-
I was just wondering, what would be a good milestone to start training a small network tabula rasa? How big will it be? 64x5? I would like to work towards that goal!
- Obviously we need a distribut…
-
@jadermcs vou colocar aqui alguns exemplos de RL, que você pode replicar :
- Este replica a expriencia do AlphaGO para outro jogo. Acho que é um bom exemplo do que vamos tentar fazer: https://git…
-
The paper introduces AlphaGo Zero, the latest evolution of AlphaGo, the first computer program to defeat a world champion at the ancient Chinese game of Go. Zero is even more powerful and is arguably …
-
In leela-zero UCTSearch.cpp:296 we see:
```
// Check whether to randomize the best move proportional
// to the playout counts, early game only.
auto movenum = int(m_rootstate.get_mov…
-
Hi,
got this error log when I tried to combine results for several (1000) repositories:
```console
panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmen…
-
In #1261 @gcp writes:
> The quality of play deteriorates rapidly with increasing handicap ("all moves lose no matter what I play?!") which is why I'm personally only interested in even games.
I'…
-
https://alphagoteach.deepmind.com/
It gives us the probabilities of the common openings. From this, we know that Black's best opening move has a 47.1% chance of winning (at Q16), which probably mea…
-
Thanks a lot for the powerful tool.
As we know, the current dual_view is using the "ABCDEFG..." to indicate different variations.
Shall we have a new interface/setting which marks the different vari…
-
In investigating https://github.com/LeelaChessZero/lc0/issues/344, I noticed that leela related self-play has never used Q=0 and had various iterations going from 1.1 to parent Q to fpu reduction, etc…
-
The Silver et al paper says that they allowed one minute (of TPU?) per chess move, which equated to about 80k positions being evaluated per second. So: once a good set of Leela-chess weights are in pl…