a
    hS                  
   @   s  d dl mZmZ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 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$m%Z% ddl&m'Z' ddl(m)Z) edG dd dej*Z+G dd dej*Z,dd Z-d2ddZ.ej/e0ej/ddd Z1d3ej*ej/ej/ej/eej/ e2e2ee! d"d#d$Z3G d%d& d&ej*Z4G d'd( d(eZ5G d)d* d*ej*Z6e"G d+d, d,eZ7e"G d-d. d.e7Z8e"G d/d0 d0e7eZ9g d1Z:dS )4    )CallableOptionalUnionN)nn   )ACT2FN)CacheDynamicCache)GenerationMixin)use_kernel_forward_from_hub)create_causal_mask)FlashAttentionKwargs)GradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargsauto_docstringcan_return_tuple)deprecate_kwarg)check_model_inputs   )BitNetConfigZRMSNormc                       s.   e Zd Zd fdd	Zdd Zdd Z  ZS )	BitNetRMSNormư>c                    s&   t    tt|| _|| _dS )z<
        BitNetRMSNorm is equivalent to T5LayerNorm
        N)super__init__r   	ParametertorchZonesweightvariance_epsilon)selfhidden_sizeeps	__class__ f/var/www/html/assistant/venv/lib/python3.9/site-packages/transformers/models/bitnet/modeling_bitnet.pyr    -   s    
zBitNetRMSNorm.__init__c                 C   sJ   |j }|tj}|djddd}|t|| j  }| j|| S )N   T)Zkeepdim)	dtypetor"   float32powmeanZrsqrtr$   r#   )r%   hidden_statesZinput_dtypeZvariancer*   r*   r+   forward5   s
    zBitNetRMSNorm.forwardc                 C   s   t | jj d| j S )Nz, eps=)tupler#   shaper$   )r%   r*   r*   r+   
extra_repr<   s    zBitNetRMSNorm.extra_repr)r   )__name__
__module____qualname__r    r4   r7   __classcell__r*   r*   r(   r+   r   +   s   r   c                       s*   e Zd Zed fddZdd Z  ZS )	BitNetMLPconfigc                    s   t    || _|j| _|j| _tj| j| jdd| _tj| j| jdd| _tj| j| jdd| _	t
|j | _t|j|jd| _d S )NFZbiasr'   )r   r    r>   r&   Zintermediate_sizer   Linear	gate_projup_proj	down_projr   Z
hidden_actact_fnr   rms_norm_epsffn_sub_normr%   r>   r(   r*   r+   r    A   s    
zBitNetMLP.__init__c              	   C   s*   |  | | | || | }|S )N)rD   rG   rE   rB   rC   )r%   xrD   r*   r*   r+   r4   L   s    &zBitNetMLP.forward)r8   r9   r:   r   r    r4   r;   r*   r*   r(   r+   r<   @   s   r<   c                 C   sH   | dd| j d d f }| d| j d d df }tj| |fddS )z*Rotates half the hidden dims of the input..Nr-   r,   dim)r6   r"   cat)rI   x1Zx2r*   r*   r+   rotate_halfQ   s    rN   c                 C   sD   | |}| |}| | t| |  }|| t||  }||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.
        position_ids (`torch.Tensor`, *optional*):
            Deprecated and unused.
        unsqueeze_dim (`int`, *optional*, defaults to 1):
            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
    Returns:
        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
    )	unsqueezerN   )qkcossinposition_idsZunsqueeze_dimZq_embedZk_embedr*   r*   r+   apply_rotary_pos_embX   s
    

rU   )r3   n_repreturnc                 C   s^   | j \}}}}|dkr| S | dddddddddf |||||} | ||| ||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)r6   expandreshape)r3   rV   batchnum_key_value_headsslenhead_dimr*   r*   r+   	repeat_kvs   s
    0r^           )modulequerykeyvalueattention_maskscalingdropoutkwargsc                 K   s   t || j}t || j}	t||dd| }
|d urf|d d d d d d d |jd f }|
| }
tjj|
dtj	d
|j}
tjj|
|| jd}
t|
|	}|dd }||
fS )Nr,   r   r-   )rK   r.   )ptrainingr   )r^   num_key_value_groupsr"   matmul	transposer6   r   Z
functionalZsoftmaxr0   r/   r.   rf   rj   
contiguous)r`   ra   rb   rc   rd   re   rf   rg   
key_statesvalue_statesattn_weightscausal_maskattn_outputr*   r*   r+   eager_attention_forward   s    
&rt   c                       s   e Zd ZdZeed fddZedddddej	e
ej	ej	f eej	 ee eej ee e
ej	eej	 f d
ddZ  ZS )BitNetAttentionz=Multi-headed attention from 'Attention Is All You Need' paperr>   	layer_idxc                    s   t    || _|| _t|d|j|j | _|j|j | _	| jd | _
|j| _d| _tj|j|j| j |jd| _tj|j|j| j |jd| _tj|j|j| j |jd| _tj|j| j |j|jd| _t|j|jd| _d S )Nr]   g      Tr?   r@   )r   r    r>   rw   getattrr&   Znum_attention_headsr]   r[   rk   re   attention_dropoutZ	is_causalr   rA   Zattention_biasq_projk_projv_projo_projr   rF   attn_sub_normr%   r>   rw   r(   r*   r+   r       s*    
zBitNetAttention.__init__past_key_valuepast_key_values4.58new_nameversionN)r3   position_embeddingsrd   r   cache_positionrg   rW   c                 K   s.  |j d d }g |d| jR }| ||dd}	| ||dd}
| ||dd}|\}}t|	|
||\}	}
|d ur|||d}||
|| j	|\}
}t
}| jjdkrt| jj }|| |	|
||f| jsdn| j| jd|\}}|jg |dR   }| |}| |}||fS )Nr-   r   r,   )rS   rR   r   eagerr_   )rf   re   )r6   r]   rz   viewrm   r{   r|   rU   updaterw   rt   r>   Z_attn_implementationr   rj   ry   re   rY   rn   r~   r}   )r%   r3   r   rd   r   r   rg   Zinput_shapeZhidden_shapeZquery_statesro   rp   rR   rS   Zcache_kwargsZattention_interfacers   rq   r*   r*   r+   r4      s:    



zBitNetAttention.forward)NN)r8   r9   r:   __doc__r   intr    r   r"   Tensorr5   r   r   
LongTensorr   r   r4   r;   r*   r*   r(   r+   ru      s     ru   c                       s   e Zd Zeed fddZedddddeje	ej e	ej
 e	e e	e e	ej
 e	eejejf  ee ejd
	ddZ  ZS )BitNetDecoderLayerrv   c                    sR   t    |j| _t||d| _t|| _t|j|jd| _	t|j|jd| _
d S )Nrv   r@   )r   r    r&   ru   	self_attnr<   mlpr   rF   input_layernormpost_attention_layernormr   r(   r*   r+   r       s    

zBitNetDecoderLayer.__init__r   r   r   r   NF)	r3   rd   rT   r   	use_cacher   r   rg   rW   c              
   K   s^   |}	|  |}| jf |||||||d|\}}
|	| }|}	| |}| |}|	| }|S )N)r3   rd   rT   r   r   r   r   )r   r   r   r   )r%   r3   rd   rT   r   r   r   r   rg   Zresidual_r*   r*   r+   r4      s&    




zBitNetDecoderLayer.forward)NNNFNN)r8   r9   r:   r   r   r    r   r"   r   r   r   r   boolr5   r   r   r4   r;   r*   r*   r(   r+   r      s&   
      r   c                       sD   e Zd ZU ejed< ded fddZe e	dd Z
  ZS )	BitNetRotaryEmbeddinginv_freqNr=   c                    s   t    t|dr:t|jtr:|jd|jd| _nd| _|j| _	|j| _
|| _t| j | _| | j|\}| _| jd|dd | j| _d S )Nrope_scaling	rope_typetypedefaultr   F)
persistent)r   r    hasattr
isinstancer   dictgetr   Zmax_position_embeddingsZmax_seq_len_cachedZoriginal_max_seq_lenr>   r   Zrope_init_fnattention_scalingZregister_bufferr   Zoriginal_inv_freq)r%   r>   devicer   r(   r*   r+   r      s    
zBitNetRotaryEmbedding.__init__c           
      C   s   | j d d d d f  |jd dd|j}|d d d d d f  }t|jjtrl|jjdkrl|jjnd}t	j
|ddV | |  dd}t	j||fdd	}| | j }| | j }	W d    n1 s0    Y  |j|jd
|	j|jd
fS )Nr   r-   r   ZmpscpuF)device_typeZenabledr,   rJ   )r.   )r   floatrX   r6   r/   r   r   r   strr"   Zautocastrm   rL   rR   r   rS   r.   )
r%   rI   rT   Zinv_freq_expandedZposition_ids_expandedr   ZfreqsZembrR   rS   r*   r*   r+   r4   %  s    0&,zBitNetRotaryEmbedding.forward)N)r8   r9   r:   r"   r   __annotations__r   r    Zno_gradr   r4   r;   r*   r*   r(   r+   r     s
   

r   c                   @   sH   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dS )BitNetPreTrainedModelr>   modelTr   r   )r3   
attentionsN)r8   r9   r:   r   r   Zbase_model_prefixZsupports_gradient_checkpointingZ_no_split_modulesZ_skip_keys_device_placementZ_supports_flash_attnZ_supports_sdpaZ_supports_flex_attnZ_can_compile_fullgraphZ_supports_attention_backendr   ru   Z_can_record_outputsr*   r*   r*   r+   r   5  s   
r   c                       st   e Zd Zed fddZeedeej	 eej
 eej	 ee eej eej	 ee ee ed	ddZ  ZS )	BitNetModelr=   c                    s   t     j| _ j| _t j j| j| _t	 fddt
 jD | _t j jd| _t d| _d| _|   d S )Nc                    s   g | ]}t  |qS r*   )r   ).0rw   r=   r*   r+   
<listcomp>Q      z(BitNetModel.__init__.<locals>.<listcomp>r@   r=   F)r   r    Zpad_token_idZpadding_idx
vocab_sizer   Z	Embeddingr&   embed_tokensZ
ModuleListrangenum_hidden_layerslayersr   rF   normr   
rotary_embZgradient_checkpointing	post_initrH   r(   r=   r+   r    J  s    zBitNetModel.__init__N)		input_idsrd   rT   r   inputs_embedsr   r   rg   rW   c              	   K   s   |d u |d uA rt d|d u r*| |}|rB|d u rBt| jd}|d u rz|d urZ| nd}	tj|	|	|jd  |jd}|d u r|	d}t
| j|||||d}
|}| ||}| jd | jj D ] }||f|
||||d|}q| |}t||dS )	Nz:You must specify exactly one of input_ids or inputs_embedsr=   r   r   )r   )r>   Zinput_embedsrd   r   r   rT   )rd   rT   r   r   r   )last_hidden_stater   )
ValueErrorr   r	   r>   Zget_seq_lengthr"   Zaranger6   r   rO   r   r   r   r   r   r   )r%   r   rd   rT   r   r   r   r   rg   Zpast_seen_tokensrr   r3   r   Zdecoder_layerr*   r*   r+   r4   Z  sP    

	

zBitNetModel.forward)NNNNNNN)r8   r9   r:   r   r    r   r   r   r"   r   r   r   FloatTensorr   r   r   r   r4   r;   r*   r*   r(   r+   r   H  s*          r   c                       s   e Zd ZdgZdZdZ fddZeed	e	e
j e	e
j e	e
j e	e e	e
j e	e
j e	e e	e
j eee
jf ee edddZ  ZS )
BitNetForCausalLMzlm_head.weightNc                    s@   t  | t|| _|j| _tj|j|jdd| _| 	  d S )NFr?   )
r   r    r   r   r   r   rA   r&   lm_headr   rH   r(   r*   r+   r      s
    
zBitNetForCausalLM.__init__r   )r   rd   rT   r   r   labelsr   r   logits_to_keeprg   rW   c
              
   K   s   | j f |||||||d|
}|j}t|	tr<t|	 dn|	}| |dd|ddf }d}|dur| jf ||| jjd|
}t	|||j
|j|jd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   rd   rT   r   r   r   r   N)logitsr   r   )lossr   r   r3   r   )r   r   r   r   slicer   Zloss_functionr>   r   r   r   r3   r   )r%   r   rd   rT   r   r   r   r   r   r   rg   outputsr3   Zslice_indicesr   r   r*   r*   r+   r4     s0    %zBitNetForCausalLM.forward)	NNNNNNNNr   )r8   r9   r:   Z_tied_weights_keysZ_tp_planZ_pp_planr    r   r   r   r"   r   r   r   r   r   r   r   r   r   r   r4   r;   r*   r*   r(   r+   r     s8   	         r   )r   r   r   )Nr   )r_   );typingr   r   r   r"   r   Zactivationsr   Zcache_utilsr   r	   Z
generationr
   Zintegrationsr   Zmasking_utilsr   Zmodeling_flash_attention_utilsr   Zmodeling_layersr   Zmodeling_outputsr   r   Zmodeling_rope_utilsr   r   Zmodeling_utilsr   r   Zprocessing_utilsr   utilsr   r   r   Zutils.deprecationr   Zutils.genericr   Zconfiguration_bitnetr   Moduler   r<   rN   rU   r   r   r^   r   rt   ru   r   r   r   r   r   __all__r*   r*   r*   r+   <module>   sX   
 J.$NP