Update app.py

app.py

@@ -8,9 +8,9 @@ nltk.download("punkt")
 
 
 model_name = "Shubham09/whisper31filescheck"
-processor = WhisperProcessor.from_pretrained(model_name)
-tokenizer = WhisperTokenizer.from_pretrained(model_name)
-model = WhisperForConditionalGeneration.from_pretrained(model_name,decoder_input_ids)
+processor = WhisperProcessor.from_pretrained(model_name,task="transcribe")
+#tokenizer = WhisperTokenizer.from_pretrained(model_name)
+model = WhisperForConditionalGeneration.from_pretrained(model_name)
 
 def load_data(input_file):
 
@@ -33,14 +33,14 @@ def asr_transcript(input_file):
 
   speech = load_data(input_file)
   #Tokenize
-  input_features = processor(speech, return_tensors="pt").input_features  #, padding="longest" , return_tensors="pt"
+  input_features = processor(speech).input_features  #, padding="longest" , return_tensors="pt"
   #input_values = tokenizer(speech, return_tensors="pt").input_values
   #Take logits
   logits = model(input_features).logits
   #Take argmax
   predicted_ids = torch.argmax(logits, dim=-1)
   #Get the words from predicted word ids
-  transcription = processor.batch_decode(predicted_ids[0])
+  transcription = processor.batch_decode(predicted_ids)
   #Correcting the letter casing
   #transcription = correct_casing(transcription.lower())
   return transcription