1.Download the raw data and extract it to the folder where the source code is located.
    2.python3 make_tfrecords.py.Processing the raw data to generate the tfrecord files for training and validating.
      In this experiment,I used 200,000 corpus to train and validate the accuracy of the model on 5000 corpus.
    3.python3 train.py.All the training results as shown below.It is not hard to find that the model eventually 
      achieved an accuracy of 0.6050 on the validation set.