• 🔗 link to the W&B project
• 🔗 link to the GitLab repository

This is the capstone project of the "Building a Reproducible Model Workflow" Udacity course within the "Machine Learning DevOps Engineer" nanodegree. Project instructions can be found in the NOTES.md file.

You are working for a property management company renting rooms and properties for short periods of time on various platforms. You need to estimate the typical price for a given property based on the price of similar properties. Your company receives new data in bulk every week. The model needs to be retrained with the same cadence, necessitating an end-to-end pipeline that can be reused.

In this project you will build such a pipeline.

I used a Taskfile to document all the tasks of this project in one YAML file, read more about Taskfiles here.
As follows, the list of tasks included:

• to execute a particular task, you need to run the following command: task <task_name>, e.g. task create_env
• to list all tasks, run task --list
• to get more details on a particular task, run task --summary <task_name>, e.g. task --summary create_env

• Before running the "Data Testing" step, make sure to add the reference tag on the clean_sample.csv W&B artifact.

• In order to find the best model,sort the runs by ascending MAE (Mean Absolute Error) and you'll find the best result at the top

• Now that we have identified the best model, click on its name, then go to its artifacts and select the "model_export" output artifact, and add a prod tag to it to mark it as "production ready".

• when your pipeline is ready for release, its artifacts graph should look like the following:

• once released, your releases page should be similar to the following:

  • GitLab

    

  • GitHub

    

note that I used GitLab to release the pipeline and pushed the same repo to both GitLab and GitHub. For some reason, GitLab's releases are shown as Tags in GitHub (not Releases).