https://github.com/indix/mlflow-gocd

GoCD plugins to work with MLFlow as model repository in a CD flow
https://github.com/indix/mlflow-gocd

continuous-delivery gocd machine-learning mlflow

Last synced: 2 months ago
JSON representation

GoCD plugins to work with MLFlow as model repository in a CD flow

• Host: GitHub
• URL: https://github.com/indix/mlflow-gocd
• Owner: indix
• License: apache-2.0
• Created: 2018-11-09T10:30:02.000Z (about 6 years ago)
• Default Branch: master
• Last Pushed: 2023-11-01T11:06:46.000Z (about 1 year ago)
• Last Synced: 2024-07-30T17:52:49.035Z (6 months ago)
• Topics: continuous-delivery, gocd, machine-learning, mlflow
• Language: Java
• Homepage:
• Size: 572 KB
• Stars: 29
• Watchers: 8
• Forks: 5
• Open Issues: 1
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# mlflow-gocd

[![Build Status](https://travis-ci.org/indix/mlflow-gocd.svg?branch=master)](https://travis-ci.org/indix/mlflow-gocd)







GoCD plugins to work with MLFlow as model repository.

The plugin works with a process where runs within an experiment are "promoted" for production use. A new build is triggered for each promoted run in an experiment and exposes the `artifact_uri` as an environment variable to the build. 

The optional fetch model plugin utilizes the artifacts uri and the given artifact pattern / name to download the model(s) (from S3 as of now) to the desired destination during the build.

Tested on `GoCD 17.2.0+` and `MLFlow 0.7.0`

## Configuring the repository







## Configuring experiments as packages







## Build comment with trackback url







## Fetching models/artifacts from mlflow

Optionally the fetch plugin can also be used in conjunction with the package plugin to fetch artifacts stored in mlflow (backed by S3.)