Large language models seem to be the future of natural language processing, yet building them is far from trivial. Not only do you have to deal with the complexity of building and cleaning datasets composed of tens-to-hundreds of billions of tokens, but you also have to establish capability to efficiently train on large GPU-accelerated systems like DGX SuperPOD.
Therefore, only a handful of people across the globe has practical experience building them and deploying to production. This panel gathers leaders in this field to discuss the key technical challenges that had to be overcome to make those models a reality.
Magnus Sahlgren, Head of Natural Language Processing, Research Institutes of Sweden (RISE)
Mohammad Shoeybi, Head of NLP, Applied Deep Learning Research, NVIDIA
Zenodia Charpy, Senior Deep Learning Solution Architect, NVIDIA
Marta Villegas, Head of Text Mining Unit, Barcelona Supercomputing Center
Thomas Wolf, Chief Science Officer, HuggingFace
Yonghui Wu, Director of Natural Language Processing, University of Florida