def main(args):
# load the dataset
dataset = mnist.get_split('test', FLAGS.data_dir)
# load batch
images, labels = load_batch(
dataset,
FLAGS.batch_size,
is_training=False)
# get the model prediction
predictions = lenet(images)
# convert prediction values for each class into single class prediction
predictions = tf.to_int64(tf.argmax(predictions, 1))
# streaming metrics to evaluate
metrics_to_values, metrics_to_updates = metrics.aggregate_metric_map({
'mse': metrics.streaming_mean_squared_error(predictions, labels),
'accuracy': metrics.streaming_accuracy(predictions, labels),
})
# write the metrics as summaries
for metric_name, metric_value in metrics_to_values.iteritems():
tf.summary.scalar(metric_name, metric_value)
# evaluate on the model saved at the checkpoint directory
# evaluate every eval_interval_secs
slim.evaluation.evaluation_loop(
'',
FLAGS.checkpoint_dir,
FLAGS.log_dir,
num_evals=FLAGS.num_evals,
eval_op=metrics_to_updates.values(),
eval_interval_secs=FLAGS.eval_interval_secs)
def main(params):
train_set, valid_set, test_set = df.datasets.mnist.data()
train_set_x, train_set_y = train_set
test_set_x, test_set_y = test_set
model = lenet()
criterion = df.ClassNLLCriterion()
optimiser = df.SGD(lr=params['lr'])
for epoch in range(100):
model.training()
train(train_set_x, train_set_y, model, optimiser, criterion, epoch, params['batch_size'], 'train')
train(train_set_x, train_set_y, model, optimiser, criterion, epoch, params['batch_size'], 'stats')
model.evaluate()
validate(test_set_x, test_set_y, model, epoch, params['batch_size'])
def main(params):
train_set, valid_set, test_set = df.datasets.mnist.data()
train_set_x, train_set_y = train_set
test_set_x, test_set_y = test_set
model = lenet()
criterion = df.ClassNLLCriterion()
optimiser = df.SGD(lr=params['lr'])
for epoch in range(100):
model.training()
train(train_set_x, train_set_y, model, optimiser, criterion, epoch,
params['batch_size'], 'train')
train(train_set_x, train_set_y, model, optimiser, criterion, epoch,
params['batch_size'], 'stats')
model.evaluate()
validate(test_set_x, test_set_y, model, epoch, params['batch_size'])
def main(args):
# load the dataset
dataset = mnist.get_split('train', FLAGS.data_dir)
# load batch of dataset
images, labels = load_batch(
dataset,
FLAGS.batch_size,
is_training=True)
# run the image through the model
predictions = lenet(images)
# get the cross-entropy loss
one_hot_labels = slim.one_hot_encoding(
labels,
dataset.num_classes)
slim.losses.softmax_cross_entropy(
predictions,
one_hot_labels)
total_loss = slim.losses.get_total_loss()
tf.summary.scalar('loss', total_loss)
# use RMSProp to optimize
optimizer = tf.train.RMSPropOptimizer(0.001, 0.9)
# create train op
train_op = slim.learning.create_train_op(
total_loss,
optimizer,
summarize_gradients=True)
# run training
slim.learning.train(
train_op,
FLAGS.log_dir,
save_summaries_secs=20)
from model import lenet
FLAGS = tf.app.flags.FLAGS
# tf.app.flags.DEFINE_integer('max_steps', 1000000,
# """Number of iterations to run for each workers.""")
tf.app.flags.DEFINE_integer(
'max_steps', 1000, """Number of iterations to run for each workers.""")
tf.app.flags.DEFINE_integer('log_frequency', 50,
"""How many steps between two runop logs.""")
tf.app.flags.DEFINE_integer('batch_size', 32, """Batch size""")
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
ops = lenet()
train_op = ops['train_op']
loss = ops['loss']
acc = ops['acc']
x = ops['images']
y = ops['labels']
is_training = ops['is_training']
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
start = time.time()
for i in range(FLAGS.max_steps):
batch = mnist.train.next_batch(FLAGS.batch_size)
_, loss_ = sess.run([train_op, loss],
feed_dict={
x: batch[0],
from model import lenet
from tensorflow.examples.tutorials.mnist import input_data
hvd.init()
FLAGS = tf.app.flags.FLAGS
tf.app.flags.DEFINE_integer(
'max_steps', 1000000, """Number of iterations to run for each workers.""")
tf.app.flags.DEFINE_integer('log_frequency', 50,
"""How many steps between two runop logs.""")
tf.app.flags.DEFINE_integer('batch_size', 32, """Batch size""")
mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
ops = lenet(only_logits=True)
logits = ops['logits']
x = ops['images']
y = ops['labels']
is_training = ops['is_training']
global_step = ops['global_step']
loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits_v2(labels=y, logits=logits))
loss += model.weight_decay * tf.losses.get_regularization_loss()
acc = tf.reduce_mean(
tf.cast(tf.equal(tf.argmax(logits, axis=1), tf.argmax(y, axis=1)),
tf.float32))
optimizer = tf.train.AdamOptimizer(learning_rate=model.learning_rate)
optimizer = hvd.DistributedOptimizer(optimizer)
def train():
mnist = input_data.read_data_sets('MNIST_data/', one_hot=True)
batch_size = cfg.BATCH_SIZE
learning_rate = cfg.LEARNING_RATE
parameter_path = cfg.PARAMETER_FILE
num_epoches = cfg.MAX_ITER
x = tf.placeholder(tf.float32, [None, 28, 28, 1])
y = tf.placeholder(tf.float32, [None, cfg.NUM_CLASSES])
keep_prob = tf.placeholder(tf.float32)
generateVariables()
model = lenet()
logits = model.build(x, keep_prob)
loss_op = loss(logits, y)
optimizer_op = optimizer(loss_op, cfg.LEARNING_RATE)
acc = accuracy(logits, y)
# Summaries
tf.summary.scalar('train_loss', loss_op)
tf.summary.scalar('train_accuracy', acc)
merged_summary = tf.summary.merge_all()
train_writer = tf.summary.FileWriter('logs/train/')
val_writer = tf.summary.FileWriter('logs/val/')
saver = tf.train.Saver()
# Get the number of training/validation steps per epoch
train_batches_per_epoch = np.floor(mnist.train.labels.size /
batch_size).astype(np.int16)
val_batches_per_epoch = np.floor(mnist.test.labels.size /
batch_size).astype(np.int16)
with tf.Session() as sess:
with tf.device("/cpu:0"):
sess.run(tf.global_variables_initializer())
train_writer.add_graph(sess.graph)
for epoch in range(num_epoches):
print("{} Epoch number: {}".format(datetime.datetime.now(),
epoch + 1))
step = 1
while step < train_batches_per_epoch:
batch = mnist.train.next_batch(batch_size)
batch_x = np.reshape(batch[0], [-1, 28, 28, 1])
batch_y = np.reshape(batch[1], [-1, cfg.NUM_CLASSES])
sess.run(optimizer_op,
feed_dict={
x: batch_x,
y: batch_y,
keep_prob: cfg.KEEP_PROB
})
# Logging
if (step + 1) % 100 == 0:
train_acc = sess.run(acc,
feed_dict={
x: batch_x,
y: batch_y,
keep_prob: 1.0
})
print('step %d, training accuracy %g' %
(step + 1, train_acc))
s = sess.run(merged_summary,
feed_dict={
x: batch_x,
y: batch_y,
keep_prob: 1.0
})
train_writer.add_summary(
s, epoch * train_batches_per_epoch + step)
step += 1
# Epoch completed, start validation
print("{} Start validation".format(datetime.datetime.now()))
test_acc = 0.
test_count = 0
for _ in range(val_batches_per_epoch):
batch = mnist.test.next_batch(batch_size)
batch_x = np.reshape(batch[0], [-1, 28, 28, 1])
batch_y = np.reshape(batch[1], [-1, cfg.NUM_CLASSES])
val_acc = sess.run(acc,
feed_dict={
x: batch_x,
y: batch_y,
keep_prob: 1.0
})
test_acc += val_acc
test_count += 1
test_acc /= test_count
s = tf.Summary(value=[
tf.Summary.Value(tag='validation_accuracy',
simple_value=test_acc)
])
val_writer.add_summary(s, epoch + 1)
print("{} Validation Accuracy = {:.4f}".format(
datetime.datetime.now(), test_acc))
checkpoint_path = os.path.join(
'checkpoint/', 'model_epoch' + str(epoch + 1) + '.ckpt')
save_path = saver.save(sess, checkpoint_path)
print("{} Model checkpoint saved at {}".format(
datetime.datetime.now(), checkpoint_path))
def __init__(self):
self.lenet = lenet()
self.sess = tf.Session()
self.parameter_path = cfg.PARAMETER_FILE