TalkChess.com

Just read on discord (dev-log channel), that Leela will play with Net 10480 in TCEC in Divison 3. If anybody want to do some testing with TCEC-Leela...

Download the Net 10480 here: http://testserver.lczero.org/networks

Because of that, I restarted my long thinking-time testrun (https://www.sp-cc.de/long-thinkingtime.htm) with that Net - lets see, how Leela will score with 10480 (300 games with 12'+5''). Result not before next Wednesday. My first testrun with a 20x256 net - Leela-ratio with that bigger net in my testrun is 0.67 (which should be a little bit better for Leela than in TCEC)
Stay tuned!

IF it qualifies

arunsoorya1309 wrote: ↑Wed Aug 08, 2018 7:18 am IF it qualifies

Of course! But after Leela's win over Wasp, I think Leela will proceed to Div 3.

Don't go popcnt-ing your chickens just yet! The top three are enjoying a nice slugfest, and there are 'enough' games yet to play...
Kudos to Wasp for keeping the NN programs more than honest.

Geonerd wrote: ↑Wed Aug 08, 2018 7:57 am Don't go popcnt-ing your chickens just yet! The top three are enjoying a nice slugfest, and there are 'enough' games yet to play...
Kudos to Wasp for keeping the NN programs more than honest.

I doubt, that DeusX will be allowed to play in Div 3.
https://blog.lczero.org/2018/08/07/mistake/

arunsoorya1309 wrote: ↑Wed Aug 08, 2018 7:18 am IF it qualifies

With plus +100 to +150 elo in updated net , there wont be any concern for stage 3 as well.
Only in stage 1, Hiccups can happen!!

pohl4711 wrote: ↑Wed Aug 08, 2018 8:03 am
I doubt, that DeusX will be allowed to play in Div 3.
https://blog.lczero.org/2018/08/07/mistake/

Interesting. While Albert's video did set off my BS detector, I thought his approach to developing weights made quite a bit of sense.

This Leela project could not confuse me more. Are there two independent groups developing networks? How is that one is at ID 500-odd, while the other is above 10000?

Gabor, it is very confusing yes. I believe they are using a 15x192 network in the main server (ID in the 500s) and a test network of 20x256 in the test server (ID in the 10000s), but I'm not totally sure.

The 20x256 network is the same size as AlphaZero so, in theory, with proper training time it will be as strong as A0, provided it will subsequently run on hardware comparable to google's 4 TPU.

The 15x192 network runs quicker and is stronger when using "lesser" hardware, e.g. nvidia geforce below 1060 or thereabouts.

pohl4711 wrote: ↑Wed Aug 08, 2018 8:03 am I doubt, that DeusX will be allowed to play in Div 3.
https://blog.lczero.org/2018/08/07/mistake/

Ok, I think that settles it. It is a bit too radical, but understandable. (I would allow DX to finish this season, but I'm not the admin lol)

Still, comparing learning approaches does have value per se. Would be nice to see how far one gets using supervised learning based on top human and engine game (of the non-selfplay variety, of course).