Hello!
I did MultiAgent Reinforcement Learning using rllib, and as a result, result.json, progress.csv, etc. were created in /ray_results.
Among them, I want to know what each column in progress.csv means. Is there an explanation for this in ray.docs? I couldn’t find it. I especially want to know about each agent’s reward. I roughly understand that episode_reward is the sum of the rewards of all agents in the episode.
Hello
The progress.csv file in the /ray_results directory logs key metrics during MultiAgent Reinforcement Learning (MARL) training using RLlib. It includes values like episode_reward_mean, showing the average reward across all episodes, episode_reward_min for the lowest reward, and episode_reward_max for the highest. Other columns track total episodes completed (episodes_total) and total timesteps (timesteps_total). While it provides an overview of performance, individual agent rewards aren’t typically separated out in this file.
Hope it helps !
Thank you
gregbowers