/Seq2Seq_Reverse_Char_Demo

Primary LanguagePython

A demo project to understand how Seq2Seq works with character sequences

code source: https://github.com/udacity/deep-learning/tree/master/seq2seq

description: input a character sequence with length <=7, output the reversed sequence. E.g. input = 'apple', output='aelpp'

modifications to source code:

  1. the source and target dictionary are not necessarily generated from training samples, using alpabet to generate a common dictionary will improve efficiency during training
  2. no attention mechanism is original code, apply attention mechanism to model

Environment: Tensorflow 1.3