# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 # This file was automatically generated from src/transformers/models/solar_open/modular_solar_open.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_solar_open.py file directly. One of our CI enforces this. # 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 # Copyright 2026 Upstage and 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 ...configuration_utils import PreTrainedConfig from ...modeling_rope_utils import RopeParameters class SolarOpenConfig(PreTrainedConfig): r""" This is the configuration class to store the configuration of a [`SolarOpenModel`]. It is used to instantiate a SolarOpen model according to the specified arguments, defining the model architecture. Configuration objects inherit from [`PreTrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PreTrainedConfig`] for more information. Instantiating a configuration defaults will yield a similar configuration to that of [upstage/Solar-Open-100B](https://huggingface.co/upstage/Solar-Open-100B) architecture. Args: vocab_size (`int`, *optional*, defaults to 196608): Vocabulary size of the SolarOpen model. hidden_size (`int`, *optional*, defaults to 4096): Dimension of the hidden representations. moe_intermediate_size (`int`, *optional*, defaults to 1280): Intermediate size of the routed expert. num_hidden_layers (`int`, *optional*, defaults to 48): Number of hidden layers in the Transformer encoder. num_attention_heads (`int`, *optional*, defaults to 64): Number of attention heads for each attention layer. num_key_value_heads (`int`, *optional*, defaults to 8): Number of key_value heads for Grouped Query Attention. n_shared_experts (`int`, *optional*, defaults to 1): Number of shared experts. n_routed_experts (`int`, *optional*, defaults to 128): Number of routed experts. head_dim (`int`, *optional*, defaults to 128): Dimension of each attention head. hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): The non-linear activation function in the decoder. max_position_embeddings (`int`, *optional*, defaults to 131072): 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. 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 to return the last key/values attentions. 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 projection layers. attention_dropout (`float`, *optional*, defaults to 0.0): The dropout ratio for the attention probabilities. num_experts_per_tok (`int`, *optional*, defaults to 8): Number of experts per token. routed_scaling_factor (`float`, *optional*, defaults to 1.0): Scaling factor for 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. norm_topk_prob (`bool`, *optional*, defaults to `True`): Whether to normalize the topk probabilities. bos_token_id (`int`, *optional*): Beginning of stream token id. eos_token_id (`int`, *optional*): End of stream token id. pad_token_id (`int`, *optional*): Padding token id. """ model_type = "solar_open" keys_to_ignore_at_inference = ["past_key_values"] # Default tensor parallel plan for base model `SolarOpenModel` 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.experts.gate_up_proj": "packed_colwise", "layers.*.mlp.experts.down_proj": "rowwise", "layers.*.mlp.experts": "moe_tp_experts", } 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", } default_theta = 1_000_000.0 def __init__( self, vocab_size: int = 196608, hidden_size: int = 4096, moe_intermediate_size: int = 1280, num_hidden_layers: int = 48, num_attention_heads: int = 64, num_key_value_heads: int = 8, n_shared_experts: int = 1, n_routed_experts: int = 128, head_dim: int = 128, hidden_act: str = "silu", max_position_embeddings: int = 131072, initializer_range: float = 0.02, rms_norm_eps: int = 1e-5, use_cache: bool = True, tie_word_embeddings: bool = False, rope_parameters: RopeParameters | None = None, attention_bias: bool = False, attention_dropout: float = 0.0, num_experts_per_tok: int = 8, routed_scaling_factor: float = 1.0, n_group: int = 1, topk_group: int = 1, norm_topk_prob: bool = True, bos_token_id: int | None = None, eos_token_id: int | None = None, pad_token_id: int | None = None, **kwargs, ): # Default partial_rotary_factor to 1.0 (instead of 0.5 in Glm4MoeConfig). # `setdefault` ensures this value is not overridden by subsequent calls. # This workaround is required due to modular inheritance limitations. kwargs.setdefault("partial_rotary_factor", 1.0) self.head_dim = head_dim 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.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.norm_topk_prob = norm_topk_prob self.tie_word_embeddings = tie_word_embeddings self.bos_token_id = bos_token_id self.eos_token_id = eos_token_id self.pad_token_id = pad_token_id super().__init__(**kwargs) __all__ = ["SolarOpenConfig"]