image-match/python/py/Lib/site-packages/transformers/models/layoutlmv2/configuration_layoutlmv2.py

# 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.
"""LayoutLMv2 model configuration"""

from huggingface_hub.dataclasses import strict

from ...configuration_utils import PreTrainedConfig
from ...utils import auto_docstring, is_detectron2_available


# soft dependency
if is_detectron2_available():
    import detectron2


@auto_docstring(checkpoint="microsoft/layoutlmv2-base-uncased")
@strict
class LayoutLMv2Config(PreTrainedConfig):
    r"""
    max_2d_position_embeddings (`int`, *optional*, defaults to 1024):
        The maximum value that the 2D position embedding might ever be used with. Typically set this to something
        large just in case (e.g., 1024).
    max_rel_pos (`int`, *optional*, defaults to 128):
        The maximum number of relative positions to be used in the self-attention mechanism.
    rel_pos_bins (`int`, *optional*, defaults to 32):
        The number of relative position bins to be used in the self-attention mechanism.
    fast_qkv (`bool`, *optional*, defaults to `True`):
        Whether or not to use a single matrix for the queries, keys, values in the self-attention layers.
    max_rel_2d_pos (`int`, *optional*, defaults to 256):
        The maximum number of relative 2D positions in the self-attention mechanism.
    rel_2d_pos_bins (`int`, *optional*, defaults to 64):
        The number of 2D relative position bins in the self-attention mechanism.
    convert_sync_batchnorm (`bool`, *optional*, defaults to `True`):
        Whether or not to convert batch normalization layers to synchronized batch normalization layers.
    image_feature_pool_shape (`list[int]`, *optional*, defaults to `[7, 7, 256]`):
        The shape of the average-pooled feature map.
    coordinate_size (`int`, *optional*, defaults to 128):
        Dimension of the coordinate embeddings.
    shape_size (`int`, *optional*, defaults to 128):
        Dimension of the width and height embeddings.
    has_relative_attention_bias (`bool`, *optional*, defaults to `True`):
        Whether or not to use a relative attention bias in the self-attention mechanism.
    has_spatial_attention_bias (`bool`, *optional*, defaults to `True`):
        Whether or not to use a spatial attention bias in the self-attention mechanism.
    has_visual_segment_embedding (`bool`, *optional*, defaults to `False`):
        Whether or not to add visual segment embeddings.
    detectron2_config_args (`dict`, *optional*):
        Dictionary containing the configuration arguments of the Detectron2 visual backbone. Refer to [this
        file](https://github.com/microsoft/unilm/blob/master/layoutlmft/layoutlmft/models/layoutlmv2/detectron2_config.py)
        for details regarding default values.

    Example:

    ```python
    >>> from transformers import LayoutLMv2Config, LayoutLMv2Model

    >>> # Initializing a LayoutLMv2 microsoft/layoutlmv2-base-uncased style configuration
    >>> configuration = LayoutLMv2Config()

    >>> # Initializing a model (with random weights) from the microsoft/layoutlmv2-base-uncased style configuration
    >>> model = LayoutLMv2Model(configuration)

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

    model_type = "layoutlmv2"

    vocab_size: int = 30522
    hidden_size: int = 768
    num_hidden_layers: int = 12
    num_attention_heads: int = 12
    intermediate_size: int = 3072
    hidden_act: str = "gelu"
    hidden_dropout_prob: float | int = 0.1
    attention_probs_dropout_prob: float | int = 0.1
    max_position_embeddings: int = 512
    type_vocab_size: int = 2
    initializer_range: float = 0.02
    layer_norm_eps: float = 1e-12
    pad_token_id: int | None = 0
    max_2d_position_embeddings: int = 1024
    max_rel_pos: int = 128
    rel_pos_bins: int = 32
    fast_qkv: bool = True
    max_rel_2d_pos: int = 256
    rel_2d_pos_bins: int = 64
    convert_sync_batchnorm: bool = True
    image_feature_pool_shape: list[int] | tuple[int, ...] = (7, 7, 256)
    coordinate_size: int = 128
    shape_size: int = 128
    has_relative_attention_bias: bool = True
    has_spatial_attention_bias: bool = True
    has_visual_segment_embedding: bool = False
    detectron2_config_args: dict | None = None

    def __post_init__(self, **kwargs):
        super().__post_init__(**kwargs)
        self.detectron2_config_args = (
            self.detectron2_config_args
            if self.detectron2_config_args is not None
            else self.get_default_detectron2_config()
        )

    @classmethod
    def get_default_detectron2_config(cls):
        return {
            "MODEL.MASK_ON": True,
            "MODEL.PIXEL_STD": [57.375, 57.120, 58.395],
            "MODEL.BACKBONE.NAME": "build_resnet_fpn_backbone",
            "MODEL.FPN.IN_FEATURES": ["res2", "res3", "res4", "res5"],
            "MODEL.ANCHOR_GENERATOR.SIZES": [[32], [64], [128], [256], [512]],
            "MODEL.RPN.IN_FEATURES": ["p2", "p3", "p4", "p5", "p6"],
            "MODEL.RPN.PRE_NMS_TOPK_TRAIN": 2000,
            "MODEL.RPN.PRE_NMS_TOPK_TEST": 1000,
            "MODEL.RPN.POST_NMS_TOPK_TRAIN": 1000,
            "MODEL.POST_NMS_TOPK_TEST": 1000,
            "MODEL.ROI_HEADS.NAME": "StandardROIHeads",
            "MODEL.ROI_HEADS.NUM_CLASSES": 5,
            "MODEL.ROI_HEADS.IN_FEATURES": ["p2", "p3", "p4", "p5"],
            "MODEL.ROI_BOX_HEAD.NAME": "FastRCNNConvFCHead",
            "MODEL.ROI_BOX_HEAD.NUM_FC": 2,
            "MODEL.ROI_BOX_HEAD.POOLER_RESOLUTION": 14,
            "MODEL.ROI_MASK_HEAD.NAME": "MaskRCNNConvUpsampleHead",
            "MODEL.ROI_MASK_HEAD.NUM_CONV": 4,
            "MODEL.ROI_MASK_HEAD.POOLER_RESOLUTION": 7,
            "MODEL.RESNETS.DEPTH": 101,
            "MODEL.RESNETS.SIZES": [[32], [64], [128], [256], [512]],
            "MODEL.RESNETS.ASPECT_RATIOS": [[0.5, 1.0, 2.0]],
            "MODEL.RESNETS.OUT_FEATURES": ["res2", "res3", "res4", "res5"],
            "MODEL.RESNETS.NUM_GROUPS": 32,
            "MODEL.RESNETS.WIDTH_PER_GROUP": 8,
            "MODEL.RESNETS.STRIDE_IN_1X1": False,
        }

    def get_detectron2_config(self):
        detectron2_config = detectron2.config.get_cfg()
        for k, v in self.detectron2_config_args.items():
            attributes = k.split(".")
            to_set = detectron2_config
            for attribute in attributes[:-1]:
                to_set = getattr(to_set, attribute)
            setattr(to_set, attributes[-1], v)

        return detectron2_config


__all__ = ["LayoutLMv2Config"]