Update handler.py

handler.py

@@ -10,7 +10,14 @@ class EndpointHandler:
         self.model = WhisperModel(model_size, device="cuda", compute_type="float16")
 
     def __call__(self, data: Dict) -> Dict[str, str]:
-        segments, info = self.model.transcribe(data["inputs"])
+        # Process the input data expected to be in 'inputs' key containing audio file bytes
+        audio_bytes = data["inputs"]
+        
+        # Convert bytes to a file-like object
+        audio_file = io.BytesIO(audio_bytes)
+        
+        # Perform transcription using the model
+        segments, info = self.model.transcribe(audio_file)
         
         # Compile the results into a text string and extract language information
         # Strip whitespace from each segment before joining them