-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
33 lines (25 loc) · 998 Bytes
/
train.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
import numpy as np
import tensorflow as tf
from tensorflow import keras
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
sns.set()
df = pd.read_csv('toxic arabic tweets classification.txt', '\t') # df columns must be [Tweet, Class]
classes = ['normal', 'abusive', 'hate']
from preprocess import Preprocessor
preprocessor = Preprocessor(df, classes)
preprocessor.map_labels()
df['Tweet'] = df['Tweet'].apply(preprocessor.normalize_text)
maxlen = 25
from model import ToxicModel
toxic_model = ToxicModel(df, maxlen)
train, val, test = toxic_model.load_dataset()
model = toxic_model.build_model()
if __name__ == '__main__':
opt = tf.keras.optimizers.Adam(1e-4)
loss = tf.keras.losses.CategoricalCrossentropy()
acc = tf.keras.metrics.CategoricalAccuracy()
model.compile(optimizer=opt, loss=loss, metrics=[acc])
history = model.fit(x=train, validation_data=val, epochs=50)
model.save('models/Bert-Multi-Dialect')