Open Standards for Machine Learning Deployment
summary
Covers practical adoption of open ML formats and concrete tradeoffs: PMML works well for classical models and metadata but lacks deep‑learning and custom ops; PFA offers flexible scoring and cross‑language function calls but misses tensor/GPU support and has limited uptake; ONNX targets deep nets and broad framework export (TensorFlow, PyTorch, Keras, scikit‑learn, Spark ML) but currently focuses on static inference not training. Recommends using a portfolio of standards rather than one silver bullet.
AI-written summary · about
chapters (12)
- 0:00Intro
- 2:22What is Machine Learning?
- 3:51Intelligent Systems
- 5:46In reality the workflow spans teams...
- 10:56What is a "model"?
- 12:09Challenges
- 15:27Containers are The Solution... right?
- 17:16Why a standard?
- 18:37Predictive Model Markup Language
- 22:11Portable Format for Analytics
- 28:09Open Neural Network Exchange ONNX
- 29:19ONNX Supported Tools
related talks
33mOpen Standards for Machine Learning Deployment
41mOpen standards for machine learning deployment
34mOpen Standards For Machine Learning Model Deployment
48mDeploying Machine Learning Models in Practice
26mIntroducing the Model Openness Framework: Achieving...
25m#6 Open Source Forum 2019 Adam Gibson
54mOne Universal Language for All Machine Learning Models
39mFrom OpenAI to Opensource AI: Navigating Between Commercial Ownership and Coll...