Web Neural Network API

This repository hosts the Web Neural Network API being worked on in the Web Machine Learning Working Group.

To join this Working Group, please see these instructions.

Development

This section describes the process for building this specification on your local computer for development and testing purposes. You may want to do this before you submit a Pull Request to catch possible build issues early.

The first step is to clone the repo:

prompt> git clone https://github.com/webmachinelearning/webnn

Next, install the makefile dependencies:

prompt> cd tools
prompt> npm install
prompt> cd ..

Now you can choose between the two options:

Install Bikeshed locally (works offline, faster build) and run:

prompt> make

Or use the Bikeshed HTTP API (requires internet, slower build):

prompt> make online=1

In both the cases, the makefile runs tools to check the source formatting and catch common errors, and then uses Bikeshed to convert the index.bs source file to an index.html output file.

If the build is clean you can open the index.html file in your browser to view it in all its glory! If you're happy with what you see, congratulations! You can proceed to submit a Pull Request.

Pull Requests

First, please read the contributing guidelines.

The spec is written using Bikeshed spec-generating tool. Please refer to the Bikeshed Documentation for the Bikeshed syntax and coding conventions specific to this specification.

To propose a change to the specification, you are only required to update the index.bs source file authored in Bikeshed syntax.

Before you submit a PR, it is recommended to test your changes locally (see Development). If your local build is clean, you can proceed to submit the PR. If you want to submit a WIP PR that still contains errors or is incomplete, please convert the PR to a draft. As a convention, please link the PR to an existing issue for non-editorial changes.

After you submit the PR, it is first automatically checked with GitHub Actions CI/CD for common issues and then reviewed by group participants. After adequate review and approvals, the PR is merged and a new version of the specification is deployed at https://www.w3.org/TR/webnn/ (history).

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!