Update handler.py

handler.py

@@ -10,17 +10,16 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 
 class EndpointHandler():
     def __init__(self, path=""):
-        model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, load_in_8bit=True, device_map='auto')
+        config = PeftConfig.from_pretrained("JeremyArancio/llm-tolkien")
+        self.model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, load_in_8bit=True, device_map='auto')
         self.tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
-        # Load the Lora model
-        self.model = PeftModel.from_pretrained(model, path)
+        
 
     def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
         """
         Args:
             data (Dict): The payload with the text prompt and generation parameters.
         """
-        LOGGER.info(f"Received data: {data}")
         # Get inputs
         prompt = data.pop("inputs", None)
         parameters = data.pop("parameters", None)
@@ -29,12 +28,10 @@ class EndpointHandler():
         # Preprocess
         input_ids = self.tokenizer(prompt, return_tensors="pt").input_ids.to(device)
         # Forward
-        LOGGER.info(f"Start generation.")
         if parameters is not None:
             output = self.model.generate(input_ids=input_ids, **parameters)
         else:
             output = self.model.generate(input_ids=input_ids)
         # Postprocess
         prediction = self.tokenizer.decode(output[0])
-        LOGGER.info(f"Generated text: {prediction}")
         return {"generated_text": prediction}