| --- |
| library_name: kernels |
| license: apache-2.0 |
| tags: |
| - kernel |
| - webgpu |
| - wgsl |
| --- |
| # ai.onnx.Pad |
|
|
| `ai.onnx` · standard ONNX operator · ONNX opset ≥ 19 |
|
|
| ## Description |
|
|
| Pads a tensor along each axis using one of four modes: `constant` (fill with a scalar value), `reflect` (mirror edge values), `edge` (replicate boundary values), or `wrap` (torus-like wrap-around). Supply the required ONNX `pads` tensor, and optional `axes` tensor when present, through the `pads` request attribute as a full-rank list. Negative counts crop the corresponding edge. |
|
|
| See the [ONNX `Pad` spec](https://onnx.ai/onnx/operators/onnx__Pad.html) for the reference semantics. |
|
|
| ## Inputs |
|
|
| | Name | Logical dtype | Rank | Shape | Description | Presence | |
| | --- | --- | --- | --- | --- | --- | |
| | `data` | `T` | — | — | Input tensor to be padded. | required | |
| | `constant_value` | `T` | `0` | — | Optional scalar value used as the fill constant when `mode` is `constant`; defaults to 0. | optional | |
|
|
| ## Outputs |
|
|
| | Name | Logical dtype | Rank | Shape | Description | Presence | |
| | --- | --- | --- | --- | --- | --- | |
| | `output` | `T` | same as `data` | — | Tensor after padding, with each axis enlarged by the corresponding begin and end pad counts. | required | |
|
|
| ## Attributes |
|
|
| Attributes and default values (overridable per request): |
|
|
| | Attribute | Default | Description | |
| | --- | --- | --- | |
| | `mode` | `"constant"` | Padding mode: `constant` (default), `reflect`, `edge`, or `wrap`. | |
| | `pads` | — | Values of the required `pads` tensor, with an optional `axes` tensor expanded to full rank, supplied as `[x1_begin, ..., xN_begin, x1_end, ..., xN_end]`. | |
|
|
| ## Type constraints |
|
|
| | Variable | Allowed dtypes | |
| | --- | --- | |
| | `T` | `float32`, `float16`, `uint32`, `int32`, `uint8`, `int8`, `bool`, `int64` | |
|
|
| ## Files |
|
|
| - [`metadata.json`](build/webgpu/metadata.json) — kernel metadata (id, digests, per-variant templates, provenance) |
| - [`manifest.json`](build/webgpu/manifest.json) — the op contract (source of truth) |
| - [`test.json`](build/webgpu/test.json) — correctness cases |
| - [`bench.json`](build/webgpu/bench.json) — benchmark cases |
| - [`pad-quad-vec4.wgsl.jinja`](build/webgpu/pad-quad-vec4.wgsl.jinja) |
| - [`pad.wgsl.jinja`](build/webgpu/pad.wgsl.jinja) |
|
|
| ## Use with `@huggingface/kernels` |
|
|
| ```sh |
| npm install --save-exact @huggingface/kernels@0.0.1-preview.3 |
| ``` |
|
|
| Outputs with inferable metadata are allocated automatically. Explicit `outputs` entries request optional results or provide metadata that cannot be inferred from the supplied inputs and attributes. |
|
|
| This example supplies explicit metadata for: |
|
|
| - `output` |
|
|
| 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. |
|
|
| ```js |
| import { getKernel } from "@huggingface/kernels"; |
| |
| const kernel = await getKernel("webgpu-kernels/ai.onnx.Pad", { version: 1 }); |
| // Explicit destinations request optional results or supply metadata that cannot be inferred. |
| const { output } = await kernel({ data: { data: dataData, shape: [2] } }, { |
| attrs: { pads: [0, 0] }, |
| outputs: { output: { shape: [2], dtype: "float32" } }, |
| }); |
| ``` |
|
|
