paddle.layer.warp_ctc 加权重训练出错
Created by: gaoyiyeah
由于实验中需要给不同的ctc_loss加权重，但发现会报错
例如：
import paddle.v2 as paddle
ctc_loss = paddle.layer.warp_ctc(input,label,size,blank) 可以正常运行
ctc_loss = 1.0 * paddle.layer.warp_ctc(input,label,size,blank) 运行则报错：
F1206 17:19:14.852391 16845 SlopeInterceptLayer.cpp:69] Check failed: size == inV->getWidth() (29 vs. 1)
* Check failure stack trace: *
@ 0x7f903d92ef2d google::LogMessage::Fail()
@ 0x7f903d9329dc google::LogMessage::SendToLog()
@ 0x7f903d92ea53 google::LogMessage::Flush()
@ 0x7f903d933eee google::LogMessageFatal::~LogMessageFatal()
@ 0x7f903d61a694 paddle::SlopeInterceptLayer::forward()
@ 0x7f903d5cf4ad paddle::NeuralNetwork::forward()
@ 0x7f903d5acc94 paddle::TrainerThread::forward()
@ 0x7f903d5adef5 paddle::TrainerThread::computeThread()
@ 0x7f9066a8d8f0 (unknown)
@ 0x7f908a3e76db start_thread
@ 0x7f908a72188f clone
@ (nil) (unknown)
Aborted (core dumped)
