TalkChess.com

Hi Ed, old Leela net 11248 finds it instantly then sticks to it after 1 sec (without any GPU - using CPU version) depth 2/4, 14 evaluated nodes LCzero 0.19-11248: 1/2 00:00 6 7 +0,26 Rd1xd3 c4xd3 2/3 00:01 12 7 -1,00 Rh1g1 Nd3xc1 Rd1xc1 2/4 00:01 14 7 -0,15 Rd1xd3 c4xd3 Bf3e4 d3d2 3/5 00:02 21 7 -0,...

Hello Larry, indeed a size balance affair, according to your CPU. Strongest on my old laptop (CPU T4500@2,3Ghz) using 1 single core is 11258-112x9-se ( distilled by Dietrich Kappe ). I know, it is an old one, but I could find no convincing one within more recent stuff. My associated setting: MiniBat...

Thanks for making us aware, Eduard!
Just bought it, and looking forward to play with it

does it run on your device? Hello Peter, yes it does (but Android 4.1.2). A last chance for you Peter: I digged out the full download from which the DD engine I use came. And know what? Source is included! :D So maybe a gentleman would be able to create a modern compile for you, not to forget other...

Indeed my Android is 4.1.2 - consistent with Aart's explanation.
Try to pick & choose from here then (thks to Gurcan)
BR, Tibono

PeterO wrote: ↑

Sun Jun 02, 2019 9:31 am

Question: Is there a android version of this engine?

Hi PeterO, yes, and I uploaded it for you here
Best regards,
Tibono

supersharp77 wrote: ↑

Wed May 29, 2019 4:18 pm

https://github.com/LeelaChessZero/lc0/f ... 6-48x5.zip

Hi, thanks for the links. This one works without any issue in my old config: Android 4.1.2, CFA 5.2.5 & DroidFish 1.53 DD.
Thanks as well to DKappe for the distilled NW that make performance acceptable!

Tibono

Hi, I also use a laptop - usually reaching no more than 72-74°C at full CPU usage; but during hot summer time I use Battle Encoder Shirazé to throttle chess engines. A very efficient tool; can also be used to tune some emulators speed, or to save some CPU available for your own use while running a t...

dkappe wrote: ↑

Sat Jan 12, 2019 12:54 am

64x6 or 64x8?

64x8 . No 64x6 hosted behind the link I provided.

dkappe wrote: ↑

Sat Jan 12, 2019 12:54 am

The 96x8-se is already released.

Thanks! Looking forward to test it!

Hi for CPU use I would suggest, worth a try: lc0 >= 0.17 to allow ponder on scs 64x8 network run1 550000 steps ( link ). Not too large for speed sake, but skilled enough. So well balanced... Cpuct MCTS 1.700000 Minibatch size 16 Max prefetch nodes 0 Scale thinking time 1.200000 NN cache size 400000 ...