BrăilaGo

Brăila, Go, restul se înţelege

RSS 2.0

12 partide profesioniste ELF OpenGo vs pro

În aprilie 2018 ELF OpenGo a jucat 14 partide împotriva a patru profesioniști coreeni, Kim Jiseok, Shin Jinseo, Park Yeonghun și Choi Cheolhan. A cîștigat toate partidele prin abandon, Mai jos sînt 12 dintre cele 14 partide. Două dintre ele nu au fost făcute publice conform dorinței celor ce le-au jucat.partidele sînt publicate pe github împreună cu cîteva versiuni executabile ale lui ELF OpenGo. Partidele sînt anonimizate, pentru că numele exact al celui ce a jucat nu este precizat.

Detaliile meciului sînt publicate în https://research.fb.com/facebook-open-sources-elf-opengo/

[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game1.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game2.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game3.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game4.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game5.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game6.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game7.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game8.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game9.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game10.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game11.sgf”]
[glift sgf=”http://brailago.ro/arhiva/partide/2018/pro-games-go-19×19-2018-04/game12.sgf”]

@jma127 jma127 released this on May 4, 2018

This archive contains 12 of the 14 games played with four KBA-associated professionals in April 2018 (see this post). ELF OpenGo won all games by resignation.

We are respecting the players’ request not to publish two of the games, and we’re also anonymizing the games. Each player is represented in at least two games contained in this archive.

The first two games were played using a weaker prerelease version of the ELF OpenGo model weights. For each move, ELF OpenGo used 2 threads with 10000 rollouts per thread (grouped into batches of 4).

All other games were played using the v0 pretrained model (publicly available for download). For each move, ELF OpenGo used 2 threads with 40000 rollouts per thread (grouped into batches of 16). This took around 50 seconds per move on a V100 GPU.

For all games, no constraints were imposed on human thinking time.

There is no source code associated with this release. All files herein are covered under this repository’s BSD-style LICENSE.

Costel


Leave a Reply

Your email address will not be published. Required fields are marked *

Recent Posts

Recent Comments