# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 # This file was automatically generated from src/transformers/models/glm4v_moe/modular_glm4v_moe.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_glm4v_moe.py file directly. One of our CI enforces this. # 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 # Copyright 2025 The ZhipuAI Inc. team and 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 ...configuration_utils import PreTrainedConfig from ...modeling_rope_utils import RopeParameters class Glm4vMoeTextConfig(PreTrainedConfig): r""" This is the configuration class to store the configuration of a [`Glm4vMoeModel`]. It is used to instantiate a GLM-4.5V model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of GLM-4.5V [zai-org/GLM-4.5V](https://huggingface.co/zai-org/GLM-4.5V). 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 151424): Vocabulary size of the Glm4vMoe model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`Glm4vMoeModel`] hidden_size (`int`, *optional*, defaults to 4096): Dimension of the hidden representations. intermediate_size (`int`, *optional*, defaults to 10944): Dimension of the MLP representations. num_hidden_layers (`int`, *optional*, defaults to 46): Number of hidden layers in the Transformer encoder. num_attention_heads (`int`, *optional*, defaults to 96): Number of attention heads for each attention layer in the Transformer encoder. num_key_value_heads (`int`, *optional*, defaults to 8): 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://huggingface.co/papers/2305.13245). If it is not specified, will default to `32`. 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 65536): 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-05): 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`. 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`, defaults to `True`, *optional*, defaults to `True`): 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. moe_intermediate_size (`int`, *optional*, defaults to 1408): Intermediate size of the routed expert. num_experts_per_tok (`int`, *optional*, defaults to 8): number of experts per token. n_shared_experts (`int`, *optional*, defaults to 1): Number of shared experts. n_routed_experts (`int`, *optional*, defaults to 128): Number of routed experts. routed_scaling_factor (`float`, *optional*, defaults to 1.0): Scaling factor or routed experts. n_group (`int`, *optional*, defaults to 1): Number of groups for routed experts. topk_group (`int`, *optional*, defaults to 1): Number of selected groups for each token(for each token, ensuring the selected experts is only within `topk_group` groups). first_k_dense_replace (`int`, *optional*, defaults to 1): Number of dense layers in shallow layers(embed->dense->dense->...->dense->moe->moe...->lm_head). \--k dense layers--/ norm_topk_prob (`bool`, *optional*, defaults to `True`): Whether to normalize the topk probabilities. pad_token_id (`int`, *optional*): Padding token id. eos_token_id (`int`, *optional*): End of stream token id. bos_token_id (`int`, *optional*): Beginning of stream token id. router_aux_loss_coef (`float`, *optional*, defaults to 0.0001): The aux loss factor for the loss. ```python >>> from transformers import Glm4vMoeTextModel, Glm4vMoeConfig >>> # Initializing a GLM-4.5V style configuration >>> configuration = Glm4vMoeConfig() >>> # Initializing a model from the GLM-4.5V style configuration >>> model = Glm4vMoeTextModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```""" model_type = "glm4v_moe_text" keys_to_ignore_at_inference = ["past_key_values"] # Default tensor parallel plan for base model `Glm4vMoe` 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", "layers.*.mlp.gate_proj": "colwise", "layers.*.mlp.up_proj": "colwise", "layers.*.mlp.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"]), } attribute_map = { "num_local_experts": "n_routed_experts", } base_config_key = "text_config" def __init__( self, vocab_size: int | None = 151424, hidden_size: int | None = 4096, intermediate_size: int | None = 10944, num_hidden_layers: int | None = 46, num_attention_heads: int | None = 96, num_key_value_heads: int | None = 8, hidden_act: str | None = "silu", max_position_embeddings: int | None = 65536, initializer_range: float | None = 0.02, rms_norm_eps: int | None = 1e-5, use_cache: bool | None = True, rope_parameters: RopeParameters | dict[str, RopeParameters] | None = None, attention_bias: bool | None = True, attention_dropout: float | None = 0.0, moe_intermediate_size: int | None = 1408, num_experts_per_tok: int | None = 8, n_shared_experts: int | None = 1, n_routed_experts: int | None = 128, routed_scaling_factor: float | None = 1.0, n_group: int | None = 1, topk_group: int | None = 1, first_k_dense_replace: int | None = 1, norm_topk_prob: bool | None = True, pad_token_id: int | None = None, eos_token_id: int | None = None, bos_token_id: int | None = None, router_aux_loss_coef: float | None = 0.0001, **kwargs, ): self.pad_token_id = pad_token_id self.eos_token_id = eos_token_id self.bos_token_id = bos_token_id 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 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.rope_parameters = rope_parameters kwargs.setdefault("partial_rotary_factor", 0.5) # assign default for BC # MoE arguments self.moe_intermediate_size = moe_intermediate_size self.num_experts_per_tok = num_experts_per_tok self.n_group = n_group self.topk_group = topk_group self.n_shared_experts = n_shared_experts self.n_routed_experts = n_routed_experts self.routed_scaling_factor = routed_scaling_factor self.first_k_dense_replace = first_k_dense_replace self.norm_topk_prob = norm_topk_prob self.router_aux_loss_coef = router_aux_loss_coef super().__init__(ignore_keys_at_rope_validation={"mrope_section"}, **kwargs) class Glm4vMoeVisionConfig(PreTrainedConfig): r""" This is the configuration class to store the configuration of a [`Glm4vMoeVisionModel`]. It is used to instantiate an Glm4vMoeVisionModel model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of GLM-4.1V-9B-Thinking [THUDM/GLM-4.1V-9B-Thinking](https://huggingface.co/THUDM/GLM-4.1V-9B-Thinking). Args: depth (`int`, *optional*, defaults to 24): Number of layers (depth) in the model. hidden_size (`int`, *optional*, defaults to 1536): Dimensionality of the encoder layers and the pooler layer. hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): The non-linear activation function (function or string) in the encoder and pooler. If string, `"gelu"`, `"relu"`, `"selu"` and `"gelu_new"` are supported. attention_bias (`bool`, *optional*, defaults to `False`): Whether to add a bias to the queries, keys and values. attention_dropout (`float`, *optional*, defaults to 0.0): Dropout probability for attention weights. num_heads (``, *optional*, defaults to 12): in_channels (``, *optional*, defaults to 3): image_size (`int` or `list[int]`, *optional*, defaults to 336): The size (resolution) of each image. patch_size (`int`, *optional*, defaults to 14): The size (resolution) of each patch. rms_norm_eps (`float`, *optional*, defaults to 1e-05): The epsilon used by the rms normalization layers. spatial_merge_size (`int`, *optional*, defaults to 2): The size used for merging spatial dimensions. temporal_patch_size (`int`, *optional*, defaults to 2): The size used for patches along the temporal dimension. out_hidden_size (`int`, *optional*, defaults to 4096): The output hidden size of the vision model. intermediate_size (`int`, *optional*, defaults to 13696): Dimensionality of the "intermediate" (i.e., feed-forward) layer in the Transformer encoder. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. Example: ```python >>> from transformers import Glm4vMoeVisionConfig, Glm4vMoeVisionModel >>> # Initializing a Glm4vMoeVisionConfig GLM-4.1V-9B style configuration >>> configuration = Glm4vMoeVisionConfig() >>> # Initializing a model (with random weights) from the GLM-4.1V-9B configuration >>> model = Glm4vMoeVisionModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```""" model_type = "glm4v_moe_vision" base_config_key = "vision_config" def __init__( self, depth=24, hidden_size=1536, hidden_act="silu", attention_bias=False, attention_dropout=0.0, num_heads=12, in_channels=3, image_size=336, patch_size=14, rms_norm_eps=1e-05, spatial_merge_size=2, temporal_patch_size=2, out_hidden_size=4096, intermediate_size=13696, initializer_range=0.02, **kwargs, ): super().__init__(**kwargs) self.depth = depth self.hidden_size = hidden_size self.hidden_act = hidden_act self.num_heads = num_heads self.in_channels = in_channels self.image_size = image_size 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.intermediate_size = intermediate_size self.initializer_range = initializer_range self.rms_norm_eps = rms_norm_eps self.attention_bias = attention_bias self.attention_dropout = attention_dropout class Glm4vMoeConfig(PreTrainedConfig): r""" This is the configuration class to store the configuration of a [`Glm4vMoeModel`]. It is used to instantiate a GLM-4.5V model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of GLM-4.5V [zai-org/GLM-4.5V](https://huggingface.co/zai-org/GLM-4.5V). 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 `Glm4vMoeTextConfig`): The config object or dictionary of the text backbone. vision_config (`Union[PreTrainedConfig, dict]`, *optional*, defaults to `Glm4vMoeVisionConfig`): The config object or dictionary of the vision backbone. image_token_id (`int`, *optional*, defaults to 151363): The image token index to encode the image prompt. video_token_id (`int`, *optional*, defaults to 151364): The video token index to encode the image prompt. image_start_token_id (`int`, *optional*, defaults to 151339): The image start token index to encode the start of image. image_end_token_id (`int`, *optional*, defaults to 151340): The image end token index to encode the end of image. video_start_token_id (`int`, *optional*, defaults to 151341): The video start token index to encode the start of video. video_end_token_id (`int`, *optional*, defaults to 151342): The video end token index to encode the end of video. tie_word_embeddings (`bool`, *optional*, defaults to `False`): Whether the model's input and output word embeddings should be tied. ```python >>> from transformers import Glm4vMoeForConditionalGeneration, Glm4vMoeConfig >>> # Initializing a GLM-4.5V style configuration >>> configuration = Glm4vMoeConfig() >>> # Initializing a model from the GLM-4.5V style configuration >>> model = Glm4vMoeForConditionalGeneration(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```""" model_type = "glm4v_moe" sub_configs = {"vision_config": Glm4vMoeVisionConfig, "text_config": Glm4vMoeTextConfig} keys_to_ignore_at_inference = ["past_key_values"] def __init__( self, text_config=None, vision_config=None, image_token_id=151363, video_token_id=151364, image_start_token_id=151339, image_end_token_id=151340, video_start_token_id=151341, video_end_token_id=151342, 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"](**kwargs) self.image_token_id = image_token_id self.video_token_id = video_token_id self.video_start_token_id = video_start_token_id self.video_end_token_id = video_end_token_id self.image_start_token_id = image_start_token_id self.image_end_token_id = image_end_token_id self.tie_word_embeddings = tie_word_embeddings super().__init__(**kwargs) __all__ = ["Glm4vMoeConfig", "Glm4vMoeVisionConfig", "Glm4vMoeTextConfig"]