openai / maddpg

Code for the MADDPG algorithm from the paper "Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments"
https://arxiv.org/pdf/1706.02275.pdf
MIT License
1.59k stars 484 forks source link

A issue about the saving the model in train.py #69

Open chengdusunny opened 2 years ago

chengdusunny commented 2 years ago

Hello, I am very interested in this project. However, although the work is almost perfect, I think there still be some amendment to this. Please refer to the line in train.py: if terminal and (len(episode_rewards) % arglist.save_rate == 0): I think your intention is that if the current episode has ended and the number of episodes is divisible by arglist.save_rate, then save the tained model and display the information. However, terminal should be replaced by (terminal or done), since the episode can also be ended if the agents are done.