Update services/strategy.py

services/strategy.py

@@ -60,7 +60,7 @@ class BestOfN(GenerationStrategy):
                 input_ids = generator.tokenizer(prompt, return_tensors="pt").input_ids.to(generator.device)
                 output = generator.model.generate(input_ids, **model_kwargs)
                 response =generator.tokenizer.decode(output[0], skip_special_tokens=True)
-                score = generator.prm_model(**generator.tokenizer(response, return_tensors="pt").to(generator.device)).logits.mean().item()
+                score = generator.prm_model.generate(**generator.tokenizer(response, return_tensors="pt").to(generator.device)).logits.mean().item()
                 scored_outputs.append((response, score))
             return max(scored_outputs, key=lambda x: x[1])[0]