Update services/strategy.py

services/strategy.py

@@ -18,7 +18,7 @@ try:
 except Exception as e:
     print("Langfuse Offline")
 
-@observe()
+
 class GenerationStrategy(ABC):
     """Base class for generation strategies."""
     
@@ -26,15 +26,15 @@ class GenerationStrategy(ABC):
     def generate(self, generator: 'BaseGenerator', prompt: str, model_kwargs: Dict[str, Any], **kwargs) -> str:
         pass
 
- 
+
 class DefaultStrategy(GenerationStrategy):
-   
+    @observe()
     def generate(self, generator: 'BaseGenerator', prompt: str, model_kwargs: Dict[str, Any], **kwargs) -> str:
         input_ids = generator.tokenizer(prompt, return_tensors="pt").input_ids.to(generator.device)
         output = generator.model.generate(input_ids, **model_kwargs)
         return generator.tokenizer.decode(output[0], skip_special_tokens=True)
 
-@observe()
+
 class MajorityVotingStrategy(GenerationStrategy):
     def generate(self, generator: 'BaseGenerator', prompt: str, model_kwargs: Dict[str, Any], num_samples: int = 5, **kwargs) -> str:
         outputs = []
@@ -44,7 +44,7 @@ class MajorityVotingStrategy(GenerationStrategy):
             outputs.append(generator.tokenizer.decode(output[0], skip_special_tokens=True))
         return max(set(outputs), key=outputs.count)
 
-@observe()
+
 class BestOfN(GenerationStrategy):
     def generate(self, generator: 'BaseGenerator', prompt: str, model_kwargs: Dict[str, Any], num_samples: int = 5, **kwargs) -> str:
             scored_outputs = []
@@ -56,7 +56,7 @@ class BestOfN(GenerationStrategy):
                 scored_outputs.append((response, score))
             return max(scored_outputs, key=lambda x: x[1])[0]
 
-@observe() 
+
 class BeamSearch(GenerationStrategy):
     def generate(self, generator: 'BaseGenerator', prompt: str, model_kwargs: Dict[str, Any], num_samples: int = 5, **kwargs) -> str:
             input_ids = self.llama_tokenizer(prompt, return_tensors="pt").input_ids.to(self.device)
@@ -68,7 +68,7 @@ class BeamSearch(GenerationStrategy):
             )
             return [self.llama_tokenizer.decode(output, skip_special_tokens=True) for output in outputs]
 
-@observe()
+
 class DVT(GenerationStrategy):
     def generate(self, generator: 'BaseGenerator', prompt: str, model_kwargs: Dict[str, Any], num_samples: int = 5, **kwargs) -> str:
             results = []
@@ -89,14 +89,14 @@ class DVT(GenerationStrategy):
                     results.append((extended_response, score))
             return max(results, key=lambda x: x[1])[0]
         
-@observe()
+
 class COT(GenerationStrategy):
     def generate(self, generator: 'BaseGenerator', prompt: str, model_kwargs: Dict[str, Any], num_samples: int = 5, **kwargs) -> str:
         #TODO implement the chain of thought strategy
         
         return "Not implemented yet"       
 
-@observe()
+
 class ReAct(GenerationStrategy):
     def generate(self, generator: 'BaseGenerator', prompt: str, model_kwargs: Dict[str, Any], num_samples: int = 5, **kwargs) -> str:
         #TODO implement the ReAct framework