Chess Challenge submission by Vincentime

README.md

@@ -0,0 +1,26 @@
+---
+library_name: transformers
+tags:
+- chess
+- llm-course
+- chess-challenge
+license: mit
+---
+
+# chess-vincentime-rook
+
+Chess model submitted to the LLM Course Chess Challenge.
+
+## Submission Info
+
+- **Submitted by**: [Vincentime](https://huggingface.co/Vincentime)
+- **Parameters**: 999,032
+- **Organization**: LLM-course
+
+## Model Details
+
+- **Architecture**: Chess Transformer (GPT-style)
+- **Vocab size**: 72
+- **Embedding dim**: 112
+- **Layers**: 8
+- **Heads**: 8

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": 112,
+  "n_head": 8,
+  "n_inner": 307,
+  "n_layer": 8,
+  "pad_token_id": 0,
+  "tie_weights": true,
+  "transformers_version": "4.57.5",
+  "vocab_size": 72
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3b22523f2f9331a77eba1642a1d84cc8d860b799bc5459245f388d6b6d009d3
+size 4004616

special_tokens_map.json

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

tokenizer.py

@@ -0,0 +1,124 @@
+from __future__ import annotations
+import json
+import os
+import re
+from typing import Dict, List, Optional
+from transformers import PreTrainedTokenizer
+
+class ChessTokenizer(PreTrainedTokenizer):
+    """
+    Tokenizer déterministe au niveau 'case' (Square-level).
+    Compatible avec les scripts de train/data du projet Chess Challenge.
+    """
+    model_input_names = ["input_ids", "attention_mask"]
+    vocab_files_names = {"vocab_file": "vocab.json"}
+    
+    # Tokens spéciaux identiques au projet original
+    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):
+        self._pad_token = self.PAD_TOKEN
+        self._bos_token = self.BOS_TOKEN
+        self._eos_token = self.EOS_TOKEN
+        self._unk_token = self.UNK_TOKEN
+
+        # Nettoyage des kwargs pour éviter les doublons lors de l'init parent
+        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_square_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 build_vocab_from_dataset(
+        cls,
+        dataset_name: str = "",
+        split: str = "",
+        column: str = "",
+        min_frequency: int = 0,
+        max_samples: Optional[int] = None,
+    ) -> "ChessTokenizer":
+        """
+        Méthode de compatibilité. 
+        Pour le SquareTokenizer, le vocabulaire est fixe, 
+        on ignore donc les arguments et on retourne une instance standard.
+        """
+        print("Square Tokenizer: Using fixed deterministic vocabulary.")
+        return cls()
+
+    def _create_square_vocab(self) -> Dict[str, int]:
+        """Crée le vocabulaire fixe de cases (64) + promos (4) + spéciaux (4)."""
+        special_tokens = [self.PAD_TOKEN, self.BOS_TOKEN, self.EOS_TOKEN, self.UNK_TOKEN]
+        files = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h']
+        ranks = ['1', '2', '3', '4', '5', '6', '7', '8']
+        squares = [f + r for f in files for r in ranks]
+        promotions = ['q', 'r', 'b', 'n']
+        
+        all_tokens = special_tokens + squares + promotions
+        return {token: idx for idx, token in enumerate(all_tokens)}
+
+    # --- MÉTHODES REQUISES POUR HUGGING FACE COMPATIBILITY ---
+    
+    @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]:
+        """Découpe 'WPe2e4' en ['e2', 'e4']."""
+        moves = text.strip().split()
+        tokens = []
+        for m in moves:
+            if m in {self.PAD_TOKEN, self.BOS_TOKEN, self.EOS_TOKEN, self.UNK_TOKEN}:
+                tokens.append(m)
+                continue
+            
+            # Nettoyage Regex : on ne garde que les coordonnées a-h, 1-8 et promos qrbn
+            clean_m = re.sub(r'[\(\)x\+\*WBPNBRQK]', '', m)
+            
+            if len(clean_m) >= 4:
+                tokens.append(clean_m[0:2]) # Case départ
+                tokens.append(clean_m[2:4]) # Case arrivée
+                if len(clean_m) == 5:
+                    tokens.append(clean_m[4]) # Promotion
+        return tokens
+
+    def _convert_token_to_id(self, token: str) -> int:
+        return self._vocab.get(token, self._vocab.get(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:
+        # Utile pour reconstruire le format texte si besoin
+        return "".join(tokens)
+
+    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, (f"{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 +1,74 @@
+{
+  "[PAD]": 0,
+  "[BOS]": 1,
+  "[EOS]": 2,
+  "[UNK]": 3,
+  "a1": 4,
+  "a2": 5,
+  "a3": 6,
+  "a4": 7,
+  "a5": 8,
+  "a6": 9,
+  "a7": 10,
+  "a8": 11,
+  "b1": 12,
+  "b2": 13,
+  "b3": 14,
+  "b4": 15,
+  "b5": 16,
+  "b6": 17,
+  "b7": 18,
+  "b8": 19,
+  "c1": 20,
+  "c2": 21,
+  "c3": 22,
+  "c4": 23,
+  "c5": 24,
+  "c6": 25,
+  "c7": 26,
+  "c8": 27,
+  "d1": 28,
+  "d2": 29,
+  "d3": 30,
+  "d4": 31,
+  "d5": 32,
+  "d6": 33,
+  "d7": 34,
+  "d8": 35,
+  "e1": 36,
+  "e2": 37,
+  "e3": 38,
+  "e4": 39,
+  "e5": 40,
+  "e6": 41,
+  "e7": 42,
+  "e8": 43,
+  "f1": 44,
+  "f2": 45,
+  "f3": 46,
+  "f4": 47,
+  "f5": 48,
+  "f6": 49,
+  "f7": 50,
+  "f8": 51,
+  "g1": 52,
+  "g2": 53,
+  "g3": 54,
+  "g4": 55,
+  "g5": 56,
+  "g6": 57,
+  "g7": 58,
+  "g8": 59,
+  "h1": 60,
+  "h2": 61,
+  "h3": 62,
+  "h4": 63,
+  "h5": 64,
+  "h6": 65,
+  "h7": 66,
+  "h8": 67,
+  "q": 68,
+  "r": 69,
+  "b": 70,
+  "n": 71
+}