ONNX Runtime Web Introduces WebGPU: Machine Learning in Web Browsers Takes a Performance Leap

Microsoft recently announced the addition of WebGPU technology to its ONNX Runtime Web, which will help improve the performance of running complex machine learning models in web browsers. Currently, Chrome and Edge browsers have begun to support this technology, albeit in an early stage.

ONNX Runtime Web is a Javascript library launched by Microsoft to help web developers deploy machine learning models directly in web browsers and provide multiple backend support based on hardware acceleration. WebGPU is an emerging network application programming interface (API) that can provide hardware acceleration for machine learning models to run more efficiently in web browsers. Previously, due to the limited computing power of web browsers, the running efficiency of complex machine learning models, especially large-scale generative models, has been limited in web browsers. However, with the introduction of WebGPU technology, this problem is expected to be solved. WebGPU unlocks the power of device GPUs so that these complex machine learning models can run efficiently in web browsers, significantly improving performance.

According to the editor’s understanding, Microsoft stated that WebGPU can handle complex machine learning workloads in a more efficient manner with the help of advanced features such as compute shaders. In addition, WebGPU's support for half-precision (FP16) also reduces GPU memory usage and bandwidth requirements, further speeding up computing. Through these optimizations, WebGPU is expected to infer more efficient and scalable machine learning applications directly in the web browser.

Currently, Chrome 113 and Edge 113 for Mac, Windows, ChromeOS, and Chrome for Android 121 WebGPU support is included by default. This means developers can try out the ONNX Runtime Web calls WebGPU acceleration resources to improve the running performance of its machine learning model in web browsers. As this technology continues to develop and improve, we are expected to see more efficient and convenient machine learning applications implemented in web browsers in the future.

The above is the detailed content of ONNX Runtime Web Introduces WebGPU: Machine Learning in Web Browsers Takes a Performance Leap. For more information, please follow other related articles on the PHP Chinese website!