Document info line aggregration

[DanijelDomazet]

I've unzipped stockfish-ubuntu-x86-64-avx2.tar and copied only the binary stockfish-ubuntu-x86-64-avx2 to /usr/games/.

I tried to use the chess engine for analysis, but it gives very wrong chess move evaluations. Then I checked with this code:

#!/usr/bin/env python3
import chess
import chess.engine

STOCKFISH_CMD = "/usr/games/stockfish-ubuntu-x86-64-avx2"

with chess.engine.SimpleEngine.popen_uci(STOCKFISH_CMD) as eng:
    # tiny search so we get an 'info string' line back
    info = eng.analyse(chess.Board(), chess.engine.Limit(depth=16))
    banner = info.get("string", "<no NNUE string found>")
    print("Loaded network:", banner)

and got the info:

Loaded network: NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))

Why is Stockfish loading small 6 MB engine instead of normal large one?

When I do
/usr/games/stockfish-ubuntu-x86-64-avx2 go depth 1
it gives me the output:

Stockfish 17.1 by the Stockfish developers (see AUTHORS file)
info string Available processors: 0-11
info string Using 1 thread
info string NNUE evaluation using nn-1c0000000000.nnue (133MiB, (22528, 3072, 15, 32, 1))
info string NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))
info depth 1 seldepth 2 multipv 1 score cp 17 nodes 20 nps 20000 hashfull 0 tbhits 0 time 1 pv e2e4
bestmove e2e4

So the big NNUE is there.

I tried the

eng.configure({"EvalFile": "nn-1c0000000000.nnue"}) 

That too does not work.

Please help, how do I force it to load proper Stockfish?

[DanijelDomazet]

Here is a better example:

import io
import chess.engine
import chess.pgn

pgn_text: str = (
    "1. e4 c5 2. Nc3 d6 3. f4 g6 4. d4 Qa5 5. Be3 Bg7 6. Qd2 Nf6 7. h3 cxd4 8. Qxd4 a6 9. b4 Qd8 10. e5 Nc6 11. Qd2 dxe5 12. Qxd8+ Kxd8 13. Rd1+ Ke8 14. fxe5 Nxe5 15. g4 Be6 16. Bg2 Nc6 17. b5 axb5 18. Nxb5 Rxa2 19. Bxc6+ bxc6 20. Nc7+ Kf8 21. Rd8+ Ne8 22. Rxe8#"
)

game = chess.pgn.read_game(io.StringIO(pgn_text))
board = game.board()

engine = chess.engine.SimpleEngine.popen_uci(
    str("/usr/games/stockfish-ubuntu-x86-64-avx2")
)

for ply, node in enumerate(game.mainline(), start=33):
    board.push(node.move)
    info = engine.analyse(board, chess.engine.Limit(depth=20))

    print(f"[ply {ply}]")
    print("score:", info["score"])
    print("nnue:", info["string"])

engine.close()

For me the this prints:

[ply 38]
score: PovScore(Cp(+16), WHITE)
nnue: NNUE evaluation using nn-37f18f62d772.nnue (6MiB, (22528, 128, 15, 32, 1))

So still the small nn-37f18f62d772.nnue engine...

[winapiadmin]

you have asked this literal question 3 sites (one at chess.SE, one at SO that's possibly closed, and now THIS?)

And it's already answered at the original page that the library code fetches line-by-line, and the info["string"] only store the LAST "info string" info!

@niklasf Shouldn't this be documented?

[DanijelDomazet]

I've resolved this issue with the help of StockFish devs.
Seems like the model used is the proper one, it's just that printout confused me.
Another thing that pointed me to wrong direction is the speed at which evaluations from StockFish were given - it's just that only the current position was evaluated, so this kind of high speed is normal. And lastly, what confused me further was the sign of evaluations; I expected the negative eval if Balck has the advantage, which is also wrong - the evals are always given from the perspective of the side to move.
All in all, several begginer errors from my side. 🥇 Thanks for help.

[winapiadmin]

I've resolved this issue with the help of StockFish devs. Seems like the model used is the proper one, it's just that printout confused me. Another thing that pointed me to wrong direction is the speed at which evaluations from StockFish were given - it's just that only the current position was evaluated, so this kind of high speed is normal. And lastly, what confused me further was the sign of evaluations; I expected the negative eval if Balck has the advantage, which is also wrong - the evals are always given from the perspective of the side to move. All in all, several begginer errors from my side. 🥇 Thanks for help.

Answered HERE

But hey, good to see if you've found the answers:) Don't forget your Chess.SE question