File size: 1,662 Bytes
3801f51
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
import gradio as gr
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import torch


# Load model and tokenizer
MODEL_NAME = "comma-project/normalization-byt5-small"

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME)


def normalize_text(text: str) -> str:
    """
    Normalize input text using ByT5.
    """

    if not text.strip():
        return ""

    # Tokenize
    inputs = tokenizer(
        text,
        return_tensors="pt",
        truncation=True,
        padding=True,
        max_length=1024,
    )

    # Generate
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_length=1024,
            num_beams=2,
            early_stopping=True,
        )

    # Decode
    normalized = tokenizer.decode(
        outputs[0],
        skip_special_tokens=True,
    )

    return normalized


# Gradio interface
demo = gr.Interface(
    fn=normalize_text,
    inputs=gr.Textbox(
        label="Input Text",
        placeholder="Enter text to normalize...",
        lines=4,
    ),
    outputs=gr.Textbox(
        label="Normalized Text",
        lines=4,
    ),
    title="Text Normalization with ByT5",
    description="Normalize noisy or non-standard text using the ByT5 model.",
    theme="soft",
    examples=[
        ["Scͥbo uobiᷤᷤ ñ pauli ł donati."],
        ["""⁊ pitie mlt' lelasce
P ities li dist. uai a ton peire
Nelaissier. """, """Uer̃ ab his qͥ ita dissert̃
q̃ri debet. qͥd ꝑ amorem dei. quidq ꝑ amorẽ
boni tẽꝑalis ueluit intellig̾e."""]
    ],
)

if __name__ == "__main__":
    demo.launch()