o
    iZ                     @   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	 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 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 Z/dej0de1dej0fddZ2	d?dej-d ej0d!ej0d"ej0d#eej0 d$e3d%e3d&e$e& fd'd(Z4d@d)d*Z5G d+d, d,ej-Z6ed-G d.d/ d/ej-Z7G d0d1 d1eZ8e'G d2d3 d3e"Z9G d4d5 d5ej-Z:e'G d6d7 d7e9Z;e'G d8d9 d9e9eZ<G d:d; d;ee9Z=G d<d= d=ee9Z>g d>Z?dS )A    )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) GenericForSequenceClassificationGenericForTokenClassificationGradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tuple)deprecate_kwarg   )
Phi3Configc                       s2   e Zd Z fddZdejdejfddZ  ZS )Phi3MLPc                    sP   t    || _tj|jd|j dd| _tj|j|jdd| _t	|j
 | _d S )N   Fbias)super__init__configr   Linearhidden_sizeintermediate_sizegate_up_proj	down_projr   
hidden_actactivation_fnselfr&   	__class__ Z/home/ubuntu/.local/lib/python3.10/site-packages/transformers/models/phi3/modeling_phi3.pyr%   3   s
   
zPhi3MLP.__init__hidden_statesreturnc                 C   s4   |  |}|jddd\}}|| | }| |S )Nr!   dim)r*   chunkr-   r+   )r/   r4   	up_statesgater2   r2   r3   forward;   s   

zPhi3MLP.forward)__name__
__module____qualname__r%   torchFloatTensorr<   __classcell__r2   r2   r0   r3   r    2   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..Nr6   r!   r7   )shaper@   cat)xx1x2r2   r2   r3   rotate_halfD   s   rH   r4   n_repr5   c                 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)rC   expandreshape)r4   rI   batchnum_key_value_headsslenhead_dimr2   r2   r3   	repeat_kvK   s
   0rP           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   r6   )r8   dtype)ptrainingr   )rP   num_key_value_groupsr@   matmul	transposerC   r   
functionalsoftmaxfloat32tor[   rX   r]   
contiguous)rR   rS   rT   rU   rV   rW   rX   rY   
key_statesvalue_statesattn_weightscausal_maskattn_outputr2   r2   r3   eager_attention_forwardW   s   
&rk   c                 C   s   | |}| |}|jd }| dd|f | d|df }}|dd|f |d|df }	}
tj|| t||  |gdd}tj|	| t|	|  |
gdd}||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.
    r6   .Nr7   )	unsqueezerC   r@   rD   rH   )qkcossinposition_idsunsqueeze_dim
rotary_dimq_rotq_passk_rotk_passq_embedk_embedr2   r2   r3   apply_rotary_pos_embq   s   


""""rz   c                       s   e Zd ZdZddede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
 eee	j
  f fddZ  ZS )Phi3Attentionz=Multi-headed attention from 'Attention Is All You Need' paperNr&   	layer_idxc                    s   t    || _|| _t|d|j|j | _|j|j | _	|j| _| jd | _
|j| _d| _|j| j d|j| j   }tj|j| j |jdd| _tj|j|dd| _d S )NrO   g      Tr!   Fr"   )r$   r%   r&   r|   getattrr(   num_attention_headsrO   rM   r^   rW   attention_dropout	is_causalr   r'   o_projqkv_proj)r/   r&   r|   op_sizer0   r2   r3   r%      s   
zPhi3Attention.__init__past_key_valuepast_key_values4.58new_nameversionr4   position_embeddingsrV   cache_positionrY   r5   c                 K   s~  |j d d }g |d| jR }| |}	| jj| j }
|	dd |
f }|	d|
|
| j| j  f }|	d|
| j| j  d f }||dd}||dd}||dd}|\}}t||||\}}|d ur~|||d}|	||| j
|\}}t}| jjdkrt| jj }|| ||||f| jsdn| j| jt| jdd d	|\}}|jg |dR   }| |}||fS )
Nr6   .r   r!   )rp   ro   r   eagerrQ   sliding_window)rX   rW   r   )rC   rO   r   r&   r~   rM   viewr`   rz   updater|   rk   _attn_implementationr   r]   r   rW   r}   rK   re   r   )r/   r4   r   rV   r   r   rY   input_shapehidden_shapeqkv	query_posquery_statesrf   rg   ro   rp   cache_kwargsattention_interfacerj   rh   r2   r2   r3   r<      sD   

	

zPhi3Attention.forwardN)NN)r=   r>   r?   __doc__r   r   intr%   r   r@   Tensortupler	   
LongTensorr   r   r<   rB   r2   r2   r0   r3   r{      s*    r{   RMSNormc                       s.   e Zd Zd fdd	Zdd Zdd Z  ZS )	Phi3RMSNormư>c                    s&   t    tt|| _|| _dS )z:
        Phi3RMSNorm is equivalent to T5LayerNorm
        N)r$   r%   r   	Parameterr@   onesweightvariance_epsilon)r/   r(   epsr0   r2   r3   r%      s   

zPhi3RMSNorm.__init__c                 C   sJ   |j }|tj}|djddd}|t|| j  }| j|| S )Nr!   r6   T)keepdim)	r[   rd   r@   rc   powmeanrsqrtr   r   )r/   r4   input_dtypevariancer2   r2   r3   r<      s
   zPhi3RMSNorm.forwardc                 C   s   t | jj d| j S )Nz, eps=)r   r   rC   r   )r/   r2   r2   r3   
extra_repr   s   zPhi3RMSNorm.extra_repr)r   )r=   r>   r?   r%   r<   r   rB   r2   r2   r0   r3   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eje	eejejf  f fddZ  ZS )Phi3DecoderLayerr&   r|   c                    st   t    |j| _t||d| _t|| _t|j|jd| _	t|j|jd| _
|| _t|j| _t|j| _d S )N)r&   r|   r   )r$   r%   r(   r{   	self_attnr    mlpr   rms_norm_epsinput_layernormpost_attention_layernormr&   r   Dropoutresid_pdropresid_attn_dropoutresid_mlp_dropout)r/   r&   r|   r0   r2   r3   r%      s   

zPhi3DecoderLayer.__init__r   r   r   r   NFr4   rV   rq   	use_cacher   r   rY   r5   c              
   K   sj   |}	|  |}| jd|||||||d|\}}
|	| | }|}	| |}| |}|	| | }|S )N)r4   rV   rq   r   r   r   r   r2   )r   r   r   r   r   r   )r/   r4   rV   rq   r   r   r   r   rY   residualself_attn_weightsr2   r2   r3   r<      s&   




zPhi3DecoderLayer.forward)NNNFNN)r=   r>   r?   r   r   r%   r   r@   r   r   r   r	   boolr   r   r   rA   r<   rB   r2   r2   r0   r3   r      s8    	
r   c                   @   sL   e Zd ZU eed< dZdZdgZdgZdZ	dZ
dZdZdZeedZdZdS )	Phi3PreTrainedModelr&   modelTr   r   )r4   
attentionsz0.0.5N)r=   r>   r?   r   __annotations__base_model_prefixsupports_gradient_checkpointing_no_split_modules_skip_keys_device_placement_supports_flash_attn_supports_sdpa_supports_flex_attn_can_compile_fullgraph_supports_attention_backendr   r{   _can_record_outputs_versionr2   r2   r2   r3   r     s   
 r   c                       sD   e Zd ZU ejed< ddef fddZe e	dd Z
  ZS )	Phi3RotaryEmbeddinginv_freqNr&   c                    s   t    t|drt|jtr|jd|jd| _nd| _|j| _	|j| _
|| _t| j | _| | j|\}| _| jd|dd | j| _d S )Nrope_scaling	rope_typetypedefaultr   F)
persistent)r$   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)r/   r&   devicer   r0   r2   r3   r%   1  s   
zPhi3RotaryEmbedding.__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   r6   r   mpscpuF)device_typeenabledr!   r7   )r[   )r   floatrJ   rC   rd   r   r   r   strr@   autocastr`   rD   ro   r   rp   r[   )
r/   rE   rq   inv_freq_expandedposition_ids_expandedr   freqsembro   rp   r2   r2   r3   r<   B  s   0&zPhi3RotaryEmbedding.forwardr   )r=   r>   r?   r@   r   r   r   r%   no_gradr   r<   rB   r2   r2   r0   r3   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 )	Phi3Modelr&   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 r2   )r   ).0r|   r&   r2   r3   
<listcomp>[  s    z&Phi3Model.__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_initr.   r0   r   r3   r%   T  s   zPhi3Model.__init__N	input_idsrV   rq   r   inputs_embedsr   r   rY   r5   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_embedsrV   r   r   rq   )rV   rq   r   r   r   r   )last_hidden_stater   )
ValueErrorr   r
   r&   get_seq_lengthr@   arangerC   r   rl   r   r   r   r   r   r   r   r   )r/   r   rV   rq   r   r   r   r   rY   past_seen_tokensmask_functionri   r4   r   decoder_layerr2   r2   r3   r<   d  sX   

	

zPhi3Model.forward)NNNNNNN)r=   r>   r?   r   r%   r   r   r   r@   r   r   r	   rA   r   r   r   r   r<   rB   r2   r2   r0   r3   r   R  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							d fdd	Z  ZS )Phi3ForCausalLMzlm_head.weightlm_headcolwise_repr4   logitsc                    s@   t  | t|| _|j| _tj|j|jdd| _| 	  d S )NFr"   )
r$   r%   r   r   r   r   r'   r(   r  r   r.   r0   r2   r3   r%     s
   
zPhi3ForCausalLM.__init__Nr   r   rV   rq   r   r   labelsr   r   logits_to_keeprY   r5   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, Phi3ForCausalLM

        >>> model = Phi3ForCausalLM.from_pretrained("meta-phi3/Phi3-2-7b-hf")
        >>> tokenizer = AutoTokenizer.from_pretrained("meta-phi3/Phi3-2-7b-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   rV   rq   r   r   r   r   N)r  r	  r   )lossr  r   r4   r   r2   )r   r   r   r   slicer  loss_functionr&   r   r   r   r4   r   )r/   r   rV   rq   r   r   r	  r   r   r
  rY   outputsr4   slice_indicesr  r  r2   r2   r3   r<     s0    zPhi3ForCausalLM.forwardTc	                    sb   |r| j jr|jd | j jd kr|d }
|
| j jkrd }t jd||||||||d|	}|S )Nr   r   )r   r   rV   r   r   rq   r   r
  r2   )r&   r   rC    original_max_position_embeddingsr$   prepare_inputs_for_generation)r/   r   r   rV   r   r   rq   r   r
  rY   past_lengthmodel_inputsr0   r2   r3   r    s*   	z-Phi3ForCausalLM.prepare_inputs_for_generation)	NNNNNNNNr   )NNNNNTN)r=   r>   r?   _tied_weights_keys_tp_plan_pp_planr%   r   r   r   r@   r   r   r	   rA   r   r   r   r   r   r   r<   r  rB   r2   r2   r0   r3   r    s^    		
=r  c                   @      e Zd ZdS )Phi3ForSequenceClassificationNr=   r>   r?   r2   r2   r2   r3   r        r  c                   @   r  )Phi3ForTokenClassificationNr  r2   r2   r2   r3   r    r  r  )r   r   r  r  r  )rQ   )Nr   )@typingr   r   r   r@   r   transformers.utils.genericr   activationsr   cache_utilsr	   r
   
generationr   integrationsr   masking_utilsr   r   modeling_flash_attention_utilsr   modeling_layersr   r   r   modeling_outputsr   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   r   utils.deprecationr   configuration_phi3r   Moduler    rH   r   r   rP   r   rk   rz   r{   r   r   r   r   r   r  r  r  __all__r2   r2   r2   r3   <module>   sl   

 F.$Or