search

starry-sky6688 / MADDPG

Pytorch implementation of the MARL algorithm, MADDPG, which correspondings to the paper "Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environments".
537 stars 83 forks source link

run是学习的部分，evaluate是评估，那在哪里训练呢？ #5

Closed xzxgg closed 3 years ago

xzxgg commented 3 years ago

load模型重载的地方好像没用到，请问在哪用呢

xzxgg commented 3 years ago

另外为啥训练出来的model_spread里面只有两个智能体的模型？

starry-sky6688 commented 3 years ago

  1. maddpg.py 里， init()函数中会判断是否存在模型，如果存在则加载。

  2. 应该是model_spread中只有两个你需要控制的agent。