---
library_name: kernels
license: apache-2.0
tags:
- kernel
- webgpu
- wgsl
---
# ai.onnx.Sub

`ai.onnx`  ·  standard ONNX operator  ·  ONNX opset ≥ 14

## Description

Performs elementwise binary subtraction (`A - B`) with multidirectional NumPy-style broadcasting support. Inputs must share a compatible numeric element type; the output has the same element type as the inputs.

See the [ONNX `Sub` spec](https://onnx.ai/onnx/operators/onnx__Sub.html) for the reference semantics.

## Inputs

| Name | Bind key | Logical dtype | Rank | Shape | Description | Presence |
| --- | --- | --- | --- | --- | --- | --- |
| `A` | `a` | `T` | — | — | First operand. | required |
| `B` | `b` | `T` | — | — | Second operand. | required |

## Outputs

| Name | Bind key | Logical dtype | Rank | Shape | Description | Presence |
| --- | --- | --- | --- | --- | --- | --- |
| `C` | `c` | `T` | derived | broadcast result of `A` and `B` | Result of the subtraction; has the same element type as the inputs. | required |

## Type constraints

| Variable | Allowed dtypes |
| --- | --- |
| `T` | `float32`, `float16`, `int32`, `uint32`, `int8`, `uint8` |

## 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
- [`binary-broadcast-vec4.wgsl.jinja`](build/webgpu/binary-broadcast-vec4.wgsl.jinja)
- [`binary-broadcast.wgsl.jinja`](build/webgpu/binary-broadcast.wgsl.jinja)
- [`binary-vec4.wgsl.jinja`](build/webgpu/binary-vec4.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.Sub", { version: 1 });
const { c } = await kernel({ a: { data: aData, shape: [3] }, b: { data: bData, shape: [3] } });
```