configuration_interns2_mobius.py

# Copyright 2025 InternS2Mobius 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 transformers.configuration_utils import PreTrainedConfig, layer_type_validation
from transformers.modeling_rope_utils import RopeParameters


class InternS2MobiusVisionConfig(PreTrainedConfig):
    model_type = "interns2_mobius"
    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,
        initializer_range=0.02,
        deepstack_visual_indexes=None,
        **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 if deepstack_visual_indexes is not None else []


class InternS2MobiusTextConfig(PreTrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`InternS2MobiusTextModel`]. It is used to instantiate a
    InternS2Mobius model according to the specified arguments, defining the model architecture.
    Instantiating a configuration with the defaults will yield a similar configuration to that of
    InternS2Mobius.

    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 248320):
            Vocabulary size of the model. Defines the number of different tokens that can be represented by the
            `inputs_ids`.
        hidden_size (`int`, *optional*, defaults to 2048):
            Dimension of the hidden representations.
        num_hidden_layers (`int`, *optional*, defaults to 40):
            Number of hidden layers in the Transformer encoder.
        num_attention_heads (`int`, *optional*, defaults to 16):
            Number of attention heads for each attention layer in the Transformer encoder.
        num_key_value_heads (`int`, *optional*, defaults to 2):
            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 checkout [this
            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `32`.
        hidden_act (`str`, *optional*, defaults to `"silu"`):
            The non-linear activation function in the decoder.
        max_position_embeddings (`int`, *optional*, defaults to 32768):
            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_parameters (`RopeParameters`, *optional*):
            Dictionary containing the configuration parameters for the RoPE embeddings. The dictionary should contain
            a value for `rope_theta` and optionally parameters used for scaling in case you want to use RoPE
            with longer `max_position_embeddings`.
        attention_bias (`bool`, *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.
        head_dim (`int`, *optional*, defaults to 256):
            Projection weights dimension in multi-head attention.
        linear_conv_kernel_dim (`int`, *optional*, defaults to 4):
            Kernel size of the convolution used in linear attention layers.
        linear_key_head_dim (`int`, *optional*, defaults to 128):
            Dimension of each key head in linear attention.
        linear_value_head_dim (`int`, *optional*, defaults to 128):
            Dimension of each value head in linear attention.
        linear_num_key_heads (`int`, *optional*, defaults to 16):
            Number of key heads used in linear attention layers.
        linear_num_value_heads (`int`, *optional*, defaults to 32):
            Number of value heads used in linear attention layers.
        moe_intermediate_size (`int`, *optional*, defaults to 512):
            Intermediate size of the routed expert.
        shared_expert_intermediate_size (`int`, *optional*, defaults to 512):
            Intermediate size of the shared expert.
        num_experts_per_tok (`int`, *optional*, defaults to 8):
            Number of selected experts.
        num_experts (`int`, *optional*, defaults to 256):
            Number of routed experts.
        output_router_logits (`bool`, *optional*, defaults to `False`):
            Whether or not the router logits should be returned by the model. Enabling this will also
            allow the model to output the auxiliary loss, including load balancing loss and router z-loss.
        router_aux_loss_coef (`float`, *optional*, defaults to 0.001):
            The aux loss factor for the total loss.
        layer_types (`list[str]`, *optional*):
            Types of each layer (attention or linear).
        pad_token_id (`int`, *optional*):
            Padding token id.
        bos_token_id (`int`, *optional*):
            Beginning of stream token id.
        eos_token_id (`int`, *optional*):
            End of stream token id.

    ```python
    >>> from transformers import InternS2MobiusTextModel, InternS2MobiusTextConfig

    >>> # Initializing a InternS2Mobius style configuration
    >>> configuration =  InternS2MobiusTextConfig()

    >>> # Initializing a model from the InternS2Mobius style configuration
    >>> model = InternS2MobiusTextModel(configuration)

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

    # NOTE: `model_type` is kept as `interns2_mobius_text` to match config.json.
    model_type = "interns2_mobius_text"
    keys_to_ignore_at_inference = ["past_key_values"]

    base_model_tp_plan = {
        "layers.*.self_attn.q_proj": "colwise",
        "layers.*.self_attn.k_proj": "colwise",
        "layers.*.self_attn.v_proj": "colwise",
        "layers.*.self_attn.o_proj": "rowwise",
        "meta_mlp.*.experts.gate_up_proj": "packed_colwise",
        "meta_mlp.*.experts.down_proj": "rowwise",
        "layers.*.mlp.shared_expert.gate_proj": "colwise",
        "layers.*.mlp.shared_expert.up_proj": "colwise",
        "layers.*.mlp.shared_expert.down_proj": "rowwise",
    }
    base_model_pp_plan = {
        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
        "norm": (["hidden_states"], ["hidden_states"]),
    }
    base_config_key = "text_config"

    def __init__(
        self,
        vocab_size=248320,
        hidden_size=2048,
        num_hidden_layers=40,
        num_attention_heads=16,
        num_key_value_heads=2,
        hidden_act="silu",
        max_position_embeddings=32768,
        initializer_range=0.02,
        rms_norm_eps=1e-6,
        use_cache=True,
        tie_word_embeddings=False,
        rope_parameters: RopeParameters | dict[str, RopeParameters] | None = None,
        attention_bias=False,
        attention_dropout=0.0,
        head_dim=256,
        linear_conv_kernel_dim=4,
        linear_key_head_dim=128,
        linear_value_head_dim=128,
        linear_num_key_heads=16,
        linear_num_value_heads=32,
        moe_intermediate_size=512,
        shared_expert_intermediate_size=512,
        num_experts_per_tok=8,
        num_experts=256,
        output_router_logits=False,
        router_aux_loss_coef=0.001,
        num_blocks=4,
        dtype=None,
        layer_types=None,
        pad_token_id: int | None = None,
        bos_token_id: int | None = None,
        eos_token_id: int | None = None,
        **kwargs,
    ):
        kwargs["ignore_keys_at_rope_validation"] = {"mrope_section", "mrope_interleaved"}
        self.num_blocks = num_blocks
        self.dtype = dtype
        self.pad_token_id = pad_token_id
        self.bos_token_id = bos_token_id
        self.eos_token_id = eos_token_id
        self.tie_word_embeddings = tie_word_embeddings
        self.vocab_size = vocab_size
        self.max_position_embeddings = max_position_embeddings
        self.hidden_size = hidden_size
        self.num_hidden_layers = num_hidden_layers
        self.num_attention_heads = num_attention_heads
        self.num_key_value_heads = num_key_value_heads
        self.hidden_act = hidden_act
        self.initializer_range = initializer_range
        self.rms_norm_eps = rms_norm_eps
        self.use_cache = use_cache
        self.attention_bias = attention_bias
        self.attention_dropout = attention_dropout
        self.head_dim = head_dim
        self.rope_parameters = rope_parameters
        kwargs.setdefault("partial_rotary_factor", 0.25)  # assign default for BC

        self.layer_types = layer_types
        if self.layer_types is None:
            interval_pattern = kwargs.get("full_attention_interval", 4)
            self.layer_types = [
                "linear_attention" if bool((i + 1) % interval_pattern) else "full_attention"
                for i in range(self.num_hidden_layers)
            ]
        layer_type_validation(self.layer_types, self.num_hidden_layers)

        # linear attention part
        self.linear_conv_kernel_dim = linear_conv_kernel_dim
        self.linear_key_head_dim = linear_key_head_dim
        self.linear_value_head_dim = linear_value_head_dim
        self.linear_num_key_heads = linear_num_key_heads
        self.linear_num_value_heads = linear_num_value_heads
        self.moe_intermediate_size = moe_intermediate_size
        self.shared_expert_intermediate_size = shared_expert_intermediate_size
        self.num_experts_per_tok = num_experts_per_tok
        self.num_experts = num_experts
        self.output_router_logits = output_router_logits
        self.router_aux_loss_coef = router_aux_loss_coef
        super().__init__(**kwargs)


class InternS2MobiusConfig(PreTrainedConfig):
    r"""
    This is the configuration class to store the configuration of a [`InternS2MobiusModel`]. It is used to instantiate a
    InternS2Mobius model according to the specified arguments, defining the model architecture. Instantiating a configuration
    with the defaults will yield a similar configuration to that of
    InternS2Mobius.

    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 `InternS2MobiusTextConfig`):
            The config object or dictionary of the text backbone.
        vision_config (`Union[PreTrainedConfig, dict]`,  *optional*, defaults to `InternS2MobiusVisionConfig`):
            The config object or dictionary of the vision backbone.
        image_token_id (`int`, *optional*, defaults to 248056):
            The image token index to encode the image prompt.
        video_token_id (`int`, *optional*, defaults to 248057):
            The video token index to encode the image prompt.
        vision_start_token_id (`int`, *optional*, defaults to 248053):
            The start token index to encode the image prompt.
        vision_end_token_id (`int`, *optional*, defaults to 248054):
            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 InternS2MobiusForConditionalGeneration, InternS2MobiusConfig

    >>> # Initializing a InternS2Mobius style configuration
    >>> configuration = InternS2MobiusConfig()

    >>> # Initializing a model from the InternS2Mobius style configuration
    >>> model = InternS2MobiusForConditionalGeneration(configuration)

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

    model_type = "interns2_mobius"
    sub_configs = {
        "vision_config": InternS2MobiusVisionConfig,
        "text_config": InternS2MobiusTextConfig,
    }
    keys_to_ignore_at_inference = ["past_key_values"]

    def __init__(
        self,
        text_config=None,
        vision_config=None,
        image_token_id=248056,
        video_token_id=248057,
        vision_start_token_id=248053,
        vision_end_token_id=248054,
        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
        self.tie_word_embeddings = tie_word_embeddings
        super().__init__(**kwargs)
        self.auto_map = {
            "AutoConfig": "configuration_interns2_mobius.InternS2MobiusConfig",
            "AutoModelForCausalLM": "modeling_interns2_mobius.InternS2MobiusForCausalLM",
            "AutoModel": "modeling_interns2_mobius.InternS2MobiusModel",
            "AutoModelForImageTextToText": "modeling_interns2_mobius.InternS2MobiusForConditionalGeneration",
            "AutoModelForMultimodalLM": "modeling_interns2_mobius.InternS2MobiusForConditionalGeneration",
        }
        self.architectures = ["InternS2MobiusForConditionalGeneration"]


__all__ = ["InternS2MobiusConfig", "InternS2MobiusTextConfig"]