feat: select default image and parse receipt with U🍩, raw output

app.py

@@ -6,7 +6,7 @@ from transformers import VisionEncoderDecoderModel, VisionEncoderDecoderConfig
 
 
 def demo_process(input_img):
-    global pretrained_model, task_prompt, task_name
+    global pretrained_model, task_prompt  # , task_name
     # input_img = Image.fromarray(input_img)
     output = pretrained_model.inference(image=input_img, prompt=task_prompt)["predictions"][0]
     return output
@@ -22,9 +22,18 @@ with st.sidebar:
     information = st.radio(
     "What information inside the are you interested in?",
     ('Receipt Summary', 'Receipt Menu Details', 'Extract all!'))
-    receipt = st.selectbox('Pick one receipt', ['1', '2', '3', '4', '5', '6'])
+    receipt = st.selectbox('Pick one receipt', ['1', '2', '3', '4', '5', '6'], index='6')
 
 st.text(f'{information} mode is ON!\nTarget receipt: {receipt}\n(opening image @:./img/receipt-{receipt}.png)')
 
 image = Image.open(f"./img/receipt-{receipt}.jpg")
-st.image(image, caption='Your target receipt')
+st.image(image, caption='Your target receipt')
+
+st.text(f'baking the 🍩...')
+pretrained_model = VisionEncoderDecoderModel.from_pretrained("unstructured/donut-base-sroie")
+pretrained_model.encoder.to(torch.bfloat16)
+pretrained_model.eval()
+
+st.text(f'parsing receipt..')
+parsed_receipt_info = demo_process(image)
+st.text(f'\nRaw output:\n{parsed_receipt_info}')