tf-models-nightly 2.15.0.dev20240208__py2.py3-none-any.whl → 2.15.0.dev20240210__py2.py3-none-any.whl

official/vision/configs/common.py

@@ -15,7 +15,7 @@
 """Common configurations."""
 
 import dataclasses
-from typing import List, Optional
+from typing import List, Optional, Sequence
 
 # Import libraries
 
@@ -100,6 +100,35 @@ class MixupAndCutmix(hyperparams.Config):
   label_smoothing: float = 0.1
 
 
+@dataclasses.dataclass
+class SSDRandomCropParam(hyperparams.Config):
+  min_object_covered: float = 0.0
+  min_box_overlap: float = 0.5
+  prob_to_apply: float = 0.85
+
+
+@dataclasses.dataclass
+class SSDRandomCrop(hyperparams.Config):
+  """Configuration for SSDRandomCrop.
+
+  Liu et al., SSD: Single shot multibox detector
+  https://arxiv.org/abs/1512.02325.
+  """
+  ssd_random_crop_params: Sequence[SSDRandomCropParam] = dataclasses.field(
+      default_factory=lambda: (
+          SSDRandomCropParam(min_object_covered=0.0),
+          SSDRandomCropParam(min_object_covered=0.1),
+          SSDRandomCropParam(min_object_covered=0.3),
+          SSDRandomCropParam(min_object_covered=0.5),
+          SSDRandomCropParam(min_object_covered=0.7),
+          SSDRandomCropParam(min_object_covered=0.9),
+          SSDRandomCropParam(min_object_covered=1.0),
+      )
+  )
+  aspect_ratio_range: tuple[float, float] = (0.5, 2.0)
+  area_range: tuple[float, float] = (0.1, 1.0)
+
+
 @dataclasses.dataclass
 class Augmentation(hyperparams.OneOfConfig):
   """Configuration for input data augmentation.
@@ -112,6 +141,9 @@ class Augmentation(hyperparams.OneOfConfig):
   type: Optional[str] = None
   randaug: RandAugment = dataclasses.field(default_factory=RandAugment)
   autoaug: AutoAugment = dataclasses.field(default_factory=AutoAugment)
+  ssd_random_crop: SSDRandomCrop = dataclasses.field(
+      default_factory=SSDRandomCrop
+  )
 
 
 @dataclasses.dataclass

official/vision/dataloaders/retinanet_input.py

@@ -155,6 +155,13 @@ class Parser(parser.Parser):
             translate_const=aug_type.randaug.translate_const,
             prob_to_apply=aug_type.randaug.prob_to_apply,
             exclude_ops=aug_type.randaug.exclude_ops)
+      elif aug_type.type == 'ssd_random_crop':
+        logging.info('Using SSD Random Crop.')
+        self._augmenter = augment.SSDRandomCrop(
+            params=aug_type.ssd_random_crop.ssd_random_crop_params,
+            aspect_ratio_range=aug_type.ssd_random_crop.aspect_ratio_range,
+            area_range=aug_type.ssd_random_crop.area_range,
+        )
       else:
         raise ValueError(f'Augmentation policy {aug_type.type} not supported.')
 

official/vision/ops/augment.py

@@ -26,14 +26,22 @@ MixupAndCutmix:
 RandomErasing, Mixup and Cutmix are inspired by
 https://github.com/rwightman/pytorch-image-models
 
+SSDRandCrop Reference:
+  - Liu et al., SSD: Single shot multibox detector:
+    https://arxiv.org/abs/1512.02325
+  - Implementation from TF Object Detection API:
+    https://github.com/tensorflow/models/
 """
+from collections.abc import Sequence
 import inspect
 import math
-from typing import Any, List, Iterable, Optional, Tuple, Union
+from typing import Any, Iterable, List, Optional, Tuple, Union
 
 import numpy as np
 import tensorflow as tf, tf_keras
 
+from official.vision.configs import common as configs
+from official.vision.ops import box_ops
 
 # This signifies the max integer that the controller RNN could predict for the
 # augmentation scheme.
@@ -184,7 +192,7 @@ def _normalize_tuple(value, n, name):
 def gaussian_filter2d(
     image: tf.Tensor,
     filter_shape: Union[List[int], Tuple[int, ...], int],
-    sigma: Union[List[float], Tuple[float], float] = 1.0,
+    sigma: Union[List[float], Tuple[float, float], float] = 1.0,
     padding: str = 'REFLECT',
     constant_values: Union[int, tf.Tensor] = 0,
     name: Optional[str] = None,
@@ -2781,3 +2789,133 @@ class MixupAndCutmix:
     labels = lam * labels_1 + (1. - lam) * labels_2
 
     return images, labels
+
+
+def filter_boxes_by_ioa(
+    bboxes: tf.Tensor, crop_box: tf.Tensor, min_box_overlap: float
+) -> tf.Tensor:
+  """Filter boxes by intersection over area (IOA).
+
+  The boxes with IOA less than min_box_overlap will be replaced by
+  (0, 0, 0, 0) so they can be filtered out later.
+
+  Args:
+    bboxes: a float tensor of shape [N, 4] representing normalized bounding box
+      coordinates.
+    crop_box: a float tensor of shape [1, 1, 4] representing the normalized crop
+      box.
+    min_box_overlap: minimum overlap of the box with the crop box to keep the
+      box.
+
+  Returns:
+    a tensor of shape [N, 4] with filtered box coordinates replaced by 0.
+  """
+  ioas = box_ops.bbox_intersection_over_area(bboxes[None, ...], crop_box)[0]
+  keep_boxes = ioas >= min_box_overlap
+  # Set coordinates to (0, 0, 0, 0) for filtered boxes
+  return bboxes * tf.cast(keep_boxes, dtype=bboxes.dtype)
+
+
+def crop_normalized_boxes(
+    bboxes: tf.Tensor,
+    ori_image_size: tf.Tensor,
+    new_image_size: tf.Tensor,
+    offset: tf.Tensor,
+) -> tf.Tensor:
+  """Crop normalized boxes.
+
+  Args:
+    bboxes: a float tensor of shape [N, 4] representing normalized box
+      coordinates.
+    ori_image_size: an int tensor of shape [2] representing the original image
+      size.
+    new_image_size: an int tensor of shape [2] representing the cropped image
+      size.
+    offset: an int tensor of shape [2] representing the offset of the crop.
+
+  Returns:
+    a tensor of shape [N, 4] representing the new normalized bounding box
+    coordinates in the new cropped image.
+  """
+  new_bboxes = box_ops.denormalize_boxes(bboxes, ori_image_size)
+  new_bboxes -= tf.tile(tf.cast(offset, dtype=tf.float32), [2])[None, ...]
+  new_bboxes = box_ops.normalize_boxes(new_bboxes, new_image_size)
+  return tf.clip_by_value(new_bboxes, 0.0, 1.0)
+
+
+class SSDRandomCrop(ImageAugment):
+  """Random crop preprocessing as in the SSD paper.
+
+  Liu et al., SSD: Single shot multibox detector
+  https://arxiv.org/abs/1512.02325.
+
+  The implementation originated from TF Object Detection API:
+  https://github.com/tensorflow/models/blob/f36581036d3346a9496de06c8fd678d23cfe2103/research/object_detection/core/preprocessor.py#L3529
+  """
+
+  def __init__(
+      self,
+      params: Sequence[configs.SSDRandomCropParam] | None = None,
+      aspect_ratio_range: tuple[float, float] = (0.5, 2.0),
+      area_range: tuple[float, float] = (0.1, 1.0),
+  ):
+    """Apply random crop to the image as in the SSD paper.
+
+    The SSD random crop will randomly select one set of the parameters.
+
+    Args:
+      params: a sequence of SSDRandomCropParam that contains:
+        min_object_covered - a float representing minimum the cropped image
+          must cover at least this fraction with at least one of the input
+          bounding boxes.
+        min_box_overlap - a float representing minimum overlap of the bounding
+          box with the cropped image to keep the box.
+        prob_to_apply - a float representing the probability to crop.
+      aspect_ratio_range: allowed range for aspect ratio of the cropped image.
+      area_range: allowed range for area ratio between cropped image and the
+        original image.
+    """
+    if params is None:
+      params = configs.SSDRandomCrop().ssd_random_crop_params
+    self.num_cases = len(params)
+    self.min_object_covered = tf.constant(
+        [param.min_object_covered for param in params], dtype=tf.float32,
+    )
+    self.min_box_overlap = tf.constant(
+        [param.min_box_overlap for param in params], dtype=tf.float32,
+    )
+    self.prob_to_apply = tf.constant(
+        [param.prob_to_apply for param in params], dtype=tf.float32,
+    )
+    self.aspect_ratio_range = aspect_ratio_range
+    self.area_range = area_range
+
+  def distort_with_boxes(
+      self, image: tf.Tensor, bboxes: tf.Tensor
+  ) -> tuple[tf.Tensor, tf.Tensor]:
+    """See base class."""
+    i_params = tf.random.uniform([], maxval=self.num_cases, dtype=tf.int32)
+
+    if tf.random.uniform(shape=[], maxval=1.0) > self.prob_to_apply[i_params]:
+      return image, bboxes
+
+    image_size = tf.shape(image)
+    bboxes = tf.clip_by_value(bboxes, 0., 1.)
+    offset, new_image_size, crop_box = tf.image.sample_distorted_bounding_box(
+        image_size=image_size,
+        bounding_boxes=bboxes[None, ...],
+        min_object_covered=self.min_object_covered[i_params],
+        aspect_ratio_range=self.aspect_ratio_range,
+        area_range=self.area_range,
+        max_attempts=100,
+        use_image_if_no_bounding_boxes=True,
+    )
+    new_image = tf.slice(image, offset, new_image_size)
+
+    new_bboxes = filter_boxes_by_ioa(
+        bboxes, crop_box, self.min_box_overlap[i_params]
+    )
+    new_bboxes = crop_normalized_boxes(
+        new_bboxes, image_size[:2], new_image_size[:2], offset[:2]
+    )
+    return new_image, new_bboxes

official/vision/ops/augment_test.py

@@ -19,11 +19,13 @@ from __future__ import division
 from __future__ import print_function
 
 import random
+from unittest import mock
 from absl.testing import parameterized
 
 import numpy as np
 import tensorflow as tf, tf_keras
 
+from official.vision.configs import common as configs
 from official.vision.ops import augment
 
 
@@ -520,5 +522,176 @@ class MixupAndCutmixTest(tf.test.TestCase, parameterized.TestCase):
     self.assertFalse(tf.math.reduce_all(images == aug_images))
 
 
+class SSDRandomCropTest(tf.test.TestCase, parameterized.TestCase):
+
+  @parameterized.named_parameters(
+      dict(
+          testcase_name='filter first one',
+          bboxes=[[0, 0, 1, 1], [0, 0, 0.5, 0.5]],
+          crop_box=[[[0, 0, 0.5, 1]]],
+          min_box_overlap=0.6,
+          expected=[[0, 0, 0, 0], [0, 0, 0.5, 0.5]],
+      ),
+      dict(
+          testcase_name='empty box list',
+          bboxes=tf.zeros([0, 4], dtype=tf.float32),
+          crop_box=[[[0, 0, 1, 1]]],
+          min_box_overlap=0.5,
+          expected=tf.zeros([0, 4], dtype=tf.float32),
+      ),
+  )
+  def test_filter_boxes_by_ioa(
+      self, bboxes, crop_box, min_box_overlap, expected
+  ):
+    new_bboxes = augment.filter_boxes_by_ioa(
+        bboxes=tf.constant(bboxes, dtype=tf.float32),
+        crop_box=tf.constant(crop_box, dtype=tf.float32),
+        min_box_overlap=min_box_overlap,
+    )
+    self.assertAllClose(expected, new_bboxes)
+
+  @parameterized.named_parameters(
+      dict(
+          testcase_name='whole image and box',
+          bboxes=[[0, 0, 1, 1], [0.1, 0.2, 0.8, 0.5]],
+          ori_image_size=[200, 600],
+          new_image_size=[100, 200],
+          offset=[70, 100],
+          expected=[[0, 0, 1, 1], [0, 0.1, 0.9, 1]],
+      ),
+      dict(
+          testcase_name='zero size boxes',
+          bboxes=tf.zeros([1, 4], dtype=tf.float32),
+          ori_image_size=[200, 600],
+          new_image_size=[100, 200],
+          offset=[70, 100],
+          expected=tf.zeros([1, 4], dtype=tf.float32),
+      ),
+      dict(
+          testcase_name='empty box list',
+          bboxes=tf.zeros([0, 4], dtype=tf.float32),
+          ori_image_size=[200, 600],
+          new_image_size=[100, 200],
+          offset=[70, 100],
+          expected=tf.zeros([0, 4], dtype=tf.float32),
+      ),
+  )
+  def test_crop_normalized_boxes(
+      self, bboxes, ori_image_size, new_image_size, offset, expected
+  ):
+    got = augment.crop_normalized_boxes(
+        bboxes=tf.constant(bboxes, dtype=tf.float32),
+        ori_image_size=tf.constant(ori_image_size, dtype=tf.int32),
+        new_image_size=tf.constant(new_image_size, dtype=tf.int32),
+        offset=tf.constant(offset, dtype=tf.int32),
+    )
+    self.assertAllClose(expected, got)
+
+  @parameterized.named_parameters(
+      dict(
+          testcase_name='uint8 image',
+          image=tf.zeros([320, 256, 3], dtype=tf.uint8),
+      ),
+      dict(
+          testcase_name='float32 image',
+          image=tf.zeros([320, 256, 3], dtype=tf.float32),
+      ),
+  )
+  def test_distort_with_boxes_output_shape(self, image):
+    bboxes = tf.constant([[0, 0, 0.5, 0.5], [0.5, 0.5, 1.0, 1.0]])
+    augmenter = augment.SSDRandomCrop()
+    new_image, new_bboxes = augmenter.distort_with_boxes(
+        image=image,
+        bboxes=bboxes,
+    )
+    self.assertDTypeEqual(new_image, image.dtype)
+    self.assertDTypeEqual(new_bboxes, bboxes.dtype)
+    self.assertShapeEqual(new_bboxes, bboxes)
+    self.assertAllGreaterEqual(new_bboxes, 0)
+    self.assertAllLessEqual(new_bboxes, 1)
+
+  def test_distort_with_empty_bboxes(self):
+    image = tf.zeros([320, 256, 3], dtype=tf.uint8)
+    bboxes = tf.zeros([0, 4], dtype=tf.float32)
+    augmenter = augment.SSDRandomCrop()
+    new_image, new_bboxes = augmenter.distort_with_boxes(
+        image=image,
+        bboxes=bboxes,
+    )
+    self.assertDTypeEqual(new_image, image.dtype)
+    self.assertDTypeEqual(new_bboxes, bboxes.dtype)
+    self.assertShapeEqual(new_bboxes, bboxes)
+
+  @parameterized.named_parameters(
+      dict(
+          testcase_name='uint8 image',
+          image=tf.zeros([320, 256, 3], dtype=tf.uint8),
+      ),
+      dict(
+          testcase_name='float32 image',
+          image=tf.zeros([320, 256, 3], dtype=tf.float32),
+      ),
+  )
+  def test_distort_with_boxes_run_as_tf_function(self, image):
+    bboxes = tf.constant([[0, 0, 0.5, 0.5], [0.5, 0.5, 1.0, 1.0]])
+    augmenter = augment.SSDRandomCrop()
+    aug_function = tf.function(augmenter.distort_with_boxes)
+    new_image, new_bboxes = aug_function(image=image, bboxes=bboxes)
+    self.assertDTypeEqual(new_image, image.dtype)
+    self.assertDTypeEqual(new_bboxes, bboxes.dtype)
+    self.assertShapeEqual(new_bboxes, bboxes)
+    self.assertAllGreaterEqual(new_bboxes, 0)
+    self.assertAllLessEqual(new_bboxes, 1)
+
+  def test_distort_with_boxes_run_as_tf_function_empty_bboxes(self):
+    image = tf.zeros([320, 256, 3], dtype=tf.uint8)
+    bboxes = tf.zeros([0, 4], dtype=tf.float32)
+    augmenter = augment.SSDRandomCrop()
+    aug_function = tf.function(augmenter.distort_with_boxes)
+    new_image, new_bboxes = aug_function(image=image, bboxes=bboxes)
+    self.assertDTypeEqual(new_image, image.dtype)
+    self.assertDTypeEqual(new_bboxes, bboxes.dtype)
+    self.assertShapeEqual(new_bboxes, bboxes)
+
+  def test_distort_with_boxes_filter_and_crop(self):
+    augmenter = augment.SSDRandomCrop(
+        params=[
+            configs.SSDRandomCropParam(
+                min_object_covered=0.0,
+                min_box_overlap=0.5,
+                prob_to_apply=1.0,
+            )
+        ],
+    )
+    image = tf.zeros([320, 256, 3], dtype=tf.uint8)
+    bboxes = tf.constant(
+        [
+            [0., 0., 1., 1.],  # filtered by low box overlap
+            [0.25, 0.75, 0.5, 2.],  # kept with box clipped
+            [0.25, 0.48, 0.5, 0.75],  # kept with box clipped
+        ],
+        dtype=tf.float32,
+    )
+    with mock.patch.object(
+        tf.image, 'sample_distorted_bounding_box', autospec=True
+    ) as mock_sample_box:
+      # crop box is an upper right box
+      offset = tf.constant([0, 128, 0], dtype=tf.int32)
+      new_image_size = tf.constant([160, 128, -1], dtype=tf.int32)
+      crop_box = tf.constant([[[0, 0.5, 0.5, 1.0]]], dtype=tf.float32)
+      mock_sample_box.return_value = offset, new_image_size, crop_box
+      new_image, new_bboxes = augmenter.distort_with_boxes(
+          image=image,
+          bboxes=bboxes,
+      )
+    self.assertAllClose(tf.zeros([160, 128, 3], dtype=tf.uint8), new_image)
+    self.assertAllClose(
+        tf.constant(
+            [[0., 0., 0., 0.], [0.5, 0.5, 1., 1.], [0.5, 0., 1., 0.5]],
+            dtype=tf.float32,
+        ),
+        new_bboxes,
+    )
+
 if __name__ == '__main__':
   tf.test.main()

{tf_models_nightly-2.15.0.dev20240208.dist-info → tf_models_nightly-2.15.0.dev20240210.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: tf-models-nightly
-Version: 2.15.0.dev20240208
+Version: 2.15.0.dev20240210
 Summary: TensorFlow Official Models
 Home-page: https://github.com/tensorflow/models
 Author: Google Inc.

{tf_models_nightly-2.15.0.dev20240208.dist-info → tf_models_nightly-2.15.0.dev20240210.dist-info}/RECORD

@@ -912,7 +912,7 @@ official/vision/train_spatial_partitioning.py,sha256=xpEusyM-fEPVGyBuzlkRgsYGaPA
 official/vision/configs/__init__.py,sha256=-iKVbGCvFMGSHMC89utzGXvZ83BhW6JnbEfS38lbW3M,1045
 official/vision/configs/backbones.py,sha256=-jVezph39bV2lQHRxI4Eu1Q0aj00epn417AgObKNoAM,5547
 official/vision/configs/backbones_3d.py,sha256=0lJsUzeYmuC5xiosOwrqlmgR1gkOa4tpSaxDbYYU7FE,3614
-official/vision/configs/common.py,sha256=YWKMEcKhb-fnr6aC4OeJ8aZ__rUwnR1gHUNfeAJGqd4,5354
+official/vision/configs/common.py,sha256=BCWBuIrVJ1QhgNZwRIUB6zY3s9ZlWBapqO9meUqeeZU,6420
 official/vision/configs/decoders.py,sha256=_wG6MH1RzYuhMrvJu5menR7gDvklEXDSxlHJwzVF6H4,2080
 official/vision/configs/image_classification.py,sha256=JiJUYKI82NtlNXEjATF0JzSmheQr7ywrfk7lpgV4xz8,24563
 official/vision/configs/image_classification_test.py,sha256=cdlkY5fPaqNZUc4-A2OOEUS1v3T-ekxqUc0Vm4IknBI,1867
@@ -942,7 +942,7 @@ official/vision/dataloaders/input_reader.py,sha256=CHojw8PJKf74jl8Q3rtH2ylwhmTYg
 official/vision/dataloaders/input_reader_factory.py,sha256=WpvSA8qyqAo3wkmme4WqXpICBVg0SuR6_nNWHZ0ECM0,1623
 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=C9Q23Zd8M1HQK8tPc9QJ2lSzvOilogoaHka_YDm74IM,17372
+official/vision/dataloaders/retinanet_input.py,sha256=JOJQ-2rC55hkrPwt1nBE-nWmr-CER4nbaOTqGRMjsSY,17729
 official/vision/dataloaders/segmentation_input.py,sha256=EV5mVYyDzmNefGtHTUw7tecMGpajTnyrfY7nV_zugnY,11838
 official/vision/dataloaders/tf_example_decoder.py,sha256=9yCT6uSLMpmw50w7zdaRR_BXy6vIvliLZntrYAgzD18,8647
 official/vision/dataloaders/tf_example_decoder_test.py,sha256=PHxneXHn5-eIMdmk1uI4IPLa178kTCifa4EF53ik2Jo,12629
@@ -1060,8 +1060,8 @@ official/vision/ops/anchor.py,sha256=ruGKslsQid6A0yfiJuG6sf_JEbJuHHpWL77D9s3SNx4
 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/augment.py,sha256=tQhas8eTKYcZQXtt56vho9KSbgiC-Hk8w-TG8DeuNDU,103080
-official/vision/ops/augment_test.py,sha256=6hWYGKxB_ejc1l5Rny13UOBVMyVT6f_00t8a9sj9QlU,19463
+official/vision/ops/augment.py,sha256=cL_g4gHRabtZrEXyDUyjBD-1d0JBib2Wilf2AOTlub0,108239
+official/vision/ops/augment_test.py,sha256=R4A5UZwVc_LCTgLCoTRf25BS8v21kATjUMAFrrc-g28,25711
 official/vision/ops/box_matcher.py,sha256=AvZd7CUUZnT4FwETLyVz3Uxb3gO-o94OwlZrvm7CtX0,9067
 official/vision/ops/box_matcher_test.py,sha256=YdxaTITaUyKBA27Pss5MZKF2ibBfSu879222c7lKngU,2438
 official/vision/ops/box_ops.py,sha256=FRKLYzVYd3zFqEYX6Ql4Pr39csvR2Vc-OQv-WxhegFM,34618
@@ -1158,9 +1158,9 @@ 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-2.15.0.dev20240208.dist-info/AUTHORS,sha256=1dG3fXVu9jlo7bul8xuix5F5vOnczMk7_yWn4y70uw0,337
-tf_models_nightly-2.15.0.dev20240208.dist-info/LICENSE,sha256=WxeBS_DejPZQabxtfMOM_xn8qoZNJDQjrT7z2wG1I4U,11512
-tf_models_nightly-2.15.0.dev20240208.dist-info/METADATA,sha256=nMGLknew6iW1rJFgbvTrMpujq0HuvjMJBl7zx0LBF70,1432
-tf_models_nightly-2.15.0.dev20240208.dist-info/WHEEL,sha256=kGT74LWyRUZrL4VgLh6_g12IeVl_9u9ZVhadrgXZUEY,110
-tf_models_nightly-2.15.0.dev20240208.dist-info/top_level.txt,sha256=gum2FfO5R4cvjl2-QtP-S1aNmsvIZaFFT6VFzU0f4-g,33
-tf_models_nightly-2.15.0.dev20240208.dist-info/RECORD,,
+tf_models_nightly-2.15.0.dev20240210.dist-info/AUTHORS,sha256=1dG3fXVu9jlo7bul8xuix5F5vOnczMk7_yWn4y70uw0,337
+tf_models_nightly-2.15.0.dev20240210.dist-info/LICENSE,sha256=WxeBS_DejPZQabxtfMOM_xn8qoZNJDQjrT7z2wG1I4U,11512
+tf_models_nightly-2.15.0.dev20240210.dist-info/METADATA,sha256=2q87qPU9_0-DLQjkqVlZmKZ8K6B4tH1ZRbOOSCAD_ek,1432
+tf_models_nightly-2.15.0.dev20240210.dist-info/WHEEL,sha256=kGT74LWyRUZrL4VgLh6_g12IeVl_9u9ZVhadrgXZUEY,110
+tf_models_nightly-2.15.0.dev20240210.dist-info/top_level.txt,sha256=gum2FfO5R4cvjl2-QtP-S1aNmsvIZaFFT6VFzU0f4-g,33
+tf_models_nightly-2.15.0.dev20240210.dist-info/RECORD,,