open_spiel
open_spiel copied to clipboard
google-deepmind
OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games.
I'm trying to build open spiel on windows 10. I've followed the installation instructions for windows, but I'm getting an error importing pyspiel when running example.py. $ python example.py Traceback...
Dear OpenSpiel community, I added a little game implementation for the board game called Stratego, a large game with imperfect information and deterministic actions. The game implementation is called Yorktown...
Hi, I am trying to create a good Ai for the backgammon game, as there are no examples for it. I tried pure PPO, but it did not give very...
The AlphaZero documentation states that the checkpoint are compatible between the C++ and Python versions. Does this still hold true for LibTorch AlphaZero? I could not find a way to...
``` printing to: cfrplus_solver_19.32756_exploitability.pickle Loading the model... OpenSpiel exception: Unknown parameter 'gamedef=GAMEDEFlimitnumPlayers = 2numRounds = 3blind = 5 10raiseSize = 10 10 20firstPlayer = 1maxRaises = 2 2 3numSuits =...
I was interested in implementing and submitting a PR for these two games: * No Thanks! — VERY simple mechanics. * Mamono Sweeper — a basic extension of Minesweeper, where...
Fix AlphaZero to expect the observation from the current player's perspective and correspondingly estimate the value function from the perspective of the current player, as discussed in the supplementary material...
I just implement the first version for armac. In fact, I am still not sure about its correctness. I hope anybody could help me review these codes. There are two...
Adding new game Parcheesi which is a variant of Ludo with two dices. The game is a good example for a perfect information game with chance element that is played...
I've been using this locally while hacking with backgammon. Lemme know if it's worth merging into the repo. Logs showing the state values of a trajectory and the training values...
Metadata
Owner
Metadata
OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games.