{"id":24824839,"url":"https://github.com/baskuit/r-nad","last_synced_at":"2025-08-19T21:21:37.808Z","repository":{"id":58755515,"uuid":"528869003","full_name":"baskuit/R-NaD","owner":"baskuit","description":"Experimentation with Regularized Nash Dynamics on a GPU accelerated game","archived":false,"fork":false,"pushed_at":"2023-04-21T19:07:38.000Z","size":882,"stargazers_count":47,"open_issues_count":0,"forks_count":5,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-04-06T19:51:22.140Z","etag":null,"topics":["deepnash","multiagent-reinforcement-learning","pytorch","reinforcement-learning","reinforcement-learning-algorithms","rnad"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/baskuit.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null}},"created_at":"2022-08-25T13:42:22.000Z","updated_at":"2025-03-29T10:41:36.000Z","dependencies_parsed_at":"2023-02-16T23:15:55.253Z","dependency_job_id":null,"html_url":"https://github.com/baskuit/R-NaD","commit_stats":{"total_commits":168,"total_committers":2,"mean_commits":84.0,"dds":"0.011904761904761862","last_synced_commit":"0d163921bc597405040c33c89e151b18da68fa6e"},"previous_names":[],"tags_count":1,"template":false,"template_full_name":null,"purl":"pkg:github/baskuit/R-NaD","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baskuit%2FR-NaD","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baskuit%2FR-NaD/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baskuit%2FR-NaD/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baskuit%2FR-NaD/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/baskuit","download_url":"https://codeload.github.com/baskuit/R-NaD/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/baskuit%2FR-NaD/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":271223856,"owners_count":24721758,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","status":"online","status_checked_at":"2025-08-19T02:00:09.176Z","response_time":63,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["deepnash","multiagent-reinforcement-learning","pytorch","reinforcement-learning","reinforcement-learning-algorithms","rnad"],"created_at":"2025-01-30T20:57:46.133Z","updated_at":"2025-08-19T21:21:37.745Z","avatar_url":"https://github.com/baskuit.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\n# RL Experiments\n\nMost people limited to consumer hardware are unable to test RL algorithms because of the cost of data-generation.\n\nThe repo is an implementation of DeepMind's R-NaD algorithm on a class of extremely vectorized imperfect information games, inspired by [Learning to Play Snake at One Million FPS](https://towardsdatascience.com/learning-to-play-snake-at-1-million-fps-4aae8d36d2f1).\n\nUsing this platform, anyone with a reasonably powered PC can experiment with the algorithm.\n\n## Setup\n\n    pip3 install -r requirements.txt\n    python3 main.py\n\n# R-NaD\n\nIntroduced here:\n\nhttps://arxiv.org/abs/2002.08456\n\nDeepNash (SOTA Stratego agent):\n\nhttps://arxiv.org/abs/2206.15378\n\nThis new regularization allows neural network policies to converge to Nash equilibrium in imperfect information games, which previously were a well-known failure case for policy gradient.\n\n# Trees\n\nThe imperfect information game implemented here is an abstract stochastic matrix tree. Many are familiar with the idea of a matrix game, like rock paper scissors. Imagine that, except sequential and with elements of chance.\n\nThe trees are randomly generated and can express wide range of depth and stochasticity using the provided numeric and functional parameters.\n\nThe default observation of a state is the matrix of expected payoff, under the assumption the rest of the game is played perfectly by both players.\n\nThis means our contrived game is well-behaved, in the sense that observations actually contain info about the optimal policy.\n\n\n![Alt text](logs.png?raw=true \"Title\")","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbaskuit%2Fr-nad","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fbaskuit%2Fr-nad","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbaskuit%2Fr-nad/lists"}