Getting Started with Gemma (By: Haleema Tallat) - Google I/O Extended 2024

Talk by Haleema Tallat (https://www.linkedin.com/in/haleema-tallat/) at Google I/O Extended 2024 by GDG Lahore.

Transcript

1. Gemma 1 Haleema Tallat
2. Agenda Introduction Gemma Colab Coding: • Getting Started Gemma with KerasNLP
3. 你好
4. Hello!
5. Introduction
6. Generative AI
7. Generative AI
8. Google is the industry pioneer in AI 2015 Google DeepMind AlphaGo defeats Go champion 2016 Google’ s DeepMi nd helps detect eye disease 2017 Google invents Transfor mer kickstarti ng LLM revolutio n 2018 Google’s groundbr eaking large language model, BERT 2019 Text-to-T ext Transfer Transfor mer LLM 10B P Model Open Sourced 2020 Google LaMDA Model Trained to converse 2022 AlphaFold predicts structures of all known proteins 2023 A conversa tional AI Service powered by PaLM2 2024 Family of multimodal LLMs & products
9. “Gemini is Google’s latest multimodality LLM model - reasoning seamlessly across text, images, video, audio and code.”
10. Gemini comes in three model sizes
11. Any Open Source LLM Model?
12. Gemma
13. “Gemma is a family of lightweight, state-of-the-art, open LLM models built from the same research and technology used to create the Gemini models.”
14. Open LLM Models Benchmark
15. Open LLM Models Benchmark
16. Gemma Model Variants
17. CodeGemma Model CodeGemma: Code completion, generation, and chat for developers and businessesHarnessing the foundation of our Gemma models, CodeGemma brings powerful yet lightweight coding capabilities to the community. CodeGemma models are available as a 7B pretrained variant that specializes in code completion and code generation tasks, a 7B instruction-tuned variant for code chat and instruction-following, and a 2B pretrained variant for fast code completion that fits on your local computer. CodeGemma models have several advantages: • Intelligent code completion and generation: Complete lines, functions, and even generate entire blocks of code – whether you're working locally or leveraging cloud resources. • Enhanced accuracy: Trained on 500 billion tokens of primarily English language data from web documents, mathematics, and code, CodeGemma models generate code that's not only more syntactically correct but also semantically meaningful, helping reduce errors and debugging time. • Multi-language proficiency: Your invaluable coding assistant for Python, JavaScript, Java, and other popular languages. • Streamlined workflows: Integrate a CodeGemma model into your development environment to write less boilerplate, and focus on interesting and differentiated code that matters – faster. This table compares the performance of CodeGemma with other similar models on both single and multi-line code completion tasks. Learn more in the technical report.
18. RecurrentGemma Model RecurrentGemma: Efficient, faster inference at higher batch sizes for researchers RecurrentGemma is a technically distinct model that leverages recurrent neural networks and local attention to improve memory efficiency. While achieving similar benchmark score performance to the Gemma 2B model, RecurrentGemma's unique architecture results in several advantages: • Reduced memory usage: Lower memory requirements allow for the generation of longer samples on devices with limited memory, such as single GPUs or CPUs. • Higher throughput: Because of its reduced memory usage, RecurrentGemma can perform inference at significantly higher batch sizes, thus generating substantially more tokens per second (especially when generating long sequences). • Research innovation: RecurrentGemma showcases a non-transformer model that achieves high performance, highlighting advancements in deep learning research. This table compares the performance of CodeGemma with other similar models on both single and multi-line code completion tasks. Learn more in the technical report.
19. Keras NLP
20. KerasNLP is a simple and powerful API for building Natural Language Processing (NLP) models within the Keras ecosystem.
22. Access Gemma Model
23. Access Gemma Model
24. Access Gemma Model in Kaggle Get started
25. Access Gemma Model in Kaggle
26. Access Gemma Model in Kaggle
27. Access Gemma Model in Kaggle
28. Access Gemma Model in Kaggle
29. Getting Started Gemma with KerasNLP Colab Coding
30. • Configure Kaggle Username and Key in notebook Getting Started Gemma with KerasNLP
31. • Install library dependencies ◦ Keras ◦ KerasNLP Getting Started Gemma with KerasNLP
32. • Select Keras backend • Import Keras packages Getting Started Gemma with KerasNLP
33. • Create a model using GemmaCausalLM with from_preset method Getting Started Gemma with KerasNLP
34. • Gemma model summary Getting Started Gemma with KerasNLP
35. • Let’s ask Gemma a question! Getting Started Gemma with KerasNLP What is Large Language Model (LLM)? A LLM is a type of neural network that can understand and generate human language. It has been trained on vast quantities of text data and is able to produce human-like written and spoken language.
36. • Let’s ask Gemma a question! Getting Started Gemma with KerasNLP What is Large Language Model (LLM)? A LLM is a type of neural network that can understand and generate human language. It has been trained on vast quantities of text data and is able to produce human-like written and spoken language.
37. • Let’s ask Gemma a question! Getting Started Gemma with KerasNLP How does the Large Language Model (LLM) work? The LLM is based on the Transformer architecture. The LLM consists of multiple Transformer blocks. A Transformer block is made up of two sub-blocks: a multi-head self-attention block, and a feed-forward block.
38. • Batched prompts using a list as input! Getting Started Gemma with KerasNLP What is Large Language Model (LLM)? Large Language Models (LLMs) are a type of artificial intelligence (AI) that can generate human-like text. They are trained on vast amounts of text data, such as books, articles, and the internet, and are able to understand and respond to natural', 'How does the Large Language Model (LLM) work?\n\nThe LLM is a large language model that is trained on a large corpus of text data. It is a type of artificial intelligence (AI) that is capable of understanding and generating human-like text.
39. Q&A
40. Thank you.