2023 Agenda

Thomas Scialom delves deep into Llama 2 and explores the world of Reinforcement Learning from Human Feedback.

In this talk, you'll gain insights into overcoming technology constraints specific to various data types, and explore the legal intricacies of intellectual property. Moderator:Jérôme Rastit, Head of Ethical Hacking, Free Pro

Discover the key role of benchmarks in guiding your ML infrastructure choices and how it continually drives innovations to overcome challenges.

Uncover real-world applications, technical challenges, and exciting announcements! Moderate: Adrienne Jan, CPO, Scaleway

Understand the practical impact of their architecture choices for their groundbreaking Mistral 7B model.

Explore innovative solutions, from hardware to software, aiming to optimize inference! Moderator: Geoffrey Cambefort, CEO, Hiremy.ai

Unravel the regulatory frameworks governing artificial intelligence. Stick around for a 15-minute Q&A. Moderator: Daphné Leprince-Ringuet, French tech reporter

Explore data sensitivity, market dynamics and business efforts. Moderator: Benoit Berthelot, Bloomberg

From POC to supporting millions of doctor consultations

Uncover how next-generation technology could address the surging demand for AI compute while prioritizing accessibility and environmental impact. Join us for a glimpse into the AI landscape of tomorrow and beyond. Moderator: Albane Bruyas, COO, Scaleway

Closed Session: Personal Invitations Required! The power and cost inefficiency of large-scale AI deployments on the most frequently chosen hardware platforms heavily impact the end user's ability to achieve desirable ROI. In response to this issue, Ampere created a new class of GPU-Free AI inference that delivers the best price/performance compared to both the GPUs themselves and the legacy architecture x86 processors. Ampere Cloud Native processors are optimized to run AI inference meeting all the performance needs as well as saving both energy and space in the data center. Join Ampere’s expert to discover how and have a glimpse at the performance these new chips can deliver!

​Closed Session: Personal Invitations Required! NVIDIA TensorRT-LLM, which will be part of NVIDIA AI Enterprise, is an open-source software that delivers state-of-the-art performance for LLM serving using NVIDIA GPUs. It consists of the TensorRT deep learning compiler and includes optimized kernels, pre- and post-processing steps, and multi-GPU/multi-node communication primitives.​ During this session, we will present TensorRT-LLM features and capabilities and we will walk you through the steps needed to build and run your model in TensorRT-LLM on both single GPU and multi-GPUs. We will also use TRT-LLM backend and Triton Inference Server for deployment.​

Closed Session: Personal Invitations Required! The power and cost inefficiency of large-scale AI deployments on the most frequently chosen hardware platforms heavily impact the end user's ability to achieve desirable ROI. In response to this issue, Ampere created a new class of GPU-Free AI inference that delivers the best price/performance compared to both the GPUs themselves and the legacy architecture x86 processors. Ampere Cloud Native processors are optimized to run AI inference meeting all the performance needs as well as saving both energy and space in the data center. Join Ampere’s expert to discover how and have a glimpse at the performance these new chips can deliver!

Closed Session: Personal Invitations Required! The NVIDIA H100 Tensor Core GPU features the fourth-generation Tensor Cores and also the new Transformer Engine with FP8 precision, which provides a great boost for training and inference over the prior generation of LLM models. During this session, we will review the latest features and capabilities of the NVIDIA H100 PCIe GPU and explore various techniques to best take advantage of its performance.