A few typos were updated in documentation.

tensorflower-gardener · tensorflower-gardener · commit 8e9f80a27e4a · 2024-02-06T22:02:00.000-08:00
Reference respective Github PR's. #10955 #10928 #10892 #11100 #11107 #11123 #11120 #11118 #11111 #11144 PiperOrigin-RevId: 604865440
diff --git a/docs/README.md b/docs/README.md
@@ -1,7 +1,7 @@
 # Public docs for TensorFlow Models
 
 This directory contains the top-level public documentation for
-[TensorFlow Models](https://github.com/tensorflow/models)
+[TensorFlow Models](https://github.com/tensorflow/models).
 
 This directory is mirrored to https://tensorflow.org/tfmodels, and is mainly
 concerned with documenting the tools provided in the `tensorflow_models` pip
diff --git a/docs/nlp/fine_tune_bert.ipynb b/docs/nlp/fine_tune_bert.ipynb
@@ -884,7 +884,7 @@
         "id": "2oHOql35k3Dd"
       },
       "source": [
-        "Note: The pretrained `TransformerEncoder` is also available on [TensorFlow Hub](https://tensorflow.org/hub). Go to the [TF Hub appendix](#hub_bert) for details."
+        "Note: The pre-trained `TransformerEncoder` is also available on [TensorFlow Hub](https://tensorflow.org/hub). Go to the [TF Hub appendix](#hub_bert) for details."
       ]
     },
     {
@@ -1148,8 +1148,8 @@
         "\n",
         "First, build a wrapper class to export the model. This wrapper does two things:\n",
         "\n",
-        "- First it packages `bert_inputs_processor` and `bert_classifier` together into a single `tf.Module`, so you can export all the functionalities.\n",
-        "- Second it defines a `tf.function` that implements the end-to-end execution of the model.\n",
+        "- First, it packages `bert_inputs_processor` and `bert_classifier` together into a single `tf.Module`, so you can export all the functionalities.\n",
+        "- Second, it defines a `tf.function` that implements the end-to-end execution of the model.\n",
         "\n",
         "Setting the `input_signature` argument of `tf.function` lets you define a fixed signature for the `tf.function`. This can be less surprising than the default automatic retracing behavior."
       ]
@@ -1189,7 +1189,7 @@
         "id": "qnxysGUfIgFQ"
       },
       "source": [
-        "Create an instance of this export-model and save it:"
+        "Create an instance of this exported model and save it:"
       ]
     },
     {
@@ -1280,7 +1280,7 @@
         "id": "CPsg7dZwfBM2"
       },
       "source": [
-        "Congratulations! You've used `tensorflow_models` to build a BERT-classifier, train it, and export for later use."
+        "Congratulations! You've used `tensorflow_models` to build a BERT-classifier, train it, and export it for later use."
       ]
     },
     {
@@ -1391,7 +1391,7 @@
         "id": "cjojn8SmLSRI"
       },
       "source": [
-        "At this point it would be simple to add a classification head yourself.\n",
+        "At this point, it would be simple to add a classification head yourself.\n",
         "\n",
         "The Model Garden `tfm.nlp.models.BertClassifier` class can also build a classifier onto the TF Hub encoder:"
       ]
@@ -1429,7 +1429,7 @@
         "id": "u_IqwXjRV1vd"
       },
       "source": [
-        "For concrete examples of this approach, refer to [Solve Glue tasks using BERT](https://www.tensorflow.org/text/tutorials/bert_glue)."
+        "For concrete examples of this approach, refer to [Solve Glue tasks using the BERT](https://www.tensorflow.org/text/tutorials/bert_glue)."
       ]
     },
     {
@@ -1494,7 +1494,7 @@
         "id": "ywn5miD_dnuh"
       },
       "source": [
-        "The advantage to using `config` objects is that they don't contain any complicated TensorFlow objects, and can be easily serialized to JSON, and rebuilt. Here's the JSON for the above `tfm.optimization.OptimizationConfig`:"
+        "The advantage of using `config` objects is that they don't contain any complicated TensorFlow objects, and can be easily serialized to JSON, and rebuilt. Here's the JSON for the above `tfm.optimization.OptimizationConfig`:"
       ]
     },
     {
diff --git a/official/README-TPU.md b/official/README-TPU.md
@@ -2,28 +2,28 @@
 
 ## Natural Language Processing
 
-*   [bert](nlp/bert): A powerful pre-trained language representation model:
+*   [bert](https://arxiv.org/abs/1810.04805): A powerful pre-trained language representation model:
     BERT, which stands for Bidirectional Encoder Representations from
     Transformers.
-    [BERT FineTuning with Cloud TPU](https://cloud.google.com/tpu/docs/tutorials/bert-2.x) provides step by step instructions on Cloud TPU training. You can look [Bert MNLI Tensorboard.dev metrics](https://tensorboard.dev/experiment/LijZ1IrERxKALQfr76gndA) for MNLI fine tuning task.
+    [BERT FineTuning with Cloud TPU](https://cloud.google.com/ai-platform/training/docs/algorithms/bert-start) provides step by step instructions on Cloud TPU training. You can look [Bert MNLI Tensorboard.dev metrics](https://tensorboard.dev/experiment/LijZ1IrERxKALQfr76gndA) for MNLI fine tuning task.
 *   [transformer](nlp/transformer): A transformer model to translate the WMT
     English to German dataset.
         [Training transformer on Cloud TPU](https://cloud.google.com/tpu/docs/tutorials/transformer-2.x) for step by step instructions on Cloud TPU training.
 
 ## Computer Vision
 
-*   [efficientnet](vision/image_classification): A family of convolutional
+*   [efficientnet](https://github.com/tensorflow/models/blob/master/official/vision/modeling/backbones/efficientnet.py): A family of convolutional
     neural networks that scale by balancing network depth, width, and
     resolution and can be used to classify ImageNet's dataset of 1000 classes.
     See [Tensorboard.dev training metrics](https://tensorboard.dev/experiment/KnaWjrq5TXGfv0NW5m7rpg/#scalars).
-*   [mnist](vision/image_classification): A basic model to classify digits
+*   [mnist](https://www.tensorflow.org/datasets/catalog/mnist): A basic model to classify digits
     from the MNIST dataset. See [Running MNIST on Cloud TPU](https://cloud.google.com/tpu/docs/tutorials/mnist-2.x) tutorial and [Tensorboard.dev metrics](https://tensorboard.dev/experiment/mIah5lppTASvrHqWrdr6NA).
-*   [mask-rcnn](vision/detection): An object detection and instance segmentation model. See [Tensorboard.dev training metrics](https://tensorboard.dev/experiment/LH7k0fMsRwqUAcE09o9kPA).
-*   [resnet](vision/image_classification): A deep residual network that can
+*   [mask-rcnn](https://www.tensorflow.org/api_docs/python/tfm/vision/configs/maskrcnn/MaskRCNN): An object detection and instance segmentation model. See [Tensorboard.dev training metrics](https://tensorboard.dev/experiment/LH7k0fMsRwqUAcE09o9kPA).
+*   [resnet]((https://www.tensorflow.org/api_docs/python/tfm/vision/configs/image_classification/image_classification_imagenet)): A deep residual network that can
     be used to classify ImageNet's dataset of 1000 classes.
     See [Training ResNet on Cloud TPU](https://cloud.google.com/tpu/docs/tutorials/resnet-2.x) tutorial and [Tensorboard.dev metrics](https://tensorboard.dev/experiment/CxlDK8YMRrSpYEGtBRpOhg).
-*   [retinanet](vision/detection): A fast and powerful object detector. See [Tensorboard.dev training metrics](https://tensorboard.dev/experiment/b8NRnWU3TqG6Rw0UxueU6Q).
-*   [shapemask](vision/detection): An object detection and instance segmentation model using shape priors. See [Tensorboard.dev training metrics](https://tensorboard.dev/experiment/ZbXgVoc6Rf6mBRlPj0JpLA).
+*   [retinanet](https://www.tensorflow.org/api_docs/python/tfm/vision/retinanet): A fast and powerful object detector. See [Tensorboard.dev training metrics](https://tensorboard.dev/experiment/b8NRnWU3TqG6Rw0UxueU6Q).
+*   [shapemask](https://cloud.google.com/tpu/docs/tutorials/shapemask-2.x): An object detection and instance segmentation model using shape priors. See [Tensorboard.dev training metrics](https://tensorboard.dev/experiment/ZbXgVoc6Rf6mBRlPj0JpLA).
 
 ## Recommendation
 *   [dlrm](recommendation/ranking): [Deep Learning Recommendation Model for
diff --git a/official/nlp/MODEL_GARDEN.md b/official/nlp/MODEL_GARDEN.md
@@ -26,7 +26,7 @@ on how to train models with this codebase.
 By default, the experiment runs on GPUs. To run on TPUs, one should overwrite
 `runtime.distribution_strategy` and set the tpu address. See [RuntimeConfig](https://github.com/tensorflow/models/blob/master/official/core/config_definitions.py) for details.
 
-In general, the experiments can run with the folloing command by setting the
+In general, the experiments can run with the following command by setting the
 corresponding `${TASK}`, `${TASK_CONFIG}`, `${MODEL_CONFIG}`.
 ```
 EXPERIMENT=???
@@ -72,7 +72,7 @@ Note that
 
 [How to Train Models](https://github.com/tensorflow/models/blob/master/official/nlp/docs/train.md)
 
-[List of Pretrained Models for finetuning](https://github.com/tensorflow/models/blob/master/official/nlp/docs/pretrained_models.md)
+[List of Pre-trained Models for finetuning](https://github.com/tensorflow/models/blob/master/official/nlp/docs/pretrained_models.md)
 
 [How to Publish Models](https://github.com/tensorflow/models/blob/master/official/nlp/docs/tfhub.md)
 
diff --git a/official/nlp/data/classifier_data_lib.py b/official/nlp/data/classifier_data_lib.py
@@ -668,7 +668,7 @@ def __init__(self,
       self._labels = list(range(info.features[self.label_key].num_classes))
 
   def _process_tfds_params_str(self, params_str):
-    """Extracts TFDS parameters from a comma-separated assignements string."""
+    """Extracts TFDS parameters from a comma-separated assignments string."""
     dtype_map = {"int": int, "float": float}
     cast_str_to_bool = lambda s: s.lower() not in ["false", "0"]
 
diff --git a/official/nlp/data/create_finetuning_data.py b/official/nlp/data/create_finetuning_data.py
@@ -165,7 +165,7 @@
     "while ALBERT uses SentencePiece tokenizer.")
 
 flags.DEFINE_string(
-    "tfds_params", "", "Comma-separated list of TFDS parameter assigments for "
+    "tfds_params", "", "Comma-separated list of TFDS parameter assignments for "
     "generic classfication data import (for more details "
     "see the TfdsProcessor class documentation).")
 
@@ -270,7 +270,7 @@ def generate_classifier_dataset():
     }
     task_name = FLAGS.classification_task_name.lower()
     if task_name not in processors:
-      raise ValueError("Task not found: %s" % (task_name))
+      raise ValueError("Task not found: %s" % (task_name,))
 
     processor = processors[task_name](process_text_fn=processor_text_fn)
     return classifier_data_lib.generate_tf_record_from_data_file(
diff --git a/official/nlp/data/create_pretraining_data.py b/official/nlp/data/create_pretraining_data.py
@@ -453,7 +453,7 @@ def _contiguous(sorted_grams):
 def _masking_ngrams(grams, max_ngram_size, max_masked_tokens, rng):
   """Create a list of masking {1, ..., n}-grams from a list of one-grams.
 
-  This is an extention of 'whole word masking' to mask multiple, contiguous
+  This is an extension of 'whole word masking' to mask multiple, contiguous
   words such as (e.g., "the red boat").
 
   Each input gram represents the token indices of a single word,
@@ -509,7 +509,7 @@ def _masking_ngrams(grams, max_ngram_size, max_masked_tokens, rng):
     rng.shuffle(v)
 
   # Create the weighting for n-gram length selection.
-  # Stored cummulatively for `random.choices` below.
+  # Stored cumulatively for `random.choices` below.
   cummulative_weights = list(
       itertools.accumulate([1./n for n in range(1, max_ngram_size+1)]))
 
@@ -519,7 +519,7 @@ def _masking_ngrams(grams, max_ngram_size, max_masked_tokens, rng):
   # Loop until we have enough masked tokens or there are no more candidate
   # n-grams of any length.
   # Each code path should ensure one or more elements from `ngrams` are removed
-  # to guarentee this loop terminates.
+  # to guarantee this loop terminates.
   while (sum(masked_tokens) < max_masked_tokens and
          sum(len(s) for s in ngrams.values())):
     # Pick an n-gram size based on our weights.
diff --git a/official/nlp/data/create_xlnet_pretraining_data.py b/official/nlp/data/create_xlnet_pretraining_data.py
@@ -271,7 +271,7 @@ def _create_a_and_b_segments(
   Args:
     tokens: The 1D input token ids. This represents an individual entry within a
       batch.
-    sentence_ids: The 1D input sentence ids. This represents an indivdual entry
+    sentence_ids: The 1D input sentence ids. This represents an individual entry
       within a batch. This should be the same length as `tokens`.
     begin_index: The reference beginning index to split data.
     total_length: The target combined length of segments A and B.
diff --git a/official/nlp/data/pretrain_dataloader.py b/official/nlp/data/pretrain_dataloader.py
@@ -143,7 +143,7 @@ class XLNetPretrainDataConfig(cfg.DataConfig):
 
   Attributes:
     input_path: See base class.
-    global_batch_size: See base calss.
+    global_batch_size: See base class.
     is_training: See base class.
     seq_length: The length of each sequence.
     max_predictions_per_seq: The number of predictions per sequence.
@@ -259,7 +259,7 @@ def _parse(self, record: Mapping[str, tf.Tensor]):
           input_mask=input_mask[:self._reuse_length])
 
       # Creates permutation mask and target mask for the rest of tokens in
-      # current example, which are concatentation of two new segments.
+      # current example, which are concatenation of two new segments.
       perm_mask_1, target_mask_1, tokens_1, masked_1 = self._get_factorization(
           inputs[self._reuse_length:], input_mask[self._reuse_length:])
 
diff --git a/official/nlp/data/squad_lib.py b/official/nlp/data/squad_lib.py
@@ -492,7 +492,7 @@ def _improve_answer_span(doc_tokens, input_start, input_end, tokenizer,
   #
   # However, this is not always possible. Consider the following:
   #
-  #   Question: What country is the top exporter of electornics?
+  #   Question: What country is the top exporter of electronics?
   #   Context: The Japanese electronics industry is the lagest in the world.
   #   Answer: Japan
   #
@@ -720,7 +720,7 @@ def postprocess_output(all_examples,
               start_logit=pred.start_logit,
               end_logit=pred.end_logit))
 
-    # if we didn't inlude the empty option in the n-best, inlcude it
+    # if we didn't include the empty option in the n-best, include it
     if version_2_with_negative and not xlnet_format:
       if "" not in seen_predictions:
         nbest.append(
@@ -815,7 +815,7 @@ def get_final_text(pred_text, orig_text, do_lower_case, verbose=False):
   # What we really want to return is "Steve Smith".
   #
   # Therefore, we have to apply a semi-complicated alignment heruistic between
-  # `pred_text` and `orig_text` to get a character-to-charcter alignment. This
+  # `pred_text` and `orig_text` to get a character-to-character alignment. This
   # can fail in certain cases in which case we just return `orig_text`.
 
   def _strip_spaces(text):
diff --git a/official/nlp/data/squad_lib_sp.py b/official/nlp/data/squad_lib_sp.py
@@ -63,7 +63,7 @@ def __repr__(self):
         tokenization.printable_text(self.question_text))
     s += ", paragraph_text: [%s]" % (" ".join(self.paragraph_text))
     if self.start_position:
-      s += ", start_position: %d" % (self.start_position)
+      s += ", start_position: %d" % (self.start_position,)
     if self.start_position:
       s += ", end_position: %d" % (self.end_position)
     if self.start_position:
@@ -776,7 +776,7 @@ def postprocess_output(all_examples,
               start_logit=pred.start_logit,
               end_logit=pred.end_logit))
 
-    # if we didn't inlude the empty option in the n-best, include it
+    # if we didn't include the empty option in the n-best, include it
     if version_2_with_negative and not xlnet_format:
       if "" not in seen_predictions:
         nbest.append(
diff --git a/official/nlp/data/wmt_dataloader.py b/official/nlp/data/wmt_dataloader.py
@@ -117,7 +117,7 @@ def _batch_examples(dataset, batch_size, max_length):
   # Validates bucket batch sizes.
   if any([batch_size <= 0 for batch_size in bucket_batch_sizes]):
     raise ValueError(
-        'The token budget, global batch size, is too small to yeild 0 bucket '
+        'The token budget, global batch size, is too small to yield 0 bucket '
         'window: %s' % str(bucket_batch_sizes))
 
   # bucket_id will be a tensor, so convert this list to a tensor as well.
diff --git a/official/nlp/modeling/ops/beam_search.py b/official/nlp/modeling/ops/beam_search.py
@@ -587,7 +587,7 @@ def _gather_beams(nested, beam_indices, batch_size, new_beam_size):
       Nested structure containing tensors with shape
         [batch_size, new_beam_size, ...]
     """
-    # Computes the i'th coodinate that contains the batch index for gather_nd.
+    # Computes the i'th coordinate that contains the batch index for gather_nd.
     # Batch pos is a tensor like [[0,0,0,0,],[1,1,1,1],..].
     batch_pos = tf.range(batch_size * new_beam_size) // new_beam_size
     batch_pos = tf.reshape(batch_pos, [batch_size, new_beam_size])
diff --git a/official/nlp/modeling/ops/segment_extractor.py b/official/nlp/modeling/ops/segment_extractor.py
@@ -27,7 +27,7 @@ def _get_random(positions, random_fn):
   return positions.with_flat_values(flat_random)
 
 
-# For every position j in a row, sample a position preceeding j or
+# For every position j in a row, sample a position preceding j or
 # a position which is [0, j-1]
 def _random_int_up_to(maxval, random_fn):
   # Need to cast because the int kernel for uniform doesn't support bcast.
@@ -87,22 +87,22 @@ def get_sentence_order_labels(sentences,
       dtype.
     random_threshold: (optional) A float threshold between 0 and 1, used to
       determine whether to extract a random, out-of-batch sentence or a
-      suceeding sentence. Higher value favors succeeding sentence.
+      succeeding sentence. Higher value favors succeeding sentence.
     random_next_threshold: (optional) A float threshold between 0 and 1, used to
       determine whether to extract either a random, out-of-batch, or succeeding
-      sentence or a preceeding sentence. Higher value favors preceeding
+      sentence or a preceding sentence. Higher value favors preceding
       sentences.
     random_fn: (optional) An op used to generate random float values.
 
   Returns:
-    a tuple of (preceeding_or_random_next, is_suceeding_or_random) where:
-      preceeding_or_random_next: a `RaggedTensor` of strings with the same shape
-        as `sentences` and contains either a preceeding, suceeding, or random
+    a tuple of (preceding_or_random_next, is_succeeding_or_random) where:
+      preceding_or_random_next: a `RaggedTensor` of strings with the same shape
+        as `sentences` and contains either a preceding, succeeding, or random
         out-of-batch sentence respective to its counterpart in `sentences` and
-        dependent on its label in `is_preceeding_or_random_next`.
-      is_suceeding_or_random: a `RaggedTensor` of bool values with the
+        dependent on its label in `is_preceding_or_random_next`.
+      is_succeeding_or_random: a `RaggedTensor` of bool values with the
         same shape as `sentences` and is True if it's corresponding sentence in
-        `preceeding_or_random_next` is a random or suceeding sentence, False
+        `preceding_or_random_next` is a random or succeeding sentence, False
         otherwise.
   """
   # Create a RaggedTensor in the same shape as sentences ([doc, (sentences)])

Original file line number	Diff line number	Diff line change
`@@ -884,7 +884,7 @@`
`884`	`884`	`"id": "2oHOql35k3Dd"`
`885`	`885`	`},`
`886`	`886`	`"source": [`
`887`		- "Note: The pretrained `TransformerEncoder` is also available on [TensorFlow Hub](https://tensorflow.org/hub). Go to the [TF Hub appendix](#hub_bert) for details."
	`887`	+ "Note: The pre-trained `TransformerEncoder` is also available on [TensorFlow Hub](https://tensorflow.org/hub). Go to the [TF Hub appendix](#hub_bert) for details."
`888`	`888`	`]`
`889`	`889`	`},`
`890`	`890`	`{`
`@@ -1148,8 +1148,8 @@`
`1148`	`1148`	`"\n",`
`1149`	`1149`	`"First, build a wrapper class to export the model. This wrapper does two things:\n",`
`1150`	`1150`	`"\n",`
`1151`		- "- First it packages `bert_inputs_processor` and `bert_classifier` together into a single `tf.Module`, so you can export all the functionalities.\n",
`1152`		- "- Second it defines a `tf.function` that implements the end-to-end execution of the model.\n",
	`1151`	+ "- First, it packages `bert_inputs_processor` and `bert_classifier` together into a single `tf.Module`, so you can export all the functionalities.\n",
	`1152`	+ "- Second, it defines a `tf.function` that implements the end-to-end execution of the model.\n",
`1153`	`1153`	`"\n",`
`1154`	`1154`	"Setting the `input_signature` argument of `tf.function` lets you define a fixed signature for the `tf.function`. This can be less surprising than the default automatic retracing behavior."
`1155`	`1155`	`]`
`@@ -1189,7 +1189,7 @@`
`1189`	`1189`	`"id": "qnxysGUfIgFQ"`
`1190`	`1190`	`},`
`1191`	`1191`	`"source": [`
`1192`		`- "Create an instance of this export-model and save it:"`
	`1192`	`+ "Create an instance of this exported model and save it:"`
`1193`	`1193`	`]`
`1194`	`1194`	`},`
`1195`	`1195`	`{`
`@@ -1280,7 +1280,7 @@`
`1280`	`1280`	`"id": "CPsg7dZwfBM2"`
`1281`	`1281`	`},`
`1282`	`1282`	`"source": [`
`1283`		- "Congratulations! You've used `tensorflow_models` to build a BERT-classifier, train it, and export for later use."
	`1283`	+ "Congratulations! You've used `tensorflow_models` to build a BERT-classifier, train it, and export it for later use."
`1284`	`1284`	`]`
`1285`	`1285`	`},`
`1286`	`1286`	`{`
`@@ -1391,7 +1391,7 @@`
`1391`	`1391`	`"id": "cjojn8SmLSRI"`
`1392`	`1392`	`},`
`1393`	`1393`	`"source": [`
`1394`		`- "At this point it would be simple to add a classification head yourself.\n",`
	`1394`	`+ "At this point, it would be simple to add a classification head yourself.\n",`
`1395`	`1395`	`"\n",`
`1396`	`1396`	"The Model Garden `tfm.nlp.models.BertClassifier` class can also build a classifier onto the TF Hub encoder:"
`1397`	`1397`	`]`
`@@ -1429,7 +1429,7 @@`
`1429`	`1429`	`"id": "u_IqwXjRV1vd"`
`1430`	`1430`	`},`
`1431`	`1431`	`"source": [`
`1432`		`- "For concrete examples of this approach, refer to [Solve Glue tasks using BERT](https://www.tensorflow.org/text/tutorials/bert_glue)."`
	`1432`	`+ "For concrete examples of this approach, refer to [Solve Glue tasks using the BERT](https://www.tensorflow.org/text/tutorials/bert_glue)."`
`1433`	`1433`	`]`
`1434`	`1434`	`},`
`1435`	`1435`	`{`
`@@ -1494,7 +1494,7 @@`
`1494`	`1494`	`"id": "ywn5miD_dnuh"`
`1495`	`1495`	`},`
`1496`	`1496`	`"source": [`
`1497`		- "The advantage to using `config` objects is that they don't contain any complicated TensorFlow objects, and can be easily serialized to JSON, and rebuilt. Here's the JSON for the above `tfm.optimization.OptimizationConfig`:"
	`1497`	+ "The advantage of using `config` objects is that they don't contain any complicated TensorFlow objects, and can be easily serialized to JSON, and rebuilt. Here's the JSON for the above `tfm.optimization.OptimizationConfig`:"
`1498`	`1498`	`]`
`1499`	`1499`	`},`
`1500`	`1500`	`{`
Original file line number	Diff line number	Diff line change
`@@ -492,7 +492,7 @@ def _improve_answer_span(doc_tokens, input_start, input_end, tokenizer,`
`492`	`492`	`#`
`493`	`493`	`# However, this is not always possible. Consider the following:`
`494`	`494`	`#`
`495`		`- # Question: What country is the top exporter of electornics?`
	`495`	`+ # Question: What country is the top exporter of electronics?`
`496`	`496`	`# Context: The Japanese electronics industry is the lagest in the world.`
`497`	`497`	`# Answer: Japan`
`498`	`498`	`#`
`@@ -720,7 +720,7 @@ def postprocess_output(all_examples,`
`720`	`720`	`start_logit=pred.start_logit,`
`721`	`721`	`end_logit=pred.end_logit))`
`722`	`722`
`723`		`- # if we didn't inlude the empty option in the n-best, inlcude it`
	`723`	`+ # if we didn't include the empty option in the n-best, include it`
`724`	`724`	`if version_2_with_negative and not xlnet_format:`
`725`	`725`	`if "" not in seen_predictions:`
`726`	`726`	`nbest.append(`
`@@ -815,7 +815,7 @@ def get_final_text(pred_text, orig_text, do_lower_case, verbose=False):`
`815`	`815`	`# What we really want to return is "Steve Smith".`
`816`	`816`	`#`
`817`	`817`	`# Therefore, we have to apply a semi-complicated alignment heruistic between`
`818`		- # `pred_text` and `orig_text` to get a character-to-charcter alignment. This
	`818`	+ # `pred_text` and `orig_text` to get a character-to-character alignment. This
`819`	`819`	# can fail in certain cases in which case we just return `orig_text`.
`820`	`820`
`821`	`821`	`def _strip_spaces(text):`