Pix2Struct

Pix2Struct

Overview

The Pix2Struct model was proposed in Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding by Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova.

The abstract from the paper is the following:

Visually-situated language is ubiquitous — sources range from textbooks with diagrams to web pages with images and tables, to mobile apps with buttons and forms. Perhaps due to this diversity, previous work has typically relied on domain-specific recipes with limited sharing of the underlying data, model architectures, and objectives. We present Pix2Struct, a pretrained image-to-text model for purely visual language understanding, which can be finetuned on tasks containing visually-situated language. Pix2Struct is pretrained by learning to parse masked screenshots of web pages into simplified HTML. The web, with its richness of visual elements cleanly reflected in the HTML structure, provides a large source of pretraining data well suited to the diversity of downstream tasks. Intuitively, this objective subsumes common pretraining signals such as OCR, language modeling, image captioning. In addition to the novel pretraining strategy, we introduce a variable-resolution input representation and a more flexible integration of language and vision inputs, where language prompts such as questions are rendered directly on top of the input image. For the first time, we show that a single pretrained model can achieve state-of-the-art results in six out of nine tasks across four domains: documents, illustrations, user interfaces, and natural images.

Tips:

Pix2Struct has been fine tuned on a variety of tasks and datasets, ranging from image captioning, visual question answering (VQA) over different inputs (books, charts, science diagrams), captioning UI components etc. The full list can be found in Table 1 of the paper. We therefore advise you to use these models for the tasks they have been fine tuned on. For instance, if you want to use Pix2Struct for UI captioning, you should use the model fine tuned on the UI dataset. If you want to use Pix2Struct for image captioning, you should use the model fine tuned on the natural images captioning dataset and so on.

If you want to use the model to perform conditional text captioning, make sure to use the processor with add_special_tokens=False.

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

Resources

Pix2StructConfig

class transformers.Pix2StructConfig

<source>

( text_config = Nonevision_config = Noneinitializer_factor = 1.0initializer_range = 0.02is_vqa = Falsetie_word_embeddings = Falseis_encoder_decoder = True**kwargs )

Parameters

  • text_config (dict, optional) — Dictionary of configuration options used to initialize Pix2StructTextConfig.

  • vision_config (dict, optional) — Dictionary of configuration options used to initialize Pix2StructVisionConfig.

  • initializer_factor (float, optional, defaults to 1.0) — Factor to multiply the initialization range with.

  • initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

  • is_vqa (bool, optional, defaults to False) — Whether the model has been fine-tuned for VQA or not.

  • kwargs (optional) — Dictionary of keyword arguments.

Pix2StructConfig is the configuration class to store the configuration of a Pix2StructForConditionalGeneration. It is used to instantiate a Pix2Struct 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 Pix2Struct-base google/pix2struct-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:

Copied

>>> from transformers import Pix2StructConfig, Pix2StructForConditionalGeneration

>>> # Initializing a Pix2StructConfig with google/pix2struct-base style configuration
>>> configuration = Pix2StructConfig()

>>> # Initializing a Pix2StructForConditionalGeneration (with random weights) from the google/pix2struct-base style configuration
>>> model = Pix2StructForConditionalGeneration(configuration)

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

>>> # We can also initialize a Pix2StructConfig from a Pix2StructTextConfig and a Pix2StructVisionConfig

>>> # Initializing a Pix2Struct text and Pix2Struct vision configuration
>>> config_text = Pix2StructTextConfig()
>>> config_vision = Pix2StructVisionConfig()

>>> config = Pix2StructConfig.from_text_vision_configs(config_text, config_vision)

from_text_vision_configs

<source>

( text_config: Pix2StructTextConfigvision_config: Pix2StructVisionConfig**kwargs ) → Pix2StructConfig

Returns

Pix2StructConfig

An instance of a configuration object

Instantiate a Pix2StructConfig (or a derived class) from pix2struct text model configuration and pix2struct vision model configuration.

Pix2StructTextConfig

class transformers.Pix2StructTextConfig

<source>

( vocab_size = 50244hidden_size = 768d_kv = 64d_ff = 2048num_layers = 12num_heads = 12relative_attention_num_buckets = 32relative_attention_max_distance = 128dropout_rate = 0.1layer_norm_epsilon = 1e-06initializer_factor = 1.0dense_act_fn = 'gelu_new'decoder_start_token_id = 0use_cache = Falsepad_token_id = 0eos_token_id = 1tie_word_embeddings = Falseis_decoder = True**kwargs )

Parameters

  • vocab_size (int, optional, defaults to 50244) — Vocabulary size of the Pix2Struct text model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling Pix2StructTextModel.

  • hidden_size (int, optional, defaults to 768) — Dimensionality of the encoder layers and the pooler layer.

  • d_kv (int, optional, defaults to 64) — Dimensionality of the key, query, value projections in each attention head.

  • d_ff (int, optional, defaults to 2048) — Dimensionality of the “intermediate” (i.e., feed-forward) layer in the Transformer encoder.

  • num_layers (int, optional, defaults to 12) — Number of hidden layers in the Transformer encoder.

  • num_heads (int, optional, defaults to 12) — Number of attention heads for each attention layer in the Transformer encoder.

  • relative_attention_num_buckets (int, optional, defaults to 32) — The number of buckets to use for each attention layer.

  • relative_attention_max_distance (int, optional, defaults to 128) — The maximum distance of the longer sequences for the bucket separation.

  • dropout_rate (float, optional, defaults to 0.1) — The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.

  • layer_norm_epsilon (float, optional, defaults to 1e-6) — The epsilon used by the layer normalization layers.

  • initializer_factor (float, optional, defaults to 1.0) — A factor for initializing all weight matrices (should be kept to 1, used internally for initialization testing).

  • dense_act_fn (Union[Callable, str], optional, defaults to "gelu_new") — The non-linear activation function (function or string).

  • decoder_start_token_id (int, optional, defaults to 0) — The id of the decoder_start_token_id token.

  • use_cache (bool, optional, defaults to False) — Whether or not the model should return the last key/values attentions (not used by all models).

  • pad_token_id (int, optional, defaults to 0) — The id of the padding token.

  • eos_token_id (int, optional, defaults to 1) — The id of the end-of-sequence token.

This is the configuration class to store the configuration of a Pix2StructTextModel. It is used to instantiate a Pix2Struct text 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 Pix2Struct text decoder used by the google/pix2struct-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:

Copied

>>> from transformers import Pix2StructTextConfig, Pix2StructTextModel

>>> # Initializing a Pix2StructTextConfig with google/pix2struct-base style configuration
>>> configuration = Pix2StructTextConfig()

>>> # Initializing a Pix2StructTextModel (with random weights) from the google/pix2struct-base style configuration
>>> model = Pix2StructTextModel(configuration)

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

Pix2StructVisionConfig

class transformers.Pix2StructVisionConfig

<source>

( hidden_size = 768patch_embed_hidden_size = 768d_ff = 2048d_kv = 64num_hidden_layers = 12num_attention_heads = 12dense_act_fn = 'gelu_new'layer_norm_eps = 1e-06dropout_rate = 0.0attention_dropout = 0.0initializer_range = 1e-10initializer_factor = 1.0seq_len = 4096relative_attention_num_buckets = 32relative_attention_max_distance = 128**kwargs )

Parameters

  • hidden_size (int, optional, defaults to 768) — Dimensionality of the encoder layers and the pooler layer.

  • patch_embed_hidden_size (int, optional, defaults to 768) — Dimensionality of the input patch_embedding layer in the Transformer encoder.

  • d_ff (int, optional, defaults to 2048) — Dimensionality of the “intermediate” (i.e., feed-forward) layer in the Transformer encoder.

  • d_kv (int, optional, defaults to 64) — Dimensionality of the key, query, value projections per attention head.

  • 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.

  • dense_act_fn (str or function, optional, defaults to "gelu_new") — The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu", "relu", "selu" and "gelu_new" `"gelu" are supported.

  • layer_norm_eps (float, optional, defaults to 1e-6) — The epsilon used by the layer normalization layers.

  • dropout_rate (float, optional, defaults to 0.0) — The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.

  • attention_dropout (float, optional, defaults to 0.0) — The dropout ratio for the attention probabilities.

  • initializer_range (float, optional, defaults to 1e-10) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

  • initializer_factor (`float“, optional, defaults to 1) — A factor for initializing all weight matrices (should be kept to 1, used internally for initialization testing).

  • seq_len (int, optional, defaults to 4096) — Maximum sequence length (here number of patches) supported by the model.

  • relative_attention_num_buckets (int, optional, defaults to 32) — The number of buckets to use for each attention layer.

  • relative_attention_max_distance (int, optional, defaults to 128) — The maximum distance (in tokens) to use for each attention layer.

This is the configuration class to store the configuration of a Pix2StructVisionModel. It is used to instantiate a Pix2Struct vision model according to the specified arguments, defining the model architecture. Instantiating a configuration defaults will yield a similar configuration to that of the Pix2Struct-base google/pix2struct-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:

Copied

>>> from transformers import Pix2StructVisionConfig, Pix2StructVisionModel

>>> # Initializing a Pix2StructVisionConfig with google/pix2struct-base style configuration
>>> configuration = Pix2StructVisionConfig()

>>> # Initializing a Pix2StructVisionModel (with random weights) from the google/pix2struct-base style configuration
>>> model = Pix2StructVisionModel(configuration)

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

Pix2StructProcessor

class transformers.Pix2StructProcessor

<source>

( image_processortokenizer )

Parameters

  • image_processor (Pix2StructImageProcessor) — An instance of Pix2StructImageProcessor. The image processor is a required input.

  • tokenizer (Union[T5TokenizerFast, T5Tokenizer]) — An instance of [‘T5TokenizerFast`] or [‘T5Tokenizer`]. The tokenizer is a required input.

Constructs a PIX2STRUCT processor which wraps a BERT tokenizer and PIX2STRUCT image processor into a single processor.

Pix2StructProcessor offers all the functionalities of Pix2StructImageProcessor and T5TokenizerFast. See the docstring of __call__() and decode() for more information.

batch_decode

<source>

( *args**kwargs )

This method forwards all its arguments to Pix2StructTokenizerFast’s batch_decode(). Please refer to the docstring of this method for more information.

decode

<source>

( *args**kwargs )

This method forwards all its arguments to Pix2StructTokenizerFast’s decode(). Please refer to the docstring of this method for more information.

Pix2StructImageProcessor

class transformers.Pix2StructImageProcessor

<source>

( do_convert_rgb: bool = Truedo_normalize: bool = Truepatch_size: typing.Dict[str, int] = Nonemax_patches: int = 2048is_vqa: bool = False**kwargs )

Parameters

  • do_convert_rgb (bool, optional, defaults to True) — Whether to convert the image to RGB.

  • do_normalize (bool, optional, defaults to True) — Whether to normalize the image. Can be overridden by the do_normalize parameter in the preprocess method. According to Pix2Struct paper and code, the image is normalized with its own mean and standard deviation.

  • patch_size (Dict[str, int], optional, defaults to {"height" -- 16, "width": 16}): The patch size to use for the image. According to Pix2Struct paper and code, the patch size is 16x16.

  • max_patches (int, optional, defaults to 2048) — The maximum number of patches to extract from the image as per the Pix2Struct paper.

  • is_vqa (bool, optional, defaults to False) — Whether or not the image processor is for the VQA task. If True and header_text is passed in, text is rendered onto the input images.

Constructs a Pix2Struct image processor.

preprocess

<source>

( images: typing.Union[ForwardRef('PIL.Image.Image'), numpy.ndarray, ForwardRef('torch.Tensor'), typing.List[ForwardRef('PIL.Image.Image')], typing.List[numpy.ndarray], typing.List[ForwardRef('torch.Tensor')]]header_text: typing.Optional[str] = Nonedo_convert_rgb: bool = Nonedo_normalize: typing.Optional[bool] = Nonemax_patches: typing.Optional[int] = Nonepatch_size: typing.Union[typing.Dict[str, int], NoneType] = Nonereturn_tensors: typing.Union[str, transformers.utils.generic.TensorType, NoneType] = Nonedata_format: ChannelDimension = <ChannelDimension.FIRST: 'channels_first'>input_data_format: typing.Union[str, transformers.image_utils.ChannelDimension, NoneType] = None**kwargs )

Parameters

  • images (ImageInput) — Image to preprocess. Expects a single or batch of images.

  • header_text (Union[List[str], str], optional) — Text to render as a header. Only has an effect if image_processor.is_vqa is True.

  • do_convert_rgb (bool, optional, defaults to self.do_convert_rgb) — Whether to convert the image to RGB.

  • do_normalize (bool, optional, defaults to self.do_normalize) — Whether to normalize the image.

  • max_patches (int, optional, defaults to self.max_patches) — Maximum number of patches to extract.

  • patch_size (dict, optional, defaults to self.patch_size) — Dictionary containing the patch height and width.

  • return_tensors (str or TensorType, optional) — The type of tensors to return. Can be one of:

    • Unset: Return a list of np.ndarray.

    • TensorType.TENSORFLOW or 'tf': Return a batch of type tf.Tensor.

    • TensorType.PYTORCH or 'pt': Return a batch of type torch.Tensor.

    • TensorType.NUMPY or 'np': Return a batch of type np.ndarray.

    • TensorType.JAX or 'jax': Return a batch of type jax.numpy.ndarray.

  • data_format (ChannelDimension or str, optional, defaults to ChannelDimension.FIRST) — The channel dimension format for the output image. Can be one of:

    • "channels_first" or ChannelDimension.FIRST: image in (num_channels, height, width) format.

    • "channels_last" or ChannelDimension.LAST: image in (height, width, num_channels) format.

    • Unset: Use the channel dimension format of the input image.

  • input_data_format (ChannelDimension or str, optional) — The channel dimension format for the input image. If unset, the channel dimension format is inferred from the input image. Can be one of:

    • "channels_first" or ChannelDimension.FIRST: image in (num_channels, height, width) format.

    • "channels_last" or ChannelDimension.LAST: image in (height, width, num_channels) format.

    • "none" or ChannelDimension.NONE: image in (height, width) format.

Preprocess an image or batch of images. The processor first computes the maximum possible number of aspect-ratio preserving patches of size patch_size that can be extracted from the image. It then pads the image with zeros to make the image respect the constraint of max_patches. Before extracting the patches the images are standardized following the tensorflow implementation of per_image_standardization (https://www.tensorflow.org/api_docs/python/tf/image/per_image_standardization).

Pix2StructTextModel

class transformers.Pix2StructTextModel

<source>

( config )

Parameters

  • config (Union[Pix2StructConfig, Pix2StructTextConfig]) — 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 standalone text decoder of Pix2Struct

The Pix2Struct model was proposed in Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding by Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova. It’s an encoder decoder transformer pre-trained in a image-to-text setting.

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

<source>

( input_ids: typing.Optional[torch.LongTensor] = Noneattention_mask: typing.Optional[torch.FloatTensor] = Noneencoder_hidden_states: typing.Optional[torch.FloatTensor] = Noneencoder_attention_mask: typing.Optional[torch.FloatTensor] = Noneinputs_embeds: typing.Optional[torch.LongTensor] = Nonehead_mask: typing.Optional[torch.FloatTensor] = Nonecross_attn_head_mask: typing.Optional[torch.Tensor] = Nonepast_key_values: typing.Optional[typing.Tuple[typing.Tuple[torch.FloatTensor]]] = Noneuse_cache: typing.Optional[bool] = Noneoutput_attentions: typing.Optional[bool] = Noneoutput_hidden_states: typing.Optional[bool] = Nonelabels: typing.Optional[torch.LongTensor] = Nonereturn_dict: typing.Optional[bool] = None**kwargs ) → transformers.modeling_outputs.CausalLMOutputWithCrossAttentions or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. Pix2StructText is a model with relative position embeddings so you should be able to pad the inputs on both the right and the left.

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

    What are input IDs?

    To know more on how to prepare input_ids for pretraining take a look a Pix2StructText Training.

  • 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?

  • decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional) — Indices of decoder input sequence tokens in the vocabulary.

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

    What are decoder input IDs?

    Pix2StructText uses the pad_token_id as the starting token for decoder_input_ids generation. If past_key_values is used, optionally only the last decoder_input_ids have to be input (see past_key_values).

    To know more on how to prepare decoder_input_ids for pretraining take a look at Pix2StructText Training.

  • decoder_attention_mask (torch.BoolTensor of shape (batch_size, target_sequence_length), optional) — Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default.

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

    • 1 indicates the head is not masked,

    • 0 indicates the head is masked.

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

    • 1 indicates the head is not masked,

    • 0 indicates the head is masked.

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

    • 1 indicates the head is not masked,

    • 0 indicates the head is masked.

  • encoder_outputs (tuple(tuple(torch.FloatTensor), optional) — Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (batch_size, sequence_length, hidden_size) is a sequence of hidden states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.

  • past_key_values (tuple(tuple(torch.FloatTensor)) of length config.n_layers with each tuple having 4 tensors of shape (batch_size, num_heads, sequence_length - 1, embed_size_per_head)) — Contains precomputed key and value hidden states of the attention layers. Can be used to speed up decoding.

    If past_key_values are used, the user can optionally input only the last decoder_input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, 1) instead of all decoder_input_ids of shape (batch_size, sequence_length).

  • 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.

  • decoder_inputs_embeds (torch.FloatTensor of shape (batch_size, target_sequence_length, hidden_size), optional) — Optionally, instead of passing decoder_input_ids you can choose to directly pass an embedded representation. If past_key_values is used, optionally only the last decoder_inputs_embeds have to be input (see past_key_values). This is useful if you want more control over how to convert decoder_input_ids indices into associated vectors than the model’s internal embedding lookup matrix.

    If decoder_input_ids and decoder_inputs_embeds are both unset, decoder_inputs_embeds takes the value of inputs_embeds.

  • use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).

  • 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.CausalLMOutputWithCrossAttentions or tuple(torch.FloatTensor)

A transformers.modeling_outputs.CausalLMOutputWithCrossAttentions 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 (Pix2StructConfig) and inputs.

  • loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Language modeling loss (for next-token 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).

  • 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 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).

    Cross attentions weights 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 torch.FloatTensor tuples of length config.n_layers, with each tuple containing the cached key, value states of the self-attention and the cross-attention layers if model is used in encoder-decoder setting. Only relevant if config.is_decoder = True.

    Contains pre-computed hidden-states (key and values in the attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

The Pix2StructTextModel 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:

Copied

>>> from transformers import AutoProcessor, Pix2StructTextModel

>>> processor = AutoProcessor.from_pretrained("google/pix2struct-textcaps-base")
>>> model = Pix2StructTextModel.from_pretrained("google/pix2struct-textcaps-base")

>>> inputs = processor(text="Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs)
>>> loss = outputs.loss

Pix2StructVisionModel

class transformers.Pix2StructVisionModel

<source>

( config: Pix2StructConfig )

Parameters

  • config (Pix2StructConfig) — 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 Pix2StructVision Model transformer outputting raw hidden-states without any specific head on top. This model is 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

<source>

( flattened_patches: typing.Optional[torch.Tensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Nonehead_mask: typing.Optional[torch.Tensor] = Noneoutput_attentions: typing.Optional[bool] = Noneoutput_hidden_states: typing.Optional[bool] = Nonereturn_dict: typing.Optional[bool] = None ) → transformers.modeling_outputs.BaseModelOutputWithPooling or tuple(torch.FloatTensor)

Parameters

  • flattened_patches (torch.FloatTensor of shape (batch_size, sequence_length, num_channels x patch_height x patch_width)) — Flattened and padded pixel values. These values can be obtained using AutoImageProcessor. See Pix2StructVisionImageProcessor.__call__ for details. Check the original paper (figure 5) for more details.

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

  • 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.

  • 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 (Pix2StructConfig) 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.

The Pix2StructVisionModel 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:

Copied

>>> import requests
>>> from PIL import Image
>>> from transformers import AutoProcessor, Pix2StructVisionModel

>>> image_processor = AutoProcessor.from_pretrained("google/pix2struct-textcaps-base")
>>> model = Pix2StructVisionModel.from_pretrained("google/pix2struct-textcaps-base")

>>> url = "https://www.ilankelman.org/stopsigns/australia.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)

>>> inputs = image_processor(images=image, return_tensors="pt")
>>> with torch.no_grad():
...     outputs = model(**inputs)

>>> last_hidden_states = outputs.last_hidden_state
>>> list(last_hidden_states.shape)
[1, 2048, 768]

Pix2StructForConditionalGeneration

class transformers.Pix2StructForConditionalGeneration

<source>

( config: Pix2StructConfig )

Parameters

  • config (Union[Pix2StructConfig, Pix2StructTextConfig]) — 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.

A conditional generation model with a language modeling head. Can be used for sequence generation tasks.

The Pix2Struct model was proposed in Pix2Struct: Screenshot Parsing as Pretraining for Visual Language Understanding by Kenton Lee, Mandar Joshi, Iulia Turc, Hexiang Hu, Fangyu Liu, Julian Eisenschlos, Urvashi Khandelwal, Peter Shaw, Ming-Wei Chang, Kristina Toutanova. It’s an encoder decoder transformer pre-trained in a image-to-text setting.

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

<source>

( flattened_patches: typing.Optional[torch.FloatTensor] = Noneattention_mask: typing.Optional[torch.FloatTensor] = Nonedecoder_input_ids: typing.Optional[torch.LongTensor] = Nonedecoder_attention_mask: typing.Optional[torch.BoolTensor] = Nonehead_mask: typing.Optional[torch.FloatTensor] = Nonedecoder_head_mask: typing.Optional[torch.FloatTensor] = Nonecross_attn_head_mask: typing.Optional[torch.Tensor] = Noneencoder_outputs: typing.Optional[typing.Tuple[typing.Tuple[torch.FloatTensor]]] = Nonepast_key_values: typing.Optional[typing.Tuple[typing.Tuple[torch.FloatTensor]]] = Nonelabels: typing.Optional[torch.LongTensor] = Nonedecoder_inputs_embeds: typing.Optional[torch.Tensor] = Noneuse_cache: typing.Optional[bool] = Noneoutput_attentions: typing.Optional[bool] = Noneoutput_hidden_states: typing.Optional[bool] = Nonereturn_dict: typing.Optional[bool] = None ) → transformers.modeling_outputs.Seq2SeqModelOutput or tuple(torch.FloatTensor)

Parameters

  • flattened_patches (torch.FloatTensor of shape (batch_size, seq_length, hidden_size)) — Flattened pixel patches. the hidden_size is obtained by the following formula: hidden_size = num_channels patch_size patch_size

    The process of flattening the pixel patches is done by Pix2StructProcessor.

  • 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?

  • decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional) — Indices of decoder input sequence tokens in the vocabulary.

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

    What are decoder input IDs?

    Pix2StructText uses the pad_token_id as the starting token for decoder_input_ids generation. If past_key_values is used, optionally only the last decoder_input_ids have to be input (see past_key_values).

    To know more on how to prepare decoder_input_ids for pretraining take a look at Pix2StructText Training.

  • decoder_attention_mask (torch.BoolTensor of shape (batch_size, target_sequence_length), optional) — Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default.

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

    • 1 indicates the head is not masked,

    • 0 indicates the head is masked.

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

    • 1 indicates the head is not masked,

    • 0 indicates the head is masked.

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

    • 1 indicates the head is not masked,

    • 0 indicates the head is masked.

  • encoder_outputs (tuple(tuple(torch.FloatTensor), optional) — Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (batch_size, sequence_length, hidden_size) is a sequence of hidden states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.

  • past_key_values (tuple(tuple(torch.FloatTensor)) of length config.n_layers with each tuple having 4 tensors of shape (batch_size, num_heads, sequence_length - 1, embed_size_per_head)) — Contains precomputed key and value hidden states of the attention layers. Can be used to speed up decoding.

    If past_key_values are used, the user can optionally input only the last decoder_input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, 1) instead of all decoder_input_ids of shape (batch_size, sequence_length).

  • decoder_inputs_embeds (torch.FloatTensor of shape (batch_size, target_sequence_length, hidden_size), optional) — Optionally, instead of passing decoder_input_ids you can choose to directly pass an embedded representation. If past_key_values is used, optionally only the last decoder_inputs_embeds have to be input (see past_key_values). This is useful if you want more control over how to convert decoder_input_ids indices into associated vectors than the model’s internal embedding lookup matrix.

    If decoder_input_ids and decoder_inputs_embeds are both unset, decoder_inputs_embeds takes the value of inputs_embeds.

  • labels (torch.LongTensor of shape (batch_size, sequence_length), optional) — Labels for computing the masked language modeling loss for the decoder.

  • use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).

  • 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.Seq2SeqModelOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.Seq2SeqModelOutput 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 (Pix2StructConfig) 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 decoder of the model.

    If past_key_values is used only the last hidden-state of the sequences of shape (batch_size, 1, hidden_size) is output.

  • 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 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 in the cross-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

  • decoder_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 decoder at the output of each layer plus the optional initial embedding outputs.

  • decoder_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 of the decoder, 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 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.

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

  • encoder_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 encoder at the output of each layer plus the optional initial embedding outputs.

  • encoder_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 of the encoder, after the attention softmax, used to compute the weighted average in the self-attention heads.

The Pix2StructForConditionalGeneration 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:

Inference:

Copied

>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, Pix2StructForConditionalGeneration

>>> processor = AutoProcessor.from_pretrained("google/pix2struct-textcaps-base")
>>> model = Pix2StructForConditionalGeneration.from_pretrained("google/pix2struct-textcaps-base")

>>> url = "https://www.ilankelman.org/stopsigns/australia.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)

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

>>> # autoregressive generation
>>> generated_ids = model.generate(**inputs, max_new_tokens=50)
>>> generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
>>> print(generated_text)
A stop sign is on a street corner.

>>> # conditional generation
>>> text = "A picture of"
>>> inputs = processor(text=text, images=image, return_tensors="pt", add_special_tokens=False)

>>> generated_ids = model.generate(**inputs, max_new_tokens=50)
>>> generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
>>> print(generated_text)
A picture of a stop sign with a red stop sign

Training:

Copied

>>> from PIL import Image
>>> import requests
>>> from transformers import AutoProcessor, Pix2StructForConditionalGeneration

>>> processor = AutoProcessor.from_pretrained("google/pix2struct-base")
>>> model = Pix2StructForConditionalGeneration.from_pretrained("google/pix2struct-base")

>>> url = "https://www.ilankelman.org/stopsigns/australia.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> text = "A stop sign is on the street corner."

>>> inputs = processor(images=image, return_tensors="pt")
>>> labels = processor(text=text, return_tensors="pt").input_ids

>>> # forward pass
>>> outputs = model(**inputs, labels=labels)
>>> loss = outputs.loss
>>> print(f"{loss.item():.5f}")
5.94282

Last updated