Spaces:
Runtime error
Runtime error
Update stri.py
Browse files
stri.py
CHANGED
|
@@ -2,6 +2,7 @@ import streamlit as st
|
|
| 2 |
import torch
|
| 3 |
import numpy as np
|
| 4 |
import pandas as pd
|
|
|
|
| 5 |
from transformers import AutoTokenizer, AutoModel
|
| 6 |
import re
|
| 7 |
import pickle
|
|
@@ -17,7 +18,6 @@ model = AutoModel.from_pretrained(model_name, output_hidden_states=True)
|
|
| 17 |
books = pd.read_csv('books_6000.csv')
|
| 18 |
books.dropna(inplace=True)
|
| 19 |
|
| 20 |
-
|
| 21 |
books = books[books['annotation'].apply(lambda x: len(x.split()) >= 10)]
|
| 22 |
books.drop_duplicates(subset='title', keep='first', inplace=True)
|
| 23 |
books = books.reset_index(drop=True)
|
|
@@ -42,24 +42,24 @@ max_len = 128
|
|
| 42 |
# Определение запроса пользователя
|
| 43 |
query = st.text_input("Введите запрос")
|
| 44 |
|
| 45 |
-
if st.button('
|
| 46 |
with open("book_embeddings.pkl", "rb") as f:
|
| 47 |
book_embeddings = pickle.load(f)
|
| 48 |
-
|
| 49 |
query_tokens = tokenizer.encode(query, add_special_tokens=True,
|
| 50 |
truncation=True, max_length=max_len)
|
| 51 |
-
|
| 52 |
query_padded = np.array(query_tokens + [0] * (max_len - len(query_tokens)))
|
| 53 |
query_mask = np.where(query_padded != 0, 1, 0)
|
| 54 |
-
|
| 55 |
# Переведем numpy массивы в тензоры PyTorch
|
| 56 |
query_padded = torch.tensor(query_padded, dtype=torch.long)
|
| 57 |
query_mask = torch.tensor(query_mask, dtype=torch.long)
|
| 58 |
-
|
| 59 |
with torch.no_grad():
|
| 60 |
query_embedding = model(query_padded.unsqueeze(0), query_mask.unsqueeze(0))
|
| 61 |
-
query_embedding = query_embedding[0][:, 0, :]
|
| 62 |
-
|
| 63 |
# Вычисление косинусного расстояния между эмбеддингом запроса и каждой аннотацией
|
| 64 |
cosine_similarities = torch.nn.functional.cosine_similarity(
|
| 65 |
query_embedding.squeeze(0),
|
|
@@ -67,8 +67,14 @@ if st.button('**Generating recommendations**'):
|
|
| 67 |
)
|
| 68 |
|
| 69 |
cosine_similarities = cosine_similarities.numpy()
|
| 70 |
-
|
| 71 |
indices = np.argsort(cosine_similarities)[::-1] # Сортировка по убыванию
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
import torch
|
| 3 |
import numpy as np
|
| 4 |
import pandas as pd
|
| 5 |
+
from PIL import Image
|
| 6 |
from transformers import AutoTokenizer, AutoModel
|
| 7 |
import re
|
| 8 |
import pickle
|
|
|
|
| 18 |
books = pd.read_csv('books_6000.csv')
|
| 19 |
books.dropna(inplace=True)
|
| 20 |
|
|
|
|
| 21 |
books = books[books['annotation'].apply(lambda x: len(x.split()) >= 10)]
|
| 22 |
books.drop_duplicates(subset='title', keep='first', inplace=True)
|
| 23 |
books = books.reset_index(drop=True)
|
|
|
|
| 42 |
# Определение запроса пользователя
|
| 43 |
query = st.text_input("Введите запрос")
|
| 44 |
|
| 45 |
+
if st.button('Сгенерировать'):
|
| 46 |
with open("book_embeddings.pkl", "rb") as f:
|
| 47 |
book_embeddings = pickle.load(f)
|
| 48 |
+
|
| 49 |
query_tokens = tokenizer.encode(query, add_special_tokens=True,
|
| 50 |
truncation=True, max_length=max_len)
|
| 51 |
+
|
| 52 |
query_padded = np.array(query_tokens + [0] * (max_len - len(query_tokens)))
|
| 53 |
query_mask = np.where(query_padded != 0, 1, 0)
|
| 54 |
+
|
| 55 |
# Переведем numpy массивы в тензоры PyTorch
|
| 56 |
query_padded = torch.tensor(query_padded, dtype=torch.long)
|
| 57 |
query_mask = torch.tensor(query_mask, dtype=torch.long)
|
| 58 |
+
|
| 59 |
with torch.no_grad():
|
| 60 |
query_embedding = model(query_padded.unsqueeze(0), query_mask.unsqueeze(0))
|
| 61 |
+
query_embedding = query_embedding[0][:, 0, :]
|
| 62 |
+
|
| 63 |
# Вычисление косинусного расстояния между эмбеддингом запроса и каждой аннотацией
|
| 64 |
cosine_similarities = torch.nn.functional.cosine_similarity(
|
| 65 |
query_embedding.squeeze(0),
|
|
|
|
| 67 |
)
|
| 68 |
|
| 69 |
cosine_similarities = cosine_similarities.numpy()
|
| 70 |
+
|
| 71 |
indices = np.argsort(cosine_similarities)[::-1] # Сортировка по убыванию
|
| 72 |
+
|
| 73 |
+
num_books_per_page = st.selectbox("Количество книг на странице:", [3, 5, 10], index=0)
|
| 74 |
+
|
| 75 |
+
for i in indices[:num_books_per_page]:
|
| 76 |
+
st.write("## " + books['title'][i])
|
| 77 |
+
st.write("**Автор:**", books['author'][i])
|
| 78 |
+
st.write("**Аннотация:**", books['annotation'][i])
|
| 79 |
+
st.image(Image.open(books['image_url'][i]))
|
| 80 |
+
st.write("---")
|