ai.onnx.Gelu

ai.onnx · standard ONNX operator · ONNX opset ≥ 20

Description

Applies the Gaussian Error Linear Unit activation elementwise: y = 0.5 * x * (1 + erf(x / sqrt(2))). When approximate is "tanh", uses the tanh-based approximation y = 0.5 * x * (1 + tanh(sqrt(2/π) * (x + 0.044715 * x³))) instead.

See the ONNX Gelu spec for the reference semantics.

Inputs

Outputs

Attributes

Default values (overridable per request):

Type constraints

Implementation variants

One implementation is selected per call from the device capabilities, the request shapes and the dtypes; these notes say what each one covers.

• same_layout_vec4_tail — Processes the vec4-aligned prefix in a bulk pass and any remaining elements in a scalar tail pass.

Files

• metadata.json — kernel metadata (id, digests, per-variant templates, provenance)
• manifest.json — the op contract (source of truth)
• test.json — correctness cases
• bench.json — benchmark + tuning cases
• unary-scalar.wgsl.jinja
• unary-vec4.wgsl.jinja

Use with @huggingface/kernels

npm install --save-exact @huggingface/kernels@0.0.1-preview.2

Required output shapes and logical data types are inferred from the supplied inputs and attributes; result tensors are allocated automatically.

The version: 1 option selects the published kernel contract; it is independent of any operator opset, contrib since_version, or model version. It follows the v1 branch as fixes land. To pin exact artifact bytes, pass a 40-character commit revision instead of version.

Replace each *Data placeholder with a typed array containing the corresponding input data.

import { getKernel } from "@huggingface/kernels";

const kernel = await getKernel("webgpu-kernels/ai.onnx.Gelu", { version: 1 });
const { y } = await kernel({ x: { data: xData, shape: [] } });