Fat Fritz 2

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Locked
Modern Times
Posts: 2769
Joined: Thu Jun 07, 2012 9:02 pm

Re: Fat Fritz 2

Post by Modern Times » Sat Feb 13, 2021 10:03 am

jshriver wrote:
Sat Feb 13, 2021 9:53 am

My $0.02. Personally I think its a money grab
I'm not so sure. Watching that hour long video, rightly or wrongly Albert came across as a bit of an egotist. So I think it is more of a personal crusade for him rather than about the money. I have no idea what chessbase sales figures are normally like, but I'd be surprised if there is a much if any profit in it after the development costs are recouped.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

MTaktikos
Posts: 16
Joined: Fri Oct 25, 2019 5:58 pm
Full name: Michael Taktikos

Re: Fat Fritz 2

Post by MTaktikos » Sat Feb 13, 2021 10:26 am

jshriver wrote:
Sat Feb 13, 2021 9:35 am
I cloned the code base when this was post and able to compiled under Linux. Actually now I think about it you have a point. Because even if you can compile it, its useless unless you have the weights file
In the meantime, the github link of DanielUranga includes also the weights file.
Although I doubt it is the same weights file offered by Chessbase, I compiled the published code from github and it works:
https://filehorst.de/d/dBDCEnDj

acepoint_de
Posts: 53
Joined: Mon Jun 10, 2013 11:14 pm

Re: Fat Fritz 2

Post by acepoint_de » Sat Feb 13, 2021 10:31 am

Dann Corbit wrote:
Sat Feb 13, 2021 7:22 am
Personally, since the source and net have been published, I am somewhat neutral about it.
Are you sure the net provided at the 11th of Feb 2021 at Github is the same net they sell? I would not put any money on it. And, apart from all the legal stuff discussion, Chessbase is selling nothing but a «Mogelpackung» as we say in Germany.

Ciao

acepoint

User avatar
Sylwy
Posts: 3600
Joined: Fri Apr 21, 2006 2:19 pm
Location: IASI (Romania) - the historical capital of MOLDOVA

Re: Fat Fritz 2

Post by Sylwy » Sat Feb 13, 2021 10:41 am

MTaktikos wrote:
Sat Feb 13, 2021 10:26 am
jshriver wrote:
Sat Feb 13, 2021 9:35 am
I cloned the code base when this was post and able to compiled under Linux. Actually now I think about it you have a point. Because even if you can compile it, its useless unless you have the weights file
In the meantime, the github link of DanielUranga includes also the weights file.
Although I doubt it is the same weights file offered by Chessbase, I compiled the published code from github and it works:
https://filehorst.de/d/dBDCEnDj
YES ! Works perfectly. Thank you ! Interesting new net- 20x512. Here's the situation in my test after just 8 games:

Image

pohl4711
Posts: 1544
Joined: Sat Sep 03, 2011 5:25 am
Location: Berlin, Germany
Contact:

Re: Fat Fritz 2

Post by pohl4711 » Sat Feb 13, 2021 10:45 am

acepoint_de wrote:
Sat Feb 13, 2021 10:31 am
Dann Corbit wrote:
Sat Feb 13, 2021 7:22 am
Personally, since the source and net have been published, I am somewhat neutral about it.
Are you sure the net provided at the 11th of Feb 2021 at Github is the same net they sell? I would not put any money on it. And, apart from all the legal stuff discussion, Chessbase is selling nothing but a «Mogelpackung» as we say in Germany.

Ciao

acepoint
The Net on GitHub is not the net, which comes with the commercial version of FF2. It is an earlier, weaker version, Albert Silver told me. Of course. Otherwise nobody had to buy FF2 anymore.
I checked the bench and it is different. And I looked into both nets with a HEX-editor and they are definitly different!

The FF2-binary is forked from Stockfish 210111, by the way.

I started a RoundRobin tournament of Stockfish 210111, FF2 and FF2 with github-net. 3000 games. To measure, how strong FF2 and FF2 with github-net really are.
Result on my website in 2-3 days.

acepoint_de
Posts: 53
Joined: Mon Jun 10, 2013 11:14 pm

Re: Fat Fritz 2

Post by acepoint_de » Sat Feb 13, 2021 10:54 am

pohl4711 wrote:
Sat Feb 13, 2021 10:45 am
The Net on GitHub is not the net, which comes with the commercial version of FF2. It is an earlier, weaker version, Albert Silver told me. Of course. Otherwise nobody had to buy FF2 anymore.
I checked the bench and it is different.
I can't say that I'm surprised ;-). My question was obviously more of a rhetorical nature. Anyhow, thanks for your answer and your testing.

Ciao

acepoint

User avatar
Sylwy
Posts: 3600
Joined: Fri Apr 21, 2006 2:19 pm
Location: IASI (Romania) - the historical capital of MOLDOVA

Re: Fat Fritz 2

Post by Sylwy » Sat Feb 13, 2021 11:04 am

pohl4711 wrote:
Sat Feb 13, 2021 10:45 am
acepoint_de wrote:
Sat Feb 13, 2021 10:31 am
Dann Corbit wrote:
Sat Feb 13, 2021 7:22 am
Personally, since the source and net have been published, I am somewhat neutral about it.
Are you sure the net provided at the 11th of Feb 2021 at Github is the same net they sell? I would not put any money on it. And, apart from all the legal stuff discussion, Chessbase is selling nothing but a «Mogelpackung» as we say in Germany.

Ciao

acepoint
The Net on GitHub is not the net, which comes with the commercial version of FF2. It is an earlier, weaker version, Albert Silver told me. Of course. Otherwise nobody had to buy FF2 anymore.
I checked the bench and it is different. And I looked into both nets with a HEX-editor and they are definitly different!

The FF2-binary is forked from Stockfish 210111, by the way.

I started a RoundRobin tournament of Stockfish 210111, FF2 and FF2 with github-net. 3000 games. To measure, how strong FF2 and FF2 with github-net really are.
Result on my website in 2-3 days.
:lol: This means that CCRL also tested an old version of the 20x512 net! :lol:

Image

FatFritz 2 x64.................3634 (493 games)
Stockfish 12 x64..............3631 (2699 games)

User avatar
jshriver
Posts: 1284
Joined: Wed Mar 08, 2006 8:41 pm
Location: Morgantown, WV, USA
Contact:

Re: Fat Fritz 2

Post by jshriver » Sat Feb 13, 2021 11:16 am

Anyone have luck using FF2 w/ the new FatFritz2_v1.bin? i recompiled after downloading the new net in place of the src default one, then tried the following:

setoption name EvalFile value ./FatFritz2_v1.bin

also

setoption name EvalFile value /home/jshriver/Chess/FF2/FatFritz2_v1.bin

No matter it still exits out with this error:
info string ERROR: If the UCI option "Use NNUE" is set to true, network evaluation parameters compatible with the engine must be available.
info string ERROR: The option is set to true, but the network file FatFritz2_v1.bin was not loaded successfully.
info string ERROR: The UCI option EvalFile might need to specify the full path, including the directory name, to the network file.
info string ERROR: The default net can be downloaded from: https://tests.stockfishchess.org/api/nn ... tz2_v1.bin
info string ERROR: The engine will be terminated now.

User avatar
jshriver
Posts: 1284
Joined: Wed Mar 08, 2006 8:41 pm
Location: Morgantown, WV, USA
Contact:

Re: Fat Fritz 2

Post by jshriver » Sat Feb 13, 2021 12:21 pm

jshriver wrote:
Sat Feb 13, 2021 11:16 am
Anyone have luck using FF2 w/ the new FatFritz2_v1.bin? i recompiled after downloading the new net in place of the src default one, then tried the following:
Disregard, needed to update to the latest codebase for it to work. But odd that the older one didnt try to embed the new one.

Related note, anyone here try to build a 512 net? Both gensfen and train are invalid commands in this release.

nio
Posts: 6
Joined: Sun Jan 17, 2021 6:27 pm
Full name: Adi Fischet

Re: Fat Fritz 2

Post by nio » Sat Feb 13, 2021 12:46 pm

So is the download source now exactly the same as the paid one?

Locked