Enabled World Economic Forum to provide an unique experience of large language translation models on Metaverse by setting up a 1440 NVIDIA A100 GPU based Cluster.
Context
WEF wanted to enrich the experience of Avatars from across the world while interacting in their own mother-tongues inside their Metaverse environment
The Metaverse avatars needed to render the translations in real-time without the need for human translators
Approach
Set up a large GPU cluster customised to run Large Language Models (LLMs) for 24 most used diverse languages from across the world
The LLMs would be constantly “learning” with new inputs, and plug into the Metaverse to deliver the actual user experience
Solution
Set up 180 servers housing 1440 NVIDIA GPUs in a cluster to meet real time translation requirements through an LLM
Storage and Infinitiband network integrated to run at 200Mbps