remove new_tokens

dan/manager/ocr.py

@@ -42,11 +42,6 @@ class OCRDatasetManager(DatasetManager):
         ):
             self.synthetic_data = self.params["config"]["synthetic_data"]
 
-        if "new_tokens" in params:
-            self.charset = sorted(
-                list(set(self.charset).union(set(params["new_tokens"])))
-            )
-
         self.tokens = {
             "pad": params["config"]["padding_token"],
         }