A starting template for continuous ML projects where data kept growing, models kept learning and services deployed continuously.

You can use this template for:

• Initiate your ML projects
• Set .dvc to track data & save models
• Set cml to automate training and track experiment by pull requests
• Set model serving to endpoint and frontend example then deploy your models continuously

My Workflow

Motivation

! Mention:

• model stuck in research
• model didn't get to production, repetitive task on modelling
• machine assignment for training, mention cheaphead machine that burns when using their own machine for training
• experiment tracking for reliability
• data and models that grows time to time that needs its own version control

Had to have a reliable getting started template for machine learning project since most of project that had data version control, experiment management, deployment as a service, serving to sample frontend and collaboration with a yaml hyperparameter tuning file to ease experiment tracking.

Deliver actions diagrams, how it works specifically and how it would help with automating the project on specific events on repository.

Using github script, definition, and how it compares to another framework

Submission Category:

DIY Deployments/Wacky Wildcards

Yaml File or Link to Code

Additional Resources / Info

[Note:] # (Be sure to link to any open source projects that are using your workflow!)

• https://github.com/sanggusti/write-docker-actions
• https://github.com/iterative/cml_dvc_case
• https://dvc.org/doc/start/data-and-model-versioning
• https://lab.github.com/

[Reminder]: # (Submissions are due on December 8th, 2021 (11:59 PM PT or 2 AM ET/6 AM UTC on December 9th).