Introducing Multimodal Llama 3.2

What You'll Learn

• Explore the features of the Llama 3.2 model, from image classification to vision reasoning and tool usage.
• Learn prompting techniques, tokenization, and how to use built-in and custom tool calling.
• Gain knowledge of the Llama Stack for building AI applications using a standardized interface.

About This Course

In this course, taught by Amit Sangani, you’ll dive into Meta’s Llama 3.2 model family, focusing on multimodal capabilities, the new Llama Stack, and advanced tool-calling features. Learn to leverage Llama 3.2's capabilities for diverse applications by building on the open Llama model architecture and Llama Stack orchestration layer.

What You’ll Do

• Understand the new model family, training processes, and features within Llama 3.2.
• Work on multimodal use cases, including image reasoning tasks like analyzing dashboards and summing totals on receipts.
• Learn about role-based prompting, and the format that identifies different roles in the Llama models.
• Explore Llama’s tokenization capabilities with a 128k vocabulary size and support for multiple languages.
• Use built-in and custom tool-calling capabilities with practical examples, such as web search and equation-solving.
• Discover Llama Stack API for customization and building agentic applications on the Llama model family.

Course Outline

• Introduction: Overview of course objectives and Llama 3.2 model capabilities.
• Overview of Llama 3.2: Learn about the new models and training enhancements.
• Multimodal Prompting: Advanced techniques for multimodal prompting with Llama.
• Multimodal Use Cases: Real-world examples like dashboard analysis and grading math homework.
• Prompt Format: Explore the prompt format for different roles in the Llama models.
• Tokenization: Understanding Llama’s tokenizer and expanded vocabulary support.
• Tool Calling: Implement both built-in and custom tool calling for various tasks.
• Llama Stack: Introduction to the Llama Stack API and its components for building applications.
• Conclusion: Recap and next steps for building on Llama 3.2.
• Appendix – Tips and Help: Additional resources and code examples.

Who Should Join?

This course is suitable for anyone with basic Python knowledge who wants to learn to build on the Llama model family and use Llama Stack for custom applications.