Cogment: Open Source Framework For Distributed Multi-actor Training, Deployment & Operations
Authors:
AI Redefined,
Sai Krishna Gottipati,
Sagar Kurandwad,
Clodéric Mars,
Gregory Szriftgiser,
François Chabot
Abstract:
Involving humans directly for the benefit of AI agents' training is getting traction thanks to several advances in reinforcement learning and human-in-the-loop learning. Humans can provide rewards to the agent, demonstrate tasks, design a curriculum, or act in the environment, but these benefits also come with architectural, functional design and engineering complexities. We present Cogment, a uni…
▽ More
Involving humans directly for the benefit of AI agents' training is getting traction thanks to several advances in reinforcement learning and human-in-the-loop learning. Humans can provide rewards to the agent, demonstrate tasks, design a curriculum, or act in the environment, but these benefits also come with architectural, functional design and engineering complexities. We present Cogment, a unifying open-source framework that introduces an actor formalism to support a variety of humans-agents collaboration typologies and training approaches. It is also scalable out of the box thanks to a distributed micro service architecture, and offers solutions to the aforementioned complexities.
△ Less
Submitted 21 June, 2021;
originally announced June 2021.
