23f6edd6bd81054510274b9045726f387009056b,tflearn/layers/recurrent.py,,gru,#,158

Before Change

                             cell.W[0])
        tf.add_to_collection(tf.GraphKeys.LAYER_VARIABLES + "/" + scope,
                             cell.W[1])
        if not restore:
            tf.add_to_collection(tf.GraphKeys.EXCL_RESTORE_VARS, cell.W[0])
            tf.add_to_collection(tf.GraphKeys.EXCL_RESTORE_VARS, cell.W[1])
        if bias:
            tf.add_to_collection(tf.GraphKeys.LAYER_VARIABLES + "/" + scope,
                                 cell.b[0])
            tf.add_to_collection(tf.GraphKeys.LAYER_VARIABLES + "/" + scope,
                                 cell.b[1])
            if not restore:
                tf.add_to_collection(tf.GraphKeys.EXCL_RESTORE_VARS, cell.b[0])
                tf.add_to_collection(tf.GraphKeys.EXCL_RESTORE_VARS, cell.b[1])
        // Track activations.
        tf.add_to_collection(tf.GraphKeys.ACTIVATIONS, outputs[-1])

    return outputs if return_seq else outputs[-1]

After Change

        tf.add_to_collection(tf.GraphKeys.ACTIVATIONS, outputs[-1])

    o = outputs if return_seq else outputs[-1]
    s = states if return_seq else states[-1]

    return (o, s) if return_states else o


def bidirectional_rnn(incoming, rnncell_fw, rnncell_bw, return_seq=False,
                      return_states=False, initial_state_fw=None,

Instances