o
    ei^W                     @   s  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	 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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' ddl(m)Z)m*Z* ddl+m,Z, ddl-m.Z. G dd dej/Z0edG dd dej/Z1G dd dej/Z2dd Z3eddEd d!Z4d"ej5d#e6d$ej5fd%d&Z7	'dFd(ej/d)ej5d*ej5d+ej5d,ej5dB d-e8d.e8d/e$e& fd0d1Z9ee4G d2d3 d3ej/Z:G d4d5 d5eZ;eG d6d7 d7e"Z<eG d8d9 d9e<Z=ed:d;G d<d= d=e<eZ>ed:d;G d>d? d?ee<Z?ed:d;G d@dA dAee<Z@ed:d;G dBdC dCee<ZAg dDZBdS )G    )Callable)OptionalN)nn)auto_docstring   )ACT2FN)CacheDynamicCache)GenerationMixin)use_kernel_forward_from_hubuse_kernel_func_from_hubuse_kernelized_func)create_causal_mask)GenericForQuestionAnswering GenericForSequenceClassificationGenericForTokenClassificationGradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)TransformersKwargscan_return_tuple)maybe_autocastmerge_with_config_defaults)capture_outputs   )ArceeConfigc                       s$   e Zd Z fddZdd Z  ZS )ArceeMLPc                    s`   t    || _|j| _|j| _tj| j| j|jd| _tj| j| j|jd| _	t
|j | _d S )Nbias)super__init__confighidden_sizeintermediate_sizer   Linearmlp_biasup_proj	down_projr   
hidden_actact_fnselfr&   	__class__ f/home/ubuntu/transcripts/venv/lib/python3.10/site-packages/transformers/models/arcee/modeling_arcee.pyr%   3   s   
zArceeMLP.__init__c                 C   s   |  | | |S N)r,   r.   r+   )r0   xr3   r3   r4   forward<   s   zArceeMLP.forward)__name__
__module____qualname__r%   r7   __classcell__r3   r3   r1   r4   r!   2   s    	r!   RMSNormc                       sF   e Zd Zddeddf fddZdejdejfdd	Zd
d Z  Z	S )ArceeRMSNormư>epsreturnNc                    s&   t    tt|| _|| _dS )z;
        ArceeRMSNorm is equivalent to T5LayerNorm
        N)r$   r%   r   	Parametertorchonesweightvariance_epsilon)r0   r'   r?   r1   r3   r4   r%   B   s   

zArceeRMSNorm.__init__hidden_statesc                 C   sJ   |j }|tj}|djddd}|t|| j  }| j|| S )N   T)keepdim)	dtypetorB   float32powmeanrsqrtrE   rD   )r0   rF   input_dtypevariancer3   r3   r4   r7   J   s
   zArceeRMSNorm.forwardc                 C   s   t | jj d| j S )Nz, eps=)tuplerD   shaperE   )r0   r3   r3   r4   
extra_reprQ   s   zArceeRMSNorm.extra_repr)r>   )
r8   r9   r:   floatr%   rB   Tensorr7   rT   r;   r3   r3   r1   r4   r=   @   s    r=   c                       s~   e Zd ZU ejed< ddef fddZe			ddedB de	d de
dB d	ed
ef fddZe edd Z  ZS )ArceeRotaryEmbeddinginv_freqNr&   c                    s   t    |j| _|j| _|| _| jjd | _| j}| jdkr$t	| j }|| j|\}| _
| jd|dd | jd| dd d S )N	rope_typedefaultrX   F)
persistentoriginal_inv_freq)r$   r%   max_position_embeddingsmax_seq_len_cachedoriginal_max_seq_lenr&   rope_parametersrY   compute_default_rope_parametersr   attention_scalingregister_bufferclone)r0   r&   devicerope_init_fnrX   r1   r3   r4   r%   X   s   


zArceeRotaryEmbedding.__init__re   ztorch.deviceseq_lenr@   ztorch.Tensorc                 C   sZ   | j d }t| ddp| j| j }d}d|tjd|dtjdj|tjd|   }||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head_dimNg      ?r   rG   rJ   )re   rJ   )	r`   getattrr'   num_attention_headsrB   arangeint64rK   rU   )r&   re   rg   basedimattention_factorrX   r3   r3   r4   ra   h   s   
&z4ArceeRotaryEmbedding.compute_default_rope_parametersc           
      C   s   | j d d d d f  |jd dd|j}|d d d d d f  }t|jjtr6|jjdkr6|jjnd}t	|dd+ | |  
dd}tj||fdd	}| | j }| | j }	W d    n1 slw   Y  |j|jd
|	j|jd
fS )Nr   rH   r   mpscpuF)device_typeenabledrG   rp   rj   )rX   rU   expandrS   rK   re   
isinstancetypestrr   	transposerB   catcosrb   sinrJ   )
r0   r6   position_idsinv_freq_expandedposition_ids_expandedrt   freqsembr}   r~   r3   r3   r4   r7      s   0&zArceeRotaryEmbedding.forwardr5   )NNN)r8   r9   r:   rB   rV   __annotations__r    r%   staticmethodr   intrR   rU   ra   no_gradr   r7   r;   r3   r3   r1   r4   rW   U   s&   
 

rW   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..NrH   rG   rv   )rS   rB   r|   )r6   x1x2r3   r3   r4   rotate_half   s   r   rotary_pos_embc                 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.
        unsqueeze_dim (`int`, *optional*, defaults to 1):
            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
    Returns:
        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
    )	unsqueezer   )qkr}   r~   unsqueeze_dimq_embedk_embedr3   r3   r4   apply_rotary_pos_emb   s
   

r   rF   n_repr@   c                 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)rS   rw   reshape)rF   r   batchnum_key_value_headsslenri   r3   r3   r4   	repeat_kv   s
   0r           modulequerykeyvalueattention_maskscalingdropoutkwargsc                 K   s   t || j}t || j}	t||dd| }
|d ur |
| }
tjj|
dtjd	|j
}
tjj|
|| jd}
t|
|	}|dd }||
fS )NrG   r   rH   )rp   rJ   )ptrainingr   )r   num_key_value_groupsrB   matmulr{   r   
functionalsoftmaxrL   rK   rJ   r   r   
contiguous)r   r   r   r   r   r   r   r   
key_statesvalue_statesattn_weightsattn_outputr3   r3   r4   eager_attention_forward   s   
r   c                       s   e Zd ZdZdedef fddZ				ddejde	ejejf dB d	ejdB d
e
dB dejdB dee de	ejejf fddZ  ZS )ArceeAttentionz=Multi-headed attention from 'Attention Is All You Need' paperr&   	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| _d S )Nri   g      Tr"   )r$   r%   r&   r   rk   r'   rl   ri   r   r   r   attention_dropout	is_causalr   r)   attention_biasq_projk_projv_projo_projr0   r&   r   r1   r3   r4   r%      s(   
zArceeAttention.__init__NrF   position_embeddingsr   past_key_valuescache_positionr   r@   c                 K   s  |j d d }g |d| jR }| ||dd}	| ||dd}
| ||dd}|\}}t|	|
||\}	}
|d urW|||d}||
|| j	|\}
}t
| jjt}|| |	|
||f| jskdn| j| jd|\}}|jg |dR   }| |}||fS )NrH   r   rG   )r~   r}   r   r   )r   r   )rS   ri   r   viewr{   r   r   r   updater   r   get_interfacer&   _attn_implementationr   r   r   r   r   r   r   )r0   rF   r   r   r   r   r   input_shapehidden_shapequery_statesr   r   r}   r~   cache_kwargsattention_interfacer   r   r3   r3   r4   r7      s8   	

zArceeAttention.forward)NNNN)r8   r9   r:   __doc__r    r   r%   rB   rV   rR   r   
LongTensorr   r   r7   r;   r3   r3   r1   r4   r      s,    r   c                       s   e Zd Zdedef fddZ						ddejdejdB d	ejdB d
e	dB de
dB dejdB deejejf dB dee dejfddZ  ZS )ArceeDecoderLayerr&   r   c                    sR   t    |j| _t||d| _t|| _t|j|jd| _	t|j|jd| _
d S )N)r&   r   r?   )r$   r%   r'   r   	self_attnr!   mlpr=   rms_norm_epsinput_layernormpost_attention_layernormr   r1   r3   r4   r%   $  s   

zArceeDecoderLayer.__init__NFrF   r   r   r   	use_cacher   r   r   r@   c              
   K   s^   |}	|  |}| jd|||||||d|\}}
|	| }|}	| |}| |}|	| }|S )N)rF   r   r   r   r   r   r   r3   )r   r   r   r   )r0   rF   r   r   r   r   r   r   r   residual_r3   r3   r4   r7   .  s&   




zArceeDecoderLayer.forward)NNNFNN)r8   r9   r:   r    r   r%   rB   rV   r   r   boolrR   r   r   r7   r;   r3   r3   r1   r4   r   #  s6    	
r   c                   @   sH   e Zd ZU eed< dZdZdgZdgZdZ	dZ
dZdZdZeedZdS )ArceePreTrainedModelr&   modelTr   r   )rF   
attentionsN)r8   r9   r:   r    r   base_model_prefixsupports_gradient_checkpointing_no_split_modules_skip_keys_device_placement_supports_flash_attn_supports_sdpa_supports_flex_attn_can_compile_fullgraph_supports_attention_backendr   r   _can_record_outputsr3   r3   r3   r4   r   P  s   
 
r   c                       s   e Zd Zdef fddZeee							ddej	dB dej
dB dej	dB dedB d	ejdB d
ej	dB dedB dee defddZ  ZS )
ArceeModelr&   c                    s   t     j| _ j| _t j j| j| _t	 fddt
 jD | _t j jd| _t d| _d| _|   d S )Nc                    s   g | ]}t  |qS r3   )r   ).0r   r&   r3   r4   
<listcomp>l  s    z'ArceeModel.__init__.<locals>.<listcomp>r   r   F)r$   r%   pad_token_idpadding_idx
vocab_sizer   	Embeddingr'   embed_tokens
ModuleListrangenum_hidden_layerslayersr=   r   normrW   
rotary_embgradient_checkpointing	post_initr/   r1   r   r4   r%   e  s   zArceeModel.__init__N	input_idsr   r   r   inputs_embedsr   r   r   r@   c              
   K   s   |d u |d uA rt d|d u r| |}|r!|d u r!t| jd}|d u r<|d ur-| nd}	tj|jd |jd|	 }|d u rE|	d}t
| j|||||d}
|}| j||d}| jd | jj D ]}||f|
|||||d|}qb| |}t||d	S )
Nz:You must specify exactly one of input_ids or inputs_embedsr   r   r   )re   )r&   r   r   r   r   r   )r   )r   r   r   r   r   r   )last_hidden_stater   )
ValueErrorr   r	   r&   get_seq_lengthrB   rm   rS   re   r   r   r   r   r   r   r   )r0   r   r   r   r   r   r   r   r   past_seen_tokenscausal_maskrF   r   decoder_layerr3   r3   r4   r7   u  sP   

	
zArceeModel.forward)NNNNNNN)r8   r9   r:   r    r%   r   r   r   rB   r   rV   r   FloatTensorr   r   r   r   r7   r;   r3   r3   r1   r4   r   c  s>    	
r   zarcee-ai/AFM-4.5B)
checkpointc                       s   e Zd ZddiZddiZddgdgfiZ fddZee																	
dde	j
d	B de	jd	B de	j
d	B ded	B de	jd	B de	j
d	B ded	B de	j
d	B dee	jB dee defddZ  ZS )ArceeForCausalLMzlm_head.weightzmodel.embed_tokens.weightlm_headcolwise_gather_outputrF   logitsc                    s@   t  | t|| _|j| _tj|j|jdd| _| 	  d S )NFr"   )
r$   r%   r   r   r   r   r)   r'   r   r   r/   r1   r3   r4   r%     s
   
zArceeForCausalLM.__init__Nr   r   r   r   r   r   labelsr   r   logits_to_keepr   r@   c
              
   K   s   | j d|||||||d|
}|j}t|	trt|	 dn|	}| |dd|ddf }d}|durB| jd||| jjd|
}t	|||j
|j|jdS )a  
        Example:

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

        >>> model = ArceeForCausalLM.from_pretrained("meta-arcee/Arcee-2-7b-hf")
        >>> tokenizer = AutoTokenizer.from_pretrained("meta-arcee/Arcee-2-7b-hf")

        >>> 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   r   r   r   r   r   r   N)r   r   r   )lossr   r   rF   r   r3   )r   r   rx   r   slicer   loss_functionr&   r   r   r   rF   r   )r0   r   r   r   r   r   r   r   r   r   r   outputsrF   slice_indicesr   r   r3   r3   r4   r7     s0    zArceeForCausalLM.forward)	NNNNNNNNr   )r8   r9   r:   _tied_weights_keys_tp_plan_pp_planr%   r   r   rB   r   rV   r   r   r   r   r   r   r   r7   r;   r3   r3   r1   r4   r     sN    		
r   c                   @      e Zd ZdS )ArceeForSequenceClassificationNr8   r9   r:   r3   r3   r3   r4   r         r  c                   @   s   e Zd ZdZdS )ArceeForQuestionAnsweringtransformerN)r8   r9   r:   r   r3   r3   r3   r4   r    s    r  c                   @   r  )ArceeForTokenClassificationNr	  r3   r3   r3   r4   r  
  r
  r  )r   r  r  r  r   r   )r   )r   )Ccollections.abcr   typingr   rB   r   transformers.utilsr   activationsr   cache_utilsr   r	   
generationr
   integrationsr   r   r   masking_utilsr   modeling_layersr   r   r   r   modeling_outputsr   r   modeling_rope_utilsr   r   modeling_utilsr   r   processing_utilsr   utilsr   r   utils.genericr   r   utils.output_capturingr   configuration_arceer    Moduler!   r=   rW   r   r   rV   r   r   rU   r   r   r   r   r   r   r  r  r  __all__r3   r3   r3   r4   <module>   sz   A
F-PK