nit

lm_eval/api/group.py

 from dataclasses import asdict, dataclass
 from inspect import getsource
-from typing import Any, Callable, List, Optional, Union
+from typing import Any, Callable, Optional, Union
 
 
 @dataclass
 class AggMetricConfig(dict):
     metric: Optional[str] = None
     aggregation: Optional[str] = "mean"
-    weight_by_size: Optional[str] = False
+    weight_by_size: bool = False
     # list of filter names which should be incorporated into the aggregated metric.
     filter_list: Optional[Union[str, list]] = "none"
 
@@ -27,7 +27,7 @@ class GroupConfig(dict):
     group_alias: Optional[str] = None
     task: Optional[Union[str, list]] = None
     aggregate_metric_list: Optional[
-        Union[List[AggMetricConfig], AggMetricConfig, dict]
+        Union[list[AggMetricConfig], AggMetricConfig, dict]
     ] = None
     version: Optional[str] = None
     metadata: Optional[dict] = (

pyproject.toml

@@ -105,8 +105,8 @@ plugins.md034.enabled = false # no-bare-urls
 
 [tool.ruff]
 target-version = "py39"
-extend-select = ["I", "UP", "E", "C419", "F", "B", "SIM"]
-ignore = ["E402", "E731", "E501", "E111", "E114", "E117"]
+lint.extend-select = ["I", "UP", "E", "C419", "F", "B", "SIM"]
+lint.ignore = ["E402", "E731", "E501", "E111", "E114", "E117"]
 
 [tool.ruff.lint.isort]
 combine-as-imports = true