Skip to content

v1.7.2-1.7.6: Strongest distributed model + packaged app for OSX
Compare
Choose a tag to compare
@sanderland sanderland released this 23 Jan 11:00
· 53 commits to master since this release
v1.7.6
34b3f92
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.

This is another minor update, improving on some issues found in 1.7.1, and including better support for downloading models.

I've also managed to create a packaged release for OSX, thanks to kivy developer @matham. This package includes a pre-compiled katago executable and should be all you need to run KaTrain on OSX. So far this has been tested on OSX 10.14.

Features:

  • You can now download the confidently strongest distributed model from general settings.
  • You can now press shift-b to navigate from a deep variation back to the main branch, thanks to @LiamHz

Bug fixes

  • Better stabiility for model downloads.
  • Queries should be properly cleared on new game, preventing an incorrect 'forever busy' engine status.
  • Fix for initial player detection using PL[]

New Theme & New YouTube Tutorial

  • See here for the first theme added.
  • See here for a new YouTube video on using the 'insert moves' feature to improve your ability to review games in situations when both players ignore an urgent move for a long time.

Installation

  • Windows users can simply download and run the KaTrain.exe file attached, or download the .zip file and run the .exe in that (which is slightly faster to start up)
  • MacOS users can use the attached KaTrainOSX.zip (unzip, mount .dmg, drag .app to applications)
    • Note that on the newer Macs with M1 processor you still need to brew install katago
  • Linux and MacOS users can install using pip3 install -U katrain or follow instructions here.

Update 1.7.3

  • Fixes a bug in downloading katago versions.
  • Shuts down faster if an external command is used.
  • Avoids using the provided exe for new M1 macs and looks for a brewed binary instead.
  • Uses an older pyinstaller to give less false positives for antivirus programs

Update 1.7.4

  • Fixed a bug in download models.

Update 1.7.6

  • Better support for M1
Assets 5
Loading