Closed ghost closed 4 years ago
That error message appears because the python-chess module is missing.
I will release an exe probably next week.
thanks
Exe file is released at https://github.com/fsmosca/chess-artist/releases/tag/v2.9
thanks..Unfortunately nothing has changed in terms of speed..For 100 FEN positions with one second, Chess-Artist needs a total of 232 seconds, with old and new exe! That means 232s-100s = 132s only for loading and starting the FENs. It is not acceptable. Please see the two log files. Was tested with Win10 and Lc0.exe (an nn). Please check again why this takes so long. Thank you..
Am Samstag, 11. Januar 2020, 09:43:04 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
Closed #70.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
2020-01-11 10:33:20,184 :: pid: 2276 :: tid: 3892 :: DEBUG :: >> uci
2020-01-11 10:33:20,200 :: pid: 2276 :: tid: 3892 :: DEBUG :: >> quit
2020-01-11 10:33:20,318 :: pid: 2276 :: tid: 3892 :: INFO :: Test engine with epd suite
2020-01-11 10:33:20,334 :: pid: 2276 :: tid: 3892 :: INFO :: epd bm: ['Qd2,', 'Bc2']
2020-01-11 10:33:20,334 :: pid: 2276 :: tid: 3892 :: DEBUG :: >> uci
2020-01-11 10:33:20,349 :: pid: 2276 :: tid: 3892 :: DEBUG :: <<
2020-01-11 10:33:20,349 :: pid: 2276 :: tid: 3892 :: DEBUG :: << | | |
2020-01-11 10:33:20,349 :: pid: 2276 :: tid: 3892 :: DEBUG :: << | | |_| v0.23.2+git.c8d9095 built Dec 31 2019
2020-01-11 10:33:20,365 :: pid: 2276 :: tid: 3892 :: DEBUG :: << id name Lc0 v0.23.2+git.c8d9095
2020-01-11 10:33:20,365 :: pid: 2276 :: tid: 3892 :: DEBUG :: << id author The LCZero Authors.
2020-01-11 10:33:20,365 :: pid: 2276 :: tid: 3892 :: DEBUG :: << option name WeightsFile type string default
thanks..Unfortunately nothing has changed in terms of speed..For 100 FEN positions with one second, Chess-Artist needs a total of 232 seconds, with old and new exe! That means 232s-100s = 132s only for loading and starting the FENs. It is not acceptable.
That is normal. Chess artist is not an ordinary analyzer, as it will analyze the move made in the actual game and save its score, after that it will also analyze the position before the actual move is made on the board and save the best move and score found by the engine. If the best move of engine and the move in the actual game are the same, Chess artist (depending on some conditions) will analyze again and try to find the threat of the move. This threat move is shown in the pgn output, like planning ... , with the idea of ..., and followed by ... comments.
Example,
7. Bc2 {+0.21} (7. g3 Bf7 8. Qe2 d6 9. h4 {+0.45}) 7... O-O-O $1 {+0.49, followed by Bf7}
8. O-O-O $6 {-0.17} (8. h4 Bxe3 9. Bxe3 Bf7 10. Qg3 {+0.49}) 8... Bf7 $1 {+0.41, planning Bh5}
9. Kb1 {-0.25} 9... Bh5 $1 {+0.46, followed by Bxd1}
10. Rc1 $5 {-0.26} 10... Rf8 $3 {+0.16, with the idea of Kb8}
If the best move of engine and actual move in the game are not the same, chess artist will calculate move NAG's, if the score of the move in the actual game is better than the score of the best move of the engine, the move in the actual game will be given a very good move !! or a good move ! annotation. Or depending on the score, and position complexity it will give !?, ?!, ? and ?? symbols to the move. It will also add pv of the engine.
Example,
18. Ka1 $6 {-0.51} (18. Qd2 Qe7 19. Qg5 h5 20. Rgf1 {-0.09})
Here white's Ka1 is dubious, see the meaning of $6 at https://en.wikipedia.org/wiki/Numeric_Annotation_Glyphs
Engine recommends Qd2 with a score of -0.09 pawn unit or -9 centipawn unit. Compared to the score of Ka1 which is only -0.51, a bit worse.
It will also try to find if one side sacrifices material.
If you use Stockfish engine, it will take additional time because it will calculate:
and would include those in the comment.
Example,
1. f4 $5 {+0.08, with a better piece mobility} 1... Nb6 $3 {+0.58, with the idea of f5}
2. Nb3 $3 {-0.27, planning e4} 2... f6 {+0.46}
3. Qg3 $1 {-0.19, planning Qxg7} 3... g6 $6 {+0.73} (3...Bf7 {-0.05})
4. e4 $5 {+0.00} 4... c6 $6 {+0.27} (4...h5 {+0.00})
5. Be3 $2 {-0.15} (5. d3 {+0.43}) 5... Bc7 $5 {+0.60}
6. Qf2 $6 {-0.35} (6. d3 {+0.73}) 6... O-O-O $6 {+0.50} (6...h5 7. d3 d5 8. Nfd2 e5 {+0.00})
7. d3 $5 {-0.27} 7... Kb8 $6 {+0.59} (7...d5 {+0.01})
8. Nfd2 {+0.04} 8... d5 $6 {+0.73} (8...h5 9. a4 Nc8 10. Be2 e5 {+0.00})
9. a4 {+0.22} (9. Be2 {+0.51}) 9... dxe4 $5 {+0.65, white had sacrificed material}
10. a5 $2 {+0.24} (10. dxe4 {+0.84}) 10... Nc8 {+0.64}
Chess artist works, that's clear to me .. I am only interested in why so much time about 1.6 seconds is needed to start for each FEN (see time interval between lines 74 and 75 in the new exe v2.9). line 74. 2020-01-11 10: 28: 57,979 :: pid: 2596 :: tid: 3176 :: DEBUG :: << GPU compute capability: 7.5 line 75. 2020-01-11 10: 28: 59,637 :: pid: 2596 :: tid: 3176 :: DEBUG :: << readyok. With stockfish (see the attached log files) the same test with the same pgn, and 100 FEN, only takes 122 seconds instead of 232 seconds at lc0..that is great for stockfish. And it should work just as quickly with lc0. Also with stockfish is all right, it's quick, but unfortunately not with lc0. Therefore you can only use chess-artist with stockfish at the moment. Time costs money .. Please try again and see what the reason is .. thanks
Am Samstag, 11. Januar 2020, 12:41:58 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
thanks..Unfortunately nothing has changed in terms of speed..For 100 FEN positions with one second, Chess-Artist needs a total of 232 seconds, with old and new exe! That means 232s-100s = 132s only for loading and starting the FENs. It is not acceptable.
That is normal. Chess artist is not an ordinary analyzer, as it will analyze the move made in the actual game and save its score, after that it will also analyze the position before the actual move is made on the board and save the best move and score found by the engine. If the best move of engine and the move in the actual game are the same, Chess artist (depending on some conditions) will analyze again and try to find the threat of the move. This threat move is shown in the pgn output, like planning ... , with the idea of ..., and followed by ... comments.
Example,
If the best move of engine and actual move in the game are not the same, chess artist will calculate move NAG's, if the score of the actual game is better than the score of the engine, the move in the actual game will be given a very good move !! or a good move ! annotation. Or depending on the score, and position complexity it will give !?, ?!, ? and ?? symbols to the move. It will also add pv of the engine.
Example,
Here white's Ka1 is dubious, see the meaning of $6 at https://en.wikipedia.org/wiki/Numeric_Annotation_Glyphs
Engine recommends Qd2 with a score of -0.09 pawn unit or 9 centipawn unit. Compared to the score of Ka1 which -0.51 which is a bit worse.
Stockfish as analyzer
If you use Stockfish engine, it will take additional time because it will calculate:
and would include those in the comment.
Example,
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
2020-01-11 11:01:41,529 :: pid: 1552 :: tid: 1044 :: DEBUG :: >> uci
2020-01-11 11:01:41,608 :: pid: 1552 :: tid: 1044 :: DEBUG :: >> quit
2020-01-11 11:01:41,608 :: pid: 1552 :: tid: 1044 :: INFO :: Test engine with epd suite
2020-01-11 11:01:41,608 :: pid: 1552 :: tid: 1044 :: INFO :: epd bm: ['Qd2,', 'Bc2']
2020-01-11 11:01:41,608 :: pid: 1552 :: tid: 1044 :: DEBUG :: >> uci
2020-01-11 11:01:41,608 :: pid: 1552 :: tid: 1044 :: DEBUG :: << Stockfish 251219 64 POPCNT by T. Romstad, M. Costalba, J. Kiiski, G. Linscott
2020-01-11 11:01:41,686 :: pid: 1552 :: tid: 1044 :: DEBUG :: << id name Stockfish 251219 64 POPCNT
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << id author T. Romstad, M. Costalba, J. Kiiski, G. Linscott
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: <<
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Debug Log File type string default
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Contempt type spin default 24 min -100 max 100
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Analysis Contempt type combo default Both var Off var White var Black var Both
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Threads type spin default 1 min 1 max 512
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Hash type spin default 16 min 1 max 131072
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Clear Hash type button
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Ponder type check default false
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name MultiPV type spin default 1 min 1 max 500
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Skill Level type spin default 20 min 0 max 20
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Move Overhead type spin default 30 min 0 max 5000
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Minimum Thinking Time type spin default 20 min 0 max 5000
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name Slow Mover type spin default 84 min 10 max 1000
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name nodestime type spin default 0 min 0 max 10000
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name UCI_Chess960 type check default false
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name UCI_AnalyseMode type check default false
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name UCI_LimitStrength type check default false
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name UCI_Elo type spin default 1350 min 1350 max 2850
2020-01-11 11:01:41,701 :: pid: 1552 :: tid: 1044 :: DEBUG :: << option name SyzygyPath type string default
its a epd not a pgn datei..
chess-artist.exe --infile 100.epd --outfile result.txt ^ --enginefile "LD2.exe --backend=cudnn-fp16" ^ --engineoptions "Hash value 128" ^ --movetime 1000 --log ^ --job test --eval search
pause
result :: EPD 100.epd TEST RESULTS :: Engine : Lc0 v0.23.2+git.c8d9095 Time/pos (sec): 1.0
Total epd lines : 100 Total tested positions: 100 Total correct : 61 Correct percentage : 61.00
Am Samstag, 11. Januar 2020, 15:09:23 MEZ hat as as <wiesas@yahoo.com> Folgendes geschrieben:
Chess artist works, that's clear to me .. I am only interested in why so much time about 1.6 seconds is needed to start for each FEN (see time interval between lines 74 and 75 in the new exe v2.9). line 74. 2020-01-11 10: 28: 57,979 :: pid: 2596 :: tid: 3176 :: DEBUG :: << GPU compute capability: 7.5 line 75. 2020-01-11 10: 28: 59,637 :: pid: 2596 :: tid: 3176 :: DEBUG :: << readyok. With stockfish (see the attached log files) the same test with the same pgn, and 100 FEN, only takes 122 seconds instead of 232 seconds at lc0..that is great for stockfish. And it should work just as quickly with lc0. Also with stockfish is all right, it's quick, but unfortunately not with lc0. Therefore you can only use chess-artist with stockfish at the moment. Time costs money .. Please try again and see what the reason is .. thanks
Am Samstag, 11. Januar 2020, 12:41:58 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
thanks..Unfortunately nothing has changed in terms of speed..For 100 FEN positions with one second, Chess-Artist needs a total of 232 seconds, with old and new exe! That means 232s-100s = 132s only for loading and starting the FENs. It is not acceptable.
That is normal. Chess artist is not an ordinary analyzer, as it will analyze the move made in the actual game and save its score, after that it will also analyze the position before the actual move is made on the board and save the best move and score found by the engine. If the best move of engine and the move in the actual game are the same, Chess artist (depending on some conditions) will analyze again and try to find the threat of the move. This threat move is shown in the pgn output, like planning ... , with the idea of ..., and followed by ... comments.
Example,
If the best move of engine and actual move in the game are not the same, chess artist will calculate move NAG's, if the score of the actual game is better than the score of the engine, the move in the actual game will be given a very good move !! or a good move ! annotation. Or depending on the score, and position complexity it will give !?, ?!, ? and ?? symbols to the move. It will also add pv of the engine.
Example,
Here white's Ka1 is dubious, see the meaning of $6 at https://en.wikipedia.org/wiki/Numeric_Annotation_Glyphs
Engine recommends Qd2 with a score of -0.09 pawn unit or 9 centipawn unit. Compared to the score of Ka1 which -0.51 which is a bit worse.
Stockfish as analyzer
If you use Stockfish engine, it will take additional time because it will calculate:
and would include those in the comment.
Example,
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
its a epd not a pgn datei.. 2. Test engine with test suite :: Annotate games chess-artist.exe --infile 100.epd --outfile result.txt ^ --enginefile "LD2.exe --backend=cudnn-fp16" ^ --engineoptions "Hash value 128" ^ --movetime 1000 --log ^ --job test --eval search pause result :: EPD 100.epd TEST RESULTS :: Engine : Lc0 v0.23.2+git.c8d9095 Time/pos (sec): 1.0 Total epd lines : 100 Total tested positions: 100 Total correct
Can you create another issue?
"Can you create another issue?" You mean "https://github.com/fsmosca/chess-artist/issues"?
Am Samstag, 11. Januar 2020, 15:28:35 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
its a epd not a pgn datei.. 2. Test engine with test suite :: Annotate games chess-artist.exe --infile 100.epd --outfile result.txt ^ --enginefile "LD2.exe --backend=cudnn-fp16" ^ --engineoptions "Hash value 128" ^ --movetime 1000 --log ^ --job test --eval search pause result :: EPD 100.epd TEST RESULTS :: Engine : Lc0 v0.23.2+git.c8d9095 Time/pos (sec): 1.0 Total epd lines : 100 Total tested positions: 100 Total correct
Can you create another issue?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
Added the issue at https://github.com/fsmosca/chess-artist/issues/71
thanks, will there be a new release (exe) after fixing the bug? Am Sonntag, 12. Januar 2020, 02:47:49 MEZ hat fsmosca notifications@github.com Folgendes geschrieben:
Added the issue at #71
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
It is not a bug. The engine is restarted in every test position so that engine memory is cleared/erased without doubt. Stockfish and other A/B engines are quick to restart but not Lc0 because of some initialization.
Will try to release an exe next week.
thanks
Am Sonntag, 12. Januar 2020, 12:10:07 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
It is not a bug. The engine is restarted in every test position so that engine memory is cleared/erased without doubt. Stockfish and other A/B engines are quick to restart but not Lc0 because of some initialization.
Will try to release an exe next week.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
A function could perhaps be added to chess-artist "2. Test engine with test suite" (if it is possible). There is the function "--depth 16 --movetime 0" and the function eg. "--movetime 500". Can you still use a function eg. Install "--nodes 1000"? This is important for lc0 because there are different nn (20x256, 10x128, 24x320). These nn take different times. The big nns are slower than the small nns and therefore have a disadvantage with "--movetime". With eg "--nodes 1000" is a little better to compare. Can you incorporate that into the exe? Thanks ..
Am Sonntag, 12. Januar 2020, 13:07:45 MEZ hat as as <wiesas@yahoo.com> Folgendes geschrieben:
thanks
Am Sonntag, 12. Januar 2020, 12:10:07 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
It is not a bug. The engine is restarted in every test position so that engine memory is cleared/erased without doubt. Stockfish and other A/B engines are quick to restart but not Lc0 because of some initialization.
Will try to release an exe next week.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
A function could perhaps be added to chess-artist "2. Test engine with test suite" (if it is possible). There is the function "--depth 16 --movetime 0" and the function eg. "--movetime 500". Can you still use a function eg. Install "--nodes 1000"? This is important for lc0 because there are different nn (20x256, 10x128, 24x320). These nn take different times. The big nns are slower than the small nns and therefore have a disadvantage with "--movetime". With eg "--nodes 1000" is a little better to compare. Can you incorporate that into the exe? Thanks ..
Create a new issue with regards to that request. Not on this issue.
ok, but first I wait until the new "v2.10.exe" is available for download.
Am Montag, 13. Januar 2020, 05:23:54 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
A function could perhaps be added to chess-artist "2. Test engine with test suite" (if it is possible). There is the function "--depth 16 --movetime 0" and the function eg. "--movetime 500". Can you still use a function eg. Install "--nodes 1000"? This is important for lc0 because there are different nn (20x256, 10x128, 24x320). These nn take different times. The big nns are slower than the small nns and therefore have a disadvantage with "--movetime". With eg "--nodes 1000" is a little better to compare. Can you incorporate that into the exe? Thanks ..
Create a new issue with regards to that request. Not on this issue.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
meanwhile i looked around because you don't seem to be able to get out with v2.10 .. too bad. There is also a simex2.exe (mea.exe) which is also interesting. However, this does not work with lc0.exe set MT = 100set HASH = 64 set THREADS = 1 set PROTOCOL = uci set EPD = epd \ example.epd set DATABASE = data \ example.data set EXE = engines \ lc0.exe set NAME = lc0. Do you know why that doesn't work?
Am Mittwoch, 15. Januar 2020, 15:02:36 MEZ hat as as <wiesas@yahoo.com> Folgendes geschrieben:
ok, but first I wait until the new "v2.10.exe" is available for download.
Am Montag, 13. Januar 2020, 05:23:54 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
A function could perhaps be added to chess-artist "2. Test engine with test suite" (if it is possible). There is the function "--depth 16 --movetime 0" and the function eg. "--movetime 500". Can you still use a function eg. Install "--nodes 1000"? This is important for lc0 because there are different nn (20x256, 10x128, 24x320). These nn take different times. The big nns are slower than the small nns and therefore have a disadvantage with "--movetime". With eg "--nodes 1000" is a little better to compare. Can you incorporate that into the exe? Thanks ..
Create a new issue with regards to that request. Not on this issue.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
I noticed a bug in mea.exe and chess-artist.exe. For example, if you enter 300ms, only 100ms are actually played, and if you enter 500ms, it is actually only 300ms, and at 200ms it is only about 7ms !! But this only applies to lc0. Stockfish adheres exactly to the time, which is great .. could you improve it, and maybe add a selection of "nodes" to the new realease (exe). Thank you Am Freitag, 17. Januar 2020, 20:44:29 MEZ hat as as wiesas@yahoo.com Folgendes geschrieben:
meanwhile i looked around because you don't seem to be able to get out with v2.10 .. too bad. There is also a simex2.exe (mea.exe) which is also interesting. However, this does not work with lc0.exe set MT = 100set HASH = 64 set THREADS = 1 set PROTOCOL = uci set EPD = epd \ example.epd set DATABASE = data \ example.data set EXE = engines \ lc0.exe set NAME = lc0. Do you know why that doesn't work?
Am Mittwoch, 15. Januar 2020, 15:02:36 MEZ hat as as <wiesas@yahoo.com> Folgendes geschrieben:
ok, but first I wait until the new "v2.10.exe" is available for download.
Am Montag, 13. Januar 2020, 05:23:54 MEZ hat fsmosca <notifications@github.com> Folgendes geschrieben:
A function could perhaps be added to chess-artist "2. Test engine with test suite" (if it is possible). There is the function "--depth 16 --movetime 0" and the function eg. "--movetime 500". Can you still use a function eg. Install "--nodes 1000"? This is important for lc0 because there are different nn (20x256, 10x128, 24x320). These nn take different times. The big nns are slower than the small nns and therefore have a disadvantage with "--movetime". With eg "--nodes 1000" is a little better to compare. Can you incorporate that into the exe? Thanks ..
Create a new issue with regards to that request. Not on this issue.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.
If you have an issue in mea, create an issue in mea repository, not in chess artist.
Unfortunately I get an error message after using cmd
"python chess-artist.py --infile wacnew.epd --outfile result_wacnew.txt --enginefile stockfish_10.exe --engineoptions" Hash value 128, Threads value 1 "--eval search --job test --movetime 500 - -log "
The error is:
Traceback (most recent call last): File "chess-artist.py", line 13, in module import chess.pgn ImportError: No module named chess.pgn
The chess-artist.py is the current one from 12/31/19. Can you maybe create a new release "chess-artist.exe"? With chess-artist.py it doesn't work for me. Thank you