Group Operations#

group ov_ops_cpp_api

OpenVINO C++ API to create operations from different opsets. Such API is used to creation models from code, write transformations and traverse the model graph

class AUGRUCell : public ov::op::util::RNNCellBase#
#include <augru_cell.hpp>

AUGRUCell operation.

Public Functions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class AUGRUSequence : public ov::op::util::RNNCellBase#
#include <augru_sequence.hpp>

AUGRUSequence operation.

Public Functions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Abs : public ov::op::util::UnaryElementwiseArithmetic#
#include <abs.hpp>

Elementwise absolute value operation.

Public Functions

Abs() = default#

Constructs an absolute value operation.

Abs(const Output<Node> &arg)#

Constructs an absolute value operation.

Output [d1, ...]

Parameters:

argOutput that produces the input tensor.[d1, ...]

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Acos : public ov::op::util::UnaryElementwiseArithmetic#
#include <acos.hpp>

Elementwise inverse cosine (arccos) operation.

Public Functions

Acos() = default#

Constructs an arccos operation.

Acos(const Output<Node> &arg)#

Constructs an arccos operation.

Output [d1, ...]

Parameters:

argOutput that produces the input tensor.[d1, ...]

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Acosh : public ov::op::util::UnaryElementwiseArithmetic#
#include <acosh.hpp>

Elementwise inverse hyperbolic cos operation.

Public Functions

Acosh() = default#

Constructs an Acosh operation.

Acosh(const Output<Node> &arg)#

Constructs an Acosh operation.

Output [d1, ...]

Parameters:

argOutput that produces the input tensor.[d1, ...]

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class AdaptiveAvgPool : public ov::op::Op#
#include <adaptive_avg_pool.hpp>

Adaptive average pooling operation.

Public Functions

AdaptiveAvgPool(const Output<Node> &data, const Output<Node> &output_shape)#

Constructs adaptive average pooling operation.

Parameters:
  • dataInput data

  • output_shape – 1D tensor describing output shape for spatial dimensions.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class AdaptiveMaxPool : public ov::op::Op#
#include <adaptive_max_pool.hpp>

Adaptive max pooling operation.

Public Functions

AdaptiveMaxPool(const Output<Node> &data, const Output<Node> &output_shape, const ov::element::Type &index_element_type = ov::element::i64)#

Constructs adaptive max pooling operation.

Parameters:
  • dataInput data

  • output_shape – 1D tensor describing output shape for spatial dimensions.

  • index_element_type – Specifies the output tensor type for indices output

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Add : public ov::op::util::BinaryElementwiseArithmetic#
#include <add.hpp>

Elementwise addition operation.

Public Functions

inline Add()#

Constructs an uninitialized addition operation.

Add(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs an addition operation.

Output [d0, ...]

Parameters:
  • arg0Output that produces the first input tensor.[d0, ...]

  • arg1Output that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification. Default is Numpy-style implicit broadcasting.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Asin : public ov::op::util::UnaryElementwiseArithmetic#
#include <asin.hpp>

Elementwise inverse sine (arcsin) operation.

Public Functions

Asin() = default#

Constructs an arcsin operation.

Asin(const Output<Node> &arg)#

Constructs an arcsin operation.

Output [d1, ...]

Parameters:

argOutput that produces the input tensor.[d1, ...]

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Assign : public ov::op::util::AssignBase#
#include <assign.hpp>

Assign operation sets an input value to the variable with variable_id

Public Functions

Assign(const Output<Node> &new_value, const std::string &variable_id)#

Constructs an Assign operation.

Parameters:
  • new_valueNode that produces the input tensor.

  • variable_id – identifier of the variable to be updated.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline virtual std::string get_variable_id() const override#

Returns the identifier of corresponding variable.

class Assign : public ov::op::util::AssignBase#
#include <assign.hpp>

Assign operation sets an input value to the variable with variable_id

Public Functions

Assign(const Output<Node> &new_value, const std::shared_ptr<util::Variable> &variable)#

Constructs an Assign operation.

Parameters:
  • new_valueNode that produces the input tensor.

  • variable – Class for storing and synchronizing element types, shapes and identifiers between pairs of Assign/ReadValue nodes.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline virtual std::string get_variable_id() const override#

Returns the identifier of corresponding variable.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Atan : public ov::op::util::UnaryElementwiseArithmetic#
#include <atan.hpp>

Elementwise inverse tangent (arctan) operation.

Public Functions

Atan() = default#

Constructs an arctan operation.

Atan(const Output<Node> &arg)#

Constructs an arctan operation.

Output [d1, ...]

Parameters:

argOutput that produces the input tensor.[d1, ...]

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Atanh : public ov::op::util::UnaryElementwiseArithmetic#
#include <atanh.hpp>

Elementwise inverse hyperbolic tangent operation.

Public Functions

Atanh() = default#

Constructs an Atanh operation.

Atanh(const Output<Node> &arg)#

Constructs an Atanh operation.

Output [d1, ...]

Parameters:

argOutput that produces the input tensor.[d1, ...]

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class AvgPool : public ov::op::util::AvgPoolBase#
#include <avg_pool.hpp>

Batched average pooling operation.

Public Functions

AvgPool() = default#

Constructs a batched average pooling operation.

AvgPool(const Output<Node> &arg, const Strides &strides, const Shape &pads_begin, const Shape &pads_end, const Shape &kernel, bool exclude_pad, RoundingType rounding_type = RoundingType::FLOOR, const PadType &auto_pad = PadType::EXPLICIT)#

Constructs a batched average pooling operation.

Parameters:
  • arg – The output producing the input data batch tensor.[d1, dn]

  • strides – The strides.[n]

  • pads_begin – The beginning of padding shape.[n]

  • pads_end – The end of padding shape.[n]

  • kernel – The kernel shape.[n]

  • exclude_pad – If false then averages include padding elements, each treated as the number zero. If true, padding elements are entirely ignored when computing averages.

  • rounding_type – Whether to use ceiling or floor rounding type while computing output shape.

  • auto_pad – Padding type to use for additional padded dimensions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class AvgPool : public ov::op::util::AvgPoolBase#
#include <avg_pool.hpp>

Batched average pooling operation.

Public Functions

AvgPool() = default#

Constructs a batched average pooling operation.

AvgPool(const Output<Node> &arg, const Strides &strides, const Shape &pads_begin, const Shape &pads_end, const Shape &kernel, bool exclude_pad, RoundingType rounding_type = RoundingType::FLOOR, const PadType &auto_pad = PadType::EXPLICIT)#

Constructs a batched average pooling operation.

Parameters:
  • arg – The output producing the input data batch tensor.[d1, dn]

  • strides – The strides.[n]

  • pads_begin – The beginning of padding shape.[n]

  • pads_end – The end of padding shape.[n]

  • kernel – The kernel shape.[n]

  • exclude_pad – If false then averages include padding elements, each treated as the number zero. If true, padding elements are entirely ignored when computing averages.

  • rounding_type – Whether to use ceiling or floor rounding type while computing output shape.

  • auto_pad – Padding type to use for additional padded dimensions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class BatchNormInference : public ov::op::Op#
#include <batch_norm.hpp>

BatchNormInference operation.

Public Functions

BatchNormInference(const Output<Node> &input, const Output<Node> &gamma, const Output<Node> &beta, const Output<Node> &mean, const Output<Node> &variance, double epsilon)#
Parameters:
  • input – [., C, …]

  • gamma – gamma scaling for normalized value. [C]

  • beta – bias added to the scaled normalized value [C]

  • mean – value for mean normalization [C]

  • variance – value for variance normalization [C]

  • epsilon – Avoids divsion by 0 if input has 0 variance

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class BatchNormInference : public ov::op::Op#
#include <batch_norm.hpp>

BatchNormInference operation.

Public Functions

BatchNormInference(const Output<Node> &input, const Output<Node> &gamma, const Output<Node> &beta, const Output<Node> &mean, const Output<Node> &variance, double epsilon)#
Parameters:
  • input – [., C, …]

  • gamma – gamma scaling for normalized value. [C]

  • beta – bias added to the scaled normalized value [C]

  • mean – value for mean normalization [C]

  • variance – value for variance normalization [C]

  • epsilon – Avoids divsion by 0 if input has 0 variance

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class BatchToSpace : public ov::op::Op#
#include <batch_to_space.hpp>

BatchToSpace permutes data from the batch dimension of the data tensor into spatial dimensions.

Note

Values from the batch dimension are moved in spatial blocks dimensions.

   Output node produces a tensor with shape:
   `[batch / (block_shape[0] * block_shape[1] * ... * block_shape[N - 1]),
    D_1 * block_shape[1] - crops_begin[1] - crops_end[1],
    D_2 * block_shape[2] - crops_begin[2] - crops_end[2], ...,
    D_{N - 1} * block_shape[N - 1] - crops_begin[N - 1] - crops_end[N - 1]`
    of the same type as `data` input.

Public Functions

BatchToSpace(const Output<Node> &data, const Output<Node> &block_shape, const Output<Node> &crops_begin, const Output<Node> &crops_end)#

Constructs a BatchToSpace operation.

Parameters:
  • dataNode producing the data tensor

  • block_shape – The sizes of the block of values to be moved

  • crops_begin – Specifies the amount to crop from the beginning along each axis of data input

  • crops_end – Specifies the amount to crop from the ending along each axis of data input.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class BinaryConvolution : public ov::op::util::ConvolutionFwdPropBase#
#include <binary_convolution.hpp>

BinaryConvolution operation.

Public Functions

BinaryConvolution() = default#

Constructs a binary convolution operation.

BinaryConvolution(const Output<Node> &data, const Output<Node> &kernel, const Strides &strides, const CoordinateDiff &pads_begin, const CoordinateDiff &pads_end, const Strides &dilations, BinaryConvolutionMode mode, float pad_value, const PadType &auto_pad = PadType::EXPLICIT)#

Constructs a binary convolution operation.

Output [N, C_OUT, R1, ... Rf]

Parameters:
  • data – The node producing the input data batch tensor.

  • kernel – The node producing the filters tensor.

  • strides – The strides.

  • pads_begin – The beginning of padding shape.

  • pads_end – The end of padding shape.

  • dilations – The dilations.

  • mode – Defines how input tensor 0/1 values and weights 0/1 are interpreted.

  • pad_value – Floating-point value used to fill pad area.

  • auto_pad – The pad type for automatically computing padding sizes.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline const BinaryConvolutionMode &get_mode() const#
Returns:

The mode of convolution.

inline float get_pad_value() const#
Returns:

The pad value.

class BitwiseAnd : public ov::op::util::BinaryElementwiseBitwise#
#include <bitwise_and.hpp>

Elementwise bitwise AND operation.

Public Functions

BitwiseAnd() = default#

Constructs a bitwise AND operation.

BitwiseAnd(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a bitwise AND operation.

Output [d0, ...]

Parameters:
  • arg0Output that produces the first input tensor.[d0, ...]

  • arg1Output that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification. Default is Numpy-style implicit broadcasting.

class BitwiseLeftShift : public ov::op::util::BinaryElementwiseBitwise#
#include <bitwise_left_shift.hpp>

Elementwise bitwise BitwiseLeftShift operation.

Public Functions

BitwiseLeftShift() = default#

Constructs a bitwise BitwiseLeftShift operation.

BitwiseLeftShift(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a bitwise BitwiseLeftShift operation.

Output [d0, ...]

Parameters:
  • arg0Node with data to be shifted. [d0, ...]

  • arg1Node with number of shifts. [d0, ...]

  • auto_broadcast – Auto broadcast specification. Default is Numpy-style implicit broadcasting.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class BitwiseNot : public ov::op::Op#
#include <bitwise_not.hpp>

Elementwise bitwise negation operation.

Public Functions

BitwiseNot() = default#

Constructs a bitwise negation operation.

BitwiseNot(const Output<Node> &arg)#

Constructs a bitwise negation operation.

Parameters:

argNode that produces the input tensor.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class BitwiseOr : public ov::op::util::BinaryElementwiseBitwise#
#include <bitwise_or.hpp>

Elementwise bitwise OR operation.

Public Functions

BitwiseOr() = default#

Constructs a bitwise OR operation.

BitwiseOr(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a bitwise OR operation.

Output [d0, ...]

Parameters:
  • arg0Output that produces the first input tensor.[d0, ...]

  • arg1Output that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification. Default is Numpy-style implicit broadcasting.

class BitwiseRightShift : public ov::op::util::BinaryElementwiseBitwise#
#include <bitwise_right_shift.hpp>

Elementwise bitwise BitwiseRightShift operation.

Public Functions

BitwiseRightShift() = default#

Constructs a bitwise BitwiseRightShift operation.

BitwiseRightShift(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a bitwise BitwiseRightShift operation.

Output [d0, ...]

Parameters:
  • arg0Node with data to be shifted. [d0, ...]

  • arg1Node with number of shifts. [d0, ...]

  • auto_broadcast – Auto broadcast specification. Default is Numpy-style implicit broadcasting.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class BitwiseXor : public ov::op::util::BinaryElementwiseBitwise#
#include <bitwise_xor.hpp>

Elementwise bitwise XOR operation.

Public Functions

BitwiseXor() = default#

Constructs a bitwise XOR operation.

BitwiseXor(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a bitwise XOR operation.

Output [d0, ...]

Parameters:
  • arg0Output that produces the first input tensor.[d0, ...]

  • arg1Output that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification. Default is Numpy-style implicit broadcasting.

class Broadcast : public ov::op::util::BroadcastBase#
#include <broadcast.hpp>

Operation which “adds” axes to an input tensor, replicating elements from the input as needed along the new axes.

Public Functions

Broadcast() = default#

Constructs a broadcast operation.

Broadcast(const Output<Node> &arg, const Output<Node> &target_shape, const Output<Node> &axes_mapping, const BroadcastModeSpec &broadcast_spec = BroadcastType::EXPLICIT)#

Constructs a broadcast operation.

Parameters:
  • arg – The input tensor to be broadcast.

  • target_shape – The shape of the output tensor.

  • axes_mapping – The axis positions (0-based) in the result that correspond to input axes. ‘Arg’ tensor is broadcast along the remaining axes. E.g., Input Shape - [3, 4], Target Shape - [3, 5, 4, 4] axes_mapping - [0, 2] => Broadcast along axes 1 and 3. axes_mapping - [0, 3] => Broadcast along axes 1 and 2.

  • broadcast_specBroadcast specification to use for determining broadcast axes. ‘axes_mapping’ should not be provided if mode other than explicit (none) is used.

Broadcast(const Output<Node> &arg, const Output<Node> &target_shape, const BroadcastModeSpec &broadcast_spec = BroadcastType::NUMPY)#

Constructs a broadcast operation.

Parameters:
  • arg – The input tensor to be broadcast.

  • target_shape – The shape of the output tensor.

  • broadcast_specBroadcast specification to use for determining broadcast axes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual std::pair<bool, AxisSet> get_broadcast_axes() const override#
Returns:

true and the AxisSet if broadcast axes can be fully determined.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Broadcast : public ov::op::util::BroadcastBase#
#include <broadcast.hpp>

Operation which “adds” axes to an input tensor, replicating elements from the input as needed along the new axes.

Public Functions

Broadcast() = default#

Constructs a broadcast operation.

Broadcast(const Output<Node> &arg, const Output<Node> &target_shape, const Output<Node> &axes_mapping, const AutoBroadcastSpec &broadcast_spec = AutoBroadcastSpec())#

Constructs a broadcast operation.

Parameters:
  • arg – The input tensor to be broadcast.

  • target_shape – The shape of the output tensor.

  • axes_mapping – The axis positions (0-based) in the result that correspond to input axes. ‘Arg’ tensor is broadcast along the remaining axes. E.g., Input Shape - [3, 4], Target Shape - [3, 5, 4, 4] axes_mapping - [0, 2] => Broadcast along axes 1 and 3. axes_mapping - [0, 3] => Broadcast along axes 1 and 2.

  • broadcast_specBroadcast specification to use for determining broadcast axes. ‘axes_mapping’ is ignored if broadcast_spec is not NONE

Broadcast(const Output<Node> &arg, const Output<Node> &target_shape, const AutoBroadcastSpec &broadcast_spec = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a broadcast operation.

Parameters:
  • arg – The input tensor to be broadcast.

  • target_shape – The shape of the output tensor.

  • broadcast_specBroadcast specification to use for determining broadcast axes

inline const AutoBroadcastSpec &get_broadcast_spec() const#
Returns:

Broadcast Specification.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Bucketize : public ov::op::Op#
#include <bucketize.hpp>

Operation that bucketizes the input based on boundaries.

Public Functions

Bucketize(const Output<Node> &data, const Output<Node> &buckets, const element::Type output_type = element::i64, const bool with_right_bound = true)#

Constructs a Bucketize node.

Parameters:
  • dataInput data to bucketize

  • buckets – 1-D of sorted unique boundaries for buckets

  • output_typeOutput tensor type, “i64” or “i32”, defaults to i64

  • with_right_bound – indicates whether bucket includes the right or left edge of interval. default true = includes right edge

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Ceiling : public ov::op::util::UnaryElementwiseArithmetic#
#include <ceiling.hpp>

Elementwise ceiling operation.

Public Functions

Ceiling() = default#

Constructs a ceiling operation.

Ceiling(const Output<Node> &arg)#

Constructs a ceiling operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Clamp : public ov::op::util::UnaryElementwiseArithmetic#
#include <clamp.hpp>

Performs a clipping operation on all elements of the input node.

All input values that are outside of the <min;max> range are set to ‘min’ or ‘max’ depending on which side of the <min;max> range they are. The values that fall into this range remain unchanged.

Public Functions

Clamp(const Output<Node> &data, const double min, const double max)#

Constructs a Clamp node.

Parameters:
  • data – - Node producing the input tensor

  • min – - the lower bound of the <min;max> range

  • max – - the upper bound of the <min;max> range

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Col2Im : public ov::op::Op#
#include <col2im.hpp>

Operator combining sliding blocks into an image tensor.

Public Functions

Col2Im(const Output<Node> &data, const Output<Node> &output_size, const Output<Node> &kernel_size, const Strides &strides = Strides{1, 1}, const Strides &dilations = Strides{1, 1}, const Shape &pads_begin = Shape{0, 0}, const Shape &pads_end = Shape{0, 0})#

Constructs a Col2Im operation.

Parameters:
  • dataInput tensor with data

  • output_sizeShape of the spatial dimensions of the output image

  • kernel_size – Size of the sliding blocks

  • strides – Stride in the sliding blocks in the input spatial dimensions

  • dilations – Local stride of the elements

  • pads_begin – Paddings at the beginning of each spatial axis, if undefined no padding is applied

  • pads_end – Paddings at the end of each spatial axis, if undefined no padding is applied

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Concat : public ov::op::Op#
#include <concat.hpp>

Concatenation operation.

Public Functions

Concat() = default#

Constructs a concatenation operation.

Concat(const OutputVector &args, int64_t axis)#

Constructs a concatenation operation.

Parameters:
  • args – The outputs producing the input tensors.

  • axis – The axis along which to concatenate the input tensors.

Concat(const NodeVector &args, int64_t axis)#

Constructs a concatenation operation.

Parameters:
  • args – The nodes producing the input tensors.

  • axis – The axis along which to concatenate the input tensors.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline int64_t get_concatenation_axis() const#
Returns:

The concatenation axis.

inline int64_t get_axis() const#
Returns:

The concatenation axis.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Constant : public ov::op::Op#
#include <constant.hpp>

Class for constants.

Public Functions

Constant(const ov::Tensor &tensor)#

Initialize a constant from ov::Tensor.

Parameters:

tensor – The ov::Tensor with data

template<typename T>
inline Constant(const element::Type &type, const Shape &shape, const std::vector<T> &values)#

Constructs a tensor constant.

Parameters:
  • type – The element type of the tensor constant.

  • shape – The shape of the tensor constant.

  • values – A vector of literals for initializing the tensor constant. The size of values must match the size of the shape.

Constant(const element::Type &type, const Shape &shape)#

Create uninitialized constant.

template<class T, class = typename std::enable_if<std::is_fundamental<T>::value>::type>
inline Constant(const element::Type &type, const Shape &shape, T value)#

Constructs a uniform tensor constant.

Parameters:
  • type – The element type of the tensor constant.

  • shape – The shape of the tensor constant.

  • value – A scalar for initializing the uniform tensor constant. The value is broadcast to the specified shape.

Constant(const element::Type &type, const Shape &shape, const std::vector<std::string> &values)#

Constructs a tensor constant This constructor is mainly to support deserialization of constants.

Parameters:
  • type – The element type of the tensor constant.

  • shape – The shape of the tensor constant.

  • values – A list of string values to use as the constant data.

Constant(const element::Type &type, const Shape &shape, const void *data)#

Constructs a tensor constant with the supplied data.

Parameters:
  • type – The element type of the tensor constant.

  • shape – The shape of the tensor constant.

  • data – A void* to constant data.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

Shape get_shape_val() const#

Returns the value of the constant node as a Shape object Can only be used on element::i64 nodes and interprets negative values as zeros.

Strides get_strides_val() const#

Returns the value of the constant node as a Strides object Can only be used on element::i64 nodes and interprets negative values as zeros.

Coordinate get_coordinate_val() const#

Returns the value of the constant node as a Coordinate object Can only be used on element::i64 nodes and interprets negative values as zeros.

CoordinateDiff get_coordinate_diff_val() const#

Returns the value of the constant node as a CoordinateDiff object Can only be used on element::i64 nodes.

AxisVector get_axis_vector_val() const#

Returns the value of the constant node as an AxisVector object Can only be used on element::i64 nodes and interprets negative values as zeros.

AxisSet get_axis_set_val() const#

Returns the value of the constant node as an AxisSet object Can only be used on element::i64 nodes and interprets negative values as zeros. Repeated values are allowed.

size_t get_byte_size() const#

Return data size in bytes.

std::vector<std::string> get_value_strings() const#
Returns:

The initialization literals for the tensor constant.

template<typename T>
inline std::vector<T> cast_vector(int64_t num_elements = -1) const#

Return the Constant’s value as a vector cast to type T.

Template Parameters:

T – Type to which data vector’s entries will be cast.

Parameters:

num_elements – (Optional) Number of elements to cast. In default case returns all elements

Returns:

Constant’s data vector.

const Tensor get_tensor_view() const#

Get view on constant data as tensor.

Returns:

OV::Tensor with constant data.

const Strides &get_strides() const#
Returns:

Constant’s strides in bytes.

Public Static Functions

template<typename T>
static inline std::shared_ptr<Constant> create(const element::Type &type, const Shape &shape, const std::vector<T> &values)#

Wrapper around constructing a shared_ptr of a Constant.

Parameters:
  • type – The element type of the tensor constant.

  • shape – The shape of the tensor constant.

  • values – A vector of values to use as the constant data.

template<typename T>
static inline std::shared_ptr<Constant> create(const element::Type &type, const Shape &shape, std::initializer_list<T> values)#

Wrapper around constructing a shared_ptr of a Constant.

Parameters:
  • type – The element type of the tensor constant.

  • shape – The shape of the tensor constant.

  • values – An initializer_list of values to use as the constant data.

static inline std::shared_ptr<Constant> create(const element::Type &type, const Shape &shape, const void *memory)#

Wrapper around constructing a shared_ptr of a Constant.

Parameters:
  • type – The element type of the tensor constant.

  • shape – The shape of the tensor constant.

  • memory – An continues memory chunk which contains the constant data.

class Convert : public ov::op::Op#
#include <convert.hpp>

Elementwise type conversion operation.

Public Functions

Convert() = default#

Constructs a conversion operation.

Convert(const Output<Node> &arg, const ov::element::Type &destination_type)#

Constructs a conversion operation.

Parameters:
  • argNode that produces the input tensor.

  • destination_type – Element type for the output tensor.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ConvertLike : public ov::op::Op#
#include <convert_like.hpp>

Elementwise type conversion operation.

Public Functions

ConvertLike() = default#

Constructs a conversion operation.

ConvertLike(const Output<Node> &data, const Output<Node> &like)#

Constructs a conversion operation.

Parameters:
  • dataNode that produces the input tensor.

  • likeNode which provides the target type information for the conversion.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class ConvertPromoteTypes : public ov::op::Op#
#include <convert_promote_types.hpp>

Elementwise operation that promote and convert input types to one common datatype.

Public Functions

ConvertPromoteTypes() = default#

Constructs operation that promote and convert input types to one common datatype.

ConvertPromoteTypes(const Output<Node> &input_0, const Output<Node> &input_1, const bool promote_unsafe = false, const bool pytorch_scalar_promotion = false, const element::Type &u64_integer_promotion_target = element::f32)#

Constructs operation that promote and convert input types to one common datatype.

Parameters:
  • input_0Node with datatype to be promoted.

  • input_1Node with datatype to be promoted.

  • promote_unsafe – Bool attribute whether to allow promotions that might result in bit-widening, precision loss and undefined behaviors.

  • pytorch_scalar_promotion – Bool attribute whether to promote scalar input to type provided by non-scalar input when number format is matching.

  • u64_integer_promotion_target – Element type attribute to select promotion result for u64 and signed integers.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

bool get_pytorch_scalar_promotion() const#

Get bool attribute whether to promote scalar input to type provided by non-scalar input when number format is matching.

void set_pytorch_scalar_promotion(bool pytorch_scalar_promotion)#

Set bool attribute whether to promote scalar input to type provided by non-scalar input when number format is matching.

bool get_promote_unsafe() const#

Get bool attribute whether to allow promotions that might result in bit-widening, precision loss and undefined behaviors.

void set_promote_unsafe(bool promote_unsafe)#

Set bool attribute whether to allow promotions that might result in bit-widening, precision loss and undefined behaviors.

const element::Type &get_u64_integer_promotion_target() const#

Get element type attribute to select promotion result for u64 and signed integers.

void set_u64_integer_promotion_target(const element::Type &u64_integer_promotion_target)#

Set element type attribute to select promotion result for u64 and signed integers.

class Convolution : public ov::op::util::ConvolutionFwdPropBase#
#include <convolution.hpp>

Batched convolution operation, with optional window dilation and stride.

Public Functions

Convolution() = default#

Constructs a batched convolution operation.

Convolution(const Output<Node> &data_batch, const Output<Node> &filters, const Strides &strides, const CoordinateDiff &pads_begin, const CoordinateDiff &pads_end, const Strides &dilations, const PadType &auto_pad = PadType::EXPLICIT)#

Constructs a batched convolution operation.

Output [N, C_OUT, R1, ... Rf]

Parameters:
  • data_batch – The node producing the input data batch tensor.[N, C_IN, D1, ... Df]

  • filters – The node producing the filters tensor.[C_OUT, C_IN, F1, ... Ff]

  • strides – The strides.[f]

  • dilations – The dilations.[f]

  • pads_begin – The beginning of padding shape.[f]

  • pads_end – The end of padding shape.[f]

  • auto_pad – The pad type for automatically computing padding sizes.[f]

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class ConvolutionBackpropData : public ov::op::util::ConvolutionBackPropBase#
#include <convolution.hpp>

Data batch backprop for batched convolution operation.

Public Functions

ConvolutionBackpropData() = default#

Constructs a batched-convolution data batch-backprop operation.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

const PartialShape get_output_shape() const#
Returns:

The output spatial dimensions shape.

class Cos : public ov::op::util::UnaryElementwiseArithmetic#
#include <cos.hpp>

Elementwise cosine operation.

Public Functions

Cos() = default#

Constructs a cosine operation.

Cos(const Output<Node> &arg)#

Constructs a cosine operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Cosh : public ov::op::util::UnaryElementwiseArithmetic#
#include <cosh.hpp>

Elementwise hyperbolic cosine (cosh) operation.

Public Functions

Cosh() = default#

Constructs a hyperbolic cosine operation.

Cosh(const Output<Node> &arg)#

Constructs a hyperbolic cosine operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class CTCGreedyDecoder : public ov::op::Op#
#include <ctc_greedy_decoder.hpp>

CTCGreedyDecoder operation.

Public Functions

CTCGreedyDecoder(const Output<Node> &input, const Output<Node> &seq_len, const bool ctc_merge_repeated)#

Constructs a CTCGreedyDecoder operation.

Parameters:
  • input – Logits on which greedy decoding is performed

  • seq_len – Sequence lengths

  • ctc_merge_repeated – Whether to merge repeated labels

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class CTCGreedyDecoderSeqLen : public ov::op::Op#
#include <ctc_greedy_decoder_seq_len.hpp>

Operator performing CTCGreedyDecoder.

Public Functions

CTCGreedyDecoderSeqLen(const Output<Node> &input, const Output<Node> &seq_len, const bool merge_repeated = true, const element::Type &classes_index_type = element::i32, const element::Type &sequence_length_type = element::i32)#

Constructs a CTCGreedyDecoderSeqLen operation.

Parameters:
  • input – 3-D tensor of logits on which greedy decoding is performed

  • seq_len – 1-D tensor of sequence lengths

  • merge_repeated – Whether to merge repeated labels

  • classes_index_type – Specifies the output classes_index tensor type

  • sequence_length_type – Specifies the output sequence_length tensor type

CTCGreedyDecoderSeqLen(const Output<Node> &input, const Output<Node> &seq_len, const Output<Node> &blank_index, const bool merge_repeated = true, const element::Type &classes_index_type = element::i32, const element::Type &sequence_length_type = element::i32)#

Constructs a CTCGreedyDecoderSeqLen operation.

Parameters:
  • input – 3-D tensor of logits on which greedy decoding is performed

  • seq_len – 1-D tensor of sequence lengths

  • blank_index – Scalar or 1-D tensor with 1 element used to mark a blank index

  • merge_repeated – Whether to merge repeated labels

  • classes_index_type – Specifies the output classes_index tensor type

  • sequence_length_type – Specifies the output sequence_length tensor type

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline bool get_merge_repeated() const#

Get merge_repeated attribute.

Returns:

Current value of merge_repeated attribute

inline void set_merge_repeated(bool merge_repeated)#

Set merge_repeated attribute.

Parameters:

merge_repeated – A new value for the attribute

inline const element::Type &get_classes_index_type() const#

Get classes_index_type attribute.

Returns:

Current value of classes_index_type attribute

inline void set_classes_index_type(const element::Type &classes_index_type)#

Set classes_index_type attribute.

Parameters:

classes_index_type – Type of classes_index

inline const element::Type &get_sequence_length_type() const#

Get sequence_length_type attribute.

Returns:

Current value of sequence_length_type attribute

inline void set_sequence_length_type(const element::Type &sequence_length_type)#

Set sequence_length_type attribute.

Parameters:

sequence_length_type – Type of sequence length

class CTCLoss : public ov::op::Op#
#include <ctc_loss.hpp>

CTCLoss operation.

Public Functions

CTCLoss(const Output<Node> &logits, const Output<Node> &logit_length, const Output<Node> &labels, const Output<Node> &label_length, const bool preprocess_collapse_repeated = false, const bool ctc_merge_repeated = true, const bool unique = false)#

Constructs a CTCLoss operation.

Parameters:
  • logits – 3-D tensor of logits

  • logit_length – 1-D tensor of length for each object from a batch

  • labels – 2-D tensor of labels for which likelyhood is estimated using logist

  • label_length – 1-D tensor of length for each label sequence

  • blank_index – Scalar used to mark a blank index

  • preprocess_collapse_repeated – Flag for preprocessing labels before loss calculation

  • ctc_merge_repeated – Flag for merging repeated characters in a potential alignment

  • unique – Flag to find unique elements in a target before matching with alignment

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class CumSum : public ov::op::Op#
#include <cum_sum.hpp>

Tensor cumulative sum operation.

Compute the cumulative sum of the input tensor along the axis specified.

Public Functions

CumSum() = default#

Constructs a cumulative summation operation.

CumSum(const Output<Node> &arg, const Output<Node> &axis, const bool exclusive = false, const bool reverse = false)#

Constructs a cumulative summation operation.

Parameters:
  • arg – The tensor to be summed.

  • axis – zero dimension tensor specifying axis position along which cumulative sum must be performed

  • exclusive – if set to true, the top element is not included

  • reverse – if set to true, will perform the sums in reverse direction

CumSum(const Output<Node> &arg, const bool exclusive = false, const bool reverse = false)#

Constructs a cumulative summation operation with axis = 0.

Parameters:

arg – The tensor to be summed

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class DeformableConvolution : public ov::op::util::DeformableConvolutionBase#
#include <deformable_convolution.hpp>

DeformableConvolution operation.

Public Functions

DeformableConvolution() = default#

Constructs a conversion operation.

DeformableConvolution(const Output<Node> &arg, const Output<Node> &offsets, const Output<Node> &filters, const Strides &strides, const CoordinateDiff &pads_begin, const CoordinateDiff &pads_end, const Strides &dilations, const PadType &auto_pad = PadType::EXPLICIT, const int64_t group = 1, const int64_t deformable_group = 1)#

Constructs a conversion operation.

Parameters:
  • argNode that produces the input tensor.

  • offsetsNode producing the deformable values tensor.

  • filtersNode producing the filters(kernels) tensor with OIZYX layout.

  • stridesConvolution strides.

  • pads_begin – Amount of padding to be added to the beginning along each axis. For example in case of a 2D input the value of (1, 2) means that 1 element will be added to the top and 2 elements to the left.

  • pads_end – Amount of padding to be added to the end along each axis.

  • dilations – The distance in width and height between the weights in the filters tensor.

  • auto_pad – Specifies how the automatic calculation of padding should be done.

  • group – The number of groups which both output and input should be split into.

  • deformable_group – The number of groups which deformable values and output should be split into along the channel axis.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class DeformableConvolution : public ov::op::util::DeformableConvolutionBase#
#include <deformable_convolution.hpp>

DeformableConvolution operation.

Public Functions

DeformableConvolution() = default#

Constructs a conversion operation.

DeformableConvolution(const Output<Node> &arg, const Output<Node> &offsets, const Output<Node> &filters, const Strides &strides, const CoordinateDiff &pads_begin, const CoordinateDiff &pads_end, const Strides &dilations, const PadType &auto_pad = PadType::EXPLICIT, const int64_t group = 1, const int64_t deformable_group = 1, const bool bilinear_interpolation_pad = false)#

Constructs a conversion operation.

Parameters:
  • argNode that produces the input tensor.

  • offsetsNode producing the deformable values tensor.

  • filtersNode producing the filters(kernels) tensor with OIZYX layout.

  • strides – Convolution strides.

  • pads_begin – Amount of padding to be added to the beginning along each axis. For example in case of a 2D input the value of (1, 2) means that 1 element will be added to the top and 2 elements to the left.

  • pads_end – Amount of padding to be added to the end along each axis.

  • dilations – The distance in width and height between the weights in the filters tensor.

  • auto_pad – Specifies how the automatic calculation of padding should be done.

  • group – The number of groups which both output and input should be split into.

  • deformable_group – The number of groups which deformable values and output should be split into along the channel axis.

  • bilinear_interpolation_pad – The flag that determines the mode of bilinear interpolation execution. If the flag is true and the sampling location is within one pixel outside of the feature map boundary, then bilinear interpolation is performed on the zero padded feature map. If the flag is false and the sampling location is within one pixel outside of the feature map boundary, then the sampling location shifts to the inner boundary of the feature map.`

DeformableConvolution(const Output<Node> &arg, const Output<Node> &offsets, const Output<Node> &filters, const Output<Node> &mask, const Strides &strides, const CoordinateDiff &pads_begin, const CoordinateDiff &pads_end, const Strides &dilations, const PadType &auto_pad = PadType::EXPLICIT, const int64_t group = 1, const int64_t deformable_group = 1, const bool bilinear_interpolation_pad = false)#

Constructs a conversion operation.

Parameters:
  • argNode that produces the input tensor.

  • offsetsNode producing the deformable values tensor.

  • filtersNode producing the filters(kernels) tensor with OIZYX layout.

  • maskNode producing the mask(mask) tensor.

  • strides – Convolution strides.

  • pads_begin – Amount of padding to be added to the beginning along each axis. For example in case of a 2D input the value of (1, 2) means that 1 element will be added to the top and 2 elements to the left.

  • pads_end – Amount of padding to be added to the end along each axis.

  • dilations – The distance in width and height between the weights in the filters tensor.

  • auto_pad – Specifies how the automatic calculation of padding should be done.

  • group – The number of groups which both output and input should be split into.

  • deformable_group – The number of groups which deformable values and output should be split into along the channel axis.

  • bilinear_interpolation_pad – The flag that determines the mode of bilinear interpolation execution. If the flag is true and the sampling location is within one pixel outside of the feature map boundary, then bilinear interpolation is performed on the zero padded feature map. If the flag is false and the sampling location is within one pixel outside of the feature map boundary, then the sampling location shifts to the inner boundary of the feature map.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class DeformablePSROIPooling : public ov::op::Op#
#include <deformable_psroi_pooling.hpp>

DeformablePSROIPooling operation.

Public Functions

DeformablePSROIPooling(const Output<Node> &input, const Output<Node> &coords, const Output<Node> &offsets, const int64_t output_dim, const float spatial_scale, const int64_t group_size = 1, const std::string mode = "bilinear_deformable", int64_t spatial_bins_x = 1, int64_t spatial_bins_y = 1, float trans_std = 1, int64_t part_size = 1)#

Constructs a DeformablePSROIPooling operation.

Parameters:
  • inputInput tensor with position sensitive score maps

  • coordsInput tensor with list of five element tuples describing ROI coordinates

  • offsetsInput tensor with transformation values

  • output_dim – Pooled output channel number

  • group_size – Number of horizontal bins per row to divide ROI area, it defines output width and height

  • spatial_scale – Multiplicative spatial scale factor to translate ROI coordinates from their input scale to the scale used when pooling

  • mode – Specifies mode for pooling.

  • spatial_bins_x – Specifies numbers of bins to divide ROI single bin over width

  • spatial_bins_y – Specifies numbers of bins to divide ROI single bin over height

  • no_trans – The flag that specifies whenever third input exists and contains transformation (offset) values

  • trans_std – The value that all transformation (offset) values are multiplied with

  • part_size – The number of parts the output tensor spatial dimensions are divided into. Basically it is the height and width of the third input

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class DepthToSpace : public ov::op::Op#
#include <depth_to_space.hpp>

DepthToSpace permutes data from the depth dimension of the input blob into spatial dimensions.

Output node produces a tensor with shape: [N, C/(blocksize * blocksize), H * blocksize, W * blocksize]

Note

Values from the depth dimension (assuming NCHW layout) are moved in spatial blocks to the height and width dimensions.

Public Functions

DepthToSpace(const Output<Node> &data, const DepthToSpaceMode &mode, std::size_t block_size = 1)#

Constructs a DepthToSpace operation.

Parameters:
  • dataNode producing the input tensor

  • mode – Specifies how the input depth dimension is split to block coordinates

  • block_size – The size of the block of values to be moved

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class DFT : public ov::op::util::FFTBase#
#include <dft.hpp>

An operation DFT that computes the discrete Fourier transformation.

Public Functions

DFT(const Output<Node> &data, const Output<Node> &axes)#

Constructs a DFT operation. DFT is performed for full size axes.

Parameters:
  • dataInput data

  • axes – Axes to perform DFT

DFT(const Output<Node> &data, const Output<Node> &axes, const Output<Node> &signal_size)#

Constructs a DFT operation.

Parameters:
  • dataInput data

  • axes – Axes to perform DFT

  • signal_size – Signal sizes for ‘axes’

class Divide : public ov::op::util::BinaryElementwiseArithmetic#
#include <divide.hpp>

Elementwise division operation.

Public Functions

inline Divide()#

Constructs a division operation.

Divide(const Output<Node> &arg0, const Output<Node> &arg1, bool pythondiv, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a division operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • pythondiv – Use Python style rounding for integral type

  • auto_broadcast – Auto broadcast specification

Divide(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a division operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Einsum : public ov::op::Op#
#include <einsum.hpp>

Einsum operation.

Public Functions

Einsum(const OutputVector &inputs, const std::string &equation)#

Constructs Einsum operation.

Parameters:
  • inputsInput nodes on which Einsum operation performs contraction

  • equation – Einstein summation convention

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

void set_equation(std::string equation)#

Set Einsum equation.

Parameters:

equation – Equation string to be set.

inline const std::string &get_equation() const#

Get an equation of Einsum operation.

Returns:

Einsum equation

Public Static Functions

static void parse_equation(const std::string &equation, std::vector<std::string> &input_subscripts, std::string &output_subscript)#

Check correctness of equation format and extract input subscripts and output subscript.

Parameters:
  • equation – Equation to be parsed and checked

  • input_subscripts – A vector of extracted input subscripts

  • output_subscript – An output subscript

static std::vector<std::string> extract_labels(const std::string &subscript)#

Extract labels (from subscript) that can be alphabetic letters or ellipsis.

Parameters:

subscript – Subscript

Returns:

A vector of extracted labels from the input subscript in the order of appearence

class Elu : public ov::op::util::UnaryElementwiseArithmetic#
#include <elu.hpp>

Exponential Linear Unit x < 0 => f(x) = alpha * (exp(x) - 1.) x >= 0 => f(x) = x.

Public Functions

Elu(const Output<Node> &data, const double alpha)#

Constructs an Elu operation.

Parameters:
  • dataInput tensor

  • alpha – Multiplier for negative values

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class EmbeddingSegmentsSum : public ov::op::Op#
#include <embedding_segments_sum.hpp>

Returns embeddings for given indices.

Public Functions

EmbeddingSegmentsSum() = default#

Constructs a EmbeddingSegmentsSum operation.

EmbeddingSegmentsSum(const Output<Node> &emb_table, const Output<Node> &indices, const Output<Node> &segment_ids, const Output<Node> &num_segments, const Output<Node> &default_index, const Output<Node> &per_sample_weights)#

Constructs a EmbeddingSegmentsSum operation.

EmbeddingSegmentsSum constructs an output tensor by replacing every index in a given input tensor with a row (from the weights matrix) at that index

Parameters:
  • 'emb_table' – tensor containing the embedding lookup table of the module of shape [num_emb, emb_dim1, emb_dim2, …] and of type T

  • 'indices' – tensor of shape [num_indices] and of type T_IND. Required

  • <tt>segment_ids</tt> – tensor of shape [num_indices] and of type T_IND with indices into the output Tensor. Values should be sorted and can be repeated. Required.

  • <tt>num_segments</tt> – scalar of type T_IND indicating the number of segments. Required.

  • 'default_index' – scalar of type T_IND containing default index in embedding table to fill empty “bags”. If not provided empty “bags” are filled with zeros. Optional.

  • 'per_sample_weights' – tensor of the same shape as indices and of type T. Each value in this tensor are multiplied with each value pooled from embedding table for each index. Optional.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class EmbeddingBagOffsets : public ov::op::util::EmbeddingBagOffsetsBase#
#include <embeddingbag_offsets.hpp>

Returns embeddings for given indices.

Public Functions

EmbeddingBagOffsets() = default#

Constructs a EmbeddingBagOffsets operation.

EmbeddingBagOffsets(const Output<Node> &emb_table, const Output<Node> &indices, const Output<Node> &offsets, const Output<Node> &default_index, const Output<Node> &per_sample_weights, const Reduction &reduction = Reduction::SUM)#

Constructs a EmbeddingBagOffsets operation.

EmbeddingBagOffsets constructs an output tensor by replacing every index in a given input tensor with a row (from the weights matrix) at that index

Parameters:
  • emb_table – tensor containing the embedding lookup table of the module of shape [num_emb, emb_dim1, emb_dim2, …] and of type T

  • indices – tensor of shape [num_indices] and of type T_IND. Required

  • offsets – tensor of shape [batch] and of type T_IND containing the starting index positions of each “bag” in indices. Required.

  • default_index – scalar of type T_IND containing default index in embedding table to fill empty “bags”. If set to value -1 or not provided, empty “bags” are filled with zeros. Reverse indexing using negative values is not supported. Optional.

  • per_sample_weights – tensor of the same shape as indices and of type T. Each value in this tensor are multiplied with each value pooled from embedding table for each index. Optional.

  • reduction – enum to select algorithm used to perform reduction of elements in bag. Optional.

class EmbeddingBagOffsetsSum : public ov::op::util::EmbeddingBagOffsetsBase#
#include <embeddingbag_offsets_sum.hpp>

Returns embeddings for given indices.

Public Functions

EmbeddingBagOffsetsSum() = default#

Constructs a EmbeddingBagOffsetsSum operation.

EmbeddingBagOffsetsSum(const Output<Node> &emb_table, const Output<Node> &indices, const Output<Node> &offsets, const Output<Node> &default_index, const Output<Node> &per_sample_weights)#

Constructs a EmbeddingBagOffsetsSum operation.

EmbeddingBagOffsetsSum constructs an output tensor by replacing every index in a given input tensor with a row (from the weights matrix) at that index

Parameters:
  • emb_table – tensor containing the embedding lookup table of the module of shape [num_emb, emb_dim1, emb_dim2, …] and of type T

  • indices – tensor of shape [num_indices] and of type T_IND. Required

  • offsets – tensor of shape [batch] and of type T_IND containing the starting index positions of each “bag” in indices. Required.

  • default_index – scalar of type T_IND containing default index in embedding table to fill empty “bags”. If set to value -1 or not provided, empty “bags” are filled with zeros. Reverse indexing using negative values is not supported. Optional.

  • per_sample_weights – tensor of the same shape as indices and of type T. Each value in this tensor are multiplied with each value pooled from embedding table for each index. Optional.

class EmbeddingBagPacked : public ov::op::util::EmbeddingBagPackedBase#
#include <embeddingbag_packed.hpp>

Returns embeddings for given indices.

Public Functions

EmbeddingBagPacked() = default#

Constructs a EmbeddingBagPacked operation.

EmbeddingBagPacked(const Output<Node> &emb_table, const Output<Node> &indices, const Output<Node> &per_sample_weights, const Reduction &reduction = Reduction::SUM)#

Constructs a EmbeddingBagPacked operation.

EmbeddingBagPacked constructs an output tensor by replacing every index in a given input tensor with a row (from the weights matrix) at that index

Parameters:
  • emb_tableTensor containing the embedding lookup table of the module of shape [num_emb, emb_dim1, emb_dim2, …] and of type T

  • indicesTensor of shape [batch, indices_per_bag] and of type T_IND. Required.

  • per_sample_weights – tensor of the same shape as indices and of type T. Each value in this tensor are multiplied with each value pooled from embedding table for each index. Optional.

  • reduction – enum to select algorithm used to perform reduction of elements in bag. Optional.

class EmbeddingBagPackedSum : public ov::op::util::EmbeddingBagPackedBase#
#include <embeddingbag_packedsum.hpp>

Returns embeddings for given indices.

Public Functions

EmbeddingBagPackedSum() = default#

Constructs a EmbeddingBagPackedSum operation.

EmbeddingBagPackedSum(const Output<Node> &emb_table, const Output<Node> &indices, const Output<Node> &per_sample_weights)#

Constructs a EmbeddingBagPackedSum operation.

EmbeddingBagPackedSum constructs an output tensor by replacing every index in a given input tensor with a row (from the weights matrix) at that index

Parameters:
  • emb_tableTensor containing the embedding lookup table of the module of shape [num_emb, emb_dim1, emb_dim2, …] and of type T

  • indicesTensor of shape [batch, indices_per_bag] and of type T_IND. Required.

  • per_sample_weigths – tensor of the same shape as indices and of type T. Each value in this tensor are multiplied with each value pooled from embedding table for each index. Optional.

class Equal : public ov::op::util::BinaryElementwiseComparison#
#include <equal.hpp>

Elementwise is-equal operation.

Inputs

Type

Description

arg0

\(E[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of any shape and element type.

arg1

\(E[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of the same shape and element type as arg0.

autob

AutoBroadcastSpec

Auto broadcast specification.

Type

Description

\(\texttt{bool}[d_1,\dots,d_n]\)

The tensor \(T\), where \(T[i_1,\dots,i_n] = 1\text{ if }\texttt{arg0}[i_1,\dots,i_n] = \texttt{arg1}[i_1,\dots,i_n]\text{, else } 0\)

Public Functions

inline Equal()#

Constructs an equal operation.

Equal(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs an equal operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Erf : public ov::op::util::UnaryElementwiseArithmetic#
#include <erf.hpp>

Elementwise erf operation.

Public Functions

Erf() = default#

Constructs a floor operation.

Erf(const Output<Node> &arg)#

Constructs a floor operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Exp : public ov::op::util::UnaryElementwiseArithmetic#
#include <exp.hpp>

Elementwise natural exponential (exp) operation.

Public Functions

Exp() = default#

Constructs an exponential operation.

Exp(const Output<Node> &arg)#

Constructs an exponential operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ExperimentalDetectronDetectionOutput : public ov::op::Op#
#include <experimental_detectron_detection_output.hpp>

An operation ExperimentalDetectronDetectionOutput performs non-maximum suppression to generate the detection output using information on location and score predictions.

Public Functions

ExperimentalDetectronDetectionOutput(const Output<Node> &input_rois, const Output<Node> &input_deltas, const Output<Node> &input_scores, const Output<Node> &input_im_info, const Attributes &attrs)#

Constructs a ExperimentalDetectronDetectionOutput operation.

Parameters:
  • input_roisInput rois

  • input_deltasInput deltas

  • input_scoresInput scores

  • input_im_infoInput image info

  • attrsAttributes attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline const Attributes &get_attrs() const#

Returns attributes of the operation ExperimentalDetectronDetectionOutput.

void set_attrs(Attributes attrs)#

Set the attributes of the operation ExperimentalDetectronDetectionOutput.

Parameters:

attrsAttributes to set.

struct Attributes#
#include <experimental_detectron_detection_output.hpp>

Structure that specifies attributes of the operation.

class ExperimentalDetectronGenerateProposalsSingleImage : public ov::op::Op#
#include <experimental_detectron_generate_proposals.hpp>

An operation ExperimentalDetectronGenerateProposalsSingleImage computes ROIs and their scores based on input data.

Public Functions

ExperimentalDetectronGenerateProposalsSingleImage(const Output<Node> &im_info, const Output<Node> &anchors, const Output<Node> &deltas, const Output<Node> &scores, const Attributes &attrs)#

Constructs a ExperimentalDetectronGenerateProposalsSingleImage operation.

Parameters:
  • im_infoInput image info

  • anchorsInput anchors

  • deltasInput deltas

  • scoresInput scores

  • attrs – Operation attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

struct Attributes#
#include <experimental_detectron_generate_proposals.hpp>

Structure that specifies attributes of the operation.

class ExperimentalDetectronPriorGridGenerator : public ov::op::Op#
#include <experimental_detectron_prior_grid_generator.hpp>

An operation ExperimentalDetectronPriorGridGenerator generates prior grids of specified sizes.

Public Functions

ExperimentalDetectronPriorGridGenerator(const Output<Node> &priors, const Output<Node> &feature_map, const Output<Node> &im_data, const Attributes &attrs)#

Constructs a ExperimentalDetectronDetectionOutput operation.

Parameters:
  • priorsInput priors

  • feature_mapInput feature map

  • im_data – Image data

  • attrs – attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline const Attributes &get_attrs() const#

Returns attributes of this operation.

void set_attrs(Attributes attrs)#

Set the attributes of the operation ExperimentalDetectronPriorGridGenerator.

Parameters:

attrsAttributes to set.

struct Attributes#
#include <experimental_detectron_prior_grid_generator.hpp>

Structure that specifies attributes of the operation.

class ExperimentalDetectronROIFeatureExtractor : public ov::op::Op#
#include <experimental_detectron_roi_feature.hpp>

An operation ExperimentalDetectronROIFeatureExtractor is the ROIAlign operation applied over a feature pyramid.

Public Functions

ExperimentalDetectronROIFeatureExtractor(const OutputVector &args, const Attributes &attrs)#

Constructs a ExperimentalDetectronROIFeatureExtractor operation.

Parameters:
ExperimentalDetectronROIFeatureExtractor(const NodeVector &args, const Attributes &attrs)#

Constructs a ExperimentalDetectronROIFeatureExtractor operation.

Parameters:
virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline const Attributes &get_attrs() const#

Returns attributes of the operation.

void set_attrs(Attributes attrs)#

Set the ExperimentalDetectronROIFeatureExtractor’s attributes.

Parameters:

attrsAttributes to set.

struct Attributes#
#include <experimental_detectron_roi_feature.hpp>

Structure that specifies attributes of the operation.

class ExperimentalDetectronTopKROIs : public ov::op::Op#
#include <experimental_detectron_topkrois.hpp>

An operation ExperimentalDetectronTopKROIs, according to the repository is TopK operation applied to probabilities of input ROIs.

Public Functions

ExperimentalDetectronTopKROIs(const Output<Node> &input_rois, const Output<Node> &rois_probs, size_t max_rois = 0)#

Constructs a ExperimentalDetectronTopKROIs operation.

Parameters:
  • input_roisInput rois

  • rois_probs – Probabilities for input rois

  • max_rois – Maximal numbers of output rois

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class ExtractImagePatches : public ov::op::Op#
#include <extractimagepatches.hpp>

ExtractImagePatches operation.

Public Functions

ExtractImagePatches(const Output<Node> &image, const Shape &sizes, const Strides &strides, const Shape &rates, const PadType &auto_pad)#

Constructs a ExtractImagePatches operation.

Parameters:
  • data – 4-D Input data to extract image patches

  • sizes – Patch size in the format of [size_rows, size_cols]

  • strides – Patch movement stride in the format of [stride_rows, stride_cols]

  • rates – Element seleciton rate for creating a patch. in the format of [rate_rows, rate_cols]

  • auto_pad – Padding type. it can be any value from valid, same_lower, same_upper

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Eye : public ov::op::Op#
#include <eye.hpp>

Tensor Eye operation.

Public Functions

Eye(const Output<Node> &num_rows, const Output<Node> &num_columns, const Output<Node> &diagonal_index, const Output<Node> &batch_shape, const ov::element::Type &out_type)#

Constructs a Eye operation.

Parameters:
  • num_rowsNode producing the tensor with row number.

  • num_columnsNode producing the tensor with column number.

  • diagonal_indexNode producing the tensor with the index of diagonal with ones.

  • batch_shapeNode producing the tensor with batch shape.

  • out_typeOutput type of the tensor.

Eye(const Output<Node> &num_rows, const Output<Node> &num_columns, const Output<Node> &diagonal_index, const ov::element::Type &out_type)#

Constructs a Eye operation without batch_shape.

Parameters:
  • num_rowsNode producing the tensor with row number.

  • num_columnsNode producing the tensor with column number.

  • diagonal_indexNode producing the tensor with the index of diagonal with ones.

  • out_typeOutput type of the tensor.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline const ov::element::Type &get_out_type() const#
Returns:

The output tensor type.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class FakeConvert : public ov::op::Op#
#include <fake_convert.hpp>

FakeConvert performs element-wise quantization of input values into a set of values corresponding to a target low-precision type.

Note

FakeConvert is an experimental operation and subject to change.

Public Functions

FakeConvert(const ov::Output<ov::Node> &data, const ov::Output<ov::Node> &scale, std::string destination_type = "f8e4m3")#

Constructs FakeConvert operation (default shift).

Parameters:
  • data – The input data tensor.

  • scaleTensor with a scale factor for the data input.

  • destination_type – The low precision type to be emulated.

FakeConvert(const ov::Output<ov::Node> &data, const ov::Output<ov::Node> &scale, const ov::Output<ov::Node> &shift, std::string destination_type = "f8e4m3")#

Constructs FakeConvert operation.

Parameters:
  • data – The input data tensor.

  • scaleTensor with a scale factor for the data input.

  • shiftTensor with a shift factor for the data input.

  • destination_type – The low precision type to be emulated.

FakeConvert(const ov::Output<ov::Node> &data, const ov::Output<ov::Node> &scale, const ov::element::Type &destination_type)#

Constructs FakeConvert operation (default shift).

Parameters:
  • data – The input data tensor.

  • scaleTensor with a scale factor for the data input.

  • destination_type – The low precision type to be emulated.

FakeConvert(const ov::Output<ov::Node> &data, const ov::Output<ov::Node> &scale, const ov::Output<ov::Node> &shift, const ov::element::Type &destination_type)#

Constructs FakeConvert operation.

Parameters:
  • data – The input data tensor.

  • scaleTensor with a scale factor for the data input.

  • shiftTensor with a shift factor for the data input.

  • destination_type – The low precision type to be emulated.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class FakeQuantize : public ov::op::Op#
#include <fake_quantize.hpp>

Class performing element-wise linear quantization.

Note

Input floating point values are quantized into a discrete set of floating point values.

Public Functions

FakeQuantize(const Output<Node> &data, const Output<Node> &input_low, const Output<Node> &input_high, const Output<Node> &output_low, const Output<Node> &output_high, std::size_t levels, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a FakeQuantize operation node.

Parameters:
  • data[in] The input data tensor.

  • input_low[in] The minimum limit for input values.

  • input_high[in] The maximum limit for input values.

  • output_low[in] The minimum quantized value.

  • output_high[in] The maximum quantized value.

  • levels[in] The number of quantization levels.

  • auto_broadcast[in] AutoBroadcast mode to be used for broadcasting limit values

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Floor : public ov::op::util::UnaryElementwiseArithmetic#
#include <floor.hpp>

Elementwise floor operation.

Public Functions

Floor() = default#

Constructs a floor operation.

Floor(const Output<Node> &arg)#

Constructs a floor operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class FloorMod : public ov::op::util::BinaryElementwiseArithmetic#
#include <floor_mod.hpp>

Elementwise FloorMod operation.

Public Functions

inline FloorMod()#

Constructs an uninitialized addition operation.

FloorMod(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastType::NUMPY)#

Constructs an Floor Mod operation.

Output [d0, ...]

Parameters:
  • arg0Output that produces the first input tensor.[d0, ...]

  • arg1Output that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Gather : public ov::op::util::GatherBase#
#include <gather.hpp>

Gather slices from axis of data according to indices.

Public Functions

Gather(const Output<Node> &params, const Output<Node> &indices, const Output<Node> &axis)#
Parameters:
  • data – The tensor from which slices are gathered

  • indicesTensor with indexes to gather

  • axis – The tensor is a dimension index to gather data from

class Gather : public ov::op::util::GatherBase#
#include <gather.hpp>

Gather slices from axis of data according to indices.

Public Functions

Gather(const Output<Node> &data, const Output<Node> &indices, const Output<Node> &axis, const int64_t batch_dims = 0)#
Parameters:
  • data – The tensor from which slices are gathered

  • indicesTensor with indexes to gather

  • axis – The tensor is a dimension index to gather data from

  • batch_dims – The number of batch dimension in data and indices tensors. If batch_dims = 0 Gather v7 is identical to Gather v1.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Gather : public ov::op::util::GatherBase#
#include <gather.hpp>

Gather slices from axis of data according to indices. Negative indices are supported and indicate reverse indexing from the end.

Subclassed by ov::op::internal::GatherCompressed

Public Functions

Gather(const Output<Node> &data, const Output<Node> &indices, const Output<Node> &axis, const int64_t batch_dims = 0)#
Parameters:
  • data – The tensor from which slices are gathered

  • indicesTensor with indexes to gather

  • axis – The tensor is a dimension index to gather data from

  • batch_dims – The number of batch dimension in data and indices tensors.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class GatherElements : public ov::op::Op#
#include <gather_elements.hpp>

GatherElements operation.

Public Functions

GatherElements(const Output<Node> &data, const Output<Node> &indices, const int64_t axis)#

Constructs a GatherElements operation.

Parameters:
  • dataNode producing data that are gathered

  • indicesNode producing indices by which the operation gathers elements

  • axis – specifies axis along which indices are specified

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class GatherND : public ov::op::util::GatherNDBase#
#include <gather_nd.hpp>

GatherND operation.

Public Functions

GatherND(const Output<Node> &data, const Output<Node> &indices, const size_t batch_dims = 0)#

Constructs a GatherND operation.

Parameters:
  • dataNode producing data that are gathered

  • indicesNode producing indices by which the operation gathers elements or slices from data

  • batch_dims – Specifies a number of batch dimensions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class GatherND : public ov::op::util::GatherNDBase#
#include <gather_nd.hpp>

GatherND operation.

Public Functions

GatherND(const Output<Node> &data, const Output<Node> &indices, const size_t batch_dims = 0)#

Constructs a GatherND operation.

Parameters:
  • dataNode producing data that are gathered

  • indicesNode producing indices by which the operation gathers elements or slices from data

  • batch_dims – Specifies a number of batch dimensions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class GatherTree : public ov::op::Op#
#include <gather_tree.hpp>

Generates the complete beams from the ids per each step and the parent beam ids.

Public Functions

GatherTree(const Output<Node> &step_ids, const Output<Node> &parent_idx, const Output<Node> &max_seq_len, const Output<Node> &end_token)#
Parameters:
  • step_idsTensor of shape [MAX_TIME, BATCH_SIZE, BEAM_WIDTH] with indices from per each step

  • parent_idxTensor of shape [MAX_TIME, BATCH_SIZE, BEAM_WIDTH] with parent beam indices

  • max_seq_lenTensor of shape [BATCH_SIZE] with maximum lengths for each sequence in the batch

  • end_tokenTensor of shape [MAX_TIME, BATCH_SIZE, BEAM_WIDTH]

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Gelu : public ov::op::util::UnaryElementwiseArithmetic#
#include <gelu.hpp>

Gaussian Error Linear Unit f(x) = 0.5 * x * (1 + erf( x / sqrt(2) )

Public Functions

Gelu(const Output<Node> &data)#

Constructs a Gelu operation.

Parameters:

dataInput tensor

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Gelu : public ov::op::util::UnaryElementwiseArithmetic#
#include <gelu.hpp>

Gaussian Error Linear Unit f(x) = 0.5 * x * (1 + erf( x / sqrt(2) ) for “approximation” = “erf” f(x) = 0.5 * x * (1 + tanh([sqrt(2 / pi)] * [x + 0.044715^3]) for “approximation” = “tanh”.

Public Functions

Gelu(const Output<Node> &data, GeluApproximationMode mode = GeluApproximationMode::ERF)#

Constructs a Gelu operation.

Parameters:
  • dataInput tensor

  • mode – Approximation mode

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Greater : public ov::op::util::BinaryElementwiseComparison#
#include <greater.hpp>

Elementwise greater-than operation.

Public Functions

inline Greater()#

Constructs a greater-than operation.

Greater(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a greater-than operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class GreaterEqual : public ov::op::util::BinaryElementwiseComparison#
#include <greater_eq.hpp>

Elementwise greater-than-or-equal operation.

Public Functions

inline GreaterEqual()#

Constructs a greater-than-or-equal operation.

GreaterEqual(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a greater-than-or-equal operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class GridSample : public ov::op::Op#
#include <grid_sample.hpp>

Operator performing interpolated sampling of the input tensor.

Public Functions

GridSample(const Output<Node> &data, const Output<Node> &grid, const Attributes &attributes)#

Constructs a GridSample operation.

Parameters:
  • dataInput data tensor (input image)

  • grid – Normalized interpolation coordinates

  • attrsGridSample attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

struct Attributes#
#include <grid_sample.hpp>

A Structure which contains all GridSample attributes.

class GroupNormalization : public ov::op::Op#
#include <group_normalization.hpp>

GroupNormalization operation over the input tensor.

Public Functions

GroupNormalization(const Output<Node> &data, const Output<Node> &scale, const Output<Node> &bias, int64_t num_groups, double epsilon)#
Parameters:
  • data – The input tensor to be normalized

  • scale – The tensor containing scale values for each channel

  • bias – The tensor containing bias values for each channel

  • num_groups – The number of groups that the channel dimension will be divided into

  • epsilon – The value that prevents divisions by zero in GroupNormalization formula

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class GRUCell : public ov::op::util::RNNCellBase#
#include <gru_cell.hpp>

Class for GRU cell node.

Note

Note this class represents only single cell and not whole GRU layer.

Public Functions

GRUCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &W, const Output<Node> &R, std::size_t hidden_size)#

Constructs GRUCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The weight tensor with shape: [gates_count * hidden_size, input_size].

  • R[in] The recurrence weight tensor with shape: [gates_count * hidden_size, hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

GRUCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &W, const Output<Node> &R, std::size_t hidden_size, const std::vector<std::string> &activations, const std::vector<float> &activations_alpha, const std::vector<float> &activations_beta, float clip, bool linear_before_reset)#

Constructs GRUCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The weight tensor with shape: [gates_count * hidden_size, input_size].

  • R[in] The recurrence weight tensor with shape: [gates_count * hidden_size, hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

GRUCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &W, const Output<Node> &R, const Output<Node> &B, std::size_t hidden_size, const std::vector<std::string> &activations = std::vector<std::string>{"sigmoid", "tanh"}, const std::vector<float> &activations_alpha = {}, const std::vector<float> &activations_beta = {}, float clip = 0.f, bool linear_before_reset = false)#

Constructs GRUCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The weight tensor with shape: [gates_count * hidden_size, input_size].

  • R[in] The recurrence weight tensor with shape: [gates_count * hidden_size, hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

  • B[in] The sum of biases (weight and recurrence) for update, reset and hidden gates. If linear_before_reset := true then biases for hidden gates are placed separately (weight and recurrence). Shape: [gates_count * hidden_size] if linear_before_reset := false Shape: [(gates_count + 1) * hidden_size] if linear_before_reset := true

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

  • linear_before_reset[in] Whether or not to apply the linear transformation before multiplying by the output of the reset gate.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class GRUSequence : public ov::op::util::RNNCellBase#
#include <gru_sequence.hpp>

GRUSequence operation.

Public Functions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class HardSigmoid : public ov::op::Op#
#include <hard_sigmoid.hpp>

Parameterized, bounded sigmoid-like, piecewise linear function. min(max(alpha*x + beta, 0), 1)

Public Functions

HardSigmoid(const Output<Node> &data, const Output<Node> &alpha, const Output<Node> &beta)#

Constructs a HardSigmoid operation.

Parameters:
  • dataInput tensor.

  • alpha[in] A scalar value representing the alpha parameter.

  • beta[in] A scalar value representing the beta parameter.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class HSigmoid : public ov::op::util::UnaryElementwiseArithmetic#
#include <hsigmoid.hpp>

A HSigmoid Activation Function f(x) = min(max(x + 3, 0), 6) / 6 or f(x) = min(ReLU(x + 3), 6) / 6.

Public Functions

HSigmoid(const Output<Node> &arg)#

Constructs a HSigmoid operation.

Parameters:

dataInput tensor

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class HSwish : public ov::op::util::UnaryElementwiseArithmetic#
#include <hswish.hpp>

A HSwish Activation Function f(x) = x * min(max(x + 3, 0), 6) / 6 or f(x) = x * min(ReLU(x + 3), 6) / 6.

Public Functions

HSwish(const Output<Node> &arg)#

Constructs a HSwish (hard version of Swish) operation.

Parameters:

dataInput tensor

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class I420toBGR : public ov::op::util::ConvertColorI420Base#
#include <i420_to_bgr.hpp>

Color conversion operation from I420 to BGR format. Input:

  • Input NV12 image can be represented in two ways: a) Single plane (as it is in the file): I420 height dimension is 1.5x bigger than image height. ‘C’ dimension shall be 1. b) Three separate planes (used this way in many physical video sources): Y, U and V. In this case b1) Y plane has height same as image height. ‘C’ dimension equals to 1 b2) U plane has dimensions: ‘H’ = image_h / 2; ‘W’ = image_w / 2; ‘C’ = 1. b3) V plane has dimensions: ‘H’ = image_h / 2; ‘W’ = image_w / 2; ‘C’ = 1.

  • Supported element types: u8 or any supported floating-point type. Output:

  • Output node will have NHWC layout and shape HxW same as image spatial dimensions.

  • Number of output channels ‘C’ will be 3, as per interleaved BGR format, first channel is B, last is R

    Conversion of each pixel from I420 (YUV) to RGB space is represented by following formulas: R = 1.164 * (Y - 16) + 1.596 * (V - 128) G = 1.164 * (Y - 16) - 0.813 * (V - 128) - 0.391 * (U - 128) B = 1.164 * (Y - 16) + 2.018 * (U - 128) Then R, G, B values are clipped to range (0, 255)

Public Functions

explicit I420toBGR(const Output<Node> &arg)#

Constructs a conversion operation from input image in I420 format As per I420 format definition, node height dimension shall be 1.5 times bigger than image height so that image (w=640, h=480) is represented by NHWC shape {N,720,640,1} (height*1.5 x width)

Parameters:

argNode that produces the input tensor. Input tensor represents image in NV12 format (YUV).

explicit I420toBGR(const Output<Node> &arg_y, const Output<Node> &arg_u, const Output<Node> &arg_v)#

Constructs a conversion operation from 2-plane input image in NV12 format In general case Y channel of image can be separated from UV channel which means that operation needs two nodes for Y and UV planes respectively. Y plane has one channel, and UV has 2 channels, both expect ‘NHWC’ layout.

Parameters:
  • arg_yNode that produces the input tensor for Y plane (NHWC layout). Shall have WxH dimensions equal to image dimensions. ‘C’ dimension equals to 1.

  • arg_uNode that produces the input tensor for U plane (NHWC layout). ‘H’ is half of image height, ‘W’ is half of image width, ‘C’ dimension equals to 1.

  • arg_vNode that produces the input tensor for V plane (NHWC layout). ‘H’ is half of image height, ‘W’ is half of image width, ‘C’ dimension equals to 1.

class I420toRGB : public ov::op::util::ConvertColorI420Base#
#include <i420_to_rgb.hpp>

Color conversion operation from I420 to RGB format. Input:

  • Input NV12 image can be represented in two ways: a) Single plane (as it is in the file): I420 height dimension is 1.5x bigger than image height. ‘C’ dimension shall be 1. b) Three separate planes (used this way in many physical video sources): Y, U and V. In this case b1) Y plane has height same as image height. ‘C’ dimension equals to 1 b2) U plane has dimensions: ‘H’ = image_h / 2; ‘W’ = image_w / 2; ‘C’ = 1. b3) V plane has dimensions: ‘H’ = image_h / 2; ‘W’ = image_w / 2; ‘C’ = 1.

  • Supported element types: u8 or any supported floating-point type. Output:

  • Output node will have NHWC layout and shape HxW same as image spatial dimensions.

  • Number of output channels ‘C’ will be 3, as per interleaved RGB format, first channel is R, last is B

    Conversion of each pixel from I420 (YUV) to RGB space is represented by following formulas: R = 1.164 * (Y - 16) + 1.596 * (V - 128) G = 1.164 * (Y - 16) - 0.813 * (V - 128) - 0.391 * (U - 128) B = 1.164 * (Y - 16) + 2.018 * (U - 128) Then R, G, B values are clipped to range (0, 255)

Public Functions

explicit I420toRGB(const Output<Node> &arg)#

Constructs a conversion operation from input image in I420 format As per I420 format definition, node height dimension shall be 1.5 times bigger than image height so that image (w=640, h=480) is represented by NHWC shape {N,720,640,1} (height*1.5 x width)

Parameters:

argNode that produces the input tensor. Input tensor represents image in NV12 format (YUV).

explicit I420toRGB(const Output<Node> &arg_y, const Output<Node> &arg_u, const Output<Node> &arg_v)#

Constructs a conversion operation from 2-plane input image in NV12 format In general case Y channel of image can be separated from UV channel which means that operation needs two nodes for Y and UV planes respectively. Y plane has one channel, and UV has 2 channels, both expect ‘NHWC’ layout.

Parameters:
  • arg_yNode that produces the input tensor for Y plane (NHWC layout). Shall have WxH dimensions equal to image dimensions. ‘C’ dimension equals to 1.

  • arg_uNode that produces the input tensor for U plane (NHWC layout). ‘H’ is half of image height, ‘W’ is half of image width, ‘C’ dimension equals to 1.

  • arg_vNode that produces the input tensor for V plane (NHWC layout). ‘H’ is half of image height, ‘W’ is half of image width, ‘C’ dimension equals to 1.

class IDFT : public ov::op::util::FFTBase#
#include <idft.hpp>

An operation IDFT that computes the inverse discrete Fourier transformation.

Public Functions

IDFT(const Output<Node> &data, const Output<Node> &axes)#

Constructs a IDFT operation. IDFT is performed for full size axes.

Parameters:
  • dataInput data

  • axes – Axes to perform IDFT

IDFT(const Output<Node> &data, const Output<Node> &axes, const Output<Node> &signal_size)#

Constructs a IDFT operation.

Parameters:
  • dataInput data

  • axes – Axes to perform IDFT

  • signal_size – Signal sizes for ‘axes’

class If : public ov::op::util::MultiSubGraphOp#
#include <if.hpp>

If operation.

Public Functions

If(const Output<Node> &execution_condition)#

Constructs If with condition.

Parameters:

execution_condition – condition node.

inline const std::shared_ptr<Model> &get_then_body() const#

gets then_body as ov::Model.

Returns:

then_body as ov::Model.

inline const std::shared_ptr<Model> &get_else_body() const#

gets else_body as ov::Model.

Returns:

else_body as ov::Model.

inline void set_then_body(const std::shared_ptr<Model> &body)#

sets new ov::Model as new then_body.

Parameters:

body – new body for ‘then’ branch.

inline void set_else_body(const std::shared_ptr<Model> &body)#

sets new ov::Model as new else_body.

Parameters:

body – new body for ‘else’ branch.

void set_input(const Output<Node> &value, const std::shared_ptr<v0::Parameter> &then_parameter, const std::shared_ptr<v0::Parameter> &else_parameter)#

sets new input to the operation associated with parameters of each sub-graphs

Parameters:
  • value – input to operation

  • then_parameter – parameter for then_body or nullptr

  • else_parameter – parameter for else_body or nullpt

Output<Node> set_output(const std::shared_ptr<v0::Result> &then_result, const std::shared_ptr<v0::Result> &else_result)#

sets new output from the operation associated with results of each sub-graphs

Parameters:
  • then_result – result from then_body

  • else_parameter – result from else_body

Returns:

output from operation

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Interpolate : public ov::op::Op#
#include <interpolate.hpp>

Layer which performs bilinear interpolation.

Public Functions

Interpolate(const Output<Node> &image, const Output<Node> &output_shape, const Attributes &attrs)#

Constructs a Interpolate operation.

Parameters:
  • imageInput image

  • output_shapeOutput shape of spatial axes

  • attrs – Interpolation attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

struct Attributes#
#include <interpolate.hpp>

Structure that specifies attributes for interpolation.

class Interpolate : public ov::op::util::InterpolateBase#
#include <interpolate.hpp>

Interpolate operation.

Public Functions

Interpolate(const Output<Node> &image, const Output<Node> &output_shape, const Output<Node> &scales, const InterpolateAttrs &attrs)#

Constructs a Interpolate operation without ‘axes’ input.

Parameters:
  • imageInput image

  • output_shapeOutput shape of spatial axes

  • scales – Scales of spatial axes, i.e. output_shape / input_shape

  • attrs – Interpolation attributes

Interpolate(const Output<Node> &image, const Output<Node> &output_shape, const Output<Node> &scales, const Output<Node> &axes, const InterpolateAttrs &attrs)#

Constructs a Interpolate operation with ‘axes’ input.

Parameters:
  • imageInput image

  • output_shapeOutput shape of spatial axes

  • scales – Scales of spatial axes, i.e. output_shape / input_shape

  • axes – Interpolation axes

  • attrs – Interpolation attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Interpolate : public ov::op::util::InterpolateBase#
#include <interpolate.hpp>

Interpolate operation.

Public Functions

Interpolate(const Output<Node> &image, const Output<Node> &scales_or_sizes, const InterpolateAttrs &attrs)#

Constructs a Interpolate operation without ‘axes’ input.

Parameters:
  • imageInput image

  • scales_or_sizes – Scales of spatial axes, i.e. output_shape / input_shape

  • attrs – Interpolation attributes

Interpolate(const Output<Node> &image, const Output<Node> &scales_or_sizes, const Output<Node> &axes, const InterpolateAttrs &attrs)#

Constructs a Interpolate operation with ‘axes’ input.

Parameters:
  • imageInput image

  • scales_or_sizes – Scales of spatial axes, i.e. output_shape / input_shape

  • axes – Interpolation axes

  • attrs – Interpolation attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Inverse : public ov::op::Op#
#include <inverse.hpp>

Inverse operation computes the inverse of the input tensor.

Public Functions

Inverse(const Output<Node> &data, const bool adjoint = false)#

Inverse operation computes the inverse of the input matrices. The inverse is computed for each MxM matrix separetely, preserving all batch dimensions.

Parameters:
  • dataInput matrices to compute the inverse for. Last two tensor dimensions must be of the same size.

  • adjoint – Boolean that determines whether to return a normal inverse or adjoint (conjugate transpose) of the input matrices.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class IsFinite : public ov::op::Op#
#include <is_finite.hpp>

Boolean mask that maps NaN and Infinity values to false and other values to true.

Public Functions

IsFinite(const Output<Node> &data)#

Constructs a IsFinite operation.

Parameters:

dataInput data tensor

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class IsInf : public ov::op::Op#
#include <is_inf.hpp>

Boolean mask that maps infinite values to true.

Public Functions

IsInf(const Output<Node> &data)#

Constructs a IsInf operation.

Parameters:

dataInput data tensor

IsInf(const Output<Node> &data, const Attributes &attributes)#

Constructs a IsInf operation.

Parameters:
  • dataInput data tensor

  • attrsIsInf attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

struct Attributes#
#include <is_inf.hpp>

A Structure which contains all IsInf attributes.

class IsNaN : public ov::op::Op#
#include <is_nan.hpp>

Boolean mask that maps NaN values to true and other values to false.

Public Functions

IsNaN(const Output<Node> &data)#

Constructs a IsNaN operation.

Parameters:

dataInput data tensor

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Less : public ov::op::util::BinaryElementwiseComparison#
#include <less.hpp>

Elementwise less-than operation.

Public Functions

inline Less()#

Constructs a less-than operation.

Less(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a less-than operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class LessEqual : public ov::op::util::BinaryElementwiseComparison#
#include <less_eq.hpp>

Elementwise less-than-or-equal operation.

Public Functions

inline LessEqual()#

Constructs a less-than-or-equal operation.

LessEqual(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a less-than-or-equal operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Log : public ov::op::util::UnaryElementwiseArithmetic#
#include <log.hpp>

Elementwise natural log operation.

Public Functions

Log() = default#

Constructs a natural log operation.

Log(const Output<Node> &arg)#

Constructs a natural log operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class LogSoftmax : public ov::op::Op#
#include <log_softmax.hpp>

LogSoftmax operation.

Public Functions

LogSoftmax(const Output<Node> &arg, const int64_t axis)#

Constructs a LogSoftmax operation.

Output [d0, ...]

Parameters:
  • argNode that produces the first input tensor.[d0, ...]

  • axis – The axis position (0-based) on which to calculate the LogSoftmax.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class LogicalAnd : public ov::op::util::BinaryElementwiseLogical#
#include <logical_and.hpp>

Elementwise logical-and operation.

Public Functions

LogicalAnd() = default#

Constructs a logical-and operation.

LogicalAnd(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a logical-and operation.

Output [d0, ...]

Parameters:
  • arg0Output that produces the first input tensor.[d0, ...]

  • arg1Output that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class LogicalNot : public ov::op::Op#
#include <logical_not.hpp>

Elementwise logical negation operation.

Public Functions

LogicalNot() = default#

Constructs a logical negation operation.

LogicalNot(const Output<Node> &arg)#

Constructs a logical negation operation.

Parameters:

argNode that produces the input tensor.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class LogicalOr : public ov::op::util::BinaryElementwiseLogical#
#include <logical_or.hpp>

Elementwise logical-or operation.

Public Functions

LogicalOr(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a logical-or operation.

Output [d0, ...]

Parameters:
  • arg0Node that produces the first input tensor.[d0, ...]

  • arg1Node that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class LogicalXor : public ov::op::util::BinaryElementwiseLogical#
#include <logical_xor.hpp>

Elementwise logical-xor operation.

Public Functions

LogicalXor(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a logical-xor operation.

Output [d0, ...]

Parameters:
  • arg0Node that produces the first input tensor.[d0, ...]

  • arg1Node that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Loop : public ov::op::util::SubGraphOp#
#include <loop.hpp>

Iterate a body over tensors, accumulating into tensors.

Public Functions

Loop() = default#

Constructs a Loop operation.

Loop(const Output<Node> &trip_count, const Output<Node> &execution_condition)#

Constructs a Loop operation.

Parameters:
  • trip_countNode specifies the maximum number of iterations.

  • execution_conditionNode determines whether to execute the first iteration or not.

virtual Output<Node> get_concatenated_slices(const Output<Node> &value, int64_t start, int64_t stride, int64_t part_size, int64_t end, int64_t axis) override#

Concatenates slices from all iterations.

Parameters:
  • value – The value supplying slice values from each iteration.

  • start – First index on axis of the slicing

  • stride – Stepping of the slice

  • part_size – Size of the slice on axis

  • end – The last index on axis of the slicing

  • axis – The axis to slice along

Returns:

The concatenated slices.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

struct SpecialBodyPorts#
#include <loop.hpp>

Allows to define the purpose of inputs/outputs in the body.

class LRN : public ov::op::Op#
#include <lrn.hpp>

Elementwise Local Response Normalization (LRN) operation.

Inputs

Type

Description

arg

\(N[n, c, d_1,\dots,d_n]~(n \geq 0)\)

A tensor of any shape and numeric element type.

Type

Description

\(N[n, c, d_1,\dots,d_n]\)

The tensor \(T\), where \(T[n, c, d_1,\dots,d_n] = \frac{N[n,i,d_1,\dots,d_n]}{ (bias + alpha * (\sum_{i=max(0,(nsize-1)/2)}^{min(C, (nsize-1)/2)+1} N[n,i,d_1,\dots,d_n]^{2}) ^ {2})}\)

Public Functions

LRN() = default#

Constructs a LRN operation.

LRN(const Output<Node> &arg, double alpha, double beta, double bias, size_t size)#

Constructs a LRN operation.

Parameters:

argNode that produces the input tensor.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class LSTMCell : public ov::op::util::RNNCellBase#
#include <lstm_cell.hpp>

Class for single lstm cell node.

See also

LSTMSequence, RNNCell, GRUCell

Note

Following implementation supports:

Note

It calculates following equations:

        it = f(Xt*(Wi^T) + Ht-1*(Ri^T) + Pi (.) Ct-1 + Wbi + Rbi)
        ft = f(Xt*(Wf^T) + Ht-1*(Rf^T) + Pf (.) Ct-1 + Wbf + Rbf)
        ct = g(Xt*(Wc^T) + Ht-1*(Rc^T) + Wbc + Rbc)
        Ct = ft (.) Ct-1 + it (.) ct
        ot = f(Xt*(Wo^T) + Ht-1*(Ro^T) + Po (.) Ct + Wbo + Rbo)
        Ht = ot (.) h(Ct)

        *       - Is a dot product,
        (.)     - is a Hadamard product (element-wise),
        f, g, h - are activation functions.

Note

This class represents only single cell (for current time step) and not the whole LSTM Sequence layer

Public Functions

LSTMCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &initial_cell_state, const Output<Node> &W, const Output<Node> &R, std::size_t hidden_size, LSTMWeightsFormat weights_format = LSTMWeightsFormat::IFCO, const std::vector<std::string> &activations = std::vector<std::string>{"sigmoid", "tanh", "tanh"}, const std::vector<float> &activations_alpha = {}, const std::vector<float> &activations_beta = {}, float clip = 0.f, bool input_forget = false)#

Constructs LSTMCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • initial_cell_state[in] The cell state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The gate weights tensor with shape: [4*hidden_size, input_size].

  • R[in] The recurrence weights tensor with shape: [4*hidden_size, hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

  • weights_format[in] The order of gates in weights tensors. The default format is IFCO since it is used by DNNL.

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

  • input_forget[in] Controls coupling input and forget gates.

LSTMCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &initial_cell_state, const Output<Node> &W, const Output<Node> &R, const Output<Node> &B, std::size_t hidden_size, LSTMWeightsFormat weights_format = LSTMWeightsFormat::IFCO, const std::vector<std::string> &activations = std::vector<std::string>{"sigmoid", "tanh", "tanh"}, const std::vector<float> &activations_alpha = {}, const std::vector<float> &activations_beta = {}, float clip = 0.f, bool input_forget = false)#

Constructs LSTMCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • initial_cell_state[in] The cell state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The weight tensor with shape: [4*hidden_size, input_size].

  • R[in] The recurrence weight tensor with shape: [4*hidden_size, hidden_size].

  • B[in] The bias tensor for gates with shape: [4*hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

  • weights_format[in] The order of gates in weights tensors. The default format is IFCO since it is used by DNNL.

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

  • input_forget[in] Controls coupling input and forget gates.

LSTMCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &initial_cell_state, const Output<Node> &W, const Output<Node> &R, const Output<Node> &B, const Output<Node> &P, std::size_t hidden_size, LSTMWeightsFormat weights_format = LSTMWeightsFormat::IFCO, const std::vector<std::string> &activations = std::vector<std::string>{"sigmoid", "tanh", "tanh"}, const std::vector<float> &activations_alpha = {}, const std::vector<float> &activations_beta = {}, float clip = 0.f, bool input_forget = false)#

Constructs LSTMCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • initial_cell_state[in] The cell state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The weight tensor with shape: [4*hidden_size, input_size].

  • R[in] The recurrence weight tensor with shape: [4*hidden_size, hidden_size].

  • B[in] The bias tensor for gates with shape: [4*hidden_size].

  • P[in] The weight tensor for peepholes with shape: [3*hidden_size] - 3 equals to only iof gates. The order is: input, output, forget gates.

  • hidden_size[in] The number of hidden units for recurrent cell.

  • weights_format[in] The order of gates in weights tensors. The default format is IFCO since it is used by DNNL.

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

  • input_forget[in] Controls coupling input and forget gates.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class LSTMCell : public ov::op::util::RNNCellBase#
#include <lstm_cell.hpp>

Class for single lstm cell node.

See also

LSTMSequence, RNNCell, GRUCell

Note

Following implementation supports:

Note

It calculates following equations:

        it = f(Xt*(Wi^T) + Ht-1*(Ri^T) + Wbi + Rbi)
        ft = f(Xt*(Wf^T) + Ht-1*(Rf^T)  + Wbf + Rbf)
        ct = g(Xt*(Wc^T) + Ht-1*(Rc^T) + Wbc + Rbc)
        Ct = ft (.) Ct-1 + it (.) ct
        ot = f(Xt*(Wo^T) + Ht-1*(Ro^T) + Wbo + Rbo)
        Ht = ot (.) h(Ct)

        *       - Is a dot product,
        (.)     - is a Hadamard product (element-wise),
        f, g, h - are activation functions.

Note

This class represents only single cell (for current time step) and not the whole LSTM Sequence layer

Public Functions

LSTMCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &initial_cell_state, const Output<Node> &W, const Output<Node> &R, std::size_t hidden_size, const std::vector<std::string> &activations = std::vector<std::string>{"sigmoid", "tanh", "tanh"}, const std::vector<float> &activations_alpha = {}, const std::vector<float> &activations_beta = {}, float clip = 0.f)#

Constructs LSTMCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • initial_cell_state[in] The cell state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The gate weights tensor with shape: [4*hidden_size, input_size].

  • R[in] The recurrence weights tensor with shape: [4*hidden_size, hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

LSTMCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &initial_cell_state, const Output<Node> &W, const Output<Node> &R, const Output<Node> &B, std::size_t hidden_size, const std::vector<std::string> &activations = std::vector<std::string>{"sigmoid", "tanh", "tanh"}, const std::vector<float> &activations_alpha = {}, const std::vector<float> &activations_beta = {}, float clip = 0.f)#

Constructs LSTMCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • initial_cell_state[in] The cell state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The weight tensor with shape: [4*hidden_size, input_size].

  • R[in] The recurrence weight tensor with shape: [4*hidden_size, hidden_size].

  • B[in] The bias tensor for gates with shape: [4*hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class LSTMSequence : public ov::op::util::RNNCellBase#
#include <lstm_sequence.hpp>

Class for lstm sequence node.

See also

LSTMCell, RNNCell, GRUCell

Note

It follows notation and equations defined as in ONNX standard: onnx/onnx

Public Functions

inline virtual size_t get_default_output_index() const override#

Returns the output of the default output, or throws if there is none.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class LSTMSequence : public ov::op::util::RNNCellBase#
#include <lstm_sequence.hpp>

Class for lstm sequence node.

See also

LSTMCell, RNNCell, GRUCell

Note

It follows notation and equations defined as in ONNX standard: onnx/onnx

Public Functions

inline virtual size_t get_default_output_index() const override#

Returns the output of the default output, or throws if there is none.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class MatMul : public ov::op::Op#
#include <matmul.hpp>

Operator performing Matrix Multiplication.

Public Functions

MatMul(const Output<Node> &A, const Output<Node> &B, const bool &transpose_a = false, const bool &transpose_b = false)#

Constructs an Matrix Multiplication operation.

Parameters:
  • A – Matrix A

  • B – Matrix B

  • transpose_a – If matrix A should be transposed.

  • transpose_b – If matrix B should be transposed.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class MatrixNms : public ov::op::Op#
#include <matrix_nms.hpp>

MatrixNms operation.

Public Functions

MatrixNms() = default#

Constructs a conversion operation.

MatrixNms(const Output<Node> &boxes, const Output<Node> &scores, const Attributes &attrs)#

Constructs a MatrixNms operation.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • attrsAttributes of the operation

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline const Attributes &get_attrs() const#

Returns attributes of the operation MatrixNms.

struct Attributes#
#include <matrix_nms.hpp>

Structure that specifies attributes of the operation.

class MaxPool : public ov::op::util::MaxPoolBase#
#include <max_pool.hpp>

Batched max pooling operation.

Public Functions

MaxPool() = default#

Constructs a batched max pooling operation.

MaxPool(const Output<Node> &arg, const Strides &strides, const Shape &pads_begin, const Shape &pads_end, const Shape &kernel, const op::RoundingType rounding_type = op::RoundingType::FLOOR, const PadType auto_pad = op::PadType::EXPLICIT)#

Constructs a batched max pooling operation.

Parameters:
  • arg – The node producing the input data batch tensor.

  • strides – The strides.

  • pads_begin – The beginning of padding shape.

  • pads_end – The end of padding shape.

  • kernel – The kernel shape.

  • rounding_type – Whether to use ceiling or floor rounding type while computing output shape.

  • auto_pad – The pad type for automatically computing padding sizes.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class MaxPool : public ov::op::util::MaxPoolBase#
#include <max_pool.hpp>

MaxPooling operation with values and indices calculated as individual outputs.

Public Functions

MaxPool() = default#

Constructs an empty MaxPool operation.

MaxPool(const Output<Node> &arg, const Strides &strides, const Strides &dilations, const Shape &pads_begin, const Shape &pads_end, const Shape &kernel, const op::RoundingType rounding_type = op::RoundingType::FLOOR, const PadType auto_pad = op::PadType::EXPLICIT, const element::Type index_element_type = element::i64, const int64_t axis = 0)#

Constructs a parametrized MaxPool operation.

Parameters:
  • argOutput of a node producing the feature tensor to be pooled.

  • strides – The strides of the pooling filter.

  • dilations – The dilations of the pooling filter.

  • pads_begin – Paddings at the beginning of each spatial axis.

  • pads_end – Paddings at the end of each spatial axis.

  • kernel – The kernel shape.

  • rounding_type – Whether to use ceiling or floor rounding type while computing the output shape.

  • auto_pad – The pad type for automatic calculation of the padding sizes.

  • index_element_type – The data type used by the second output tensor containing the selected indices.

  • axis – Indicates a dimension in the input data shape which should be used as a starting point for calculation of the upper bound of allowed values of the indices output.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

const Strides &get_dilations() const noexcept#
Returns:

The pooling filter’s dilations.

element::Type get_index_element_type() const noexcept#
Returns:

The data type of the second output tensor (indices).

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class MaxPool : public ov::op::util::MaxPoolBase#
#include <max_pool.hpp>

MaxPooling operation with values and indices calculated as individual outputs.

Public Functions

MaxPool() = default#

Constructs an empty MaxPool operation.

MaxPool(const Output<Node> &arg, const Strides &strides, const Strides &dilations, const Shape &pads_begin, const Shape &pads_end, const Shape &kernel, const op::RoundingType rounding_type = op::RoundingType::FLOOR, const PadType auto_pad = op::PadType::EXPLICIT, const element::Type index_element_type = element::i64, const int64_t axis = 0)#

Constructs a parametrized MaxPool operation.

Parameters:
  • argOutput of a node producing the feature tensor to be pooled.

  • strides – The strides of the pooling filter.

  • dilations – The dilations of the pooling filter.

  • pads_begin – Paddings at the beginning of each spatial axis.

  • pads_end – Paddings at the end of each spatial axis.

  • kernel – The kernel shape.

  • rounding_type – Whether to use ceiling or floor rounding type while computing the output shape.

  • auto_pad – The pad type for automatic calculation of the padding sizes.

  • index_element_type – The data type used by the second output tensor containing the selected indices.

  • axis – Indicates a dimension in the input data shape which should be used as a starting point for calculation of the upper bound of allowed values of the indices output.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

const Strides &get_dilations() const noexcept#
Returns:

The pooling filter’s dilations.

element::Type get_index_element_type() const noexcept#
Returns:

The data type of the second output tensor (indices).

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Maximum : public ov::op::util::BinaryElementwiseArithmetic#
#include <maximum.hpp>

Elementwise maximum operation.

Public Functions

inline Maximum()#

Constructs a maximum operation.

Maximum(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a maximum operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Minimum : public ov::op::util::BinaryElementwiseArithmetic#
#include <minimum.hpp>

Elementwise minimum operation.

Public Functions

inline Minimum()#

Constructs a minimum operation.

Minimum(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a minimum operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Mish : public ov::op::util::UnaryElementwiseArithmetic#
#include <mish.hpp>

A Self Regularized Non-Monotonic Neural Activation Function f(x) = x * tanh(log(exp(x) + 1.))

Public Functions

Mish(const Output<Node> &arg)#

Constructs an Mish operation.

Parameters:

dataInput tensor

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Mod : public ov::op::util::BinaryElementwiseArithmetic#
#include <mod.hpp>

Mod returns an element-wise division reminder with two given tensors applying multi-directional broadcast rules.

Public Functions

inline Mod()#

Constructs a Mod node.

Mod(const Output<Node> &A, const Output<Node> &B, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#
Parameters:
  • A – - Dividend tensor

  • B – - Divisor tensor

  • auto_broadcast – Auto broadcast specification

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Multinomial : public ov::op::Op#
#include <multinomial.hpp>

Multinomial operation creates a sequence of indices of classes sampled from the multinomial distribution.

Public Functions

Multinomial(const Output<Node> &input, const Output<Node> &num_samples, const ov::element::Type_t convert_type, const bool with_replacement, const bool log_probs, const uint64_t global_seed = 0, const uint64_t op_seed = 0)#

Multinomial operation creates a sequence of indices of classes sampled from the multinomial distribution.

Parameters:
  • probsInput tensor containing at each index poisition probability/log probability of sampling a given class. Any floating-point precision values are allowed.

  • num_samples – Scalar or 1D tensor with a single value that determines the number of samples to generate per batch. Values should be of an integer type.

  • convert_type – Data type to which to convert the output class indices. Allowed values: i32/i64

  • with_replacement – Boolean that determines whether a sampled class can appear more than once in the output.

  • log_probs – Boolean that determines whether to treat input probabilities as log probabilities.

  • global_seed – First seed value (key) of Philox random number generation algorithm. (See RandomUniform for details)

  • op_seed – Second seed value (counter) of Philox random number generation algorithm. (See RandomUniform for details)

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Multiply : public ov::op::util::BinaryElementwiseArithmetic#
#include <multiply.hpp>

Elementwise multiplication operation.

Public Functions

inline Multiply()#

Constructs a multiplication operation.

Multiply(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a multiplication operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class MVN : public ov::op::Op#
#include <mvn.hpp>

Operator performing Mean Variance Normalization.

Public Functions

MVN(const Output<Node> &data, bool across_channels = true, bool normalize_variance = true, double eps = 1e-9)#

Constructs an MVN operation.

Parameters:
  • dataInput tensor with data

  • normalize_variance – flag that denotes whether to perform variance normalization.

  • across_channels – flag that denotes if mean values are shared across channels.

  • eps – the number to be added to the variance to avoid division by zero when normalizing the value

MVN(const Output<Node> &data, AxisSet reduction_axes, bool normalize_variance = true, double eps = 1e-9)#

Constructs an MVN operation.

Parameters:
  • dataInput tensor with data

  • reduction_axes – A list of axes, along which to reduce.

  • normalize_variance – flag that denotes whether to perform variance normalization.

  • eps – the number to be added to the variance to avoid division by zero when normalizing the value

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class MVN : public ov::op::Op#
#include <mvn.hpp>

Operator performing Mean Variance Normalization.

Public Functions

MVN(const Output<Node> &data, const Output<Node> &reduction_axes, bool normalize_variance, float eps, MVNEpsMode eps_mode)#

Constructs an MVN operation.

Parameters:
  • dataInput tensor with data

  • reduction_axes – A list of axes, along which to reduce.

  • normalize_variance – flag that denotes whether to perform variance normalization.

  • eps – the number to be added to the variance to avoid division by zero when normalizing the value

  • eps_mode – the mode of applying epsilon

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool evaluate(ov::TensorVector &output_values, const ov::TensorVector &input_values) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Negative : public ov::op::util::UnaryElementwiseArithmetic#
#include <negative.hpp>

Elementwise negative operation.

Public Functions

Negative() = default#

Constructs a negative operation.

Negative(const Output<Node> &arg)#

Constructs a negative operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class NonMaxSuppression : public ov::op::Op#
#include <non_max_suppression.hpp>

Elementwise addition operation.

Public Functions

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const Output<Node> &max_output_boxes_per_class, const Output<Node> &iou_threshold, const Output<Node> &score_threshold, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true)#

Constructs a NonMaxSuppression operation.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • max_output_boxes_per_classNode producing maximum number of boxes to be selected per class

  • iou_thresholdNode producing intersection over union threshold

  • score_thresholdNode producing minimum score threshold

  • box_encoding – Specifies the format of boxes data encoding

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true)#

Constructs a NonMaxSuppression operation with default values for the last 3 inputs.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box coordinates

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class NonMaxSuppression : public ov::op::Op#
#include <non_max_suppression.hpp>

NonMaxSuppression operation.

Subclassed by ov::op::v4::NonMaxSuppression

Public Functions

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const Output<Node> &max_output_boxes_per_class, const Output<Node> &iou_threshold, const Output<Node> &score_threshold, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • max_output_boxes_per_classNode producing maximum number of boxes to be selected per class

  • iou_thresholdNode producing intersection over union threshold

  • score_thresholdNode producing minimum score threshold

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation with default values for the last 3 inputs.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box coordinates

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class NonMaxSuppression : public ov::op::v3::NonMaxSuppression#
#include <non_max_suppression.hpp>

NonMaxSuppression operation.

Public Functions

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const Output<Node> &max_output_boxes_per_class, const Output<Node> &iou_threshold, const Output<Node> &score_threshold, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • max_output_boxes_per_classNode producing maximum number of boxes to be selected per class

  • iou_thresholdNode producing intersection over union threshold

  • score_thresholdNode producing minimum score threshold

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation with default values for the last 3 inputs.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box coordinates

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class NonMaxSuppression : public ov::op::Op#
#include <non_max_suppression.hpp>

NonMaxSuppression operation.

Public Functions

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation with default values in the last 4 inputs.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const Output<Node> &max_output_boxes_per_class, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation with default values in the last. 3 inputs.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • max_output_boxes_per_classNode producing maximum number of boxes to be selected per class

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const Output<Node> &max_output_boxes_per_class, const Output<Node> &iou_threshold, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation with default values in the last. 2 inputs.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • max_output_boxes_per_classNode producing maximum number of boxes to be selected per class

  • iou_thresholdNode producing intersection over union threshold

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const Output<Node> &max_output_boxes_per_class, const Output<Node> &iou_threshold, const Output<Node> &score_threshold, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation with default value in the last. input.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • max_output_boxes_per_classNode producing maximum number of boxes to be selected per class

  • iou_thresholdNode producing intersection over union threshold

  • score_thresholdNode producing minimum score threshold

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

NonMaxSuppression(const Output<Node> &boxes, const Output<Node> &scores, const Output<Node> &max_output_boxes_per_class, const Output<Node> &iou_threshold, const Output<Node> &score_threshold, const Output<Node> &soft_nms_sigma, const BoxEncodingType box_encoding = BoxEncodingType::CORNER, const bool sort_result_descending = true, const ov::element::Type &output_type = ov::element::i64)#

Constructs a NonMaxSuppression operation.

Parameters:
  • boxesNode producing the box coordinates

  • scoresNode producing the box scores

  • max_output_boxes_per_classNode producing maximum number of boxes to be selected per class

  • iou_thresholdNode producing intersection over union threshold

  • score_thresholdNode producing minimum score threshold

  • soft_nms_sigmaNode specifying the sigma parameter for Soft-NMS

  • box_encoding – Specifies the format of boxes data encoding

  • sort_result_descending – Specifies whether it is necessary to sort selected boxes across batches

  • output_type – Specifies the output tensor type

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class NonZero : public ov::op::Op#
#include <non_zero.hpp>

NonZero operation returning indices of non-zero elements in the input tensor.

Note

The indices are returned by-dimension in row-major order. For example the following output contains 3 indices of a 3D input tensor elements: [[0, 0, 2], [0, 1, 1], [0, 1, 2]] The values point to input elements at [0,0,0], [0,1,1] and [2,1,2]

Public Functions

NonZero() = default#

Constructs a NonZero operation.

NonZero(const Output<Node> &arg)#

Constructs a NonZero operation.

Note

The output type is int64.

Parameters:

argNode that produces the input tensor.

NonZero(const Output<Node> &arg, const std::string &output_type)#

Constructs a NonZero operation.

Parameters:
  • argNode that produces the input tensor.

  • output_type – produce indices. Currently, only ‘int64’ or ‘int32’ are supported

NonZero(const Output<Node> &arg, const element::Type &output_type)#

Constructs a NonZero operation.

Parameters:
  • argNode that produces the input tensor.

  • output_type – produce indices. Currently, only int64 or int32 are supported

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class NormalizeL2 : public ov::op::Op#
#include <normalize_l2.hpp>

Normalization with L2 norm.

Public Functions

NormalizeL2(const Output<Node> &data, const Output<Node> &axes, float eps, EpsMode eps_mode)#

Constructs a NormalizeL2 operation.

Parameters:
  • data – - Node producing the input tensor

  • axes – - Node indicating axes along which reduction is calculated

  • eps – - The epsilon added to L2 norm.

  • eps_mode – - Specifies how eps is combined with L2 value calculated before division

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class NotEqual : public ov::op::util::BinaryElementwiseComparison#
#include <not_equal.hpp>

Elementwise not-equal operation.

Public Functions

inline NotEqual()#

Constructs a not-equal operation.

NotEqual(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a not-equal operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class NV12toBGR : public ov::op::util::ConvertColorNV12Base#
#include <nv12_to_bgr.hpp>

Color conversion operation from NV12 to RGB format. Input:

  • Input NV12 image can be represented in two ways: a) Single plane (as it is in the file): NV12 height dimension is 1.5x bigger than image height. ‘C’ dimension shall be 1. b) Two separate planes (used this way in many physical video sources): Y and UV. In this case b1) Y plane has height same as image height. ‘C’ dimension equals to 1 b2) UV plane has dimensions: ‘H’ = image_h / 2; ‘W’ = image_w / 2; ‘C’ = 2.

  • Supported element types: u8 or any supported floating-point type. Output:

  • Output node will have NHWC layout and shape HxW same as image spatial dimensions.

  • Number of output channels ‘C’ will be 3, as per interleaved RGB format, first channel is B, last is R

    Conversion of each pixel from NV12 (YUV) to RGB space is represented by following formulas: R = 1.164 * (Y - 16) + 1.596 * (V - 128) G = 1.164 * (Y - 16) - 0.813 * (V - 128) - 0.391 * (U - 128) B = 1.164 * (Y - 16) + 2.018 * (U - 128) Then R, G, B values are clipped to range (0, 255)

Public Functions

explicit NV12toBGR(const Output<Node> &arg)#

Constructs a conversion operation from input image in NV12 format As per NV12 format definition, node height dimension shall be 1.5 times bigger than image height so that image (w=640, h=480) is represented by NHWC shape {N,720,640,1} (height*1.5 x width)

Parameters:

argNode that produces the input tensor. Input tensor represents image in NV12 format (YUV).

explicit NV12toBGR(const Output<Node> &arg_y, const Output<Node> &arg_uv)#

Constructs a conversion operation from 2-plane input image in NV12 format In general case Y channel of image can be separated from UV channel which means that operation needs two nodes for Y and UV planes respectively. Y plane has one channel, and UV has 2 channels, both expect ‘NHWC’ layout.

Parameters:
  • arg_yNode that produces the input tensor for Y plane (NHWC layout). Shall have WxH dimensions equal to image dimensions. ‘C’ dimension equals to 1.

  • arg_uvNode that produces the input tensor for UV plane (NHWC layout). ‘H’ is half of image height, ‘W’ is half of image width, ‘C’ dimension equals to 2. Channel 0 represents ‘U’, channel 1 represents ‘V’ channel

class NV12toRGB : public ov::op::util::ConvertColorNV12Base#
#include <nv12_to_rgb.hpp>

Color conversion operation from NV12 to RGB format. Input:

  • Input NV12 image can be represented in two ways: a) Single plane (as it is in the file): NV12 height dimension is 1.5x bigger than image height. ‘C’ dimension shall be 1. b) Two separate planes (used this way in many physical video sources): Y and UV. In this case b1) Y plane has height same as image height. ‘C’ dimension equals to 1 b2) UV plane has dimensions: ‘H’ = image_h / 2; ‘W’ = image_w / 2; ‘C’ = 2.

  • Supported element types: u8 or any supported floating-point type. Output:

  • Output node will have NHWC layout and shape HxW same as image spatial dimensions.

  • Number of output channels ‘C’ will be 3, as per interleaved RGB format, first channel is R, last is B

    Conversion of each pixel from NV12 (YUV) to RGB space is represented by following formulas: R = 1.164 * (Y - 16) + 1.596 * (V - 128) G = 1.164 * (Y - 16) - 0.813 * (V - 128) - 0.391 * (U - 128) B = 1.164 * (Y - 16) + 2.018 * (U - 128) Then R, G, B values are clipped to range (0, 255)

Public Functions

explicit NV12toRGB(const Output<Node> &arg)#

Constructs a conversion operation from input image in NV12 format As per NV12 format definition, node height dimension shall be 1.5 times bigger than image height so that image (w=640, h=480) is represented by NHWC shape {N,720,640,1} (height*1.5 x width)

Parameters:

argNode that produces the input tensor. Input tensor represents image in NV12 format (YUV).

NV12toRGB(const Output<Node> &arg_y, const Output<Node> &arg_uv)#

Constructs a conversion operation from 2-plane input image in NV12 format In general case Y channel of image can be separated from UV channel which means that operation needs two nodes for Y and UV planes respectively. Y plane has one channel, and UV has 2 channels, both expect ‘NHWC’ layout.

Parameters:
  • arg_yNode that produces the input tensor for Y plane (NHWC layout). Shall have WxH dimensions equal to image dimensions. ‘C’ dimension equals to 1.

  • arg_uvNode that produces the input tensor for UV plane (NHWC layout). ‘H’ is half of image height, ‘W’ is half of image width, ‘C’ dimension equals to 2. Channel 0 represents ‘U’, channel 1 represents ‘V’ channel

class OneHot : public ov::op::Op#
#include <one_hot.hpp>

OneHot operation.

Public Functions

OneHot() = default#

Constructs a one-hot operation.

OneHot(const Output<Node> &indices, const Output<Node> &depth, const Output<Node> &on_value, const Output<Node> &off_value, int64_t axis)#

Constructs a one-hot operation.

Parameters:
  • indicesInput tensor containing indices.

  • depth – Specifies number of classes and the size of one-hot dimension.

  • on_value – Specifies value that the locations in output tensor represented by indices in input take.

  • off_value – Specifies value that the locations in output tensor not represented by indices in input take.

  • axis – Axis along which one-hot representation in added.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

inline const int64_t &get_axis() const#
Returns:

The index of the one-hot axis.

class Op : public ov::Node#
#include <op.hpp>

Root of all actual ops.

Subclassed by ov::exec_model_info::ExecutionNode, ov::op::Sink, ov::op::internal::DynamicQuantize, ov::op::internal::NonMaxSuppressionIEInternal, ov::op::internal::RMS, ov::op::internal::RoPE, ov::op::util::AvgPoolBase, ov::op::util::BinaryElementwiseArithmetic, ov::op::util::BinaryElementwiseBitwise, ov::op::util::BinaryElementwiseComparison, ov::op::util::BinaryElementwiseLogical, ov::op::util::BroadcastBase, ov::op::util::ConvertColorI420Base, ov::op::util::ConvertColorNV12Base, ov::op::util::ConvolutionBase, ov::op::util::DetectionOutputBase, ov::op::util::EmbeddingBagOffsetsBase, ov::op::util::EmbeddingBagPackedBase, ov::op::util::FFTBase, ov::op::util::GatherBase, ov::op::util::GatherNDBase, ov::op::util::IndexReduction, ov::op::util::InterpolateBase, ov::op::util::MaxPoolBase, ov::op::util::MulticlassNmsBase, ov::op::util::PadBase, ov::op::util::RNNCellBase, ov::op::util::ROIAlignBase, ov::op::util::ReadValueBase, ov::op::util::ReductionBase, ov::op::util::ScatterBase, ov::op::util::ScatterElementsUpdateBase, ov::op::util::ScatterNDBase, ov::op::util::ShapeOfBase, ov::op::util::TopKBase, ov::op::util::UnaryElementwiseArithmetic, ov::op::v0::BatchNormInference, ov::op::v0::CTCGreedyDecoder, ov::op::v0::Concat, ov::op::v0::Constant, ov::op::v0::Convert, ov::op::v0::CumSum, ov::op::v0::DepthToSpace, ov::op::v0::FakeQuantize, ov::op::v0::HardSigmoid, ov::op::v0::Interpolate, ov::op::v0::LRN, ov::op::v0::MVN, ov::op::v0::MatMul, ov::op::v0::NormalizeL2, ov::op::v0::PRelu, ov::op::v0::PSROIPooling, ov::op::v0::Parameter, ov::op::v0::PriorBox, ov::op::v0::PriorBoxClustered, ov::op::v0::Proposal, ov::op::v0::ROIPooling, ov::op::v0::Range, ov::op::v0::RegionYolo, ov::op::v0::ReorgYolo, ov::op::v0::Result, ov::op::v0::ReverseSequence, ov::op::v0::Selu, ov::op::v0::ShuffleChannels, ov::op::v0::SpaceToDepth, ov::op::v0::Squeeze, ov::op::v0::Tile, ov::op::v0::Unsqueeze, ov::op::v10::IsFinite, ov::op::v10::IsInf, ov::op::v10::IsNaN, ov::op::v10::Unique, ov::op::v12::GroupNormalization, ov::op::v13::BitwiseNot, ov::op::v13::FakeConvert, ov::op::v13::Multinomial, ov::op::v13::NMSRotated, ov::op::v13::ScaledDotProductAttention, ov::op::v14::ConvertPromoteTypes, ov::op::v14::Inverse, ov::op::v15::Col2Im, ov::op::v15::SliceScatter, ov::op::v15::StringTensorPack, ov::op::v15::StringTensorUnpack, ov::op::v1::BatchToSpace, ov::op::v1::ConvertLike, ov::op::v1::DeformablePSROIPooling, ov::op::v1::GatherTree, ov::op::v1::LogicalNot, ov::op::v1::NonMaxSuppression, ov::op::v1::OneHot, ov::op::v1::Reshape, ov::op::v1::Reverse, ov::op::v1::Select, ov::op::v1::Softmax, ov::op::v1::SpaceToBatch, ov::op::v1::Split, ov::op::v1::StridedSlice, ov::op::v1::Transpose, ov::op::v1::VariadicSplit, ov::op::v3::Bucketize, ov::op::v3::EmbeddingSegmentsSum, ov::op::v3::ExtractImagePatches, ov::op::v3::NonMaxSuppression, ov::op::v3::NonZero, ov::op::v4::CTCLoss, ov::op::v4::Range, ov::op::v4::Swish, ov::op::v5::BatchNormInference, ov::op::v5::LogSoftmax, ov::op::v5::NonMaxSuppression, ov::op::v6::CTCGreedyDecoderSeqLen, ov::op::v6::ExperimentalDetectronDetectionOutput, ov::op::v6::ExperimentalDetectronGenerateProposalsSingleImage, ov::op::v6::ExperimentalDetectronPriorGridGenerator, ov::op::v6::ExperimentalDetectronROIFeatureExtractor, ov::op::v6::ExperimentalDetectronTopKROIs, ov::op::v6::GatherElements, ov::op::v6::MVN, ov::op::v7::Einsum, ov::op::v7::Roll, ov::op::v8::AdaptiveAvgPool, ov::op::v8::AdaptiveMaxPool, ov::op::v8::MatrixNms, ov::op::v8::PriorBox, ov::op::v8::RandomUniform, ov::op::v8::Slice, ov::op::v8::Softmax, ov::op::v9::Eye, ov::op::v9::GenerateProposals, ov::op::v9::GridSample, ov::op::v9::NonMaxSuppression

Public Functions

inline virtual const ::ov::Node::type_info_t &get_type_info() const override#

Returns the NodeTypeInfo for the node’s class. During transition to type_info, returns a dummy type_info for Node if the class has not been updated yet.

class Pad : public ov::op::util::PadBase#
#include <pad.hpp>

Generic padding operation.

Public Functions

Pad() = default#

Constructs a Pad-1 operation.

Pad(const Output<Node> &arg, const Output<Node> &pads_begin, const Output<Node> &pads_end, const Output<Node> &arg_pad_value, PadMode pad_mode)#

Constructs a Pad-1 operation.

Parameters:
  • arg – The output producing input tensor to be padded.

  • pads_begin – The output which specifies the number of padding elements added before position 0 on each axis of arg.

  • pads_end – The output which specifies the number of padding elements after the last element on each axis.

  • arg_pad_value – The scalar output with the value used for padding if pad_mode is CONSTANT

  • pad_mode – The padding mode: CONSTANT, EDGE, REFLECT or SYMMETRIC. CONSTANT initializes new elements with arg_pad_value, EDGE uses the nearest value from arg. REFLECT and SYMMETRIC tile the background by flipping arg at the edge (SYMMETRIC) or on the last row/column/etc. (REFLECT).

Pad(const Output<Node> &arg, const Output<Node> &pads_begin, const Output<Node> &pads_end, PadMode pad_mode)#

Constructs a Pad-1 operation.

Parameters:
  • arg – The output producing input tensor to be padded.

  • pads_begin – The output which specifies the number of padding elements added

  • pads_end – The output which specifies the number of padding elements after the last element on each axis.

  • pad_mode – The padding mode: CONSTANT, EDGE, REFLECT or SYMMETRIC.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

class Pad : public ov::op::util::PadBase#
#include <pad.hpp>

Generic padding operation.

Public Functions

Pad() = default#

Constructs a Pad-12 operation.

Pad(const Output<Node> &arg, const Output<Node> &pads_begin, const Output<Node> &pads_end, PadMode pad_mode)#

Constructs a Pad-12 operation.

Parameters:
  • arg – The output producing input tensor to be padded.

  • pads_begin – The output which specifies the number of padding elements to add (or remove) before position 0 on each axis of arg.

  • pads_end – The output which specifies the number of padding elements to add (or remove) after the last element on each axis.

  • pad_mode – The padding mode: CONSTANT, EDGE, REFLECT or SYMMETRIC.

Pad(const Output<Node> &arg, const Output<Node> &pads_begin, const Output<Node> &pads_end, const Output<Node> &arg_pad_value, PadMode pad_mode)#

Constructs a Pad-12 operation.

Parameters:
  • arg – The output producing input tensor to be padded.

  • pads_begin – The output which specifies the number of padding elements to add (or remove) before position 0 on each axis of arg.

  • pads_end – The output which specifies the number of padding elements to add (or remove) after the last element on each axis.

  • arg_pad_value – The scalar output with the value used for padding if pad_mode is CONSTANT

  • pad_mode – The padding mode: CONSTANT, EDGE, REFLECT or SYMMETRIC.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

class Parameter : public ov::op::Op#
#include <parameter.hpp>

A model parameter.

Parameters are nodes that represent the arguments that will be passed to user-defined models. Model creation requires a sequence of parameters. Basic graph operations do not need parameters attached to a model.

Public Functions

Parameter() = default#

Constructions a tensor-typed parameter node.

Parameter(const ov::element::Type &element_type, const PartialShape &pshape)#

Constructions a tensor-typed parameter node.

Parameters:
  • element_type – The element type of the parameter.

  • pshape – The partial shape of the parameter.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

Layout get_layout() const#

Returns current layout, or empty Layout if it is not set.

void set_layout(const Layout &layout)#

Sets layout runtime information to tensor.

Parameters:

layoutLayout to set. If empty (default constructed), layout runtime information is erased.

class Power : public ov::op::util::BinaryElementwiseArithmetic#
#include <power.hpp>

Elementwise exponentiation operation.

Inputs

Type

Description

arg0

\(N[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of any shape and numeric element type.

arg1

\(N[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of the same shape and element type as arg0.

Type

Description

\(N[d_1,\dots,d_n]\)

The tensor \(T\), where \(T[i_1,\dots,i_n] = \texttt{arg0}[i_1,\dots,i_n]^{\texttt{arg1}[i_1,\dots,i_n]}\)

Public Functions

Power(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs an exponentiation operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class PRelu : public ov::op::Op#
#include <prelu.hpp>

Parametrized Relu x < 0 => f(x) = x * slope x >= 0 => f(x) = x.

Public Functions

PRelu(const Output<Node> &data, const Output<Node> &slope)#

Constructs a PRelu operation.

Parameters:
  • dataInput tensor

  • slope – Multipliers for negative values

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class PriorBox : public ov::op::Op#
#include <prior_box.hpp>

Layer which generates prior boxes of specified sizes normalized to input image size.

Public Functions

PriorBox(const Output<Node> &layer_shape, const Output<Node> &image_shape, const Attributes &attrs)#

Constructs a PriorBox operation.

Parameters:
  • layer_shapeShape of layer for which prior boxes are computed

  • image_shapeShape of image to which prior boxes are scaled

  • attrsPriorBox attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

struct Attributes#
#include <prior_box.hpp>
class PriorBox : public ov::op::Op#
#include <prior_box.hpp>

Layer which generates prior boxes of specified sizes normalized to input image size.

Public Functions

PriorBox(const Output<Node> &layer_shape, const Output<Node> &image_shape, const Attributes &attrs)#

Constructs a PriorBox operation.

Parameters:
  • layer_shapeShape of layer for which prior boxes are computed

  • image_shapeShape of image to which prior boxes are scaled

  • attrsPriorBox attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

struct Attributes#
#include <prior_box.hpp>
class PriorBoxClustered : public ov::op::Op#
#include <prior_box_clustered.hpp>

Layer which generates prior boxes of specified sizes normalized to input image size.

Public Functions

PriorBoxClustered(const Output<Node> &layer_shape, const Output<Node> &image_shape, const Attributes &attrs)#

Constructs a PriorBoxClustered operation.

Parameters:
  • layer_shapeShape of layer for which prior boxes are computed

  • image_shapeShape of image to which prior boxes are scaled

  • attrsPriorBoxClustered attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

struct Attributes#
#include <prior_box_clustered.hpp>
class Proposal : public ov::op::Op#
#include <proposal.hpp>

Proposal operation.

Subclassed by ov::op::v4::Proposal

Public Functions

Proposal(const Output<Node> &class_probs, const Output<Node> &bbox_deltas, const Output<Node> &image_shape, const Attributes &attrs)#

Constructs a Proposal operation.

Parameters:
  • class_probs – Class probability scores

  • bbox_deltas – Prediction of bounding box deltas

  • image_shapeShape of image

  • attrsProposal op attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

void set_attrs(Attributes attrs)#

Set the Proposal operator attributes.

Parameters:

attrsAttributes to be set.

struct Attributes#
#include <proposal.hpp>
class Proposal : public ov::op::v0::Proposal#
#include <proposal.hpp>

Proposal operation.

Public Functions

Proposal(const Output<Node> &class_probs, const Output<Node> &bbox_deltas, const Output<Node> &image_shape, const Attributes &attrs)#

Constructs a Proposal operation.

Parameters:
  • class_probs – Class probability scores

  • bbox_deltas – Prediction of bounding box deltas

  • image_shapeShape of image

  • attrsProposal op attributes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class PSROIPooling : public ov::op::Op#
#include <psroi_pooling.hpp>

PSROIPooling operation.

Public Functions

PSROIPooling(const Output<Node> &input, const Output<Node> &coords, const size_t output_dim, const size_t group_size, const float spatial_scale, int spatial_bins_x, int spatial_bins_y, const std::string &mode)#

Constructs a PSROIPooling operation.

Parameters:
  • inputInput feature map {N, C, …}

  • coords – Coordinates of bounding boxes

  • output_dimOutput channel number

  • group_size – Number of groups to encode position-sensitive scores

  • spatial_scale – Ratio of input feature map over input image size

  • spatial_bins_x – Numbers of bins to divide the input feature maps over width

  • spatial_bins_y – Numbers of bins to divide the input feature maps over height

  • mode – Mode of pooling - Avg or Bilinear

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

void set_output_dim(size_t output_dim)#

Set the output channel dimension size.

Parameters:

output_dim – Channel dimension size.

void set_group_size(size_t group_size)#

Set the output groups number.

Parameters:

group_size – Number of groups.

void set_spatial_scale(float scale)#

Set the spatial scale.

Parameters:

scale – Spatial scale value.

void set_spatial_bins_x(int x)#

Set the number of bins over image width.

Parameters:

x – Number of bins over width (x) axis.

void set_spatial_bins_y(int y)#

Set the number of bins over image height.

Parameters:

y – Number of bins over height (y) axis.

void set_mode(std::string mode)#

Set the pooling mode.

Parameters:

mode – Pooling mode name.

class RandomUniform : public ov::op::Op#
#include <random_uniform.hpp>

Tensor RandomUniform operation.

Public Functions

RandomUniform(const Output<Node> &out_shape, const Output<Node> &min_val, const Output<Node> &max_val, const ov::element::Type &out_type, uint64_t global_seed = 0, uint64_t op_seed = 0, ov::op::PhiloxAlignment alignment = ov::op::PhiloxAlignment::TENSORFLOW)#

Constructs a RandomUniform operation.

Parameters:
  • out_shapeNode producing the tensor with output shape.

  • min_valNode producing the tensor with minimum value.

  • max_valNode producing the tensor with maximum value.

  • out_typeOutput type of the tensor.

  • global_seed – Global seed value.

  • op_seed – Operational seed value.

  • alignment – Alignment of numbers generated by Philox algorithm based on provided seed.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool constant_fold(OutputVector &output_values, const OutputVector &inputs_values) override#
Returns:

Turns off constant folding for RandomUniform operation.

const ov::element::Type &get_out_type() const#
Returns:

The output tensor type.

uint64_t get_global_seed() const#
Returns:

The global seed value.

uint64_t get_op_seed() const#
Returns:

The operational seed value.

std::pair<uint64_t, uint64_t> get_state() const#
Returns:

The state value.

ov::op::PhiloxAlignment get_alignment() const#
Returns:

The alignment mode.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Range : public ov::op::Op#
#include <range.hpp>

Range operation, analogous to arange() in Numpy.

Public Functions

Range() = default#

Constructs an unitialized range operation.

Range(const Output<Node> &start, const Output<Node> &stop, const Output<Node> &step, element::Type output_type)#

Constructs a range operation.

Parameters:
  • start – The tensor producing the start value. Must be a scalar of numeric element type.

  • stop – The tensor producing the stop value. Must be a scalar of numeric element type.

  • step – The tensor producing the step value. Must be a scalar of numeric element type.

  • output_type – The type of the output.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Range : public ov::op::Op#
#include <range.hpp>

Range operation, analogous to range() in Python.

Public Functions

Range() = default#

Constructs an unitialized range operation.

Range(const Output<Node> &start, const Output<Node> &stop, const Output<Node> &step)#

Constructs a range operation.

Parameters:
  • start – The tensor producing the start value. Must be a scalar of integer element type, and same element type as stop and step.

  • stop – The tensor producing the stop value. Must be a scalar of integer element type, and same element type as start and step.

  • step – The tensor producing the step value. Must be a scalar of integer element type, and same element type as start and stop.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReadValue : public ov::op::util::ReadValueBase#
#include <read_value.hpp>

ReadValue operation creates the variable with variable_id and returns value of this variable.

Public Functions

ReadValue(const Output<Node> &init_value, const std::string &variable_id)#

Constructs a ReadValue operation.

Parameters:
  • init_valueNode that produces the input tensor.

  • variable_id – identificator of the variable to create.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline virtual std::string get_variable_id() const override#

Returns the identifier of corresponding variable.

class ReadValue : public ov::op::util::ReadValueBase#
#include <read_value.hpp>

ReadValue operation gets an input value from the variable with variable_id and returns it as an output.

Public Functions

explicit ReadValue(const std::shared_ptr<util::Variable> &variable)#

Constructs a ReadValue operation.

Parameters:

variable – Class for storing and synchronizing element types, shapes and identifiers between pairs of Assign/ReadValue nodes.

ReadValue(const Output<Node> &init_value, const std::shared_ptr<util::Variable> &variable)#

Constructs a ReadValue operation.

Parameters:
  • init_valueNode that produces the input tensor.

  • variable – Class for storing and synchronizing element types, shapes and identifiers between pairs of Assign/ReadValue nodes.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline virtual std::string get_variable_id() const override#

Returns the identifier of corresponding variable.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceL1 : public ov::op::util::ArithmeticReductionKeepDims#
#include <reduce_l1.hpp>

Reduction operation using L1 norm: L1(x) = sum(abs(x)) if all dimensions are specified for the normalisation.

Reduces the tensor, eliminating the specified reduction axes by taking the L1-norm.

Public Functions

ReduceL1() = default#

Constructs a reducet L1-norm operation.

ReduceL1(const Output<Node> &arg, const Output<Node> &reduction_axes, bool keep_dims = false)#

Constructs a reduce L1-norm operation.

Parameters:
  • arg – The tensor to be reduced.

  • reduction_axes – The axis positions (0-based) to be eliminated.

  • keep_dims – If set to true it holds axes that are used for reduction.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceL2 : public ov::op::util::ArithmeticReductionKeepDims#
#include <reduce_l2.hpp>

Reduction operation using L2 norm:

Reduces the tensor, eliminating the specified reduction axes by taking the L2-norm.

Public Functions

ReduceL2() = default#

Constructs a reducet L2-norm operation.

ReduceL2(const Output<Node> &arg, const Output<Node> &reduction_axes, bool keep_dims = false)#

Constructs a reduce L2-norm operation.

Parameters:
  • arg – The tensor to be reduced.

  • reduction_axes – The axis positions (0-based) to be eliminated.

  • keep_dims – If set to true it holds axes that are used for reduction.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceLogicalAnd : public ov::op::util::LogicalReductionKeepDims#
#include <reduce_logical_and.hpp>

Performs a reduction using “logical and”.

The reduction is performed over slices of the first input. The slices shape depends on the values passed to the second input - the axes.

Public Functions

ReduceLogicalAnd(const Output<Node> &data, const Output<Node> &reduction_axes, const bool keep_dims = false)#

Constructs a ReduceLogicalAnd node.

Parameters:
  • data – - The input tensor with data to be reduced

  • reduction_axes – - The input tensor with information about axes over which the first tensor should be sliced prior to the reduction operation

  • keep_dims – - Indicates if the axes used for reduction should be held/kept

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceLogicalOr : public ov::op::util::LogicalReductionKeepDims#
#include <reduce_logical_or.hpp>

Performs a reduction using “logical or”.

The reduction is performed over slices of the first input. The slices shape depends on the values passed to the second input - the axes.

Public Functions

ReduceLogicalOr(const Output<Node> &data, const Output<Node> &reduction_axes, const bool keep_dims = false)#

Constructs a ReduceLogicalOr node.

Parameters:
  • data – - The input tensor with data to be reduced

  • reduction_axes – - The input tensor with information about axes over which the first tensor should be sliced prior to the reduction operation

  • keep_dims – - Indicates if the axes used for reduction should be held/kept

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceMax : public ov::op::util::ArithmeticReductionKeepDims#
#include <reduce_max.hpp>

ReduceMax operation.

Public Functions

ReduceMax() = default#

Constructs a summation operation.

ReduceMax(const Output<Node> &arg, const Output<Node> &reduction_axes, bool keep_dims = false)#

Constructs a summation operation.

Parameters:
  • arg – The tensor to be summed.

  • reduction_axes – The axis positions (0-based) to be eliminated.

  • keep_dims – If set to 1 it holds axes that are used for reduction.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceMean : public ov::op::util::ArithmeticReductionKeepDims#
#include <reduce_mean.hpp>

ReduceMean operation.

Public Functions

ReduceMean(const Output<Node> &arg, const Output<Node> &reduction_axes, bool keep_dims = false)#
Parameters:
  • arg – The tensor to be summed.

  • reduction_axes – The axis positions (0-based) to be eliminated.

  • keep_dims – If set to 1 it holds axes that are used for reduction.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceMin : public ov::op::util::ArithmeticReductionKeepDims#
#include <reduce_min.hpp>

ReduceMin operation.

Public Functions

ReduceMin() = default#

Constructs a summation operation.

ReduceMin(const Output<Node> &arg, const Output<Node> &reduction_axes, bool keep_dims = false)#

Constructs a summation operation.

Parameters:
  • arg – The tensor to be summed.

  • reduction_axes – The axis positions (0-based) to be eliminated.

  • keep_dims – If set to 1 it holds axes that are used for reduction.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceProd : public ov::op::util::ArithmeticReductionKeepDims#
#include <reduce_prod.hpp>

Product reduction operation.

Reduces the tensor, eliminating the specified reduction axes by taking the product.

Public Functions

ReduceProd() = default#

Constructs a product reduction operation.

ReduceProd(const Output<Node> &arg, const Output<Node> &reduction_axes, bool keep_dims = false)#

Constructs a product reduction operation.

Parameters:
  • arg – The tensor to be reduced.

  • reduction_axes – The axis positions (0-based) to be eliminated.

  • keep_dims – If set to true it holds axes that are used for reduction.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReduceSum : public ov::op::util::ArithmeticReductionKeepDims#
#include <reduce_sum.hpp>

Tensor sum operation.

Element-wise sums the input tensor, eliminating the specified reduction axes. For example:

\[\begin{split} \mathit{sum}\left(\{0\}, \left[ \begin{array}{ccc} 1 & 2 \\ 3 & 4 \\ 5 & 6 \end{array} \right]\right) = \left[ (1 + 3 + 5), (2 + 4 + 6) \right] = \left[ 9, 12 \right]~~~\text{(dimension 0 (rows) is eliminated)} \end{split}\]

\[\begin{split} \mathit{sum}\left(\{1\}, \left[ \begin{array}{ccc} 1 & 2 \\ 3 & 4 \\ 5 & 6 \end{array} \right]\right) = \left[ (1 + 2), (3 + 4), (5 + 6) \right] = \left[ 3, 7, 11 \right]~~~\text{(dimension 1 (columns) is eliminated)} \end{split}\]

\[\begin{split} \mathit{sum}\left(\{0,1\}, \left[ \begin{array}{ccc} 1 & 2 \\ 3 & 4 \\ 5 & 6 \end{array} \right]\right) = (1 + 2) + (3 + 4) + (5 + 6) = 21~~~\text{(both dimensions (rows and columns) are eliminated)} \end{split}\]

Parameters

Description

reduction_axes

The axes to eliminate through summation.

keep_dims

If set to 1 it holds axes that are used for reduction.

Inputs

Type

Description

arg

\(N[d_1,\dots,d_n]~(n \geq 0)\)

An input tensor of any shape and numeric element type.

Type

Description

\(N[\textit{delete}(A,d_1,\dots,d_n)]\)

The tensor \(T\), where \(T\) is the input tensor with the reduction_axes \(A\) eliminated by summation.

Public Functions

ReduceSum() = default#

Constructs a summation operation.

ReduceSum(const Output<Node> &arg, const Output<Node> &reduction_axes, bool keep_dims = false)#

Constructs a summation operation.

Parameters:
  • arg – The tensor to be summed.

  • reduction_axes – The axis positions (0-based) to be eliminated.

  • keep_dims – If set to 1 it holds axes that are used for reduction.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class RegionYolo : public ov::op::Op#
#include <region_yolo.hpp>

RegionYolo operation.

Public Functions

RegionYolo(const Output<Node> &input, const size_t coords, const size_t classes, const size_t regions, const bool do_softmax, const std::vector<int64_t> &mask, const int axis, const int end_axis, const std::vector<float> &anchors = std::vector<float>{})#

Constructs a RegionYolo operation.

Parameters:
  • input[in] Input

  • coords[in] Number of coordinates for each region

  • classes[in] Number of classes for each region

  • regions[in] Number of regions

  • do_softmax[in] Compute softmax

  • mask[in] Mask

  • axis[in] Axis to begin softmax on

  • end_axis[in] Axis to end softmax on

  • anchors[in] A flattened list of pairs [width, height] that describes prior box sizes.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Relu : public ov::op::util::UnaryElementwiseArithmetic#
#include <relu.hpp>

Elementwise Relu operation.

Public Functions

Relu(const Output<ov::Node> &arg)#

Constructs a Relu operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReorgYolo : public ov::op::Op#
#include <reorg_yolo.hpp>

ReorgYolo operation.

Public Functions

ReorgYolo(const Output<Node> &input, const size_t stride)#

Constructs a ReorgYolo operation.

Parameters:
  • inputInput

  • stride – Stride to reorganize input by

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Reshape : public ov::op::Op#
#include <reshape.hpp>

Tensor dynamic reshape operation.

“Converts” an input tensor into a new shape with the same number of elements. This op does not touch the actual data. If needed, use Transpose for that purpose.

Public Functions

Reshape(const Output<Node> &arg, const Output<Node> &shape_pattern, bool special_zero)#

Constructs a dynamic reshape operation. This operation does not perform transpose.

Parameters:
  • arg – The tensor to be reshaped.

  • shape_pattern – The node that defines output shape shape_pattern. If the input shape is \((a_0,\dots,a_{k-1})\) then the output shape must be of the form \((b_0,\dots,b_{j-1})\) where \(\Pi(a_i) = \Pi(b_i)\). A value of -1 is allowed for at most one dimension, in which case the dimension size is inferred based on element count of input tensor.

  • special_zero – Treats zeros in shape_pattern as wildcard flags indicating a copy from input shape at the same index.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Result : public ov::op::Op#
#include <result.hpp>

Result operation.

Public Functions

Result() = default#

Allows a value to be used as a function result.

Result(const Output<Node> &arg)#

Allows a value to be used as a function result.

Parameters:

argNode that produces the input tensor.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

Layout get_layout() const#

Returns current layout, or empty Layout if it is not set.

void set_layout(const Layout &layout)#

Sets layout runtime information to tensor.

Parameters:

layoutLayout to set. If empty (default constructed), layout runtime information is erased.

class Reverse : public ov::op::Op#
#include <reverse.hpp>

Reverse operation.

Public Functions

Reverse(const Output<Node> &data, const Output<Node> &reversed_axes, const std::string &mode)#

Constructs a reverse operation.

Parameters:
  • data – The input tensor, some of whose axes are to be reversed.

  • reversed_axes – The axes to reverse in a form of a set of indices or boolean mask.

  • mode – The way reversed_axes should be interpreted - a set or a mask.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline Mode get_mode() const#
Returns:

The second input data interpretation mode.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ReverseSequence : public ov::op::Op#
#include <reverse_sequence.hpp>

ReverseSequence operation.

Public Functions

ReverseSequence(const Output<Node> &arg, const Output<Node> &seq_lengths, int64_t batch_axis = 0, int64_t seq_axis = 1)#

Constructs a ReverseSequence operation.

Parameters:
  • arg – tensor with input data to reverse

  • seq_lengths – 1D tensor of integers with sequence lengths in the input tensor.

  • batch_axis – index of the batch dimension.

  • seq_axis – index of the sequence dimension.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class RNNCell : public ov::op::util::RNNCellBase#
#include <rnn_cell.hpp>

Class for single RNN cell node.

See also

LSTMSequence, LSTMCell, GRUCell

Note

It follows notation and equations defined as in ONNX standard: onnx/onnx

Note

It calculates following equations:

        Ht = f(Xt*(Wi^T) + Ht-1*(Ri^T) + Wbi + Rbi)

        *       - Is a dot product,
        f       - is activation functions.

Note

This class represents only single cell (for current time step) and not the whole RNN Sequence layer

Public Functions

RNNCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &W, const Output<Node> &R, std::size_t hidden_size, const std::vector<std::string> &activations = std::vector<std::string>{"tanh"}, const std::vector<float> &activations_alpha = {}, const std::vector<float> &activations_beta = {}, float clip = 0.f)#

Constructs RNNCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The weight tensor with shape: [hidden_size, input_size].

  • R[in] The recurrence weight tensor with shape: [hidden_size, hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

RNNCell(const Output<Node> &X, const Output<Node> &initial_hidden_state, const Output<Node> &W, const Output<Node> &R, const Output<Node> &B, std::size_t hidden_size, const std::vector<std::string> &activations = std::vector<std::string>{"tanh"}, const std::vector<float> &activations_alpha = {}, const std::vector<float> &activations_beta = {}, float clip = 0.f)#

Constructs RNNCell node.

Parameters:
  • X[in] The input tensor with shape: [batch_size, input_size].

  • initial_hidden_state[in] The hidden state tensor at current time step with shape: [batch_size, hidden_size].

  • W[in] The weight tensor with shape: [hidden_size, input_size].

  • R[in] The recurrence weight tensor with shape: [hidden_size, hidden_size].

  • B[in] The bias tensor for input gate with shape: [hidden_size].

  • hidden_size[in] The number of hidden units for recurrent cell.

  • activations[in] The vector of activation functions used inside recurrent cell.

  • activations_alpha[in] The vector of alpha parameters for activation functions in order respective to activation list.

  • activations_beta[in] The vector of beta parameters for activation functions in order respective to activation list.

  • clip[in] The value defining clipping range [-clip, clip] on input of activation functions.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class RNNSequence : public ov::op::util::RNNCellBase#
#include <rnn_sequence.hpp>

RNNSequence operation.

Public Functions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class ROIAlign : public ov::op::util::ROIAlignBase#
#include <roi_align.hpp>

ROIAlign operation.

Public Functions

ROIAlign(const Output<Node> &input, const Output<Node> &rois, const Output<Node> &batch_indices, const int pooled_h, const int pooled_w, const int sampling_ratio, const float spatial_scale, const std::string &mode)#

Constructs a ROIAlign node matching the ONNX ROIAlign specification Check util::ROIAlignBase for description of common params.

Parameters:

mode – Method of pooling - ‘avg’ or ‘max’

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ROIAlignRotated : public ov::op::util::ROIAlignBase#
#include <roi_align_rotated.hpp>

ROIAlignRotated operation.

Public Functions

ROIAlignRotated(const Output<Node> &input, const Output<Node> &rois, const Output<Node> &batch_indices, const int pooled_h, const int pooled_w, const int sampling_ratio, const float spatial_scale, const bool clockwise_mode)#

Constructs a ROIAlignRotated operation.

Parameters:
  • inputInput feature map {N, C, H, W}

  • rois – Regions of interest to pool over

  • batch_indices – Indices of images in the batch matching the number or ROIs

  • pooled_h – Height of the ROI output features

  • pooled_w – Width of the ROI output features

  • sampling_ratio – Number of sampling points used to compute an output element

  • spatial_scale – Spatial scale factor used to translate ROI coordinates

  • clockwise_mode – If true, rotation angle is interpreted as clockwise, otherwise as counterclockwise

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class ROIPooling : public ov::op::Op#
#include <roi_pooling.hpp>

ROIPooling operation.

Public Functions

ROIPooling(const Output<Node> &input, const Output<Node> &coords, const Shape &output_size, const float spatial_scale, const std::string &method = "max")#

Constructs a ROIPooling operation.

Parameters:
  • inputInput feature map {N, C, H, W}

  • coords – Coordinates of bounding boxes

  • output_size – Height/Width of ROI output features

  • spatial_scale – Ratio of input feature map over input image size

  • method – Method of pooling - Max or Bilinear

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

void set_output_roi(Shape output_size)#

Set the output ROI feature map (pooled_h, pooled_w).

Parameters:

output_sizeShape with pooling attributes pooled_h and pooled_w sizes.

const Shape &get_output_roi() const#

Get the output ROI feature map shape (H x W)

Returns:

Shape with pooled_h and pooled_w attributes.

void set_spatial_scale(float scale)#

Set the spatial scale value.

Parameters:

scale – Scale value to set.

void set_method(std::string method_name)#

Set the method of pooling.

Parameters:

method_name – Pooling method name.

class Roll : public ov::op::Op#
#include <roll.hpp>

Tensor roll operation.

Public Functions

Roll(const Output<Node> &data, const Output<Node> &shift, const Output<Node> &axes)#

Constructs a roll operation.

Parameters:
  • dataNode producing the tensor to be shifted.

  • shiftNode producing the 0D or 1D tensor which specifies the number of places by which the elements are shifted.

  • axesNode producing the 0D or 1D tensor which specifies axes along which elements are shifted.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Round : public ov::op::util::UnaryElementwiseArithmetic#
#include <round.hpp>

Elementwise round operation. The output is round to the nearest integer for each value. In case of halfs, the rule is defined in attribute ‘mode’: ‘HALF_TO_EVEN’ - round halfs to the nearest even integer. ‘HALF_AWAY_FROM_ZERO’: - round in such a way that the result heads away from zero.

Public Functions

Round() = default#

Constructs a round operation.

Round(const Output<Node> &arg, const RoundMode mode)#

Constructs a round operation.

Parameters:
  • argNode that produces the input tensor.

  • mode – Rule to resolve halfs

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ScaledDotProductAttention : public ov::op::Op#
#include <scaled_dot_product_attention.hpp>

Scaled dot product attention operation from PyTorch.

Public Functions

ScaledDotProductAttention() = default#

Constructs a ScaledDotProductAttention operation.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class ScatterElementsUpdate : public ov::op::util::ScatterElementsUpdateBase#
#include <scatter_elements_update.hpp>

ScatterElementsUpdate operation.

Public Functions

ScatterElementsUpdate(const Output<Node> &data, const Output<Node> &indices, const Output<Node> &updates, const Output<Node> &axis)#

Constructs a ScatterElementsUpdate node.

Parameters:
  • dataInput data

  • indices – Data entry index that will be updated

  • updates – Update values

  • axis – Axis to scatter on

class ScatterNDUpdate : public ov::op::util::ScatterNDBase#
#include <scatter_nd_update.hpp>

Add updates to slices from inputs addressed by indices.

Public Functions

inline ScatterNDUpdate(const Output<Node> &inputs, const Output<Node> &indices, const Output<Node> &updates)#
Parameters:
virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ScatterNDUpdate : public ov::op::util::ScatterNDBase#
#include <scatter_nd_update.hpp>

Add updates to slices from inputs addressed by indices.

Public Types

enum class Reduction#

Lists the supported reduction types for this version of the operator. See the specification for the description of how reduction works with ScatterNDUpdate.

Values:

enumerator NONE#
enumerator SUM#
enumerator SUB#
enumerator PROD#
enumerator MIN#
enumerator MAX#

Public Functions

ScatterNDUpdate(const Output<Node> &inputs, const Output<Node> &indices, const Output<Node> &updates, const Reduction reduction = Reduction::NONE)#
Parameters:
  • inputsTensor

  • indices – Index tensor: Data type must be element::i32 or element::i64

  • updatesTensor: Must have same type as inputs

  • reduction – Reduction: Type of operation to perform on inputs

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ScatterUpdate : public ov::op::util::ScatterBase#
#include <scatter_update.hpp>

Set new values to slices from data addressed by indices.

Public Functions

ScatterUpdate(const Output<Node> &data, const Output<Node> &indices, const Output<Node> &updates, const Output<Node> &axis)#

Constructs ScatterUpdate operator object.

Parameters:
  • data – The input tensor to be updated.

  • indices – The tensor with indexes which will be updated.

  • updates – The tensor with update values.

  • axis[in] The axis at which elements will be updated.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Select : public ov::op::Op#
#include <select.hpp>

Elementwise selection operation.

Inputs

Type

Description

arg0

\(\texttt{bool}[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of any shape, with element bool.

arg1

\(E[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of a shape that is broadcast-compatible with arg0, with any element type.

arg2

\(E[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of a shape that is broadcast-compatible with arg0, and same element type as arg1.

auto_broadcast

AutoBroadcastSpec

Auto broadcast specification.

Type

Description

\(E[d_1,\dots,d_n]\)

The tensor \(T\), where \(T[i_1,\dots,i_n] = \texttt{arg1}[i_1,\dots,i_n]\text{ if }\texttt{arg0}[i_1,\dots,i_n] \neq 0\text{, else }\texttt{arg2}[i_1,\dots,i_n]\)

Public Functions

inline Select()#

Constructs a selection operation.

Select(const Output<Node> &arg0, const Output<Node> &arg1, const Output<Node> &arg2, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a selection operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • arg2Node that produces the third input tensor.

  • auto_broadcast – Auto broadcast specification. Default is Numpy-style implicit broadcasting.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline virtual const AutoBroadcastSpec &get_autob() const override#
Returns:

the autobroadcasr spec

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Selu : public ov::op::Op#
#include <selu.hpp>

Performs a SELU activation function on all elements of the input node.

Public Functions

Selu(const Output<Node> &data, const Output<Node> &alpha, const Output<Node> &lambda)#

Constructs a Selu node.

Parameters:
  • data – - Node producing the input tensor

  • alpha – - Alpha coefficient of SELU operation

  • lambda – - Lambda coefficient of SELU operation

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class ShapeOf : public ov::op::util::ShapeOfBase#
#include <shape_of.hpp>

Operation that returns the shape of its input argument as a tensor.

Public Functions

ShapeOf(const Output<Node> &arg, const element::Type output_type = element::i64)#

Constructs a shape-of operation.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ShapeOf : public ov::op::util::ShapeOfBase#
#include <shape_of.hpp>

Operation that returns the shape of its input argument as a tensor.

Public Functions

ShapeOf(const Output<Node> &arg)#

Constructs a shape-of operation.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class ShuffleChannels : public ov::op::Op#
#include <shuffle_channels.hpp>

Permutes data in the channel dimension of the input.

Public Functions

ShuffleChannels(const Output<Node> &data, const int64_t axis = 1, const int64_t group = 1)#

Constructs a ShuffleChannels node.

Parameters:
  • dataNode producing the input tensor.

  • axis – Channel dimension index in the data tensor. A negative value means that the index should be calculated from the back of the input data shape.

  • group – Number of group the channel dimension should be split into.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Sigmoid : public ov::op::util::UnaryElementwiseArithmetic#
#include <sigmoid.hpp>

Sigmoid operation.

Public Functions

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Sign : public ov::op::util::UnaryElementwiseArithmetic#
#include <sign.hpp>

Elementwise sign operation.

Public Functions

Sign(const Output<Node> &arg)#

Constructs an elementwise sign operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Sin : public ov::op::util::UnaryElementwiseArithmetic#
#include <sin.hpp>

Elementwise sine operation.

Inputs

Type

Description

arg

\(N[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of any shape and numeric element type.

Type

Description

\(N[d_1,\dots,d_n]\)

The tensor \(T\), where \(T[i_1,\dots,i_n] = \sin(\texttt{arg}[i_1,\dots,i_n])\)

Public Functions

Sin(const Output<Node> &arg)#

Constructs a sine operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Sinh : public ov::op::util::UnaryElementwiseArithmetic#
#include <sinh.hpp>

Elementwise hyperbolic sine (sinh) operation.

Public Functions

Sinh(const Output<Node> &arg)#

Constructs a hyperbolic sine operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Sink : public ov::op::Op#
#include <sink.hpp>

Root of nodes that can be sink nodes.

Subclassed by ov::op::util::AssignBase, ov::op::util::MultiSubGraphOp

class Slice : public ov::op::Op#
#include <slice.hpp>

Slice operation.

Public Functions

Slice(const Output<Node> &data, const Output<Node> &start, const Output<Node> &stop, const Output<Node> &step)#

Constructs Slice operation (default axes).

Parameters:
  • data – The tensor to be sliced.

  • start – 1D tensor with start indices of the slice.

  • stop – 1D tensor with end indices of the slice.

  • step – 1D tensor specifies the increment to use in slicing along corresponding axes.

Slice(const Output<Node> &data, const Output<Node> &start, const Output<Node> &stop, const Output<Node> &step, const Output<Node> &axes)#

Constructs Slice operation.

Parameters:
  • data – The tensor to be sliced.

  • start – 1D tensor with start indices of the slice.

  • stop – 1D tensor with end indices of the slice.

  • step – 1D tensor specifies the increment to use in slicing along corresponding axes.

  • axes – 1D tensor indicating which dimensions the values in the start and stop apply to.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class SliceScatter : public ov::op::Op#
#include <slice_scatter.hpp>

SliceScatter operation.

Public Functions

SliceScatter(const Output<Node> &data, const Output<Node> &updates, const Output<Node> &start, const Output<Node> &stop, const Output<Node> &step)#

Constructs SliceScatter operation (default axes).

Parameters:
  • data – The tensor to be updated.

  • updatesTensor containing update values.

  • start – 1D tensor with start indices of the update slice.

  • stop – 1D tensor with end indices of the update slice.

  • step – 1D tensor specifies the increment to use in slicing along corresponding axes.

SliceScatter(const Output<Node> &data, const Output<Node> &updates, const Output<Node> &start, const Output<Node> &stop, const Output<Node> &step, const Output<Node> &axes)#

Constructs SliceScatter operation.

Parameters:
  • data – The tensor to be updated.

  • updatesTensor containing update values.

  • start – 1D tensor with start indices of the update slice.

  • stop – 1D tensor with end indices of the update slice.

  • step – 1D tensor specifies the increment to use in slicing along corresponding axes.

  • axes – 1D tensor indicating which dimensions the values in the start and stop apply to.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Softmax : public ov::op::Op#
#include <softmax.hpp>

Softmax operation.

Public Functions

Softmax(const Output<Node> &arg, const size_t axis = 1)#

Constructs a softmax operation.

Output [d0, ...]

Parameters:
  • argNode that produces the first input tensor.[d0, ...]

  • axis – The axis position (0-based) on which to calculate the softmax.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Softmax : public ov::op::Op#
#include <softmax.hpp>

Softmax operation with negative axis values.

Public Functions

Softmax(const Output<Node> &arg, const int64_t axis = 1)#

Constructs a softmax operation.

Output [d0, ...]

Parameters:
  • argNode that produces the first input tensor.[d0, ...]

  • axis – The axis position (0-based) in range [-rank(arg), rank(arg) - 1] on which to calculate the softmax.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class SoftPlus : public ov::op::util::UnaryElementwiseArithmetic#
#include <softplus.hpp>

A Self Regularized Non-Monotonic Neural Activation Function f(x) = ln(exp(x) + 1.)

Public Functions

SoftPlus(const Output<Node> &arg)#

Constructs an SoftPlus operation.

Parameters:

dataInput tensor

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class SpaceToBatch : public ov::op::Op#
#include <space_to_batch.hpp>

SpaceToBatch permutes data tensor blocks of spatial data into batch dimension.

Note

Values from spatial blocks dimensions are moved in the batch dimension.

   Output node produces a tensor with shape: tensor with shape
   `[batch * block_shape[0] * block_shape[1] * ... * block_shape[N - 1],
    (pads_begin[1] + D_1 + pads_end[1]) / block_shape[1],
    (pads_begin[2] + D_2 + pads_end[2]) / block_shape[2], ...,
    (pads_begin[N - 1] + D_{N - 1} + pads_end[N - 1]) / block_shape[N - 1]`
    of the same type as `data` input.

Public Functions

SpaceToBatch(const Output<Node> &data, const Output<Node> &block_shape, const Output<ov::Node> &pads_begin, const Output<ov::Node> &pads_end)#

Constructs a SpaceToBatch operation.

Parameters:
  • dataNode producing the data tensor

  • block_shape – The sizes of the block of values to be moved

  • pads_begin – Specifies the padding for the beginning along each axis of data input

  • pads_end – Specifies the padding for the ending along each axis of data input.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class SpaceToDepth : public ov::op::Op#
#include <space_to_depth.hpp>

SpaceToDepth permutes input tensor blocks of spatial data into depth dimension.

Note

Values from the height and width dimensions are moved to the depth dimension.

   Output node produces a tensor with shape:
   [N, C * blocksize * blocksize, H / blocksize, W / blocksize]

Public Functions

SpaceToDepth(const Output<Node> &data, const SpaceToDepthMode &mode, std::size_t block_size = 1)#

Constructs a SpaceToDepth operation.

Parameters:
  • data – - Node producing the input tensor

  • mode – Specifies how the output depth dimension is gathered from block coordinates and the old depth dimension.

  • block_size – - the size of the block of values to be moved

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Split : public ov::op::Op#
#include <split.hpp>

Splits the input tensor into a list of equal sized tensors.

Public Functions

Split() = default#

Constructs a split operation.

Split(const Output<Node> &data, const Output<Node> &axis, const size_t num_splits)#

Constructs a split operation.

Parameters:
  • data – The tensor to be split.

  • axis – The index of an axis in “data” along which to perform the split.

  • num_splits – The number of pieces that the data tensor should be split into.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Sqrt : public ov::op::util::UnaryElementwiseArithmetic#
#include <sqrt.hpp>

Elementwise square root operation.

Inputs

Type

Description

arg

\(N[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of any shape and numeric element type.

Type

Description

\(N[d_1,\dots,d_n]\)

The tensor \(T\), where \(T[i_1,\dots,i_n] = \sqrt{\texttt{arg}[i_1,\dots,i_n]}\)

Public Functions

Sqrt(const Output<Node> &arg)#

Constructs a square operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class SquaredDifference : public ov::op::util::BinaryElementwiseArithmetic#
#include <squared_difference.hpp>

Calculates an element-wise squared difference between two tensors.

y[i] = (x1[i] - x2[i])^2

Public Functions

inline SquaredDifference()#

Constrcuts an uninitialized squared difference operation.

SquaredDifference(const Output<Node> &x1, const Output<Node> &x2, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs the squared difference operation.

Parameters:
  • x1 – First input tensor

  • x2 – Second input tensor

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Squeeze : public ov::op::Op#
#include <squeeze.hpp>

Squeeze operation.

Public Functions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class StridedSlice : public ov::op::Op#
#include <strided_slice.hpp>

Takes a slice of an input tensor, i.e., the sub-tensor that resides within a bounding box, optionally with stride.

Public Functions

StridedSlice(const Output<Node> &data, const Output<Node> &begin, const Output<Node> &end, const Output<Node> &strides, const std::vector<int64_t> &begin_mask, const std::vector<int64_t> &end_mask, const std::vector<int64_t> &new_axis_mask = std::vector<int64_t>{}, const std::vector<int64_t> &shrink_axis_mask = std::vector<int64_t>{}, const std::vector<int64_t> &ellipsis_mask = std::vector<int64_t>{})#

Constructs a dynamic tensor strided slice operation.

Parameters:
  • data – The tensor to be sliced.

  • begin – 1D tensor with begin indexes for input blob slicing.

  • end – 1D tensor with end indexes for input blob slicing.

  • strides – The slicing strides; for example, strides of {n,m} means to take every nth row and every mth column of the input matrix.

  • begin_mask – When begin_mask[i] equal to 1 means that the corresponding dimension of the begin input is ignored.

  • end_mask – When end_mask[i] is 1, the corresponding dimension of the end input is ignored.

  • new_axis_mask – If new_axis_mask[i] is 1, a length 1 dimension is inserted on the i-th position.

  • shrink_axis_mask – If shrink_axis_mask[i] is 1, the dimension on the i-th position is deleted.

  • ellipsis_mask – It inserts missing dimensions on a position of a non-zero bit.

StridedSlice(const Output<Node> &data, const Output<Node> &begin, const Output<Node> &end, const std::vector<int64_t> &begin_mask, const std::vector<int64_t> &end_mask, const std::vector<int64_t> &new_axis_mask = std::vector<int64_t>{}, const std::vector<int64_t> &shrink_axis_mask = std::vector<int64_t>{}, const std::vector<int64_t> &ellipsis_mask = std::vector<int64_t>{})#

Constructs a dynamic tensor strided slice operation.

Parameters:
  • data – The tensor to be sliced.

  • begin – 1D tensor with begin indexes for input blob slicing.

  • end – 1D tensor with end indexes for input blob slicing.

  • begin_mask – When begin_mask[i] equal to 1 means that the corresponding dimension of the begin input is ignored.

  • end_mask – When end_mask[i] is 1, the corresponding dimension of the end input is ignored.

  • new_axis_mask – If new_axis_mask[i] is 1, a length 1 dimension is inserted on the i-th position.

  • shrink_axis_mask – If shrink_axis_mask[i] is 1, the dimension on the i-th position is deleted.

  • ellipsis_mask – It inserts missing dimensions on a position of a non-zero bit.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class StringTensorPack : public ov::op::Op#
#include <string_tensor_pack.hpp>

Operator packing a concatenated batch of strings into a batched string tensor.

Public Functions

StringTensorPack(const Output<Node> &begins, const Output<Node> &ends, const Output<Node> &symbols)#

Constructs a StringTensorPack operation.

Parameters:
  • begins – Indices of each string’s beginnings

  • ends – Indices of each string’s endings

  • symbols – Concatenated input strings encoded in utf-8 bytes

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class StringTensorUnpack : public ov::op::Op#
#include <string_tensor_unpack.hpp>

Operator unpacking a batch of strings into three tensors.

Public Functions

StringTensorUnpack(const Output<Node> &data)#

Constructs a StringTensorUnpack operation.

Parameters:

dataInput of type element::string

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Subtract : public ov::op::util::BinaryElementwiseArithmetic#
#include <subtract.hpp>

Elementwise subtraction operation.

Public Functions

Subtract(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec(AutoBroadcastType::NUMPY))#

Constructs a subtraction operation.

Parameters:
  • arg0Node that produces the first input tensor.

  • arg1Node that produces the second input tensor.

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Swish : public ov::op::Op#
#include <swish.hpp>

A Swish Activation Function f(x) = x / (1.0 + exp(-beta * x)) or f(x) = x * sigmoid(beta * x)

Public Functions

Swish(const Output<Node> &arg, const Output<Node> &beta)#

Constructs an Swish operation.

Parameters:
  • dataInput tensor

  • beta – Scalar with beta value. If the argument is not specified then use the default value 1.0

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Tan : public ov::op::util::UnaryElementwiseArithmetic#
#include <tan.hpp>

Elementwise tangent operation.

Inputs

Type

Description

arg

\(N[d_1,\dots,d_n]~(n \geq 0)\)

A tensor of any shape and numeric element type.

Type

Description

\(N[d_1,\dots,d_n]\)

The tensor \(T\), where \(T[i_1,\dots,i_n] = \tan(\texttt{arg}[i_1,\dots,i_n])\)

Public Functions

Tan(const Output<Node> &arg)#

Constructs a tangent operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Tanh : public ov::op::util::UnaryElementwiseArithmetic#
#include <tanh.hpp>

Elementwise hyperbolic tangent operation.

Public Functions

Tanh(const Output<Node> &arg)#

Constructs a hyperbolic tangent operation.

Parameters:

argNode that produces the input tensor.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class TensorIterator : public ov::op::util::SubGraphOp#
#include <tensor_iterator.hpp>

Iterate a body over tensors, accumulating into tensors.

Public Functions

inline std::shared_ptr<Model> get_body() const#
Returns:

the body of the iteration

inline void set_body(const std::shared_ptr<Model> &body)#
Parameters:

body – set the body of the iteration

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Tile : public ov::op::Op#
#include <tile.hpp>

Dynamic Tiling operation which repeats a tensor multiple times along each dimension.

Public Functions

Tile(const Output<Node> &data, const Output<Node> &repeats)#

Perform dynamic padding of a tensor.

Parameters:
  • data – The node producing input tensor to be padded.

  • repeats – The node producing the per-dimension replication factor

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

class TopK : public ov::op::util::TopKBase#
#include <topk.hpp>

Computes indices and values of the k maximum/minimum values for each slice along specified axis.

Public Functions

TopK() = default#

Constructs a TopK operation.

TopK(const Output<Node> &data, const Output<Node> &k, const int64_t axis, const std::string &mode, const std::string &sort, const element::Type &index_element_type = element::i32)#

Constructs a TopK operation with two outputs: values and indices. By default the indices output is described by i32 data type.

Parameters:
  • data – The input tensor

  • k – Specifies how many maximum/minimum elements should be computed (note: scalar input tensor)

  • axis – The axis along which to compute top k indices

  • mode – Specifies which operation (min or max) is used to select the biggest element of two.

  • sort – Specifies order of output elements and/or indices Accepted values: none, index, value

  • index_element_type – Specifies type of produced indices

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class TopK : public ov::op::util::TopKBase#
#include <topk.hpp>

Computes indices and values of the k maximum/minimum values for each slice along specified axis.

Public Functions

TopK() = default#

Constructs a TopK operation.

TopK(const Output<Node> &data, const Output<Node> &k, const int64_t axis, const std::string &mode, const std::string &sort, const element::Type &index_element_type = element::i32)#

Constructs a TopK operation with two outputs: values and indices. By default the indices output is described by i32 data type.

Parameters:
  • data – The input tensor

  • k – Specifies how many maximum/minimum elements should be computed (note: scalar input tensor)

  • axis – The axis along which to compute top k indices

  • mode – Specifies which operation (min or max) is used to select the biggest element of two.

  • sort – Specifies order of output elements and/or indices Accepted values: none, index, value

  • index_element_type – Specifies type of produced indices

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class TopK : public ov::op::util::TopKBase#
#include <topk.hpp>

Computes the top K elements of a given tensor along the specified axis.

Public Functions

TopK() = default#

Constructs a TopK operation.

TopK(const Output<Node> &data, const Output<Node> &k, const int64_t axis, const std::string &mode, const std::string &sort, const element::Type &index_element_type = element::i32, const bool stable = false)#

Constructs a TopK operation with two outputs: values and indices.

Parameters:
  • data – The input tensor

  • k – Specifies how many maximum/minimum elements should be computed

  • axis – The axis along which the TopK operation should be executed

  • mode – Specifies whether TopK selects the largest or the smallest elements from each slice

  • sort – Specifies the order of corresponding elements of the output tensor

  • index_element_type – Specifies the data type of the elements in the ‘indices’ output tensor.

  • stable – Specifies whether the equivalent elements should maintain their relative order from the input tensor during sorting.

TopK(const Output<Node> &data, const Output<Node> &k, const int64_t axis, const TopKMode mode, const TopKSortType sort, const element::Type &index_element_type = element::i32, const bool stable = false)#

Constructs a TopK operation with two outputs: values and indices.

Parameters:
  • data – The input tensor

  • k – Specifies how many maximum/minimum elements should be computed

  • axis – The axis along which the TopK operation should be executed

  • mode – Specifies whether TopK selects the largest or the smallest elements from each slice

  • sort – Specifies the order of corresponding elements of the output tensor

  • index_element_type – Specifies the data type of the elements in the ‘indices’ output tensor.

  • stable – Specifies whether the equivalent elements should maintain their relative order from the input tensor during sorting.

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Transpose : public ov::op::Op#
#include <transpose.hpp>

Tensor transpose operation.

Public Types

enum Ins#

Inputs indexes and count.

Values:

enumerator ARG#
enumerator ORDER#
enumerator IN_COUNT#
enum Outs#

Outputs indexes and count.

Values:

enumerator ARG_T#
enumerator OUT_COUNT#

Public Functions

Transpose(const Output<Node> &arg, const Output<Node> &input_order)#

Constructs a transpose operation.

Parameters:
  • argNode producing the tensor to be transposed.

  • input_orderNode producing the permutation to apply to the axes of the input shape. Must be a vector with shape [n], where n is the rank of arg. The tensor’s value must contain every integer in the range [0, n-1].

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Unique : public ov::op::Op#
#include <unique.hpp>

Operator which selects and returns unique elements or unique slices of the input tensor.

Public Functions

Unique(const Output<Node> &data, const bool sorted = true, const element::Type &index_element_type = element::i64, const element::Type &count_element_type = element::i64)#

Constructs a Unique operation.

Parameters:
  • dataInput data tensor

  • sorted – Controls the order of the returned unique values (sorts ascendingly when true)

  • index_element_type – The data type for outputs containing indices

  • count_element_type – The data type for output containing repetition count

Unique(const Output<Node> &data, const Output<Node> &axis, const bool sorted = true, const element::Type &index_element_type = element::i64, const element::Type &count_element_type = element::i64)#

Constructs a Unique operation.

Parameters:
  • dataInput data tensor

  • axis – An input tensor containing the axis value

  • sorted – Controls the order of the returned unique values (sorts ascendingly when true)

  • index_element_type – The data type for outputs containing indices

  • count_element_type – The data type for output containing repetition count

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

class Unsqueeze : public ov::op::Op#
#include <unsqueeze.hpp>

Unsqueeze operation.

Public Functions

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

virtual bool evaluate(ov::TensorVector &outputs, const ov::TensorVector &inputs) const override#

Evaluates the op on input_values putting results in output_values.

Parameters:
  • output_values – Tensors for the outputs to compute. One for each result

  • input_values – Tensors for the inputs. One for each inputs.

Returns:

true if successful

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class VariadicSplit : public ov::op::Op#
#include <variadic_split.hpp>

VariadicSplit operation splits an input tensor into pieces along some axis. The pieces may have variadic lengths depending on “split_lengths” attribute.

Public Functions

VariadicSplit() = default#

Constructs a variadic split operation.

VariadicSplit(const Output<Node> &data, const Output<Node> &axis, const Output<Node> &split_lengths)#

Constructs a variadic split operation.

outputs. The sum of split_lengths must match data.shape[axis]

Parameters:
  • data – The tensor to be split.

  • axis – The index of an axis in “data” along which to perform the split.

  • split_lengths – A list containing the sizes of each output tensor along the split “axis”. Size of “split_lengths” should be equal to the number of

virtual void validate_and_infer_types() override#

Verifies that attributes and inputs are consistent and computes output shapes and element types. Must be implemented by concrete child classes so that it can be run any number of times.

Throws if the node is invalid.

inline virtual size_t get_default_output_index() const override#

Returns the output of the default output, or throws if there is none.

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.

class Xor : public ov::op::util::BinaryElementwiseLogical#
#include <xor.hpp>

Elementwise logical-xor operation.

Public Functions

Xor(const Output<Node> &arg0, const Output<Node> &arg1, const AutoBroadcastSpec &auto_broadcast = AutoBroadcastSpec())#

Constructs a logical-xor operation.

Output [d0, ...]

Parameters:
  • arg0Node that produces the first input tensor.[d0, ...]

  • arg1Node that produces the second input tensor.[d0, ...]

  • auto_broadcast – Auto broadcast specification

virtual bool has_evaluate() const override#

Allows to get information about availability of evaluate method for the current operation.