# Vision Encoder Decoder Models

## Vision Encoder Decoder Models

### Overview

The [VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel) can be used to initialize an image-to-text model with any pretrained Transformer-based vision model as the encoder (*e.g.* [ViT](https://huggingface.co/docs/transformers/model_doc/vit), [BEiT](https://huggingface.co/docs/transformers/model_doc/beit), [DeiT](https://huggingface.co/docs/transformers/model_doc/deit), [Swin](https://huggingface.co/docs/transformers/model_doc/swin)) and any pretrained language model as the decoder (*e.g.* [RoBERTa](https://huggingface.co/docs/transformers/model_doc/roberta), [GPT2](https://huggingface.co/docs/transformers/model_doc/gpt2), [BERT](https://huggingface.co/docs/transformers/model_doc/bert), [DistilBERT](https://huggingface.co/docs/transformers/model_doc/distilbert)).

The effectiveness of initializing image-to-text-sequence models with pretrained checkpoints has been shown in (for example) [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) by Minghao Li, Tengchao Lv, Lei Cui, Yijuan Lu, Dinei Florencio, Cha Zhang, Zhoujun Li, Furu Wei.

After such a [VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel) has been trained/fine-tuned, it can be saved/loaded just like any other models (see the examples below for more information).

An example application is image captioning, in which the encoder is used to encode the image, after which an autoregressive language model generates the caption. Another example is optical character recognition. Refer to [TrOCR](https://huggingface.co/docs/transformers/model_doc/trocr), which is an instance of [VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel).

### Randomly initializing VisionEncoderDecoderModel from model configurations.

[VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel) can be randomly initialized from an encoder and a decoder config. In the following example, we show how to do this using the default [ViTModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vit#transformers.ViTModel) configuration for the encoder and the default `BertForCausalLM` configuration for the decoder.

Copied

```
>>> from transformers import BertConfig, ViTConfig, VisionEncoderDecoderConfig, VisionEncoderDecoderModel

>>> config_encoder = ViTConfig()
>>> config_decoder = BertConfig()

>>> config = VisionEncoderDecoderConfig.from_encoder_decoder_configs(config_encoder, config_decoder)
>>> model = VisionEncoderDecoderModel(config=config)
```

### Initialising VisionEncoderDecoderModel from a pretrained encoder and a pretrained decoder.

[VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel) can be initialized from a pretrained encoder checkpoint and a pretrained decoder checkpoint. Note that any pretrained Transformer-based vision model, *e.g.* [Swin](https://huggingface.co/docs/transformers/model_doc/swin), can serve as the encoder and both pretrained auto-encoding models, *e.g.* BERT, pretrained causal language models, *e.g.* GPT2, as well as the pretrained decoder part of sequence-to-sequence models, *e.g.* decoder of BART, can be used as the decoder. Depending on which architecture you choose as the decoder, the cross-attention layers might be randomly initialized. Initializing [VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel) from a pretrained encoder and decoder checkpoint requires the model to be fine-tuned on a downstream task, as has been shown in [the *Warm-starting-encoder-decoder blog post*](https://huggingface.co/blog/warm-starting-encoder-decoder). To do so, the `VisionEncoderDecoderModel` class provides a [VisionEncoderDecoderModel.from\_encoder\_decoder\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel.from_encoder_decoder_pretrained) method.

Copied

```
>>> from transformers import VisionEncoderDecoderModel

>>> model = VisionEncoderDecoderModel.from_encoder_decoder_pretrained(
...     "microsoft/swin-base-patch4-window7-224-in22k", "bert-base-uncased"
... )
```

### Loading an existing VisionEncoderDecoderModel checkpoint and perform inference.

To load fine-tuned checkpoints of the `VisionEncoderDecoderModel` class, [VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel) provides the `from_pretrained(...)` method just like any other model architecture in Transformers.

To perform inference, one uses the `generate` method, which allows to autoregressively generate text. This method supports various forms of decoding, such as greedy, beam search and multinomial sampling.

Copied

```
>>> import requests
>>> from PIL import Image

>>> from transformers import GPT2TokenizerFast, ViTImageProcessor, VisionEncoderDecoderModel

>>> # load a fine-tuned image captioning model and corresponding tokenizer and image processor
>>> model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
>>> tokenizer = GPT2TokenizerFast.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
>>> image_processor = ViTImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")

>>> # let's perform inference on an image
>>> url = "http://images.cocodataset.org/val2017/000000039769.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw)
>>> pixel_values = image_processor(image, return_tensors="pt").pixel_values

>>> # autoregressively generate caption (uses greedy decoding by default)
>>> generated_ids = model.generate(pixel_values)
>>> generated_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
>>> print(generated_text)
a cat laying on a blanket next to a cat laying on a bed
```

### Loading a PyTorch checkpoint into TFVisionEncoderDecoderModel .

`TFVisionEncoderDecoderModel.from_pretrained()` currently doesn’t support initializing the model from a PyTorch checkpoint. Passing `from_pt=True` to this method will throw an exception. If there are only PyTorch checkpoints for a particular vision encoder-decoder model, a workaround is:

Copied

```
>>> from transformers import VisionEncoderDecoderModel, TFVisionEncoderDecoderModel

>>> _model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")

>>> _model.encoder.save_pretrained("./encoder")
>>> _model.decoder.save_pretrained("./decoder")

>>> model = TFVisionEncoderDecoderModel.from_encoder_decoder_pretrained(
...     "./encoder", "./decoder", encoder_from_pt=True, decoder_from_pt=True
... )
>>> # This is only for copying some specific attributes of this particular model.
>>> model.config = _model.config
```

### Training

Once the model is created, it can be fine-tuned similar to BART, T5 or any other encoder-decoder model on a dataset of (image, text) pairs. As you can see, only 2 inputs are required for the model in order to compute a loss: `pixel_values` (which are the images) and `labels` (which are the `input_ids` of the encoded target sequence).

Copied

```
>>> from transformers import ViTImageProcessor, BertTokenizer, VisionEncoderDecoderModel
>>> from datasets import load_dataset

>>> image_processor = ViTImageProcessor.from_pretrained("google/vit-base-patch16-224-in21k")
>>> tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
>>> model = VisionEncoderDecoderModel.from_encoder_decoder_pretrained(
...     "google/vit-base-patch16-224-in21k", "bert-base-uncased"
... )

>>> model.config.decoder_start_token_id = tokenizer.cls_token_id
>>> model.config.pad_token_id = tokenizer.pad_token_id

>>> dataset = load_dataset("boincai/cats-image")
>>> image = dataset["test"]["image"][0]
>>> pixel_values = image_processor(image, return_tensors="pt").pixel_values

>>> labels = tokenizer(
...     "an image of two cats chilling on a couch",
...     return_tensors="pt",
... ).input_ids

>>> # the forward function automatically creates the correct decoder_input_ids
>>> loss = model(pixel_values=pixel_values, labels=labels).loss
```

This model was contributed by [nielsr](https://github.com/nielsrogge). This model’s TensorFlow and Flax versions were contributed by [ydshieh](https://github.com/ydshieh).

### VisionEncoderDecoderConfig

#### class transformers.VisionEncoderDecoderConfig

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/configuration_vision_encoder_decoder.py#L33)

( \*\*kwargs )

Parameters

* **kwargs** (*optional*) — Dictionary of keyword arguments. Notably:
  * **encoder** ([PretrainedConfig](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/configuration#transformers.PretrainedConfig), *optional*) — An instance of a configuration object that defines the encoder config.
  * **decoder** ([PretrainedConfig](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/configuration#transformers.PretrainedConfig), *optional*) — An instance of a configuration object that defines the decoder config.

[VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig) is the configuration class to store the configuration of a [VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel). It is used to instantiate a Vision-Encoder-Text-Decoder model according to the specified arguments, defining the encoder and decoder configs.

Configuration objects inherit from [PretrainedConfig](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/configuration#transformers.PretrainedConfig) and can be used to control the model outputs. Read the documentation from [PretrainedConfig](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/configuration#transformers.PretrainedConfig) for more information.

Examples:

Copied

```
>>> from transformers import BertConfig, ViTConfig, VisionEncoderDecoderConfig, VisionEncoderDecoderModel

>>> # Initializing a ViT & BERT style configuration
>>> config_encoder = ViTConfig()
>>> config_decoder = BertConfig()

>>> config = VisionEncoderDecoderConfig.from_encoder_decoder_configs(config_encoder, config_decoder)

>>> # Initializing a ViTBert model (with random weights) from a ViT & bert-base-uncased style configurations
>>> model = VisionEncoderDecoderModel(config=config)

>>> # Accessing the model configuration
>>> config_encoder = model.config.encoder
>>> config_decoder = model.config.decoder
>>> # set decoder config to causal lm
>>> config_decoder.is_decoder = True
>>> config_decoder.add_cross_attention = True

>>> # Saving the model, including its configuration
>>> model.save_pretrained("my-model")

>>> # loading model and config from pretrained folder
>>> encoder_decoder_config = VisionEncoderDecoderConfig.from_pretrained("my-model")
>>> model = VisionEncoderDecoderModel.from_pretrained("my-model", config=encoder_decoder_config)
```

**from\_encoder\_decoder\_configs**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/configuration_vision_encoder_decoder.py#L99)

( encoder\_config: PretrainedConfigdecoder\_config: PretrainedConfig\*\*kwargs ) → [VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig)

Returns

[VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig)

An instance of a configuration object

Instantiate a [VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig) (or a derived class) from a pre-trained encoder model configuration and decoder model configuration.

### VisionEncoderDecoderModel

#### class transformers.VisionEncoderDecoderModel

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_vision_encoder_decoder.py#L151)

( config: typing.Optional\[transformers.configuration\_utils.PretrainedConfig] = Noneencoder: typing.Optional\[transformers.modeling\_utils.PreTrainedModel] = Nonedecoder: typing.Optional\[transformers.modeling\_utils.PreTrainedModel] = None )

Parameters

* **config** ([VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.PreTrainedModel.from_pretrained) method to load the model weights.

This class can be used to initialize an image-to-text-sequence model with any pretrained vision autoencoding model as the encoder and any pretrained text autoregressive model as the decoder. The encoder is loaded via [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.FlaxAutoModelForVision2Seq.from_pretrained) function and the decoder is loaded via [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.FlaxAutoModelForVision2Seq.from_pretrained) function. Cross-attention layers are automatically added to the decoder and should be fine-tuned on a downstream generative task, like image captioning.

The effectiveness of initializing sequence-to-sequence models with pretrained checkpoints for sequence generation tasks was shown in [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn. Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu.

Additionally, in [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) it is shown how leveraging large pretrained vision models for optical character recognition (OCR) yields a significant performance improvement.

After such a Vision-Encoder-Text-Decoder model has been trained/fine-tuned, it can be saved/loaded just like any other models (see the examples for more information).

This model inherits from [PreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.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](https://pytorch.org/docs/stable/nn.html#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.

[VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel) is a generic model class that will be instantiated as a transformer architecture with one of the base vision model classes of the library as encoder and another one as decoder when created with the :met&#x68;*\~transformers.AutoModel.from\_pretrained* class method for the encoder and :met&#x68;*\~transformers.AutoModelForCausalLM.from\_pretrained* class method for the decoder.

**forward**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_vision_encoder_decoder.py#L519)

( pixel\_values: typing.Optional\[torch.FloatTensor] = Nonedecoder\_input\_ids: typing.Optional\[torch.LongTensor] = Nonedecoder\_attention\_mask: typing.Optional\[torch.BoolTensor] = Noneencoder\_outputs: typing.Optional\[typing.Tuple\[torch.FloatTensor]] = Nonepast\_key\_values: typing.Optional\[typing.Tuple\[typing.Tuple\[torch.FloatTensor]]] = Nonedecoder\_inputs\_embeds: typing.Optional\[torch.FloatTensor] = Nonelabels: typing.Optional\[torch.LongTensor] = Noneuse\_cache: typing.Optional\[bool] = Noneoutput\_attentions: typing.Optional\[bool] = Noneoutput\_hidden\_states: typing.Optional\[bool] = Nonereturn\_dict: typing.Optional\[bool] = None\*\*kwargs ) → [transformers.modeling\_outputs.Seq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.Seq2SeqLMOutput) or `tuple(torch.FloatTensor)`

Parameters

* **pixel\_values** (`torch.FloatTensor` of shape `(batch_size, num_channels, height, width)`) — Pixel values. Pixel values can be obtained using an image processor (e.g. if you use ViT as the encoder, you should use [AutoImageProcessor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoImageProcessor)). See [ViTImageProcessor.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/deit#transformers.DeiTFeatureExtractor.__call__) for details.
* **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 [PreTrainedTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details.

  [What are input IDs?](https://huggingface.co/docs/transformers/glossary#input-ids)

  If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see `past_key_values`).

  For training, `decoder_input_ids` are automatically created by the model by shifting the `labels` to the right, replacing -100 by the `pad_token_id` and prepending them with the `decoder_start_token_id`.
* **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.
* **encoder\_outputs** (`tuple(torch.FloatTensor)`, *optional*) — This tuple must consist of (`last_hidden_state`, *optional*: `hidden_states`, *optional*: `attentions`) `last_hidden_state` (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`) is a tensor 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 blocks. 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. 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.
* **labels** (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*) — Labels for computing the masked language modeling loss for the decoder. 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]`
* **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*) — If set to `True`, the model will return a `~utils.Seq2SeqLMOutput` instead of a plain tuple.
* **kwargs** (*optional*) — Remaining dictionary of keyword arguments. Keyword arguments come in two flavors:
  * Without a prefix which will be input as `**encoder_kwargs` for the encoder forward function.
  * With a *decoder\_* prefix which will be input as `**decoder_kwargs` for the decoder forward function.

Returns

[transformers.modeling\_outputs.Seq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.Seq2SeqLMOutput) or `tuple(torch.FloatTensor)`

A [transformers.modeling\_outputs.Seq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_outputs.Seq2SeqLMOutput) 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 ([VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig)) and inputs.

* **loss** (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided) — Language modeling loss.
* **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).
* **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 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 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 [VisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderModel) 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:

Copied

```
>>> from transformers import AutoProcessor, VisionEncoderDecoderModel
>>> import requests
>>> from PIL import Image
>>> import torch

>>> processor = AutoProcessor.from_pretrained("microsoft/trocr-base-handwritten")
>>> model = VisionEncoderDecoderModel.from_pretrained("microsoft/trocr-base-handwritten")

>>> # load image from the IAM dataset
>>> url = "https://fki.tic.heia-fr.ch/static/img/a01-122-02.jpg"
>>> image = Image.open(requests.get(url, stream=True).raw).convert("RGB")

>>> # training
>>> model.config.decoder_start_token_id = processor.tokenizer.cls_token_id
>>> model.config.pad_token_id = processor.tokenizer.pad_token_id
>>> model.config.vocab_size = model.config.decoder.vocab_size

>>> pixel_values = processor(image, return_tensors="pt").pixel_values
>>> text = "hello world"
>>> labels = processor.tokenizer(text, return_tensors="pt").input_ids
>>> outputs = model(pixel_values=pixel_values, labels=labels)
>>> loss = outputs.loss

>>> # inference (generation)
>>> generated_ids = model.generate(pixel_values)
>>> generated_text = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
```

**from\_encoder\_decoder\_pretrained**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_vision_encoder_decoder.py#L365)

( encoder\_pretrained\_model\_name\_or\_path: str = Nonedecoder\_pretrained\_model\_name\_or\_path: str = None\*model\_args\*\*kwargs )

Parameters

* **encoder\_pretrained\_model\_name\_or\_path** (`str`, *optional*) — Information necessary to initiate the image encoder. Can be either:
  * A string, the *model id* of a pretrained model hosted inside a model repo on boincai.com. An example is `google/vit-base-patch16-224-in21k`.
  * A path to a *directory* containing model weights saved using [save\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.PreTrainedModel.save_pretrained), e.g., `./my_model_directory/`.
  * A path or url to a *tensorflow index checkpoint file* (e.g, `./tf_model/model.ckpt.index`). In this case, `from_tf` should be set to `True` and a configuration object should be provided as `config` argument. This loading path is slower than converting the TensorFlow checkpoint in a PyTorch model using the provided conversion scripts and loading the PyTorch model afterwards.
* **decoder\_pretrained\_model\_name\_or\_path** (`str`, *optional*, defaults to `None`) — Information necessary to initiate the text decoder. Can be either:
  * A string, the *model id* of a pretrained model hosted inside a model repo on boincai.com. Valid model ids can be located at the root-level, like `bert-base-uncased`, or namespaced under a user or organization name, like `dbmdz/bert-base-german-cased`.
  * A path to a *directory* containing model weights saved using [save\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.PreTrainedModel.save_pretrained), e.g., `./my_model_directory/`.
  * A path or url to a *tensorflow index checkpoint file* (e.g, `./tf_model/model.ckpt.index`). In this case, `from_tf` should be set to `True` and a configuration object should be provided as `config` argument. This loading path is slower than converting the TensorFlow checkpoint in a PyTorch model using the provided conversion scripts and loading the PyTorch model afterwards.
* **model\_args** (remaining positional arguments, *optional*) — All remaning positional arguments will be passed to the underlying model’s `__init__` method.
* **kwargs** (remaining dictionary of keyword arguments, *optional*) — Can be used to update the configuration object (after it being loaded) and initiate the model (e.g., `output_attentions=True`).

  * To update the encoder configuration, use the prefix *encoder\_* for each configuration parameter.
  * To update the decoder configuration, use the prefix *decoder\_* for each configuration parameter.
  * To update the parent model configuration, do not use a prefix for each configuration parameter.

  Behaves differently depending on whether a `config` is provided or automatically loaded.

Instantiate an encoder and a decoder from one or two base classes of the library from pretrained model checkpoints.

The model is set in evaluation mode by default using `model.eval()` (Dropout modules are deactivated). To train the model, you need to first set it back in training mode with `model.train()`.

Example:

Copied

```
>>> from transformers import VisionEncoderDecoderModel

>>> # initialize a vit-bert from a pretrained ViT and a pretrained BERT model. Note that the cross-attention layers will be randomly initialized
>>> model = VisionEncoderDecoderModel.from_encoder_decoder_pretrained(
...     "google/vit-base-patch16-224-in21k", "bert-base-uncased"
... )
>>> # saving model after fine-tuning
>>> model.save_pretrained("./vit-bert")
>>> # load fine-tuned model
>>> model = VisionEncoderDecoderModel.from_pretrained("./vit-bert")
```

### TFVisionEncoderDecoderModel

#### class transformers.TFVisionEncoderDecoderModel

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_tf_vision_encoder_decoder.py#L176)

( \*args\*\*kwargs )

Parameters

* **config** ([VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.TFPreTrainedModel.from_pretrained) method to load the model weights.

This class can be used to initialize an image-to-text-sequence model with any pretrained vision autoencoding model as the encoder and any pretrained text autoregressive model as the decoder. The encoder is loaded via [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.FlaxAutoModelForVision2Seq.from_pretrained) function and the decoder is loaded via [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.FlaxAutoModelForVision2Seq.from_pretrained) function. Cross-attention layers are automatically added to the decoder and should be fine-tuned on a downstream generative task, like image captioning.

The effectiveness of initializing sequence-to-sequence models with pretrained checkpoints for sequence generation tasks was shown in [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn. Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu.

Additionally, in [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) it is shown how leveraging large pretrained vision models for optical character recognition (OCR) yields a significant performance improvement.

After such a Vision-Encoder-Text-Decoder model has been trained/fine-tuned, it can be saved/loaded just like any other models (see the examples for more information).

This model inherits from [TFPreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.TFPreTrainedModel). 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 [tf.keras.Model](https://www.tensorflow.org/api_docs/python/tf/keras/Model) subclass. Use it as a regular TF 2.0 Keras Model and refer to the TF 2.0 documentation for all matter related to general usage and behavior.

[TFVisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.TFVisionEncoderDecoderModel) is a generic model class that will be instantiated as a transformer architecture with one of the base vision model classes of the library as encoder and another one of the base model classes as decoder when created with the [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.FlaxAutoModelForVision2Seq.from_pretrained) class method for the encoder and [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.FlaxAutoModelForVision2Seq.from_pretrained) class method for the decoder.

**call**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_tf_vision_encoder_decoder.py#L486)

( pixel\_values: np.ndarray | tf.Tensor | None = Nonedecoder\_input\_ids: np.ndarray | tf.Tensor | None = Nonedecoder\_attention\_mask: np.ndarray | tf.Tensor | None = Noneencoder\_outputs: Optional\[Union\[Tuple, TFBaseModelOutput]] = Nonepast\_key\_values: Optional\[Tuple\[Tuple\[Union\[np.ndarray, tf.Tensor]]]] = Nonedecoder\_inputs\_embeds: np.ndarray | tf.Tensor | None = Nonelabels: np.ndarray | tf.Tensor | None = Noneuse\_cache: Optional\[bool] = Noneoutput\_attentions: Optional\[bool] = Noneoutput\_hidden\_states: Optional\[bool] = Nonereturn\_dict: Optional\[bool] = Nonetraining: bool = False\*\*kwargs ) → [transformers.modeling\_tf\_outputs.TFSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqLMOutput) or `tuple(tf.Tensor)`

Parameters

* **pixel\_values** (`np.ndarray`, `tf.Tensor`, `List[tf.Tensor]` \``Dict[str, tf.Tensor]` or `Dict[str, np.ndarray]` and each example must have the shape `(batch_size, num_channels, height, width)`) — Pixel values. Pixel values can be obtained using the vision’s model’s image processor. For example, using [AutoImageProcessor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoImageProcessor). See [ViTImageProcessor.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/deit#transformers.DeiTFeatureExtractor.__call__) for details.
* **decoder\_input\_ids** (`np.ndarray` or `tf.Tensor` of shape `(batch_size, target_sequence_length)`, *optional*) — Indices of decoder input sequence tokens in the vocabulary.

  Indices can be obtained using [PreTrainedTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details.

  [What are input IDs?](https://huggingface.co/docs/transformers/glossary#input-ids)

  If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see `past_key_values`).

  Provide for sequence to sequence training to the decoder. Indices can be obtained using [PreTrainedTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details.
* **decoder\_attention\_mask** (`np.ndarray` or `tf.Tensor` 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.
* **encoder\_outputs** (`tuple(tuple(tf.Tensor)`, *optional*) — This tuple must consist of (`last_hidden_state`, *optional*: `hidden_states`, *optional*: `attentions`) `last_hidden_state` (`tf.Tensor` of shape `(batch_size, sequence_length, hidden_size)`) is a tensor 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(tf.Tensor))` 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 blocks. 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** (`np.ndarray` or `tf.Tensor` 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. 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.
* **labels** (`np.ndarray` or `tf.Tensor` of shape `(batch_size, sequence_length)`, *optional*) — Labels for computing the masked language modeling loss for the decoder. 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]`
* **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*) — If set to `True`, the model will return a `~utils.Seq2SeqLMOutput` instead of a plain tuple.
* **training** (`bool`, *optional*, defaults to `False`) — Whether or not to use the model in training mode (some modules like dropout modules have different behaviors between training and evaluation).
* **kwargs** (*optional*) — Remaining dictionary of keyword arguments. Keyword arguments come in two flavors:
  * Without a prefix which will be input as `**encoder_kwargs` for the encoder forward function.
  * With a *decoder\_* prefix which will be input as `**decoder_kwargs` for the decoder forward function.

Returns

[transformers.modeling\_tf\_outputs.TFSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqLMOutput) or `tuple(tf.Tensor)`

A [transformers.modeling\_tf\_outputs.TFSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_tf_outputs.TFSeq2SeqLMOutput) or a tuple of `tf.Tensor` (if `return_dict=False` is passed or when `config.return_dict=False`) comprising various elements depending on the configuration ([VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig)) and inputs.

* **loss** (`tf.Tensor` of shape `(n,)`, *optional*, where n is the number of non-masked labels, returned when `labels` is provided) — Language modeling loss.
* **logits** (`tf.Tensor` of shape `(batch_size, sequence_length, config.vocab_size)`) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
* **past\_key\_values** (`List[tf.Tensor]`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`) — List of `tf.Tensor` of length `config.n_layers`, with each tensor of shape `(2, batch_size, num_heads, sequence_length, embed_size_per_head)`).

  Contains pre-computed hidden-states (key and values in the attention blocks) of the decoder that can be used (see `past_key_values` input) to speed up sequential decoding.
* **decoder\_hidden\_states** (`tuple(tf.Tensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `tf.Tensor` (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 decoder at the output of each layer plus the initial embedding outputs.
* **decoder\_attentions** (`tuple(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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** (`tf.Tensor` 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(tf.Tensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `tf.Tensor` (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 encoder at the output of each layer plus the initial embedding outputs.
* **encoder\_attentions** (`tuple(tf.Tensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `tf.Tensor` (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 [TFVisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.TFVisionEncoderDecoderModel) 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:

Copied

```
>>> from transformers import AutoImageProcessor, AutoTokenizer, TFVisionEncoderDecoderModel
>>> from PIL import Image
>>> import requests

>>> image_processor = AutoImageProcessor.from_pretrained("google/vit-base-patch16-224-in21k")
>>> decoder_tokenizer = AutoTokenizer.from_pretrained("gpt2")

>>> # initialize a bert2gpt2 from a pretrained BERT and GPT2 models. Note that the cross-attention layers will be randomly initialized
>>> model = TFVisionEncoderDecoderModel.from_encoder_decoder_pretrained(
...     "google/vit-base-patch16-224-in21k", "gpt2"
... )

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

>>> # forward
>>> pixel_values = image_processor(images=img, return_tensors="tf").pixel_values  # Batch size 1
>>> decoder_input_ids = decoder_tokenizer("Linda Davis", return_tensors="tf").input_ids  # Batch size 1
>>> outputs = model(pixel_values=pixel_values, decoder_input_ids=decoder_input_ids)

>>> # training
>>> outputs = model(pixel_values=pixel_values, decoder_input_ids=decoder_input_ids, labels=decoder_input_ids)
>>> loss, logits = outputs.loss, outputs.logits

>>> # save and load from pretrained
>>> model.save_pretrained("vit-gpt2")
>>> model = TFVisionEncoderDecoderModel.from_pretrained("vit-gpt2")

>>> # generation
>>> generated = model.generate(pixel_values, decoder_start_token_id=model.config.decoder.bos_token_id)
```

**from\_encoder\_decoder\_pretrained**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_tf_vision_encoder_decoder.py#L338)

( encoder\_pretrained\_model\_name\_or\_path: str = Nonedecoder\_pretrained\_model\_name\_or\_path: str = None\*model\_args\*\*kwargs )

Parameters

* **encoder\_pretrained\_model\_name\_or\_path** (`str`, *optional*) — Information necessary to initiate the encoder. Can be either:
  * A string, the *model id* of a pretrained model hosted inside a model repo on boincai.com. An example is `google/vit-base-patch16-224-in21k`.
  * A path to a *directory* containing model weights saved using [save\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.TFPreTrainedModel.save_pretrained), e.g., `./my_model_directory/`.
  * A path or url to a *pytorch index checkpoint file* (e.g, `./pt_model/`). In this case, `encoder_from_pt` should be set to `True`.
* **decoder\_pretrained\_model\_name\_or\_path** (`str`, *optional*, defaults to *None*) — Information necessary to initiate the decoder. Can be either:
  * A string, the *model id* of a pretrained model hosted inside a model repo on boincai.com. Valid model ids can be located at the root-level, like `bert-base-uncased`, or namespaced under a user or organization name, like `dbmdz/bert-base-german-cased`.
  * A path to a *directory* containing model weights saved using [save\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.TFPreTrainedModel.save_pretrained), e.g., `./my_model_directory/`.
  * A path or url to a *pytorch checkpoint file* (e.g, `./pt_model/`). In this case, `decoder_from_pt` should be set to `True`.
* **model\_args** (remaining positional arguments, *optional*) — All remaning positional arguments will be passed to the underlying model’s `__init__` method.
* **kwargs** (remaining dictionary of keyword arguments, *optional*) — Can be used to update the configuration object (after it being loaded) and initiate the model (e.g., `output_attentions=True`).

  * To update the encoder configuration, use the prefix *encoder\_* for each configuration parameter.
  * To update the decoder configuration, use the prefix *decoder\_* for each configuration parameter.
  * To update the parent model configuration, do not use a prefix for each configuration parameter.

  Behaves differently depending on whether a `config` is provided or automatically loaded.

Instantiate an encoder and a decoder from one or two base classes of the library from pretrained model checkpoints.

Example:

Copied

```
>>> from transformers import TFVisionEncoderDecoderModel

>>> # initialize a vit-bert from a pretrained ViT and a pretrained BERT model. Note that the cross-attention layers will be randomly initialized
>>> model = TFVisionEncoderDecoderModel.from_encoder_decoder_pretrained(
...     "google/vit-base-patch16-224-in21k", "bert-base-uncased"
... )
>>> # saving model after fine-tuning
>>> model.save_pretrained("./vit-bert")
>>> # load fine-tuned model
>>> model = TFVisionEncoderDecoderModel.from_pretrained("./vit-bert")
```

### FlaxVisionEncoderDecoderModel

#### class transformers.FlaxVisionEncoderDecoderModel

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_flax_vision_encoder_decoder.py#L268)

( config: VisionEncoderDecoderConfiginput\_shape: typing.Optional\[typing.Tuple] = Noneseed: int = 0dtype: dtype = \<class 'jax.numpy.float32'>\_do\_init: bool = True\*\*kwargs )

Parameters

* **config** ([VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig)) — 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()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.from_pretrained) method to load the model weights.
* **dtype** (`jax.numpy.dtype`, *optional*, defaults to `jax.numpy.float32`) — The data type of the computation. Can be one of `jax.numpy.float32`, `jax.numpy.float16` (on GPUs) and `jax.numpy.bfloat16` (on TPUs).

  This can be used to enable mixed-precision training or half-precision inference on GPUs or TPUs. If specified all the computation will be performed with the given `dtype`.

  **Note that this only specifies the dtype of the computation and does not influence the dtype of model parameters.**

  If you wish to change the dtype of the model parameters, see [to\_fp16()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.to_fp16) and [to\_bf16()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.to_bf16).

This class can be used to initialize an image-to-text-sequence model with any pretrained vision autoencoding model as the encoder and any pretrained text autoregressive model as the decoder. The encoder is loaded via [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.FlaxAutoModelForVision2Seq.from_pretrained) function and the decoder is loaded via [from\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.FlaxAutoModelForVision2Seq.from_pretrained) function. Cross-attention layers are automatically added to the decoder and should be fine-tuned on a downstream generative task, like image captioning.

The effectiveness of initializing sequence-to-sequence models with pretrained checkpoints for sequence generation tasks was shown in [Leveraging Pre-trained Checkpoints for Sequence Generation Tasks](https://arxiv.org/abs/1907.12461) by Sascha Rothe, Shashi Narayan, Aliaksei Severyn. Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu.

Additionally, in [TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models](https://arxiv.org/abs/2109.10282) it is shown how leveraging large pretrained vision models for optical character recognition (OCR) yields a significant performance improvement.

After such a Vision-Encoder-Text-Decoder model has been trained/fine-tuned, it can be saved/loaded just like any other models (see the examples for more information).

This model inherits from [FlaxPreTrainedModel](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel). 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 Flax Linen [flax.nn.Module](https://flax.readthedocs.io/en/latest/_autosummary/flax.nn.module.html) subclass. Use it as a regular Flax Module and refer to the Flax documentation for all matter related to general usage and behavior.

[FlaxVisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.FlaxVisionEncoderDecoderModel) is a generic model class that will be instantiated as a transformer architecture with the module (flax.nn.Module) of one of the base vision model classes of the library as encoder module and another one as decoder module when created with the :met&#x68;*\~transformers.FlaxAutoModel.from\_pretrained* class method for the encoder and :met&#x68;*\~transformers.FlaxAutoModelForCausalLM.from\_pretrained* class method for the decoder.

**\_\_call\_\_**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_flax_vision_encoder_decoder.py#L598)

( pixel\_values: Arraydecoder\_input\_ids: typing.Optional\[jax.Array] = Nonedecoder\_attention\_mask: typing.Optional\[jax.Array] = Nonedecoder\_position\_ids: typing.Optional\[jax.Array] = Noneoutput\_attentions: typing.Optional\[bool] = Noneoutput\_hidden\_states: typing.Optional\[bool] = Nonereturn\_dict: typing.Optional\[bool] = Nonetrain: bool = Falseparams: dict = Nonedropout\_rng: PRNGKey = None ) → [transformers.modeling\_flax\_outputs.FlaxSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqLMOutput) or `tuple(torch.FloatTensor)`

Parameters

* **pixel\_values** (`jnp.ndarray` of shape `(batch_size, num_channels, height, width)`) — Pixel values. Pixel values can be obtained using the vision model’s image processor. For example, using [AutoImageProcessor](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/auto#transformers.AutoImageProcessor). See [ViTImageProcessor.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/deit#transformers.DeiTFeatureExtractor.__call__) for details.
* **decoder\_input\_ids** (`jnp.ndarray` of shape `(batch_size, target_sequence_length)`, *optional*) — Indices of decoder input sequence tokens in the vocabulary.

  Indices can be obtained using [PreTrainedTokenizer](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizer). See [PreTrainedTokenizer.encode()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/tokenizer#transformers.PreTrainedTokenizerFast.encode) and [PreTrainedTokenizer.**call**()](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vits#transformers.VitsTokenizer.__call__) for details.

  [What are decoder input IDs?](https://huggingface.co/docs/transformers/glossary#decoder-input-ids)
* **decoder\_attention\_mask** (`jnp.ndarray` 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.
* **decoder\_position\_ids** (`jnp.ndarray` of shape `(batch_size, sequence_length)`, *optional*) — Indices of positions of each decoder input sequence tokens in the position embeddings. Selected in the range `[0, config.decoder.max_position_embeddings - 1]`.
* **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*) — If set to `True`, the model will return a `~utils.FlaxSeq2SeqLMOutput` instead of a plain tuple.

Returns

[transformers.modeling\_flax\_outputs.FlaxSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqLMOutput) or `tuple(torch.FloatTensor)`

A [transformers.modeling\_flax\_outputs.FlaxSeq2SeqLMOutput](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/output#transformers.modeling_flax_outputs.FlaxSeq2SeqLMOutput) 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 ([VisionEncoderDecoderConfig](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.VisionEncoderDecoderConfig)) and inputs.

* **logits** (`jnp.ndarray` of shape `(batch_size, sequence_length, config.vocab_size)`) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
* **past\_key\_values** (`tuple(tuple(jnp.ndarray))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`) — Tuple of `tuple(jnp.ndarray)` 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(jnp.ndarray)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `jnp.ndarray` (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 decoder at the output of each layer plus the initial embedding outputs.
* **decoder\_attentions** (`tuple(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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** (`jnp.ndarray` 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(jnp.ndarray)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`) — Tuple of `jnp.ndarray` (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 encoder at the output of each layer plus the initial embedding outputs.
* **encoder\_attentions** (`tuple(jnp.ndarray)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`) — Tuple of `jnp.ndarray` (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 [FlaxVisionEncoderDecoderModel](https://huggingface.co/docs/transformers/v4.34.1/en/model_doc/vision-encoder-decoder#transformers.FlaxVisionEncoderDecoderModel) 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:

Copied

```
>>> from transformers import FlaxVisionEncoderDecoderModel, AutoImageProcessor, AutoTokenizer
>>> from PIL import Image
>>> import requests

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

>>> image_processor = AutoImageProcessor.from_pretrained("google/vit-base-patch16-224-in21k")

>>> # load output tokenizer
>>> tokenizer_output = AutoTokenizer.from_pretrained("gpt2")

>>> # initialize a vit-gpt2 from pretrained ViT and GPT2 models. Note that the cross-attention layers will be randomly initialized
>>> model = FlaxVisionEncoderDecoderModel.from_encoder_decoder_pretrained(
...     "google/vit-base-patch16-224-in21k", "gpt2"
... )

>>> pixel_values = image_processor(images=image, return_tensors="np").pixel_values

>>> # use GPT2's eos_token as the pad as well as eos token
>>> model.config.eos_token_id = model.config.decoder.eos_token_id
>>> model.config.pad_token_id = model.config.eos_token_id

>>> # generation
>>> sequences = model.generate(pixel_values, num_beams=4, max_length=12).sequences

>>> captions = tokenizer_output.batch_decode(sequences, skip_special_tokens=True)
```

**from\_encoder\_decoder\_pretrained**

[\<source>](https://github.com/huggingface/transformers/blob/v4.34.1/src/transformers/models/vision_encoder_decoder/modeling_flax_vision_encoder_decoder.py#L723)

( encoder\_pretrained\_model\_name\_or\_path: typing.Union\[str, os.PathLike, NoneType] = Nonedecoder\_pretrained\_model\_name\_or\_path: typing.Union\[str, os.PathLike, NoneType] = None\*model\_args\*\*kwargs )

Parameters

* **encoder\_pretrained\_model\_name\_or\_path** (`Union[str, os.PathLike]`, *optional*) — Information necessary to initiate the encoder. Can be either:
  * A string, the *model id* of a pretrained model hosted inside a model repo on boincai.com. An example is `google/vit-base-patch16-224-in21k`.
  * A path to a *directory* containing model weights saved using [save\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.save_pretrained), e.g., `./my_model_directory/`.
* **decoder\_pretrained\_model\_name\_or\_path** (`Union[str, os.PathLike]`, *optional*, defaults to `None`) — Information necessary to initiate the decoder. Can be either:
  * A string, the *model id* of a pretrained model hosted inside a model repo on boincai.com. Valid model ids can be located at the root-level, like `bert-base-uncased`, or namespaced under a user or organization name, like `dbmdz/bert-base-german-cased`.
  * A path to a *directory* containing model weights saved using [save\_pretrained()](https://huggingface.co/docs/transformers/v4.34.1/en/main_classes/model#transformers.FlaxPreTrainedModel.save_pretrained), e.g., `./my_model_directory/`.
* **model\_args** (remaining positional arguments, *optional*) — All remaning positional arguments will be passed to the underlying model’s `__init__` method.
* **kwargs** (remaining dictionary of keyword arguments, *optional*) — Can be used to update the configuration object (after it being loaded) and initiate the model (e.g., `output_attentions=True`).

  * To update the encoder configuration, use the prefix *encoder\_* for each configuration parameter.
  * To update the decoder configuration, use the prefix *decoder\_* for each configuration parameter.
  * To update the parent model configuration, do not use a prefix for each configuration parameter.

  Behaves differently depending on whether a `config` is provided or automatically loaded.

Instantiate an encoder and a decoder from one or two base classes of the library from pretrained model checkpoints.

Example:

Copied

```
>>> from transformers import FlaxVisionEncoderDecoderModel

>>> # initialize a vit-gpt2 from a pretrained ViT and a pretrained GPT2 model. Note that the cross-attention layers will be randomly initialized
>>> model = FlaxVisionEncoderDecoderModel.from_encoder_decoder_pretrained(
...     "google/vit-base-patch16-224-in21k", "gpt2"
... )
>>> # saving model after fine-tuning
>>> model.save_pretrained("./vit-gpt2")
>>> # load fine-tuned model
>>> model = FlaxVisionEncoderDecoderModel.from_pretrained("./vit-gpt2")
```


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://boinc-ai.gitbook.io/transformers/api/models/multimodal-models/vision-encoder-decoder-models.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.