a
    huT                  
   @   s<  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	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$m%Z%m&Z&m'Z' ddl(m)Z) ddl*m+Z+ ddl,m-Z- G dd dej.Z/dd Z0d8ddZ1ej2e3ej2dddZ4d9ej.ej2ej2ej2eej2 e5e5e#e% dd d!Z6G d"d# d#ej.Z7ed$G d%d& d&ej.Z8G d'd( d(eZ9e&G d)d* d*e!Z:G d+d, d,ej.Z;e&G d-d. d.e:Z<e&G d/d0 d0e:eZ=G d1d2 d2ee:Z>G d3d4 d4ee:Z?G d5d6 d6ee:Z@g d7ZAdS ):    )CallableOptionalUnionN)nn   )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)check_model_inputs   )Qwen2Configc                       s$   e Zd Z fddZdd Z  ZS )Qwen2MLPc                    sr   t    || _|j| _|j| _tj| j| jdd| _tj| j| jdd| _tj| j| jdd| _	t
|j | _d S NFZbias)super__init__confighidden_sizeZintermediate_sizer   Linear	gate_projup_proj	down_projr   Z
hidden_actact_fnselfr&   	__class__ d/var/www/html/assistant/venv/lib/python3.9/site-packages/transformers/models/qwen2/modeling_qwen2.pyr%   #   s    
zQwen2MLP.__init__c                 C   s$   |  | | || | }|S )N)r+   r,   r)   r*   )r.   xr+   r1   r1   r2   forward-   s     zQwen2MLP.forward)__name__
__module____qualname__r%   r4   __classcell__r1   r1   r/   r2   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)r3   x1Zx2r1   r1   r2   rotate_half2   s    rA   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.
    )	unsqueezerA   )qkcossinposition_idsZunsqueeze_dimZq_embedZk_embedr1   r1   r2   apply_rotary_pos_emb9   s
    

rH   )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)r=   expandreshape)rI   rJ   batchnum_key_value_headsslenhead_dimr1   r1   r2   	repeat_kvT   s
    0rR           )modulequerykeyvalueattention_maskscalingdropoutkwargsc                 K   s   t || j}t || j}	t||dd| }
|d urf|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   r9   )r<   dtype)ptrainingr   )rR   num_key_value_groupsr>   matmul	transposer=   r   Z
functionalZsoftmaxfloat32tor]   rZ   r_   
contiguous)rT   rU   rV   rW   rX   rY   rZ   r[   
key_statesvalue_statesattn_weightsZcausal_maskattn_outputr1   r1   r2   eager_attention_forward`   s    
&rj   c                       s   e Zd ZdZeed fddZedddddej	e
ej	ej	f eej	 ee eej ee e
ej	eej	 f d
ddZ  ZS )Qwen2Attentionz=Multi-headed attention from 'Attention Is All You Need' paperr&   	layer_idxc                    s   t    || _|| _t|d|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| _|j| dkr|jnd | _d S )NrQ   g      Tr#   Fsliding_attention)r$   r%   r&   rm   getattrr'   Znum_attention_headsrQ   rO   r`   rY   attention_dropoutZ	is_causalr   r(   q_projk_projv_projo_projlayer_typessliding_windowr.   r&   rm   r/   r1   r2   r%   }   s    
zQwen2Attention.__init__past_key_valuepast_key_values4.58new_nameversionN)rI   position_embeddingsrX   ry   cache_positionr[   rK   c                 K   s(  |j d d }g |d| jR }| ||dd}	| ||dd}
| ||dd}|\}}t|	|
||\}	}
|d ur|||d}||
|| j	|\}
}t
}| jjdkrt| jj }|| |	|
||f| jsdn| j| j| jd|\}}|jg |dR   }| |}||fS )Nr9   r   r:   )rF   rE   r   eagerrS   )rZ   rY   rv   )r=   rQ   rq   viewrb   rr   rs   rH   updaterm   rj   r&   Z_attn_implementationr   r_   rp   rY   rv   rM   re   rt   )r.   rI   r~   rX   ry   r   r[   Zinput_shapeZhidden_shapeZquery_statesrf   rg   rE   rF   Zcache_kwargsZattention_interfaceri   rh   r1   r1   r2   r4      s:    
	

zQwen2Attention.forward)NN)r5   r6   r7   __doc__r    intr%   r   r>   Tensortupler   r   
LongTensorr   r   r4   r8   r1   r1   r/   r2   rk   z   s     rk   ZRMSNormc                       sB   e Zd Zdedd fddZejejdddZd	d
 Z  Z	S )Qwen2RMSNormư>N)epsrK   c                    s&   t    tt|| _|| _dS )z;
        Qwen2RMSNorm is equivalent to T5LayerNorm
        N)r$   r%   r   	Parameterr>   Zonesweightvariance_epsilon)r.   r'   r   r/   r1   r2   r%      s    
zQwen2RMSNorm.__init__)rI   rK   c                 C   sJ   |j }|tj}|djddd}|t|| j  }| j|| S )Nr:   r9   T)Zkeepdim)	r]   rd   r>   rc   powmeanZrsqrtr   r   )r.   rI   Zinput_dtypeZvariancer1   r1   r2   r4      s
    zQwen2RMSNorm.forwardc                 C   s   t | jj d| j S )Nz, eps=)r   r   r=   r   )r.   r1   r1   r2   
extra_repr   s    zQwen2RMSNorm.extra_repr)r   )
r5   r6   r7   floatr%   r>   r   r4   r   r8   r1   r1   r/   r2   r      s   r   c                       s   e Zd Zeed fddZedddddeje	ej e	ej
 e	e e	e e	ej
 e	eejejf  ee ejd
	ddZ  ZS )Qwen2DecoderLayerrl   c                    s^   t    |j| _t||d| _t|| _t|j|jd| _	t|j|jd| _
|j| | _d S )Nrl   r   )r$   r%   r'   rk   	self_attnr!   mlpr   rms_norm_epsinput_layernormpost_attention_layernormru   attention_typerw   r/   r1   r2   r%      s    

zQwen2DecoderLayer.__init__rx   ry   rz   r{   NF)	rI   rX   rG   ry   	use_cacher   r~   r[   rK   c              
   K   s^   |}	|  |}| jf |||||||d|\}}
|	| }|}	| |}| |}|	| }|S )N)rI   rX   rG   ry   r   r   r~   )r   r   r   r   )r.   rI   rX   rG   ry   r   r   r~   r[   Zresidual_r1   r1   r2   r4      s&    




zQwen2DecoderLayer.forward)NNNFNN)r5   r6   r7   r    r   r%   r   r>   r   r   r   r   boolr   r   r   r4   r8   r1   r1   r/   r2   r      s&         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 )Qwen2PreTrainedModelr&   modelTr   ry   )rI   
attentionsN)r5   r6   r7   r    __annotations__base_model_prefixZsupports_gradient_checkpointingZ_no_split_modulesZ_skip_keys_device_placementZ_supports_flash_attnZ_supports_sdpaZ_supports_flex_attnZ_can_compile_fullgraphZ_supports_attention_backendr   rk   Z_can_record_outputsr1   r1   r1   r2   r      s   
r   c                       sD   e Zd ZU ejed< ded fddZe e	dd Z
  ZS )	Qwen2RotaryEmbeddinginv_freqNr&   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$   r%   hasattr
isinstancer   dictgetr   Zmax_position_embeddingsZmax_seq_len_cachedZoriginal_max_seq_lenr&   r   Zrope_init_fnattention_scalingZregister_bufferr   Zoriginal_inv_freq)r.   r&   devicer   r/   r1   r2   r%     s    
zQwen2RotaryEmbedding.__init__c           
      C   s   | j d d d d f  |jd dd|j}|d d d d d f  }t|jjtrl|jjdkrl|jjnd}t	j
|ddV | |  dd}t	j||fdd	}| | j }| | j }	W d    n1 s0    Y  |j|jd
|	j|jd
fS )Nr   r9   r   ZmpscpuF)device_typeZenabledr:   r;   )r]   )r   r   rL   r=   rd   r   r   r   strr>   Zautocastrb   r?   rE   r   rF   r]   )
r.   r3   rG   Zinv_freq_expandedZposition_ids_expandedr   ZfreqsZembrE   rF   r1   r1   r2   r4   %  s    0&,zQwen2RotaryEmbedding.forward)N)r5   r6   r7   r>   r   r   r    r%   Zno_gradr   r4   r8   r1   r1   r/   r2   r     s
   

r   c                       st   e Zd Zed fddZeedeej	 eej
 eej	 ee eej ee eej	 ee ed	ddZ  ZS )	
Qwen2Modelr   c                    s   t     j| _ j| _t j j| j| _t	 fddt
 jD | _t j jd| _t d| _d| _d| jjv | _|   d S )Nc                    s   g | ]}t  |qS r1   )r   ).0rm   r   r1   r2   
<listcomp>>      z'Qwen2Model.__init__.<locals>.<listcomp>r   r   Frn   )r$   r%   Zpad_token_idZpadding_idx
vocab_sizer   Z	Embeddingr'   embed_tokensZ
ModuleListrangenum_hidden_layerslayersr   r   normr   
rotary_embZgradient_checkpointingr&   ru   has_sliding_layers	post_initr-   r/   r   r2   r%   7  s    zQwen2Model.__init__N)		input_idsrX   rG   ry   inputs_embedsr   r   r[   rK   c              
   K   sD  |d u |d uA rt d|d u r*| |}|rB|d u rBt| jd}|d u rz|d urZ| nd}	tj|	|	|jd  |jd}|d u r|	d}t
| }
ts| j|||||d}dtf i |i}
| jrtf i ||
d< |}| ||}| jd | jj D ](}||f|
|j |||||d	|}q| |}t||r<|nd d
S )Nz:You must specify exactly one of input_ids or inputs_embedsr   r   r   )r   )r&   Zinput_embedsrX   r   ry   rG   Zfull_attentionrn   )rX   rG   ry   r   r   r~   )last_hidden_statery   )
ValueErrorr   r	   r&   Zget_seq_lengthr>   Zaranger=   r   rB   r   r   r   r   r   r   r   r   r   r   r   )r.   r   rX   rG   ry   r   r   r   r[   Zpast_seen_tokensZcausal_mask_mappingZmask_kwargsrI   r~   Zdecoder_layerr1   r1   r2   r4   H  sZ    



zQwen2Model.forward)NNNNNNN)r5   r6   r7   r    r%   r   r   r   r>   r   r   r   FloatTensorr   r   r   r   r4   r8   r1   r1   r/   r2   r   5  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e	e
j e	e
j e	e
j e	e e	e
j e	e
j e	e e	e
j eee
jf ee ed
ddZ  ZS )Qwen2ForCausalLMzlm_head.weightlm_headZcolwise_reprI   logitsc                    s@   t  | t|| _|j| _tj|j|jdd| _| 	  d S r"   )
r$   r%   r   r   r   r   r(   r'   r   r   r-   r/   r1   r2   r%     s
    
zQwen2ForCausalLM.__init__Nr   )r   rX   rG   ry   r   labelsr   r   logits_to_keepr[   rK   c
              
   K   s   | j f |||||||d|
}|j}t|	tr<t|	 dn|	}| |dd|ddf }d}|dur| jf ||| jjd|
}t	|||j
|j|jdS )a  
        Example:

        ```python
        >>> from transformers import AutoTokenizer, Qwen2ForCausalLM

        >>> model = Qwen2ForCausalLM.from_pretrained("meta-qwen2/Qwen2-2-7b-hf")
        >>> tokenizer = AutoTokenizer.from_pretrained("meta-qwen2/Qwen2-2-7b-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   rX   rG   ry   r   r   r   N)r   r   r   )lossr   ry   rI   r   )r   r   r   r   slicer   Zloss_functionr&   r   r   ry   rI   r   )r.   r   rX   rG   ry   r   r   r   r   r   r[   outputsrI   Zslice_indicesr   r   r1   r1   r2   r4     s0     zQwen2ForCausalLM.forward)	NNNNNNNNr   )r5   r6   r7   Z_tied_weights_keysZ_tp_planZ_pp_planr%   r   r   r   r>   r   r   r   r   r   r   r   r   r   r   r4   r8   r1   r1   r/   r2   r     s8   	         r   c                   @   s   e Zd ZdS )Qwen2ForSequenceClassificationNr5   r6   r7   r1   r1   r1   r2   r     s   r   c                   @   s   e Zd ZdS )Qwen2ForTokenClassificationNr   r1   r1   r1   r2   r     s   r   c                   @   s   e Zd ZdZdS )Qwen2ForQuestionAnsweringZtransformerN)r5   r6   r7   r   r1   r1   r1   r2   r     s   r   )r   r   r   r   r   r   r   )Nr   )rS   )Btypingr   r   r   r>   r   Zactivationsr   Zcache_utilsr   r	   Z
generationr
   Zintegrationsr   Zmasking_utilsr   r   Zmodeling_flash_attention_utilsr   Zmodeling_layersr   r   r   r   Zmodeling_outputsr   r   Zmodeling_rope_utilsr   r   Zmodeling_utilsr   r   Zprocessing_utilsr   utilsr   r   r   Zutils.deprecationr   Zutils.genericr   Zconfiguration_qwen2r    Moduler!   rA   rH   r   r   rR   r   rj   rk   r   r   r   r   r   r   r   r   r   __all__r1   r1   r1   r2   <module>   s^   
 @/$\K