U
    4Af$                     @   s   d dl Z d dlZd dlmZmZ d dlmZ d dlmZm	Z	m
Z
mZ d dlZd dlmZ d dlmZ ddlmZ dd	lmZ dd
lmZ ddlmZmZmZmZ eeZee  Z!e"dd e!D Z#eG dd dZ$G dd deZ%G dd deZ&dS )    N)	dataclassfield)Enum)DictListOptionalUnion)FileLock)Dataset   )$MODEL_FOR_QUESTION_ANSWERING_MAPPING)PreTrainedTokenizer)logging   )SquadFeaturesSquadV1ProcessorSquadV2Processor"squad_convert_examples_to_featuresc                 c   s   | ]}|j V  qd S N)
model_type).0conf r   D/tmp/pip-unpacked-wheel-zw5xktn0/transformers/data/datasets/squad.py	<genexpr>"   s     r   c                   @   s<  e Zd ZU dZedddde idZee	d< edddidZ
ee	d	< ed
ddidZee	d< ed
ddidZee	d< edddidZee	d< edddidZee	d< edddidZee	d< edddidZee	d< edddidZee	d< edddidZee	d< eddd idZee	d!< ed"dd#idZee	d$< dS )%SquadDataTrainingArgumentszb
    Arguments pertaining to what data we are going to input our model for training and eval.
    Nhelpz!Model type selected in the list: z, )defaultmetadatar   zFThe input data dir. Should contain the .json files for the SQuAD task.data_dir   zThe maximum total input sequence length after tokenization. Sequences longer than this will be truncated, sequences shorter will be padded.max_seq_lengthzVWhen splitting up a long document into chunks, how much stride to take between chunks.
doc_stride@   zkThe maximum number of tokens for the question. Questions longer than this will be truncated to this length.max_query_length   zThe maximum length of an answer that can be generated. This is needed because the start and end predictions are not conditioned on one another.max_answer_lengthFz1Overwrite the cached training and evaluation setsoverwrite_cachezDIf true, the SQuAD examples contain some that do not have an answer.version_2_with_negativeg        zIIf null_score - best_non_null is greater than the threshold predict null.null_score_diff_threshold   n_best_sizer   zjlanguage id of input for language-specific xlm models (see tokenization_xlm.PRETRAINED_INIT_CONFIGURATION)lang_id   z3multiple threads for converting example to featuresthreads)__name__
__module____qualname____doc__r   joinMODEL_TYPESr   str__annotations__r   r!   intr"   r$   r&   r'   boolr(   r)   floatr+   r,   r.   r   r   r   r   r   %   sl   
  			    	r   c                   @   s   e Zd ZdZdZdS )SplittraindevN)r/   r0   r1   r;   r<   r   r   r   r   r:   h   s   r:   c                	   @   s   e Zd ZU dZeed< ee ed< eed< e	ed< dej
dddfeeee eeef ee	 ee ee d	d
dZdd Zeeejf dddZdS )SquadDatasetzH
    This will be superseded by a framework-agnostic approach soon.
    argsfeaturesmodeis_language_sensitiveNFpt)r>   	tokenizerlimit_lengthr@   rA   	cache_dirdataset_formatc                 C   s  || _ || _|jrt nt | _t|trTzt| }W n t	k
rR   t	dY nX || _
|jrddnd}tj|d k	rz|n|jd|j d|jj d|j d| }	|	d }
t|
F tj|	rV|jsVt }t|	| _| jd | _| jdd | _| jd	d | _td
|	 dt |  | jd ksB| jd krt d|	 d n|tj!krt| j"|j| _n| j#|j| _t$| j||j|j%|j&|tj'k|j(|d\| _| _t }t)| j| j| jd|	 td|	 dt | dd W 5 Q R X d S )Nzmode is not a valid split nameZv2Zv1Zcached__z.lockr?   datasetexamplesz"Loading features from cached file z [took %.3f s]zDeleting cached file z; will allow dataset and examples to be cached in future run)rI   rC   r!   r"   r$   Zis_trainingr.   Zreturn_dataset)r?   rH   rI   z!Saving features into cached file z [took z.3fz s])*r>   rA   r(   r   r   	processor
isinstancer5   r:   KeyErrorr@   ospathr3   r   value	__class__r/   r!   r	   existsr'   timetorchloadZold_featuresr?   getrH   rI   loggerinfowarningr<   Zget_dev_examplesZget_train_examplesr   r"   r$   r;   r.   save)selfr>   rC   rD   r@   rA   rE   rF   Zversion_tagZcached_features_fileZ	lock_pathstartr   r   r   __init__w   sf    

"
 

zSquadDataset.__init__c                 C   s
   t | jS r   )lenr?   )rZ   r   r   r   __len__   s    zSquadDataset.__len__)returnc                 C   s6  | j | }tj|jtjd}tj|jtjd}tj|jtjd}tj|jtjd}tj|jtj	d}tj|j
tj	d}|||d}	| jjdkr|	d= | jjdkr|	||d | jjr|	d|i | jr|	dtj|jtjd| jj i | jtjkr2tj|jtjd}
tj|jtjd}|	|
|d	 |	S )
N)Zdtype)	input_idsattention_masktoken_type_ids)xlmZrobertaZ
distilbertZ	camembertrb   )Zxlnetrc   )	cls_indexp_maskis_impossibleZlangs)start_positionsend_positions)r?   rS   Ztensorr`   longra   rb   rd   re   r9   rf   r>   r   updater(   rA   ZonesshapeZint64r,   r@   r:   r;   Zstart_positionZend_position)rZ   iZfeaturer`   ra   rb   rd   re   rf   inputsrg   rh   r   r   r   __getitem__   s0    
$zSquadDataset.__getitem__)r/   r0   r1   r2   r   r6   r   r   r:   r8   r;   r   r   r7   r   r5   r\   r^   r   rS   ZTensorrn   r   r   r   r   r=   m   s*   

Kr=   )'rM   rR   Zdataclassesr   r   enumr   typingr   r   r   r   rS   Zfilelockr	   Ztorch.utils.datar
   Zmodels.auto.modeling_autor   Ztokenization_utilsr   utilsr   Zprocessors.squadr   r   r   r   Z
get_loggerr/   rV   listkeysZMODEL_CONFIG_CLASSEStupler4   r   r:   r=   r   r   r   r   <module>   s$   
B