o
    iQ                     @   sP  d dl mZmZmZ d dlZd dlmZ d dlmZ ddlm	Z	 ddl
mZmZ ddlmZ dd	lmZ dd
lmZmZ ddlmZ ddlmZmZmZmZ ddlmZmZ ddlmZm Z  ddl!m"Z"m#Z# ddl$m%Z% ddl&m'Z'm(Z(m)Z) ddl*m+Z+ ddl,m-Z- G dd dej.Z/dd Z0dAddZ1dej2de3dej2fddZ4	 dBd!ej.d"ej2d#ej2d$ej2d%eej2 d&e5d'e5d(e%e' fd)d*Z6G d+d, d,ej.Z7ed-G d.d/ d/ej.Z8G d0d1 d1eZ9e(G d2d3 d3e#Z:G d4d5 d5ej.Z;e(G d6d7 d7e:Z<e(G d8d9 d9e:eZ=G d:d; d;ee:Z>G d<d= d=ee:Z?G d>d? d?ee:Z@g d@ZAdS )C    )CallableOptionalUnionN)nn)check_model_inputs   )ACT2FN)CacheDynamicCache)GenerationMixin)use_kernel_forward_from_hub)create_causal_mask!create_sliding_window_causal_mask)FlashAttentionKwargs)GenericForQuestionAnswering GenericForSequenceClassificationGenericForTokenClassificationGradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tuple)deprecate_kwarg   )MistralConfigc                       s$   e Zd Z fddZdd Z  ZS )
MistralMLPc                    sr   t    || _|j| _|j| _tj| j| jdd| _tj| j| jdd| _tj| j| jdd| _	t
|j | _d S NFbias)super__init__confighidden_sizeintermediate_sizer   Linear	gate_projup_proj	down_projr   
hidden_actact_fnselfr'   	__class__ `/home/ubuntu/.local/lib/python3.10/site-packages/transformers/models/mistral/modeling_mistral.pyr&   $   s   
zMistralMLP.__init__c                 C   s$   |  | | || | }|S N)r-   r/   r+   r,   )r1   xr-   r4   r4   r5   forward.   s    zMistralMLP.forward)__name__
__module____qualname__r&   r8   __classcell__r4   r4   r2   r5   r!   #   s    
r!   c                 C   sH   | dd| j d d f }| d| j d d df }tj| |fddS )z*Rotates half the hidden dims of the input..N   dim)shapetorchcat)r7   x1x2r4   r4   r5   rotate_half3   s   rF   c                 C   sD   | |}| |}| | t| |  }|| t||  }||fS )a  Applies Rotary Position Embedding to the query and key tensors.

    Args:
        q (`torch.Tensor`): The query tensor.
        k (`torch.Tensor`): The key tensor.
        cos (`torch.Tensor`): The cosine part of the rotary embedding.
        sin (`torch.Tensor`): The sine part of the rotary embedding.
        position_ids (`torch.Tensor`, *optional*):
            Deprecated and unused.
        unsqueeze_dim (`int`, *optional*, defaults to 1):
            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
    Returns:
        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
    )	unsqueezerF   )qkcossinposition_idsunsqueeze_dimq_embedk_embedr4   r4   r5   apply_rotary_pos_emb:   s
   

rP   hidden_statesn_repreturnc                 C   s^   | j \}}}}|dkr| S | dddddddddf |||||} | ||| ||S )z
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    r   N)rA   expandreshape)rQ   rR   batchnum_key_value_headsslenhead_dimr4   r4   r5   	repeat_kvU   s
   0rZ           modulequerykeyvalueattention_maskscalingdropoutkwargsc                 K   s   t || j}t || j}	t||dd| }
|d ur3|d d d d d d d |jd f }|
| }
tjj|
dtj	d
|j}
tjj|
|| jd}
t|
|	}|dd }||
fS )Nr>   r   r=   )r@   dtype)ptrainingr   )rZ   num_key_value_groupsrB   matmul	transposerA   r   
functionalsoftmaxfloat32tore   rb   rg   
contiguous)r\   r]   r^   r_   r`   ra   rb   rc   
key_statesvalue_statesattn_weightscausal_maskattn_outputr4   r4   r5   eager_attention_forwarda   s   
&ru   c                       s   e Zd ZdZdedef fddZedddd		
	
ddej	de
ej	ej	f deej	 dee deej dee de
ej	eej	 f fddZ  ZS )MistralAttentionz=Multi-headed attention from 'Attention Is All You Need' paperr'   	layer_idxc                    s   t    || _|| _t|dd p|j|j | _|j|j | _	| jd | _
|j| _d| _tj|j|j| j dd| _tj|j|j| j dd| _tj|j|j| j dd| _tj|j| j |jdd| _d S )NrY   g      TFr#   )r%   r&   r'   rw   getattrr(   num_attention_headsrY   rW   rh   ra   attention_dropout	is_causalr   r*   q_projk_projv_projo_projr1   r'   rw   r2   r4   r5   r&   ~   s   
 zMistralAttention.__init__past_key_valuepast_key_values4.58new_nameversionNrQ   position_embeddingsr`   cache_positionrc   rS   c                 K   s0  |j d d }g |d| jR }| ||dd}	| ||dd}
| ||dd}|\}}t|	|
||\}	}
|d urW|||d}||
|| j	|\}
}t
}| jjdkret| jj }|| |	|
||f| jsqdn| j| jt| jdd d|\}}|jg |dR   }| |}||fS )	Nr=   r   r>   )rK   rJ   r   eagerr[   sliding_window)rb   ra   r   )rA   rY   r|   viewrj   r}   r~   rP   updaterw   ru   r'   _attn_implementationr   rg   rz   ra   rx   rU   ro   r   )r1   rQ   r   r`   r   r   rc   input_shapehidden_shapequery_statesrp   rq   rJ   rK   cache_kwargsattention_interfacert   rr   r4   r4   r5   r8      s:   
	

zMistralAttention.forward)NN)r9   r:   r;   __doc__r    intr&   r   rB   Tensortupler   r	   
LongTensorr   r   r8   r<   r4   r4   r2   r5   rv   {   s*    rv   RMSNormc                       s.   e Zd Zd fdd	Zdd Zdd Z  ZS )	MistralRMSNormư>c                    s&   t    tt|| _|| _dS )z=
        MistralRMSNorm is equivalent to T5LayerNorm
        N)r%   r&   r   	ParameterrB   onesweightvariance_epsilon)r1   r(   epsr2   r4   r5   r&      s   

zMistralRMSNorm.__init__c                 C   sJ   |j }|tj}|djddd}|t|| j  }| j|| S )Nr>   r=   T)keepdim)	re   rn   rB   rm   powmeanrsqrtr   r   )r1   rQ   input_dtypevariancer4   r4   r5   r8      s
   zMistralRMSNorm.forwardc                 C   s   t | jj d| j S )Nz, eps=)r   r   rA   r   )r1   r4   r4   r5   
extra_repr   s   zMistralRMSNorm.extra_repr)r   )r9   r:   r;   r&   r8   r   r<   r4   r4   r2   r5   r      s    r   c                       s   e Zd Zdedef fddZedddd							
				ddejde	ej de	ej
 de	e de	e de	ej
 de	eejejf  dee dejfddZ  ZS )MistralDecoderLayerr'   rw   c                    sR   t    |j| _t||d| _t|| _t|j|jd| _	t|j|jd| _
d S )N)r'   rw   r   )r%   r&   r(   rv   	self_attnr!   mlpr   rms_norm_epsinput_layernormpost_attention_layernormr   r2   r4   r5   r&      s   

zMistralDecoderLayer.__init__r   r   r   r   NFrQ   r`   rL   	use_cacher   r   rc   rS   c              
   K   s^   |}	|  |}| jd|||||||d|\}}
|	| }|}	| |}| |}|	| }|S )N)rQ   r`   rL   r   r   r   r   r4   )r   r   r   r   )r1   rQ   r`   rL   r   r   r   r   rc   residual_r4   r4   r5   r8      s&   




zMistralDecoderLayer.forward)NNNFNN)r9   r:   r;   r    r   r&   r   rB   r   r   r   r	   boolr   r   r   r8   r<   r4   r4   r2   r5   r      s8    	
r   c                   @   sH   e Zd ZU eed< dZdZdgZdgZdZ	dZ
dZdZdZeedZdS )MistralPreTrainedModelr'   modelTr   r   )rQ   
attentionsN)r9   r:   r;   r    __annotations__base_model_prefixsupports_gradient_checkpointing_no_split_modules_skip_keys_device_placement_supports_flash_attn_supports_sdpa_supports_flex_attn_can_compile_fullgraph_supports_attention_backendr   rv   _can_record_outputsr4   r4   r4   r5   r      s   
 
r   c                       sD   e Zd ZU ejed< ddef fddZe e	dd Z
  ZS )	MistralRotaryEmbeddinginv_freqNr'   c                    s   t    t|drt|jtr|jd|jd| _nd| _|j| _	|j| _
|| _t| j | _| | j|\}| _| jd|dd | j| _d S )Nrope_scaling	rope_typetypedefaultr   F)
persistent)r%   r&   hasattr
isinstancer   dictgetr   max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenr'   r   rope_init_fnattention_scalingregister_bufferr   original_inv_freq)r1   r'   devicer   r2   r4   r5   r&     s   
zMistralRotaryEmbedding.__init__c           
      C   s   | j d d d d f  |jd dd|j}|d d d d d f  }t|jjtr6|jjdkr6|jjnd}t	j
|dd+ | |  dd}t	j||fdd	}| | j }| | j }	W d    n1 smw   Y  |j|jd
|	j|jd
fS )Nr   r=   r   mpscpuF)device_typeenabledr>   r?   )re   )r   floatrT   rA   rn   r   r   r   strrB   autocastrj   rC   rJ   r   rK   re   )
r1   r7   rL   inv_freq_expandedposition_ids_expandedr   freqsembrJ   rK   r4   r4   r5   r8   "  s   0&zMistralRotaryEmbedding.forwardr6   )r9   r:   r;   rB   r   r   r    r&   no_gradr   r8   r<   r4   r4   r2   r5   r     s   
 
r   c                       s   e Zd Zdef fddZee							ddeej	 deej
 deej	 dee d	eej d
ee deej	 dee defddZ  ZS )MistralModelr'   c                    s   t     j| _ j| _t j j| j| _t	 fddt
 jD | _t j jd| _t d| _d| _|   d S )Nc                    s   g | ]}t  |qS r4   )r   ).0rw   r'   r4   r5   
<listcomp>;  s    z)MistralModel.__init__.<locals>.<listcomp>r   r   F)r%   r&   pad_token_idpadding_idx
vocab_sizer   	Embeddingr(   embed_tokens
ModuleListrangenum_hidden_layerslayersr   r   normr   
rotary_embgradient_checkpointing	post_initr0   r2   r   r5   r&   4  s   zMistralModel.__init__N	input_idsr`   rL   r   inputs_embedsr   r   rc   rS   c              
   K   s  |d u |d uA rt d|d u r| |}|r!|d u r!t| jd}|d u r=|d ur-| nd}	tj|	|	|jd  |jd}|d u rF|	d}| jj
d u rNtnt}
|
| j|||||d}|}| ||}| jd | jj D ]}||f||||||d|}ql| |}t||r|dS d dS )	Nz:You must specify exactly one of input_ids or inputs_embedsr   r   r   )r   )r'   input_embedsr`   r   r   rL   )r`   rL   r   r   r   r   )last_hidden_stater   )
ValueErrorr   r
   r'   get_seq_lengthrB   arangerA   r   rG   r   r   r   r   r   r   r   r   )r1   r   r`   rL   r   r   r   r   rc   past_seen_tokensmask_functionrs   rQ   r   decoder_layerr4   r4   r5   r8   D  sX   

	

zMistralModel.forward)NNNNNNN)r9   r:   r;   r    r&   r   r   r   rB   r   r   r	   FloatTensorr   r   r   r   r8   r<   r4   r4   r2   r5   r   2  s<    	
r   c                       s   e Zd ZdgZddiZddgdgfiZ fddZee										dd
e	e
j de	e
j de	e
j de	e de	e
j de	e
j de	e de	e
j deee
jf dee defddZ  ZS )MistralForCausalLMzlm_head.weightlm_headcolwise_reprQ   logitsc                    s@   t  | t|| _|j| _tj|j|jdd| _| 	  d S r"   )
r%   r&   r   r   r   r   r*   r(   r   r   r0   r2   r4   r5   r&     s
   
zMistralForCausalLM.__init__Nr   r   r`   rL   r   r   labelsr   r   logits_to_keeprc   rS   c
              
   K   s   | j d|||||||d|
}|j}t|	trt|	 dn|	}| |dd|ddf }d}|durB| jd||| jjd|
}t	|||j
|j|jdS )a  
        Example:

        ```python
        >>> from transformers import AutoTokenizer, MistralForCausalLM

        >>> model = MistralForCausalLM.from_pretrained("meta-mistral/Mistral-2-7b-hf")
        >>> tokenizer = AutoTokenizer.from_pretrained("meta-mistral/Mistral-2-7b-hf")

        >>> prompt = "Hey, are you conscious? Can you talk to me?"
        >>> inputs = tokenizer(prompt, return_tensors="pt")

        >>> # Generate
        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
        ```)r   r`   rL   r   r   r   r   N)r   r   r   )lossr   r   rQ   r   r4   )r   r   r   r   slicer   loss_functionr'   r   r   r   rQ   r   )r1   r   r`   rL   r   r   r   r   r   r   rc   outputsrQ   slice_indicesr   r  r4   r4   r5   r8     s0    zMistralForCausalLM.forward)	NNNNNNNNr   )r9   r:   r;   _tied_weights_keys_tp_plan_pp_planr&   r   r   r   rB   r   r   r	   r   r   r   r   r   r   r   r8   r<   r4   r4   r2   r5   r     sN    		
r   c                   @      e Zd ZdS )MistralForTokenClassificationNr9   r:   r;   r4   r4   r4   r5   r
        r
  c                   @   r	  ) MistralForSequenceClassificationNr  r4   r4   r4   r5   r    r  r  c                   @   r	  )MistralForQuestionAnsweringNr  r4   r4   r4   r5   r    s    r  )r   r  r   r   r  r
  )Nr   )r[   )Btypingr   r   r   rB   r   transformers.utils.genericr   activationsr   cache_utilsr	   r
   
generationr   integrationsr   masking_utilsr   r   modeling_flash_attention_utilsr   modeling_layersr   r   r   r   modeling_outputsr   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   r   utils.deprecationr   configuration_mistralr    Moduler!   rF   rP   r   r   rZ   r   ru   rv   r   r   r   r   r   r   r
  r  r  __all__r4   r4   r4   r5   <module>   sn   

?,$OK