ONNX Runtime
summary
ONNX Runtime is an open‑source, high‑performance scoring engine on GitHub that runs ONNX models on CPU and GPU across Linux, Windows, and Mac, and it supports custom hardware accelerators with contributions from Nvidia and Intel. The demo converts a Core ML “tiny yellow” object‑detection model from the ONNX model zoo (convertible from TensorFlow, Keras, Scikit‑Learn, PyTorch, Core ML), registers and deploys via Azure ML into an image/ACS or AKS service, claiming measurable latency and throughput gains and noting support from over 20 companies.
AI-written summary · about
chapters (3)
related talks
30mScalable ML acceleration with ONNX Runtime
45mLeverage Power of Machine Learning with ONNX
33mWhat on earth is ONNX and ONNXRUNTIME
41mLeverage Power of Machine Learning with ONNX
30mThe Evolution of ONNX: Driving Interoperability in AI...
51mMaking neural networks run in browser with ONNX
05mONNX-Go, neural networks made easy
21m#10 Open Source Forum 2019 Shintarou Okada