ngraph.opset5.ops Namespace Reference

Functions
Node	batch_norm_inference (NodeInput data, NodeInput gamma, NodeInput beta, NodeInput mean, NodeInput variance, float epsilon, Optional[str] name=None)
	Perform layer normalizes a input tensor by mean and variance with appling scale and offset. More...
Node	gather_nd (NodeInput data, NodeInput indices, Optional[int] batch_dims=0, Optional[str] name=None)
	Return a node which performs GatherND. More...
Node	log_softmax (NodeInput data, int axis, Optional[str] name=None)
	Apply LogSoftmax operation on each element of input tensor. More...
Node	non_max_suppression (NodeInput boxes, NodeInput scores, Optional[NodeInput] max_output_boxes_per_class=None, Optional[NodeInput] iou_threshold=None, Optional[NodeInput] score_threshold=None, Optional[NodeInput] soft_nms_sigma=None, str box_encoding="corner", bool sort_result_descending=True, str output_type="i64", Optional[str] name=None)
	Return a node which performs NonMaxSuppression. More...
Node	round (NodeInput data, str mode="half_to_even", Optional[str] name=None)
	Apply Round operation on each element of input tensor. More...
Node	lstm_sequence (NodeInput X, NodeInput initial_hidden_state, NodeInput initial_cell_state, NodeInput sequence_lengths, NodeInput W, NodeInput R, NodeInput B, int hidden_size, str direction, List[str] activations=None, List[float] activations_alpha=None, List[float] activations_beta=None, float clip=0.0, Optional[str] name=None)
	Return a node which performs LSTMSequence operation. More...
Node	hsigmoid (NodeInput data, Optional[str] name=None)
	Return a node which performs HSigmoid. More...
Node	gru_sequence (NodeInput X, NodeInput initial_hidden_state, NodeInput sequence_lengths, NodeInput W, NodeInput R, NodeInput B, int hidden_size, str direction, List[str] activations=None, List[float] activations_alpha=None, List[float] activations_beta=None, float clip=0.0, bool linear_before_reset=False, Optional[str] name=None)
	Return a node which performs GRUSequence operation. More...
Node	rnn_sequence (NodeInput X, NodeInput initial_hidden_state, NodeInput sequence_lengths, NodeInput W, NodeInput R, NodeInput B, int hidden_size, str direction, List[str] activations=None, List[float] activations_alpha=None, List[float] activations_beta=None, float clip=0.0, Optional[str] name=None)
	Return a node which performs RNNSequence operation. More...
Node	loop (NodeInput trip_count, NodeInput execution_condition, List[Node] inputs, GraphBody graph_body, List[TensorIteratorSliceInputDesc] slice_input_desc, List[TensorIteratorMergedInputDesc] merged_input_desc, List[TensorIteratorInvariantInputDesc] invariant_input_desc, List[TensorIteratorBodyOutputDesc] body_output_desc, List[TensorIteratorConcatOutputDesc] concat_output_desc, int body_condition_output_idx, int current_iteration_input_idx=-1, Optional[str] name=None)
	Perform recurrent execution of the network described in the body, iterating through the data. More...

Function Documentation

batch_norm_inference()

Node ngraph.opset5.ops.batch_norm_inference	(	NodeInput	data,
		NodeInput	gamma,
		NodeInput	beta,
		NodeInput	mean,
		NodeInput	variance,
		float	epsilon,
		Optional[str]	name = None
	)

Perform layer normalizes a input tensor by mean and variance with appling scale and offset.

@param data: The input tensor with data for normalization.
@param gamma: The scalar scaling for normalized value.
@param beta: The bias added to the scaled normalized value.
@param mean: The value for mean normalization.
@param variance: The value for variance normalization.
@param epsilon: The  number to be added to the variance to avoid division
                by zero when normalizing a value.
@param name: The optional name of the output node.
@return: The new node which performs BatchNormInference.

gather_nd()

Node ngraph.opset5.ops.gather_nd	(	NodeInput	data,
		NodeInput	indices,
		Optional[int]	batch_dims = 0,
		Optional[str]	name = None
	)

Return a node which performs GatherND.

@param data:       N-D tensor with data for gathering
@param indices:    K-D tensor of tuples with indices by which data is gathered
@param batch_dims: Scalar value of batch dimensions
@return: The new node which performs GatherND

gru_sequence()

Node ngraph.opset5.ops.gru_sequence	(	NodeInput	X,
		NodeInput	initial_hidden_state,
		NodeInput	sequence_lengths,
		NodeInput	W,
		NodeInput	R,
		NodeInput	B,
		int	hidden_size,
		str	direction,
		List[str]	activations = None,
		List[float]	activations_alpha = None,
		List[float]	activations_beta = None,
		float	clip = 0.0,
		bool	linear_before_reset = False,
		Optional[str]	name = None
	)

Return a node which performs GRUSequence operation.

@param X: The input tensor. Shape: [batch_size, seq_length, input_size].
@param initial_hidden_state:    The hidden state tensor.
                                Shape: [batch_size, num_directions, hidden_size].
@param sequence_lengths:        Specifies real sequence lengths for each batch element.
                                Shape: [batch_size]. Integer type.
@param W: Tensor with weights for matrix multiplication operation with input portion of data.
          Shape: [num_directions, 3*hidden_size, input_size].
@param R: The tensor with weights for matrix multiplication operation with hidden state.
          Shape: [num_directions, 3*hidden_size, hidden_size].
@param B: The sum of biases (weight and recurrence).
          For linear_before_reset set True the shape is [num_directions, 4*hidden_size].
          Otherwise the shape is [num_directions, 3*hidden_size].
@param hidden_size: Specifies hidden state size.
@param direction: Specifies if the RNN is forward, reverse, or bidirectional.
@param activations: The list of three activation functions for gates.
@param activations_alpha: The list of alpha parameters for activation functions.
@param activations_beta: The list of beta parameters for activation functions.
@param clip: Specifies bound values [-C, C] for tensor clipping performed before activations.
@param linear_before_reset: Flag denotes if the layer behaves according to the modification
                            of GRU described in the formula in the ONNX documentation.
@param name: An optional name of the output node.

@return: The new node represents GRUSequence. Node outputs count: 2.

hsigmoid()

Node ngraph.opset5.ops.hsigmoid	(	NodeInput	data,
		Optional[str]	name = None
	)

Return a node which performs HSigmoid.

@param data: Tensor with input data floating point type.
@return: The new node which performs HSigmoid

log_softmax()

Node ngraph.opset5.ops.log_softmax	(	NodeInput	data,
		int	axis,
		Optional[str]	name = None
	)

Apply LogSoftmax operation on each element of input tensor.

@param data: The tensor providing input data.
@param axis: An axis along which LogSoftmax should be calculated
@return: The new node with LogSoftmax operation applied on each element.

loop()

Node ngraph.opset5.ops.loop	(	NodeInput	trip_count,
		NodeInput	execution_condition,
		List[Node]	inputs,
		GraphBody	graph_body,
		List[TensorIteratorSliceInputDesc]	slice_input_desc,
		List[TensorIteratorMergedInputDesc]	merged_input_desc,
		List[TensorIteratorInvariantInputDesc]	invariant_input_desc,
		List[TensorIteratorBodyOutputDesc]	body_output_desc,
		List[TensorIteratorConcatOutputDesc]	concat_output_desc,
		int	body_condition_output_idx,
		int	current_iteration_input_idx = -1,
		Optional[str]	name = None
	)

Perform recurrent execution of the network described in the body, iterating through the data.

@param trip_count: A scalar or 1D tensor with 1 element specifying
    maximum number of iterations.
@param execution_condition: A scalar or 1D tensor with 1 element
    specifying whether to execute the first iteration or not.
@param      inputs:                The provided to TensorIterator operator.
@param      graph_body:            The graph representing the body we execute.
@param      slice_input_desc:      The descriptors describing sliced inputs, that is nodes
                                   representing tensors we iterate through, processing single
                                   data slice in one iteration.
@param      merged_input_desc:     The descriptors describing merged inputs, that is nodes
                                   representing variables with initial value at first iteration,
                                   which may be changing through iterations.
@param      invariant_input_desc:  The descriptors describing invariant inputs, that is nodes
                                   representing variable with persistent value through all
                                   iterations.
@param      body_output_desc:      The descriptors describing body outputs from specified
                                   iteration.
@param      concat_output_desc:    The descriptors describing specified output values through
                                   all the iterations concatenated into one node.
@param      body_condition_output_idx:    Determines the purpose of the corresponding result in
                                          the graph_body. This result will determine the dynamic
                                          exit condition. If the value of this result is False,
                                          then iterations stop.
@param      current_iteration_input_idx:  Determines the purpose of the corresponding parameter
                                          in the graph_body. This parameter will be used as
                                          an iteration counter. Optional.
@return: The new node which performs Loop.

lstm_sequence()

Node ngraph.opset5.ops.lstm_sequence	(	NodeInput	X,
		NodeInput	initial_hidden_state,
		NodeInput	initial_cell_state,
		NodeInput	sequence_lengths,
		NodeInput	W,
		NodeInput	R,
		NodeInput	B,
		int	hidden_size,
		str	direction,
		List[str]	activations = None,
		List[float]	activations_alpha = None,
		List[float]	activations_beta = None,
		float	clip = 0.0,
		Optional[str]	name = None
	)

Return a node which performs LSTMSequence operation.

@param X: The input tensor. Shape: [batch_size, seq_length, input_size].
@param initial_hidden_state:    The hidden state tensor.
                                Shape: [batch_size, num_directions, hidden_size].
@param initial_cell_state:      The cell state tensor.
                                Shape: [batch_size, num_directions, hidden_size].
@param sequence_lengths:        Specifies real sequence lengths for each batch element.
                                Shape: [batch_size]. Integer type.
@param W: Tensor with weights for matrix multiplication operation with input portion of data.
          Expected format: fico
          Shape: [num_directions, 4*hidden_size, input_size].
@param R: The tensor with weights for matrix multiplication operation with hidden state.
          Expected format: fico
          Shape: [num_directions, 4*hidden_size, hidden_size].
@param B: The sum of biases (weight and recurrence). Expected format: fico
          Shape: [num_directions, 4*hidden_size].
@param hidden_size: Specifies hidden state size.
@param direction: Specifies if the RNN is forward, reverse, or bidirectional.
@param activations: The list of three activation functions for gates.
@param activations_alpha: The list of alpha parameters for activation functions.
@param activations_beta: The list of beta parameters for activation functions.
@param clip: Specifies bound values [-C, C] for tensor clipping performed before activations.
@param name: An optional name of the output node.

@return: The new node represents LSTMSequence. Node outputs count: 3.

non_max_suppression()

Node ngraph.opset5.ops.non_max_suppression	(	NodeInput	boxes,
		NodeInput	scores,
		Optional[NodeInput]	max_output_boxes_per_class = None,
		Optional[NodeInput]	iou_threshold = None,
		Optional[NodeInput]	score_threshold = None,
		Optional[NodeInput]	soft_nms_sigma = None,
		str	box_encoding = "corner",
		bool	sort_result_descending = True,
		str	output_type = "i64",
		Optional[str]	name = None
	)

Return a node which performs NonMaxSuppression.

@param boxes: Tensor with box coordinates.
@param scores: Tensor with box scores.
@param max_output_boxes_per_class: Tensor Specifying maximum number of boxes
                                    to be selected per class.
@param iou_threshold: Tensor specifying intersection over union threshold
@param score_threshold: Tensor specifying minimum score to consider box for the processing.
@param soft_nms_sigma: Tensor specifying the sigma parameter for Soft-NMS.
@param box_encoding: Format of boxes data encoding.
@param sort_result_descending: Flag that specifies whenever it is necessary to sort selected
                               boxes across batches or not.
@param output_type: Output element type.
@return: The new node which performs NonMaxSuppression

rnn_sequence()

Node ngraph.opset5.ops.rnn_sequence	(	NodeInput	X,
		NodeInput	initial_hidden_state,
		NodeInput	sequence_lengths,
		NodeInput	W,
		NodeInput	R,
		NodeInput	B,
		int	hidden_size,
		str	direction,
		List[str]	activations = None,
		List[float]	activations_alpha = None,
		List[float]	activations_beta = None,
		float	clip = 0.0,
		Optional[str]	name = None
	)

Return a node which performs RNNSequence operation.

@param X: The input tensor. Shape: [batch_size, seq_length, input_size].
@param initial_hidden_state:    The hidden state tensor.
                                Shape: [batch_size, num_directions, hidden_size].
@param sequence_lengths:        Specifies real sequence lengths for each batch element.
                                Shape: [batch_size]. Integer type.
@param W: Tensor with weights for matrix multiplication operation with input portion of data.
          Shape: [num_directions, hidden_size, input_size].
@param R: The tensor with weights for matrix multiplication operation with hidden state.
          Shape: [num_directions, hidden_size, hidden_size].
@param B: The sum of biases (weight and recurrence).
          Shape: [num_directions, hidden_size].
@param hidden_size: Specifies hidden state size.
@param direction: Specifies if the RNN is forward, reverse, or bidirectional.
@param activations: The list of three activation functions for gates.
@param activations_alpha: The list of alpha parameters for activation functions.
@param activations_beta: The list of beta parameters for activation functions.
@param clip: Specifies bound values [-C, C] for tensor clipping performed before activations.
@param name: An optional name of the output node.

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

round()

Node ngraph.opset5.ops.round	(	NodeInput	data,
		str	mode = "half_to_even",
		Optional[str]	name = None
	)

Apply Round operation on each element of input tensor.

@param data: The tensor providing input data.
@param mode: Rule to round halfway cases. If set to 'half_to_even' then halfs round to the nearest even
    integer or rounding in such a way that the result heads away from zero if `mode` attribute is
    'half_away_from_zero`.
@param name: An optional name of the output node.
@return: The new node with Round operation applied on each element.