
    iX                     v   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 dd	lmZ dd
lmZ ddlmZmZmZ ddlmZmZ ddlmZmZ ddlm Z m!Z! ddl"m#Z# ddl$m%Z%m&Z&m'Z' ddl(m)Z)m*Z* ddl+m,Z, ddl-m.Z.  G d dej^                        Z0 G d de      Z1dejd                  de3dejd                  fdZ4	 d:dej^                  dejd                  dejd                  d ejd                  d!ejd                  dz  d"e5d#e5d$e#e%   fd%Z6d& Z7d;d'Z8 ee8       G d( d)ej^                               Z9 G d* d+ej^                        Z: ed,       G d- d.ej^                               Z;e& G d/ d0e!             Z<e& G d1 d2e<             Z=e& G d3 d4e<e             Z> G d5 d6ee<      Z? G d7 d8ee<      Z@g d9ZAy)<    )Callable)OptionalN   )ACT2FN)CacheDynamicCache)GenerationMixin)use_kernel_forward_from_hubuse_kernelized_func)create_causal_mask)FlashAttentionKwargs) GenericForSequenceClassificationGenericForTokenClassification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   )
Glm4Configc                   V     e Zd Z fdZdej
                  dej
                  fdZ xZS )Glm4MLPc                 *   t         |           || _        t        j                  |j
                  d|j                  z  d      | _        t        j                  |j                  |j
                  d      | _        t        |j                     | _        y )N   Fbias)super__init__confignnLinearhidden_sizeintermediate_sizegate_up_proj	down_projr   
hidden_actactivation_fnselfr(   	__class__s     w/var/www/vps2.regionflexible.com/Desarrollo/venv/lib/python3.12/site-packages/transformers/models/glm4/modeling_glm4.pyr'   zGlm4MLP.__init__2   sp    IIf&8&8!f>V>V:V]bc6#;#;V=O=OV[\#F$5$56    hidden_statesreturnc                     | j                  |      }|j                  dd      \  }}|| j                  |      z  }| j                  |      S )Nr#   dim)r-   chunkr0   r.   )r2   r6   	up_statesgates       r4   forwardzGlm4MLP.forward:   sL    %%m4	#//!/4i 2 24 88	~~i((r5   )__name__
__module____qualname__r'   torchFloatTensorr?   __classcell__r3   s   @r4   r!   r!   1   s'    7)U%6%6 )5;L;L )r5   r!   c                   D    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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 )Glm4DecoderLayerr(   	layer_idxc                    t         |           |j                  | _        t        ||      | _        t        |      | _        t        |j                  |j                        | _	        t        |j                  |j                        | _
        t        |j                  |j                        | _        t        |j                  |j                        | _        y )N)r(   rI   eps)r&   r'   r+   Glm4Attention	self_attnr!   mlpGlm4RMSNormrms_norm_epsinput_layernormpost_attention_layernormpost_self_attn_layernormpost_mlp_layernormr2   r(   rI   r3   s      r4   r'   zGlm4DecoderLayer.__init__D   s    !--&f	J6?*6+=+=6CVCVW(3F4F4FFL_L_(`%(3F4F4FFL_L_(`%"-f.@.@fFYFY"Zr5   Nr6   attention_maskposition_idspast_key_values	use_cacheposition_embeddingskwargsr7   c           
         |}| j                  |      } | j                  d||||||d|\  }}	| j                  |      }||z   }|}| j                  |      }| j	                  |      }| j                  |      }||z   }|S )N)r6   rW   rX   rY   rZ   r[    )rR   rN   rT   rS   rO   rU   )
r2   r6   rW   rX   rY   rZ   r[   r\   residual_s
             r4   r?   zGlm4DecoderLayer.forwardO   s     !,,];)4>> 
')%+ 3
 
q 55mD =0 55mD///> =0r5   )NNNFN)r@   rA   rB   r   intr'   rC   Tensor
LongTensorr   booltupler   r   rD   r?   rE   rF   s   @r4   rH   rH   C   s    	[z 	[c 	[ /304(,!&HL|| t+ &&-	
  $; #5<<#=>E -. 
u  %(9(95;L;L(L"MPT"TT	Ur5   rH   r6   n_repr7   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)shapeexpandreshape)r6   rf   batchnum_key_value_headsslenhead_dims         r4   	repeat_kvro   q   so    
 2?1D1D.Ehz!!Qa"23::5BUW\^bdlmM  (;e(CT8TTr5   modulequerykeyvaluerW   scalingdropoutr\   c                    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 )Nr#   r   r9   )r;   dtype)ptrainingr   )ro   num_key_value_groupsrC   matmul	transposer)   
functionalsoftmaxfloat32torw   ru   ry   
contiguous)rp   rq   rr   rs   rW   rt   ru   r\   
key_statesvalue_statesattn_weightsattn_outputs               r4   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$$r5   c                 |    | ddddf   }| ddddf   }t        j                  | |fd      j                  d      S )	z*Rotates half the hidden dims of the input..r   Nr#   r   r9   r:   )rC   stackflatten)xx1x2s      r4   rotate_halfr      sJ    	
319B	
319B;;Ryb)11"55r5   c                    |j                  |      }|j                  |      }|dd|j                  d   dz  f   j                  dd      }|dd|j                  d   dz  f   j                  dd      }|j                  d   }| dd|f   | d|df   }}|dd|f   |d|df   }	}||z  t        |      |z  z   }
||z  t        |      |z  z   }t	        j
                  |
|gd      }
t	        j
                  ||	gd      }|
|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.
    .Nr9   r#   r:   )	unsqueezerh   repeat_interleaver   rC   cat)qkcossinunsqueeze_dim
rotary_dimq_rotq_passk_rotk_passq_embedk_embeds               r4   apply_rotary_pos_embr      sD   $ --
&C
--
&C c'SYYr]a'''
(
:
:1"
:
EC
c'SYYr]a'''
(
:
:1"
:
EC 2Jc;J;&'3
+;)<6Ec;J;&'3
+;)<6E s{{51C78Gs{{51C78G ii&)r2Gii&)r2GGr5   c                        e Zd ZdZddededz  f fdZ	 	 	 ddej                  de	ej                  ej                  f   dz  dej                  dz  d	e
dz  d
ee   de	ej                  ej                  f   fdZ xZS )rM   z=Multi-headed attention from 'Attention Is All You Need' paperNr(   rI   c                 P   t         |           || _        || _        t	        |d|j
                  |j                  z        | _        |j                  |j                  z  | _	        | j                  dz  | _
        |j                  | _        d| _        t        j                  |j
                  |j                  | j                  z  |j                        | _        t        j                  |j
                  |j                  | j                  z  |j                        | _        t        j                  |j
                  |j                  | j                  z  |j                        | _        t        j                  |j                  | j                  z  |j
                  d      | _        y )Nrn   g      Tr$   F)r&   r'   r(   rI   getattrr+   num_attention_headsrn   rl   rz   rt   attention_dropout	is_causalr)   r*   attention_biasq_projk_projv_projo_projrV   s      r4   r'   zGlm4Attention.__init__   sD   "
F4F4F&JdJd4de$*$>$>&B\B\$\!}}d*!'!9!9ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii : :T]] JQWQfQf
 ii : :T]] JFL^L^ejkr5   r6   r[   rW   rY   r\   r7   c                 
   |j                   d d }g |d| j                  }| j                  |      j                  |      j	                  dd      }| j                  |      j                  |      j	                  dd      }	| j                  |      j                  |      j	                  dd      }
|\  }}t        ||	||      \  }}	| |j                  |	|
| j                        \  }	}
t        j                  | j                  j                  t              } || ||	|
|f| j                  sdn| j                   | j"                  d|\  }} |j$                  g |d j'                         }| j)                  |      }||fS )Nr9   r   r#           )ru   rt   )rh   rn   r   viewr|   r   r   r   updaterI   r   get_interfacer(   _attn_implementationr   ry   r   rt   rj   r   r   )r2   r6   r[   rW   rY   r\   input_shapehidden_shapequery_statesr   r   r   r   attention_interfacer   r   s                   r4   r?   zGlm4Attention.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&'6'='=j,X\XfXf'g$J(?(M(MKK,,.E)
 %8	%
  $}}C$2H2HLL	%
 	%
!\ *k));;;;FFHkk+.L((r5   NNNN)r@   rA   rB   __doc__r   ra   r'   rC   rb   re   r   r   r   r?   rE   rF   s   @r4   rM   rM      s    Glz lcDj l0 IM.2(,&)||&) #5<<#=>E&) t+	&)
 &) +,&) 
u||U\\)	*&)r5   rM   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 )Glm4RotaryEmbeddinginv_freqNr(   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_lenr(   rope_parametersr   compute_default_rope_parametersr   attention_scalingregister_bufferclone)r2   r(   devicerope_init_fnr   r3   s        r4   r'   zGlm4RotaryEmbedding.__init__
  s    "("@"@$*$B$B!44[A!%!E!E>>Y&.t~~>L+7V+L($(ZeD0(..2BuUr5   r   ztorch.deviceseq_lenr7   ztorch.Tensorc                 n   | j                   d   }| j                   j                  dd      }t        | dd      xs | j                  | j                  z  }t        ||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_thetapartial_rotary_factorg      ?rn   Nr   r#   rw   )r   rw   )r   getr   r+   r   ra   rC   arangeint64r   float)	r(   r   r   baser   rn   r;   attention_factorr   s	            r4   r   z3Glm4RotaryEmbedding.compute_default_rope_parameters  s    & %%l3 & 6 6 : :;RTW X6:t4h8J8JfNhNh8h(223 U\\!S!5;;?BB&X]XcXcBdgjjk
 )))r5   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   r9   r   mpscpuF)device_typeenabledr#   r:   r   )r   r   ri   rh   r   r   
isinstancetypestrr   r|   rC   r   r   r   r   rw   )
r2   r   rX   inv_freq_expandedposition_ids_expandedr   freqsembr   r   s
             r4   r?   zGlm4RotaryEmbedding.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$r   r   )r@   rA   rB   rC   rb   __annotations__r   r'   staticmethodr   ra   re   r   r   no_gradr   r?   rE   rF   s   @r4   r   r     s    llVz V  $(+/"*T!*(* t* 
~u$	%	* *> U]]_<  <r5   r   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 )	rP   rL   r7   Nc                     t         |           t        j                  t	        j
                  |            | _        || _        y)z:
        Glm4RMSNorm is equivalent to T5LayerNorm
        N)r&   r'   r)   	ParameterrC   onesweightvariance_epsilon)r2   r+   rL   r3   s      r4   r'   zGlm4RMSNorm.__init__L  s1     	ll5::k#:; #r5   r6   c                 "   |j                   }|j                  t        j                        }|j	                  d      j                  dd      }|t        j                  || j                  z         z  }| j                  |j                  |      z  S )Nr#   r9   T)keepdim)	rw   r   rC   r   powmeanrsqrtr   r   )r2   r6   input_dtypevariances       r4   r?   zGlm4RMSNorm.forwardT  sy    #))%((7 $$Q',,R,>%Ht?T?T4T(UU{{]--k:::r5   c                 ^    t        | j                  j                         d| j                   S )Nz, eps=)re   r   rh   r   )r2   s    r4   
extra_reprzGlm4RMSNorm.extra_repr[  s*    ))*+6$2G2G1HIIr5   )gư>)
r@   rA   rB   r   r'   rC   rb   r?   r   rE   rF   s   @r4   rP   rP   J  s7    $ $$ $;U\\ ;ell ;Jr5   rP   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)Glm4PreTrainedModelr(   modelTrH   rY   )r6   
attentionsN)r@   rA   rB   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_backendrH   rM   _can_record_outputsr^   r5   r4   r   r   _  sQ    &*#+,#4"5N!"&)#r5   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dz  d
ee   defd                     Z xZS )	Glm4Modelr(   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 )NrK   r(   F)r&   r'   pad_token_idpadding_idx
vocab_sizer)   	Embeddingr+   embed_tokens
ModuleListrangenum_hidden_layersrH   layersrP   rQ   normr   
rotary_embgradient_checkpointing	post_initrV   s      r4   r'   zGlm4Model.__init__t  s     !.. ++LL):):F<N<NPTP`P`ammBGH`H`BabYfi0b
   2 28K8KL	-V<&+# 	 cs   DN	input_idsrW   rX   rY   inputs_embedsrZ   r\   r7   c           
      <   |d u |d uz  rt        d      || j                  |      }|r|t        | j                        }|V||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   )r(   r  rW   rY   rX   )rX   )rW   r[   rX   rY   rZ   )last_hidden_staterY   )
ValueErrorr
  r   r(   get_seq_lengthrC   r   rh   r   r   r   r  r  r  r  r   )r2   r  rW   rX   rY   r  rZ   r\   past_seen_tokenscausal_maskr6   r[   decoder_layers                r4   r?   zGlm4Model.forward  sL    -t";<YZZ *.*;*;I*FM0*$++>OCRC^==?de <<(;(;A(>}G[G[\_ooL'11!4L(;;')+%
 &"oom,oW![[)H4;;+H+HI 		M)*$7) /# M		 		-0&++
 	
r5   )NNNNNN)r@   rA   rB   r   r'   r   r   r   rC   rc   rb   r   rD   rd   r   r   r   r?   rE   rF   s   @r4   r  r  r  s    z     .2.204(,26!%2
##d*2
 t+2
 &&-	2

 2
 ((4/2
 $;2
 +,2
 
!2
    2
r5   r  c                   H    e Zd ZddiZddiZddgdgfiZ fdZee	 	 	 	 	 	 	 	 dd	e	j                  dz  d
e	j                  dz  de	j                  dz  dedz  de	j                  dz  de	j                  dz  dedz  dee	j                  z  dee   deez  fd              Z xZS )Glm4ForCausalLMzlm_head.weightzmodel.embed_tokens.weightlm_headcolwise_gather_outputr6   logitsc                     t         |   |       t        |      | _        |j                  | _        t        j                  |j                  |j                  d      | _        | j                          y )NFr$   )
r&   r'   r  r   r  r)   r*   r+   r  r  r1   s     r4   r'   zGlm4ForCausalLM.__init__  sU     v&
 ++yy!3!3V5F5FUS 	r5   Nr  rW   rX   rY   r  labelsrZ   logits_to_keepr\   r7   c	           
      x    | 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 )ah  
        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, Glm4ForCausalLM

        >>> model = Glm4ForCausalLM.from_pretrained("THUDM/GLM-4-9B-0414")
        >>> tokenizer = AutoTokenizer.from_pretrained("THUDM/GLM-4-9B-0414")

        >>> 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."
        ```)r  rW   rX   rY   r  rZ   N)r   r"  r  )lossr   rY   r6   r   r^   )r   r  r   ra   slicer  loss_functionr(   r  r   rY   r6   r   )r2   r  rW   rX   rY   r  r"  rZ   r#  r\   outputsr6   slice_indicesr   r%  s                  r4   r?   zGlm4ForCausalLM.forward  s    H ,64:: ,
)%+',
 ,
  118B>SV8W~ot4]kmA}a,?@A%4%%pVFt{{OeOepiopD%#33!//))
 	
r5   )NNNNNNNr   )r@   rA   rB   _tied_weights_keys_tp_plan_pp_planr'   r   r   rC   rc   rb   r   rD   rd   ra   r   r   re   r   r?   rE   rF   s   @r4   r  r    s   *,GH23H_-z:;H  .2.204(,26*.!%-.;
##d*;
 t+;
 &&-	;

 ;
 ((4/;
   4';
 $;;
 ell*;
 +,;
 
'	';
  ;
r5   r  c                       e Zd Zy)Glm4ForSequenceClassificationNr@   rA   rB   r^   r5   r4   r.  r.        r5   r.  c                       e Zd Zy)Glm4ForTokenClassificationNr/  r^   r5   r4   r2  r2    r0  r5   r2  )r   r  r  r.  r2  )r   )r   )Bcollections.abcr   typingr   rC   torch.nnr)   activationsr   cache_utilsr   r   
generationr	   integrationsr
   r   masking_utilsr   modeling_flash_attention_utilsr   modeling_layersr   r   r   modeling_outputsr   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   r   utils.genericr   r   utils.output_capturingr   configuration_glm4r   Moduler!   rH   rb   ra   ro   r   r   r   r   rM   r   rP   r   r  r  r.  r2  __all__r^   r5   r4   <module>rG     s  , %    ! . ) L / B 
 P K F & I I G 5 *)bii )$+1 +\	UU\\ 	U# 	U%,, 	U& %II%<<% 
% <<	%
 LL4'% % % '(%26%P )*>)BII >) +>)B@<")) @<F Y'J")) J (J( /  $ F
# F
 F
R K
)? K
 K
\	$DFY 		!>@S 	r5   