Chess Challenge submission by swdo

README.md

@@ -0,0 +1,26 @@
+---
+library_name: transformers
+tags:
+- chess
+- llm-course
+- chess-challenge
+license: mit
+---
+
+# chess_swdo_up
+
+Chess model submitted to the LLM Course Chess Challenge.
+
+## Submission Info
+
+- **Submitted by**: [swdo](https://huggingface.co/swdo)
+- **Parameters**: 871,168
+- **Organization**: LLM-course
+
+## Model Details
+
+- **Architecture**: Chess Transformer (GPT-style)
+- **Vocab size**: 88
+- **Embedding dim**: 128
+- **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": 128,
+  "n_head": 4,
+  "n_inner": 384,
+  "n_layer": 5,
+  "pad_token_id": 0,
+  "tie_weights": true,
+  "transformers_version": "4.57.3",
+  "vocab_size": 88
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c5a58fe6555f95d4157289e7462a9152909962f3c251f806f621e8bed568a42
+size 3490096

special_tokens_map.json

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

tokenizer.py

@@ -0,0 +1,199 @@
+
+
+"""
+Custom Chess Tokenizer - Decomposed Move Representation.
+
+Decomposes chess moves into components: Color, Piece, From square, To square, Modifiers.
+Uses a fixed vocabulary of ~88 tokens.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import re
+from typing import Dict, List, Optional
+
+from transformers import PreTrainedTokenizer
+
+
+class ChessTokenizer(PreTrainedTokenizer):
+    model_input_names = ["input_ids", "attention_mask"]
+    
+    PAD_TOKEN = "[PAD]"
+    BOS_TOKEN = "[BOS]"
+    EOS_TOKEN = "[EOS]"
+    UNK_TOKEN = "[UNK]"
+    
+    COLORS = ["[W]", "[B]"]
+    PIECES = ["P", "N", "B", "R", "Q", "K"]
+    SQUARES = [f + r for f in "abcdefgh" for r in "12345678"]
+    MODIFIERS = ["x", "+", "#", "+*", "=Q", "=R", "=B", "=N", "O-O", "O-O-O", "o", "O"]
+    
+    MOVE_PATTERN = re.compile(
+        r'^([WB])([PNBRQK])([a-h][1-8])([a-h][1-8])(=[QRBN])?(\([xoO+*]+\))?$'
+    )
+    
+    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_default_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,
+        )
+    
+    def _create_default_vocab(self) -> Dict[str, int]:
+        tokens = [self.PAD_TOKEN, self.BOS_TOKEN, self.EOS_TOKEN, self.UNK_TOKEN]
+        tokens.extend(self.COLORS)
+        tokens.extend(self.PIECES)
+        tokens.extend(self.SQUARES)
+        tokens.extend(self.MODIFIERS)
+        return {token: idx for idx, token in enumerate(tokens)}
+    
+    def _parse_move(self, move: str) -> List[str]:
+        """Parse a single move into component tokens."""
+        match = self.MOVE_PATTERN.match(move)
+        if not match:
+            return [self.UNK_TOKEN]
+        
+        tokens = [f"[{match.group(1)}]", match.group(2), match.group(3), match.group(4)]
+        
+        if match.group(5):
+            tokens.append(match.group(5))
+        
+        if match.group(6):
+            suffix_content = match.group(6)[1:-1]
+            if "x" in suffix_content:
+                tokens.append("x")
+            if "+*" in suffix_content:
+                tokens.append("+*")
+            elif "+" in suffix_content:
+                tokens.append("+")
+            if suffix_content == "o":
+                tokens.append("o")
+            elif suffix_content == "O":
+                tokens.append("O")
+        
+        return tokens
+    
+    def _tokenize(self, text: str) -> List[str]:
+        """Tokenize a string of moves into component tokens."""
+        tokens = []
+        for move in text.strip().split():
+            tokens.extend(self._parse_move(move))
+        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:
+        """Reconstruct moves from component tokens."""
+        special = {self.PAD_TOKEN, self.BOS_TOKEN, self.EOS_TOKEN, self.UNK_TOKEN}
+        result = []
+        current_move = []
+        
+        for token in tokens:
+            if token in special:
+                if current_move:
+                    result.append(self._reconstruct_move(current_move))
+                    current_move = []
+                continue
+            
+            current_move.append(token)
+            if self._is_complete_move(current_move):
+                result.append(self._reconstruct_move(current_move))
+                current_move = []
+        
+        if current_move:
+            result.append(self._reconstruct_move(current_move))
+        
+        return " ".join(result)
+    
+    def _is_complete_move(self, tokens: List[str]) -> bool:
+        """Check if tokens form a complete move."""
+        if len(tokens) < 4:
+            return False
+        
+        if (tokens[0] in self.COLORS and tokens[1] in self.PIECES and
+            tokens[2] in self.SQUARES and tokens[3] in self.SQUARES):
+            if len(tokens) == 4:
+                return True
+            for t in tokens[4:]:
+                if t in self.COLORS or (t not in self.MODIFIERS and not t.startswith("=")):
+                    return True
+            return True
+        return False
+    
+    def _reconstruct_move(self, tokens: List[str]) -> str:
+        """Reconstruct a move string from component tokens."""
+        if not tokens or len(tokens) < 4:
+            return "".join(tokens)
+        
+        color = tokens[0][1] if tokens[0] in self.COLORS else tokens[0]
+        move = color + "".join(tokens[1:4])
+        
+        suffixes = []
+        for t in tokens[4:]:
+            if t.startswith("="):
+                move += t
+            elif t in ["x", "+", "+*", "o", "O"]:
+                suffixes.append(t)
+        
+        if suffixes:
+            move += "(" + "".join(suffixes) + ")"
+        
+        return move
+    
+    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,)
+    
+    @classmethod
+    def build_vocab_from_iterator(cls, iterator, min_frequency=1):
+        return cls()
+    
+    @classmethod
+    def build_vocab_from_dataset(cls, **kwargs):
+        return cls()
+
+    @property
+    def vocab_size(self) -> int:
+        return len(self._vocab)
+    
+    def get_vocab(self) -> Dict[str, int]:
+        return dict(self._vocab)

tokenizer_config.json

@@ -0,0 +1,44 @@
+{
+  "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
+    }
+  },
+  "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 +1,90 @@
+{
+  "[PAD]": 0,
+  "[BOS]": 1,
+  "[EOS]": 2,
+  "[UNK]": 3,
+  "[W]": 4,
+  "[B]": 5,
+  "P": 6,
+  "N": 7,
+  "B": 8,
+  "R": 9,
+  "Q": 10,
+  "K": 11,
+  "a1": 12,
+  "a2": 13,
+  "a3": 14,
+  "a4": 15,
+  "a5": 16,
+  "a6": 17,
+  "a7": 18,
+  "a8": 19,
+  "b1": 20,
+  "b2": 21,
+  "b3": 22,
+  "b4": 23,
+  "b5": 24,
+  "b6": 25,
+  "b7": 26,
+  "b8": 27,
+  "c1": 28,
+  "c2": 29,
+  "c3": 30,
+  "c4": 31,
+  "c5": 32,
+  "c6": 33,
+  "c7": 34,
+  "c8": 35,
+  "d1": 36,
+  "d2": 37,
+  "d3": 38,
+  "d4": 39,
+  "d5": 40,
+  "d6": 41,
+  "d7": 42,
+  "d8": 43,
+  "e1": 44,
+  "e2": 45,
+  "e3": 46,
+  "e4": 47,
+  "e5": 48,
+  "e6": 49,
+  "e7": 50,
+  "e8": 51,
+  "f1": 52,
+  "f2": 53,
+  "f3": 54,
+  "f4": 55,
+  "f5": 56,
+  "f6": 57,
+  "f7": 58,
+  "f8": 59,
+  "g1": 60,
+  "g2": 61,
+  "g3": 62,
+  "g4": 63,
+  "g5": 64,
+  "g6": 65,
+  "g7": 66,
+  "g8": 67,
+  "h1": 68,
+  "h2": 69,
+  "h3": 70,
+  "h4": 71,
+  "h5": 72,
+  "h6": 73,
+  "h7": 74,
+  "h8": 75,
+  "x": 76,
+  "+": 77,
+  "#": 78,
+  "+*": 79,
+  "=Q": 80,
+  "=R": 81,
+  "=B": 82,
+  "=N": 83,
+  "O-O": 84,
+  "O-O-O": 85,
+  "o": 86,
+  "O": 87
+}