Ripped Machine Learning with Distributed Computing
summary
Shows how BentoML packages a scikit‑learn iris model into a Docker image, auto‑generates a Dockerfile and REST/Swagger endpoint, and pushes the image to a private registry. Builds use multi‑stage Docker to reduce bloat but the example still produces a 1.6 GB image that is pushed and deployed to a Kubernetes cluster where the deployment runs three replicas successfully and serves predictions; adaptive microbatching, probes and metrics are included.
AI-written summary · about
chapters (7)
- 0:00Introduction and overview
- 3:12Polyglot and poly-model world
- 7:03Distillation pattern for containers
- 14:30Kubernetes as distributed OS
- 24:28Introducing Bento ML framework
- 33:25Hands-on Bento ML demo
- 53:49Summary and Q&A
related talks
24mDocker for Machine Learning!
24mBuilding an ML Application Platform from the Ground Up
41musing-django-docker-and-scikit-learn-to-bootstrap-your-machine-learning-project/
16mThe quandary of containerized machine learning
30mBuild consistent and portable ML environments with containers
46mAccelerating Machine Learning DevOps with Kubeflow
35mAccelerate deep learning in the cloud with custom ML environments
24mFrom Model to Microservice - Machine Learning at Scale