The ViTMAE model was proposed in Masked Autoencoders Are Scalable Vision Learners by Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, Ross Girshick. The paper shows that, by pre-training a Vision Transformer (ViT) to reconstruct pixel values for masked patches, one can get results after fine-tuning that outperform supervised pre-training.
The abstract from the paper is the following:
This paper shows that masked autoencoders (MAE) are scalable self-supervised learners for computer vision. Our MAE approach is simple: we mask random patches of the input image and reconstruct the missing pixels. It is based on two core designs. First, we develop an asymmetric encoder-decoder architecture, with an encoder that operates only on the visible subset of patches (without mask tokens), along with a lightweight decoder that reconstructs the original image from the latent representation and mask tokens. Second, we find that masking a high proportion of the input image, e.g., 75%, yields a nontrivial and meaningful self-supervisory task. Coupling these two designs enables us to train large models efficiently and effectively: we accelerate training (by 3x or more) and improve accuracy. Our scalable approach allows for learning high-capacity models that generalize well: e.g., a vanilla ViT-Huge model achieves the best accuracy (87.8%) among methods that use only ImageNet-1K data. Transfer performance in downstream tasks outperforms supervised pre-training and shows promising scaling behavior.
MAE architecture. Taken from the original paper. This model was contributed by nielsr. TensorFlow version of the model was contributed by sayakpaul and ariG23498 (equal contribution). The original code can be found here.
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with ViTMAE.
If you’re interested in submitting a resource to be included here, please feel free to open a Pull Request and we’ll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
( hidden_size = 768 num_hidden_layers = 12 num_attention_heads = 12 intermediate_size = 3072 hidden_act = 'gelu' hidden_dropout_prob = 0.0 attention_probs_dropout_prob = 0.0 initializer_range = 0.02 layer_norm_eps = 1e-12 image_size = 224 patch_size = 16 num_channels = 3 qkv_bias = True decoder_num_attention_heads = 16 decoder_hidden_size = 512 decoder_num_hidden_layers = 8 decoder_intermediate_size = 2048 mask_ratio = 0.75 norm_pix_loss = False **kwargs )
Parameters
int, optional, defaults to 768) —
Dimensionality of the encoder layers and the pooler layer. int, optional, defaults to 12) —
Number of hidden layers in the Transformer encoder. int, optional, defaults to 12) —
Number of attention heads for each attention layer in the Transformer encoder. int, optional, defaults to 3072) —
Dimensionality of the “intermediate” (i.e., feed-forward) layer in the Transformer encoder. str or function, optional, defaults to "gelu") —
The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu",
"relu", "selu" and "gelu_new" are supported. float, optional, defaults to 0.0) —
The dropout probability for all fully connected layers in the embeddings, encoder, and pooler. float, optional, defaults to 0.0) —
The dropout ratio for the attention probabilities. float, optional, defaults to 0.02) —
The standard deviation of the truncated_normal_initializer for initializing all weight matrices. float, optional, defaults to 1e-12) —
The epsilon used by the layer normalization layers. int, optional, defaults to 224) —
The size (resolution) of each image. int, optional, defaults to 16) —
The size (resolution) of each patch. int, optional, defaults to 3) —
The number of input channels. bool, optional, defaults to True) —
Whether to add a bias to the queries, keys and values. int, optional, defaults to 16) —
Number of attention heads for each attention layer in the decoder. int, optional, defaults to 512) —
Dimensionality of the decoder. int, optional, defaults to 8) —
Number of hidden layers in the decoder. int, optional, defaults to 2048) —
Dimensionality of the “intermediate” (i.e., feed-forward) layer in the decoder. float, optional, defaults to 0.75) —
The ratio of the number of masked tokens in the input sequence. bool, optional, defaults to False) —
Whether or not to train with normalized pixels (see Table 3 in the paper). Using normalized pixels improved
representation quality in the experiments of the authors. This is the configuration class to store the configuration of a ViTMAEModel. It is used to instantiate an ViT MAE 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 ViT facebook/vit-mae-base architecture.
Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.
Example:
>>> from transformers import ViTMAEConfig, ViTMAEModel
>>> # Initializing a ViT MAE vit-mae-base style configuration
>>> configuration = ViTMAEConfig()
>>> # Initializing a model (with random weights) from the vit-mae-base style configuration
>>> model = ViTMAEModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config( config )
Parameters
The bare ViTMAE 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.
( pixel_values: Optional = None noise: Optional = None head_mask: Optional = None output_attentions: Optional = None output_hidden_states: Optional = None return_dict: Optional = None ) → transformers.models.vit_mae.modeling_vit_mae.ViTMAEModelOutput or tuple(torch.FloatTensor)
Parameters
torch.FloatTensor of shape (batch_size, num_channels, height, width)) —
Pixel values. Pixel values can be obtained using AutoImageProcessor. See ViTImageProcessor.call()
for details. 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]:
bool, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions under returned
tensors for more detail. bool, optional) —
Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for
more detail. bool, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. Returns
transformers.models.vit_mae.modeling_vit_mae.ViTMAEModelOutput or tuple(torch.FloatTensor)
A transformers.models.vit_mae.modeling_vit_mae.ViTMAEModelOutput 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 (ViTMAEConfig) and inputs.
torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)) — Sequence of hidden-states at the output of the last layer of the model.torch.FloatTensor of shape (batch_size, sequence_length)) — Tensor indicating which patches are masked (1) and which are not (0).torch.LongTensor of shape (batch_size, sequence_length)) — Tensor containing the original index of the (shuffled) masked patches.tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of
shape (batch_size, sequence_length, hidden_size). Hidden-states of the model at the output of each layer
plus the initial embedding outputs.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 ViTMAEModel forward method, overrides the __call__ special method.
Although the recipe for forward pass needs to be defined within this function, one should call the Module
instance afterwards instead of this since the former takes care of running the pre and post processing steps while
the latter silently ignores them.
Examples:
>>> from transformers import AutoImageProcessor, ViTMAEModel
>>> 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("facebook/vit-mae-base")
>>> model = ViTMAEModel.from_pretrained("facebook/vit-mae-base")
>>> inputs = image_processor(images=image, return_tensors="pt")
>>> outputs = model(**inputs)
>>> last_hidden_states = outputs.last_hidden_state( config )
Parameters
The ViTMAE Model transformer with the decoder on top for self-supervised pre-training.
Note that we provide a script to pre-train this model on custom data in our examples directory.
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.
( pixel_values: Optional = None noise: Optional = None head_mask: Optional = None output_attentions: Optional = None output_hidden_states: Optional = None return_dict: Optional = None ) → transformers.models.vit_mae.modeling_vit_mae.ViTMAEForPreTrainingOutput or tuple(torch.FloatTensor)
Parameters
torch.FloatTensor of shape (batch_size, num_channels, height, width)) —
Pixel values. Pixel values can be obtained using AutoImageProcessor. See ViTImageProcessor.call()
for details. 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]:
bool, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions under returned
tensors for more detail. bool, optional) —
Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for
more detail. bool, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. Returns
transformers.models.vit_mae.modeling_vit_mae.ViTMAEForPreTrainingOutput or tuple(torch.FloatTensor)
A transformers.models.vit_mae.modeling_vit_mae.ViTMAEForPreTrainingOutput 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 (ViTMAEConfig) and inputs.
torch.FloatTensor of shape (1,)) — Pixel reconstruction loss.torch.FloatTensor of shape (batch_size, sequence_length, patch_size ** 2 * num_channels)) — Pixel reconstruction logits.torch.FloatTensor of shape (batch_size, sequence_length)) — Tensor indicating which patches are masked (1) and which are not (0).torch.LongTensor of shape (batch_size, sequence_length)) — Tensor containing the original index of the (shuffled) masked patches.tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of
shape (batch_size, sequence_length, hidden_size). Hidden-states of the model at the output of each layer
plus the initial embedding outputs.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 ViTMAEForPreTraining forward method, overrides the __call__ special method.
Although the recipe for forward pass needs to be defined within this function, one should call the Module
instance afterwards instead of this since the former takes care of running the pre and post processing steps while
the latter silently ignores them.
Examples:
>>> from transformers import AutoImageProcessor, ViTMAEForPreTraining
>>> 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("facebook/vit-mae-base")
>>> model = ViTMAEForPreTraining.from_pretrained("facebook/vit-mae-base")
>>> inputs = image_processor(images=image, return_tensors="pt")
>>> outputs = model(**inputs)
>>> loss = outputs.loss
>>> mask = outputs.mask
>>> ids_restore = outputs.ids_restore( config: ViTMAEConfig *inputs **kwargs )
Parameters
The bare ViTMAE Model transformer outputting raw hidden-states without any specific head on top. This model inherits from 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 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.
TensorFlow models and layers in transformers accept two formats as input:
The reason the second format is supported is that Keras methods prefer this format when passing inputs to models
and layers. Because of this support, when using methods like model.fit() things should “just work” for you - just
pass your inputs and labels in any format that model.fit() supports! If, however, you want to use the second
format outside of Keras methods like fit() and predict(), such as when creating your own layers or models with
the Keras Functional API, there are three possibilities you can use to gather all the input Tensors in the first
positional argument:
pixel_values only and nothing else: model(pixel_values)model([pixel_values, attention_mask]) or model([pixel_values, attention_mask, token_type_ids])model({"pixel_values": pixel_values, "token_type_ids": token_type_ids})Note that when creating models and layers with subclassing then you don’t need to worry about any of this, as you can just pass inputs like you would to any other Python function!
( pixel_values: TFModelInputType | None = None noise: tf.Tensor = None head_mask: np.ndarray | tf.Tensor | None = None output_attentions: Optional[bool] = None output_hidden_states: Optional[bool] = None return_dict: Optional[bool] = None training: bool = False ) → transformers.models.vit_mae.modeling_tf_vit_mae.TFViTMAEModelOutput or tuple(tf.Tensor)
Parameters
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 AutoImageProcessor. See ViTImageProcessor.call()
for details. np.ndarray or tf.Tensor 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]:
bool, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions under returned
tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
config will be used instead. bool, optional) —
Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for
more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
used instead. bool, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. This argument can be used
in eager mode, in graph mode the value will always be set to True. 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). Returns
transformers.models.vit_mae.modeling_tf_vit_mae.TFViTMAEModelOutput or tuple(tf.Tensor)
A transformers.models.vit_mae.modeling_tf_vit_mae.TFViTMAEModelOutput 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 (ViTMAEConfig) and inputs.
tf.Tensor of shape (batch_size, sequence_length, hidden_size)) — Sequence of hidden-states at the output of the last layer of the model.tf.Tensor of shape (batch_size, sequence_length)) — Tensor indicating which patches are masked (1) and which are not (0).tf.Tensor of shape (batch_size, sequence_length)) — Tensor containing the original index of the (shuffled) masked patches.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 model at the output of each layer plus
the initial embedding outputs.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 after the attention softmax, used to compute the weighted average in
the self-attention heads.The TFViTMAEModel forward method, overrides the __call__ special method.
Although the recipe for forward pass needs to be defined within this function, one should call the Module
instance afterwards instead of this since the former takes care of running the pre and post processing steps while
the latter silently ignores them.
Examples:
>>> from transformers import AutoImageProcessor, TFViTMAEModel
>>> 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("facebook/vit-mae-base")
>>> model = TFViTMAEModel.from_pretrained("facebook/vit-mae-base")
>>> inputs = image_processor(images=image, return_tensors="tf")
>>> outputs = model(**inputs)
>>> last_hidden_states = outputs.last_hidden_state( config )
Parameters
The ViTMAE Model transformer with the decoder on top for self-supervised pre-training. This model inherits from 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 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.
TensorFlow models and layers in transformers accept two formats as input:
The reason the second format is supported is that Keras methods prefer this format when passing inputs to models
and layers. Because of this support, when using methods like model.fit() things should “just work” for you - just
pass your inputs and labels in any format that model.fit() supports! If, however, you want to use the second
format outside of Keras methods like fit() and predict(), such as when creating your own layers or models with
the Keras Functional API, there are three possibilities you can use to gather all the input Tensors in the first
positional argument:
pixel_values only and nothing else: model(pixel_values)model([pixel_values, attention_mask]) or model([pixel_values, attention_mask, token_type_ids])model({"pixel_values": pixel_values, "token_type_ids": token_type_ids})Note that when creating models and layers with subclassing then you don’t need to worry about any of this, as you can just pass inputs like you would to any other Python function!
( pixel_values: TFModelInputType | None = None noise: tf.Tensor = None head_mask: np.ndarray | tf.Tensor | None = None output_attentions: Optional[bool] = None output_hidden_states: Optional[bool] = None return_dict: Optional[bool] = None training: bool = False ) → transformers.models.vit_mae.modeling_tf_vit_mae.TFViTMAEForPreTrainingOutput or tuple(tf.Tensor)
Parameters
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 AutoImageProcessor. See ViTImageProcessor.call()
for details. np.ndarray or tf.Tensor 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]:
bool, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions under returned
tensors for more detail. This argument can be used only in eager mode, in graph mode the value in the
config will be used instead. bool, optional) —
Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for
more detail. This argument can be used only in eager mode, in graph mode the value in the config will be
used instead. bool, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. This argument can be used
in eager mode, in graph mode the value will always be set to True. 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). Returns
transformers.models.vit_mae.modeling_tf_vit_mae.TFViTMAEForPreTrainingOutput or tuple(tf.Tensor)
A transformers.models.vit_mae.modeling_tf_vit_mae.TFViTMAEForPreTrainingOutput 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 (ViTMAEConfig) and inputs.
tf.Tensor of shape (1,)) — Pixel reconstruction loss.tf.Tensor of shape (batch_size, sequence_length, patch_size ** 2 * num_channels)) — Pixel reconstruction logits.tf.Tensor of shape (batch_size, sequence_length)) — Tensor indicating which patches are masked (1) and which are not (0).tf.Tensor of shape (batch_size, sequence_length)) — Tensor containing the original index of the (shuffled) masked patches.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 model at the output of each layer plus
the initial embedding outputs.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 after the attention softmax, used to compute the weighted average in
the self-attention heads.The TFViTMAEForPreTraining forward method, overrides the __call__ special method.
Although the recipe for forward pass needs to be defined within this function, one should call the Module
instance afterwards instead of this since the former takes care of running the pre and post processing steps while
the latter silently ignores them.
Examples:
>>> from transformers import AutoImageProcessor, TFViTMAEForPreTraining
>>> 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("facebook/vit-mae-base")
>>> model = TFViTMAEForPreTraining.from_pretrained("facebook/vit-mae-base")
>>> inputs = image_processor(images=image, return_tensors="pt")
>>> outputs = model(**inputs)
>>> loss = outputs.loss
>>> mask = outputs.mask
>>> ids_restore = outputs.ids_restore