qi*ddlZddlmZddlmZddlmZddlmZddl m Z dd l m Z dd l mZdd lmZmZmZmZmZmZdd lmZe j0eZGd de ZGddeZGddeZGddeZGddeZGddeZ GddeeZ!GddeZ"GddeZ#gdZ$y) N)nn)initialization)RopeParameters)PreTrainedModel)logging)DeepseekV3Config)DeepseekV3Attention)LlamaDecoderLayerLlamaForCausalLM LlamaModelLlamaPreTrainedModel LlamaRMSNormLlamaRotaryEmbedding)Qwen3MLPc4eZdZdZdZddddZiZ d#dedzdedzd edzd edzd edzd edzd edzdedzdedzdedzdedzdedzdedzde dzde dzdedzde dzdedzdedzdedzde dzde e ee fzde dzde dzde dzf2fd Z d$d!edzfd"ZxZS)% YoutuConfiga This is the configuration class to store the configuration of a [`YoutuModel`]. It is used to instantiate an Youtu model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the Youtu-LLM-2B. e.g. [tencent/Youtu-LLM-2B](https://huggingface.co/tencent/Youtu-LLM-2B) 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 128256): Vocabulary size of the Deep model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`YoutuModel`] hidden_size (`int`, *optional*, defaults to 2048): Dimension of the hidden representations. intermediate_size (`int`, *optional*, defaults to 6144): Dimension of the MLP representations. num_hidden_layers (`int`, *optional*, defaults to 32): Number of hidden layers in the Transformer decoder. num_attention_heads (`int`, *optional*, defaults to 16): Number of attention heads for each attention layer in the Transformer decoder. num_key_value_heads (`int`, *optional*, defaults to 16): In MLA, num_key_value_heads=num_attention_heads. kv_lora_rank (`int`, *optional*, defaults to 512): Rank of the LoRA matrices for key and value projections. q_lora_rank (`int`, *optional*, defaults to 1536): Rank of the LoRA matrices for query projections. qk_rope_head_dim (`int`, *optional*, defaults to 64): Dimension of the query/key heads that use rotary position embeddings. v_head_dim (`int`, *optional*, defaults to 128): Dimension of the value heads. qk_nope_head_dim (`int`, *optional*, defaults to 128): Dimension of the query/key heads that don't use rotary position embeddings. 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 131072): The maximum sequence length that this model might ever be used with. initializer_range (`float`, *optional*): The standard deviation of the truncated_normal_initializer for initializing all weight matrices, except embedding matrices. embedding_initializer_range (`float`, *optional*): The standard deviation of the truncated_normal_initializer for initializing all embedding 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`. pad_token_id (`int`, *optional*): Padding token id. bos_token_id (`int`, *optional*, defaults to 128000): Beginning of stream token id. eos_token_id (`int`, *optional*, defaults to 128001): End of stream token id. tie_word_embeddings (`bool`, *optional*, defaults to `True`): Whether to tie weight embeddings 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`. rope_interleave (`bool`, *optional*, defaults to `True`): Whether to interleave the rotary position embeddings. 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 YoutuModel, YoutuConfig >>> # Initializing a Youtu-LLM-2B style configuration >>> configuration = YoutuConfig() >>> # Accessing the model configuration >>> configuration = model.config ```youtucolwiserowwise)zlayers.*.mlp.gate_projzlayers.*.mlp.up_projzlayers.*.mlp.down_projN vocab_size hidden_sizeintermediate_sizenum_hidden_layersnum_attention_headsnum_key_value_heads kv_lora_rank q_lora_rankqk_rope_head_dim v_head_dimqk_nope_head_dim hidden_actmax_position_embeddingsinitializer_rangeembedding_initializer_range rms_norm_eps use_cache pad_token_id bos_token_id eos_token_idtie_word_embeddingsrope_parametersrope_interleaveattention_biasattention_dropoutc t|did|d|d|d|d|d|d|d|d | d | d | d | d | d|d|d|d|d|d|d|d|d|d|||`|`|`|`|`|`|`|` |` |` |j1|jdk7rdd|jzdzz |_ nd|_ |d|jz|_y||_y)Nrrrrrrrrr r!r"r#r$r'r(r)r*r+r,r-r.r/r0rg@g@g?{Gz?)super__init__n_shared_expertsn_routed_expertsrouted_scaling_factorn_group topk_groupnum_experts_per_tokfirst_k_dense_replacenorm_topk_probpretraining_tpmoe_intermediate_sizer%rr&)selfrrrrrrrrr r!r"r#r$r%r&r'r(r)r*r+r,r-r.r/r0kwargs __class__s Y/opt/pipecat/venv/lib/python3.12/site-packages/transformers/models/youtu/modular_youtu.pyr5zYoutuConfig.__init__|s:  ! # 0 0  !4  !4  & $ . " . " %< &   &! "&# $&% &!4' (,) *,+ ,*- .01 8  !  !  & L O  $  &      &  ! ! )1$),d6F6F0F3/N)N&)-& ' ./2T5K5K/KD ,/JD ,ignore_keys_at_rope_validationc td)Nz$Not overwritten for the Youtu model!)AttributeError)r@rErAs rCconvert_rope_params_to_dictz'YoutuConfig.convert_rope_params_to_dictsCDDrD)iii rJii@rLsiluiNNgư>TNiiTNTF)N)__name__ __module__ __qualname____doc__ model_typebase_model_tp_plan attribute_mapintstrfloatboolrdictr5setrH __classcell__)rBs@rCrr+s*F PJ"+ )"+ M"("&(,(**,*,#&"&')!$'*!'.4*.48#'!%#'#)#)+/FJ'+&+*-5OK$JOK4ZOK: OK : OK !4Z OK!4ZOKDjOK4ZOK*OK$JOK*OK$JOK"%tOK!4<OK &+T\!OK"Dj#OK$$;%OK&Dj'OK(Dj)OK*Dj+OK,"D[-OK.($sN/B*CC/OK01OK2t 3OK4!4<5OKbE#PT*ErDrc eZdZy) YoutuRMSNormNrOrPrQr3rDrCr^r^rDr^c eZdZy)YoutuRotaryEmbeddingNr_r3rDrCrbrbr`rDrbc eZdZy)YoutuMLPNr_r3rDrCrdrdr`rDrdc eZdZy)YoutuAttentionNr_r3rDrCrfrfr`rDrfc eZdZy)YoutuDecoderLayerNr_r3rDrCrhrhr`rDrhc:eZdZejdZy)YoutuPreTrainedModelctj||t|jdd}t|jdd|z}t |t j rftj|jd||j7tj|jj|jyyy)Nr%r2r&r rN)meanstd) r _init_weightsgetattrconfig isinstancer Embeddinginitnormal_weight padding_idxzeros_data)r@modulerm embed_stds rCrnz"YoutuPreTrainedModel._init_weightss%%dF3dkk#6=DKK)FCP fbll + LLSi @!!- FMM..v/A/ABC. ,rDN)rOrPrQtorchno_gradrnr3rDrCrjrjsU]]_DDrDrjc eZdZy) YoutuModelNr_r3rDrCr~r~r`rDr~c eZdZy)YoutuForCausalLMNr_r3rDrCrrr`rDr)rrjr~r)%r{rrrsmodeling_rope_utilsrmodeling_utilsrutilsr%deepseek_v3.configuration_deepseek_v3r deepseek_v3.modeling_deepseek_v3r llama.modeling_llamar r rrrrqwen3.modeling_qwen3r get_loggerrOloggerrr^rbrdrfrhrjr~r__all__r3rDrCrs( &1-DB,   H %cE"cEL <  /  x  (  )  D/ D   '  rD