leela-zero 0.17-1build2 source package in Ubuntu

Changelog

leela-zero (0.17-1build2) focal; urgency=medium

  * No change rebuild against new boost1.71 ABI

 -- Dimitri John Ledkov <email address hidden>  Mon, 03 Feb 2020 20:41:37 +0000

Upload details

Uploaded by:
Dimitri John Ledkov on 2020-02-03
Uploaded to:
Focal
Original maintainer:
Ubuntu Developers
Architectures:
any
Section:
misc
Urgency:
Medium Urgency

See full publishing history Publishing

Series Pocket Published Component Section
Groovy release on 2020-04-24 universe misc
Focal release on 2020-02-04 universe misc

Downloads

File Size SHA-256 Checksum
leela-zero_0.17.orig.tar.gz 1.7 MiB fc2cc2ae9739aeab058dfceee8e4e67fdcfc0f7bfe2777e565a2ba17a5579267
leela-zero_0.17-1build2.debian.tar.xz 6.3 KiB 053e1aaec39b2e2c96240220215a6900066214cdf2381ea7a131d24dda077d3a
leela-zero_0.17-1build2.dsc 2.2 KiB cf4142aa327f5dd3c4de6538572975190034d69285b3ff4f5c88a6aabcfa5a49

Available diffs

View changes file

Binary packages built by this source

leela-zero: Go engine with no human-provided knowledge, modeled after the AlphaGo Zero paper

 A Go program with no human provided knowledge. Using MCTS (but without Monte
 Carlo playouts) and a deep residual convolutional neural network stack.
 .
 This is a fairly faithful reimplementation of the system described in the
 Alpha Go Zero paper "Mastering the Game of Go without Human Knowledge". For
 all intents and purposes, it is an open source AlphaGo Zero.
 .
 https://deepmind.com/documents/119/agz_unformatted_nature.pdf
 .
 No network weights are in this repository. If you manage to obtain the AlphaGo
 Zero weights, this program will be about as strong, provided you also obtain a
 few Tensor Processing Units. Lacking those TPUs, the author recommends a top
 of the line GPU - it's not exactly the same, but the result would still be an
 engine that is far stronger than the top humans.
 .
 Recomputing the AlphaGo Zero weights will take about 1700 years on commodity
 hardware. Upstream is running a public, distributed effort to repeat this
 work. Working together, and especially when starting on a smaller scale, it
 will take less than 1700 years to get a good network (which you can feed into
 this program, suddenly making it strong). To help with this effort, run the
 leelaz-autogtp binary provided in this package. The best-known network weights
 file is at http://zero.sjeng.org/best-network

leela-zero-dbgsym: debug symbols for leela-zero