ngraph.rnn_sequence

ngraph.rnn_sequence(X: Union[_pyngraph.Node, int, float, numpy.ndarray], initial_hidden_state: Union[_pyngraph.Node, int, float, numpy.ndarray], sequence_lengths: Union[_pyngraph.Node, int, float, numpy.ndarray], W: Union[_pyngraph.Node, int, float, numpy.ndarray], R: Union[_pyngraph.Node, int, float, numpy.ndarray], B: Union[_pyngraph.Node, int, float, numpy.ndarray], hidden_size: int, direction: str, activations: Optional[List[str]] = None, activations_alpha: Optional[List[float]] = None, activations_beta: Optional[List[float]] = None, clip: float = 0.0, name: Optional[str] = None) → _pyngraph.Node

Return a node which performs RNNSequence operation.

Parameters

  • X – The input tensor. Shape: [batch_size, seq_length, input_size].

  • initial_hidden_state – The hidden state tensor. Shape: [batch_size, num_directions, hidden_size].

  • sequence_lengths – Specifies real sequence lengths for each batch element. Shape: [batch_size]. Integer type.

  • W – Tensor with weights for matrix multiplication operation with input portion of data. Shape: [num_directions, hidden_size, input_size].

  • R – The tensor with weights for matrix multiplication operation with hidden state. Shape: [num_directions, hidden_size, hidden_size].

  • B – The sum of biases (weight and recurrence). Shape: [num_directions, hidden_size].

  • hidden_size – Specifies hidden state size.

  • direction – Specifies if the RNN is forward, reverse, or bidirectional.

  • activations – The list of three activation functions for gates.

  • activations_alpha – The list of alpha parameters for activation functions.

  • activations_beta – The list of beta parameters for activation functions.

  • clip – Specifies bound values [-C, C] for tensor clipping performed before activations.

  • name – An optional name of the output node.

Returns

The new node represents RNNSequence. Node outputs count: 2.