What Is ONNX? (And Why Transformers.js Uses It)
Transformers.js uses ONNX as an open standard to run models portably across devices and backends.
“ONNX says what to compute. ONNX runtime decides how to compute it.”
An educational explainer from Hugging Face describing how ONNX (Open Neural Network Exchange) provides a portable model format and how ONNX Runtime's execution providers let Transformers.js run models across WebGPU, WebAssembly, and other backends. It clarifies the architecture/weights split and the runtime abstraction but contains no new announcement or industry-moving signal.