Spaces:
Sleeping
Sleeping
pochti final
Browse files
bot/bot.py
CHANGED
|
@@ -29,7 +29,7 @@ start_keyboard = ReplyKeyboardMarkup(
|
|
| 29 |
@lru_cache(maxsize=1)
|
| 30 |
def load_model():
|
| 31 |
model = BERTClassifier()
|
| 32 |
-
weights_path = 'bot/
|
| 33 |
state_dict = torch.load(weights_path, map_location=device)
|
| 34 |
model.load_state_dict(state_dict)
|
| 35 |
model.to(device)
|
|
|
|
| 29 |
@lru_cache(maxsize=1)
|
| 30 |
def load_model():
|
| 31 |
model = BERTClassifier()
|
| 32 |
+
weights_path = 'bot/model_weights_new.pth'
|
| 33 |
state_dict = torch.load(weights_path, map_location=device)
|
| 34 |
model.load_state_dict(state_dict)
|
| 35 |
model.to(device)
|
images/toxity_metrics.png
DELETED
|
Binary file (50.2 kB)
|
|
|
models/model2/__pycache__/model.cpython-310.pyc
CHANGED
|
Binary files a/models/model2/__pycache__/model.cpython-310.pyc and b/models/model2/__pycache__/model.cpython-310.pyc differ
|
|
|
models/model2/__pycache__/preprocess_text.cpython-310.pyc
CHANGED
|
Binary files a/models/model2/__pycache__/preprocess_text.cpython-310.pyc and b/models/model2/__pycache__/preprocess_text.cpython-310.pyc differ
|
|
|
models/model2/model_weights.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0b84f9c8041dd44751288c4777723fb4ff4b3886423f9f6efca37e43c6492429
|
| 3 |
-
size 47712485
|
|
|
|
|
|
|
|
|
|
|
|
pages/comments.py
DELETED
|
@@ -1,70 +0,0 @@
|
|
| 1 |
-
import streamlit as st
|
| 2 |
-
import torch
|
| 3 |
-
import sys
|
| 4 |
-
from pathlib import Path
|
| 5 |
-
import requests
|
| 6 |
-
import time
|
| 7 |
-
import cv2
|
| 8 |
-
import numpy as np
|
| 9 |
-
from transformers import AutoTokenizer
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
st.write("# Оценка степени токсичности пользовательского сообщения")
|
| 13 |
-
# st.write("Здесь вы можете загрузить картинку со своего устройства, либо при помощи ссылки")
|
| 14 |
-
|
| 15 |
-
# Добавление пути к проекту и моделям
|
| 16 |
-
project_root = Path(__file__).resolve().parents[1]
|
| 17 |
-
models_path = project_root / 'models'
|
| 18 |
-
sys.path.append(str(models_path))
|
| 19 |
-
from models.model2.preprocess_text import TextPreprocessorBERT
|
| 20 |
-
from models.model2.model import BERTClassifier
|
| 21 |
-
|
| 22 |
-
device = 'cpu'
|
| 23 |
-
|
| 24 |
-
# Загрузка модели и словаря
|
| 25 |
-
@st.cache_resource
|
| 26 |
-
def load_model():
|
| 27 |
-
model = BERTClassifier()
|
| 28 |
-
weights_path = models_path / 'model2' / 'model_weights_new.pth'
|
| 29 |
-
state_dict = torch.load(weights_path, map_location=device)
|
| 30 |
-
model.load_state_dict(state_dict)
|
| 31 |
-
model.to(device)
|
| 32 |
-
model.eval()
|
| 33 |
-
return model
|
| 34 |
-
|
| 35 |
-
@st.cache_resource
|
| 36 |
-
def load_tokenizer():
|
| 37 |
-
return AutoTokenizer.from_pretrained('cointegrated/rubert-tiny-toxicity')
|
| 38 |
-
|
| 39 |
-
model = load_model()
|
| 40 |
-
tokenizer = load_tokenizer()
|
| 41 |
-
|
| 42 |
-
input_text = st.text_area('Введите текст сообщения')
|
| 43 |
-
|
| 44 |
-
if st.button('Предсказать'):
|
| 45 |
-
# Применяем предобработку
|
| 46 |
-
preprocessor = TextPreprocessorBERT()
|
| 47 |
-
preprocessed_text = preprocessor.transform(input_text)
|
| 48 |
-
|
| 49 |
-
# Токенизация
|
| 50 |
-
tokens = tokenizer.encode_plus(
|
| 51 |
-
preprocessed_text,
|
| 52 |
-
add_special_tokens=True,
|
| 53 |
-
truncation=True,
|
| 54 |
-
max_length=100,
|
| 55 |
-
padding='max_length',
|
| 56 |
-
return_tensors='pt'
|
| 57 |
-
)
|
| 58 |
-
|
| 59 |
-
# Получаем input_ids и attention_mask из токенов
|
| 60 |
-
input_ids = tokens['input_ids'].to(device)
|
| 61 |
-
attention_mask = tokens['attention_mask'].to(device)
|
| 62 |
-
|
| 63 |
-
# Предсказание
|
| 64 |
-
with torch.no_grad():
|
| 65 |
-
output = model(input_ids, attention_mask=attention_mask)
|
| 66 |
-
|
| 67 |
-
# Интерпретация результата
|
| 68 |
-
prediction = torch.sigmoid(output).item()
|
| 69 |
-
st.write(f'Предсказанный класс токсичности: {prediction:.4f}')
|
| 70 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
pages/policlinic.py
DELETED
|
@@ -1,17 +0,0 @@
|
|
| 1 |
-
import streamlit as st
|
| 2 |
-
import joblib
|
| 3 |
-
import pandas as pd
|
| 4 |
-
from models.model1.Custom_class import TextPreprocessor
|
| 5 |
-
|
| 6 |
-
# Load the trained pipeline
|
| 7 |
-
pipeline = joblib.load('models/model1/logistic_regression_pipeline.pkl')
|
| 8 |
-
|
| 9 |
-
# Streamlit application
|
| 10 |
-
st.title('Классификация отзывов на русском языке')
|
| 11 |
-
|
| 12 |
-
input_text = st.text_area('Введите текст отзыва')
|
| 13 |
-
|
| 14 |
-
if st.button('Предсказать'):
|
| 15 |
-
prediction = pipeline.predict(pd.Series([input_text]))
|
| 16 |
-
st.write(f'Предсказанный класс с помощью логрег: {prediction[0]}')
|
| 17 |
-
st.write(f'1 - negative, 0 - positive')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|