Building AutoML Pipelines With Argo Workflows and Katib
Talkat Kubernetes AI Day Europe 2022 · on Monday, 2022-05-16· at 11:30· Pavilion 3, Room D | Level 2 | Central Forum·
summary
Shows a concrete Katib+Argo setup that runs each Katib trial as an Argo Workflow, with the trial template carrying a full Argo YAML, the metrics collector injected as a sidecar only into the training pod, and a reproducible example that runs 8 trials with 2 parallel workflows. Demonstrates a working two-step workflow (preprocess → train) where Argo passes parameters to training, metrics are written to the DB, Katib selects best learning rates, and the pipeline runs locally on kind or on-cluster.
AI-written summary · about
chapters (10)
- 0:00Start
- 0:28Automated Machine Learning Goal
- 1:03Landscape of AutoML
- 2:14What is Katib?
- 4:31Katib Architecture
- 6:26Argo Worldflows Ecosystem
- 7:17Using Argo Workflows in Katib
- 8:44Steps to integrate Argo Workflows to Katib
- 10:00Argo Workflow Example
- 20:54Join Argo Workflows and Katib Community
related talks
27mManaging Thousands of Automatic Machine Learning Experiments with...
10mScaling Time Series Analysis With Argo Workflows: Patterns and Practi...
38mMachine Learning Pipelines with K3s and Argo – Michael Lohan
25mLeveraging Argo Workflows and Argo Events for Cloud Agnostic Enter...
38mMachine Learning Pipelines with K3s and Argo – Sergio Mendez
36mMachine learning pipeline Kubeflow, let's start
39mBuild and Deploy Your Application with Argo on Kubernetes
36mAutomating Cloud-native Spark Jobs with Argo Workflows