HandwritingRecognition-with-MultiDimensionalRecurrentNeuralNetworks
HandwritingRecognition-with-MultiDimensionalRecurrentNeuralNetworks copied to clipboard
A FloatingPointError
Hi, I recently read your program and encountered a problem.Can you give me some suggestions?
running resluts:
D:\tjt_app\python3.5\python.exe E:/多维RNN/HandwritingRecognition-with-MultiDimensionalRecurrentNeuralNetworks-master/train.py
..................Start Training..................
Training data =80000 Validation data = 15000
Epoc no = 0 Loss is = 79.159587 and Prev Cer = 0.000000 Loss is = 79.013560 and Prev Cer = 97.948718 Loss is = 78.771661 and Prev Cer = 97.948718 Loss is = 78.485006 and Prev Cer = 97.948718
Inside exception clause.....
multiprocessing.pool.RemoteTraceback: """ Traceback (most recent call last): File "E:\多维RNN\HandwritingRecognition-with-MultiDimensionalRecurrentNeuralNetworks-master\ctc_loss.py", line 109, in ctc_loss_mass grad = params - grad / (params * absum) FloatingPointError: invalid value encountered in true_divide