Add file
Browse files
app.py
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import re
|
3 |
+
import gradio as gr
|
4 |
+
from transformers import DonutProcessor, VisionEncoderDecoderModel
|
5 |
+
import torch
|
6 |
+
import traceback
|
7 |
+
|
8 |
+
# 1) Load pretrained Donut model and processor
|
9 |
+
MODEL_NAME = "naver-clova-ix/donut-base-finetuned-cord-v2"
|
10 |
+
processor = DonutProcessor.from_pretrained(MODEL_NAME)
|
11 |
+
model = VisionEncoderDecoderModel.from_pretrained(MODEL_NAME)
|
12 |
+
|
13 |
+
# 2) Set device and move model
|
14 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
15 |
+
model.to(device)
|
16 |
+
|
17 |
+
# 3) Inference function with debugging
|
18 |
+
def ocr_donut(image):
|
19 |
+
try:
|
20 |
+
if image is None:
|
21 |
+
return {"error": "No image provided."}
|
22 |
+
|
23 |
+
# Prepare prompt and inputs
|
24 |
+
task_prompt = "<s_cord-v2>"
|
25 |
+
decoder_input_ids = processor.tokenizer(
|
26 |
+
task_prompt,
|
27 |
+
add_special_tokens=False,
|
28 |
+
return_tensors="pt"
|
29 |
+
).input_ids.to(device)
|
30 |
+
|
31 |
+
# Convert to tensor
|
32 |
+
pixel_values = processor(image.convert("RGB"), return_tensors="pt").pixel_values.to(device)
|
33 |
+
|
34 |
+
# Generate outputs
|
35 |
+
outputs = model.generate(
|
36 |
+
pixel_values,
|
37 |
+
decoder_input_ids=decoder_input_ids,
|
38 |
+
max_length=model.config.decoder.max_position_embeddings,
|
39 |
+
pad_token_id=processor.tokenizer.pad_token_id,
|
40 |
+
eos_token_id=processor.tokenizer.eos_token_id,
|
41 |
+
use_cache=True,
|
42 |
+
bad_words_ids=[[processor.tokenizer.unk_token_id]],
|
43 |
+
return_dict_in_generate=True,
|
44 |
+
)
|
45 |
+
|
46 |
+
# Decode and clean up
|
47 |
+
sequence = processor.batch_decode(outputs.sequences)[0]
|
48 |
+
sequence = sequence.replace(processor.tokenizer.eos_token, "").replace(processor.tokenizer.pad_token, "")
|
49 |
+
sequence = re.sub(r"<.*?>", "", sequence, count=1).strip()
|
50 |
+
json_output = processor.token2json(sequence)
|
51 |
+
|
52 |
+
return {"result": json_output}
|
53 |
+
|
54 |
+
except Exception:
|
55 |
+
tb = traceback.format_exc()
|
56 |
+
print(tb)
|
57 |
+
return {"error": tb}
|
58 |
+
|
59 |
+
# 4) Build Gradio interface
|
60 |
+
demo = gr.Interface(
|
61 |
+
fn=ocr_donut,
|
62 |
+
inputs=gr.Image(type="pil", label="Upload Document Image"),
|
63 |
+
outputs=gr.JSON(label="Output"),
|
64 |
+
title="Donut OCR Gradio App",
|
65 |
+
description="Upload a document image and get structured JSON output. Errors will be shown for debugging."
|
66 |
+
)
|
67 |
+
|
68 |
+
# 5) Launch for Spaces
|
69 |
+
demo.launch(
|
70 |
+
server_name="0.0.0.0",
|
71 |
+
server_port=int(os.environ.get("PORT", 7860)),
|
72 |
+
debug=True
|
73 |
+
)
|