Skip to content

Latest commit

 

History

History
99 lines (59 loc) · 8 KB

ml_in_production.md

File metadata and controls

99 lines (59 loc) · 8 KB

Machine Learning In Production

First Thing First - What does it mean by ML In Production?

  • How do you take a machine learning model to production? Håkon Hapnes Strand's answer in Quora - Here & There

  • When Models Go Rogue: Hard Earned Lessons About Using Machine Learning in Production - Blog, Slides, Talk, Talk at Strata Data Conference 2017

  • What does your production machine learning pipeline look like? - Hacker News Thread

  • What is the process of deploying machine learning models in production (For any ML library)? - Reddit Thread

  • Quest to understand Machine Learning in Production & Notes - Towards Data Science Blog - Part 1 and Part 2

Use Cases & Papers

  • Michelangelo: Uber’s Machine Learning Platform - Blog, Paper

  • TFX: A TensorFlow-Based Production-Scale Machine Learning Platform (by Google) - Paper

  • Meson: Workflow Orchestration for Netflix Recommendations - Blog, Video

  • Airbnb's End-to-End Machine Learning Infrastructure - Slides, Video

Best Practices

  • Rules of Machine Learning: Best Practices for ML Engineering - Document, Video, google course

  • What’s your ML Test Score? A rubric for ML production systems - Paper, Slides, Video

  • Hidden Technical Debt in Machine Learning Systems - Paper

  • Machine Learning: The High Interest Credit Card of Technical Debt - Paper

  • Practical Methodology from Deep Learning Book - Chapter

Courses

  • Deployment of Machine Learning Models (Udemy) - Link

  • The Facebook Field Guide to Machine Learning - Videos

  • Machine Learning Crash Course (by Google)

    • Production ML Systems - Link
    • Static vs. Dynamic Training - Link
    • Static vs. Dynamic Inference - Link
    • Data Dependencies - Link

End to End Machine Learning (As a Service)

  • Amazon Sagemaker - Feature List, Documentation

  • Azure Machine Learning Service (Not Azure Machine Studio) - Link

    • Experimentation service
    • Model management
    • Workbench

  • Google Cloud Machine Learning Engine - Link

  • Domino's DataLab - Available as a service & on premise Link, Demo Video

Machine Learing Server / Model Servers

Model Management

  • How do you version control models? - Quora Answer by Anand Sampat, Co-Founder & CEO @ Datmo - Link

  • MLFlow (Alpha) Tracking Module - Python APIs can be used with any Machine Learning library for Model Management Link

  • H2O Steam - Not under active development. Open Source Version is still available under AGPL license (viral?) - Documentation, github

  • ModelDB - Developed as a part of PHD research project at MIT. Supports spark-ml, scikit-learn out of the box. Can be used with any ML environment via the ModelDB Light API. Not under active development(?) - Link

  • comet.ml - Available as a service

  • studio.ml - Link

  • Michelangelo: Uber’s Machine Learning Platform - Blog, Paper

Feature Store

  • logicalclocks Blog

Data Versioning

  • Data Version Control - Blog
  • pachyderm - Link

Quick Start

  • Blog : How to Deploy Machine Learning Models A Guide - Link

  • Blog : A Guide to Scaling Machine Learning Models in Production - Link

  • Blog : A guide to deploying Machine/Deep Learning model(s) in Production - Link

Performance Benchmarking For Machine Learning Libraries

  • benchm-ml : Benchmark for scalability, speed, accuracy of commonly used open source implementations of the top machine learning algorithms (using binary classification) - github, (talks)[https://github.com/szilard/talks]

  • MLPerf - ML benchmark suite for measuring performance of ML software frameworks, ML hardware accelerators, and ML cloud platforms - Link