Update app.py

app.py

@@ -5,60 +5,32 @@ MIT License
 
 https://github.com/clovaai/donut
 """
-import argparse
-
 import gradio as gr
 import torch
 from PIL import Image
 
 from donut import DonutModel
 
-
-def demo_process_vqa(input_img, question):
-    global pretrained_model, task_prompt, task_name
-    input_img = Image.fromarray(input_img)
-    user_prompt = task_prompt.replace("{user_input}", question)
-    output = pretrained_model.inference(input_img, prompt=user_prompt)["predictions"][0]
-    return output
-
-
 def demo_process(input_img):
     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
 
+task_prompt = f"<s_cord-v2>"
 
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser()
-    parser.add_argument("--task", type=str, default="cord-v2")
-    parser.add_argument("--pretrained_path", type=str, default="naver-clova-ix/donut-base-finetuned-cord-v2")
-    parser.add_argument("--port", type=int, default=None)
-    parser.add_argument("--url", type=str, default=None)
-    parser.add_argument("--sample_img_path", type=str, default="./sample_image_cord_test_receipt_00004.jpg")
-    args, left_argv = parser.parse_known_args()
+image = Image.open("./sample_image_cord_test_receipt_00004.jpg")
+image.save("cord_sample_receipt.jpg")
 
-    task_name = args.task
-    if "docvqa" == task_name:
-        task_prompt = "<s_docvqa><s_question>{user_input}</s_question><s_answer>"
-    else:  # rvlcdip, cord, ...
-        task_prompt = f"<s_{task_name}>"
-    
-    example_sample = []
-    if args.sample_img_path:
-        image = Image.open(args.sample_img_path)
-        image.save("cord_sample_receipt.jpg")
-        example_sample.append("cord_sample_receipt.jpg")
+pretrained_model = DonutModel.from_pretrained("naver-clova-ix/donut-base-finetuned-cord-v2")
+pretrained_model.encoder.to(torch.bfloat16)
+pretrained_model.eval()
 
-    pretrained_model = DonutModel.from_pretrained(args.pretrained_path)
-    pretrained_model.encoder.to(torch.bfloat16)
-    pretrained_model.eval()
-    
-    demo = gr.Interface(
-        fn=demo_process_vqa if task_name == "docvqa" else demo_process,
-        inputs=["image", "text"] if task_name == "docvqa" else "image",
-        outputs="json",
-        title=f"Donut 🍩 demonstration for `{task_name}` task",
-        examples=[example_sample] if example_sample else None,
-    )
-    demo.launch(server_name=args.url, server_port=args.port)
+demo = gr.Interface(
+    fn=demo_process,
+    inputs=gr.inputs.Image(type="pil"),
+    outputs="json",
+    title=f"Donut 🍩 demonstration for `{task_name}` task",
+    examples=[["cord_sample_receipt.jpg"]],
+)
+demo.launch(debug=True)