Merge branch 'tensorflow:master' into panoptic-deeplab-modeling

official/legacy/image_classification/resnet/resnet_runnable.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/image_classification/resnet/tfhub_export.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/image_classification/test_utils.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/image_classification/vgg/__init__.py

+# Copyright 2022 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.
+
+

official/legacy/image_classification/vgg/vgg_config.py

+# Copyright 2022 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.
+
+# Lint as: python3
+"""Configuration definitions for VGG losses, learning rates, and optimizers."""
+
+import dataclasses
+from official.legacy.image_classification.configs import base_configs
+from official.modeling.hyperparams import base_config
+
+
+@dataclasses.dataclass
+class VGGModelConfig(base_configs.ModelConfig):
+  """Configuration for the VGG model."""
+  name: str = 'VGG'
+  num_classes: int = 1000
+  model_params: base_config.Config = dataclasses.field(default_factory=lambda: {   # pylint:disable=g-long-lambda
+      'num_classes': 1000,
+      'batch_size': None,
+      'use_l2_regularizer': True
+  })
+  loss: base_configs.LossConfig = base_configs.LossConfig(
+      name='sparse_categorical_crossentropy')
+  optimizer: base_configs.OptimizerConfig = base_configs.OptimizerConfig(
+      name='momentum', epsilon=0.001, momentum=0.9, moving_average_decay=None)
+  learning_rate: base_configs.LearningRateConfig = (
+      base_configs.LearningRateConfig(
+          name='stepwise',
+          initial_lr=0.01,
+          examples_per_epoch=1281167,
+          boundaries=[30, 60],
+          warmup_epochs=0,
+          scale_by_batch_size=1. / 256.,
+          multipliers=[0.01 / 256, 0.001 / 256, 0.0001 / 256]))

official/legacy/image_classification/vgg/vgg_model.py

+# Copyright 2022 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.
+
+"""VGG16 model for Keras.
+
+Adapted from tf.keras.applications.vgg16.VGG16().
+
+Related papers/blogs:
+- https://arxiv.org/abs/1409.1556
+"""
+
+import tensorflow as tf
+
+layers = tf.keras.layers
+
+
+def _gen_l2_regularizer(use_l2_regularizer=True, l2_weight_decay=1e-4):
+  return tf.keras.regularizers.L2(
+      l2_weight_decay) if use_l2_regularizer else None
+
+
+def vgg16(num_classes,
+          batch_size=None,
+          use_l2_regularizer=True,
+          batch_norm_decay=0.9,
+          batch_norm_epsilon=1e-5):
+  """Instantiates the VGG16 architecture.
+
+  Args:
+    num_classes: `int` number of classes for image classification.
+    batch_size: Size of the batches for each step.
+    use_l2_regularizer: whether to use L2 regularizer on Conv/Dense layer.
+    batch_norm_decay: Moment of batch norm layers.
+    batch_norm_epsilon: Epsilon of batch borm layers.
+
+  Returns:
+    A Keras model instance.
+
+  """
+  input_shape = (224, 224, 3)
+  img_input = layers.Input(shape=input_shape, batch_size=batch_size)
+
+  x = img_input
+
+  if tf.keras.backend.image_data_format() == 'channels_first':
+    x = layers.Permute((3, 1, 2))(x)
+    bn_axis = 1
+  else:  # channels_last
+    bn_axis = 3
+  # Block 1
+  x = layers.Conv2D(
+      64, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block1_conv1')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv1')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Conv2D(
+      64, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block1_conv2')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv2')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block1_pool')(x)
+
+  # Block 2
+  x = layers.Conv2D(
+      128, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block2_conv1')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv3')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Conv2D(
+      128, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block2_conv2')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv4')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block2_pool')(x)
+
+  # Block 3
+  x = layers.Conv2D(
+      256, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block3_conv1')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv5')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Conv2D(
+      256, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block3_conv2')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv6')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Conv2D(
+      256, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block3_conv3')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv7')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block3_pool')(x)
+
+  # Block 4
+  x = layers.Conv2D(
+      512, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block4_conv1')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv8')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Conv2D(
+      512, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block4_conv2')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv9')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Conv2D(
+      512, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block4_conv3')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv10')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block4_pool')(x)
+
+  # Block 5
+  x = layers.Conv2D(
+      512, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block5_conv1')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv11')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Conv2D(
+      512, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block5_conv2')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv12')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Conv2D(
+      512, (3, 3),
+      padding='same',
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='block5_conv3')(
+          x)
+  x = layers.BatchNormalization(
+      axis=bn_axis,
+      momentum=batch_norm_decay,
+      epsilon=batch_norm_epsilon,
+      name='bn_conv13')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.MaxPooling2D((2, 2), strides=(2, 2), name='block5_pool')(x)
+
+  x = layers.Flatten(name='flatten')(x)
+  x = layers.Dense(
+      4096,
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='fc1')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Dropout(0.5)(x)
+  x = layers.Dense(
+      4096,
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='fc2')(
+          x)
+  x = layers.Activation('relu')(x)
+  x = layers.Dropout(0.5)(x)
+  x = layers.Dense(
+      num_classes,
+      kernel_regularizer=_gen_l2_regularizer(use_l2_regularizer),
+      name='fc1000')(
+          x)
+
+  x = layers.Activation('softmax', dtype='float32')(x)
+
+  # Create model.
+  return tf.keras.Model(img_input, x, name='vgg16')

official/legacy/transformer/__init__.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/attention_layer.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/beam_search_v1.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/compute_bleu.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/compute_bleu_test.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/data_download.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/data_pipeline.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/embedding_layer.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/ffn_layer.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/metrics.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/misc.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/model_params.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/model_utils.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.

official/legacy/transformer/model_utils_test.py

-# Copyright 2021 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2022 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.