Skip to content

GitLab

Commit d4eedbb9 authored by Mark Sandler's avatar Mark Sandler Committed by Hongkun Yu
Browse files

Merged commit includes the following changes: (#8077) 



Internal cleanup (py2->py3) plus the following changes:

285513318  by Sergio Guadarrama:

    Adds a script for post-training quantization

284222305  by Sergio Guadarrama:

    Modified squeeze-excite operation to accommodate tensors of undefined (Nonetype) H/W.

282028343  by Sergio Guadarrama:

    Add MobilenetV3 and MobilenetEdgeTPU to the slim/nets_factory.

PiperOrigin-RevId: 289455329
Co-authored-by: default avatarSergio Guadarrama <sguada@gmail.com>
parent 0e0a94a6
Hide whitespace changes
Inline Side-by-side
Showing with 533 additions and 304 deletions
research/slim/nets/nets_factory.py
View file @ d4eedbb9
......@@ -32,84 +32,98 @@ from nets import resnet_v2
from nets import s3dg
from nets import vgg
from nets.mobilenet import mobilenet_v2
from nets.mobilenet import mobilenet_v3
from nets.nasnet import nasnet
from nets.nasnet import pnasnet
slim = contrib_slim
networks_map = {'alexnet_v2': alexnet.alexnet_v2,
'cifarnet': cifarnet.cifarnet,
'overfeat': overfeat.overfeat,
'vgg_a': vgg.vgg_a,
'vgg_16': vgg.vgg_16,
'vgg_19': vgg.vgg_19,
'inception_v1': inception.inception_v1,
'inception_v2': inception.inception_v2,
'inception_v3': inception.inception_v3,
'inception_v4': inception.inception_v4,
'inception_resnet_v2': inception.inception_resnet_v2,
'i3d': i3d.i3d,
's3dg': s3dg.s3dg,
'lenet': lenet.lenet,
'resnet_v1_50': resnet_v1.resnet_v1_50,
'resnet_v1_101': resnet_v1.resnet_v1_101,
'resnet_v1_152': resnet_v1.resnet_v1_152,
'resnet_v1_200': resnet_v1.resnet_v1_200,
'resnet_v2_50': resnet_v2.resnet_v2_50,
'resnet_v2_101': resnet_v2.resnet_v2_101,
'resnet_v2_152': resnet_v2.resnet_v2_152,
'resnet_v2_200': resnet_v2.resnet_v2_200,
'mobilenet_v1': mobilenet_v1.mobilenet_v1,
'mobilenet_v1_075': mobilenet_v1.mobilenet_v1_075,
'mobilenet_v1_050': mobilenet_v1.mobilenet_v1_050,
'mobilenet_v1_025': mobilenet_v1.mobilenet_v1_025,
'mobilenet_v2': mobilenet_v2.mobilenet,
'mobilenet_v2_140': mobilenet_v2.mobilenet_v2_140,
'mobilenet_v2_035': mobilenet_v2.mobilenet_v2_035,
'nasnet_cifar': nasnet.build_nasnet_cifar,
'nasnet_mobile': nasnet.build_nasnet_mobile,
'nasnet_large': nasnet.build_nasnet_large,
'pnasnet_large': pnasnet.build_pnasnet_large,
'pnasnet_mobile': pnasnet.build_pnasnet_mobile,
}
networks_map = {
'alexnet_v2': alexnet.alexnet_v2,
'cifarnet': cifarnet.cifarnet,
'overfeat': overfeat.overfeat,
'vgg_a': vgg.vgg_a,
'vgg_16': vgg.vgg_16,
'vgg_19': vgg.vgg_19,
'inception_v1': inception.inception_v1,
'inception_v2': inception.inception_v2,
'inception_v3': inception.inception_v3,
'inception_v4': inception.inception_v4,
'inception_resnet_v2': inception.inception_resnet_v2,
'i3d': i3d.i3d,
's3dg': s3dg.s3dg,
'lenet': lenet.lenet,
'resnet_v1_50': resnet_v1.resnet_v1_50,
'resnet_v1_101': resnet_v1.resnet_v1_101,
'resnet_v1_152': resnet_v1.resnet_v1_152,
'resnet_v1_200': resnet_v1.resnet_v1_200,
'resnet_v2_50': resnet_v2.resnet_v2_50,
'resnet_v2_101': resnet_v2.resnet_v2_101,
'resnet_v2_152': resnet_v2.resnet_v2_152,
'resnet_v2_200': resnet_v2.resnet_v2_200,
'mobilenet_v1': mobilenet_v1.mobilenet_v1,
'mobilenet_v1_075': mobilenet_v1.mobilenet_v1_075,
'mobilenet_v1_050': mobilenet_v1.mobilenet_v1_050,
'mobilenet_v1_025': mobilenet_v1.mobilenet_v1_025,
'mobilenet_v2': mobilenet_v2.mobilenet,
'mobilenet_v2_140': mobilenet_v2.mobilenet_v2_140,
'mobilenet_v2_035': mobilenet_v2.mobilenet_v2_035,
'mobilenet_v3_small': mobilenet_v3.small,
'mobilenet_v3_large': mobilenet_v3.large,
'mobilenet_v3_small_minimalistic': mobilenet_v3.small_minimalistic,
'mobilenet_v3_large_minimalistic': mobilenet_v3.large_minimalistic,
'mobilenet_edgetpu': mobilenet_v3.edge_tpu,
'mobilenet_edgetpu_075': mobilenet_v3.edge_tpu_075,
'nasnet_cifar': nasnet.build_nasnet_cifar,
'nasnet_mobile': nasnet.build_nasnet_mobile,
'nasnet_large': nasnet.build_nasnet_large,
'pnasnet_large': pnasnet.build_pnasnet_large,
'pnasnet_mobile': pnasnet.build_pnasnet_mobile,
}
arg_scopes_map = {'alexnet_v2': alexnet.alexnet_v2_arg_scope,
'cifarnet': cifarnet.cifarnet_arg_scope,
'overfeat': overfeat.overfeat_arg_scope,
'vgg_a': vgg.vgg_arg_scope,
'vgg_16': vgg.vgg_arg_scope,
'vgg_19': vgg.vgg_arg_scope,
'inception_v1': inception.inception_v3_arg_scope,
'inception_v2': inception.inception_v3_arg_scope,
'inception_v3': inception.inception_v3_arg_scope,
'inception_v4': inception.inception_v4_arg_scope,
'inception_resnet_v2':
inception.inception_resnet_v2_arg_scope,
'i3d': i3d.i3d_arg_scope,
's3dg': s3dg.s3dg_arg_scope,
'lenet': lenet.lenet_arg_scope,
'resnet_v1_50': resnet_v1.resnet_arg_scope,
'resnet_v1_101': resnet_v1.resnet_arg_scope,
'resnet_v1_152': resnet_v1.resnet_arg_scope,
'resnet_v1_200': resnet_v1.resnet_arg_scope,
'resnet_v2_50': resnet_v2.resnet_arg_scope,
'resnet_v2_101': resnet_v2.resnet_arg_scope,
'resnet_v2_152': resnet_v2.resnet_arg_scope,
'resnet_v2_200': resnet_v2.resnet_arg_scope,
'mobilenet_v1': mobilenet_v1.mobilenet_v1_arg_scope,
'mobilenet_v1_075': mobilenet_v1.mobilenet_v1_arg_scope,
'mobilenet_v1_050': mobilenet_v1.mobilenet_v1_arg_scope,
'mobilenet_v1_025': mobilenet_v1.mobilenet_v1_arg_scope,
'mobilenet_v2': mobilenet_v2.training_scope,
'mobilenet_v2_035': mobilenet_v2.training_scope,
'mobilenet_v2_140': mobilenet_v2.training_scope,
'nasnet_cifar': nasnet.nasnet_cifar_arg_scope,
'nasnet_mobile': nasnet.nasnet_mobile_arg_scope,
'nasnet_large': nasnet.nasnet_large_arg_scope,
'pnasnet_large': pnasnet.pnasnet_large_arg_scope,
'pnasnet_mobile': pnasnet.pnasnet_mobile_arg_scope,
}
arg_scopes_map = {
'alexnet_v2': alexnet.alexnet_v2_arg_scope,
'cifarnet': cifarnet.cifarnet_arg_scope,
'overfeat': overfeat.overfeat_arg_scope,
'vgg_a': vgg.vgg_arg_scope,
'vgg_16': vgg.vgg_arg_scope,
'vgg_19': vgg.vgg_arg_scope,
'inception_v1': inception.inception_v3_arg_scope,
'inception_v2': inception.inception_v3_arg_scope,
'inception_v3': inception.inception_v3_arg_scope,
'inception_v4': inception.inception_v4_arg_scope,
'inception_resnet_v2': inception.inception_resnet_v2_arg_scope,
'i3d': i3d.i3d_arg_scope,
's3dg': s3dg.s3dg_arg_scope,
'lenet': lenet.lenet_arg_scope,
'resnet_v1_50': resnet_v1.resnet_arg_scope,
'resnet_v1_101': resnet_v1.resnet_arg_scope,
'resnet_v1_152': resnet_v1.resnet_arg_scope,
'resnet_v1_200': resnet_v1.resnet_arg_scope,
'resnet_v2_50': resnet_v2.resnet_arg_scope,
'resnet_v2_101': resnet_v2.resnet_arg_scope,
'resnet_v2_152': resnet_v2.resnet_arg_scope,
'resnet_v2_200': resnet_v2.resnet_arg_scope,
'mobilenet_v1': mobilenet_v1.mobilenet_v1_arg_scope,
'mobilenet_v1_075': mobilenet_v1.mobilenet_v1_arg_scope,
'mobilenet_v1_050': mobilenet_v1.mobilenet_v1_arg_scope,
'mobilenet_v1_025': mobilenet_v1.mobilenet_v1_arg_scope,
'mobilenet_v2': mobilenet_v2.training_scope,
'mobilenet_v2_035': mobilenet_v2.training_scope,
'mobilenet_v2_140': mobilenet_v2.training_scope,
'mobilenet_v3_small': mobilenet_v3.training_scope,
'mobilenet_v3_large': mobilenet_v3.training_scope,
'mobilenet_v3_small_minimalistic': mobilenet_v3.training_scope,
'mobilenet_v3_large_minimalistic': mobilenet_v3.training_scope,
'mobilenet_edgetpu': mobilenet_v3.training_scope,
'mobilenet_edgetpu_075': mobilenet_v3.training_scope,
'nasnet_cifar': nasnet.nasnet_cifar_arg_scope,
'nasnet_mobile': nasnet.nasnet_mobile_arg_scope,
'nasnet_large': nasnet.nasnet_large_arg_scope,
'pnasnet_large': pnasnet.pnasnet_large_arg_scope,
'pnasnet_mobile': pnasnet.pnasnet_mobile_arg_scope,
}
def get_network_fn(name, num_classes, weight_decay=0.0, is_training=False):
......
research/slim/nets/nets_factory_test.py
View file @ d4eedbb9
......@@ -36,8 +36,7 @@ class NetworksTest(tf.test.TestCase):
# Most networks use 224 as their default_image_size
image_size = getattr(net_fn, 'default_image_size', 224)
if net not in ['i3d', 's3dg']:
inputs = tf.random_uniform(
(batch_size, image_size, image_size, 3))
inputs = tf.random.uniform((batch_size, image_size, image_size, 3))
logits, end_points = net_fn(inputs)
self.assertTrue(isinstance(logits, tf.Tensor))
self.assertTrue(isinstance(end_points, dict))
......@@ -53,8 +52,7 @@ class NetworksTest(tf.test.TestCase):
# Most networks use 224 as their default_image_size
image_size = getattr(net_fn, 'default_image_size', 224)
if net not in ['i3d', 's3dg']:
inputs = tf.random_uniform(
(batch_size, image_size, image_size, 3))
inputs = tf.random.uniform((batch_size, image_size, image_size, 3))
logits, end_points = net_fn(inputs)
self.assertTrue(isinstance(logits, tf.Tensor))
self.assertTrue(isinstance(end_points, dict))
......@@ -69,8 +67,7 @@ class NetworksTest(tf.test.TestCase):
net_fn = nets_factory.get_network_fn(net, num_classes=num_classes)
# Most networks use 224 as their default_image_size
image_size = getattr(net_fn, 'default_image_size', 224) // 2
inputs = tf.random_uniform(
(batch_size, 10, image_size, image_size, 3))
inputs = tf.random.uniform((batch_size, 10, image_size, image_size, 3))
logits, end_points = net_fn(inputs)
self.assertTrue(isinstance(logits, tf.Tensor))
self.assertTrue(isinstance(end_points, dict))
......
research/slim/nets/overfeat.py
View file @ d4eedbb9
......@@ -35,14 +35,17 @@ import tensorflow as tf
from tensorflow.contrib import slim as contrib_slim
slim = contrib_slim
trunc_normal = lambda stddev: tf.truncated_normal_initializer(0.0, stddev)
# pylint: disable=g-long-lambda
trunc_normal = lambda stddev: tf.compat.v1.truncated_normal_initializer(
0.0, stddev)
def overfeat_arg_scope(weight_decay=0.0005):
with slim.arg_scope([slim.conv2d, slim.fully_connected],
activation_fn=tf.nn.relu,
weights_regularizer=slim.l2_regularizer(weight_decay),
biases_initializer=tf.zeros_initializer()):
biases_initializer=tf.compat.v1.zeros_initializer()):
with slim.arg_scope([slim.conv2d], padding='SAME'):
with slim.arg_scope([slim.max_pool2d], padding='VALID') as arg_sc:
return arg_sc
......@@ -88,7 +91,7 @@ def overfeat(inputs,
None).
end_points: a dict of tensors with intermediate activations.
"""
with tf.variable_scope(scope, 'overfeat', [inputs]) as sc:
with tf.compat.v1.variable_scope(scope, 'overfeat', [inputs]) as sc:
end_points_collection = sc.original_name_scope + '_end_points'
# Collect outputs for conv2d, fully_connected and max_pool2d
with slim.arg_scope([slim.conv2d, slim.fully_connected, slim.max_pool2d],
......@@ -104,9 +107,10 @@ def overfeat(inputs,
net = slim.max_pool2d(net, [2, 2], scope='pool5')
# Use conv2d instead of fully_connected layers.
with slim.arg_scope([slim.conv2d],
weights_initializer=trunc_normal(0.005),
biases_initializer=tf.constant_initializer(0.1)):
with slim.arg_scope(
[slim.conv2d],
weights_initializer=trunc_normal(0.005),
biases_initializer=tf.compat.v1.constant_initializer(0.1)):
net = slim.conv2d(net, 3072, [6, 6], padding='VALID', scope='fc6')
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout6')
......@@ -115,16 +119,19 @@ def overfeat(inputs,
end_points = slim.utils.convert_collection_to_dict(
end_points_collection)
if global_pool:
net = tf.reduce_mean(net, [1, 2], keep_dims=True, name='global_pool')
net = tf.reduce_mean(
input_tensor=net, axis=[1, 2], keepdims=True, name='global_pool')
end_points['global_pool'] = net
if num_classes:
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
scope='dropout7')
net = slim.conv2d(net, num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
biases_initializer=tf.zeros_initializer(),
scope='fc8')
net = slim.conv2d(
net,
num_classes, [1, 1],
activation_fn=None,
normalizer_fn=None,
biases_initializer=tf.compat.v1.zeros_initializer(),
scope='fc8')
if spatial_squeeze:
net = tf.squeeze(net, [1, 2], name='fc8/squeezed')
end_points[sc.name + '/fc8'] = net
......
research/slim/nets/overfeat_test.py
View file @ d4eedbb9
......@@ -32,7 +32,7 @@ class OverFeatTest(tf.test.TestCase):
height, width = 231, 231
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = overfeat.overfeat(inputs, num_classes)
self.assertEquals(logits.op.name, 'overfeat/fc8/squeezed')
self.assertListEqual(logits.get_shape().as_list(),
......@@ -43,7 +43,7 @@ class OverFeatTest(tf.test.TestCase):
height, width = 281, 281
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = overfeat.overfeat(inputs, num_classes, spatial_squeeze=False)
self.assertEquals(logits.op.name, 'overfeat/fc8/BiasAdd')
self.assertListEqual(logits.get_shape().as_list(),
......@@ -54,7 +54,7 @@ class OverFeatTest(tf.test.TestCase):
height, width = 281, 281
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = overfeat.overfeat(inputs, num_classes, spatial_squeeze=False,
global_pool=True)
self.assertEquals(logits.op.name, 'overfeat/fc8/BiasAdd')
......@@ -66,7 +66,7 @@ class OverFeatTest(tf.test.TestCase):
height, width = 231, 231
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
_, end_points = overfeat.overfeat(inputs, num_classes)
expected_names = ['overfeat/conv1',
'overfeat/pool1',
......@@ -87,7 +87,7 @@ class OverFeatTest(tf.test.TestCase):
height, width = 231, 231
num_classes = None
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
net, end_points = overfeat.overfeat(inputs, num_classes)
expected_names = ['overfeat/conv1',
'overfeat/pool1',
......@@ -108,7 +108,7 @@ class OverFeatTest(tf.test.TestCase):
height, width = 231, 231
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
overfeat.overfeat(inputs, num_classes)
expected_names = ['overfeat/conv1/weights',
'overfeat/conv1/biases',
......@@ -135,11 +135,11 @@ class OverFeatTest(tf.test.TestCase):
height, width = 231, 231
num_classes = 1000
with self.test_session():
eval_inputs = tf.random_uniform((batch_size, height, width, 3))
eval_inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = overfeat.overfeat(eval_inputs, is_training=False)
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
predictions = tf.argmax(logits, 1)
predictions = tf.argmax(input=logits, axis=1)
self.assertListEqual(predictions.get_shape().as_list(), [batch_size])
def testTrainEvalWithReuse(self):
......@@ -149,29 +149,29 @@ class OverFeatTest(tf.test.TestCase):
eval_height, eval_width = 281, 281
num_classes = 1000
with self.test_session():
train_inputs = tf.random_uniform(
train_inputs = tf.random.uniform(
(train_batch_size, train_height, train_width, 3))
logits, _ = overfeat.overfeat(train_inputs)
self.assertListEqual(logits.get_shape().as_list(),
[train_batch_size, num_classes])
tf.get_variable_scope().reuse_variables()
eval_inputs = tf.random_uniform(
tf.compat.v1.get_variable_scope().reuse_variables()
eval_inputs = tf.random.uniform(
(eval_batch_size, eval_height, eval_width, 3))
logits, _ = overfeat.overfeat(eval_inputs, is_training=False,
spatial_squeeze=False)
self.assertListEqual(logits.get_shape().as_list(),
[eval_batch_size, 2, 2, num_classes])
logits = tf.reduce_mean(logits, [1, 2])
predictions = tf.argmax(logits, 1)
logits = tf.reduce_mean(input_tensor=logits, axis=[1, 2])
predictions = tf.argmax(input=logits, axis=1)
self.assertEquals(predictions.get_shape().as_list(), [eval_batch_size])
def testForward(self):
batch_size = 1
height, width = 231, 231
with self.test_session() as sess:
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = overfeat.overfeat(inputs)
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(logits)
self.assertTrue(output.any())
......
research/slim/nets/pix2pix.py
View file @ d4eedbb9
......@@ -58,7 +58,8 @@ def pix2pix_arg_scope():
[layers.conv2d, layers.conv2d_transpose],
normalizer_fn=layers.instance_norm,
normalizer_params=instance_norm_params,
weights_initializer=tf.random_normal_initializer(0, 0.02)) as sc:
weights_initializer=tf.compat.v1.random_normal_initializer(0,
0.02)) as sc:
return sc
......@@ -80,13 +81,14 @@ def upsample(net, num_outputs, kernel_size, method='nn_upsample_conv'):
Raises:
ValueError: if `method` is not recognized.
"""
net_shape = tf.shape(net)
net_shape = tf.shape(input=net)
height = net_shape[1]
width = net_shape[2]
if method == 'nn_upsample_conv':
net = tf.image.resize_nearest_neighbor(
net, [kernel_size[0] * height, kernel_size[1] * width])
net = tf.image.resize(
net, [kernel_size[0] * height, kernel_size[1] * width],
method=tf.image.ResizeMethod.NEAREST_NEIGHBOR)
net = layers.conv2d(net, num_outputs, [4, 4], activation_fn=None)
elif method == 'conv2d_transpose':
net = layers.conv2d_transpose(
......@@ -166,7 +168,7 @@ def pix2pix_generator(net,
###########
# Encoder #
###########
with tf.variable_scope('encoder'):
with tf.compat.v1.variable_scope('encoder'):
with contrib_framework.arg_scope([layers.conv2d],
kernel_size=[4, 4],
stride=2,
......@@ -194,7 +196,7 @@ def pix2pix_generator(net,
reversed_blocks = list(blocks)
reversed_blocks.reverse()
with tf.variable_scope('decoder'):
with tf.compat.v1.variable_scope('decoder'):
# Dropout is used at both train and test time as per 'Image-to-Image',
# Section 2.1 (last paragraph).
with contrib_framework.arg_scope([layers.dropout], is_training=True):
......@@ -210,7 +212,7 @@ def pix2pix_generator(net,
net = layers.dropout(net, keep_prob=block.decoder_keep_prob)
end_points['decoder%d' % block_id] = net
with tf.variable_scope('output'):
with tf.compat.v1.variable_scope('output'):
# Explicitly set the normalizer_fn to None to override any default value
# that may come from an arg_scope, such as pix2pix_arg_scope.
logits = layers.conv2d(
......@@ -249,11 +251,11 @@ def pix2pix_discriminator(net, num_filters, padding=2, pad_mode='REFLECT',
def padded(net, scope):
if padding:
with tf.variable_scope(scope):
with tf.compat.v1.variable_scope(scope):
spatial_pad = tf.constant(
[[0, 0], [padding, padding], [padding, padding], [0, 0]],
dtype=tf.int32)
return tf.pad(net, spatial_pad, pad_mode)
return tf.pad(tensor=net, paddings=spatial_pad, mode=pad_mode)
else:
return net
......
research/slim/nets/pix2pix_test.py
View file @ d4eedbb9
......@@ -42,7 +42,7 @@ class GeneratorTest(tf.test.TestCase):
upsample_method='nn_upsample_conv')
with self.test_session() as session:
session.run(tf.global_variables_initializer())
session.run(tf.compat.v1.global_variables_initializer())
np_outputs = session.run(logits)
self.assertListEqual([batch_size, height, width, num_outputs],
list(np_outputs.shape))
......@@ -59,7 +59,7 @@ class GeneratorTest(tf.test.TestCase):
upsample_method='conv2d_transpose')
with self.test_session() as session:
session.run(tf.global_variables_initializer())
session.run(tf.compat.v1.global_variables_initializer())
np_outputs = session.run(logits)
self.assertListEqual([batch_size, height, width, num_outputs],
list(np_outputs.shape))
......
research/slim/nets/post_training_quantization.py 0 → 100644
View file @ d4eedbb9
# Copyright 2019 The TensorFlow Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
"""Export quantized tflite model from a trained checkpoint."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import functools
from absl import app
from absl import flags
import tensorflow as tf
import tensorflow_datasets as tfds
from nets import nets_factory
from preprocessing import preprocessing_factory
flags.DEFINE_string("model_name", None,
"The name of the architecture to quantize.")
flags.DEFINE_string("checkpoint_path", None, "Path to the training checkpoint.")
flags.DEFINE_string("dataset_name", "imagenet2012",
"Name of the dataset to use for quantization calibration.")
flags.DEFINE_string("dataset_dir", None, "Dataset location.")
flags.DEFINE_string(
"dataset_split", "train",
"The dataset split (train, validation etc.) to use for calibration.")
flags.DEFINE_string("output_tflite", None, "Path to output tflite file.")
flags.DEFINE_boolean(
"use_model_specific_preprocessing", False,
"When true, uses the preprocessing corresponding to the model as specified "
"in preprocessing factory.")
flags.DEFINE_boolean("enable_ema", True,
"Load exponential moving average version of variables.")
flags.DEFINE_integer(
"num_steps", 1000,
"Number of post-training quantization calibration steps to run.")
flags.DEFINE_integer("image_size", 224, "Size of the input image.")
flags.DEFINE_integer("num_classes", 1001,
"Number of output classes for the model.")
FLAGS = flags.FLAGS
# Mean and standard deviation used for normalizing the image tensor.
_MEAN_RGB = 127.5
_STD_RGB = 127.5
def _preprocess_for_quantization(image_data, image_size, crop_padding=32):
"""Crops to center of image with padding then scales, normalizes image_size.
Args:
image_data: A 3D Tensor representing the RGB image data. Image can be of
arbitrary height and width.
image_size: image height/width dimension.
crop_padding: the padding size to use when centering the crop.
Returns:
A decoded and cropped image Tensor. Image is normalized to [-1,1].
"""
shape = tf.shape(image_data)
image_height = shape[0]
image_width = shape[1]
padded_center_crop_size = tf.cast(
(image_size * 1.0 / (image_size + crop_padding)) *
tf.cast(tf.minimum(image_height, image_width), tf.float32), tf.int32)
offset_height = ((image_height - padded_center_crop_size) + 1) // 2
offset_width = ((image_width - padded_center_crop_size) + 1) // 2
image = tf.image.crop_to_bounding_box(
image_data,
offset_height=offset_height,
offset_width=offset_width,
target_height=padded_center_crop_size,
target_width=padded_center_crop_size)
image = tf.image.resize([image], [image_size, image_size],
method=tf.image.ResizeMethod.BICUBIC)[0]
image = tf.cast(image, tf.float32)
image -= tf.constant(_MEAN_RGB)
image /= tf.constant(_STD_RGB)
return image
def restore_model(sess, checkpoint_path, enable_ema=True):
"""Restore variables from the checkpoint into the provided session.
Args:
sess: A tensorflow session where the checkpoint will be loaded.
checkpoint_path: Path to the trained checkpoint.
enable_ema: (optional) Whether to load the exponential moving average (ema)
version of the tensorflow variables. Defaults to True.
"""
if enable_ema:
ema = tf.train.ExponentialMovingAverage(decay=0.0)
ema_vars = tf.trainable_variables() + tf.get_collection("moving_vars")
for v in tf.global_variables():
if "moving_mean" in v.name or "moving_variance" in v.name:
ema_vars.append(v)
ema_vars = list(set(ema_vars))
var_dict = ema.variables_to_restore(ema_vars)
else:
var_dict = None
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver(var_dict, max_to_keep=1)
saver.restore(sess, checkpoint_path)
def _representative_dataset_gen():
"""Gets a python generator of numpy arrays for the given dataset."""
image_size = FLAGS.image_size
dataset = tfds.builder(FLAGS.dataset_name, data_dir=FLAGS.dataset_dir)
dataset.download_and_prepare()
data = dataset.as_dataset()[FLAGS.dataset_split]
iterator = tf.compat.v1.data.make_one_shot_iterator(data)
if FLAGS.use_model_specific_preprocessing:
preprocess_fn = functools.partial(
preprocessing_factory.get_preprocessing(name=FLAGS.model_name),
output_height=image_size,
output_width=image_size)
else:
preprocess_fn = functools.partial(
_preprocess_for_quantization, image_size=image_size)
features = iterator.get_next()
image = features["image"]
image = preprocess_fn(image)
image = tf.reshape(image, [1, image_size, image_size, 3])
for _ in range(FLAGS.num_steps):
yield [image.eval()]
def main(_):
with tf.Graph().as_default(), tf.Session() as sess:
network_fn = nets_factory.get_network_fn(
FLAGS.model_name, num_classes=FLAGS.num_classes, is_training=False)
image_size = FLAGS.image_size
images = tf.placeholder(
tf.float32, shape=(1, image_size, image_size, 3), name="images")
logits, _ = network_fn(images)
output_tensor = tf.nn.softmax(logits)
restore_model(sess, FLAGS.checkpoint_path, enable_ema=FLAGS.enable_ema)
converter = tf.lite.TFLiteConverter.from_session(sess, [images],
[output_tensor])
converter.representative_dataset = tf.lite.RepresentativeDataset(
_representative_dataset_gen)
converter.optimizations = [tf.lite.Optimize.DEFAULT]
converter.inference_input_type = tf.int8
converter.inference_output_type = tf.int8
converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]
tflite_buffer = converter.convert()
with tf.gfile.GFile(FLAGS.output_tflite, "wb") as output_tflite:
output_tflite.write(tflite_buffer)
print("tflite model written to %s" % FLAGS.output_tflite)
if __name__ == "__main__":
flags.mark_flag_as_required("model_name")
flags.mark_flag_as_required("checkpoint_path")
flags.mark_flag_as_required("dataset_dir")
flags.mark_flag_as_required("output_tflite")
app.run(main)
research/slim/nets/resnet_utils.py
View file @ d4eedbb9
......@@ -117,8 +117,9 @@ def conv2d_same(inputs, num_outputs, kernel_size, stride, rate=1, scope=None):
pad_total = kernel_size_effective - 1
pad_beg = pad_total // 2
pad_end = pad_total - pad_beg
inputs = tf.pad(inputs,
[[0, 0], [pad_beg, pad_end], [pad_beg, pad_end], [0, 0]])
inputs = tf.pad(
tensor=inputs,
paddings=[[0, 0], [pad_beg, pad_end], [pad_beg, pad_end], [0, 0]])
return slim.conv2d(inputs, num_outputs, kernel_size, stride=stride,
rate=rate, padding='VALID', scope=scope)
......@@ -180,7 +181,7 @@ def stack_blocks_dense(net, blocks, output_stride=None,
rate = 1
for block in blocks:
with tf.variable_scope(block.scope, 'block', [net]) as sc:
with tf.compat.v1.variable_scope(block.scope, 'block', [net]) as sc:
block_stride = 1
for i, unit in enumerate(block.args):
if store_non_strided_activations and i == len(block.args) - 1:
......@@ -188,7 +189,7 @@ def stack_blocks_dense(net, blocks, output_stride=None,
block_stride = unit.get('stride', 1)
unit = dict(unit, stride=1)
with tf.variable_scope('unit_%d' % (i + 1), values=[net]):
with tf.compat.v1.variable_scope('unit_%d' % (i + 1), values=[net]):
# If we have reached the target output_stride, then we need to employ
# atrous convolution with stride=1 and multiply the atrous rate by the
# current unit's stride for use in subsequent layers.
......@@ -220,13 +221,14 @@ def stack_blocks_dense(net, blocks, output_stride=None,
return net
def resnet_arg_scope(weight_decay=0.0001,
batch_norm_decay=0.997,
batch_norm_epsilon=1e-5,
batch_norm_scale=True,
activation_fn=tf.nn.relu,
use_batch_norm=True,
batch_norm_updates_collections=tf.GraphKeys.UPDATE_OPS):
def resnet_arg_scope(
weight_decay=0.0001,
batch_norm_decay=0.997,
batch_norm_epsilon=1e-5,
batch_norm_scale=True,
activation_fn=tf.nn.relu,
use_batch_norm=True,
batch_norm_updates_collections=tf.compat.v1.GraphKeys.UPDATE_OPS):
"""Defines the default ResNet arg scope.
TODO(gpapan): The batch-normalization related default values above are
......
research/slim/nets/resnet_v1.py
View file @ d4eedbb9
......@@ -109,7 +109,7 @@ def bottleneck(inputs,
Returns:
The ResNet unit's output.
"""
with tf.variable_scope(scope, 'bottleneck_v1', [inputs]) as sc:
with tf.compat.v1.variable_scope(scope, 'bottleneck_v1', [inputs]) as sc:
depth_in = slim.utils.last_dimension(inputs.get_shape(), min_rank=4)
if depth == depth_in:
shortcut = resnet_utils.subsample(inputs, stride, 'shortcut')
......@@ -219,7 +219,8 @@ def resnet_v1(inputs,
Raises:
ValueError: If the target output_stride is not valid.
"""
with tf.variable_scope(scope, 'resnet_v1', [inputs], reuse=reuse) as sc:
with tf.compat.v1.variable_scope(
scope, 'resnet_v1', [inputs], reuse=reuse) as sc:
end_points_collection = sc.original_name_scope + '_end_points'
with slim.arg_scope([slim.conv2d, bottleneck,
resnet_utils.stack_blocks_dense],
......@@ -242,7 +243,8 @@ def resnet_v1(inputs,
if global_pool:
# Global average pooling.
net = tf.reduce_mean(net, [1, 2], name='pool5', keep_dims=True)
net = tf.reduce_mean(
input_tensor=net, axis=[1, 2], name='pool5', keepdims=True)
end_points['global_pool'] = net
if num_classes:
net = slim.conv2d(net, num_classes, [1, 1], activation_fn=None,
......
research/slim/nets/resnet_v1_test.py
View file @ d4eedbb9
......@@ -27,6 +27,8 @@ from nets import resnet_v1
slim = contrib_slim
tf.compat.v1.disable_resource_variables()
def create_test_input(batch_size, height, width, channels):
"""Create test input tensor.
......@@ -43,28 +45,29 @@ def create_test_input(batch_size, height, width, channels):
constant `Tensor` with the mesh grid values along the spatial dimensions.
"""
if None in [batch_size, height, width, channels]:
return tf.placeholder(tf.float32, (batch_size, height, width, channels))
return tf.compat.v1.placeholder(tf.float32,
(batch_size, height, width, channels))
else:
return tf.to_float(
return tf.cast(
np.tile(
np.reshape(
np.reshape(np.arange(height), [height, 1]) +
np.reshape(np.arange(width), [1, width]),
[1, height, width, 1]),
[batch_size, 1, 1, channels]))
[1, height, width, 1]), [batch_size, 1, 1, channels]),
dtype=tf.float32)
class ResnetUtilsTest(tf.test.TestCase):
def testSubsampleThreeByThree(self):
x = tf.reshape(tf.to_float(tf.range(9)), [1, 3, 3, 1])
x = tf.reshape(tf.cast(tf.range(9), dtype=tf.float32), [1, 3, 3, 1])
x = resnet_utils.subsample(x, 2)
expected = tf.reshape(tf.constant([0, 2, 6, 8]), [1, 2, 2, 1])
with self.test_session():
self.assertAllClose(x.eval(), expected.eval())
def testSubsampleFourByFour(self):
x = tf.reshape(tf.to_float(tf.range(16)), [1, 4, 4, 1])
x = tf.reshape(tf.cast(tf.range(16), dtype=tf.float32), [1, 4, 4, 1])
x = resnet_utils.subsample(x, 2)
expected = tf.reshape(tf.constant([0, 2, 8, 10]), [1, 2, 2, 1])
with self.test_session():
......@@ -80,32 +83,29 @@ class ResnetUtilsTest(tf.test.TestCase):
w = create_test_input(1, 3, 3, 1)
w = tf.reshape(w, [3, 3, 1, 1])
tf.get_variable('Conv/weights', initializer=w)
tf.get_variable('Conv/biases', initializer=tf.zeros([1]))
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable('Conv/weights', initializer=w)
tf.compat.v1.get_variable('Conv/biases', initializer=tf.zeros([1]))
tf.compat.v1.get_variable_scope().reuse_variables()
y1 = slim.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
y1_expected = tf.to_float([[14, 28, 43, 26],
[28, 48, 66, 37],
[43, 66, 84, 46],
[26, 37, 46, 22]])
y1_expected = tf.cast([[14, 28, 43, 26], [28, 48, 66, 37], [43, 66, 84, 46],
[26, 37, 46, 22]],
dtype=tf.float32)
y1_expected = tf.reshape(y1_expected, [1, n, n, 1])
y2 = resnet_utils.subsample(y1, 2)
y2_expected = tf.to_float([[14, 43],
[43, 84]])
y2_expected = tf.cast([[14, 43], [43, 84]], dtype=tf.float32)
y2_expected = tf.reshape(y2_expected, [1, n2, n2, 1])
y3 = resnet_utils.conv2d_same(x, 1, 3, stride=2, scope='Conv')
y3_expected = y2_expected
y4 = slim.conv2d(x, 1, [3, 3], stride=2, scope='Conv')
y4_expected = tf.to_float([[48, 37],
[37, 22]])
y4_expected = tf.cast([[48, 37], [37, 22]], dtype=tf.float32)
y4_expected = tf.reshape(y4_expected, [1, n2, n2, 1])
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
self.assertAllClose(y1.eval(), y1_expected.eval())
self.assertAllClose(y2.eval(), y2_expected.eval())
self.assertAllClose(y3.eval(), y3_expected.eval())
......@@ -121,22 +121,20 @@ class ResnetUtilsTest(tf.test.TestCase):
w = create_test_input(1, 3, 3, 1)
w = tf.reshape(w, [3, 3, 1, 1])
tf.get_variable('Conv/weights', initializer=w)
tf.get_variable('Conv/biases', initializer=tf.zeros([1]))
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable('Conv/weights', initializer=w)
tf.compat.v1.get_variable('Conv/biases', initializer=tf.zeros([1]))
tf.compat.v1.get_variable_scope().reuse_variables()
y1 = slim.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
y1_expected = tf.to_float([[14, 28, 43, 58, 34],
[28, 48, 66, 84, 46],
[43, 66, 84, 102, 55],
[58, 84, 102, 120, 64],
[34, 46, 55, 64, 30]])
y1_expected = tf.cast(
[[14, 28, 43, 58, 34], [28, 48, 66, 84, 46], [43, 66, 84, 102, 55],
[58, 84, 102, 120, 64], [34, 46, 55, 64, 30]],
dtype=tf.float32)
y1_expected = tf.reshape(y1_expected, [1, n, n, 1])
y2 = resnet_utils.subsample(y1, 2)
y2_expected = tf.to_float([[14, 43, 34],
[43, 84, 55],
[34, 55, 30]])
y2_expected = tf.cast([[14, 43, 34], [43, 84, 55], [34, 55, 30]],
dtype=tf.float32)
y2_expected = tf.reshape(y2_expected, [1, n2, n2, 1])
y3 = resnet_utils.conv2d_same(x, 1, 3, stride=2, scope='Conv')
......@@ -146,7 +144,7 @@ class ResnetUtilsTest(tf.test.TestCase):
y4_expected = y2_expected
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
self.assertAllClose(y1.eval(), y1_expected.eval())
self.assertAllClose(y2.eval(), y2_expected.eval())
self.assertAllClose(y3.eval(), y3_expected.eval())
......@@ -154,7 +152,7 @@ class ResnetUtilsTest(tf.test.TestCase):
def _resnet_plain(self, inputs, blocks, output_stride=None, scope=None):
"""A plain ResNet without extra layers before or after the ResNet blocks."""
with tf.variable_scope(scope, values=[inputs]):
with tf.compat.v1.variable_scope(scope, values=[inputs]):
with slim.arg_scope([slim.conv2d], outputs_collections='end_points'):
net = resnet_utils.stack_blocks_dense(inputs, blocks, output_stride)
end_points = slim.utils.convert_collection_to_dict('end_points')
......@@ -191,9 +189,9 @@ class ResnetUtilsTest(tf.test.TestCase):
def _stack_blocks_nondense(self, net, blocks):
"""A simplified ResNet Block stacker without output stride control."""
for block in blocks:
with tf.variable_scope(block.scope, 'block', [net]):
with tf.compat.v1.variable_scope(block.scope, 'block', [net]):
for i, unit in enumerate(block.args):
with tf.variable_scope('unit_%d' % (i + 1), values=[net]):
with tf.compat.v1.variable_scope('unit_%d' % (i + 1), values=[net]):
net = block.unit_fn(net, rate=1, **unit)
return net
......@@ -221,7 +219,7 @@ class ResnetUtilsTest(tf.test.TestCase):
for output_stride in [1, 2, 4, 8, None]:
with tf.Graph().as_default():
with self.test_session() as sess:
tf.set_random_seed(0)
tf.compat.v1.set_random_seed(0)
inputs = create_test_input(1, height, width, 3)
# Dense feature extraction followed by subsampling.
output = resnet_utils.stack_blocks_dense(inputs,
......@@ -234,10 +232,10 @@ class ResnetUtilsTest(tf.test.TestCase):
output = resnet_utils.subsample(output, factor)
# Make the two networks use the same weights.
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable_scope().reuse_variables()
# Feature extraction at the nominal network rate.
expected = self._stack_blocks_nondense(inputs, blocks)
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output, expected = sess.run([output, expected])
self.assertAllClose(output, expected, atol=1e-4, rtol=1e-4)
......@@ -264,7 +262,7 @@ class ResnetUtilsTest(tf.test.TestCase):
for output_stride in [1, 2, 4, 8, None]:
with tf.Graph().as_default():
with self.test_session() as sess:
tf.set_random_seed(0)
tf.compat.v1.set_random_seed(0)
inputs = create_test_input(1, height, width, 3)
# Subsampling at the last unit of the block.
......@@ -276,7 +274,7 @@ class ResnetUtilsTest(tf.test.TestCase):
'output')
# Make the two networks use the same weights.
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable_scope().reuse_variables()
# Subsample activations at the end of the blocks.
expected = resnet_utils.stack_blocks_dense(
......@@ -286,7 +284,7 @@ class ResnetUtilsTest(tf.test.TestCase):
expected_end_points = slim.utils.convert_collection_to_dict(
'expected')
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
# Make sure that the final output is the same.
output, expected = sess.run([output, expected])
......@@ -477,7 +475,7 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
with slim.arg_scope(resnet_utils.resnet_arg_scope()):
with tf.Graph().as_default():
with self.test_session() as sess:
tf.set_random_seed(0)
tf.compat.v1.set_random_seed(0)
inputs = create_test_input(2, 81, 81, 3)
# Dense feature extraction followed by subsampling.
output, _ = self._resnet_small(inputs, None, is_training=False,
......@@ -489,11 +487,11 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
factor = nominal_stride // output_stride
output = resnet_utils.subsample(output, factor)
# Make the two networks use the same weights.
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable_scope().reuse_variables()
# Feature extraction at the nominal network rate.
expected, _ = self._resnet_small(inputs, None, is_training=False,
global_pool=False)
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
self.assertAllClose(output.eval(), expected.eval(),
atol=1e-4, rtol=1e-4)
......@@ -513,7 +511,7 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
[None, 1, 1, num_classes])
images = create_test_input(batch, height, width, 3)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(logits, {inputs: images.eval()})
self.assertEqual(output.shape, (batch, 1, 1, num_classes))
......@@ -528,7 +526,7 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
[batch, None, None, 32])
images = create_test_input(batch, height, width, 3)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(output, {inputs: images.eval()})
self.assertEqual(output.shape, (batch, 3, 3, 32))
......@@ -547,7 +545,7 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
[batch, None, None, 32])
images = create_test_input(batch, height, width, 3)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(output, {inputs: images.eval()})
self.assertEqual(output.shape, (batch, 9, 9, 32))
......
research/slim/nets/resnet_v2.py
View file @ d4eedbb9
......@@ -84,7 +84,7 @@ def bottleneck(inputs, depth, depth_bottleneck, stride, rate=1,
Returns:
The ResNet unit's output.
"""
with tf.variable_scope(scope, 'bottleneck_v2', [inputs]) as sc:
with tf.compat.v1.variable_scope(scope, 'bottleneck_v2', [inputs]) as sc:
depth_in = slim.utils.last_dimension(inputs.get_shape(), min_rank=4)
preact = slim.batch_norm(inputs, activation_fn=tf.nn.relu, scope='preact')
if depth == depth_in:
......@@ -181,7 +181,8 @@ def resnet_v2(inputs,
Raises:
ValueError: If the target output_stride is not valid.
"""
with tf.variable_scope(scope, 'resnet_v2', [inputs], reuse=reuse) as sc:
with tf.compat.v1.variable_scope(
scope, 'resnet_v2', [inputs], reuse=reuse) as sc:
end_points_collection = sc.original_name_scope + '_end_points'
with slim.arg_scope([slim.conv2d, bottleneck,
resnet_utils.stack_blocks_dense],
......@@ -211,7 +212,8 @@ def resnet_v2(inputs,
if global_pool:
# Global average pooling.
net = tf.reduce_mean(net, [1, 2], name='pool5', keep_dims=True)
net = tf.reduce_mean(
input_tensor=net, axis=[1, 2], name='pool5', keepdims=True)
end_points['global_pool'] = net
if num_classes:
net = slim.conv2d(net, num_classes, [1, 1], activation_fn=None,
......
research/slim/nets/resnet_v2_test.py
View file @ d4eedbb9
......@@ -27,6 +27,8 @@ from nets import resnet_v2
slim = contrib_slim
tf.compat.v1.disable_resource_variables()
def create_test_input(batch_size, height, width, channels):
"""Create test input tensor.
......@@ -43,28 +45,29 @@ def create_test_input(batch_size, height, width, channels):
constant `Tensor` with the mesh grid values along the spatial dimensions.
"""
if None in [batch_size, height, width, channels]:
return tf.placeholder(tf.float32, (batch_size, height, width, channels))
return tf.compat.v1.placeholder(tf.float32,
(batch_size, height, width, channels))
else:
return tf.to_float(
return tf.cast(
np.tile(
np.reshape(
np.reshape(np.arange(height), [height, 1]) +
np.reshape(np.arange(width), [1, width]),
[1, height, width, 1]),
[batch_size, 1, 1, channels]))
[1, height, width, 1]), [batch_size, 1, 1, channels]),
dtype=tf.float32)
class ResnetUtilsTest(tf.test.TestCase):
def testSubsampleThreeByThree(self):
x = tf.reshape(tf.to_float(tf.range(9)), [1, 3, 3, 1])
x = tf.reshape(tf.cast(tf.range(9), dtype=tf.float32), [1, 3, 3, 1])
x = resnet_utils.subsample(x, 2)
expected = tf.reshape(tf.constant([0, 2, 6, 8]), [1, 2, 2, 1])
with self.test_session():
self.assertAllClose(x.eval(), expected.eval())
def testSubsampleFourByFour(self):
x = tf.reshape(tf.to_float(tf.range(16)), [1, 4, 4, 1])
x = tf.reshape(tf.cast(tf.range(16), dtype=tf.float32), [1, 4, 4, 1])
x = resnet_utils.subsample(x, 2)
expected = tf.reshape(tf.constant([0, 2, 8, 10]), [1, 2, 2, 1])
with self.test_session():
......@@ -80,32 +83,29 @@ class ResnetUtilsTest(tf.test.TestCase):
w = create_test_input(1, 3, 3, 1)
w = tf.reshape(w, [3, 3, 1, 1])
tf.get_variable('Conv/weights', initializer=w)
tf.get_variable('Conv/biases', initializer=tf.zeros([1]))
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable('Conv/weights', initializer=w)
tf.compat.v1.get_variable('Conv/biases', initializer=tf.zeros([1]))
tf.compat.v1.get_variable_scope().reuse_variables()
y1 = slim.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
y1_expected = tf.to_float([[14, 28, 43, 26],
[28, 48, 66, 37],
[43, 66, 84, 46],
[26, 37, 46, 22]])
y1_expected = tf.cast([[14, 28, 43, 26], [28, 48, 66, 37], [43, 66, 84, 46],
[26, 37, 46, 22]],
dtype=tf.float32)
y1_expected = tf.reshape(y1_expected, [1, n, n, 1])
y2 = resnet_utils.subsample(y1, 2)
y2_expected = tf.to_float([[14, 43],
[43, 84]])
y2_expected = tf.cast([[14, 43], [43, 84]], dtype=tf.float32)
y2_expected = tf.reshape(y2_expected, [1, n2, n2, 1])
y3 = resnet_utils.conv2d_same(x, 1, 3, stride=2, scope='Conv')
y3_expected = y2_expected
y4 = slim.conv2d(x, 1, [3, 3], stride=2, scope='Conv')
y4_expected = tf.to_float([[48, 37],
[37, 22]])
y4_expected = tf.cast([[48, 37], [37, 22]], dtype=tf.float32)
y4_expected = tf.reshape(y4_expected, [1, n2, n2, 1])
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
self.assertAllClose(y1.eval(), y1_expected.eval())
self.assertAllClose(y2.eval(), y2_expected.eval())
self.assertAllClose(y3.eval(), y3_expected.eval())
......@@ -121,22 +121,20 @@ class ResnetUtilsTest(tf.test.TestCase):
w = create_test_input(1, 3, 3, 1)
w = tf.reshape(w, [3, 3, 1, 1])
tf.get_variable('Conv/weights', initializer=w)
tf.get_variable('Conv/biases', initializer=tf.zeros([1]))
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable('Conv/weights', initializer=w)
tf.compat.v1.get_variable('Conv/biases', initializer=tf.zeros([1]))
tf.compat.v1.get_variable_scope().reuse_variables()
y1 = slim.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
y1_expected = tf.to_float([[14, 28, 43, 58, 34],
[28, 48, 66, 84, 46],
[43, 66, 84, 102, 55],
[58, 84, 102, 120, 64],
[34, 46, 55, 64, 30]])
y1_expected = tf.cast(
[[14, 28, 43, 58, 34], [28, 48, 66, 84, 46], [43, 66, 84, 102, 55],
[58, 84, 102, 120, 64], [34, 46, 55, 64, 30]],
dtype=tf.float32)
y1_expected = tf.reshape(y1_expected, [1, n, n, 1])
y2 = resnet_utils.subsample(y1, 2)
y2_expected = tf.to_float([[14, 43, 34],
[43, 84, 55],
[34, 55, 30]])
y2_expected = tf.cast([[14, 43, 34], [43, 84, 55], [34, 55, 30]],
dtype=tf.float32)
y2_expected = tf.reshape(y2_expected, [1, n2, n2, 1])
y3 = resnet_utils.conv2d_same(x, 1, 3, stride=2, scope='Conv')
......@@ -146,7 +144,7 @@ class ResnetUtilsTest(tf.test.TestCase):
y4_expected = y2_expected
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
self.assertAllClose(y1.eval(), y1_expected.eval())
self.assertAllClose(y2.eval(), y2_expected.eval())
self.assertAllClose(y3.eval(), y3_expected.eval())
......@@ -154,7 +152,7 @@ class ResnetUtilsTest(tf.test.TestCase):
def _resnet_plain(self, inputs, blocks, output_stride=None, scope=None):
"""A plain ResNet without extra layers before or after the ResNet blocks."""
with tf.variable_scope(scope, values=[inputs]):
with tf.compat.v1.variable_scope(scope, values=[inputs]):
with slim.arg_scope([slim.conv2d], outputs_collections='end_points'):
net = resnet_utils.stack_blocks_dense(inputs, blocks, output_stride)
end_points = slim.utils.convert_collection_to_dict('end_points')
......@@ -191,9 +189,9 @@ class ResnetUtilsTest(tf.test.TestCase):
def _stack_blocks_nondense(self, net, blocks):
"""A simplified ResNet Block stacker without output stride control."""
for block in blocks:
with tf.variable_scope(block.scope, 'block', [net]):
with tf.compat.v1.variable_scope(block.scope, 'block', [net]):
for i, unit in enumerate(block.args):
with tf.variable_scope('unit_%d' % (i + 1), values=[net]):
with tf.compat.v1.variable_scope('unit_%d' % (i + 1), values=[net]):
net = block.unit_fn(net, rate=1, **unit)
return net
......@@ -221,7 +219,7 @@ class ResnetUtilsTest(tf.test.TestCase):
for output_stride in [1, 2, 4, 8, None]:
with tf.Graph().as_default():
with self.test_session() as sess:
tf.set_random_seed(0)
tf.compat.v1.set_random_seed(0)
inputs = create_test_input(1, height, width, 3)
# Dense feature extraction followed by subsampling.
output = resnet_utils.stack_blocks_dense(inputs,
......@@ -234,10 +232,10 @@ class ResnetUtilsTest(tf.test.TestCase):
output = resnet_utils.subsample(output, factor)
# Make the two networks use the same weights.
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable_scope().reuse_variables()
# Feature extraction at the nominal network rate.
expected = self._stack_blocks_nondense(inputs, blocks)
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output, expected = sess.run([output, expected])
self.assertAllClose(output, expected, atol=1e-4, rtol=1e-4)
......@@ -394,7 +392,7 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
with slim.arg_scope(resnet_utils.resnet_arg_scope()):
with tf.Graph().as_default():
with self.test_session() as sess:
tf.set_random_seed(0)
tf.compat.v1.set_random_seed(0)
inputs = create_test_input(2, 81, 81, 3)
# Dense feature extraction followed by subsampling.
output, _ = self._resnet_small(inputs, None,
......@@ -407,12 +405,12 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
factor = nominal_stride // output_stride
output = resnet_utils.subsample(output, factor)
# Make the two networks use the same weights.
tf.get_variable_scope().reuse_variables()
tf.compat.v1.get_variable_scope().reuse_variables()
# Feature extraction at the nominal network rate.
expected, _ = self._resnet_small(inputs, None,
is_training=False,
global_pool=False)
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
self.assertAllClose(output.eval(), expected.eval(),
atol=1e-4, rtol=1e-4)
......@@ -432,7 +430,7 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
[None, 1, 1, num_classes])
images = create_test_input(batch, height, width, 3)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(logits, {inputs: images.eval()})
self.assertEqual(output.shape, (batch, 1, 1, num_classes))
......@@ -448,7 +446,7 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
[batch, None, None, 32])
images = create_test_input(batch, height, width, 3)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(output, {inputs: images.eval()})
self.assertEqual(output.shape, (batch, 3, 3, 32))
......@@ -467,7 +465,7 @@ class ResnetCompleteNetworkTest(tf.test.TestCase):
[batch, None, None, 32])
images = create_test_input(batch, height, width, 3)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(output, {inputs: images.eval()})
self.assertEqual(output.shape, (batch, 9, 9, 32))
......
research/slim/nets/s3dg.py
View file @ d4eedbb9
......@@ -30,7 +30,9 @@ from tensorflow.contrib import layers as contrib_layers
from nets import i3d_utils
trunc_normal = lambda stddev: tf.truncated_normal_initializer(0.0, stddev)
# pylint: disable=g-long-lambda
trunc_normal = lambda stddev: tf.compat.v1.truncated_normal_initializer(
0.0, stddev)
conv3d_spatiotemporal = i3d_utils.conv3d_spatiotemporal
inception_block_v1_3d = i3d_utils.inception_block_v1_3d
......@@ -209,7 +211,7 @@ def s3dg_base(inputs,
raise ValueError('depth_multiplier is not greater than zero.')
depth = lambda d: max(int(d * depth_multiplier), min_depth)
with tf.variable_scope(scope, 'InceptionV1', [inputs]):
with tf.compat.v1.variable_scope(scope, 'InceptionV1', [inputs]):
with arg_scope([layers.conv3d], weights_initializer=trunc_normal(0.01)):
with arg_scope(
[layers.conv3d, layers.max_pool3d, conv3d_spatiotemporal],
......@@ -556,7 +558,7 @@ def s3dg(inputs,
"""
assert data_format in ['NDHWC', 'NCDHW']
# Final pooling and prediction
with tf.variable_scope(
with tf.compat.v1.variable_scope(
scope, 'InceptionV1', [inputs, num_classes], reuse=reuse) as scope:
with arg_scope(
[layers.batch_norm, layers.dropout], is_training=is_training):
......@@ -570,9 +572,9 @@ def s3dg(inputs,
depth_multiplier=depth_multiplier,
data_format=data_format,
scope=scope)
with tf.variable_scope('Logits'):
with tf.compat.v1.variable_scope('Logits'):
if data_format.startswith('NC'):
net = tf.transpose(net, [0, 2, 3, 4, 1])
net = tf.transpose(a=net, perm=[0, 2, 3, 4, 1])
kernel_size = i3d_utils.reduced_kernel_size_3d(net, [2, 7, 7])
net = layers.avg_pool3d(
net,
......@@ -589,7 +591,7 @@ def s3dg(inputs,
data_format='NDHWC',
scope='Conv2d_0c_1x1')
# Temporal average pooling.
logits = tf.reduce_mean(logits, axis=1)
logits = tf.reduce_mean(input_tensor=logits, axis=1)
if spatial_squeeze:
logits = tf.squeeze(logits, [1, 2], name='SpatialSqueeze')
......
research/slim/nets/s3dg_test.py
View file @ d4eedbb9
......@@ -31,7 +31,7 @@ class S3DGTest(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
inputs = tf.random_uniform((batch_size, num_frames, height, width, 3))
inputs = tf.random.uniform((batch_size, num_frames, height, width, 3))
logits, end_points = s3dg.s3dg(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV1/Logits'))
self.assertListEqual(logits.get_shape().as_list(),
......@@ -45,7 +45,7 @@ class S3DGTest(tf.test.TestCase):
num_frames = 64
height, width = 224, 224
inputs = tf.random_uniform((batch_size, num_frames, height, width, 3))
inputs = tf.random.uniform((batch_size, num_frames, height, width, 3))
mixed_6c, end_points = s3dg.s3dg_base(inputs)
self.assertTrue(mixed_6c.op.name.startswith('InceptionV1/Mixed_5c'))
self.assertListEqual(mixed_6c.get_shape().as_list(),
......@@ -68,7 +68,7 @@ class S3DGTest(tf.test.TestCase):
'Mixed_5c']
for index, endpoint in enumerate(endpoints):
with tf.Graph().as_default():
inputs = tf.random_uniform((batch_size, num_frames, height, width, 3))
inputs = tf.random.uniform((batch_size, num_frames, height, width, 3))
out_tensor, end_points = s3dg.s3dg_base(
inputs, final_endpoint=endpoint, gating_startat=None)
print(endpoint, out_tensor.op.name)
......@@ -81,7 +81,7 @@ class S3DGTest(tf.test.TestCase):
num_frames = 64
height, width = 224, 224
inputs = tf.random_uniform((batch_size, num_frames, height, width, 3))
inputs = tf.random.uniform((batch_size, num_frames, height, width, 3))
_, end_points = s3dg.s3dg_base(inputs,
final_endpoint='Mixed_5c')
endpoints_shapes = {'Conv2d_1a_7x7': [5, 32, 112, 112, 64],
......@@ -112,7 +112,7 @@ class S3DGTest(tf.test.TestCase):
num_frames = 64
height, width = 112, 112
inputs = tf.random_uniform((batch_size, num_frames, height, width, 3))
inputs = tf.random.uniform((batch_size, num_frames, height, width, 3))
mixed_5c, _ = s3dg.s3dg_base(inputs)
self.assertTrue(mixed_5c.op.name.startswith('InceptionV1/Mixed_5c'))
self.assertListEqual(mixed_5c.get_shape().as_list(),
......@@ -123,7 +123,7 @@ class S3DGTest(tf.test.TestCase):
num_frames = 10
height, width = 224, 224
inputs = tf.random_uniform((batch_size, num_frames, height, width, 3))
inputs = tf.random.uniform((batch_size, num_frames, height, width, 3))
mixed_5c, _ = s3dg.s3dg_base(inputs)
self.assertTrue(mixed_5c.op.name.startswith('InceptionV1/Mixed_5c'))
self.assertListEqual(mixed_5c.get_shape().as_list(),
......@@ -135,13 +135,13 @@ class S3DGTest(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
eval_inputs = tf.random_uniform((batch_size, num_frames, height, width, 3))
eval_inputs = tf.random.uniform((batch_size, num_frames, height, width, 3))
logits, _ = s3dg.s3dg(eval_inputs, num_classes,
is_training=False)
predictions = tf.argmax(logits, 1)
predictions = tf.argmax(input=logits, axis=1)
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size,))
......
research/slim/nets/vgg.py
View file @ d4eedbb9
......@@ -59,7 +59,7 @@ def vgg_arg_scope(weight_decay=0.0005):
with slim.arg_scope([slim.conv2d, slim.fully_connected],
activation_fn=tf.nn.relu,
weights_regularizer=slim.l2_regularizer(weight_decay),
biases_initializer=tf.zeros_initializer()):
biases_initializer=tf.compat.v1.zeros_initializer()):
with slim.arg_scope([slim.conv2d], padding='SAME') as arg_sc:
return arg_sc
......@@ -105,7 +105,7 @@ def vgg_a(inputs,
or the input to the logits layer (if num_classes is 0 or None).
end_points: a dict of tensors with intermediate activations.
"""
with tf.variable_scope(scope, 'vgg_a', [inputs], reuse=reuse) as sc:
with tf.compat.v1.variable_scope(scope, 'vgg_a', [inputs], reuse=reuse) as sc:
end_points_collection = sc.original_name_scope + '_end_points'
# Collect outputs for conv2d, fully_connected and max_pool2d.
with slim.arg_scope([slim.conv2d, slim.max_pool2d],
......@@ -129,7 +129,8 @@ def vgg_a(inputs,
# Convert end_points_collection into a end_point dict.
end_points = slim.utils.convert_collection_to_dict(end_points_collection)
if global_pool:
net = tf.reduce_mean(net, [1, 2], keep_dims=True, name='global_pool')
net = tf.reduce_mean(
input_tensor=net, axis=[1, 2], keepdims=True, name='global_pool')
end_points['global_pool'] = net
if num_classes:
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
......@@ -186,7 +187,8 @@ def vgg_16(inputs,
or the input to the logits layer (if num_classes is 0 or None).
end_points: a dict of tensors with intermediate activations.
"""
with tf.variable_scope(scope, 'vgg_16', [inputs], reuse=reuse) as sc:
with tf.compat.v1.variable_scope(
scope, 'vgg_16', [inputs], reuse=reuse) as sc:
end_points_collection = sc.original_name_scope + '_end_points'
# Collect outputs for conv2d, fully_connected and max_pool2d.
with slim.arg_scope([slim.conv2d, slim.fully_connected, slim.max_pool2d],
......@@ -210,7 +212,8 @@ def vgg_16(inputs,
# Convert end_points_collection into a end_point dict.
end_points = slim.utils.convert_collection_to_dict(end_points_collection)
if global_pool:
net = tf.reduce_mean(net, [1, 2], keep_dims=True, name='global_pool')
net = tf.reduce_mean(
input_tensor=net, axis=[1, 2], keepdims=True, name='global_pool')
end_points['global_pool'] = net
if num_classes:
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
......@@ -268,7 +271,8 @@ def vgg_19(inputs,
None).
end_points: a dict of tensors with intermediate activations.
"""
with tf.variable_scope(scope, 'vgg_19', [inputs], reuse=reuse) as sc:
with tf.compat.v1.variable_scope(
scope, 'vgg_19', [inputs], reuse=reuse) as sc:
end_points_collection = sc.original_name_scope + '_end_points'
# Collect outputs for conv2d, fully_connected and max_pool2d.
with slim.arg_scope([slim.conv2d, slim.fully_connected, slim.max_pool2d],
......@@ -292,7 +296,8 @@ def vgg_19(inputs,
# Convert end_points_collection into a end_point dict.
end_points = slim.utils.convert_collection_to_dict(end_points_collection)
if global_pool:
net = tf.reduce_mean(net, [1, 2], keep_dims=True, name='global_pool')
net = tf.reduce_mean(
input_tensor=net, axis=[1, 2], keepdims=True, name='global_pool')
end_points['global_pool'] = net
if num_classes:
net = slim.dropout(net, dropout_keep_prob, is_training=is_training,
......
research/slim/nets/vgg_test.py
View file @ d4eedbb9
......@@ -32,7 +32,7 @@ class VGGATest(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_a(inputs, num_classes)
self.assertEquals(logits.op.name, 'vgg_a/fc8/squeezed')
self.assertListEqual(logits.get_shape().as_list(),
......@@ -43,7 +43,7 @@ class VGGATest(tf.test.TestCase):
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_a(inputs, num_classes, spatial_squeeze=False)
self.assertEquals(logits.op.name, 'vgg_a/fc8/BiasAdd')
self.assertListEqual(logits.get_shape().as_list(),
......@@ -54,7 +54,7 @@ class VGGATest(tf.test.TestCase):
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_a(inputs, num_classes, spatial_squeeze=False,
global_pool=True)
self.assertEquals(logits.op.name, 'vgg_a/fc8/BiasAdd')
......@@ -66,7 +66,7 @@ class VGGATest(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
_, end_points = vgg.vgg_a(inputs, num_classes)
expected_names = ['vgg_a/conv1/conv1_1',
'vgg_a/pool1',
......@@ -92,7 +92,7 @@ class VGGATest(tf.test.TestCase):
height, width = 224, 224
num_classes = None
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
net, end_points = vgg.vgg_a(inputs, num_classes)
expected_names = ['vgg_a/conv1/conv1_1',
'vgg_a/pool1',
......@@ -118,7 +118,7 @@ class VGGATest(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
vgg.vgg_a(inputs, num_classes)
expected_names = ['vgg_a/conv1/conv1_1/weights',
'vgg_a/conv1/conv1_1/biases',
......@@ -151,11 +151,11 @@ class VGGATest(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
eval_inputs = tf.random_uniform((batch_size, height, width, 3))
eval_inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_a(eval_inputs, is_training=False)
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
predictions = tf.argmax(logits, 1)
predictions = tf.argmax(input=logits, axis=1)
self.assertListEqual(predictions.get_shape().as_list(), [batch_size])
def testTrainEvalWithReuse(self):
......@@ -165,29 +165,29 @@ class VGGATest(tf.test.TestCase):
eval_height, eval_width = 256, 256
num_classes = 1000
with self.test_session():
train_inputs = tf.random_uniform(
train_inputs = tf.random.uniform(
(train_batch_size, train_height, train_width, 3))
logits, _ = vgg.vgg_a(train_inputs)
self.assertListEqual(logits.get_shape().as_list(),
[train_batch_size, num_classes])
tf.get_variable_scope().reuse_variables()
eval_inputs = tf.random_uniform(
tf.compat.v1.get_variable_scope().reuse_variables()
eval_inputs = tf.random.uniform(
(eval_batch_size, eval_height, eval_width, 3))
logits, _ = vgg.vgg_a(eval_inputs, is_training=False,
spatial_squeeze=False)
self.assertListEqual(logits.get_shape().as_list(),
[eval_batch_size, 2, 2, num_classes])
logits = tf.reduce_mean(logits, [1, 2])
predictions = tf.argmax(logits, 1)
logits = tf.reduce_mean(input_tensor=logits, axis=[1, 2])
predictions = tf.argmax(input=logits, axis=1)
self.assertEquals(predictions.get_shape().as_list(), [eval_batch_size])
def testForward(self):
batch_size = 1
height, width = 224, 224
with self.test_session() as sess:
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_a(inputs)
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(logits)
self.assertTrue(output.any())
......@@ -199,7 +199,7 @@ class VGG16Test(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_16(inputs, num_classes)
self.assertEquals(logits.op.name, 'vgg_16/fc8/squeezed')
self.assertListEqual(logits.get_shape().as_list(),
......@@ -210,7 +210,7 @@ class VGG16Test(tf.test.TestCase):
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_16(inputs, num_classes, spatial_squeeze=False)
self.assertEquals(logits.op.name, 'vgg_16/fc8/BiasAdd')
self.assertListEqual(logits.get_shape().as_list(),
......@@ -221,7 +221,7 @@ class VGG16Test(tf.test.TestCase):
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_16(inputs, num_classes, spatial_squeeze=False,
global_pool=True)
self.assertEquals(logits.op.name, 'vgg_16/fc8/BiasAdd')
......@@ -233,7 +233,7 @@ class VGG16Test(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
_, end_points = vgg.vgg_16(inputs, num_classes)
expected_names = ['vgg_16/conv1/conv1_1',
'vgg_16/conv1/conv1_2',
......@@ -264,7 +264,7 @@ class VGG16Test(tf.test.TestCase):
height, width = 224, 224
num_classes = None
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
net, end_points = vgg.vgg_16(inputs, num_classes)
expected_names = ['vgg_16/conv1/conv1_1',
'vgg_16/conv1/conv1_2',
......@@ -295,7 +295,7 @@ class VGG16Test(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
vgg.vgg_16(inputs, num_classes)
expected_names = ['vgg_16/conv1/conv1_1/weights',
'vgg_16/conv1/conv1_1/biases',
......@@ -338,11 +338,11 @@ class VGG16Test(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
eval_inputs = tf.random_uniform((batch_size, height, width, 3))
eval_inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_16(eval_inputs, is_training=False)
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
predictions = tf.argmax(logits, 1)
predictions = tf.argmax(input=logits, axis=1)
self.assertListEqual(predictions.get_shape().as_list(), [batch_size])
def testTrainEvalWithReuse(self):
......@@ -352,29 +352,29 @@ class VGG16Test(tf.test.TestCase):
eval_height, eval_width = 256, 256
num_classes = 1000
with self.test_session():
train_inputs = tf.random_uniform(
train_inputs = tf.random.uniform(
(train_batch_size, train_height, train_width, 3))
logits, _ = vgg.vgg_16(train_inputs)
self.assertListEqual(logits.get_shape().as_list(),
[train_batch_size, num_classes])
tf.get_variable_scope().reuse_variables()
eval_inputs = tf.random_uniform(
tf.compat.v1.get_variable_scope().reuse_variables()
eval_inputs = tf.random.uniform(
(eval_batch_size, eval_height, eval_width, 3))
logits, _ = vgg.vgg_16(eval_inputs, is_training=False,
spatial_squeeze=False)
self.assertListEqual(logits.get_shape().as_list(),
[eval_batch_size, 2, 2, num_classes])
logits = tf.reduce_mean(logits, [1, 2])
predictions = tf.argmax(logits, 1)
logits = tf.reduce_mean(input_tensor=logits, axis=[1, 2])
predictions = tf.argmax(input=logits, axis=1)
self.assertEquals(predictions.get_shape().as_list(), [eval_batch_size])
def testForward(self):
batch_size = 1
height, width = 224, 224
with self.test_session() as sess:
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_16(inputs)
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(logits)
self.assertTrue(output.any())
......@@ -386,7 +386,7 @@ class VGG19Test(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_19(inputs, num_classes)
self.assertEquals(logits.op.name, 'vgg_19/fc8/squeezed')
self.assertListEqual(logits.get_shape().as_list(),
......@@ -397,7 +397,7 @@ class VGG19Test(tf.test.TestCase):
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_19(inputs, num_classes, spatial_squeeze=False)
self.assertEquals(logits.op.name, 'vgg_19/fc8/BiasAdd')
self.assertListEqual(logits.get_shape().as_list(),
......@@ -408,7 +408,7 @@ class VGG19Test(tf.test.TestCase):
height, width = 256, 256
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_19(inputs, num_classes, spatial_squeeze=False,
global_pool=True)
self.assertEquals(logits.op.name, 'vgg_19/fc8/BiasAdd')
......@@ -420,7 +420,7 @@ class VGG19Test(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
_, end_points = vgg.vgg_19(inputs, num_classes)
expected_names = [
'vgg_19/conv1/conv1_1',
......@@ -455,7 +455,7 @@ class VGG19Test(tf.test.TestCase):
height, width = 224, 224
num_classes = None
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
net, end_points = vgg.vgg_19(inputs, num_classes)
expected_names = [
'vgg_19/conv1/conv1_1',
......@@ -490,7 +490,7 @@ class VGG19Test(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
vgg.vgg_19(inputs, num_classes)
expected_names = [
'vgg_19/conv1/conv1_1/weights',
......@@ -540,11 +540,11 @@ class VGG19Test(tf.test.TestCase):
height, width = 224, 224
num_classes = 1000
with self.test_session():
eval_inputs = tf.random_uniform((batch_size, height, width, 3))
eval_inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_19(eval_inputs, is_training=False)
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
predictions = tf.argmax(logits, 1)
predictions = tf.argmax(input=logits, axis=1)
self.assertListEqual(predictions.get_shape().as_list(), [batch_size])
def testTrainEvalWithReuse(self):
......@@ -554,29 +554,29 @@ class VGG19Test(tf.test.TestCase):
eval_height, eval_width = 256, 256
num_classes = 1000
with self.test_session():
train_inputs = tf.random_uniform(
train_inputs = tf.random.uniform(
(train_batch_size, train_height, train_width, 3))
logits, _ = vgg.vgg_19(train_inputs)
self.assertListEqual(logits.get_shape().as_list(),
[train_batch_size, num_classes])
tf.get_variable_scope().reuse_variables()
eval_inputs = tf.random_uniform(
tf.compat.v1.get_variable_scope().reuse_variables()
eval_inputs = tf.random.uniform(
(eval_batch_size, eval_height, eval_width, 3))
logits, _ = vgg.vgg_19(eval_inputs, is_training=False,
spatial_squeeze=False)
self.assertListEqual(logits.get_shape().as_list(),
[eval_batch_size, 2, 2, num_classes])
logits = tf.reduce_mean(logits, [1, 2])
predictions = tf.argmax(logits, 1)
logits = tf.reduce_mean(input_tensor=logits, axis=[1, 2])
predictions = tf.argmax(input=logits, axis=1)
self.assertEquals(predictions.get_shape().as_list(), [eval_batch_size])
def testForward(self):
batch_size = 1
height, width = 224, 224
with self.test_session() as sess:
inputs = tf.random_uniform((batch_size, height, width, 3))
inputs = tf.random.uniform((batch_size, height, width, 3))
logits, _ = vgg.vgg_19(inputs)
sess.run(tf.global_variables_initializer())
sess.run(tf.compat.v1.global_variables_initializer())
output = sess.run(logits)
self.assertTrue(output.any())
......
research/slim/preprocessing/preprocessing_factory.py
View file @ d4eedbb9
......@@ -56,6 +56,12 @@ def get_preprocessing(name, is_training=False, use_grayscale=False):
'mobilenet_v1': inception_preprocessing,
'mobilenet_v2': inception_preprocessing,
'mobilenet_v2_035': inception_preprocessing,
'mobilenet_v3_small': inception_preprocessing,
'mobilenet_v3_large': inception_preprocessing,
'mobilenet_v3_small_minimalistic': inception_preprocessing,
'mobilenet_v3_large_minimalistic': inception_preprocessing,
'mobilenet_edgetpu': inception_preprocessing,
'mobilenet_edgetpu_075': inception_preprocessing,
'mobilenet_v2_140': inception_preprocessing,
'nasnet_mobile': inception_preprocessing,
'nasnet_large': inception_preprocessing,
......
research/slim/train_image_classifier.py
View file @ d4eedbb9
......@@ -35,6 +35,10 @@ tf.app.flags.DEFINE_string(
tf.app.flags.DEFINE_string(
'train_dir', '/tmp/tfmodel/',
'Directory where checkpoints and event logs are written to.')
tf.app.flags.DEFINE_float(
'warmup_epochs', 0,
'Linearly warmup learning rate from 0 to learning_rate over this '
'many epochs.')
tf.app.flags.DEFINE_integer('num_clones', 1,
'Number of model clones to deploy. Note For '
......@@ -252,33 +256,42 @@ def _configure_learning_rate(num_samples_per_epoch, global_step):
# Note: when num_clones is > 1, this will actually have each clone to go
# over each epoch FLAGS.num_epochs_per_decay times. This is different
# behavior from sync replicas and is expected to produce different results.
decay_steps = int(num_samples_per_epoch * FLAGS.num_epochs_per_decay /
FLAGS.batch_size)
steps_per_epoch = num_samples_per_epoch / FLAGS.batch_size
if FLAGS.sync_replicas:
decay_steps /= FLAGS.replicas_to_aggregate
steps_per_epoch /= FLAGS.replicas_to_aggregate
decay_steps = int(steps_per_epoch * FLAGS.num_epochs_per_decay)
if FLAGS.learning_rate_decay_type == 'exponential':
return tf.train.exponential_decay(FLAGS.learning_rate,
global_step,
decay_steps,
FLAGS.learning_rate_decay_factor,
staircase=True,
name='exponential_decay_learning_rate')
learning_rate = tf.train.exponential_decay(
FLAGS.learning_rate,
global_step,
decay_steps,
FLAGS.learning_rate_decay_factor,
staircase=True,
name='exponential_decay_learning_rate')
elif FLAGS.learning_rate_decay_type == 'fixed':
return tf.constant(FLAGS.learning_rate, name='fixed_learning_rate')
learning_rate = tf.constant(FLAGS.learning_rate, name='fixed_learning_rate')
elif FLAGS.learning_rate_decay_type == 'polynomial':
return tf.train.polynomial_decay(FLAGS.learning_rate,
global_step,
decay_steps,
FLAGS.end_learning_rate,
power=1.0,
cycle=False,
name='polynomial_decay_learning_rate')
learning_rate = tf.train.polynomial_decay(
FLAGS.learning_rate,
global_step,
decay_steps,
FLAGS.end_learning_rate,
power=1.0,
cycle=False,
name='polynomial_decay_learning_rate')
else:
raise ValueError('learning_rate_decay_type [%s] was not recognized' %
FLAGS.learning_rate_decay_type)
if FLAGS.warmup_epochs:
warmup_lr = (
FLAGS.learning_rate * tf.cast(global_step, tf.float32) /
(steps_per_epoch * FLAGS.warmup_epochs))
learning_rate = tf.minimum(warmup_lr, learning_rate)
return learning_rate
def _configure_optimizer(learning_rate):
"""Configures the optimizer used for training.
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment