Elu

Elu - 22

Version

• name: Elu (GitHub)

• domain: main

• since_version: 22

• function: True

• support_level: SupportType.COMMON

• shape inference: True

This version of the operator has been available since version 22.

Summary

Elu takes one input data (Tensor<T>) and produces one output data (Tensor<T>) where the function f(x) = alpha * (exp(x) - 1.) for x < 0, f(x) = x for x >= 0., is applied to the tensor elementwise.

Function Body

The function definition for this operator.

<
  domain: "",
  opset_import: ["" : 18]
>
Elu <alpha>(X) => (Y)
{
   Alpha = Constant <value_float: float = @alpha> ()
   AlphaCast = CastLike (Alpha, X)
   Zero = Constant <value: tensor = float {0}> ()
   ZeroCast = CastLike (Zero, X)
   One = Constant <value: tensor = float {1}> ()
   OneCast = CastLike (One, X)
   XLessThanZero = Less (X, ZeroCast)
   ExpX = Exp (X)
   ExpXSubOne = Sub (ExpX, OneCast)
   AlphaMulExpXSubOne = Mul (AlphaCast, ExpXSubOne)
   Y = Where (XLessThanZero, AlphaMulExpXSubOne, X)
}

Attributes

• alpha - FLOAT (default is 1.0):

  Coefficient of ELU.

Inputs

• X (heterogeneous) - T:

  Input tensor

Outputs

• Y (heterogeneous) - T:

  Output tensor

Type Constraints

• T in ( tensor(bfloat16), tensor(double), tensor(float), tensor(float16) ):

  Constrain input and output types to float tensors.

Examples

default

import numpy as np
import onnx

node = onnx.helper.make_node("Elu", inputs=["x"], outputs=["y"], alpha=2.0)

x = np.array([-1, 0, 1]).astype(np.float32)
# expected output [-1.2642411, 0., 1.]
y = np.clip(x, 0, np.inf) + np.expm1(np.clip(x, -np.inf, 0)) * 2.0
expect(node, inputs=[x], outputs=[y], name="test_elu_example")

x = np.random.randn(3, 4, 5).astype(np.float32)
y = np.clip(x, 0, np.inf) + np.expm1(np.clip(x, -np.inf, 0)) * 2.0
expect(node, inputs=[x], outputs=[y], name="test_elu")

_elu_default

import numpy as np
import onnx

default_alpha = 1.0
node = onnx.helper.make_node(
    "Elu",
    inputs=["x"],
    outputs=["y"],
)
x = np.random.randn(3, 4, 5).astype(np.float32)
y = np.clip(x, 0, np.inf) + np.expm1(np.clip(x, -np.inf, 0)) * default_alpha
expect(node, inputs=[x], outputs=[y], name="test_elu_default")

• Elu - 6 vs 22

Elu - 6

Version

• name: Elu (GitHub)

• domain: main

• since_version: 6

• function: True

• support_level: SupportType.COMMON

• shape inference: True

This version of the operator has been available since version 6.

Summary

Elu takes one input data (Tensor<T>) and produces one output data (Tensor<T>) where the function f(x) = alpha * (exp(x) - 1.) for x < 0, f(x) = x for x >= 0., is applied to the tensor elementwise.

Function Body

The function definition for this operator.

<
  domain: "",
  opset_import: ["" : 18]
>
Elu <alpha>(X) => (Y)
{
   Alpha = Constant <value_float: float = @alpha> ()
   AlphaCast = CastLike (Alpha, X)
   Zero = Constant <value: tensor = float {0}> ()
   ZeroCast = CastLike (Zero, X)
   One = Constant <value: tensor = float {1}> ()
   OneCast = CastLike (One, X)
   XLessThanZero = Less (X, ZeroCast)
   ExpX = Exp (X)
   ExpXSubOne = Sub (ExpX, OneCast)
   AlphaMulExpXSubOne = Mul (AlphaCast, ExpXSubOne)
   Y = Where (XLessThanZero, AlphaMulExpXSubOne, X)
}

Attributes

• alpha - FLOAT (default is 1.0):

  Coefficient of ELU.

Inputs

• X (heterogeneous) - T:

  Input tensor

Outputs

• Y (heterogeneous) - T:

  Output tensor

Type Constraints

• T in ( tensor(double), tensor(float), tensor(float16) ):

  Constrain input and output types to float tensors.

• Elu - 1 vs 22
• Elu - 1 vs 6

Elu - 1

Version

• name: Elu (GitHub)

• domain: main

• since_version: 1

• function: False

• support_level: SupportType.COMMON

• shape inference: False

This version of the operator has been available since version 1.

Summary

Elu takes one input data (Tensor<T>) and produces one output data (Tensor<T>) where the function f(x) = alpha * (exp(x) - 1.) for x < 0, f(x) = x for x >= 0., is applied to the tensor elementwise.

Attributes

• alpha - FLOAT (default is 1.0):

  Coefficient of ELU default to 1.0.

• consumed_inputs - INTS :

  legacy optimization attribute.

Inputs

• X (heterogeneous) - T:

  Input tensor

Outputs

• Y (heterogeneous) - T:

  Output tensor

Type Constraints

• T in ( tensor(double), tensor(float), tensor(float16) ):

  Constrain input and output types to float tensors.