Seoyeon Park’s Post

🚀 Day 2 of #30DaysOfFLCode !  👉 30DaysOfFLCode.com Federated Learning isn't just a buzzword – it's a game-changing approach tackling privacy, scalability, and heterogeneity head-on. Today, I dove deep into the core challenges (spoiler: communication costs and non-IID data are just the tip of the iceberg) and how they shape the future of decentralized AI. 🤖✨ Curious to know how we can address these challenges? Check out the detailed breakdown in my GitHub repo: https://lnkd.in/e5hDUWra #FederatedLearning #MachineLearning #DataPrivacy #30DaysOfFLCode

🎉Hi, We participated in the Google Machine Learning Bootcamp 2024💻 and worked on the Gemma Sprint project! This project aims to develop an AI-powered system for the automatic classification of emergency calls using the advanced language understanding capabilities of the Gemma language model and the Whisper speech recognition model. By accurately categorizing calls based on the type of emergency (fire, medical, etc.), the system will enable faster and more targeted dispatch of emergency services, potentially saving lives. We have written the details in the tutorial. Model: https://lnkd.in/gqhhRN-c Tutorial code: https://lnkd.in/g3K4eURA Team Members 이현준, 최상헌 Thanks to the Google Korea and everyone involved in the MLB for making this great opportunity. #GemmaSplint #GoogleMLBootcamp

Discover how Adam uses EduProtocols. Also discover how Adam organizes and sequences content to make it accessible to all students.

Gemma 2 released! Google just released the next iteration of its open LLM! Gemma 2 comes in two sizes, 9B & 27B, trained on 13T tokens. Gemma 2 27B approaches Meta Llama 3 70B performance! First Chatbot Arena evals place Gemma2 27B around Anthropic Claude 3 Sonnet, Llama 3 70B, and OpenAI GPT-4. What's new with Gemma 2: 9B & 27B Instruction and base version with 8192 context window!! Trained on 13T tokens (27B) and 8T tokens (9B) NEW Sliding window attention, logit soft-capping and Grouped-Query Attention (GQA) 9B scores 71.3 MMLU; 52.8 AGIEval; 40.2 Human Eval 27B scores 75.2 MMLU; 55.1 AGIEval; 51.8 Human Eval Commercial use allowed!! Used SFT, Distillation, RLHF & Model Merging. Trained on Google TPUv5e Models: https://lnkd.in/eVW34g55

💡 This is an excellent free 1-hour course to help you get started with Llama 3.2 and make full use of its multimodal capabilities! Llama 3.2 is Meta's first multimodal LLM, packed with a bunch of new features. This hands-on intro course by DeepLearning.AI is perfect for understanding how to use it for your use cases! 📕 What's covered: - Details on how the Llama models are trained - Multimodal prompting for some real world use-cases - Tool calling and agentic workflows - Using the open-source Llama Stack API which is a standardized interface for building LLM applications. Link: https://lnkd.in/ee_VvJG6

Just finished the course “Advanced Prompt Engineering Techniques” by Morten Rand-Hendriksen! Check it out: https://lnkd.in/dUy4bPaE #artificialintelligence #promptengineering #generativeai

Talking to the AIs in natural language, doesn't mean it necessarily comes naturally... Check the course “Prompt Engineering: How to Talk to the AIs” by Xavier Amatriain I just finished for mote insights!  https://lnkd.in/dGD2jvbg #generativeai #largelanguagemodels #promptengineering.

Google DeepMind just unveiled Code Gemma, a groundbreaking collection of open code models. Gemma can now code! 🤯 Here's what you need to know: -- Code Gemma comes in 2B & 7B sizes, boasting an impressive 8192 context. -- These models are initialized from Gemma Base and further trained on a whopping 500B additional Tokens (web, code & math). -- Fine-tuned with SFT & RLHF for superior performance. -- The 2B model achieves 27% on HumanEval, while the 7B model surpasses expectations with a staggering 52%. -- Commercial use is allowed, making it perfect for industry applications. -- Optimized for on-device code completion, enhancing efficiency and productivity. -- You can access these cutting-edge models on Hugging Face. Get all the details: 🔗 Models: https://lnkd.in/gmsKgTWE 🔗 Google Blog: https://lnkd.in/gcFGufaf

Is Pedro Domingos’ The Master Algorithm still relevant today? I just revisited this fascinating book, where he explores the five tribes of AI: Symbolists, Connectionists, Evolutionaries, Bayesians, and Analogizers. It’s incredible to see how these paradigms shaped AI’s foundation and how they contrast with the LLM-driven boom we’re seeing now. If you’re interested in AI’s evolution, this book is a must-read! What do you think—are these five tribes still relevant in the age of generative AI? https://lnkd.in/e4GEVyeN

At Awsome LLC, we're fine-tuning large language models (LLMs) to make our energy utility talkbot smarter and more efficient. This article covers the essential math behind the fine-tuning process, breaking it down into simple terms.   Our team at Awsome LLC is dedicated to leveraging these techniques to set new standards in customer service and efficiency. Want to learn more about the process? Check out the article.