def evaluate():
"""Eval CIFAR-10 for a number of steps."""
with tf.Graph().as_default() as g:
# Get images and labels for CIFAR-10.
eval_data = FLAGS.eval_data == 'test'
images, labels = resnet18.inputs(eval_data=eval_data)
# Build a Graph that computes the logits predictions from the
# inference model.
logits = resnet18.inference(images)
# Calculate predictions.
top_k_op = tf.nn.in_top_k(logits, labels, 5)
# Restore the moving average version of the learned variables for eval.
variable_averages = tf.train.ExponentialMovingAverage(
resnet18.MOVING_AVERAGE_DECAY)
variables_to_restore = variable_averages.variables_to_restore()
saver = tf.train.Saver(variables_to_restore)
# Build the summary operation based on the TF collection of Summaries.
summary_op = tf.summary.merge_all()
summary_writer = tf.summary.FileWriter(FLAGS.eval_dir, g)
while True:
eval_once(saver, summary_writer, top_k_op, summary_op)
if FLAGS.run_once:
break
time.sleep(FLAGS.eval_interval_secs)
def evaluate(k=5):
"""Eval CIFAR-10 for a number of steps."""
with tf.Graph().as_default() as g:
# Get images and labels for CIFAR-10.
images, labels = resnet18.inputs(eval_data=True)
# Build a Graph that computes the logits predictions from the
# inference model.
W1 = tf.placeholder(tf.float32, [7, 7, 3, 64])
W2_1_b2a = tf.placeholder(tf.float32, [3, 3, 64, 64])
W2_1_b2b = tf.placeholder(tf.float32, [3, 3, 64, 64])
W2_2_b2a = tf.placeholder(tf.float32, [3, 3, 64, 64])
W2_2_b2b = tf.placeholder(tf.float32, [3, 3, 64, 64])
W3_1_b1 = tf.placeholder(tf.float32, [1, 1, 64, 128])
W3_1_b2a = tf.placeholder(tf.float32, [3, 3, 64, 128])
W3_1_b2b = tf.placeholder(tf.float32, [3, 3, 128, 128])
W3_2_b2a = tf.placeholder(tf.float32, [3, 3, 128, 128])
W3_2_b2b = tf.placeholder(tf.float32, [3, 3, 128, 128])
W4_1_b1 = tf.placeholder(tf.float32, [1, 1, 128, 256])
W4_1_b2a = tf.placeholder(tf.float32, [3, 3, 128, 256])
W4_1_b2b = tf.placeholder(tf.float32, [3, 3, 256, 256])
W4_2_b2a = tf.placeholder(tf.float32, [3, 3, 256, 256])
W4_2_b2b = tf.placeholder(tf.float32, [3, 3, 256, 256])
W5_1_b1 = tf.placeholder(tf.float32, [1, 1, 256, 512])
W5_1_b2a = tf.placeholder(tf.float32, [3, 3, 256, 512])
W5_1_b2b = tf.placeholder(tf.float32, [3, 3, 512, 512])
W5_2_b2a = tf.placeholder(tf.float32, [3, 3, 512, 512])
W5_2_b2b = tf.placeholder(tf.float32, [3, 3, 512, 512])
logits = resnet18.inference(images, W1, W2_1_b2a, W2_1_b2b, W2_2_b2a,
W2_2_b2b, W3_1_b1, W3_1_b2a, W3_1_b2b,
W3_2_b2a, W3_2_b2b, W4_1_b1, W4_1_b2a,
W4_1_b2b, W4_2_b2a, W4_2_b2b, W5_1_b1,
W5_1_b2a, W5_1_b2b, W5_2_b2a, W5_2_b2b)
# Calculate predictions.
top_k_op = tf.nn.in_top_k(logits, labels, k)
# Restore the moving average version of the learned variables for eval.
# variable_averages = tf.train.ExponentialMovingAverage(
# resnet18.MOVING_AVERAGE_DECAY)
# variables_to_restore = variable_averages.variables_to_restore()
# saver = tf.train.Saver(variables_to_restore)
saver = tf.train.Saver()
# Build the summary operation based on the TF collection of Summaries.
summary_op = tf.summary.merge_all()
summary_writer = tf.summary.FileWriter(FLAGS.eval_dir, g)
while True:
precision = eval_once(saver, summary_writer, top_k_op, summary_op,
k, W1, W2_1_b2a, W2_1_b2b, W2_2_b2a,
W2_2_b2b, W3_1_b1, W3_1_b2a, W3_1_b2b,
W3_2_b2a, W3_2_b2b, W4_1_b1, W4_1_b2a,
W4_1_b2b, W4_2_b2a, W4_2_b2b, W5_1_b1,
W5_1_b2a, W5_1_b2b, W5_2_b2a, W5_2_b2b)
if FLAGS.run_once:
break
time.sleep(FLAGS.eval_interval_secs)
return precision
def tower_loss(scope, images, labels):
"""Calculate the total loss on a single tower running the CIFAR model.
Args:
scope: unique prefix string identifying the CIFAR tower, e.g. 'tower_0'
images: Images. 4D tensor of shape [batch_size, height, width, 3].
labels: Labels. 1D tensor of shape [batch_size].
Returns:
Tensor of shape [] containing the total loss for a batch of data
"""
# Build inference Graph.
logits = resnet18.inference(images, train=True)
# Build the portion of the Graph calculating the losses. Note that we will
# assemble the total_loss using a custom function below.
_ = resnet18.loss(logits, labels)
# Assemble all of the losses for the current tower only.
losses = tf.get_collection('losses', scope)
# Calculate the total loss for the current tower.
total_loss = tf.add_n(losses, name='total_loss')
# Attach a scalar summary to all individual losses and the total loss; do the
# same for the averaged version of the losses.
for l in losses + [total_loss]:
# Remove 'tower_[0-9]/' from the name in case this is a multi-GPU training
# session. This helps the clarity of presentation on tensorboard.
loss_name = re.sub('%s_[0-9]*/' % resnet18.TOWER_NAME, '', l.op.name)
tf.summary.scalar(loss_name, l)
return total_loss
def tower_loss(scope, images, labels):
# Build inference Graph.
logits = resnet18.inference(images,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
None,
train=True)
# Build the portion of the Graph calculating the losses. Note that we will
# assemble the total_loss using a custom function below.
_ = resnet18.loss(logits, labels)
# Assemble all of the losses for the current tower only.
losses = tf.get_collection('losses', scope)
# Calculate the total loss for the current tower.
total_loss = tf.add_n(losses, name='total_loss')
# Attach a scalar summary to all individual losses and the total loss; do the
# same for the averaged version of the losses.
for l in losses + [total_loss]:
# Remove 'tower_[0-9]/' from the name in case this is a multi-GPU training
# session. This helps the clarity of presentation on tensorboard.
loss_name = re.sub('%s_[0-9]*/' % resnet18.TOWER_NAME, '', l.op.name)
tf.summary.scalar(loss_name, l)
return total_loss