25 Feb 17:21

lanctot

2f2e0e7

OpenSpiel 0.3.0

This release aims to address the problem that the new Python games were not compatible with the old version of OpenSpiel used by our pip package since it is too far behind (i.e. a fix to #503).

New Functionality

Games

Add Dark Hex
Add Kuhn (new Python game)
Tic-Tac-Toe (Python game, updated to new API)
Liar's Dice: new bidding variant, and configurable number of faces
Trade Comm: add info state string
Backgammon: expos action conversion functions available to Python

Algorithms

Deep CFR (PyTorch)
EVA (PyTorch)
Policy gradients (PyTorch)
Tabular Sarsa (C++)
Tabular Q-Learning (C++)
Sequence-form linear programming (C++)
Variance Reduction baselines (VR-MCCFR) in Python

Metrics

Distance to correlated and coarse-correlated equilibrium, for extensive-form games): CEDist and CCEDist (C++)

Examples

Poker "fold, call, pot, all-in" (FCPA) abstracted no-limit example (Python)
Generating multiple equilibria using CFR with random initial regrets and MCCFR

Process

Move from Travis to Github Actions for continuous integration (CI)
Update versions: TF, Jax, Julia

Misc.

Allow registration of observers
Distinction between perfect recall and imperfect recall example
Information state trees
JupyterLab environment (two Dockerfiles)
Allow Python games from C++, now fully compatible with main C++ API
Expose repr (Python)

Fixes and Documentation updates

Fix network construction in Exploitability Descent example
Fix GameParameters template compiler causing issues on older compilers
Fix Observation consistency in Trade Comm
Fix joining processes in Python AlphaZero
Fix HistoryString as identifier in HistoryTree
Improved documentation for the new observation API
Many other smaller fixes

Thanks

Thanks to DeepMind for continued support of development and maintenance of OpenSpiel.

Thanks to all of our contributors:

Core Team: https://github.com/deepmind/open_spiel/blob/master/docs/authors.md
All Contributors: https://github.com/deepmind/open_spiel/graphs/contributors

Files

The open_spiel-source-0.3.0.tar.gz bundles the necessary and some optional dependencies along with the core code (pybind11, Hanabi, ACPC, etc.) and should be able to be built directly without any additional downloads

Assets 3

07 Dec 09:54

lanctot

v0.2.0

ffe263b

OpenSpiel 0.2.0

This is our second release, mainly a stable version to use for our first entry in pip/PyPI.

Changes since last release:

Availability in PyPI as a source distribtuion. See installation page for instructions.

New Games

Battleship
Blackjack
Clobber
Hearts
Klondike Solitaire
Lewis Signaling
Oh Hell
Sheriff
Slovenian Tarok
Stones & Gems

New Algorithms

CorrDist correlated equilibrium gap metrics for agent-form (AFCE, AFCCE)
DQN (PyTorch)
Deep CFR (TF2)
External Sampling MCCFR (Python)
Online Outcome Sampling (Lisy et al, 2015)
NFSP (PyTorch)
Policy Iteration
Regression CFR (PyTorch) (Waugh et al. 2015, Morrill 2016)
ResponseGraphUCB (Rowland et al. 2019)

New Support

Build as a shared library using BUILD_SHARED_LIB
Eigen
Libtorch (PyTorch's C++ API)
PyTorch (including 3 algorithms)

Bots

Roshambo (rock paper scissors competition bots), optional using BUILD_WITH_ROSHAMBO
xinxin (Nathan Sturtevant's Hearts bot), optional using BUILD_WITH_XINIXIN

Misc

C++ LP solving via OR-Tools
Correlated equilibrium solver via LP solver (C++)
HyperGammon option to Backgammon
Exact correlation device builder (for extensive-form correlated equilibria)
Gambit normal-form game (.nfg) writer and reader
Game Transform to start at a specific place in a game
Observation framework
Public State API and factored observation games (FOGs): public observations, action-observation histories, visualizations (Kovarik et al.)
Repeated games
Single-player turns in simultaneous move games
Serialization and checkpointing in CFR, MCCFR, CFR-BR
Serialization for kSampledStochastic chance modes
Suggested alpha function for alpha-rank
Supervised learning example for Hearts (based on xinxin self-play)

New utilities

ActionView
Random interface
down_cast
combinatorial helpers