how to make decoder in rnn, tensorflow, feeded previous output

Question

I'm wondering how to make decoder in tensorflow rnn, feed it's i th output to (i+1)th input

my inputs have 20 sequence and 3680 dementions and my outputs have 39 sequence and 3680 dementions all data is 0~1 number

here is my model

with tf.variable_scope('encoder'):    
    enc_input = tf.placeholder(tf.float32,[None, input_sequence_length, input_dim])

    enc_cell = tf.contrib.rnn.BasicLSTMCell(num_units = input_sequence_length)

    _ , encoder_states = tf.nn.dynamic_rnn(enc_cell, enc_input ,  dtype=tf.float32)

with tf.variable_scope('decoder'):
    dec_input = tf.placeholder(tf.float32,[None, output_sequence_length, output_dim])
    dec_output = tf.placeholder(tf.float32,[None, output_sequence_length, output_dim])

    dec_cell = tf.contrib.rnn.BasicLSTMCell(num_units = output_sequence_length)

    outputs , _ = tf.nn.dynamic_rnn(dec_cell, dec_input,dtype = tf.float32,
                                    initial_state = encoder_states)

how can i make decoder model that feed previous outputs to next input?

P.S

I make my self-answer code like this

with tf.variable_scope('decoder'):
    dec_input = tf.placeholder(tf.float32,[None, 1, output_dim])
    dec_output = tf.placeholder(tf.float32,[None, output_sequence_length, output_dim])

    outputs = []
    state = encoder_states

    dec_cell = tf.contrib.rnn.BasicLSTMCell(num_units = dec_hidden_size)

    for i in range(output_sequence_length):
        if i==0:
            output , state = tf.nn.dynamic_rnn(dec_cell, dec_input, initial_state = state, dtype = tf.float32)
            outputs.append(output)
        else:
            output , state = tf.nn.dynamic_rnn(dec_cell, 
                                                 output, 
                                                 initial_state = state, 
                                                 dtype = tf.float32)
            outputs.append(output)

outputs = tf.reshape(outputs,[-1,output_dim])
outputs = tf.reshape(outputs,[-1,output_sequence_length,output_dim])

I think this code's output is different from upper code's output but I'm not sure it worked properly.

so still I wonder how to make decoder that has loop function((i)output->(i+1)input) with tensorflow method, because it takes more memory allocation than the upper code. (in my thought it has same cell count)

how to make decoder in rnn, tensorflow, feeded previous output

0 Answers0