Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,115 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import streamlit as st
|
2 |
+
import numpy as np
|
3 |
+
import pandas as pd
|
4 |
+
import torch
|
5 |
+
import transformers
|
6 |
+
from transformers import AutoTokenizer, AutoModelForSequenceClassification
|
7 |
+
from scipy.special import softmax
|
8 |
+
|
9 |
+
|
10 |
+
|
11 |
+
def load_model():
|
12 |
+
model = AutoModelForSequenceClassification.from_pretrained('model_distilbert_trained')
|
13 |
+
tokenizer = AutoTokenizer.from_pretrained(
|
14 |
+
'distilbert-base-cased', do_lower_case=True)
|
15 |
+
model.eval()
|
16 |
+
return model, tokenizer
|
17 |
+
|
18 |
+
|
19 |
+
def get_predictions(logits, indexes):
|
20 |
+
sum = 0
|
21 |
+
ind = []
|
22 |
+
probs = []
|
23 |
+
for i in indexes:
|
24 |
+
sum += logits[i]
|
25 |
+
ind.append(i)
|
26 |
+
probs.append(indexes[i])
|
27 |
+
if sum >= 0.95:
|
28 |
+
return ind, probs
|
29 |
+
|
30 |
+
|
31 |
+
def return_pred_name(name_dict, ind):
|
32 |
+
out = []
|
33 |
+
for i in ind:
|
34 |
+
out.append(name_dict[i])
|
35 |
+
return out
|
36 |
+
|
37 |
+
|
38 |
+
def predict(title, summary, model, tokenizer):
|
39 |
+
text = title + '.' + summary
|
40 |
+
tokens = tokenizer.encode(text)
|
41 |
+
with torch.no_grad():
|
42 |
+
logits = model(torch.as_tensor([tokens]))[0]
|
43 |
+
probs = torch.softmax(logits[-1, :], dim=-1).data.cpu().numpy()
|
44 |
+
|
45 |
+
classes = np.flip(np.argsort(probs))
|
46 |
+
sum_probs = 0
|
47 |
+
ind = 0
|
48 |
+
prediction = []
|
49 |
+
prediction_probs = []
|
50 |
+
while sum_probs < 0.95:
|
51 |
+
prediction.append(name_dict[classes[ind]])
|
52 |
+
prediction_probs.append(str("{:.2f}".format(100 * probs[classes[ind]])) + "%")
|
53 |
+
sum_probs += probs[classes[ind]]
|
54 |
+
ind += 1
|
55 |
+
|
56 |
+
return prediction, prediction_probs
|
57 |
+
|
58 |
+
|
59 |
+
def get_results(prediction, prediction_probs):
|
60 |
+
frame = pd.DataFrame({'Category': prediction, 'Confidence': prediction_probs})
|
61 |
+
frame.index = np.arange(1, len(frame) + 1)
|
62 |
+
return frame
|
63 |
+
|
64 |
+
name_dict = {4: 'cs',
|
65 |
+
19: 'stat',
|
66 |
+
1: 'astro-ph',
|
67 |
+
16: 'q-bio',
|
68 |
+
6: 'eess',
|
69 |
+
3: 'cond-mat',
|
70 |
+
12: 'math',
|
71 |
+
15: 'physics',
|
72 |
+
18: 'quant-ph',
|
73 |
+
17: 'q-fin',
|
74 |
+
7: 'gr-qc',
|
75 |
+
13: 'nlin',
|
76 |
+
2: 'cmp-lg',
|
77 |
+
5: 'econ',
|
78 |
+
8: 'hep-ex',
|
79 |
+
11: 'hep-th',
|
80 |
+
14: 'nucl-th',
|
81 |
+
10: 'hep-ph',
|
82 |
+
9: 'hep-lat',
|
83 |
+
0: 'adap-org'}
|
84 |
+
|
85 |
+
|
86 |
+
|
87 |
+
st.title("Find out the topic of the article without reading!")
|
88 |
+
st.markdown("<h1 style='text-align: center;'><img width=320px src = 'https://upload.wikimedia.org/wikipedia/ru/8/81/Sheldon_cooper.jpg'>",
|
89 |
+
unsafe_allow_html=True)
|
90 |
+
# ^-- можно показывать пользователю текст, картинки, ограниченное подмножество html - всё как в jupyter
|
91 |
+
|
92 |
+
title = st.text_area(label='Title',
|
93 |
+
value='',
|
94 |
+
height=30,
|
95 |
+
help='If you know a title type it here')
|
96 |
+
|
97 |
+
|
98 |
+
summary = st.text_area(label='Summary',
|
99 |
+
value='',
|
100 |
+
height=200,
|
101 |
+
help='If you have a summary enter it here')
|
102 |
+
|
103 |
+
|
104 |
+
button = st.button(label='Get the theme!')
|
105 |
+
|
106 |
+
if button:
|
107 |
+
if (title == '' and summary == ''):
|
108 |
+
st.write('There is nothing to analyze...')
|
109 |
+
st.write('Fill at list one of the fields')
|
110 |
+
else:
|
111 |
+
model, tokenizer = load_model()
|
112 |
+
prediction, prediction_probs = predict(title, summary, model, tokenizer)
|
113 |
+
ans = get_results(prediction, prediction_probs)
|
114 |
+
st.write('Result')
|
115 |
+
st.write(ans)
|