Fixed Several Typos

orbit/actions/conditional_action.py

@@ -47,7 +47,7 @@ def __init__(
     """Initializes the instance.
 
     Args:
-      condition: A callable accepting train or eval outputs and returing a bool.
+      condition: A callable accepting train or eval outputs and returning a bool.
       action: The action (or optionally sequence of actions) to perform when
         `condition` is met.
     """

orbit/actions/new_best_metric.py

@@ -40,7 +40,7 @@ class NewBestMetric:
   if it is achieved. These separate methods enable the same `NewBestMetric`
   instance to be reused as a condition multiple times, and can also provide
   additional preemption/failure safety. For example, to avoid updating the best
-  metric if a model export fails or is pre-empted:
+  metric if a model export fails or is pre-emptied:
 
       new_best_metric = orbit.actions.NewBestMetric(
         'accuracy', filename='/model/dir/best_metric')

orbit/actions/new_best_metric_test.py

@@ -61,7 +61,7 @@ def test_json_persisted_value(self):
     tempfile = self.create_tempfile().full_path
     value = {'a': 1, 'b': 2}
     persisted_value = actions.JSONPersistedValue(value, tempfile)
-    # The inital value is used since tempfile is empty.
+    # The initial value is used since tempfile is empty.
     self.assertEqual(persisted_value.read(), value)
     persisted_value = actions.JSONPersistedValue('ignored', tempfile)
     # Initial value of 'ignored' is ignored, since there's a value in tempfile.

research/adversarial_text/graphs.py

@@ -45,7 +45,7 @@
 flags.DEFINE_integer('batch_size', 64, 'Size of the batch.')
 flags.DEFINE_integer('num_timesteps', 100, 'Number of timesteps for BPTT')
 
-# Model architechture
+# Model architecture
 flags.DEFINE_bool('bidir_lstm', False, 'Whether to build a bidirectional LSTM.')
 flags.DEFINE_bool('single_label', True, 'Whether the sequence has a single '
                   'label, for optimization.')

research/attention_ocr/python/model.py

@@ -579,7 +579,7 @@ def label_smoothing_regularization(self, chars_labels, weight=0.1):
     Uses the same method as in https://arxiv.org/abs/1512.00567.
 
     Args:
-      chars_labels: ground truth ids of charactes, shape=[batch_size,
+      chars_labels: ground truth ids of characters, shape=[batch_size,
         seq_length];
       weight: label-smoothing regularization weight.
 

research/attention_ocr/python/model_export.py

@@ -80,7 +80,7 @@ def export_model(export_dir,
                  crop_image_height=None):
   """Exports a model to the named directory.
 
-  Note that --datatset_name and --checkpoint are required and parsed by the
+  Note that --dataset_name and --checkpoint are required and parsed by the
   underlying module common_flags.
 
   Args:

research/attention_ocr/python/model_test.py

@@ -180,7 +180,7 @@ def test_sequence_loss_function_without_label_smoothing(self):
     self.assertEqual(loss_np.shape, tuple())
 
   def encode_coordinates_alt(self, net):
-    """An alternative implemenation for the encoding coordinates.
+    """An alternative implementation for the encoding coordinates.
 
     Args:
       net: a tensor of shape=[batch_size, height, width, num_features]

research/audioset/vggish/README.md

@@ -23,7 +23,7 @@ VGGish depends on the following Python packages:
 
 These are all easily installable via, e.g., `pip install numpy` (as in the
 sample installation session below). Any reasonably recent version of these
-packages shold work.
+packages should work.
 
 VGGish also requires downloading two data files:
 

research/audioset/vggish/vggish_slim.py

@@ -111,7 +111,7 @@ def load_vggish_slim_checkpoint(session, checkpoint_path):
 
   This function can be used as an initialization function (referred to as
   init_fn in TensorFlow documentation) which is called in a Session after
-  initializating all variables. When used as an init_fn, this will load
+  initializing all variables. When used as an init_fn, this will load
   a pre-trained checkpoint that is compatible with the VGGish model
   definition. Only variables defined by VGGish will be loaded.
 

research/audioset/vggish/vggish_smoke_test.py

@@ -17,7 +17,7 @@
 
 This is a simple smoke test of a local install of VGGish and its associated
 downloaded files. We create a synthetic sound, extract log mel spectrogram
-features, run them through VGGish, post-process the embedding ouputs, and
+features, run them through VGGish, post-process the embedding outputs, and
 check some simple statistics of the results, allowing for variations that
 might occur due to platform/version differences in the libraries we use.
 

tensorflow_models/tensorflow_models_pypi.ipynb

@@ -49,7 +49,7 @@
         "\n",
         "## Colab environment setup. To use a stable TF release version\n",
         "##  because of the possible breakage in tf-nightly.\n",
-        "# !pip3 install -U numpy\u003e=1.20\n",
+        "# !pip3 install -U numpy>=1.20\n",
         "# !pip3 install -q tensorflow==2.8.0"
       ]
     },
@@ -101,7 +101,7 @@
         "## Check out modules\n",
         "\n",
         "**Note: As the TensorFlow Models (NLP + Vision) 2.9 release which is tested for this notebook, we partially exported selected modules but the APIs are not stable. Also be aware that, the\n",
-        "modeling libraries are advancing very fast, so we generally don't guarantee compatability between versions.** "
+        "modeling libraries are advancing very fast, so we generally don't guarantee compatibility between versions.** "
       ]
     },
     {
@@ -245,7 +245,7 @@
           "output_type": "stream",
           "text": [
             "spine_net\n",
-            "{'4': \u003cKerasTensor: shape=(1, 8, 8, 128) dtype=float32 (created by layer 'spine_net')\u003e, '5': \u003cKerasTensor: shape=(1, 4, 4, 128) dtype=float32 (created by layer 'spine_net')\u003e, '6': \u003cKerasTensor: shape=(1, 2, 2, 128) dtype=float32 (created by layer 'spine_net')\u003e}\n"
+            "{'4': <KerasTensor: shape=(1, 8, 8, 128) dtype=float32 (created by layer 'spine_net')>, '5': <KerasTensor: shape=(1, 4, 4, 128) dtype=float32 (created by layer 'spine_net')>, '6': <KerasTensor: shape=(1, 2, 2, 128) dtype=float32 (created by layer 'spine_net')>}\n"
           ]
         }
       ],