#31 AI Series: HuggingFace - N. Tazi

Join us for an insightful session featuring Nouamane Tazi, a Research Engineer at Hugging Face! 📈

Event Highlights:

• 🗣️ Topic: "Scaling LLM Training to Thousands of GPUs"
• ⏰ Duration: Approximately 45 minutes
• 🍕 Enjoy complimentary drinks and pizza while networking with fellow AI enthusiasts after the talk!

Important: Doors close by 7:15 PM, so arrive early! RSVP on Meetup to secure your spot.

Who Should Attend?

This event welcomes everyone passionate about cutting-edge AI research, especially:

• 🎓 Students
• 📖 PhD Candidates
• 💼 Academic Researchers
• 🔍 Industry Professionals

Abstract

This presentation delves into the complexities of training large language models at scale, addressing various system bottlenecks encountered with increased GPU count. Attendees will learn about:

• 🔄 Effective parallelism strategies (and their limitations)
• ⚙️ Maintaining efficient and stable training
• ⚖️ The engineering trade-offs essential for modern pretraining infrastructure

Nouamane aims to make this talk accessible to a broad ML audience, drawing on real-world experiences from Hugging Face’s large-scale open-source training.

Speaker Bio

Nouamane Tazi specializes in training and scaling large language models at Hugging Face. He is a co-author of SmolLM3 and The Ultra Scale Playbook, with research interests spanning NLP, deep learning, and scalable AI infrastructure.

We're BLISS e.V., the AI organization in Berlin connecting like-minded individuals! 🎉