
    qi_                        d dl mZ d dl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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-m.Z. ddl/m0Z0 ddl1m2Z2  G d dejf                        Z4 G d dejf                        Z5 G d dejf                        Z6d Z7 ed      d;d       Z8dejr                  d e:d!ejr                  fd"Z;	 	 	 d<d#ejf                  d$ejr                  d%ejr                  d&ejr                  d'ejr                  dz  d(e<d)e<dz  d*e<dz  d!e=ejr                  ejr                  f   fd+Z> ee8       G d, d-ejf                               Z? G d. d/e      Z@e* G d0 d1e%             ZAe* G d2 d3eA             ZBe* G d4 d5eAe             ZC G d6 d7eeA      ZD G d8 d9eeA      ZEg d:ZFy)=    )Callable)OptionalN   )initialization)ACT2FN)CacheDynamicCache)GenerationMixin)use_kernel_func_from_hubuse_kernelized_func)create_causal_mask!create_sliding_window_causal_mask)FlashAttentionKwargs) GenericForSequenceClassificationGenericForTokenClassificationGradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tuple)maybe_autocastmerge_with_config_defaults)capture_outputs   )Gemma2Configc                   <     e Zd Zddedef fdZd Zd Zd Z xZ	S )Gemma2RMSNormdimepsc                     t         |           || _        t        j                  t        j                  |            | _        y N)super__init__r%   nn	Parametertorchzerosweight)selfr$   r%   	__class__s      \/opt/pipecat/venv/lib/python3.12/site-packages/transformers/models/gemma2/modeling_gemma2.pyr)   zGemma2RMSNorm.__init__2   s.    ll5;;s#34    c                     |t        j                  |j                  d      j                  dd      | j                  z         z  S )N   T)keepdim)r,   rsqrtpowmeanr%   )r/   xs     r1   _normzGemma2RMSNorm._norm7   s4    5;;quuQx}}R}>IJJJr2   c                     | j                  |j                               }|d| j                  j                         z   z  }|j                  |      S )N      ?)r;   floatr.   type_as)r/   r:   outputs      r1   forwardzGemma2RMSNorm.forward:   sC    AGGI& 3!2!2!445~~a  r2   c                 ^    t        | j                  j                         d| j                   S )Nz, eps=)tupler.   shaper%   )r/   s    r1   
extra_reprzGemma2RMSNorm.extra_reprA   s'    ))*+6$((<<r2   )gư>)
__name__
__module____qualname__intr>   r)   r;   rA   rE   __classcell__r0   s   @r1   r#   r#   1   s&    5C 5e 5
K!=r2   r#   c                   $     e Zd Z fdZd Z xZS )	Gemma2MLPc                    t         |           || _        |j                  | _        |j                  | _        t        j                  | j                  | j                  d      | _        t        j                  | j                  | j                  d      | _        t        j                  | j                  | j                  d      | _	        t        |j                     | _        y NFbias)r(   r)   confighidden_sizeintermediate_sizer*   Linear	gate_projup_proj	down_projr   hidden_activationact_fnr/   rR   r0   s     r1   r)   zGemma2MLP.__init__F   s    !--!'!9!94#3#3T5K5KRWXyy!1!143I3IPUV4#9#94;K;KRWXV556r2   c                     | j                  | j                  | j                  |            | j                  |      z        }|S r'   )rX   rZ   rV   rW   )r/   r:   rX   s      r1   rA   zGemma2MLP.forwardP   s6    NN4;;t~~a/@#ADLLQRO#ST	r2   )rF   rG   rH   r)   rA   rJ   rK   s   @r1   rM   rM   E   s    7r2   rM   c                        e Zd ZU ej                  ed<   ddef fdZe	 	 	 ddedz  de	d   de
dz  ded	ef   fd
       Z ej                         ed               Z xZS )Gemma2RotaryEmbeddinginv_freqNrR   c                    t         |           |j                  | _        |j                  | _        || _        | j
                  j                  d   | _        | j                  }| j                  dk7  rt        | j                     } || j
                  |      \  }| _
        | j                  d|d       | j                  d|j                         d       y )N	rope_typedefaultr_   F)
persistentoriginal_inv_freq)r(   r)   max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenrR   rope_parametersra   compute_default_rope_parametersr   attention_scalingregister_bufferclone)r/   rR   devicerope_init_fnr_   r0   s        r1   r)   zGemma2RotaryEmbedding.__init__X   s    "("@"@$*$B$B!44[A!%!E!E>>Y&.t~~>L+7V+L($(ZeD0(..2BuUr2   rm   ztorch.deviceseq_lenreturnztorch.Tensorc                    | j                   d   }t        | dd      xs | j                  | j                  z  }d}d|t	        j
                  d|dt        j                        j                  |t        j                        |z  z  z  }||fS )	a  
        Computes the inverse frequencies according to the original RoPE implementation
        Args:
            config ([`~transformers.PreTrainedConfig`]):
                The model configuration.
            device (`torch.device`):
                The device to use for initialization of the inverse frequencies.
            seq_len (`int`, *optional*):
                The current sequence length. Unused for this type of RoPE.
        Returns:
            Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the
            post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE).
        
rope_thetahead_dimNr=   r   r4   dtype)rm   ru   )	rh   getattrrS   num_attention_headsr,   arangeint64tor>   )rR   rm   ro   baser$   attention_factorr_   s          r1   ri   z5Gemma2RotaryEmbedding.compute_default_rope_parametersh   s    & %%l3fj$/c63E3EIcIc3c U\\!S!5;;?BB&X]XcXcBdgjjk
 )))r2   c                 N   | j                   d d d d f   j                         j                  |j                  d   dd      j	                  |j
                        }|d d d d d f   j                         }t        |j
                  j                  t              r/|j
                  j                  dk7  r|j
                  j                  nd}t        |d      5  |j                         |j                         z  j                  dd      }t        j                  ||fd	      }|j                         | j                  z  }|j                         | j                  z  }	d d d        j	                  |j                   
      	j	                  |j                   
      fS # 1 sw Y   AxY w)Nr   r5   r    mpscpuF)device_typeenabledr4   r$   rt   )r_   r>   expandrD   rz   rm   
isinstancetypestrr   	transposer,   catcosrj   sinru   )
r/   r:   position_idsinv_freq_expandedposition_ids_expandedr   freqsembr   r   s
             r1   rA   zGemma2RotaryEmbedding.forward   sR    !MM$4-8>>@GGHZHZ[\H]_acdehhijiqiqr ,QaZ 8 > > @'1!((--'E!((--[`J`ahhmmfkUC 	5&,,.1F1L1L1NNYYZ[]^_E))UEN3C'')d444C'')d444C		5 vvAGGv$cff177f&;;;	5 	5s   BFF$r'   )NNN)rF   rG   rH   r,   Tensor__annotations__r!   r)   staticmethodr   rI   rC   r>   ri   no_gradr   rA   rJ   rK   s   @r1   r^   r^   U   s    llV| V  &*+/"*t#*(* t* 
~u$	%	* *: U]]_<  <r2   r^   c                     | dd| j                   d   dz  f   }| d| j                   d   dz  df   }t        j                  | |fd      S )z*Rotates half the hidden dims of the input..Nr5   r4   r   )rD   r,   r   )r:   x1x2s      r1   rotate_halfr      sZ    	
3"!''"+"""	#B	
3q ""	#B99rc2YB''r2   rotary_pos_embc                     |j                  |      }|j                  |      }| |z  t        |       |z  z   }||z  t        |      |z  z   }||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.
        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kr   r   unsqueeze_dimq_embedk_embeds          r1   apply_rotary_pos_embr      sY    & --
&C
--
&C3w;q>C/0G3w;q>C/0GGr2   hidden_statesn_reprp   c                     | j                   \  }}}}|dk(  r| S | dddddddddf   j                  |||||      } | j                  |||z  ||      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)rD   r   reshape)r   r   batchnum_key_value_headsslenrs   s         r1   	repeat_kvr      so    
 2?1D1D.Ehz!!Qa"23::5BUW\^bdlmM  (;e(CT8TTr2   modulequerykeyvalueattention_maskdropoutscalingsoftcapc                 |   || j                   dz  }t        || j                        }	t        || j                        }
t        j                  ||	j                  dd            |z  }|||z  }t        j                  |      }||z  }|||z   }t        j                  j                  |dt        j                        j                  |j                        }t        j                  j                  ||| j                        }t        j                  ||
      }|j                  dd      j                         }||fS )N      r4   r   r5   )r$   ru   )ptrainingr    )rs   r   num_key_value_groupsr,   matmulr   tanhr*   
functionalsoftmaxfloat32rz   ru   r   r   
contiguous)r   r   r   r   r   r   r   r   kwargs
key_statesvalue_statesattn_weightsattn_outputs                r1   eager_attention_forwardr      s    //4'3 ; ;<JUF$?$?@L<<z';';Aq'ABWLL#g-zz,/#g-!#n4 ==((2U]](SVVW\WbWbcL==((6??([L,,|\:K''1-88:K$$r2   c                   @    e Zd ZdZdedef fdZ	 	 	 	 ddej                  de	ej                  ej                  f   dz  dej                  dz  d	e
dz  d
ej                  dz  dee   de	ej                  ej                  dz  e	ej                     dz  f   fdZ xZS )Gemma2Attentionz=Multi-headed attention from 'Attention Is All You Need' paperrR   	layer_idxc                 Z   t         |           t        |d      r|j                  |   nd | _        || _        || _        t        |d|j                  |j                  z        | _
        |j                  |j                  z  | _        |j                  dz  | _        | j
                  j                  | _        t        |dd       | _        t#        j$                  |j                  |j                  | j                  z  |j&                        | _        t#        j$                  |j                  |j                  | j                  z  |j&                        | _        t#        j$                  |j                  |j                  | j                  z  |j&                        | _        t#        j$                  |j                  | j                  z  |j                  |j&                        | _        | j
                  j0                  | _        | j                  dk(  r|j2                  | _        y d | _        y )Nlayer_typesrs   r   use_bidirectional_attentionFrP   sliding_attention)r(   r)   hasattrr   
layer_typerR   r   rv   rS   rw   rs   r   r   query_pre_attn_scalarr   attention_dropout	is_causalr*   rU   attention_biasq_projk_projv_projo_projattn_logit_softcappingsliding_windowr/   rR   r   r0   s      r1   r)   zGemma2Attention.__init__   s   ;B6=;Y&,,Y7_c"
F4F4F&JdJd4de$*$>$>&B\B\$\!33T9!%!>!>$V-JERRii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii&&68J8JQWQfQf
 '+kk&H&H#7;J]7]f33cgr2   Nr   position_embeddingsr   past_key_valuescache_positionr   rp   c                 D   |j                   d d }g |d| j                  }| j                  |      j                  |      j	                  dd      }	| j                  |      j                  |      j	                  dd      }
| j                  |      j                  |      j	                  dd      }|\  }}t        |	|
||      \  }	}
|'|||d}|j                  |
|| j                  |      \  }
}t        j                  | j                  j                  t              } || |	|
||f| j                  r| j                   nd| j"                  | j$                  | j&                  d|\  }} |j(                  g |d j+                         }| j-                  |      }||fS )Nr5   r    r4   )r   r   r           )r   r   r   r   )rD   rs   r   viewr   r   r   r   updater   r   get_interfacerR   _attn_implementationr   r   r   r   r   r   r   r   r   )r/   r   r   r   r   r   r   input_shapehidden_shapequery_statesr   r   r   r   cache_kwargsattention_interfacer   r   s                     r1   rA   zGemma2Attention.forward  s    $))#2.88b8$--8{{=166|DNNqRST[[/44\BLLQPQR
{{=166|DNNqRST&S#7jRUWZ#[ j&#&snUL'6'='=j,X\XfXfht'u$J(?(M(MKK,,.E)
 %8%
 /3mmD**LL..//%
 %
!\ *k));;;;FFHkk+.L((r2   )NNNN)rF   rG   rH   __doc__r!   rI   r)   r,   r   rC   r   
LongTensorr   r   rA   rJ   rK   s   @r1   r   r      s    Gh| h h: IM.2(,26+)||+) #5<<#=>E+) t+	+)
 +) ((4/+) -.+) 
u||U\\D0%2E2LL	M+)r2   r   c                   N    e Zd Zdedef fdZ	 	 	 	 	 ddej                  deej                  ej                  f   dz  dej                  dz  dej                  dz  d	e
dz  d
ej                  dz  deej                  eej                  ej                  f   dz  f   fdZ xZS )Gemma2DecoderLayerrR   r   c                    t         |           |j                  | _        || _        |j                  |   | _        t        ||      | _        t        |      | _	        t        |j                  |j                        | _        t        |j                  |j                        | _        t        |j                  |j                        | _        t        |j                  |j                        | _        y )N)rR   r   r%   )r(   r)   rS   rR   r   attention_typer   	self_attnrM   mlpr#   rms_norm_epsinput_layernormpost_attention_layernormpre_feedforward_layernormpost_feedforward_layernormr   s      r1   r)   zGemma2DecoderLayer.__init__2  s    !--$00;()LV$,V-?-?VEXEXY(5f6H6HfNaNa(b%)6v7I7IvObOb)c&*78J8JPVPcPc*d'r2   Nr   r   r   r   r   r   rp   c           
         |}| j                  |      } | j                  d||||||d|\  }}	| j                  |      }||z   }|}| j                  |      }| j	                  |      }| j                  |      }||z   }|S )N)r   r   r   r   r   r    )r   r   r   r   r   r   )
r/   r   r   r   r   r   r   r   residual_s
             r1   rA   zGemma2DecoderLayer.forward?  s     !,,]; *4>> 
' 3)%+)
 
q 55mD =0 66}E/77F =0r2   )NNNNN)rF   rG   rH   r!   rI   r)   r,   r   rC   r   r   FloatTensorrA   rJ   rK   s   @r1   r   r   1  s    e| e e  IM.204(,26!||! #5<<#=>E! t+	!
 &&-! ! ((4/! 
u  %(9(95;L;L(L"MPT"TT	U!r2   r   c                        e Zd ZU eed<   dZdZdgZdgZdZ	dZ
dZdZdZeedZ ej$                          fd       Z xZS )Gemma2PreTrainedModelrR   modelTr   r   )r   
attentionsc                     t         |   |       d|j                  j                  v r t	        j
                  |j                         y y )NRMSNorm)r(   _init_weightsr0   rF   initzeros_r.   )r/   r   r0   s     r1   r   z#Gemma2PreTrainedModel._init_weightsu  s9    f%((111KK& 2r2   )rF   rG   rH   r!   r   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   rJ   rK   s   @r1   r   r   c  sn    &*#-.#4"5N!"&+%
 U]]_' 'r2   r   c                       e Zd Zdef fdZeee	 	 	 	 	 	 	 ddej                  dz  dej                  dz  dej                  dz  dedz  dej                  dz  d	edz  d
ej                  dz  dee   defd                     Z xZS )Gemma2ModelrR   c           	      
   t         |   |       |j                  | _        |j                  | _        t        j                  |j                  |j                  | j                        | _        t        j                  t        |j                        D cg c]  }t        ||       c}      | _        t        |j                  |j                        | _        t#        |      | _        d| _        | j)                          y c c}w )Nr   F)r(   r)   pad_token_idpadding_idx
vocab_sizer*   	EmbeddingrS   embed_tokens
ModuleListrangenum_hidden_layersr   layersr#   r   normr^   
rotary_embgradient_checkpointing	post_initr   s      r1   r)   zGemma2Model.__init__  s     !.. ++LL):):F<N<NPTP`P`ammDI&JbJbDcdy	2d
 "&"4"4&:M:MN	/7&+# 	 es   D N	input_idsr   r   r   inputs_embeds	use_cacher   r   rp   c           
         |d u |d uz  rt        d      || j                  |      }|r|t        | j                        }|F||j	                         nd}	t        j                  |	|	|j                  d   z   |j                        }||j                  d      }t        |x}
t              s*| j                  |||||d}t        di |t        di |d}
|}| j                  ||      }t        j                  | j                  j                   dz  |j"                  	      }||z  }| j$                  d | j                  j&                   D ]  } ||f|
|j(                     ||||d
|}  | j+                  |      }t-        ||      S )Nz:You must specify exactly one of input_ids or inputs_embeds)rR   r   r    )rm   )rR   r  r   r   r   r   )full_attentionr   g      ?rt   )r   r   r   r   r   )last_hidden_stater   r   )
ValueErrorr  r	   rR   get_seq_lengthr,   rx   rD   rm   r   r   dictr   r   r  tensorrS   ru   r  r  r   r  r   )r/   r  r   r   r   r  r  r   r   past_seen_tokenscausal_mask_mappingmask_kwargsr   r   
normalizerdecoder_layers                   r1   rA   zGemma2Model.forward  s    -t";<YZZ *.*;*;I*FM0*$++>O!CRC^==?de"\\ "2]5H5H5K"KTaThThN )33A6L ?-F ++!."0"0#2 ,K #5"C{"C%F%U%U# &"oom\J
 \\$++"9"93">mFYFYZ
%
2![[)H4;;+H+HI 		M)2=3O3OP$7) /- M		 		-0&++
 	
r2   )NNNNNNN)rF   rG   rH   r!   r)   r   r   r   r,   r   r   r   r   boolr   r   r   rA   rJ   rK   s   @r1   r  r  }  s    |     .2.204(,26!%26H
##d*H
 t+H
 &&-	H

 H
 ((4/H
 $;H
 ((4/H
 +,H
 
!H
    H
r2   r  c                   b    e Zd ZddiZddiZddgdgfiZ fdZee	 	 	 	 	 	 	 	 	 dd	e	j                  dz  d
e	j                  dz  de	j                  dz  dedz  de	j                  dz  de	j                  dz  dedz  de	j                  dz  dee	j                  z  dee   defd              Z xZS )Gemma2ForCausalLMzlm_head.weightzmodel.embed_tokens.weightlm_headcolwise_gather_outputr   logitsc                     t         |   |       t        |      | _        |j                  | _        t        j                  |j                  |j                  d      | _        | j                          y rO   )
r(   r)   r  r   r  r*   rU   rS   r,  r  r[   s     r1   r)   zGemma2ForCausalLM.__init__  sU      (
 ++yy!3!3V5F5FUS 	r2   Nr  r   r   r   r  labelsr  r   logits_to_keepr   rp   c
                     | j                   d|||||||d|
}|j                  }t        |	t              rt	        |	 d      n|	}| j                  |dd|ddf         }| j                  j                  G|| j                  j                  z  }t        j                  |      }|| j                  j                  z  }d}| | j                  ||| j                  fi |
}t        |||j                  |j                  |j                        S )a  
        Example:

        ```python
        >>> from transformers import AutoTokenizer, Gemma2ForCausalLM

        >>> model = Gemma2ForCausalLM.from_pretrained("google/gemma-2-9b")
        >>> tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-9b")

        >>> 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   r   r   r  r  r   N)lossr.  r   r   r   r   )r   r  r   rI   slicer,  rR   final_logit_softcappingr,   r   loss_functionr  r   r   r   r   )r/   r  r   r   r   r  r0  r  r   r1  r   outputsr   slice_indicesr.  r3  s                   r1   rA   zGemma2ForCausalLM.forward  s   B ,64:: 	,
)%+')	,
 	,
  118B>SV8W~ot4]kmA}a,?@A;;..:dkkAAAFZZ'FdkkAAAF%4%%ffdooPPD%#33!//))
 	
r2   )	NNNNNNNNr   )rF   rG   rH   _tied_weights_keys_tp_plan_pp_planr)   r   r   r,   r   r   r   r   r)  rI   r   r   r   rA   rJ   rK   s   @r1   r+  r+    s/   *,GH23H_-z:;H  .2.204(,26*.!%26-.=
##d*=
 t+=
 &&-	=

 =
 ((4/=
   4'=
 $;=
 ((4/=
 ell*=
 +,=
 
 =
  =
r2   r+  c                       e Zd Zy)Gemma2ForSequenceClassificationNrF   rG   rH   r   r2   r1   r=  r=  .      r2   r=  c                       e Zd Zy)Gemma2ForTokenClassificationNr>  r   r2   r1   rA  rA  2  r?  r2   rA  )r+  r  r   r=  rA  )r    )r   NN)Gcollections.abcr   typingr   r,   torch.nnr*    r   r   activationsr   cache_utilsr   r	   
generationr
   integrationsr   r   masking_utilsr   r   modeling_flash_attention_utilsr   modeling_layersr   r   r   modeling_outputsr   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   r   utils.genericr   r   utils.output_capturingr   configuration_gemma2r!   Moduler#   rM   r^   r   r   r   rI   r   r>   rC   r   r   r   r   r  r+  r=  rA  __all__r   r2   r1   <module>rW     s  * %    & ! . ) I R B 
 P K F & I I G 5 .=BII =(		  ><BII ><B( *+ ,2	UU\\ 	U# 	U%,, 	U$   %II%<<% 
% <<	%
 LL4'% % T\% T\% 5<<%&%D )*H)bii H) +H)V/3 /d 'O ' '2 \
' \
 \
~ M
- M
 M
`	&FH] 		#@BW 	r2   