configuration_mibot.py

# coding=utf-8
# Copyright (C) 2026 Xiaomi Corporation.
# Copyright 2025 The Qwen Team 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.
from typing import Optional

from transformers.configuration_utils import PretrainedConfig
from transformers.modeling_rope_utils import rope_config_validation


class Qwen3VLVisionConfig(PretrainedConfig):
    model_type = "qwen3_vl"
    base_config_key = "vision_config"

    def __init__(
        self,
        depth=27,
        hidden_size=1152,
        hidden_act="gelu_pytorch_tanh",
        intermediate_size=4304,
        num_heads=16,
        in_channels=3,
        patch_size=16,
        spatial_merge_size=2,
        temporal_patch_size=2,
        out_hidden_size=3584,
        num_position_embeddings=2304,
        deepstack_visual_indexes=[8, 16, 24],
        initializer_range=0.02,
        **kwargs,
    ):
        super().__init__(**kwargs)

        self.depth = depth
        self.hidden_size = hidden_size
        self.hidden_act = hidden_act
        self.intermediate_size = intermediate_size
        self.num_heads = num_heads
        self.in_channels = in_channels
        self.patch_size = patch_size
        self.spatial_merge_size = spatial_merge_size
        self.temporal_patch_size = temporal_patch_size
        self.out_hidden_size = out_hidden_size
        self.num_position_embeddings = num_position_embeddings
        self.initializer_range = initializer_range
        self.deepstack_visual_indexes = deepstack_visual_indexes


class Qwen3VLTextConfig(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`Qwen3VLTextModel`]. It is used to instantiate a
    Qwen3-VL model according to the specified arguments, defining the model architecture. Instantiating a configuration
    with the defaults will yield a similar configuration to that of
    Qwen3-VL-4B-Instruct [Qwen/Qwen3-VL-4B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct).

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

    Args:
        vocab_size (`int`, *optional*, defaults to 151936):
            Vocabulary size of the Qwen3VL model. Defines the number of different tokens that can be represented by the
            `inputs_ids` passed when calling [`Qwen3VLModel`]
        hidden_size (`int`, *optional*, defaults to 4096):
            Dimension of the hidden representations.
        intermediate_size (`int`, *optional*, defaults to 22016):
            Dimension of the MLP representations.
        num_hidden_layers (`int`, *optional*, defaults to 32):
            Number of hidden layers in the Transformer encoder.
        num_attention_heads (`int`, *optional*, defaults to 32):
            Number of attention heads for each attention layer in the Transformer encoder.
        num_key_value_heads (`int`, *optional*, defaults to 32):
            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
            `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When
            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
            by meanpooling all the original heads within that group. For more details, check out [this
            paper](https://huggingface.co/papers/2305.13245). If it is not specified, will default to `32`.
        head_dim (`int`, *optional*, defaults to 128):
            The dimension of the head. If not specified, will default to `hidden_size // num_attention_heads`.
        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
            The non-linear activation function (function or string) in the decoder.
        max_position_embeddings (`int`, *optional*, defaults to 128000):
            The maximum sequence length that this model might ever be used with.
        initializer_range (`float`, *optional*, defaults to 0.02):
            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
        rms_norm_eps (`float`, *optional*, defaults to 1e-06):
            The epsilon used by the rms normalization layers.
        use_cache (`bool`, *optional*, defaults to `True`):
            Whether or not the model should return the last key/values attentions (not used by all models). Only
            relevant if `config.is_decoder=True`.
        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
            Whether the model's input and output word embeddings should be tied.
        rope_theta (`float`, *optional*, defaults to 5000000.0):
            The base period of the RoPE embeddings.
        rope_scaling (`Dict`, *optional*):
            Dictionary containing the scaling configuration for the RoPE embeddings. NOTE: if you apply new rope type
            and you expect the model to work on longer `max_position_embeddings`, we recommend you to update this value
            accordingly.
            Expected contents:
                `rope_type` (`str`):
                    The sub-variant of RoPE to use. Can be one of ['default', 'linear', 'dynamic', 'yarn', 'longrope',
                    'llama3'], with 'default' being the original RoPE implementation.
                `factor` (`float`, *optional*):
                    Used with all rope types except 'default'. The scaling factor to apply to the RoPE embeddings. In
                    most scaling types, a `factor` of x will enable the model to handle sequences of length x *
                    original maximum pre-trained length.
                `original_max_position_embeddings` (`int`, *optional*):
                    Used with 'dynamic', 'longrope' and 'llama3'. The original max position embeddings used during
                    pretraining.
                `attention_factor` (`float`, *optional*):
                    Used with 'yarn' and 'longrope'. The scaling factor to be applied on the attention
                    computation. If unspecified, it defaults to value recommended by the implementation, using the
                    `factor` field to infer the suggested value.
                `beta_fast` (`float`, *optional*):
                    Only used with 'yarn'. Parameter to set the boundary for extrapolation (only) in the linear
                    ramp function. If unspecified, it defaults to 32.
                `beta_slow` (`float`, *optional*):
                    Only used with 'yarn'. Parameter to set the boundary for interpolation (only) in the linear
                    ramp function. If unspecified, it defaults to 1.
                `short_factor` (`list[float]`, *optional*):
                    Only used with 'longrope'. The scaling factor to be applied to short contexts (<
                    `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
                    size divided by the number of attention heads divided by 2
                `long_factor` (`list[float]`, *optional*):
                    Only used with 'longrope'. The scaling factor to be applied to long contexts (<
                    `original_max_position_embeddings`). Must be a list of numbers with the same length as the hidden
                    size divided by the number of attention heads divided by 2
                `low_freq_factor` (`float`, *optional*):
                    Only used with 'llama3'. Scaling factor applied to low frequency components of the RoPE
                `high_freq_factor` (`float`, *optional*):
                    Only used with 'llama3'. Scaling factor applied to high frequency components of the RoPE
        attention_bias (`bool`, defaults to `False`, *optional*, defaults to `False`):
            Whether to use a bias in the query, key, value and output projection layers during self-attention.
        attention_dropout (`float`, *optional*, defaults to 0.0):
            The dropout ratio for the attention probabilities.

    ```python
    >>> from transformers import Qwen3VLTextModel, Qwen3VLTextConfig

    >>> # Initializing a Qwen3VL style configuration
    >>> configuration = Qwen3VLTextConfig()

    >>> # Initializing a model from the Qwen3-VL-7B style configuration
    >>> model = Qwen3VLTextModel(configuration)

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

    model_type = "qwen3_vl_text"
    base_config_key = "text_config"

    def __init__(
        self,
        vocab_size=151936,
        hidden_size=4096,
        intermediate_size=22016,
        num_hidden_layers=32,
        num_attention_heads=32,
        num_key_value_heads=32,
        head_dim=128,
        hidden_act="silu",
        max_position_embeddings=128000,
        initializer_range=0.02,
        rms_norm_eps=1e-6,
        use_cache=True,
        tie_word_embeddings=False,
        rope_theta=5000000.0,
        rope_scaling=None,
        attention_bias=False,
        attention_dropout=0.0,
        **kwargs,
    ):
        self.vocab_size = vocab_size
        self.max_position_embeddings = max_position_embeddings
        self.hidden_size = hidden_size
        self.intermediate_size = intermediate_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads

        # for backward compatibility
        if num_key_value_heads is None:
            num_key_value_heads = num_attention_heads

        self.num_key_value_heads = num_key_value_heads
        self.head_dim = head_dim
        self.hidden_act = hidden_act
        self.initializer_range = initializer_range
        self.rms_norm_eps = rms_norm_eps
        self.use_cache = use_cache
        self.rope_theta = rope_theta
        self.rope_scaling = rope_scaling
        self.attention_bias = attention_bias
        self.attention_dropout = attention_dropout

        rope_config_validation(self, ignore_keys={"mrope_section", "mrope_interleaved"})

        super().__init__(tie_word_embeddings=tie_word_embeddings, **kwargs)


class Qwen3VLConfig(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`Qwen3VLModel`]. It is used to instantiate a
    Qwen3-VL model according to the specified arguments, defining the model architecture. Instantiating a configuration
    with the defaults will yield a similar configuration to that of
    Qwen3-VL-4B-Instruct [Qwen/Qwen3-VL-4B-Instruct](https://huggingface.co/Qwen/Qwen3-VL-4B-Instruct).

    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 (`Union[PretrainedConfig, dict]`, *optional*, defaults to `Qwen3VLTextConfig`):
            The config object or dictionary of the text backbone.
        vision_config (`Union[PretrainedConfig, dict]`,  *optional*, defaults to `Qwen3VLVisionConfig`):
            The config object or dictionary of the vision backbone.
        image_token_id (`int`, *optional*, defaults to 151655):
            The image token index to encode the image prompt.
        video_token_id (`int`, *optional*, defaults to 151656):
            The video token index to encode the image prompt.
        vision_start_token_id (`int`, *optional*, defaults to 151652):
            The start token index to encode the image prompt.
        vision_end_token_id (`int`, *optional*, defaults to 151653):
            The end token index to encode the image prompt.
        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
            Whether to tie the word embeddings.

    ```python
    >>> from transformers import Qwen3VLForConditionalGeneration, Qwen3VLConfig

    >>> # Initializing a Qwen3-VL style configuration
    >>> configuration = Qwen3VLConfig()

    >>> # Initializing a model from the Qwen3-VL-4B style configuration
    >>> model = Qwen3VLForConditionalGeneration(configuration)

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

    model_type = "qwen3_vl"
    sub_configs = {"vision_config": Qwen3VLVisionConfig, "text_config": Qwen3VLTextConfig}
    keys_to_ignore_at_inference = ["past_key_values"]

    def __init__(
        self,
        text_config=None,
        vision_config=None,
        image_token_id=151655,
        video_token_id=151656,
        vision_start_token_id=151652,
        vision_end_token_id=151653,
        tie_word_embeddings=False,
        **kwargs,
    ):
        if isinstance(vision_config, dict):
            self.vision_config = self.sub_configs["vision_config"](**vision_config)
        elif vision_config is None:
            self.vision_config = self.sub_configs["vision_config"]()

        if isinstance(text_config, dict):
            self.text_config = self.sub_configs["text_config"](**text_config)
        elif text_config is None:
            self.text_config = self.sub_configs["text_config"]()

        self.image_token_id = image_token_id
        self.video_token_id = video_token_id
        self.vision_start_token_id = vision_start_token_id
        self.vision_end_token_id = vision_end_token_id
        super().__init__(**kwargs, tie_word_embeddings=tie_word_embeddings)


class DiTConfig(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`DiT`] model. It is used to instantiate a DiT
    model according to the specified arguments, defining the model architecture.

    Args:
        hidden_size (`int`, *optional*, defaults to 768):
            Dimension of the hidden representations.
        num_hidden_layers (`int`, *optional*, defaults to 8):
            Number of decoder layers in the DiT.
        head_dim (`int`, *optional*, defaults to 128):
            Dimension of each attention head.
        num_key_value_heads (`int`, *optional*, defaults to 2):
            Number of key-value heads for grouped query attention.
        is_causal (`bool`, *optional*, defaults to `True`):
            Whether the model uses causal (autoregressive) attention.
        attention_dropout (`float`, *optional*, defaults to 0.0):
            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.
    """

    model_type = "dit"

    def __init__(
        self,
        hidden_size: int = 1024,
        num_hidden_layers: int = 16,
        head_dim: int = 128,
        num_key_value_heads: int = 8,
        **kwargs,
    ):
        super().__init__(**kwargs)
        self.hidden_size = hidden_size
        self.num_hidden_layers = num_hidden_layers
        self.head_dim = head_dim
        self.num_key_value_heads = num_key_value_heads

class MiBoTConfig(PretrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`MiBoTModel`]. It is used to instantiate a MiBoT
    model according to the specified arguments, defining the model architecture. The MiBoT model consists of a
    vision-language model (Qwen3-VL) and a policy head (DiT).

    Args:
        vlm_config (`Union[PretrainedConfig, dict]`, *optional*):
            The config object or dictionary for the vision-language model (Qwen3-VL).
        dit_config (`Union[PretrainedConfig, dict]`, *optional*):
            The config object or dictionary for the DiT policy head.
        state_length (`int`, *optional*, defaults to 1):
            Number of historical states to include as input.
        state_dim (`int`, *optional*, defaults to 16):
            Dimension of each state vector.
        action_dim (`int`, *optional*, defaults to 32):
            Dimension of each action vector.
        **kwargs: Additional keyword arguments passed to PretrainedConfig.
    """

    model_type = "mibot"
    sub_configs = {
        "vlm_config": Qwen3VLConfig,
        "dit_config": DiTConfig,
    }
    keys_to_ignore_at_inference = ["past_key_values"]

    def __init__(
        self,
        vlm_config: Optional[dict] = None,
        dit_config: Optional[dict] = None,
        state_length: int = 1,
        state_dim: int = 32,
        action_dim: int = 32,
        **kwargs,
    ):
        if isinstance(vlm_config, dict):
            self.vlm_config = self.sub_configs["vlm_config"](**vlm_config)
        else:
            self.vlm_config = self.sub_configs["vlm_config"]()

        if isinstance(dit_config, dict):
            self.dit_config = self.sub_configs["dit_config"](**dit_config)
        else:
            self.dit_config = self.sub_configs["dit_config"]()

        self.state_length = state_length
        self.state_dim = state_dim
        self.action_dim = action_dim

        super().__init__(**kwargs)

__all__ = ["MiBoTConfig", "DiTConfig"]