Bad error flow in input_data.py

[OrDavidi]

Hi When trying to evaluate on the latest version of speech_data ( http://download.tensorflow.org/data/speech_commands_v0.02.tar.gz )

some of the data is probably corrupted or too short and i got this fatal error :

tensorflow/core/framework/op_kernel.cc:1202] OP_REQUIRES failed at decode_wav_op.cc:55 : Invalid argument: Data too short when trying to read RIFF exception in data

I suggest to wrap this line : data[i - offset, :] = sess.run(self.mfcc_, feed_dict=input_dict).flatten()

with try,except block to avoid train script crash on a single corrupted wav file.

Thanks

[fmbao]

I meet same ERRO , How to solve it by try catch?