We have introduced several backend implementations. They are Vanilla CPU, WebGL, and Node.js. These backends were provided from the very beginning of TensorFlow.js, but the community keeps developing new backend implementations to support many more use cases and platforms. Bear in mind that most of them are still experimental and they do not fully support all operations. Let's take a look at how and when to use these backends in our applications.
Experimental backend implementations
WebGPU – a new standard for accelerated graphics and computations
WebGPU is a new web standard for accelerated graphics processing and computation. It provides a brand-new JavaScript API that allows us to flexibly achieve higher-performance...
