cCiX.SSKJr SSKrSSKJr SSKJrJr SSKJr SSK J r J r SSK J r SS KJr SS KJrJr SS KJr S S KJr S SKJrJrJrJrJrJrJrJrJ r J!r!J"r" "SS\\5r#"SS\5r$"SS\5r%"SS\!5r&"SS\5r'"SS\ 5r("SS\"5r)"SS\5r*"SS \5r+"S!S"\5r,"S#S$\5r-"S%S&\5r./S'Qr/g)()OptionalN)nn)Cache DynamicCache)PretrainedConfig)create_causal_mask!create_sliding_window_causal_mask)BaseModelOutputWithPast)Unpack)TransformersKwargsauto_docstring)check_model_inputs) MistralConfig) Qwen2AttentionQwen2DecoderLayerQwen2ForCausalLMQwen2ForQuestionAnsweringQwen2ForSequenceClassificationQwen2ForTokenClassificationQwen2MLP Qwen2ModelQwen2PreTrainedModel Qwen2RMSNormQwen2RotaryEmbeddingcN\rSrSrSrSrSSjrSrg)MinistralConfiga This is the configuration class to store the configuration of a [`MinistralModel`]. It is used to instantiate an Ministral 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 Ministral-8B-Instruct-2410. [mistralai/Ministral-8B-Instruct-2410](https://huggingface.co/mistralai/Ministral-8B-Instruct-2410) [mistralai/Ministral-8B-Instruct-2410](https://huggingface.co/mistralai/Ministral-8B-Instruct-2410) 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 32000): Vocabulary size of the Ministral model. Defines the number of different tokens that can be represented by the `inputs_ids` passed when calling [`MinistralModel`] hidden_size (`int`, *optional*, defaults to 4096): Dimension of the hidden representations. intermediate_size (`int`, *optional*, defaults to 14336): Dimension of the MLP representations. num_hidden_layers (`int`, *optional*, defaults to 32): Number of hidden layers in the Transformer encoder. num_attention_heads (`int`, *optional*, defaults to 32): Number of attention heads for each attention layer in the Transformer encoder. num_key_value_heads (`int`, *optional*, defaults to 8): This is the number of key_value heads that should be used to implement Grouped Query Attention. If `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed by meanpooling all the original heads within that group. For more details, check out [this paper](https://huggingface.co/papers/2305.13245). If it is not specified, will default to `8`. head_dim (`int`, *optional*, defaults to `hidden_size // num_attention_heads`): The attention head dimension. 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 `4096*32`): The maximum sequence length that this model might ever be used with. Ministral's sliding window attention allows sequence of up to 4096*32 tokens. initializer_range (`float`, *optional*, defaults to 0.02): The standard deviation of the truncated_normal_initializer for initializing all weight 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*): The id of the padding token. bos_token_id (`int`, *optional*, defaults to 1): The id of the "beginning-of-sequence" token. eos_token_id (`int`, *optional*, defaults to 2): The id of the "end-of-sequence" token. tie_word_embeddings (`bool`, *optional*, defaults to `False`): Whether the model's input and output word embeddings should be tied. rope_theta (`float`, *optional*, defaults to 10000.0): The base period of the RoPE embeddings. sliding_window (`int`, *optional*, defaults to 4096): Sliding window attention window size. If not specified, will default to `4096`. attention_dropout (`float`, *optional*, defaults to 0.0): The dropout ratio for the attention probabilities. layer_types (`list`, *optional*): Attention pattern for each layer. ```python >>> from transformers import MinistralModel, MinistralConfig >>> # Initializing a Ministral 8B style configuration >>> configuration = MinistralConfig() >>> # Initializing a model from the Ministral 8B style configuration >>> model = MinistralModel(configuration) >>> # Accessing the model configuration >>> configuration = model.config ``` ministralNc ^[R"U4U UUUS.UD6 XlXlX lX0lX@lXPlUUlXpl UcUnX`l Xl Xl Xl XlUUlUUlUUlUR"cURbSOS/U-Ulgg)N) pad_token_id bos_token_id eos_token_idtie_word_embeddingssliding_attentionfull_attention)r__init__ vocab_sizemax_position_embeddings hidden_sizeintermediate_sizenum_hidden_layersnum_attention_headssliding_windowhead_dimnum_key_value_heads hidden_actinitializer_range rms_norm_eps use_cache rope_thetaattention_dropout layer_types)selfr)r+r,r-r.r1r0r2r*r3r4r5r"r#r$r%r6r/r7r8kwargss i/home/james-whalen/.local/lib/python3.13/site-packages/transformers/models/ministral/modular_ministral.pyr(MinistralConfig.__init__ks0 !!  %%% 3    %'>$&!2!2#6 ,   &"5 #6 $!2("$!2&    #'+':':'F#L\ ! "D  $)r7r0r2r+r3r,r8r*r.r-r1r4r6r/r5r))i}i8 r?Nsiluig{Gz?gư>TNrFg@r>gN)__name__ __module__ __qualname____firstlineno____doc__ model_typer(__static_attributes__r=r;rrsSI VJ )!+9"r=rc\rSrSrSrg) MinistralMLPrJNrCrDrErFrIrJr=r;rLrLr=rLc0^\rSrSrS\4U4SjjrSrU=r$)MinistralAttention layer_idxc>[TU]X5 [R"URUR UR -SS9Ul[R"URURUR -SS9Ul [R"URURUR -SS9Ul g)NF)bias) superr(rLinearr+r.r0q_projr1k_projv_proj)r9configrS __class__s r;r(MinistralAttention.__init__s +ii 2 2F4N4NQUQ^Q^4^ejk ii 2 2F4N4NQUQ^Q^4^ejk ii 2 2F4N4NQUQ^Q^4^ejk r=)rYrXrZ)rCrDrErFintr(rI __classcell__r\s@r;rQrQsl#llr=rQc\rSrSrSrg)MinistralRMSNormrJNrNrJr=r;rbrbrOr=rbc\rSrSrSrg)MinistralDecoderLayerrJNrNrJr=r;rererOr=rec\rSrSrSrg)MinistralPreTrainedModelrJNrNrJr=r;rhrhrOr=rhc\rSrSrSrg)MinistralRotaryEmbeddingrJNrNrJr=r;rkrkrOr=rkc"^\rSrSrS\4U4Sjjr\"5\SS\\ RS\\ RS\\ RS\\ S\\ RS \\S \\ RS \\S \4S jj55rSrU=r$)MinistralModelr[c(>[TU]U5 U?g)N)rVr(has_sliding_layers)r9r[r\s r;r(MinistralModel.__init__s   #r= input_idsattention_mask position_idspast_key_values inputs_embedsr5cache_positionr:returnc USLUSL- (a [S5eUcURU5nU(aUc[URS9nUcDUbUR 5OSn [ R "XURS-URS9nUcURS5n[U=n [5(d*URUUUUUS.n [S 0U D6[S 0U D6S.n Un URX5n URSURR HnU"U 4XR"UUUUU S.UD6n M! UR%U 5n ['U U(aUS 9$SS 9$) Nz:You must specify exactly one of input_ids or inputs_embeds)r[rrB)device)r[ input_embedsrtrxrvru)r'r&)rtrurvr5rxposition_embeddings)last_hidden_statervrJ) ValueError embed_tokensrr[get_seq_lengthtorcharangeshaper{ unsqueeze isinstancedictr r rotary_emblayersr-attention_typenormr )r9rsrtrurvrwr5rxr:past_seen_tokenscausal_mask_mapping mask_kwargs hidden_statesr} decoder_layers r;forwardMinistralModel.forwards  -t"; <YZ Z   --i8M 0*$++>O  !CRC^==?de "\\ ]5H5H5K"KTaThThN  )33A6L?-FF++ -"0"0#2 , K#5"C{"C%F%U%U#  & #oomJ![[)H4;;+H+HIM) 23O3OP) /#-$7  MJ -0 &+/8O  >B  r=rJ)NNNNNNN)rCrDrErFrr(rrrr LongTensorTensorr FloatTensorboolr r r rrIr_r`s@r;rnrns$$151537+/59$(59C E,,-C !.C u//0 C "% C   1 12 C D>C !!1!12C +,C  !C C r=rnc\rSrSrSrg)MinistralForCausalLMirJNrNrJr=r;rrrOr=rc\rSrSrSrg)"MinistralForSequenceClassificationirJNrNrJr=r;rrrOr=rc\rSrSrSrg)MinistralForTokenClassificationirJNrNrJr=r;rrrOr=rc\rSrSrSrg)MinistralForQuestionAnsweringirJNrNrJr=r;rrrOr=r)rrhrnrrrr)0typingrrr cache_utilsrrconfiguration_utilsr masking_utilsr r modeling_outputsr processing_utilsr utilsr r utils.genericrmistral.configuration_mistralrqwen2.modeling_qwen2rrrrrrrrrrrrrLrQrbrerhrkrnrrrr__all__rJr=r;rs .3R7&7/9    G"m%5G"T 8 ll |  -  3  3 J ZJ Z +  )G  &A  $=  r=