Page 1 of 2

Leela version 0.23.0 for download

Posted: Mon Dec 02, 2019 9:29 pm
by corres
You can download the last Leela (version 0.23.0) from
http://lczero.org
or directly
https://github.com/leelachesszero/lc0/releases
Note
Its source is uncleared and has some bugs.

Re: Leela version 0.23.0 for download

Posted: Mon Dec 02, 2019 10:22 pm
by supersharp77
corres wrote: Mon Dec 02, 2019 9:29 pm You can download the last Leela (version 0.23.0) from
http://lczero.org
or directly
https://github.com/leelachesszero/lc0/releases
Note
Its source is uncleared and has some bugs.
A very new Development...........

"Introduced DNNL BLAS build for modern CPUs in addition to OpenBLAS."......Available in v0.23.0
and In v0.23.0 rc-2

Re: Leela version 0.23.0 for download

Posted: Mon Dec 02, 2019 10:55 pm
by crem

Re: Leela version 0.23.0 for download

Posted: Mon Dec 02, 2019 11:17 pm
by corres
crem wrote: Mon Dec 02, 2019 10:55 pm There is a blog post at https://blog.lczero.org/2019/12/lc0-v02 ... eased.html
As I wrote bugs are in the source code.
I had to repair/modify it to get binary.

Re: Leela version 0.23.0 for download

Posted: Tue Dec 03, 2019 9:36 am
by crem
corres wrote: Mon Dec 02, 2019 11:17 pm
crem wrote: Mon Dec 02, 2019 10:55 pm There is a blog post at https://blog.lczero.org/2019/12/lc0-v02 ... eased.html
As I wrote bugs are in the source code.
I had to repair/modify it to get binary.
Could you be more specific? I can build it just fine.
Alternatively, you can you join our discord at https://discord.gg/pKujYxD, there are more people who can help.

Re: Leela version 0.23.0 for download

Posted: Tue Dec 03, 2019 10:05 am
by corres
crem wrote: Tue Dec 03, 2019 9:36 am
corres wrote: Mon Dec 02, 2019 11:17 pm
crem wrote: Mon Dec 02, 2019 10:55 pm There is a blog post at https://blog.lczero.org/2019/12/lc0-v02 ... eased.html
As I wrote bugs are in the source code.
I had to repair/modify it to get binary.
Could you be more specific? I can build it just fine.
Alternatively, you can you join our discord at https://discord.gg/pKujYxD, there are more people who can help.
Sorry, I no need help.
I used for compilation the same system what I used for lco ver. 0.20-0.21-0.22 and opposite to the present case the course of compilation was faultless. If you also use the earlier compilation system (wrote down in the README.md) and the source what is on Github.com you will find the same faults what I found.

Re: Leela version 0.23.0 for download

Posted: Tue Dec 03, 2019 10:27 am
by zullil
crem wrote: Tue Dec 03, 2019 9:36 am
corres wrote: Mon Dec 02, 2019 11:17 pm
crem wrote: Mon Dec 02, 2019 10:55 pm There is a blog post at https://blog.lczero.org/2019/12/lc0-v02 ... eased.html
As I wrote bugs are in the source code.
I had to repair/modify it to get binary.
Could you be more specific? I can build it just fine.
Alternatively, you can you join our discord at https://discord.gg/pKujYxD, there are more people who can help.
I built from source on Linux, using gcc-9 and g++-9. Simply followed the instructions at https://github.com/LeelaChessZero/lc0, as I did when I built the previous release version. Build was successful, though there were many warnings, mostly repetitions of the following two:

Code: Select all

lc0_lib@sha/proto/net.pb.h:253:13: warning: In the GNU C Library, "minor" is defined
 by <sys/sysmacros.h>. For historical compatibility, it is
 currently defined by <sys/types.h> as well, but we plan to
 remove this soon. To use "minor", include <sys/sysmacros.h>
 directly. If you did not intend to use a system-defined macro
 "minor", you should undefine it after including <sys/types.h>.
  253 |   ::google::protobuf::uint32 minor() const;
      |             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~                                                                                                                                                                                                                                                                                                                                         
lc0_lib@sha/proto/net.pb.h:1763:13: warning: In the GNU C Library, "major" is defined
 by <sys/sysmacros.h>. For historical compatibility, it is
 currently defined by <sys/types.h> as well, but we plan to
 remove this soon. To use "major", include <sys/sysmacros.h>
 directly. If you did not intend to use a system-defined macro
 "major", you should undefine it after including <sys/types.h>.
 1763 | inline ::google::protobuf::uint32 EngineVersion::major() const {
      |             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~                                                                                                                                         

Re: Leela version 0.23.0 for download

Posted: Tue Dec 03, 2019 10:50 am
by zullil
Some folks might find the new Win-Draw-Loss score interesting:
Lc0 now reports WDL score. In addition to showing position score in centipawns, also show probability of winning, losing, and drawing. To enable it, use --show-wdl command-line argument or UCI_ShowWDL UCI option.
For example, for the current position of a certain ongoing game in the Tournaments and Matches forum:

Code: Select all

info depth 19 seldepth 54 time 552264 nodes 14394110 score cp -203 wdl 39 236 725 

Re: Leela version 0.23.0 for download

Posted: Tue Dec 03, 2019 12:02 pm
by corres
I use VS 2015 and auxiliaries for compilation.
Maybe the current src is not portable opposite to the earlier.
My compiler shows faults in version.h, \neural\encoder.cc, \utils\esc_codes.cc (undeclared identifiers, fault in number of arguments, unresolved externals, etc. )

Re: Leela version 0.23.0 for download

Posted: Tue Dec 03, 2019 2:50 pm
by crem
corres wrote: Tue Dec 03, 2019 12:02 pm I use VS 2015 and auxiliaries for compilation.
Maybe the current src is not portable opposite to the earlier.
My compiler shows faults in version.h, \neural\encoder.cc, \utils\esc_codes.cc (undeclared identifiers, fault in number of arguments, unresolved externals, etc. )
Could you provide more data? Like, compiler error messages?

In any case, I suggest joining the discord. It would be much easier to investigate why you have problems building the code this way.