There are limited amount of documentation on Sequence to Sequence CustomHelper
helper = tf.contrib.seq2seq.CustomHelper(initialize_fn = initialize_fn,sample_fn = sample_fn, next_inputs_fn = next_inputs_fn)
in Tensorflow.
Would anyone explain the inputs of the Custom-helper, in terms of the input data
X = tf.placeholder(tf.float32, [batch_size x time_steps x features])
and encoder,
encoder_cell = tf.contrib.rnn.BasicLSTMCell(hidden_size)
initial_state = encoder_cell.zero_state(batch_size, dtype=tf.float32)
and/or possibly
rnn_output, rnn_states = tf.nn.dynamic_rnn(encoder_cell, X, dtype=tf.float32)
?
