g+dZddlmZmZddlmZddlmZeje Z GddeZ Gdd eZ Gd d eZ d S) zDBRX model configuration)AnyOptional)PretrainedConfig)loggingc reZdZdZ ddedeeded ed ef fd Ze d e d ed dfdZ xZ S)DbrxAttentionConfiga_Configuration class for Dbrx Attention. [`DbrxAttention`] class. It is used to instantiate attention layers according to the specified arguments, defining the layers architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: attn_pdrop (`float`, *optional*, defaults to 0.0): The dropout probability for the attention layers. clip_qkv (`float`, *optional*): If set, clip the queries, keys, and values in the attention layer to this value. kv_n_heads (`int`, *optional*, defaults to 1): For grouped_query_attention only, allow user to specify number of kv heads. rope_theta (`float`, *optional*, defaults to 10000.0): The base frequency for rope. N@ attn_pdropclip_qkv kv_n_heads rope_thetakwargsc tjdi|||_||_||_||_dD]}||vr||t|dkrtd|dS)N model_typerFound unknown kwargs=) super__init__r rrrpoplen ValueError)selfr rrrrk __class__s g/var/www/html/ai-engine/env/lib/python3.11/site-packages/transformers/models/dbrx/configuration_dbrx.pyrzDbrxAttentionConfig.__init__,s ""6"""$  $$  AF{{ 1 v;;!  7f7788 8  pretrained_model_name_or_pathreturnrc T|||j|fi|\}}|ddkr|d}d|vrPt|dr@|d|jkr/t d|dd|jdz|j|fi|S)Nrdbrx attn_configYou are using a model of type to instantiate a model of type N. This is not supported for all configurations of models and can yield errors._set_token_in_kwargsget_config_dictgethasattrrloggerwarning from_dictclsr!r config_dicts rfrom_pretrainedz#DbrxAttentionConfig.from_pretrained@s   (((1c12OZZSYZZ V ??< ( (F 2 2%m4K ; & &73 +E+E &+VbJcgjguJuJu NNl\1Jlll^ssst    s}[33F333r )r Nr r ) __name__ __module__ __qualname____doc__floatrintrr classmethodstrr4 __classcell__rs@rr r s& $(# 9995/9 9  9  999999(4C434Se444[44444r r ceZdZdZ dded ed ed ed eed edeedeffd Z e de deddfdZ xZ S) DbrxFFNConfiga|Configuration class for Dbrx FFN. [`DbrxFFN`] class. It is used to instantiate feedforward layers according to the specified arguments, defining the layers architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: ffn_act_fn (`dict`, *optional*, defaults to `None`): A dict specifying activation function for the FFN. The dict should have a key 'name' with the value being the name of the activation function along with any additional keyword arguments. If `None`, then set to `{"name": "silu"}`. ffn_hidden_size (`int`, *optional*, defaults to 3584): The hidden size of the feedforward network. moe_num_experts (`int`, *optional*, defaults to 4): The number of experts in the mixture of experts layer. moe_top_k (`int`, *optional*, defaults to 1): The number of experts to use in the mixture of experts layer. moe_jitter_eps (`float`, *optional*, defaults to `None`): If not `None`, the jitter epsilon for the mixture of experts layer. moe_loss_weight (`float`, *optional*, defaults to 0.01): The loss weight for the mixture of experts layer. moe_normalize_expert_weights (`float`, *optional*, defaults to 1.0): The normalization factor for the expert weights. Nr {Gz?? ffn_act_fnffn_hidden_sizemoe_num_experts moe_top_kmoe_jitter_epsmoe_loss_weightmoe_normalize_expert_weightsrc <t|ddi}||_||_||_||_||_||_||_dD]} | |vr| | t|dkrtd|dS)Nnamesilurrr) rrrErFrGrHrIrJrKrrr) rrErFrGrHrIrJrKrrrs rrzDbrxFFNConfig.__init__gs    &)J$..",.,H)  AF{{ 1 v;;!  7f7788 8  r r!r"rc T|||j|fi|\}}|ddkr|d}d|vrPt|dr@|d|jkr/t d|dd|jdz|j|fi|S)Nrr$ ffn_configr&r'r(r)r1s rr4zDbrxFFNConfig.from_pretraineds   (((1c12OZZSYZZ V ??< ( (F 2 2%l3K ; & &73 +E+E &+VbJcgjguJuJu NNl\1Jlll^ssst    s}[33F333r )NrArBr NrCrD)r5r6r7r8dictr:rr9rrr;r<r4r=r>s@rr@r@Rs, # *.!%8;9999 9  9 ! 99'/uo999999984C434Se444[44444r r@ceZdZdZdZdddddZ ddedededededededee dee de dede de ffd Z xZS) DbrxConfiga. This is the configuration class to store the configuration of a [`DbrxModel`]. It is used to instantiate a Dbrx model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a different configuration to that of the [databricks/dbrx-instruct](https://huggingface.co/databricks/dbrx-instruct) architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: d_model (`int`, *optional*, defaults to 2048): Dimensionality of the embeddings and hidden states. n_heads (`int`, *optional*, defaults to 16): Number of attention heads for each attention layer in the Transformer encoder. n_layers (`int`, *optional*, defaults to 24): Number of hidden layers in the Transformer encoder. max_seq_len (`int`, *optional*, defaults to 2048): The maximum sequence length of the model. vocab_size (`int`, *optional*, defaults to 32000): Vocabulary size of the Dbrx model. Defines the maximum number of different tokens that can be represented by the `inputs_ids` passed when calling [`DbrxModel`]. resid_pdrop (`float`, *optional*, defaults to 0.0): The dropout probability applied to the attention output before combining with residual. emb_pdrop (`float`, *optional*, defaults to 0.0): The dropout probability for the embedding layer. attn_config (`dict`, *optional*): A dictionary used to configure the model's attention module. ffn_config (`dict`, *optional*): A dictionary used to configure the model's FFN module. use_cache (`bool`, *optional*, defaults to `True`): Whether or not the model should return the last key/values attentions (not used by all models). initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight matrices. 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. See [here]() for more details. Example: ```python >>> from transformers import DbrxConfig, DbrxModel >>> # Initializing a Dbrx configuration >>> configuration = DbrxConfig(n_layers=2, d_model=256, n_heads=8, vocab_size=128) >>> # Initializing a model (with random weights) from the configuration >>> model = DbrxModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ``` r$n_headsd_modeln_layers max_seq_len)num_attention_heads hidden_sizenum_hidden_layersmax_position_embeddings}r NT{Gz?F vocab_size resid_pdrop emb_pdropr%rP use_cacheinitializer_rangeoutput_router_logitsrc  J|t|_n.t|trtdi||_n||_| t |_n.t| trt di| |_n| |_||_||_||_||_ ||_ ||_ ||_ | |_ | |_| |_|jj|_| dd}|rt'dt)jdd|i| dS)Ntie_word_embeddingsFz5tie_word_embeddings is not supported for DBRX models.r)r r% isinstancerQr@rPrUrTrVrWrarbrcrdrerfrnum_key_value_headsrrrr)rrUrTrVrWrarbrcr%rPrdrerfrrhrs rrzDbrxConfig.__init__s@  244D    T * * +2AA[AAD  *D   +ooDOO  D ) ) )+99j99DOO(DO    &$&""!2$8!#'#3#> $jj)>FF  VTUU UKK-@KFKKKKKr ) r\r]r^r\r_r r NNTr`F)r5r6r7r8r attribute_mapr:r9rr r@boolrrr=r>s@rrSrSs544lJ( '#0 M 59.2#'%*.L.L.L.L .L  .L  .L.L.L12.L]+.L.L!.L#.L.L.L.L.L.L.L.L.L.L.Lr rSN)r8typingrrconfiguration_utilsrutilsr get_loggerr5r.r r@rSrr rrqs 333333  H % %5454545454*545454p@4@4@4@4@4$@4@4@4FmLmLmLmLmL!mLmLmLmLmLr