MARLlib icon indicating copy to clipboard operation
MARLlib copied to clipboard

Published 20 hours ago • verified publisher icon Replicable-MARL

Backpropagation through time for PPO

Open fulacse opened this issue 1 year ago • 1 comments

PPO + LSTM have a extral hyperparameter what is bptt horizon. Is possible I set up it?

fulacse avatar Jan 06 '24 16:01 fulacse

Seems like in RLlib version 1.8.0, there is no native support for configuring the BPTT when using an RNN structure as the model base. Maybe consider trying the latest version of Ray for potential updates on this?

Theohhhu avatar Jan 18 '24 00:01 Theohhhu