o
    ZhE                     @   sZ  d dl Z d dlmZ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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" ddl#m$Z$m%Z%m&Z&m'Z'm(Z( ddl)m*Z* e' rd dl+m,Z, ddl-m.Z. e(/e0Z1G dd dej2Z3G dd dej2Z4G dd dej2Z5dej6de7dej6fddZ8	d?dej2d ej6d!ej6d"ej6d#eej6 d$e9d%e9fd&d'Z:d(d) Z;d@d*d+Z<G d,d- d-ej2Z=G d.d/ d/eZ>e%G d0d1 d1e Z?e%G d2d3 d3e?Z@G d4d5 d5ee$ZAe%G d6d7 d7e?eZBe%d8d9G d:d; d;e?ZCe%G d<d= d=e?ZDg d>ZEdS )A    N)CallableOptionalTupleUnion   )ACT2FN)CacheDynamicCache)GenerationMixin)AttentionMaskConverter)FlashAttentionKwargs)GradientCheckpointingLayer)BaseModelOutputWithPastCausalLMOutputWithPast SequenceClassifierOutputWithPastTokenClassifierOutput)ROPE_INIT_FUNCTIONSdynamic_rope_update)ALL_ATTENTION_FUNCTIONSPreTrainedModel)Unpack)
LossKwargsauto_docstringcan_return_tupleis_torch_flex_attn_availablelogging   )HeliumConfig)	BlockMask)make_flex_block_causal_maskc                       s.   e Zd Zd fdd	Zdd Zdd Z  ZS )	HeliumRMSNormư>c                    s&   t    tt|| _|| _d S N)super__init__nn	ParametertorchZonesweightvariance_epsilon)selfhidden_sizeeps	__class__ Y/var/www/auris/lib/python3.10/site-packages/transformers/models/helium/modeling_helium.pyr$   9   s   

zHeliumRMSNorm.__init__c                 C   sR   |j }|tj}|djddd}|t|| j  }| jtj| |S )N   T)Zkeepdim)	dtypetor'   float32powmeanZrsqrtr)   r(   )r*   hidden_statesZinput_dtypeZvariancer/   r/   r0   forward>   s
   zHeliumRMSNorm.forwardc                 C   s   t | jj d| j S )Nz, eps=)tupler(   shaper)   r*   r/   r/   r0   
extra_reprE   s   zHeliumRMSNorm.extra_repr)r!   )__name__
__module____qualname__r$   r9   r=   __classcell__r/   r/   r-   r0   r    8   s    r    c                       s8   e Zd Zddef fddZe edd Z  Z	S )HeliumRotaryEmbeddingNconfigc                    s   t    t|dr|jd u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defaultinv_freqF)
persistent)r#   r$   hasattrrD   getrE   Zmax_position_embeddingsZmax_seq_len_cachedZoriginal_max_seq_lenrC   r   Zrope_init_fnattention_scalingZregister_bufferrH   Zoriginal_inv_freq)r*   rC   devicerH   r-   r/   r0   r$   J   s   
zHeliumRotaryEmbedding.__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r6|jjdkr6|jjnd}t	j
|dd+ | |  dd}t	j||fdd	}| | j }| | j }	W d    n1 smw   Y  |j|jd
|	j|jd
fS )Nr   r2   r   ZmpscpuF)device_typeenabledr1   dim)r3   )rH   floatexpandr;   r4   rM   
isinstancerF   strr'   Zautocast	transposecatcosrL   sinr3   )
r*   xposition_idsZinv_freq_expandedZposition_ids_expandedrO   ZfreqsZembrY   rZ   r/   r/   r0   r9   [   s   0&zHeliumRotaryEmbedding.forwardr"   )
r>   r?   r@   r   r$   r'   Zno_gradr   r9   rA   r/   r/   r-   r0   rB   I   s
    rB   c                       s$   e Zd Z fddZdd Z  ZS )	HeliumMLPc                    sx   t    || _|j| _|j| _tj| j| j|jd| _tj| j| j|jd| _	tj| j| j|jd| _
t|j | _d S )Nbias)r#   r$   rC   r+   Zintermediate_sizer%   LinearZmlp_bias	gate_projup_proj	down_projr   Z
hidden_actact_fnr*   rC   r-   r/   r0   r$   l   s   
zHeliumMLP.__init__c                 C   s$   |  | | || | }|S r"   )rc   rd   ra   rb   )r*   r[   rc   r/   r/   r0   r9   v   s    zHeliumMLP.forward)r>   r?   r@   r$   r9   rA   r/   r/   r-   r0   r]   k   s    
r]   r8   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)r;   rT   reshape)r8   rf   batchnum_key_value_headsslenhead_dimr/   r/   r0   	repeat_kv{   s
   0rm           modulequerykeyvalueattention_maskscalingdropoutc                 K   s   t || j}t || j}	t||dd| }
|d ur3|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 )Nr1   r   r2   )rR   r3   )ptrainingr   )rm   num_key_value_groupsr'   matmulrW   r;   r%   Z
functionalZsoftmaxr5   r4   r3   ru   rx   
contiguous)ro   rp   rq   rr   rs   rt   ru   kwargs
key_statesvalue_statesattn_weightscausal_maskattn_outputr/   r/   r0   eager_attention_forward   s   
&r   c                 C   s>   | ddddf }| ddddf }t j| |fdddS )	z*Rotates half the hidden dims of the input..r   Nr1   r   r2   rQ   rv   )r'   stackflatten)r[   x1Zx2r/   r/   r0   rotate_half   s   r   c                 C   s   | |}| |}|dd|jd d f jddd}|dd|jd d f jddd}| | 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.
    .Nr2   r1   rQ   )	unsqueezer;   Zrepeat_interleaver   )qkrY   rZ   r\   Zunsqueeze_dimZq_embedZk_embedr/   r/   r0   apply_rotary_pos_emb   s   

$$r   c                       s   e Zd ZdZddedee f fddZ		ddej	de
ej	ej	f d	eej	 d
ee deej dee de
ej	eej	 ee
ej	  f fddZ  ZS )HeliumAttentionz=Multi-headed attention from 'Attention Is All You Need' paperNrC   	layer_idxc                    s   t    || _|| _t|d|j|j | _|j|j | _	dt
| 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|jdd| _d S )Nrl   r   Tr^   F)r#   r$   rC   r   getattrr+   Znum_attention_headsrl   rj   ry   mathsqrtrt   attention_dropoutZ	is_causalr%   r`   Zattention_biasq_projk_projv_projo_projr*   rC   r   r-   r/   r0   r$      s$   
zHeliumAttention.__init__r8   position_embeddingsrs   past_key_valuecache_positionr|   rg   c                 K   sH  |j d d }g |d| jR }| ||dd}	| ||dd}
| ||dd}|\}}t|	|
||\}	}
|d urW|||d}||
|| j	|\}
}t
}| jjdkrw| jjdkrq|ddrqtd	 nt| jj }|| |	|
||f| jsd
n| j| jd|\}}|jg |dR   }| |}||fS )Nr2   r   r1   )rZ   rY   r   eagersdpaoutput_attentionsFz`torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to eager attention. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.rn   )ru   rt   )r;   rl   r   viewrW   r   r   r   updater   r   rC   _attn_implementationrK   loggerwarning_oncer   rx   r   rt   rh   r{   r   )r*   r8   r   rs   r   r   r|   Zinput_shapeZhidden_shapeZquery_statesr}   r~   rY   rZ   Zcache_kwargsZattention_interfacer   r   r/   r/   r0   r9      s@   	

zHeliumAttention.forwardr"   )NN)r>   r?   r@   __doc__r   r   intr$   r'   Tensorr   r   
LongTensorr   r   r9   rA   r/   r/   r-   r0   r      s(    r   c                       s   e Zd Zddedee f fddZ							ddejdeej d	eej	 d
ee
 dee dee deej	 deeejejf  dee deejeeejejf  f fddZ  ZS )HeliumDecoderLayerNrC   r   c                    sR   t    |j| _t||d| _t|| _t|j|jd| _	t|j|jd| _
d S )N)rC   r   r,   )r#   r$   r+   r   	self_attnr]   mlpr    rms_norm_epsinput_layernormpost_attention_layernormr   r-   r/   r0   r$     s   

zHeliumDecoderLayer.__init__Fr8   rs   r\   r   r   	use_cacher   r   r|   rg   c	                 K   st   |}
|  |}| jd||||||||d|	\}}|
| }|}
| |}| |}|
| }|f}|r8||f7 }|S )N)r8   rs   r\   r   r   r   r   r   r/   )r   r   r   r   )r*   r8   rs   r\   r   r   r   r   r   r|   ZresidualZself_attn_weightsoutputsr/   r/   r0   r9     s.   
	



zHeliumDecoderLayer.forwardr"   )NNNFFNN)r>   r?   r@   r   r   r   r$   r'   r   r   r   boolr   r   r   FloatTensorr9   rA   r/   r/   r-   r0   r     s<    	
r   c                   @   sH   e Zd ZeZdZdZdgZdgZdZ	dZ
dZdZdZdZdZdd ZdS )HeliumPreTrainedModelmodelTr   past_key_valuesc                 C   s   | j j}t|tjr"|jjjd|d |jd ur |jj	  d S d S t|tj
rC|jjjd|d |jd urA|jj|j 	  d S d S t|trQ|jjd d S d S )Nrn   )r7   stdg      ?)rC   Zinitializer_rangerU   r%   r`   r(   dataZnormal_r_   Zzero_	Embeddingpadding_idxr    Zfill_)r*   ro   r   r/   r/   r0   _init_weightsX  s   


z#HeliumPreTrainedModel._init_weightsN)r>   r?   r@   r   Zconfig_classZbase_model_prefixZsupports_gradient_checkpointingZ_no_split_modulesZ_skip_keys_device_placementZ_supports_flash_attn_2Z_supports_sdpaZ_supports_flex_attnZ_supports_cache_classZ_supports_quantized_cacheZ_supports_static_cacheZ_supports_attention_backendr   r/   r/   r/   r0   r   I  s    r   c                       s  e Zd Zdef fddZdd Zdd Zee									d!d	e	e
j d
e	e
j de	e
j de	e de	e
j de	e de	e de	e de	e
j dee defddZ	d"d
ee
jdf de
jde
jdedef
ddZed
e
jdedede
jde
jdefdd Z  ZS )#HeliumModelrC   c                    s   t     j| _ j| _t j j| j| _t	 fddt
 jD | _t j jd| _t | _d| _|   d S )Nc                    s   g | ]}t  |qS r/   )r   ).0r   rC   r/   r0   
<listcomp>o  s    z(HeliumModel.__init__.<locals>.<listcomp>r   F)r#   r$   pad_token_idr   
vocab_sizer%   r   r+   embed_tokensZ
ModuleListrangenum_hidden_layerslayersr    r   normrB   
rotary_embgradient_checkpointing	post_initre   r-   r   r0   r$   h  s   
zHeliumModel.__init__c                 C      | j S r"   r   r<   r/   r/   r0   get_input_embeddingsx     z HeliumModel.get_input_embeddingsc                 C   
   || _ d S r"   r   r*   rr   r/   r/   r0   set_input_embeddings{     
z HeliumModel.set_input_embeddingsN	input_idsrs   r\   r   inputs_embedsr   r   output_hidden_statesr   flash_attn_kwargsrg   c
                 K   s  |d ur|n| j j}|d ur|n| j j}|d ur|n| j j}|d u |d uA r*td| jr9| jr9|r9td d}t	|t
d tfsFtd|d u rO| |}|rX|d u rXt }|	d u rt|d urd| nd}tj|||jd  |jd}	|d u r}|	d}| |||	||}|}| ||}|rdnd }|rdnd }| jd | j j D ]&}|r||f7 }||f||||||	|d	|
}|d }|r||d f7 }q| |}|r||f7 }t||r|nd ||d
S )Nz:You must specify exactly one of input_ids or inputs_embedszX`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.FzBThe `past_key_values` should be either a `Cache` object or `None`.r   r   rM   r/   )rs   r\   r   r   r   r   r   )last_hidden_stater   r8   
attentions)rC   r   r   r   
ValueErrorr   rx   r   r   rU   rF   r   r   r	   get_seq_lengthr'   aranger;   rM   r   _update_causal_maskr   r   r   r   r   )r*   r   rs   r\   r   r   r   r   r   r   r   past_seen_tokensr   r8   r   Zall_hidden_statesZall_self_attnsZdecoder_layerZlayer_outputsr/   r/   r0   r9   ~  sx   



	


zHeliumModel.forwardFr   input_tensorc                 C   s:  | j jdkr|d ur|dk r|S d S | j jdkr&t|tjr$t|}|S |d ur.| nd}|d ur7|jnd}| j jdkrO|sO|sOt	j
|||| jdrOd S |j}|jd }	|r^| }
nt|tjri|jd	 n||	 d }
| j||	|
|||jd d
}| j jdkr|d ur|jjdv r|st|j}t	||}|S )NZflash_attention_2rn   Zflex_attentionr   Fr   )r   Zpast_key_values_lengthZis_trainingr   r2   )sequence_lengthtarget_lengthr3   r   
batch_size)cudaZxpuZnpu)rC   r   anyrU   r'   r   r   r   Zis_compileabler   Z_ignore_causal_mask_sdparx   r3   r;   Zget_max_cache_shape5_prepare_4d_causal_attention_mask_with_cache_positionrM   rF   finfominZ_unmask_unattended)r*   rs   r   r   r   r   r   Zusing_compilable_cacher3   r   r   r   	min_dtyper/   r/   r0   r     sT   




zHeliumModel._update_causal_maskr   r   r3   r   c                 K   sD  | dur|   dkr| }|S t|j}tj||f|||jd}|dkr+tj|dd}|tj||jd|ddk9 }|ddddddf 	|ddd}| dur|
 }| jd }	|ddddddd|	f | ddddddf |j }
|
dk}
|ddddddd|	f |
||ddddddd|	f< |S )	aM  
        Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape
        `(batch_size, key_value_length)`, or if the input `attention_mask` is already 4D, do nothing.

        Args:
            attention_mask (`torch.Tensor`):
                A 2D attention mask of shape `(batch_size, key_value_length)` or a 4D attention mask of shape
                `(batch_size, 1, query_length, key_value_length)`.
            sequence_length (`int`):
                The sequence length being processed.
            target_length (`int`):
                The target length: when generating with static cache, the mask should be as long as the static cache,
                to account for the 0 padding, the part of the cache that is not filled yet.
            dtype (`torch.dtype`):
                The dtype to use for the 4D attention mask.
            cache_position (`torch.Tensor`):
                Indices depicting the position of the input sequence tokens in the sequence.
            batch_size (`torch.Tensor`):
                Batch size.
        N   )Z
fill_valuer3   rM   r   )Zdiagonalr   r2   r   )rR   r'   r   r   fullrM   Ztriur   rh   rT   cloner;   r4   Zmasked_fill)rs   r   r   r3   r   r   r|   r   r   Zmask_lengthZpadding_maskr/   r/   r0   r   "  s,    $
6  zAHeliumModel._prepare_4d_causal_attention_mask_with_cache_position	NNNNNNNNN)F)r>   r?   r@   r   r$   r   r   r   r   r   r'   r   r   r   r   r   r   r   r   r9   r   r   staticmethodr   r3   r   rA   r/   r/   r-   r0   r   f  s    	
d
Dr   c                   @   s   e Zd ZdS )KwargsForCausalLMN)r>   r?   r@   r/   r/   r/   r0   r   Z  s    r   c                       s  e Zd ZdgZddiZddgdgfiZdef fddZd	d
 Zdd Z	dd Z
dd Zdd Zdd Zee											d&deej deej deej dee deej deej dee dee dee d eej d!eeejf d"ee d#efd$d%Z  ZS )'HeliumForCausalLMzlm_head.weightlm_headZcolwise_repr8   logitsrC   c                    s@   t  | t|| _|j| _tj|j|jdd| _| 	  d S NFr^   )
r#   r$   r   r   r   r%   r`   r+   r   r   re   r-   r/   r0   r$   c  s
   
zHeliumForCausalLM.__init__c                 C      | j jS r"   r   r   r<   r/   r/   r0   r   l     z&HeliumForCausalLM.get_input_embeddingsc                 C      || j _d S r"   r   r   r/   r/   r0   r   o     z&HeliumForCausalLM.set_input_embeddingsc                 C   r   r"   r   r<   r/   r/   r0   get_output_embeddingsr  r   z'HeliumForCausalLM.get_output_embeddingsc                 C   r   r"   r   )r*   Znew_embeddingsr/   r/   r0   set_output_embeddingsu  r   z'HeliumForCausalLM.set_output_embeddingsc                 C   r   r"   r   )r*   decoderr/   r/   r0   set_decoderx  r   zHeliumForCausalLM.set_decoderc                 C   r   r"   r   r<   r/   r/   r0   get_decoder{  r   zHeliumForCausalLM.get_decoderNr   r   rs   r\   r   r   labelsr   r   r   r   logits_to_keepr|   rg   c                 K   s   |dur|n| j j}|	dur|	n| j j}	| jd||||||||	|
d	|}|j}t|tr4t| dn|}| |dd|ddf }d}|durX| j	d||| 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, ...,
            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, HeliumForCausalLM

        >>> model = HeliumForCausalLM.from_pretrained("google/helium-7b")
        >>> tokenizer = AutoTokenizer.from_pretrained("google/helium-7b")

        >>> prompt = "What is your favorite condiment?"
        >>> 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]
        "What is your favorite condiment?"
        ```N)	r   rs   r\   r   r   r   r   r   r   )r   r   r   lossr   r   r8   r   r/   )rC   r   r   r   r   rU   r   slicer   loss_functionr   r   r   r8   r   )r*   r   rs   r\   r   r   r   r   r   r   r   r   r|   r   r8   Zslice_indicesr   r   r/   r/   r0   r9   ~  s:   '
zHeliumForCausalLM.forward)NNNNNNNNNNr   )r>   r?   r@   Z_tied_weights_keysZ_tp_planZ_pp_planr   r$   r   r   r   r   r   r   r   r   r   r'   r   r   r   r   r   r   r   r   r   r   r9   rA   r/   r/   r-   r0   r   ]  sf    		
r   a  
    The Helium Model transformer with a sequence classification head on top (linear layer).

    [`HeliumForSequenceClassification`] uses the last token in order to do the classification, as other causal models
    (e.g. GPT-2) do.

    Since it does classification on the last token, it requires to know the position of the last token. If a
    `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If
    no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the
    padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in
    each row of the batch).
    )Zcustom_introc                          e Zd Zdef fddZdd Zdd Zee									dd	e	e
j d
e	e
j de	e
j de	e de	e
j de	e
j de	e de	e de	e defddZ  ZS )HeliumForSequenceClassificationrC   c                    s@   t  | |j| _t|| _tj|j| jdd| _| 	  d S r   )
r#   r$   
num_labelsr   r   r%   r`   r+   scorer   re   r-   r/   r0   r$     s
   
z(HeliumForSequenceClassification.__init__c                 C   r   r"   r   r<   r/   r/   r0   r     r   z4HeliumForSequenceClassification.get_input_embeddingsc                 C   r   r"   r   r   r/   r/   r0   r     r   z4HeliumForSequenceClassification.set_input_embeddingsNr   rs   r\   r   r   r   r   r   r   rg   c
              
   C   s(  | j ||||||||	d}
|
j}| |}|dur|jd }n|jd }| jjdu r2|dkr2td| jjdu r;d}n1|dur`|| jjk|jt	j
}t	j|jd |jt	j
d}|| d}nd}t| jj d |t	j||jd	|f }d}|dur| j|||| jd
}t|||
j|
j|
jdS )  
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        rs   r\   r   r   r   r   r   Nr   r   z=Cannot handle batch sizes > 1 if no padding token is defined.r2   )rM   r3   z will not detect padding tokens in `inputs_embeds`. Results may be unexpected if using padding tokens in conjunction with `inputs_embeds.`r   )r   r   pooled_logitsrC   r   )r   r   r  r;   rC   r   r   r4   rM   r'   Zint32r   Zargmaxr   r   r.   r>   r   r   r   r8   r   )r*   r   rs   r\   r   r   r   r   r   r   Ztransformer_outputsr8   r   r   Zlast_non_pad_tokenZnon_pad_maskZtoken_indicesr  r   r/   r/   r0   r9     sL   


z'HeliumForSequenceClassification.forwardr   )r>   r?   r@   r   r$   r   r   r   r   r   r'   r   r   r   r   r   r   r9   rA   r/   r/   r-   r0   r    sH    		
r  c                       r   )HeliumForTokenClassificationrC   c                    s|   t  | |j| _t|| _t|dd d ur|j}nt|dd d ur'|j}nd}t	|| _
t|j|j| _|   d S )Nclassifier_dropouthidden_dropoutg?)r#   r$   r  r   r   r   r  r	  r%   ZDropoutru   r`   r+   r  r   )r*   rC   r  r-   r/   r0   r$   0  s   
z%HeliumForTokenClassification.__init__c                 C   r   r"   r   r<   r/   r/   r0   r   @  r   z1HeliumForTokenClassification.get_input_embeddingsc                 C   r   r"   r   r   r/   r/   r0   r   C  r   z1HeliumForTokenClassification.set_input_embeddingsNr   rs   r\   r   r   r   r   r   r   rg   c
              
   C   sd   | j ||||||||	d}
|
j}| |}| |}d}|dur(| ||| j}t|||
j|
jdS )r  r  N)r   r   r8   r   )	r   r   ru   r  r   rC   r   r8   r   )r*   r   rs   r\   r   r   r   r   r   r   r   Zsequence_outputr   r   r/   r/   r0   r9   F  s,   


z$HeliumForTokenClassification.forwardr   )r>   r?   r@   r   r$   r   r   r   r   r   r'   r   r   r   r   r   r   r9   rA   r/   r/   r-   r0   r  .  sH    	
r  )r   r   r   r  r  )rn   )Nr   )Fr   typingr   r   r   r   r'   Ztorch.nnr%   Zactivationsr   Zcache_utilsr   r	   Z
generationr
   Zmodeling_attn_mask_utilsr   Zmodeling_flash_attention_utilsr   Zmodeling_layersr   Zmodeling_outputsr   r   r   r   Zmodeling_rope_utilsr   r   Zmodeling_utilsr   r   Zprocessing_utilsr   utilsr   r   r   r   r   Zconfiguration_heliumr   Z!torch.nn.attention.flex_attentionr   Zintegrations.flex_attentionr   Z
get_loggerr>   r   Moduler    rB   r]   r   r   rm   rS   r   r   r   r   r   r   r   r   r   r  r  __all__r/   r/   r/   r0   <module>   sv   
"

!K5 tlVF