Transformers documentation

ALIGN

Join the Hugging Face community

to get started

# ALIGN

## Overview

The ALIGN model was proposed in Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision by Chao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen, Zarana Parekh, Hieu Pham, Quoc V. Le, Yunhsuan Sung, Zhen Li, Tom Duerig. ALIGN is a multi-modal vision and language model. It can be used for image-text similarity and for zero-shot image classification. ALIGN features a dual-encoder architecture with EfficientNet as its vision encoder and BERT as its text encoder, and learns to align visual and text representations with contrastive learning. Unlike previous work, ALIGN leverages a massive noisy dataset and shows that the scale of the corpus can be used to achieve SOTA representations with a simple recipe.

The abstract from the paper is the following:

Pre-trained representations are becoming crucial for many NLP and perception tasks. While representation learning in NLP has transitioned to training on raw text without human annotations, visual and vision-language representations still rely heavily on curated training datasets that are expensive or require expert knowledge. For vision applications, representations are mostly learned using datasets with explicit class labels such as ImageNet or OpenImages. For vision-language, popular datasets like Conceptual Captions, MSCOCO, or CLIP all involve a non-trivial data collection (and cleaning) process. This costly curation process limits the size of datasets and hence hinders the scaling of trained models. In this paper, we leverage a noisy dataset of over one billion image alt-text pairs, obtained without expensive filtering or post-processing steps in the Conceptual Captions dataset. A simple dual-encoder architecture learns to align visual and language representations of the image and text pairs using a contrastive loss. We show that the scale of our corpus can make up for its noise and leads to state-of-the-art representations even with such a simple learning scheme. Our visual representation achieves strong performance when transferred to classification tasks such as ImageNet and VTAB. The aligned visual and language representations enables zero-shot image classification and also set new state-of-the-art results on Flickr30K and MSCOCO image-text retrieval benchmarks, even when compared with more sophisticated cross-attention models. The representations also enable cross-modality search with complex text and text + image queries.

## Usage

ALIGN uses EfficientNet to get visual features and BERT to get the text features. Both the text and visual features are then projected to a latent space with identical dimension. The dot product between the projected image and text features is then used as a similarity score.

AlignProcessor wraps EfficientNetImageProcessor and BertTokenizer into a single instance to both encode the text and preprocess the images. The following example shows how to get the image-text similarity scores using AlignProcessor and AlignModel.

import requests
import torch
from PIL import Image
from transformers import AlignProcessor, AlignModel

processor = AlignProcessor.from_pretrained("kakaobrain/align-base")
model = AlignModel.from_pretrained("kakaobrain/align-base")

url = "http://images.cocodataset.org/val2017/000000039769.jpg"
image = Image.open(requests.get(url, stream=True).raw)
candidate_labels = ["an image of a cat", "an image of a dog"]

inputs = processor(text=candidate_labels, images=image, return_tensors="pt")

outputs = model(**inputs)

# this is the image-text similarity score
logits_per_image = outputs.logits_per_image

# we can take the softmax to get the label probabilities
probs = logits_per_image.softmax(dim=1)
print(probs)

This model was contributed by Alara Dirik. The original code is not released, this implementation is based on the Kakao Brain implementation based on the original paper.

## Resources

A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with ALIGN.

If you’re interested in submitting a resource to be included here, please feel free to open a Pull Request and we will review it. The resource should ideally demonstrate something new instead of duplicating an existing resource.

## AlignConfig

### class transformers.AlignConfig

< >

( text_config = None vision_config = None projection_dim = 640 temperature_init_value = 1.0 initializer_range = 0.02 **kwargs )

Parameters

• text_config (dict, optional) — Dictionary of configuration options used to initialize AlignTextConfig.
• vision_config (dict, optional) — Dictionary of configuration options used to initialize AlignVisionConfig.
• projection_dim (int, optional, defaults to 640) — Dimentionality of text and vision projection layers.
• temperature_init_value (float, optional, defaults to 1.0) — The inital value of the temperature paramter. Default is used as per the original ALIGN implementation.
• initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
• kwargs (optional) — Dictionary of keyword arguments.

AlignConfig is the configuration class to store the configuration of a AlignModel. It is used to instantiate a ALIGN model according to the specified arguments, defining the text model and vision model configs. Instantiating a configuration with the defaults will yield a similar configuration to that of the ALIGN kakaobrain/align-base 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 AlignConfig, AlignModel

>>> # Initializing a AlignConfig with kakaobrain/align-base style configuration
>>> configuration = AlignConfig()

>>> # Initializing a AlignModel (with random weights) from the kakaobrain/align-base style configuration
>>> model = AlignModel(configuration)

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

>>> # We can also initialize a AlignConfig from a AlignTextConfig and a AlignVisionConfig
>>> from transformers import AlignTextConfig, AlignVisionConfig

>>> # Initializing ALIGN Text and Vision configurations
>>> config_text = AlignTextConfig()
>>> config_vision = AlignVisionConfig()

>>> config = AlignConfig.from_text_vision_configs(config_text, config_vision)

#### from_text_vision_configs

< >

( text_config: AlignTextConfig vision_config: AlignVisionConfig **kwargs ) AlignConfig

Returns

AlignConfig

An instance of a configuration object

Instantiate a AlignConfig (or a derived class) from align text model configuration and align vision model configuration.

## AlignTextConfig

### class transformers.AlignTextConfig

< >

( vocab_size = 30522 hidden_size = 768 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 pad_token_id = 0 position_embedding_type = 'absolute' use_cache = True **kwargs )

Parameters

• vocab_size (int, optional, defaults to 30522) — Vocabulary size of the Align Text model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling AlignTextModel.
• hidden_size (int, optional, defaults to 768) — Dimensionality of the encoder layers and the pooler layer.
• 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” (often named feed-forward) layer in the Transformer encoder.
• hidden_act (str or Callable, optional, defaults to "gelu") — The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu", "relu", "silu" and "gelu_new" are supported.
• hidden_dropout_prob (float, optional, defaults to 0.1) — The dropout probability 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 AlignTextModel.
• 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.
• position_embedding_type (str, optional, defaults to "absolute") — Type of position embedding. Choose one of "absolute", "relative_key", "relative_key_query". For positional embeddings use "absolute". For more information on "relative_key", please refer to Self-Attention with Relative Position Representations (Shaw et al.). For more information on "relative_key_query", please refer to Method 4 in Improve Transformer Models with Better Relative Position Embeddings (Huang et al.).
• use_cache (bool, optional, defaults to True) — Whether or not the model should return the last key/values attentions (not used by all models). Only relevant if config.is_decoder=True.
• pad_token_id (int, optional, defaults to 0) — Padding token id.

This is the configuration class to store the configuration of a AlignTextModel. It is used to instantiate a ALIGN text encoder according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the text encoder of the ALIGN kakaobrain/align-base architecture. The default values here are copied from BERT.

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 AlignTextConfig, AlignTextModel

>>> # Initializing a AlignTextConfig with kakaobrain/align-base style configuration
>>> configuration = AlignTextConfig()

>>> # Initializing a AlignTextModel (with random weights) from the kakaobrain/align-base style configuration
>>> model = AlignTextModel(configuration)

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

## AlignVisionConfig

### class transformers.AlignVisionConfig

< >

( num_channels: int = 3 image_size: int = 600 width_coefficient: float = 2.0 depth_coefficient: float = 3.1 depth_divisor: int = 8 kernel_sizes: typing.List[int] = [3, 3, 5, 3, 5, 5, 3] in_channels: typing.List[int] = [32, 16, 24, 40, 80, 112, 192] out_channels: typing.List[int] = [16, 24, 40, 80, 112, 192, 320] depthwise_padding: typing.List[int] = [] strides: typing.List[int] = [1, 2, 2, 2, 1, 2, 1] num_block_repeats: typing.List[int] = [1, 2, 2, 3, 3, 4, 1] expand_ratios: typing.List[int] = [1, 6, 6, 6, 6, 6, 6] squeeze_expansion_ratio: float = 0.25 hidden_act: str = 'swish' hidden_dim: int = 2560 pooling_type: str = 'mean' initializer_range: float = 0.02 batch_norm_eps: float = 0.001 batch_norm_momentum: float = 0.99 dropout_rate: float = 0.5 drop_connect_rate: float = 0.2 **kwargs )

Parameters

• num_channels (int, optional, defaults to 3) — The number of input channels.
• image_size (int, optional, defaults to 600) — The input image size.
• width_coefficient (float, optional, defaults to 2.0) — Scaling coefficient for network width at each stage.
• depth_coefficient (float, optional, defaults to 3.1) — Scaling coefficient for network depth at each stage.
• depth_divisor int, optional, defaults to 8) — A unit of network width.
• kernel_sizes (List[int], optional, defaults to [3, 3, 5, 3, 5, 5, 3]) — List of kernel sizes to be used in each block.
• in_channels (List[int], optional, defaults to [32, 16, 24, 40, 80, 112, 192]) — List of input channel sizes to be used in each block for convolutional layers.
• out_channels (List[int], optional, defaults to [16, 24, 40, 80, 112, 192, 320]) — List of output channel sizes to be used in each block for convolutional layers.
• depthwise_padding (List[int], optional, defaults to []) — List of block indices with square padding. strides — (List[int], optional, defaults to [1, 2, 2, 2, 1, 2, 1]): List of stride sizes to be used in each block for convolutional layers.
• num_block_repeats (List[int], optional, defaults to [1, 2, 2, 3, 3, 4, 1]) — List of the number of times each block is to repeated.
• expand_ratios (List[int], optional, defaults to [1, 6, 6, 6, 6, 6, 6]) — List of scaling coefficient of each block.
• squeeze_expansion_ratio (float, optional, defaults to 0.25) — Squeeze expansion ratio.
• hidden_act (str or function, optional, defaults to "silu") — The non-linear activation function (function or string) in each block. If string, "gelu", "relu", "selu", “gelu_new”, “silu”and“mish” are supported.
• hiddem_dim (int, optional, defaults to 1280) — The hidden dimension of the layer before the classification head.
• pooling_type (str or function, optional, defaults to "mean") — Type of final pooling to be applied before the dense classification head. Available options are ["mean", "max"]
• initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
• batch_norm_eps (float, optional, defaults to 1e-3) — The epsilon used by the batch normalization layers.
• batch_norm_momentum (float, optional, defaults to 0.99) — The momentum used by the batch normalization layers.
• dropout_rate (float, optional, defaults to 0.5) — The dropout rate to be applied before final classifier layer.
• drop_connect_rate (float, optional, defaults to 0.2) — The drop rate for skip connections.

This is the configuration class to store the configuration of a AlignVisionModel. It is used to instantiate a ALIGN vision encoder according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the vision encoder of the ALIGN kakaobrain/align-base architecture. The default values are copied from EfficientNet (efficientnet-b7)

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 AlignVisionConfig, AlignVisionModel

>>> # Initializing a AlignVisionConfig with kakaobrain/align-base style configuration
>>> configuration = AlignVisionConfig()

>>> # Initializing a AlignVisionModel (with random weights) from the kakaobrain/align-base style configuration
>>> model = AlignVisionModel(configuration)

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

## AlignProcessor

### class transformers.AlignProcessor

< >

( image_processor tokenizer )

Parameters

• image_processor (EfficientNetImageProcessor) — The image processor is a required input.
• tokenizer ([BertTokenizer, BertTokenizerFast]) — The tokenizer is a required input.

Constructs an ALIGN processor which wraps EfficientNetImageProcessor and BertTokenizer/BertTokenizerFast into a single processor that interits both the image processor and tokenizer functionalities. See the __call__() and decode() for more information.

#### batch_decode

< >

( *args **kwargs )

This method forwards all its arguments to BertTokenizerFast’s batch_decode(). Please refer to the docstring of this method for more information.

#### decode

< >

( *args **kwargs )

This method forwards all its arguments to BertTokenizerFast’s decode(). Please refer to the docstring of this method for more information.

## AlignModel

### class transformers.AlignModel

< >

( config: AlignConfig )

Parameters

• config (AlignConfig) — 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.

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: typing.Optional[torch.LongTensor] = None pixel_values: typing.Optional[torch.FloatTensor] = None attention_mask: typing.Optional[torch.Tensor] = None token_type_ids: typing.Optional[torch.Tensor] = None position_ids: typing.Optional[torch.Tensor] = None head_mask: typing.Optional[torch.Tensor] = None inputs_embeds: typing.Optional[torch.Tensor] = None return_loss: typing.Optional[bool] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None )

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide it.

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

What are input IDs? attention_mask (torch.Tensor 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? 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? token_type_ids (torch.LongTensor of shape ({0}), 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? 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]:

inputs_embeds (torch.FloatTensor of shape ({0}, 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. pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)): Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using AutoImageProcessor. See EfficientNetImageProcessor.call() for details. return_loss (bool, optional): Whether or not to return the contrastive loss. 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.

The AlignModel 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.

#### get_text_features

< >

( input_ids: typing.Optional[torch.Tensor] = None attention_mask: typing.Optional[torch.Tensor] = None token_type_ids: typing.Optional[torch.Tensor] = None position_ids: typing.Optional[torch.Tensor] = None head_mask: typing.Optional[torch.Tensor] = None inputs_embeds: typing.Optional[torch.Tensor] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) text_features (torch.FloatTensor of shape (batch_size, output_dim)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide it.

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

What are input IDs?

• attention_mask (torch.Tensor 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.

• 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?

• token_type_ids (torch.LongTensor of shape ({0}), 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?

• 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]:

• inputs_embeds (torch.FloatTensor of shape ({0}, 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.
• 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

text_features (torch.FloatTensor of shape (batch_size, output_dim)

The text embeddings obtained by applying the projection layer to the pooled output of AlignTextModel.

The AlignModel 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.

Examples:

>>> from transformers import AutoTokenizer, AlignModel

>>> model = AlignModel.from_pretrained("kakaobrain/align-base")
>>> tokenizer = AutoTokenizer.from_pretrained("kakaobrain/align-base")

>>> inputs = tokenizer(["a photo of a cat", "a photo of a dog"], padding=True, return_tensors="pt")
>>> text_features = model.get_text_features(**inputs)

#### get_image_features

< >

( pixel_values: typing.Optional[torch.FloatTensor] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) image_features (torch.FloatTensor of shape (batch_size, output_dim)

Parameters

• pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using AutoImageProcessor. See EfficientNetImageProcessor.call() for details.
• 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

image_features (torch.FloatTensor of shape (batch_size, output_dim)

The image embeddings obtained by applying the projection layer to the pooled output of AlignVisionModel.

The AlignModel 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.

Examples:

>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, AlignModel

>>> model = AlignModel.from_pretrained("kakaobrain/align-base")
>>> processor = AutoProcessor.from_pretrained("kakaobrain/align-base")

>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)

>>> inputs = processor(images=image, return_tensors="pt")

>>> image_features = model.get_image_features(**inputs)

## AlignTextModel

### class transformers.AlignTextModel

< >

( config: AlignTextConfig add_pooling_layer: bool = True )

Parameters

• config (AlignConfig) — 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 text model from ALIGN without any head or projection 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.

#### forward

< >

( input_ids: typing.Optional[torch.Tensor] = None attention_mask: typing.Optional[torch.Tensor] = None token_type_ids: typing.Optional[torch.Tensor] = None position_ids: typing.Optional[torch.Tensor] = None head_mask: typing.Optional[torch.Tensor] = None inputs_embeds: typing.Optional[torch.Tensor] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) transformers.modeling_outputs.BaseModelOutputWithPoolingAndCrossAttentions or tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide it.

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

What are input IDs?

• attention_mask (torch.Tensor 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.

• 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?

• token_type_ids (torch.LongTensor of shape ({0}), 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?

• 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]:

• inputs_embeds (torch.FloatTensor of shape ({0}, 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.
• 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.BaseModelOutputWithPoolingAndCrossAttentions or tuple(torch.FloatTensor)

A transformers.modeling_outputs.BaseModelOutputWithPoolingAndCrossAttentions 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 (<class 'transformers.models.align.configuration_align.AlignTextConfig'>) 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, if the model has an embedding layer, + 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 optional 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.

• cross_attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True and config.add_cross_attention=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 of the decoder’s cross-attention layer, after the attention softmax, used to compute the weighted average in the cross-attention heads.

• past_key_values (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) — Tuple of tuple(torch.FloatTensor) of length config.n_layers, with each tuple having 2 tensors of shape (batch_size, num_heads, sequence_length, embed_size_per_head)) and optionally if config.is_encoder_decoder=True 2 additional tensors of shape (batch_size, num_heads, encoder_sequence_length, embed_size_per_head).

Contains pre-computed hidden-states (key and values in the self-attention blocks and optionally if config.is_encoder_decoder=True in the cross-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

The AlignTextModel 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.

Examples:

>>> from transformers import AutoTokenizer, AlignTextModel

>>> model = AlignTextModel.from_pretrained("kakaobrain/align-base")
>>> tokenizer = AutoTokenizer.from_pretrained("kakaobrain/align-base")

>>> inputs = tokenizer(["a photo of a cat", "a photo of a dog"], padding=True, return_tensors="pt")

>>> outputs = model(**inputs)
>>> last_hidden_state = outputs.last_hidden_state
>>> pooled_output = outputs.pooler_output  # pooled (EOS token) states

## AlignVisionModel

### class transformers.AlignVisionModel

< >

( config: AlignVisionConfig )

Parameters

• config (AlignConfig) — 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 vision model from ALIGN without any head or projection 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.

#### forward

< >

( pixel_values: typing.Optional[torch.FloatTensor] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) transformers.modeling_outputs.BaseModelOutputWithPoolingAndNoAttention or tuple(torch.FloatTensor)

Parameters

• pixel_values (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — Pixel values. Padding will be ignored by default should you provide it. Pixel values can be obtained using AutoImageProcessor. See EfficientNetImageProcessor.call() for details.
• 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.BaseModelOutputWithPoolingAndNoAttention or tuple(torch.FloatTensor)

A transformers.modeling_outputs.BaseModelOutputWithPoolingAndNoAttention 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 (<class 'transformers.models.align.configuration_align.AlignVisionConfig'>) and inputs.

• last_hidden_state (torch.FloatTensor of shape (batch_size, num_channels, height, width)) — 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 after a pooling operation on the spatial dimensions.

• 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, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, num_channels, height, width).

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

The AlignVisionModel 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.

Examples:

>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, AlignVisionModel

>>> model = AlignVisionModel.from_pretrained("kakaobrain/align-base")
>>> processor = AutoProcessor.from_pretrained("kakaobrain/align-base")

>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)

>>> inputs = processor(images=image, return_tensors="pt")

>>> outputs = model(**inputs)
>>> last_hidden_state = outputs.last_hidden_state
>>> pooled_output = outputs.pooler_output  # pooled CLS states`