[Bug] Extraneous back slashes

[iuliaturc]

When scraping https://huggingface.co/docs/transformers/main_classes/pipelines, I'm seeing a lot of back slashes:

Firecrawl Markdown:

### FillMaskPipeline\
\
### classtransformers.FillMaskPipeline\
\
[<source>](https://github.com/huggingface/transformers/blob/v4.21.2/src/transformers/pipelines/fill_mask.py#L34)\
\
(model: typing.Union\[ForwardRef('PreTrainedModel'), ForwardRef('TFPreTrainedModel')\]tokenizer: typing.Optional\[transformers.tokenization\_utils.PreTrainedTokenizer\] = Nonefeature\_extractor: typing.Optional\[ForwardRef('SequenceFeatureExtractor')\] = Nonemodelcard: typing.Optional\[transformers.modelcard.ModelCard\] = Noneframework: typing.Optional\[str\] = Nonetask: str = ''args\_parser: ArgumentHandler = Nonedevice: int = -1binary\_output: bool = False\*\*kwargs)\
\
Parameters\
\
- **model** ( [PreTrainedModel](/docs/transformers/v4.21.2/en/main_classes/model#transformers.PreTrainedModel) or [TFPreTrainedModel](/docs/transformers/v4.21.2/en/main_classes/model#transformers.TFPreTrainedModel)) —\
The model that will be used by the pipeline to make predictions. This needs to be a model inheriting from\
[PreTrainedModel](/docs/transformers/v4.21.2/en/main_classes/model#transformers.PreTrainedModel) for PyTorch and [TFPreTrainedModel](/docs/transformers/v4.21.2/en/main_classes/model#transformers.TFPreTrainedModel) for TensorFlow.\

Note these back slashes don't always show up. For instance, when I scrape https://huggingface.co/transformers/main_classes/tokenizer.html#transformers, I get cleaner Markdown:

## PreTrainedModel

### classtransformers.PreTrainedModel

[<source>](https://github.com/huggingface/transformers/blob/v4.44.2/src/transformers/modeling_utils.py#L1297)

(config: PretrainedConfig\*inputs\*\*kwargs)

Base class for all models.

[PreTrainedModel](/docs/transformers/v4.44.2/en/main_classes/model#transformers.PreTrainedModel) takes care of storing the configuration of the models and handles methods for loading,
downloading and saving models as well as a few methods common to all models to: