| |
|
EVENT DETAILS |
The Meetup will take place at GitHub's offices in San Francisco. Note that pre-registration is mandatory.
88 Colin P Kelly Jr St, San Francisco, CA 94107
Lessons Learned fine-tuning Llama2 for Autonomous Agents
In this talk, Rahul Parundekar, Founder of A.I. Hero does a deep dive into the practicalities & nuances of making LLMs more effective & efficient. He'll share hard-earned lessons from the trenches of LLMOps on Kubernetes, covering everything from the critical importance of data quality to the choice of fine-tuning techniques like LoRA & QLoRA. Rahul will share insights into the quirks of fine-tuning LLMs like Llama2, the need for looking beyond loss metrics & benchmarks for model performance, & the pivotal role of iterative improvement through user feedback - all learned through his work on fine-tuning an LLM for retrieval-augmented generation & autonomous agents. Whether you're a seasoned AI professional or just starting, this talk will equip you with the knowledge of when & why you should fine-tune, to the long-term strategies to push the boundaries of what's possible with LLMs, to building a performant framework on top of Kubernetes for fine-tuning at scale.
Speaker: Rahul Parundekar is the founder of A.I. Hero, a seasoned engineer, & architect with over 15 years of experience in AI development, focusing on Machine Learning & Large Language Model Operations (MLOps & LLMOps). AI Hero automates mundane enterprise tasks through agents, utilizing a framework for fine-tuning LLMs with both open & closed-source models to enhance agent autonomy.
Multi-Modal Visual Question Answering (VQA) using UForm tiny models with Milvus vector database
UForm is a multimodal AI library that will help you understand & search visual & textual content across various languages. UForm not only supports RAG chat use-cases, but is also capable of Visual Question Answering (VQA). Compact custom pre-trained transformer models can run anywhere from your server farm down to your laptop. I'll be giving a demo of RAG & VQA using Milvus vector database.
Speaker: Christy Bergman is a passionate Developer Advocate at Zilliz. She previously worked in distributed computing at Anyscale & as a Specialist AI/ML Solutions Architect at AWS.
Speaker: Ash Vardanian is the Founder of Unum Cloud. With a background in Astrophysics, his work today primarily lies in the intersection of Theoretical Computer Science, High-Performance Computing, & AI Systems Design.
Combining Hugging Face Transformer Models & Image Data with FiftyOne
Datasets & Models are the two pillars of modern machine learning, but connecting the two can be cumbersome & time-consuming. In this lightning talk, you will learn how the seamless integration between Hugging Face & FiftyOne simplifies this complexity, enabling more effective data-model co-development. By the end of the talk, you will be able to download & visualize datasets from the Hugging Face hub with FiftyOne, apply state-of-the-art transformer models directly to your data, & effortlessly share your datasets with others.
Speaker: Jacob Marks, PhD is a Machine Learning Engineer & Developer Evangelist at Voxel51, where he leads open source efforts in vector search, semantic search, & generative AI for the FiftyOne data-centric AI toolkit.
Prior to joining Voxel51, Jacob worked at Google X, Samsung Research, & Wolfram Research.
Strategies for Enhancing the Adoption of Open Source Libraries: A Case Study on Albumentations.ai
In this presentation, we explore key strategies for boosting the adoption of open-source libraries, using Albumentations.ai as a case study. We will cover the importance of community engagement, continuous innovation, & comprehensive documentation in driving a project's success. Through the lens of Albumentations.ai's growth, attendees will gain insights into effective practices for promoting their open source projects within the machine learning & broader developer communities.
Speaker: Vladimir Iglovikov, PhD is a co-creator of Albumentations.ai, a Kaggle Grandmaster.
|
|
|
|
|
|
