Chess Challenge submission by totobobo1111

README.md

@@ -0,0 +1,26 @@
+---
+library_name: transformers
+tags:
+- chess
+- llm-course
+- chess-challenge
+license: mit
+---
+
+# chess-troy
+
+Chess model submitted to the LLM Course Chess Challenge.
+
+## Submission Info
+
+- **Submitted by**: [totobobo1111](https://huggingface.co/totobobo1111)
+- **Parameters**: 923,472
+- **Organization**: LLM-course
+
+## Model Details
+
+- **Architecture**: Chess Transformer (GPT-style)
+- **Vocab size**: 78
+- **Embedding dim**: 132
+- **Layers**: 5
+- **Heads**: 4

config.json

@@ -0,0 +1,20 @@
+{
+  "architectures": [
+    "ChessForCausalLM"
+  ],
+  "bos_token_id": 1,
+  "dropout": 0.1,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "chess_transformer",
+  "n_ctx": 256,
+  "n_embd": 132,
+  "n_head": 4,
+  "n_inner": 396,
+  "n_layer": 5,
+  "pad_token_id": 0,
+  "tie_weights": true,
+  "transformers_version": "4.57.6",
+  "vocab_size": 78
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e081e90c1b60660d8f3f6149216c53601f447a30e215b0ba0643796b18304d8
+size 3699312

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "[BOS]",
+  "eos_token": "[EOS]",
+  "pad_token": "[PAD]",
+  "unk_token": "[UNK]"
+}

tokenizer.py

@@ -0,0 +1,190 @@
+from __future__ import annotations
+
+import json
+import os
+from typing import Dict, List, Optional
+import re
+
+from transformers import PreTrainedTokenizer
+
+
+class ChessTokenizer(PreTrainedTokenizer):
+    """
+    Chess tokenizer with structured move tokens:
+    Each move is split into: [side][piece][from][to][suffixes].
+    Example:
+        "WPe2e4 BNg8xf6+" -> [W][P][e2][e4] [B][N][g8][f6][x][+]
+    """
+    
+    model_input_names = ["input_ids", "attention_mask"]
+    vocab_files_names = {"vocab_file": "vocab.json"}
+    
+    # Special tokens
+    PAD_TOKEN = "[PAD]"
+    BOS_TOKEN = "[BOS]"
+    EOS_TOKEN = "[EOS]"
+    UNK_TOKEN = "[UNK]"
+    
+    MOVE_RE = re.compile(
+        r"^(?P<side>[WB])"
+        r"(?P<piece>[PNBRQK])"
+        r"(?P<src>[a-h][1-8])"
+        r"(?P<dst>[a-h][1-8])"
+        r"(?P<suffix>.*)$"
+    )
+    
+    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
+
+        # Remove duplicates from kwargs
+        kwargs.pop("pad_token", None)
+        kwargs.pop("bos_token", None)
+        kwargs.pop("eos_token", None)
+        kwargs.pop("unk_token", None)
+        
+        # Load or create vocab
+        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_default_vocab()
+        
+        # Reverse mapping
+        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,
+        )
+    
+    def _create_default_vocab(self) -> Dict[str, int]:
+        special = [self.PAD_TOKEN, self.BOS_TOKEN, self.EOS_TOKEN, self.UNK_TOKEN]
+
+        sides = ["[W]", "[B]"]
+        pieces = ["[P]", "[N]", "[B]", "[R]", "[Q]", "[K]"]
+        squares = [f"[{f}{r}]" for f in "abcdefgh" for r in "12345678"]
+        suffixes = ["[x]", "[+]", "[#]", "[O-O]", "[O-O-O]",
+                    "[prom_Q]", "[prom_R]", "[prom_B]", "[prom_N]"]
+
+        vocab_list = special + sides + pieces + squares + suffixes
+        return {tok: i for i, tok in enumerate(vocab_list)}
+    
+    @classmethod
+    def build_vocab_from_iterator(cls, iterator, min_frequency: int = 1) -> "ChessTokenizer":
+        from collections import Counter
+
+        token_counts = Counter()
+        tokenizer = cls()
+
+        for game in iterator:
+            tokens = tokenizer._tokenize(game)
+            token_counts.update(tokens)
+
+        # Keep tokens meeting frequency threshold
+        tokens = [t for t, c in token_counts.items() if c >= min_frequency]
+        tokens = sorted(tokens)
+
+        special = [cls.PAD_TOKEN, cls.BOS_TOKEN, cls.EOS_TOKEN, cls.UNK_TOKEN]
+        vocab = {tok: i for i, tok in enumerate(special + tokens)}
+
+        return cls(vocab=vocab)
+    
+    @classmethod
+    def build_vocab_from_dataset(
+        cls,
+        dataset_name: str = "dlouapre/lichess_2025-01_1M",
+        split: str = "train",
+        column: str = "text",
+        min_frequency: int = 500,
+        max_samples: Optional[int] = 100000,
+    ) -> "ChessTokenizer":
+        from datasets import load_dataset
+
+        dataset = load_dataset(dataset_name, split=split)
+        if max_samples is not None:
+            dataset = dataset.select(range(min(max_samples, len(dataset))))
+
+        def game_iterator():
+            for example in dataset:
+                yield example[column]
+
+        return cls.build_vocab_from_iterator(game_iterator(), min_frequency=min_frequency)
+    
+    @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]:
+        tokens: List[str] = []
+
+        moves = text.strip().split()
+        for move in moves:
+            # Castling
+            if "O-O-O" in move:
+                tokens.append("[W]" if move.startswith("W") else "[B]")
+                tokens.append("[O-O-O]")
+                continue
+            if "O-O" in move:
+                tokens.append("[W]" if move.startswith("W") else "[B]")
+                tokens.append("[O-O]")
+                continue
+
+            m = self.MOVE_RE.match(move)
+            if not m:
+                tokens.append(self.UNK_TOKEN)
+                continue
+
+            tokens.append(f"[{m.group('side')}]")
+            tokens.append(f"[{m.group('piece')}]")
+            tokens.append(f"[{m.group('src')}]")
+            tokens.append(f"[{m.group('dst')}]")
+
+            suffix = m.group("suffix")
+            if "x" in suffix:
+                tokens.append("[x]")
+            if "+" in suffix:
+                tokens.append("[+]")
+            if "*" in suffix:
+                tokens.append("[#]")
+            if "=" in suffix:
+                promo = suffix.split("=")[-1].upper()
+                tokens.append(f"[prom_{promo}]")
+
+        return tokens
+    
+    def _convert_token_to_id(self, token: str) -> int:
+        return self._vocab.get(token, self._vocab.get(self.UNK_TOKEN, 0))
+    
+    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:
+        special = {self.PAD_TOKEN, self.BOS_TOKEN, self.EOS_TOKEN, self.UNK_TOKEN}
+        return " ".join(t for t in tokens if t not in special)
+    
+    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,)

tokenizer_config.json

@@ -0,0 +1,50 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "[PAD]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "[BOS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "[EOS]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "[UNK]",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "auto_map": {
+    "AutoTokenizer": [
+      "tokenizer.ChessTokenizer",
+      null
+    ]
+  },
+  "bos_token": "[BOS]",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "[EOS]",
+  "extra_special_tokens": {},
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "[PAD]",
+  "tokenizer_class": "ChessTokenizer",
+  "unk_token": "[UNK]"
+}

vocab.json

@@ -0,0 +1,80 @@
+{
+  "[PAD]": 0,
+  "[BOS]": 1,
+  "[EOS]": 2,
+  "[UNK]": 3,
+  "[#]": 4,
+  "[+]": 5,
+  "[B]": 6,
+  "[K]": 7,
+  "[N]": 8,
+  "[P]": 9,
+  "[Q]": 10,
+  "[R]": 11,
+  "[W]": 12,
+  "[a1]": 13,
+  "[a2]": 14,
+  "[a3]": 15,
+  "[a4]": 16,
+  "[a5]": 17,
+  "[a6]": 18,
+  "[a7]": 19,
+  "[a8]": 20,
+  "[b1]": 21,
+  "[b2]": 22,
+  "[b3]": 23,
+  "[b4]": 24,
+  "[b5]": 25,
+  "[b6]": 26,
+  "[b7]": 27,
+  "[b8]": 28,
+  "[c1]": 29,
+  "[c2]": 30,
+  "[c3]": 31,
+  "[c4]": 32,
+  "[c5]": 33,
+  "[c6]": 34,
+  "[c7]": 35,
+  "[c8]": 36,
+  "[d1]": 37,
+  "[d2]": 38,
+  "[d3]": 39,
+  "[d4]": 40,
+  "[d5]": 41,
+  "[d6]": 42,
+  "[d7]": 43,
+  "[d8]": 44,
+  "[e1]": 45,
+  "[e2]": 46,
+  "[e3]": 47,
+  "[e4]": 48,
+  "[e5]": 49,
+  "[e6]": 50,
+  "[e7]": 51,
+  "[e8]": 52,
+  "[f1]": 53,
+  "[f2]": 54,
+  "[f3]": 55,
+  "[f4]": 56,
+  "[f5]": 57,
+  "[f6]": 58,
+  "[f7]": 59,
+  "[f8]": 60,
+  "[g1]": 61,
+  "[g2]": 62,
+  "[g3]": 63,
+  "[g4]": 64,
+  "[g5]": 65,
+  "[g6]": 66,
+  "[g7]": 67,
+  "[g8]": 68,
+  "[h1]": 69,
+  "[h2]": 70,
+  "[h3]": 71,
+  "[h4]": 72,
+  "[h5]": 73,
+  "[h6]": 74,
+  "[h7]": 75,
+  "[h8]": 76,
+  "[x]": 77
+}