u:i0SrSSKJrJr SSKJrJrJrJrJ r J r J r J r J r SSKrSSKrSSKJr SSKJr SSKJr SS KJrJrJr SS KJr S S KJr S \R@S\RB4Sjr"\"SS5r#\#"SSSSSS5r$S\ \ \RBS\ \RB4Sjr%S\\RLS\\RL\'\(/S4SS4Sjr)"SS\5r*S\RLS\4S jr+S!\\'\ \RB4S"\ \,S#\\'\4S$\ \S%\\'\4S\4 S&jr-S\\RLS'\ \ \'S"\ \,S(\(S)\\'\4S*\(S+\(S\ \\ \44S,jr.S1S-\S!\S.\ \S/\(S\RB4 S0jjr/g)2z*Utilities for processing spark partitions.) defaultdict namedtuple) AnyCallableDictIteratorListOptionalSequenceTupleUnionN) csr_matrix) ArrayLikeconcat)DataIterDMatrixQuantileDMatrix)XGBModel) get_loggerseriesreturncPURSS9n[R"U5nU$)zStack a series of arrays.F)copy)to_numpynpstack)rarrays U/srv/projetos/modelo_ml_acdoc/venv/lib/python3.13/site-packages/xgboost/spark/data.py stack_seriesr"s$ OOO 'E HHUOE LAlias)datalabelweightmarginvalidqidvaluesr&r' baseMarginvalidationIndicatorr*seqc(U(a [U5$g)z&Concatenate the data if it's not None.Nr)r.s r!concat_or_noner0s c{ r#iteratorappendc^S[RS[SS4U4SjjnSnUHnUc[RUR ;nUSLa [RUR ;deU(aJUR U[R)SS24nUR U[RSS24nOUSpeU"US5 UcMU"US5 M g)zjExtract partitions from pyspark iterator. `append` is a user defined function for accepting new partition.partis_validrNc>T"U[RU5 T"U[RU5 T"U[RU5 T"U[RU5 T"U[R U5 g)N)aliasr%r&r'r(r*)r4r5r2s r! make_blob#cache_partitions..make_blob+sStUZZ*tU[[(+tU\\8,tU\\8,tUYY)r#TF)pd DataFrameboolr7r)columnsloc)r1r2r8has_validationr4trainr)s ` r!cache_partitionsrA%s * ***&*N  !"[[DLL8N T !;;$,,. .. HHd5;;//23EHHT%++.12E5%   eT "r#c^\rSrSrSrS\\\4S\\ S\ SS4U4Sjjr S\\ \ RS\\ R4S jrS \S\4S jrSS jrS rU=r$)PartIterEz7Iterator for creating Quantile DMatrix from partitions.r% device_idkwargsrNc R>SUlX lXlX0l[TU]SS9 g)NrT) release_data)_iter _device_id_data_kwargssuper__init__)selfr%rErF __class__s r!rNPartIter.__init__Hs+ #   d+r#cU(dgURbTSSKnSSKnURRR UR5 UR XR5$XR$Nr)rJcudfcupycudaruntime setDevicer;rI)rOr%rTcps r!_fetchPartIter._fetchRsW ?? &   GGOO % %doo 6>>$zz"23 3JJr# input_datac UR[UR[R5:XagU"SUR UR[R5UR URR [RS55UR URR [RS55UR URR [RS55UR URR [RS55S.URD6 U=RS- slg)NFr%r&r' base_marginr*rT) rIlenrKr7r%rZgetr&r'r(r*rL)rOr\s r!next PartIter.nextas ::TZZ 34 4 TZZ 34++djjnnU[[$?@;;tzz~~ellDAB DJJNN5<<$FG DJJNN599d;<  ll   a r#cSUlgrS)rI)rOs r!resetPartIter.resetos  r#)rKrJrIrL)rN)__name__ __module__ __qualname____firstlineno____doc__rstrr r intrrNr r:r;rZrr<rcrf__static_attributes__ __classcell__)rPs@r!rCrCEs~A,dO,08 ,IL, ,  8HR\\$:;  @V   x D r#rCr4cr/S//p2nSn[URURURUR5HupVpxUS:Xa[ U5n Un Un O0[ U5n [R"U [RS9n Un US:XaU nXI:XdeURU 5 URUS[ U 5-5 URU 5 M [R"U5n [R"U5n [R"U5n[XU 4[ U5U4S9$)Nr)dtype)shape)zipfeatureVectorTypefeatureVectorSizefeatureVectorIndicesfeatureVectorValuesrnrararangeint32r2r concatenater)r4csr_indices_listcsr_indptr_listcsr_values_list n_featuresvec_type vec_size_ vec_indices vec_valuesvec_size csr_indices csr_valuescsr_indptr_arrcsr_indices_arrcsr_values_arrs r!)_read_csr_matrix_from_unwrapped_spark_vecrss+9;aS"J8;   !!   94[ q=9~H%K#J :H))HBHH=K#J ?!J%%% ,r2S5EEFz*598XXo.Nnn%56O^^O4N  .9#d)ZAX r#r% dev_ordinalmetarefparamscU(d[[R"S5US9$[X40UD6n[U40UDSU0D6nU$)z+Handle empty partition for QuantileDMatrix.rr)rr)rremptyrC)r%rrrritms r!make_qdmrsE rxx/S99 $ ,t ,B.f.#.A Hr# feature_colsuse_qdmrFenable_sparse_data_optimhas_validation_colc ^^^^^[[5m[[5mSmS[RS[S[ SS4UUUU4SjjnS[RS[S[ SS4UUU4SjjnS [ [[[R4S [ [[4S[4S jn U(aUn S T;a TS S :XdeOUn S[[ [[4[ [[[[[ 4444U4Sjjn U "5upTb!U(a[!X 5 [#TX,SU 5nOWTbU(d[!X 5 U "TT5nO8Tc!U(a[!X 5 [#TX,SU 5nO[!X 5 U "TT5nU(a(U(a[#TX,X5nOU(a U "TT5OSnOSnUb$UR%5UR%5:XdeX4$)a.Create DMatrix from spark data partitions. Parameters ---------- iterator : Pyspark partition iterator. feature_cols: A sequence of feature names, used only when rapids plugin is enabled. dev_ordinal: Device ordinal, used when GPU is enabled. use_qdm : Whether QuantileDMatrix should be used instead of DMatrix. kwargs : Metainfo for DMatrix. enable_sparse_data_optim : Whether sparse data should be unwrapped has_validation: Whether there's validation data. Returns ------- Training DMatrix and an optional validation DMatrix. rr4namer5rNc>U[R:XdXR;aU[R:XaTbUTRSS:aUTnO;XRSS:a$XnU[R:Xa [ U5nOSnU[R:Xa-Ub*TS:XaURSmTURS:XdeUcgU(aTUR U5 gTUR U5 ggNrr)r7r%r=rtr"r2)r4rr5r rr train_data valid_datas r!append_m0create_dmatrix_from_partitions..append_ms 5:: !5 " ,&,,Q/!3.2<.@!!!$q( 5::%(/Euzz!e&7?!&QJ!U[[^333}4 ''.4 ''.5"6r#cJ>U[R:XdXR;aU[R:Xa6[U5nTS:XaURSmTURS:XdeOXnU(aTUR U5 gTUR U5 ggr)r7r%r=rrtr2)r4rr5r rrrs r!append_m_sparse7create_dmatrix_from_partitions..append_m_sparses 5:: !5uzz!A$G?!&QJ!U[[^333 4 ''.4 ''."6r#r+rFc 4[U5S:Xa;[S5RS5 [SS[R "S50UD6$[ U[R5n[ UR[RS55n[ UR[RS55n[ UR[RS55n[ UR[RS55n[SX#XEUS.UD6$)NrXGBoostPySparkz_Detected an empty partition in the training data. Consider to enable repartition_random_shuffler%rr^r`)rarwarningrrrr0r7r%rbr&r'r(r*)r+rFr%r&r'r(r*s r!make,create_dmatrix_from_partitions..makes v;!  ' ( 0 0.  ; 0;F; ;fUZZ01vzz%++t<= 5<< >? 5<< >?VZZ 489 63 RX  r#missinggcf>Sn0n0nTR5Hup4X0;aXAU'MXBU'M X!4$)N)max_binrsilentnthreadenable_categorical)items) non_data_keysnon_data_paramsrkvrFs r! split_params4create_dmatrix_from_partitions..split_paramssH  LLNDA!%&"Q # $$r#)rlistr:r;rmr<rr rndarrayrrr r rnfloatrArnum_col)r1rrrrFrrrrr append_fnrrrdtraindvalidrrrs ` ` @@@r!create_dmatrix_from_partitionsrsD/:$.?J.9$.?JJ/r||/3/$/4//</bll/#//$//" T#tBJJ//0 $sCx. W $ # F"vi'8C'???'? %%S#XS%UD@P:Q5Q0R RS%* >LDG-":{$O  !'-j&)  '-*kvF-j&) (0Kv)F2DT*f-F ~~6>>#3333 >r#modelr_ strict_shapec URS5n[UUURURURUR UR S9nUR5RUSSUUS9$)z4Predict contributions with data with the full model.N)r_rr feature_typesfeature_weightsrTF) pred_contribsvalidate_featuresiteration_ranger) _get_iteration_rangerrn_jobsrrr get_boosterpredict)rr%r_rr data_dmatrixs r!rrUs006O   ))-- 33L     & &'! ' r#)NF)0rl collectionsrrtypingrrrrr r r r r numpyrpandasr: scipy.sparser_typingrcompatrcorerrrsklearnrutilsrSeriesrr"r$r7r0r;rmr<rArCrrnrrrr`r#r!rs20/XXX#55rzz 7QR h<9NPUV"**!568BJJ;O#r||$#.6 c47PRV7V.W# #@+x+\(BLL(Z(V  sD$$ %  #   sCx.   '   cN    er||$e8C=)e# e  e cN e#ee 7HW% %&eV(,   )$  ZZ r#