Update handler.py

handler.py

@@ -7,7 +7,7 @@ class EndpointHandler:
         # Set model size, assuming installation has been done with appropriate model files and setup
         model_size = "medium" if model_dir is None else model_dir
         # Change to 'cuda' to use the GPU, and set compute_type for faster computation
-        self.model = WhisperModel(model_size, device="cuda", compute_type="fp16")
+        self.model = WhisperModel(model_size, device="cuda", compute_type="float16")
 
     def __call__(self, data: Dict) -> Dict[str, str]:
         # Process the input data expected to be in 'inputs' key containing audio file bytes