<ul><li>An incentive system called Gauntlet has been developed for distributed deep learning of foundational models where peers are rewarded for contributions.</li><li>Gauntlet has been deployed on the bittensor blockchain and used to train a 1.2B LLM with completely permissionless contributions of pseudo-gradients.</li><li>The system can be applied to any synchronous distributed training scheme that relies on aggregating updates or pseudo-gradients.</li><li>The project involves a mechanism for filtering peer uptime and reliability, an OpenSkill rating system, and a unique computation mechanism to ensure peer contributions are distinct.</li></ul>

Incentivizing Permissionless Distributed Learning of LLMs

Discover more