some typo
jszhujun2010 opened this issue · comments
learned a lot from this, some typo found either.
line 177, shape should be [batch, seq_len, target_vocab]
# Make the prediction out of the decoder output.<br/>
logits = tf.layers.dense(dec_out, target_vocab) # [batch, target_vocab]
line 433, should be num_dec_layers
for i in range(self.num_enc_layers):
out = self.decoder_layer(out, enc_out, input_mask, target_mask, f'dec_{i}')