
    qiX                     X   d dl mZ d dlmZ d dlZd dlmZ ddlmZ ddlm	Z	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 ddlmZmZ ddlmZmZ ddlmZmZ ddl m!Z! ddl"m#Z#m$Z$m%Z% ddl&m'Z'm(Z( ddl)m*Z* ddl+m,Z,  ed       G d dejZ                               Z. G d dejZ                        Z/d Z0 ed      d8d       Z1dejd                  de3d ejd                  fd!Z4	 d9d"ejZ                  d#ejd                  d$ejd                  d%ejd                  d&ejd                  dz  d'e5d(e5d)e!e#   fd*Z6 ee1       G d+ d,ejZ                               Z7 G d- d.e      Z8 G d/ d0ejZ                        Z9e$ G d1 d2e             Z:e$ G d3 d4e:             Z;e$ G d5 d6e:e             Z<g d7Z=y):    )Callable)OptionalN)nn   )ACT2FN)CacheDynamicCache)GenerationMixin)use_kernel_forward_from_hubuse_kernel_func_from_hubuse_kernelized_func)create_causal_mask)FlashAttentionKwargs)GradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tuple)maybe_autocastmerge_with_config_defaults)capture_outputs   )BitNetConfigRMSNormc                   h     e Zd Zddeddf fdZdej                  dej                  fdZd Z xZ	S )	BitNetRMSNormepsreturnNc                     t         |           t        j                  t	        j
                  |            | _        || _        y)z<
        BitNetRMSNorm is equivalent to T5LayerNorm
        N)super__init__r   	Parametertorchonesweightvariance_epsilon)selfhidden_sizer#   	__class__s      \/opt/pipecat/venv/lib/python3.12/site-packages/transformers/models/bitnet/modeling_bitnet.pyr'   zBitNetRMSNorm.__init__-   s1     	ll5::k#:; #    hidden_statesc                 "   |j                   }|j                  t        j                        }|j	                  d      j                  dd      }|t        j                  || j                  z         z  }| j                  |j                  |      z  S )N   T)keepdim)	dtypetor)   float32powmeanrsqrtr,   r+   )r-   r2   input_dtypevariances       r0   forwardzBitNetRMSNorm.forward5   sy    #))%((7 $$Q',,R,>%Ht?T?T4T(UU{{]--k:::r1   c                 ^    t        | j                  j                         d| j                   S )Nz, eps=)tupler+   shaper,   )r-   s    r0   
extra_reprzBitNetRMSNorm.extra_repr<   s*    ))*+6$2G2G1HIIr1   )gư>)
__name__
__module____qualname__floatr'   r)   Tensorr?   rC   __classcell__r/   s   @r0   r"   r"   +   s7    $ $$ $;U\\ ;ell ;Jr1   r"   c                   *     e Zd Zdef fdZd Z xZS )	BitNetMLPconfigc                    t         |           || _        |j                  | _        |j                  | _        t        j                  | j                  | j                  d      | _        t        j                  | j                  | j                  d      | _        t        j                  | j                  | j                  d      | _	        t        |j                     | _        t        |j                  |j                        | _        y )NFbiasr#   )r&   r'   rM   r.   intermediate_sizer   Linear	gate_projup_proj	down_projr   
hidden_actact_fnr"   rms_norm_epsffn_sub_normr-   rM   r/   s     r0   r'   zBitNetMLP.__init__A   s    !--!'!9!94#3#3T5K5KRWXyy!1!143I3IPUV4#9#94;K;KRWXV../)&*B*BH[H[\r1   c           	          | j                  | j                  | j                  | j                  |            | j	                  |      z              }|S N)rV   rZ   rX   rT   rU   )r-   xrV   s      r0   r?   zBitNetMLP.forwardL   sF    NN4#4#4T[[PQAR5SVZVbVbcdVe5e#fg	r1   )rD   rE   rF   r   r'   r?   rI   rJ   s   @r0   rL   rL   @   s    	]| 	]r1   rL   c                     | dd| j                   d   dz  f   }| d| j                   d   dz  df   }t        j                  | |fd      S )z*Rotates half the hidden dims of the input..Nr5   r4   dim)rB   r)   cat)r^   x1x2s      r0   rotate_halfre   Q   sZ    	
3"!''"+"""	#B	
3q ""	#B99rc2YB''r1   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.
    )	unsqueezere   )qkcossinunsqueeze_dimq_embedk_embeds          r0   apply_rotary_pos_embrp   X   sY    & --
&C
--
&C3w;q>C/0G3w;q>C/0GGr1   r2   n_repr$   c                     | j                   \  }}}}|dk(  r| S | dddddddddf   j                  |||||      } | j                  |||z  ||      S )z
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    r   N)rB   expandreshape)r2   rq   batchnum_key_value_headsslenhead_dims         r0   	repeat_kvry   r   so    
 2?1D1D.Ehz!!Qa"23::5BUW\^bdlmM  (;e(CT8TTr1   modulequerykeyvalueattention_maskscalingdropoutkwargsc                    t        || j                        }t        || j                        }	t        j                  ||j	                  dd            |z  }
||
|z   }
t
        j                  j                  |
dt        j                        j                  |j                        }
t
        j                  j                  |
|| j                        }
t        j                  |
|	      }|j	                  dd      j                         }||
fS )Nr4   r   r5   )ra   r7   )ptrainingr   )ry   num_key_value_groupsr)   matmul	transposer   
functionalsoftmaxr9   r8   r7   r   r   
contiguous)rz   r{   r|   r}   r~   r   r   r   
key_statesvalue_statesattn_weightsattn_outputs               r0   eager_attention_forwardr   ~   s     3 ; ;<JUF$?$?@L<<z';';Aq'ABWLL!#n4==((2U]](SVVW\WbWbcL==((6??([L,,|\:K''1-88:K$$r1   c                       e Zd ZdZdedef fdZ	 	 ddej                  de	ej                  ej                  f   dej                  dz  d	e
dz  d
ej                  dz  dee   de	ej                  ej                  dz  f   fdZ xZS )BitNetAttentionz=Multi-headed attention from 'Attention Is All You Need' paperrM   	layer_idxc                    t         |           || _        || _        t	        |d|j
                  |j                  z        | _        |j                  |j                  z  | _	        | j                  dz  | _
        |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                        | _        t)        |j
                  |j*                        | _        y )Nrx   g      TrO   rQ   )r&   r'   rM   r   getattrr.   num_attention_headsrx   rv   r   r   attention_dropout	is_causalr   rS   attention_biasq_projk_projv_projo_projr"   rY   attn_sub_normr-   rM   r   r/   s      r0   r'   zBitNetAttention.__init__   sj   "
F4F4F&JdJd4de$*$>$>&B\B\$\!}}d*!'!9!9ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii&&68J8JQWQfQf
 +6+=+=6CVCVWr1   Nr2   position_embeddingsr~   past_key_valuescache_positionr   r$   c                 :   |j                   d d }g |d| j                  }| j                  |      j                  |      j	                  dd      }	| j                  |      j                  |      j	                  dd      }
| j                  |      j                  |      j	                  dd      }|\  }}t        |	|
||      \  }	}
|'|||d}|j                  |
|| j                  |      \  }
}t        j                  | j                  j                  t              } || |	|
||f| j                  sdn| j                   | j"                  d|\  }} |j$                  g |d j'                         }| j)                  |      }| j+                  |      }||fS )Nr5   r   r4   )rl   rk   r           )r   r   )rB   rx   r   viewr   r   r   rp   updater   r   get_interfacerM   _attn_implementationr   r   r   r   rt   r   r   r   )r-   r2   r   r~   r   r   r   input_shapehidden_shapequery_statesr   r   rk   rl   cache_kwargsattention_interfacer   r   s                     r0   r?   zBitNetAttention.forward   s    $))#2.88b8$--8{{=166|DNNqRST[[/44\BLLQPQR
{{=166|DNNqRST&S#7jRUWZ#[ j&#&snUL'6'='=j,X\XfXfht'u$J(?(M(MKK,,.E)
 %8	%
  $}}C$2H2HLL	%
 	%
!\ *k));;;;FFH((5kk+.L((r1   )NN)rD   rE   rF   __doc__r   intr'   r)   rH   rA   r   
LongTensorr   r   r?   rI   rJ   s   @r0   r   r      s    GX| X X: )-26*)||*) #5<<#=>*) t+	*)
 *) ((4/*) -.*) 
u||U\\D00	1*)r1   r   c                   "    e Zd Zdedef fdZ	 	 	 	 	 	 ddej                  dej                  dz  dej                  dz  de	dz  d	e
dz  d
ej                  dz  deej                  ej                  f   dz  dee   dej                  fdZ xZS )BitNetDecoderLayerrM   r   c                     t         |           |j                  | _        t        ||      | _        t        |      | _        t        |j                  |j                        | _	        t        |j                  |j                        | _
        y )N)rM   r   rQ   )r&   r'   r.   r   	self_attnrL   mlpr"   rY   input_layernormpost_attention_layernormr   s      r0   r'   zBitNetDecoderLayer.__init__   sl    !--()LV$,V-?-?VEXEXY(5f6H6HfNaNa(b%r1   Nr2   r~   position_idsr   	use_cacher   r   r   r$   c                     |}	| j                  |      } | j                  d|||||||d|\  }}
|	|z   }|}	| j                  |      }| j                  |      }|	|z   }|S )N)r2   r~   r   r   r   r   r    )r   r   r   r   )r-   r2   r~   r   r   r   r   r   r   residual_s              r0   r?   zBitNetDecoderLayer.forward   s     !,,];)4>> 	
')%+) 3	
 	
q !=0 !55mD/ =0r1   )NNNFNN)rD   rE   rF   r   r   r'   r)   rH   r   r   boolrA   r   r   r?   rI   rJ   s   @r0   r   r      s    c| c c /304(,!&26HL|| t+ &&-	
  $; ((4/ #5<<#=>E +, 
r1   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 )BitNetRotaryEmbeddinginv_freqNrM   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_lenrM   rope_parametersr   compute_default_rope_parametersr   attention_scalingregister_bufferclone)r-   rM   devicerope_init_fnr   r/   s        r0   r'   zBitNetRotaryEmbedding.__init__  s    "("@"@$*$B$B!44[A!%!E!E>>Y&.t~~>L+7V+L($(ZeD0(..2BuUr1   r   ztorch.deviceseq_lenr$   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_thetarx   Ng      ?r   r4   r7   )r   r7   )	r   r   r.   r   r)   arangeint64r8   rG   )rM   r   r   basera   attention_factorr   s          r0   r   z5BitNetRotaryEmbedding.compute_default_rope_parameters   s    & %%l3fj$/c63E3EIcIc3c U\\!S!5;;?BB&X]XcXcBdgjjk
 )))r1   c                 N   | j                   d d d d f   j                         j                  |j                  d   dd      j	                  |j
                        }|d d d d d f   j                         }t        |j
                  j                  t              r/|j
                  j                  dk7  r|j
                  j                  nd}t        |d      5  |j                         |j                         z  j                  dd      }t        j                  ||fd	      }|j                         | j                  z  }|j                         | j                  z  }	d d d        j	                  |j                   
      	j	                  |j                   
      fS # 1 sw Y   AxY w)Nr   r5   r   mpscpuF)device_typeenabledr4   r`   r   )r   rG   rs   rB   r8   r   
isinstancetypestrr   r   r)   rb   rk   r   rl   r7   )
r-   r^   r   inv_freq_expandedposition_ids_expandedr   freqsembrk   rl   s
             r0   r?   zBitNetRotaryEmbedding.forward>  sR    !MM$4-8>>@GGHZHZ[\H]_acdehhijiqiqr ,QaZ 8 > > @'1!((--'E!((--[`J`ahhmmfkUC 	5&,,.1F1L1L1NNYYZ[]^_E))UEN3C'')d444C'')d444C		5 vvAGGv$cff177f&;;;	5 	5s   BFF$r]   )NNN)rD   rE   rF   r)   rH   __annotations__r   r'   staticmethodr   r   rA   rG   r   no_gradr   r?   rI   rJ   s   @r0   r   r     s    llV| V  &*+/"*t#*(* t* 
~u$	%	* *: U]]_<  <r1   r   c                   J    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y)BitNetPreTrainedModelrM   modelTr   r   )r2   
attentionsN)rD   rE   rF   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   r1   r0   r   r   N  sQ    &*#-.#4"5N!"&+%r1   r   c                       e Zd Zdef fdZeee	 	 	 	 	 	 	 ddej                  dz  dej                  dz  dej                  dz  dedz  dej                  dz  d	ej                  dz  d
edz  dee   defd                     Z xZS )BitNetModelrM   c           	         t         |   |       |j                  | _        |j                  | _        t        j                  |j                  |j                  | j                        | _        t        j                  t        |j                        D cg c]  }t        ||       c}      | _        t        |j                  |j                        | _        t#        |      | _        d| _        | j)                          y c c}w )NrQ   rM   F)r&   r'   pad_token_idpadding_idx
vocab_sizer   	Embeddingr.   embed_tokens
ModuleListrangenum_hidden_layersr   layersr"   rY   normr   
rotary_embgradient_checkpointing	post_initr   s      r0   r'   zBitNetModel.__init__c  s     !.. ++LL):):F<N<NPTP`P`ammDI&JbJbDcdy	2d
 "&"4"4&:M:MN	/v>&+# 	 es   DN	input_idsr~   r   r   inputs_embedsr   r   r   r$   c                 D   |d u |d uz  rt        d      || j                  |      }|r|t        | j                        }|E||j	                         nd}	t        j                  |j                  d   |j                        |	z   }||j                  d      }t        | j                  |||||      }
|}| j                  ||      }| j                  d | j                  j                   D ]  } ||f|
|||||d|} | j                  |      }t        ||	      S )
Nz:You must specify exactly one of input_ids or inputs_embedsr   r   r   )r   )rM   r  r~   r   r   r   )r   )r~   r   r   r   r   r   )last_hidden_stater   )
ValueErrorr   r	   rM   get_seq_lengthr)   r   rB   r   rh   r   r  r  r   r  r   )r-   r  r~   r   r   r  r   r   r   past_seen_tokenscausal_maskr2   r   decoder_layers                 r0   r?   zBitNetModel.forwards  s]    -t";<YZZ *.*;*;I*FM0*$++>O!CRC^==?de]003M<P<PQTdd  )33A6L(;;'))+%
 &"oom,oW![[)H4;;+H+HI 
	M)	*$7) /#-	 	M
	 		-0&++
 	
r1   )NNNNNNN)rD   rE   rF   r   r'   r   r   r   r)   r   rH   r   FloatTensorr   r   r   r   r?   rI   rJ   s   @r0   r   r   a  s    |     .2.204(,2626!%9
##d*9
 t+9
 &&-	9

 9
 ((4/9
 ((4/9
 $;9
 +,9
 
!9
    9
r1   r   c                   R    e Zd ZddiZdZdZ fdZee	 	 	 	 	 	 	 	 	 dde	j                  dz  de	j                  dz  de	j                  dz  dedz  d	e	j                  dz  d
e	j                  dz  dedz  de	j                  dz  dee	j                  z  dee   defd              Z xZS )BitNetForCausalLMzlm_head.weightzmodel.embed_tokens.weightNc                     t         |   |       t        |      | _        |j                  | _        t        j                  |j                  |j                  d      | _        | j                          y )NFrO   )
r&   r'   r   r   r   r   rS   r.   lm_headr  r[   s     r0   r'   zBitNetForCausalLM.__init__  sU      (
 ++yy!3!3V5F5FUS 	r1   r  r~   r   r   r  labelsr   r   logits_to_keepr   r$   c
                 z    | j                   d|||||||d|
}|j                  }t        |	t              rt	        |	 d      n|	}| j                  |dd|ddf         }d}|* | j                  d||| j                  j                  d|
}t        |||j                  |j                  |j                        S )a$  
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the masked language modeling loss. Indices should either be in `[0, transformers.,
            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, transformers., config.vocab_size]`.

        Example:

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

        >>> model = BitNetForCausalLM.from_pretrained("microsoft/bitnet-b1.58-2B-4T")
        >>> tokenizer = AutoTokenizer.from_pretrained("microsoft/bitnet-b1.58-2B-4T")

        >>> prompt = f'<|begin_of_text|>User: Hey, are you conscious? Can you talk to me?<|eot_id|>Assistant: '
        >>> inputs = tokenizer(prompt, return_tensors="pt")

        >>> # Generate
        >>> generate_ids = model.generate(inputs.input_ids, max_length=100)
        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        "User: Hey, are you conscious? Can you talk to me?Assistant: No, I'm not conscious. I'm an artificial intelligence designed to assist with information and tasks. How can I help you today?"
        ```)r  r~   r   r   r  r   r   N)logitsr  r   )lossr  r   r2   r   r   )r   r	  r   r   slicer  loss_functionrM   r   r   r   r2   r   )r-   r  r~   r   r   r  r  r   r   r  r   outputsr2   slice_indicesr  r  s                   r0   r?   zBitNetForCausalLM.forward  s    J ,64:: 	,
)%+')	,
 	,
  118B>SV8W~ot4]kmA}a,?@A%4%%pVFt{{OeOepiopD%#33!//))
 	
r1   )	NNNNNNNNr   )rD   rE   rF   _tied_weights_keys_tp_plan_pp_planr'   r   r   r)   r   rH   r   r  r   r   r   r   r   r?   rI   rJ   s   @r0   r  r    s   *,GHHH  .2.204(,26*.!%26-.=
##d*=
 t+=
 &&-	=

 =
 ((4/=
   4'=
 $;=
 ((4/=
 ell*=
 +,=
 
 =
  =
r1   r  )r  r   r   )r   )r   )>collections.abcr   typingr   r)   r   activationsr   cache_utilsr   r	   
generationr
   integrationsr   r   r   masking_utilsr   modeling_flash_attention_utilsr   modeling_layersr   modeling_outputsr   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   r   utils.genericr   r   utils.output_capturingr   configuration_bitnetr   Moduler"   rL   re   rp   rH   r   ry   rG   r   r   r   r   r   r   r  __all__r   r1   r0   <module>r3     s  ( %    ! . ) f f / B 9 O K F & I I G 5 . Y'JBII J (J(		 "( *+ ,2	UU\\ 	U# 	U%,, 	U& %II%<<% 
% <<	%
 LL4'% % % '(%2 )*E)bii E) +E)P*3 *Z><BII ><B O  $ M
' M
 M
` M
- M
 M
` Hr1   