df experiment

app.py

@@ -1,6 +1,7 @@
 import re
 import gradio
 import torch
+import pandas as pd
 
 from PIL import Image
 from transformers import DonutProcessor, VisionEncoderDecoderModel
@@ -38,9 +39,10 @@ def process_document(image):
     # postprocess
     sequence = processor.batch_decode(outputs.sequences)[0]
     sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
-    sequence = re.sub(r"<.*?>", "", sequence).strip()  # remove first task start token
+    sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()  # remove first task start token
     
-    return {'output': sequence}
+    js = processor.token2json(sequence)
+    return pd.json_normalize(js)
 
 demo = gradio.Interface(
     fn=process_document,