Re: TCEC season 13, 2 NN engines will be participating, Leela and Deus X
Posted: Fri Aug 10, 2018 8:53 am
Hosted by Your Move Chess & Games - chessusa.com
Is withdrawing really their only option? Can't they if they really want e.g. get together with Wasp & lodge a protest?Gian-Carlo Pascutto wrote: ↑Fri Aug 10, 2018 8:47 amIt's up to (all) the other authors to vote with their feet. The core Leela Chess Zero people made a firm statement now that is a bit less conciliatory than the previous one. Will they withdraw Leela Zero from TCEC if DX plays? If they don't, IMHO the statement is without teeth. By playing in the tournament you are supporting the status quo.frankp wrote: ↑Fri Aug 10, 2018 8:00 amObviously DX could not exist without leela. Such arguments have been rehashed many times but have had zero impact as far as I can tell. The clone status is so obvious and this has been known by TCEC for some time now, so I do not expect it be banned at this late stage.
Just a bit of my oil to throw on the fire.
Of course it is a matter for TCEC what they do and do not allow, as I said.Gian-Carlo Pascutto wrote: ↑Fri Aug 10, 2018 8:47 amTCEC has always allowed engines that are derivative works of others, haven't they?frankp wrote: ↑Fri Aug 10, 2018 8:00 amNot seen any official announcement yet, but I imagine the answer is no. It was probably always a 'done deal' when it was accepted in the apparently secretive way it all happened at the last minute.
Obviously a matter for TCEC and their reputation. leela-DX is an interesting project. But if I was the wasp author I would be unhappy about not getting promoted because of a clone - or indeed a program that could not compete because its place was taken in Div 4 by a clone.
The whole situation is so bizarre that I am still waiting for the 'real' plan to be revealed. AS is not a programmer to my knowledge, has never submitted a chess program to a competition before and did not write or contribute to leela in any way, simply produced a weights file (interesting though that is) which is 'powered by leela' as he rather insultingly now says. The weights file is even produced by unaltered leela training code. Obviously DX could not exist without leela. Such arguments have been rehashed many times but have had zero impact as far as I can tell. The clone status is so obvious and this has been known by TCEC for some time now, so I do not expect it be banned at this late stage.
It's up to (all) the other authors to vote with their feet. The core Leela Chess Zero people made a firm statement now that is a bit less conciliatory than the previous one. Will they withdraw Leela Zero from TCEC if DX plays? If they don't, IMHO the statement is without teeth. By playing in the tournament you are supporting the status quo.
Just a bit of my oil to throw on the fire.
Even that solution does not sit right with me. Seems more like a consolation prize. Why not just set a extra seat at the table for Div 3?whereagles wrote: ↑Fri Aug 10, 2018 10:23 amJust let DX finish this season. I doubt it will promote to division 2 anyway. Next season Wasp gets a direct entry to division 3 and it should be fine.
It is a fun engine to put in-between seasons for showdowns. Plus, it can be trained with engine games for improvements.
Price-wise Nvidia 1060 is a pretty fair comparison to Ryzen 7 1700. A bit lower price for GPU, but higher consumption power-wise. I still think that Lc0 on GTX 1060 performs somewhat better against regular engines on 8 core Ryzen 7 1700 than what TCEC shows. I think that two CPU cores allocation for Lc0 is a bit bogus, one can allocate 2 threads of the same core with very little loss, a large part of CPU cycles Lc0 needs being anyway garbage. In fact I might check the GPU load with affinites set to 2 threads of the same CPU core.Milos wrote: ↑Wed Aug 08, 2018 11:23 amPricewise only because 22-core Xeon is ridiculously expensive compared to its performance, i.e. performance/price ratio is terrible. It's a bit the case like V100.Laskos wrote: ↑Wed Aug 08, 2018 10:41 amAh, some newer, more power efficient 22 core per CPU Xeon? In that case, only price-wise there is a disbalance. What I can say, is that my i7 4 core CPU and GTX 1060 GPU is a pretty balanced combination, and the results I get Lc0 versus regular engine are somewhat better than what TCEC seems to show now.
1060 has the best performance/price ratio within GPUs. Its equivalent within CPUs would be an 8 core Ryzen 7 like 1700. However that one is at least 2x better power wise. And that doesn't include another at least 2 core CPU that you need in order to run Lc0 on 1060, which raises power ratio to 3x and increases the price of Lc0 run system.
Latest SFdev on 16 threads Ryzen 7 1700 is at least 500Elo stronger of the best test (and overall) net in CCRL40/4 equivalent TC on 1060 and all that with 3x less power and cheaper price overall.