CI/CD for Machine Learning
2 likes395 views
InfoQ is a global community and news site for over 1,000,000 software developers, offering a variety of content including podcasts, articles, and conference materials. The document discusses the purpose and strategy of QCon, a practitioner-driven conference aimed at promoting knowledge and innovation in software development, specifically focusing on Machine Learning Operations (MLOps) and Continuous Integration/Continuous Delivery (CI/CD) practices. It highlights key topics in MLOps, such as model management, reproducibility, and methodologies to improve machine learning workflows.
CI/CD for Machine Learning
- 2. InfoQ.com: News & Community Site • Over 1,000,000 software developers, architects and CTOs read the site world- wide every month • 250,000 senior developers subscribe to our weekly newsletter • Published in 4 languages (English, Chinese, Japanese and Brazilian Portuguese) • Post content from our QCon conferences • 2 dedicated podcast channels: The InfoQ Podcast, with a focus on Architecture and The Engineering Culture Podcast, with a focus on building • 96 deep dives on innovative topics packed as downloadable emags and minibooks • Over 40 new content items per week Watch the video with slide synchronization on InfoQ.com! https://www.infoq.com/presentations/ ci-cd-ml/
- 3. Purpose of QCon - to empower software development by facilitating the spread of knowledge and innovation Strategy - practitioner-driven conference designed for YOU: influencers of change and innovation in your teams - speakers and topics driving the evolution and innovation - connecting and catalyzing the influencers and innovators Highlights - attended by more than 12,000 delegates since 2007 - held in 9 cities worldwide Presented at QCon San Francisco www.qconsf.com
- 4. MLOps CI/CD for Machine Learning SASHA ROSENBAUM
- 5. Sasha Rosenbaum Sr. Program Manager @DivineOps
- 7. Agenda §Machine Learning 101 §ML CI/CD Pipeline Overview §Potential implementation §Demo
- 9. What is MLOps and WHY should you care?
- 10. Machine Learning (ML) Is the science of getting computers to act Without being explicitly programmed
- 12. Machine Learning vs DevOps Google searches
- 14. OK, but why should YOU care?
- 15. 13
- 16. Data Scientists just want to Data Science
- 17. Deep Learning Some ML training algorithms are complex
- 18. Deep Learning - Backpropagation Some ML training algorithms are complex
- 20. We’ve got the notebook into source control!
- 26. How do we put the model in production?
- 27. What is an ML model?
- 28. Linear Regression – Housing Prices The training finds a and b such that Y = a+bX+ϵ
- 29. Deep Learning The input and output may be vectors !𝑋, !𝑌
- 31. ML Model A definition of the mathematical formula with a number of parameters that are learned from the data
- 32. Isn’t this just an API endpoint?!
- 33. Do models really change that often?
- 34. Models must be improved continuously
- 44. The model predictions depend on what it has “seen”
- 45. => Dataset is part of the model version!
- 46. FBLearner FlowTensorFlow Extended Uber’s Michelangelo Microsoft Aether
- 47. But I don’t work at a big company with thousands of ML engineers!
- 48. How do we iterate?
- 50. DevOps/SREData Scientist • Quick iteration • Versioning • Reuse • Great tools • Ease of management • Unlimited scale • Eliminating drift • Quick iteration • Versioning • Reuse • Compliance • Observability • Uptime • UpdatesFriends?
- 51. App developer MLOps Workflow Build appCollaborate Test app Release app Monitor app Model reproducibility Model retrainingModel deploymentModel validation Data scientist
- 52. Code, dataset, and environment versioning Model reproducibility Model retrainingModel deploymentModel validation Build appCollaborate Test app Release app Monitor app MLOps Workflow App developer Data scientist
- 53. MLOps Workflow Model reproducibility Model retrainingModel deploymentModel validation Automated ML ML Pipelines Hyperparameter tuning Train model Build appCollaborate Test app Release app Monitor app App developer Data scientist
- 54. MLOps Workflow Model validation & certification Model reproducibility Model retrainingModel deploymentModel validation Train model Validate model Build appCollaborate Test app Release app Monitor app App developer Data scientist
- 55. MLOps Workflow Model packaging Simple deployment Model reproducibility Model retrainingModel deploymentModel validation Train model Validate model Deploy model Build appCollaborate Test app Release app Monitor app App developer Data scientist
- 56. MLOps Workflow Model management & monitoring Model performance analysis Model reproducibility Model retrainingModel deploymentModel validation Train model Validate model Deploy model Monitor model Retrain model Build appCollaborate Test app Release app Monitor app App developer Data scientist
- 57. Build Your Own MLOps Platform + +
- 58. ML Pipeline A reusable, scaleable ML workflow template Kubeflow pipeline A reusable, scalable ML workflow template that runs on containers
- 59. Azure ML • Prep data • Train • Test • Deploy • Manage
- 60. Demo
- 61. Even a simple CI/CD pipeline is better than none!
- 62. DevOps Because change is the only constant in life
- 63. AI Ethics
- 64. Bias is a property of information
- 65. We must build AI responsibly
- 68. Questions?
- 69. Resources GitHub repo https://www.kubeflow.org/docs/azure/azureendtoend/ Deploy Kubeflow on Azure https://www.kubeflow.org/docs/azure/deploy/install-kubeflow/ Example Kubeflow Azure Pipeline https://www.kubeflow.org/docs/azure/azureendtoend/ Release pipeline https://dev.azure.com/sasrose/kubeflow/_release
- 70. Watch the video with slide synchronization on InfoQ.com! https://www.infoq.com/presentations/ ci-cd-ml/