Make more of compilation numeric

gptc/compiler.py

@@ -28,8 +28,8 @@ def compile(
 
     """
 
-    word_counts: Dict[int, Dict[str, int]] = {}
+    word_counts: Dict[int, Dict[int, int]] = {}
-    category_lengths: Dict[str, int] = {}
+    category_lengths: Dict[int, int] = {}
     names: List[str] = []
 
     for portion in raw_model:
@@ -37,10 +37,12 @@ def compile(
             gptc.tokenizer.tokenize(portion["text"], max_ngram_length),
             hash_algorithm,
         )
-        category = portion["category"]
+        category_name = portion["category"]
 
-        if not category in names:
+        if not category_name in names:
-            names.append(category)
+            names.append(category_name)
+
+        category = names.index(category_name)
 
         category_lengths[category] = category_lengths.get(category, 0) + len(
             text
@@ -64,7 +66,7 @@ def compile(
             }
             total = sum(weights.values())
             new_weights: List[int] = []
-            for category in names:
+            for category in range(len(names)):
                 new_weights.append(
                     round((weights.get(category, 0) / total) * 65535)
                 )