Update handler.py

handler.py

@@ -28,7 +28,8 @@ class EndpointHandler():
         processor = WhisperProcessor.from_pretrained(peft_config.base_model_name_or_path, language=language, task=task)
         feature_extractor = processor.feature_extractor
         self.forced_decoder_ids = processor.get_decoder_prompt_ids(language=language, task=task)
-        self.pipeline = AutomaticSpeechRecognitionPipeline(model=model, tokenizer=tokenizer, feature_extractor=feature_extractor)
+        # self.pipeline = AutomaticSpeechRecognitionPipeline(model=model, tokenizer=tokenizer, feature_extractor=feature_extractor)
+        self.pipeline = pipeline(task= "automatic-speech-recognition", model=path, tokenizer=tokenizer, feature_extractor = feature_extractor)
     
     def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
         """
@@ -42,11 +43,12 @@ class EndpointHandler():
         
         # run normal prediction
         inputs = data.pop("inputs", data)
-        print(inputs)
-        print(inputs, file=sys.stderr)
-        print(inputs, file=sys.stdout)
-        prediction = self.pipeline(inputs, generate_kwargs={"forced_decoder_ids": self.forced_decoder_ids}, max_new_tokens=255)["text"]
-        print(prediction)
-        print(predcition, file=sys.stderr)
-        print(predcition, file=sys.stdout)
+        print("a1", inputs)
+        print("a2", inputs, file=sys.stderr)
+        print("a3", inputs, file=sys.stdout)
+        # prediction = self.pipeline(inputs, generate_kwargs={"forced_decoder_ids": self.forced_decoder_ids}, max_new_tokens=255)["text"]
+        prediction = self.pipeline(inputs, return_timestamps=False)
+        print("b1", prediction)
+        print("b2", predcition, file=sys.stderr)
+        print("b3", predcition, file=sys.stdout)
         return prediction