Merge branch 'master' of https://github.com/tensorflow/models into run_superglue

official/vision/beta/__init__.py

-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -11,7 +11,7 @@
 # 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.
-# ==============================================================================
+
 """Vision package definition."""
 # Lint as: python3
 # pylint: disable=unused-import

official/vision/beta/configs/__init__.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Configs package definition."""
 
 from official.vision.beta.configs import image_classification

official/vision/beta/configs/backbones.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Backbones configurations."""
 from typing import Optional, List
 
@@ -69,6 +69,15 @@ class SpineNet(hyperparams.Config):
   stochastic_depth_drop_rate: float = 0.0
 
 
+@dataclasses.dataclass
+class SpineNetMobile(hyperparams.Config):
+  """SpineNet config."""
+  model_id: str = '49'
+  stochastic_depth_drop_rate: float = 0.0
+  se_ratio: float = 0.2
+  expand_ratio: int = 6
+
+
 @dataclasses.dataclass
 class RevNet(hyperparams.Config):
   """RevNet config."""
@@ -81,12 +90,13 @@ class Backbone(hyperparams.OneOfConfig):
   """Configuration for backbones.
 
   Attributes:
-    type: 'str', type of backbone be used, one the of fields below.
+    type: 'str', type of backbone be used, one of the fields below.
     resnet: resnet backbone config.
     dilated_resnet: dilated resnet backbone for semantic segmentation config.
     revnet: revnet backbone config.
     efficientnet: efficientnet backbone config.
     spinenet: spinenet backbone config.
+    spinenet_mobile: mobile spinenet backbone config.
     mobilenet: mobilenet backbone config.
   """
   type: Optional[str] = None
@@ -95,4 +105,5 @@ class Backbone(hyperparams.OneOfConfig):
   revnet: RevNet = RevNet()
   efficientnet: EfficientNet = EfficientNet()
   spinenet: SpineNet = SpineNet()
+  spinenet_mobile: SpineNetMobile = SpineNetMobile()
   mobilenet: MobileNet = MobileNet()

official/vision/beta/configs/backbones_3d.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """3D Backbones configurations."""
 from typing import Optional, Tuple
 
@@ -65,7 +65,7 @@ class Backbone3D(hyperparams.OneOfConfig):
   """Configuration for backbones.
 
   Attributes:
-    type: 'str', type of backbone be used, on the of fields below.
+    type: 'str', type of backbone be used, one of the fields below.
     resnet: resnet3d backbone config.
   """
   type: Optional[str] = None

official/vision/beta/configs/common.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Common configurations."""
 
 # Import libraries

official/vision/beta/configs/decoders.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Decoders configurations."""
 from typing import Optional, List
 
@@ -58,7 +58,7 @@ class Decoder(hyperparams.OneOfConfig):
   """Configuration for decoders.
 
   Attributes:
-    type: 'str', type of decoder be used, on the of fields below.
+    type: 'str', type of decoder be used, one of the fields below.
     fpn: fpn config.
   """
   type: Optional[str] = None

official/vision/beta/configs/experiments/retinanet/coco_spinenet49_mobile_tpu.yaml

+runtime:
+  distribution_strategy: 'tpu'
+  mixed_precision_dtype: 'bfloat16'
+task:
+  losses:
+    l2_weight_decay: 3.0e-05
+  model:
+    anchor:
+      anchor_size: 3
+      aspect_ratios: [0.5, 1.0, 2.0]
+      num_scales: 3
+    backbone:
+      spinenet_mobile:
+        stochastic_depth_drop_rate: 0.2
+        model_id: '49'
+        se_ratio: 0.2
+      type: 'spinenet_mobile'
+    decoder:
+      type: 'identity'
+    head:
+      num_convs: 4
+      num_filters: 48
+      use_separable_conv: true
+    input_size: [384, 384, 3]
+    max_level: 7
+    min_level: 3
+    norm_activation:
+      activation: 'swish'
+      norm_epsilon: 0.001
+      norm_momentum: 0.99
+      use_sync_bn: true
+  train_data:
+    dtype: 'bfloat16'
+    global_batch_size: 256
+    is_training: true
+    parser:
+      aug_rand_hflip: true
+      aug_scale_max: 2.0
+      aug_scale_min: 0.1
+  validation_data:
+    dtype: 'bfloat16'
+    global_batch_size: 8
+    is_training: false
+trainer:
+  checkpoint_interval: 462
+  optimizer_config:
+    learning_rate:
+      stepwise:
+        boundaries: [265650, 272580]
+        values: [0.32, 0.032, 0.0032]
+      type: 'stepwise'
+    warmup:
+      linear:
+        warmup_learning_rate: 0.0067
+        warmup_steps: 2000
+  steps_per_loop: 462
+  train_steps: 277200
+  validation_interval: 462
+  validation_steps: 625

official/vision/beta/configs/experiments/retinanet/coco_spinenet49s_mobile_tpu.yaml

+runtime:
+  distribution_strategy: 'tpu'
+  mixed_precision_dtype: 'bfloat16'
+task:
+  losses:
+    l2_weight_decay: 3.0e-05
+  model:
+    anchor:
+      anchor_size: 3
+      aspect_ratios: [0.5, 1.0, 2.0]
+      num_scales: 3
+    backbone:
+      spinenet_mobile:
+        stochastic_depth_drop_rate: 0.2
+        model_id: '49S'
+        se_ratio: 0.2
+      type: 'spinenet_mobile'
+    decoder:
+      type: 'identity'
+    head:
+      num_convs: 4
+      num_filters: 40
+      use_separable_conv: true
+    input_size: [384, 384, 3]
+    max_level: 7
+    min_level: 3
+    norm_activation:
+      activation: 'swish'
+      norm_epsilon: 0.001
+      norm_momentum: 0.99
+      use_sync_bn: true
+  train_data:
+    dtype: 'bfloat16'
+    global_batch_size: 256
+    is_training: true
+    parser:
+      aug_rand_hflip: true
+      aug_scale_max: 2.0
+      aug_scale_min: 0.1
+  validation_data:
+    dtype: 'bfloat16'
+    global_batch_size: 8
+    is_training: false
+trainer:
+  checkpoint_interval: 462
+  optimizer_config:
+    learning_rate:
+      stepwise:
+        boundaries: [265650, 272580]
+        values: [0.32, 0.032, 0.0032]
+      type: 'stepwise'
+    warmup:
+      linear:
+        warmup_learning_rate: 0.0067
+        warmup_steps: 2000
+  steps_per_loop: 462
+  train_steps: 277200
+  validation_interval: 462
+  validation_steps: 625

official/vision/beta/configs/experiments/retinanet/coco_spinenet49xs_mobile_tpu.yaml

+runtime:
+  distribution_strategy: 'tpu'
+  mixed_precision_dtype: 'bfloat16'
+task:
+  losses:
+    l2_weight_decay: 3.0e-05
+  model:
+    anchor:
+      anchor_size: 3
+      aspect_ratios: [0.5, 1.0, 2.0]
+      num_scales: 3
+    backbone:
+      spinenet_mobile:
+        stochastic_depth_drop_rate: 0.2
+        model_id: '49XS'
+        se_ratio: 0.2
+      type: 'spinenet_mobile'
+    decoder:
+      type: 'identity'
+    head:
+      num_convs: 4
+      num_filters: 24
+      use_separable_conv: false
+    input_size: [256, 256, 3]
+    max_level: 7
+    min_level: 3
+    norm_activation:
+      activation: 'swish'
+      norm_epsilon: 0.001
+      norm_momentum: 0.99
+      use_sync_bn: true
+  train_data:
+    dtype: 'bfloat16'
+    global_batch_size: 256
+    is_training: true
+    parser:
+      aug_rand_hflip: true
+      aug_scale_max: 2.0
+      aug_scale_min: 0.1
+  validation_data:
+    dtype: 'bfloat16'
+    global_batch_size: 8
+    is_training: false
+trainer:
+  checkpoint_interval: 462
+  optimizer_config:
+    learning_rate:
+      stepwise:
+        boundaries: [265650, 272580]
+        values: [0.32, 0.032, 0.0032]
+      type: 'stepwise'
+    warmup:
+      linear:
+        warmup_learning_rate: 0.0067
+        warmup_steps: 2000
+  steps_per_loop: 462
+  train_steps: 277200
+  validation_interval: 462
+  validation_steps: 625

official/vision/beta/configs/image_classification.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Image classification configuration definition."""
 import os
 from typing import List, Optional

official/vision/beta/configs/image_classification_test.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Tests for image_classification."""
 # pylint: disable=unused-import
 from absl.testing import parameterized

official/vision/beta/configs/maskrcnn.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Mask R-CNN configuration definition."""
 
 import os

official/vision/beta/configs/retinanet.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """RetinaNet configuration definition."""
 
 import os
@@ -300,3 +300,85 @@ def retinanet_spinenet_coco() -> cfg.ExperimentConfig:
       ])
 
   return config
+
+
+@exp_factory.register_config_factory('retinanet_spinenet_mobile_coco')
+def retinanet_spinenet_mobile_coco() -> cfg.ExperimentConfig:
+  """COCO object detection with RetinaNet using Mobile SpineNet backbone."""
+  train_batch_size = 256
+  eval_batch_size = 8
+  steps_per_epoch = COCO_TRAIN_EXAMPLES // train_batch_size
+  input_size = 384
+
+  config = cfg.ExperimentConfig(
+      runtime=cfg.RuntimeConfig(mixed_precision_dtype='float32'),
+      task=RetinaNetTask(
+          annotation_file=os.path.join(COCO_INPUT_PATH_BASE,
+                                       'instances_val2017.json'),
+          model=RetinaNet(
+              backbone=backbones.Backbone(
+                  type='spinenet_mobile',
+                  spinenet_mobile=backbones.SpineNetMobile(
+                      model_id='49', stochastic_depth_drop_rate=0.2)),
+              decoder=decoders.Decoder(
+                  type='identity', identity=decoders.Identity()),
+              head=RetinaNetHead(num_filters=48, use_separable_conv=True),
+              anchor=Anchor(anchor_size=3),
+              norm_activation=common.NormActivation(
+                  use_sync_bn=True, activation='swish'),
+              num_classes=91,
+              input_size=[input_size, input_size, 3],
+              min_level=3,
+              max_level=7),
+          losses=Losses(l2_weight_decay=3e-5),
+          train_data=DataConfig(
+              input_path=os.path.join(COCO_INPUT_PATH_BASE, 'train*'),
+              is_training=True,
+              global_batch_size=train_batch_size,
+              parser=Parser(
+                  aug_rand_hflip=True, aug_scale_min=0.1, aug_scale_max=2.0)),
+          validation_data=DataConfig(
+              input_path=os.path.join(COCO_INPUT_PATH_BASE, 'val*'),
+              is_training=False,
+              global_batch_size=eval_batch_size)),
+      trainer=cfg.TrainerConfig(
+          train_steps=600 * steps_per_epoch,
+          validation_steps=COCO_VAL_EXAMPLES // eval_batch_size,
+          validation_interval=steps_per_epoch,
+          steps_per_loop=steps_per_epoch,
+          summary_interval=steps_per_epoch,
+          checkpoint_interval=steps_per_epoch,
+          optimizer_config=optimization.OptimizationConfig({
+              'optimizer': {
+                  'type': 'sgd',
+                  'sgd': {
+                      'momentum': 0.9
+                  }
+              },
+              'learning_rate': {
+                  'type': 'stepwise',
+                  'stepwise': {
+                      'boundaries': [
+                          575 * steps_per_epoch, 590 * steps_per_epoch
+                      ],
+                      'values': [
+                          0.32 * train_batch_size / 256.0,
+                          0.032 * train_batch_size / 256.0,
+                          0.0032 * train_batch_size / 256.0
+                      ],
+                  }
+              },
+              'warmup': {
+                  'type': 'linear',
+                  'linear': {
+                      'warmup_steps': 2000,
+                      'warmup_learning_rate': 0.0067
+                  }
+              }
+          })),
+      restrictions=[
+          'task.train_data.is_training != None',
+          'task.validation_data.is_training != None'
+      ])
+
+  return config

official/vision/beta/configs/semantic_segmentation.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Semantic segmentation configuration definition."""
 import os
 from typing import List, Optional, Union
@@ -405,7 +405,9 @@ def seg_deeplabv3plus_cityscapes() -> cfg.ExperimentConfig:
   config = cfg.ExperimentConfig(
       task=SemanticSegmentationTask(
           model=SemanticSegmentationModel(
-              num_classes=20,
+              # Cityscapes uses only 19 semantic classes for train/evaluation.
+              # The void (background) class is ignored in train and evaluation.
+              num_classes=19,
               input_size=[None, None, 3],
               backbone=backbones.Backbone(
                   type='dilated_resnet', dilated_resnet=backbones.DilatedResNet(

official/vision/beta/configs/semantic_segmentation_test.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Tests for semantic_segmentation."""
 
 # pylint: disable=unused-import

official/vision/beta/configs/video_classification.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Video classification configuration definition."""
 from typing import Optional, Tuple
 import dataclasses

official/vision/beta/configs/video_classification_test.py

-# Lint as: python3
-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -12,7 +11,8 @@
 # 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
 """Tests for video_classification."""
 
 # pylint: disable=unused-import

official/vision/beta/data/__init__.py

+# Copyright 2021 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/vision/beta/data/create_coco_tf_record.py

-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -11,7 +11,7 @@
 # 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.
-# ==============================================================================
+
 r"""Convert raw COCO dataset to TFRecord format.
 
 This scripts follows the label map decoder format and supports detection

official/vision/beta/data/tfrecord_lib.py

-# Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+# Copyright 2021 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.
@@ -11,7 +11,7 @@
 # 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.
-# ==============================================================================
+
 """Helper functions for creating TFRecord datasets."""
 
 import hashlib