SHARE
TWEET

Untitled

a guest Sep 18th, 2019 97 Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
  1. class Decoder(tf.keras.Model):
  2.   def __init__(self, vocab_size, embedding_dim, dec_units, batch_sz):
  3.     super(Decoder, self).__init__()
  4.     self.batch_sz = batch_sz
  5.     self.dec_units = dec_units
  6.     self.embedding = tf.keras.layers.Embedding(vocab_size, embedding_dim)
  7.     self.gru = tf.keras.layers.GRU(self.dec_units,
  8.                                    return_sequences=True,
  9.                                    return_state=True,
  10.                                    recurrent_initializer='glorot_uniform')
  11.     self.fc = tf.keras.layers.Dense(vocab_size)
  12.  
  13.     # used for attention
  14.     self.attention = BahdanauAttention(self.dec_units)
  15.  
  16.   def call(self, x, hidden, enc_output):
  17.     # enc_output shape == (batch_size, max_length, hidden_size)
  18.     context_vector, attention_weights = self.attention(hidden, enc_output)
  19.  
  20.     # x shape after passing through embedding == (batch_size, 1, embedding_dim)
  21.     x = self.embedding(x)
  22.  
  23.     # x shape after concatenation == (batch_size, 1, embedding_dim + hidden_size)
  24.     x = tf.concat([tf.expand_dims(context_vector, 1), x], axis=-1)
  25.  
  26.     # passing the concatenated vector to the GRU
  27.     output, state = self.gru(x)
  28.  
  29.     # output shape == (batch_size * 1, hidden_size)
  30.     output = tf.reshape(output, (-1, output.shape[2]))
  31.  
  32.     # output shape == (batch_size, vocab)
  33.     x = self.fc(output)
  34.  
  35.     return x, state, attention_weights
  36.  
  37.  
  38. decoder = Decoder(vocab_out_size, embedding_dim, units, BATCH_SIZE)
RAW Paste Data
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy. OK, I Understand
 
Top