geek-ai/MAgent: A Platform for Many-agent ... - GitHub

MAgent is a research platform for many-agent reinforcement learning. Unlike previous research platforms that focus on reinforcement learning research with a ... Skiptocontent {{message}} geek-ai / MAgent Public Notifications Fork 301 Star 1.4k APlatformforMany-agentReinforcementLearning MITLicense 1.4k stars 301 forks Star Notifications Code Issues 26 Pullrequests 1 Actions Projects 0 Wiki Security Insights More Code Issues Pullrequests Actions Projects Wiki Security Insights Thiscommitdoesnotbelongtoanybranchonthisrepository,andmaybelongtoaforkoutsideoftherepository. master Branches Tags Couldnotloadbranches Nothingtoshow {{refName}} default Couldnotloadtags Nothingtoshow {{refName}} default 3 branches 0 tags Code Latestcommit merrymercy Mergepullrequest#79fromjustinkterry/patch-1 … b1fce27 May28,2020 Mergepullrequest#79fromjustinkterry/patch-1 Mentionmaintainedforkinreadme b1fce27 Gitstats 134 commits Files Permalink Failedtoloadlatestcommitinformation. Type Name Latestcommitmessage Committime data     doc     examples     python/magent     scripts     src     .gitignore     CMakeLists.txt     LICENSE     README.md     build.sh     Viewcode Thisprojectisnolongermaintained MAgent Requirement InstallonLinux InstallonOSX Docs Examples Train Play BaselineAlgorithms Acknowledgement README.md Thisprojectisnolongermaintained Pleaseseehttps://github.com/PettingZoo-Team/MAgentforamaintainedforkofthisprojectthat'sinstallablewithpip. MAgent MAgentisaresearchplatformformany-agentreinforcementlearning. Unlikepreviousresearchplatformsthatfocusonreinforcementlearningresearchwithasingleagentoronlyfewagents, MAgentaimsatsupportingreinforcementlearningresearchthatscalesupfromhundredstomillionsofagents. AAAI2018demopaper:MAgent:AMany-AgentReinforcementLearningPlatformforArtificialCollectiveIntelligence Watchourdemovideoforsomeinterestingshowcases. Herearetwoimmediatedemoforthebattlecase. Requirement MAgentsupportsLinuxandOSXrunningPython2.7orpython3. Wemakenoassumptionsaboutthestructureofyouragents. Youcanwriterule-basedalgorithmsorusedeeplearningframeworks. InstallonLinux [email protected]:geek-ai/MAgent.git cdMAgent sudoapt-getinstallcmakelibboost-system-devlibjsoncpp-devlibwebsocketpp-dev bashbuild.sh exportPYTHONPATH=$(pwd)/python:$PYTHONPATH InstallonOSX Note:Thereisanissuewithhomebrewforinstallingwebsocketpp,pleasereferto#17 [email protected]:geek-ai/MAgent.git cdMAgent [email protected] brewinstalljsoncppargp-standalone brewtapdavid-icracked/homebrew-websocketpp brewinstall--HEADdavid-icracked/websocketpp/websocketpp [email protected] bashbuild.sh exportPYTHONPATH=$(pwd)/python:$PYTHONPATH Docs Getstarted Examples Thetrainingtimeoffollowingtasksisabout1dayonaGTX1080-Ticard. Ifout-of-memoryerrorsoccur,youcantuneinfer_batch_sizesmallerinmodels. Note:Youshouldrunfollowingexamplesintherootdirectoryofthisrepo.Donotcdtoexamples/. Train Threeexamplesshownintheabovevideo. Videofileswillbesavedevery10rounds.Youcanuserendertowatchthem. pursuit pythonexamples/train_pursuit.py--train gathering pythonexamples/train_gather.py--train battle pythonexamples/train_battle.py--train Play Aninteractivegametoplaywithbattleagents.Youwillactasageneralanddispatchyoursoldiers. battlegame pythonexamples/show_battle_game.py BaselineAlgorithms Thebaselinealgorithmsparameter-sharingDQN,DRQN,a2careimplementedinTensorflowandMXNet. DQNperformsbestinourlargenumbersharingandgridworldsettings. Acknowledgement ManythankstoTianqiChenforthehelpfulsuggestions. About APlatformforMany-agentReinforcementLearning Topics reinforcement-learning deep-learning multi-agent Resources Readme License MITLicense Stars 1.4k stars Watchers 69 watching Forks 301 forks Releases Noreleasespublished Packages0 Nopackagespublished Contributors9 Languages Python 45.7% C++ 44.4% JavaScript 7.3% HTML 1.5% C 0.6% CMake 0.4% Shell 0.1% Youcan’tperformthatactionatthistime. Yousignedinwithanothertaborwindow.Reloadtorefreshyoursession. Yousignedoutinanothertaborwindow.Reloadtorefreshyoursession.