fix formatting

Author: Your Name

vllm/beam/filtering.py

@@ -21,9 +21,7 @@ def format_filter(filter_params_list):
     return [dict(zip(keys, vals)) for vals in filter_params_list]
 
 
-DEFAULT_CHAR_SERVER_FILTER = format_filter(
-    [{"name": "annotations_porn", "threshold": 0.5156}, {"name": "annotations_racist", "threshold": 0.9763}, {"name": "annotations_disturbing", "threshold": 0.5472}, {"name": "annotations_harmful_promotes_selfharm", "threshold": 0.0657}]
-)
+DEFAULT_CHAR_SERVER_FILTER = [{"name": "annotations_porn", "threshold": 0.5156}, {"name": "annotations_racist", "threshold": 0.9763}, {"name": "annotations_disturbing", "threshold": 0.5472}, {"name": "annotations_harmful_promotes_selfharm", "threshold": 0.0657}]
 
 MAX_GENERATIONS = 10
 _CHUNK_SIZE = 16
@@ -113,6 +111,8 @@ def get_filtered_classifiers(self, prob_C, filter_params) -> list[str]:
             if p["name"] in self.classifier_names
         ]
 
+        print(f"prob_C: {prob_C}")
+        print(f"relevant_filters: {relevant_filters}")
         if not relevant_filters:
             return []
 

vllm/beam/penalty.py

@@ -646,7 +646,7 @@ def format_penalty(penalty_params_list):
     return [dict(zip(keys, vals)) for vals in penalty_params_list]
 
 
-MODEL_SERVER_PENALTY = format_penalty([{"name": "annotations_sexually_suggestive", "penalty": 15000, "final_threshold": 0.4484, "per_unit_penalty": 0}, {"name": "annotations_porn", "penalty": 15000, "final_threshold": 0.4378, "per_unit_penalty": 0}, {"name": "annotations_porn", "penalty": 20000, "final_threshold": 0.7773, "per_unit_penalty": 1}, {"name": "annotations_racist", "penalty": 20000, "final_threshold": 0.9763, "per_unit_penalty": 1}, {"name": "annotations_disturbing", "penalty": 15000, "final_threshold": 0.4156, "per_unit_penalty": 1}, {"name": "annotations_harmful_promotes_selfharm", "penalty": 10000, "final_threshold": 0.0657, "per_unit_penalty": 1}, {"name": "annotations_harmful_promotes_selfharm", "penalty": 10001, "final_threshold": 0.0017, "per_unit_penalty": 1}, {"name": "repetitive", "penalty": 1000, "final_threshold": 0.8164, "per_unit_penalty": 1}, {"name": "annotations_nonsense_gd", "penalty": 1000, "final_threshold": 0.2539, "per_unit_penalty": 1}, {"name": "annotations_looping", "penalty": 1000, "final_threshold": 0.1944, "per_unit_penalty": 1}, {"name": "annotations_out_of_character_gd", "penalty": 1000, "final_threshold": 0.3631, "per_unit_penalty": 1}])
+MODEL_SERVER_PENALTY = [{"name": "annotations_sexually_suggestive", "penalty": 15000, "final_threshold": 0.4484, "per_unit_penalty": 0}, {"name": "annotations_porn", "penalty": 15000, "final_threshold": 0.4378, "per_unit_penalty": 0}, {"name": "annotations_porn", "penalty": 20000, "final_threshold": 0.7773, "per_unit_penalty": 1}, {"name": "annotations_racist", "penalty": 20000, "final_threshold": 0.9763, "per_unit_penalty": 1}, {"name": "annotations_disturbing", "penalty": 15000, "final_threshold": 0.4156, "per_unit_penalty": 1}, {"name": "annotations_harmful_promotes_selfharm", "penalty": 10000, "final_threshold": 0.0657, "per_unit_penalty": 1}, {"name": "annotations_harmful_promotes_selfharm", "penalty": 10001, "final_threshold": 0.0017, "per_unit_penalty": 1}, {"name": "repetitive", "penalty": 1000, "final_threshold": 0.8164, "per_unit_penalty": 1}, {"name": "annotations_nonsense_gd", "penalty": 1000, "final_threshold": 0.2539, "per_unit_penalty": 1}, {"name": "annotations_looping", "penalty": 1000, "final_threshold": 0.1944, "per_unit_penalty": 1}, {"name": "annotations_out_of_character_gd", "penalty": 1000, "final_threshold": 0.3631, "per_unit_penalty": 1}]
 
 
 class PenaltyComputer: