Transformers.js: State-of-the-art Machine Learning for the web
summary
Library brings Hugging Face Transformers parity to browsers using ONNX and Optimum, with WebGPU support and WebAssembly execution; over 1,200 converted models across ~120 architectures are available from the Hugging Face hub and used via a simple pipeline API. Reported metrics include ~750,000 monthly users, 2.4M NPM downloads, and measured speedups like 64x on BERT and ~90 tokens/sec for a 53.5-mini LLM on an RTX 4090; Whisper WebGPU gives ~10x real‑time on M3 Max. Models are cached per site, currently capped by a 4GB Wasm limit.
AI-written summary · about
chapters (4)
related talks
26mDeploying PyTorch Models To the Browser and Beyond With Transformers.js
33mUsing Transformers.js to Bridge the Gap Between AI Research and
57mMachine Learning with JavaScript
22mBoring but Effective: WebAI That Ships
15mMachine learning for next gen web apps with TensorFlow.js
20mTensorFlow.js Bringing Machine Learning to the Web and Beyond
41mEscaping the Cloud: High-Performance AI in your Browser
29mSuperpowers for next gen web apps: Machine learning