Pads the `video` with the specified (height, width) `padding` and `mode`. Args: video (`np.ndarray`): The video to pad.

Parse command-line args into instances of the specified dataclass types. This relies on argparse's `ArgumentParser.parse_known_args`

Expands 2-dimensional `image` to 3 dimensions. Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Tensor`):

Get the value of the key if the key exists in the cache, otherwise return None. Move the key to the end of the cache to show that it

Will save the model, so you can reload it using `from_pretrained()`. Will only save from the main process.

Serializes this instance to a Python dictionary. Override the default [`~PreTrainedConfig.to_dict`]. Returns: `dict[str,

Returns the sequence length of the cached states. A layer index can be optionally passed.

Maybe initializes weights. If using a custom `PreTrainedModel`, you need to implement any initialization logic in `_init_weights`.

r""" Generates sequences of token ids for models with a language modeling head. <Tip warning={true}> Most generation-contro

r""" Save a generation configuration object to the directory `save_directory`, so that it can be re-loaded using the [`~GenerationConf

Create a sliding window causal mask based on the attention implementation used (stored in the config). This type of attention pattern was mos

Invert an attention mask (e.g., switches 0. and 1.). Args: encoder_attention_mask (`torch.Tensor`): An attention mask.

Accepts a library version and returns True if the current version of the library is greater than or equal to the given version. If `accept_de

Checks whether `char` is a control character.

Prepare image-like inputs for processing. Args: images (`ImageInput`): The image-like inputs to process.

Check that `layer_types` is correctly defined.

r""" encoder (`PreTrainedModel`, *optional*): The encoder model to use. decoder (`PreTrainedModel`, *optional*):

Attend to all tokens in masked rows from the expanded attention mask, for example the relevant first rows when using left padding. Th

Makes broadcastable attention and causal masks so that future and masked tokens are ignored. Arguments: word_attention_m

Find the correct padding/truncation strategy

Traces `root` and returns the corresponding FX `torch.fx.Graph` representation. `root` can either be a `torch.nn.Module` instance or

Checks that number of special tokens in text and processed text is same. The count can be different if tokenized text was truncated,

Return whether the cache data is initialized

Resizes `image`. Enforces conversion of input to PIL.Image. Args: image (`PIL.Image.Image` or `np.ndarray` or `torch.Ten

IMPORTANT NOTICE: This function is deprecated in favor of using the mask primitives in `masking_utils.py`, and will be removed in a future ve

Reads a tab separated value file.

Get the `out_features` and `out_indices` so that they are aligned. The logic is as follows: - `out_features` not set, `out_indices`

Loads the backbone model from a config object. If the config is from the backbone model itself, then we return a backbone model with randoml

Scikit / Keras interface to transformers' pipelines. This method will forward to __call__().

This class computes the losses using the class predictions, mask predictions and the contrastive queries. Oneformer calculates the c

r""" current_modality (`str`, *optional*, defaults to `"text"`): Default modality. Used to initialize the model.

Return a list mapping the tokens to the id of their original sentences: - `None` for special tokens added around or between sequ

Start the background generation thread.

Tokenize and prepare for the model a list of sequences or a list of pairs of sequences. Args: batch_text_or_text_pairs (

Returns maximum sequence length of the cache object. Dynamic caches do not have a maximum length.

Checks if `torch_npu` is installed and potentially if a NPU is in the environment

Temporarily updates the `os.environ` dictionary in-place. Similar to mockenv The `os.environ` dictionary is updated in-place so that the mod

Alternative helper method that does not use `argparse` at all, instead loading a json file and populating the dataclass types.

Reconstructs a list of videos in the original order.

Recursively reset all layers tensors

r""" current_modality (`str`, *optional*, defaults to `"text"`): Default modality. Used to initialize the model.

Detects whether the optional user-specified attention_mask & the automatically created causal mask can be ignored in case PyTorch's S

Converts an image to RGB format. Only converts if the image is of type PIL.Image.Image, otherwise returns the image as is. Args:

Pads the inputs to same length and returns attention_mask. Args: inputs (`BatchFeature`): Processed audi

Returns score indicating how similar two instances of the `Properties` tuple are. Rules are as follows: * Matching `type`

Calculates a spectrogram over one waveform using the Short-Time Fourier Transform. This function can create the following kinds of spectrogr

Get the index of the token in the encoded output comprising a character in the original string for a sequence of the batch.

Checks if XPU acceleration is available either via native PyTorch (>=2.6), `intel_extension_for_pytorch` or via stock PyTorch (>=2.4) and pot

Load adapter weights from file or remote Hub folder. If you are not familiar with adapters and PEFT methods, we invite you to read mo

Ensure that the output is a nested list of images. Args: images (`Union[list[ImageInput], ImageInput]`): The input image.

Return a copy of the `os.environ` object that sets up `PYTHONPATH` correctly, depending on the test suite it's invoked from. This is