onnxrt

OCaml bindings to ONNX Runtime Web for browser-based ML inference via js_of_ocaml.

Supports WebAssembly (CPU) and WebGPU (GPU) execution providers, typed tensors, and session management. Models are loaded from .onnx files and run asynchronously using Lwt.

Examples

• Tensor addition — minimal example of creating tensors and running a model
• Sentiment analysis — text classification using a transformer model

API

The main entry points are:

• Onnxrt.Session — load models and run inference
• Onnxrt.Tensor — create and inspect typed tensors
• Onnxrt.Env — configure execution providers (WASM threads, WebGPU, SIMD)