sgQ$dZddlZddlmZddlZddlmcmZddlmZddl m Z ddlm Z m Z dd l mZd d lmZe j"eZGd d e ZdZe deGddeZy)z RetriBERT model N)Optional)nn)PreTrainedModel)add_start_docstringslogging) BertModel)RetriBertConfigc"eZdZdZeZdZdZdZy)RetriBertPreTrainedModelz An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. N retribertct|tjrm|jjj d|j j|j%|jjjyyt|tjrz|jjj d|j j|j2|jj|jjyyt|tjrJ|jjj|jjjdyy)zInitialize the weightsg)meanstdNg?) isinstancerLinearweightdatanormal_configinitializer_rangebiaszero_ Embedding padding_idx LayerNormfill_)selfmodules n/var/www/html/venv/lib/python3.12/site-packages/transformers/models/deprecated/retribert/modeling_retribert.py _init_weightsz&RetriBertPreTrainedModel._init_weights.s fbii ( MM   & &CT[[5R5R & S{{&   &&('  - MM   & &CT[[5R5R & S!!- ""6#5#56<<>.  - KK   " " $ MM   $ $S ).) __name__ __module__ __qualname____doc__r config_classload_tf_weightsbase_model_prefixr#r$r"rr$s #LO# *r$raC This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.) This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior. Parameters: config ([`RetriBertConfig`]): Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights. zEBert Based model to embed queries or document for document retrieval.c eZdZdeddffd Z ddZ ddZ ddZ ddejd e ejd ejd e ejd e dejf d Z xZS)RetriBertModelrreturnNct|||j|_t||_|j rdn t||_tj|j|_ tj|j|jd|_ tj|j|jd|_tjd|_|j#y)NF)rr) reduction)super__init__projection_dimr bert_queryshare_encodersbert_docrDropouthidden_dropout_probdropoutr hidden_size project_query project_docCrossEntropyLossce_loss post_init)r r __class__s r"r3zRetriBertModel.__init__Ss  $33#F+ & 5 59V;L zz&"<"<= YYv'9'96;P;PW\]99V%7%79N9NUZ[**V<  r$c|dks|jd|kr ||dS|j}|j}tj|tj |}dgj jzj||}fd} j|d|d} g} ttj|jd|z D]F} | | |z| dz|z} || |z| dz|z}tj| | |}| j|Htj| dS)Nr)attention_maskr )dtypedevicechj|d|d}|d}j|}|S)Nrr )rC head_mask)encoderpooler)inputsencoder_outputssequence_output pooled_outputrG sent_encoders r"partial_encodezCRetriBertModel.embed_sentences_checkpointed..partial_encodewsK"."6"61I#)!9'#7# #2!"4 , 3 3O D $$r$) input_ids position_idstoken_type_ids inputs_embeds)dim)shaperEsizetorchzeroslongrnum_hidden_layersget_extended_attention_mask embeddingsrangemathceil checkpointappendcat)r rPrCrNcheckpoint_batch_sizerE input_shaperRextended_attention_maskrOembedding_outputpooled_output_listbb_embedding_outputb_attention_maskrMrGs ` @r"embed_sentences_checkpointedz+RetriBertModel.embed_sentences_checkpointedbs` !1 $ (:=R(R .I!L L%%F#..*K"[[EJJvVN!4!4!F!FFI4@4\4\ 5 #  % ,66#$~ei 7  "$ 499Y__Q%7:O%OPQ 9%5a:O6OSTWXSX\qRq%r"#:1?T;TXY\]X]avWv#w * 5 5nFXZj k "))-8  9 99/Q7 7r$c`|j|||j|}|j|SN)rkr5r<)r rPrCrcq_repss r"embed_questionszRetriBertModel.embed_questionss7 22   OO !   !!&))r$c|j|||j |jn |j|}|j|Srm)rkr7r5r=)r rPrCrca_repss r" embed_answerszRetriBertModel.embed_answerssF 22  #}}4DOO$-- !   ''r$input_ids_queryattention_mask_query input_ids_docattention_mask_docrcc|j}|j|||}|j|||}tj||j } |j | tj| jdj|} |j | j tj| jdj|} | | zdz } | S)a Args: input_ids_query (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence tokens in the vocabulary for the queries in a batch. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) attention_mask_query (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. [What are attention masks?](../glossary#attention-mask) input_ids_doc (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence tokens in the vocabulary for the documents in a batch. attention_mask_doc (`torch.FloatTensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to avoid performing attention on documents padding token indices. checkpoint_batch_size (`int`, *optional*, defaults to `-1`): If greater than 0, uses gradient checkpointing to only compute sequence representation on `checkpoint_batch_size` examples at a time on the GPU. All query representations are still compared to all document representations in the batch. Return: `torch.FloatTensor``: The bidirectional cross-entropy loss obtained while trying to match each query to its corresponding document and each document to its corresponding query in the batch r r) rErorrrWmmtr?arangerUto) r rsrtrurvrcrErnrqcompare_scoresloss_qaloss_aqlosss r"forwardzRetriBertModel.forwardsJ!''%%o7KMbc##M3EG\]&&((*5,,~u||Nrs|  ++.3+4   H %**2"OD-D Dr$