[fix] modify the GPT evaluation model (#668)

* [fix] modify the GPT evaluation model

* [fix] modify the GPT evaluation model

Author: jujeongho0

lmms_eval/models/qwen2_5_vl.py

@@ -195,18 +195,18 @@ def _collate(x):
             split = split[0]
             visual_list = [doc_to_visual[0](self.task_dict[task][split][ids]) for ids in doc_id]
             gen_kwargs = all_gen_kwargs[0]
-            
+
             # Set default until or update values from gen_kwargs if present
             until = gen_kwargs.get("until", [self.tokenizer.decode(self.eot_token_id)])
-            
+
             if isinstance(until, str):
                 until = [until]
             elif not isinstance(until, list):
                 raise ValueError(f"Expected `gen_kwargs['until']` to be of type Union[str, list], but got {type(until)}")
-            
+
             # Avoid using '\n\n' as a stopper for Qwen2.5VL to prevent truncation, which can lead to incorrect results
             until = [item for item in until if item != "\n\n"]
-            
+
             if isinstance(contexts, tuple):
                 contexts = list(contexts)
 

lmms_eval/models/vllm.py

@@ -53,7 +53,7 @@ def __init__(
 
         # Convert any string arguments that start with { and end with } to dictionaries
         for key, value in kwargs.items():
-            if isinstance(value, str) and value.strip().startswith('{') and value.strip().endswith('}'):
+            if isinstance(value, str) and value.strip().startswith("{") and value.strip().endswith("}"):
                 try:
                     kwargs[key] = json.loads(value)
                 except json.JSONDecodeError:
@@ -184,7 +184,7 @@ def generate_until(self, requests) -> List[str]:
                 batched_messages.append(messages)
 
             sampling_params = SamplingParams(**params)
-            
+
             if self.chat_template is not None:
                 with open(self.chat_template, "r") as f:
                     chat_template = f.read()