Standardize docstring (keras-team#1516)

STYLE_GUIDE.md

@@ -116,7 +116,7 @@ class PositionEmbedding(keras.layers.Layer):
  Args:
  sequence_length: The maximum length of the dynamic sequence.
 
- Examples:
+ Example:
 
  Direct call.
  >>> layer = keras_nlp.layers.PositionEmbedding(sequence_length=10)

keras_nlp/layers/modeling/alibi_bias.py

@@ -43,7 +43,7 @@ class AlibiBias(keras.layers.Layer):
  multi-head attention layer of the transformer to add alibi bias to
  it. With shape `(batch_size, num_heads, query_length, key_length)`.
 
- Examples:
+ Example:
  ```python
  query_length = 10
  key_length = 10

keras_nlp/layers/modeling/f_net_encoder.py

@@ -50,7 +50,7 @@ class FNetEncoder(keras.layers.Layer):
  **kwargs: other keyword arguments passed to `keras.layers.Layer`,
  including `name`, `trainable`, `dtype` etc.
 
- Examples:
+ Example:
 
  ```python
  # Create a single FNet encoder layer.

keras_nlp/layers/modeling/masked_lm_head.py

@@ -62,7 +62,7 @@ class MaskedLMHead(keras.layers.Layer):
  **kwargs: other keyword arguments passed to `keras.layers.Layer`,
  including `name`, `trainable`, `dtype` etc.
 
- Examples:
+ Example:
 
  ```python
  batch_size = 16

keras_nlp/layers/modeling/position_embedding.py

@@ -45,7 +45,7 @@ class PositionEmbedding(keras.layers.Layer):
  compute the position embedding from. This is useful during cached
  decoding, where each position is predicted separately in a loop.
 
- Examples:
+ Example:
 
  Called directly on input.
  >>> layer = keras_nlp.layers.PositionEmbedding(sequence_length=10)

keras_nlp/layers/modeling/reversible_embedding.py

@@ -61,7 +61,7 @@ class ReversibleEmbedding(keras.layers.Embedding):
  from `output_dim` to `input_dim`, instead of a normal embedding
  call. Default to `False`.
 
- Examples:
+ Example:
  ```python
  batch_size = 16
  vocab_size = 100

keras_nlp/layers/modeling/sine_position_encoding.py

@@ -44,7 +44,7 @@ class SinePositionEncoding(keras.layers.Layer):
  compute the encoding from. This is useful during cached decoding,
  where each position is predicted separately in a loop.
 
- Examples:
+ Example:
  ```python
  # create a simple embedding layer with sinusoidal positional encoding
  seq_len = 100

keras_nlp/layers/modeling/token_and_position_embedding.py

@@ -49,7 +49,7 @@ class TokenAndPositionEmbedding(keras.layers.Layer):
  **kwargs: other keyword arguments passed to `keras.layers.Layer`,
  including `name`, `trainable`, `dtype` etc.
 
- Examples:
+ Example:
  ```python
  inputs = np.ones(shape=(1, 50), dtype="int32")
  embedding_layer = keras_nlp.layers.TokenAndPositionEmbedding(

keras_nlp/layers/modeling/transformer_decoder.py

@@ -72,7 +72,7 @@ class TransformerDecoder(keras.layers.Layer):
  **kwargs: other keyword arguments passed to `keras.layers.Layer`,
  including `name`, `trainable`, `dtype` etc.
 
- Examples:
+ Example:
  ```python
  # Create a single transformer decoder layer.
  decoder = keras_nlp.layers.TransformerDecoder(

keras_nlp/layers/modeling/transformer_encoder.py

@@ -61,7 +61,7 @@ class TransformerEncoder(keras.layers.Layer):
  **kwargs: other keyword arguments passed to `keras.layers.Layer`,
  including `name`, `trainable`, `dtype` etc.
 
- Examples:
+ Example:
 
  ```python
  # Create a single transformer encoder layer.

keras_nlp/models/albert/albert_backbone.py

@@ -77,7 +77,7 @@ class AlbertBackbone(Backbone):
  such as softmax and layer normalization, will always be done at
  float32 precision regardless of dtype.
 
- Examples:
+ Example:
  ```python
  input_data = {
  "token_ids": np.ones(shape=(1, 12), dtype="int32"),

keras_nlp/models/albert/albert_masked_lm.py

@@ -52,7 +52,7 @@ class AlbertMaskedLM(Task):
  `None`. If `None`, this model will not apply preprocessing, and
  inputs should be preprocessed before calling the model.
 
- Example usage:
+ Examples:
 
  Raw string data.
  ```python

keras_nlp/models/bert/bert_masked_lm.py

@@ -51,7 +51,7 @@ class BertMaskedLM(Task):
  `None`. If `None`, this model will not apply preprocessing, and
  inputs should be preprocessed before calling the model.
 
- Example usage:
+ Examples:
 
  Raw string data.
  ```python

keras_nlp/models/bloom/bloom_backbone.py

@@ -58,7 +58,7 @@ class BloomBackbone(Backbone):
  such as softmax and layer normalization, will always be done at
  float32 precision regardless of dtype.
 
- Examples:
+ Example:
  ```python
  input_data = {
  "token_ids": np.ones(shape=(1, 12), dtype="int32"),

keras_nlp/models/deberta_v3/deberta_v3_masked_lm.py

@@ -55,7 +55,7 @@ class DebertaV3MaskedLM(Task):
  `None`. If `None`, this model will not apply preprocessing, and
  inputs should be preprocessed before calling the model.
 
- Example usage:
+ Examples:
 
  Raw string data.
  ```python

keras_nlp/models/distil_bert/distil_bert_masked_lm.py

@@ -55,7 +55,7 @@ class DistilBertMaskedLM(Task):
  `None`. If `None`, this model will not apply preprocessing, and
  inputs should be preprocessed before calling the model.
 
- Example usage:
+ Examples:
 
  Raw string data.
  ```python

keras_nlp/models/electra/electra_backbone.py

@@ -63,7 +63,7 @@ class ElectraBackbone(Backbone):
  such as softmax and layer normalization, will always be done at
  float32 precision regardless of dtype.
 
- Examples:
+ Example:
  ```python
  input_data = {
  "token_ids": np.ones(shape=(1, 12), dtype="int32"),

keras_nlp/models/f_net/f_net_masked_lm.py

@@ -51,7 +51,7 @@ class FNetMaskedLM(Task):
  `None`. If `None`, this model will not apply preprocessing, and
  inputs should be preprocessed before calling the model.
 
- Example usage:
+ Examples:
 
  Raw string data.
  ```python

keras_nlp/models/gemma/gemma_backbone.py

@@ -60,7 +60,7 @@ class GemmaBackbone(Backbone):
  computations, such as softmax and layer normalization will always
  be done a float32 precision regardless of dtype.
 
- Example usage:
+ Example:
  ```python
  input_data = {
  "token_ids": np.ones(shape=(1, 12), dtype="int32"),
@@ -205,7 +205,7 @@ def get_layout_map(
  backbone weights, so that you can use it to distribute weights across
  the accelerators.
 
- Sample usage:
+ Example:
  ```
  # Feel free to change the mesh shape to balance data and model parallel
  mesh = keras.distribution.DeviceMesh(

keras_nlp/models/gemma/gemma_causal_lm.py

@@ -359,7 +359,7 @@ def score(
  <float>[batch_size, num_tokens, vocab_size] in "logits" mode, or
  <float>[batch_size, num_tokens] in "loss" mode.
 
- Examples:
+ Example:
 
  Compute gradients between embeddings and loss scores with TensorFlow:
  ```python

keras_nlp/models/preprocessor.py

@@ -75,7 +75,7 @@ def from_preset(
  Args:
  preset: string. Must be one of "{{preset_names}}".
 
- Examples:
+ Example:
  ```python
  # Load a preprocessor layer from a preset.
  preprocessor = keras_nlp.models.{{preprocessor_name}}.from_preset(

keras_nlp/models/xlm_roberta/xlm_roberta_masked_lm.py

@@ -53,7 +53,7 @@ class XLMRobertaMaskedLM(Task):
  `None`. If `None`, this model will not apply preprocessing, and
  inputs should be preprocessed before calling the model.
 
- Example usage:
+ Examples:
 
  Raw string inputs and pretrained backbone.
  ```python

keras_nlp/models/xlnet/xlnet_backbone.py

@@ -65,7 +65,7 @@ class XLNetBackbone(Backbone):
  padding_mask: Mask to avoid performing attention on padding token indices
  of shape `[batch_size, sequence_length]`.
 
- Examples:
+ Example:
  ```python
  import numpy as np
  from keras_nlp.models import XLNetBackbone

keras_nlp/samplers/sampler.py

@@ -36,7 +36,7 @@ class Sampler:
  computes the next token based on a probability distribution over all
  possible vocab entries.
 
- Examples:
+ Example:
 
  ```python
  causal_lm = keras_nlp.models.GPT2CausalLM.from_preset("gpt2_base_en")

keras_nlp/tokenizers/byte_pair_tokenizer.py

@@ -156,7 +156,7 @@ class BytePairTokenizerCache(tf.Module):
  The cache key is string tensor or python strings, and the value is split
  tokens joined by whitespace. For example, "dragonfly" => "dragon fly"
 
- Examples:
+ Example:
  ```
  cache = BytePairTokenizerCache()
  cache.insert(["butterfly", "dragonfly"], ["but ter fly", "dragon fly"])
@@ -665,7 +665,7 @@ def from_preset(
  Args:
  preset: string. Must be one of "{{preset_names}}".
 
- Examples:
+ Example:
  ```python
  # Load a preset tokenizer.
  tokenizer = {{model_name}}.from_preset("{{example_preset_name}}")

keras_nlp/tokenizers/sentence_piece_tokenizer.py

@@ -275,7 +275,7 @@ def from_preset(
  Args:
  preset: string. Must be one of "{{preset_names}}".
 
- Examples:
+ Example:
  ```python
  # Load a preset tokenizer.
  tokenizer = {{model_name}}.from_preset("{{example_preset_name}}")

keras_nlp/tokenizers/tokenizer.py

@@ -40,7 +40,7 @@ class Tokenizer(PreprocessingLayer):
  "vocab free" tokenizers, such as a whitespace splitter show below, these
  methods do not apply and can be skipped.
 
- Examples:
+ Example:
 
  ```python
  class WhitespaceSplitterTokenizer(keras_nlp.tokenizers.Tokenizer):

keras_nlp/tokenizers/word_piece_tokenizer.py

@@ -480,7 +480,7 @@ def from_preset(
  Args:
  preset: string. Must be one of "{{preset_names}}".
 
- Examples:
+ Example:
  ```python
  # Load a preset tokenizer.
  tokenizer = {{model_name}}.from_preset("{{example_preset_name}}")