
    <h                     f   S SK JrJrJrJr  S SKrS SKJs  Jr	  S SKJr  SSK
Jr  SSKJrJr  SSKJr  SSKJr  SS	KJr  SS
KJrJrJr  SSKJrJr  SSKJrJr  SSKJ r   SSK!J"r"J#r#J$r$  SSK%J&r&  \#" 5       (       a  S SK'J(r(  SSK)J*r*  \$RV                  " \,5      r- " S S\SS9r. " S S\R^                  5      r0 " S S\R^                  5      r1 " S S\R^                  5      r2 " S S\R^                  5      r3 " S S \R^                  5      r4S! r5S?S" jr6S#\Rn                  S$\8S%\Rn                  4S& jr9 S@S'\R^                  S(\Rn                  S)\Rn                  S*\Rn                  S+\\Rn                     S,\:S-\:4S. jjr; " S/ S0\R^                  5      r< " S1 S2\5      r=\" " S3 S4\5      5       r> " S5 S6\R^                  5      r?\" " S7 S8\>5      5       r@   SAS9\\Rn                  \A\Rn                     S4   S:\\8   S+\\Rn                     S%\\Rn                  \84   4S; jjrB " S< S=\>\5      rC/ S>QrDg)B    )CallableOptional	TypedDictUnionN)nn   )ACT2FN)CacheDynamicCache)GenerationMixin)AttentionMaskConverter)GradientCheckpointingLayer)BaseModelOutputWithPastMoeCausalLMOutputWithPastMoeModelOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)auto_docstringis_torch_flex_attn_availablelogging   )GraniteMoeSharedConfig)	BlockMask)make_flex_block_causal_maskc                       \ rS rSr% Sr\R                  \S'   \R                  \S'   \\S'   \\S'   \R                  \S'   Sr
g	)
GraniteFlashAttentionKwargs2   aR  
Keyword arguments for advanced Flash Attention, causal-conv1d, and mamba_ssm kernel usage.
Use cases include padding-free training and fewer `torch.compile` graph breaks.

Attributes:
    cu_seq_lens_q (`torch.LongTensor`)
        Gets cumulative sequence length for query state.
    cu_seq_lens_k (`torch.LongTensor`)
        Gets cumulative sequence length for key state.
    max_length_q (`int`):
        Maximum sequence length for query state.
    max_length_k (`int`):
        Maximum sequence length for key state.
    seq_idx (`torch.IntTensor):
        Index of each packed sequence.
cu_seq_lens_qcu_seq_lens_kmax_length_qmax_length_kseq_idx N)__name__
__module____qualname____firstlineno____doc__torch
LongTensor__annotations__int	IntTensor__static_attributes__r&       v/var/www/html/shao/venv/lib/python3.13/site-packages/transformers/models/granitemoeshared/modeling_granitemoeshared.pyr   r   2   s7    " ######__r2   r   F)totalc                   n   ^  \ rS rSrSrS\4U 4S jjrS\R                  S\R                  4S jr	Sr
U =r$ )	GraniteMoeSharedMLPK   zj
MLP layer for shared experts

Args:
    config:
        Configuration object with model hyperparameters.
configc                 X  > [         TU ]  5         UR                  U l        UR                  U l        [
        UR                     U l        [        R                  " U R                  U R                  S-  SS9U l
        [        R                  " U R                  U R                  SS9U l        g )N   Fbias)super__init__hidden_size
input_sizeshared_intermediate_sizer	   
hidden_act
activationr   Linearinput_linearoutput_linearselfr8   	__class__s     r3   r>   GraniteMoeSharedMLP.__init__T   s     ,,!:: !2!23IIdoot7G7G!7KRWXYYt'7'7uUr2   hidden_statesreturnc                     U R                  U5      nUR                  SSS9nU R                  US   5      US   -  nU R                  U5      nU$ )Nr:   dimr   r   )rE   chunkrC   rF   )rH   rK   chunked_hidden_statess      r3   forwardGraniteMoeSharedMLP.forward]   s^    ))-8 - 3 3A2 3 >(=a(@ADYZ[D\\**=9r2   )rC   r?   rE   r@   rF   )r'   r(   r)   r*   r+   r   r>   r,   TensorrS   r1   __classcell__rI   s   @r3   r6   r6   K   s7    V5 VU\\ ell  r2   r6   c                   8   ^  \ rS rSrSU 4S jjrS rS rSrU =r$ )GraniteMoeSharedRMSNorme   c                    > [         TU ]  5         [        R                  " [        R
                  " U5      5      U l        X l        g)z6
GraniteMoeSharedRMSNorm is equivalent to T5LayerNorm
N)r=   r>   r   	Parameterr,   onesweightvariance_epsilon)rH   r?   epsrI   s      r3   r>    GraniteMoeSharedRMSNorm.__init__f   s/     	ll5::k#:; #r2   c                    UR                   nUR                  [        R                  5      nUR	                  S5      R                  SSS9nU[        R                  " X0R                  -   5      -  nU R                  UR                  U5      -  $ )Nr:   rN   T)keepdim)	dtypetor,   float32powmeanrsqrtr_   r^   )rH   rK   input_dtypevariances       r3   rS   GraniteMoeSharedRMSNorm.forwardn   sw    #))%((7 $$Q',,R,>%H?T?T4T(UU{{]--k:::r2   c                 ^    [        U R                  R                  5       SU R                   3$ )Nz, eps=)tupler^   shaper_   rH   s    r3   
extra_repr"GraniteMoeSharedRMSNorm.extra_repru   s*    ))*+6$2G2G1HIIr2   )r_   r^   )gư>)	r'   r(   r)   r*   r>   rS   rq   r1   rV   rW   s   @r3   rY   rY   e   s    $;J Jr2   rY   c                   B   ^  \ rS rSrS\S\S\SS4U 4S jjrS rS	rU =r$ )
GraniteMoeSharedParallelExpertsy   num_expertsr@   output_sizerL   Nc                    > [         TU ]  5         [        R                  " [        R
                  " XU5      5      U l        Xl        X l        X0l	        g)a]  
Initialize the GraniteMoeSharedParallelExperts module.
The experts weights are stored in [num_experts, output_size, input_size] format. Such that it's compatible with
many MoE libraries, such as [Megablock](https://github.com/databricks/megablocks) and
[ScatterMoE](https://github.com/shawntan/scattermoe), as well as the
[MoE kernel](https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/layers/fused_moe/fused_moe.py)
used in vllm.

Args:
    num_experts (int):
        Number of experts.
    input_size (int):
        Size of the input.
    output_size (int):
        Size of the output.
N)
r=   r>   r   r\   r,   emptyr^   rv   r@   rw   )rH   rv   r@   rw   rI   s       r3   r>   (GraniteMoeSharedParallelExperts.__init__z   s<    " 	ll5;;{#TU&$&r2   c                     UR                  USS9n/ n[        U R                  5       H8  nUR                  [        R
                  " X5   U R                  U   5      5        M:     [        R                  " USS9nU$ )z
Forward pass of the GraniteMoeSharedParallelExperts module.

Args:
    inputs (Tensor):
        Input tensor.
    expert_size:
        Expert size information.

Returns:
    Tensor: Output tensor.
r   rO   )	splitrangerv   appendFlinearr^   r,   cat)rH   inputsexpert_size
input_listoutput_listiresultss          r3   rS   'GraniteMoeSharedParallelExperts.forward   sh     \\+1\5
t''(Aqxx
t{{1~FG )))KQ/r2   )r@   rv   rw   r^   	r'   r(   r)   r*   r/   r>   rS   r1   rV   rW   s   @r3   rt   rt   y   s.    'C 'S 's 't '. r2   rt   c                   >   ^  \ rS rSrS\S\S\4U 4S jjrS rSrU =r$ )GraniteMoeSharedTopKGating   r@   rv   top_kc                 z   > [         TU ]  5         X l        Xl        X0l        [
        R                  " XSS9U l        g)z
Initialize the top-k gating mechanism.
Args:
    input_size (`int`):
        Size of the input.
    num_experts (`int`):
        Number of experts.
    top_k (`int`):
        Number of top experts to select.
Fr;   N)r=   r>   rv   r@   r   r   rD   layer)rH   r@   rv   r   rI   s       r3   r>   #GraniteMoeSharedTopKGating.__init__   s2     	&$
YYzUC
r2   c                 z   U R                  U5      R                  5       nUR                  U R                  SS9u  p4[        R
                  " USS9R                  U5      n[        R                  " UR                  S5      U R                  /UR                  UR                  S9nUR                  SUS5      nUR                  5       R                  S5      nUR                  5       nUR!                  5       n	U	R#                  S5      u  pUR%                  U R                  SS9nUR!                  5       nX[   nXXU4$ )Nr   rO   r   rd   devicetrunc)rounding_mode)r   floattopkr   r,   softmaxtype_aszerossizerv   rd   r   scatterlongsumtolistflattensortdiv)rH   rK   logitstop_k_logitstop_k_indicestop_k_gatesr   gatesr   top_k_experts_index_sorted_expertsbatch_indexbatch_gatess                 r3   rS   "GraniteMoeSharedTopKGating.forward   s"   M*002&,kk$**!k&D#mmLa8@@O a $"2"23;;L;LU`UgUg
 a2jjl&&q) "((* &--/"/"4"4Q"7*..tzz.Q "))+!7#+FRRr2   )r@   r   rv   r   r   rW   s   @r3   r   r      s-    D3 DS D D&S Sr2   r   c                   :   ^  \ rS rSrSrS\4U 4S jjrS rSrU =r	$ )GraniteMoeSharedMoE   z
A Sparsely gated mixture of experts layer with 1-layer Feed-Forward networks as experts.

Args:
    config:
        Configuration object with model hyperparameters.
r8   c                   > [         TU ]  5         UR                  U l        UR                  U l        [
        UR                     U l        [        UR                  U R                  U R                  S-  5      U l
        [        UR                  U R                  U R                  5      U l        [        U R                  UR                  UR                  S9U l        g )Nr:   )r@   rv   r   )r=   r>   r?   r@   intermediate_sizer	   rB   rC   rt   num_local_expertsrE   rF   r   num_experts_per_tokrouterrG   s     r3   r>   GraniteMoeSharedMoE.__init__   s     ,,!33 !2!23;$$doot7G7G!7K
 =$$d&6&6
 100,,
r2   c                    UR                  5       u  p#nUR                  SU5      nU R                  U5      u  pVpxn	X   n
U R                  X5      nUR	                  SSS9nU R                  US   5      US   -  nU R                  X5      nXSS2S4   -  n[        R                  " X#-  U R                  4UR                  UR                  S9nUR                  SXm5      nUR                  X#U R                  5      nX4$ )z
Forward pass of the mixture of experts layer.

Args:
    layer_input (Tensor):
        Input tensor.

Returns:
    Tensor:
        Output tensor.
    Tensor:
        Router logits.
rN   r:   rO   r   r   Nr   )r   reshaper   rE   rQ   rC   rF   r,   r   r@   rd   r   	index_addview)rH   layer_inputbszlengthemb_sizer   r   r   r   router_logitsexpert_inputsrK   rR   expert_outputsr   layer_outputs                   r3   rS   GraniteMoeSharedMoE.forward   s    !, 0 0 2X!))"h7BF++kBZ?-#0))-E - 3 3A2 3 >(=a(@ADYZ[D\\++MG'ag*>>S\4??;>CWCW`n`u`uvq+F#((dooF**r2   )rC   r?   rE   r@   rF   r   )
r'   r(   r)   r*   r+   r   r>   rS   r1   rV   rW   s   @r3   r   r      s    
5 
&+ +r2   r   c                     U SSU R                   S   S-  24   nU SU R                   S   S-  S24   n[        R                  " U* U4SS9$ )z*Rotates half the hidden dims of the input..NrN   r:   rO   )ro   r,   r   )xx1x2s      r3   rotate_halfr     sZ    	
3"!''"+"""	#B	
3q ""	#B99rc2YB''r2   c                     UR                  U5      nUR                  U5      nX-  [        U 5      U-  -   nX-  [        U5      U-  -   nXg4$ )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kcossinposition_idsunsqueeze_dimq_embedk_embeds           r3   apply_rotary_pos_embr     sS    ( --
&C
--
&Cw;q>C/0Gw;q>C/0Gr2   rK   n_reprL   c                     U R                   u  p#pEUS:X  a  U $ U SS2SS2SSS2SS24   R                  X#XU5      n U R                  X#U-  XE5      $ )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)ro   expandr   )rK   r   batchnum_key_value_headsslenhead_dims         r3   	repeat_kvr   4  s_    
 2?1D1D.Ez!!Qa"23::5W\dlmM  e(CTTTr2   modulequerykeyvalueattention_maskscalingdropoutc                 @   [        X R                  5      n[        X0R                  5      n	[        R                  " XR	                  SS5      5      U-  n
Ub"  US S 2S S 2S S 2S UR
                  S   24   nX-   n
[        R                  R                  U
S[        R                  S9R                  UR                  5      n
[        R                  R                  XU R                  S9n
[        R                  " X5      nUR	                  SS5      R                  5       nX4$ )Nr:   r   rN   )rP   rd   )ptrainingr   )r   num_key_value_groupsr,   matmul	transposero   r   
functionalr   rf   re   rd   r   r   
contiguous)r   r   r   r   r   r   r   kwargs
key_statesvalue_statesattn_weightscausal_maskattn_outputs                r3   eager_attention_forwardr   @  s     3 ; ;<JU$?$?@L<<';';Aq'ABWLL!$Q1.D
0@0@0D.D%DE#1 ==((2U]](SVVW\WbWbcL==((6??([L,,|:K''1-88:K$$r2   c                   v  ^  \ rS rSrSrSS\S\\   4U 4S jjjr      SS\	R                  S\\	R                     S\\	R                     S	\\   S
\S\\	R                     S\\\	R                  \	R                  4      S\\	R                  \\	R                     \\\	R                        4   4S jjrSrU =r$ )GraniteMoeSharedAttentioni]  z=Multi-headed attention from 'Attention Is All You Need' paperr8   	layer_idxc                 z  > [         TU ]  5         Xl        X l        Uc-  [        R                  SU R                  R                   S35        UR                  U l        UR                  U l	        UR                  U l        U R                  U R                  -  U l        UR                  U l        U R                  U R                  -  U l        SU l        UR                   U l        U R                  U R                  -  U R                  :w  a&  [%        SU R                   SU R                   S35      e[&        R(                  " U R                  U R                  U R                  -  UR*                  S9U l        [&        R(                  " U R                  U R                  U R                  -  UR*                  S9U l        [&        R(                  " U R                  U R                  U R                  -  UR*                  S9U l        [&        R(                  " U R                  U R                  UR*                  S9U l        g )NzInstantiating z without passing a `layer_idx` is not recommended and will lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` when creating this class.Tz?hidden_size must be divisible by num_heads (got `hidden_size`: z and `num_heads`: z).r;   )r=   r>   r8   r   loggerwarning_oncerI   r'   attention_dropoutr?   num_attention_heads	num_headsr   r   r   	is_causalattention_multiplierr   
ValueErrorr   rD   attention_biasq_projk_projv_projo_projrH   r8   r   rI   s      r3   r>   "GraniteMoeSharedAttention.__init__`  s   " !8!8 9 :, , "(!9!9!--33((DNN:#)#=#= $(NNd6N6N$N!22MMDNN*t/?/??QRVRbRbQc$T^^$4B8 
 ii 0 0$..4==2PW]WlWlmii 0 0$2J2JT]]2Zagavavwii 0 0$2J2JT]]2Zagavavwii 0 0$2B2BI^I^_r2   rK   r   r   past_key_value	use_cachecache_positionposition_embeddingsrL   c                    UR                  5       u  pnU R                  U5      nU R                  U5      nU R                  U5      nUR	                  XU R
                  U R                  5      R                  SS5      nUR	                  XU R                  U R                  5      R                  SS5      nUR	                  XU R                  U R                  5      R                  SS5      nUb  UOSu  nnUb  [        XUU5      u  pUb$  UXS.nUR                  XU R                  U5      u  p[        nU R                  R                  S:w  a  [        U R                  R                     nU" U UUUU4U R                   (       d  SOU R"                  U R$                  S.UD6u  nnUR	                  XS5      nU R'                  U5      nUU4$ )	Nr   r:   )NN)r   r   r  eager        )r   r   rN   )r   r   r  r  r   r   r   r   r   r   updater   r   r8   _attn_implementationr   r   r   r   r  )rH   rK   r   r   r  r  r  r	  r   r   q_lenr   query_statesr   r   r   r   cache_kwargsattention_interfacer   r   s                        r3   rS   !GraniteMoeSharedAttention.forward  s    &**,A{{=1[[/
{{=1#((T^^T]]S]]^_abc__S1I1I4==Yccdeghi
#((T5M5Mt}}]gghiklm*=*I&|S*';LVY[^'_$L%#&sUL'5'<'<ZW[WeWegs't$J(?;;++w6"9$++:Z:Z"[$7	%
  $}}C$2H2HLL	%
 	%
!\ "&&s26kk+.L((r2   )r   r8   r   r?   r   r  r   r   r   r   r  r   r   r  N)NNNFNN)r'   r(   r)   r*   r+   r   r   r/   r>   r,   rU   r-   r
   boolrn   rS   r1   rV   rW   s   @r3   r   r   ]  s    G`5 `(3- ` `F 2637*.59KO0)||0) !.0) u//0	0)
 !0) 0) !!1!120) &eELL%,,,F&GH0) 
u||Xell3XeELL>Q5RR	S0) 0)r2   r   c                     ^  \ rS rSrS\S\4U 4S jjr        SS\R                  S\	\R                     S\	\R                     S\	\   S	\	\   S
\	\   S\	\R                     S\	\   S\	\\R                  \R                  4      S\\   S\\R                   \	\\R                   \R                   4      4   4S jjrSrU =r$ )GraniteMoeSharedDecoderLayeri  r8   r   c                   > [         TU ]  5         UR                  U l        [        XS9U l        UR
                  S:  a  [        U5      U l        [        UR                  UR                  S9U l
        [        UR                  UR                  S9U l        UR                  U l        UR                  S:X  a  S U l        g [        U5      U l        g )N)r8   r   r   r`   )r=   r>   r?   r   	self_attnr   r   block_sparse_moerY   rms_norm_epsinput_layernormpost_attention_layernormresidual_multiplierrA   r6   
shared_mlpr  s      r3   r>   %GraniteMoeSharedDecoderLayer.__init__  s    !--2&V##a'$7$?D!6v7I7IvObObc(?@R@RX^XkXk(l%#)#=#= "("A"AQ"F$L_`fLgr2   rK   r   r   r  output_attentionsr  r  output_router_logitsr	  r   rL   c
                 r   UnU R                  U5      nU R                  " SUUUUUUUU	S.U
D6u  pXU R                  -  -   nUnU R                  U5      nU R	                  U5      u  pU R
                  c  UnOXR                  U5      -   nXU R                  -  -   nU4nU(       a  X4-  nU(       a  X4-  nU$ )a  
Args:
    hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
    attention_mask (`torch.FloatTensor`, *optional*):
        attention mask of size `(batch_size, sequence_length)` if flash attention is used or `(batch_size, 1,
        query_sequence_length, key_sequence_length)` if default attention is used.
    output_attentions (`bool`, *optional*):
        Whether or not to return the attentions tensors of all attention layers. See `attentions` under
        returned tensors for more detail.
    use_cache (`bool`, *optional*):
        If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
        (see `past_key_values`).
    past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
    cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
        Indices depicting the position of the input sequence tokens in the sequence
    output_router_logits (`bool`, *optional*):
        Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
        should not be returned during inference.
    position_embeddings (`tuple[torch.FloatTensor, torch.FloatTensor]`, *optional*):
        Tuple containing the cosine and sine positional embeddings of shape `(batch_size, seq_len, head_dim)`,
        with `head_dim` being the embedding dimension of each attention head.
    kwargs (`dict`, *optional*):
        Arbitrary kwargs. Can be used to provide `GraniteFlashAttentionKwargs` for
        padding-free training and/or improve torch.compile performance.
)rK   r   r   r  r"  r  r  r	  r&   )r  r  r  r  r  r   )rH   rK   r   r   r  r"  r  r  r#  r	  r   residualself_attn_weightsmoe_hidden_statesr   outputss                   r3   rS   $GraniteMoeSharedDecoderLayer.forward  s    L !,,]; ,0>> 
,
')%)/) 3
,
 
,
( !43K3K#KK !55mD+/+@+@+O(??"-M-0NNM 43K3K#KK "++G''Gr2   )r  r?   r  r  r  r  r   )NNNFFNFN)r'   r(   r)   r*   r   r/   r>   r,   rU   r   r-   r
   r  rn   r   r   FloatTensorrS   r1   rV   rW   s   @r3   r  r    s2   h5 h# h  2637*.,1$)59/4KOM||M !.M u//0	M
 !M $D>M D>M !!1!12M 'tnM &eELL%,,,F&GHM 45M 
u  (51B1BEDUDU1U+V"WW	XM Mr2   r  c                   T   ^  \ rS rSr% \\S'   SrSrS/rS/r	Sr
SrSrU 4S jrS	rU =r$ )
GraniteMoeSharedPreTrainedModeli  r8   modelTr  past_key_valuesFc                    > [         TU ]  U5        [        U[        5      (       a9  UR                  R
                  R                  SU R                  R                  S9  g g )Nr  )rh   std)	r=   _init_weights
isinstancert   r^   datanormal_r8   initializer_range)rH   r   rI   s     r3   r1  -GraniteMoeSharedPreTrainedModel._init_weights  sJ    f%f=>>MM&&CT[[5R5R&S ?r2   r&   )r'   r(   r)   r*   r   r.   base_model_prefixsupports_gradient_checkpointing_no_split_modules_skip_keys_device_placement_supports_flash_attn_supports_sdpa_can_compile_fullgraphr1  r1   rV   rW   s   @r3   r,  r,    sD    ""&*#78#4"5N"T Tr2   r,  c                   l   ^  \ rS rSrSS\4U 4S jjjr\R                  " 5       \S 5       5       r	Sr
U =r$ )GraniteMoeSharedRotaryEmbeddingi#  r8   c                   > [         TU ]  5         [        US5      (       aZ  [        UR                  [
        5      (       a;  UR                  R                  SUR                  R                  S5      5      U l        OSU l        UR                  U l	        UR                  U l
        Xl        [        U R                     U l        U R                  U R                  U5      u  o0l        U R                  SUSS9  U R                   U l        g )Nrope_scaling	rope_typetypedefaultinv_freqF)
persistent)r=   r>   hasattrr2  rA  dictgetrB  max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenr8   r   rope_init_fnattention_scalingregister_bufferrE  original_inv_freq)rH   r8   r   rE  rI   s       r3   r>   (GraniteMoeSharedRotaryEmbedding.__init__$  s    6>**z&:M:Mt/T/T#0044[&BUBUBYBYZ`BabDN&DN"("@"@$*$B$B!/?+/+<+<T[[&+Q((ZeD!%r2   c                 b   U R                   S S S 2S 4   R                  5       R                  UR                  S   SS5      R	                  UR
                  5      nUS S 2S S S 24   R                  5       n[        UR
                  R                  [        5      (       a0  UR
                  R                  S:w  a  UR
                  R                  OSn[        R                  " USS9   UR                  5       UR                  5       -  R                  SS5      n[        R                  " Xf4SS	9nUR                  5       U R                  -  nUR                  5       U R                  -  n	S S S 5        WR	                  UR                   S
9W	R	                  UR                   S
94$ ! , (       d  f       N@= f)Nr   rN   r   mpscpuF)device_typeenabledr:   rO   )rd   )rE  r   r   ro   re   r   r2  rC  strr,   autocastr   r   r   rN  r   rd   )
rH   r   r   inv_freq_expandedposition_ids_expandedrU  freqsembr   r   s
             r3   rS   'GraniteMoeSharedRotaryEmbedding.forward5  sR    !MM$4-8>>@GGHZHZ[\H]_acdehhijiqiqr ,QaZ 8 > > @'1!((--'E'E!((--[`J`ahhmmfk^^UC&,,.1F1L1L1NNYYZ[]^_E))UN3C'')d444C'')d444C	 D vvAGGv$cff177f&;;; DCs   $BF  
F.)rN  r8   rK  rP  rL  rM  rB  r  )r'   r(   r)   r*   r   r>   r,   no_gradr   rS   r1   rV   rW   s   @r3   r?  r?  #  s7    /5 / /" ]]_<  <r2   r?  c                   4  ^  \ rS rSrS\4U 4S jjr\           SS\\R                     S\\R                     S\\R                     S\\\\\R                     4      S\\R                     S	\\   S
\\   S\\   S\\   S\\   S\\R                     S\\\4   4S jj5       r SS\\R                  S4   S\R                  S\R                  S\S
\4
S jjr\S\R                  S\S\S\R,                  S\R                  S\4S j5       rSrU =r$ )GraniteMoeSharedModeliE  r8   c           	      6  > [         TU ]  U5        UR                  U l        UR                  U l        [
        R                  " UR                  UR                  U R                  5      U l        [
        R                  " [        UR                  5       Vs/ sH  n[        X5      PM     sn5      U l        [        UR                  UR                  S9U l        SU l        UR$                  U l        UR                  U l        UR&                  U l        U R                  U R(                  -  U l        UR,                  U l        UR.                  U l        UR0                  U l        U R0                  S:X  a  [3        U5      OS U l        U R7                  5         g s  snf )Nr  Frope)r=   r>   pad_token_idpadding_idx
vocab_sizer   	Embeddingr?   embed_tokens
ModuleListr}   num_hidden_layersr  layersrY   r  normgradient_checkpointingembedding_multiplierr   r   r   rJ  
rope_thetaposition_embedding_typer?  
rotary_emb	post_initr  s      r3   r>   GraniteMoeSharedModel.__init__G  sE    !.. ++LL):):F<N<NPTP`P`ammNSTZTlTlNmnNm)&<Nmn
 ,F,>,>FDWDWX	&+#$*$?$?!!--33((DNN:'-'E'E$ ++'-'E'E$EIEaEaekEk9&Aqu 	! os   F	input_idsr   r   r.  inputs_embedsr  r"  output_hidden_statesr#  return_dictr  rL   c                    Ub  UOU R                   R                  nUb  UOU R                   R                  nUb  UOU R                   R                  nU
b  U
OU R                   R                  n
US L US L-  (       a  [        S5      eU R                  (       a/  U R                  (       a  U(       a  [        R                  S5        SnUc  U R                  U5      nXPR                  -  n[        U[        S 5      [        45      (       d  [        S5      eU(       a  Uc
  [        5       nUcD  Ub  UR!                  5       OSn["        R$                  " XUR&                  S   -   UR(                  S9nUc  UR+                  S5      nU R-                  X%XU5      nUnS nU R.                  b  U R/                  X5      nU(       a  SOS nU(       a  SOS nU	(       a  SOS nU R0                   HE  nU(       a  UU4-  nU" UUUUUUUU	US	9	nUS   nU(       a	  UUS   4-  nU	(       d  M<  UUS
   4-  nMG     U R3                  U5      nU(       a  UU4-  nU
(       d  [5        S XUU4 5       5      $ [7        UUUUUS9$ )Nz:You must specify exactly one of input_ids or inputs_embedszX`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.FzBThe `past_key_values` should be either a `Cache` object or `None`.r   r   r   r&   )r   r   r  r"  r  r  r#  r	  rN   c              3   ,   #    U H  oc  M  Uv   M     g 7fr  r&   ).0vs     r3   	<genexpr>0GraniteMoeSharedModel.forward.<locals>.<genexpr>  s      ^a^s   	)last_hidden_stater.  rK   
attentionsr   )r8   r"  ru  r  use_return_dictr   rl  r   r   r   rg  rm  r2  rC  r
   r   get_seq_lengthr,   arangero   r   r   _update_causal_maskrp  rj  rk  rn   r   )rH   rs  r   r   r.  rt  r  r"  ru  r#  rv  r  r   past_seen_tokensr   rK   r	  all_hidden_statesall_self_attnsall_router_logitsdecoder_layerlayer_outputss                         r3   rS   GraniteMoeSharedModel.forward`  s     2C1N-TXT_T_TqTq$8$D $++JjJj 	 "+!6IDKK<Q<Q	%0%<k$++B]B]-t";<YZZ&&4==Yj I  --i8M%(A(AA /DJ+>??abb0*nO!CRC^==?de"\\ ]5H5H5K"KTaThThN )33A6L..>L]

 &"??&"&//-"N #7BD0d"6BD![[M#!m%55!)*)."3#-%9$7
M *!,M =#3"55##!mB&7%99!- )0 		-0  -!11 )<M~^   &+++%+
 	
r2   r   input_tensorc           	         U R                   R                  S:X  a  Ub  US:H  R                  5       (       a  U$ g U R                   R                  S:X  a,  [        U[        R
                  5      (       a  [        U5      nU$ Ub  UR                  5       OSnUb  UR                  OSnU R                   R                  S:X  a5  U(       d.  U(       d'  [        R                  " UUUU R                  S9(       a  g UR                  nUR                  S   n	U(       a  UR                  5       n
O5[        U[        R
                  5      (       a  UR                  S	   OXi-   S-   n
U R                  UU	U
UUUR                  S   S
9nU R                   R                  S:X  aZ  UbW  UR                   R"                  S;   a=  U(       d6  [        R$                  " U5      R&                  n[        R(                  " X5      nU$ )Nflash_attention_2r  flex_attentionr   Fsdpa)rt  past_key_values_lengthis_trainingr   rN   )sequence_lengthtarget_lengthrd   r  
batch_size)cudaxpunpu)r8   r  anyr2  r,   rU   r   r  is_compileabler   _ignore_causal_mask_sdpar   rd   ro   get_max_cache_shape5_prepare_4d_causal_attention_mask_with_cache_positionr   rC  finfomin_unmask_unattended)rH   r   r  r  r.  r"  r  using_compilable_cacherd   r  r  r   	min_dtypes                r3   r  )GraniteMoeSharedModel._update_causal_mask  s    ;;++/BB)~/D.I.I.K.K%%;;++/??.%,,77!<^!L!!
 @O?Z?99;`aCRC^!?!?di ;;++v5>T]n%>>*'7 MM	 ""&,,Q/!+??AM nell;; $$R(%7!;  PP+')#))!, Q 
 KK,,6*%%**.DD%
 E*..I0CCK[Kr2   r  r  rd   r  c                    U b  U R                  5       S:X  a  U nU$ [        R                  " U5      R                  n[        R                  " X4XUR
                  S9nUS:w  a  [        R                  " USS9nU[        R                  " X$R
                  S9UR                  SS5      :  -  nUSSSS2SS24   R                  USSS5      nU b  UR                  5       nU R                  S   n	USS2SS2SS2SU	24   U SS2SSSS24   R                  UR
                  5      -   n
U
S:H  n
USS2SS2SS2SU	24   R                  X5      USS2SS2SS2SU	24'   U$ )	a  
Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape
`(batch_size, key_value_length)`, or if the input `attention_mask` is already 4D, do nothing.

Args:
    attention_mask (`torch.Tensor`):
        A 2D attention mask of shape `(batch_size, key_value_length)` or a 4D attention mask of shape
        `(batch_size, 1, query_length, key_value_length)`.
    sequence_length (`int`):
        The sequence length being processed.
    target_length (`int`):
        The target length: when generating with static cache, the mask should be as long as the static cache,
        to account for the 0 padding, the part of the cache that is not filled yet.
    dtype (`torch.dtype`):
        The dtype to use for the 4D attention mask.
    cache_position (`torch.Tensor`):
        Indices depicting the position of the input sequence tokens in the sequence.
    batch_size (`torch.Tensor`):
        Batch size.
N   )
fill_valuerd   r   r   )diagonalrx  rN   r   )rP   r,   r  r  fullr   triur  r   r   clonero   re   masked_fill)r   r  r  rd   r  r  r   r   r  mask_lengthpadding_masks              r3   r  KGraniteMoeSharedModel._prepare_4d_causal_attention_mask_with_cache_position  s}   < %.*<*<*>!*C(K* ' E*..I** 0Y\j\q\qK !##jjqA5<<>S>STWeWmWmnprsWtttK%dD!Q&67>>z1bRTUK))//1,2226*1aL[L+@ANSTVZ\`bcScDdDgDg&&E    ,q05@Aq,;,AV5W5c5c 6Aq!\k\12 r2   )rg  rm  rl  r   r?   rj  rJ  rk  r   rd  ro  rn  rp  re  )NNNNNNNNNNN)F)r'   r(   r)   r*   r   r>   r   r   r,   r-   rU   r   r
   listr*  r  rn   r   rS   r  staticmethodr/   rd   r  r1   rV   rW   s   @r3   r`  r`  E  s   5 2  151537KO59$(,0/3/3&*59l
E,,-l
 !.l
 u//0	l

 "%tE4E4E/F(F"GHl
   1 12l
 D>l
 $D>l
 'tnl
 'tnl
 d^l
 !!1!12l
 
u--	.l
 l
h #(BellK78B llB 	B
 B  BH 444 4 {{	4
 4 4 4r2   r`  gate_logitsrv   c                 d   U b  [        U [        5      (       d  g[        U [        5      (       aB  U S   R                  n[        R                  " U  Vs/ sH  oUR                  U5      PM     snSS9n[        R                  R                  R                  WSS9n[        R                  " XrSS9u  p[        R                  R                  R                  X5      n
Uc:  [        R                  " U
R                  5       SS9n[        R                  " USS9nGO"UR                  u  pUR                  S   X-  -  nUSSS2SS2SS4   R                  XXU45      R                  SX!5      R                  W5      n[        R                   " U
R                  5       U-  SS9[        R                   " USS9-  nUSSS2SS2S4   R                  XXR                  S   45      R                  SUR                  S   5      R                  U5      n[        R                   " UU-  SS9[        R                   " USS9-  nUR                  S   [#        UR                  R$                  5      -  n[        R                   " USS2UUUR                  S   -   24   UR'                  S5      -  5      nUU-  $ s  snf )ax  
Computes auxiliary load balancing loss as in Switch Transformer - implemented in Pytorch.

See Switch Transformer (https://huggingface.co/papers/2101.03961) for more details. This function implements the loss
function presented in equations (4) - (6) of the paper. It aims at penalizing cases where the routing between
experts is too unbalanced.

Args:
    gate_logits:
        Logits from the `gate`, should be a tuple of model.config.num_hidden_layers tensors of
        shape [batch_size X sequence_length, num_experts].
    num_experts:
        Number of experts
    top_k:
        The number of experts to route per-token, can be also interpreted as the `top-k` routing
        parameter.
    attention_mask (`torch.Tensor`, *optional*):
        The attention_mask used in forward function
        shape [batch_size X sequence_length] if not None.

Returns:
    The auxiliary loss.
Nr   rO   rN   r   )r2  rn   r   r,   r   re   r   r   r   r   one_hotrh   r   ro   r   r   r   r/   indexr   )r  rv   r   r   compute_device
layer_gateconcatenated_gate_logitsrouting_weightsr   selected_expertsexpert_masktokens_per_expertrouter_prob_per_expertr  r  ri  expert_attention_mask router_per_expert_attention_maskrankoverall_losss                       r3   load_balancing_loss_funcr  K  s   : *[%"@"@+u%%$Q..#(99^i-j^iPZmmN.K^i-jpq#r hh))112JPR1SO**_DA((%%--.>LK!JJ{'8'8':B "'O!C&4&:&:#
4::1=*B^_ 4AtT12V&OKXYWR,R	 	 "IIk&7&7&9<Q&QWXY\a\e\e!q]
 
 4At+,V&OEZEZ[\E]^_WR..q12R	 	) "'?=]+]cd!ehmhqhq,!i
 "
   #c/*@*@*F*F&GGD99!TD?+@+@+C$CCCDG]GgGghiGjjL +%%a .ks   J-c                      ^  \ rS rSrS/rS\4U 4S jjrS rS r\	             SS\
\R                     S\
\R                     S	\
\R                     S
\
\\\\R"                     4      S\
\R"                     S\
\R                     S\
\   S\
\   S\
\   S\
\   S\
\   S\
\R                     S\\\R                  4   S\\\4   4S jj5       rSrU =r$ )GraniteMoeSharedForCausalLMi  zlm_head.weightr8   c                 J  > [         TU ]  U5        [        U5      U l        UR                  U l        [
        R                  " UR                  UR                  SS9U l        UR                  U l	        UR                  U l        UR                  U l        U R                  5         g )NFr;   )r=   r>   r`  r-  re  r   rD   r?   lm_headrouter_aux_loss_coefr   rv   r   rq  rG   s     r3   r>   $GraniteMoeSharedForCausalLM.__init__  s     *62
 ++yy!3!3V5F5FUS$*$?$?!!33#)#=#=  	r2   c                     Xl         g r  r-  )rH   decoders     r3   set_decoder'GraniteMoeSharedForCausalLM.set_decoder  s    
r2   c                     U R                   $ r  r  rp   s    r3   get_decoder'GraniteMoeSharedForCausalLM.get_decoder  s    zzr2   rs  r   r   r.  rt  labelsr  r"  ru  r#  rv  r  logits_to_keeprL   c                    Ub  UOU R                   R                  nU
b  U
OU R                   R                  n
U	b  U	OU R                   R                  n	Ub  UOU R                   R                  nU R
                  " SUUUUUUUU	U
UUS.UD6nUS   n[        U[        5      (       a  [        U* S5      OUnU R                  USS2USS24   5      nUU R                   R                  -  nSnUb:  UR                  5       nU R                  " UU4SU R                   R                  0UD6nSnU
(       af  [        U(       a  UR                  OUS   U R                   U R"                  U5      nUb+  UU R$                  UR'                  UR(                  5      -  -  nU(       d!  U4USS -   nU
(       a  U4U-   nUb  U4U-   $ U$ [+        UUUUR,                  UR.                  UR0                  UR                  S9$ )	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.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.vocab_size]`.

Example:

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

>>> model = GraniteMoeSharedForCausalLM.from_pretrained("ibm/PowerMoE-3b")
>>> tokenizer = AutoTokenizer.from_pretrained("ibm/PowerMoE-3b")

>>> prompt = "Hey, are you conscious? Can you talk to me?"
>>> inputs = tokenizer(prompt, return_tensors="pt")

>>> # Generate
>>> generate_ids = model.generate(inputs.input_ids, max_length=30)
>>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
"Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
```N)rs  r   r   r.  rt  r  r"  ru  r#  rv  r  r   re  rN   r   )lossaux_lossr   r.  rK   r  r   r&   )r8   r"  r#  ru  r  r-  r2  r/   slicer  logits_scalingr   loss_functionre  r  r   rv   r   r  re   r   r   r.  rK   r  )rH   rs  r   r   r.  rt  r  r  r"  ru  r#  rv  r  r  r   r(  rK   slice_indicesr   r  r  outputs                         r3   rS   #GraniteMoeSharedForCausalLM.forward  s)   P 2C1N-TXT_T_TqTq$8$D $++JjJj 	 %9$D $++JjJj 	 &1%<k$++B]B] ** 
)%+'/!5!5#)
 
   
8B>SV8W8W~ot4]kmA}a,?@A$++444\\^F%%  ;;11 	D /)4%%'"+  ((	H !11HKK4LLLY,F#"v-'+'7D7V#CVC(#33!//))!//
 	
r2   )r  r-  rv   r   r  re  )NNNNNNNNNNNNr   )r'   r(   r)   r*   _tied_weights_keysr   r>   r  r  r   r   r,   r-   rU   r   r
   r  r*  r  r/   rn   r   rS   r1   rV   rW   s   @r3   r  r    s   *+5   151537KO59-1$(,0/3/3&*5934k
E,,-k
 !.k
 u//0	k

 "%tE4E4E/F(F"GHk
   1 12k
 ))*k
 D>k
 $D>k
 'tnk
 'tnk
 d^k
 !!1!12k
 c5<</0k
  
u//	0!k
 k
r2   r  )r  r`  r,  )Nr   )r  )Nr:   N)Etypingr   r   r   r   r,   torch.nn.functionalr   r   r   activationsr	   cache_utilsr
   r   
generationr   modeling_attn_mask_utilsr   modeling_layersr   modeling_outputsr   r   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   r   configuration_granitemoesharedr   !torch.nn.attention.flex_attentionr   integrations.flex_attentionr   
get_loggerr'   r   r   Moduler6   rY   rt   r   r   r   r   rU   r/   r   r   r   r   r  r,  r?  r`  rn   r  r  __all__r&   r2   r3   <module>r     sW  , 8 7     ! . ) > 9 j j K F & J J B  !!;J 
		H	%)5 2")) 4Jbii J(*bii *Z-S -S`9+")) 9+x(6	UU\\ 	U# 	U%,, 	U& %II%<<% 
% <<	%
 U\\*% % %:S)		 S)l[#= [| To T T"<bii <D B; B BN "&
-1	R&u||U5<<%8$>?R&#R& U\\*	R&
 5<<R&jB
"A? B
J fr2   