o idW@sdZddlZddlmZddlmZddlmZddlZddl Z ddl m Z m Z ddl mZddlmZd d lmZd d lmZmZmZmZmZmZmZmZmZmZmZmZm Z m!Z!m"Z"d d l#m$Z$d d l%m&Z&m'Z'm(Z(m)Z)eeeeeeeee e!eieee"edZ*e)+e,Z-dZ.iddddddddddddddddd d!d"d#d$d%d&d'd(d)d*d+d,d-d.d/Z/gd0Z0d1d2Z1d3d4Z2d5d6Z3d7d8Z4d9d:Z5d;d<Z6eGd=d>d>Z7d?d@Z8dMdBdCZ9dDdEZ:dFdGZ;dHdIZdS)Nz'Configuration base class and utilities.N) dataclass)Path)Any)is_offline_mode model_info)OfflineModeIsEnabled)HFValidationError) __version__),MODEL_FOR_AUDIO_CLASSIFICATION_MAPPING_NAMES!MODEL_FOR_CAUSAL_LM_MAPPING_NAMESMODEL_FOR_CTC_MAPPING_NAMES,MODEL_FOR_IMAGE_CLASSIFICATION_MAPPING_NAMES*MODEL_FOR_IMAGE_SEGMENTATION_MAPPING_NAMES*MODEL_FOR_IMAGE_TEXT_TO_TEXT_MAPPING_NAMES!MODEL_FOR_MASKED_LM_MAPPING_NAMES(MODEL_FOR_OBJECT_DETECTION_MAPPING_NAMES*MODEL_FOR_QUESTION_ANSWERING_MAPPING_NAMES,MODEL_FOR_SEQ_TO_SEQ_CAUSAL_LM_MAPPING_NAMES/MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES(MODEL_FOR_SPEECH_SEQ_2_SEQ_MAPPING_NAMES0MODEL_FOR_TABLE_QUESTION_ANSWERING_MAPPING_NAMES,MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES6MODEL_FOR_ZERO_SHOT_IMAGE_CLASSIFICATION_MAPPING_NAMES) ParallelMode)is_datasets_availableis_tokenizers_availableis_torch_availablelogging)text-generationimage-classificationimage-segmentation fill-maskobject-detectionquestion-answeringtext2text-generationtext-classificationtable-question-answeringtoken-classificationaudio-classificationautomatic-speech-recognitionzzero-shot-image-classificationzimage-text-to-textz r"zMasked Language Modelingr zImage Classificationr!zImage Segmentationzmultiple-choicezMultiple Choicer#zObject Detectionr$zQuestion Answering summarization Summarizationr'zTable Question Answeringr&zText ClassificationrzCausal Language Modelingr%z&Sequence-to-sequence Language Modelingr(zToken Classification translation Translationzzero-shot-classificationzZero Shot Classificationr*zAutomatic Speech Recognitionr)zAudio Classification) accuracybleuf1matthews_correlationpearsonr precisionrecallrouge sacrebleu spearmanrwercCs |durgSt|tr|gS|SN) isinstancestr)objr>X/sda-disk/www/egybert/egybert_env/lib/python3.10/site-packages/transformers/modelcard.py_listifyss  r@cCsF|dur|St|tr|g}dd|D}t|dkr|S|||<|S)NcSsg|]}|dur|qSr:r>.0vr>r>r? sz*_insert_values_as_list..r)r;r<len)metadatanamevaluesr>r>r?_insert_values_as_list|s  rIcCs\|duriSi}|D]!}|ddtvr!|||dd<q |dkr+||d<q |S)N _rouge1r6)lowerreplace METRIC_TAGS) eval_resultsresultkeyr>r>r?#infer_metric_tags_from_eval_resultss rScCs|dur|S|||<|Sr:r>)rFrGvaluer>r>r? _insert_valuesrUcCs(tsdSddlm}m}t|||fS)NFr)DatasetIterableDataset)rdatasetsrVrWr;)datasetrVrWr>r>r? is_hf_datasetsrZcCs<g}|D]}t|ttfr|t|7}q||q|Sr:)rHr;tuplelistappend)mappingrQrCr>r>r?_get_mapping_valuess   r_c@steZdZUeed<dZeeeBdBed<dZedBed<dZeeeBdBed<dZ edBed<dZ eeeBdBed<dZ eeeBdBed<dZ eeeBdBed <dZ eeeBdBed <dZeeefdBed <dZeeefdBed <dZeedBed <dZeeefdBed<dZedBed<ddZddZddZddZe          dddZdS)TrainingSummary model_nameNlanguagelicensetagsfinetuned_fromtasksrY dataset_tags dataset_argsdataset_metadatarP eval_lineshyperparameterstrainersourcec Cs|jdur=ts?|jdurAt|jdkrCzt|j}|jD]}|dr+|dd|_qWdStjt t fy<YdSwdSdSdSdS)Nrzlicense:) rcrrerErrd startswithhttpx HTTPErrorrr)selfinfotagr>r>r? __post_init__s$     zTrainingSummary.__post_init__csd|ji}t|j}t|j}t|j}t|j}t|t|kr.|dgt|t|}tt||tt||}tt||}ddt|j D} g|d<t| dkr`tdkr`|gSt| dkrjddi} tdkrtddifdd| D} | D]s\} } i} | dur| | | d| d <| dur| | i}| | d|| d <|| dur|| | d d <t|dkrg| d <| D]\}}| d  |||j |d qd | vrd | vrd | vr|d | qtd| q|gS)NrGcSsi|] }|tvr|t|qSr>)TASK_TAG_TO_NAME_MAPPING)rBtaskr>r>r? sz6TrainingSummary.create_model_index..resultsrcsg|] }D]}||fqqSr>r>)rBtask_tagds_tagdataset_mappingr>r?rDsz6TrainingSummary.create_model_index..)rGtyperwrYargsmetrics)rGr~rTzLDropping the following result as it does not have all the necessary fields: )rar@rYrgrhrirEdictziprfgetitemsr]rPloggerrs)rrmetric_mapping model_index dataset_namesrgrhridataset_arg_mappingdataset_metadata_mapping task_mappingall_possibilitiesrzr{rQrF metric_tag metric_namer>r|r?create_model_indexs^            z"TrainingSummary.create_model_indexcCst|j}i}t|dd}t|d|j}t|d|j}|jdur4t|jtr4t |jdkr4t|d|j}t|d|j }t|d|j }t|d t | }|||d <|S) N library_name transformersrbrcr base_modelrdrXrz model-index)rSrPrUrIrbrcrer;r<rErdrgr\keysr)rrrrFr>r>r?create_metadatas  $zTrainingSummary.create_metadatacCsd}tj|dd}t|dkrd|d}|jdkr |t7}|d|jd7}|jdur3|d 7}n |d |jd |jd 7}|jdusRt |jt rWt|jdkrW|d 7}nEt |jt rg|d|jd7}n5t |jt t frt|jdkr|d|jdd7}n|d dd|jddDd|jdd7}|jdur|d7}|d dd|jD7}|d7}|d7}|d7}|d7}|d7}|d7}|jdur|d7}|d d d|jD7}|d7}n|d!7}|jdur|d"7}|t|j7}|d7}|d#7}|d$td7}|jdkr%tr%ddl}|d%|jd7}tr6ddl}|d&|jd7}trGddl}|d'|jd7}|S)(NF) sort_keysrz--- rlz # z z'This model was trained from scratch on z'This model is a fine-tuned version of [z](https://huggingface.co/z) on zan unknown dataset.the z dataset.r z, cSsg|]}d|qS)rr>)rBdsr>r>r?rD<sz1TrainingSummary.to_model_card..z and the z datasets.z: It achieves the following results on the evaluation set:  cSs$g|]\}}d|dt|qSz- z:  _maybe_roundrBrGrTr>r>r?rDAs$z/ ## Model description More information needed z9 ## Intended uses & limitations More information needed z: ## Training and evaluation data More information needed z ## Training procedure z ### Training hyperparameters z: The following hyperparameters were used during training: cSs g|] \}}d|d|qSrr>rr>r>r?rDLs z More information needed z ### Training results z ### Framework versions z- Transformers z - Pytorch z - Datasets z - Tokenizers )yamldumprrErmAUTOGENERATED_TRAINER_COMMENTrarerYr;r\r<r[joinrPrrkrjmake_markdown_tabler rtorchrrXr tokenizers)rr model_cardrFrrXrr>r>r? to_model_cardsn     $  .    zTrainingSummary.to_model_cardc Cs|jdur|jn|j} t| r?|dus| dus| dur?| j} | dvr?| dur0| jt| jdg} |dur7| g}| dur?| jg} | durI|durI|} |durbt|jj drbt j |jj j sb|jj j }|dur||jjj}tD] \}}|t|vr{|}qo|durt|jjj}t|dkr|}|durdg}nt|tr|dkr|dg}n d|vr|dt|jj\}}}t|}|||||||| || | |||d S)N)csvjsonpandasparquettext)configsplit _name_or_pathrgenerated_from_trainer) rbrcrdrarerfrYrgrhrirPrjrk) eval_dataset train_datasetrZ builder_name config_namer<rhasattrmodelrospathisdirr __class____name__ TASK_MAPPINGrr_rr output_dirrGrEr;r]parse_log_historystate log_history$extract_hyperparameters_from_trainer)clsrlrbrcrdrarerfrgrirYrh one_dataset default_tagmodel_class_namerwr^rKrjrPrkr>r>r? from_trainerhsf        zTrainingSummary.from_trainer) NNNNNNNNNN)r __module__ __qualname__r<__annotations__rbr\rcrdrerfrYrgrhrirrrPfloatrjrkrmrurrr classmethodrr>r>r>r?r`s> ?Jr`cCsBd}|t|krd||vr|d7}|t|krd||vs|t|krM|d8}|dkr@d||vr@|d8}|dkr@d||vs2|dkrKdd||fSdS||}g}d}t|D]r}d||vrg||d}d||vr||}|d d}|d d}|d d} |d d}|d d}|dd}||| d} |D]$\} } | dkr| | d<q| d} ddd| ddD}| | |<q|| qYt|d}|dkrd||vr|d8}|dkrd||vs|dkri}||D]\}}|d}|dvrddd|dD}|||<q|||fS||dfS)zd Parse the `log_history` of a Trainer to get the intermediate and final evaluation results. r train_runtimer eval_lossN)NNNzNo logloss total_flosepochstep eval_runtimeeval_samples_per_secondeval_steps_per_second)z Training LossEpochStepzValidation LossrKrJcSg|]}|qSr> capitalizerBpartr>r>r?rDz%parse_log_history..eval_)runtimesamples_per_secondsteps_per_secondrrcSrr>rrr>r>r?rDr) rErangecopypoprrrr] removeprefix)ridx train_loglines training_lossirrKrrrHkrCsplitsrGrPrRrTcamel_cased_keyr>r>r?rsb                       rcCsRt|tr%tt|ddkr%tt|dd|kr%|d|dSt|S)N.r f)r;rrEr<r)rCdecimalsr>r>r?rs:rcCs"ddt||D}d|dS)NcSs.g|]\}}d|d|t|dqS)z| rJr )rE)rBrCwr>r>r?rDs.z'_regular_table_line..r| )rr)rH col_widthsvalues_with_spacer>r>r?_regular_table_linesrcCsdd|D}d|dS)NcSsg|] }dd|dqS)z|:-:r>)rBrr>r>r?rDsz&_second_table_line..rr)r)rrHr>r>r?_second_table_linesrcCs|dus t|dkr dSdd|dD}|D]}|D]\}}||tt|kr3tt|||<qqtt|dt|}|tt|7}|D]}|tdd|Dt|7}qP|S)zC Create a nice Markdown table from the results in `lines`. NrrcSsi|] }|tt|qSr>)rEr<)rBrRr>r>r?rx z'make_markdown_table..cSsg|]}t|qSr>rrAr>r>r?rDrz'make_markdown_table..)rErrrr\rrHr)rrlinerRrTtabler>r>r?rs&r) learning_ratetrain_batch_sizeeval_batch_sizeseedc sfddtD}jjtjtjfvr#jjtjkrdnjjj|d<jjdkr/jj|d<jj dkr;jj |d<jj jjjj }||dkrQ||d <jj jj}||d krc||d <jj rjj }jj rsjj nd }d |vrd|djjdjjdjjd| |d<n d|d||d<jjj|d<jjdkrjj|d<jjdkrjj|d<njj|d<jjrd|d<jjdkrوjj|d<|S)Ncsi|] }|tj|qSr>)getattrr)rBrrlr>r?rxrz8extract_hyperparameters_from_trainer..z multi-GPUdistributed_typer num_devicesgradient_accumulation_stepsrtotal_train_batch_sizertotal_eval_batch_sizez!No additional optimizer argumentsadamzUse z with betas=(,z) and epsilon=z and optimizer_args= optimizerz and the args are: lr_scheduler_typeglr_scheduler_warmup_stepsrtraining_steps num_epochsz Native AMPmixed_precision_traininglabel_smoothing_factor)_TRAINING_ARGS_KEYSr parallel_moder NOT_PARALLELNOT_DISTRIBUTED DISTRIBUTEDrT world_sizerrroptim optim_argsrM adam_beta1 adam_beta2 adam_epsilonr  warmup_steps max_stepsnum_train_epochsfp16r)rlrkrroptimizer_nameoptimizer_argsr>rr?rsJ             r)r)?__doc__r dataclassesrpathlibrtypingrrprhuggingface_hubrrhuggingface_hub.errorsrhuggingface_hub.utilsrrr models.auto.modeling_autor r r rrrrrrrrrrrr training_argsrutilsrrrrr get_loggerrrrrvrOr@rIrSrUrZr_r`rrrrrrrr>r>r>r?s       D               <