groups.google.com

Hiroshi Yamashita

unread,

Mar 19, 2016, 6:32:58 AM3/19/16

to compu...@computer-go.org

There are 32 participants, include one guest GNU GO.
After 7 swiss round,

1st CGI 7-0
2nd CrazyStone 6-1
3rd Zen 6-1
4th Aya 6-1
5th Gonanza 5-2
6th Ray 5-2
7th DolBaram 5-2
8th darkforest 5-2

CrazyStone lost against Zen.
Zen lost against CGI.
DolBaram lost against Ray and CGI.
darkforest lost against Zen and CrazyStone.

Top 16 programs will play tommorow tournament.
All top 8 prigrams except DolBaram? use DCNN.

Thanks,
Hiroshi Yamashita


_______________________________________________
Computer-go mailing list
Compu...@computer-go.org
http://computer-go.org/mailman/listinfo/computer-go

James Guo

unread,

Mar 19, 2016, 7:13:34 AM3/19/16

to compu...@computer-go.org

Show the official web site link?

Rémi Coulom

unread,

Mar 19, 2016, 7:17:38 AM3/19/16

to compu...@computer-go.org

"Ingo Althöfer"

unread,

Mar 19, 2016, 7:40:09 AM3/19/16

to compu...@computer-go.org

Hello Hiroshi,

thanks for the information.

WHo are the peolbe behind CGI Go?

Which hardware is used by the top participants in UEC cup?

Ingo.

> Gesendet: Samstag, 19. März 2016 um 14:32 Uhr
> Von: "Hiroshi Yamashita" <y...@bd.mbn.or.jp>
> An: compu...@computer-go.org
> Betreff: [Computer-go] UEC cup 1st day result

Nick Wedd

unread,

Mar 19, 2016, 7:44:36 AM3/19/16

to compu...@computer-go.org

Petr Baudis

unread,

Mar 19, 2016, 7:44:40 AM3/19/16

to compu...@computer-go.org

I think they are

https://github.com/CGI-LAB
http://www.aigames.nctu.edu.tw/

e.g. NiceGo, Amigo programs in the past.

I'm curious if they are mainly reimplementing the AlphaGo paper or doing
something else.

--
Petr Baudis
If you have good ideas, good data and fast computers,
you can do almost anything. -- Geoffrey Hinton

Hiroshi Yamashita

unread,

Mar 19, 2016, 8:11:45 AM3/19/16

to compu...@computer-go.org

Hello Ingo,

As far as I know,
CrazyStone used a machine 24 cores, no GPUs.
Aya used a Amazon EC2 g2.8xlarge 16 cores, 4 GPUS.
Gonanza used a mahicne 16 cores with a GTX 980 ti.
Ray used a mahicne with a GPU.
Zen and CrazyStone uses DCNN without GPU.

I'll use 4 g2.8xlarge by root parallelization.
g2.8xlarge has 16 cores, but I use only 8 cores.
GPUs are slow, so creating node is not enough.
Too many playouts looks harm for tree result.
16 threads is weaker than 8 threads by my 1sec/move test.

Thanks,
Hiroshi Yamashita

Wu Ti-Rong

unread,

Mar 19, 2016, 9:27:49 AM3/19/16

to compu...@computer-go.org

The current version in UEC is CGI 2.0 (or DeepCGI) as described in this page (also including the contributors).

The main contribution for Amigo (nothing to do with nicego) is to design a general MCTS framework that many game programs can directly use.

CGI 2.0 is still based on the framework.

> I'm curious if they are mainly reimplementing the AlphaGo paper or doing
> something else.

We developed DCNN from Dec 2015. That means that we have not read AlphaGo’s paper at that time yet.

But, we are still inspired by their paper.

The hardware used for UEC is 4 GPUs + 40 threads.

Best Regards,

Ti-Rong Wu

Hiroshi Yamashita

unread,

Mar 19, 2016, 3:21:15 PM3/19/16

to compu...@computer-go.org

Today's tournament table is
https://twitter.com/math26/status/711190391098187776
(Thanks Shinoda-san).

Holmes got Originality award. Holmes does both Policy Network and playout
on 8 GPUs. Playouts are done by 256 mini-batches. 256 playouts are
executed simultaneously.

Reply all

Reply to author

Forward

Read the original on groups.google.com ↗