forked from leohsuofnthu/Tensorflow-YOLACT
-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
260 lines (230 loc) · 10.9 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
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
import datetime
import contextlib
import tensorflow as tf
# it s recommanded to use absl for tf 2.0
from absl import app
from absl import flags
from absl import logging
import yolact
from data import dataset_coco
from loss import loss_yolact
from utils import learning_rate_schedule
tf.random.set_seed(1234)
FLAGS = flags.FLAGS
flags.DEFINE_string('tfrecord_dir', './data/coco',
'directory of tfrecord')
flags.DEFINE_string('weights', './weights',
'path to store weights')
flags.DEFINE_integer('train_iter', 800000,
'iteraitons')
flags.DEFINE_integer('batch_size', 8,
'batch size')
flags.DEFINE_float('lr', 1e-3,
'learning rate')
flags.DEFINE_float('momentum', 0.9,
'momentum')
flags.DEFINE_float('weight_decay', 5 * 1e-4,
'weight_decay')
flags.DEFINE_float('print_interval', 10,
'number of iteration between printing loss')
flags.DEFINE_float('save_interval', 10000,
'number of iteration between saving model(checkpoint)')
flags.DEFINE_float('valid_iter', 5000,
'number of iteration between saving validation weights')
@tf.function
def train_step(model,
loss_fn,
metrics,
optimizer,
image,
labels):
# training using tensorflow gradient tape
with tf.GradientTape() as tape:
output = model(image, training=True)
loc_loss, conf_loss, mask_loss, seg_loss, total_loss = loss_fn(output, labels, 91)
grads = tape.gradient(total_loss, model.trainable_variables)
optimizer.apply_gradients(zip(grads, model.trainable_variables))
metrics.update_state(total_loss)
return loc_loss, conf_loss, mask_loss, seg_loss
@tf.function
def valid_step(model,
loss_fn,
metrics,
image,
labels):
output = model(image, training=False)
loc_loss, conf_loss, mask_loss, seg_loss, total_loss = loss_fn(output, labels, 91)
metrics.update_state(total_loss)
return loc_loss, conf_loss, mask_loss, seg_loss
def main(argv):
# set up Grappler for graph optimization
# Ref: https://www.tensorflow.org/guide/graph_optimization
@contextlib.contextmanager
def options(options):
old_opts = tf.config.optimizer.get_experimental_options()
tf.config.optimizer.set_experimental_options(options)
try:
yield
finally:
tf.config.optimizer.set_experimental_options(old_opts)
# -----------------------------------------------------------------
# Creating dataloaders for training and validation
logging.info("Creating the dataloader from: %s..." % FLAGS.tfrecord_dir)
train_dataset = dataset_coco.prepare_dataloader(tfrecord_dir=FLAGS.tfrecord_dir,
batch_size=FLAGS.batch_size,
subset='train')
valid_dataset = dataset_coco.prepare_dataloader(tfrecord_dir=FLAGS.tfrecord_dir,
batch_size=1,
subset='val')
# -----------------------------------------------------------------
# Creating the instance of the model specified.
logging.info("Creating the model instance of YOLACT")
model = yolact.Yolact(input_size=550,
fpn_channels=256,
feature_map_size=[69, 35, 18, 9, 5],
num_class=91,
num_mask=32,
aspect_ratio=[1, 0.5, 2],
scales=[24, 48, 96, 192, 384])
# add weight decay
for layer in model.layers:
if isinstance(layer, tf.keras.layers.Conv2D) or isinstance(layer, tf.keras.layers.Dense):
layer.add_loss(lambda: tf.keras.regularizers.l2(FLAGS.weight_decay)(layer.kernel))
if hasattr(layer, 'bias_regularizer') and layer.use_bias:
layer.add_loss(lambda: tf.keras.regularizers.l2(FLAGS.weight_decay)(layer.bias))
# -----------------------------------------------------------------
# Choose the Optimizor, Loss Function, and Metrics, learning rate schedule
lr_schedule = learning_rate_schedule.Yolact_LearningRateSchedule(warmup_steps=500, warmup_lr=1e-4,
initial_lr=FLAGS.lr)
logging.info("Initiate the Optimizer and Loss function...")
optimizer = tf.keras.optimizers.SGD(learning_rate=lr_schedule, momentum=FLAGS.momentum)
criterion = loss_yolact.YOLACTLoss()
train_loss = tf.keras.metrics.Mean('train_loss', dtype=tf.float32)
valid_loss = tf.keras.metrics.Mean('valid_loss', dtype=tf.float32)
loc = tf.keras.metrics.Mean('loc_loss', dtype=tf.float32)
conf = tf.keras.metrics.Mean('conf_loss', dtype=tf.float32)
mask = tf.keras.metrics.Mean('mask_loss', dtype=tf.float32)
seg = tf.keras.metrics.Mean('seg_loss', dtype=tf.float32)
v_loc = tf.keras.metrics.Mean('vloc_loss', dtype=tf.float32)
v_conf = tf.keras.metrics.Mean('vconf_loss', dtype=tf.float32)
v_mask = tf.keras.metrics.Mean('vmask_loss', dtype=tf.float32)
v_seg = tf.keras.metrics.Mean('vseg_loss', dtype=tf.float32)
# -----------------------------------------------------------------
# Setup the TensorBoard for better visualization
# Ref: https://www.tensorflow.org/tensorboard/get_started
logging.info("Setup the TensorBoard...")
current_time = datetime.datetime.now().strftime("%Y%m%d-%H%M%S")
train_log_dir = './logs/gradient_tape/' + current_time + '/train'
test_log_dir = './logs/gradient_tape/' + current_time + '/test'
train_summary_writer = tf.summary.create_file_writer(train_log_dir)
test_summary_writer = tf.summary.create_file_writer(test_log_dir)
# -----------------------------------------------------------------
# Start the Training and Validation Process
logging.info("Start the training process...")
# setup checkpoints manager
checkpoint = tf.train.Checkpoint(step=tf.Variable(1), optimizer=optimizer, model=model)
manager = tf.train.CheckpointManager(
checkpoint, directory="./checkpoints", max_to_keep=5
)
# restore from latest checkpoint and iteration
status = checkpoint.restore(manager.latest_checkpoint)
if manager.latest_checkpoint:
logging.info("Restored from {}".format(manager.latest_checkpoint))
else:
logging.info("Initializing from scratch.")
best_val = 1e10
iterations = checkpoint.step.numpy()
for image, labels in train_dataset:
# check iteration and change the learning rate
if iterations > FLAGS.train_iter:
break
checkpoint.step.assign_add(1)
iterations += 1
with options({'constant_folding': True,
'layout_optimize': True,
'loop_optimization': True,
'arithmetic_optimization': True,
'remapping': True}):
loc_loss, conf_loss, mask_loss, seg_loss = train_step(model, criterion, train_loss, optimizer, image,
labels)
loc.update_state(loc_loss)
conf.update_state(conf_loss)
mask.update_state(mask_loss)
seg.update_state(seg_loss)
with train_summary_writer.as_default():
tf.summary.scalar('Total loss', train_loss.result(), step=iterations)
tf.summary.scalar('Loc loss', loc.result(), step=iterations)
tf.summary.scalar('Conf loss', conf.result(), step=iterations)
tf.summary.scalar('Mask loss', mask.result(), step=iterations)
tf.summary.scalar('Seg loss', seg.result(), step=iterations)
if iterations and iterations % FLAGS.print_interval == 0:
logging.info("Iteration {}, LR: {}, Total Loss: {}, B: {}, C: {}, M: {}, S:{} ".format(
iterations,
optimizer._decayed_lr(var_dtype=tf.float32),
train_loss.result(), loc.result(),
conf.result(),
mask.result(),
seg.result()
))
if iterations and iterations % FLAGS.save_interval == 0:
# save checkpoint
save_path = manager.save()
logging.info("Saved checkpoint for step {}: {}".format(int(checkpoint.step), save_path))
# validation
valid_iter = 0
for valid_image, valid_labels in valid_dataset:
if valid_iter > FLAGS.valid_iter:
break
# calculate validation loss
with options({'constant_folding': True,
'layout_optimize': True,
'loop_optimization': True,
'arithmetic_optimization': True,
'remapping': True}):
valid_loc_loss, valid_conf_loss, valid_mask_loss, valid_seg_loss = valid_step(model,
criterion,
valid_loss,
valid_image,
valid_labels)
v_loc.update_state(valid_loc_loss)
v_conf.update_state(valid_conf_loss)
v_mask.update_state(valid_mask_loss)
v_seg.update_state(valid_seg_loss)
valid_iter += 1
with test_summary_writer.as_default():
tf.summary.scalar('V Total loss', valid_loss.result(), step=iterations)
tf.summary.scalar('V Loc loss', v_loc.result(), step=iterations)
tf.summary.scalar('V Conf loss', v_conf.result(), step=iterations)
tf.summary.scalar('V Mask loss', v_mask.result(), step=iterations)
tf.summary.scalar('V Seg loss', v_seg.result(), step=iterations)
train_template = 'Iteration {}, Train Loss: {}, Loc Loss: {}, Conf Loss: {}, Mask Loss: {}, Seg Loss: {}'
valid_template = 'Iteration {}, Valid Loss: {}, V Loc Loss: {}, V Conf Loss: {}, V Mask Loss: {}, Seg Loss: {}'
logging.info(train_template.format(iterations + 1,
train_loss.result(),
loc.result(),
conf.result(),
mask.result(),
seg.result()))
logging.info(valid_template.format(iterations + 1,
valid_loss.result(),
v_loc.result(),
v_conf.result(),
v_mask.result(),
v_seg.result()))
if valid_loss.result() < best_val:
# Saving the weights:
best_val = valid_loss.result()
model.save_weights('./weights/weights_' + str(valid_loss.result().numpy()) + '.h5')
# reset the metrics
train_loss.reset_states()
loc.reset_states()
conf.reset_states()
mask.reset_states()
seg.reset_states()
valid_loss.reset_states()
v_loc.reset_states()
v_conf.reset_states()
v_mask.reset_states()
v_seg.reset_states()
if __name__ == '__main__':
app.run(main)