TalkChess.com

Testrun finished: Lc0 0.22.0 with huge 384x30-t40-1097 S.Vieri Net. How many nodes per second were you getting? On my mobile RTX 2060 (with reduced speed by afterburner-tool, in order to get a Leela-Ratio of 1.3, because the mobile Hexacore CPU i7-8750H is not so fast (with turbo-boost switched off...

Leo wrote: ↑

Fri Nov 15, 2019 4:52 pm

Stefan, are you yourself training the Leela Cerebellum book?

No. All work for the Cerebellums (Lc0 and Stockfish) is done by Thomas Zipproth. I do the testruns, only.

Testrun finished: BrainFish-2 191007 (same patch-level as the best Stockfish in my testings so far) with brand new Cerebellums (Leela-Cerebellum Release 3 and Stockfish-Cerebellum Release 175).

https://www.sp-cc.de

(Perhaps you have to clear your browsercache or reload the website)

Testrun finished: Lc0 0.22.0 with Dark Queen 2.0 Net (first testrun of a net built out of human (online) games, only)

https://www.sp-cc.de

(Perhaps you have to clear your browsercache or reload the website)

What is the difference between "Fritz17Setup_x64" and "Fritz17Setup_Win10" ? I run on Windows 10 64-bit. I was wondering the same thing, I have only setup and am testing the Fat Fritz Network right now. I would like to save my Fritz 17 key for now as they just renewed my Playchess subscription. Not...

crem wrote: ↑

Tue Nov 12, 2019 8:26 am

It's not compressed, they just changed the header for Lc0 not to be able to load it.

OK, Thanx for information.

AdminX wrote: ↑

Tue Nov 12, 2019 7:53 am

Please forgive me if I missed it, but what is the size of the neural network of Fat Fritz?

It is a 256x20 Net. But the file on HDD is compressed - it can no be read by lc0 0.22.0, only by the FatFritz-Engine.exe

Next NN-testrun: Lc0 0.22.0 with DarkQueen 2.0 Net (first testrun of a net built out of human (online) games, only)

I doubt, that human games are good enough to build a really strong net out of them. But I want to test that...

For me, it becomes more and more clear, that it is the wrong way to build nets, which are bigger size, than the net-size in the training run, which delievered the learning-data. IMHO the other direction is the right way: Building nets, which are same or smaller size, than the net-size in the trainin...

Testrun finished: Lc0 0.22.0 with huge 384x30-t40-1097 S.Vieri Net.

https://www.sp-cc.de

(Perhaps you have to clear your browsercache or reload the website)