-
Notifications
You must be signed in to change notification settings - Fork 1
/
dataset.py
38 lines (27 loc) · 1.02 KB
/
dataset.py
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
from torch.utils.data import Dataset
from embedder import SentenceBERTModel
import pandas as pd
import numpy as np
import config as cfg
class IntentDataset(Dataset):
def __init__(self, csv_file=cfg.DATA):
# One-Hot encoding labels
LABELS = cfg.ONE_HOT_LABELS
SentenceModel = SentenceBERTModel()
data = pd.read_csv(csv_file, delimiter=';')
target = data['intent']
data = data.drop(['intent'], axis=1)
lst = []
# Data preprocess
for i in range(len(data)):
predict = SentenceModel(data.loc[i]['text']).detach().numpy()
lst.append(predict[0])
self.train_data = np.array(lst)
# One-Hot Encoding
for i in range(len(target)):
target.loc[i] = LABELS[target.loc[i]]
self.encoding_target = np.array(pd.get_dummies(target))
def __len__(self):
return self.encoding_target.shape[0]
def __getitem__(self, index):
return self.train_data[index], self.encoding_target[index]