Docs: formatting nits (#32247)

* doc formatting nits

* ignore non-autodocs

* Apply suggestions from code review
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* Update src/transformers/models/esm/modeling_esm.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* Update src/transformers/models/esm/modeling_esm.py
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* make fixup

---------
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

src/transformers/models/roberta_prelayernorm/modeling_roberta_prelayernorm.py

@@ -1075,7 +1075,7 @@ class RobertaPreLayerNormForMaskedLM(RobertaPreLayerNormPreTrainedModel):
             Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
             config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
             loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
-        kwargs (`Dict[str, any]`, optional, defaults to *{}*):
+        kwargs (`Dict[str, any]`, *optional*, defaults to `{}`):
             Used to hide legacy arguments that have been deprecated.
         """
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict

src/transformers/models/roc_bert/modeling_roc_bert.py

@@ -1151,7 +1151,7 @@ class RoCBertForPreTraining(RoCBertPreTrainedModel):
                  ignored (masked), the loss is only computed for the tokens with labels in `[0, ...,
                  config.vocab_size]`
 
-            kwargs (`Dict[str, any]`, optional, defaults to *{}*):
+            kwargs (`Dict[str, any]`, *optional*, defaults to *{}*):
                 Used to hide legacy arguments that have been deprecated.
 
         Returns:

src/transformers/models/seggpt/modeling_seggpt.py

@@ -59,7 +59,7 @@ class SegGptEncoderOutput(ModelOutput):
         attentions (`Tuple[torch.FloatTensor]`, `optional`, returned when `config.output_attentions=True`):
             Tuple of *torch.FloatTensor* (one for each layer) of shape
             `(batch_size, num_heads, seq_len, seq_len)`.
-        intermediate_hidden_states (`Tuple[torch.FloatTensor]`, `optional`, returned when `config.intermediate_hidden_state_indices` is set):
+        intermediate_hidden_states (`Tuple[torch.FloatTensor]`, *optional*, returned when `config.intermediate_hidden_state_indices` is set):
             Tuple of `torch.FloatTensor` of shape `(batch_size, patch_height, patch_width, hidden_size)`.
             Each element in the Tuple corresponds to the output of the layer specified in `config.intermediate_hidden_state_indices`.
             Additionaly, each feature passes through a LayerNorm.
@@ -77,7 +77,7 @@ class SegGptImageSegmentationOutput(ModelOutput):
     Output type of [`SegGptImageSegmentationOutput`].
 
     Args:
-        loss (`torch.FloatTensor`, `optional`, returned when `labels` is provided):
+        loss (`torch.FloatTensor`, *optional*, returned when `labels` is provided):
             The loss value.
         pred_masks (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`):
             The predicted masks.

src/transformers/models/sew_d/modeling_sew_d.py

@@ -745,10 +745,10 @@ class DisentangledSelfAttention(nn.Module):
                 sequence length in which element [i,j] = *1* means the *i* th token in the input can attend to the *j*
                 th token.
 
-            output_attentions (`bool`, optional):
+            output_attentions (`bool`, *optional*):
                 Whether return the attention matrix.
 
-            query_states (`torch.FloatTensor`, optional):
+            query_states (`torch.FloatTensor`, *optional*):
                 The *Q* state in *Attention(Q,K,V)*.
 
             relative_pos (`torch.LongTensor`):

src/transformers/models/speech_to_text/feature_extraction_speech_to_text.py

@@ -220,7 +220,7 @@ class Speech2TextFeatureExtractor(SequenceFeatureExtractor):
             sampling_rate (`int`, *optional*):
                 The sampling rate at which the `raw_speech` input was sampled. It is strongly recommended to pass
                 `sampling_rate` at the forward call to prevent silent errors.
-            padding_value (`float`, defaults to 0.0):
+            padding_value (`float`, *optional*, defaults to 0.0):
                 The value that is used to fill the padding values / vectors.
         """
 

src/transformers/models/t5/modeling_t5.py

@@ -181,7 +181,7 @@ PARALLELIZE_DOCSTRING = r"""
     it will evenly distribute blocks across all devices.
 
     Args:
-        device_map (`Dict[int, list]`, optional, defaults to None):
+        device_map (`Dict[int, list]`, *optional*):
             A dictionary that maps attention modules to devices. Note that the embedding module and LMHead are always
             automatically mapped to the first device (for esoteric reasons). That means that the first device should
             have fewer attention modules mapped to it than other devices. For reference, the t5 models have the

src/transformers/models/tapas/tokenization_tapas.py

@@ -1249,7 +1249,7 @@ class TapasTokenizer(PreTrainedTokenizer):
                 Total number of table columns
             max_length (`int`):
                 Total maximum length.
-            truncation_strategy (`str` or [`TapasTruncationStrategy`]):
+            truncation_strategy (`str` or [`TapasTruncationStrategy]`):
                 Truncation strategy to use. Seeing as this method should only be called when truncating, the only
                 available strategy is the `"drop_rows_to_fit"` strategy.
 

src/transformers/models/udop/tokenization_udop.py

@@ -833,7 +833,7 @@ class UdopTokenizer(PreTrainedTokenizer):
         </Tip>
 
         Args:
-            text (`str`, `List[str]` or `List[int]` (the latter only for not-fast tokenizers)):
+            text (`str`, `List[str]` or (for non-fast tokenizers) `List[int]`):
                 The first sequence to be encoded. This can be a string, a list of strings (tokenized string using the
                 `tokenize` method) or a list of integers (tokenized string ids using the `convert_tokens_to_ids`
                 method).

src/transformers/models/udop/tokenization_udop_fast.py

@@ -814,7 +814,7 @@ class UdopTokenizerFast(PreTrainedTokenizerFast):
         </Tip>
 
         Args:
-            text (`str`, `List[str]` or `List[int]` (the latter only for not-fast tokenizers)):
+            text (`str`, `List[str]` or (for non-fast tokenizers) `List[int]`):
                 The first sequence to be encoded. This can be a string, a list of strings (tokenized string using the
                 `tokenize` method) or a list of integers (tokenized string ids using the `convert_tokens_to_ids`
                 method).

src/transformers/models/vilt/image_processing_vilt.py

@@ -243,7 +243,7 @@ class ViltImageProcessor(BaseImageProcessor):
                 Image to resize.
             size (`Dict[str, int]`):
                 Controls the size of the output image. Should be of the form `{"shortest_edge": int}`.
-            size_divisor (`int`, defaults to 32):
+            size_divisor (`int`, *optional*, defaults to 32):
                 The image is resized to a size that is a multiple of this value.
             resample (`PILImageResampling` filter, *optional*, defaults to `PILImageResampling.BICUBIC`):
                 Resampling filter to use when resiizing the image.

src/transformers/models/vitdet/modeling_vitdet.py

@@ -182,7 +182,7 @@ def add_decomposed_relative_positions(attn, queries, rel_pos_h, rel_pos_w, q_siz
             Relative position embeddings (Lw, num_channels) for width axis.
         q_size (`Tuple[int]`):
             Spatial sequence size of query q with (queries_height, queries_width).
-        k_size (`Tuple[int]`]):
+        k_size (`Tuple[int]`):
             Spatial sequence size of key k with (keys_height, keys_width).
 
     Returns:

src/transformers/models/wav2vec2/feature_extraction_wav2vec2.py

@@ -36,11 +36,11 @@ class Wav2Vec2FeatureExtractor(SequenceFeatureExtractor):
     most of the main methods. Users should refer to this superclass for more information regarding those methods.
 
     Args:
-        feature_size (`int`, defaults to 1):
+        feature_size (`int`, *optional*, defaults to 1):
             The feature dimension of the extracted features.
-        sampling_rate (`int`, defaults to 16000):
+        sampling_rate (`int`, *optional*, defaults to 16000):
             The sampling rate at which the audio files should be digitalized expressed in hertz (Hz).
-        padding_value (`float`, defaults to 0.0):
+        padding_value (`float`, *optional*, defaults to 0.0):
             The value that is used to fill the padding values.
         do_normalize (`bool`, *optional*, defaults to `True`):
             Whether or not to zero-mean unit-variance normalize the input. Normalizing can help to significantly
@@ -166,7 +166,7 @@ class Wav2Vec2FeatureExtractor(SequenceFeatureExtractor):
             sampling_rate (`int`, *optional*):
                 The sampling rate at which the `raw_speech` input was sampled. It is strongly recommended to pass
                 `sampling_rate` at the forward call to prevent silent errors.
-            padding_value (`float`, defaults to 0.0):
+            padding_value (`float`, *optional*, defaults to 0.0):
         """
 
         if sampling_rate is not None:

src/transformers/models/wav2vec2_conformer/configuration_wav2vec2_conformer.py

@@ -184,9 +184,9 @@ class Wav2Vec2ConformerConfig(PretrainedConfig):
             If `"rotary"` position embeddings are used, defines the size of the embedding base.
         max_source_positions (`int`, *optional*, defaults to 5000):
             if `"relative"` position embeddings are used, defines the maximum source input positions.
-        conv_depthwise_kernel_size (`int`, defaults to 31):
+        conv_depthwise_kernel_size (`int`, *optional*, defaults to 31):
             Kernel size of convolutional depthwise 1D layer in Conformer blocks.
-        conformer_conv_dropout (`float`, defaults to 0.1):
+        conformer_conv_dropout (`float`, *optional*, defaults to 0.1):
             The dropout probability for all convolutional layers in Conformer blocks.
 
     Example:

src/transformers/models/whisper/feature_extraction_whisper.py

@@ -44,16 +44,16 @@ class WhisperFeatureExtractor(SequenceFeatureExtractor):
     Fourier Transform` which should match pytorch's `torch.stft` equivalent.
 
     Args:
-        feature_size (`int`, defaults to 80):
+        feature_size (`int`, *optional*, defaults to 80):
             The feature dimension of the extracted features.
-        sampling_rate (`int`, defaults to 16000):
+        sampling_rate (`int`, *optional*, defaults to 16000):
             The sampling rate at which the audio files should be digitalized expressed in hertz (Hz).
-        hop_length (`int`, defaults to 160):
+        hop_length (`int`, *optional*, defaults to 160):
             Length of the overlaping windows for the STFT used to obtain the Mel Frequency coefficients.
-        chunk_length (`int`, defaults to 30):
+        chunk_length (`int`, *optional*, defaults to 30):
             The maximum number of chuncks of `sampling_rate` samples used to trim and pad longer or shorter audio
             sequences.
-        n_fft (`int`, defaults to 400):
+        n_fft (`int`, *optional*, defaults to 400):
             Size of the Fourier transform.
         padding_value (`float`, *optional*, defaults to 0.0):
             Padding value used to pad the audio. Should correspond to silences.
@@ -231,7 +231,7 @@ class WhisperFeatureExtractor(SequenceFeatureExtractor):
                 The sampling rate at which the `raw_speech` input was sampled. It is strongly recommended to pass
                 `sampling_rate` at the forward call to prevent silent errors and allow automatic speech recognition
                 pipeline.
-            padding_value (`float`, defaults to 0.0):
+            padding_value (`float`, *optional*, defaults to 0.0):
                 The value that is used to fill the padding values / vectors.
             do_normalize (`bool`, *optional*, defaults to `False`):
                 Whether or not to zero-mean unit-variance normalize the input. Normalizing can help to significantly

src/transformers/models/whisper/generation_whisper.py

@@ -1368,7 +1368,7 @@ class WhisperGenerationMixin:
                 priority: 1) from the `generation_config.json` model file, if it exists; 2) from the model
                 configuration. Please note that unspecified parameters will inherit [`~generation.GenerationConfig`]'s
                 default values, whose documentation should be checked to parameterize generation.
-            num_segment_frames (`int`, defaults to 3000):
+            num_segment_frames (`int`, *optional*, defaults to 3000):
                 The number of log-mel frames the model expects
 
         Return:

src/transformers/models/whisper/tokenization_whisper.py

@@ -565,7 +565,7 @@ class WhisperTokenizer(PreTrainedTokenizer):
         Args:
             token_ids (`Union[int, List[int], np.ndarray, torch.Tensor, tf.Tensor]`):
                 List of tokenized input ids. Can be obtained using the `__call__` method.
-            time_precision (`float`, `optional`, defaults to 0.02):
+            time_precision (`float`, *optional*, defaults to 0.02):
                 The time ratio to convert from token to time.
         """
         offsets = []
@@ -615,7 +615,7 @@ class WhisperTokenizer(PreTrainedTokenizer):
         Compute the timestamp token ids for a given precision and save to least-recently used (LRU) cache.
 
         Args:
-            time_precision (`float`, `optional`, defaults to 0.02):
+            time_precision (`float`, *optional*, defaults to 0.02):
                 The time ratio to convert from token to time.
         """
         return self.convert_tokens_to_ids([("<|%.2f|>" % (i * time_precision)) for i in range(1500 + 1)])
@@ -671,7 +671,7 @@ class WhisperTokenizer(PreTrainedTokenizer):
             output_offsets (`bool`, *optional*, defaults to `False`):
                 Whether or not to output the offsets of the tokens. This should only be set if the model predicted
                 timestamps.
-            time_precision (`float`, `optional`, defaults to 0.02):
+            time_precision (`float`, *optional*, defaults to 0.02):
                 The time ratio to convert from token to time.
             decode_with_timestamps (`bool`, *optional*, defaults to `False`):
                 Whether or not to decode with timestamps included in the raw text.

src/transformers/models/whisper/tokenization_whisper_fast.py

@@ -207,7 +207,7 @@ class WhisperTokenizerFast(PreTrainedTokenizerFast):
         Args:
             token_ids (`Union[int, List[int], np.ndarray, torch.Tensor, tf.Tensor]`):
                 List of tokenized input ids. Can be obtained using the `__call__` method.
-            time_precision (`float`, `optional`, defaults to 0.02):
+            time_precision (`float`, *optional*, defaults to 0.02):
                 The time ratio to convert from token to time.
         """
         offsets = []
@@ -258,7 +258,7 @@ class WhisperTokenizerFast(PreTrainedTokenizerFast):
         Compute the timestamp token ids for a given precision and save to least-recently used (LRU) cache.
 
         Args:
-            time_precision (`float`, `optional`, defaults to 0.02):
+            time_precision (`float`, *optional*, defaults to 0.02):
                 The time ratio to convert from token to time.
         """
         return self.convert_tokens_to_ids([("<|%.2f|>" % (i * time_precision)) for i in range(1500 + 1)])
@@ -317,7 +317,7 @@ class WhisperTokenizerFast(PreTrainedTokenizerFast):
             output_offsets (`bool`, *optional*, defaults to `False`):
                 Whether or not to output the offsets of the tokens. This should only be set if the model predicted
                 timestamps.
-            time_precision (`float`, `optional`, defaults to 0.02):
+            time_precision (`float`, *optional*, defaults to 0.02):
                 The time ratio to convert from token to time.
             decode_with_timestamps (`bool`, *optional*, defaults to `False`):
                 Whether or not to decode with timestamps included in the raw text.

src/transformers/models/xlm_roberta/modeling_xlm_roberta.py

@@ -1081,7 +1081,7 @@ class XLMRobertaForMaskedLM(XLMRobertaPreTrainedModel):
             Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
             config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
             loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
-        kwargs (`Dict[str, any]`, optional, defaults to *{}*):
+        kwargs (`Dict[str, any]`, *optional*, defaults to `{}`):
             Used to hide legacy arguments that have been deprecated.
         """
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict

src/transformers/models/xlm_roberta_xl/modeling_xlm_roberta_xl.py

@@ -1039,7 +1039,7 @@ class XLMRobertaXLForMaskedLM(XLMRobertaXLPreTrainedModel):
             Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
             config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
             loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
-        kwargs (`Dict[str, any]`, optional, defaults to *{}*):
+        kwargs (`Dict[str, any]`, *optional*, defaults to `{}`):
             Used to hide legacy arguments that have been deprecated.
         """
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict

src/transformers/models/xmod/modeling_xmod.py

@@ -1173,7 +1173,7 @@ class XmodForMaskedLM(XmodPreTrainedModel):
             Labels for computing the masked language modeling loss. Indices should be in `[-100, 0, ...,
             config.vocab_size]` (see `input_ids` docstring) Tokens with indices set to `-100` are ignored (masked), the
             loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`
-        kwargs (`Dict[str, any]`, optional, defaults to *{}*):
+        kwargs (`Dict[str, any]`, *optional*, defaults to *{}*):
             Used to hide legacy arguments that have been deprecated.
         """
         return_dict = return_dict if return_dict is not None else self.config.use_return_dict