open_spiel
open_spiel copied to clipboard
OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games.
This PR would bring bindings for the infostate tree class to the python module pyspiel. The following classes would be bound by this PR: InfostateTree InfostateTreeNode InfostateNodeType (enum) TreeplexVector< double...
Hi, I have one question about the usages of Dirichlet noises used in AlphaZero that I have been confused for a while. I understand it is used for exploration at...
Hi all. I've added the implementation of a game called Squadro. It's a two-player, perfect information game. I believe it has some similarities with Quoridor, which is already present in...
Hi, I met few problems when I try to compile with librtorch. I can use this Open Spiel without libtorch correctly, but when I: 1. Turn on 'BUILD_WITH_LIBTORCH' and 'BUILD_WITH_LIBNOP'...
Now can get/set scores and set a bonus reward for winning/losing overall game to returns.
The `player_id` argument is not need, but still needs to be present in order to implement `Policy`. For example without it PolicyBot breaks as it passes three arguments (self, the...
I am training RNaDSolver on Kuhn Poker with this config: ``` config = rnad.RNaDConfig( game_name="kuhn_poker", trajectory_max=10, state_representation=rnad.StateRepresentation.INFO_SET, policy_network_layers=(128,), batch_size=256, learning_rate=1e-6, seed=42 ) ``` 2 things I cannot explain: 1. There...
Hi there, I'm trying to make a new implementation of Spades and am looking for some help along the way as I'm pretty new to RL. Ultimately, I'm hoping to...
I noticed that when downloading the OpenSpiel package using pip, the code obtained is inconsistent with the one on GitHub. For example, the code downloaded via pip does not include...