Life In 19x19

The attempt to build leela-zero resulted in various incomprehensible errors, of
which the last is shown below. A leelaz binary is left behind, but cannot be
tested.

Disable Werror in /gtest/googletest/cmake/internal_utils.cmake

# set(cxx_base_flags "-Wall -Wshadow -Werror")




The attempt ...

I am interested in a different question, which is whether RAVE makes leela zero a better analyst for reviews. One problem with good players as analysts is sharp policy. It is not at all unusual for humans to come up with good candidate moves that sharp policy overlooks, but which arguably a good ...

RAVE wasn't good for leela zero. It seems that RAVE improves value a bit in few POs, but it spoils sharp policy.
https://github.com/zakki/leela-zero/commits/rave

Sometime RLO plays PASS at losing game, because they cannot find any good move, but win rate is bigger than a resign threshold.

@zakki, can you explain how using Ray with LZ weights is different from using LZ with the same weights?

RLO is experimental branch to understand how AlphaGo Zero style policy and value network behave, and to experiment integrating monte carlo playout to AGZ's dual network.

It has some advantage ...

Ray
I haven't been able to crack that one neither, but the latest version does not use CNTK/CUDA any more and now uses OpenCL. This was confirmed by the developer . The make file only needs to be updated apparently.
Unfortunately that is not actually true. I've just downloaded the most recent ...

I usually use Windows, and Rn has no maintainer on Linux.
Pull requests is welcomed.

Rn requires CNTK 2.1 now.

AlphaGo uses 30000000 games.
But now 1600000 human games without random moves and 1440000 aya games with random move are available.
Rayn (and other programs) needs high quality 10x amount of games.

I think architecture like Folding@home is suitable to generate millions of self-play game, rather than training neural network.

Perhaps g++-4.8 or is needed.
https://github.com/zakki/Ray/issues/8

Hey, I see Rayon can organize its timing with "--time" flag. Can you implement this as GTP commands, so we can use the GUIs (Drago here) interface for setting time controls? Thx for the engine.

A comparison Leela/Rayon, how can matches be fair? I'd like to have them calculate a fixed time per move ...

Perhaps you need to link libeval like .

Rn.3.3.1-4c0g doesn't use gpu. http://www.yss-aya.com/cgos/19x19/cross ... -4c0g.html
Other Rn.xx-4c uses i7-2600K (or 6700K) + GTX980ti