Deploy your ML Workloads on Kubernetes with Kubeflow
summary
Kubeflow claims to make ML on Kubernetes simple, portable and scalable and shows concrete tooling: Kubeflow Pipelines, TFX components, kfctl install and KFServing for predict and explain. The demo wires ExampleGen, StatisticsGen, SchemaGen, Trainer, Evaluator, ModelValidator and Pusher into an end to end pipeline on ≈5,000 rows, compiles to a tarball, runs with green checkboxes, exposes tensorboard and a model lineage explorer, and pushes a blessed model to production.
AI-written summary · about
chapters (21)
- 0:00Introduction
- 1:40ML Challenges
- 4:53Data Science
- 8:20Core capabilities
- 11:35Components
- 12:10Installation
- 15:10Predict and Explain
- 17:18Pipelines
- 20:48Custom Components
- 23:54Component Implementation
- 25:39User Interface
- 27:29Standalone Kubernetes
- 28:17Pipeline Demo
- 29:27Pipeline Run
- 31:22Data View
- 32:14Schema Gem
- 33:01Example Validator
- 33:51Training
- 34:55Evaluation
- 35:49Continuous Deployment
- 36:59Artifacts
related talks
14mKubeflow: Where ML Meets Kubernetes
29mML Engineering Kubeflow
30mKubeflow, Your ML Tools on Kubernetes
46mDeploying scalable Machine Learning models with Kubernetes
51mMachine Learning using Kubeflow and Kubernetes
31mKubernetes for Machine Learning: Managing and Scaling AI Workloads
32mE2E ML Platform on Kubernetes with just a few clicks
54mScaling Machine Learning Pipelines in Cloud