o
i' @ s d dl Z d dlZd dlmZ ddlmZ ddlmZ ddlm Z ddl
m Z m Z ddl
mZmZ dd lmZ d
d lmZ e rId dlZd dlmZ d Zd
ZdZdZdZ dddZ G dd dZ dS ) N)Image )create_causal_mask)_get_model_class)
AutoConfig) MODEL_FOR_PRETRAINING_MAPPING
MODEL_MAPPING)PROCESSOR_MAPPING_NAMES
AutoProcessor)
AutoTokenizer )is_torch_availablez[92mz[93mz[0mu ■u ⬚![]()
c
st jdkrdddddf jdkr%ddddddf ttdd D }d}g }tD ]6\ } | krL|sL }d f< |dkrp| ksZ d krp d krd d7 d| | f< d}q:ddurfd d
tD d fd dtD }
|dur|dk} t| d
d |
| d
dt d|d |} tj | dt
t t dt t t d}
|d |
d |d dt|
d }dur|d7 }|| g }tD ]3\}||f dkr|dd
tt|ttD q|tt|tt qtttt| }|D ] }|durO|d d d | d d | nd q4tD ]\\ t|}v rpt | t n|}d fddtD }d}durd fddtD }|| dt d d | d| qXd|S )z
Generates an attention matrix from a given attention mask.
Optionally applies a sliding window mask (e.g., for Gemma2/3) and
marks regions where image tokens occur based on the specified `img_token`.
r N c s s | ] }t t|V qd S N)lenrepr).0word r i/sda-disk/www/egybert/egybert_env/lib/python3.10/site-packages/transformers/utils/attention_visualizer.py 8 s z6generate_attention_matrix_from_mask..r r c s$ g | ] fd dt D qS )c s0 g | ]}d | krk rn ndnd qS )r r r r j)isliding_windowr r
I s 0 zBgenerate_attention_matrix_from_mask...)range)r )nr )r r r I s $ z7generate_attention_matrix_from_mask.. c 3 sX | ]'} d |f rt t t n|d kr t t t n d |f r&tntV qdS r NYELLOW BLACK_SQUARERESETGREEN WHITE_SQUAREr )maskr r r K s
)
boundariesz: i == j (diagonal) z: token_type_idszAttention MatrixzSliding Window Maskc S s g | ]
}t | t qS r )r# r% )r kr r r r i s z | c 3 sl | ]1}| v r |f r v r t t t n |kr(t t t n |f r0tntV qd S r r" r )r img_tokenr( r wordsr r r x s
c 3 sx | ]7}| v r"v r"d f d |f kr"t t t n |kr.t t t n | r6tntV qdS r! r" r )r r. sliding_window_masktoken_type_bucketsr r/ r r r s ,
z:
)intndimr max enumerater jointorchwherecumsumboolarange bucketizer& r$ r% r# appendljustliststrrjustmapzipr )r/ r( r. r token_type_idsimage_seq_lengthmax_word_length
first_img_idxoutputr, row_dummy
is_specialr+ legendf_stringvertical_headeridxrow word_repr colored_word row_displaysliding_window_rowr ) r r. r( r r r0 r1 r r/ r #generate_attention_matrix_from_mask( s~
,"
$
.
rU c @ s: e Zd ZdefddZd defddZd defdd Zd
S ) AttentionMaskVisualizer
model_namec C s t |}d| _t| drt| dd | _zt|t}W n t y- t|t
}Y nw |d u r:t d| d|| _ G dd d|t
j}|||| _| j|j || _|| _d S )Nr r z Model name z- is not supported for attention visualizationc @ s e Zd Zdd ZdS )z7AttentionMaskVisualizer.__init__.._ModelWrapperc S s$ t j| t dd| _|| _d S )Nr )nnModule__init__Linear dummy_moduleconfig)selfr] rW r r r rZ s
z@AttentionMaskVisualizer.__init__.._ModelWrapper.__init__N)__name__
__module__ __qualname__rZ r r r r
_ModelWrapper s rb )r from_pretrained image_tokenhasattrget_text_configgetattrr r r Exceptionr
ValueError
mapped_clsrX rY modeltodtyperepo_idr] )r^ rW r] rj rb r r r rZ s"
z AttentionMaskVisualizer.__init__r- input_sentencec C s | j ||d d S )N)suffix)visualize_attention_mask)r^ ro rp r r r __call__ s z AttentionMaskVisualizer.__call__c C sV | j }i }d }| jjtv rnd}tttj |ddj
}d}t j | j
|d}t|dr0|j}n
|j|jgd }|rB|d|}||||d d
} |j|jgd | _| d }
d | v rc| d |d < |j| d
d } n!t | j
} | |} | |d dd }
|
d u rtd| jj dd|j_| |
j\}
}tj|
||jj f| j j d}t |}t |j||
|d d}|d ur| }
n|
!d!d"|
d||}
dt#d| jj d| j$ d }d}t%d| t%dd| jj d| j
d | j$j& 't#| d | t%| t(| |
| jt)| jd d | d |d }t%| t%| d S ) Nzchttps://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg?download=trueT)follow_redirectsr* )rF rd r r pt)imagestextrp return_tensorsattention_maskrE input_ids)rw z Model type z) does not support attention visualizationeager)rm )r]
inputs_embedsrx cache_positionpast_key_valuesr z##z Attention visualization for z | r r2 z" Attention visualization for [1m:z[0m z r )r. r rE rF )*rk r]
model_typer r openioBytesIOhttpxgetcontentr
rc rn re rd tokenizerconvert_ids_to_tokensimage_token_idreplacer tokenizeri _attn_implementationtrainshaper8 zeros hidden_sizerm r< r r; unsqueezeexpandr rj printr_ centerrU rg )r^ ro rp rk kwargsrF img processorrd inputsrx tokensr
batch_size
seq_lengthr{ r| causal_masktop_bottom_border side_borderrM r r r rq s
z0AttentionMaskVisualizer.visualize_attention_maskN)r- )r_ r` ra rA rZ rr rq r r r r rV s rV )r NNN) r r PILr
masking_utilsr models.auto.auto_factoryr models.auto.configuration_autor models.auto.modeling_autor r models.auto.processing_autor r
models.auto.tokenization_autor import_utilsr
r8 torch.nnrX r& r# r% r$ r' rU rV r r r r s*
l