Ig+dZddlmZmZddlmZddlmZeje Z GddeZ Gdd eZ Gd d eZ y ) zDBRX model configuration)AnyOptional)PretrainedConfig)loggingc feZdZdZ d dedeedededef fd Ze de ded d fd Z 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. attn_pdropclip_qkv kv_n_heads rope_thetakwargsc t|di|||_||_||_||_dD]}||vs|j |t|dk7rtd|y)N model_typerFound unknown kwargs=) super__init__r r r r poplen ValueError)selfr r r r rk __class__s h/var/www/html/answerous/venv/lib/python3.12/site-packages/transformers/models/dbrx/configuration_dbrx.pyrzDbrxAttentionConfig.__init__,sx "6"$  $$ AF{ 1   v;! 5fY78 8 pretrained_model_name_or_pathreturnrc D|j||j|fi|\}}|jddk(r|d}d|vrJt|dr>|d|jk7r,t j d|dd|jdz|j|fi|S)Nrdbrx attn_configYou are using a model of type to instantiate a model of type N. This is not supported for all configurations of models and can yield errors._set_token_in_kwargsget_config_dictgethasattrrloggerwarning from_dictclsrr config_dicts rfrom_pretrainedz#DbrxAttentionConfig.from_pretrained@s   (1c112OZSYZ V ??< (F 2%m4K ; &73 +E+VbJcgjguguJu NN0\1J0KKkl^^$$rst  s}}[3F33r)Ng@) __name__ __module__ __qualname____doc__floatrintrr classmethodstrr1 __classcell__rs@rr r sy& $(# 995/9 9  9  9(4C434Se44rr c~eZdZdZ ddededededeededeed effd Z e d e d ed d fdZ xZ S) DbrxFFNConfiga|Configuration class for Dbrx FFN. [`DbrxFFN`] class. It is used to instantiate feedforward layers according to the specified arguments, defining the layers architecture. Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the documentation from [`PretrainedConfig`] for more information. Args: ffn_act_fn (`dict`, *optional*, defaults to `None`): A dict specifying activation function for the FFN. The dict should have a key 'name' with the value being the name of the activation function along with any additional keyword arguments. If `None`, then set to `{"name": "silu"}`. ffn_hidden_size (`int`, *optional*, defaults to 3584): The hidden size of the feedforward network. moe_num_experts (`int`, *optional*, defaults to 4): The number of experts in the mixture of experts layer. moe_top_k (`int`, *optional*, defaults to 1): The number of experts to use in the mixture of experts layer. moe_jitter_eps (`float`, *optional*, defaults to `None`): If not `None`, the jitter epsilon for the mixture of experts layer. moe_loss_weight (`float`, *optional*, defaults to 0.01): The loss weight for the mixture of experts layer. moe_normalize_expert_weights (`float`, *optional*, defaults to 1.0): The normalization factor for the expert weights. ffn_act_fnffn_hidden_sizemoe_num_experts moe_top_kmoe_jitter_epsmoe_loss_weightmoe_normalize_expert_weightsrc t ||ddi}||_||_||_||_||_||_||_dD]} | |vs|j| t|dk7rtd|y)Nnamesilurrr) rrr@rArBrCrDrErFrrr) rr@rArBrCrDrErFrrrs rrzDbrxFFNConfig.__init__gs    &)J$..",.,H) AF{ 1   v;! 5fY78 8 rrrrc D|j||j|fi|\}}|jddk(r|d}d|vrJt|dr>|d|jk7r,t j d|dd|jdz|j|fi|S)Nrr! ffn_configr#r$r%r&r.s rr1zDbrxFFNConfig.from_pretraineds   (1c112OZSYZ V ??< (F 2%l3K ; &73 +E+VbJcgjguguJu NN0\1J0KKkl^^$$rst  s}}[3F33r)Nir3Ng{Gz?g?)r4r5r6r7dictr9rr8rrr:r;r1r<r=s@rr?r?Rs, # *.!%8;999 9  9 ! 99'/uo9984C434Se44rr?ceZdZdZdZdddddZ ddededededed ed ed ee d ee d e 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 vocab_size resid_pdrop emb_pdropr"rK use_cacheinitializer_rangeoutput_router_logitsrc  |t|_n(t|trtdi||_n||_| t |_n(t| trt di| |_n| |_||_||_||_||_ ||_ ||_ ||_ | |_ | |_| |_|jj |_| j%dd}|r t'dt)|Tdd|i| y)Ntie_word_embeddingsFz5tie_word_embeddings is not supported for DBRX models.r)r r" isinstancerMr?rKrQrPrRrSrXrYrZr[r\r]r num_key_value_headsrrrr)rrQrPrRrSrXrYrZr"rKr[r\r]rr_rs rrzDbrxConfig.__init__s  24D   T *2A[AD *D   +oDO  D )+9j9DO(DO    &$&""!2$8!#'#3#3#>#> $jj)>F TU U K-@KFKr) rbi}r2r2NNTg{Gz?F)r4r5r6r7r attribute_mapr9r8rr r?boolrrr<r=s@rrOrOs4lJ( '#0 M 59.2#'%*.L.L.L .L  .L  .L.L.L12.L]+.L.L!.L#.L.L.LrrON)r7typingrrconfiguration_utilsrutilsr get_loggerr4r+r r?rOrrrrksV 3   H %54*54p@4$@4FmL!mLr