Skip to content

Complete Muzero with Training & API. Reinforcement learning with neural networks to navigate Rust Monte Carlo for playing games in open gym environments. Forked From Muzero-General. This system aims to be able to play any turn based game at a high level. Is it luck or was it skill?

License

Notifications You must be signed in to change notification settings

AnOnionAi/destino

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

The Destiny Of Mu0

alt text

To get started with Muz. Use Poetry.

poetry install 

If you do not have Poetry, run this command. (Linux Only. Destino does not work on Windows)

curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python -

To Run mu

poetry run python main.py

Save models in the Cloud (Will Have to Setup. Uses Google Cloud)

Change cloud_save equal to True in games/config/game_name.yaml.

To Run the Ai as an API.

poetry run uvicorn app:app --reload

To See Graphs On Tensorboard

poetry run tensorboard --logdir ./results

Google Colab

Configure only use 2 workers in game yaml file. 1 Self Player and 1 Reanalyze Worker 1 Worker per 1 Thread.

Copy the vars.env file into Google Cloud Drive. Enter the credentials for your Github Account. Open destino.ipynp on Google Co-lab to use with GPU for free. Two additional notebooks are available for educational purposes.

Python version need to match up between all programs for ray, poetry & colab to work.

How to Create a Game Wheel - Python Package

Zarena Repo For Rust Versions of Gato, Chess, Checkers, Blackjack and Poker.

Checkout notebooks/build_game_wheel.ipynp

!poetry install
!apt install rustc
!poetry run maturin build --release

The Build System

Rust Only

Cargo.toml -> Rust Build

Uses Cargo

cargo run cargo check cargo build cargo build --release

Python & Rust Hybrid - Building the Games For Destino

Maturin

Will need to change pyproject.toml build system to "maturin>=0.10,<0.11"

When a game package is made with maturin. We want the tar.gz file in games/gym Insert into the pyproject.toml file like this:

gym_chess = { path = "games/gyms/gym_chess-0.3.0.tar.gz" }

To specify python dependencies, add a list requires-dist in a [package.metadata.maturin] section in the Cargo.toml. This list is equivalent to install_requires in setuptools:

Rust&Python - Dev

maturin build

Rust&Python - Production

maturin build --release 

Maturin builds the wheels and stores them in a folder (target/wheels by default) Maturin develop builds the crate and installs it as a python module directly in the current virtualenv.

Diagnose Models

The following Debian packages must be installed in order to use this diagnose models tool.

sudo apt-get install graphviz
sudo apt-get install xdg-utils

Thanks to DeepMind, Werner Duvaud and Aurèle Hainaut

About

Complete Muzero with Training & API. Reinforcement learning with neural networks to navigate Rust Monte Carlo for playing games in open gym environments. Forked From Muzero-General. This system aims to be able to play any turn based game at a high level. Is it luck or was it skill?

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published