
    i                        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 ddlmZ dd	lmZmZ dd
lmZ ddlmZmZmZ ddlmZ ddlmZ ddlmZ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) ddl*m+Z+m,Z,m-Z- ddl.m/Z/  G d ded      Z0 G d dejb                        Z2 ed       G d dejb                               Z3 G d d ejb                        Z4 G d! d"ejb                        Z5 G d# d$ejb                        Z6d% Z7 ed&      dEd'       Z8d(ejr                  d)e:d*ejr                  fd+Z;	 dFd,ejb                  d-ejr                  d.ejr                  d/ejr                  d0ejr                  dz  d1e<d2e<d3e&e(   fd4Z= ee8       G d5 d6ejb                               Z> G d7 d8e      Z?e) G d9 d:e$             Z@ G d; d<ejb                        ZAe) G d= d>e@             ZB	 	 	 dGd?ejr                  eCejr                     z  dz  d@e:dz  d0ejr                  dz  d*ejr                  e:z  fdAZDe) G dB dCe@e             ZEg dDZFy)H    )Callable)Optional	TypedDictN)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check_model_inputsmaybe_autocast   )GraniteMoeSharedConfigc                       e Zd ZU dZej
                  ed<   ej
                  ed<   eed<   eed<   ej                  ed<   y)GraniteFlashAttentionKwargsaT  
    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.

    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_idxN)	__name__
__module____qualname____doc__torch
LongTensor__annotations__int	IntTensor     /home/obispo/Crisostomo_bridge/mision_env/lib/python3.12/site-packages/transformers/models/granitemoeshared/modeling_granitemoeshared.pyr"   r"   ,   s7      ######__r2   r"   F)totalc                   `     e Zd ZdZdef fdZdej                  dej                  fdZ xZ	S )GraniteMoeSharedMLPz~
    MLP layer for shared experts

    Args:
        config:
            Configuration object with model hyperparameters.
    configc                 `   t         |           |j                  | _        |j                  | _        t
        |j                     | _        t        j                  | j                  | j                  dz  d      | _
        t        j                  | j                  | j                  d      | _        y )N   Fbias)super__init__hidden_size
input_sizeshared_intermediate_sizer
   
hidden_act
activationr   Linearinput_linearoutput_linearselfr7   	__class__s     r3   r=   zGraniteMoeSharedMLP.__init__M   s     ,,!:: !2!23IIdoot7G7G!7KRWXYYt'7'7uUr2   hidden_statesreturnc                     | j                  |      }|j                  dd      }| j                  |d         |d   z  }| j                  |      }|S )Nr9   dimr   r   )rD   chunkrB   rE   )rG   rI   chunked_hidden_statess      r3   forwardzGraniteMoeSharedMLP.forwardV   s^    ))-8 - 3 3A2 3 >(=a(@ADYZ[D\\**=9r2   )
r(   r)   r*   r+   r    r=   r,   TensorrQ   __classcell__rH   s   @r3   r6   r6   D   s2    V5 VU\\ ell r2   r6   RMSNormc                   ,     e Zd Zd fd	Zd Zd Z xZS )GraniteMoeSharedRMSNormc                     t         |           t        j                  t	        j
                  |            | _        || _        y)zF
        GraniteMoeSharedRMSNorm is equivalent to T5LayerNorm
        N)r<   r=   r   	Parameterr,   onesweightvariance_epsilon)rG   r>   epsrH   s      r3   r=   z GraniteMoeSharedRMSNorm.__init__`   s1     	ll5::k#:; #r2   c                 "   |j                   }|j                  t        j                        }|j	                  d      j                  dd      }|t        j                  || j                  z         z  }| j                  |j                  |      z  S )Nr9   rL   T)keepdim)	dtypetor,   float32powmeanrsqrtr\   r[   )rG   rI   input_dtypevariances       r3   rQ   zGraniteMoeSharedRMSNorm.forwardh   sy    #))%((7 $$Q',,R,>%Ht?T?T4T(UU{{]--k:::r2   c                 ^    t        | j                  j                         d| j                   S )Nz, eps=)tupler[   shaper\   )rG   s    r3   
extra_reprz"GraniteMoeSharedRMSNorm.extra_repro   s*    ))*+6$2G2G1HIIr2   )gư>)r(   r)   r*   r=   rQ   rk   rS   rT   s   @r3   rW   rW   ^   s    $;Jr2   rW   c                   6     e Zd Zdedededdf fdZd Z xZS )GraniteMoeSharedParallelExpertsnum_expertsr?   output_sizerJ   Nc                     t         |           t        j                  t	        j
                  |||            | _        || _        || _        || _	        y)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   rY   r,   emptyr[   rn   r?   ro   )rG   rn   r?   ro   rH   s       r3   r=   z(GraniteMoeSharedParallelExperts.__init__t   sD    " 	ll5;;{K#TU&$&r2   c                     |j                  |d      }g }t        | j                        D ]7  }|j                  t	        j
                  ||   | j                  |                9 t        j                  |d      }|S )a  
        Forward pass of the GraniteMoeSharedParallelExperts module.

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

        Returns:
            Tensor: Output tensor.
        r   rM   )	splitrangern   appendFlinearr[   r,   cat)rG   inputsexpert_size
input_listoutput_listiresultss          r3   rQ   z'GraniteMoeSharedParallelExperts.forward   sq     \\+1\5
t''( 	HAqxx
1t{{1~FG	H))KQ/r2   r(   r)   r*   r/   r=   rQ   rS   rT   s   @r3   rm   rm   s   s)    'C 'S 's 't '.r2   rm   c                   2     e Zd Zdededef fdZd Z xZS )GraniteMoeSharedTopKGatingr?   rn   top_kc                     t         |           || _        || _        || _        t        j                  ||d      | _        y)a  
        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=   rn   r?   r   r   rC   layer)rG   r?   rn   r   rH   s       r3   r=   z#GraniteMoeSharedTopKGating.__init__   s:     	&$
YYz;UC
r2   c                    | j                  |      j                         }|j                  | j                  d      \  }}t	        j
                  |d      j                  |      }t	        j                  |j                  d      | j                  g|j                  |j                        }|j                  d|d      }|j                         j                  d      }|j                         }|j!                         }	|	j#                  d      \  }
}|j%                  | j                  d      }|j!                         }||   }|||||fS )Nr   rM   r   r`   devicetrunc)rounding_mode)r   floattopkr   r,   softmaxtype_aszerossizern   r`   r   scatterlongsumtolistflattensortdiv)rG   rI   logitstop_k_logitstop_k_indicestop_k_gatesr   gatesrz   top_k_experts_index_sorted_expertsbatch_indexbatch_gatess                 r3   rQ   z"GraniteMoeSharedTopKGating.forward   s.   M*002&,kk$**!k&D#mmmLa8@@O a $"2"23;;L;LU`UgUg
 a2jjl&&q) "((* &--/"/"4"4Q"7*..tzz.Q "))+!"67#[+{FRRr2   r   rT   s   @r3   r   r      s'    D3 DS D D(Sr2   r   c                   .     e Zd ZdZdef fdZd Z xZS )GraniteMoeSharedMoEz
    A Sparsely gated mixture of experts layer with 1-layer Feed-Forward networks as experts.

    Args:
        config:
            Configuration object with model hyperparameters.
    r7   c                    t         |           |j                  | _        |j                  | _        t
        |j                     | _        t        |j                  | j                  | j                  dz        | _
        t        |j                  | j                  | j                        | _        t        | j                  |j                  |j                        | _        y )Nr9   )r?   rn   r   )r<   r=   r>   r?   intermediate_sizer
   rA   rB   rm   num_local_expertsrD   rE   r   num_experts_per_tokrouterrF   s     r3   r=   zGraniteMoeSharedMoE.__init__   s     ,,!33 !2!23;$$doot7G7G!7K
 =$$d&6&6
 100,,
r2   c                    |j                         \  }}}|j                  d|      }| j                  |      \  }}}}}||   }	| j                  |	|      }
|
j	                  dd      }| j                  |d         |d   z  }
| j                  |
|      }||d d d f   z  }t        j                  ||z  | j                  f|j                  |j                        }|j                  d||      }|j                  ||| j                        }|S )NrL   r9   rM   r   r   r   )r   reshaper   rD   rO   rB   rE   r,   r   r?   r`   r   	index_addview)rG   layer_inputbszlengthemb_sizer   r   r   rz   expert_inputsrI   rP   expert_outputsr   layer_outputs                  r3   rQ   zGraniteMoeSharedMoE.forward   s    + 0 0 2VX!))"h76:kk+6N3;[!#K0))-E - 3 3A2 3 >(=a(@ADYZ[D\\++M;G'+ag*>>S6\4??;>CWCW`n`u`uvq+~F#((fdooFr2   )r(   r)   r*   r+   r    r=   rQ   rS   rT   s   @r3   r   r      s    
5 
&r2   r   c                     | dd| j                   d   dz  f   }| d| j                   d   dz  df   }t        j                  | |fd      S )z*Rotates half the hidden dims of the input..NrL   r9   rM   )rj   r,   rx   )xx1x2s      r3   rotate_halfr      sZ    	
3"!''"+"""	#B	
3q ""	#B99rc2YB''r2   rotary_pos_embc                     |j                  |      }|j                  |      }| |z  t        |       |z  z   }||z  t        |      |z  z   }||fS )a  Applies Rotary Position Embedding to the query and key tensors.

    Args:
        q (`torch.Tensor`): The query tensor.
        k (`torch.Tensor`): The key tensor.
        cos (`torch.Tensor`): The cosine part of the rotary embedding.
        sin (`torch.Tensor`): The sine part of the rotary embedding.
        unsqueeze_dim (`int`, *optional*, defaults to 1):
            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
    Returns:
        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
    )	unsqueezer   )qkcossinunsqueeze_dimq_embedk_embeds          r3   apply_rotary_pos_embr     sY    & --
&C
--
&C3w;q>C/0G3w;q>C/0GGr2   rI   n_reprJ   c                     | j                   \  }}}}|dk(  r| S | dddddddddf   j                  |||||      } | j                  |||z  ||      S )z
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    r   N)rj   expandr   )rI   r   batchnum_key_value_headsslenhead_dims         r3   	repeat_kvr   !  so    
 2?1D1D.Ehz!!Qa"23::5BUW\^bdlmM  (;e(CT8TTr2   modulequerykeyvalueattention_maskscalingdropoutkwargsc                 T   t        || j                        }t        || j                        }	t        j                  ||j	                  dd            |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 )Nr9   r   rL   )rN   r`   )ptrainingr   )r   num_key_value_groupsr,   matmul	transposerj   r   r   r   rb   ra   r`   r   r   
contiguous)r   r   r   r   r   r   r   r   
key_statesvalue_statesattn_weightscausal_maskattn_outputs                r3   eager_attention_forwardr   -  s    3 ; ;<JUF$?$?@L<<z';';Aq'ABWLL!$Q1.D
0@0@0D.D%DE#k1==((2U]](SVVW\WbWbcL==((6??([L,,|\:K''1-88:K$$r2   c                       e Zd ZdZdedef fdZ	 	 	 	 ddej                  de	ej                  ej                  f   dz  dej                  dz  d	e
dz  d
ej                  dz  dee   de	ej                  ej                  f   fdZ xZS )GraniteMoeSharedAttentionz=Multi-headed attention from 'Attention Is All You Need' paperr7   	layer_idxc                 ^   t         |           || _        || _        t	        |d|j
                  |j                  z        | _        |j                  |j                  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                         | _        y )Nr   Tr:   )r<   r=   r7   r   getattrr>   num_attention_headsr   r   r   attention_multiplierr   attention_dropout	is_causalr   rC   attention_biasq_projk_projv_projo_projrG   r7   r   rH   s      r3   r=   z"GraniteMoeSharedAttention.__init__K  sJ   "
F4F4F&JdJd4de$*$>$>&B\B\$\!22!'!9!9ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii&&68J8JQWQfQf
r2   NrI   position_embeddingsr   past_key_valuescache_positionr   rJ   c                    |j                   d d }g |d| j                  }| j                  |      j                  |      j	                  dd      }	| j                  |      j                  |      j	                  dd      }
| j                  |      j                  |      j	                  dd      }|\  }}t        |	|
||      \  }	}
|'|||d}|j                  |
|| j                  |      \  }
}t        j                  | j                  j                  t              } || |	|
||f| j                  sdn| j                   | j"                  d|\  }} |j$                  g |d j'                         }| j)                  |      }||fS )NrL   r   r9   )r   r   r           )r   r   )rj   r   r   r   r   r   r   r   updater   r   get_interfacer7   _attn_implementationr   r   r   r   r   r   r   )rG   rI   r   r   r   r   r   input_shapehidden_shapequery_statesr   r   r   r   cache_kwargsattention_interfacer   r   s                     r3   rQ   z!GraniteMoeSharedAttention.forwardb  s    $))#2.88b8$--8{{=166|DNNqRST[[/44\BLLQPQR
{{=166|DNNqRST&S#7jRUWZ#[ j&#&snUL'6'='=j,X\XfXfht'u$J(?(M(MKK,,.E)
 %8	%
  $}}C$2H2HLL	%
 	%
!\ *k));;;;FFHkk+.L((r2   )NNNN)r(   r)   r*   r+   r    r/   r=   r,   rR   ri   r   r-   r   r   rQ   rS   rT   s   @r3   r   r   G  s    G
5 
# 
4 IM.2(,26))||)) #5<<#=>E)) t+	))
 )) ((4/)) +,)) 
u||U\\)	*))r2   r   c                   p    e Zd Zdedef fdZ	 	 	 	 	 	 	 ddej                  dej                  dz  dej                  dz  de	dz  d	e
dz  d
e
dz  dej                  dz  deej                  ej                  f   dz  dee   deej                  eej                  ej                  f   dz  f   fdZ xZS )GraniteMoeSharedDecoderLayerr7   r   c                    t         |           |j                  | _        t        ||      | _        t        |j                  |j                        | _        t        |j                  |j                        | _        t        |      | _
        |j                  | _        |j                  dk(  rd | _        y t        |      | _        y )N)r7   r   r]   r   )r<   r=   r>   r   	self_attnrW   rms_norm_epsinput_layernormpost_attention_layernormr   block_sparse_moeresidual_multiplierr@   r6   
shared_mlpr   s      r3   r=   z%GraniteMoeSharedDecoderLayer.__init__  s    !--2&IV6v7I7IvObObc(?@R@RX^XkXk(l% 3F ;#)#=#= "("A"AQ"F$L_`fLgr2   NrI   r   position_idsr   output_attentions	use_cacher   r   r   rJ   c	                 >   |}
| j                  |      } | j                  d||||||||d|	\  }}|
|| j                  z  z   }|}
| j                  |      }| j	                  |      }| j
                  |}n|| j                  |      z   }|
|| j                  z  z   }|S )N)rI   r   r  r   r	  r
  r   r   r1   )r  r  r  r  r  r  )rG   rI   r   r  r   r	  r
  r   r   r   residualr   moe_hidden_statess                r3   rQ   z$GraniteMoeSharedDecoderLayer.forward  s     !,,]; *4>> 

')%+/) 3

 

q !=43K3K#KK 55mD 11-@??"-M-0NNM =43K3K#KKr2   )NNNFFNN)r(   r)   r*   r    r/   r=   r,   rR   r-   r   boolri   r   r"   FloatTensorrQ   rS   rT   s   @r3   r   r     s   h5 h# h /304(,).!&26HL'||' t+' &&-	'
 '  $;' $;' ((4/' #5<<#=>E' 45' 
u  %(9(95;L;L(L"MPT"TT	U'r2   r   c                        e Zd ZU eed<   dZdZdgZdgZdZ	dZ
dZdZdZeedZ ej$                          fd       Z xZS )	GraniteMoeSharedPreTrainedModelr7   modelTr   r   F)rI   
attentionsc                     t         |   |       t        |t              r7t	        j
                  |j                  d| j                  j                         y y )Nr   )rd   std)	r<   _init_weights
isinstancerm   initnormal_r[   r7   initializer_range)rG   r   rH   s     r3   r  z-GraniteMoeSharedPreTrainedModel._init_weights  s>    f%f=>LLSdkk6S6ST ?r2   )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,   no_gradr  rS   rT   s   @r3   r  r    sp    ""&*#78#4"5N""&5/
 U]]_U Ur2   r  c                        e Zd ZU ej                  ed<   ddef fdZe	 	 	 ddedz  de	d   de
dz  ded	ef   fd
       Z ej                         ed               Z xZS )GraniteMoeSharedRotaryEmbeddinginv_freqNr7   c                    t         |           |j                  | _        |j                  | _        || _        | j
                  j                  d   | _        | j                  }| j                  dk7  rt        | j                     } || j
                  |      \  }| _
        | j                  d|d       | j                  d|j                         d       y )N	rope_typedefaultr(  F)
persistentoriginal_inv_freq)r<   r=   max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenr7   rope_parametersr*  compute_default_rope_parametersr   attention_scalingregister_bufferclone)rG   r7   r   rope_init_fnr(  rH   s        r3   r=   z(GraniteMoeSharedRotaryEmbedding.__init__  s    "("@"@$*$B$B!44[A!%!E!E>>Y&.t~~>L+7V+L($(ZeD0(..2BuUr2   r   ztorch.deviceseq_lenrJ   ztorch.Tensorc                    | j                   d   }t        | dd      xs | j                  | j                  z  }d}d|t	        j
                  d|dt        j                        j                  |t        j                        |z  z  z  }||fS )	a  
        Computes the inverse frequencies according to the original RoPE implementation
        Args:
            config ([`~transformers.PreTrainedConfig`]):
                The model configuration.
            device (`torch.device`):
                The device to use for initialization of the inverse frequencies.
            seq_len (`int`, *optional*):
                The current sequence length. Unused for this type of RoPE.
        Returns:
            Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the
            post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE).
        
rope_thetar   Ng      ?r   r9   r`   )r   r`   )	r1  r   r>   r   r,   arangeint64ra   r   )r7   r   r7  baserN   attention_factorr(  s          r3   r2  z?GraniteMoeSharedRotaryEmbedding.compute_default_rope_parameters  s    & %%l3fj$/c63E3EIcIc3c U\\!S!5;;?BB&X]XcXcBdgjjk
 )))r2   c                 N   | j                   d d d d f   j                         j                  |j                  d   dd      j	                  |j
                        }|d d d d d f   j                         }t        |j
                  j                  t              r/|j
                  j                  dk7  r|j
                  j                  nd}t        |d      5  |j                         |j                         z  j                  dd      }t        j                  ||fd	      }|j                         | j                  z  }|j                         | j                  z  }	d d d        j	                  |j                   
      	j	                  |j                   
      fS # 1 sw Y   AxY w)Nr   rL   r   mpscpuF)device_typeenabledr9   rM   r:  )r(  r   r   rj   ra   r   r  typestrr   r   r,   rx   r   r3  r   r`   )
rG   r   r  inv_freq_expandedposition_ids_expandedrB  freqsembr   r   s
             r3   rQ   z'GraniteMoeSharedRotaryEmbedding.forward  sR    !MM$4-8>>@GGHZHZ[\H]_acdehhijiqiqr ,QaZ 8 > > @'1!((--'E!((--[`J`ahhmmfkUC 	5&,,.1F1L1L1NNYYZ[]^_E))UEN3C'')d444C'')d444C		5 vvAGGv$cff177f&;;;	5 	5s   BFF$)N)NNN)r(   r)   r*   r,   rR   r.   r    r=   staticmethodr   r/   ri   r   r2  r%  r   rQ   rS   rT   s   @r3   r'  r'    s    llV5 V  04+/"*&-*(* t* 
~u$	%	* *: U]]_<  <r2   r'  c                       e Zd Zdef fdZee	 	 	 	 	 	 	 ddej                  dz  dej                  dz  dej                  dz  de
dz  dej                  dz  d	edz  d
ej                  dz  dee   defd              Z xZS )GraniteMoeSharedModelr7   c           	      .   t         |   |       |j                  | _        |j                  | _        t        j                  |j                  |j                  | j                        | _        t        j                  t        |j                        D cg c]  }t        ||       c}      | _        t        |j                  |j                        | _        t#        |      | _        d| _        |j(                  | _        | j+                          y c c}w )Nr   r7   F)r<   r=   pad_token_idpadding_idx
vocab_sizer   	Embeddingr>   embed_tokens
ModuleListrt   num_hidden_layersr   layersrW   r  normr'  
rotary_embgradient_checkpointingembedding_multiplier	post_initr   s      r3   r=   zGraniteMoeSharedModel.__init__  s     !.. ++LL):):F<N<NPTP`P`ammNSTZTlTlNmn)&)<n
 ,F,>,>FDWDWX	9H&+#$*$?$?! 	 os   DN	input_idsr   r  r   inputs_embedsr
  r   r   rJ   c                 b   |d u |d uz  rt        d      |r|t        | j                        }|| j                  |      }|F||j	                         nd}	t        j                  |	|	|j                  d   z   |j                        }||j                  d      }t        | j                  |||||      }
|| j                  z  }|}| j                  ||      }| j                  d | j                  j                   D ]  } ||f||
||||d|} | j                  |      }t!        ||      S )	Nz:You must specify exactly one of input_ids or inputs_embedsrN  r   r   )r   )r7   input_embedsr   r   r   r  )r   r   r  r   r
  r   )last_hidden_stater   )
ValueErrorr   r7   rS  get_seq_lengthr,   r;  rj   r   r   r   rZ  rX  rV  rU  rW  r   )rG   r\  r   r  r   r]  r
  r   r   past_seen_tokensr   rI   r   decoder_layers                 r3   rQ   zGraniteMoeSharedModel.forward/  sp    -t";<YZZ0*$++>O  --i8M!CRC^==?de"\\ "2]5H5H5K"KTaThThN )33A6L(;;&))+%
 &(A(AA% #oom\J![[)H4;;+H+HI 
	M)	$7*) /#-	 	M
	 		-0%++
 	
r2   )NNNNNNN)r(   r)   r*   r    r=   r   r   r,   r-   rR   r   r  r  r   r   r   rQ   rS   rT   s   @r3   rL  rL    s    5 "  .2.204(,26!%26;
##d*;
 t+;
 &&-	;

 ;
 ((4/;
 $;;
 ((4/;
 +,;
 
 ;
  ;
r2   rL  gate_logitsrn   c                    | t        | t              syt        | t              rC| d   j                  }t        j                  | D cg c]  }|j                  |       c}d      }t        j                  j                  j                  d      }t        j                  ||d      \  }}	t        j                  j                  j                  |	|      }
|>t        j                  |
j                         d      }t        j                  |d      }n|j                  \  }}|j                  d   ||z  z  }|dddddddf   j                  |||||f      j                  d||      j                        }t        j                   |
j                         |z  d      t        j                   |d      z  }|ddddddf   j                  ||||f      j                  d|      j                  |      }t        j                   ||z  d      t        j                   |d      z  }t        j                   ||j#                  d      z        }||z  S c c}w )a  
    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   rM   rL   )r  ri   r   r,   rx   ra   r   r   r   r   one_hotrd   r   rj   r   r   r   r   )re  rn   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                      r3   load_balancing_loss_funcru  o  s9   : *[%"@+u%$Q..#(99^i-jPZjmmN.K-jpq#r hh))112JPR1SO**_eDA((%%--.>LK!JJ{'8'8':B "'O!C&4&:&:#
O4::1=*B^_ 4AtT12V&
OUKXYWR,R	 	 "IIk&7&7&9<Q&QWXY\a\e\e!q]
 
 4At+,V&
O[QRWR%R	 	) "'?=]+]cd!ehmhqhq,!i
 "
 99.1G1Q1QRS1TTUL+%%[ .ks   Ic                   d    e Zd ZddiZddiZddgdgfiZdef fdZee		 	 	 	 	 	 	 	 	 dd
e
j                  d	z  de
j                  d	z  de
j                  d	z  ded	z  de
j                  d	z  de
j                  d	z  ded	z  de
j                  d	z  dee
j                  z  deez  fd              Z xZS )GraniteMoeSharedForCausalLMzlm_head.weightzmodel.embed_tokens.weightlm_headcolwise_gather_outputrI   r   r7   c                 p   t         |   |       t        |      | _        |j                  | _        t        j                  |j                  |j                  d      | _        |j                  | _	        |j                  | _        |j                  | _        |j                  | _        | j                          y )NFr:   )r<   r=   rL  r  rQ  r   rC   r>   rx  router_aux_loss_coefr   rn   r   logits_scalingr[  rF   s     r3   r=   z$GraniteMoeSharedForCausalLM.__init__  s     *62
 ++yy!3!3V5F5FUS$*$?$?!!33#)#=#= $33 	r2   Nr\  r   r  r   r]  labelsoutput_router_logitsr   logits_to_keeprJ   c
           
         ||n| j                   j                  } | j                  d||||||d|
}|j                  }t	        |	t
              rt        |	 d      n|	}| j                  |dd|ddf         }|| j                   j                  z  }d}|* | j                  ||fd| j                   j                  i|
}d}|rYt        |j                  | j                  | j                  |      }|+|| j                  |j!                  |j"                        z  z  }t%        ||||j&                  |j(                  |j*                  |j                        S )ax  
        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)r\  r   r  r   r]  r   rQ  )lossaux_lossr   r   rI   r  router_logitsr1   )r7   r~  r  r`  r  r/   slicerx  r|  loss_functionrQ  ru  r  rn   r   r{  ra   r   r   r   rI   r  )rG   r\  r   r  r   r]  r}  r~  r   r  r   outputsrI   slice_indicesr   r  r  s                    r3   rQ   z#GraniteMoeSharedForCausalLM.forward  s   L %9$D $++JjJj 	 $** 
)%+')
 
  118B>SV8W~ot4]kmA}a,?@A$++444%4%%  ;;11 	D /%%  ((	H !11HKK4LLL(#33!//))!//
 	
r2   )	NNNNNNNNr   )r(   r)   r*   _tied_weights_keys_tp_plan_pp_planr    r=   r   r   r,   r-   rR   r   r  r  r/   ri   r   rQ   rS   rT   s   @r3   rw  rw    s9   *,GH23H_-z:;H5   .2.204(,26*.,026-.S
##d*S
 t+S
 &&-	S

 S
 ((4/S
   4'S
 #TkS
 ((4/S
 ell*S
 
*	*S
  S
r2   rw  )rw  rL  r  )r   )r   )Nr9   N)Gcollections.abcr   typingr   r   r,   r   torch.nnr   rv    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   configuration_granitemoesharedr    r"   Moduler6   rW   rm   r   r   r   r   rR   r/   r   r   r   r   r   r  r'  rL  ri   ru  rw  __all__r1   r2   r3   <module>r     s  * % &   $ & ! . ) f f / 9 Q K F & 7 Q Q B)5 0")) 4 Y'Jbii J (J(*bii *Z.S .Sb,")) ,^( *+ ,2	UU\\ 	U# 	U%,, 	U& %II%<<% 
% <<	%
 LL4'% % % '(%4 )*C)		 C) +C)L2#= 2j Uo U U.><bii ><B O
; O
 O
h #
*.	O&ell 33d:O&tO& LL4'	O&
 \\CO&d g
"A? g
 g
T fr2   