AttributeHasValue#

AttributeHasValue - 18 #

Version

name: AttributeHasValue (GitHub)
domain: main
since_version: 18
function: False
support_level: SupportType.COMMON
shape inference: True

This version of the operator has been available since version 18.

Summary

Returns true if at least one of the attribute-value is specified.

Attributes

value_float: The float attribute.
value_floats: The floats attribute.
value_graph: The graph attribute.
value_graphs: The graphs attribute.
value_int: The int attribute.
value_ints: The ints attribute.
value_sparse_tensor: The sparse_tensor attribute.
value_sparse_tensors: The sparse_tensors attribute.
value_string: The string attribute.
value_strings: The strings attribute.
value_tensor: The tensor attribute.
value_tensors: The tensors attribute.
value_type_proto: The type_proto attribute.
value_type_protos: The type_protos attribute.

Outputs

output (heterogeneous) - B: A scalar boolean tensor. If true, it indicates that an attribute is provided.

Type Constraints

B in ( tensor(bool) ): Constrain output to a boolean tensor.

Examples

default

import numpy as np
import onnx

def test_one_attribute(name: str, **kwargs: Any) -> None:
    node = onnx.helper.make_node(
        "AttributeHasValue",
        inputs=[],
        outputs=["output"],
    )

    output = np.array(False)
    expect(
        node,
        inputs=[],
        outputs=[output],
        name=f"test_attribute_has_{name}_false",
    )

    node = onnx.helper.make_node(
        "AttributeHasValue",
        inputs=[],
        outputs=["output"],
        **kwargs,
    )

    output = np.array(True)
    expect(
        node,
        inputs=[],
        outputs=[output],
        name=f"test_attribute_has_{name}_true",
    )

value_float = 0.1
test_one_attribute("value_float", value_float=value_float)

value_int = 1
test_one_attribute("value_int", value_int=value_int)

value_string = "test"
test_one_attribute("value_string", value_string=value_string)

tensor_values = np.random.randn(5, 5).astype(np.float32)
value_tensor = onnx.helper.make_tensor(
    name="const_tensor",
    data_type=onnx.TensorProto.FLOAT,
    dims=tensor_values.shape,
    vals=tensor_values.flatten().astype(float),
)
test_one_attribute("value_tensor", value_tensor=value_tensor)

value_graph = onnx.parser.parse_graph("agraph (X) => (Y) {Y = Identity(X)}")
test_one_attribute("value_graph", value_graph=value_graph)

value_sparse_tensor = onnx.helper.make_sparse_tensor(
    onnx.helper.make_tensor(
        name="",
        data_type=onnx.TensorProto.FLOAT,
        dims=(5,),
        vals=[1.1, 2.2, 3.3, 4.4, 5.5],
    ),
    onnx.helper.make_tensor(
        name="",
        data_type=onnx.TensorProto.INT64,
        dims=(5,),
        vals=[1, 3, 5, 7, 9],
    ),
    [10],
)

test_one_attribute(
    "value_sparse_tensor", value_sparse_tensor=value_sparse_tensor
)

value_type_proto = onnx.helper.make_tensor_type_proto(
    onnx.TensorProto.FLOAT, shape=[5]
)
test_one_attribute("value_type_proto", value_type_proto=value_type_proto)

value_floats = [0.0, 1.1]
test_one_attribute("value_floats", value_floats=value_floats)

value_ints = [0, 1]
test_one_attribute("value_ints", value_ints=value_ints)

value_strings = ["test strings"]
test_one_attribute("value_strings", value_strings=value_strings)

value_tensors = [value_tensor, value_tensor]
test_one_attribute("value_tensors", value_tensors=value_tensors)

value_graphs = [value_graph, value_graph]
test_one_attribute("value_graphs", value_graphs=value_graphs)

value_sparse_tensors = [value_sparse_tensor, value_sparse_tensor]
test_one_attribute(
    "value_sparse_tensors", value_sparse_tensors=value_sparse_tensors
)

value_type_protos = [value_type_proto, value_type_proto]
test_one_attribute("value_type_protos", value_type_protos=value_type_protos)

AttributeHasValue#

AttributeHasValue - 18#

AttributeHasValue - 18 #