Update handler.py

handler.py

@@ -3,31 +3,26 @@ from faster_whisper import WhisperModel
 
 class EndpointHandler:
     def __init__(self):
-        # Initialize WhisperModel
-        self.model = WhisperModel("large-v2")
+        # Initialize WhisperModel; assume that "large-v2" model files are in /repository
+        self.model = WhisperModel("large-v2", device="cpu")
 
     def __call__(self, data: Dict) -> Dict:
-        # Get the audio file bytes from the request data
+        # Process the input data expected to be in 'inputs' key containing audio file bytes
         audio_bytes = data["inputs"]
         
-        # Perform transcription
-        results = []
+        # Perform transcription using the model
         segments, info = self.model.transcribe(audio_bytes)
-        for segment in segments:
-            result = {
-                "start": segment.start,
-                "end": segment.end,
-                "text": segment.text
-            }
-            results.append(result)
         
-        # Return the transcribed text along with language data
-        language_code, language_prob = info.language, info.language_probability
-        response = {
-            "transcription": results,
-            "language": {
-                "code": language_code,
-                "probability": language_prob
-            }
+        # Compile the results into a text string and extract language information
+        text = " ".join(segment.text for segment in segments)
+        language_code = info.language
+        language_prob = info.language_probability
+
+        # Compile the response dictionary
+        result = {
+            "text": text,
+            "language": language_code,
+            "language_probability": language_prob
         }
-        return response
+        
+        return result