Chess Challenge submission by graug

README.md

@@ -14,13 +14,13 @@ Chess model submitted to the LLM Course Chess Challenge.
 ## Submission Info
 
 - **Submitted by**: [graug](https://huggingface.co/graug)
-- **Parameters**: 704,000
+- **Parameters**: 703,232
 - **Organization**: LLM-course
 
 ## Model Details
 
 - **Architecture**: Chess Transformer (GPT-style)
-- **Vocab size**: 74
+- **Vocab size**: 68
 - **Embedding dim**: 128
 - **Layers**: 4
 - **Heads**: 4

config.json

@@ -20,5 +20,5 @@
   "pad_token_id": 0,
   "tie_weights": true,
   "transformers_version": "4.57.5",
-  "vocab_size": 74
+  "vocab_size": 68
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:59b83e6a044a255f31154ccfed055dc4cc9ff42d3ceba06068732294fa6ce406
-size 2820400
+oid sha256:9f4ad5a4f30494574e79f1987ece70a031bd032836128bad310028653d2a0def
+size 2817328

tokenizer.py

@@ -183,7 +183,7 @@ class ChessTokenizer(PreTrainedTokenizer):
     @classmethod
     def build_vocab_uci(cls):
         
-        pièces = ['P', 'N', 'B', 'R', 'Q', 'K']
+        
         files = ['a', 'b', 'c', 'd', 'e', 'f', 'g', 'h']
         ranks = ['1', '2', '3', '4', '5', '6', '7', '8']
         
@@ -192,7 +192,7 @@ class ChessTokenizer(PreTrainedTokenizer):
             for rank in ranks:
                 cases.append(f"{file}{rank}")
                 
-        vocab_list = pièces + cases 
+        vocab_list = cases 
         return cls.build_vocab_from_iterator(vocab_list)
     
   
@@ -229,27 +229,30 @@ class ChessTokenizer(PreTrainedTokenizer):
                 tokens.append(word)
                 continue
             
-            # Remove W/B color prefix
+            # Remove W/B color prefix 
             if len(word) > 0 and word[0] in 'WB':
                 move = word[1:]
             else:
                 move = word
+            #remove piece
+            if word[1] in 'PNBRQK':
+                move = move[1:]
+                
             
             # Remove suffixes like (x), (+), (+*), (o), (O)
             for symbol in ['(x+*)', '(x+)', '(x)', '(+*)', '(+)', '(o)', '(O)']:
                 move = move.replace(symbol, '')
             
-            # Handle promotion: Pe7e8=Q -> extract piece, from, to (ignore promotion for now)
+            # suppress promotion for now
             if '=' in move:
                 move = move.split('=')[0]
             
-            # Now move should be like "Pe2e4" (5 chars) or longer
-            if len(move) >= 5:
-                piece = move[0]       # 'P', 'N', 'B', 'R', 'Q', 'K'
-                from_sq = move[1:3]   # 'e2'
-                to_sq = move[3:5]     # 'e4'
-                tokens.extend([piece, from_sq, to_sq])
-        
+            
+            # Now move should be in format from_square + to_square (e.g., e2e4)
+            if len(move) == 4:
+                from_sq = move[:2]
+                to_sq = move[2:]
+                tokens.extend([from_sq, to_sq])
         return tokens
     
         
@@ -269,19 +272,13 @@ class ChessTokenizer(PreTrainedTokenizer):
         
         tokens = [t for t in tokens if t not in special]
         
-        #Grouper les tokens par 3
-        grouped_tokens = ["".join(tokens[i:i+3]) for i in range(0, len(tokens), 3)]
+        #Grouper les tokens par 2 (from_square, to_square)
+        grouped_tokens = ["".join(tokens[i:i+2]) for i in range(0, len(tokens), 2)]
+        
         
-        # mettre une fois sur deux B avant les coups noirs sinon W
-        final_tokens = []
         
-        for i, token in enumerate(grouped_tokens):
-            if i % 2 == 0:
-                final_tokens.append('W' + token)
-            else:
-                final_tokens.append('B' + token)
                 
-        return " ".join(final_tokens)
+        return " ".join(grouped_tokens)
     
     def save_vocabulary(
         self,

vocab.json

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