DISCLAIMER:

• Please note that registration is needed to enter the event - We will not be able to grant access on the day without a ticket. Please show proof of ticket confirmation upon arrival.
• Please ensure you register with your full legal name - We are unable to accept registrations without surnames or name abbreviations.
• Due to Google's office security, we are unable to give +1 access.

Hands-On with Gemma 3: What's New & 5 Ways to Deploy

Unlock the potential of Google's family of open models, in our focused 2-hour workshop on Gemma 3. Go beyond the basics to understand the key advancements and learn practical techniques for running the model both locally and on server infrastructure.

In the first part, we'll dissect what makes Gemma 3 a significant leap forward from Gemma 2. Discover the versatility offered by the new model sizes (1B, 4B, 12B, 27B). We'll explore game-changing capabilities like the extended 128k context window, integrated multimodal (vision) processing for richer applications, and enhanced multilingual support for broader reach.

Transition from theory to practice in the second part: '5 Ways to Deploy Gemma 3.' This hands-on session provides actionable insights and step-by-step guidance across diverse deployment scenarios:

1. Accessing Gemma through AI Studio.
2. Local execution using Ollama.
3. Deployment on Cloud Run with vLLM.
4. Mobile deployment using MediaPipe and LiteRT.
5. In-browser/web implementation using Transformers.js.

Follow along live on your own laptop (some steps may require specific software to be installed), or take detailed notes and utilise our online guides at your own pace!

This workshop is tailored for developers, data scientists, and technical professionals aiming to integrate Gemma 3 into their workflows. You'll leave with a clear understanding of Gemma 3's architecture and the practical skills needed to deploy it effectively across various platforms, from local machines to the cloud and edge devices.