File size: 4,647 Bytes
d31451e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e55d629
d31451e
 
 
 
 
 
 
 
 
 
e55d629
 
d31451e
e55d629
d31451e
 
 
 
 
 
e55d629
d31451e
 
 
 
 
 
 
 
e55d629
d31451e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ef7bc7f
d31451e
 
 
 
24e04ee
d31451e
 
 
 
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
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
import torch
import torch.nn as TNN
import pandas as pd
from tqdm import tqdm
from torch.utils.data import Dataset as set, DataLoader as DL
from torch import cuda
import streamlit as st
from transformers import BertTokenizer as BT, BertModel as BM

device = 'cuda' if cuda.is_available() else 'cpu'

# Defined variables for later use
MAX_LEN = 128
TRAIN_BATCH_SIZE = 4
VALID_BATCH_SIZE = 4
LEARNING_RATE = 5e-05

modName = 'bert-base-uncased' # Pre-trained model
categories = ['toxic', 'severe_toxic', 'obscene', 'threat', 'insult', 'identity_hate'] # Labels

data = pd.read_csv('./train.csv')
data.drop(['id'], inplace=True, axis=1)

new = pd.DataFrame()
new['text'] = data['comment_text']
new['labels'] = data.iloc[:,1].values.tolist()

tokenizer = BT.from_pretrained(modName, truncation=True, do_lower_case=True)

class MultiLabelDataset(set):
    def __init__(self, df, tokenizer, max_len):
        self.tokenizer = tokenizer
        self.data = df
        self.text = df.text
        self.targets = self.data.labels
        self.max_len = max_len

    def __len__(self):
        return len(self.targets)

    def __getitem__(self, idx):
        text = str(self.text[idx])
        text = " ".join(text.split())

        ins = self.tokenizer.encode_plus(
            text,
            None,
            add_special_tokens=True,
            max_length=self.max_len,
            pad_to_max_length=True,
            return_token_type_ids=True
        )
        input_ids = ins['input_ids']
        attention_mask = ins['attention_mask']
        token_type_ids = ins["token_type_ids"]

        #st.write("Input Keys: ", ins.keys()) # was used for debugging
        return {
            'input_ids': torch.tensor(input_ids, dtype=torch.long),
            'attention_mask': torch.tensor(attention_mask, dtype=torch.long),
            'token_type_ids': torch.tensor(token_type_ids, dtype=torch.long),
            'targets': torch.tensor(self.targets[idx], dtype=torch.float)
        }

trainSize = 0.8
trainData = new.sample(frac=trainSize,random_state=200)
testData = new.drop(trainData.index).reset_index(drop=True)
trainData = trainData.reset_index(drop=True)

trainSet = MultiLabelDataset(trainData, tokenizer, MAX_LEN)
testSet = MultiLabelDataset(testData, tokenizer, MAX_LEN)

training_loader = DL(trainSet, batch_size=TRAIN_BATCH_SIZE, shuffle=True)
testing_loader = DL(testSet, batch_size=VALID_BATCH_SIZE, shuffle=True)

# neural network
class BERTClass(TNN.Module):
    def __init__(self):
        super(BERTClass, self).__init__()
        self.l1 = BM.from_pretrained(modName)
        self.pre_classifier = TNN.Linear(768, 768)
        self.dropout = TNN.Dropout(0.1)
        self.classifier = TNN.Linear(768, 6)

    def forward(self, input_ids, attention_mask, token_type_ids):
        out = self.l1(input_ids=input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids)
        hidden_state = out[0]
        po = hidden_state[:, 0]
        po = self.pre_classifier(po)
        po = TNN.Tanh()(po)
        po = self.dropout(po)
        outs = self.classifier(po)
        return outs

mod = BERTClass()
mod.to(device)

# Loss function and Optimizer
def lossFN(outs, targets):
    targets = targets.unsqueeze(1).expand_as(outs)
    return TNN.BCEWithLogitsLoss()(outs, targets)

opt = torch.optim.Adam(mod.parameters(), lr=LEARNING_RATE)

# Training and Finetuning
def train(mod, training_loader):
    mod.train()
    for _, data in tqdm(enumerate(training_loader, 0)):
        input_ids = data['input_ids'].to(device, dtype=torch.long)
        attention_mask = data['attention_mask'].to(device, dtype=torch.long)
        token_type_ids = data['token_type_ids'].to(device, dtype=torch.long)
        targets = data['targets'].to(device, dtype=torch.float)

        outs = mod(input_ids, attention_mask, token_type_ids)

        opt.zero_grad()
        loss = lossFN(outs, targets)
        loss.backward()
        opt.step()

# StreamLit Table of Results
st.title("Finetuned Model for Toxicity")
st.subheader("Model: bert-base-uncased")

def predict(tweets):
    mod.eval()
    res = []
    with torch.no_grad():
        for ins in tweets:
            outs = mod(input_ids=ins['input_ids'].to(device), attention_mask=ins['attention_mask'].to(device), token_type_ids=ins['token_type_ids'].to(device))
            probs = torch.softmax(outs[0], dim=-1)
            preds = torch.argmax(probs, dim=-1)
            for i in range(len(tweets)):
                res.append({'TWEETS': tweets, 'LABEL': preds[i].item(), 'PROBABILITY': probs[i][preds[i].item()].item()})
    return res

res = predict(testing_loader)
st.table(res) # table