Update app.py

app.py

@@ -1,6 +1,6 @@
 import gradio as gr
 import torch
-from transformers import pipeline, AutoModelForSpeechSeq2Seq, AutoProcessor
+from transformers import WhisperProcessor, WhisperForConditionalGeneration, pipeline
 from jiwer import wer
 
 # Load models
@@ -16,24 +16,22 @@ whisper_pipeline_2 = pipeline(
 )
 
 # Set up openai/whisper-medium for Urdu transcription
-model_3 = AutoModelForSpeechSeq2Seq.from_pretrained("openai/whisper-medium")
-processor_3 = AutoProcessor.from_pretrained("openai/whisper-medium")
-
-# Set forced decoder IDs for Urdu
-language = "<|ur|>"  # Urdu language token
-task = "<|transcribe|>"  # Transcription task token
-forced_decoder_ids = processor_3.get_decoder_prompt_ids(language=language, task=task)
+processor = WhisperProcessor.from_pretrained("openai/whisper-medium")
+model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-medium")
+forced_decoder_ids = processor.get_decoder_prompt_ids(language="urdu", task="transcribe")
 
 def transcribe_with_whisper_medium(audio_path):
     """
     Transcribe audio using the openai/whisper-medium model with forced language settings for Urdu.
     """
-    inputs = processor_3(audio_path, return_tensors="pt", sampling_rate=16000)
+    inputs = processor(audio_path, return_tensors="pt", sampling_rate=16000)
     with torch.no_grad():
-        outputs = model_3.generate(
+        # Generate the transcription using the forced decoder IDs for Urdu
+        outputs = model.generate(
             inputs["input_features"], forced_decoder_ids=forced_decoder_ids
         )
-    return processor_3.batch_decode(outputs, skip_special_tokens=True)[0]
+    # Decode the outputs to text
+    return processor.batch_decode(outputs, skip_special_tokens=True)[0]
 
 def transcribe_and_compare(audio_path, original_transcription=None):
     """