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.
Google colab containing preliminary implementation of ARMAC in Jax
Hi, I find it interesting to model prices from the point of view of marketing in an entrepreneurial setting. This is relevant because the product is sold on the basis...
As the title says, I'm wondering which game is the simplest (as in smallest, fastest to train on) for 2 players and perfect information. It looks like it should be...
[`best_response.BestResponsePolicy`](https://github.com/deepmind/open_spiel/blob/master/open_spiel/python/algorithms/best_response.py#L87) (which is being called by my code via [`exploitability.nash_conv`](https://github.com/deepmind/open_spiel/blob/master/open_spiel/python/algorithms/exploitability.py#L154)) does not work when policy probabilities are JAX abstract tracer values: ``` Traceback (most recent call last): [...] File "/usr/local/lib/python3.10/site-packages/open_spiel/python/algorithms/best_response.py",...
Update alpha_zero_torch C++ mlp model construction to match the python version.
…, which otherwise would not run. Also, added build commands in cmake for shared library in examples directory. Tested both Linux and Windows.
As per @richdewey we have added liar's poker to the games folder.
https://github.com/deepmind/open_spiel/blob/f4ae28033cc1f92ca4cc9f8bc87477ba9c8c4237/open_spiel/python/algorithms/policy_gradient.py Hi, The docstring states the critic learning rate is 0.001, but it really defaults to 0.01. It has the same issue in PPO, etc., I guess it's related to...
Hi, I am planning to contribute a JAX implementation of [Learning with Opponent-Learning Awareness" (Foerster et al.)](https://arxiv.org/abs/1709.04326) if no one currently works on it and if it is still appreciated...
Hi, here is Thorsten from Gravon (www.gravon.de). I implemented MAEDN (Mensch Ärgere Dich Nicht) in OpenSpiel. If you have any comments or would like to propose changes, you're very welcome....