c2509 PyEval_EvalCode
@ 0x4f1def (unknown)
@ 0x4ec652 PyRun_FileExFlags
@ 0x4eae31 PyRun_SimpleFileExFlags
@ 0x49e14a Py_Main
@ 0x7fcca388b830 __libc_start_main
@ 0x49d9d9 _start
@ (nil) (unknown)