Microsoft Developer (featuring Bethany Jepchumba) explains supervised fine-tuning in Microsoft Foundry, when to use fine-tuning, and how to review checkpoints, logs, and evaluation metrics, with a simple tone fine-tuning demo run directly in Foundry.

Model Optimization in Microsoft Foundry: Supervised Fine-Tuning

This video covers the fundamentals of model optimization using Microsoft Foundry (Azure AI Foundry), focusing on supervised fine-tuning: what it is, when it makes sense, and what to look at when evaluating a fine-tuning job.

Agenda and timestamps

• 00:03 Welcome and scenario
• 01:23 What is fine-tuning and when to use it
• 01:57 Fine-tuning options in Microsoft Foundry
• 02:15 Demo: supervised fine-tuning
• 05:23 Showcase: checkpoints, logs, and metrics

What fine-tuning is (and when to use it)

The session explains:

• What fine-tuning means in practice
• Common scenarios for deciding why and when to fine-tune (vs. alternatives)

Fine-tuning options in Microsoft Foundry

A walkthrough of the different fine-tuning options available in Microsoft Foundry.

Demo: supervised fine-tuning (tone fine-tuning)

A simple, practical demo of running a tone fine-tuning job directly in Foundry.

Evaluating your fine-tuning job

The video closes with what to inspect to understand how your job is performing:

• Checkpoints
• Logs
• Essential evaluation metrics

Links

• Microsoft Foundry (fine-tuning): https://aka.ms/foundry-ft
• Foundry fine-tuning demos on GitHub: https://aka.ms/ft-demos

Presenter links

• Bethany Jepchumba on X: https://twitter.com/bethanyjep
• Bethany Jepchumba on LinkedIn: https://www.linkedin.com/in/bethany-jep/
• Bethany Jepchumba on GitHub: https://github.com/bethanyjep