Update app.py

app.py

@@ -25,12 +25,13 @@ print(dataset, "and at 0[audio][array] ", dataset[0]["audio"]["array"], type(dat
 
 def transcribe(audio):
     text = pipe(audio)["text"]
-    return text
+    test = evalWhisper(model, dataset)
+    return text, test
 
 iface = gr.Interface(
     fn=transcribe, 
     inputs=gr.Audio(source="microphone", type="filepath"), 
-    outputs="text",
+    outputs="text", "text"
     title="Whisper Small Miso Test",
 )
 
@@ -39,50 +40,7 @@ iface.launch()
 def evalWhisper(model, dataset):
     model.eval()
     print("model.eval ", model.eval())
-    
-    # Define a list to store the print statements
-    log_texts = []
-    
-    with torch.no_grad():
-        outputs = model(**input_data)  # Define input_data appropriately
-        print("outputs ", outputs)
-        log_texts.append(f"outputs: {outputs}")
-    
-    # Convert predicted token IDs back to text
-    predicted_text = tokenizer.batch_decode(outputs.logits.argmax(dim=-1), skip_special_tokens=True)
-    
-    # Get ground truth labels from the dataset
-    labels = dataset["audio"]  # Replace "labels" with the appropriate key in your dataset
-    print("labels are ", labels)
-    log_texts.append(f"labels: {labels}")
-    
-    # Compute WER
-    wer_score = wer(labels, predicted_text)  # Define wer function
-    
-    # Print or return WER score
-    wer_message = f"Word Error Rate (WER): {wer_score}"
-    print(wer_message)
-    log_texts.append(wer_message)
-
-    print(log_texts)
-    
-    return log_texts
-
-# Call evalWhisper and get the log texts
-log_texts = evalWhisper(model, dataset)
 
-# Display the log texts using gr.Interface
-log_text = "\n".join(log_texts)
-log_interface = gr.Interface(
-    fn=lambda: log_text,
-    inputs=None,
-    outputs="text",
-    title="EvalWhisper Log",
-)
-log_interface.launch()
-
-
-'''
     # Evaluate the model
     model.eval()
     print("model.eval ", model.eval())
@@ -102,7 +60,7 @@ log_interface.launch()
     
     # Print or return WER score
     print(f"Word Error Rate (WER): {wer_score}")
-'''
+
 '''
 print("check check")
 print(inputs)