search

eleurent / rl-agents

Implementations of Reinforcement Learning and Planning algorithms
MIT License
591 stars 153 forks source link

The problem of reward performance #115

Open Louise599 opened 6 months ago

Louise599 commented 6 months ago

Hello, thank you very much for open sourcing such a great project. I am running the code: python experiments.py evaluate configs/IntersectionEnv/env.json using the command \ configs/IntersectionEnv/agents/DQNAgent/baseline.json \ --train --episodes=4000 --name-from-config, the reward graph I get is unstable. I hope to get your help, thanks a lot! 屏幕截图 2024-05-08 111045

kongxincaizi commented 5 months ago

hi，I also encountered the same problem. In tensorboard, all of my curves did not converge. Have you solved this problem now？

kongxincaizi commented 5 months ago

Perhaps I have found a solution. Adjust the smoothness index in tensorboard