'AttentionWrapper' object has no attribute 'zero_state'

[Neel125]

def _build_decoder_cell(self, hparams, encoder_outputs, encoder_state,
                        source_sequence_length):
    """Build a RNN cell with attention mechanism that can be used by decoder."""
    # No Attention
    if not self.has_attention:
        return super(AttentionModel, self)._build_decoder_cell(
            hparams, encoder_outputs, encoder_state, source_sequence_length)
    elif hparams["attention_architecture"] != "standard":
        raise ValueError(
            "Unknown attention architecture %s" % hparams["attention_architecture"])

    num_units = hparams["num_units"]
    num_layers = self.num_decoder_layers
    num_residual_layers = self.num_decoder_residual_layers
    infer_mode = hparams["infer_mode"]

    dtype = tf.float32

    # Ensure memory is batch-major
    if self.time_major:
        memory = tf.transpose(encoder_outputs, [1, 0, 2])
    else:
        memory = encoder_outputs

    if (self.mode == tf.estimator.ModeKeys.PREDICT and
            infer_mode == "beam_search"):
        memory, source_sequence_length, encoder_state, batch_size = (
            self._prepare_beam_search_decoder_inputs(
                hparams["beam_width"], memory, source_sequence_length,
                encoder_state))
    else:
        batch_size = self.batch_size

    # Attention
    attention_mechanism = self.attention_mechanism_fn(
        hparams["attention"], num_units, memory, source_sequence_length, self.mode)

    cell = model_helper.create_rnn_cell(
        unit_type=hparams["unit_type"],
        num_units=num_units,
        num_layers=num_layers,
        num_residual_layers=num_residual_layers,
        forget_bias=hparams["forget_bias"],
        dropout=hparams["dropout"],
        num_gpus=self.num_gpus,
        mode=self.mode,
        single_cell_fn=self.single_cell_fn)

    # Only generate alignment in greedy INFER mode.
    alignment_history = (self.mode == tf.estimator.ModeKeys.PREDICT and
                         infer_mode != "beam_search")
    cell = tfa.seq2seq.AttentionWrapper(
        cell,
        attention_mechanism,
        attention_layer_size=num_units,
        alignment_history=alignment_history,
        output_attention=hparams["output_attention"],
        name="attention")

    # TODO(thangluong): do we need num_layers, num_gpus?
    device = tf.device(model_helper.get_device_str(num_layers-1, self.num_gpus))

    cell = tf.nn.rnn_cell.DeviceWrapper(cell,
                                        device)
    cell = tf.nn.rnn_cell.DropoutWrapper(cell, input_keep_prob=0.8)
    if hparams["pass_hidden_state"]:
        decoder_initial_state = cell.zero_state(batch_size=batch_size*hparams["beam_width"], dtype=dtype).clone(
            cell_state=encoder_state)
    else:
        decoder_initial_state = cell.zero_state(batch_size=batch_size*hparams["beam_width"], dtype=dtype)

    return cell, decoder_initial_state

Error: File "/home/ml-ai4/Neel-dev023/ChatBot/nmt-chatbot/nmt/nmt/attention_model.py", line 144, in _build_decoder_cell decoder_initial_state = cell.zero_state(batch_size=batch_size*hparams["beam_width"], dtype=dtype).clone( File "/home/ml-ai4/Neel-dev023/ChatBot/nmt-chatbot/venv/lib/python3.6/site-packages/tensorflow_core/python/ops/rnn_cell_wrapper_impl.py", line 199, in zero_state return self.cell.zero_state(batch_size, dtype) File "/home/ml-ai4/Neel-dev023/ChatBot/nmt-chatbot/venv/lib/python3.6/site-packages/tensorflow_core/python/ops/rnn_cell_wrapper_impl.py", line 431, in zero_state return self.cell.zero_state(batch_size, dtype) AttributeError: 'AttentionWrapper' object has no attribute 'zero_state'

[Abonia1]

Facing the same issue while try to use tensorflow_addons with tf V2.X

[princebaretto99]

Facing the same issue with tensorflow version 2.x

[princebaretto99]

Got the solution : Just replace zero_state with get_inital_state, because the function get_initial_state returns an AttentionWrapperState tuple containing zeroed out tensors same as zero_state

[Abonia1]

Hello @princebaretto99 I have already found this solution the same day I encountered this issue but really sorry because I forget to update it here in github.Zero_state issue is resolved by using get_initial_state; Thank you for your solution.