-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
35 lines (25 loc) · 1.07 KB
/
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
34
35
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
import pandas as pd
import numpy as np
from utils import read_csv,read_glove_vecs,sentences_to_indices,convert_to_one_hot
from buildModel import define_model, pretrained_embedding_layer
word_to_index, index_to_word, word_to_vec_map = read_glove_vecs('glove.6B.50d.txt')
X_train, Y_train = read_csv('train_emoji.csv')
X_test, Y_test = read_csv('test.csv')
maxLen = len(max(X_train, key=len).split())
model = define_model((maxLen,), word_to_vec_map, word_to_index)
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
X_train_indices = sentences_to_indices(X_train, word_to_index, maxLen)
Y_train_one_hot = convert_to_one_hot(Y_train, C = 5)
print("===" * 20)
print("MODEL TRAINING STARTED")
history = model.fit(X_train_indices, Y_train_one_hot, epochs = 50, batch_size=32, shuffle=True)
print("MODEL TRAINING FINISHED")
print("===" * 20)
print("SAVING THE MODEL")
model_filename = 'emojify.h5'
model.save(model_filename)
print("===" * 20)
print("MODEL SAVED AS: ", model_filename)
print("===" * 20)