Charset should only include training characters

Closes #190 (closed)

dan/datasets/extract/__init__.py

@@ -113,6 +113,12 @@ def add_extract_parser(subcommands) -> None:
         """,
         required=False,
     )
+    parser.add_argument(
+        "--unknown-token",
+        type=str,
+        help="Token to use to replace character in the validation/test sets that is not included in the training set.",
+        default="?",
+    )
     parser.add_argument(
         "--tokens",
         type=pathlib.Path,