Created by: jamestang0219

Hello, I have trained a LSTM model, and now I'm using PaddlePaddle's prediction python api to test the samples, but durning the prediction, the error occurred. Check failed: index[i] < (int)tableSize, does it means the value is out of boundary? The same sample can be used to train this model, but cannot be predicted, why? thanks. here is the log:

I1103 07:12:24.671309 100820 Util.cpp:144] commandline:  --use_gpu=0
I1103 07:12:24.671407 100820 Util.cpp:113] Calling runInitFunctions
I1103 07:12:24.672067 100820 Util.cpp:126] Call runInitFunctions done.
I1103 07:12:29.378867 100820 GradientMachine.cpp:123] Loading parameters from ../data/pass-00004-208/
F1103 07:13:14.820614 100820 Matrix.cpp:2317] Check failed: index[i] < (int)tableSize (332123 vs. 332123)
*** Check failure stack trace: ***
    @     0x7fa7924dcdaa  (unknown)
    @     0x7fa7924dcce4  (unknown)
    @     0x7fa7924dc6e6  (unknown)
    @     0x7fa7924df687  (unknown)
    @     0x7fa7934aca49  paddle::CpuMatrix::selectRowsImp<>()
    @     0x7fa7934aac66  paddle::CpuMatrix::selectRows()
    @     0x7fa7933e2b37  paddle::TableProjection::forward()
    @     0x7fa7933bd266  paddle::MixedLayer::forward()
    @     0x7fa793358a8c  paddle::NeuralNetwork::forward()
    @     0x7fa793307d25  _wrap_GradientMachine_forward