echo_words.py 文件源码

def buildCharEncDec(hidden, RNN, layers, maxlen, chars, dropout=.3):
    print('Build model...')
    model = Sequential()
    # "Encode" the input sequence using an RNN,
    # producing an output of HIDDEN_SIZE.
    # Note: In a situation where your input sequences have a variable length,
    # use input_shape=(None, nb_feature).


    # model.add(RNN(hidden, input_shape=(maxlen, len(chars)), 
    #               name="encoder-rnn"))

    model.add(Dropout(dropout, input_shape=(maxlen, len(chars)),
                       noise_shape=(1, maxlen, 1)))
    model.add(RNN(hidden, name="encoder-rnn"))

    # As the decoder RNN's input, repeatedly provide with the 
    # last hidden state of
    # RNN for each time step. Repeat 'DIGITS + 1' times as that's the maximum
    # length of output, e.g., when DIGITS=3, max output is 999+999=1998.
    model.add(RepeatVector(maxlen, name="encoding"))
    # The decoder RNN could be multiple layers stacked or a single layer.
    for ii in range(layers):
        # By setting return_sequences to True, return not only the last output 
        # but all the outputs so far in the form of (nb_samples, timesteps,
        # output_dim). This is necessary as TimeDistributed in the below
        # expects the first dimension to be the timesteps.
        model.add(RNN(hidden, return_sequences=True, 
                      name="decoder%i" % ii))

    # Apply a dense layer to the every temporal slice of an input.
    # For each step
    # of the output sequence, decide which character should be chosen.
    model.add(TimeDistributed(Dense(len(chars), name="dense"), name="td"))
    model.add(Activation('softmax', name="softmax"))
    model.compile(loss='categorical_crossentropy',
                  optimizer='adam',
                  metrics=['accuracy'])
    model.summary()

    return model