tokenizer.py

# src/tokenizer.py
from __future__ import annotations

import json
import os
import re
from typing import Dict, List, Optional

from transformers import PreTrainedTokenizer


class ChessTokenizer(PreTrainedTokenizer):
    """
    Ultra-simple square tokenizer.

    Vocab (68 tokens):
      - 4 specials: [PAD] [BOS] [EOS] [UNK]
      - 64 squares: a1..h8

    Tokenization:
      - Any text containing two squares -> emits those squares as tokens
      - Accepts:
          "WPe2e4(x+)" , "e2e4" , "e2 e4"  -> ["e2","e4"]
      - For longer histories, extracts ALL squares in order.

    Decoding:
      - Joins square tokens with spaces => evaluator regex sees them easily.
    """

    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]"

    _SQUARE_PATTERN = r"[a-h][1-8]"
    _SQUARE_RE = re.compile(_SQUARE_PATTERN)

    def __init__(
        self,
        vocab_file: Optional[str] = None,
        vocab: Optional[Dict[str, int]] = None,
        **kwargs,
    ):
        self._pad_token = self.PAD_TOKEN
        self._bos_token = self.BOS_TOKEN
        self._eos_token = self.EOS_TOKEN
        self._unk_token = self.UNK_TOKEN

        kwargs.pop("pad_token", None)
        kwargs.pop("bos_token", None)
        kwargs.pop("eos_token", None)
        kwargs.pop("unk_token", None)

        if vocab is not None:
            self._vocab = vocab
        elif vocab_file is not None and os.path.exists(vocab_file):
            with open(vocab_file, "r", encoding="utf-8") as f:
                self._vocab = json.load(f)
        else:
            self._vocab = self._create_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,
        )

    @classmethod
    def _create_fixed_vocab(cls) -> Dict[str, int]:
        specials = [cls.PAD_TOKEN, cls.BOS_TOKEN, cls.EOS_TOKEN, cls.UNK_TOKEN]
        files = "abcdefgh"
        ranks = "12345678"
        squares = [f + r for r in ranks for f in files]  # a1..h8
        tokens = specials + squares
        return {tok: i for i, tok in enumerate(tokens)}

    @classmethod
    def build_vocab_from_iterator(cls, iterator, **kwargs) -> "ChessTokenizer":
        return cls(vocab=cls._create_fixed_vocab())

    @classmethod
    def build_vocab_from_dataset(cls, *args, **kwargs) -> "ChessTokenizer":
        return cls(vocab=cls._create_fixed_vocab())

    @property
    def vocab_size(self) -> int:
        return len(self._vocab)

    def get_vocab(self) -> Dict[str, int]:
        return dict(self._vocab)

    def _tokenize(self, text: str) -> List[str]:
        text = text.strip()
        if not text:
            return []

        # Keep BOS/EOS tokens if they appear as standalone strings
        # (rare, but safe)
        if text in {self.PAD_TOKEN, self.BOS_TOKEN, self.EOS_TOKEN, self.UNK_TOKEN}:
            return [text]

        # Extract all squares in order from the text
        squares = self._SQUARE_RE.findall(text)
        if not squares:
            # if nothing parsable, return UNK token
            return [self.UNK_TOKEN]

        # Filter to vocab squares only (should always be true)
        out = [sq for sq in squares if sq in self._vocab]
        return out if out else [self.UNK_TOKEN]

    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:
        # Drop special tokens; join squares with spaces so evaluator can parse.
        special = {self.PAD_TOKEN, self.BOS_TOKEN, self.EOS_TOKEN, self.UNK_TOKEN}
        toks = [t for t in tokens if t not in special]
        return " ".join(toks)

    def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> tuple:
        if not os.path.isdir(save_directory):
            os.makedirs(save_directory, exist_ok=True)

        vocab_file = os.path.join(
            save_directory,
            (filename_prefix + "-" if filename_prefix else "") + "vocab.json",
        )
        with open(vocab_file, "w", encoding="utf-8") as f:
            json.dump(self._vocab, f, ensure_ascii=False, indent=2)

        return (vocab_file,)