Update handler.py

handler.py

@@ -1,7 +1,13 @@
+from typing import Dict, Any
+import logging
+
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from peft import PeftConfig, PeftModel
-
 import torch.cuda
+
+
+LOGGER = logging.getLogger(__name__)
+logging.basicConfig(level=logging.INFO)
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 
@@ -13,25 +19,26 @@ class EndpointHandler():
         # 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.
-            """
-            # Get inputs
-            prompt = data.pop("inputs", None)
-            parameters = data.pop("parameters", None)
-            if prompt is None:
-                raise ValueError("Missing prompt.")
-            # Preprocess
-            input_ids = self.tokenizer(prompt, return_tensors="pt").input_ids.to(device)
-            # Forward
-            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])
-            return {"generated_text": prediction}
+        """
+        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)
+        if prompt is None:
+            raise ValueError("Missing prompt.")
+        # 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}