| --- |
| library_name: kernels |
| license: apache-2.0 |
| tags: |
| - kernel |
| - webgpu |
| - wgsl |
| --- |
| # ai.onnx.InstanceNormalization |
|
|
| `ai.onnx` · standard ONNX operator · ONNX opset ≥ 6 |
|
|
| ## Description |
|
|
| Applies instance normalization to the input: `y = scale * (x - mean) / sqrt(variance + epsilon) + B`, where `mean` and `variance` are computed per instance per channel over the spatial dimensions. Equivalent to batch normalization with a batch size of one per channel. |
|
|
| See the [ONNX `InstanceNormalization` spec](https://onnx.ai/onnx/operators/onnx__InstanceNormalization.html) for the reference semantics. |
|
|
| ## Inputs |
|
|
| | Name | Bind key | Logical dtype | Rank | Shape | Description | Presence | |
| | --- | --- | --- | --- | --- | --- | --- | |
| | `input` | `input` | `T` | — | — | Input tensor of shape `(N x C x D1 x ... x Dn)`; at least 3-D. | required | |
| | `scale` | `scale` | `T` | `1` | — | 1-D scale tensor of size C, one scale factor per channel. | required | |
| | `B` | `b` | `T` | `1` | — | 1-D bias tensor of size C, one bias value per channel. | required | |
|
|
| ## Outputs |
|
|
| | Name | Bind key | Logical dtype | Rank | Shape | Description | Presence | |
| | --- | --- | --- | --- | --- | --- | --- | |
| | `output` | `output` | `T` | same as `input` | same as `input` | Normalized output tensor; same shape as the input. | required | |
|
|
| ## Attributes |
|
|
| Default values (overridable per request): |
|
|
| | Attribute | Default | Description | |
| | --- | --- | --- | |
| | `epsilon` | `0.00001` | Small constant added to the variance before taking the square root to avoid division by zero. | |
|
|
| ## Type constraints |
|
|
| | Variable | Allowed dtypes | |
| | --- | --- | |
| | `T` | `float32`, `float16` | |
|
|
| ## Device requirements |
|
|
| Some implementation variants require `subgroups`. These are route-specific capabilities, not package-wide requirements; availability also depends on the request shape and dtype. |
|
|
| ## Files |
|
|
| - [`metadata.json`](build/webgpu/metadata.json) — kernel metadata (id, digests, 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 + tuning cases |
| - [`instance-normalization-apply.wgsl.jinja`](build/webgpu/instance-normalization-apply.wgsl.jinja) |
| - [`instance-normalization-batched-planes-vec4.wgsl.jinja`](build/webgpu/instance-normalization-batched-planes-vec4.wgsl.jinja) |
| - [`instance-normalization-splitk-combine.wgsl.jinja`](build/webgpu/instance-normalization-splitk-combine.wgsl.jinja) |
| - [`instance-normalization-splitk-partials.wgsl.jinja`](build/webgpu/instance-normalization-splitk-partials.wgsl.jinja) |
| - [`norm-row-stats.wgsl.jinja`](build/webgpu/norm-row-stats.wgsl.jinja) |
|
|
| ## Use with `@huggingface/kernels` |
|
|
| The loader derives every required output's shape and logical dtype from the manifest contract and this call. |
| It then allocates the result tensors automatically. |
|
|
| The `version: 1` option selects the published kernel contract; it is independent of any operator opset, contrib `since_version`, or model 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.InstanceNormalization", { version: 1 }); |
| const { output } = await kernel({ |
| input: { data: inputData, shape: [1, 2, 1, 3] }, |
| scale: { data: scaleData, shape: [2] }, |
| b: { data: bData, shape: [2] }, |
| }); |
| ``` |
|
|
