leitro/simplest_seq2seq

The simplest seq2seq implementation based on encoder-decoder using TensorFlow 1.3

simplest_seq2seq's Issues

images

Figure 1. Architecture of the model

Figure 2. CER of the testing datasets

what is the shape of 'out_data' ?

def seq2seq(features, labels, mode, params):
in_data = features['input']
out_data = features['output']
# (batch_size, seq_length, embed_dim)
in_embed = layers.embed_sequence(in_data, vocab_size=vocab_size, embed_dim=embed_dim, scope='embed')
out_embed = layers.embed_sequence(out_data, vocab_size=vocab_size, embed_dim=embed_dim, scope='embed', reuse=True)