
    Z j{                     X   S SK Jr  S SKJr  S SKrS SKJr  S SK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JrJr  SSKJr  SSKJr  SSKJrJr  SSKJrJ r   SSK!J"r"J#r#  SSK$J%r%  SSK&J'r'J(r(  SSK)J*r*J+r+J,r,  SSK-J.r.  SSK/J0r0  \" S5       " S S\Rb                  5      5       r2 " S S\Rb                  5      r3 " S S\Rb                  5      r4 " S S\Rb                  5      r5 " S  S!\Rb                  5      r6S" r7\" S#5      S@S$ j5       r8S%\Rr                  S&\:S'\Rr                  4S( jr; SAS)\Rb                  S*\Rr                  S+\Rr                  S,\Rr                  S-\Rr                  S-  S.\<S/\<S0\%\'   4S1 jjr=\" \85       " S2 S3\Rb                  5      5       r> " S4 S5\5      r?\( " S6 S7\#5      5       r@\( " S8 S9\@5      5       rA   SBS:\Rr                  \B\Rr                     -  S-  S;\:S-  S-\Rr                  S-  S'\Rr                  \:-  4S< jjrC\( " S= S>\@\5      5       rD/ S?QrEg)C    )Callable)OptionalN)nn)
functional   )initialization)ACT2FN)CacheDynamicCache)GenerationMixin)use_kernel_forward_from_hubuse_kernel_func_from_hubuse_kernelized_func)create_causal_mask)GradientCheckpointingLayer)MoeCausalLMOutputWithPastMoeModelOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstring)can_return_tuplemaybe_autocastmerge_with_config_defaults)capture_outputs   )GraniteMoeConfigRMSNormc                   x   ^  \ rS rSrS
S\SS4U 4S jjjrS\R                  S\R                  4S jrS r	S	r
U =r$ )GraniteMoeRMSNorm.   epsreturnNc                    > [         TU ]  5         [        R                  " [        R
                  " U5      5      U l        X l        g)z0
GraniteMoeRMSNorm is equivalent to T5LayerNorm
N)super__init__r   	Parametertorchonesweightvariance_epsilon)selfhidden_sizer%   	__class__s      ڃ/root/GenerationalWealth/GenerationalWealth/venv/lib/python3.13/site-packages/transformers/models/granitemoe/modeling_granitemoe.pyr)   GraniteMoeRMSNorm.__init__0   s/     	ll5::k#:; #    hidden_statesc                    UR                   nUR                  [        R                  5      nUR	                  S5      R                  SSS9nU[        R                  " X0R                  -   5      -  nU R                  UR                  U5      -  $ )N   T)keepdim)	dtypetor+   float32powmeanrsqrtr.   r-   )r/   r5   input_dtypevariances       r2   forwardGraniteMoeRMSNorm.forward8   sw    #))%((7 $$Q',,R,>%H?T?T4T(UU{{]--k:::r4   c                 ^    [        U R                  R                  5       SU R                   3$ )Nz, eps=)tupler-   shaper.   )r/   s    r2   
extra_reprGraniteMoeRMSNorm.extra_repr?   s*    ))*+6$2G2G1HIIr4   )r.   r-   )gư>)__name__
__module____qualname____firstlineno__floatr)   r+   TensorrB   rG   __static_attributes____classcell__r1   s   @r2   r#   r#   .   sB    $ $$ $ $;U\\ ;ell ;J Jr4   r#   c                      ^  \ rS rSr% \R
                  \S'   SS\4U 4S jjjr\	   SS\S-  S\
S   S\S-  S	\S
\4   4S jj5       r\R                  " 5       \S 5       5       rSrU =r$ )GraniteMoeRotaryEmbeddingC   inv_freqNconfigc                   > [         TU ]  5         UR                  U l        UR                  U l        Xl        U R
                  R                  S   U l        U R                  nU R                  S:w  a  [        U R                     nU" U R
                  U5      u  o@l
        U R                  SUSS9  U R                  SUR                  5       SS9  g )N	rope_typedefaultrU   F)
persistentoriginal_inv_freq)r(   r)   max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenrV   rope_parametersrX   compute_default_rope_parametersr   attention_scalingregister_bufferclone)r/   rV   devicerope_init_fnrU   r1   s        r2   r)   "GraniteMoeRotaryEmbedding.__init__F   s    "("@"@$*$B$B!44[A!%!E!E>>Y&.t~~>L+7V+L((ZeD0(..2BuUr4   rd   ztorch.deviceseq_lenr&   ztorch.Tensorc           	         U R                   S   n[        U SS5      =(       d    U R                  U R                  -  nSnSU[        R
                  " SUS[        R                  S9R                  U[        R                  S9U-  -  -  nXe4$ )	aH  
Computes the inverse frequencies according to the original RoPE implementation
Args:
    config ([`~transformers.PreTrainedConfig`]):
        The model configuration.
    device (`torch.device`):
        The device to use for initialization of the inverse frequencies.
    seq_len (`int`, *optional*):
        The current sequence length. Unused for this type of RoPE.
Returns:
    Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the
    post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE).

rope_thetahead_dimNg      ?r   r7   r:   )rd   r:   )	r_   getattrr0   num_attention_headsr+   arangeint64r;   rM   )rV   rd   rg   basedimattention_factorrU   s          r2   r`   9GraniteMoeRotaryEmbedding.compute_default_rope_parametersV   s    & %%l3fj$/c63E3EIcIc3c U\\!S!5;;?BB&X]XcXcBdgjjk
 ))r4   c                 L   U R                   S S S 2S 4   R                  5       R                  UR                  S   SS5      R	                  UR
                  5      nUS S 2S S S 24   R                  5       n[        UR
                  R                  [        5      (       a0  UR
                  R                  S:w  a  UR
                  R                  OSn[        USS9   UR                  5       UR                  5       -  R                  SS5      n[        R                  " Xf4SS	9nUR                  5       U R                  -  nUR                  5       U R                  -  n	S S S 5        WR	                  UR                   S
9W	R	                  UR                   S
94$ ! , (       d  f       N@= f)Nr   r8   r   mpscpuF)device_typeenabledr7   rq   rk   )rU   rM   expandrF   r;   rd   
isinstancetypestrr   	transposer+   catcosra   sinr:   )
r/   xposition_idsinv_freq_expandedposition_ids_expandedrw   freqsembr   r   s
             r2   rB   !GraniteMoeRotaryEmbedding.forwardt   sN    !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#)ra   rV   r]   r^   rX   )NNNN)rI   rJ   rK   rL   r+   rN   __annotations__r    r)   staticmethodr   intrE   rM   r`   no_gradr   rB   rO   rP   rQ   s   @r2   rS   rS   C   s    llV/ V V  *.+/"* 4'*(* t* 
~u$	%	* *: ]]_<  <r4   rS   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$ )
GraniteMoeParallelExperts   num_experts
input_sizeoutput_sizer&   Nc                    > [         TU ]  5         [        R                  " [        R
                  " XU5      5      U l        Xl        X l        X0l	        g)aW  
Initialize the GraniteMoeParallelExperts 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-   r   r   r   )r/   r   r   r   r1   s       r2   r)   "GraniteMoeParallelExperts.__init__   s<    " 	ll5;;{#TU&$&r4   c                     UR                  USS9n/ n[        U R                  5       H8  nUR                  [        R
                  " X5   U R                  U   5      5        M:     [        R                  " USS9nU$ )z
Forward pass of the GraniteMoeParallelExperts module.

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

Returns:
    Tensor: Output tensor.
r   ry   )	splitranger   appendFlinearr-   r+   r   )r/   inputsexpert_size
input_listoutput_listiresultss          r2   rB   !GraniteMoeParallelExperts.forward   sh     \\+1\5
t''(Aqxx
t{{1~FG )))KQ/r4   )r   r   r   r-   	rI   rJ   rK   rL   r   r)   rB   rO   rP   rQ   s   @r2   r   r      s.    'C 'S 's 't '. r4   r   c                   >   ^  \ rS rSrS\S\S\4U 4S jjrS rSrU =r$ )GraniteMoeTopKGating   r   r   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.
FbiasN)r(   r)   r   r   r   r   Linearlayer)r/   r   r   r   r1   s       r2   r)   GraniteMoeTopKGating.__init__   s2     	&$
YYzUC
r4   c                 z   U R                  U5      R                  5       nUR                  U R                  SS9u  p4[        R
                  " USS9R                  U5      n[        R                  " UR                  S5      U R                  /UR                  UR                  S9nUR                  SUS5      nUR                  5       R                  S5      nUR                  5       nUR!                  5       n	U	R#                  S5      u  pUR%                  U R                  SS9nUR!                  5       nX[   nXXU4$ )Nr   ry   r   r:   rd   trunc)rounding_mode)r   rM   topkr   r+   softmaxtype_aszerossizer   r:   rd   scatterlongsumtolistflattensortdiv)r/   r5   logitstop_k_logitstop_k_indicestop_k_gatesr   gatesr   top_k_experts_index_sorted_expertsbatch_indexbatch_gatess                 r2   rB   GraniteMoeTopKGating.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Rr4   )r   r   r   r   r   rQ   s   @r2   r   r      s-    D3 DS D D(S Sr4   r   c                   :   ^  \ rS rSrSrS\4U 4S jjrS rSrU =r	$ )GraniteMoeMoE   z
A Sparsely gated mixture of experts layer with 1-layer Feed-Forward networks as experts.

Args:
    config:
        Configuration object with model hyperparameters.
rV   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 )Nr7   )r   r   r   )r(   r)   r0   r   intermediate_sizer	   
hidden_act
activationr   num_local_expertsinput_linearoutput_linearr   num_experts_per_tokrouterr/   rV   r1   s     r2   r)   GraniteMoeMoE.__init__   s     ,,!33 !2!235f6N6NPTP_P_aeaqaqtuauv6v7O7OQUQaQacgcrcrs*00,,
r4   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Xl5      nUR                  X#U R                  5      nU$ )Nr8   r7   ry   r   r   r   )r   reshaper   r   chunkr   r   r+   r   r   r:   rd   	index_addview)r/   layer_inputbszlengthemb_sizer   r   r   r   expert_inputsr5   chunked_hidden_statesexpert_outputsr   layer_outputs                  r2   rB   GraniteMoeMoE.forward   s    + 0 0 2X!))"h76:kk+6N3!#0))-E - 3 3A2 3 >(=a(@ADYZ[D\\++MG'ag*>>S\4??;>CWCW`n`u`uvq+F#((dooFr4   )r   r0   r   r   r   r   )
rI   rJ   rK   rL   __doc__r    r)   rB   rO   rP   rQ   s   @r2   r   r      s    
/ 
 r4   r   c                     U SSU R                   S   S-  24   nU SU R                   S   S-  S24   n[        R                  " U* U4SS9$ )z*Rotates half the hidden dims of the input..Nr8   r7   ry   )rF   r+   r   )r   x1x2s      r2   rotate_halfr     sZ    	
3"!''"+"""	#B	
3q ""	#B99rc2YB''r4   rotary_pos_embc                     UR                  U5      nUR                  U5      nX-  [        U 5      U-  -   nX-  [        U5      U-  -   nXV4$ )aI  Applies Rotary Position Embedding to the query and key tensors.

Args:
    q (`torch.Tensor`): The query tensor.
    k (`torch.Tensor`): The key tensor.
    cos (`torch.Tensor`): The cosine part of the rotary embedding.
    sin (`torch.Tensor`): The sine part of the rotary embedding.
    unsqueeze_dim (`int`, *optional*, defaults to 1):
        The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
        sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
        that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
        k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
        cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
        the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
Returns:
    `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
)	unsqueezer   )qkr   r   unsqueeze_dimq_embedk_embeds          r2   apply_rotary_pos_embr     sS    & --
&C
--
&Cw;q>C/0Gw;q>C/0Gr4   r5   n_repr&   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)rF   rz   r   )r5   r   batchnum_key_value_headsslenrj   s         r2   	repeat_kvr   .  s_    
 2?1D1D.Ez!!Qa"23::5W\dlmM  e(CTTTr4   modulequerykeyvalueattention_maskscalingdropoutkwargsc                    [        X R                  5      n[        X0R                  5      n	[        R                  " XR	                  SS5      5      U-  n
Ub  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$ )Nr7   r   r8   )rq   r:   )ptrainingr   )r   num_key_value_groupsr+   matmulr~   r   r   r   r<   r;   r:   r   r  
contiguous)r   r   r   r   r   r   r   r  
key_statesvalue_statesattn_weightsattn_outputs               r2   eager_attention_forwardr  :  s     3 ; ;<JU$?$?@L<<';';Aq'ABWLL!#4==((2U]](SVVW\WbWbcL==((6??([L,,|:K''1-88:K$$r4   c                     ^  \ rS rSrSrS\S\4U 4S jjr   SS\R                  S\
\R                  \R                  4   S-  S	\R                  S-  S
\S-  S\\   S\
\R                  \R                  4   4S jjrSrU =r$ )GraniteMoeAttentioniS  z=Multi-headed attention from 'Attention Is All You Need' paperrV   	layer_idxc                 J  > [         TU ]  5         Xl        X l        [	        USUR
                  UR                  -  5      U l        UR                  UR                  -  U l	        UR                  U l        UR                  U l        S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                  -  UR                   S9U l        [        R                  " UR                  U R                  -  UR
                  UR                   S9U l        g )Nrj   Tr   )r(   r)   rV   r  rl   r0   rm   rj   r   r  attention_multiplierr   attention_dropout	is_causalr   r   attention_biasq_projk_projv_projo_projr/   rV   r  r1   s      r2   r)   GraniteMoeAttention.__init__W  sF   "
F4F4F&JdJd4de$*$>$>&B\B\$\!22!'!9!9ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii&&68J8JQWQfQf
r4   Nr5   position_embeddingsr   past_key_valuesr  r&   c                    UR                   S S n/ UQSPU R                  P7nU R                  U5      R                  U5      R	                  SS5      nU R                  U5      R                  U5      R	                  SS5      n	U R                  U5      R                  U5      R	                  SS5      n
Uu  p[        XX5      u  pUb  UR                  XU R                  5      u  p[        R                  " U R                  R                  [        5      nU" U UU	U
U4U R                  (       d  SOU R                   U R"                  S.UD6u  pUR$                  " / UQSP76 R'                  5       nU R)                  U5      nX4$ )Nr8   r   r7           )r   r   )rF   rj   r  r   r~   r  r  r   updater  r   get_interfacerV   _attn_implementationr  r  r  r   r   r  r  )r/   r5   r  r   r  r  input_shapehidden_shapequery_statesr  r	  r   r   attention_interfacer  r
  s                   r2   rB   GraniteMoeAttention.forwardn  s~    $))#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&#7RU#[ &'6'='=jX\XfXf'g$J(?(M(MKK,,.E)
 %8	%
  $}}C$2H2HLL	%
 	%
! "));;;;FFHkk+.((r4   )r  rV   rj   r  r  r  r  r  r  r   r  r   )rI   rJ   rK   rL   r   r    r   r)   r+   rN   rE   r
   r   r   rB   rO   rP   rQ   s   @r2   r  r  S  s    G
/ 
C 
4 IM.2(,&)||&) #5<<#=>E&) t+	&)
 &) +,&) 
u||U\\)	*&) &)r4   r  c                      ^  \ rS rSrS\S\4U 4S jjr   SS\R                  S\R                  S-  S\	S-  S	\
\R                  \R                  4   S-  S
\R                  4
S jjrSrU =r$ )GraniteMoeDecoderLayeri  rV   r  c                 .  > [         TU ]  5         UR                  U l        [        XS9U l        [        UR                  UR                  S9U l        [        UR                  UR                  S9U l        [        U5      U l
        UR                  U l        g )N)rV   r  r%   )r(   r)   r0   r  	self_attnr#   rms_norm_epsinput_layernormpost_attention_layernormr   block_sparse_moeresidual_multiplierr  s      r2   r)   GraniteMoeDecoderLayer.__init__  sz    !--,FP01C1CI\I\](9&:L:LRXReRe(f% -f 5#)#=#= r4   Nr5   r   r  r  r&   c                     UnU R                  U5      nU R                  " SUUUUS.UD6u  pXaU R                  -  -   nUnU R                  U5      nU R	                  U5      nXaU R                  -  -   nU$ )N)r5   r   r  r   )r-  r+  r0  r.  r/  )r/   r5   r   r  r  r  residualr   s           r2   rB   GraniteMoeDecoderLayer.forward  s     !,,];>> 
')+ 3	

 
 !43K3K#KK 55mD--m< 43K3K#KKr4   )r/  r0   r-  r.  r0  r+  r   )rI   rJ   rK   rL   r    r   r)   r+   rN   r
   rE   rB   rO   rP   rQ   s   @r2   r(  r(    s    >/ >C > /3(,HL|| t+ 	
 #5<<#=>E 
 r4   r(  c                      ^  \ rS rSr% \\S'   SrSrS/rS/r	Sr
SrSrSrSr\\S.r\R&                  " 5       U 4S	 j5       rS
rU =r$ )GraniteMoePreTrainedModeli  rV   modelTr(  r  F)r5   
attentionsc                    > [         TU ]  U5        [        U[        5      (       a5  [        R
                  " UR                  SU R                  R                  S9  g g )Nr  )r>   std)	r(   _init_weightsr{   r   initnormal_r-   rV   initializer_range)r/   r   r1   s     r2   r<  'GraniteMoePreTrainedModel._init_weights  sA    f%f788LLSdkk6S6ST 9r4   r3  )rI   rJ   rK   rL   r    r   base_model_prefixsupports_gradient_checkpointing_no_split_modules_skip_keys_device_placement_supports_flash_attn_supports_sdpa_supports_flex_attn_can_compile_fullgraph_supports_attention_backendr(  r  _can_record_outputsr+   r   r<  rO   rP   rQ   s   @r2   r7  r7    sn    &*#12#4"5N""&/)
 ]]_U Ur4   r7  c                     ^  \ rS rSrS\4U 4S jjr\\\      SS\	R                  S-  S\	R                  S-  S\	R                  S-  S\S-  S	\	R                  S-  S
\S-  S\\   S\4S jj5       5       5       rSrU =r$ )GraniteMoeModeli  rV   c           	      *  > [         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        [#        US9U l        SU l        UR(                  U l        U R+                  5         g s  snf )Nr*  rV   F)r(   r)   pad_token_idpadding_idx
vocab_sizer   	Embeddingr0   embed_tokens
ModuleListr   num_hidden_layersr(  layersr#   r,  normrS   
rotary_embgradient_checkpointingembedding_multiplier	post_initr  s      r2   r)   GraniteMoeModel.__init__  s     !.. ++LL):):F<N<NPTP`P`ammHMfNfNfHghHg9#F6Hgh
 &f&8&8f>Q>QR	36B&+#$*$?$?! 	 is   DN	input_idsr   r   r  inputs_embeds	use_cacher  r&   c           
      ^   US L US L-  (       a  [        S5      eU(       a  Uc  [        U R                  S9nUc  U R                  U5      nUcU  Ub  UR	                  5       OSn[
        R                  " UR                  S   UR                  S9U-   nUR                  S5      n[        U R                  UUUUS9n	XPR                  -  nUn
U R                  X5      nU R                  S U R                  R                    H  nU" U
4UU	UUUS.UD6n
M     U R                  U
5      n
[!        U
US9$ )	Nz:You must specify exactly one of input_ids or inputs_embedsrN  r   r   )rd   )rV   r^  r   r  r   )r  r   r   r  r_  )last_hidden_stater  )
ValueErrorr   rV   rS  get_seq_lengthr+   rn   rF   rd   r   r   rZ  rX  rV  rU  rW  r   )r/   r]  r   r   r  r^  r_  r  past_seen_tokenscausal_maskr5   r  decoder_layers                r2   rB   GraniteMoeModel.forward  sT    -t";<YZZ0*$++>O  --i8MCRC^==?de <<(;(;A(>}G[G[\_ooL'11!4L(;;')+%
 &(A(AA% #oomJ![[)H4;;+H+HIM)$7*) /# M J 		-0%++
 	
r4   )rS  rZ  rY  rV  rW  rP  rX  rQ  )NNNNNN)rI   rJ   rK   rL   r    r)   r   r   r   r+   
LongTensorrN   r
   FloatTensorboolr   r   r   rB   rO   rP   rQ   s   @r2   rL  rL    s    / "   .2.204(,26!%5
##d*5
 t+5
 &&-	5

 5
 ((4/5
 $;5
 +,5
 
 5
    5
r4   rL  gate_logitsr   c                    U b  [        U [        5      (       d  g[        U [        5      (       aC  U S   R                  n[        R                  " U  Vs/ s H  oUR                  U5      PM     snSS9n[        R                  R                  R                  WSS9n[        R                  " XrSS9u  p[        R                  R                  R                  X5      n
Uc:  [        R                  " U
R                  5       SS9n[        R                  " USS9nGOUR                  u  pUR                  S   X-  -  nUSSS2SS2SS4   R                  XXU45      R                  SX!5      R                  W5      n[        R                   " U
R                  5       U-  SS9[        R                   " USS9-  nUSSS2SS2S4   R                  XX45      R                  SU5      R                  U5      n[        R                   " UU-  SS9[        R                   " USS9-  n[        R                   " XR#                  S5      -  5      nUU-  $ 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   ry   r8   )r{   rE   rd   r+   r   r;   r   r   r   r   one_hotr>   rM   rF   rz   r   r   r   )rk  r   r   r   compute_device
layer_gateconcatenated_gate_logitsrouting_weightsr   selected_expertsexpert_masktokens_per_expertrouter_prob_per_expert
batch_sizesequence_lengthrU  expert_attention_mask router_per_expert_attention_maskoverall_losss                      r2   load_balancing_loss_funcr{     s+   : *[%"@"@+u%%$Q..#(99^i-j^iPZmmN.K^i-jpq#r hh))112JPR1SO**_DA((%%--.>LK!JJ{'8'8':B "'O!C&4&:&:#
4::1=*B^_ 4AtT12V&OKXYWR,R	 	 "IIk&7&7&9<Q&QWXY\a\e\e!q]
 
 4At+,V&OQRWR%R	 	) "'?=]+]cd!ehmhqhq,!i
 "
 99.1Q1QRS1TTUL+%%[ .ks   Ic                   T  ^  \ rS rSrSS0rSS0rSS/S/40rS\4U 4S	 jjr\	\
        SS\R                  S
-  S\R                  S
-  S\R                  S
-  S\S
-  S\R                  S
-  S\R                  S
-  S\S
-  S\\R                  -  S\\-  4S jj5       5       rSrU =r$ )GraniteMoeForCausalLMir  zlm_head.weightzmodel.embed_tokens.weightlm_headcolwise_gather_outputr5   r   rV   c                 l  > [         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                  U l        U R                  5         g )NFr   )r(   r)   rL  r8  rQ  r   r   r0   r~  router_aux_loss_coefr   r   r   logits_scalingr[  r   s     r2   r)   GraniteMoeForCausalLM.__init__x  s     $V,
 ++yy!3!3V5F5FUS$*$?$?!!33#)#=#= $33 	r4   Nr]  r   r   r  r^  labelsoutput_router_logitslogits_to_keepr&   c	           
         Ub  UOU R                   R                  nU R                  " SUUUUUS.U	D6n
U
R                  n[	        U[
        5      (       a  [        U* S5      OUnU R                  USS2USS24   5      nXR                   R                  -  nSnUb*  U R                  " UU4SU R                   R                  0U	D6nSnU(       aY  [        U
R                  U R                  U R                  U5      nUb*  XR                  UR!                  UR"                  5      -  -  n[%        UUUU
R&                  U
R(                  U
R*                  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, GraniteMoeForCausalLM

>>> model = GraniteMoeForCausalLM.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)r]  r   r   r  r^  rQ  )lossaux_lossr   r  r5   r9  router_logitsr3  )rV   r  r8  ra  r{   r   slicer~  r  loss_functionrQ  r{  r  r   r   r  r;   rd   r   r  r5   r9  )r/   r]  r   r   r  r^  r  r  r  r  outputsr5   slice_indicesr   r  r  s                   r2   rB   GraniteMoeForCausalLM.forward  sw   J %9$D $++JjJj 	 ** 
)%+'
 
  118B>SV8W8W~ot4]kmA}a,?@A++444%%  ;;11 	D /%%  ((	H !11HKK4LLL(#33!//))!//
 	
r4   )r~  r  r8  r   r   r  rQ  )NNNNNNNr   )rI   rJ   rK   rL   _tied_weights_keys_tp_plan_pp_planr    r)   r   r   r+   rh  rN   r
   ri  rj  r   rE   r   rB   rO   rP   rQ   s   @r2   r}  r}  r  s    *,GH23H_-z:;H/   .2.204(,26*.,0-.Q
##d*Q
 t+Q
 &&-	Q

 Q
 ((4/Q
   4'Q
 #TkQ
 ell*Q
 
*	*Q
  Q
r4   r}  )r}  rL  r7  )r   )r  )Nr7   N)Fcollections.abcr   typingr   r+   r   torch.nnr   r    r   r=  activationsr	   cache_utilsr
   r   
generationr   integrationsr   r   r   masking_utilsr   modeling_layersr   modeling_outputsr   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   utils.genericr   r   r   utils.output_capturingr   configuration_granitemoer    Moduler#   rS   r   r   r   r   r   rN   r   r   rM   r  r  r(  r7  rL  rE   r{  r}  __all__r3  r4   r2   <module>r     sv  , %    $ & ! . ) f f / 9 Q K F & 7 Y Y 5 6 Y'J		 J (J(><		 ><B*		 *Z.S299 .Sb(BII (V( *+ ,2	UU\\ 	U# 	U%,, 	U& %II%<<% 
% <<	%
 LL4'% % % '(%2 )*@)")) @) +@)F 7  F U U U. J
/ J
 J
^ #
*.	O&ell 33d:O&tO& LL4'	O&
 \\CO&d e
5 e
 e
P Tr4   