Spaces:
Running
Running
Andrey Moskalenko
commited on
Commit
•
3d38624
1
Parent(s):
c94270d
Upload Train_fakenews_detector.ipynb
Browse files- Train_fakenews_detector.ipynb +1465 -0
Train_fakenews_detector.ipynb
ADDED
@@ -0,0 +1,1465 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"cells": [
|
3 |
+
{
|
4 |
+
"cell_type": "markdown",
|
5 |
+
"metadata": {},
|
6 |
+
"source": [
|
7 |
+
"# Data Preparation"
|
8 |
+
]
|
9 |
+
},
|
10 |
+
{
|
11 |
+
"cell_type": "markdown",
|
12 |
+
"metadata": {},
|
13 |
+
"source": [
|
14 |
+
"Я нашел три датасета на kaggle по классификации фейков. Они все на английском, поэтому для поддержки русскуязычных статей будем использовать специально обученную для перевода новостей модель wmt19-ru-en. \n",
|
15 |
+
"\n",
|
16 |
+
"Выбранные датасеты:\n",
|
17 |
+
"* https://www.kaggle.com/c/fake-news/data\n",
|
18 |
+
"* https://www.kaggle.com/c/fakenewskdd2020/data\n",
|
19 |
+
"* https://www.kaggle.com/c/classifying-the-fake-news/data"
|
20 |
+
]
|
21 |
+
},
|
22 |
+
{
|
23 |
+
"cell_type": "code",
|
24 |
+
"execution_count": 95,
|
25 |
+
"metadata": {},
|
26 |
+
"outputs": [],
|
27 |
+
"source": [
|
28 |
+
"import pandas as pd\n",
|
29 |
+
"\n",
|
30 |
+
"df1_train = pd.read_csv('./data1/train.csv')"
|
31 |
+
]
|
32 |
+
},
|
33 |
+
{
|
34 |
+
"cell_type": "code",
|
35 |
+
"execution_count": 96,
|
36 |
+
"metadata": {},
|
37 |
+
"outputs": [
|
38 |
+
{
|
39 |
+
"data": {
|
40 |
+
"text/html": [
|
41 |
+
"<div>\n",
|
42 |
+
"<style scoped>\n",
|
43 |
+
" .dataframe tbody tr th:only-of-type {\n",
|
44 |
+
" vertical-align: middle;\n",
|
45 |
+
" }\n",
|
46 |
+
"\n",
|
47 |
+
" .dataframe tbody tr th {\n",
|
48 |
+
" vertical-align: top;\n",
|
49 |
+
" }\n",
|
50 |
+
"\n",
|
51 |
+
" .dataframe thead th {\n",
|
52 |
+
" text-align: right;\n",
|
53 |
+
" }\n",
|
54 |
+
"</style>\n",
|
55 |
+
"<table border=\"1\" class=\"dataframe\">\n",
|
56 |
+
" <thead>\n",
|
57 |
+
" <tr style=\"text-align: right;\">\n",
|
58 |
+
" <th></th>\n",
|
59 |
+
" <th>id</th>\n",
|
60 |
+
" <th>title</th>\n",
|
61 |
+
" <th>author</th>\n",
|
62 |
+
" <th>text</th>\n",
|
63 |
+
" <th>label</th>\n",
|
64 |
+
" </tr>\n",
|
65 |
+
" </thead>\n",
|
66 |
+
" <tbody>\n",
|
67 |
+
" <tr>\n",
|
68 |
+
" <th>0</th>\n",
|
69 |
+
" <td>0</td>\n",
|
70 |
+
" <td>House Dem Aide: We Didn’t Even See Comey’s Let...</td>\n",
|
71 |
+
" <td>Darrell Lucus</td>\n",
|
72 |
+
" <td>House Dem Aide: We Didn’t Even See Comey’s Let...</td>\n",
|
73 |
+
" <td>1</td>\n",
|
74 |
+
" </tr>\n",
|
75 |
+
" <tr>\n",
|
76 |
+
" <th>1</th>\n",
|
77 |
+
" <td>1</td>\n",
|
78 |
+
" <td>FLYNN: Hillary Clinton, Big Woman on Campus - ...</td>\n",
|
79 |
+
" <td>Daniel J. Flynn</td>\n",
|
80 |
+
" <td>Ever get the feeling your life circles the rou...</td>\n",
|
81 |
+
" <td>0</td>\n",
|
82 |
+
" </tr>\n",
|
83 |
+
" <tr>\n",
|
84 |
+
" <th>2</th>\n",
|
85 |
+
" <td>2</td>\n",
|
86 |
+
" <td>Why the Truth Might Get You Fired</td>\n",
|
87 |
+
" <td>Consortiumnews.com</td>\n",
|
88 |
+
" <td>Why the Truth Might Get You Fired October 29, ...</td>\n",
|
89 |
+
" <td>1</td>\n",
|
90 |
+
" </tr>\n",
|
91 |
+
" <tr>\n",
|
92 |
+
" <th>3</th>\n",
|
93 |
+
" <td>3</td>\n",
|
94 |
+
" <td>15 Civilians Killed In Single US Airstrike Hav...</td>\n",
|
95 |
+
" <td>Jessica Purkiss</td>\n",
|
96 |
+
" <td>Videos 15 Civilians Killed In Single US Airstr...</td>\n",
|
97 |
+
" <td>1</td>\n",
|
98 |
+
" </tr>\n",
|
99 |
+
" <tr>\n",
|
100 |
+
" <th>4</th>\n",
|
101 |
+
" <td>4</td>\n",
|
102 |
+
" <td>Iranian woman jailed for fictional unpublished...</td>\n",
|
103 |
+
" <td>Howard Portnoy</td>\n",
|
104 |
+
" <td>Print \\nAn Iranian woman has been sentenced to...</td>\n",
|
105 |
+
" <td>1</td>\n",
|
106 |
+
" </tr>\n",
|
107 |
+
" <tr>\n",
|
108 |
+
" <th>...</th>\n",
|
109 |
+
" <td>...</td>\n",
|
110 |
+
" <td>...</td>\n",
|
111 |
+
" <td>...</td>\n",
|
112 |
+
" <td>...</td>\n",
|
113 |
+
" <td>...</td>\n",
|
114 |
+
" </tr>\n",
|
115 |
+
" <tr>\n",
|
116 |
+
" <th>20795</th>\n",
|
117 |
+
" <td>20795</td>\n",
|
118 |
+
" <td>Rapper T.I.: Trump a ’Poster Child For White S...</td>\n",
|
119 |
+
" <td>Jerome Hudson</td>\n",
|
120 |
+
" <td>Rapper T. I. unloaded on black celebrities who...</td>\n",
|
121 |
+
" <td>0</td>\n",
|
122 |
+
" </tr>\n",
|
123 |
+
" <tr>\n",
|
124 |
+
" <th>20796</th>\n",
|
125 |
+
" <td>20796</td>\n",
|
126 |
+
" <td>N.F.L. Playoffs: Schedule, Matchups and Odds -...</td>\n",
|
127 |
+
" <td>Benjamin Hoffman</td>\n",
|
128 |
+
" <td>When the Green Bay Packers lost to the Washing...</td>\n",
|
129 |
+
" <td>0</td>\n",
|
130 |
+
" </tr>\n",
|
131 |
+
" <tr>\n",
|
132 |
+
" <th>20797</th>\n",
|
133 |
+
" <td>20797</td>\n",
|
134 |
+
" <td>Macy’s Is Said to Receive Takeover Approach by...</td>\n",
|
135 |
+
" <td>Michael J. de la Merced and Rachel Abrams</td>\n",
|
136 |
+
" <td>The Macy’s of today grew from the union of sev...</td>\n",
|
137 |
+
" <td>0</td>\n",
|
138 |
+
" </tr>\n",
|
139 |
+
" <tr>\n",
|
140 |
+
" <th>20798</th>\n",
|
141 |
+
" <td>20798</td>\n",
|
142 |
+
" <td>NATO, Russia To Hold Parallel Exercises In Bal...</td>\n",
|
143 |
+
" <td>Alex Ansary</td>\n",
|
144 |
+
" <td>NATO, Russia To Hold Parallel Exercises In Bal...</td>\n",
|
145 |
+
" <td>1</td>\n",
|
146 |
+
" </tr>\n",
|
147 |
+
" <tr>\n",
|
148 |
+
" <th>20799</th>\n",
|
149 |
+
" <td>20799</td>\n",
|
150 |
+
" <td>What Keeps the F-35 Alive</td>\n",
|
151 |
+
" <td>David Swanson</td>\n",
|
152 |
+
" <td>David Swanson is an author, activist, journa...</td>\n",
|
153 |
+
" <td>1</td>\n",
|
154 |
+
" </tr>\n",
|
155 |
+
" </tbody>\n",
|
156 |
+
"</table>\n",
|
157 |
+
"<p>20800 rows × 5 columns</p>\n",
|
158 |
+
"</div>"
|
159 |
+
],
|
160 |
+
"text/plain": [
|
161 |
+
" id title \\\n",
|
162 |
+
"0 0 House Dem Aide: We Didn’t Even See Comey’s Let... \n",
|
163 |
+
"1 1 FLYNN: Hillary Clinton, Big Woman on Campus - ... \n",
|
164 |
+
"2 2 Why the Truth Might Get You Fired \n",
|
165 |
+
"3 3 15 Civilians Killed In Single US Airstrike Hav... \n",
|
166 |
+
"4 4 Iranian woman jailed for fictional unpublished... \n",
|
167 |
+
"... ... ... \n",
|
168 |
+
"20795 20795 Rapper T.I.: Trump a ’Poster Child For White S... \n",
|
169 |
+
"20796 20796 N.F.L. Playoffs: Schedule, Matchups and Odds -... \n",
|
170 |
+
"20797 20797 Macy’s Is Said to Receive Takeover Approach by... \n",
|
171 |
+
"20798 20798 NATO, Russia To Hold Parallel Exercises In Bal... \n",
|
172 |
+
"20799 20799 What Keeps the F-35 Alive \n",
|
173 |
+
"\n",
|
174 |
+
" author \\\n",
|
175 |
+
"0 Darrell Lucus \n",
|
176 |
+
"1 Daniel J. Flynn \n",
|
177 |
+
"2 Consortiumnews.com \n",
|
178 |
+
"3 Jessica Purkiss \n",
|
179 |
+
"4 Howard Portnoy \n",
|
180 |
+
"... ... \n",
|
181 |
+
"20795 Jerome Hudson \n",
|
182 |
+
"20796 Benjamin Hoffman \n",
|
183 |
+
"20797 Michael J. de la Merced and Rachel Abrams \n",
|
184 |
+
"20798 Alex Ansary \n",
|
185 |
+
"20799 David Swanson \n",
|
186 |
+
"\n",
|
187 |
+
" text label \n",
|
188 |
+
"0 House Dem Aide: We Didn’t Even See Comey’s Let... 1 \n",
|
189 |
+
"1 Ever get the feeling your life circles the rou... 0 \n",
|
190 |
+
"2 Why the Truth Might Get You Fired October 29, ... 1 \n",
|
191 |
+
"3 Videos 15 Civilians Killed In Single US Airstr... 1 \n",
|
192 |
+
"4 Print \\nAn Iranian woman has been sentenced to... 1 \n",
|
193 |
+
"... ... ... \n",
|
194 |
+
"20795 Rapper T. I. unloaded on black celebrities who... 0 \n",
|
195 |
+
"20796 When the Green Bay Packers lost to the Washing... 0 \n",
|
196 |
+
"20797 The Macy’s of today grew from the union of sev... 0 \n",
|
197 |
+
"20798 NATO, Russia To Hold Parallel Exercises In Bal... 1 \n",
|
198 |
+
"20799 David Swanson is an author, activist, journa... 1 \n",
|
199 |
+
"\n",
|
200 |
+
"[20800 rows x 5 columns]"
|
201 |
+
]
|
202 |
+
},
|
203 |
+
"execution_count": 96,
|
204 |
+
"metadata": {},
|
205 |
+
"output_type": "execute_result"
|
206 |
+
}
|
207 |
+
],
|
208 |
+
"source": [
|
209 |
+
"df1_train"
|
210 |
+
]
|
211 |
+
},
|
212 |
+
{
|
213 |
+
"cell_type": "code",
|
214 |
+
"execution_count": 97,
|
215 |
+
"metadata": {},
|
216 |
+
"outputs": [],
|
217 |
+
"source": [
|
218 |
+
"df1_train['text'] = df1_train.apply(lambda x: str(x.title) + '. ' + str(x.text), axis=1)\n",
|
219 |
+
"df1_train = df1_train[['text', 'label']]"
|
220 |
+
]
|
221 |
+
},
|
222 |
+
{
|
223 |
+
"cell_type": "code",
|
224 |
+
"execution_count": 98,
|
225 |
+
"metadata": {},
|
226 |
+
"outputs": [],
|
227 |
+
"source": [
|
228 |
+
"df2_train = pd.read_csv('./data2/train.csv', sep='\\t')"
|
229 |
+
]
|
230 |
+
},
|
231 |
+
{
|
232 |
+
"cell_type": "code",
|
233 |
+
"execution_count": 99,
|
234 |
+
"metadata": {},
|
235 |
+
"outputs": [],
|
236 |
+
"source": [
|
237 |
+
"# Битая строка\n",
|
238 |
+
"df2_train = df2_train.drop([1615])"
|
239 |
+
]
|
240 |
+
},
|
241 |
+
{
|
242 |
+
"cell_type": "code",
|
243 |
+
"execution_count": 100,
|
244 |
+
"metadata": {},
|
245 |
+
"outputs": [
|
246 |
+
{
|
247 |
+
"data": {
|
248 |
+
"text/html": [
|
249 |
+
"<div>\n",
|
250 |
+
"<style scoped>\n",
|
251 |
+
" .dataframe tbody tr th:only-of-type {\n",
|
252 |
+
" vertical-align: middle;\n",
|
253 |
+
" }\n",
|
254 |
+
"\n",
|
255 |
+
" .dataframe tbody tr th {\n",
|
256 |
+
" vertical-align: top;\n",
|
257 |
+
" }\n",
|
258 |
+
"\n",
|
259 |
+
" .dataframe thead th {\n",
|
260 |
+
" text-align: right;\n",
|
261 |
+
" }\n",
|
262 |
+
"</style>\n",
|
263 |
+
"<table border=\"1\" class=\"dataframe\">\n",
|
264 |
+
" <thead>\n",
|
265 |
+
" <tr style=\"text-align: right;\">\n",
|
266 |
+
" <th></th>\n",
|
267 |
+
" <th>text</th>\n",
|
268 |
+
" <th>label</th>\n",
|
269 |
+
" </tr>\n",
|
270 |
+
" </thead>\n",
|
271 |
+
" <tbody>\n",
|
272 |
+
" <tr>\n",
|
273 |
+
" <th>0</th>\n",
|
274 |
+
" <td>Get the latest from TODAY Sign up for our news...</td>\n",
|
275 |
+
" <td>1</td>\n",
|
276 |
+
" </tr>\n",
|
277 |
+
" <tr>\n",
|
278 |
+
" <th>1</th>\n",
|
279 |
+
" <td>2d Conan On The Funeral Trump Will Be Invited...</td>\n",
|
280 |
+
" <td>1</td>\n",
|
281 |
+
" </tr>\n",
|
282 |
+
" <tr>\n",
|
283 |
+
" <th>2</th>\n",
|
284 |
+
" <td>It’s safe to say that Instagram Stories has fa...</td>\n",
|
285 |
+
" <td>0</td>\n",
|
286 |
+
" </tr>\n",
|
287 |
+
" <tr>\n",
|
288 |
+
" <th>3</th>\n",
|
289 |
+
" <td>Much like a certain Amazon goddess with a lass...</td>\n",
|
290 |
+
" <td>0</td>\n",
|
291 |
+
" </tr>\n",
|
292 |
+
" <tr>\n",
|
293 |
+
" <th>4</th>\n",
|
294 |
+
" <td>At a time when the perfect outfit is just one ...</td>\n",
|
295 |
+
" <td>0</td>\n",
|
296 |
+
" </tr>\n",
|
297 |
+
" <tr>\n",
|
298 |
+
" <th>...</th>\n",
|
299 |
+
" <td>...</td>\n",
|
300 |
+
" <td>...</td>\n",
|
301 |
+
" </tr>\n",
|
302 |
+
" <tr>\n",
|
303 |
+
" <th>4982</th>\n",
|
304 |
+
" <td>The storybook romance of WWE stars John Cena a...</td>\n",
|
305 |
+
" <td>0</td>\n",
|
306 |
+
" </tr>\n",
|
307 |
+
" <tr>\n",
|
308 |
+
" <th>4983</th>\n",
|
309 |
+
" <td>The actor told friends he’s responsible for en...</td>\n",
|
310 |
+
" <td>0</td>\n",
|
311 |
+
" </tr>\n",
|
312 |
+
" <tr>\n",
|
313 |
+
" <th>4984</th>\n",
|
314 |
+
" <td>Sarah Hyland is getting real. The Modern Fami...</td>\n",
|
315 |
+
" <td>0</td>\n",
|
316 |
+
" </tr>\n",
|
317 |
+
" <tr>\n",
|
318 |
+
" <th>4985</th>\n",
|
319 |
+
" <td>Production has been suspended on the sixth and...</td>\n",
|
320 |
+
" <td>0</td>\n",
|
321 |
+
" </tr>\n",
|
322 |
+
" <tr>\n",
|
323 |
+
" <th>4986</th>\n",
|
324 |
+
" <td>A jury ruled against Bill Cosby in his sexual ...</td>\n",
|
325 |
+
" <td>0</td>\n",
|
326 |
+
" </tr>\n",
|
327 |
+
" </tbody>\n",
|
328 |
+
"</table>\n",
|
329 |
+
"<p>4986 rows × 2 columns</p>\n",
|
330 |
+
"</div>"
|
331 |
+
],
|
332 |
+
"text/plain": [
|
333 |
+
" text label\n",
|
334 |
+
"0 Get the latest from TODAY Sign up for our news... 1\n",
|
335 |
+
"1 2d Conan On The Funeral Trump Will Be Invited... 1\n",
|
336 |
+
"2 It’s safe to say that Instagram Stories has fa... 0\n",
|
337 |
+
"3 Much like a certain Amazon goddess with a lass... 0\n",
|
338 |
+
"4 At a time when the perfect outfit is just one ... 0\n",
|
339 |
+
"... ... ...\n",
|
340 |
+
"4982 The storybook romance of WWE stars John Cena a... 0\n",
|
341 |
+
"4983 The actor told friends he’s responsible for en... 0\n",
|
342 |
+
"4984 Sarah Hyland is getting real. The Modern Fami... 0\n",
|
343 |
+
"4985 Production has been suspended on the sixth and... 0\n",
|
344 |
+
"4986 A jury ruled against Bill Cosby in his sexual ... 0\n",
|
345 |
+
"\n",
|
346 |
+
"[4986 rows x 2 columns]"
|
347 |
+
]
|
348 |
+
},
|
349 |
+
"execution_count": 100,
|
350 |
+
"metadata": {},
|
351 |
+
"output_type": "execute_result"
|
352 |
+
}
|
353 |
+
],
|
354 |
+
"source": [
|
355 |
+
"df2_train"
|
356 |
+
]
|
357 |
+
},
|
358 |
+
{
|
359 |
+
"cell_type": "code",
|
360 |
+
"execution_count": 104,
|
361 |
+
"metadata": {},
|
362 |
+
"outputs": [],
|
363 |
+
"source": [
|
364 |
+
"df3_train = pd.read_csv('./data3/training.csv')"
|
365 |
+
]
|
366 |
+
},
|
367 |
+
{
|
368 |
+
"cell_type": "code",
|
369 |
+
"execution_count": 105,
|
370 |
+
"metadata": {},
|
371 |
+
"outputs": [],
|
372 |
+
"source": [
|
373 |
+
"df3_train['text'] = df3_train.apply(lambda x: str(x.title) + '. ' + str(x.text), axis=1)\n",
|
374 |
+
"df3_train = df3_train[['text', 'label']]"
|
375 |
+
]
|
376 |
+
},
|
377 |
+
{
|
378 |
+
"cell_type": "code",
|
379 |
+
"execution_count": 106,
|
380 |
+
"metadata": {},
|
381 |
+
"outputs": [],
|
382 |
+
"source": [
|
383 |
+
"all_data_train = df1_train.append(df2_train).append(df3_train)\n",
|
384 |
+
"all_data_train.to_csv('./train.csv', index=False)"
|
385 |
+
]
|
386 |
+
},
|
387 |
+
{
|
388 |
+
"cell_type": "markdown",
|
389 |
+
"metadata": {},
|
390 |
+
"source": [
|
391 |
+
"# Training"
|
392 |
+
]
|
393 |
+
},
|
394 |
+
{
|
395 |
+
"cell_type": "code",
|
396 |
+
"execution_count": 1,
|
397 |
+
"metadata": {
|
398 |
+
"id": "zriTdjauH8iQ"
|
399 |
+
},
|
400 |
+
"outputs": [],
|
401 |
+
"source": [
|
402 |
+
"#!pip install transformers\n",
|
403 |
+
"import transformers"
|
404 |
+
]
|
405 |
+
},
|
406 |
+
{
|
407 |
+
"cell_type": "code",
|
408 |
+
"execution_count": 2,
|
409 |
+
"metadata": {
|
410 |
+
"id": "TFh3upySL3XG"
|
411 |
+
},
|
412 |
+
"outputs": [],
|
413 |
+
"source": [
|
414 |
+
"from transformers import Trainer, TrainingArguments, LineByLineTextDataset"
|
415 |
+
]
|
416 |
+
},
|
417 |
+
{
|
418 |
+
"cell_type": "code",
|
419 |
+
"execution_count": 3,
|
420 |
+
"metadata": {
|
421 |
+
"id": "H2Ym6YhyNfON"
|
422 |
+
},
|
423 |
+
"outputs": [],
|
424 |
+
"source": [
|
425 |
+
"import pandas as pd"
|
426 |
+
]
|
427 |
+
},
|
428 |
+
{
|
429 |
+
"cell_type": "code",
|
430 |
+
"execution_count": 4,
|
431 |
+
"metadata": {
|
432 |
+
"id": "ueRyDnvgNgpW"
|
433 |
+
},
|
434 |
+
"outputs": [],
|
435 |
+
"source": [
|
436 |
+
"from datasets import Dataset"
|
437 |
+
]
|
438 |
+
},
|
439 |
+
{
|
440 |
+
"cell_type": "code",
|
441 |
+
"execution_count": 5,
|
442 |
+
"metadata": {
|
443 |
+
"id": "HVBCtqyjNhLn"
|
444 |
+
},
|
445 |
+
"outputs": [],
|
446 |
+
"source": [
|
447 |
+
"df = pd.read_csv('./train.csv')"
|
448 |
+
]
|
449 |
+
},
|
450 |
+
{
|
451 |
+
"cell_type": "code",
|
452 |
+
"execution_count": 6,
|
453 |
+
"metadata": {
|
454 |
+
"colab": {
|
455 |
+
"base_uri": "https://localhost:8080/",
|
456 |
+
"height": 424
|
457 |
+
},
|
458 |
+
"id": "f7j8fEl1Nogb",
|
459 |
+
"outputId": "3b5b13a0-4c34-412c-9718-5b0decb855cc"
|
460 |
+
},
|
461 |
+
"outputs": [
|
462 |
+
{
|
463 |
+
"data": {
|
464 |
+
"text/html": [
|
465 |
+
"<div>\n",
|
466 |
+
"<style scoped>\n",
|
467 |
+
" .dataframe tbody tr th:only-of-type {\n",
|
468 |
+
" vertical-align: middle;\n",
|
469 |
+
" }\n",
|
470 |
+
"\n",
|
471 |
+
" .dataframe tbody tr th {\n",
|
472 |
+
" vertical-align: top;\n",
|
473 |
+
" }\n",
|
474 |
+
"\n",
|
475 |
+
" .dataframe thead th {\n",
|
476 |
+
" text-align: right;\n",
|
477 |
+
" }\n",
|
478 |
+
"</style>\n",
|
479 |
+
"<table border=\"1\" class=\"dataframe\">\n",
|
480 |
+
" <thead>\n",
|
481 |
+
" <tr style=\"text-align: right;\">\n",
|
482 |
+
" <th></th>\n",
|
483 |
+
" <th>text</th>\n",
|
484 |
+
" <th>label</th>\n",
|
485 |
+
" </tr>\n",
|
486 |
+
" </thead>\n",
|
487 |
+
" <tbody>\n",
|
488 |
+
" <tr>\n",
|
489 |
+
" <th>0</th>\n",
|
490 |
+
" <td>House Dem Aide: We Didn’t Even See Comey’s Let...</td>\n",
|
491 |
+
" <td>1</td>\n",
|
492 |
+
" </tr>\n",
|
493 |
+
" <tr>\n",
|
494 |
+
" <th>1</th>\n",
|
495 |
+
" <td>FLYNN: Hillary Clinton, Big Woman on Campus - ...</td>\n",
|
496 |
+
" <td>0</td>\n",
|
497 |
+
" </tr>\n",
|
498 |
+
" <tr>\n",
|
499 |
+
" <th>2</th>\n",
|
500 |
+
" <td>Why the Truth Might Get You Fired.Why the Trut...</td>\n",
|
501 |
+
" <td>1</td>\n",
|
502 |
+
" </tr>\n",
|
503 |
+
" <tr>\n",
|
504 |
+
" <th>3</th>\n",
|
505 |
+
" <td>15 Civilians Killed In Single US Airstrike Hav...</td>\n",
|
506 |
+
" <td>1</td>\n",
|
507 |
+
" </tr>\n",
|
508 |
+
" <tr>\n",
|
509 |
+
" <th>4</th>\n",
|
510 |
+
" <td>Iranian woman jailed for fictional unpublished...</td>\n",
|
511 |
+
" <td>1</td>\n",
|
512 |
+
" </tr>\n",
|
513 |
+
" <tr>\n",
|
514 |
+
" <th>...</th>\n",
|
515 |
+
" <td>...</td>\n",
|
516 |
+
" <td>...</td>\n",
|
517 |
+
" </tr>\n",
|
518 |
+
" <tr>\n",
|
519 |
+
" <th>57209</th>\n",
|
520 |
+
" <td>CHICAGO TRUMP RALLY CANCELLED: Radicals And BL...</td>\n",
|
521 |
+
" <td>1</td>\n",
|
522 |
+
" </tr>\n",
|
523 |
+
" <tr>\n",
|
524 |
+
" <th>57210</th>\n",
|
525 |
+
" <td>Trump supports completion of Dakota Access Pip...</td>\n",
|
526 |
+
" <td>0</td>\n",
|
527 |
+
" </tr>\n",
|
528 |
+
" <tr>\n",
|
529 |
+
" <th>57211</th>\n",
|
530 |
+
" <td>Obama Can’t Stop Winning As New Jobs Report S...</td>\n",
|
531 |
+
" <td>1</td>\n",
|
532 |
+
" </tr>\n",
|
533 |
+
" <tr>\n",
|
534 |
+
" <th>57212</th>\n",
|
535 |
+
" <td>Turkey bank regulator dismisses 'rumors' after...</td>\n",
|
536 |
+
" <td>0</td>\n",
|
537 |
+
" </tr>\n",
|
538 |
+
" <tr>\n",
|
539 |
+
" <th>57213</th>\n",
|
540 |
+
" <td>California mayors ask for governor's support f...</td>\n",
|
541 |
+
" <td>0</td>\n",
|
542 |
+
" </tr>\n",
|
543 |
+
" </tbody>\n",
|
544 |
+
"</table>\n",
|
545 |
+
"<p>57214 rows × 2 columns</p>\n",
|
546 |
+
"</div>"
|
547 |
+
],
|
548 |
+
"text/plain": [
|
549 |
+
" text label\n",
|
550 |
+
"0 House Dem Aide: We Didn’t Even See Comey’s Let... 1\n",
|
551 |
+
"1 FLYNN: Hillary Clinton, Big Woman on Campus - ... 0\n",
|
552 |
+
"2 Why the Truth Might Get You Fired.Why the Trut... 1\n",
|
553 |
+
"3 15 Civilians Killed In Single US Airstrike Hav... 1\n",
|
554 |
+
"4 Iranian woman jailed for fictional unpublished... 1\n",
|
555 |
+
"... ... ...\n",
|
556 |
+
"57209 CHICAGO TRUMP RALLY CANCELLED: Radicals And BL... 1\n",
|
557 |
+
"57210 Trump supports completion of Dakota Access Pip... 0\n",
|
558 |
+
"57211 Obama Can’t Stop Winning As New Jobs Report S... 1\n",
|
559 |
+
"57212 Turkey bank regulator dismisses 'rumors' after... 0\n",
|
560 |
+
"57213 California mayors ask for governor's support f... 0\n",
|
561 |
+
"\n",
|
562 |
+
"[57214 rows x 2 columns]"
|
563 |
+
]
|
564 |
+
},
|
565 |
+
"execution_count": 6,
|
566 |
+
"metadata": {},
|
567 |
+
"output_type": "execute_result"
|
568 |
+
}
|
569 |
+
],
|
570 |
+
"source": [
|
571 |
+
"df"
|
572 |
+
]
|
573 |
+
},
|
574 |
+
{
|
575 |
+
"cell_type": "code",
|
576 |
+
"execution_count": 7,
|
577 |
+
"metadata": {
|
578 |
+
"id": "L0ET6Z83Pcxu"
|
579 |
+
},
|
580 |
+
"outputs": [],
|
581 |
+
"source": [
|
582 |
+
"df['labels'] = df['label']"
|
583 |
+
]
|
584 |
+
},
|
585 |
+
{
|
586 |
+
"cell_type": "code",
|
587 |
+
"execution_count": 8,
|
588 |
+
"metadata": {
|
589 |
+
"id": "39Zv6HBJPgEt"
|
590 |
+
},
|
591 |
+
"outputs": [],
|
592 |
+
"source": [
|
593 |
+
"df = df[['text', 'labels']]"
|
594 |
+
]
|
595 |
+
},
|
596 |
+
{
|
597 |
+
"cell_type": "code",
|
598 |
+
"execution_count": 9,
|
599 |
+
"metadata": {
|
600 |
+
"id": "bPGVPY17NI7x"
|
601 |
+
},
|
602 |
+
"outputs": [],
|
603 |
+
"source": [
|
604 |
+
"dataset = Dataset.from_pandas(df)"
|
605 |
+
]
|
606 |
+
},
|
607 |
+
{
|
608 |
+
"cell_type": "code",
|
609 |
+
"execution_count": 10,
|
610 |
+
"metadata": {
|
611 |
+
"colab": {
|
612 |
+
"base_uri": "https://localhost:8080/"
|
613 |
+
},
|
614 |
+
"id": "3LTGwWrINmZq",
|
615 |
+
"outputId": "177d8749-68cf-4f81-a91b-1097bf155478"
|
616 |
+
},
|
617 |
+
"outputs": [
|
618 |
+
{
|
619 |
+
"data": {
|
620 |
+
"text/plain": [
|
621 |
+
"Dataset({\n",
|
622 |
+
" features: ['text', 'labels'],\n",
|
623 |
+
" num_rows: 57214\n",
|
624 |
+
"})"
|
625 |
+
]
|
626 |
+
},
|
627 |
+
"execution_count": 10,
|
628 |
+
"metadata": {},
|
629 |
+
"output_type": "execute_result"
|
630 |
+
}
|
631 |
+
],
|
632 |
+
"source": [
|
633 |
+
"dataset"
|
634 |
+
]
|
635 |
+
},
|
636 |
+
{
|
637 |
+
"cell_type": "code",
|
638 |
+
"execution_count": 11,
|
639 |
+
"metadata": {
|
640 |
+
"colab": {
|
641 |
+
"base_uri": "https://localhost:8080/"
|
642 |
+
},
|
643 |
+
"id": "3DrWrMiDd7e-",
|
644 |
+
"outputId": "d331ebe6-5ed4-4fef-8a8d-41d25ed4b638"
|
645 |
+
},
|
646 |
+
"outputs": [],
|
647 |
+
"source": [
|
648 |
+
"import torch\n",
|
649 |
+
"from transformers import AutoTokenizer, AutoModel, pipeline\n",
|
650 |
+
"\n",
|
651 |
+
"model_name = 'distilbert-base-uncased-finetuned-sst-2-english'\n",
|
652 |
+
"tokenizer = AutoTokenizer.from_pretrained(model_name)"
|
653 |
+
]
|
654 |
+
},
|
655 |
+
{
|
656 |
+
"cell_type": "code",
|
657 |
+
"execution_count": 12,
|
658 |
+
"metadata": {
|
659 |
+
"id": "dRJOO2c5PT3V"
|
660 |
+
},
|
661 |
+
"outputs": [],
|
662 |
+
"source": [
|
663 |
+
"def preprocess_function(examples):\n",
|
664 |
+
" return tokenizer(examples[\"text\"], padding=True, truncation=True)"
|
665 |
+
]
|
666 |
+
},
|
667 |
+
{
|
668 |
+
"cell_type": "code",
|
669 |
+
"execution_count": 13,
|
670 |
+
"metadata": {
|
671 |
+
"colab": {
|
672 |
+
"base_uri": "https://localhost:8080/",
|
673 |
+
"height": 49,
|
674 |
+
"referenced_widgets": [
|
675 |
+
"5b49dc833234406da3da7435b9045fd2",
|
676 |
+
"300b70ed57dd493997afb0b3f25f4245",
|
677 |
+
"c03cc68b079c4e23b339e9de5ba38d29",
|
678 |
+
"57c3794731c84c42bb49618482b6b8cc",
|
679 |
+
"e306828f6d7444ddafce604e9a170467",
|
680 |
+
"9e11898bc51e483d91301387099368a4",
|
681 |
+
"a43574fa5fdf47ba9d5598b2b31f2082",
|
682 |
+
"482bae742d2a461cad525888e6ee8b91",
|
683 |
+
"e9c56275d73545a6961efe5704308ede",
|
684 |
+
"d604380b5e444f62ad36c4598230c561",
|
685 |
+
"c52ad745acb3423494b4ea5af5a934c7"
|
686 |
+
]
|
687 |
+
},
|
688 |
+
"id": "hCxs-HasPQ7s",
|
689 |
+
"outputId": "be4f8483-316c-4677-f804-12c78f358fac"
|
690 |
+
},
|
691 |
+
"outputs": [
|
692 |
+
{
|
693 |
+
"data": {
|
694 |
+
"application/vnd.jupyter.widget-view+json": {
|
695 |
+
"model_id": "67689f0c8fb842b2969c4fc584fa3a4b",
|
696 |
+
"version_major": 2,
|
697 |
+
"version_minor": 0
|
698 |
+
},
|
699 |
+
"text/plain": [
|
700 |
+
" 0%| | 0/58 [00:00<?, ?ba/s]"
|
701 |
+
]
|
702 |
+
},
|
703 |
+
"metadata": {},
|
704 |
+
"output_type": "display_data"
|
705 |
+
}
|
706 |
+
],
|
707 |
+
"source": [
|
708 |
+
"dataset = dataset.map(preprocess_function, batched=True)"
|
709 |
+
]
|
710 |
+
},
|
711 |
+
{
|
712 |
+
"cell_type": "code",
|
713 |
+
"execution_count": 14,
|
714 |
+
"metadata": {},
|
715 |
+
"outputs": [],
|
716 |
+
"source": [
|
717 |
+
"dataset_splitted = dataset.shuffle(1337).train_test_split(0.1)"
|
718 |
+
]
|
719 |
+
},
|
720 |
+
{
|
721 |
+
"cell_type": "code",
|
722 |
+
"execution_count": 15,
|
723 |
+
"metadata": {},
|
724 |
+
"outputs": [
|
725 |
+
{
|
726 |
+
"data": {
|
727 |
+
"text/plain": [
|
728 |
+
"DatasetDict({\n",
|
729 |
+
" train: Dataset({\n",
|
730 |
+
" features: ['text', 'labels', 'input_ids', 'attention_mask'],\n",
|
731 |
+
" num_rows: 51492\n",
|
732 |
+
" })\n",
|
733 |
+
" test: Dataset({\n",
|
734 |
+
" features: ['text', 'labels', 'input_ids', 'attention_mask'],\n",
|
735 |
+
" num_rows: 5722\n",
|
736 |
+
" })\n",
|
737 |
+
"})"
|
738 |
+
]
|
739 |
+
},
|
740 |
+
"execution_count": 15,
|
741 |
+
"metadata": {},
|
742 |
+
"output_type": "execute_result"
|
743 |
+
}
|
744 |
+
],
|
745 |
+
"source": [
|
746 |
+
"dataset_splitted"
|
747 |
+
]
|
748 |
+
},
|
749 |
+
{
|
750 |
+
"cell_type": "code",
|
751 |
+
"execution_count": 16,
|
752 |
+
"metadata": {
|
753 |
+
"id": "NyHknkwcYi6L"
|
754 |
+
},
|
755 |
+
"outputs": [],
|
756 |
+
"source": [
|
757 |
+
"from transformers import AutoModelForSequenceClassification"
|
758 |
+
]
|
759 |
+
},
|
760 |
+
{
|
761 |
+
"cell_type": "code",
|
762 |
+
"execution_count": 23,
|
763 |
+
"metadata": {
|
764 |
+
"colab": {
|
765 |
+
"base_uri": "https://localhost:8080/"
|
766 |
+
},
|
767 |
+
"id": "gv_fYzmEYlUm",
|
768 |
+
"outputId": "7a97df03-8f7b-4d54-f8d7-6a6b71d4c8c4"
|
769 |
+
},
|
770 |
+
"outputs": [
|
771 |
+
{
|
772 |
+
"name": "stderr",
|
773 |
+
"output_type": "stream",
|
774 |
+
"text": [
|
775 |
+
"loading configuration file https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english/resolve/main/config.json from cache at C:\\Users\\andry/.cache\\huggingface\\transformers\\4e60bb8efad3d4b7dc9969bf204947c185166a0a3cf37ddb6f481a876a3777b5.9f8326d0b7697c7fd57366cdde57032f46bc10e37ae81cb7eb564d66d23ec96b\n",
|
776 |
+
"Model config DistilBertConfig {\n",
|
777 |
+
" \"_name_or_path\": \"distilbert-base-uncased-finetuned-sst-2-english\",\n",
|
778 |
+
" \"activation\": \"gelu\",\n",
|
779 |
+
" \"architectures\": [\n",
|
780 |
+
" \"DistilBertForSequenceClassification\"\n",
|
781 |
+
" ],\n",
|
782 |
+
" \"attention_dropout\": 0.1,\n",
|
783 |
+
" \"dim\": 768,\n",
|
784 |
+
" \"dropout\": 0.1,\n",
|
785 |
+
" \"finetuning_task\": \"sst-2\",\n",
|
786 |
+
" \"hidden_dim\": 3072,\n",
|
787 |
+
" \"id2label\": {\n",
|
788 |
+
" \"0\": \"NEGATIVE\",\n",
|
789 |
+
" \"1\": \"POSITIVE\"\n",
|
790 |
+
" },\n",
|
791 |
+
" \"initializer_range\": 0.02,\n",
|
792 |
+
" \"label2id\": {\n",
|
793 |
+
" \"NEGATIVE\": 0,\n",
|
794 |
+
" \"POSITIVE\": 1\n",
|
795 |
+
" },\n",
|
796 |
+
" \"max_position_embeddings\": 512,\n",
|
797 |
+
" \"model_type\": \"distilbert\",\n",
|
798 |
+
" \"n_heads\": 12,\n",
|
799 |
+
" \"n_layers\": 6,\n",
|
800 |
+
" \"output_past\": true,\n",
|
801 |
+
" \"pad_token_id\": 0,\n",
|
802 |
+
" \"qa_dropout\": 0.1,\n",
|
803 |
+
" \"seq_classif_dropout\": 0.2,\n",
|
804 |
+
" \"sinusoidal_pos_embds\": false,\n",
|
805 |
+
" \"tie_weights_\": true,\n",
|
806 |
+
" \"transformers_version\": \"4.17.0\",\n",
|
807 |
+
" \"vocab_size\": 30522\n",
|
808 |
+
"}\n",
|
809 |
+
"\n",
|
810 |
+
"loading weights file https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english/resolve/main/pytorch_model.bin from cache at C:\\Users\\andry/.cache\\huggingface\\transformers\\8d04c767d9d4c14d929ce7ad8e067b80c74dbdb212ef4c3fb743db4ee109fae0.9d268a35da669ead745c44d369dc9948b408da5010c6bac414414a7e33d5748c\n",
|
811 |
+
"All model checkpoint weights were used when initializing DistilBertForSequenceClassification.\n",
|
812 |
+
"\n",
|
813 |
+
"All the weights of DistilBertForSequenceClassification were initialized from the model checkpoint at distilbert-base-uncased-finetuned-sst-2-english.\n",
|
814 |
+
"If your task is similar to the task the model of the checkpoint was trained on, you can already use DistilBertForSequenceClassification for predictions without further training.\n"
|
815 |
+
]
|
816 |
+
}
|
817 |
+
],
|
818 |
+
"source": [
|
819 |
+
"model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=2)"
|
820 |
+
]
|
821 |
+
},
|
822 |
+
{
|
823 |
+
"cell_type": "code",
|
824 |
+
"execution_count": 24,
|
825 |
+
"metadata": {
|
826 |
+
"id": "YqcdtMXZelbm"
|
827 |
+
},
|
828 |
+
"outputs": [],
|
829 |
+
"source": [
|
830 |
+
"for name, param in model.named_parameters():\n",
|
831 |
+
" if name in ['classifier.weight', 'classifier.bias']:\n",
|
832 |
+
" param.requires_grad = True\n",
|
833 |
+
" else:\n",
|
834 |
+
" param.requires_grad = False"
|
835 |
+
]
|
836 |
+
},
|
837 |
+
{
|
838 |
+
"cell_type": "code",
|
839 |
+
"execution_count": 25,
|
840 |
+
"metadata": {},
|
841 |
+
"outputs": [],
|
842 |
+
"source": [
|
843 |
+
"from sklearn.metrics import accuracy_score\n",
|
844 |
+
"\n",
|
845 |
+
"def compute_metrics(pred):\n",
|
846 |
+
" labels = pred.label_ids\n",
|
847 |
+
" preds = pred.predictions.argmax(-1)\n",
|
848 |
+
" acc = accuracy_score(labels, preds)\n",
|
849 |
+
" return {'accuracy': acc}"
|
850 |
+
]
|
851 |
+
},
|
852 |
+
{
|
853 |
+
"cell_type": "code",
|
854 |
+
"execution_count": 26,
|
855 |
+
"metadata": {
|
856 |
+
"colab": {
|
857 |
+
"base_uri": "https://localhost:8080/",
|
858 |
+
"height": 608
|
859 |
+
},
|
860 |
+
"id": "DkBWiEiyIgnV",
|
861 |
+
"outputId": "07f58180-8005-4f7e-fd72-62a5d2c78717",
|
862 |
+
"scrolled": false
|
863 |
+
},
|
864 |
+
"outputs": [
|
865 |
+
{
|
866 |
+
"name": "stderr",
|
867 |
+
"output_type": "stream",
|
868 |
+
"text": [
|
869 |
+
"PyTorch: setting up devices\n",
|
870 |
+
"The default value for the training argument `--report_to` will change in v5 (from all installed integrations to none). In v5, you will need to use `--report_to all` to get the same behavior as now. You should start updating your code and make this info disappear :-).\n",
|
871 |
+
"The following columns in the training set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
872 |
+
"***** Running training *****\n",
|
873 |
+
" Num examples = 51492\n",
|
874 |
+
" Num Epochs = 10\n",
|
875 |
+
" Instantaneous batch size per device = 64\n",
|
876 |
+
" Total train batch size (w. parallel, distributed & accumulation) = 64\n",
|
877 |
+
" Gradient Accumulation steps = 1\n",
|
878 |
+
" Total optimization steps = 8050\n"
|
879 |
+
]
|
880 |
+
},
|
881 |
+
{
|
882 |
+
"data": {
|
883 |
+
"text/html": [
|
884 |
+
"\n",
|
885 |
+
" <div>\n",
|
886 |
+
" \n",
|
887 |
+
" <progress value='8050' max='8050' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
|
888 |
+
" [8050/8050 1:31:55, Epoch 10/10]\n",
|
889 |
+
" </div>\n",
|
890 |
+
" <table border=\"1\" class=\"dataframe\">\n",
|
891 |
+
" <thead>\n",
|
892 |
+
" <tr style=\"text-align: left;\">\n",
|
893 |
+
" <th>Epoch</th>\n",
|
894 |
+
" <th>Training Loss</th>\n",
|
895 |
+
" <th>Validation Loss</th>\n",
|
896 |
+
" <th>Accuracy</th>\n",
|
897 |
+
" </tr>\n",
|
898 |
+
" </thead>\n",
|
899 |
+
" <tbody>\n",
|
900 |
+
" <tr>\n",
|
901 |
+
" <td>1</td>\n",
|
902 |
+
" <td>1.124500</td>\n",
|
903 |
+
" <td>0.655170</td>\n",
|
904 |
+
" <td>0.631423</td>\n",
|
905 |
+
" </tr>\n",
|
906 |
+
" <tr>\n",
|
907 |
+
" <td>2</td>\n",
|
908 |
+
" <td>0.635900</td>\n",
|
909 |
+
" <td>0.616928</td>\n",
|
910 |
+
" <td>0.696435</td>\n",
|
911 |
+
" </tr>\n",
|
912 |
+
" <tr>\n",
|
913 |
+
" <td>3</td>\n",
|
914 |
+
" <td>0.617400</td>\n",
|
915 |
+
" <td>0.592879</td>\n",
|
916 |
+
" <td>0.727019</td>\n",
|
917 |
+
" </tr>\n",
|
918 |
+
" <tr>\n",
|
919 |
+
" <td>4</td>\n",
|
920 |
+
" <td>0.591200</td>\n",
|
921 |
+
" <td>0.577941</td>\n",
|
922 |
+
" <td>0.734533</td>\n",
|
923 |
+
" </tr>\n",
|
924 |
+
" <tr>\n",
|
925 |
+
" <td>5</td>\n",
|
926 |
+
" <td>0.577100</td>\n",
|
927 |
+
" <td>0.564665</td>\n",
|
928 |
+
" <td>0.747466</td>\n",
|
929 |
+
" </tr>\n",
|
930 |
+
" <tr>\n",
|
931 |
+
" <td>6</td>\n",
|
932 |
+
" <td>0.569300</td>\n",
|
933 |
+
" <td>0.556096</td>\n",
|
934 |
+
" <td>0.749913</td>\n",
|
935 |
+
" </tr>\n",
|
936 |
+
" <tr>\n",
|
937 |
+
" <td>7</td>\n",
|
938 |
+
" <td>0.563200</td>\n",
|
939 |
+
" <td>0.551389</td>\n",
|
940 |
+
" <td>0.755330</td>\n",
|
941 |
+
" </tr>\n",
|
942 |
+
" <tr>\n",
|
943 |
+
" <td>8</td>\n",
|
944 |
+
" <td>0.559900</td>\n",
|
945 |
+
" <td>0.546756</td>\n",
|
946 |
+
" <td>0.754981</td>\n",
|
947 |
+
" </tr>\n",
|
948 |
+
" <tr>\n",
|
949 |
+
" <td>9</td>\n",
|
950 |
+
" <td>0.554800</td>\n",
|
951 |
+
" <td>0.544496</td>\n",
|
952 |
+
" <td>0.759000</td>\n",
|
953 |
+
" </tr>\n",
|
954 |
+
" <tr>\n",
|
955 |
+
" <td>10</td>\n",
|
956 |
+
" <td>0.554000</td>\n",
|
957 |
+
" <td>0.543604</td>\n",
|
958 |
+
" <td>0.760398</td>\n",
|
959 |
+
" </tr>\n",
|
960 |
+
" </tbody>\n",
|
961 |
+
"</table><p>"
|
962 |
+
],
|
963 |
+
"text/plain": [
|
964 |
+
"<IPython.core.display.HTML object>"
|
965 |
+
]
|
966 |
+
},
|
967 |
+
"metadata": {},
|
968 |
+
"output_type": "display_data"
|
969 |
+
},
|
970 |
+
{
|
971 |
+
"name": "stderr",
|
972 |
+
"output_type": "stream",
|
973 |
+
"text": [
|
974 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
975 |
+
"***** Running Evaluation *****\n",
|
976 |
+
" Num examples = 5722\n",
|
977 |
+
" Batch size = 64\n",
|
978 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-805\n",
|
979 |
+
"Configuration saved in ./my_saved_model\\checkpoint-805\\config.json\n",
|
980 |
+
"Model weights saved in ./my_saved_model\\checkpoint-805\\pytorch_model.bin\n",
|
981 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
982 |
+
"***** Running Evaluation *****\n",
|
983 |
+
" Num examples = 5722\n",
|
984 |
+
" Batch size = 64\n",
|
985 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-1610\n",
|
986 |
+
"Configuration saved in ./my_saved_model\\checkpoint-1610\\config.json\n",
|
987 |
+
"Model weights saved in ./my_saved_model\\checkpoint-1610\\pytorch_model.bin\n",
|
988 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
989 |
+
"***** Running Evaluation *****\n",
|
990 |
+
" Num examples = 5722\n",
|
991 |
+
" Batch size = 64\n",
|
992 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-2415\n",
|
993 |
+
"Configuration saved in ./my_saved_model\\checkpoint-2415\\config.json\n",
|
994 |
+
"Model weights saved in ./my_saved_model\\checkpoint-2415\\pytorch_model.bin\n",
|
995 |
+
"Deleting older checkpoint [my_saved_model\\checkpoint-805] due to args.save_total_limit\n",
|
996 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
997 |
+
"***** Running Evaluation *****\n",
|
998 |
+
" Num examples = 5722\n",
|
999 |
+
" Batch size = 64\n",
|
1000 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-3220\n",
|
1001 |
+
"Configuration saved in ./my_saved_model\\checkpoint-3220\\config.json\n",
|
1002 |
+
"Model weights saved in ./my_saved_model\\checkpoint-3220\\pytorch_model.bin\n",
|
1003 |
+
"Deleting older checkpoint [my_saved_model\\checkpoint-1610] due to args.save_total_limit\n",
|
1004 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
1005 |
+
"***** Running Evaluation *****\n",
|
1006 |
+
" Num examples = 5722\n",
|
1007 |
+
" Batch size = 64\n",
|
1008 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-4025\n",
|
1009 |
+
"Configuration saved in ./my_saved_model\\checkpoint-4025\\config.json\n",
|
1010 |
+
"Model weights saved in ./my_saved_model\\checkpoint-4025\\pytorch_model.bin\n",
|
1011 |
+
"Deleting older checkpoint [my_saved_model\\checkpoint-2415] due to args.save_total_limit\n",
|
1012 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
1013 |
+
"***** Running Evaluation *****\n",
|
1014 |
+
" Num examples = 5722\n",
|
1015 |
+
" Batch size = 64\n",
|
1016 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-4830\n",
|
1017 |
+
"Configuration saved in ./my_saved_model\\checkpoint-4830\\config.json\n",
|
1018 |
+
"Model weights saved in ./my_saved_model\\checkpoint-4830\\pytorch_model.bin\n",
|
1019 |
+
"Deleting older checkpoint [my_saved_model\\checkpoint-3220] due to args.save_total_limit\n",
|
1020 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
1021 |
+
"***** Running Evaluation *****\n",
|
1022 |
+
" Num examples = 5722\n",
|
1023 |
+
" Batch size = 64\n",
|
1024 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-5635\n",
|
1025 |
+
"Configuration saved in ./my_saved_model\\checkpoint-5635\\config.json\n",
|
1026 |
+
"Model weights saved in ./my_saved_model\\checkpoint-5635\\pytorch_model.bin\n",
|
1027 |
+
"Deleting older checkpoint [my_saved_model\\checkpoint-4025] due to args.save_total_limit\n",
|
1028 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
1029 |
+
"***** Running Evaluation *****\n",
|
1030 |
+
" Num examples = 5722\n",
|
1031 |
+
" Batch size = 64\n",
|
1032 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-6440\n",
|
1033 |
+
"Configuration saved in ./my_saved_model\\checkpoint-6440\\config.json\n",
|
1034 |
+
"Model weights saved in ./my_saved_model\\checkpoint-6440\\pytorch_model.bin\n",
|
1035 |
+
"Deleting older checkpoint [my_saved_model\\checkpoint-4830] due to args.save_total_limit\n",
|
1036 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
1037 |
+
"***** Running Evaluation *****\n",
|
1038 |
+
" Num examples = 5722\n",
|
1039 |
+
" Batch size = 64\n",
|
1040 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-7245\n",
|
1041 |
+
"Configuration saved in ./my_saved_model\\checkpoint-7245\\config.json\n",
|
1042 |
+
"Model weights saved in ./my_saved_model\\checkpoint-7245\\pytorch_model.bin\n",
|
1043 |
+
"Deleting older checkpoint [my_saved_model\\checkpoint-5635] due to args.save_total_limit\n",
|
1044 |
+
"The following columns in the evaluation set don't have a corresponding argument in `DistilBertForSequenceClassification.forward` and have been ignored: text. If text are not expected by `DistilBertForSequenceClassification.forward`, you can safely ignore this message.\n",
|
1045 |
+
"***** Running Evaluation *****\n",
|
1046 |
+
" Num examples = 5722\n",
|
1047 |
+
" Batch size = 64\n",
|
1048 |
+
"Saving model checkpoint to ./my_saved_model\\checkpoint-8050\n",
|
1049 |
+
"Configuration saved in ./my_saved_model\\checkpoint-8050\\config.json\n",
|
1050 |
+
"Model weights saved in ./my_saved_model\\checkpoint-8050\\pytorch_model.bin\n",
|
1051 |
+
"Deleting older checkpoint [my_saved_model\\checkpoint-6440] due to args.save_total_limit\n",
|
1052 |
+
"\n",
|
1053 |
+
"\n",
|
1054 |
+
"Training completed. Do not forget to share your model on huggingface.co/models =)\n",
|
1055 |
+
"\n",
|
1056 |
+
"\n",
|
1057 |
+
"Loading best model from ./my_saved_model\\checkpoint-8050 (score: 0.543603777885437).\n"
|
1058 |
+
]
|
1059 |
+
},
|
1060 |
+
{
|
1061 |
+
"data": {
|
1062 |
+
"text/plain": [
|
1063 |
+
"TrainOutput(global_step=8050, training_loss=0.6166538418598057, metrics={'train_runtime': 5516.6092, 'train_samples_per_second': 93.34, 'train_steps_per_second': 1.459, 'total_flos': 6.821011291594752e+16, 'train_loss': 0.6166538418598057, 'epoch': 10.0})"
|
1064 |
+
]
|
1065 |
+
},
|
1066 |
+
"execution_count": 26,
|
1067 |
+
"metadata": {},
|
1068 |
+
"output_type": "execute_result"
|
1069 |
+
}
|
1070 |
+
],
|
1071 |
+
"source": [
|
1072 |
+
"from transformers import Trainer, TrainingArguments\n",
|
1073 |
+
"\n",
|
1074 |
+
"trainer = Trainer(\n",
|
1075 |
+
" model=model, train_dataset=dataset_splitted['train'], \n",
|
1076 |
+
" eval_dataset=dataset_splitted['test'],\n",
|
1077 |
+
" compute_metrics=compute_metrics,\n",
|
1078 |
+
" args=TrainingArguments(\n",
|
1079 |
+
" load_best_model_at_end=True,\n",
|
1080 |
+
" output_dir=\"./my_saved_model\", overwrite_output_dir=True,\n",
|
1081 |
+
" num_train_epochs=10, per_device_train_batch_size=64, \n",
|
1082 |
+
" per_device_eval_batch_size=64,\n",
|
1083 |
+
" evaluation_strategy = \"epoch\",\n",
|
1084 |
+
" save_strategy = \"epoch\",\n",
|
1085 |
+
" save_steps=10_000, save_total_limit=2),\n",
|
1086 |
+
")\n",
|
1087 |
+
"\n",
|
1088 |
+
"trainer.train()"
|
1089 |
+
]
|
1090 |
+
}
|
1091 |
+
],
|
1092 |
+
"metadata": {
|
1093 |
+
"accelerator": "GPU",
|
1094 |
+
"colab": {
|
1095 |
+
"collapsed_sections": [],
|
1096 |
+
"name": "Копия блокнота \"ysda_2022.03.07.ipynb\"",
|
1097 |
+
"provenance": []
|
1098 |
+
},
|
1099 |
+
"kernelspec": {
|
1100 |
+
"display_name": "Python 3 (ipykernel)",
|
1101 |
+
"language": "python",
|
1102 |
+
"name": "python3"
|
1103 |
+
},
|
1104 |
+
"language_info": {
|
1105 |
+
"codemirror_mode": {
|
1106 |
+
"name": "ipython",
|
1107 |
+
"version": 3
|
1108 |
+
},
|
1109 |
+
"file_extension": ".py",
|
1110 |
+
"mimetype": "text/x-python",
|
1111 |
+
"name": "python",
|
1112 |
+
"nbconvert_exporter": "python",
|
1113 |
+
"pygments_lexer": "ipython3",
|
1114 |
+
"version": "3.7.12"
|
1115 |
+
},
|
1116 |
+
"widgets": {
|
1117 |
+
"application/vnd.jupyter.widget-state+json": {
|
1118 |
+
"300b70ed57dd493997afb0b3f25f4245": {
|
1119 |
+
"model_module": "@jupyter-widgets/controls",
|
1120 |
+
"model_module_version": "1.5.0",
|
1121 |
+
"model_name": "HTMLModel",
|
1122 |
+
"state": {
|
1123 |
+
"_dom_classes": [],
|
1124 |
+
"_model_module": "@jupyter-widgets/controls",
|
1125 |
+
"_model_module_version": "1.5.0",
|
1126 |
+
"_model_name": "HTMLModel",
|
1127 |
+
"_view_count": null,
|
1128 |
+
"_view_module": "@jupyter-widgets/controls",
|
1129 |
+
"_view_module_version": "1.5.0",
|
1130 |
+
"_view_name": "HTMLView",
|
1131 |
+
"description": "",
|
1132 |
+
"description_tooltip": null,
|
1133 |
+
"layout": "IPY_MODEL_9e11898bc51e483d91301387099368a4",
|
1134 |
+
"placeholder": "",
|
1135 |
+
"style": "IPY_MODEL_a43574fa5fdf47ba9d5598b2b31f2082",
|
1136 |
+
"value": "100%"
|
1137 |
+
}
|
1138 |
+
},
|
1139 |
+
"482bae742d2a461cad525888e6ee8b91": {
|
1140 |
+
"model_module": "@jupyter-widgets/base",
|
1141 |
+
"model_module_version": "1.2.0",
|
1142 |
+
"model_name": "LayoutModel",
|
1143 |
+
"state": {
|
1144 |
+
"_model_module": "@jupyter-widgets/base",
|
1145 |
+
"_model_module_version": "1.2.0",
|
1146 |
+
"_model_name": "LayoutModel",
|
1147 |
+
"_view_count": null,
|
1148 |
+
"_view_module": "@jupyter-widgets/base",
|
1149 |
+
"_view_module_version": "1.2.0",
|
1150 |
+
"_view_name": "LayoutView",
|
1151 |
+
"align_content": null,
|
1152 |
+
"align_items": null,
|
1153 |
+
"align_self": null,
|
1154 |
+
"border": null,
|
1155 |
+
"bottom": null,
|
1156 |
+
"display": null,
|
1157 |
+
"flex": null,
|
1158 |
+
"flex_flow": null,
|
1159 |
+
"grid_area": null,
|
1160 |
+
"grid_auto_columns": null,
|
1161 |
+
"grid_auto_flow": null,
|
1162 |
+
"grid_auto_rows": null,
|
1163 |
+
"grid_column": null,
|
1164 |
+
"grid_gap": null,
|
1165 |
+
"grid_row": null,
|
1166 |
+
"grid_template_areas": null,
|
1167 |
+
"grid_template_columns": null,
|
1168 |
+
"grid_template_rows": null,
|
1169 |
+
"height": null,
|
1170 |
+
"justify_content": null,
|
1171 |
+
"justify_items": null,
|
1172 |
+
"left": null,
|
1173 |
+
"margin": null,
|
1174 |
+
"max_height": null,
|
1175 |
+
"max_width": null,
|
1176 |
+
"min_height": null,
|
1177 |
+
"min_width": null,
|
1178 |
+
"object_fit": null,
|
1179 |
+
"object_position": null,
|
1180 |
+
"order": null,
|
1181 |
+
"overflow": null,
|
1182 |
+
"overflow_x": null,
|
1183 |
+
"overflow_y": null,
|
1184 |
+
"padding": null,
|
1185 |
+
"right": null,
|
1186 |
+
"top": null,
|
1187 |
+
"visibility": null,
|
1188 |
+
"width": null
|
1189 |
+
}
|
1190 |
+
},
|
1191 |
+
"57c3794731c84c42bb49618482b6b8cc": {
|
1192 |
+
"model_module": "@jupyter-widgets/controls",
|
1193 |
+
"model_module_version": "1.5.0",
|
1194 |
+
"model_name": "HTMLModel",
|
1195 |
+
"state": {
|
1196 |
+
"_dom_classes": [],
|
1197 |
+
"_model_module": "@jupyter-widgets/controls",
|
1198 |
+
"_model_module_version": "1.5.0",
|
1199 |
+
"_model_name": "HTMLModel",
|
1200 |
+
"_view_count": null,
|
1201 |
+
"_view_module": "@jupyter-widgets/controls",
|
1202 |
+
"_view_module_version": "1.5.0",
|
1203 |
+
"_view_name": "HTMLView",
|
1204 |
+
"description": "",
|
1205 |
+
"description_tooltip": null,
|
1206 |
+
"layout": "IPY_MODEL_d604380b5e444f62ad36c4598230c561",
|
1207 |
+
"placeholder": "",
|
1208 |
+
"style": "IPY_MODEL_c52ad745acb3423494b4ea5af5a934c7",
|
1209 |
+
"value": " 58/58 [02:02<00:00, 1.83s/ba]"
|
1210 |
+
}
|
1211 |
+
},
|
1212 |
+
"5b49dc833234406da3da7435b9045fd2": {
|
1213 |
+
"model_module": "@jupyter-widgets/controls",
|
1214 |
+
"model_module_version": "1.5.0",
|
1215 |
+
"model_name": "HBoxModel",
|
1216 |
+
"state": {
|
1217 |
+
"_dom_classes": [],
|
1218 |
+
"_model_module": "@jupyter-widgets/controls",
|
1219 |
+
"_model_module_version": "1.5.0",
|
1220 |
+
"_model_name": "HBoxModel",
|
1221 |
+
"_view_count": null,
|
1222 |
+
"_view_module": "@jupyter-widgets/controls",
|
1223 |
+
"_view_module_version": "1.5.0",
|
1224 |
+
"_view_name": "HBoxView",
|
1225 |
+
"box_style": "",
|
1226 |
+
"children": [
|
1227 |
+
"IPY_MODEL_300b70ed57dd493997afb0b3f25f4245",
|
1228 |
+
"IPY_MODEL_c03cc68b079c4e23b339e9de5ba38d29",
|
1229 |
+
"IPY_MODEL_57c3794731c84c42bb49618482b6b8cc"
|
1230 |
+
],
|
1231 |
+
"layout": "IPY_MODEL_e306828f6d7444ddafce604e9a170467"
|
1232 |
+
}
|
1233 |
+
},
|
1234 |
+
"9e11898bc51e483d91301387099368a4": {
|
1235 |
+
"model_module": "@jupyter-widgets/base",
|
1236 |
+
"model_module_version": "1.2.0",
|
1237 |
+
"model_name": "LayoutModel",
|
1238 |
+
"state": {
|
1239 |
+
"_model_module": "@jupyter-widgets/base",
|
1240 |
+
"_model_module_version": "1.2.0",
|
1241 |
+
"_model_name": "LayoutModel",
|
1242 |
+
"_view_count": null,
|
1243 |
+
"_view_module": "@jupyter-widgets/base",
|
1244 |
+
"_view_module_version": "1.2.0",
|
1245 |
+
"_view_name": "LayoutView",
|
1246 |
+
"align_content": null,
|
1247 |
+
"align_items": null,
|
1248 |
+
"align_self": null,
|
1249 |
+
"border": null,
|
1250 |
+
"bottom": null,
|
1251 |
+
"display": null,
|
1252 |
+
"flex": null,
|
1253 |
+
"flex_flow": null,
|
1254 |
+
"grid_area": null,
|
1255 |
+
"grid_auto_columns": null,
|
1256 |
+
"grid_auto_flow": null,
|
1257 |
+
"grid_auto_rows": null,
|
1258 |
+
"grid_column": null,
|
1259 |
+
"grid_gap": null,
|
1260 |
+
"grid_row": null,
|
1261 |
+
"grid_template_areas": null,
|
1262 |
+
"grid_template_columns": null,
|
1263 |
+
"grid_template_rows": null,
|
1264 |
+
"height": null,
|
1265 |
+
"justify_content": null,
|
1266 |
+
"justify_items": null,
|
1267 |
+
"left": null,
|
1268 |
+
"margin": null,
|
1269 |
+
"max_height": null,
|
1270 |
+
"max_width": null,
|
1271 |
+
"min_height": null,
|
1272 |
+
"min_width": null,
|
1273 |
+
"object_fit": null,
|
1274 |
+
"object_position": null,
|
1275 |
+
"order": null,
|
1276 |
+
"overflow": null,
|
1277 |
+
"overflow_x": null,
|
1278 |
+
"overflow_y": null,
|
1279 |
+
"padding": null,
|
1280 |
+
"right": null,
|
1281 |
+
"top": null,
|
1282 |
+
"visibility": null,
|
1283 |
+
"width": null
|
1284 |
+
}
|
1285 |
+
},
|
1286 |
+
"a43574fa5fdf47ba9d5598b2b31f2082": {
|
1287 |
+
"model_module": "@jupyter-widgets/controls",
|
1288 |
+
"model_module_version": "1.5.0",
|
1289 |
+
"model_name": "DescriptionStyleModel",
|
1290 |
+
"state": {
|
1291 |
+
"_model_module": "@jupyter-widgets/controls",
|
1292 |
+
"_model_module_version": "1.5.0",
|
1293 |
+
"_model_name": "DescriptionStyleModel",
|
1294 |
+
"_view_count": null,
|
1295 |
+
"_view_module": "@jupyter-widgets/base",
|
1296 |
+
"_view_module_version": "1.2.0",
|
1297 |
+
"_view_name": "StyleView",
|
1298 |
+
"description_width": ""
|
1299 |
+
}
|
1300 |
+
},
|
1301 |
+
"c03cc68b079c4e23b339e9de5ba38d29": {
|
1302 |
+
"model_module": "@jupyter-widgets/controls",
|
1303 |
+
"model_module_version": "1.5.0",
|
1304 |
+
"model_name": "FloatProgressModel",
|
1305 |
+
"state": {
|
1306 |
+
"_dom_classes": [],
|
1307 |
+
"_model_module": "@jupyter-widgets/controls",
|
1308 |
+
"_model_module_version": "1.5.0",
|
1309 |
+
"_model_name": "FloatProgressModel",
|
1310 |
+
"_view_count": null,
|
1311 |
+
"_view_module": "@jupyter-widgets/controls",
|
1312 |
+
"_view_module_version": "1.5.0",
|
1313 |
+
"_view_name": "ProgressView",
|
1314 |
+
"bar_style": "success",
|
1315 |
+
"description": "",
|
1316 |
+
"description_tooltip": null,
|
1317 |
+
"layout": "IPY_MODEL_482bae742d2a461cad525888e6ee8b91",
|
1318 |
+
"max": 58,
|
1319 |
+
"min": 0,
|
1320 |
+
"orientation": "horizontal",
|
1321 |
+
"style": "IPY_MODEL_e9c56275d73545a6961efe5704308ede",
|
1322 |
+
"value": 58
|
1323 |
+
}
|
1324 |
+
},
|
1325 |
+
"c52ad745acb3423494b4ea5af5a934c7": {
|
1326 |
+
"model_module": "@jupyter-widgets/controls",
|
1327 |
+
"model_module_version": "1.5.0",
|
1328 |
+
"model_name": "DescriptionStyleModel",
|
1329 |
+
"state": {
|
1330 |
+
"_model_module": "@jupyter-widgets/controls",
|
1331 |
+
"_model_module_version": "1.5.0",
|
1332 |
+
"_model_name": "DescriptionStyleModel",
|
1333 |
+
"_view_count": null,
|
1334 |
+
"_view_module": "@jupyter-widgets/base",
|
1335 |
+
"_view_module_version": "1.2.0",
|
1336 |
+
"_view_name": "StyleView",
|
1337 |
+
"description_width": ""
|
1338 |
+
}
|
1339 |
+
},
|
1340 |
+
"d604380b5e444f62ad36c4598230c561": {
|
1341 |
+
"model_module": "@jupyter-widgets/base",
|
1342 |
+
"model_module_version": "1.2.0",
|
1343 |
+
"model_name": "LayoutModel",
|
1344 |
+
"state": {
|
1345 |
+
"_model_module": "@jupyter-widgets/base",
|
1346 |
+
"_model_module_version": "1.2.0",
|
1347 |
+
"_model_name": "LayoutModel",
|
1348 |
+
"_view_count": null,
|
1349 |
+
"_view_module": "@jupyter-widgets/base",
|
1350 |
+
"_view_module_version": "1.2.0",
|
1351 |
+
"_view_name": "LayoutView",
|
1352 |
+
"align_content": null,
|
1353 |
+
"align_items": null,
|
1354 |
+
"align_self": null,
|
1355 |
+
"border": null,
|
1356 |
+
"bottom": null,
|
1357 |
+
"display": null,
|
1358 |
+
"flex": null,
|
1359 |
+
"flex_flow": null,
|
1360 |
+
"grid_area": null,
|
1361 |
+
"grid_auto_columns": null,
|
1362 |
+
"grid_auto_flow": null,
|
1363 |
+
"grid_auto_rows": null,
|
1364 |
+
"grid_column": null,
|
1365 |
+
"grid_gap": null,
|
1366 |
+
"grid_row": null,
|
1367 |
+
"grid_template_areas": null,
|
1368 |
+
"grid_template_columns": null,
|
1369 |
+
"grid_template_rows": null,
|
1370 |
+
"height": null,
|
1371 |
+
"justify_content": null,
|
1372 |
+
"justify_items": null,
|
1373 |
+
"left": null,
|
1374 |
+
"margin": null,
|
1375 |
+
"max_height": null,
|
1376 |
+
"max_width": null,
|
1377 |
+
"min_height": null,
|
1378 |
+
"min_width": null,
|
1379 |
+
"object_fit": null,
|
1380 |
+
"object_position": null,
|
1381 |
+
"order": null,
|
1382 |
+
"overflow": null,
|
1383 |
+
"overflow_x": null,
|
1384 |
+
"overflow_y": null,
|
1385 |
+
"padding": null,
|
1386 |
+
"right": null,
|
1387 |
+
"top": null,
|
1388 |
+
"visibility": null,
|
1389 |
+
"width": null
|
1390 |
+
}
|
1391 |
+
},
|
1392 |
+
"e306828f6d7444ddafce604e9a170467": {
|
1393 |
+
"model_module": "@jupyter-widgets/base",
|
1394 |
+
"model_module_version": "1.2.0",
|
1395 |
+
"model_name": "LayoutModel",
|
1396 |
+
"state": {
|
1397 |
+
"_model_module": "@jupyter-widgets/base",
|
1398 |
+
"_model_module_version": "1.2.0",
|
1399 |
+
"_model_name": "LayoutModel",
|
1400 |
+
"_view_count": null,
|
1401 |
+
"_view_module": "@jupyter-widgets/base",
|
1402 |
+
"_view_module_version": "1.2.0",
|
1403 |
+
"_view_name": "LayoutView",
|
1404 |
+
"align_content": null,
|
1405 |
+
"align_items": null,
|
1406 |
+
"align_self": null,
|
1407 |
+
"border": null,
|
1408 |
+
"bottom": null,
|
1409 |
+
"display": null,
|
1410 |
+
"flex": null,
|
1411 |
+
"flex_flow": null,
|
1412 |
+
"grid_area": null,
|
1413 |
+
"grid_auto_columns": null,
|
1414 |
+
"grid_auto_flow": null,
|
1415 |
+
"grid_auto_rows": null,
|
1416 |
+
"grid_column": null,
|
1417 |
+
"grid_gap": null,
|
1418 |
+
"grid_row": null,
|
1419 |
+
"grid_template_areas": null,
|
1420 |
+
"grid_template_columns": null,
|
1421 |
+
"grid_template_rows": null,
|
1422 |
+
"height": null,
|
1423 |
+
"justify_content": null,
|
1424 |
+
"justify_items": null,
|
1425 |
+
"left": null,
|
1426 |
+
"margin": null,
|
1427 |
+
"max_height": null,
|
1428 |
+
"max_width": null,
|
1429 |
+
"min_height": null,
|
1430 |
+
"min_width": null,
|
1431 |
+
"object_fit": null,
|
1432 |
+
"object_position": null,
|
1433 |
+
"order": null,
|
1434 |
+
"overflow": null,
|
1435 |
+
"overflow_x": null,
|
1436 |
+
"overflow_y": null,
|
1437 |
+
"padding": null,
|
1438 |
+
"right": null,
|
1439 |
+
"top": null,
|
1440 |
+
"visibility": null,
|
1441 |
+
"width": null
|
1442 |
+
}
|
1443 |
+
},
|
1444 |
+
"e9c56275d73545a6961efe5704308ede": {
|
1445 |
+
"model_module": "@jupyter-widgets/controls",
|
1446 |
+
"model_module_version": "1.5.0",
|
1447 |
+
"model_name": "ProgressStyleModel",
|
1448 |
+
"state": {
|
1449 |
+
"_model_module": "@jupyter-widgets/controls",
|
1450 |
+
"_model_module_version": "1.5.0",
|
1451 |
+
"_model_name": "ProgressStyleModel",
|
1452 |
+
"_view_count": null,
|
1453 |
+
"_view_module": "@jupyter-widgets/base",
|
1454 |
+
"_view_module_version": "1.2.0",
|
1455 |
+
"_view_name": "StyleView",
|
1456 |
+
"bar_color": null,
|
1457 |
+
"description_width": ""
|
1458 |
+
}
|
1459 |
+
}
|
1460 |
+
}
|
1461 |
+
}
|
1462 |
+
},
|
1463 |
+
"nbformat": 4,
|
1464 |
+
"nbformat_minor": 1
|
1465 |
+
}
|