steffen/marl-factory-grid
1
0
Fork 0
mirror of https://github.com/illiumst/marl-factory-grid.git synced 2025-06-18 18:52:52 +02:00
Code Issues Packages Projects Releases Wiki Activity

Results resolved. Small ADjustments

Browse Source
This commit is contained in:
Steffen Illium
2023-11-16 15:29:07 +01:00
parent cb76972a5f
commit c3c434a97e
17 changed files with 236 additions and 93 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
marl_factory_grid/algorithms/marl/mappo.py
View File

@ -46,7 +46,7 @@ class LoopMAPPO(LoopSNAC):
# monte carlo returns
mc_returns = self.monte_carlo_returns(batch[nms.REWARD], batch[nms.DONE], gamma)
mc_returns = (mc_returns - mc_returns.mean()) / (mc_returns.std() + 1e-8) # todo: norm across agent ok?
advantages = mc_returns - out[nms.CRITIC][:, :-1]
advantages = mc_returns - out[nms.CRITIC][:, :-1]
# policy loss
log_ap = torch.log_softmax(logits, -1)