
    rh                     (   d dl Z d dlmZ d dlmZ d dl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 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' ddl(m)Z)m*Z*m+Z+m,Z,m-Z- ddl.m/Z/ ddl0m1Z1 ddl2m3Z3m4Z4  e-jj                  e6      Z7e e+d       G d de                    Z8e e+d       G d de)                    Z9 G d de
jt                        Z; G d  d!e
jx                        Z= G d" d#e
jx                        Z> G d$ d%e
jx                        Z?d& Z@dNd'ZAd(ej                  d)eCd*ej                  fd+ZD	 	 	 dOd,e
jx                  d-ej                  d.ej                  d/ej                  d0eej                     d1eEd2eeE   d3eeE   d*eFej                  ej                  f   fd4ZG G d5 d6e
jx                        ZH G d7 d8e      ZIe+ G d9 d:e%             ZJe+ G d; d<eJ             ZKe+ G d= d>eJe             ZL G d? d@e
jx                        ZMdAeej                     dBeej                     dCeCd*ee   fdDZN e+dE       G dF dGeJ             ZO e+dH       G dI dJeJe             ZP G dK dLeJ      ZQg dMZRy)P    N)Callable)	dataclass)OptionalUnion   )ACT2FN)CacheDynamicCache)PretrainedConfig)GenerationMixin)create_causal_maskcreate_masks_for_generate!create_sliding_window_causal_mask)FlashAttentionKwargs)GradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast SequenceClassifierOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)ModelOutputTransformersKwargsauto_docstringcan_return_tuplelogging)check_model_inputs   )	AutoModel   )Gemma3ConfigGemma3TextConfigzK
    Base class for Gemma3 outputs, with hidden states and attentions.
    )custom_introc                   :    e Zd ZU dZdZeej                     ed<   y)Gemma3ModelOutputWithPasta  
    past_key_values (`Cache`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
        Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
        `(batch_size, num_heads, sequence_length, embed_size_per_head)`)

        Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
        `past_key_values` input) to speed up sequential decoding.
    image_hidden_states (`torch.FloatTensor`, *optional*):
        A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`.
        image_hidden_states of the model produced by the vision encoder and after projecting the last hidden state.
    Nimage_hidden_states)	__name__
__module____qualname____doc__r(   r   torchFloatTensor__annotations__     }/var/www/html/ai-insurance-compliance-backend/venv/lib/python3.12/site-packages/transformers/models/gemma3/modeling_gemma3.pyr'   r'   2   s    
 8<%"3"34;r1   r'   zR
    Base class for Gemma3 causal language model (or autoregressive) outputs.
    c                      e Zd ZU dZdZeej                     ed<   dZ	eej                     ed<   dZ
eeeej                     ef      ed<   dZeeej                        ed<   dZeeej                        ed<   dZeej                     ed<   y)	Gemma3CausalLMOutputWithPastaa  
    loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided):
        Language modeling loss (for next-token prediction).
    logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.text_config.vocab_size)`):
        Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
    past_key_values (`Cache`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
        Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
        `(batch_size, num_heads, sequence_length, embed_size_per_head)`)

        Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
        `past_key_values` input) to speed up sequential decoding.
    image_hidden_states (`torch.FloatTensor`, *optional*):
        A `torch.FloatTensor` of size `(batch_size, num_images, sequence_length, hidden_size)`.
        image_hidden_states of the model produced by the vision encoder after projecting last hidden state.
    Nlosslogitspast_key_valueshidden_states
attentionsr(   )r)   r*   r+   r,   r5   r   r-   r.   r/   r6   r7   r   listr	   r8   tupler9   r(   r0   r1   r2   r4   r4   H   s      )-D(5$$
%,*.FHU&&'.GKOXeD):):$;U$BCDK8<M8E%"3"345<59Ju001297;%"3"34;r1   r4   c            	       Z     e Zd ZdZd	dedededef fdZdej                  f fdZ	 xZ
S )
Gemma3TextScaledWordEmbeddingz\
    This module overrides nn.Embeddings' forward by multiplying with embeddings scale.
    num_embeddingsembedding_dimpadding_idxembed_scalec                 v    t         |   |||       | j                  dt        j                  |      d       y )NrA   F
persistent)super__init__register_bufferr-   tensor)selfr>   r?   r@   rA   	__class__s        r2   rF   z&Gemma3TextScaledWordEmbedding.__init__l   s3    D]ELL,ERWXr1   	input_idsc                     t         |   |      | j                  j                  | j                  j
                        z  S N)rE   forwardrA   toweightdtype)rI   rK   rJ   s     r2   rN   z%Gemma3TextScaledWordEmbedding.forwardp   s2    wy)D,<,<,?,?@Q@Q,RRRr1   )      ?)r)   r*   r+   r,   intfloatrF   r-   TensorrN   __classcell__rJ   s   @r2   r=   r=   g   sG    Ys Y3 YS Y_d YS S Sr1   r=   c                   *     e Zd Zdef fdZd Z xZS )	Gemma3MLPconfigc                    t         |           || _        |j                  | _        |j                  | _        t        j                  | j                  | j                  d      | _        t        j                  | j                  | j                  d      | _        t        j                  | j                  | j                  d      | _	        t        |j                     | _        y NFbias)rE   rF   rZ   hidden_sizeintermediate_sizennLinear	gate_projup_proj	down_projr   hidden_activationact_fnrI   rZ   rJ   s     r2   rF   zGemma3MLP.__init__u   s    !--!'!9!94#3#3T5K5KRWXyy!1!143I3IPUV4#9#94;K;KRWXV556r1   c                     | j                  | j                  | j                  |            | j                  |      z        }|S rM   )re   rg   rc   rd   )rI   xre   s      r2   rN   zGemma3MLP.forward   s6    NN4;;t~~a/@#ADLLQRO#ST	r1   )r)   r*   r+   r$   rF   rN   rV   rW   s   @r2   rY   rY   t   s    7/ 7r1   rY   c                   <     e Zd Zddedef fdZd Zd Zd Z xZ	S )Gemma3RMSNormdimepsc                     t         |           || _        t        j                  t        j                  |            | _        y rM   )rE   rF   rn   ra   	Parameterr-   zerosrP   )rI   rm   rn   rJ   s      r2   rF   zGemma3RMSNorm.__init__   s.    ll5;;s#34r1   c                     |t        j                  |j                  d      j                  dd      | j                  z         z  S )Nr    T)keepdim)r-   rsqrtpowmeanrn   )rI   rj   s     r2   _normzGemma3RMSNorm._norm   s4    5;;quuQx}}R}>IJJJr1   c                     | j                  |j                               }|d| j                  j                         z   z  }|j                  |      S )NrR   )rx   rT   rP   type_as)rI   rj   outputs      r2   rN   zGemma3RMSNorm.forward   sC    AGGI& 3!2!2!445~~a  r1   c                 ^    t        | j                  j                         d| j                   S )Nz, eps=)r;   rP   shapern   rI   s    r2   
extra_reprzGemma3RMSNorm.extra_repr   s'    ))*+6$((<<r1   )gư>)
r)   r*   r+   rS   rT   rF   rx   rN   r   rV   rW   s   @r2   rl   rl      s&    5C 5e 5
K!=r1   rl   c                   ^     e Zd Zddef fdZ ej                         ed               Z xZ	S )Gemma3RotaryEmbeddingrZ   c                    t         |           t        |d      rUt        |j                  t
              r;|j                  j                  d|j                  j                  d            | _        nd| _        |j                  | _	        |j                  | _
        || _        t        | j                     | _        | j                  | j                  |      \  }| _        | j                  d|d       | j                   | _        y )Nrope_scaling	rope_typetypedefaultinv_freqFrC   )rE   rF   hasattr
isinstancer   dictgetr   max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenrZ   r   rope_init_fnattention_scalingrG   r   original_inv_freq)rI   rZ   devicer   rJ   s       r2   rF   zGemma3RotaryEmbedding.__init__   s    6>*z&:M:Mt/T#0044[&BUBUBYBYZ`BabDN&DN"("@"@$*$B$B!/?+/+<+<T[[&+Q($(ZeD!%r1   c                 b   | 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        j                  |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   rs   r"   mpscpuF)device_typeenabledr    rm   )rQ   )r   rT   expandr}   rO   r   r   r   strr-   autocast	transposecatcosr   sinrQ   )
rI   rj   position_idsinv_freq_expandedposition_ids_expandedr   freqsembr   r   s
             r2   rN   zGemma3RotaryEmbedding.forward   sV    !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.rM   )
r)   r*   r+   r$   rF   r-   no_gradr   rN   rV   rW   s   @r2   r   r      s4    // /" U]]_<  <r1   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..Nrs   r    r   )r}   r-   r   )rj   x1x2s      r2   rotate_halfr      sZ    	
3"!''"+"""	#B	
3q ""	#B99rc2YB''r1   c                     |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.
        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kr   r   r   unsqueeze_dimq_embedk_embeds           r2   apply_rotary_pos_embr      sY    ( --
&C
--
&C3w;q>C/0G3w;q>C/0GGr1   r8   n_repreturnc                     | 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)r}   r   reshape)r8   r   batchnum_key_value_headsslenhead_dims         r2   	repeat_kvr      so    
 2?1D1D.Ehz!!Qa"23::5BUW\^bdlmM  (;e(CT8TTr1   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  }|#|d d d d d d d |	j                  d   f   }||z   }t        j                  j                  |dt        j                        j                  |j                        }t        j                  j                  ||| j                        }t        j                  ||
      }|j                  dd      j!                         }||fS )	N      r    r   rs   )rm   rQ   )ptrainingr"   )r   r   num_key_value_groupsr-   matmulr   tanhr}   ra   
functionalsoftmaxfloat32rO   rQ   r   r   
contiguous)r   r   r   r   r   r   r   r   kwargs
key_statesvalue_statesattn_weightscausal_maskattn_outputs                 r2   eager_attention_forwardr      sA    //4'3 ; ;<JUF$?$?@L<<z';';Aq'ABWLL#g-zz,/#g-!$Q1.D
0@0@0D.D%DE#k1 ==((2U]](SVVW\WbWbcL==((6??([L,,|\:K''1-88:K$$r1   c                       e Zd ZdZdedef fdZ	 	 ddej                  dej                  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Z xZS )Gemma3Attentionz=Multi-headed attention from 'Attention Is All You Need' paperrZ   	layer_idxc                    t         |           |j                  |   dk(  | _        || _        || _        t        |d|j                  |j                  z        | _	        |j                  |j                  z  | _        |j                  dz  | _        | j                  j                  | _        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                  j.                  | _        | j                  r|j0                  nd | _        t3        |j                  |j4                        | _        t3        |j                  |j4                        | _        y )Nsliding_attentionr   r   Tr]   )rm   rn   )rE   rF   layer_types
is_slidingrZ   r   getattrr_   num_attention_headsr   r   r   query_pre_attn_scalarr   attention_dropout	is_causalra   rb   attention_biasq_projk_projv_projo_projattn_logit_softcappingsliding_windowrl   rms_norm_epsq_normk_normrI   rZ   r   rJ   s      r2   rF   zGemma3Attention.__init__  s    ,,Y7;NN"
F4F4F&JdJd4de$*$>$>&B\B\$\!33T9!%!>!>ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii&&68J8JQWQfQf
 '+kk&H&H#7;f33D#V=P=PQ#V=P=PQr1   r8   position_embeddingsr   past_key_valuecache_positionr   r   c                    |j                   d d }g |d| j                  }| j                  |      j                  |      j	                  dd      }	| j                  |      j                  |      j	                  dd      }
| j                  |      j                  |      j	                  dd      }| j                  |	      }	| j                  |
      }
|\  }}t        |	|
||      \  }	}
|'|||d}|j                  |
|| j                  |      \  }
}t        }| j                  j                  dk7  rt        | j                  j                     } || |	|
||f| j                   r| j"                  nd| j$                  | j&                  d|\  }} |j(                  g |d j+                         }| j-                  |      }||fS )Nrs   r"   r    )r   r   r   eager        )r   r   r   )r}   r   r   viewr   r   r   r   r   r   updater   r   rZ   _attn_implementationr   r   r   r   r   r   r   r   )rI   r8   r   r   r   r   r   input_shapehidden_shapequery_statesr   r   r   r   cache_kwargsattention_interfacer   r   s                     r2   rN   zGemma3Attention.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{{<0[[,
&S#7jRUWZ#[ j%#&snUL'5'<'<ZW[WeWegs't$J(?;;++w6"9$++:Z:Z"[$7
%
 /3mmD**LL..
%
 
%
!\ *k));;;;FFHkk+.L((r1   )NN)r)   r*   r+   r,   r$   rS   rF   r-   rU   r   r	   
LongTensorr   r   r;   rN   rV   rW   s   @r2   r   r     s    GR/ RC RD +/59-)||-) #\\-) !.	-)
 !-) !!1!12-) -.-) 
u||Xell3XeELL>Q5RR	S-)r1   r   c                   X    e Zd Zdedef fdZ	 	 	 	 	 	 ddej                  dej                  dej                  deej                     deej                     d	ee
   d
ee   dee   deej                     deej                  eeej                  ej                  f      f   fdZ xZS )Gemma3DecoderLayerrZ   r   c                    t         |           || _        |j                  | _        || _        |j
                  |   | _        t        ||      | _        t        |      | _
        t        | j                  |j                        | _        t        | j                  |j                        | _        t        | j                  |j                        | _        t        | j                  |j                        | _        y )N)rZ   r   rn   )rE   rF   rZ   r_   r   r   attention_typer   	self_attnrY   mlprl   r   input_layernormpost_attention_layernormpre_feedforward_layernormpost_feedforward_layernormr   s      r2   rF   zGemma3DecoderLayer.__init__\  s    !--"$00;()LV$,T-=-=6CVCVW(5d6F6FFL_L_(`%)6t7G7GVM`M`)a&*78H8HfNaNa*b'r1   r8   position_embeddings_globalposition_embeddings_localr   r   r   output_attentions	use_cacher   r   c
                 T   |}| j                  |      }| j                  j                  r|}n|} | j                  d||||||||	d|
\  }}| j                  |      }||z   }|}| j	                  |      }| j                  |      }| j                  |      }||z   }|f}|r||fz  }|S )N)r8   r   r   r   r   r	  r
  r   r0   )r  r  r   r  r  r  r  )rI   r8   r  r  r   r   r   r	  r
  r   r   residualr   self_attn_weightsoutputss                  r2   rN   zGemma3DecoderLayer.forwardi  s     !,,]; >>$$";"<+94>> 
,
' 3)%)/)
,
 
,
(( 55mD =0 66}E/77F =0 ")++Gr1   )NNNFFN)r)   r*   r+   r$   rS   rF   r-   rU   r   r   r	   boolr;   r.   rN   rV   rW   s   @r2   r   r   [  s    c/ cC c$ 2637*.,1$)590||0 %*LL0 $)<<	0
 !.0 u//00 !0 $D>0 D>0 !!1!120 
u  (51B1BEDUDU1U+V"WW	X0r1   r   c                   ^     e Zd ZU eed<   dZdZg dZdgZdZ	dZ
dZdZdZeedZ fdZ xZS )Gemma3PreTrainedModelrZ    T)r   SiglipVisionEmbeddingsSiglipEncoderLayer#SiglipMultiheadAttentionPoolingHeadr7   )r8   r9   c                     t         |   |       t        |t              r%|j                  j
                  j                          y y rM   )rE   _init_weightsr   Gemma3MultiModalProjectormm_input_projection_weightdatazero_)rI   r   rJ   s     r2   r  z#Gemma3PreTrainedModel._init_weights  s8    f%f78--2288: 9r1   )r)   r*   r+   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  rV   rW   s   @r2   r  r    s]    &*# $5"5N!"&+%
; ;r1   r  c                   ,    e Zd ZU eed<   def fdZee	 	 	 	 	 	 	 	 	 ddee	j                     dee	j                     dee	j                     dee   dee	j                     dee   d	ee   d
ee   dee	j                     dee   defd              Z xZS )Gemma3TextModelrZ   c           	         t         |   |       |j                  | _        |j                  | _        t        |j                  |j                  | j                  | j                  j                  dz        | _        t        j                  t        |j                        D cg c]  }t        ||       c}      | _        t        |j                  |j                         | _        t%        |      | _        d| _        t+        j,                  |      }|j.                  |_        ddi|_        t%        |      | _        | j7                          y c c}w )N      ?)rA   r   rZ   Fr   r   )rE   rF   pad_token_idr@   
vocab_sizer=   r_   rZ   embed_tokensra   
ModuleListrangenum_hidden_layersr   layersrl   r   normr   
rotary_embgradient_checkpointingcopydeepcopyrope_local_base_freq
rope_thetar   rotary_emb_local	post_initr   s      r2   rF   zGemma3TextModel.__init__  s    !.. ++ :v1143C3CQUQ\Q\QhQhjmQm
 mmDI&JbJbDcdy	2d
 "&"4"4&:M:MN	/v>&+# v&"77*I6 5V D 	 es   "ErK   r   r   r7   inputs_embedsr
  r	  output_hidden_statesr   r   r   c
                 $   ||n| j                   j                  }||n| j                   j                  }||n| j                   j                  }|d u |d uz  rt	        d      | j
                  r%| j                  r|rt        j                  d       d}|| j                  |      }|r|| j                  s
t               }|	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+                  ||      }| j-                  ||      }|rd	nd }|rd	nd }| j.                  d | j                   j0                   D ]:  }|r||fz  } ||f||||j2                     |||||	d
|
}|d   }|s2||d   fz  }< | j5                  |      }|r||fz  }t7        ||||      S )N:You must specify exactly one of input_ids or inputs_embedszX`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.Fr   r"   r   rZ   input_embedsr   r   r7   r   full_attentionr   r0   )r  r  r   r   r   r	  r
  r   )last_hidden_stater7   r8   r9   )rZ   r	  r<  r
  
ValueErrorr4  r   loggerwarning_oncer-  r
   get_seq_lengthr-   aranger}   r   r   r   r   r   r   r3  r9  r1  r0  r   r2  r   )rI   rK   r   r   r7   r;  r
  r	  r<  r   r   past_seen_tokenscausal_mask_mappingmask_kwargsr8   r  r  all_hidden_statesall_self_attnsdecoder_layerlayer_outputss                        r2   rN   zGemma3TextModel.forward  s    2C1N-TXT_T_TqTq$8$D $++JjJj 	 "+!6IDKK<Q<Q	-t";<YZZ&&4==Yj I  --i8M0*nO!CRC^==?de"\\  =#6#6q#99$++N )33A6L ?-F ++ -"0"0#2 ,K #5"C{"C%F%U%U# & &*__]L%Q"$($9$9-$V! #7BD0d![[)H4;;+H+HI 	6M#!m%55!)+E*C2=3O3OP)."3#- M *!,M =#3"55)	6, 		-0-!11&+++%	
 	
r1   	NNNNNNNNN)r)   r*   r+   r$   r/   rF   r   r   r   r-   r   rU   r	   r.   r  r   r   r   rN   rV   rW   s   @r2   r'  r'    s   / 4  151537+/59$(,0/359i
E,,-i
 !.i
 u//0	i

 "%i
   1 12i
 D>i
 $D>i
 'tni
 !!1!12i
 +,i
 
!i
  i
r1   r'  c                       e Zd ZU dgZddiZddgdgfiZeed<   dZdef fdZ	d	 Z
d
 Zee	 	 	 	 	 	 	 	 	 	 	 ddeej                      deej"                     deej                      dee   deej&                     deej                      dee   dee   dee   deej                      deeej"                  f   defd              Z xZS )Gemma3ForCausalLMlm_head.weightlm_headcolwise_repr8   r6   rZ   language_modelc                     t         |   |       t        |      | _        |j                  | _        t        j                  |j                  |j                  d      | _        | j                          y r\   )
rE   rF   r'  modelr,  ra   rb   r_   rU  r:  rh   s     r2   rF   zGemma3ForCausalLM.__init__M  sU     $V,
 ++yy!3!3V5F5FUS 	r1   c                     || _         y rM   rY  rI   decoders     r2   set_decoderzGemma3ForCausalLM.set_decoderV  s	    
r1   c                     | j                   S rM   r[  r~   s    r2   get_decoderzGemma3ForCausalLM.get_decoderY  s    zzr1   rK   r   r   r7   r;  labelsr
  r	  r<  r   logits_to_keepr   c                 .   | j                   rF| j                  j                  dk7  r-t        j	                  d| j                  j                   d       ||n| j                  j
                  }|	|	n| j                  j                  }	 | 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, Gemma3ForCausalLM

        >>> model = Gemma3ForCausalLM.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   zhIt is strongly recommended to train Gemma3 models with the `eager` attention implementation instead of `zp`. Use `eager` with `AutoModelForCausalLM.from_pretrained('<path-to-checkpoint>', attn_implementation='eager')`.N)	rK   r   r   r7   r;  r
  r	  r<  r   r5   r6   r7   r8   r9   r0   )r   rZ   r   rF  rG  r	  r<  rY  rD  r   rS   slicerU  final_logit_softcappingr-   r   loss_functionr,  r   r7   r8   r9   )rI   rK   r   r   r7   r;  ra  r
  r	  r<  r   rb  r   r  r8   slice_indicesr6   r5   s                     r2   rN   zGemma3ForCausalLM.forward\  s   F ==T[[==H#{{??@  Aqr 2C1N-TXT_T_TqTq$8$D $++JjJj 	 ,64:: ,
)%+'/!5),
 ,
  118B>SV8W~ot4]kmA}a,?@A;;..:dkkAAAFZZ'FdkkAAAF%4%%ffdooPPD%#33!//))
 	
r1   )NNNNNNNNNNr   )r)   r*   r+   _tied_weights_keys_tp_plan_pp_planr$   r/   r  rF   r^  r`  r   r   r   r-   r   rU   r	   r.   r  r   rS   r   rN   rV   rW   s   @r2   rS  rS  E  sn   *+=)H_-z:;H(/   151537+/59-1$(,0/35934K
E,,-K
 !.K
 u//0	K

 "%K
   1 12K
 ))*K
 D>K
 $D>K
 'tnK
 !!1!12K
 c5<</0K
 
 K
  K
r1   rS  c                   D     e Zd Zdef fdZdej                  fdZ xZS )r  rZ   c                    t         |           t        j                  t	        j
                  |j                  j                  |j                  j                              | _	        t        |j                  j                  |j                  j                        | _        t        |j                  j                  |j                  j                  z        | _        t        |j"                  dz        | _        | j                   | j$                  z  | _        t        j(                  | j&                  | j&                        | _        y )Nr   r)  )kernel_sizestride)rE   rF   ra   rp   r-   rq   vision_configr_   text_configr  rl   layer_norm_epsmm_soft_emb_normrS   
image_size
patch_sizepatches_per_imagemm_tokens_per_imagetokens_per_sidern  	AvgPool2davg_poolrh   s     r2   rF   z"Gemma3MultiModalProjector.__init__  s    *,,,KK,,88&:L:L:X:XY+
' !.  ,,&2F2F2U2U!
 "%V%9%9%D%DH\H\HgHg%g!h"6#=#=s#BC11T5I5II1A1A$JZJZ[r1   vision_outputsc                    |j                   \  }}}|j                  dd      }|j                  ||| j                  | j                        }|j	                         }| j                  |      }|j                  d      }|j                  dd      }| j                  |      }t        j                  || j                        }|j                  |      S )Nr"   r    )r}   r   r   rv  r   rz  flattenrs  r-   r   r  rz   )	rI   r{  
batch_size_
seq_lengthreshaped_vision_outputspooled_vision_outputsnormed_vision_outputsprojected_vision_outputss	            r2   rN   z!Gemma3MultiModalProjector.forward  s    $2$8$8!
Az"0":":1a"@"9"A"A
D$:$:D<R<R#
 #:"D"D"F $.E F 5 = =a @ 5 ? ?1 E $ 5 56K L#(<<0EtGfGf#g '//??r1   )	r)   r*   r+   r#   rF   r-   rU   rN   rV   rW   s   @r2   r  r    s#    \| \ @ell @r1   r  token_type_idsimage_group_idstokens_per_imagec           
      Z      ydt         dt         dt         dt         dt        f
 fd}|S )z
    This function adds the correct offsets to the `q_idx` and `kv_idx` as the torch API can only accept lengths,
    not start and end indices.
    N	batch_idxhead_idxq_idxkv_idxr   c                 H   t        j                  |
j                  d   k  |d      }
| |f   }t        j                  |
j                  d   k  |d      }	| |f   }t        j                  |	j                  d   k  |d      }
| |f   dk(  |dk(  z  }	| |f   |k(  }||z  S )Nr"   r   rs   )r-   wherer}   )r  r  r  r  safe_idxtoken_type_ids_at_kv_idximage_group_ids_at_kv_idxis_image_blocksame_image_blockr  r  s            r2   
inner_maskz0token_type_ids_mask_function.<locals>.inner_mask  s     ;;v(<(<Q(??K#1)X2E#F #(;;v8L8LQ8O/OQikl#m $3Ix4G$H!$)KK9N9Nq9Q0QSlnp$q!(E)9:a?D\`aDab*9e+;<@YY  000r1   )rS   r  )r  r  r  r  s   ``  r2   token_type_ids_mask_functionr    s>     1c 1S 1 1c 1d 1" r1   zx
    The Base Gemma3 model which consists of a vision backbone and a language model withou language modeling head.,
    c            !       P    e Zd ZddiZdZdef fdZd Zd Zd Z	d	 Z
d
ej                  dej                  fdZdej                  dej                  dej                  fdZee	 	 	 	 	 	 	 	 	 	 	 	 	 ddej                  d
ej                  deej                     deej                     deeeej                     ef      deej                     deej                     deej                     deej                     dee   dee   dee   dee   deeef   fd              Z xZS )Gemma3Modelzlanguage_model.modelrW  FrZ   c                    t         |   |       t        j                  |j                        | _        t        |      | _        |j                  j                  | _	        t        j                  |j                        }|| _
        | j                  j                  | j                  j                  nd| _        | j                          y )Nr*  rs   )rE   rF   r!   from_configrp  vision_towerr  multi_modal_projectorrq  r,  rW  rZ   r+  r:  )rI   rZ   rW  rJ   s      r2   rF   zGemma3Model.__init__  s     %119M9MN%>v%F" ,,77"..f6H6HI,8<8P8P8\DKK44bdr1   c                 6    | j                   j                         S rM   )rW  get_input_embeddingsr~   s    r2   r  z Gemma3Model.get_input_embeddings  s    ""7799r1   c                 :    | j                   j                  |       y rM   )rW  set_input_embeddingsrI   r   s     r2   r  z Gemma3Model.set_input_embeddings
  s    007r1   c                     || _         y rM   rW  r\  s     r2   r^  zGemma3Model.set_decoder  s
    %r1   c                     | j                   S rM   r  r~   s    r2   r`  zGemma3Model.get_decoder  s    """r1   pixel_valuesr   c                 `    | j                  |      j                  }| j                  |      }|S )a  
        Projects the last hidden state from the vision model into language model space.

        Args:
            pixel_values (`torch.FloatTensor]` of shape `(batch_size, channels, height, width)`)
               The tensors corresponding to the input images.
        Returns:
            image_features (`torch.Tensor`): Image feature tensor of shape `(num_images, image_length, embed_dim)`).
        )r  )r  rD  r  )rI   r  r{  image_featuress       r2   get_image_featureszGemma3Model.get_image_features  s3     ***EWW33NCr1   rK   r;  r  c                 P   |m| | j                         t        j                  | j                  j                  t        j
                  |j                              k(  }|j                  d      }n|| j                  j                  k(  }|j                         }|j                  d      j                  |      j                  |j                        }|j                  d   |j                  d   z  }||   j                         |j                         k7  rt        d| d|       |S )z
        Obtains multimodal placeholdr mask from `input_ids` or `inputs_embeds`, and checks that the placeholder token count is
        equal to the length of multimodal features. If the lengths are different, an error is raised.
        )rQ   r   rs   r   r"   z6Image features and image tokens do not match: tokens: z, features )r  r-   rH   rZ   image_token_idlongr   allsumr   	expand_asrO   r}   numelrE  )rI   rK   r;  r  special_image_maskn_image_tokensn_image_featuress          r2   get_placeholder_maskz Gemma3Model.get_placeholder_mask!  s    !.2M$2K2K2MT[[77uzzR_RfRfg3 " "4!7!7!;!*dkk.H.H!H+//1/99"=GGVYYZgZnZno)//2^5I5I!5LL+,2248L8L8NNHHXXcdtcuv  "!r1   r   r   r7   r  r   ra  r
  r	  r<  return_dictc                 *   |du |duz  rt        d      ||n| j                  j                  }||n| j                  j                  }||n| j                  j                  }|R| j                  j
                  | j                  k\  r/|| j                  j
                  k(  }|j                         }d||<   n|}| | j                         |      }|F||j                         nd}t        j                  |||j                  d   z   |j                        }|]| j                  |      }|j                  |j                  |j                         }| j#                  |||      }|j%                  ||      }t'        |x}t(              s)| j                  j+                         |||||d}||j                  d   dk7  r|dk(  j                  |j                        }|t,        j.                  j1                  |dd	      dddd
f    z  }t        j2                  |j5                         d      dz
  }t        j6                  ||t        j8                  |d
            }t;        |j                  |j                        || j                  j<                        |d<   t?        di |tA        di |d} | jB                  d|||||
||d|d	|}tE        |jF                  |
r|jH                  nd|jJ                  |jL                  |      S d      S )a]  
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
            config.text_config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
            (masked), the loss is only computed for the tokens with labels in `[0, ..., config.text_config.vocab_size]`.

        Example:

        ```python
        >>> from PIL import Image
        >>> import requests
        >>> from transformers import AutoProcessor, Gemma3ForConditionalGeneration

        >>> model = Gemma3ForConditionalGeneration.from_pretrained("google/gemma32-3b-mix-224")
        >>> processor = AutoProcessor.from_pretrained("google/gemma32-3b-mix-224")

        >>> prompt = "Where is the cat standing?"
        >>> url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg"
        >>> image = Image.open(requests.get(url, stream=True).raw)

        >>> inputs = processor(images=image, text=prompt,  return_tensors="pt")

        >>> # Generate
        >>> generate_ids = model.generate(**inputs,)
        >>> processor.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        "Where is the cat standing?\nsnow"
        ```Nr>  r   r"   r?  )r;  r  r@  r"   r   r   rs   r   or_mask_functionrB  T)	r   r   r7   r;  r
  r	  r<  r  r   )rD  r7   r8   r9   r(   r0   )'rE  rZ   r	  r<  use_return_dictr  r,  cloner  rH  r-   rI  r}   r   r  rO   rQ   r  masked_scatterr   r   get_text_configra   r   padcumsumrS   r  	full_liker  rw  r   r   rW  r'   rD  r7   r8   r9   )rI   rK   r  r   r   r7   r  r   r;  ra  r
  r	  r<  r  	lm_kwargsr  llm_input_idsrJ  r  rK  rL  is_imagenew_image_startr  r  s                            r2   rN   zGemma3Model.forward9  sO   \ -t";<YZZ1B1N-TXT_T_TqTq$8$D $++JjJj 	 &1%<k$++B]B]  T[[%?%?4??%R!*dkk.H.H!H%OO-M01M,-%M 7D557FM!CRC^==?de"\\ "2]5H5H5K"KTaThThN
 #!44\BN+..}/C/C]EXEXYN!%!:!:~ "; " *889K^\M ?-F ++557 -"0"0#2 ,K )m.A.A!.D.I
 +a/33N4I4IJ"*bmm.?.?&XY.?.Z[\^a_a^a[a.b-b"b"',,/B/B/D!"Lq"P"'++hYgikIl"m2N"%%n&;&;<ot{{OnOn3./ #5"C{"C%F%U%U#
 &$%% 
.%+'/!5)
 
 )%777@G33d!//))2>2J
 	

 QU
 	
r1   )NNNNNNNNNNNNN)r)   r*   r+   _checkpoint_conversion_mappingaccepts_loss_kwargsr#   rF   r  r  r^  r`  r-   rU   r  r   r.   r  r   r   r   r   r:   r	   r  r;   r'   rN   rV   rW   s   @r2   r  r    s    '=>N%O"
| 
:8&#u||  "))":?:K:K"]b]n]n"0  '+*.1537KO595959-1$(,0/3&*@
##@
 ''@
 !.	@

 u//0@
 "%U->->(?(F"GH@
 !!1!12@
 !!1!12@
   1 12@
 ))*@
 D>@
 $D>@
 'tn@
 d^@
  
u//	0!@
  @
r1   r  zy
    The Base Gemma3 model which consists of a vision backbone and a language model without language modeling head.,
    c            "           e Zd ZdddddZdgZdef fdZd	 Zd
 Zd Z	d Z
d Zed        Zed        Zed        Ze	 	 	 	 	 	 	 	 	 	 	 	 	 	 d$dej$                  dej&                  deej*                     deej$                     deeeej&                     ef      deej$                     deej$                     deej&                     deej$                     dee   dee   dee   dee   deeej*                  f   deeef   fd        Z	 	 	 	 	 	 	 	 	 	 d% fd!	Ze	 d&de d"ej*                  deej*                     dej*                  dee   deej*                     deej*                     de!fd#       Z" xZ#S )'Gemma3ForConditionalGenerationzmodel.language_modelzmodel.vision_towerzmodel.multi_modal_projectorrU  )z^language_model.modelz^vision_towerz^multi_modal_projectorz^language_model.lm_headrT  rZ   c                     t         |   |       t        |      | _        t	        j
                  |j                  j                  |j                  j                  d      | _	        | j                          y r\   )rE   rF   r  rY  ra   rb   rq  r_   r,  rU  r:  rh   s     r2   rF   z'Gemma3ForConditionalGeneration.__init__  sS      (
yy!3!3!?!?ASASA^A^ejkr1   c                 6    | j                   j                         S rM   rY  r  r~   s    r2   r  z3Gemma3ForConditionalGeneration.get_input_embeddings      zz..00r1   c                 :    | j                   j                  |       y rM   rY  r  r  s     r2   r  z3Gemma3ForConditionalGeneration.set_input_embeddings      

''.r1   c                 :    | j                   j                  |       y rM   )rY  r^  r\  s     r2   r^  z*Gemma3ForConditionalGeneration.set_decoder  s    

w'r1   c                 6    | j                   j                         S rM   )rY  r`  r~   s    r2   r`  z*Gemma3ForConditionalGeneration.get_decoder  s    zz%%''r1   c                 8    | j                   j                  |      S rM   )rY  r  )rI   r  s     r2   r  z1Gemma3ForConditionalGeneration.get_image_features  s    zz,,\::r1   c                 .    | j                   j                  S rM   )rY  rW  r~   s    r2   rW  z-Gemma3ForConditionalGeneration.language_model  s    zz(((r1   c                 .    | j                   j                  S rM   )rY  r  r~   s    r2   r  z+Gemma3ForConditionalGeneration.vision_tower  s    zz&&&r1   c                 .    | j                   j                  S rM   )rY  r  r~   s    r2   r  z4Gemma3ForConditionalGeneration.multi_modal_projector  s    zz///r1   rK   r  r   r   r7   r  r   r;  ra  r
  r	  r<  r  rb  r   c                    ||n| j                   j                  }||n| j                   j                  }||n| j                   j                  } | j                  d||||||||
|	||||d|}|d   }t        |t              rt        | d      n|}| j                  |dd|ddf         }d}|	O|j                         }|dddddf   }|	dddf   }||dd|j                  d    df   j                  |j                        }||j                  |j                        dk7     j                         }||j                  |j                        dk7     j                         }n |j                         }|j                         }t        j                         }|j!                  d| j                   j"                  j$                        }|j!                  d      j                  |j                        } |||      }|s|f|dd z   }||f|z   S |S t'        |||j(                  |j*                  |j,                  |j.                        S )	a  
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
            config.text_config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
            (masked), the loss is only computed for the tokens with labels in `[0, ..., config.text_config.vocab_size]`.

        Example:

        ```python
        >>> from PIL import Image
        >>> import requests
        >>> from transformers import AutoProcessor, Gemma3ForConditionalGeneration

        >>> model = Gemma3ForConditionalGeneration.from_pretrained("google/gemma-3-4b-it")
        >>> processor = AutoProcessor.from_pretrained("google/gemma-3-4b-it")

        >>> messages = [
        ...     {
        ...         "role": "system",
        ...         "content": [
        ...             {"type": "text", "text": "You are a helpful assistant."}
        ...         ]
        ...     },
        ...     {
        ...         "role": "user", "content": [
        ...             {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg"},
        ...             {"type": "text", "text": "Where is the cat standing?"},
        ...         ]
        ...     },
        ... ]

        >>> inputs = processor.apply_chat_template(
        ...     messages,
        ...     tokenize=True,
        ...     return_dict=True,
        ...     return_tensors="pt",
        ...     add_generation_prompt=True
        ... )
        >>> # Generate
        >>> generate_ids = model.generate(**inputs)
        >>> processor.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        "user\nYou are a helpful assistant.\n\n\n\n\n\nWhere is the cat standing?\nmodel\nBased on the image, the cat is standing in a snowy area, likely outdoors. It appears to"
        ```
        N)rK   r  r  r   r   r7   r;  r
  ra  r	  r<  r  r   r   .rs   r"   )r5   r6   r7   r8   r9   r(   r0   )rZ   r	  r<  r  rY  r   rS   re  rU  rT   r}   rO   r   r   ra   CrossEntropyLossr   rq  r,  r4   r7   r8   r9   r(   )rI   rK   r  r   r   r7   r  r   r;  ra  r
  r	  r<  r  rb  r  r  r8   rh  r6   r5   shift_logitsshift_labelsshift_attention_maskloss_fctflat_logitsflat_labelsr{   s                               r2   rN   z&Gemma3ForConditionalGeneration.forward  s}   @ 2C1N-TXT_T_TqTq$8$D $++JjJj 	 &1%<k$++B]B]$** 
%))%+'/!5#)
 
"  
8B>SV8W~ot4]kmA}a,?@A\\^F!#ssA+.L!#qr'?L) (6a,:L:LQ:O9O9Q6Q'R'U'UV\VcVc'd$+,@,C,CFMM,RVW,WXcce+,@,C,CLDWDW,X\],]^iik+668+668**,H&++B0G0G0R0RSK&++B/22<3F3FGKK5DY,F'+'7D7V#CVC+#33!//)) ' ; ;
 	
r1   c                 T    t        |   |f||||||	|
|d|}|d   dk(  r||d<   |S )N)r7   r;  r   r   r   r
  rb  r  r   r  )rE   prepare_inputs_for_generation)rI   rK   r7   r;  r   r   r  r   r  r
  rb  ra  r   model_inputsrJ   s                 r2   r  z<Gemma3ForConditionalGeneration.prepare_inputs_for_generationm  s]      w<
+')%)))
 
 !!+7L(r1   rA  c                    | j                         |||||d}||j                  d   dk7  r|dk(  j                  |j                        }	|	t        j
                  j                  |	dd      d d d df    z  }
t        j                  |
j                         d      dz
  }t        j                  |	|t        j                  |d            }t        |j                  |j                        || j                        |d<   t        d	i |S )
Nr@  r"   r  r   r  rs   r   r  r0   )r  r}   rO   r   ra   r   r  r-   r  rS   r  r  r  rw  r   )rZ   rA  r   r   r7   r   r  r   rL  r  r  r  s               r2   r   z8Gemma3ForConditionalGeneration.create_masks_for_generate  s    ,,.(,,.(
 %,*<*<Q*?1*D
 '!+//0E0EFH&"--*;*;HfTU*;*VWXZ][]Z]W]*^)^^O#ll?+>+>+@aH1LO#kk(OU__UcegEhiO.J!!."7"78/6KeKe/K*+ )7;77r1   )NNNNNNNNNNNNNr   )
NNNNNNNTNNrM   )$r)   r*   r+   r  ri  r#   rF   r  r  r^  r`  r  propertyrW  r  r  r   r-   r   r.   r   rU   r   r:   r	   r  rS   r;   r4   rN   r  staticmethodr   r   r   rV   rW   s   @r2   r  r    s    "8-"?#,	&" ++| 1/((; ) ) ' ' 0 0  '+*.1537KO595959-1$(,0/3&*34|
##|
 ''|
 !.	|

 u//0|
 "%U->->(?(F"GH|
 !!1!12|
 !!1!12|
   1 12|
 ))*|
 D>|
 $D>|
 'tn|
 d^|
 c5<</0|
" 
u22	3#|
 |
B "H  26!8 !8ll!8 !.!8 	!8
 "%!8 u||,!8 !.!8 
!8 !8r1   r  c                   H    e Zd Z fdZd Zd Zee	 	 	 	 	 	 	 	 	 ddej                  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j                     de
e   dee   defd              Z xZS )Gemma3ForSequenceClassificationc                     t         |   |       |j                  | _        t        |      | _        t        j                  |j                  j                  | j                  d      | _	        | j                          y r\   )rE   rF   
num_labelsr  rY  ra   rb   rq  r_   scorer:  rh   s     r2   rF   z(Gemma3ForSequenceClassification.__init__  sZ      ++ (
YYv11==tUZ[
 	r1   c                 6    | j                   j                         S rM   r  r~   s    r2   r  z4Gemma3ForSequenceClassification.get_input_embeddings  r  r1   c                 :    | j                   j                  |       y rM   r  r  s     r2   r  z4Gemma3ForSequenceClassification.set_input_embeddings  r  r1   rK   r  r   r   r7   r;  r  ra  r
  r   r   c
                     | j                   |f|||||||	d|
}|j                  }| j                  |      }||j                  d   }n|j                  d   }| j                  j
                  j                  |dk7  rt        d      | j                  j
                  j                  d}n||| j                  j
                  j                  k7  j                  |j                  t        j                        }t        j                  |j                  d   |j                  t        j                        }||z  j                  d      }n.d}t        j                  | j                   j"                   d       |t        j                  ||j                  	      |f   }d}|| j%                  |||| j                  
      }t'        |||j(                  |j*                  |j,                        S )a  
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        )r   r  r   r7   r;  r  r
  Nr   r"   z=Cannot handle batch sizes > 1 if no padding token is defined.rs   )r   rQ   z will not detect padding tokens in `inputs_embeds`. Results may be unexpected if using padding tokens in conjunction with `inputs_embeds.`r?  )r6   ra  pooled_logitsrZ   rd  )rY  rD  r  r}   rZ   rq  r+  rE  rO   r   r-   int32rI  argmaxrF  rG  rJ   r)   rg  r   r7   r8   r9   )rI   rK   r  r   r   r7   r;  r  ra  r
  r   transformer_outputsr8   r6   r~  last_non_pad_tokennon_pad_masktoken_indicesr  r5   s                       r2   rN   z'Gemma3ForSequenceClassification.forward  s   , )djj

)%%+')

 

 ,==M* "+J&,,Q/J;;""//7J!O\]];;""//7!#"%)@)@)M)MMQQRXR_R_afalalmL!LL)<V]]Z_ZeZefM"/,">!F!Fr!J!#>>**+ ,Z Z
 u||Jv}}MOaab%%VFR_hlhshs%tD/ /??-;;*55
 	
r1   rQ  )r)   r*   r+   rF   r  r  r   r   r-   r   r   r.   rU   r	   r  r   r   r   rN   rV   rW   s   @r2   r  r    s   1/  '+481537+/5959-1$(C
##C
 u001C
 !.	C

 u//0C
 "%C
   1 12C
 !!1!12C
 ))*C
 D>C
 +,C
 
*C
  C
r1   r  )r  r'  rS  r  r  r  )Nr"   )r   NN)Sr5  collections.abcr   dataclassesr   typingr   r   r-   torch.nnra   activationsr   cache_utilsr	   r
   configuration_utilsr   
generationr   masking_utilsr   r   r   modeling_flash_attention_utilsr   modeling_layersr   modeling_outputsr   r   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   r   r   r   utils.genericr   autor!   configuration_gemma3r#   r$   
get_loggerr)   rF  r'   r4   	Embeddingr=   ModulerY   rl   r   r   r   rU   rS   r   rT   r;   r   r   r   r  r'  rS  r  r  r  r  r  __all__r0   r1   r2   <module>r     s
  ,  $ ! "   ! . 3 ) m m B 9 q q K F & _ _ /  @ 
		H	% 
< 7 < <  
<; < <2
SBLL 
S		  =BII =(<BII <D(6	UU\\ 	U# 	U%,, 	U$ ## %II %<< % 
 % <<	 %
 U\\* %  % e_ % e_ % 5<<%& %FM)bii M)`>3 >B ;O ; ;8 H
+ H
 H
V c
- c
 c
L!@		 !@HU\\*ell+  h	B 
E
' E

E
P 
p8%:O p8
p8fU
&; U
pr1   