File size: 2,624 Bytes
56670e1
 
 
5f5dede
56670e1
2223ef6
 
 
 
 
 
 
 
 
 
5f9bfe0
 
 
 
5f5dede
 
 
 
 
 
 
 
 
 
56670e1
 
81ceb92
 
 
56670e1
a95cf91
 
 
 
 
ddb6473
56670e1
81ceb92
5f5dede
 
 
 
 
 
 
 
 
 
 
 
 
dacd217
5f5dede
 
 
 
 
 
 
 
56670e1
 
 
2223ef6
 
 
 
5f5dede
56670e1
 
 
46f08aa
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
76
77
78
import gradio as gr
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import re

models = {
    "RUSpam/spam_deberta_v4": "RUSpam/spam_deberta_v4",
    "RUSpam/spamNS_v1": "RUSpam/spamNS_v1"
}

tokenizers = {}
model_instances = {}

for name, path in models.items():
    tokenizers[name] = AutoTokenizer.from_pretrained(path)
    model_instances[name] = AutoModelForSequenceClassification.from_pretrained(path)

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model_instances["RUSpam/spamNS_v1"] = model_instances["RUSpam/spamNS_v1"].to(device).eval()

def clean_text(text):
    text = re.sub(r'http\S+', '', text)
    text = re.sub(r'[^А-Яа-я0-9 ]+', ' ', text)
    text = text.lower().strip()
    return text

def predict_spam_deberta(text):
    tokenizer = tokenizers["RUSpam/spam_deberta_v4"]
    model = model_instances["RUSpam/spam_deberta_v4"]

    inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=256)
    input_ids = inputs['input_ids'].to(device)
    attention_mask = inputs['attention_mask'].to(device)

    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        predicted_class = torch.argmax(logits, dim=1).item()
    
    result = "Спам" if predicted_class == 1 else "Не спам"
    return result


def predict_spam_spamns(text):
    tokenizer = tokenizers["RUSpam/spamNS_v1"]
    model = model_instances["RUSpam/spamNS_v1"]

    text = clean_text(text)
    encoding = tokenizer(text, padding='max_length', truncation=True, max_length=128, return_tensors='pt')
    input_ids = encoding['input_ids'].to(device)
    attention_mask = encoding['attention_mask'].to(device)

    with torch.no_grad():
        outputs = model(input_ids, attention_mask=attention_mask).logits
        pred = torch.sigmoid(outputs).cpu().numpy()[0][0]

    result = "Спам" if pred >= 0.5 else "Не спам"
    return result

def predict_spam(text, model_choice):
    if model_choice == "RUSpam/spam_deberta_v4":
        return predict_spam_deberta(text)
    elif model_choice == "RUSpam/spamNS_v1":
        return predict_spam_spamns(text)

# Создание интерфейса Gradio
iface = gr.Interface(
    fn=predict_spam,
    inputs=[
        gr.Textbox(lines=5, label="Введите текст"),
        gr.Radio(choices=list(models.keys()), label="Выберите модель", value="RUSpam/spam_deberta_v4")
    ],
    outputs=gr.Label(label="Результат"),
    title="Определение спама в русскоязычных текстах"
)

iface.launch()