Hugging Face's logo

Transformers documentation

VisualBERT

You are viewing v4.16.2 version. A newer version v4.46.3 is available.
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

VisualBERT

Overview

The VisualBERT model was proposed in VisualBERT: A Simple and Performant Baseline for Vision and Language by Liunian Harold Li, Mark Yatskar, Da Yin, Cho-Jui Hsieh, Kai-Wei Chang. VisualBERT is a neural network trained on a variety of (image, text) pairs.

The abstract from the paper is the following:

We propose VisualBERT, a simple and flexible framework for modeling a broad range of vision-and-language tasks. VisualBERT consists of a stack of Transformer layers that implicitly align elements of an input text and regions in an associated input image with self-attention. We further propose two visually-grounded language model objectives for pre-training VisualBERT on image caption data. Experiments on four vision-and-language tasks including VQA, VCR, NLVR2, and Flickr30K show that VisualBERT outperforms or rivals with state-of-the-art models while being significantly simpler. Further analysis demonstrates that VisualBERT can ground elements of language to image regions without any explicit supervision and is even sensitive to syntactic relationships, tracking, for example, associations between verbs and image regions corresponding to their arguments.

Tips:

  1. Most of the checkpoints provided work with the VisualBertForPreTraining configuration. Other checkpoints provided are the fine-tuned checkpoints for down-stream tasks - VQA (‘visualbert-vqa’), VCR (‘visualbert-vcr’), NLVR2 (‘visualbert-nlvr2’). Hence, if you are not working on these downstream tasks, it is recommended that you use the pretrained checkpoints.

  2. For the VCR task, the authors use a fine-tuned detector for generating visual embeddings, for all the checkpoints. We do not provide the detector and its weights as a part of the package, but it will be available in the research projects, and the states can be loaded directly into the detector provided.

Usage

VisualBERT is a multi-modal vision and language model. It can be used for visual question answering, multiple choice, visual reasoning and region-to-phrase correspondence tasks. VisualBERT uses a BERT-like transformer to prepare embeddings for image-text pairs. Both the text and visual features are then projected to a latent space with identical dimension.

To feed images to the model, each image is passed through a pre-trained object detector and the regions and the bounding boxes are extracted. The authors use the features generated after passing these regions through a pre-trained CNN like ResNet as visual embeddings. They also add absolute position embeddings, and feed the resulting sequence of vectors to a standard BERT model. The text input is concatenated in the front of the visual embeddings in the embedding layer, and is expected to be bound by [CLS] and a [SEP] tokens, as in BERT. The segment IDs must also be set appropriately for the textual and visual parts.

The BertTokenizer is used to encode the text. A custom detector/feature extractor must be used to get the visual embeddings. The following example notebooks show how to use VisualBERT with Detectron-like models:

The following example shows how to get the last hidden state using VisualBertModel:

>>> import torch
>>> from transformers import BertTokenizer, VisualBertModel

>>> model = VisualBertModel.from_pretrained("uclanlp/visualbert-vqa-coco-pre")
>>> tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")

>>> inputs = tokenizer("What is the man eating?", return_tensors="pt")
>>> # this is a custom function that returns the visual embeddings given the image path
>>> visual_embeds = get_visual_embeddings(image_path)

>>> visual_token_type_ids = torch.ones(visual_embeds.shape[:-1], dtype=torch.long)
>>> visual_attention_mask = torch.ones(visual_embeds.shape[:-1], dtype=torch.float)
>>> inputs.update(
...     {
...         "visual_embeds": visual_embeds,
...         "visual_token_type_ids": visual_token_type_ids,
...         "visual_attention_mask": visual_attention_mask,
...     }
... )
>>> outputs = model(**inputs)
>>> last_hidden_state = outputs.last_hidden_state

This model was contributed by gchhablani. The original code can be found here.

VisualBertConfig

class transformers.VisualBertConfig < >

( vocab_size = 30522 hidden_size = 768 visual_embedding_dim = 512 num_hidden_layers = 12 num_attention_heads = 12 intermediate_size = 3072 hidden_act = 'gelu' hidden_dropout_prob = 0.1 attention_probs_dropout_prob = 0.1 max_position_embeddings = 512 type_vocab_size = 2 initializer_range = 0.02 layer_norm_eps = 1e-12 bypass_transformer = False special_visual_initialize = True pad_token_id = 1 bos_token_id = 0 eos_token_id = 2 **kwargs )

Parameters

  • vocab_size (int, optional, defaults to 30522) — Vocabulary size of the VisualBERT model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling VisualBertModel. Vocabulary size of the model. Defines the different tokens that can be represented by the inputs_ids passed to the forward method of VisualBertModel.
  • hidden_size (int, optional, defaults to 768) — Dimensionality of the encoder layers and the pooler layer.
  • visual_embedding_dim (int, optional, defaults to 512) — Dimensionality of the visual embeddings to be passed to the model.
  • num_hidden_layers (int, optional, defaults to 12) — Number of hidden layers in the Transformer encoder.
  • num_attention_heads (int, optional, defaults to 12) — Number of attention heads for each attention layer in the Transformer encoder.
  • intermediate_size (int, optional, defaults to 3072) — Dimensionality of the “intermediate” (i.e., feed-forward) layer in the Transformer encoder.
  • hidden_act (str or function, optional, defaults to "gelu") — The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu", "relu", "selu" and "gelu_new" are supported.
  • hidden_dropout_prob (float, optional, defaults to 0.1) — The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.
  • attention_probs_dropout_prob (float, optional, defaults to 0.1) — The dropout ratio for the attention probabilities.
  • max_position_embeddings (int, optional, defaults to 512) — The maximum sequence length that this model might ever be used with. Typically set this to something large just in case (e.g., 512 or 1024 or 2048).
  • type_vocab_size (int, optional, defaults to 2) — The vocabulary size of the token_type_ids passed when calling VisualBertModel.
  • initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
  • layer_norm_eps (float, optional, defaults to 1e-12) — The epsilon used by the layer normalization layers.
  • bypass_transformer (bool, optional, defaults to False) — Whether or not the model should bypass the transformer for the visual embeddings. If set to True, the model directly concatenates the visual embeddings from VisualBertEmbeddings with text output from transformers, and then pass it to a self-attention layer.
  • special_visual_initialize (bool, optional, defaults to True) — Whether or not the visual token type and position type embedding weights should be initialized the same as the textual token type and positive type embeddings. When set to True, the weights of the textual token type and position type embeddings are copied to the respective visual embedding layers.

This is the configuration class to store the configuration of a VisualBertModel. It is used to instantiate an VisualBERT model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the VisualBERT visualbert-vqa-coco-pre architecture.

Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.

Example:

>>> from transformers import VisualBertModel, VisualBertConfig

>>> # Initializing a VisualBERT visualbert-vqa-coco-pre style configuration
>>> configuration = VisualBertConfig.from_pretrained("visualbert-vqa-coco-pre")

>>> # Initializing a model from the visualbert-vqa-coco-pre style configuration
>>> model = VisualBertModel(configuration)

>>> # Accessing the model configuration
>>> configuration = model.config

VisualBertModel

class transformers.VisualBertModel < >

( config add_pooling_layer = True )

Parameters

  • config (VisualBertConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The bare VisualBert Model transformer outputting raw hidden-states without any specific head on top. This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

The model can behave as an encoder (with only self-attention) following the architecture described in Attention is all you need by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.

forward < >

( input_ids = None attention_mask = None token_type_ids = None position_ids = None head_mask = None inputs_embeds = None visual_embeds = None visual_attention_mask = None visual_token_type_ids = None image_text_alignment = None output_attentions = None output_hidden_states = None return_dict = None ) → transformers.modeling_outputs.BaseModelOutputWithPooling or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary.

    Indices can be obtained using BertTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,
    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

    • 1 indicates the head is not masked,
    • 0 indicates the head is masked.
  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • visual_embeds (torch.FloatTensor of shape (batch_size, visual_seq_length, visual_embedding_dim), optional) — The embedded representation of the visual inputs, generally derived using using an object detector.
  • visual_attention_mask (torch.FloatTensor of shape (batch_size, visual_seq_length), optional) — Mask to avoid performing attention on visual embeddings. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • visual_token_type_ids (torch.LongTensor of shape (batch_size, visual_seq_length), optional) — Segment token indices to indicate different portions of the visual embeds.

    What are token type IDs? The authors of VisualBERT set the visual_token_type_ids to 1 for all tokens.

  • image_text_alignment (torch.LongTensor of shape (batch_size, visual_seq_length, alignment_number), optional) — Image-Text alignment uses to decide the position IDs of the visual embeddings.
  • output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
  • output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
  • return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.

Returns

transformers.modeling_outputs.BaseModelOutputWithPooling or tuple(torch.FloatTensor)

A transformers.modeling_outputs.BaseModelOutputWithPooling or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (VisualBertConfig) and inputs.

  • last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Sequence of hidden-states at the output of the last layer of the model.

  • pooler_output (torch.FloatTensor of shape (batch_size, hidden_size)) — Last layer hidden-state of the first token of the sequence (classification token) after further processing through the layers used for the auxiliary pretraining task. E.g. for BERT-family of models, this returns the classification token after processing through a linear layer and a tanh activation function. The linear layer weights are trained from the next sentence prediction (classification) objective during pretraining.

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The VisualBertModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

# Assumption: *get_visual_embeddings(image)* gets the visual embeddings of the image.
from transformers import BertTokenizer, VisualBertModel
import torch

tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
model = VisualBertModel.from_pretrained("uclanlp/visualbert-vqa-coco-pre")

inputs = tokenizer("The capital of France is Paris.", return_tensors="pt")
visual_embeds = get_visual_embeddings(image).unsqueeze(0)
visual_token_type_ids = torch.ones(visual_embeds.shape[:-1], dtype=torch.long)
visual_attention_mask = torch.ones(visual_embeds.shape[:-1], dtype=torch.float)

inputs.update(
    {
        "visual_embeds": visual_embeds,
        "visual_token_type_ids": visual_token_type_ids,
        "visual_attention_mask": visual_attention_mask,
    }
)

outputs = model(**inputs)

last_hidden_states = outputs.last_hidden_state

VisualBertForPreTraining

class transformers.VisualBertForPreTraining < >

( config )

Parameters

  • config (VisualBertConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

VisualBert Model with two heads on top as done during the pretraining: a masked language modeling head and a sentence-image prediction (classification) head.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None attention_mask = None token_type_ids = None position_ids = None head_mask = None inputs_embeds = None visual_embeds = None visual_attention_mask = None visual_token_type_ids = None image_text_alignment = None output_attentions = None output_hidden_states = None return_dict = None labels = None sentence_image_labels = None ) → transformers.models.visual_bert.modeling_visual_bert.VisualBertForPreTrainingOutputor tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary.

    Indices can be obtained using BertTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,
    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

    • 1 indicates the head is not masked,
    • 0 indicates the head is masked.
  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • visual_embeds (torch.FloatTensor of shape (batch_size, visual_seq_length, visual_embedding_dim), optional) — The embedded representation of the visual inputs, generally derived using using an object detector.
  • visual_attention_mask (torch.FloatTensor of shape (batch_size, visual_seq_length), optional) — Mask to avoid performing attention on visual embeddings. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • visual_token_type_ids (torch.LongTensor of shape (batch_size, visual_seq_length), optional) — Segment token indices to indicate different portions of the visual embeds.

    What are token type IDs? The authors of VisualBERT set the visual_token_type_ids to 1 for all tokens.

  • image_text_alignment (torch.LongTensor of shape (batch_size, visual_seq_length, alignment_number), optional) — Image-Text alignment uses to decide the position IDs of the visual embeddings.
  • output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
  • output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
  • return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
  • labels (torch.LongTensor of shape (batch_size, total_sequence_length), optional) — 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]
  • sentence_image_labels (torch.LongTensor of shape (batch_size,), optional) — Labels for computing the sentence-image prediction (classification) loss. Input should be a sequence pair (see input_ids docstring) Indices should be in [0, 1]:

    • 0 indicates sequence B is a matching pair of sequence A for the given image,
    • 1 indicates sequence B is a random sequence w.r.t A for the given image.

Returns

transformers.models.visual_bert.modeling_visual_bert.VisualBertForPreTrainingOutputor tuple(torch.FloatTensor)

A transformers.models.visual_bert.modeling_visual_bert.VisualBertForPreTrainingOutputor a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (VisualBertConfig) and inputs.

  • loss (optional, returned when labels is provided, torch.FloatTensor of shape (1,)) — Total loss as the sum of the masked language modeling loss and the sentence-image prediction (classification) loss.

  • prediction_logits (torch.FloatTensor of shape (batch_size, sequence_length, config.vocab_size)) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).

  • seq_relationship_logits (torch.FloatTensor of shape (batch_size, 2)) — Prediction scores of the sentence-image prediction (classification) head (scores of True/False continuation before SoftMax).

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The VisualBertForPreTraining forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

# Assumption: *get_visual_embeddings(image)* gets the visual embeddings of the image in the batch.
from transformers import BertTokenizer, VisualBertForPreTraining

tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
model = VisualBertForPreTraining.from_pretrained("uclanlp/visualbert-vqa-coco-pre")

inputs = tokenizer("The capital of France is {mask}.", return_tensors="pt")
visual_embeds = get_visual_embeddings(image).unsqueeze(0)
visual_token_type_ids = torch.ones(visual_embeds.shape[:-1], dtype=torch.long)
visual_attention_mask = torch.ones(visual_embeds.shape[:-1], dtype=torch.float)

inputs.update(
    {
        "visual_embeds": visual_embeds,
        "visual_token_type_ids": visual_token_type_ids,
        "visual_attention_mask": visual_attention_mask,
    }
)
max_length = inputs["input_ids"].shape[-1] + visual_embeds.shape[-2]
labels = tokenizer(
    "The capital of France is Paris.", return_tensors="pt", padding="max_length", max_length=max_length
)["input_ids"]
sentence_image_labels = torch.tensor(1).unsqueeze(0)  # Batch_size


outputs = model(**inputs, labels=labels, sentence_image_labels=sentence_image_labels)
loss = outputs.loss
prediction_logits = outputs.prediction_logits
seq_relationship_logits = outputs.seq_relationship_logits

VisualBertForQuestionAnswering

class transformers.VisualBertForQuestionAnswering < >

( config )

Parameters

  • config (VisualBertConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

VisualBert Model with a classification/regression head on top (a dropout and a linear layer on top of the pooled output) for VQA.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None attention_mask = None token_type_ids = None position_ids = None head_mask = None inputs_embeds = None visual_embeds = None visual_attention_mask = None visual_token_type_ids = None image_text_alignment = None output_attentions = None output_hidden_states = None return_dict = None labels = None ) → transformers.modeling_outputs.SequenceClassifierOutput or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary.

    Indices can be obtained using BertTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,
    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

    • 1 indicates the head is not masked,
    • 0 indicates the head is masked.
  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • visual_embeds (torch.FloatTensor of shape (batch_size, visual_seq_length, visual_embedding_dim), optional) — The embedded representation of the visual inputs, generally derived using using an object detector.
  • visual_attention_mask (torch.FloatTensor of shape (batch_size, visual_seq_length), optional) — Mask to avoid performing attention on visual embeddings. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • visual_token_type_ids (torch.LongTensor of shape (batch_size, visual_seq_length), optional) — Segment token indices to indicate different portions of the visual embeds.

    What are token type IDs? The authors of VisualBERT set the visual_token_type_ids to 1 for all tokens.

  • image_text_alignment (torch.LongTensor of shape (batch_size, visual_seq_length, alignment_number), optional) — Image-Text alignment uses to decide the position IDs of the visual embeddings.
  • output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
  • output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
  • return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
  • labels (torch.LongTensor of shape (batch_size, total_sequence_length), optional) — Labels for computing the sequence classification/regression loss. Indices should be in [0, ..., config.num_labels - 1]. A KLDivLoss is computed between the labels and the returned logits.

Returns

transformers.modeling_outputs.SequenceClassifierOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.SequenceClassifierOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (VisualBertConfig) and inputs.

  • loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Classification (or regression if config.num_labels==1) loss.

  • logits (torch.FloatTensor of shape (batch_size, config.num_labels)) — Classification (or regression if config.num_labels==1) scores (before SoftMax).

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The VisualBertForQuestionAnswering forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

# Assumption: *get_visual_embeddings(image)* gets the visual embeddings of the image in the batch.
from transformers import BertTokenizer, VisualBertForQuestionAnswering
import torch

tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
model = VisualBertForQuestionAnswering.from_pretrained("uclanlp/visualbert-vqa")

text = "Who is eating the apple?"
inputs = tokenizer(text, return_tensors="pt")
visual_embeds = get_visual_embeddings(image).unsqueeze(0)
visual_token_type_ids = torch.ones(visual_embeds.shape[:-1], dtype=torch.long)
visual_attention_mask = torch.ones(visual_embeds.shape[:-1], dtype=torch.float)

inputs.update(
    {
        "visual_embeds": visual_embeds,
        "visual_token_type_ids": visual_token_type_ids,
        "visual_attention_mask": visual_attention_mask,
    }
)

labels = torch.tensor([[0.0, 1.0]]).unsqueeze(0)  # Batch size 1, Num labels 2

outputs = model(**inputs, labels=labels)
loss = outputs.loss
scores = outputs.logits

VisualBertForMultipleChoice

class transformers.VisualBertForMultipleChoice < >

( config )

Parameters

  • config (VisualBertConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

VisualBert Model with a multiple choice classification head on top (a linear layer on top of the pooled output and a softmax) e.g. for VCR tasks.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None attention_mask = None token_type_ids = None position_ids = None head_mask = None inputs_embeds = None visual_embeds = None visual_attention_mask = None visual_token_type_ids = None image_text_alignment = None output_attentions = None output_hidden_states = None return_dict = None labels = None ) → transformers.modeling_outputs.MultipleChoiceModelOutput or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, num_choices, sequence_length)) — Indices of input sequence tokens in the vocabulary.

    Indices can be obtained using BertTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.FloatTensor of shape (batch_size, num_choices, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, num_choices, sequence_length), optional) — Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,
    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, num_choices, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

    • 1 indicates the head is not masked,
    • 0 indicates the head is masked.
  • inputs_embeds (torch.FloatTensor of shape (batch_size, num_choices, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • visual_embeds (torch.FloatTensor of shape (batch_size, visual_seq_length, visual_embedding_dim), optional) — The embedded representation of the visual inputs, generally derived using using an object detector.
  • visual_attention_mask (torch.FloatTensor of shape (batch_size, visual_seq_length), optional) — Mask to avoid performing attention on visual embeddings. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • visual_token_type_ids (torch.LongTensor of shape (batch_size, visual_seq_length), optional) — Segment token indices to indicate different portions of the visual embeds.

    What are token type IDs? The authors of VisualBERT set the visual_token_type_ids to 1 for all tokens.

  • image_text_alignment (torch.LongTensor of shape (batch_size, visual_seq_length, alignment_number), optional) — Image-Text alignment uses to decide the position IDs of the visual embeddings.
  • output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
  • output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
  • return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
  • labels (torch.LongTensor of shape (batch_size,), optional) — Labels for computing the multiple choice classification loss. Indices should be in [0, ..., num_choices-1] where num_choices is the size of the second dimension of the input tensors. (See input_ids above)

Returns

transformers.modeling_outputs.MultipleChoiceModelOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.MultipleChoiceModelOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (VisualBertConfig) and inputs.

  • loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Classification loss.

  • logits (torch.FloatTensor of shape (batch_size, num_choices)) — num_choices is the second dimension of the input tensors. (see input_ids above).

    Classification scores (before SoftMax).

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The VisualBertForMultipleChoice forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

# Assumption: *get_visual_embeddings(image)* gets the visual embeddings of the image in the batch.
from transformers import BertTokenizer, VisualBertForMultipleChoice
import torch

tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
model = VisualBertForMultipleChoice.from_pretrained("uclanlp/visualbert-vcr")

prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
choice0 = "It is eaten with a fork and a knife."
choice1 = "It is eaten while held in the hand."

visual_embeds = get_visual_embeddings(image)
# (batch_size, num_choices, visual_seq_length, visual_embedding_dim)
visual_embeds = visual_embeds.expand(1, 2, *visual_embeds.shape)
visual_token_type_ids = torch.ones(visual_embeds.shape[:-1], dtype=torch.long)
visual_attention_mask = torch.ones(visual_embeds.shape[:-1], dtype=torch.float)

labels = torch.tensor(0).unsqueeze(0)  # choice0 is correct (according to Wikipedia ;)), batch size 1

encoding = tokenizer([[prompt, prompt], [choice0, choice1]], return_tensors="pt", padding=True)
# batch size is 1
inputs_dict = {k: v.unsqueeze(0) for k, v in encoding.items()}
inputs_dict.update(
    {
        "visual_embeds": visual_embeds,
        "visual_attention_mask": visual_attention_mask,
        "visual_token_type_ids": visual_token_type_ids,
        "labels": labels,
    }
)
outputs = model(**inputs_dict)

loss = outputs.loss
logits = outputs.logits

VisualBertForVisualReasoning

class transformers.VisualBertForVisualReasoning < >

( config )

Parameters

  • config (VisualBertConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

VisualBert Model with a sequence classification head on top (a dropout and a linear layer on top of the pooled output) for Visual Reasoning e.g. for NLVR task.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None attention_mask = None token_type_ids = None position_ids = None head_mask = None inputs_embeds = None visual_embeds = None visual_attention_mask = None visual_token_type_ids = None image_text_alignment = None output_attentions = None output_hidden_states = None return_dict = None labels = None ) → transformers.modeling_outputs.SequenceClassifierOutput or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary.

    Indices can be obtained using BertTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,
    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

    • 1 indicates the head is not masked,
    • 0 indicates the head is masked.
  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • visual_embeds (torch.FloatTensor of shape (batch_size, visual_seq_length, visual_embedding_dim), optional) — The embedded representation of the visual inputs, generally derived using using an object detector.
  • visual_attention_mask (torch.FloatTensor of shape (batch_size, visual_seq_length), optional) — Mask to avoid performing attention on visual embeddings. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • visual_token_type_ids (torch.LongTensor of shape (batch_size, visual_seq_length), optional) — Segment token indices to indicate different portions of the visual embeds.

    What are token type IDs? The authors of VisualBERT set the visual_token_type_ids to 1 for all tokens.

  • image_text_alignment (torch.LongTensor of shape (batch_size, visual_seq_length, alignment_number), optional) — Image-Text alignment uses to decide the position IDs of the visual embeddings.
  • output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
  • output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
  • return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
  • labels (torch.LongTensor of shape (batch_size,), optional) — Labels for computing the sequence classification/regression loss. Indices should be in [0, ..., config.num_labels - 1]. A classification loss is computed (Cross-Entropy) against these labels.

Returns

transformers.modeling_outputs.SequenceClassifierOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.SequenceClassifierOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (VisualBertConfig) and inputs.

  • loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Classification (or regression if config.num_labels==1) loss.

  • logits (torch.FloatTensor of shape (batch_size, config.num_labels)) — Classification (or regression if config.num_labels==1) scores (before SoftMax).

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The VisualBertForVisualReasoning forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

# Assumption: *get_visual_embeddings(image)* gets the visual embeddings of the image in the batch.
from transformers import BertTokenizer, VisualBertForVisualReasoning
import torch

tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
model = VisualBertForVisualReasoning.from_pretrained("uclanlp/visualbert-nlvr2")

text = "Who is eating the apple?"
inputs = tokenizer(text, return_tensors="pt")
visual_embeds = get_visual_embeddings(image).unsqueeze(0)
visual_token_type_ids = torch.ones(visual_embeds.shape[:-1], dtype=torch.long)
visual_attention_mask = torch.ones(visual_embeds.shape[:-1], dtype=torch.float)

inputs.update(
    {
        "visual_embeds": visual_embeds,
        "visual_token_type_ids": visual_token_type_ids,
        "visual_attention_mask": visual_attention_mask,
    }
)

labels = torch.tensor(1).unsqueeze(0)  # Batch size 1, Num choices 2

outputs = model(**inputs, labels=labels)
loss = outputs.loss
scores = outputs.logits

VisualBertForRegionToPhraseAlignment

class transformers.VisualBertForRegionToPhraseAlignment < >

( config )

Parameters

  • config (VisualBertConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

VisualBert Model with a Masked Language Modeling head and an attention layer on top for Region-to-Phrase Alignment e.g. for Flickr30 Entities task.

This model inherits from PreTrainedModel. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.)

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None attention_mask = None token_type_ids = None position_ids = None head_mask = None inputs_embeds = None visual_embeds = None visual_attention_mask = None visual_token_type_ids = None image_text_alignment = None output_attentions = None output_hidden_states = None return_dict = None region_to_phrase_position = None labels = None ) → transformers.modeling_outputs.SequenceClassifierOutput or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary.

    Indices can be obtained using BertTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

    What are input IDs?

  • attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) — Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]:

    • 0 corresponds to a sentence A token,
    • 1 corresponds to a sentence B token.

    What are token type IDs?

  • position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

    What are position IDs?

  • head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

    • 1 indicates the head is not masked,
    • 0 indicates the head is masked.
  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • visual_embeds (torch.FloatTensor of shape (batch_size, visual_seq_length, visual_embedding_dim), optional) — The embedded representation of the visual inputs, generally derived using using an object detector.
  • visual_attention_mask (torch.FloatTensor of shape (batch_size, visual_seq_length), optional) — Mask to avoid performing attention on visual embeddings. Mask values selected in [0, 1]:

    • 1 for tokens that are not masked,
    • 0 for tokens that are masked.

    What are attention masks?

  • visual_token_type_ids (torch.LongTensor of shape (batch_size, visual_seq_length), optional) — Segment token indices to indicate different portions of the visual embeds.

    What are token type IDs? The authors of VisualBERT set the visual_token_type_ids to 1 for all tokens.

  • image_text_alignment (torch.LongTensor of shape (batch_size, visual_seq_length, alignment_number), optional) — Image-Text alignment uses to decide the position IDs of the visual embeddings.
  • output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
  • output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
  • return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
  • region_to_phrase_position (torch.LongTensor of shape (batch_size, total_sequence_length), optional) — The positions depicting the position of the image embedding corresponding to the textual tokens.
  • labels (torch.LongTensor of shape (batch_size, total_sequence_length, visual_sequence_length), optional) — Labels for computing the masked language modeling loss. KLDivLoss is computed against these labels and the outputs from the attention layer.

Returns

transformers.modeling_outputs.SequenceClassifierOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.SequenceClassifierOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (VisualBertConfig) and inputs.

  • loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Classification (or regression if config.num_labels==1) loss.

  • logits (torch.FloatTensor of shape (batch_size, config.num_labels)) — Classification (or regression if config.num_labels==1) scores (before SoftMax).

  • hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

    Hidden-states of the model at the output of each layer plus the initial embedding outputs.

  • attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

    Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The VisualBertForRegionToPhraseAlignment forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

# Assumption: *get_visual_embeddings(image)* gets the visual embeddings of the image in the batch.
from transformers import BertTokenizer, VisualBertForRegionToPhraseAlignment
import torch

tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
model = VisualBertForRegionToPhraseAlignment.from_pretrained("uclanlp/visualbert-vqa-coco-pre")

text = "Who is eating the apple?"
inputs = tokenizer(text, return_tensors="pt")
visual_embeds = get_visual_embeddings(image).unsqueeze(0)
visual_token_type_ids = torch.ones(visual_embeds.shape[:-1], dtype=torch.long)
visual_attention_mask = torch.ones(visual_embeds.shape[:-1], dtype=torch.float)
region_to_phrase_position = torch.ones((1, inputs["input_ids"].shape[-1] + visual_embeds.shape[-2]))

inputs.update(
    {
        "region_to_phrase_position": region_to_phrase_position,
        "visual_embeds": visual_embeds,
        "visual_token_type_ids": visual_token_type_ids,
        "visual_attention_mask": visual_attention_mask,
    }
)

labels = torch.ones(
    (1, inputs["input_ids"].shape[-1] + visual_embeds.shape[-2], visual_embeds.shape[-2])
)  # Batch size 1

outputs = model(**inputs, labels=labels)
loss = outputs.loss
scores = outputs.logits
←ViTMAE Wav2Vec2→