net_shape = net.get_shape().as_list()
net = tf.reshape(net, [-1, net_shape[1] * net_shape[2] * net_shape[3]])
# Compute logits (1 per class)
logits = tf.layers.dense(net, n_target_classes, activation=None, name='logits')
cross_entropy = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=y_))
train_step = tf.train.AdamOptimizer(1e-5,
name='train_step').minimize(cross_entropy)
correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(y_, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32),
name='accuracy')
model_file = os.path.dirname(
os.path.realpath(__file__)) + '/' + os.path.basename(__file__)
trainer = Trainer(data_path=data_path,
model_file=model_file,
s3_bucket=s3_bucket,
total_epochs=epochs,
max_sample_records=100,
show_speed=show_speed,
s3_sync=s3_sync)
trainer.train(sess=sess,
x=x,
y_=y_,
optimization=accuracy,
train_step=train_step,
train_feed_dict={},
test_feed_dict={})
from ai.Trainer import Trainer, parse_args
args = parse_args()
trainer = Trainer(data_path=args["data_path"],
postgres_host=args["postgres_host"],
port=args['port'],
model_base_directory=args['model_base_directory'],
total_epochs=args["epochs"],
image_scale=args['image_scale'],
crop_percent=args['crop_percent'])
trainer.train()
Regarding tf.control_dependencies:
with g.control_dependencies([a, b, c]):
# `d` and `e` will only run after `a`, `b`, and `c` have executed.
d = ...
e = ...
'''
update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS)
with tf.control_dependencies(update_ops):
train_step = tf.train.AdamOptimizer(1e-5).minimize(rmse)
model_file = os.path.dirname(
os.path.realpath(__file__)) + '/' + os.path.basename(__file__)
trainer = Trainer(data_path=data_path,
model_file=model_file,
s3_bucket=s3_bucket,
total_epochs=epochs,
max_sample_records=100,
show_speed=show_speed,
s3_sync=s3_sync,
save_to_disk=save_to_disk,
image_scale=image_scale)
trainer.train(sess=sess,
x=x,
y_=y_,
optimization=rmse,
train_step=train_step,
train_feed_dict={'phase:0': True},
test_feed_dict={})
from ai.Trainer import Trainer, parse_args
args = parse_args()
trainer = Trainer(overfit=args['overfit'],
data_path=args["data_path"],
batch_size=int(args['batch_size']),
postgres_host=args["postgres_host"],
port=args['port'],
model_base_directory=args['model_base_directory'],
model_id=args['model_id'],
total_epochs=args["epochs"],
image_scale=args['image_scale'],
crop_percent=args['crop_percent'])
trainer.train()