OpenSpiel + Swift for TensorFlow

This directory is a port of OpenSpiel using Swift for TensorFlow. Swift OpenSpiel explores using a single programming language for the entire OpenSpiel environment, from game implementations to algorithms and deep learning models.

Swift OpenSpiel is intended for serious research use. As the Swift for TensorFlow platform matures and gains additional capabilities (e.g. distributed training), significantly more algorithms will become expressible and tractable to train.

All contributions are welcome: both games and algorithms! If you run into issues (or would like to share your successes), please reach out to the Swift for TensorFlow community at [email protected].

Building

To use Swift OpenSpiel, download a recent Swift for TensorFlow toolchain following these installation instructions (available for macOS and Ubuntu currently). Swift OpenSpiel currently builds with the latest stable toolchains.

Using the toolchain, build and test Swift OpenSpiel like a normal SwiftPM package:

cd swift
swift build # Build the OpenSpiel library.
swift test  # Run tests.

A tour through the code

• Spiel.swift contains common game abstractions, including GameProtocol and StateProtocol.
• Games are implemented in separate subdirectories. Perfect information games include TicTacToe and Breakthrough. Imperfect information games include KuhnPoker and LeducPoker.
• Available algorithms include tabular exploitability and exploitability descent.

Join the community!

If you have any questions about Swift for TensorFlow (or would like to share your work or research with the community), please join our mailing list [email protected].