configuration_HunyuanViT.py

#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
#           This file was automatically generated from src/transformers/models/HunyuanViT/modular_HunyuanViT.py.
#               Do NOT edit this file manually as any edits will be overwritten by the generation of
#             the file from the modular. If any change should be done, please apply the change to the
#                          modular_HunyuanViT.py file directly. One of our CI enforces this.
#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# coding=utf-8
# Copyright 2025 The HuggingFace Inc. team.
#
# 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.

from transformers.configuration_utils import PretrainedConfig
from transformers.utils import logging


logger = logging.get_logger(__name__)

class HunyuanViTVisionConfig(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`HunyuanViTVisionModel`]. It is used to instantiate a
    HunyuanViT vision encoder according to the specified arguments, defining the model architecture. Instantiating a
    configuration with the defaults will yield a similar configuration to that of the vision encoder of the HunyuanViT
    [google/HunyuanViT-base-patch16-naflex](https://huggingface.co/google/HunyuanViT-base-patch16-naflex) architecture.

    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
    documentation from [`PretrainedConfig`] for more information.

    Args:
        hidden_size (`int`, *optional*, defaults to 768):
            Dimensionality of the encoder layers and the pooler layer.
        intermediate_size (`int`, *optional*, defaults to 3072):
            Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder.
        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.
        num_channels (`int`, *optional*, defaults to 3):
            Number of channels in the input images.
        num_patches (`int`, *optional*, defaults to 256):
            The number of patches in the image with the size of (`patch_size`, `patch_size`).
            The image is resized to fill maximum of this number of patches, and to preserve
            the aspect ratio. In case the resulted number of patches is lower, the image is
            padded in "patch" dimension.
        patch_size (`int`, *optional*, defaults to 16):
            The size (resolution) of each patch.
        hidden_act (`str` or `function`, *optional*, defaults to `"gelu_pytorch_tanh"`):
            The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`,
            `"relu"`, `"selu"` and `"gelu_new"` `"quick_gelu"` are supported.
        layer_norm_eps (`float`, *optional*, defaults to 1e-06):
            The epsilon used by the layer normalization layers.
        attention_dropout (`float`, *optional*, defaults to 0.0):
            The dropout ratio for the attention probabilities.

    Example:

    ```python
    >>> from transformers import HunyuanViTVisionConfig, HunyuanViTVisionModel

    >>> # Initializing a HunyuanViTVisionConfig with google/HunyuanViT-base-patch16-naflex style configuration
    >>> configuration = HunyuanViTVisionConfig()

    >>> # Initializing a HunyuanViTVisionModel (with random weights) from the google/HunyuanViT-base-patch16-naflex style configuration
    >>> model = HunyuanViTVisionModel(configuration)

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

    model_type = "HunyuanViT_vision_model"
    base_config_key = "vision_config"

    def __init__(
        self,
        hidden_size=768,
        intermediate_size=3072,
        num_hidden_layers=12,
        num_attention_heads=12,
        num_channels=3,
        num_patches=256,
        patch_size=16,
        hidden_act="gelu_pytorch_tanh",
        layer_norm_eps=1e-6,
        attention_dropout=0.0,
        **kwargs,
    ):
        super().__init__(**kwargs)

        self.hidden_size = hidden_size
        self.intermediate_size = intermediate_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.num_channels = num_channels
        self.patch_size = patch_size
        self.attention_dropout = attention_dropout
        self.layer_norm_eps = layer_norm_eps
        self.hidden_act = hidden_act
        self.num_patches = num_patches


class HunyuanViTConfig(PretrainedConfig):
    r"""
    [`HunyuanViTConfig`] is the configuration class to store the configuration of a [`HunyuanViTModel`]. It is used to
    instantiate a HunyuanViT model according to the specified arguments, defining the text model and vision model configs.
    Instantiating a configuration with the defaults will yield a similar configuration to that of the HunyuanViT
    [google/HunyuanViT-base-patch16-224](https://huggingface.co/google/HunyuanViT-base-patch16-224) architecture.

    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
    documentation from [`PretrainedConfig`] for more information.

    Args:
        text_config (`dict`, *optional*):
            Dictionary of configuration options used to initialize [`HunyuanViTTextConfig`].
        vision_config (`dict`, *optional*):
            Dictionary of configuration options used to initialize [`HunyuanViTVisionConfig`].
        kwargs (*optional*):
            Dictionary of keyword arguments.

    Example:

    ```python
    >>> from transformers import HunyuanViTConfig, HunyuanViTModel

    >>> # Initializing a HunyuanViTConfig with google/HunyuanViT-base-patch16-224 style configuration
    >>> configuration = HunyuanViTConfig()

    >>> # Initializing a HunyuanViTModel (with random weights) from the google/HunyuanViT-base-patch16-224 style configuration
    >>> model = HunyuanViTModel(configuration)

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

    >>> # We can also initialize a HunyuanViTConfig from a HunyuanViTTextConfig and a HunyuanViTVisionConfig
    >>> from transformers import HunyuanViTTextConfig, HunyuanViTVisionConfig

    >>> # Initializing a HunyuanViTText and HunyuanViTVision configuration
    >>> config_text = HunyuanViTTextConfig()
    >>> config_vision = HunyuanViTVisionConfig()

    >>> config = HunyuanViTConfig.from_text_vision_configs(config_text, config_vision)
    ```"""

    model_type = "HunyuanViT"
    sub_configs = {"vision_config": HunyuanViTVisionConfig}

    def __init__(self, text_config=None, vision_config=None, **kwargs):
        super().__init__(**kwargs)

        if vision_config is None:
            vision_config = {}
            logger.info("`vision_config` is `None`. initializing the `HunyuanViTVisionConfig` with default values.")

        self.vision_config = HunyuanViTVisionConfig(**vision_config)

        self.initializer_factor = 1.0

    @classmethod
    def from_text_vision_configs(cls, vision_config: HunyuanViTVisionConfig, **kwargs):
        r"""
        Instantiate a [`HunyuanViTConfig`] (or a derived class) from HunyuanViT text model configuration and HunyuanViT vision
        model configuration.

        Returns:
            [`HunyuanViTConfig`]: An instance of a configuration object
        """

        return cls(vision_config=vision_config.to_dict(), **kwargs)


__all__ = ["HunyuanViTConfig", "HunyuanViTVisionConfig"]