DoruC/Grounded-Segment-Anything

1# coding=utf-82# Copyright 2021 The HuggingFace Inc. team. All rights reserved.3#4# Licensed under the Apache License, Version 2.0 (the "License");5# you may not use this file except in compliance with the License.6# You may obtain a copy of the License at7#8#     http://www.apache.org/licenses/LICENSE-2.09#10# Unless required by applicable law or agreed to in writing, software11# distributed under the License is distributed on an "AS IS" BASIS,12# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.13# See the License for the specific language governing permissions and14# limitations under the License.15""" VisualBERT model configuration"""16 17from ...configuration_utils import PretrainedConfig18from ...utils import logging19 20 21logger = logging.get_logger(__name__)22 23VISUAL_BERT_PRETRAINED_CONFIG_ARCHIVE_MAP = {24    "uclanlp/visualbert-vqa": "https://huggingface.co/uclanlp/visualbert-vqa/resolve/main/config.json",25    "uclanlp/visualbert-vqa-pre": "https://huggingface.co/uclanlp/visualbert-vqa-pre/resolve/main/config.json",26    "uclanlp/visualbert-vqa-coco-pre": (27        "https://huggingface.co/uclanlp/visualbert-vqa-coco-pre/resolve/main/config.json"28    ),29    "uclanlp/visualbert-vcr": "https://huggingface.co/uclanlp/visualbert-vcr/resolve/main/config.json",30    "uclanlp/visualbert-vcr-pre": "https://huggingface.co/uclanlp/visualbert-vcr-pre/resolve/main/config.json",31    "uclanlp/visualbert-vcr-coco-pre": (32        "https://huggingface.co/uclanlp/visualbert-vcr-coco-pre/resolve/main/config.json"33    ),34    "uclanlp/visualbert-nlvr2": "https://huggingface.co/uclanlp/visualbert-nlvr2/resolve/main/config.json",35    "uclanlp/visualbert-nlvr2-pre": "https://huggingface.co/uclanlp/visualbert-nlvr2-pre/resolve/main/config.json",36    "uclanlp/visualbert-nlvr2-coco-pre": (37        "https://huggingface.co/uclanlp/visualbert-nlvr2-coco-pre/resolve/main/config.json"38    )39    # See all VisualBERT models at https://huggingface.co/models?filter=visual_bert40}41 42 43class VisualBertConfig(PretrainedConfig):44    r"""45    This is the configuration class to store the configuration of a [`VisualBertModel`]. It is used to instantiate an46    VisualBERT model according to the specified arguments, defining the model architecture. Instantiating a47    configuration with the defaults will yield a similar configuration to that of the VisualBERT48    [uclanlp/visualbert-vqa-coco-pre](https://huggingface.co/uclanlp/visualbert-vqa-coco-pre) architecture.49 50    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the51    documentation from [`PretrainedConfig`] for more information.52 53 54    Args:55        vocab_size (`int`, *optional*, defaults to 30522):56            Vocabulary size of the VisualBERT model. Defines the number of different tokens that can be represented by57            the `inputs_ids` passed when calling [`VisualBertModel`]. Vocabulary size of the model. Defines the58            different tokens that can be represented by the `inputs_ids` passed to the forward method of59            [`VisualBertModel`].60        hidden_size (`int`, *optional*, defaults to 768):61            Dimensionality of the encoder layers and the pooler layer.62        visual_embedding_dim (`int`, *optional*, defaults to 512):63            Dimensionality of the visual embeddings to be passed to the model.64        num_hidden_layers (`int`, *optional*, defaults to 12):65            Number of hidden layers in the Transformer encoder.66        num_attention_heads (`int`, *optional*, defaults to 12):67            Number of attention heads for each attention layer in the Transformer encoder.68        intermediate_size (`int`, *optional*, defaults to 3072):69            Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.70        hidden_act (`str` or `function`, *optional*, defaults to `"gelu"`):71            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,72            `"relu"`, `"selu"` and `"gelu_new"` are supported.73        hidden_dropout_prob (`float`, *optional*, defaults to 0.1):74            The dropout probabilitiy for all fully connected layers in the embeddings, encoder, and pooler.75        attention_probs_dropout_prob (`float`, *optional*, defaults to 0.1):76            The dropout ratio for the attention probabilities.77        max_position_embeddings (`int`, *optional*, defaults to 512):78            The maximum sequence length that this model might ever be used with. Typically set this to something large79            just in case (e.g., 512 or 1024 or 2048).80        type_vocab_size (`int`, *optional*, defaults to 2):81            The vocabulary size of the `token_type_ids` passed when calling [`VisualBertModel`].82        initializer_range (`float`, *optional*, defaults to 0.02):83            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.84        layer_norm_eps (`float`, *optional*, defaults to 1e-12):85            The epsilon used by the layer normalization layers.86        bypass_transformer (`bool`, *optional*, defaults to `False`):87            Whether or not the model should bypass the transformer for the visual embeddings. If set to `True`, the88            model directly concatenates the visual embeddings from [`VisualBertEmbeddings`] with text output from89            transformers, and then pass it to a self-attention layer.90        special_visual_initialize (`bool`, *optional*, defaults to `True`):91            Whether or not the visual token type and position type embedding weights should be initialized the same as92            the textual token type and positive type embeddings. When set to `True`, the weights of the textual token93            type and position type embeddings are copied to the respective visual embedding layers.94 95 96    Example:97 98    ```python99    >>> from transformers import VisualBertConfig, VisualBertModel100 101    >>> # Initializing a VisualBERT visualbert-vqa-coco-pre style configuration102    >>> configuration = VisualBertConfig.from_pretrained("uclanlp/visualbert-vqa-coco-pre")103 104    >>> # Initializing a model (with random weights) from the visualbert-vqa-coco-pre style configuration105    >>> model = VisualBertModel(configuration)106 107    >>> # Accessing the model configuration108    >>> configuration = model.config109    ```"""110 111    model_type = "visual_bert"112 113    def __init__(114        self,115        vocab_size=30522,116        hidden_size=768,117        visual_embedding_dim=512,118        num_hidden_layers=12,119        num_attention_heads=12,120        intermediate_size=3072,121        hidden_act="gelu",122        hidden_dropout_prob=0.1,123        attention_probs_dropout_prob=0.1,124        max_position_embeddings=512,125        type_vocab_size=2,126        initializer_range=0.02,127        layer_norm_eps=1e-12,128        bypass_transformer=False,129        special_visual_initialize=True,130        pad_token_id=1,131        bos_token_id=0,132        eos_token_id=2,133        **kwargs,134    ):135        super().__init__(pad_token_id=pad_token_id, bos_token_id=bos_token_id, eos_token_id=eos_token_id, **kwargs)136 137        self.vocab_size = vocab_size138        self.max_position_embeddings = max_position_embeddings139        self.hidden_size = hidden_size140        self.visual_embedding_dim = visual_embedding_dim141        self.num_hidden_layers = num_hidden_layers142        self.num_attention_heads = num_attention_heads143        self.intermediate_size = intermediate_size144        self.hidden_act = hidden_act145        self.hidden_dropout_prob = hidden_dropout_prob146        self.attention_probs_dropout_prob = attention_probs_dropout_prob147        self.initializer_range = initializer_range148        self.type_vocab_size = type_vocab_size149        self.layer_norm_eps = layer_norm_eps150        self.bypass_transformer = bypass_transformer151        self.special_visual_initialize = special_visual_initialize152