Update app.py

app.py

@@ -27,7 +27,9 @@ def generate_caption(image):
     # autoregressively complete prompt
     output = model.generate(**inputs, max_new_tokens=200)
 
-    return processor.decode(output[0], skip_special_tokens=True)
+    prompt_len = inputs["input_ids"].shape[1]
+    decoded_text = processor.batch_decode(output[:, prompt_len:])[0]
+    return decoded_text
 
 
 # Define the Gradio interface