-
_[Edit: This issue has become clarified in the discussion/comments below, which identifies the existence of the gzip.exe quirk/bug.]_
I tried to follow these straightforward instructions from the r…
-
so that we know if one of the color is stronger?
for example
Start Date | Network Hashes | Wins(W/b) / Losses | Games | SPRT
-- | -- | -- | -- | --
2017-12-18 23:45 | ce1aa643 VS 0eb8b23b | 257(…
-
https://github.com/gcp/leela-zero/blob/94a345f7cfcc47fff8a066e68d95376ccef9cde4/src/UCTNode.cpp#L288
In normal /next we overwrite the fpu value if the node has visits:
```
auto winrate …
-
currently, when using phoenixgo, unless i didnt find the option, we cant see the variations of the next moves (PV)
full options list here : https://github.com/wonderingabout/PhoenixGo/blob/faqv2-ba…
-
I use [this weight](https://leela.online-go.com/zero/best_v1.txt.zip)
But it don't work, and prints `Weights file is the wrong version.`
-
i think it involves supporting a custom GTP command; might involve being in minigo
-
Per V3 Eval graphs https://cloudygo.com/leela-zero-v3/eval-graphs , LZ199 is 81 ELO stronger than LZ network 201. Let's re-use 199 to create match games and be the new best network.
-
For some background, see https://github.com/gcp/leela-zero/issues/2007#issuecomment-439654245.
In order to be able to distribute Leela Zero binaries with cuDNN/CUDA/TensorRT support together with t…
-
Could a table be put on zero.sjeng.org, beside the graph showing the top 10 networks ratings. (the graph is hard to see with so many nets together).
thanks.
-
It would be nice if instead Original strength graph (Mostly obsolete) one could see a magnification of the last fraction of the main graph (like 10 times bigger for example with last 500.000 games)
A…