In the evening session, Lily (Xiaoxuan) Liu, presented vLLM, the most popular high-performance library for LLM inference, with a deep dive on speculative decoding. More about the event: The CUDA Mode server met for the first-time for an IRL hackathon: 6 Keynotes from the First CUDA Mode IRL Hackathon: Tri Dao: : Supriya Rao: PyTorch Insights: Andrej Karpathy: Eureka Labs and llm.c: Lily (Xiaoxuan) Liu: vLLM: Tim Dettmers: The Power of Open Source: Wen-mei Hwu: How to Pick a Hard Problem: Accel: Accel is a global venture capital firm that is the first partner to exceptional teams everywhere, from inception through all phases of private company growth. Atlassian, Bumble, CrowdStrike, Fiverr, Flipkart, Freshworks, Qualtrics, Scale, Segment, Slack, Spotify, Squarespace, Tenable, and UiPath are among the companies Accel has backed over the past 40+ years. We help ambitious entrepreneurs build iconic global businesses. Connect With Us: Website: Linkedin: Twitter: #CUDAMode #GPUMode #hackathon #vllm #llm #opensource #opensourceai