SicSSKJr SSKrSSKJrJr SSKrSSKr\(aSSK J r "SS\ 5r "SS\ 5r "S S \5r"S S 5r"S S5rS Sjr\"5r"SS5r"SS5rS!SjrS"Sjr"SS5r"SS\R25r"SS\5r"SS\5r"SS\5rg)#) annotationsN) TYPE_CHECKINGAny)Sequencec\rSrSrSrSrg)RuntimeTypeErrorz/Raised when a type of a variable is unexpected.N__name__ __module__ __qualname____firstlineno____doc____static_attributes__r T/var/www/html/ai-image-ml/venv/lib/python3.13/site-packages/onnx/reference/op_run.pyrrs9rrc\rSrSrSrSrg)RuntimeContextErrorzfRaised when the context is missing but an context dependent implementation is defined for an operator.r Nr r rrrrsprrc\rSrSrSrSrg)RuntimeImplementationErrorz8Raised when no implementation was found for an operator.r Nr r rrrrsBrrc\rSrSrSrSrg) DefaultNonezjDefault value for parameters when the parameter is not set but the operator has a default behavior for it.r Nr r rrrrstrrc,\rSrSrSrSSjrSSjrSrg) RefAttrName!zqImplements a link between a parameter of a function and an attribute in node. Args: name: name of the input cXlgNname)selfr#s r__init__RefAttrName.__init__(s rcPURRSUR<S3$)N()) __class__r r#r$s r__repr__RefAttrName.__repr__+s$..))*!DII=::rr"N)r#strreturnr.)r r rrrr%r,rr rrrr!s ;rrc @0n[RRRR5HnURU;aURXRR:waB[ SUR<SUR<SXRR<S35eUR XRR :aXUR'OXUR'XURS-[UR 5-'M U$)Nz7This function assumes every operator has a unique name z even across multiple domains z and ._)onnxdefsget_all_schemas_with_historyr#domainNotImplementedError since_versionr.)resschemas r_build_schemasr</s)+C))..%%BBD ;;# }}KK 0 7 77)Mfkk_]44:MM3DE#kkJZJaJaIddeg##c++&6&D&DD#)FKK % =CFKK# F$8$8 9 9:E Jrc(\rSrSrSSjrSSjrSrg)OnnxTypeEc~[U[R5(d[S[ U5S35eXlg)Nz type_proto z must be of type onnx.TypeProto.) isinstancer4 TypeProto TypeErrortype type_proto)r$rEs rr%OnnxType.__init__Fs:*dnn55d:.//OP %rc$SUR<S3$)Nz OnnxType(r)rEr+s rr,OnnxType.__repr__Ms4??-Q//rrHN)rEzonnx.TypeProtor/)r r rrr%r,rr rrr>r>Es %0rr>cF\rSrSrSrSSjr\SSj5rSrg) SparseTensorQz[Simple representation of a sparse tensor. It is based on numpy but does not require scipy. c(XlX lX0lgr!)valuesindicesshape)r$rNrOrPs rr%SparseTensor.__init__Vs   rc.URR$r!)rNdtyper+s rrSSparseTensor.dtype]s{{   r)rOrPrNN)rN np.ndarrayrOrUrPz tuple[int]r0None)r0r) r r rrrr%propertyrSrr rrrKrKQs@ +5>H !!rrKc[SUR55n[[RR UR 5[RR UR5U5$)zHosts a sparse tensor.c3$# UHov M g7fr!r ).0ds r #to_sparse_tensor..ds&X!Xs)tupledimsrKr4 numpy_helperto_arrayrNrO)attrPs rto_sparse_tensorrcbsV &SXX& &E  ""3::. ""3;;/  rc[RRS[RRS[RRS[RR S[RR S[RRS[RRS[RRS[RRS [RRS [RRS [RRS [RRS [RRS0U$)NcB[R"UR5$r!)npfloat32frbs r0_attribute_conversion_function..nsrzz#%%/@rcnURVs/sHn[R"U5PM sn$s snfr!)floatsrfrg)rbrhs rrjrkos" 0S 1A 0S0Ss 2c,[UR5$r!)Graphgris rrjrkps uSUU|rcXURVs/sHn[U5PM sn$s snfr!)graphsro)rbrps rrjrkqs3::0N:aq:0N0N'c,[UR5$r!)intiris rrjrkrs SZrcXURVs/sHn[U5PM sn$s snfr!)intsru)rbrvs rrjrksssxx.Hx!s1vx.H.Hrsc,[UR5$r!)rc sparse_tensorris rrjrkts7G   8 rcXURVs/sHn[U5PM sn$s snfr!)sparse_tensorsrcrbts rrjrkws+),););9 );A Q );9 9 rsc8URRS5$Nzutf-8)sdecoderis rrjrkzs W0ErcbURVs/sHoRS5PM sn$s snfr)stringsr)rbrs rrjrk{s)'*{{2 '2!HHW {2 2 s,cT[RRUR5$r!)r4r`rar~ris rrjrk~s0A0A0J0J3550QrcURVs/sH"n[RRU5PM$ sn$s snfr!)tensorsr4r`rar}s rrjrks336;;2 3>aD   & &q );2 2 s);c,[UR5$r!)r>tpris rrjrks HSVV4DrcXURVs/sHn[U5PM sn$s snfr!) type_protosr>r}s rrjrks&!$6 !0AHQK6 6 rs)r4AttributeProtoFLOATFLOATSGRAPHGRAPHSINTINTS SPARSE_TENSORSPARSE_TENSORSSTRINGSTRINGSTENSORTENSORS TYPE_PROTO TYPE_PROTOS) attr_types r_attribute_conversion_functionrls$ !!#@ ""$S !!#; ""$N !7   "H )),  **-  ""$E ##&  ""$Q ##&  &&(D ''* - 23 rc"\rSrSrSrSSjrSrg)rorpcXlgr!r)r$rps rr%Graph.__init__srN)rponnx.GraphProtor0rV)r r rr __slots__r%rr rrroros Irroc\rSrSrSrSrSSSjjrSrSSSjjr\ S5r SS jr \ S S j5r \ S!S j5r\ S!S j5r\ S"S j5r\ S"Sj5rS#SjrS"Sjr\R*S5rS$SjrSSS.Sjr\S5r\S%S&Sjj5r\S'S(Sjj5r\SSS.S)Sjj5rSrg)*OpRunaAncestor to all operators in this subfolder. Args: onnx_node: `onnx` node run_params: additional parameters such as `verbose`, `opsets` (it can be more than one if the operator has a subgraph), `log` for a logging function schema: operator schema Nc [U[5(d[S[U5S35eSH$nXB;dM [ SU<S[ U5S35e SU;a [ S5eXlX lUc[URS 5(aURRUl O|URR[;a#[URRUl O;UR[;a[URUl OSUl OX0l S UlUR#5 g) Nz$run_params must be a dictionary not r2)opsetsnew_ops Attribute z must be in run_params, only z was found.logz#run_params must contains key 'log'. op_schemaF)rAdictrCrD RuntimeErrorsortedKeyError onnx_node run_paramshasattrr*r_schemar _schemasop_type has_subgraph_load_attributes)r$rrr;rbs rr%OpRun.__init__s*d++B4 CSBTTUVW W(C$" 'Dj)*+7)  "@A A"$ >t~~{33#~~77 ((H4'(?(?@ ""h.' (9(9: # !L! rc0URS"U/UQ76 g)Nr)r)r$patternargss r_log OpRun._logs w..rcUR[RR:XaURR SS5nSUR;a'[ URSR55nOSnURR SS5nUc SU35eU"URURS[SURR SS5S - 5UcSUS 9$[ UR55US 9$[UR5nUbU"U5$Uc[[S UR<S UR<S URR<SURR<SUS3 5e[SUR<S UR<S URR<SURR<SUSUS3 5e)z0Converts an attribute value into a python value.rNexisting_functions evaluator_clsz0evaluator_cls must be specified to evaluate att=rrverbose)rrr functionszUnable to convert attribute z type z from node type z , domain  r2z$Unable to convert default value for )rDr4rrrgetlistrNrpmaxrAttributeErrorr#rrr7)r$rbref_attrrrconversion_functions r_extract_attribute_valueOpRun._extract_attribute_values 88t**00 0oo)))T:G#t6 1E!F!M!M!OP   OO//FM , B3%H ,!x0At229a@1DE '#  6:'..:J5K#  =SXXF  *&s+ + ? .sxxl& M""&.."8"8!;<..//2"SE<  27<<2B& U"nn4478nn++.bRy C  rc6URSU=(d 0US9$)N attributes)runcontextvaluers r_evaluate_subgraphOpRun._evaluate_subgraphsyyw}"yDDrcSUl/nURRHnURnUR(a[ UR5nSUlOUR U5n[XU5 URU5 UR[RR:XdMSUl U=RUR-sl[USUR3US4Sj5 M UR(Ga;URRS;Ga URR R#5HupV[%X5(aM['USS5(a([)SU<S URR<S 35e[%US 5(dMdUR*RS :Xd>UR*RS :Xa'UR*R,R.S :XaSnOUR UR*U5n[XU5 URU5 M [1U5Ulg)zChecks and loads attributes.FT_run_Nc.[RXU5$r!)rrrs rrj(OpRun._load_attributes..s%*%=%= &r>Constantrequiredrz3 is expected based on ONNX specifications for node r2 default_valuer)has_linked_attributer attributer# ref_attr_namerrsetattrappendrDr4rrrrrritemsrgetattrrrr~ data_typesetattributes_names_)r$added_attributesrbr#rkvs rrOpRun._load_attributess$)!>>++C88D  #C$5$56,0)55c: D &  # #D )xx4..444$(!))U-G-GG)CHH:&#,, <<(>'ADq/22??//14OO00A5 ! 1 1 ; ;q @%)E$($A$A!//ST$UE/(//2#8$"%%5!6rc[U[R5(d[S[ U5<S35e[ 5n[ 5nUR HnURUR5 M URHnURUR5 M URHnURUR5 M URHRnURHnURU5 M URHnX;dM URU5 M MT [U5$)zReturns all variables not registered as inputs and not produced by an node inside the graph. This inputs are part of the context existing in the graph calling this one. zUnexpected type r2)rAr4 GraphProtorCrDr initializeraddr#sparse_initializerinputnodeoutputr) graphlocalknowninit sparse_initinprorvs rimplicit_inputsOpRun.implicit_inputss %11.tE{oQ?@ @%%D IIdii & 33K IIk&& '4;;C IIchh JJD[[ ! !ZZ>IIaL  E{rc.URR$)zReturns node attribute `input`.)rrr+s rr OpRun.input/s~~###rc.URR$)z Returns node attribute `output`.)rrr+s rr OpRun.output4~~$$$rc.URR$)z!Returns node attribute `op_type`.)rrr+s rr OpRun.op_type9s~~%%%rc.URR$)z Returns node attribute `domain`.)rr7r+s rr7 OpRun.domain>rrcg)zTells the runtime if this node needs the context (all the results produced so far) as it may silently access one of them (operator Scan, If, Loop). The default answer is `False`. Fr r+s r need_contextOpRun.need_contextCs rcURRS-SURR3/n[ UR R 55HFup#US;aM SUSs=::aS::dMO M#USS:wdM.URS US US 35 MH URS 5 S RU5$)Nr(z op_type=>descrarzr3z =,r)r) r*r rrr__dict__rrjoin)r$attsrrs r__str__ OpRun.__str__Ks''#-dnn>T>T=U/VW4==..01DA))ad!c!!aesl d1#QqcO, 2 CyyrcJ[SURR<S35e)a@Should be overwritten. Args: *args: operator inputs **kwargs: optional inputs and overridden attributes, an attribute may be overridden if it belongs to a function, in this case, the same instance of OpRun can be called with different values of the same attribute. Returns: outputs z/Method '_run' must be overwritten for operator r2)r8r*r r$rkwargss r_run OpRun._runUs)"=dnn>U>U=XXY Z  rc[U[5(d0[SURR<S[ U5S35eU(d$[ SURR<S35e[SU55(aCUVs/sHn[ U5PM nn[SURR<SU<S35e[S U55n[S U55(aCUVs/sHn[ U5PM nn[SURR<S U<S35eU$s snfs snf) z-Checks the output are from the expected type.zMethod '_run' of class z does not return a tuple but 'z'.z does not return any result.c3B# UHn[U[5v M g7fr!)rAr^rZr~s rr\/OpRun._check_and_fix_outputs..qs1Sz!U##Ssz6One of the results returned by method '_run' of class zg is a tuple, this is no ONNX corresponding type (Map, List, Tensor, SparseTensor). All returned types: r2c3# UH8n[R"U5(a[R"U5OUv M: g7fr!)rfisscalararray)rZxs rr\rxs, )linked_attributesrc >UR5(aUc[S[U5S35eOUb[S[U5S35eUR(aUc[ S[U5S35eUR(d'Ub$[ S[U5S[ U5S35e0nUR(aUc[ S [U5S35eURHen[X5n[U[5(dM%URU;a [ S U<S U<S [U5S35eXRXE'Mg URS URR[U55 0nURHOnXT;aM [!X5(d([#SU<SURR<S35e[X5Xu'MQ UR$(a4UR(aU(d[S[U5S35eXS'UbX'S'U(aUR&"U0UDUD6nOUR&"U0UD6nURSURRUb [U5OS5 UR-U5$![([ 4aen [)SUV s/sHn [U 5PM Os sn fsn S[+U5S[+U5SURR<S3 5U eSn A ff=f)anCalls method ``_run``, catches exceptions, displays a longer error message. Args: *args: inputs linked_attributes: used if this has an attribute linked to the attribute of the function it belongs to context: if this node is part of the subgraph, `context` is a dictionary with the values this node may use Returns: tuple of results NzThis node if type z needs context to be filled.z+ does not need any contextbut one is given.z This node zK has linked attributes but None are given in parameter 'linked_attributes'.zM has no linked attribute but some are given in parameter 'linked_attributes' r2zBOne attribute is linked but no linked value is provided, in class z,Unable to find a value for linked attribute z in z in node z-- begin %s.run(%d inputs)rz is missing in operator z6A subgraph has linked attribute but none was given to rrzIssues with types z and attributes z and linked attributes=z (operator ).z-- done %s.run -> %d outputsr)r rrDrr'rrrrrArr#rr*r lenr NameErrorrrrCrr*) r$r,rroverridden_attributesrbrrr:er3s rr OpRun.runsG     "(d 4PQ $T$ZL0[\   $ $):)BT$ZL(st ((->-JT$ZL)''*+<'=&>aA !#  $ $ ($ $T |1.--D&a--vv%66(J3'QUVgUjk''+Dzl!52C661J).. .0G0GTS))C+4%% '?@W@W?ZZ[\"$,FK*   ((1B"LTRVZLXYZ$5<   '9  $iiI)>I&Iii00 * NN # #CHQ **3//>* $t%""9&AV:W9XY!^^447r;   s+2J'J''L7 LK ALLc URnSU;a U[RR54$UR SS5up[ U5nX4$![ a- UR[RR54s$f=f)Nr3)r r4r5onnx_opset_versionrsplitrur')clsr#versi_verss r infer_nameOpRun.infer_names|| d?5577 7[[a(  @YF| @<<!=!=!?? ? @s A4B  B c UR5upEURnSnUc/Uc [RR XEU5nUR nUc/Uc [RR XEU5nUR n[U5Vs/sHnSU3PM n n[U5Vs/sHnSU3PM n n[RR"XIU 40UD6$s snfs snf)atCreates an ONNX node for this class based on the given information. Args: n_inputs: number of inputs (default is defined by the operator schema) n_outputs: number of outputs (default is defined by the operator schema) verbose: verbosity **kwargs: node attributes Returns: NodeProto Method :meth:`eval ` creates an onnx node returned by method :meth:`make_node `. .. exec_code:: import numpy as np from onnx.reference.ops._op_list import Celu onnx_node = Celu.make_node(alpha=0.5) print(onnx_node) Nr#y) r; op_domainr4r5 get_schema min_input min_outputrangehelper make_node) r8n_inputs n_outputsrropsetr7r;rvnames_in names_outs rrEOpRun.make_nodes>)  ~--gfE''H  ~--gfE))I%*8_5_asG_5&+I&67&6qW&6 7{{$$W LVLL67s C#0C(rc ^SU4SjjnUR"X40UD6nTUSS[RR50S.nU"Xg5$)a4Instantiates this class based on the given information. Args: n_inputs: number of inputs (default is defined by the operator schema) n_outputs: number of outputs (default is defined by the operator schema) verbose: verbosity **kwargs: node attributes Returns: NodeProto cB>TS:a[U[U5-5 gg)Nr5)printr^)rrrs r log_function"OpRun.create..log_function-s {gd +,rNr)rrrr)rr.rrr0rV)rEr4r5r6)r8rFrGrrrOrrs ` rcreate OpRun.createsO, -}}X;F;499779:  4$$rrGrcUR"[U54XS.UD6nUR"U6n[U5S:XaUS$U$)zEvaluates this operator. Args: *args: inputs n_outputs: number of outputs (default is defined by the operator schema) verbose: verbosity **kwargs: node attributes Returns: NodeProto rSr5r)rQr/r)r8rGrrrinstr:s reval OpRun.eval:sF(zz#d)TyTVThho s8q=q6M r)rrrrrrr!)ronnx.NodeProtorzdict[str, Any]r;r)rbonnx.AttributeProtorzonnx.AttributeProto | Noner0r)r0rV)rrr0z list[str])r0z Sequence[str]r/)r0bool)r:tuple[Any, ...]r0r[NN)rF int | NonerGr]rrr0rX)NNr) rFr]rGr]rrurrr0r) rz list[Any]rGr]rrurrr0r)r r rrrr?r%rr staticmethodrrrrWrrrr7r rabcabstractmethodrr*r classmethodr;rErQrVrr rrrrs ITX ' 5C MP :/OS% &% 1K% % NEE-7^.$$%%&&%%   "@,0S0j   $ $,M,M,M ,M  ,M,M\ $ $ % % % %  % % %D!%     rrc$\rSrSrSrSrSrSrg) OpRunExpandiUz.Class any operator to avoid must inherit from.c2[S[U5S35eNz5The reference implementation must not use this node (r.rrDrs rr%OpRunExpand.__init__XCDJ t  !S\ 1A#g,P669$:K:K6L5MN LL+=qJ  W~r)rrrsr\) rrXrdict[str, Any] | Nonertrrrr0rV) r r rrrr%rrwrr rrrorocsL! ,0  ! *   *    *=rroc8\rSrSrSrSSSjjrSrSrg) OpFunctionContextDependantizThe function can be instantiated but only at execution time. An instance of OpFunction is created everytime to node is executed. This is needed when the schema of an operator defines a context dependent function. Nc[RXX 0S9 X0lURURn[ R RURXAR5Ul g)N)rtr) ror%parentrr7r4r5r@rschema_)r$rrrversions rr%#OpFunctionContextDependant.__init__sY DZrR -- 0 01yy++   w(8(8  rc/nUHdn[RRUR5nUR [RR XTR 55 Mf URRURU5nU"URUR5nUR"UR/UQ70UD6$r!) r4rDnp_dtype_to_tensor_dtyperSrmake_tensor_type_protorPr _load_implrrrwrs)r$rkrtypesr~rSclrUs rrOpFunctionContextDependant._runsAKK88AE LL;;E77K L[[ # #DNNE :$..$//2~~djj<6rKrcrroABCrrcrorr rrrs# % (:|:q,qC!4Cuu ; ;&   0 0!!":BCGGBJ  %  ,,^==r