Fix some typos in official/.

official/legacy/image_classification/efficientnet/tfhub_export.py

@@ -43,8 +43,8 @@ def export_tfhub(model_path, hub_destination, model_name):
   image_input = tf.keras.layers.Input(
       shape=(None, None, 3), name="image_input", dtype=tf.float32)
   x = image_input * 255.0
-  ouputs = efficientnet_model.efficientnet(x, config)
-  hub_model = tf.keras.Model(image_input, ouputs)
+  outputs = efficientnet_model.efficientnet(x, config)
+  hub_model = tf.keras.Model(image_input, outputs)
   ckpt = tf.train.Checkpoint(model=hub_model)
   ckpt.restore(model_path).assert_existing_objects_matched()
   hub_model.save(

official/legacy/transformer/data_download.py

@@ -188,7 +188,7 @@ def download_and_extract(path, url, input_filename, target_filename):
     Full paths to extracted input and target files.
 
   Raises:
-    OSError: if the the download/extraction fails.
+    OSError: if the download/extraction fails.
   """
   # Check if extracted files already exist in path
   input_file = find_file(path, input_filename)

official/modeling/optimization/lr_schedule.py

@@ -23,7 +23,7 @@ import tensorflow as tf
 def _make_offset_wrapper(new_class_name: str, base_lr_class):
   """Generates a offset wrapper of learning rate schedule.
 
-  It will returns a subclass of the the `base_lr_class`, the subclass takes an
+  It will returns a subclass of the `base_lr_class`, the subclass takes an
   `offset` argument in the constructor. When the new class instance is called,
   the behavior is:
     new_class_object(step) = base_lr_class_object(step - offset)

official/nlp/data/train_sentencepiece.py

@@ -36,7 +36,7 @@ from sentencepiece import SentencePieceTrainer
 
 FLAGS = flags.FLAGS
 flags.DEFINE_string("output_model_path", None,
-                    "Path to save the the sentencepiece model.")
+                    "Path to save the sentencepiece model.")
 flags.mark_flag_as_required("output_model_path")
 
 flags.DEFINE_string("tfds_dir", None, "Directory of the tfds.")

official/nlp/docs/train.md

@@ -105,7 +105,7 @@ pip3 install --user -r official/requirements.txt
 
 <details>
 
-This example fine-tunes BERT-base from TF-Hub on the the Multi-Genre Natural
+This example fine-tunes BERT-base from TF-Hub on the Multi-Genre Natural
 Language Inference (MultiNLI) corpus using TPUs.
 
 Firstly, you can prepare the fine-tuning data using

official/nlp/modeling/layers/README.md

@@ -13,7 +13,7 @@ assemble new `tf.keras` layers or models.
     ["Big Bird: Transformers for Longer Sequences"](https://arxiv.org/abs/2007.14062).
 
 *   [CachedAttention](attention.py) implements an attention layer with cache
-    used for auto-agressive decoding.
+    used for auto-aggressive decoding.
 
 *   [KernelAttention](kernel_attention.py) implements a group of attention
     mechansim that express the self-attention as a linear dot-product of

official/nlp/modeling/layers/gaussian_process.py

@@ -226,7 +226,7 @@ class RandomFeatureGaussianProcess(tf.keras.layers.Layer):
     """Resets covariance matrix of the GP layer.
 
     This function is useful for reseting the model's covariance matrix at the
-    begining of a new epoch.
+    beginning of a new epoch.
     """
     self._gp_cov_layer.reset_precision_matrix()
 
@@ -380,7 +380,7 @@ class LaplaceRandomFeatureCovariance(tf.keras.layers.Layer):
     """Resets precision matrix to its initial value.
 
     This function is useful for reseting the model's covariance matrix at the
-    begining of a new epoch.
+    beginning of a new epoch.
     """
     precision_matrix_reset_op = self.precision_matrix.assign(
         self.initial_precision_matrix)

official/nlp/modeling/layers/reuse_attention.py

@@ -465,7 +465,7 @@ class ReuseMultiHeadAttention(tf.keras.layers.Layer):
     """Builds multi-head dot-product attention computations.
 
     This function builds attributes necessary for `_compute_attention` to
-    costomize attention computation to replace the default dot-product
+    customize attention computation to replace the default dot-product
     attention.
 
     Args:

official/nlp/modeling/ops/beam_search.py

@@ -204,7 +204,7 @@ class SequenceBeamSearch(tf.Module):
       candidate_log_probs = _log_prob_from_logits(logits)
 
       # Calculate new log probabilities if each of the alive sequences were
-      # extended # by the the candidate IDs.
+      # extended # by the candidate IDs.
       # Shape [batch_size, beam_size, vocab_size]
       log_probs = candidate_log_probs + tf.expand_dims(alive_log_probs, axis=2)
 

official/projects/edgetpu/nlp/modeling/edgetpu_layers.py

@@ -123,7 +123,7 @@ class EdgeTPUMultiHeadAttention(tf.keras.layers.MultiHeadAttention):
     """Builds multi-head dot-product attention computations.
 
     This function builds attributes necessary for `_compute_attention` to
-    costomize attention computation to replace the default dot-product
+    customize attention computation to replace the default dot-product
     attention.
 
     Args:

official/projects/edgetpu/vision/tasks/image_classification.py

@@ -265,7 +265,7 @@ class EdgeTPUTask(base_task.Task):
     """Does forward and backward.
 
     Args:
-      inputs: A tuple of of input tensors of (features, labels).
+      inputs: A tuple of input tensors of (features, labels).
       model: A tf.keras.Model instance.
       optimizer: The optimizer for this training step.
       metrics: A nested structure of metrics objects.
@@ -319,7 +319,7 @@ class EdgeTPUTask(base_task.Task):
     """Runs validatation step.
 
     Args:
-      inputs: A tuple of of input tensors of (features, labels).
+      inputs: A tuple of input tensors of (features, labels).
       model: A tf.keras.Model instance.
       metrics: A nested structure of metrics objects.
 

official/projects/video_ssl/modeling/video_ssl_model.py

@@ -53,7 +53,7 @@ class VideoSSLModel(tf.keras.Model):
       hidden_dim: `int` number of hidden units in MLP.
       hidden_layer_num: `int` number of hidden layers in MLP.
       hidden_norm_args: `dict` for batchnorm arguments in MLP.
-      projection_dim: `int` number of ouput dimension for MLP.
+      projection_dim: `int` number of output dimension for MLP.
       input_specs: `tf.keras.layers.InputSpec` specs of the input tensor.
       dropout_rate: `float` rate for dropout regularization.
       aggregate_endpoints: `bool` aggregate all end ponits or only use the

official/projects/yt8m/dataloaders/yt8m_input.py

@@ -33,7 +33,7 @@ from official.vision.dataloaders import parser
 
 
 def resize_axis(tensor, axis, new_size, fill_value=0):
-  """Truncates or pads a tensor to new_size on on a given axis.
+  """Truncates or pads a tensor to new_size on a given axis.
 
   Truncate or extend tensor such that tensor.shape[axis] == new_size. If the
   size increases, the padding will be performed at the end, using fill_value.

official/vision/beta/projects/centernet/README.md

@@ -22,7 +22,7 @@ heatmaps (one heatmap for each class) is needed to predict the object. CenterNet
 proves that this can be done without a significant difference in accuracy.
 
 
-## Enviroment setup
+## Environment setup
 
 The code can be run on multiple GPUs or TPUs with different distribution
 strategies. See the TensorFlow distributed training

official/vision/beta/projects/simclr/README.md

@@ -10,7 +10,7 @@
   An illustration of SimCLR (from <a href="https://ai.googleblog.com/2020/04/advancing-self-supervised-and-semi.html">our blog here</a>).
 </div>
 
-## Enviroment setup
+## Environment setup
 
 The code can be run on multiple GPUs or TPUs with different distribution
 strategies. See the TensorFlow distributed training

official/vision/beta/projects/yolo/losses/yolo_loss.py

@@ -323,7 +323,7 @@ class DarknetLoss(YoloLossBase):
     grid_points = tf.stop_gradient(grid_points)
     anchor_grid = tf.stop_gradient(anchor_grid)
 
-    # Split all the ground truths to use as seperate items in loss computation.
+    # Split all the ground truths to use as separate items in loss computation.
     (true_box, ind_mask, true_class) = tf.split(y_true, [4, 1, 1], axis=-1)
     true_conf = tf.squeeze(true_conf, axis=-1)
     true_class = tf.squeeze(true_class, axis=-1)

official/vision/beta/projects/yolo/modeling/layers/nn_blocks.py

@@ -823,7 +823,7 @@ class CSPStack(tf.keras.layers.Layer):
   make it a cross stage partial. Added for ease of use. you should be able
   to wrap any layer stack with a CSP independent of wether it belongs
   to the Darknet family. if filter_scale = 2, then the blocks in the stack
-  passed into the the CSP stack should also have filters = filters/filter_scale
+  passed into the CSP stack should also have filters = filters/filter_scale
   Cross Stage Partial networks (CSPNets) were proposed in:
 
   [1] Chien-Yao Wang, Hong-Yuan Mark Liao, I-Hau Yeh, Yueh-Hua Wu,

official/vision/data/create_coco_tf_record.py

@@ -112,7 +112,7 @@ def generate_coco_panoptics_masks(segments_info, mask_path,
       represent "stuff" and "things" classes respectively.
 
   Returns:
-    A dict with with keys: [u'semantic_segmentation_mask', u'category_mask',
+    A dict with keys: [u'semantic_segmentation_mask', u'category_mask',
       u'instance_mask']. The dict contains 'category_mask' and 'instance_mask'
       only if `include_panoptic_eval_masks` is set to True.
   """

official/vision/examples/starter/example_task.py

@@ -138,7 +138,7 @@ class ExampleTask(base_task.Task):
     between output from Parser and input used here.
 
     Args:
-      inputs: A tuple of of input tensors of (features, labels).
+      inputs: A tuple of input tensors of (features, labels).
       model: A tf.keras.Model instance.
       optimizer: The optimizer for this training step.
       metrics: A nested structure of metrics objects.
@@ -186,7 +186,7 @@ class ExampleTask(base_task.Task):
     """Runs validatation step.
 
     Args:
-      inputs: A tuple of of input tensors of (features, labels).
+      inputs: A tuple of input tensors of (features, labels).
       model: A tf.keras.Model instance.
       metrics: A nested structure of metrics objects.
 

official/vision/ops/box_ops.py

@@ -616,7 +616,7 @@ def bbox_overlap(boxes, gt_boxes):
         tf.transpose(gt_invalid_mask, [0, 2, 1]))
     iou = tf.where(padding_mask, -tf.ones_like(iou), iou)
 
-    # Fills -1 for for invalid (-1) boxes.
+    # Fills -1 for invalid (-1) boxes.
     boxes_invalid_mask = tf.less(
         tf.reduce_max(boxes, axis=-1, keepdims=True), 0.0)
     iou = tf.where(boxes_invalid_mask, -tf.ones_like(iou), iou)