GELU- Gaussian Error Linear Unit

Versioned name: Gelu-2

Category: Activation

Short description: Reference

Detailed description: Reference

Attributes: Gelu operation has no attributes.

Mathematical Formulation Gelu(x)=x*Φ(x), where Φ(x) is the Cumulative Distribution Function for Gaussian Distribution. The following equivalent combination is recognized and fused into single Gelu op:

\[ Gelu(x) = 0.5*x*(1.0 + erf((x) / \sqrt{2}) \]

Similarly, the following Gelu approximation (typical for the TensorFlow*) is recognized and fused into single Gelu op

\[ Gelu(x) \approx 0.5x(1.0 + tanh(\sqrt{2.0/pi} * (x + 0.044715 * x ^ 3)) \]

Inputs:

• 1: Multidimensional input tensor. Required.

Example

<layer ... type="Gelu">
    <input>
        <port id="0">
            <dim>1</dim>
            <dim>128</dim>
        </port>
    </input>
    <output>
        <port id="1">
            <dim>1</dim>
            <dim>128</dim>
        </port>
    </output>
</layer>