-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
155 lines (121 loc) · 7.06 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
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
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
from model_architectures.bilstm import BiLSTM
from model_architectures.res_bilstm import ResBiLSTM
from model_architectures.sentence_pair import SentencePair
from ReadData import ReadData
from keras.optimizers import Adam, RMSprop
from keras.models import Model
from keras.callbacks import TensorBoard, ModelCheckpoint, ReduceLROnPlateau, EarlyStopping
import tensorflow as tf
from tqdm import tqdm
import numpy as np
import os
import argparse
class TrainValTensorBoard(TensorBoard):
def __init__(self, log_dir='./logs', **kwargs):
# Make the original `TensorBoard` log to a subdirectory 'training'
training_log_dir = os.path.join(log_dir, 'training')
super(TrainValTensorBoard, self).__init__(training_log_dir, **kwargs)
# Log the validation metrics to a separate subdirectory
self.val_log_dir = os.path.join(log_dir, 'validation')
def set_model(self, model):
# Setup writer for validation metrics
self.val_writer = tf.summary.FileWriter(self.val_log_dir)
super(TrainValTensorBoard, self).set_model(model)
def on_epoch_end(self, epoch, logs=None):
# Pop the validation logs and handle them separately with
# `self.val_writer`. Also rename the keys so that they can
# be plotted on the same figure with the training metrics
logs = logs or {}
val_logs = {k.replace('val_', ''): v for k, v in logs.items() if k.startswith('val_')}
for name, value in val_logs.items():
summary = tf.Summary()
summary_value = summary.value.add()
summary_value.simple_value = value.item()
summary_value.tag = name
self.val_writer.add_summary(summary, epoch)
self.val_writer.flush()
# Pass the remaining logs to `TensorBoard.on_epoch_end`
logs = {k: v for k, v in logs.items() if not k.startswith('val_')}
super(TrainValTensorBoard, self).on_epoch_end(epoch, logs)
def on_train_end(self, logs=None):
super(TrainValTensorBoard, self).on_train_end(logs)
self.val_writer.close()
parser = argparse.ArgumentParser()
parser.add_argument('--model', '-m', required=True, help='Name of model to train [bilstm, resbilstm, sentence_pair]')
parser.add_argument('--dataset', '-d', default='Participants_Data_News_category/Data_Train.xlsx', help='Path to dataset')
parser.add_argument('--embedding_path', '-ep', default='fasttext-embedding/skipgram-256-news-classification.fasttext',
help='Path to Embedding Model | Default: fasttext-embedding/skipgram-256-news-classification.fasttext')
parser.add_argument('--embedding_type', '-et', default='fasttext', help='Embedding type [fasttext] | Default: fasttext')
parser.add_argument('--batch_size', '-b', default=64, help='Batch Size | Default: 64', type=int)
parser.add_argument('--epochs', '-e', default=50, help='No of Epochs | Default: 50', type=int)
parser.add_argument('--logs', '-l', help='Path to Logs (weights, tensorboard) | Default: [model_name]', type=str)
parser.add_argument('--no_classes', '-c', default=4, help='Number of Classes | Default: 4', type=int)
parser.add_argument('--hidden_size', '-hs', default=256, help='Hidden Size of LSTM Cell | Default: 256', type=int)
parser.add_argument('--learning_rate', '-lr', default=0.001, help='Learning Rate | Default: 0.001', type=float)
parser.add_argument('--train_val_split', '-tvs', default=0.2, help='Train vs Validation Split | Default: 0.2', type=float)
parser.add_argument('--check_build', action='store_true', help='Check Model Build')
parser.add_argument('--use_attention', '-a', action='store_true', help='Whether to add Attention Layer')
args = parser.parse_args()
hidden_size = args.hidden_size
if args.model == 'bilstm':
inputs = (400, 256)
model_instance = BiLSTM(hidden_size=hidden_size, no_classes=args.no_classes, use_attention=args.use_attention)
elif args.model == 'resbilstm':
inputs = (400, 256)
model_instance = ResBiLSTM(hidden_size=hidden_size, no_classes=args.no_classes, use_attention=args.use_attention)
elif args.model == 'sentence_pair':
inputs = [(400, 256), (256,)]
model_instance = SentencePair(hidden_size=hidden_size, no_classes=args.no_classes, use_attention=args.use_attention)
model = model_instance.build(inputs)
optimizer = RMSprop(args.learning_rate)
if args.no_classes > 1:
model.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])
else:
model.compile(loss='binary_crossentropy', optimizer=optimizer, metrics=['accuracy'])
model.summary()
if args.check_build:
exit()
embedding = {'type': args.embedding_type, 'path': args.embedding_path}
if args.model == 'sentence_pair':
reader = ReadData(path_file=args.dataset, embedding_config=embedding, data_shape=inputs, train_val_split=args.train_val_split, sentence_pair=True)
else:
reader = ReadData(path_file=args.dataset, embedding_config=embedding, data_shape=inputs, train_val_split=args.train_val_split, sentence_pair=False)
print('Reading Validation Data ..')
val_x, val_y = reader.read_val()
train_generator = reader.generator()
log_dir = args.model
logging = TrainValTensorBoard(log_dir=log_dir)
checkpoint = ModelCheckpoint(os.path.join(log_dir, 'ep{epoch:03d}-val_loss{val_loss:.3f}-val_acc{val_acc:.3f}.h5'),
monitor='val_loss', save_weights_only=True, save_best_only=True, period=1)
reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=3, verbose=1)
early_stopping = EarlyStopping(monitor='val_loss', min_delta=0, patience=5, verbose=1)
if not args.model == 'sentence_pair':
model.fit_generator(generator=train_generator, steps_per_epoch=int(reader.train_size/args.batch_size),
validation_data=(val_x, val_y), epochs=args.epochs,
callbacks=[logging, checkpoint, reduce_lr, early_stopping])
else:
if not os.path.exists(log_dir):
os.mkdir(log_dir)
for epoch in range(args.epochs):
num_batches = int(reader.train_size/args.batch_size)
start_index = 0
epoch_loss, epoch_acc = [], []
for i in range(num_batches):
start_index = i*args.batch_size
epoch_x, epoch_y = reader.get_next_batch(start_index, args.batch_size)
[loss, acc] = model.train_on_batch(epoch_x, epoch_y)
epoch_loss.append(loss)
epoch_acc.append(acc)
num_batches = int(reader.val_size/args.batch_size)
start_index = 0
val_loss, val_acc = [], []
for i in range(num_batches):
start_index = i*args.batch_size
epoch_x, epoch_y = reader.get_next_val_batch(start_index, args.batch_size)
[loss, acc] = model.test_on_batch(epoch_x, epoch_y)
val_loss.append(loss)
val_acc.append(acc)
print('Epoch {}/{}: loss: {}, acc: {}, val_loss: {}, val_acc: {}'.format(
epoch+1, args.epochs, np.average(epoch_loss), np.average(epoch_acc),
np.average(val_loss), np.average(val_acc)))
model.save_weights(os.path.join(log_dir, 'ep{}-val_loss{}-val_acc{}.h5'.format(epoch+1, np.average(val_loss), np.average(val_acc))))