| --- |
| library_name: kernels |
| license: apache-2.0 |
| tags: |
| - kernel |
| - webgpu |
| - wgsl |
| --- |
| # ai.onnx.HannWindow |
|
|
| `ai.onnx` · standard ONNX operator · ONNX opset ≥ 17 |
|
|
| ## Description |
|
|
| Generates a Hann window of a given length using the formula `0.5 - 0.5 * cos(2π * n / N)`, where `N` is `size` for periodic mode or `size - 1` for symmetric mode. The output is a 1-D tensor of length `size`. |
|
|
| See the [ONNX `HannWindow` spec](https://onnx.ai/onnx/operators/onnx__HannWindow.html) for the reference semantics. |
|
|
| ## Inputs |
|
|
| | Name | Logical dtype | Rank | Shape | Description | Presence | |
| | --- | --- | --- | --- | --- | --- | |
| | `size` | `T1` | `0` | — | Scalar indicating the number of elements in the output window. | required | |
|
|
| ## Outputs |
|
|
| | Name | Upstream name | Logical dtype | Rank | Shape | Description | Presence | |
| | --- | --- | --- | --- | --- | --- | --- | |
| | `y` | `output` | `T2` | `1` | — | 1-D Hann window tensor of shape `[size]`. | required | |
|
|
| ## Attributes |
|
|
| Default values (overridable per request): |
|
|
| | Attribute | Default | Description | |
| | --- | --- | --- | |
| | `output_datatype` | `1` | Data type of the output tensor, encoded as a DataType enum value from TensorProto (default `1` = FLOAT). | |
| | `periodic` | `1` | If `1` (default), returns a periodic window of length `size`; if `0`, returns a symmetric window of length `size`. | |
|
|
| ## Type constraints |
|
|
| | Variable | Allowed dtypes | |
| | --- | --- | |
| | `T1` | `int32` | |
| | `T2` | `float32`, `float16`, `uint32`, `int32`, `uint8`, `int8`, `int16` | |
|
|
| ## 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 |
| - [`window-cosine-sum.wgsl.jinja`](build/webgpu/window-cosine-sum.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: |
|
|
| - `y` |
|
|
| 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.HannWindow", { version: 1 }); |
| // Explicit destinations request optional results or supply metadata that cannot be inferred. |
| const { y } = await kernel({ size: { data: sizeData, shape: [] } }, { |
| outputs: { y: { shape: [1], dtype: "float32" } }, |
| }); |
| ``` |
|
|
