Sample represent some arbitary data. All datasets in antmmf must return an object of type ``Sample``. Args: init_dict (dict): Diction

Args: config: path of the yaml file or dict of the configuration Usage: .. code-block:: python # construct from any col

:class:`SampleList` is used to collate a list of :class:`Sample` into a batch during batch preparation. It can be thought of as a merger of list o

A graph encoder register for visual encoder, all other details can be seen from :class:`antmmf.modules.module_registry.ModuleRegistry`.

A class that keeps all conversation history.

Subclass torch's LayerNorm to handle fp16.

This structure stores a list of boxes as a Nx4 :external:py:class:`Tensor <torch.Tensor>`. It supports some common methods about boxes (`area

Subclass torch's LayerNorm to handle fp16.

Rescales the input PIL.Image to the given 'size'. 'size' will be the size of the smaller edge. For example, if height > width, then image will

Construct the embeddings from word, position and token_type embeddings.

A metric manager which can record the history states of your interested metrics, such as losses, accuracies, and memory usage, etc. The state

Wrapper class used by antmmf to initialized processor based on their ``type`` as passed in configuration. It retrieves the processor class reg

r""" :class:`~transformers.BertConfig` is the configuration class to store the configuration of a `BertModel`. Arguments: vocab_

Internally used by antmmf, Metrics acts as wrapper for handling calculation of metrics over various metrics specified by the model in the conf

It is inspired from Detr experimental results in Table. 3. PositionEnhancedTransformer passes positional encodings directly in attention

Converts a PIL.Image (RGB) or numpy.ndarray (H x W x C) in the range [0, 255] to a torch.FloatTensor of shape (C x H x W) in the range [0.0, 1.0]

Vision Transformer with support for patch or hybrid CNN input stage

Encode bbox position with LayoutLM 2d positon embedding style, which is quite different from Uniter/VilBERT style. Visual Embedding for

r""" CompGCN convolutional layer Args: in_channels (int): Input dimension out_channels (int): Output dimension num_rel

Drop paths (Stochastic Depth) per sample (when applied in main path of residual blocks).

r""" From: https://arxiv.org/pdf/1707.07998.pdf nonlinear_layer (f_a) : x\in R^m => y \in R^n # noqa \tilda{y} = tanh(Wx + b) g = sigm

Randomly horizontally flips the given PIL.Image with a probability of 0.5

This class computes an assignment between the targets and the predictions of the network For efficiency reasons, the targets don't include the no_

r""" Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs: **last_hidden_state**: ``torch.Float

Internal antmmf helper and wrapper class for all Loss classes. It makes sure that the value returned from a Loss class is a dict and contain p

r""" Constructs a Bailing2 image processor that dynamically resizes images based on the original images. Args: do_resize (`bool`, *op

A basic Swin Transformer layer for one stage. Args: dim (int): Number of feature channels depth (int): Depths of this stage.

Construct the embeddings from word, position and token_type embeddings.

r""" Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs: **last_hidden_state**: ``torch.Float

A classification layer for any model, all details can be seen from :class:`antmmf.modules.module_registry.ModuleRegistry`. We registered :cla

Runs end-to-end tokenziation.

r"""The graph isomorphism operator from the `"How Powerful are Graph Neural Networks?" <https://arxiv.org/abs/1810.00826>`_ paper .. math::

Graph attention layer for building GAT network. Version for adjacency matrix. Args: in_features(int): dimension of input tensor.

Classifier perform Hierarchical Softmax. Hierarchical Softmax is an alternative to softmax that is faster to evaluate: it is time to eva

r""" This is the configuration class to store the configuration of a [`Llama3dModel`]. It is used to instantiate an LLaMA model according to t

Mobile Inverted Residual Bottleneck Block. Args: block_args (namedtuple): BlockArgs, defined in utils.py. global_params (namedtup

dataset for Region & OCR & Image joint modelling

dataset for Region & OCR & Image joint modelling

dataset for Region & OCR & Image joint modelling

dataset for Region & OCR & Image joint modelling