onnxruntime/docs/tutorials/web/ep-webnn.md

title	description	parent	grand_parent	has_children	nav_order
Using WebNN	Using WebNN	Web	Tutorials	false	4

{::options toc_levels="2..4" /}

Using the WebNN Execution Provider

{: .no_toc }

This document explains how to use the WebNN execution provider in ONNX Runtime.

Contents

{: .no_toc}

• TOC {:toc}

Basics

What is WebNN? Should I use it?

Web Neural Network (WebNN) API is a new web standard that allows web apps and frameworks to accelerate deep neural networks with on-device hardware such as GPUs, CPUs, or purpose-built AI accelerators(NPUs).

WebNN is available in latest versions of Chrome and Edge on Windows, Linux, macOS, Android and ChromeOS behind a "Enables WebNN API" flag. Check WebNN status for the latest implementation status.

Refer to the WebNN operators for the most recent status of operator support in the WebNN execution provider. If the WebNN execution provider supports most of the operators in your model (with unsupported operators falling back to the WASM EP), and you wish to achieve power-efficient, faster processing and smoother performance by utilizing on-device accelerators, consider using the WebNN execution provider.

How to use WebNN EP in ONNX Runtime Web

This section assumes you have already set up your web application with ONNX Runtime Web. If you haven't, you can follow the Get Started for some basic info.

To use WebNN EP, you just need to make 3 small changes:

1. Update your import statement:

   • For HTML script tag, change ort.min.js to ort.all.min.js:

     <script src="https://example.com/path/ort.all.min.js"></script>
     

   • For JavaScript import statement, change onnxruntime-web to onnxruntime-web/all:

     import * as ort from 'onnxruntime-web/all';
     

   See Conditional Importing for details.

2. Specify 'webnn' EP explicitly in session options:

   const session = await ort.InferenceSession.create(modelPath, { ..., executionProviders: ['webnn'] });
   

   WebNN EP also offers a set of options for creating diverse types of WebNN MLContext.

   • deviceType: 'cpu'|'gpu'|'npu'(default value is 'cpu'), specifies the preferred type of device to be used for the MLContext.
   • powerPreference: 'default'|'low-power'|'high-performance'(default value is 'default'), specifies the preferred type of power consumption to be used for the MLContext.
   • numThreads: type of number, allows users to specify the number of multi-threads for 'cpu' device type.
   • context: type of MLContext, allows users to pass a pre-created MLContext to WebNN EP, it is required in IO binding feature. If this option is provided, the other options will be ignored.

   Example of using WebNN EP options:

   const options = {
      executionProviders: [
        {
          name: 'webnn',
          deviceType: 'gpu',
          powerPreference: "default",
        },
      ],
   }
   

3. If it is dynamic shape model, ONNX Runtime Web offers freeDimensionOverrides session option to override the free dimensions of the model. See freeDimensionOverrides introduction for more details.

WebNN API and WebNN EP are in actively development, you might consider installing the latest nightly build version of ONNX Runtime Web (onnxruntime-web@dev) to benefit from the latest features and improvements.

Keep tensor data on WebNN MLBuffer (IO binding)

By default, a model's inputs and outputs are tensors that hold data in CPU memory. When you run a session with WebNN EP with 'gpu' or 'npu' device type, the data is copied to GPU or NPU memory, and the results are copied back to CPU memory. Memory copy between different devices as well as different sessions will bring much overhead to the inference time, WebNN provides a new opaque device-specific storage type MLBuffer to address this issue. If you get your input data from a MLBuffer, or you want to keep the output data on MLBuffer for further processing, you can use IO binding to keep the data on MLBuffer. This will be especially helpful when running transformer based models, which usually runs a single model multiple times with previous output as the next input.

For model input, if your input data is a WebNN storage MLBuffer, you can create a MLBuffer tensor and use it as input tensor.

For model output, there are 2 ways to use the IO binding feature:

• Use pre-allocated MLBuffer tensors
• Specify the output data location

Please also check the following topic:

• MLBuffer tensor life cycle management

Note: The MLBuffer necessitates a shared MLContext for IO binding. This implies that the MLContext should be pre-created as a WebNN EP option and utilized across all sessions.

Create input tensor from a MLBuffer

If your input data is a WebNN storage MLBuffer, you can create a MLBuffer tensor and use it as input tensor:

const mlContext = await navigator.ml.createContext({deviceType, ...});
const inputMLBuffer = await mlContext.createBuffer({
  dataType: 'float32',
  dimensions: [1, 3, 224, 224],
  usage: MLBufferUsage.WRITE_TO,
});

mlContext.writeBuffer(mlBuffer, inputArrayBuffer);
const inputTensor = ort.Tensor.fromMLBuffer(mlBuffer, {
  dataType: 'float32',
  dims: [1, 3, 224, 224]
});

Use this tensor as model inputs(feeds) so that the input data will be kept on MLBuffer.

Use pre-allocated MLBuffer tensors

If you know the output shape in advance, you can create a MLBuffer tensor and use it as output tensor:

// Create a pre-allocated buffer and the corresponding tensor. Assuming that the output shape is [10, 1000].
const mlContext = await navigator.ml.createContext({deviceType, ...});
const myPreAllocatedBuffer = await mlContext.createBuffer({
  dataType: 'float32',
  dimensions: [10, 1000],
  usage: MLBufferUsage.READ_FROM,
});

const myPreAllocatedOutputTensor = ort.Tensor.fromMLBuffer(myPreAllocatedBuffer, {
  dataType: 'float32',
  dims: [10, 1000]
});

// ...

// Run the session with fetches
const feeds = { 'input_0': myInputTensor };
const fetches = { 'output_0': myPreAllocatedOutputTensor };
const results = await mySession.run(feeds, fetches);

By specifying the output tensor in the fetches, ONNX Runtime Web will use the pre-allocated buffer as the output buffer. If there is a shape mismatch, the run() call will fail.

Specify the output data location

If you don't want to use pre-allocated MLBuffer tensors for outputs, you can also specify the output data location in the session options:

const mySessionOptions1 = {
  ...,
  // keep all output data on MLBuffer
  preferredOutputLocation: 'ml-buffer'
};

const mySessionOptions2 = {
  ...,
  // alternatively, you can specify the output location for each output tensor
  preferredOutputLocation: {
    'output_0': 'cpu',         // keep output_0 on CPU. This is the default behavior.
    'output_1': 'ml-buffer'   // keep output_1 on MLBuffer buffer
  }
};

By specifying the config preferredOutputLocation, ONNX Runtime Web will keep the output data on the specified device.

See API reference: preferredOutputLocation for more details.

Notes

MLBuffer tensor life cycle management

It is important to understand how the underlying MLBuffer is managed so that you can avoid memory leaks and improve buffer usage efficiency.

A MLBuffer tensor is created either by user code or by ONNX Runtime Web as model's output.

• When it is created by user code, it is always created with an existing MLBuffer using Tensor.fromMLBuffer(). In this case, the tensor does not "own" the MLBuffer.

  • It is user's responsibility to make sure the underlying buffer is valid during the inference, and call mlBuffer.destroy() to dispose the buffer when it is no longer needed.
  • Avoid calling tensor.getData() and tensor.dispose(). Use the MLBuffer directly.
  • Using a MLBuffer tensor with a destroyed MLBuffer will cause the session run to fail.

• When it is created by ONNX Runtime Web as model's output (not a pre-allocated MLBuffer tensor), the tensor "owns" the buffer.

  • You don't need to worry about the case that the buffer is destroyed before the tensor is used.
  • Call tensor.getData() to download the data from the MLBuffer to CPU and get the data as a typed array.
  • Call tensor.dispose() explicitly to destroy the underlying MLBuffer when it is no longer needed.