| --- |
| library_name: kernels |
| license: apache-2.0 |
| tags: |
| - kernel |
| - webgpu |
| - wgsl |
| --- |
| # ai.onnx.Pow |
|
|
| `ai.onnx` · standard ONNX operator · ONNX opset ≥ 15 |
|
|
| ## Description |
|
|
| Computes elementwise exponentiation `Z = X ^ Y`, applying `f(x) = x^y` to each element. Supports multidirectional (NumPy-style) broadcasting between `X` and `Y`. |
|
|
| See the [ONNX `Pow` spec](https://onnx.ai/onnx/operators/onnx__Pow.html) for the reference semantics. |
|
|
| ## Inputs |
|
|
| | Name | Upstream name | Logical dtype | Rank | Shape | Description | Presence | |
| | --- | --- | --- | --- | --- | --- | --- | |
| | `x` | `X` | `T` | — | — | Base tensor; first operand of the exponentiation. | required | |
| | `y` | `Y` | `U` | — | — | Exponent tensor; power applied to each base element. | required | |
|
|
| ## Outputs |
|
|
| | Name | Upstream name | Logical dtype | Rank | Shape | Description | Presence | |
| | --- | --- | --- | --- | --- | --- | --- | |
| | `z` | `Z` | `T` | derived | broadcast result of `x` and `y` | Output tensor containing `X` raised to the power `Y`, same type as `X`. | required | |
|
|
| ## Runtime arguments |
|
|
| | Name | Kind | Description | Presence | |
| | --- | --- | --- | --- | |
| | `constantY` | `f32` | Optional host-known constant exponent. When the exponent tensor `y` holds one integer value in `[-16, 16]` that the caller also passes here, a specialized integer-power kernel is selected. | optional | |
|
|
| ## Type constraints |
|
|
| | Variable | Allowed dtypes | |
| | --- | --- | |
| | `T` | `float32`, `float16`, `int32` | |
| | `U` | `float32`, `float16`, `int32`, `int16`, `uint32`, `int8`, `uint8` | |
|
|
| ## 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 |
| - [`pow-scalar-exponent-vec4.wgsl.jinja`](build/webgpu/pow-scalar-exponent-vec4.wgsl.jinja) |
| - [`pow-vec4.wgsl.jinja`](build/webgpu/pow-vec4.wgsl.jinja) |
| - [`pow.wgsl.jinja`](build/webgpu/pow.wgsl.jinja) |
|
|
| ## Use with `@huggingface/kernels` |
|
|
| ```sh |
| npm install --save-exact @huggingface/kernels@0.0.1-preview.3 |
| ``` |
|
|
| 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. |
|
|
| ```js |
| import { getKernel } from "@huggingface/kernels"; |
| |
| const kernel = await getKernel("webgpu-kernels/ai.onnx.Pow", { version: 1 }); |
| const { z } = await kernel({ x: { data: xData, shape: [1] }, y: { data: yData, shape: [1] } }); |
| ``` |
|
|
