Open Standards For Machine Learning Model Deployment
summary
Compares three concrete model-exchange standards and tooling: PMML XML model+transformations as a portable, schema‑based way to record feature transforms and scoring logic; PFA for function‑level predictive workflows with composable steps; and ONNX + ONNX Runtime with converters from TensorFlow/Spark/scikit‑learn and production runtimes that several companies run today. Shows exported examples, conversion commands, where round‑trip fidelity fails, and which runtimes successfully serve models at scale.
AI-written summary · about
chapters (7)
- 0:00Introduction to open standards
- 4:01Challenges in model deployment
- 10:16Transformation formats and dictionaries
- 17:40Open source tools and products
- 24:10Portability and extensions
- 29:07Industry adoption and support
- 33:10Conclusion and Q&A
related talks
41mOpen standards for machine learning deployment
33mOpen Standards for Machine Learning Deployment
33mOpen Standards for Machine Learning Deployment
48mDeploying Machine Learning Models in Practice
46mProduction Model Deployment
54mOne Universal Language for All Machine Learning Models
24mMaking Spark ML Models Portable Know Your Options
45mMaking Machine Learning Applications Fast and Simple with...