# coding=utf-8 # Copyright Microsoft Research and The HuggingFace Inc. team. All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """BEiT model configuration""" import warnings from collections import OrderedDict from typing import Mapping from packaging import version from ...configuration_utils import PretrainedConfig from ...onnx import OnnxConfig from ...utils.backbone_utils import BackboneConfigMixin, get_aligned_output_features_output_indices class BeitConfig(BackboneConfigMixin, PretrainedConfig): r""" This is the configuration class to store the configuration of a [`BeitModel`]. It is used to instantiate an BEiT 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 BEiT [microsoft/beit-base-patch16-224-pt22k](https://huggingface.co/microsoft/beit-base-patch16-224-pt22k) architecture. Args: vocab_size (`int`, *optional*, defaults to 8192): Vocabulary size of the BEiT model. Defines the number of different image tokens that can be used during pre-training. hidden_size (`int`, *optional*, defaults to 768): Dimensionality of the encoder layers and the pooler layer. num_hidden_layers (`int`, *optional*, defaults to 12): Number of hidden layers in the Transformer encoder. num_attention_heads (`int`, *optional*, defaults to 12): Number of attention heads for each attention layer in the Transformer encoder. intermediate_size (`int`, *optional*, defaults to 3072): Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder. hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`): The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`, `"relu"`, `"selu"` and `"gelu_new"` are supported. hidden_dropout_prob (`float`, *optional*, defaults to 0.0): The dropout probability for all fully connected layers in the embeddings, encoder, and pooler. attention_probs_dropout_prob (`float`, *optional*, defaults to 0.0): The dropout ratio for the attention probabilities. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. layer_norm_eps (`float`, *optional*, defaults to 1e-12): The epsilon used by the layer normalization layers. image_size (`int`, *optional*, defaults to 224): The size (resolution) of each image. patch_size (`int`, *optional*, defaults to 16): The size (resolution) of each patch. num_channels (`int`, *optional*, defaults to 3): The number of input channels. use_mask_token (`bool`, *optional*, defaults to `False`): Whether to use a mask token for masked image modeling. use_absolute_position_embeddings (`bool`, *optional*, defaults to `False`): Whether to use BERT-style absolute position embeddings. use_relative_position_bias (`bool`, *optional*, defaults to `False`): Whether to use T5-style relative position embeddings in the self-attention layers. use_shared_relative_position_bias (`bool`, *optional*, defaults to `False`): Whether to use the same relative position embeddings across all self-attention layers of the Transformer. layer_scale_init_value (`float`, *optional*, defaults to 0.1): Scale to use in the self-attention layers. 0.1 for base, 1e-5 for large. Set 0 to disable layer scale. drop_path_rate (`float`, *optional*, defaults to 0.1): Stochastic depth rate per sample (when applied in the main path of residual layers). use_mean_pooling (`bool`, *optional*, defaults to `True`): Whether to mean pool the final hidden states of the patches instead of using the final hidden state of the CLS token, before applying the classification head. pool_scales (`Tuple[int]`, *optional*, defaults to `[1, 2, 3, 6]`): Pooling scales used in Pooling Pyramid Module applied on the last feature map. use_auxiliary_head (`bool`, *optional*, defaults to `True`): Whether to use an auxiliary head during training. auxiliary_loss_weight (`float`, *optional*, defaults to 0.4): Weight of the cross-entropy loss of the auxiliary head. auxiliary_channels (`int`, *optional*, defaults to 256): Number of channels to use in the auxiliary head. auxiliary_num_convs (`int`, *optional*, defaults to 1): Number of convolutional layers to use in the auxiliary head. auxiliary_concat_input (`bool`, *optional*, defaults to `False`): Whether to concatenate the output of the auxiliary head with the input before the classification layer. semantic_loss_ignore_index (`int`, *optional*, defaults to 255): The index that is ignored by the loss function of the semantic segmentation model. out_features (`List[str]`, *optional*): If used as backbone, list of features to output. Can be any of `"stem"`, `"stage1"`, `"stage2"`, etc. (depending on how many stages the model has). If unset and `out_indices` is set, will default to the corresponding stages. If unset and `out_indices` is unset, will default to the last stage. Must be in the same order as defined in the `stage_names` attribute. out_indices (`List[int]`, *optional*): If used as backbone, list of indices of features to output. Can be any of 0, 1, 2, etc. (depending on how many stages the model has). If unset and `out_features` is set, will default to the corresponding stages. If unset and `out_features` is unset, will default to the last stage. Must be in the same order as defined in the `stage_names` attribute. add_fpn (`bool`, *optional*, defaults to `False`): Whether to add a FPN as part of the backbone. Only relevant for [`BeitBackbone`]. reshape_hidden_states (`bool`, *optional*, defaults to `True`): Whether to reshape the feature maps to 4D tensors of shape `(batch_size, hidden_size, height, width)` in case the model is used as backbone. If `False`, the feature maps will be 3D tensors of shape `(batch_size, seq_len, hidden_size)`. Only relevant for [`BeitBackbone`]. Example: ```python >>> from transformers import BeitConfig, BeitModel >>> # Initializing a BEiT beit-base-patch16-224-pt22k style configuration >>> configuration = BeitConfig() >>> # Initializing a model (with random weights) from the beit-base-patch16-224-pt22k style configuration >>> model = BeitModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```""" model_type = "beit" def __init__( self, vocab_size=8192, 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, use_mask_token=False, use_absolute_position_embeddings=False, use_relative_position_bias=False, use_shared_relative_position_bias=False, layer_scale_init_value=0.1, drop_path_rate=0.1, use_mean_pooling=True, pool_scales=[1, 2, 3, 6], use_auxiliary_head=True, auxiliary_loss_weight=0.4, auxiliary_channels=256, auxiliary_num_convs=1, auxiliary_concat_input=False, semantic_loss_ignore_index=255, out_features=None, out_indices=None, add_fpn=False, reshape_hidden_states=True, **kwargs, ): super().__init__(**kwargs) self.vocab_size = vocab_size self.hidden_size = hidden_size self.num_hidden_layers = num_hidden_layers self.num_attention_heads = num_attention_heads self.intermediate_size = intermediate_size self.hidden_act = hidden_act self.hidden_dropout_prob = hidden_dropout_prob self.attention_probs_dropout_prob = attention_probs_dropout_prob self.initializer_range = initializer_range self.layer_norm_eps = layer_norm_eps self.image_size = image_size self.patch_size = patch_size self.num_channels = num_channels self.use_mask_token = use_mask_token self.use_absolute_position_embeddings = use_absolute_position_embeddings self.use_relative_position_bias = use_relative_position_bias self.use_shared_relative_position_bias = use_shared_relative_position_bias self.layer_scale_init_value = layer_scale_init_value self.drop_path_rate = drop_path_rate self.use_mean_pooling = use_mean_pooling # decode head attributes (semantic segmentation) self.pool_scales = pool_scales # auxiliary head attributes (semantic segmentation) self.use_auxiliary_head = use_auxiliary_head self.auxiliary_loss_weight = auxiliary_loss_weight self.auxiliary_channels = auxiliary_channels self.auxiliary_num_convs = auxiliary_num_convs self.auxiliary_concat_input = auxiliary_concat_input self.semantic_loss_ignore_index = semantic_loss_ignore_index # handle backwards compatibility if "segmentation_indices" in kwargs: warnings.warn( "The `segmentation_indices` argument is deprecated and will be removed in a future version, use `out_indices` instead.", FutureWarning, ) out_indices = kwargs.pop("segmentation_indices") # backbone attributes self.stage_names = ["stem"] + [f"stage{idx}" for idx in range(1, self.num_hidden_layers + 1)] self._out_features, self._out_indices = get_aligned_output_features_output_indices( out_features=out_features, out_indices=out_indices, stage_names=self.stage_names ) self.add_fpn = add_fpn self.reshape_hidden_states = reshape_hidden_states # Copied from transformers.models.vit.configuration_vit.ViTOnnxConfig class BeitOnnxConfig(OnnxConfig): torch_onnx_minimum_version = version.parse("1.11") @property def inputs(self) -> Mapping[str, Mapping[int, str]]: return OrderedDict( [ ("pixel_values", {0: "batch", 1: "num_channels", 2: "height", 3: "width"}), ] ) @property def atol_for_validation(self) -> float: return 1e-4 __all__ = ["BeitConfig", "BeitOnnxConfig"]