TalkChess.com

Eduard wrote: ↑

Mon Feb 04, 2019 3:29 pm

Thank you! I will test of course. It is fun. Such fast networks are optimal for mobile devices. Is there a chance that Lc0 will eventually run on Android?

I know someone was tinkering with an iOS version, not sure about android.

There are a few things that might help. One of them is basic certainty propagation (MCTS Solver) that will help in tactical and endgame positions and overall in self-play. Should be merged soon into an upcoming version of lc0.

11258-32x4-se distilled network released. Links and details on the wiki page: https://github.com/dkappe/leela-chess-w ... d-Networks

120x10 cooking now. Maybe 144x11 after.

Again … if someone could just figure out a way to add a "sanity checker" which uses a simple stripped down AB executable using the unused CPU resources this would be a very quick shortcut fix until this problem is solved by other means using 100% NN. The idea would be to have Lc0 run on the GPU and...

Of course I don't have A0, but I do have 100 games it played against SF. So I went through some games with Lc0 and it was remarkable that almost every move was a match with Lc0. The only moves that did not match were tactical shots that Lc0 could not see. So you could say that once Lc0 sorts out th...

M ANSARI wrote: ↑

Thu Jan 31, 2019 9:01 am

Can you explain what a "distilled" network is. Is this just a stripped down version of the original? I am guessing stripping down would cause the engine to be faster and might be better for less powerful GPU's?

It’s all explained on the wiki page linked above.

11258-104x9-se distilled network released. See here: https://github.com/dkappe/leela-chess-w ... d-Networks

32x4-se is next. The hunt continues.

The 24x3-se network has been released. See the distilled networks page. https://github.com/dkappe/leela-chess-weights/wiki/Distilled-Networks 24x3-se is ~300 elo stronger than 16x2-se, just around winter in strength. If someone has a very slow machine, like a raspberry pi, this might be the one for ...

OK, but these papers do not tell us your own works and the results depend on what you actually did. So I think an interpretation of your works would be needed. Naturally LC0 has the both head. But there are NN with separated value and policy head and there are NN in which the both are in the same s...

OK, thanks. But about the lost of information during the process there is no any data/opinion. I think in the case of GPUs the playing power also a trade off between speed and knowledge. The recent 20x256 net size is not enough to overcharge an RTX 2080Ti. Maybe a 40x256 can do it. I think you know...