Update services/strategy.py

services/strategy.py

@@ -70,7 +70,7 @@ class BestOfN(GenerationStrategy):
             response_inputs = generator.tokenizer(response, return_tensors="pt").to(generator.device)
 
             # Pass the response inputs correctly to the PRM model
-            prm_output = generator.prm_model(response_inputs, **model_kwargs)  # Pass the inputs correctly to the model
+            prm_output = generator.prm_model(response_inputs)  # Pass the inputs correctly to the model
             
             # Check the expected output structure for prm_model and use it accordingly
             score = prm_output.logits.mean().item() if hasattr(prm_output, 'logits') else 0.0