Fix missing name code logging in q-learning.

slippi_ai/train_q_lib.py

@@ -240,12 +240,10 @@ def train(config: Config):
     json.dump(name_map, f)
   wandb.save(name_map_path, policy='now')
 
-  missing_name_code = len(name_map)
-  num_codes = missing_name_code + 1
-
-  def encode_name(name: str) -> np.uint8:
-    return np.uint8(name_map.get(name, missing_name_code))
-  batch_encode_name = np.vectorize(encode_name)
+  num_codes = nametags.max_name_code(name_map) + 1
+  encode_name = nametags.name_encoder(name_map)
+  encode_name_uint8 = lambda name: np.uint8(encode_name(name))
+  batch_encode_name = np.vectorize(encode_name_uint8)
 
   # Create data sources for train and test.
   data_config = dict(