Short promotional/transcript excerpt for a new short course introducing Multimodal Llama 3.2, produced in partnership with Meta and taught by Amit Sanghani. The course covers Llama 3.2’s multimodal text+image capabilities, four new model sizes (1B, 3B, 11B, 90B), the Llama Stack orchestration layer, built-in and user-defined tool-calling (including an IPython role), and hands-on lessons for vision and on-device use cases.
Questions this video answers
Amit Sanghani (director of pun engineering for Meta's Llama team).
Llama 3.2.
1B, 3B, 11B, and 90B.
The 11B and 90B models (vision-enabled).
The smaller 1B and 3B models.
Chapters
I'm thrilled to announce our new course introducing Multimodal Llama 3.2, built in partnership with Meta and taught by Amit Sanghani, who's the director of Pun engineering for Meta's Llama team.
Very glad to be here with you Andrew and to teach this course.
The Llama open family of models has been a game changer for AI and Llama 3.2 released new multimodal capabilities so you can understand both text and images. This course will show you how you can take advantage of the Llama model as well as Meta's new Llama Stack, which is an open source orchestration layer for building on top of the Llama family of models. The Llama models are open so anyone can download, customize, fine tune and build new applications on top of them. Meta's been a leader in developing open models, which is why many developers have found them incredibly useful.
That's right. At Meta, we believe that openness drives innovation and is the right path forward. That is why we continue to share our research and collaborate with our partners and the developer community.
Llamas are already powering many applications, for example, including these Meta Ray Bans that I'm wearing. Hey Meta, tell me a joke about llamas. Why did the llama refuse to play poker? Because it always got fleeced. It almost feels like it was just yesterday that we were here together presenting a short course on llama 2 and now you have llama 3.2 out. What can learners look forward to seeing in this course?
This course will introduce you to the new models and features in 3.2 release and allow you to try them out in building many interesting and cool use cases in different lessons. In this release we have provided four new models. With the 11B and 90B models we have added vision capabilities and to support on device applications we have provided the smaller 1B and 3B models in the lessons. In this course you will explore the vision capabilities of the 11B and 90B models ranging from image classification to combining the vision and tool use. You might have heard of the user and assistant roles in LLMs. Llama introduces new roles such as the IPython role. This is a tool calling role giving the Llama family built in and user defined tool calling capabilities. And you will explore the Llama Stack API. This is a standardized interface for common tool chain components like fine tuning or synthetic data generation to customize Llama models and build agentic applications.
All that sounds fantastic. So I hope you join us for this course and I don't know Ahmed, if the model family keeps growing, we might need a bigger desk.
