| | """
|
| | Role-marked square tokenizer for Chess Challenge.
|
| | Each move is represented as: <from_square>_f <to_square>_t [promo?] [EOS]
|
| |
|
| | Examples:
|
| | WPe2e4 -> e2_f e4_t [EOS]
|
| | BPe7e8=Q -> e7_f e8_t q [EOS]
|
| | """
|
| | from __future__ import annotations
|
| | import json
|
| | import os
|
| | import re
|
| | from typing import Dict, List, Optional, Tuple, Any
|
| | from transformers import PreTrainedTokenizer
|
| |
|
| | _MOVE_RE = re.compile(r"^([WB])([PNBRQK])([a-h][1-8])([a-h][1-8])(.*)$")
|
| | _PROMO_RE = re.compile(r"=([QRBNqrbn])")
|
| |
|
| | SQUARES = [f"{f}{r}" for r in "12345678" for f in "abcdefgh"]
|
| | PROMOS = ["q", "r", "b", "n"]
|
| |
|
| | class ChessTokenizer(PreTrainedTokenizer):
|
| | model_input_names = ["input_ids", "attention_mask"]
|
| | vocab_files_names = {"vocab_file": "vocab.json"}
|
| |
|
| | PAD_TOKEN = "[PAD]"
|
| | BOS_TOKEN = "[BOS]"
|
| | EOS_TOKEN = "[EOS]"
|
| | UNK_TOKEN = "[UNK]"
|
| |
|
| | def __init__(
|
| | self,
|
| | vocab_file: Optional[str] = None,
|
| | vocab: Optional[Dict[str, int]] = None,
|
| | **kwargs: Any,
|
| | ):
|
| | self._pad_token = self.PAD_TOKEN
|
| | self._bos_token = self.BOS_TOKEN
|
| | self._eos_token = self.EOS_TOKEN
|
| | self._unk_token = self.UNK_TOKEN
|
| |
|
| | for k in ["pad_token", "bos_token", "eos_token", "unk_token"]:
|
| | kwargs.pop(k, None)
|
| |
|
| | if vocab is not None:
|
| | self._vocab = vocab
|
| | elif vocab_file is not None and os.path.isfile(vocab_file):
|
| | with open(vocab_file, "r", encoding="utf-8") as f:
|
| | self._vocab = json.load(f)
|
| | else:
|
| | self._vocab = self._build_fixed_vocab()
|
| |
|
| | self._ids_to_tokens = {v: k for k, v in self._vocab.items()}
|
| |
|
| | super().__init__(
|
| | pad_token=self._pad_token,
|
| | bos_token=self._bos_token,
|
| | eos_token=self._eos_token,
|
| | unk_token=self._unk_token,
|
| | **kwargs,
|
| | )
|
| |
|
| | @staticmethod
|
| | def _build_fixed_vocab() -> Dict[str, int]:
|
| | tokens: List[str] = [
|
| | ChessTokenizer.PAD_TOKEN,
|
| | ChessTokenizer.BOS_TOKEN,
|
| | ChessTokenizer.EOS_TOKEN,
|
| | ChessTokenizer.UNK_TOKEN,
|
| | ]
|
| | tokens += [f"{sq}_f" for sq in SQUARES]
|
| | tokens += [f"{sq}_t" for sq in SQUARES]
|
| | tokens += PROMOS
|
| | return {tok: i for i, tok in enumerate(tokens)}
|
| |
|
| | @property
|
| | def vocab_size(self) -> int:
|
| | return len(self._vocab)
|
| |
|
| | def get_vocab(self) -> Dict[str, int]:
|
| | return dict(self._vocab)
|
| |
|
| | @classmethod
|
| | def build_vocab_from_dataset(cls, *args: Any, **kwargs: Any) -> "ChessTokenizer":
|
| | return cls()
|
| |
|
| | @classmethod
|
| | def build_vocab_from_iterator(cls, *args: Any, **kwargs: Any) -> "ChessTokenizer":
|
| | return cls()
|
| |
|
| | def _tokenize(self, text: str) -> List[str]:
|
| | """Tokenize a space-separated list of dataset moves into role-marked tokens."""
|
| | text = (text or "").strip()
|
| | if not text:
|
| | return []
|
| |
|
| | out: List[str] = []
|
| | for move in text.split():
|
| |
|
| | if move in self._vocab:
|
| | out.append(move)
|
| | continue
|
| |
|
| |
|
| | m = _MOVE_RE.match(move)
|
| | if not m:
|
| |
|
| | if re.fullmatch(r"[a-h][1-8][a-h][1-8][qrbn]?", move):
|
| | src, dst = move[:2], move[2:4]
|
| | out.append(f"{src}_f")
|
| | out.append(f"{dst}_t")
|
| | if len(move) == 5:
|
| | out.append(move[4])
|
| | out.append(self.EOS_TOKEN)
|
| | continue
|
| |
|
| |
|
| | out.append(self.UNK_TOKEN)
|
| | out.append(self.EOS_TOKEN)
|
| | continue
|
| |
|
| |
|
| | _side, _piece, src, dst, suffix = m.groups()
|
| | out.append(f"{src}_f")
|
| | out.append(f"{dst}_t")
|
| |
|
| | promo = None
|
| | pm = _PROMO_RE.search(suffix or "")
|
| | if pm:
|
| | promo = pm.group(1).lower()
|
| |
|
| | if promo in PROMOS:
|
| | out.append(promo)
|
| |
|
| | out.append(self.EOS_TOKEN)
|
| |
|
| | return out
|
| |
|
| | def _convert_token_to_id(self, token: str) -> int:
|
| | return self._vocab.get(token, self._vocab[self.UNK_TOKEN])
|
| |
|
| | def _convert_id_to_token(self, index: int) -> str:
|
| | return self._ids_to_tokens.get(index, self.UNK_TOKEN)
|
| |
|
| | def convert_tokens_to_string(self, tokens: List[str]) -> str:
|
| | return " ".join(tokens)
|
| |
|
| | def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
|
| | os.makedirs(save_directory, exist_ok=True)
|
| | name = "vocab.json" if not filename_prefix else f"{filename_prefix}-vocab.json"
|
| | path = os.path.join(save_directory, name)
|
| | with open(path, "w", encoding="utf-8") as f:
|
| | json.dump(self._vocab, f, indent=2, ensure_ascii=False)
|
| | return (path,) |
