- change handler

handler.py

@@ -1,7 +1,10 @@
 from typing import Dict, List, Any
 
 import torch as torch
-from transformers import pipeline
+from transformers import pipeline, WhisperProcessor
+
+from scipy.io.wavfile import read
+
 
 
 class EndpointHandler():
@@ -16,7 +19,8 @@ class EndpointHandler():
             chunk_length_s=30,
             device=device,
         )
-        self.pipe.model.config.forced_decoder_ids = self.pipe.tokenizer.get_decoder_prompt_ids(language="nl", task="transcribe")
+        processor = WhisperProcessor.from_pretrained("openai/whisper-large")
+        self.pipe.model.config.forced_decoder_ids = processor.get_decoder_prompt_ids(language="nl", task="transcribe")
 
     def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
         """
@@ -29,7 +33,8 @@ class EndpointHandler():
         #print request
         print("request")
         print(data)
-        # get inputs
+        # audio_data = read(io.BytesIO(data))
+        # get inputs, inputs in request body is possible equal to wav or mp3 file
         inputs = data.pop("inputs", data)
 
         text = self.pipe(inputs)["text"]