def __init__(self, FLAGS):
"""
Initializes the ATLAS model.
Inputs:
- FLAGS: A _FlagValuesWrapper object.
"""
self.FLAGS = FLAGS
with tf.variable_scope("ATLASModel"):
self.add_placeholders()
self.build_graph()
self.add_loss()
print('Finished add_placeholders, build_graph, add_loss')
# Defines the trainable parameters, gradient, gradient norm, and clip by gradient norm
params = tf.trainable_variables()
# print('Number of trainable parameters:',np.sum([np.prod(v.get_shape().as_list()) for v in tf.trainable_variables()]))
# pprint([n.name for n in tf.get_default_graph().as_graph_def().node]);exit()
gradients = tf.gradients(self.loss, params)
self.gradient_norm = tf.global_norm(gradients)
clipped_gradients, _ = tf.clip_by_global_norm(gradients,
FLAGS.max_gradient_norm)
self.param_norm = tf.global_norm(params)
# Defines optimizer and updates; {self.updates} needs to be fetched in
# sess.run to do a gradient update
self.global_step_op = tf.Variable(0,
name="global_step",
trainable=False)
opt = tf.train.AdamOptimizer(learning_rate=FLAGS.learning_rate)
self.updates = opt.apply_gradients(zip(clipped_gradients, params),
global_step=self.global_step_op)
# Adds a summary to write examples of images to TensorBoard
utils.add_summary_image_triplet(
self.inputs_op,
self.target_masks_op,
self.predicted_masks_op,
num_images=self.FLAGS.num_summary_images,
use_volumetric=self.FLAGS.use_volumetric)
# Defines savers (for checkpointing) and summaries (for tensorboard)
self.saver = tf.train.Saver(tf.global_variables(),
max_to_keep=FLAGS.keep)
self.summaries = tf.summary.merge_all()
def __init__(self, FLAGS):
"""
Initializes the ATLAS model.
Inputs:
- FLAGS: A _FlagValuesWrapper object.
"""
self.FLAGS = FLAGS
with tf.variable_scope("MetaUNetATLASModel"):
self.add_placeholders()
self.build_graph()
self.add_loss()
# Defines the trainable parameters, gradient, gradient norm, and clip by
# gradient norm
params = tf.trainable_variables()
gradients = tf.gradients(self.loss, params)
self.gradient_norm = tf.global_norm(gradients)
clipped_gradients, _ = tf.clip_by_global_norm(gradients,
FLAGS.max_gradient_norm)
self.param_norm = tf.global_norm(params)
# Defines optimizer and updates; {self.updates} needs to be fetched in
# sess.run to do a gradient update
self.global_step_op = tf.Variable(0,
name="global_step",
trainable=False)
opt = tf.train.AdamOptimizer(learning_rate=FLAGS.learning_rate)
self.updates = opt.apply_gradients(zip(clipped_gradients, params),
global_step=self.global_step_op)
# Adds a summary to write examples of images to TensorBoard
utils.add_summary_image_triplet(
self.inputs_op[:, :, :, 0],
self.target_masks_op,
self.predicted_masks_op,
num_images=self.FLAGS.num_summary_images)
# Defines savers (for checkpointing) and summaries (for tensorboard)
self.saver = tf.train.Saver(tf.global_variables(),
max_to_keep=FLAGS.keep)
self.summaries = tf.summary.merge_all()