Update app.py

app.py

@@ -18,7 +18,7 @@ MODEL_CONFIGS = {
     "Among Us": 13, "Warframe": 13, "Call of Duty": 11, "Forza Horizon": 10,
     "Halo": 14, "Overwatch": 9, "Subnautica": 14, "Fantasy": 16,
     
-    # New Models in version 1.4
+    # New Models
     "Animal Crossing": 14, "Civilization VI": 22, "Control": 22, "Cuphead": 24,
     "Dead Space": 18, "Diablo": 20, "Dota 2": 27, "EVE Online": 24, "GTA": 23,
     "Hades": 20, "Metroid": 28, "Portal": 28, "Resident Evil": 21, "RimWorld": 23,
@@ -106,17 +106,25 @@ def generate_random_name(interpreter, vocab_size, sp, max_length=10, temperature
         if next_token == '' or len(decoded_name) > max_length:
             break
 
+    # --- TEXT NORMALIZATION BLOCK ---
+    
+    # 1. Strip out unwanted tokens (including <s>)
     decoded_name = decoded_name.replace("▁", " ")
     decoded_name = decoded_name.replace("</s>", "")
     decoded_name = decoded_name.replace("<unk>", "")
-    decoded_name = decoded_name.replace("<s>", "")
-    generated_name = decoded_name.strip().capitalize()
+    decoded_name = decoded_name.replace("<s>", "") 
+
+    # 2 & 3. Normalize spacing and apply Title Case to every word
+    # .split() inherently destroys all irregular spacing (multiple spaces, tabs, etc.)
+    words = decoded_name.split()
+    normalized_name = " ".join([word.capitalize() for word in words])
 
-    parts = generated_name.split()
+    # 4. Split the name and check the last part length rule
+    parts = normalized_name.split()
     if parts and len(parts[-1]) < 3:
-        generated_name = " ".join(parts[:-1])
+        normalized_name = " ".join(parts[:-1])
         
-    return generated_name.strip()
+    return normalized_name.strip()
 
 def generateNames(game_type, amount, max_length=30, temperature=0.5, seed_text=""):
     hate_speech = detect_hate_speech(seed_text)
@@ -156,6 +164,11 @@ def generateNames(game_type, amount, max_length=30, temperature=0.5, seed_text="
             temperature=temperature, max_seq_len=max_seq_len
         )
         stripped = generated_name.strip()
+        
+        # Don't pass empty strings to validators if generation completely failed
+        if not stripped:
+            continue
+            
         item_hate_speech = detect_hate_speech(stripped)
         item_profanity = detect_profanity([stripped], language='All')
         name = ''
@@ -170,7 +183,8 @@ def generateNames(game_type, amount, max_length=30, temperature=0.5, seed_text="
             elif item_hate_speech == ['No Hate and Offensive Speech']:
                 name = stripped
         
-        names.append(name)
+        if name: # Only append if we actually have a name
+            names.append(name)
         
     return pd.DataFrame(names, columns=['Names'])