Introduce FuseML to wider community, educate about its purpose and architecture.
Welcome, state the goals of the meeting and agenda (very briefly)
We could include same Demo Stefan/Alessandro already offered in different meetings to get people on the same page.
Quick demonstration of the installer:
- mention that it is a fork of carrier project (meanwhile renamed to Epinio)
- explain that it needs Kubernetes cluster with specific setup
- list the components parts it is installing
- show the installation
(Unless it was already part of initial presentation point)
- Client-server architecture
- How it fits into MLOps mindset
- dependency on other components (mentioned by installer)
- show current components (more details to follow)
- "sets of code", basically the Data Scientist's project
- currently implemented as git(ea) repositories
- show commands live:
- create a codeset (+ show gitea web UI too)
- list codesets
- delete codeset
- the result of the workflow/pipeline
- show commands live
- show list of applications
- test prediction service using the output URL from application
- (show the web app?)
- delete an application
- This is the main component, driving MLOps work routine
- explain DSL, describe example yaml file
- currently implemented as Tekton pipelines
- explain relation with
- codesets
- workflow runs
- applications
- show commands live
- create workflow
- list codesets and assign workflow to codesets
- show workflow runs (+ show Tekton dashboard)
- it is possible (though unlikely) some people will be there
-
point to
- github repository
- issues for feedback
- Slack channel