.`i(^dZddlmZddlmZejeZGddeZdS)zJAIS configuration)PretrainedConfig)loggingcteZdZdZdZdgZdddddZ dfd ZdZxZ S) JAISConfigaw This is the configuration class to store the configuration of a [`JAISModel`]. It is used to instantiate a JAIS 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. Args: vocab_size (`int`, *optional*, defaults to 50257): Vocabulary size of the JAIS model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`JAISModel`]. n_positions (`int`, *optional*, defaults to 1024): The maximum sequence length that this model might ever be used with. Typically set this to something large just in case (e.g., 512 or 1024 or 2048). n_embd (`int`, *optional*, defaults to 768): Dimensionality of the embeddings and hidden states. n_layer (`int`, *optional*, defaults to 12): Number of hidden layers in the Transformer encoder. n_head (`int`, *optional*, defaults to 12): Number of attention heads for each attention layer in the Transformer encoder. n_inner (`int`, *optional*, defaults to None): Dimensionality of the inner feed-forward layers. `None` will set it to 4 times n_embd activation_function (`str`, *optional*, defaults to `"gelu"`): Activation function, to be selected in the list `["relu", "silu", "gelu", "tanh", "gelu_new", "swiglu"]`. resid_pdrop (`float`, *optional*, defaults to 0.1): The dropout probability for all fully connected layers in the embeddings, encoder, and pooler. embd_pdrop (`float`, *optional*, defaults to 0.1): The dropout ratio for the embeddings. attn_pdrop (`float`, *optional*, defaults to 0.1): The dropout ratio for the attention. layer_norm_epsilon (`float`, *optional*, defaults to 1e-5): The epsilon to use in the layer normalization layers. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. scale_attn_weights (`bool`, *optional*, defaults to `True`): Scale attention weights by dividing by sqrt(hidden_size).. use_cache (`bool`, *optional*, defaults to `True`): Whether or not the model should return the last key/values attentions (not used by all models). scale_attn_by_inverse_layer_idx (`bool`, *optional*, default `True`): Whether to additionally scale attention weights by `1 / layer_idx + 1`. reorder_and_upcast_attn (`bool`, *optional*, defaults to `False`): Whether to scale keys (K) prior to computing attention (dot-product) and upcast attention dot-product/softmax to float() when training with mixed precision. position_embedding_type (`str`, *optional*, defaults to `"learned"`): Positional embedding can be either `"alibi"` or `"learned"`. mup_width_scale (`float`, *optional*, defaults to 1.0): muP parameter to scale learning rate and initializers. Calculated as (`d_model,0 / d_model`), where `d_model` is the model's width and `d_model,0` is the proxy model's width. mup_embeddings_scale (`float`, *optional*, defaults to 1.0): muP parameter to scale token and position embeddings. mup_output_alpha (`float`, *optional*, defaults to 1.0): muP parameter to scale output logits (`output_logits_scale = mup_output_alpha * mup_width_scale`). mup_scale_qk_dot_by_d (`bool`, *optional*, defaults to `False`): Scale attention weights by dividing by hidden_size instead of sqrt(hidden_size). Need to set scale_attn_weights to `True` as well. alibi_scaling (`dict`, *optional*): Dictionary containing the scaling configuration for ALiBi embeddings. Currently only supports linear scaling strategy. Can specify either the scaling `factor` (must be a float greater than 1) for fixed scaling or `train_seq_len` for dynamic scaling on input samples with sequence length > `train_seq_len`. The expected formats are `{"type": strategy name, "factor": scaling factor}` or `{"type": strategy name, "train_seq_len": training sequence length}`. architectures (`list`, *optional*, defaults to ['JAISLMHeadModel']): architecture names for Jais. Example: ```python >>> from transformers import JAISConfig, JAISModel >>> # Initializing a JAIS configuration >>> configuration = JAISConfig() >>> # Initializing a model (with random weights) from the configuration >>> model = JAISModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ```jaispast_key_valuesn_embd n_positionsn_headn_layer) hidden_sizemax_position_embeddingsnum_attention_headsnum_hidden_layersQ Ngelu_new皙?h㈵>{Gz?TPFlearned?c ||_||_||_||_||_||_||_||_| |_| |_ | |_ | |_ | |_ ||_ ||_||_||_||_||_||_||_||_||_||_||dg}t3jd|||d|dS)NJAISLMHeadModel) bos_token_id eos_token_id architectures) vocab_sizer r r r n_inneractivation_function resid_pdrop embd_pdrop attn_pdroplayer_norm_epsiloninitializer_rangescale_attn_weights use_cachescale_attn_by_inverse_layer_idxreorder_and_upcast_attnrrposition_embedding_typemup_width_scalemup_embeddings_scalemup_output_alphamup_scale_qk_dot_by_d alibi_scaling_alibi_scaling_validationsuper__init__)selfr"r r r r r#r$r%r&r'r(r)r*r+rrr,r-r.r/r0r1r2r3r kwargs __class__s x/home/jaya/work/projects/VOICE-AGENT/VIET/agent-env/lib/python3.11/site-packages/vllm/transformers_utils/configs/jais.pyr6zJAISConfig.__init__s:%&    #6 &$$"4!2"4"/N,'>$(('>$.$8! 0%:"* &&(((  ./M %%'        cH|jdSt|jtrt|jdkrt d|j|jdd}|jdd}|jdd}||dkrt d||t|t r||d krt d ||t|tr||d krt d |dSdS) z= Validate the `alibi_scaling` configuration. Nzm`alibi_scaling` must be a dictionary with two fields, `type` and `factor` or `type` and `train_seq_len`, got typefactor train_seq_lenlinearz3`alibi_scaling`'s type field must be 'linear', got rz:`alibi_scaling`'s factor field must be a float > 1.0, got zD`alibi_scaling`'s `train_seq_len` field must be an integer > 1, got )r3 isinstancedictlen ValueErrorgetfloatint)r7alibi_scaling_typealibi_scaling_factoralibi_dynamic_scalings r:r4z$JAISConfig._alibi_scaling_validations   % F$,d33 s4;M7N7NRS7S7S,),,  "/33FDAA#155hEE $ 2 6 6 M M  %);x)G)G,),,  ! ,3U;; -$05IS5P5P.+..  " -4c:: .%16Kq6P6P<$9<< 216P6Pr;)rrrrrNrrrrrrTTrrFFrrrrFNN) __name__ __module__ __qualname____doc__ model_typekeys_to_ignore_at_inference attribute_mapr6r4 __classcell__)r9s@r:rrsc c JJ#4"5#0'& M&(- % ) #5A A A A A A F&&&&&&&r;rN) rP transformers.configuration_utilsrtransformers.utilsr get_loggerrMloggerrr!r;r:rYs&======&&&&&&  H % %XXXXX!XXXXXr;