o
    	۷iS                     @   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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$m%Z% ddl&m'Z' G dd dej(Z)G dd dej(Z*G dd dej(Z+dd Z,d<ddZ-dej.de/dej.fd d!Z0	"d=d#ej(d$ej.d%ej.d&ej.d'eej. d(e1d)e1d*ee fd+d,Z2G d-d. d.ej(Z3G d/d0 d0eZ4e G d1d2 d2eZ5e G d3d4 d4e5Z6e G d5d6 d6e5eZ7G d7d8 d8ee5Z8G d9d: d:ee5Z9g d;Z:dS )>    )CallableOptionalUnionN)nn   )ACT2FN)CacheDynamicCache)GenerationMixin)create_causal_mask) 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   )GemmaConfigc                       s@   e Zd Zddedef fddZdd Zdd	 Zd
d Z  Z	S )GemmaRMSNormư>dimepsc                    s&   t    || _tt|| _d S N)super__init__r    r   	Parametertorchzerosweight)selfr   r    	__class__ ^/home/ubuntu/vllm_env/lib/python3.10/site-packages/transformers/models/gemma/modeling_gemma.pyr#   /   s   
zGemmaRMSNorm.__init__c                 C   s$   |t |djddd| j  S )N   T)keepdim)r%   rsqrtpowmeanr    )r(   xr+   r+   r,   _norm4   s   $zGemmaRMSNorm._normc                 C   s*   |  | }|d| j   }||S )Ng      ?)r4   floatr'   type_as)r(   r3   outputr+   r+   r,   forward7   s   
zGemmaRMSNorm.forwardc                 C   s   t | jj d| j S )Nz, eps=)tupler'   shaper    )r(   r+   r+   r,   
extra_repr>   s   zGemmaRMSNorm.extra_repr)r   )
__name__
__module____qualname__intr5   r#   r4   r8   r;   __classcell__r+   r+   r)   r,   r   .   s
    r   c                       s$   e Zd Z fddZdd Z  ZS )GemmaMLPc                    sr   t    || _|j| _|j| _tj| j| jdd| _tj| j| jdd| _tj| j| jdd| _	t
|j | _d S NFbias)r"   r#   confighidden_sizeintermediate_sizer   Linear	gate_projup_proj	down_projr   
hidden_actact_fnr(   rE   r)   r+   r,   r#   C   s   
zGemmaMLP.__init__c                 C   s$   |  | | || | }|S r!   )rK   rM   rI   rJ   )r(   r3   rK   r+   r+   r,   r8   M   s    zGemmaMLP.forward)r<   r=   r>   r#   r8   r@   r+   r+   r)   r,   rA   B   s    
rA   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 )	GemmaRotaryEmbeddinginv_freqNrE   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defaultrP   F)
persistent)r"   r#   hasattr
isinstancerQ   dictgetrR   max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenrE   r   rope_init_fnattention_scalingregister_bufferrP   original_inv_freq)r(   rE   devicerP   r)   r+   r,   r#   U   s   
zGemmaRotaryEmbedding.__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   dtype)rP   r5   expandr:   tora   rW   rS   strr%   autocast	transposecatcosr^   sinrh   )
r(   r3   position_idsinv_freq_expandedposition_ids_expandedrd   freqsembro   rp   r+   r+   r,   r8   f   s   0&zGemmaRotaryEmbedding.forwardr!   )r<   r=   r>   r%   Tensor__annotations__r   r#   no_gradr   r8   r@   r+   r+   r)   r,   rO   R   s   
 
rO   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..Nr.   r-   rf   )r:   r%   rn   )r3   x1x2r+   r+   r,   rotate_halfv   s   r{   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.
    )	unsqueezer{   )qkro   rp   rq   unsqueeze_dimq_embedk_embedr+   r+   r,   apply_rotary_pos_emb}   s
   

r   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:   ri   reshape)r   r   batchnum_key_value_headsslenhead_dimr+   r+   r,   	repeat_kv   s
   0r           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   rh   )ptrainingr   )r   num_key_value_groupsr%   matmulrm   r:   r   
functionalsoftmaxfloat32rj   rh   r   r   
contiguous)r   r   r   r   r   r   r   r   
key_statesvalue_statesattn_weightscausal_maskattn_outputr+   r+   r,   eager_attention_forward   s   
&r   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j	f fddZ  ZS )GemmaAttentionz=Multi-headed attention from 'Attention Is All You Need' paperrE   	layer_idxc                    s   t    || _|| _t|d|j|j | _|j|j | _	| jd | _
|j| _d| _tj|j|j| j |jd| _tj|j|j| j |jd| _tj|j|j| j |jd| _tj|j| j |j|jd| _d S )Nr   g      TrC   )r"   r#   rE   r   getattrrF   num_attention_headsr   r   r   r   attention_dropout	is_causalr   rH   attention_biasq_projk_projv_projo_projr(   rE   r   r)   r+   r,   r#      s(   
zGemmaAttention.__init__past_key_valuepast_key_values4.58new_nameversionNr   position_embeddingsr   cache_positionr   r   c                 K   s$  |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d|\}}|jg |dR   }| |}||fS )Nr.   r   r-   )rp   ro   r   eagerr   )r   r   )r:   r   r   viewrm   r   r   r   updater   r   rE   _attn_implementationr   r   r   r   r   r   r   )r(   r   r   r   r   r   r   input_shapehidden_shapequery_statesr   r   ro   rp   cache_kwargsattention_interfacer   r   r+   r+   r,   r8      s8   


zGemmaAttention.forward)NN)r<   r=   r>   __doc__r   r?   r#   r   r%   rv   r9   r   r   
LongTensorr   r   r8   r@   r+   r+   r)   r,   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 )GemmaDecoderLayerrE   r   c                    sR   t    |j| _t||d| _t|| _t|j|jd| _	t|j|jd| _
d S )N)rE   r   r    )r"   r#   rF   r   	self_attnrA   mlpr   rms_norm_epsinput_layernormpost_attention_layernormr   r)   r+   r,   r#     s   

zGemmaDecoderLayer.__init__r   r   r   r   NFr   r   rq   	use_cacher   r   r   r   c              
   K   s^   |}	|  |}| jd|||||||d|\}}
|	| }|}	| |}| |}|	| }|S )N)r   r   rq   r   r   r   r   r+   )r   r   r   r   )r(   r   r   rq   r   r   r   r   r   residual_r+   r+   r,   r8     s&   




zGemmaDecoderLayer.forward)NNNFNN)r<   r=   r>   r   r?   r#   r   r%   rv   r   r   r   boolr9   r   r   r8   r@   r+   r+   r)   r,   r     s8    
	
r   c                       sX   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 fddZ  ZS )	GemmaPreTrainedModelrE   modelTr   r   )r   
attentionsc                    s,   t  | d|jjv r|jj  d S d S )NRMSNorm)r"   _init_weightsr*   r<   r'   datazero_)r(   r   r)   r+   r,   r   E  s   z"GemmaPreTrainedModel._init_weights)r<   r=   r>   r   rw   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_outputsr   r@   r+   r+   r)   r,   r   3  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 )
GemmaModelrE   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 r+   )r   ).0r   rE   r+   r,   
<listcomp>V  s    z'GemmaModel.__init__.<locals>.<listcomp>r   r   F)r"   r#   pad_token_idpadding_idx
vocab_sizer   	EmbeddingrF   embed_tokens
ModuleListrangenum_hidden_layerslayersr   r   normrO   
rotary_embgradient_checkpointing	post_initrN   r)   r   r,   r#   O  s   zGemmaModel.__init__N	input_idsr   rq   r   inputs_embedsr   r   r   r   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}t
| j|||||d}
|}| ||}tj| jjd |jd}|| }| jd | jj D ]}||f|
|||||d	|}qr| |}t||r|d
S d d
S )Nz:You must specify exactly one of input_ids or inputs_embedsr   r   r   )ra   )rE   input_embedsr   r   r   rq   g      ?rg   )r   rq   r   r   r   r   )last_hidden_stater   )
ValueErrorr   r	   rE   get_seq_lengthr%   aranger:   ra   r|   r   r   tensorrF   rh   r   r   r   r   )r(   r   r   rq   r   r   r   r   r   past_seen_tokensr   r   r   
normalizerdecoder_layerr+   r+   r,   r8   _  sZ   




zGemmaModel.forward)NNNNNNN)r<   r=   r>   r   r#   r   r   r   r%   r   rv   r   FloatTensorr   r   r   r   r8   r@   r+   r+   r)   r,   r   M  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 )GemmaForCausalLMzlm_head.weightlm_headcolwise_repr   logitsc                    s@   t  | t|| _|j| _tj|j|jdd| _| 	  d S rB   )
r"   r#   r   r   r   r   rH   rF   r   r   rN   r)   r+   r,   r#     s
   
zGemmaForCausalLM.__init__Nr   r   r   rq   r   r   labelsr   r   logits_to_keepr   r   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, GemmaForCausalLM

        >>> model = GemmaForCausalLM.from_pretrained("google/gemma-7b")
        >>> tokenizer = AutoTokenizer.from_pretrained("google/gemma-7b")

        >>> prompt = "What is your favorite condiment?"
        >>> 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]
        "What is your favorite condiment?"
        ```)r   r   rq   r   r   r   r   N)r   r   r   )lossr   r   r   r   r+   )r   r   rW   r?   slicer   loss_functionrE   r   r   r   r   r   )r(   r   r   rq   r   r   r   r   r   r  r   outputsr   slice_indicesr   r  r+   r+   r,   r8     s0    zGemmaForCausalLM.forward)	NNNNNNNNr   )r<   r=   r>   _tied_weights_keys_tp_plan_pp_planr#   r   r   r   r%   r   rv   r   r   r   r   r?   r   r   r   r8   r@   r+   r+   r)   r,   r     sN    		
r   c                   @      e Zd ZdS )GemmaForSequenceClassificationNr<   r=   r>   r+   r+   r+   r,   r        r  c                   @   r
  )GemmaForTokenClassificationNr  r+   r+   r+   r,   r    r  r  )r   r   r  r  r   )Nr   )r   );typingr   r   r   r%   r   activationsr   cache_utilsr   r	   
generationr
   masking_utilsr   modeling_layersr   r   r   modeling_outputsr   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   r   utils.deprecationr   utils.genericr   configuration_gemmar   Moduler   rA   rO   r{   r   rv   r?   r   r5   r   r   r   r   r   r   r  r  __all__r+   r+   r+   r,   <module>   sf   $

G.WK