Comments (3)
Which version of CrazyAra are you using?
Both the latest C++ release, CrazyAra 0.8.0, and the last python release, CrazyAra 0.5.1, have a routine to skip the search if only a single move is legal:
Does this also occur if you set Moves:
to 0
in CuteChess' game configuration window which means that both engines have 10 minutes available for the whole match?
from crazyara.
I'm using CrazyAra 0.8.0 for MacOS. This issue occurs even if I set Moves
to 0
in CuteChess' game configuration window. You can test it yourself with the following FEN: 3r1rk1/ppp1Nppp/3p4/2bPp3/2Bn2b1/3P1N2/PPP2PPP/R2Q1RK1/Qbn b - - 2 11
.
With 10 minutes available for the whole match, CrazyAra is taking about 20 seconds on that move on my laptop.
from crazyara.
Thank you @DeFilippis for providing an example FEN and the version number.
I was able to replicate this issue.
It seems that single move are only played instantly if parts of the former tree are reused.
If CrazyAra is run with a GPU back-end this is the default case (e.g. move 25. Rxg1 {+21.61/10 0.016s}
in the following PGN).
[Event "My Tournament"]
[Site "?"]
[Date "2021.01.14"]
[Round "5"]
[White "CrazyAra-0.8.0-risev2-BSize-16-FP16"]
[Black "CrazyAra-0.8.0-risev2-BSize-16-FP16"]
[Result "1-0"]
[GameDuration "00:00:12"]
[GameEndTime "2021-01-14T17:12:43.380 CET"]
[GameStartTime "2021-01-14T17:12:30.854 CET"]
[PlyCount "59"]
[TimeControl "10+0.1"]
[Variant "crazyhouse"]
1. e4 {book} c5 {book} 2. Nf3 {book} e6 {book} 3. d4 {book} cxd4 {book}
4. Nxd4 {book} Nc6 {book} 5. Nc3 {book} Qc7 {book} 6. Be3 {book} a6 {book}
7. Bd3 {book} b5 {book} 8. O-O {book} Bb7 {book} 9. Qe2 {book} Nf6 {book}
10. f4 {book} Nxd4 {book} 11. Bxd4 {+5.61/22 0.31s} N@h5 {-5.14/21 0.30s}
12. e5 {+7.22/23 0.31s} P@g3 {-6.66/22 0.31s} 13. P@b6 {+6.95/18 0.31s}
Qc6 {-6.70/16 0.31s} 14. N@e4 {+8.75/21 0.32s} Nxe4 {-8.32/21 0.32s}
15. Bxe4 {+9.89/22 0.32s} Qc5 {-9.32/18 0.32s} 16. N@c7+ {+10.80/20 0.32s}
Kd8 {-10.30/18 0.32s} 17. Bxc5 {+11.29/17 0.33s} Bxc5+ {-10.97/20 0.32s}
18. Kh1 {+11.62/17 0.33s} gxh2 {-11.30/16 0.33s} 19. Q@h4+ {+12.51/15 0.33s}
B@e7 {-12.34/16 0.33s} 20. Qexh5 {+14.43/14 0.34s} Bxh4 {-13.63/14 0.34s}
21. Qxh4+ {+15.08/11 0.34s} f6 {-14.06/13 0.34s} 22. Nxa8 {+17.41/15 0.35s}
Bxa8 {-14.23/12 0.35s} 23. R@b8+ {+20.80/13 0.35s} N@c8 {-20.64/12 0.35s}
24. exf6 {+28.11/12 0.35s} Q@g1+ {-23.34/11 0.35s} 25. Rxg1 {+21.61/10 0.016s}
hxg1=Q+ {-24.10/10 0.36s} 26. Rxg1 {+18.77/8 0.004s} R@h2+ {-21.27/8 0.36s}
27. Kxh2 {+36.09/8 0.41s} Bxg1+ {-28.24/7 0.37s} 28. Kxg1 {+35.80/6 0.42s}
P@f2+ {-25.28/5 0.37s} 29. Qxf2 {+41.35/5 0.43s} R@e8 {-75.09/2 0.38s}
30. Q@c7# {+M1/1 0.014s, White mates} 1-0
I plan to fix this bug for the upcoming release.
The current NN-backend for Mac also doesn't use Intel-MKL yet which should roughly double the number of nodes per second.
Moreover, I managed to calibrate INT8 weights for the Intel-MKL backend which will be provided in the next release.
from crazyara.
Related Issues (20)
- Implement loss condition when repeatedly checking in Xiangqi
- Could not install Classic Ara on Both Linux and Windows HOT 13
- 3-fold Repetition Bug
- lichess.org bot issues HOT 2
- High memory usage (original: Tree doesn't grow beyond 6 million Nodes) HOT 11
- Mistaken Evaluations by Classic Ara HOT 3
- ClassicAra does not load using the current model. HOT 2
- Crash for ClassicAra 0.9.5 with Int8 Inference
- Still Losses on Time for ClassicAra 0.9.5 HOT 1
- Not able to execute ClassicAra 0.9.5.post0 HOT 2
- Classic Ara stops working after a few moves and needs re-opening in Chessbase 16 HOT 5
- TensorRT Memory Free Error HOT 4
- Hello, can you make a Chinese chess release HOT 6
- Classic Ara 0.9.9 threads use HOT 1
- ClassicAra fails at winning five piece end-game with table bases
- About the transformer. HOT 5
- Wrong replay memory inclusion in fileio.py HOT 1
- Question about the training of the released classic ara model HOT 1
- Incorrect thesis title? HOT 1
- Engine sometimes accepts draws in winning position
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from crazyara.