Page 3 of 3

Re: Sayuri

Posted: Sat Nov 18, 2023 2:39 pm
by And
Sayuri - Zenith Go
Zenith Go 3 dan
Sayuri p 1


Re: Sayuri

Posted: Sat Nov 18, 2023 3:08 pm
by And
help

Code: Select all

Arguments:
        --quiet, -q
                Disable all diagnostic verbose.

        --analysis-verbose, -a
                Dump the search verbose.

        --ponder
                Thinking on opponent's time.

        --reuse-tree
                Will reuse the sub-tree.

        --early-symm-cache
                Accelerate the search on the opening stage.

        --friendly-pass
                Do pass move if the engine wins the game.

        --capture-all-dead
                Try to remove all dead strings before pass. May be not safe for
game.

        --cache-memory-mib <integer>
                Set the NN cache size in MiB.

        --playouts, -p <integer>
                The number of maximum playouts.

        --const-time <integer>
                Const time of search in seconds.

        --gpu, -g <integer>
                Select a specific GPU device. Default is all devices.

        --threads, -t <integer>
                The number of threads used. Set 0 will select a reasonable numbe
r.

        --batch-size, -b <integer>
                The number of batches for a single evaluation. Set 0 will select
 a reasonable number.

        --lag-buffer <float>
                Safety margin for time usage in seconds.

        --score-utility-factor <float>
                Score utility heuristic value.

        --lcb-reduction <float>
                Reduce the LCB weights. Set 1 will select the most visits node a
s the best move in MCTS.

        --resign-threshold, -r <float>
                Resign when winrate is less than x. Default is 0.1.

        --weights, -w <weight file name>
                File with network weights.

        --book <book file name>
                File with opening book.

        --logfile, -l <log file name>
                File to log input/output to.

Re: Sayuri

Posted: Mon Nov 20, 2023 1:28 am
by And
Sayuri v0.5.0 with a network zero-80k.bin, 1 playouts plays about 15k
CS Zero 15k - Sayuri


Re: Sayuri

Posted: Mon Nov 20, 2023 11:52 am
by And
Sayuri - CS Zero 15k


Re: Sayuri

Posted: Sun Nov 26, 2023 10:47 pm
by And
CS Zero 13k - Sayuri


Re: Sayuri

Posted: Sun Nov 26, 2023 10:51 pm
by And
Sayuri - CS Zero 13k


Re: Sayuri

Posted: Mon Nov 27, 2023 11:40 pm
by And
do you think Sayuri's game is similar to a human game?

Re: Sayuri

Posted: Mon Nov 27, 2023 11:43 pm
by And
Zenith Go - Sayuri
Zenith Go 9 dan
Sayuri p 100, zero-swa-2200k.bin.txt


Re: Sayuri

Posted: Mon Nov 27, 2023 11:47 pm
by And
Sayuri - Zenith Go
Zenith Go 9 dan
Sayuri p 100


Re: Sayuri

Posted: Sat Mar 02, 2024 10:01 pm
by And
Release v0.7.0
https://github.com/CGLemon/Sayuri/releases/

Since v0.7.0, we will provide the full Windows version exe file. The exe file should work well on the Windows 10/11. If you want to compile the code on you local device, please see this section
  • Changes in v0.7.0
  • Support Japanese-like rule. Add the flag --scoring-rule territory to enable the rule. But we don't have the executable weights now. The weights will be coming soon.
  • Support the rank AI style. Add the flag --relative-rank <int> to adjustment engine strength. The ideal is from the katrain.
  • Some bugfixes, the self-play process should be better.

    The 3rd Run
    Coming soon.

Re: Sayuri

Posted: Sun Mar 03, 2024 10:19 pm
by And
About --relative-rank?
https://github.com/CGLemon/Sayuri/issues/18

"The range of --relative-rank should be between 0~25. The higher value should be stronger and only work on the 19x19 game. Here is how it works.

We transform the relative-rank into the number of active children at the root node. The root will randomly select this number of children and prune other non-selected children. For example, the root randomly select 23 children at the first move if the relative-rank is 0.

The number of playouts option will significantly affect the strength, because the root may prune all high priority move. The policy network does not work in the case. The tree search could keep the quality of moves."

Re: Sayuri

Posted: Fri Jul 26, 2024 2:17 pm
by And

Re: Sayuri

Posted: Thu Sep 12, 2024 8:44 pm
by And