Package Details: lc0 0.30.0-1

Git Clone URL: https://aur.archlinux.org/lc0.git (read-only, click to copy)
Package Base: lc0
Description: UCI-compliant chess engine designed to play chess via neural network, specifically those of the LeelaChessZero project.
Upstream URL: https://lczero.org/
Keywords: chess engine lc0 lczero leela neural-network uci
Licenses: GPL3
Submitter: ivanmlerner
Maintainer: ivanmlerner
Last Packager: ivanmlerner
Votes: 5
Popularity: 0.009297
First Submitted: 2019-08-04 07:47 (UTC)
Last Updated: 2023-09-10 20:20 (UTC)

Pinned Comments

ivanmlerner commented on 2019-08-04 18:25 (UTC) (edited on 2023-09-10 20:23 (UTC) by ivanmlerner)

At least one backend is needed. The most common are listed as optional dependencies.

Latest Comments

« First ‹ Previous 1 2 3 4 Next › Last »

ivanmlerner commented on 2022-11-15 15:52 (UTC) (edited on 2022-11-15 16:05 (UTC) by ivanmlerner)

@tdtooke You don't need to tweak the build to use other weights files, you can just remove the default one and place yours on /usr/lib/lc0 or if you use multiple weights files you can select what you want to use in lc0.config as explained in lc0 --help or directly in your UCI interface. Yeah, I believe it is normal. Picking the best network is not about picking the biggest one. Such big networks should only be used for long analysis with top hardware. Here they explain how to choose the proper network for your situation.

mrzeznicki commented on 2022-11-15 09:25 (UTC)

@ivanmlerner Yes, thanks. That really helps for people who don't have bash as sh, like me. Thanks!

ivanmlerner commented on 2022-11-14 20:31 (UTC)

@mrzeznicki I changed the PKGBUILD according to your suggestion, is that what you meant?

mrzeznicki commented on 2022-10-15 11:12 (UTC)

Does this work for you with the latest NVIDIA drivers 520.56.06? I am getting Error building OpenCL kernels - after downgrading to 515.76 it works again.

tdtooke commented on 2022-10-14 09:23 (UTC)

I usually use Dragon or Stockfish and recently decided to give this one a go. I tweaked your build a little to download a more beefy weights file. My question is, with the larger networks is it normal for an analysis to take 'literally' all night long? Other than the ridiculous amount of time it takes to do an analysis, on my build anyways, I can say that my CUDA build does everything it's supposed to do.

mrzeznicki commented on 2022-10-13 01:11 (UTC)

build.sh is NOT sh-compatible (it contains popd). You should not do sh build.sh in PKGBUILD, but rather execute the file directly.

FabioLolix commented on 2022-04-02 13:50 (UTC)

@Tyrin.price run the command updpkgsums in the directory with the pkgbuild.

The main issue is that using master.tar.gz as source is sub-optimal and lead at this kind of issues

Tyrin.price commented on 2022-04-02 13:47 (UTC) (edited on 2022-04-02 13:52 (UTC) by Tyrin.price)

Anyone know how to get past this error?

==> Validating source files with sha256sums... 

lc0 ... Passed

lc0-0.28.2.tar.gz ... Passed

lczero-common-master.tar.gz ... FAILED

weights_hanse-69722-vf2.gz ... Passed

==> ERROR: One or more files did not pass the validity check!

rvess commented on 2022-02-05 07:26 (UTC)

==> ERROR: Failure while downloading https://www.comp.nus.edu.sg/~sergio-v/t40/256x20/256x20-t40-1541.pb.gz.

« First ‹ Previous 1 2 3 4 Next › Last »