o
    ei8                     @   sF  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
 ddl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 d
dlmZmZmZmZmZmZ eeZ G dd deZ!G dd deZ"G dd deZ#dd Z$G dd deZ%G dd deZ&G dd deZ'G dd deZ(G dd deZ)g d Z*dS )!    )CallableN)TransformersKwargs   )Cache)RopeParameters)ALL_ATTENTION_FUNCTIONS)Unpack)logging   )LlamaPreTrainedModelLlamaRMSNormeager_attention_forward)
OlmoConfig)OlmoAttentionOlmoDecoderLayerOlmoForCausalLM	OlmoModelOlmoRotaryEmbeddingapply_rotary_pos_embc                &       s"  e Zd ZdZdZddddddddZdgd	gfd
dgd
gfd
gd
gfdZ																		d/dedB dedB dedB dedB dedB d edB d!edB d"edB d#e	dB d$e
dB d%edB d&edB d'edB d(e
dB d)eeeef B dB d*e
dB d+e	dB d,edB f$ fd-d.Z  ZS )0Olmo2Configa  
    This is the configuration class to store the configuration of a [`Olmo2Model`]. It is used to instantiate an OLMo2
    model according to the specified arguments, defining the model architecture. Instantiating a configuration with the
    defaults will yield a similar configuration to that of the [allenai/Olmo2-7B-1124-hf](https://huggingface.co/allenai/Olmo2-7B-1124-hf).

    Configuration objects inherit from [`PreTrainedConfig`] and can be used to control the model outputs. Read the
    documentation from [`PreTrainedConfig`] for more information.


    Args:
        vocab_size (`int`, *optional*, defaults to 50304):
            Vocabulary size of the Olmo2 model. Defines the number of different tokens that can be represented by the
            `inputs_ids` passed when calling [`Olmo2Model`]
        hidden_size (`int`, *optional*, defaults to 4096):
            Dimension of the hidden representations.
        intermediate_size (`int`, *optional*, defaults to 11008):
            Dimension of the MLP representations.
        num_hidden_layers (`int`, *optional*, defaults to 32):
            Number of hidden layers in the Transformer decoder.
        num_attention_heads (`int`, *optional*, defaults to 32):
            Number of attention heads for each attention layer in the Transformer decoder.
        num_key_value_heads (`int`, *optional*):
            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
            `num_key_value_heads=1` the model will use Multi Query Attention (MQA) otherwise GQA is used. When
            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
            by meanpooling all the original heads within that group. For more details, check out [this
            paper](https://huggingface.co/papers/2305.13245). If it is not specified, will default to
            `num_attention_heads`.
        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
            The non-linear activation function (function or string) in the decoder.
        max_position_embeddings (`int`, *optional*, defaults to 2048):
            The maximum sequence length that this model might ever be used with.
        initializer_range (`float`, *optional*, defaults to 0.02):
            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
        use_cache (`bool`, *optional*, defaults to `True`):
            Whether or not the model should return the last key/values attentions (not used by all models). Only
            relevant if `config.is_decoder=True`.
        pad_token_id (`int`, *optional*, defaults to 1):
            Padding token id.
        bos_token_id (`int`, *optional*):
            Beginning of stream token id.
        eos_token_id (`int`, *optional*, defaults to 50279):
            End of stream token id.
        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
            Whether to tie weight embeddings
        rope_parameters (`RopeParameters`, *optional*):
            Dictionary containing the configuration parameters for the RoPE embeddings. The dictionary should contain
            a value for `rope_theta` and optionally parameters used for scaling in case you want to use RoPE
            with longer `max_position_embeddings`.
        attention_bias (`bool`, defaults to `False`, *optional*, defaults to `False`):
            Whether to use a bias in the query, key, value and output projection layers during self-attention.
        attention_dropout (`float`, *optional*, defaults to 0.0):
            The dropout ratio for the attention probabilities.
        rms_norm_eps (`float`, *optional*, defaults to 1e-05):
            The epsilon used by the rms normalization layers.

    ```python
    >>> from transformers import Olmo2Model, Olmo2Config

    >>> # Initializing a Olmo2 7B style configuration
    >>> configuration = Olmo2Config()

    >>> # Initializing a model from the Olmo2 7B style configuration
    >>> model = Olmo2Model(configuration)

    >>> # Accessing the model configuration
    >>> configuration = model.config
    ```
    olmo2colwise_gather_outputrowwise_split_inputcolwiserowwise)zlayers.*.self_attn.q_projzlayers.*.self_attn.k_projzlayers.*.self_attn.v_projzlayers.*.self_attn.o_projzlayers.*.mlp.gate_projzlayers.*.mlp.up_projzlayers.*.mlp.down_proj	input_idsinputs_embedshidden_statesattention_mask)embed_tokenslayersnorm      +      Nsilu   {Gz?T   g  F        h㈵>
vocab_sizehidden_sizeintermediate_sizenum_hidden_layersnum_attention_headsnum_key_value_heads
hidden_actmax_position_embeddingsinitializer_range	use_cachepad_token_idbos_token_ideos_token_idtie_word_embeddingsrope_parametersattention_biasattention_dropoutrms_norm_epsc                    s   t  jdi d|d|d|d|d|d|d|d|d	|	d
|
d|d|d|d|d|d|d|| || _| `d S )Nr-   r.   r/   r0   r1   r2   r3   r4   r5   r6   r7   r8   r9   r:   r;   r<   r=    )super__init__r>   clip_qkv)selfr-   r.   r/   r0   r1   r2   r3   r4   r5   r6   r7   r8   r9   r:   r;   r<   r=   r>   kwargs	__class__r?   e/home/ubuntu/transcripts/venv/lib/python3.10/site-packages/transformers/models/olmo2/modular_olmo2.pyrA      sL   	
zOlmo2Config.__init__)r"   r#   r$   r%   r%   Nr&   r'   r(   Tr)   Nr*   FNFr+   r,   )__name__
__module____qualname____doc__
model_typebase_model_tp_planbase_model_pp_planintstrfloatboolr   dictrA   __classcell__r?   r?   rE   rG   r   /   s    G


	
r   c                   @   s   e Zd Zdd ZdS )Olmo2RMSNormc                 C   sJ   |j }|tj}|djddd}|t|| j  }| j| |S )Nr
   T)keepdim)	dtypetotorchfloat32powmeanrsqrtvariance_epsilonweight)rC   r   input_dtypevariancer?   r?   rG   forward   s
   zOlmo2RMSNorm.forwardN)rH   rI   rJ   rc   r?   r?   r?   rG   rU      s    rU   c                   @      e Zd ZdS )Olmo2RotaryEmbeddingNrH   rI   rJ   r?   r?   r?   rG   re          re   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..NrV   r
   )dim)shaperZ   cat)xx1x2r?   r?   rG   rotate_half   s   rn   c                       s   e Zd ZddededB f fddZ		ddejdeejejf dejdB d	e	dB d
ej
dB dee deejejdB f fddZ  ZS )Olmo2AttentionNconfig	layer_idxc                    s@   t  j||d t|j| j |j| _t|j| j |j| _d S )Nrq   )	r@   rA   rU   r1   head_dimr>   q_normr2   k_normrC   rp   rq   rE   r?   rG   rA      s   zOlmo2Attention.__init__r   position_embeddingsr   past_key_valuescache_positionrD   returnc                 K   s0  |j d d }g |d| jR }| | |}	| | |}
| |}|	|dd}	|
|dd}
||dd}|\}}t	|	|
||\}	}
|d urc|||d}|
|
|| j|\}
}t| jjt}|| |	|
||f| jswdn| j| jd|\}}|jg |dR   }| |}||fS )NrV   r)   r
   )sincosry   r+   )dropoutscaling)ri   rs   rt   q_projru   k_projv_projview	transposer   updaterq   r   get_interfacerp   _attn_implementationr   trainingr=   r~   reshape
contiguouso_proj)rC   r   rw   r   rx   ry   rD   input_shapehidden_shapequery_states
key_statesvalue_statesr|   r{   cache_kwargsattention_interfaceattn_outputattn_weightsr?   r?   rG   rc      s>   	


zOlmo2Attention.forward)N)NN)rH   rI   rJ   r   rO   rA   rZ   Tensortupler   
LongTensorr   r   rc   rT   r?   r?   rE   rG   ro      s&    
ro   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 )Olmo2DecoderLayerrp   rq   c                    sJ   t  j||d t|j|jd| _t|j|jd| _t||d| _| `	d S )Nrr   eps)rp   rq   )
r@   rA   rU   r.   r>   post_attention_layernormpost_feedforward_layernormro   	self_attninput_layernormrv   rE   r?   rG   rA   	  s
   zOlmo2DecoderLayer.__init__NFr   r   position_idsrx   r6   ry   rw   rD   rz   c              
   K   s^   |}	| j d|||||||d|\}}
| |}|	| }|}	| |}| |}|	| }|S )N)r   r   r   rx   r6   ry   rw   r?   )r   r   mlpr   )rC   r   r   r   rx   r6   ry   rw   rD   residual_r?   r?   rG   rc     s&   




zOlmo2DecoderLayer.forward)NNNFNN)rH   rI   rJ   r   rO   rA   rZ   r   r   r   rR   r   r   r   rc   rT   r?   r?   rE   rG   r     s6    
	
r   c                   @   rd   )Olmo2PreTrainedModelNrf   r?   r?   r?   rG   r   1  rg   r   c                       s"   e Zd Zdef fddZ  ZS )
Olmo2Modelrp   c                    sB   t    t j jd| _t fddt j	D | _
d S )Nr   c                    s   g | ]}t  |qS r?   )r   ).0rq   rp   r?   rG   
<listcomp><  s    z'Olmo2Model.__init__.<locals>.<listcomp>)r@   rA   rU   r.   r>   r!   nn
ModuleListranger0   r    )rC   rp   rE   r   rG   rA   8  s
   
zOlmo2Model.__init__)rH   rI   rJ   r   rA   rT   r?   r?   rE   rG   r   7  s    r   c                   @   rd   )Olmo2ForCausalLMNrf   r?   r?   r?   rG   r   A  rg   r   )r   r   r   r   )+collections.abcr   rZ   torch.nnr   transformers.utils.genericr   cache_utilsr   modeling_rope_utilsr   modeling_utilsr   processing_utilsr   utilsr	   llama.modeling_llamar   r   r   olmo.configuration_olmor   olmo.modeling_olmor   r   r   r   r   r   
get_loggerrH   loggerr   rU   re   rn   ro   r   r   r   r   __all__r?   r?   r?   rG   <module>   s0    

 
	
9)
