krzysztofjordan

The tensorflow people switched out the order of labels and inputs change this line to look like this: return tf.nn.sampled_softmax_loss(w_t, b, labels, inputs, num_samples, self.target_vocab_size) in seq2seq_model.py