metadata

license: mit
datasets:
  - yp-edu/stockfish-debug
name: yp-edu/gpt2-stockfish-debug
results:
  - task: train
    metrics:
      - name: train-loss
        type: loss
        value: 0.151
        verified: false
      - name: eval-loss
        type: loss
        value: 0.138
        verified: false
widget:
  - text: |-
      FEN: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1
      MOVE:
    example_title: Init Board
  - text: |-
      FEN: r2q1rk1/1p3ppp/4bb2/p2p4/5B2/1P1P4/1PPQ1PPP/R3R1K1 w - - 1 17
      MOVE:
    example_title: Middle Board
  - text: |-
      FEN: 4r1k1/1p1b1ppp/8/8/3P4/2P5/1q3PPP/6K1 b - - 0 28
      MOVE:
    example_title: Checkmate Possible

Model Card for gpt2-stockfish-debug

See my blog post for additional details.

Training Details

The model was trained during 1 epoch on the yp-edu/stockfish-debug dataset (no hyperparameter tuning done). The samples are:

{"prompt":"FEN: {fen}\nMOVE:", "completion": " {move}"}

Two possible simple extensions:

Expand the FEN string: r2qk3/... -> r11qk111/... or equivalent
Condition with the result (ELO not available in the dataset):

{"prompt":"RES: {res}\nFEN: {fen}\nMOVE:", "completion": " {move}"}

Use the Model

The following code requires python-chess (in addition to transformers) which you can install using pip install python-chess.

import chess
from transformers import AutoModelForCausalLM, AutoTokenizer


def next_move(model, tokenizer, fen):
    input_ids = tokenizer(f"FEN: {fen}\nMOVE:", return_tensors="pt")
    input_ids = {k: v.to(model.device) for k, v in input_ids.items()}
    out = model.generate(
        **input_ids,
        max_new_tokens=10,
        pad_token_id=tokenizer.eos_token_id,
        do_sample=True,
        temperature=0.1,
    )
    out_str = tokenizer.batch_decode(out)[0]
    return out_str.split("MOVE:")[-1].replace("<|endoftext|>", "").strip()


board = chess.Board()
model = AutoModelForCausalLM.from_pretrained("yp-edu/gpt2-stockfish-debug")
tokenizer = AutoTokenizer.from_pretrained("yp-edu/gpt2-stockfish-debug")  # or "gpt2"
tokenizer.pad_token = tokenizer.eos_token
for i in range(100):
    fen = board.fen()
    move_uci = next_move(model, tokenizer, fen)
    try:
        print(move_uci)
        move = chess.Move.from_uci(move_uci)
        if move not in board.legal_moves:
            raise chess.IllegalMoveError
        board.push(move)
        outcome = board.outcome()
        if outcome is not None:
            print(board)
            print(outcome.result())
            break
    except chess.IllegalMoveError:
        print(board)
        print("Illegal move", i)
        break
else:
    print(board)