tf-models-nightly 2.17.0.dev20240313__py2.py3-none-any.whl → 2.17.0.dev20240320__py2.py3-none-any.whl

official/nlp/modeling/layers/__init__.py

@@ -46,6 +46,7 @@ from official.nlp.modeling.layers.moe import MoeLayerWithBackbone
 from official.nlp.modeling.layers.multi_channel_attention import *
 from official.nlp.modeling.layers.on_device_embedding import OnDeviceEmbedding
 from official.nlp.modeling.layers.pack_optimization import PackBertEmbeddings
+from official.nlp.modeling.layers.pack_optimization import StridedReZeroTransformer
 from official.nlp.modeling.layers.pack_optimization import StridedTransformerEncoderBlock
 from official.nlp.modeling.layers.pack_optimization import StridedTransformerScaffold
 from official.nlp.modeling.layers.per_dim_scale_attention import PerDimScaleAttention

official/vision/configs/semantic_segmentation.py

@@ -91,6 +91,10 @@ class DataConfig(cfg.DataConfig):
   )
   additional_dense_features: List[DenseFeatureConfig] = dataclasses.field(
       default_factory=list)
+  # If `centered_crop` is set to True, then resized crop
+  # (if smaller than padded size) is place in the center of the image.
+  # Default behaviour is to place it at left top corner.
+  centered_crop: bool = False
 
 
 @dataclasses.dataclass

official/vision/dataloaders/segmentation_input.py

@@ -25,48 +25,54 @@ from official.vision.ops import preprocess_ops
 class Decoder(decoder.Decoder):
   """A tf.Example decoder for segmentation task."""
 
-  def __init__(self,
-               image_feature=config_lib.DenseFeatureConfig(),
-               additional_dense_features=None):
+  def __init__(
+      self,
+      image_feature=config_lib.DenseFeatureConfig(),
+      additional_dense_features=None,
+  ):
     self._keys_to_features = {
-        'image/encoded':
-            tf.io.FixedLenFeature((), tf.string, default_value=''),
-        'image/height':
-            tf.io.FixedLenFeature((), tf.int64, default_value=0),
-        'image/width':
-            tf.io.FixedLenFeature((), tf.int64, default_value=0),
-        'image/segmentation/class/encoded':
-            tf.io.FixedLenFeature((), tf.string, default_value=''),
-        image_feature.feature_name:
-            tf.io.FixedLenFeature((), tf.string, default_value='')
+        'image/encoded': tf.io.FixedLenFeature((), tf.string, default_value=''),
+        'image/height': tf.io.FixedLenFeature((), tf.int64, default_value=0),
+        'image/width': tf.io.FixedLenFeature((), tf.int64, default_value=0),
+        'image/segmentation/class/encoded': tf.io.FixedLenFeature(
+            (), tf.string, default_value=''
+        ),
+        image_feature.feature_name: tf.io.FixedLenFeature(
+            (), tf.string, default_value=''
+        ),
     }
     if additional_dense_features:
       for feature in additional_dense_features:
         self._keys_to_features[feature.feature_name] = tf.io.FixedLenFeature(
-            (), tf.string, default_value='')
+            (), tf.string, default_value=''
+        )
 
   def decode(self, serialized_example):
-    return tf.io.parse_single_example(serialized_example,
-                                      self._keys_to_features)
+    return tf.io.parse_single_example(
+        serialized_example, self._keys_to_features
+    )
 
 
 class Parser(parser.Parser):
   """Parser to parse an image and its annotations into a dictionary of tensors."""
 
-  def __init__(self,
-               output_size,
-               crop_size=None,
-               resize_eval_groundtruth=True,
-               gt_is_matting_map=False,
-               groundtruth_padded_size=None,
-               ignore_label=255,
-               aug_rand_hflip=False,
-               preserve_aspect_ratio=True,
-               aug_scale_min=1.0,
-               aug_scale_max=1.0,
-               dtype='float32',
-               image_feature=config_lib.DenseFeatureConfig(),
-               additional_dense_features=None):
+  def __init__(
+      self,
+      output_size,
+      crop_size=None,
+      resize_eval_groundtruth=True,
+      gt_is_matting_map=False,
+      groundtruth_padded_size=None,
+      ignore_label=255,
+      aug_rand_hflip=False,
+      preserve_aspect_ratio=True,
+      aug_scale_min=1.0,
+      aug_scale_max=1.0,
+      dtype='float32',
+      image_feature=config_lib.DenseFeatureConfig(),
+      additional_dense_features=None,
+      centered_crop=False,
+  ):
     """Initializes parameters for parsing annotations in the dataset.
 
     Args:
@@ -100,13 +106,18 @@ class Parser(parser.Parser):
         dataset mean/stddev.
       additional_dense_features: `list` of DenseFeatureConfig for additional
         dense features.
+      centered_crop: If `centered_crop` is set to True, then resized crop (if
+        smaller than padded size) is place in the center of the image. Default
+        behaviour is to place it at left top corner.
     """
     self._output_size = output_size
     self._crop_size = crop_size
     self._resize_eval_groundtruth = resize_eval_groundtruth
     if (not resize_eval_groundtruth) and (groundtruth_padded_size is None):
-      raise ValueError('groundtruth_padded_size ([height, width]) needs to be'
-                       'specified when resize_eval_groundtruth is False.')
+      raise ValueError(
+          'groundtruth_padded_size ([height, width]) needs to be'
+          'specified when resize_eval_groundtruth is False.'
+      )
     self._gt_is_matting_map = gt_is_matting_map
     self._groundtruth_padded_size = groundtruth_padded_size
     self._ignore_label = ignore_label
@@ -122,6 +133,12 @@ class Parser(parser.Parser):
 
     self._image_feature = image_feature
     self._additional_dense_features = additional_dense_features
+    self._centered_crop = centered_crop
+    if self._centered_crop and not self._resize_eval_groundtruth:
+      raise ValueError(
+          'centered_crop is only supported when resize_eval_groundtruth is'
+          ' True.'
+      )
 
   def _prepare_image_and_label(self, data):
     """Prepare normalized image and label."""
@@ -129,21 +146,25 @@ class Parser(parser.Parser):
     width = data['image/width']
 
     label = tf.io.decode_image(
-        data['image/segmentation/class/encoded'], channels=1)
+        data['image/segmentation/class/encoded'], channels=1
+    )
     label = tf.reshape(label, (1, height, width))
     label = tf.cast(label, tf.float32)
 
     image = tf.io.decode_image(
         data[self._image_feature.feature_name],
         channels=self._image_feature.num_channels,
-        dtype=tf.uint8)
+        dtype=tf.uint8,
+    )
     image = tf.reshape(image, (height, width, self._image_feature.num_channels))
     # Normalizes the image feature with mean and std values, which are divided
     # by 255 because an uint8 image are re-scaled automatically. Images other
     # than uint8 type will be wrongly normalized.
     image = preprocess_ops.normalize_image(
-        image, [mean / 255.0 for mean in self._image_feature.mean],
-        [stddev / 255.0 for stddev in self._image_feature.stddev])
+        image,
+        [mean / 255.0 for mean in self._image_feature.mean],
+        [stddev / 255.0 for stddev in self._image_feature.stddev],
+    )
 
     if self._additional_dense_features:
       input_list = [image]
@@ -151,11 +172,14 @@ class Parser(parser.Parser):
         feature = tf.io.decode_image(
             data[feature_cfg.feature_name],
             channels=feature_cfg.num_channels,
-            dtype=tf.uint8)
+            dtype=tf.uint8,
+        )
         feature = tf.reshape(feature, (height, width, feature_cfg.num_channels))
         feature = preprocess_ops.normalize_image(
-            feature, [mean / 255.0 for mean in feature_cfg.mean],
-            [stddev / 255.0 for stddev in feature_cfg.stddev])
+            feature,
+            [mean / 255.0 for mean in feature_cfg.mean],
+            [stddev / 255.0 for stddev in feature_cfg.stddev],
+        )
         input_list.append(feature)
       concat_input = tf.concat(input_list, axis=2)
     else:
@@ -164,7 +188,8 @@ class Parser(parser.Parser):
     if not self._preserve_aspect_ratio:
       label = tf.reshape(label, [data['image/height'], data['image/width'], 1])
       concat_input = tf.image.resize(
-          concat_input, self._output_size, method='bilinear')
+          concat_input, self._output_size, method='bilinear'
+      )
       label = tf.image.resize(label, self._output_size, method='nearest')
       label = tf.reshape(label[:, :, -1], [1] + self._output_size)
 
@@ -195,14 +220,16 @@ class Parser(parser.Parser):
 
       image_mask = tf.concat([image, label], axis=2)
       image_mask_crop = tf.image.random_crop(
-          image_mask, self._crop_size + [tf.shape(image_mask)[-1]])
+          image_mask, self._crop_size + [tf.shape(image_mask)[-1]]
+      )
       image = image_mask_crop[:, :, :-1]
       label = tf.reshape(image_mask_crop[:, :, -1], [1] + self._crop_size)
 
     # Flips image randomly during training.
     if self._aug_rand_hflip:
       image, _, label = preprocess_ops.random_horizontal_flip(
-          image, masks=label)
+          image, masks=label
+      )
 
     train_image_size = self._crop_size if self._crop_size else self._output_size
     # Resizes and crops image.
@@ -211,7 +238,9 @@ class Parser(parser.Parser):
         train_image_size,
         train_image_size,
         aug_scale_min=self._aug_scale_min,
-        aug_scale_max=self._aug_scale_max)
+        aug_scale_max=self._aug_scale_max,
+        centered_crop=self._centered_crop,
+    )
 
     # Resizes and crops boxes.
     image_scale = image_info[2, :]
@@ -221,11 +250,17 @@ class Parser(parser.Parser):
     # The label is first offset by +1 and then padded with 0.
     label += 1
     label = tf.expand_dims(label, axis=3)
-    label = preprocess_ops.resize_and_crop_masks(label, image_scale,
-                                                 train_image_size, offset)
+    label = preprocess_ops.resize_and_crop_masks(
+        label,
+        image_scale,
+        train_image_size,
+        offset,
+        centered_crop=self._centered_crop,
+    )
     label -= 1
     label = tf.where(
-        tf.equal(label, -1), self._ignore_label * tf.ones_like(label), label)
+        tf.equal(label, -1), self._ignore_label * tf.ones_like(label), label
+    )
     label = tf.squeeze(label, axis=0)
     valid_mask = tf.not_equal(label, self._ignore_label)
 
@@ -255,30 +290,58 @@ class Parser(parser.Parser):
 
     # Resizes and crops image.
     image, image_info = preprocess_ops.resize_and_crop_image(
-        image, self._output_size, self._output_size)
+        image,
+        self._output_size,
+        self._output_size,
+        centered_crop=self._centered_crop,
+    )
 
     if self._resize_eval_groundtruth:
       # Resizes eval masks to match input image sizes. In that case, mean IoU
       # is computed on output_size not the original size of the images.
       image_scale = image_info[2, :]
       offset = image_info[3, :]
-      label = preprocess_ops.resize_and_crop_masks(label, image_scale,
-                                                   self._output_size, offset)
+      label = preprocess_ops.resize_and_crop_masks(
+          label,
+          image_scale,
+          self._output_size,
+          offset,
+          centered_crop=self._centered_crop,
+      )
     else:
-      label = tf.image.pad_to_bounding_box(label, 0, 0,
-                                           self._groundtruth_padded_size[0],
-                                           self._groundtruth_padded_size[1])
+      if self._centered_crop:
+        label_size = tf.cast(tf.shape(label)[0:2], tf.int32)
+        label = tf.image.pad_to_bounding_box(
+            label,
+            tf.maximum(
+                (self._groundtruth_padded_size[0] - label_size[0]) // 2, 0
+            ),
+            tf.maximum(
+                (self._groundtruth_padded_size[1] - label_size[1]) // 2, 0
+            ),
+            self._groundtruth_padded_size[0],
+            self._groundtruth_padded_size[1],
+        )
+      else:
+        label = tf.image.pad_to_bounding_box(
+            label,
+            0,
+            0,
+            self._groundtruth_padded_size[0],
+            self._groundtruth_padded_size[1],
+        )
 
     label -= 1
     label = tf.where(
-        tf.equal(label, -1), self._ignore_label * tf.ones_like(label), label)
+        tf.equal(label, -1), self._ignore_label * tf.ones_like(label), label
+    )
     label = tf.squeeze(label, axis=0)
 
     valid_mask = tf.not_equal(label, self._ignore_label)
     labels = {
         'masks': label,
         'valid_masks': valid_mask,
-        'image_info': image_info
+        'image_info': image_info,
     }
 
     # Cast image as self._dtype

official/vision/ops/anchor.py

@@ -22,7 +22,6 @@ from typing import Dict, Optional, Tuple
 
 import tensorflow as tf, tf_keras
 
-from official.vision.ops import anchor_generator
 from official.vision.ops import box_matcher
 from official.vision.ops import iou_similarity
 from official.vision.ops import target_gather
@@ -32,7 +31,38 @@ from official.vision.utils.object_detection import faster_rcnn_box_coder
 
 
 class Anchor(object):
-  """Anchor class for anchor-based object detectors."""
+  """Anchor class for anchor-based object detectors.
+
+  Example:
+  ```python
+  anchor_boxes = Anchor(
+      min_level=3,
+      max_level=4,
+      num_scales=2,
+      aspect_ratios=[0.5, 1., 2.],
+      anchor_size=4.,
+      image_size=[256, 256],
+  ).multilevel_boxes
+  ```
+
+  Attributes:
+    min_level: integer number of minimum level of the output feature pyramid.
+    max_level: integer number of maximum level of the output feature pyramid.
+    num_scales: integer number representing intermediate scales added on each
+      level. For instances, num_scales=2 adds one additional intermediate
+      anchor scales [2^0, 2^0.5] on each level.
+    aspect_ratios: list of float numbers representing the aspect ratio anchors
+      added on each level. The number indicates the ratio of width to height.
+      For instances, aspect_ratios=[1.0, 2.0, 0.5] adds three anchors on each
+      scale level.
+    anchor_size: float number representing the scale of size of the base
+      anchor to the feature stride 2^level.
+    image_size: a list of integer numbers or Tensors representing [height,
+      width] of the input image size.
+    multilevel_boxes: an OrderedDict from level to the generated anchor boxes of
+      shape [height_l, width_l, num_anchors_per_location * 4].
+    anchors_per_location: number of anchors per pixel location.
+  """
 
   def __init__(
       self,
@@ -43,57 +73,40 @@ class Anchor(object):
       anchor_size,
       image_size,
   ):
-    """Constructs multi-scale anchors.
-
-    Args:
-      min_level: integer number of minimum level of the output feature pyramid.
-      max_level: integer number of maximum level of the output feature pyramid.
-      num_scales: integer number representing intermediate scales added on each
-        level. For instances, num_scales=2 adds one additional intermediate
-        anchor scales [2^0, 2^0.5] on each level.
-      aspect_ratios: list of float numbers representing the aspect ratio anchors
-        added on each level. The number indicates the ratio of width to height.
-        For instances, aspect_ratios=[1.0, 2.0, 0.5] adds three anchors on each
-        scale level.
-      anchor_size: float number representing the scale of size of the base
-        anchor to the feature stride 2^level.
-      image_size: a list of integer numbers or Tensors representing [height,
-        width] of the input image size.The image_size should be divided by the
-        largest feature stride 2^max_level.
-    """
+    """Initializes the instance."""
     self.min_level = min_level
     self.max_level = max_level
     self.num_scales = num_scales
     self.aspect_ratios = aspect_ratios
     self.anchor_size = anchor_size
     self.image_size = image_size
-    self.boxes = self._generate_boxes()
+    self.multilevel_boxes = self._generate_multilevel_boxes()
 
-  def _generate_boxes(self) -> tf.Tensor:
+  def _generate_multilevel_boxes(self) -> Dict[str, tf.Tensor]:
     """Generates multi-scale anchor boxes.
 
     Returns:
-      a Tensor of shape [N, 4], representing anchor boxes of all levels
-      concatenated together.
+      An OrderedDict from level to anchor boxes of shape [height_l, width_l,
+      num_anchors_per_location * 4].
     """
-    boxes_all = []
+    multilevel_boxes = collections.OrderedDict()
     for level in range(self.min_level, self.max_level + 1):
       boxes_l = []
-      feat_size = math.ceil(self.image_size[0] / 2**level)
-      stride = tf.cast(self.image_size[0] / feat_size, tf.float32)
+      feat_size_y = math.ceil(self.image_size[0] / 2**level)
+      feat_size_x = math.ceil(self.image_size[1] / 2**level)
+      stride_y = tf.cast(self.image_size[0] / feat_size_y, tf.float32)
+      stride_x = tf.cast(self.image_size[1] / feat_size_x, tf.float32)
+      x = tf.range(stride_x / 2, self.image_size[1], stride_x)
+      y = tf.range(stride_y / 2, self.image_size[0], stride_y)
+      xv, yv = tf.meshgrid(x, y)
       for scale in range(self.num_scales):
         for aspect_ratio in self.aspect_ratios:
-          intermidate_scale = 2 ** (scale / float(self.num_scales))
-          base_anchor_size = self.anchor_size * stride * intermidate_scale
+          intermidate_scale = 2 ** (scale / self.num_scales)
+          base_anchor_size = self.anchor_size * 2**level * intermidate_scale
           aspect_x = aspect_ratio**0.5
           aspect_y = aspect_ratio**-0.5
           half_anchor_size_x = base_anchor_size * aspect_x / 2.0
           half_anchor_size_y = base_anchor_size * aspect_y / 2.0
-          x = tf.range(stride / 2, self.image_size[1], stride)
-          y = tf.range(stride / 2, self.image_size[0], stride)
-          xv, yv = tf.meshgrid(x, y)
-          xv = tf.cast(tf.reshape(xv, [-1]), dtype=tf.float32)
-          yv = tf.cast(tf.reshape(yv, [-1]), dtype=tf.float32)
           # Tensor shape Nx4.
           boxes = tf.stack(
               [
@@ -102,41 +115,18 @@ class Anchor(object):
                   yv + half_anchor_size_y,
                   xv + half_anchor_size_x,
               ],
-              axis=1,
+              axis=-1,
           )
           boxes_l.append(boxes)
-      # Concat anchors on the same level to tensor shape NxAx4.
-      boxes_l = tf.stack(boxes_l, axis=1)
-      boxes_l = tf.reshape(boxes_l, [-1, 4])
-      boxes_all.append(boxes_l)
-    return tf.concat(boxes_all, axis=0)
-
-  def unpack_labels(self, labels: tf.Tensor) -> Dict[str, tf.Tensor]:
-    """Unpacks an array of labels into multi-scales labels."""
-    unpacked_labels = collections.OrderedDict()
-    count = 0
-    for level in range(self.min_level, self.max_level + 1):
-      feat_size_y = tf.cast(
-          math.ceil(self.image_size[0] / 2**level), tf.int32
-      )
-      feat_size_x = tf.cast(
-          math.ceil(self.image_size[1] / 2**level), tf.int32
-      )
-      steps = feat_size_y * feat_size_x * self.anchors_per_location
-      unpacked_labels[str(level)] = tf.reshape(
-          labels[count : count + steps], [feat_size_y, feat_size_x, -1]
-      )
-      count += steps
-    return unpacked_labels
+      # Concat anchors on the same level to tensor shape HxWx(Ax4).
+      boxes_l = tf.concat(boxes_l, axis=-1)
+      multilevel_boxes[str(level)] = boxes_l
+    return multilevel_boxes
 
   @property
-  def anchors_per_location(self):
+  def anchors_per_location(self) -> int:
     return self.num_scales * len(self.aspect_ratios)
 
-  @property
-  def multilevel_boxes(self):
-    return self.unpack_labels(self.boxes)
-
 
 class AnchorLabeler(object):
   """Labeler for dense object detector."""
@@ -420,24 +410,68 @@ class RpnAnchorLabeler(AnchorLabeler):
     return score_targets_dict, box_targets_dict
 
 
+class AnchorGeneratorv2:
+  """Utility to generate anchors for a multiple feature maps.
+
+  Attributes:
+    min_level: integer number of minimum level of the output feature pyramid.
+    max_level: integer number of maximum level of the output feature pyramid.
+    num_scales: integer number representing intermediate scales added on each
+      level. For instances, num_scales=2 adds one additional intermediate
+      anchor scales [2^0, 2^0.5] on each level.
+    aspect_ratios: list of float numbers representing the aspect ratio anchors
+      added on each level. The number indicates the ratio of width to height.
+      For instances, aspect_ratios=[1.0, 2.0, 0.5] adds three anchors on each
+      scale level.
+    anchor_size: float number representing the scale of size of the base
+      anchor to the feature stride 2^level.
+  """
+
+  def __init__(
+      self,
+      min_level,
+      max_level,
+      num_scales,
+      aspect_ratios,
+      anchor_size,
+    ):
+    """Initializes the instance."""
+    self.min_level = min_level
+    self.max_level = max_level
+    self.num_scales = num_scales
+    self.aspect_ratios = aspect_ratios
+    self.anchor_size = anchor_size
+
+  def __call__(self, image_size):
+    """Generate multilevel anchor boxes.
+
+    Args:
+      image_size: a list of integer numbers or Tensors representing [height,
+        width] of the input image size.
+    Returns:
+      An ordered dictionary from level to anchor boxes of shape [height_l,
+      width_l, num_anchors_per_location * 4].
+    """
+    return Anchor(
+        min_level=self.min_level,
+        max_level=self.max_level,
+        num_scales=self.num_scales,
+        aspect_ratios=self.aspect_ratios,
+        anchor_size=self.anchor_size,
+        image_size=image_size,
+    ).multilevel_boxes
+
+
 def build_anchor_generator(
     min_level, max_level, num_scales, aspect_ratios, anchor_size
 ):
   """Build anchor generator from levels."""
-  anchor_sizes = collections.OrderedDict()
-  strides = collections.OrderedDict()
-  scales = []
-  for scale in range(num_scales):
-    scales.append(2 ** (scale / float(num_scales)))
-  for level in range(min_level, max_level + 1):
-    stride = 2**level
-    strides[str(level)] = stride
-    anchor_sizes[str(level)] = anchor_size * stride
-  anchor_gen = anchor_generator.AnchorGenerator(
-      anchor_sizes=anchor_sizes,
-      scales=scales,
+  anchor_gen = AnchorGeneratorv2(
+      min_level=min_level,
+      max_level=max_level,
+      num_scales=num_scales,
       aspect_ratios=aspect_ratios,
-      strides=strides,
+      anchor_size=anchor_size,
   )
   return anchor_gen
 

official/vision/ops/anchor_generator.py

@@ -109,7 +109,7 @@ class _SingleAnchorGenerator:
     return tf.reshape(result, [shape[0], shape[1], shape[2] * shape[3]])
 
 
-class AnchorGenerator():
+class AnchorGeneratorv1():
   """Utility to generate anchors for a multiple feature maps.
 
   Example:

official/vision/ops/anchor_generator_test.py

@@ -77,7 +77,7 @@ class MultiScaleAnchorGeneratorTest(parameterized.TestCase, tf.test.TestCase):
     levels = range(min_level, max_level + 1)
     anchor_sizes = [2**(level + 1) for level in levels]
     strides = [2**level for level in levels]
-    anchor_gen = anchor_generator.AnchorGenerator(
+    anchor_gen = anchor_generator.AnchorGeneratorv1(
         anchor_sizes=anchor_sizes,
         scales=[1.],
         aspect_ratios=aspect_ratios,
@@ -98,7 +98,7 @@ class MultiScaleAnchorGeneratorTest(parameterized.TestCase, tf.test.TestCase):
     levels = range(min_level, max_level + 1)
     anchor_sizes = [2**(level + 1) for level in levels]
     strides = [2**level for level in levels]
-    anchor_gen = anchor_generator.AnchorGenerator(
+    anchor_gen = anchor_generator.AnchorGeneratorv1(
         anchor_sizes=anchor_sizes,
         scales=[1.],
         aspect_ratios=aspect_ratios,
@@ -122,7 +122,7 @@ class MultiScaleAnchorGeneratorTest(parameterized.TestCase, tf.test.TestCase):
     levels = range(min_level, max_level + 1)
     anchor_sizes = dict((str(level), 2**(level + 1)) for level in levels)
     strides = dict((str(level), 2**level) for level in levels)
-    anchor_gen = anchor_generator.AnchorGenerator(
+    anchor_gen = anchor_generator.AnchorGeneratorv1(
         anchor_sizes=anchor_sizes,
         scales=[1.],
         aspect_ratios=aspect_ratios,

official/vision/ops/anchor_test.py

@@ -58,40 +58,52 @@ class AnchorTest(parameterized.TestCase, tf.test.TestCase):
     self.assertEqual(negatives, expected_negatives)
 
   @parameterized.parameters(
-      # Single scale anchor.
-      (5, 5, 1, [1.0], 2.0,
-       [[-16, -16, 48, 48], [-16, 16, 48, 80],
-        [16, -16, 80, 48], [16, 16, 80, 80]]),
-      # Multi scale anchor.
-      (5, 6, 1, [1.0], 2.0,
-       [[-16, -16, 48, 48], [-16, 16, 48, 80],
-        [16, -16, 80, 48], [16, 16, 80, 80], [-32, -32, 96, 96]]),
-      # # Multi aspect ratio anchor.
-      (6, 6, 1, [1.0, 4.0, 0.25], 2.0,
-       [[-32, -32, 96, 96], [-0, -96, 64, 160], [-96, -0, 160, 64]]),
-
+      # Single scale anchor
+      (5, 5, 1, [1.0], 2.0, [64, 64],
+       {'5': [[[-16, -16, 48, 48], [-16, 16, 48, 80]],
+              [[16, -16, 80, 48], [16, 16, 80, 80]]]}),
+      # Multi scale anchor
+      (5, 6, 1, [1.0], 2.0, [64, 64],
+       {'5': [[[-16, -16, 48, 48], [-16, 16, 48, 80]],
+              [[16, -16, 80, 48], [16, 16, 80, 80]]],
+        '6': [[[-32, -32, 96, 96]]]}),
+      # Multi aspect ratio anchor
+      (6, 6, 1, [1.0, 4.0, 0.25], 2.0, [64, 64],
+       {'6': [[[-32, -32, 96, 96, -0, -96, 64, 160, -96, -0, 160, 64]]]}),
+      # Intermidate scales
+      (5, 5, 2, [1.0], 1.0, [32, 32],
+       {'5': [[[0, 0, 32, 32,
+                16 - 16 * 2**0.5, 16 - 16 * 2**0.5,
+                16 + 16 * 2**0.5, 16 + 16 * 2**0.5]]]}),
+      # Non-square
+      (5, 5, 1, [1.0], 1.0, [64, 32],
+       {'5': [[[0, 0, 32, 32]],
+              [[32, 0, 64, 32]]]}),
+      # Indivisible by 2^level
+      (5, 5, 1, [1.0], 1.0, [40, 32],
+       {'5': [[[-6, 0, 26, 32]],
+              [[14, 0, 46, 32]]]}),
   )
   def testAnchorGeneration(self, min_level, max_level, num_scales,
-                           aspect_ratios, anchor_size, expected_boxes):
-    image_size = [64, 64]
+                           aspect_ratios, anchor_size, image_size,
+                           expected_boxes):
     anchors = anchor.Anchor(min_level, max_level, num_scales, aspect_ratios,
                             anchor_size, image_size)
-    boxes = anchors.boxes.numpy()
-    self.assertEqual(expected_boxes, boxes.tolist())
+    self.assertAllClose(expected_boxes, anchors.multilevel_boxes)
 
   @parameterized.parameters(
       # Single scale anchor.
       (5, 5, 1, [1.0], 2.0,
-       [[-16, -16, 48, 48], [-16, 16, 48, 80],
-        [16, -16, 80, 48], [16, 16, 80, 80]]),
+       {'5': [[[-16, -16, 48, 48], [-16, 16, 48, 80]],
+              [[16, -16, 80, 48], [16, 16, 80, 80]]]}),
       # Multi scale anchor.
       (5, 6, 1, [1.0], 2.0,
-       [[-16, -16, 48, 48], [-16, 16, 48, 80],
-        [16, -16, 80, 48], [16, 16, 80, 80], [-32, -32, 96, 96]]),
-      # # Multi aspect ratio anchor.
+       {'5': [[[-16, -16, 48, 48], [-16, 16, 48, 80]],
+              [[16, -16, 80, 48], [16, 16, 80, 80]]],
+        '6': [[[-32, -32, 96, 96]]]}),
+      # Multi aspect ratio anchor.
       (6, 6, 1, [1.0, 4.0, 0.25], 2.0,
-       [[-32, -32, 96, 96], [-0, -96, 64, 160], [-96, -0, 160, 64]]),
-
+       {'6': [[[-32, -32, 96, 96, -0, -96, 64, 160, -96, -0, 160, 64]]]}),
   )
   def testAnchorGenerationWithImageSizeAsTensor(self,
                                                 min_level,
@@ -103,8 +115,25 @@ class AnchorTest(parameterized.TestCase, tf.test.TestCase):
     image_size = tf.constant([64, 64], tf.int32)
     anchors = anchor.Anchor(min_level, max_level, num_scales, aspect_ratios,
                             anchor_size, image_size)
-    boxes = anchors.boxes.numpy()
-    self.assertEqual(expected_boxes, boxes.tolist())
+    self.assertAllClose(expected_boxes, anchors.multilevel_boxes)
+
+  @parameterized.parameters(
+      (6, 8, 2, [1.0, 2.0, 0.5], 3.0, [320, 256]),
+  )
+  def testAnchorGenerationAreCentered(self, min_level, max_level, num_scales,
+                                      aspect_ratios, anchor_size, image_size):
+    anchors = anchor.Anchor(min_level, max_level, num_scales, aspect_ratios,
+                            anchor_size, image_size)
+    multilevel_boxes = anchors.multilevel_boxes
+    image_size = np.array(image_size)
+    for boxes in multilevel_boxes.values():
+      boxes = boxes.numpy()
+      box_centers = boxes.mean(axis=0).mean(axis=0)
+      box_centers = [
+          (box_centers[0] + box_centers[2]) / 2,
+          (box_centers[1] + box_centers[3]) / 2,
+      ]
+      self.assertAllClose(image_size / 2, box_centers)
 
   @parameterized.parameters(
       (3, 6, 2, [1.0], 2.0, False),
@@ -164,6 +193,7 @@ class AnchorTest(parameterized.TestCase, tf.test.TestCase):
       (3, 7, [.5, 1., 2.], 2, 8, (256, 256)),
       (3, 8, [1.], 3, 32, (512, 512)),
       (3, 3, [1.], 2, 4, (32, 32)),
+      (4, 8, [.5, 1., 2.], 2, 3, (320, 256)),
   )
   def testEquivalentResult(self, min_level, max_level, aspect_ratios,
                            num_scales, anchor_size, image_size):

official/vision/ops/preprocess_ops.py

@@ -168,6 +168,7 @@ def resize_and_crop_image(
     seed=1,
     method=tf.image.ResizeMethod.BILINEAR,
     keep_aspect_ratio=True,
+    centered_crop=False,
 ):
   """Resizes the input image to output size (RetinaNet style).
 
@@ -195,6 +196,9 @@ def resize_and_crop_image(
     seed: seed for random scale jittering.
     method: function to resize input image to scaled image.
     keep_aspect_ratio: whether or not to keep the aspect ratio when resizing.
+    centered_crop: If `centered_crop` is set to True, then resized crop (if
+      smaller than padded size) is place in the center of the image. Default
+      behaviour is to place it at left top corner.
 
   Returns:
     output_image: `Tensor` of shape [height, width, 3] where [height, width]
@@ -266,9 +270,19 @@ def resize_and_crop_image(
 
     output_image = scaled_image
     if padded_size is not None:
-      output_image = tf.image.pad_to_bounding_box(
-          scaled_image, 0, 0, padded_size[0], padded_size[1]
-      )
+      if centered_crop:
+        scaled_image_size = tf.cast(tf.shape(scaled_image)[0:2], tf.int32)
+        output_image = tf.image.pad_to_bounding_box(
+            scaled_image,
+            tf.maximum((padded_size[0] - scaled_image_size[0]) // 2, 0),
+            tf.maximum((padded_size[1] - scaled_image_size[1]) // 2, 0),
+            padded_size[0],
+            padded_size[1],
+        )
+      else:
+        output_image = tf.image.pad_to_bounding_box(
+            scaled_image, 0, 0, padded_size[0], padded_size[1]
+        )
 
     image_info = tf.stack([
         image_size,
@@ -686,7 +700,9 @@ def resize_and_crop_boxes(boxes, image_scale, output_size, offset):
     return boxes
 
 
-def resize_and_crop_masks(masks, image_scale, output_size, offset):
+def resize_and_crop_masks(
+    masks, image_scale, output_size, offset, centered_crop: bool = False
+):
   """Resizes boxes to output size with scale and offset.
 
   Args:
@@ -697,6 +713,9 @@ def resize_and_crop_masks(masks, image_scale, output_size, offset):
       output image size.
     offset: 2D `Tensor` representing top-left corner [y0, x0] to crop scaled
       boxes.
+    centered_crop: If `centered_crop` is set to True, then resized crop (if
+      smaller than padded size) is place in the center of the image. Default
+      behaviour is to place it at left top corner.
 
   Returns:
     masks: `Tensor` of shape [N, H, W, C] representing the scaled masks.
@@ -719,6 +738,7 @@ def resize_and_crop_masks(masks, image_scale, output_size, offset):
     scaled_masks = tf.image.resize(
         masks, scaled_size, method=tf.image.ResizeMethod.NEAREST_NEIGHBOR
     )
+
     offset = tf.cast(offset, tf.int32)
     scaled_masks = scaled_masks[
         :,
@@ -727,9 +747,20 @@ def resize_and_crop_masks(masks, image_scale, output_size, offset):
         :,
     ]
 
-    output_masks = tf.image.pad_to_bounding_box(
-        scaled_masks, 0, 0, output_size[0], output_size[1]
-    )
+    if centered_crop:
+      scaled_mask_size = tf.cast(tf.shape(scaled_masks)[1:3], tf.int32)
+      output_masks = tf.image.pad_to_bounding_box(
+          scaled_masks,
+          tf.maximum((output_size[0] - scaled_mask_size[0]) // 2, 0),
+          tf.maximum((output_size[1] - scaled_mask_size[1]) // 2, 0),
+          output_size[0],
+          output_size[1],
+      )
+    else:
+      output_masks = tf.image.pad_to_bounding_box(
+          scaled_masks, 0, 0, output_size[0], output_size[1]
+      )
+
     # Remove padding.
     output_masks = output_masks[1::]
     return output_masks

official/vision/ops/preprocess_ops_test.py

@@ -482,6 +482,46 @@ class InputUtilsTest(parameterized.TestCase, tf.test.TestCase):
     self.assertShapeEqual(input_image, aug_image)
     self.assertDTypeEqual(aug_image, np.uint8)
 
+  @parameterized.parameters(0.25, 0.5, 0.75, 1, 1.25, 1.5)
+  def test_resize_and_crop_image_and_masks(self, scale):
+    image = tf.convert_to_tensor(np.random.rand(1024, 2048, 3))
+    label = tf.convert_to_tensor(np.ones((1, 1024, 2048, 1), dtype=np.int32))
+    image, image_info = preprocess_ops.resize_and_crop_image(
+        image, (256, 256), (256, 256), scale, scale, centered_crop=True
+    )
+    image_scale = image_info[2, :]
+    offset = image_info[3, :]
+    label = preprocess_ops.resize_and_crop_masks(
+        label, image_scale, (256, 256), offset, centered_crop=True
+    )
+    self.assertEqual(image.shape[0:2], label.shape[1:3])
+    image_arr = image.numpy()
+    label_arr = np.squeeze(label.numpy())
+
+    scaled_height = round(1024 * 256 * scale / 2048)
+    scaled_width = round(2048 * 256 * scale / 2048)
+    height_offset = max((256 - scaled_height) // 2, 0)
+    width_offset = max((256 - scaled_width) // 2, 0)
+
+    self.assertEqual(
+        label_arr[
+            height_offset : 256 - height_offset,
+            width_offset : 256 - width_offset,
+        ].mean(),
+        1,
+    )
+    self.assertEqual(label_arr[0:height_offset, :].mean(), 0)
+    self.assertEqual(image_arr[0:height_offset, :, :].mean(), 0)
+    self.assertEqual(label_arr[256 - height_offset :, :].mean(), 0)
+    self.assertEqual(image_arr[256 - height_offset :, :, :].mean(), 0)
+    if width_offset > 0:
+      self.assertEqual(label_arr[height_offset, 0:width_offset].mean(), 0)
+      self.assertEqual(label_arr[height_offset, 256 - width_offset :].mean(), 0)
+      self.assertEqual(image_arr[height_offset, 0:width_offset, :].mean(), 0)
+      self.assertEqual(
+          image_arr[height_offset, 256 - width_offset :, :].mean(), 0
+      )
+
 
 if __name__ == '__main__':
   tf.test.main()

official/vision/tasks/semantic_segmentation.py

@@ -114,7 +114,8 @@ class SemanticSegmentationTask(base_task.Task):
         preserve_aspect_ratio=params.preserve_aspect_ratio,
         dtype=params.dtype,
         image_feature=params.image_feature,
-        additional_dense_features=params.additional_dense_features)
+        additional_dense_features=params.additional_dense_features,
+        centered_crop=params.centered_crop)
 
     reader = input_reader_factory.input_reader_generator(
         params,

{tf_models_nightly-2.17.0.dev20240313.dist-info → tf_models_nightly-2.17.0.dev20240320.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tf-models-nightly
-Version: 2.17.0.dev20240313
+Version: 2.17.0.dev20240320
 Summary: TensorFlow Official Models
 Home-page: https://github.com/tensorflow/models
 Author: Google Inc.

{tf_models_nightly-2.17.0.dev20240313.dist-info → tf_models_nightly-2.17.0.dev20240320.dist-info}/RECORD

@@ -298,7 +298,7 @@ official/nlp/metrics/__init__.py,sha256=7oiypy0N82PDw9aSdcJBLVoGTd_oRSUOdvuJhMv4
 official/nlp/metrics/bleu.py,sha256=XOTTbjC3B9lt8-MLvNX02tjA94wfsUVse6KJ5CWPzfk,6587
 official/nlp/metrics/bleu_test.py,sha256=0j4pZ1MSIcndvUNZa25oXCu4UFOE367KaL7oRNCzLCI,2508
 official/nlp/modeling/__init__.py,sha256=SQozaRl78tYS6xvGCfM3msABe2VL20x_mL2vIln1Sn0,1062
-official/nlp/modeling/layers/__init__.py,sha256=ylm8koecxcv2OtyYxZWr2asrGi1IM9d1seYntRanN4g,4780
+official/nlp/modeling/layers/__init__.py,sha256=no0uyA68hsEDa_UuUdhC4jXeqb4lj-3dp6j0GTDzrIQ,4864
 official/nlp/modeling/layers/attention.py,sha256=3-jG3m_L9Y41BY35c4uTFG_Ywlfk4SOwUEtmqfSoKkk,3906
 official/nlp/modeling/layers/attention_test.py,sha256=c7KezuYUze8PWAPuwYow8KTQNRyuuJgwICSsFTyJ2nQ,3536
 official/nlp/modeling/layers/bigbird_attention.py,sha256=dzutgRoQt2DFsYMpMILv_QF0O_FMDbiLQ3T-7c1Zpcs,21111
@@ -965,7 +965,7 @@ official/vision/configs/maskrcnn.py,sha256=yL8kggxXaCTIpSkcozAV2UudO7UqVcEh1_-rM
 official/vision/configs/maskrcnn_test.py,sha256=Wfkbz30h2qxPcpuu6CEpQsf8I_2df6y10-4bRLsWlj8,1733
 official/vision/configs/retinanet.py,sha256=oCKinkh4IyPslmI1pakwi6dVziwjkZ2cIcpSoGRjqnM,17806
 official/vision/configs/retinanet_test.py,sha256=ffS3QufQMLF8FZhKNmi7Yr1RDTnIyZ1XKQ9agr2EyW8,1699
-official/vision/configs/semantic_segmentation.py,sha256=wN9ejI_KTOS4SasCHJTRRdV_3DWW0eiqvVWtArevP8o,30616
+official/vision/configs/semantic_segmentation.py,sha256=4ZAyLWKcFYReyrEWBc5b7wld3mMcuH0RcaRe_4J2RrA,30831
 official/vision/configs/semantic_segmentation_test.py,sha256=va-ZG6CtBKcs0NicZe6WmJvHxPxxih7nB0orNtrRiEA,1867
 official/vision/configs/video_classification.py,sha256=tf2XJhD_7c1Ned3eS93Sc0qrQ8U3M_zVJy09KI-9em8,14513
 official/vision/configs/video_classification_test.py,sha256=I1HSamxRQ3-f4-YHIeUChnT5CtHCxFQdiL0zy6RRUXU,1879
@@ -988,7 +988,7 @@ official/vision/dataloaders/input_reader_factory.py,sha256=WpvSA8qyqAo3wkmme4WqX
 official/vision/dataloaders/maskrcnn_input.py,sha256=iCc08yYD-7mvIPojgBjm_nSvoQACXWCIeZNZN8CfXSs,16822
 official/vision/dataloaders/parser.py,sha256=nMXnhigMa_ascSJ2OK88xi4HdE9xvfL3G4oMrHau-t4,2315
 official/vision/dataloaders/retinanet_input.py,sha256=joxJL4hQVPw-FW5iUc7RsxP60N7iYGRuVFpU3gC5flE,18291
-official/vision/dataloaders/segmentation_input.py,sha256=EV5mVYyDzmNefGtHTUw7tecMGpajTnyrfY7nV_zugnY,11838
+official/vision/dataloaders/segmentation_input.py,sha256=Klg5KAChYZDRvqzZfyIzdPy54rTlWYZp2AotolD3WX8,12934
 official/vision/dataloaders/tf_example_decoder.py,sha256=9yCT6uSLMpmw50w7zdaRR_BXy6vIvliLZntrYAgzD18,8647
 official/vision/dataloaders/tf_example_decoder_test.py,sha256=PHxneXHn5-eIMdmk1uI4IPLa178kTCifa4EF53ik2Jo,12629
 official/vision/dataloaders/tf_example_label_map_decoder.py,sha256=EHu6ZQvYxqjUliOlsN_f4okYt9Hdpydv_lM_dQwrklU,2598
@@ -1101,10 +1101,10 @@ official/vision/modeling/layers/roi_generator.py,sha256=ZL-m5IObg8JsXJ_DZ0QQusv2
 official/vision/modeling/layers/roi_sampler.py,sha256=uajmniFsZrmAR_-ojva_dVm9xZU3wze5EqCryP9gUoQ,10006
 official/vision/modeling/models/__init__.py,sha256=9tSOgQ3WTOmg91vIOdiAzpcHCvMJ76nFqCDyQR095k4,1020
 official/vision/ops/__init__.py,sha256=7oiypy0N82PDw9aSdcJBLVoGTd_oRSUOdvuJhMv4leQ,609
-official/vision/ops/anchor.py,sha256=ruGKslsQid6A0yfiJuG6sf_JEbJuHHpWL77D9s3SNx4,19627
-official/vision/ops/anchor_generator.py,sha256=293RH42qBk8zMtfiJovtBJ6Hn9yi4Ln-Az-sGXNAOR0,7274
-official/vision/ops/anchor_generator_test.py,sha256=grVbHuxlu1W7xbjMErV3q0ARxCesHN6q-7LvLcCi-_4,5296
-official/vision/ops/anchor_test.py,sha256=OQPiWHNUkQi3KaMWzOYd9HuqJa9vQ6Za4_6tf79X0qY,7633
+official/vision/ops/anchor.py,sha256=awu5WOXwTK8H5j2MppnRQcvPHNXSCiBnKcJCnISNfnE,20631
+official/vision/ops/anchor_generator.py,sha256=cN2k9rw4EpHPGrtSm8Ea51QERPy6WB1XGwYKjAkP8Uw,7276
+official/vision/ops/anchor_generator_test.py,sha256=qL4LCRr7UzwWlXmWNuS_yVwXe0ygNdMwYOz5T6Mrar4,5302
+official/vision/ops/anchor_test.py,sha256=PP2UNanwQjdiGxGaL9FZjU0vqTUbOp_Ni_1Kedga9aU,9015
 official/vision/ops/augment.py,sha256=20K-OWV_GLbpzv8sY5rBe6x3H5MIRbTlDWOPTfNlGeY,108290
 official/vision/ops/augment_test.py,sha256=9QISeqJ79JgkI2-Hn4pMxPRRin3mbOd7mSJ4v7S8W6E,25881
 official/vision/ops/box_matcher.py,sha256=AvZd7CUUZnT4FwETLyVz3Uxb3gO-o94OwlZrvm7CtX0,9067
@@ -1115,10 +1115,10 @@ official/vision/ops/iou_similarity_test.py,sha256=x5jlcMqMCUYC5cRgdbR0VlAW67AoXo
 official/vision/ops/mask_ops.py,sha256=cZLpIowzEA57bXPDbVXa6mktZVHvGSH-TQ1CxHjpQXw,10270
 official/vision/ops/mask_ops_test.py,sha256=D3xbbbleJd4HkpWOSDSEy6hNihsRBY93BqPF6JP-dJk,2835
 official/vision/ops/nms.py,sha256=bKYDAtyV5j6PG7g-RGF2ZccCI5V1xVvuajNblCy1TGs,8125
-official/vision/ops/preprocess_ops.py,sha256=RAouiPkapg_TJHRh_dj1mwcULTBdEvHf6wHPSL_S-Nk,41306
+official/vision/ops/preprocess_ops.py,sha256=tDW9apUocwSzGnKkgSrKGwqbcZpb6-WQ8nGnW1_ds20,42569
 official/vision/ops/preprocess_ops_3d.py,sha256=K2583ynA9Zt9pOBNoWSD8KtQR1fwRYBoylJ9NusIBtI,16110
 official/vision/ops/preprocess_ops_3d_test.py,sha256=1MmygNfRBnQYGszsrKBGqP_GrPlP4_UGuddCbkYcIms,8364
-official/vision/ops/preprocess_ops_test.py,sha256=ilCDXPyGI7bYWIXWumTcqNJJ9fVEYmGloW-95au0rOo,15483
+official/vision/ops/preprocess_ops_test.py,sha256=FY3EUvQIb82fYqYJPmzkE7pmkhXIQrd7JNLGFPB6SXI,17213
 official/vision/ops/sampling_ops.py,sha256=1jywCA_E4qDUFWsykCLUwZsWtQSR0QREXdJhvP5qCvc,16072
 official/vision/ops/spatial_transform_ops.py,sha256=PVEJGAn0ygtsrid84vD5GgV0jsjyWoNn14RBzreMxM4,38389
 official/vision/ops/target_gather.py,sha256=Ir3X76yXYEVFSYX5h-yfS8SMkY37GYuypBP2C8ykggo,3965
@@ -1150,7 +1150,7 @@ official/vision/tasks/__init__.py,sha256=qfhL5xyDrjZez_zjw613TyciLkqtWm-INFeES7G
 official/vision/tasks/image_classification.py,sha256=Oh4tH-SAD0-MmVfglE3pFXI6-LXVJr2C-WwbudnOdSk,16699
 official/vision/tasks/maskrcnn.py,sha256=iC8-OIFyYcQWpdbBODCXEag2R3YUNdQcZsn_wYAd8f0,25569
 official/vision/tasks/retinanet.py,sha256=EFILc2YPDeLqWcH7QtzN2k5sT5KdKQwioh12NKVOIqg,18261
-official/vision/tasks/semantic_segmentation.py,sha256=jpKzMSrhgXXD1y-HqvUtrcktCOEfL99C35uiaZval3U,14243
+official/vision/tasks/semantic_segmentation.py,sha256=hQBxyT1qZ1SQd4xHCWETiVNcQs96mYxPEXzdTMY8zTI,14287
 official/vision/tasks/video_classification.py,sha256=F4RnG_OvnayPDeWb8khEp8lFyM6CRWi_FlUaBOgsQjk,14318
 official/vision/utils/__init__.py,sha256=7oiypy0N82PDw9aSdcJBLVoGTd_oRSUOdvuJhMv4leQ,609
 official/vision/utils/ops_test.py,sha256=GqsKndrbfbpsfR6Bhs4gnMelXMVnO66CN5eNna-Wj7Y,4394
@@ -1203,9 +1203,9 @@ tensorflow_models/__init__.py,sha256=etxw45SHxuwFCRX5qGxGMP83II0JfJulzNl5GSNJvhw
 tensorflow_models/tensorflow_models_test.py,sha256=AxUYUdiQn416UR7jg0h6rmv688esvlKDfpyDCIQkF18,1395
 tensorflow_models/nlp/__init__.py,sha256=4tA5Pf4qaFwT-fIFOpX7x7FHJpnyJT-5UgOeFYTyMlc,807
 tensorflow_models/vision/__init__.py,sha256=zBorY_v5xva1uI-qxhZO3Qh-Dii-Suq6wEYh6hKHDfc,833
-tf_models_nightly-2.17.0.dev20240313.dist-info/AUTHORS,sha256=1dG3fXVu9jlo7bul8xuix5F5vOnczMk7_yWn4y70uw0,337
-tf_models_nightly-2.17.0.dev20240313.dist-info/LICENSE,sha256=WxeBS_DejPZQabxtfMOM_xn8qoZNJDQjrT7z2wG1I4U,11512
-tf_models_nightly-2.17.0.dev20240313.dist-info/METADATA,sha256=XJi11Uap1W1o_JQQ5OlBn7Ty-1FNcy_jAjkSeu3F6dE,1432
-tf_models_nightly-2.17.0.dev20240313.dist-info/WHEEL,sha256=kGT74LWyRUZrL4VgLh6_g12IeVl_9u9ZVhadrgXZUEY,110
-tf_models_nightly-2.17.0.dev20240313.dist-info/top_level.txt,sha256=gum2FfO5R4cvjl2-QtP-S1aNmsvIZaFFT6VFzU0f4-g,33
-tf_models_nightly-2.17.0.dev20240313.dist-info/RECORD,,
+tf_models_nightly-2.17.0.dev20240320.dist-info/AUTHORS,sha256=1dG3fXVu9jlo7bul8xuix5F5vOnczMk7_yWn4y70uw0,337
+tf_models_nightly-2.17.0.dev20240320.dist-info/LICENSE,sha256=WxeBS_DejPZQabxtfMOM_xn8qoZNJDQjrT7z2wG1I4U,11512
+tf_models_nightly-2.17.0.dev20240320.dist-info/METADATA,sha256=xPZzhngyKemP0x15q_EQ3oPSZAv8zboLm-n68yaKW6I,1432
+tf_models_nightly-2.17.0.dev20240320.dist-info/WHEEL,sha256=kGT74LWyRUZrL4VgLh6_g12IeVl_9u9ZVhadrgXZUEY,110
+tf_models_nightly-2.17.0.dev20240320.dist-info/top_level.txt,sha256=gum2FfO5R4cvjl2-QtP-S1aNmsvIZaFFT6VFzU0f4-g,33
+tf_models_nightly-2.17.0.dev20240320.dist-info/RECORD,,