bcgm%rddlmZddlZddlZddlmZddlmZddlm Z ddl m Z dd l m Z Gd d eZy) )print_functionN) CatBoostError)get_eval_logger)make_dirs_if_not_exists)CaseEvaluationResult) FoldModelceZdZdZ dZedZdZdZdZ edZ dZ d Z d Z ed Zed Zedd Zy)FoldModelsHandlerzS Class that is responsible for learning models and computing their metrics modelsc tjjtjr$tj tjyy#t $r,}tjt|Yd}~yd}~wwxYwN) ospathexistsr _FoldModelsHandler__MODEL_DIRrmdirOSErrorrwarningstr)errs ]/var/www/html/bid-api/venv/lib/python3.12/site-packages/catboost/eval/_fold_models_handler.py_remove_model_dirz#FoldModelsHandler._remove_model_dirs^ 0ww~~/;;<*667= 0   % %c#h / / 0sAA B "BB c||_||_t|_|jD]}t|j|<||_||_||_d|_y)zy Args: :param remove_models: Set true if you want models to be removed after applying them. N)_cases_metricsdict _case_results _thread_count _eval_step_flag_remove_models_metric_descriptions)selfmetricscases thread_count eval_step remove_modelscases r__init__zFoldModelsHandler.__init__s]   !VKKD'+vD  t $ *##0 $(!c||_|jD]=}|j|}|jD]}t|||j||<?y)N)r()r#rrrr!)r$metric_descriptionsr* case_resultmetric_descriptions r_init_case_resultsz$FoldModelsHandler._init_case_results0sY$7!KKD,,T2K&*&?&?"2FtGYQUQ`Q`3b ./'@ r,c i}|jD]\}}t||<|D]} | j||j|j} ||j | |j |j| jk|j | jk7stdt||z|zD]o\} } tj| |j} |jD]7\}}||}t|D]\}}| |k7s ||j| !9q|jD]r\}}||}|j|}t||D]J\}}|j!}|j D]&}||j#||j%|(Lty)N)r(r'z(Error: metric names should be consistent)itemslistcreate_metrics_calcerr!r appendr#r1r.r enumerater _create_pooladdrzip eval_metrics_add get_result)r$r%grouped_by_case_models learn_folds skipped_folds rest_foldsmetric_calcersr* case_models case_model metric_calcerfile_num fold_filepoolcalcers model_nummodel case_resultscalcerscoresmetrics r_compute_metricsz"FoldModelsHandler._compute_metrics9s!7!=!=!? D+#'6N4 ) * @ @KO??NRN`N`!A!b t$++M:,,4++M,M,M,OP..-2S2S2UU'(RSS*"@$-[=-H:-U#V Hi$11)T=O=OPD%;%A%A%C!k(.(1+(>$Iu9, *..t4)?&D$W"8!=!=!? D+$T*G--d3L!$Wk!: ,,."77F (--eV5F5Fv5NO8";"@r,cddlm}ttj|j }d|j vrF|j d}t|r(t||k\rtdj|tjdjt||tj}||j }|j||j!|tjdjtj|z t#|||S) Nr)CatBoostignored_featureszError: input parameter contains feature indices wich are not available in pool: {} Check eval_feature set and ignored features optionszLearn model {} on fold #{})params)fnamez Operation was done in {} seconds)rRrr rnum_col get_paramslenmaxrformatrdebugrtimefit save_modelr ) rHr*fold_id model_pathrR feature_countrScur_timeinstances r _fit_modelzFoldModelsHandler._fit_model\s 1 = =>  !2 2#01CD #$-=)>-)O#%ZZ`Z`aqZrtt  < C CCIw WX99;4??#45 T*- B I I$))+X`J` abz733r,c ttji}|jD]}t ||<t |D] \}}tj ||j}||z}|jD]}tjjtjtj||} tjdjt||| |j!|||| } tj#djt||||j%| |S)a  Train models for each algorithm and learn dataset(folds). Than return them. Args: :param learn_files: Entities of FoldStorage for learning models. :return: Dictionary of models where the key is case and the value is models on learn folds z#For model {} on fold #{} path is {}zModel {} on fold #{} was fitted)rr rrr4r7r8r rrjoin_create_model_namerr\r[rreinfor6) r$ learn_files fold_id_biasr r*rF learn_filerHr`ra fold_models r _fit_modelszFoldModelsHandler._fit_modelsrs" 1 = =>KKD6F4L %.k$: Hj$11*d>P>PQD"X-G WW\\*;*G*G*;*N*NtU\*]_ !''(M(T(TUXY]U^`gis(tu!__T4*M !&&'H'O'OPSTXPY[b'cdt ##J/ $ %; r,c |j||}|j||}|}|D]\}} } t|dk(rg} |j||} | j D] \} }| |z } t|}t j dj|||z|j|j| || | t j dj|||z||z }|j|j|  |j||jS#|j|j| wxYw#|j|wxYw)a9 Run all processes to gain stats. It applies algorithms to fold files that gains from learning. It keeps stats inside models and models are stored in DataFrame. Columns are matched to the different algos and rows to the folds. Args: :param splitter: Splitter entity. :param fold_size: The size of fold. :param folds_count: Count of golds. :param fold_offset: The offset (count of folds that we want to skip). :return: return dict: keys metric to CaseEvaluationResult rz+Start metric computation for folds [{}, {})z7Computation of metrics for folds [{}, {}) is completed)create_fold_setsfold_groups_files_generatorrYrnr3rrir[rPr clean_folds_remove_models_cleanr)r$splitter fold_size folds_count fold_offset folds_setsfold_groups_filesrkr?r@rA list_modelspermutation_modelsr*rClearn_folds_counts rproceedzFoldModelsHandler.proceeds "!229kJJ ( D DZEP!R &L:K6 ]J{#q( 5)-)9)9+|)T&-?-E-E-G)k#{2 .H),K(8%#%**+X,2F<PaAa,bd))$--*<*5}jR$%**+d,2F<PaAa,bd $55L((*'' 4+;L2 KK !!!!((*'' 4 KK !s%AE B6D:9#E :#EE E3c\|jr|j|jyr)r"rclean)r$rus rrtzFoldModelsHandler._cleans!  # #  " " $r,ctjdjt||D]}|j y)NzRemove models {})rr\r[rdelete)r{rKs rrsz FoldModelsHandler._remove_modelss7 2 9 9#k:J KL E LLN!r,cddl}t|jjdd}dj |j ||}|S)Nr-_z&model_{_name}_fold_{_fold}_{_uuid}.bin)_name_fold_uuid)uuidruuid1replacer[ get_label) model_casefoldrid_str model_names rrhz$FoldModelsHandler._create_model_namesRTZZ\"**34=DD:K_K_KaKOKQES r,c~ddlm}||j|j|j |}|S)Nr)Pool)column_description delimiterr')rVrrr get_separator)rGr'r data_pools rr8zFoldModelsHandler._create_pools;),5,H,H,J#,#:#:#<&24 r,N))__name__ __module__ __qualname____doc__r staticmethodrr+r1rPrernr~rtrsrhr8r,rr r s K00)$b!PF44*8/"b  r,r ) __future__rrr]rVreval.log_configr eval.utilsreval.evaluation_resultr _fold_modelr objectr rr,rrs+% -09"MMr,