+ j&Q +^RIt^RIHtHt^RIHt^RIHtHt^RI H t H t ^RI t ^RI Ht^RIt ^RI HtHt^RIHt^RIHt^R IHt^R IHt^R IHt^R I HtHtHtH t H!t!H"t"H#t#H$t$H%t%H&t&H't'H(t(^R I)H*t*^RI+H,t,H-t-^RI.H/t/H0t0^RI1H2t2H3t3H4t4] PjPmRR4t7] PjPmRRR4t8] PjPmRRR4t9] PjPmRRR4t:] PjPmRRR4t;.RNRNRNRNRNRNRNRNR NR!NR"NR#NR$NR%NR&NR'NR(NR)NR*NR+NR,NR-NR.NR/NR0NR1NR2NR3NR4NR5NR6NR7NR8NR9NR:NR;NRNR?NR@NRANRBNRCNRDNRENRFNRGNRHNRINRJNRKNRLNRMNRNNRONRPNRQNRRNRSNRTNRUNRVNRWNRXNRYNRZNR[NR\NR]NR^NR_NR`NRaNRbNRcNRdNReNRfNRgNRhNRiNRjNRkNRlNRmNRnNRoNRpNRqNRrNRsNRtNRuNRvNRwNRxNRyNRzNR{NR|NR}NR~NRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNRNt!RR]4t?RR/RRllt@RtARRltBRRltCRtD]B!R] PR]?PR7tE]B!R] PR]?PR7tG]B!R] PR]?PR7tI]B!R] PR]?PR7tJ]B!R] PR]?PR7tK]B!R] PR]?PR7tL]B!R] PR]?PR7tM]B!R ] PR]?PR7tN]B!R!] PR]?PR7tO]B!R&] PPR]?PR7tQ]B!R'] PPR]?PR7tR]B!R"] PR]?PR7tT]B!R#] PPR]?PR7tV]B!R$] PPR]?PR7tX]B!R%] PPR]?PR7tZ]B!R(] PR]?PR7t[RRlt\]B!R)]\R]?PR7t]]B!R*] PR]?PR7t^RRlt_]>!R]_] PR]#PR7tbR] P/RRlltd]>!R]d] PR]#PRR7te]B!R,] PR]?PR7tf]B!R-] PR]?PR7tg]B!R.] PPR]?PR7ti]B!R/] PPR]?PR7tj]B!R0] PPR]?PR7tk]B!R1] PR]?PR7tl]B!R2] PPR]?PR7tm]B!R3] PPR]?PR7tnRRlto]>!R]#P]o] PRR7tp]B!R5] PR]?PR7tq]>!R]!]A]?PR7]#P] PRR7ts]B!R7] PR]?PR7tt]B!R8] PR]?PR7tv]B!R9] PR]?PR7tw]B!R:] PR]?PR7tx]B!R;] PR]?PR7ty]B!R<] PR]?PR7tz]>!R]!]A]?PR7]#P] PRR7t{]B!R@] PR]?PR7t|]B!R=] PPR]?PR7t}]B!R>] PR]?PR7t~]B!RA] PR]?PR7t]B!Re] EPR]?PR7t]B!RB] EPR]?PR7t]B!RC] EPR]?PR7t]B!RD] EPR]?PR7t]B!RE] EPR]?PR7t]B!RF] PEP R]?PR7t]B!RG] EP R]?PR7t]B!RH] EPR]?PR7t]B!RI] EPR]?PR7t]B!RJ] EPR]?PR7t]C!RK] EPR]?PR7t]C!RL] EPR]?PR7t]C!RM] EPR]?PR7t]C!RN] EPR]?PR7t]C!RO] EPR]?PR7tRt]C!RP]R]?PR7t]C!RQ] EP"R]?PR7t]C!RR] EP$R]?PR7t]C!RS] EP&R]?PR7t]C!RT] EP(R]?PR7t]C!RV] EP*R]?PR7t]C!RW] EP,R]?PR7t]C!RX] EP.R]?PR7t]C!RY] EP0R]?PR7t]C!RZ] PEP2R]?PR7t]C!R[] PEP6R]?PR7t]C!R\] EP:R]?PR7t]C!R]] EP<R]?PR7tRRlt]C!R^]R]?PR7tRRlt]C!R_]R]?PR7t]C!R`] EPFR]?PR7t]C!Ra] EPHR]?PR7t]C!Rb] EPJR]?PR7t]C!Rc] EPLR]?PR7t]C!Rd] EPNR]?PR7t]C!Rf] EPPR]?PR7t]C!Rg] EPTR]?PR7t]Dt]C!Ri] EPZR]?PR7t]C!Rj] PEP\R]?PR7tRRltRRltRt]>!R]]]#P]R7tRRltRtRt]>!R]]]#P]R7tRRltRRltRRltRRltRRltRt]>!R]]]#P]R7tRRltRt]>!R]] EP]#P]R7tERRRlltRRltRRltRt]>!R]]]#P]R7tRRltRt]>!R]] EP]#P]R7tRRltRRltRt]>!R]]]#P]R7tRRltRRltRt]>!R]]]#P]R7tERERltERERltERt]>!ER]]]#P]R7tERERltERt]>!ER ]] EP]#P]R7tER ER ltER ER ltERt]>!ER]]]#P]R7tERERltERERltERt]>!ER]]]#P]R7tERERltERERltERt]>!ER]]]#P]R7tERERltERt]>!ER]] EP]#P]R7tER ER!ltER"t]>!ER#]] EP]#P]R7tER$ER%ltER&ER'ltER(t]>!ER)]]]#P]] EPEP3ER*7tERER+ER,lltERER-ER.lltER/t]>!ER0]]]#P]R7tER1ER2ltER3tER4t]>!ER5]]]#P]R7tER6ER7ltER8ER9ltER:t]>!ER;]]]#P]R7tER<ER=ltER>ER?ltER@t]>!ERA]]]#P]R7EtERBERClEtERDERElEtERFEt]>!ERGE]E]]#EPE]RERH7EtERIERJlEtERKERLlEtERMEt]>!ERNE]E]]#PE]R7Et EROERPlEt ERQERRlEt ERSEt ]>!ERTE] E] ]#EPE] R7Et ERUR/ERVlEtERWEtERXEtERYEtERZEtER[EtER\EtER]EtER^ER_lEtER`ERalEtE]!R] EP0E]ERb7EtRR/ERcERdllEtE]!RE]E]ERb7EtRR/EReERfllEtE]!RE]E]ERb7EtERERglEtE]!RE]E]ERb7EtE]!R] EP>E]ERb7EtE]!R] EP@E]ERb7Et ERhEt!ERiERjlEt"ERkERllEt#]>!ERm]#PE]"E]#E]!R7Et$ERnERolEt%ERpERqlEt&ERrEt']>!ERsE]%E]&]#PE]'R7Et(ERtERulEt)ERvEt*]>!ERw]#PE])] EPVE]*R7Et+ERxERylEt,ERzEt-]>!ER{]#PE],] EP\E]-R7Et.ER|ER}lEt/ER~ERlEt0EREt1]>!ERE]/E]0]#PE]1R7Et2ERERlEt3ERERlEt4EREt5]>!ERE]3E]4]#PE]5R7Et6ERERlEt7ERERlEt8EREt9]>!ERE]7E]8]#PE]9R7Et:ERERlEt;ERERlEt!ERE];E]<]#P]#P]#P3E]=R7Et>ERR/ERERllEt?ERR/ERERllEt@EREtA]>!ER]#PE]?E]@E]AR7EtBERR/ERERllEtCERR/ERERllEtDEREtE]>!ER]#PE]CE]DE]ER7EtFERERlEtGERERlEtHEREtI]>!ERE]GE]H]#PE]IR7EtJERERlEtKERERlEtLEREtM]>!ERE]KE]L]#PE]MR7EtNERERlEtOERERlEtPEREtQ]>!ERE]OE]P]#PE]QR7EtRERERlEtS]>!ERE]S]#P]#P3] EPRR7EtTERERlEtU]>!ERE]U]#PE]UERR7EtVERERlEtW]>!ERE]W]#EPE]WERR7EtY] EPEPEPE]Y4]!4]!4R#(N)CallableSequence)Enum)partialreduce)OptionalUnion) sym_floatTensor)_get_default_devicenew_token_tensor)is_functional_schema)register_debug_prims)register_rng_prims) DimDimsSequenceTypeDimsTypeIntLikeNumber NumberType RETURN_TYPE ShapeType StrideType TensorLikeTensorLikeType type_to_dtypebackwards_not_supported) FakeTensorFakeTensorMode)handle_torch_functionhas_torch_function) tree_flattentree_maptree_unflattenprimsDEFIMPLCompositeExplicitAutograd BackendSelectAutogradMetarabsacosacoshasinasinhatanatanhcoscosh bessel_i0 bessel_i0e bessel_i1 bessel_i1e bessel_j0 bessel_j1 bitwise_notcbrtceil conj_physicaldigammaerferf_inverfcerfcxexpexpm1exp2fillfloorimagisfinitelgammaloglog1plog2log10ndtrinegreal reciprocalroundsignsignbitsinsinhspherical_bessel_j0sqrttantanhtruncaddatan2 bitwise_and bitwise_or bitwise_xordiveqfmaxfminfmodfrexpgcdgegthypotigammaigammacleltmaximumminimummulne nextafterpow remainderrsqrt shift_leftshift_right_arithmeticshift_right_logicalsubzeta as_stridedbroadcast_in_dim collapse_viewconj expand_dimsslice split_dimsqueeze transposeview_ofview_element_typeas_strided_scattercollapsecatreshaperevwherecloneconvert_element_type device_putitem maximum_value minimum_value copy_stridedcopy_toresizeamaxaminprodsumxor_sumvar empty_stridedempty_permuted scalar_tensoriotasvdnormal_uniform_helperfft_r2cfft_c2cfft_c2r _make_token _sink_tokensshapestridesdtypedevicec V^8dQhR\\P,R,R\R,R\R,R\P R,R\P \,R,/#) tensorlikeNrrrr)rtorchr rrrrstr)formats"V/home/luca/screen-checking/.venv/lib/python3.14/site-packages/torch/_prims/__init__.py __annotate__rsr5K5KU\\)D05K t 5K$  5K ;;  5K LL3  % 5Kc\V\4'dV'd.\V\4'g\R\ V4 24hV'd.\V\4'g\R\ V4 24hR pR p\ \ V44p\ P!R4pMVe\V\ P4'g\R\ V4 24h\VP4p\VP44pVPpVPpM !RS S#mNuE")/uW~G#mNE &_FF&#f%   uU JJrtagsuse_old_custom_ops_apiFregister_conj_neg_fallthroughcV^8dQhR\R\\\R3,,R\R\R\R\\ P ,R,R \R \/#) rschema return_type.meta impl_atendocrNrr)rrrrrrTagbool)rs"rrrsnn nu[#%566n  n  n n 599  $n!n$(nrc( aaaaaaVV3RloV3RlpVV3Rlp VPR^R7^,p V\V 4Rp\PP W,4p V'g\ V 4'gy\ PW,\PPR7\PV S4\PW4\PV S4EMV PU u.uF=p V PfKV PP 'gK1V P"NK? p p \P$P'RV ,S\)V 4VR 7pVP+S4V\,P.8Xg V'dkVP0PV \P$P2R 4VP0PV \P$P2R 4\5\P6P8P:V 4pVP<oV'd VSnEM\5\P8P@V R4;p'dVPC4Uu.uFpVPDNK pp\GV^,4oSPH!VR ,!SPK\PPL4SPK\PPN4\(;QJd.V3R lV^,4FNK 5M!V3R lV^,44Sn^RI(H)o\T;QJd3V3RlSPVP4F 'gK RM( RM$!V3RlSPVP44'd\YS4R8Xd\ZPW4VS3F-pVVn.VVn/VVn0SVn SVn SVn1K/ \PPeSPVP"SPVPf4S#uup iuupi)z! Creates a primitive operation. c$<S!V/VBS!V/VB#Nr)argskwargsrrs*,r _prim_impl_make_prim.._prim_impl(s" df$)&))rc&<\S4!V/VB#rr)rr_prims*,r_autograd_impl"_make_prim.._autograd_impl3s&u-t>v>>rc<VPR4'd!VR,PR8Xd S!V/VB#\;QJdRV4F 'gK RM RM !RV44'd S!V/VB#S!V/VB#)rrc3"TF6p\V\P4;'dVPR8HxK8 R#5i)rN)rrrr).0xs& r ;_make_prim.._backend_select_impl..9s/NAz!U\\*??qvv/??s %AATF)getrany)rrrrs*,r_backend_select_impl(_make_prim.._backend_select_impl6su ::h  F8$4$9$9V$C(( ( 3NN333NN N N(( (t.v. .r()maxsplitN)rzprims::) mutates_argsr ConjugateNegative:NNc38<"TFqS9gK VxK R#5irr)rttags_intersections& rr_make_prim..ksR'7!@Q;QAA'7s )contains_tensor_typesc3H<"TFpS!VP4xK R#5ir)r)rars& rrrps O7N!%aff--7Ns"TFzprims.device_put.default)4splitlenr_C parse_schemarprimdefinerpt2_compliant_tag prim_implimplprim_autograd_implprim_meta_impl arguments alias_infois_writenamelibrary custom_opr register_fakerVIEW_libfallthrough_kernelgetattr_opsopsr'default_tagsaten op_overloadsrsetintersection_updatediscardcoredata_dependent_outputtorch._subclasses.fake_tensorrr_schemarprim_backend_select_impl__doc__rrr!_fake_dispatch_register_prim_meta overload_name)rrrrrrrrrrr cpp_schemaargrprim_def _prim_packet aten_packetoverload overload_tagsprrrrs$$dd$$$$ @@@@r _make_primr#sL *?/ <<a< ( +D CIK F&&t}5J%9*%E%E DM (C(C DtZ(5D$'"++ +~~ .1nn.E.E CHH+  ==**   |, +  t$ +** *.K MM  tU]]%E%E{ S MM  tU]]%E%Ez R5::>>//6L  E   d; ; ;7B7O7O7QR7Q87Q R a 01--}R/@A !!%))..1 !!%))"A"ABeR}Q'7ReeR}Q'7RR C COu}}7N7NOCCCOu}}7N7NO O O    &  & !%%dAE " #    # HH.. EMM77 L{ .SsR R :R !Rc&]tRtRtRtRtRtRtRtR#)$ELEMENTWISE_PRIM_TYPE_PROMOTION_KINDirNr)r))) __name__ __module__ __qualname____firstlineno__DEFAULT INT_TO_FLOAT ALWAYS_BOOLCOMPLEX_TO_FLOAT__static_attributes__rrrr%r%sGLKrr%args_with_fixed_dtypescdV^8dQhR\R\\R3,R,R\/#)rtype_promotionr2.Nreturn)r%rrr )rs"rrrs<__8_".#"56=_ _rcJ\V4^8Xd \R4h\P!V!\ V4pVe\ V4V,p\P !VRR/\P !VRR/\P!V!wrE\P!VRR/pRpRpVFtp \V \4'd9\P!V 4'gV PpM5V PpKQ\V \4'gKi\V 4pKv VfVe\P!V4pRp Rp VFwp \V \4'd?\P!V 4'dV fV P p KGKIV P p M$\V \4'gKoV eKuT p Ky V EeVf \R4hV\"P$8Xd\&P(pMV\"P*8XdO\P,!V4'g\P.!V4'd\&P0!4pMGV\"P28Xd3\P4!V4'd\P6!V4pVf \R4h\&P8!WdWR7#Rp \V \&P:\&P<34'dVF~p \V \>\@\&P:\&P<34'g\R \V 4 24hT ;'g!\V \@\&P<34p K V 'd \CV 4p \EV 4#) z Meta function for elementwise operations that produce outputs in the same dtype as their inputs. Stride logic is currently incorrect. z4elementwise operation requires at least one argumentNallow_cpu_scalar_tensorsTz.dtype must not be None when device is not Nonez.shape must not be None when device is not None)rrFz.Expected int, float, SymInt, or SymFloat, got )#rrutilscheck_same_dtypelistcheck_same_devicecheck_same_shape3compute_elementwise_output_logical_to_physical_perm extract_shaperris_cpu_scalar_tensorrrrrrr%r/rrr.is_integer_dtypeis_boolean_dtypeget_default_dtyper0is_complex_dtypecorresponding_real_dtyperSymIntSymFloatintfloatr r)r4r2rargs_l2p_perm_rr scalar_typerrnumber seen_floatrs$$* r_prim_elementwise_metarOs 4yA~STT D! JE)+,u4 UBTB EADAKKUSKH    F FE EK c: & &--c22   V $ $s)K }0##K0F F c: & &))#..> ZZF" V $ $~" = !QR R AMM MJJE CPP P%%e,,0F0Fu0M0M//1 CTT T%%e,,66u= = !QR R##EFPP J&5<<899Aa#uellENN!KLL$DT!WIN$MMz!eU^^5L'MJ  v&F f rc\P!\P!V^,P4R4\ V/VB#)rcR#)zOnly complex dtype is supportedrrrr0_complex_only_elementwise_meta..s7Xr)r_checkr8rCrrOrrs*,r_complex_only_elementwise_metarVs9 LL tAw}}-/X "4 26 22rc0V^8dQhR\R\/#rrr4rr%)rs"rrr    "F rc d\RRV R2R\\VR7R\P/VB#)z$ Creates an elementwise unary prim. rz(Tensor self) -> Tensorrr4rrr#rrOrNEWrr4rs&$,r_make_elementwise_unary_primr`sI  ./  +N K  OO    rc0V^8dQhR\R\/#rXrY)rs"rrr rZrc d\RRV R2R\\VR7R\P/VB#)z% Creates an elementwise binary prim. rz%(Tensor self, Tensor other) -> Tensorrr\rrr]r_s&$,r_make_elementwise_binary_primrc sI  <=  +N K  OO    rc\hrNotImplementedErrorrUs*,r _not_implrgs r)rrr4cDV^8dQhR\PR\/#rrr5)rr )rs"rrrs 8 8%,, 86 8rc\P!VP4'*R4\P!\P!VP 4R4V4#)cR#)zJcbrt: Complex inputs not supported. Consider calling torch.pow(a, 1.0/3.0)rrrrrR_cbrt_aten..s\rgUUUUUU?)rrT is_complexcopysignrxr.rs&r _cbrt_atenrqsA LL LLN\ >>%))AEEGU3Q 77rc0V^8dQhR\R\/#)rinputr5r)rs"rrrs..~...rcVPP'g \R4h\P!V4p\ WR7#)z6prims.conj_physical is only defined for complex dtypes)r)rrn RuntimeErrorr8"compute_elementwise_output_stridesr)rsrs& r_conj_physical_metarxs9 ;; ! ! !STT66u=G e --rz$conj_physical(Tensor self) -> Tensorz4Returns the physical conjugation of a complex tensor)rrrrr memory_formatcPV^8dQhR\R\PR\/#)rrsryr5)rrry)rs"rrrs*   -2-@-@  rcdV\P8wdD\P!VPVPVP VP VR7#\P!V4p\P!VPVVPVP VP R7#))rlayoutrry)rr|r) rpreserve_formatemptyrrr|rr8rwr)rsrycomputed_strides&$ r _clone_metars---{{ KK++<<<<'    BB5I"" KK ++<<<<   rzAclone(Tensor self, *, MemoryFormat? memory_format=None) -> TensorzReturns the copy of a tensorT)rrrrrrc<V^8dQhR\R\R\/#)rrvaluer5rr)rs"rrrs!.rc8\V\PR7#)r\rOr%r-)rrs&&r _fill_metars ! >FF rz)fill(Tensor self, Scalar value) -> Tensor)rrrrrz!imag(Tensor(a) self) -> Tensor(a)r\)rrrrrz!real(Tensor(a) self) -> Tensor(a))rrrr4cL\V\\\P34;'gC\V\P 4;'d!\ P!VP4pV'd\P!WRR7#\P!W4#)r_) rounding_mode) rrrGrrEr r8r@rre true_divide)rb is_integrals&& r _div_atenrsrQsELL 9:1ell#GG(>(>qww(GyyW55  &&rchV^8dQhR\\,R\\,R\/#rrrr5r)rs"rrrl. "'5 'Brcd\V\4'd9\V\4'd#\WPVP R7pMM\V\4'd8\V\4'd"\WPVP R7p\ P!W4#r)rrrrrrrrsrrs&&r _maximum_atenrlm!Z  Z6%:%: !77188 < Az " "z!V'<'< !77188 < == rchV^8dQhR\\,R\\,R\/#rr)rs"rrrrrcd\V\4'd9\V\4'd#\WPVP R7pMM\V\4'd8\V\4'd"\WPVP R7p\ P!W4#r)rrrrrrrrtrs&&r _minimum_atenrrrc TV^8dQhR\R\R\R\R\/#rrsizerstorage_offsetr5rrrrG)rs"rrrs4===&=0:=LO==rc\V4\V48wd$\R\V4 R\V4 24hV^8d\RV 24h\P!V4\P!V4\ \ PV4^8XdMF\V\P4'd'\P!VP4WV4\P!WW#4#)z len(size)=z != len(stride)=z!storage_offset must be >= 0, got )rrr8validate_stridesvalidate_shaperoperatorrurrr check_in_bounds_for_storage_typed_storagerrrrrs&&&&r_as_strided_metars 4yCKz#d)4DS[MRSS@@PQRR 6"  hllD!Q& Au|| $ $ ))   n    AV < Tensor(a!)rrrrrcRV^8dQhR\R\R\\,/#)rrrbroadcast_dimensions)rrrrG)rs"rrrs2T@T@T@'T@?G}T@rc aaaa a ^RIHpHpHp\ S\ 4'g\ R\S4 24h\ S\4'g\ R\S4 24h\ V\4'g\ R\V4 24hSP\V48wd&\ RSP R\V4 R24h\S4SP8d&\ R\S4 R SP R24hV3R lp\WbR 4\V4FZwoo \P!V!SPS,^8HSS ,SPS,8H4VVV V3R l4K\ .p^o \!\S44EFoSV9dV!SPS ,^8H4'dcV!SPS ,SS,8H4'd(VP#SP%4S ,4MqVP#^4M_\P!SPS ,SS,8HVVV V3R l4VP#SP%4S ,4S ^,o KV!SS,^8g4'dVP#^4EK%S SP8XdVP#^4EKJVP#SP%4S ,SP'4S ,,4EK SP)SVSP+44#)r)guard_or_false guard_or_truesym_ora must be TensorLike, got zshape must be a Sequence, got z-broadcast_dimensions must be a Sequence, got za.ndim (z ) != len(broadcast_dimensions) ()z len(shape) (z) must be >= a.ndim (c<\V\4'g\R\V4 24hW8:d\RV RV 24hV\ S48d\RV R\ S4 24hV#)z.broadcast_dimensions element must be Dim, got z1broadcast_dimensions must be strictly ascending: z <= zbroadcast_dimension z# out of bounds for shape of length )rrrrr)accrrs&&r_greater_than_reduce4_broadcast_in_dim_meta.._greater_than_reduce s!S!! @a J  8 CA3d3%P  E ? &qc)LSQVZLY rcB<SPS, RSS, 2#)z must be broadcastable to r)ridxnew_idxrsrrR(_broadcast_in_dim_meta..6sqwws|n$>uW~>NOrcD<RSPS, RSS, 2#)z#non-broadcasting semantics require z == r)rr original_idxrsrrRrGs&A!'',BWAXX\]bcf]g\hir)%torch.fx.experimental.symbolic_shapesrrrrrrrrndimrr enumeraterrTrrangeappendrrrr) rrrrrrr new_stridesrrrs ff& @@@r_broadcast_in_dim_metars a $ $9$q'CDD eX & &=d5k]KLL *H 5 5;DAUsCW?X>YYZ [   5zAFF|CJ<7LQVVHTUVWW   r:""67 W 1773<1$eGn &D E O 8 KLSZ  & &aggl3q899!!'',"75:"EFF&&qxxz,'?@&&q) GGL)U3Z7i""188:l#;<'!+LU3Z1_--""1%'""1%""188:l#;affh|>T#TU/!2 <<{A,<,<,> ??rc\V4pVFpRW4&K Tp\V4FwrgVR8wgKVPV4pK! VPV4#rr)r:r unsqueezeexpand)rrrsbroadcast_dimensionvrrs&&& r_broadcast_in_dim_atenrWsZ U A3!# 4 AA, 7 C A 88E?raD Creates a view of a with the specified shape. Allows adding dimensions of any length and broadcasting dimensions of length one in a to any length. The location of the broadcast dimensions must be specified using the broadcast_dimensions argument. Changing the relative order of dimensions is not supported. zVbroadcast_in_dim(Tensor(a) a, SymInt[] shape, int[] broadcast_dimensions) -> Tensor(a)c@V^8dQhR\R\R\RR/#)rrstartendr5Nr rG)rs"rrrxs(  v c   rcaa\^VP44p\P!VS4\P!VS4\P !SS8VV3Rl4R#)rc<RS RS R2#)z Attempting to collapse but end, z, is less than start, !r)rrsrrR)_validate_collapse_args..s23%7MeWTUVrN)maxdimr8 validate_idxr _check_value)rrrrs&ff r_validate_collapse_argsrxsN q!%%'?D tU# tS!  u Vrc bV^8dQhR\R\R\R\\R3,/#)rrrrr5.)rrGr)rs"rrrs0 = =I =c = =c3h =rc\V4^8XdRM \V4p^pWV^,F pW4,pK V^VV3,W^,R,#)zR Returns the shape of a with dims in [start, end) merged into a single dimension. Nr)rr)rrr dim_lengthrs&&& r_collapsed_shapers[ J!ODuEJ 37 #^ $ 5>ZM )E'),< <.s}rNrr)NN)rrrrrrrrrrrrrnumelrrrTminr8make_contiguous_strides_for)rrrrrrrrrrvalid_oprrlength new_shapers&&&f r_collapse_view_helperrs] a $ $9$q'CDDAc* vv{((*vv{s|~Haggi1n%%q%!)R0CJ!O'Na'LG!G$4u1W~$EEHh%/001h Q/H X45h'' \FS1Weai, :?.F -ZFVq[!!q%!)R0CejAo..Cj(F 1fu  )E'),< D# DC DN Drc\WVR4wr4Vf \R4hVf \R4hVPW4VP44#)z?Attempting to view a collapsed tensor, but no such view exists!z:new_strides should not be None after _collapse_view_helperz8new_shape should not be None after _collapse_view_helper)rrrr)rrrrrs&&& r_collapse_view_metars]2 #XI H  WXX << 0@0@0B CCrcHV^8dQhR\R\R\R\/#rr)rs"rrrs(6#CFrcP\VPW4pVPV4#r)rrviewrrrrs&&& r_collapse_view_atenrs! %5I 66) ra Creates a view of a with the dimensions between start (inclusive) and end (exclusive) merged into a single dimension. If it's not possible to take such a view then an error is thrown. See collapse instead. The dimensions can be merged if and only if they are all "nested" with each other. That is, they all have the property that stride[i] = stride[i+1] * shape[i+1] for all i in [start, end - 1). z;collapse_view(Tensor(a) a, int start, int end) -> Tensor(a)c0V^8dQhR\R\/#rjrt)rs"rrrs.^rc*VPP'g \R4hVPVPVP 4VP 44p\PPWP4'*4V#)z$Expected complex dtype in prims.conj) rrnrvrrrrrr _set_conjis_conj)routs& r _conj_metarsb 77   ABB ,,qww A,<,<,> ?C HHs O, Jrz2 Returns a conjugated view of the original tensor zconj(Tensor(a) a) -> Tensor(a)c<V^8dQhR\R\R\/#rr dimensionsr5rr)rs"rrr-s)@@@#3@@rcVe!\\P!W!44pM*\\P!VPV44p\ \ V44\ V48wdR\ V4 2p\V4h\VP4pVFpVPV^4K \\ V44Uu.uF qfV9gK VNK pp\WV4#uupi)z Creates a view of a with a.ndim + len(dimensions) dimensions, with new dimensions of length one at the dimensions specified by dimensions. z+Received duplicate dimensions to expand in ) sortedr8canonicalize_dimsrrrr ValueErrorr:rinsertrr)rrrdimsmsgrrrs&&& rrr-s e--d?@e--affjAB 3t9~T";C O;LMoQWW Ia S^,,:0E, A*> ??s C1C1cHV^8dQhR\R\R\R\/#rrr outer_lengthr5r)rs"rrrGs.DD~DCDsD~DrcR\V\4'g\R\V4 24h\P !VP V4\P!V4VPV,V,pVPV,V,^8wd&RVPV, RV R2p\V4h.p.p\VP 4FpWq8XdYVPW#34VPVP4V,V,VP4V,34KaVPVPV,4VPVP4V,4K VPWVVP44#)rz(Attempting to split dimension of length z, but outer length of z divides it with a remainder!)rrrrr8rrvalidate_dim_lengthrrrextendrrrr)rrr inner_lengthrrrrs&&& r_split_dim_metar GsM a $ $9$q'CDD qvvs# l+773<</L  |#)6qwws|nE##/.0M O oIKQVV} :   l9 :    3, > 3P Q   QWWS\ *   qxxz# /  << 0@0@0B CCrcHV^8dQhR\R\R\R\/#rr)rs"rrrds(vCsvrcVPV,V,pVP^VW#3,VPV^,R,pVPV4#)rN)rr)rrrr rs&&& r_split_dim_atenrdsN773<</L#,!==a @RRI 66) rz Creates a view of a with the given dimension (of length l) split into two dimensions, with the outer of the two having length outer_length and the inner of the two having computed length inner_length such outer_length * inner_length = l. zAsplit_dim(Tensor(a) a, int dim, SymInt outer_length) -> Tensor(a)c<V^8dQhR\R\R\/#r)rr)rs"rrr}s&DD^DDnDrcX\V\4'g\R\V4 24hVF_p\P !VP V4VPV,^8wgK>\RV RVPV, R24h .p.p\\VP44FSpW!9dK VPVPV,4VPVP4V,4KU VPW4VP44#)rzCannot squeeze dimension z with size z (must be 1))rrrrr8rrrrrrrrr)rrrrrs&& r _squeeze_metar}s a $ $9$q'CDD 1663' 773<1  +C5 AGGCL>V IKS\"   &188:c?+ # << 0@0@0B CCrz~ Creates a view of the tensor with the specified dimensions removed. The removed dimensions must each have length one. z3squeeze(Tensor(a) a, int[] dimensions) -> Tensor(a)c<V^8dQhR\R\R\/#rr permutationr5r)rs"rrrs'RR~R4DRRrc.VP\V48wd(RVP R\V4 R2p\V4h\P!VPV4'gRV R2p\V4h^.VP,p^.VP,p\ V4F3wrVVP V,W5&VP4V,WE&K5 VP\V4\V4VP44#)z'Attempting to permute a tensor of rank z', but received a permutation of length rz!Received an invalid permutation, ) rrrr8is_valid_permutationrrrrrr)rrrrrrrs&& r_transpose_metarsvv[!!7x?fgjkvgwfxxyzo  % %affk : :1+a@oaff I#,Kk* 88:c? + <<i(% *N>N>P QQrc<V^8dQhR\R\R\/#r)r r)rs"rrrs"))v),<))rc.\P!W4#r)rpermute)rrs&&r_transpose_atenrs == ((rz Creates a view of the tensor with its dimensions permuted. The length of the permutation must be the rank of the tensor, and each element of the permutation specifies the new order for the corresponding dimension. z6transpose(Tensor(a) a, int[] permutation) -> Tensor(a)c0V^8dQhR\R\/#rjrt)rs"rrrsAA^AArctVPVPVP4VP44#r)rrrrrps&r _view_of_metar s( <<Q-=-=-? @@rc0V^8dQhR\R\/#rjr )rs"rrrsVrc8VPVP4#r)rrrps&r _view_of_atenr$s 66!''?rz' Creates a view of the tensor. z!view_of(Tensor(a) a) -> Tensor(a)cPV^8dQhR\R\PR\/#rrrr5rrr)rs"rrrs%~ekknrc$VPV4#rrrrs&&r_view_element_type_metar+ 66%=rcPV^8dQhR\R\PR\/#r&r rr)rs"rrrs%vekkfrc$VPV4#rr)r*s&&r_view_element_type_atenr0r,rz> Creates a view of the tensor with a different dtype. z9view_of_dtype(Tensor(a) a, ScalarType dtype) -> Tensor(a)c `V^8dQhR\R\R\R\R\R\/#)rrssrcrrrr5r)rs"rrrsD// / / /  /  /  /rcaaaaaa\P!S4\P!S4\P!SSS4o\P !SP 4S8VVVVV3Rl4\P !\P!SPS4VV3Rl4\P!S4#)c<RS RS RS RSP4 RSSP4, RSP4SP4, 2 #)zas_strided_scatter: sizes z , strides z, storage offset z and itemsize z requiring a storage size of z' are out of bounds for storage of size ) element_sizer)rs required_sizerrrsrrR*_as_strided_scatter_meta..sj(j@QR`Qab"//122Ou113345##(;;=53E3E3G#G"H Jrc(<RSP RS 2#)zCexpected src to have a size equal to the slice of self. src size = z, slice size = r)rr2srrRr7sUVYV_V_U``optouvr) r8rrcompute_required_storage_lengthrrTr is_same_shaperclone_preserve_strides)rsr2rrrr6sfffff@r_as_strided_scatter_metar<s  6"99$WM LL &   LL CIIt,v  ' ' ..rz Creates a new tensor equivalent to ``out = input.clone()`` after mutation by ``out.as_strided(size, stride, storage_offset).copy_(src)``. zlas_strided_scatter(Tensor self, Tensor src, SymInt[] size, SymInt[] stride, SymInt storage_offset) -> TensorcHV^8dQhR\R\R\R\/#rr)rs"rrr's(""f"S"s"v"rch\WV4\VPW4pVPV4#r)rrr new_emptyrs&&& r_collapse_metar@'s+Ac* %5I ;;y !!rcHV^8dQhR\R\R\R\/#rr)rs"rrr.s(fSsvrc\VPW4pVPV4p\P!4;_uu_4VP V4P V4RRR4V# +'giT#;ir)rrr?rno_gradview_ascopy_)rrrrrs&&& r_collapse_atenrF.sX %5I ++i C  AQ  J  Js !A00 B zn Collapse a span of neighboring dimensions into one. See collapse_view for the corresponding view operation. z0collapse(Tensor a, int start, int end) -> TensorcRV^8dQhR\\,R\R\/#)rtensorsrr5)rrrG)rs"rrrGs&x/cnrc aaaaa S^8d\RS 24hV^,Pp.p\V4Fwo p\V4\VP48wd1\R\V4 RS R\VP4 24h\\ W$P44FEwowooSS8XdVP S4K#\ P!SS8HVVVVV 3Rl4KG K \V^,P4P4p\ P!V4VS&\V^,V\P!V4R7#)rzdim must be non-negative, got z>All tensors must have the same number of dimensions. Expected z but tensor z has c (<RS RS RS RS RS R2 #)z0Sizes of tensors must match except in dimension z . Expected z in dimension z but got z for tensor number z in the listr) common_lengthrrr tensor_idxsrrR_cat_meta..Ys0NseT -nSE6(Re!l,0rrr)rrrrziprrrTr:copysym_sumrr8r) rHrr sym_sum_argstensorrrKrrrLs &f @@@@r _cat_metarTGs? Qw=cUCDD AJ  EL'0 F u:V\\* * J<|J Tensorc0V^8dQhR\R\/#rrrrr)rs"rrrzs X X^ XI Xrcj\V\4'g\R\V4 24h\P !V4\ \PV4pW P48wd#RVP4 RV R2p\V4h\W\P!V4R7#)rz$Attempting to reshape a tensor with z elements to a shape with elements!rN) rrrrr8rrrrurrrr)rrrrs&& r _reshape_metar]zs a $ $9$q'CDD  8<< 'E  4QWWYK?YZ_Y``jko ae.O.OPU.V WWrc<V^8dQhR\R\R\/#rrrr5r r)rs"rrrs&IIVIII&Irc`VPV4P\PR7#)ry)rrrcontiguous_formatrrs&&r _reshape_atenres% 99U  ! !0G0G ! HHrz` Creates a contiguous tensor with the specified shape containing a copy of the data in a. z+reshape(Tensor a, SymInt[] shape) -> Tensorc<V^8dQhR\R\R\/#)rrrr5r)rs"rrrs'DDD'7DNDrc\P!VPV4\P!V\P R7#rb)r8validate_dimension_indicesrr empty_liker})rrs&&r _rev_metarjs/ $$QVVT2   AU-B-B CCrzD Reverses the order of elements along the given dimensions. z#rev(Tensor a, int[] dims) -> TensorcHV^8dQhR\R\R\R\/#)rpredrrr5rt)rs"rrrs, +0>rc>\VV\PV3R7#))r4r2r)rlrrs&&&r _where_metarns% "  ;CC $w  rz Selects elements from a and b according to pred. Where pred is true the result contains the element from a, and where pred is false the result contains the element from b. z0where(Tensor pred, Tensor a, Tensor b) -> TensorcPV^8dQhR\R\PR\/#r&r')rs"rrrs% 7 7. 7 7 7rc|\V\4'g\R\V4 24h\V\P 4'g\R\V4 24h\P PV4'dVP4pM\P!V4p\WVR7#)rzdtype must be torch.dtype, got )rr) rrrrrr _prims_common%is_non_overlapping_and_dense_or_falserr8rwr)rrrs&& r_convert_element_type_metarss a $ $9$q'CDD eU[[ ) )>tE{mLMM @@CC((*::1= a 66rcPV^8dQhR\R\PR\/#r&r.)rs"rrrs%""&"""rcR\P!V4'gRpM VPp\P !WP WR7p\P!4;_uu_4\W04uuRRR4# \dRpLdi;i +'giR#;i)F)rr requires_gradN) r8 is_grad_dtyperv ExceptionrrirrCr)rrrvresults&& r_convert_element_type_atenrzs}   u % %  "OOM   ((%F v!   "!M " s B, B BB B& z6 Creates a copy of a tensor with the given dtype. z:convert_element_type(Tensor a, ScalarType dtype) -> Tensor)rrrrrrcfV^8dQhR\R\\P,R\/#rrrr5)rrrr)rs"rrrs2 C C C"U\\1 C Crcr\V\4'g\R\V4 24h\V\\ P 34'g\R\V4 24h\V\4'g\R\V4 24h\V\P!V4R7#)rz(device must be str or torch.device, got znon_blocking must be bool, got )r) rrrrrrrrrr8canonicalize_devicerr non_blockings&&&r_device_put_metars a $ $9$q'CDD fsELL1 2 2GV ~VWW lD ) )>tL?Q>RSTT a 9 9& A BBrcfV^8dQhR\R\\P,R\/#r|)r rrr)rs"rrr s-33 3U\\)3 3rc&VPWR7#))r)tors&&&r_device_put_atenr s 444 22rz5 Creates a copy of a tensor on the given device. zFdevice_put(Tensor a, Device device, bool non_blocking=False) -> Tensorc0V^8dQhR\R\/#rj)rr )rs"rrr s''.'Z'rcd\P!VP4p\V!R44#r)r8 dtype_to_typerr)r number_types& r _item_metar s%%%agg.K k"o &&rz< Converts a tensor with one element to a Python number. c\PPP4;_uu_4\PP !V/VBuuRRR4# +'giR#;ir)r _dispatchpythonno_python_dispatcherr rrUs*,r_item_aten_no_python_dispatcherr-sA    4 4 6 6||  $1&1 7 6 6 6s A A- zitem(Tensor a) -> ScalarcDV^8dQhR\PR\/#rrr5rrr )rs"rrr@''u{{'z'rcP\P!V4p\V!R44#rr8rrrrs& r_maximum_value_metar@!%%e,K k"o &&rc8V^8dQhR\P/#rrrr)rs"rrrE&&u{{&rcV\P8XdR#VP'gVP'd!\P!V4P #\P !V4P #)T)rrrnis_floating_pointfinforiinfors&r_maximum_value_atenrEsR      U444{{5!%%%{{5!%%%rz2 Return the maximum finite value for a dtype. z)maximum_value(ScalarType dtype) -> ScalarcDV^8dQhR\PR\/#rr)rs"rrr`rrcP\P!V4p\V!R44#rrrs& r_minimum_value_metar`rrc8V^8dQhR\P/#rr)rs"rrrerrcV\P8XdR#VP'gVP'd!\P!V4P #\P !V4P #F)rrrnrrrrrs&r_minimum_value_atenresR      U444{{5!%%%{{5!%%%rz2 Return the minimum finite value for a dtype. z)minimum_value(ScalarType dtype) -> Scalarc0V^8dQhR\R\/#)rrrrt)rs"rrrs  ^  rc`\V\4'g\R\V4 24h\V\4'g\R\V4 24hVP 4VP 48wd1RVP 4 RVP 4 R2p\ V4hV#)rzb must be TensorLike, got zAttempting to copy z elements to a tensor with r\)rrrrrrv)rrrs&& r _copy_to_metars a $ $9$q'CDD a $ $9$q'CDD wwyAGGI#AGGI;.I!'')T^_3 Hrc<V^8dQhR\R\R\/#rr")rs"rrrs!V6rc$VPV4#r)rErs&&r _copy_to_atenrs 771:rz; Copies the data in b to a and returns the modified a. z-copy_to(Tensor(a!) a, Tensor b) -> Tensor(a!))rrrrrrc0V^8dQhR\R\/#)rrrrZ)rs"rrrs  . ) rc \V\4'g\R\V4 24h\P !VP VVPVPVPVPR7#)r)rr|rrv) rrrrrrrrr|rrv)rrs&&r_copy_strided_metars_ a $ $9$q'CDD    ggxxxxoo  rc<V^8dQhR\R\R\/#)rrrr5r`)rs"rrrs!  & )  rc \P!VP4VVPVPVP VP R7pVPV4V#))rrr|rrv)rrrrr|rrvrE)rrrs&& r_copy_strided_atenrsL    ggxxxxoo  CIIaL Jrzp Copies the data in a to a new tensor, the new tensor has same shape with a size, but has different stride. z1copy_strided(Tensor a, SymInt[] stride) -> Tensorc0V^8dQhR\R\/#rYrZ)rs"rrrsN9rc$VPV4#rresize_rds&&r _resize_metar 99U rc<V^8dQhR\R\R\/#r_r`)rs"rrrs!F9rc$VPV4#rrrds&&r _resize_atenrrrz Gives a tensor with no elements a new shape, returning the modified tensor. The tensor's strides are contiguous and its values are uninitialized. z2resize(Tensor(a!) a, SymInt[] shape) -> Tensor(a!) output_dtypec\V\4'g\R\V4 24hVf VPp\ P !VPV4p\V\ P!V4VVPR7#)zP Meta function for single output reduction operations Stride logic is incorrect zinp must be TensorLike, got rrrr) rrrrrr8compute_reduction_output_shaperrrr)inprr output_shapes&&$ r_reduction_metarsu c: & &;DI;GHHyy 77 4HL 11,?zz  rc\P!VP4'd"\P!VP4pM VPp\ WVR7#))r)r8rCrrDr)rr correctionrs&&& r_var_reduction_metarsA cii((55cii@ yy 3< @@rzx Computes the sum of elements in the input tensor over the list of dimensions specified in the dim argument z| Computes the xor sum of elements in the input tensor over the list of dimensions specified in the dim argument z| Computes the product of elements in the input tensor over the list of dimensions specified in the dim argument z Computes the maximum value of elements in the input tensor over the list of dimensions specified in the dim argument z Computes the minimum value of elements in the input tensor over the list of dimensions specified in the dim argument ze Computes the biased variance of x over the list of dimensions specified in the dim argument c$V^8dQhR\/#rrr)rs"rrr ssrcL\V R2\V\PVR7#)Creates a reduction prim.zE(Tensor inp, int[]? dims, *, ScalarType? output_dtype=None) -> Tensorr)r#rrr^rrrs&&&r_make_reduction_primr s* \] OO   rc$V^8dQhR\/#rr)rs"rrr( s3rcL\V R2\V\PVR7#)rzZ(Tensor inp, int[]? dims, float? correction=1, *, ScalarType? output_dtype=None) -> Tensorr)r#rrr^rs&&&r_make_var_reduction_primr( s* qr OO   rrcxV^8dQhR\R\R,R\PR,R\/#rrrNrr5rrrrr )rs"rrr: sDHH H T !H ;;  H  Hrc\R4h)z&xor_sum only implemented with inductorre)rrrs&&$r _xor_sum_atenr: s F GGrcxV^8dQhR\R\R,R\PR,R\/#rr)rs"rrrJ s>,, , T !, ;;  ,  ,rcVed\V4^8XdVP4#\VRR7F0pV^8d\RV 24h\P !WVR7pK2 V#\P !WR7#)NT)reversez$dimension must be non-negative, got r)rrrrrr)rrrds&&$ r _prod_atenrJ sr   t9>99; d+A1u$'KA3%OPP**S51C, zz#++rc 8\P!V3RVRV/VB#)rr)rr)rsrrrs&&&,r torch_varrd s 99U E E Ef EErzC Constructs a 1-D tensor t where ``t[i] == start + i * step``. cV^8dQhR\R\R\R\PR\PR\R\ /#rrrsteprrrvr5rGrrrrr)rs"rrr sX     ;;  LL rc\P!\P!V4R4\P!V^8gR4\P!VVVVR7#)cR#)z'prims.iota only supports integer dtypesrrrrrR_iota_meta.. s9rcR#)zstep must be nonzerorrrrrRr s$:rrrrv)rrTr8r@r~)rrrrrrvs&$$$$$r _iota_metar sS LL u%9 LL:; ;;#  rcV^8dQhR\R\R\R\PR\PR\R\ /#rr)rs"rrr sX         ;;  LL   rc RWV,,p\P!WW#WER7#r)rarange)rrrrrrvrs&$$$$$ r _iota_atenr s( 4- C << Df rzpiota(SymInt length, *, SymInt start, SymInt step, ScalarType dtype, Device device, bool requires_grad) -> Tensorc |V^8dQhR\R\PR\PR\R\ /#rrrrrvr5)rrrrrr)rs"rrr sCPP P % P5:\\PRVPPrcH\P!V4p\WWR7#rr8rr)rrrrvrs&$$$ r _empty_metar s!//6G E% OOrc |V^8dQhR\R\PR\PR\R\ /#r)rrrrrr )rs"rrr sCWW W % W5:\\WRVW Wrc2\P!WW#R7#r)rr~)rrrrvs&$$$r _empty_atenr s ;;u& VVrz\ Creates a tensor with uninitialized values and the specified shape, dtype, and device. zWempty(SymInt[] shape, *, ScalarType dtype, Device device, bool requires_grad) -> Tensorc V^8dQhR\R\R\PR\PR\ R\ /#)rrrrrrvr5)rrrrrrr)rs"rrr sTPP P P ;; P LL P  PPrc\WW#R7#rr)rrrrrvs&&$$$r_empty_strided_metar s E% OOrz1 Creates a tensor with uninitialized values. zqempty_strided(SymInt[] shape, SymInt[] strides, *, ScalarType dtype, Device device, bool requires_grad) -> Tensorc V^8dQhR\R\R\PR\PR\ R\ /#)rrphysical_layoutrrrvr5)rrrrrrr)rs"rrr sL%% %%% ;; % LL %  %%rc&aaa a \P!SUu.uF qPV,NK up4p\V4o \P!\S4S 8HV V3Rl4^.\V4,p\ 4p\ S4Fowo o\P!^Su;8*;'dS 8MuV VV 3Rl4\P!SV9R4VS ,VS&VPS4Kq \VVVVR7#uupi)c&<RS R\S4 2#)zlNumber of dimensions in the tensor input does not match the length of the physical layout; i.e. len(size) = z( is not equal to len(physical_layout) = )r)rrsrrR&_empty_permuted_meta.. s$ ??BeD669/6J5K Mrc*<RS^, RS RS R2#)z5Dimension out of range (expected to be between 0 and z , but got z at index zL). NB: negative dims not currently supported; file an issue if you want it.r)rlr"srrRr s(GayPZ#Zs#IIrcR#)zDuplicate dim not allowedrrrrrRr s1Lrr) r8rrrrTrrr`r) rrrrrvr p_stridesr seen_dimsrr"s &f$$$` @@r_empty_permuted_metar s11_2U_88_2UVI e*C LL O# cCJGI/*1 LLSL    Qi')LMq\  a+   13VsDz Creates a tensor with uninitialized values according to some physical layout, that is guaranteed to be non-overlapping and dense. zwempty_permuted(SymInt[] shape, int[] physical_layout, *, ScalarType dtype, Device device, bool requires_grad) -> Tensorc V^8dQhR\R\R\PR\PR\ R\ /#rr fill_valuerrrvr5)rrrrrrr)rs"rrr sT P P  P P ;; P LL P  P PrcH\P!V4p\WW#R7#rr)rrrrrvrs&&$$$ r _full_metar  s!//6G E% OOrc V^8dQhR\R\R\PR\PR\ R\ /#r)rrrrrrr )rs"rrr( sL  ;;  LL    rc8\P!VVVVVR7#r)rfull)rrrrrvs&&$$$r _full_atenr( s$ :: #  rzi Creates a tensor filled with the given fill value, and with the specified shape, dtype, and device. zifull(SymInt[] shape, Scalar fill_value, *, ScalarType dtype, Device device, bool requires_grad) -> Tensorc V^8dQhR\R\R\PR\PR\ R\/#rrrrrrvr5)rrrrrr)rs"rrrH sT F F F F ;; F LL F  F Frc\P!V4pVP4^8XdVP4p\ WW#R7#)r)rrr)r8rwrrr)rrrrrvrs&&$$$ r_full_like_metarH s966q9GwwyA~((* a EErc V^8dQhR\R\R\PR\PR\ R\/#r)r rrrrr)rs"rrrW sL  ;;  LL    rc8\P!VVVVVR7#r)r full_like)rrrrrvs&&$$$r_full_like_atenrW s$ ?? #  rz Creates a tensor filled with the given fill value, and the same shape, dtype, and device as the given tensor by default. The dtype and device settings can be overridden by specifying them explicitly. zhfull_like(Tensor a, Scalar fill_value, *, ScalarType dtype, Device device, bool requires_grad) -> TensorcpV^8dQhR\R\PR\PR\/#rscalarrrr5)rrrrr)rs"rrrx s@XX X ;;X LL X  XrcN.p\P!V4p\WWAVR7#rr)rrrrrs&$$ r_scalar_tensor_metarx s) E//6G f7PV WWrcpV^8dQhR\R\PR\PR\/#r)rrrrr )rs"rrr s@ C C  C ;; C LL C  Crc\V\4'd,Ve\P!V4'g \ R4h\ P !WVR7#)Nz-Complex scalar requires complex tensor dtype.r)rcomplexr8rC TypeErrorrr)rrrs&$$r_scalar_tensor_atenr  sE &'"" U33E::GHH   v6 BBrzG Wraps a Number into a Tensor with the specified dtype and device. zQscalar_tensor(Scalar s, *, ScalarType? dtype=None, Device? device=None) -> Tensorc hV^8dQhR\R\R\\\\3,/#rA full_matricesr5)rrr)rs"rrr s0""")-" >>> 9:"rc\P!VR4\P!VPRRR7VPpVRRpVRRwrE\ WE4pY4V'dTMT3,p\P !VRR7p\WxVPVPR7p W63,p \P !V 4p \T T VP4'd!\P!VP4M VPVPR7p Y1'dTMTV3,p VPPR8Hp\P !WR7p\WVPVPR7pVP4^8wdKVP4'd5\PP4'dVP!4pWV3#)z linalg.svdF)allow_low_precision_dtypesN) row_majorrcuda)r8check_is_matrixcheck_fp_or_complexrrrrrrrnrDrrrr( is_availabler)r#r$A_shapebatchmnkshape_U strides_UUshape_S strides_SSshape_Vhis_cuda strides_VhVhs&$ r _svd_metar< sn !\* agg|PUVggG CRLE 23 (Tensor U, Tensor S, Tensor Vh) generatorcV^8dQhR\R\\,R\R\PR\P R\ R\PR,R \/# rrmeanstdrrrvrANr5) rrHrrrrr Generatorr)rs"rrr syPP P '/P  P ;; P LL PP%PPrc$aa\P!SR8V3Rl4\P!\P!S4;'g\P!S4V3Rl4\P !V4p\ WSVR7#)gc<RS 2#)z6expected non-negative standard deviation, but got std=r)rEsrrR_normal_meta.. s HNrc<RS 2#)z:expected a floating-point or complex dtype, but got dtype=rrsrrRrI s LUGTrr)rrTr8is_float_dtyperCrr)rrDrErrrvrArs&$dd$$$ r _normal_metarL sn LL s N  LL U#DDu'='=e'DT //6G E% OOrcV^8dQhR\R\\,R\R\PR\P R\ R\PR,R \/#rC) rrHrrrrrrFr )rs"rrr so    '/    ;;  LL   %   rc\P!WWER7p\P!4;_uu_4VPWVR7RRR4V# +'giT#;i)rrAN)rr~rCnormal_)rrDrErrrvrArs&$$$$$$ r _normal_atenrQ sJ  EvSA  $y 1  H  Hs A A% z Constructs a tensor filled with values drawn from a normal distribution with the specified mean and standard deviation. Only supports floating-point types. znormal(SymInt[] shape, *, Scalar mean, Scalar std, ScalarType dtype, Device device, bool requires_grad, Generator? generator=None) -> TensorcV^8dQhR\R\R\R\PR\PR\R\P R,R \ /# rrlowhighrrrrANr5)rrHrrrrFr)rs"rrr su O O  O  O  O ;; O LL O  O% O Orc\WW4R7#rr)rrTrUrrrrAs&$$$$$$r _uniform_metarW s E NNrcV^8dQhR\R\R\R\PR\PR\R\P R,R \ /#rS)rrHrrrrFr )rs"rrr+ sk           ;;  LL     %     rc\\P!WW4R7pVPWVR7V#))rrrrO)rruniform_)rrTrUrrrrArs&$$$$$$ r _uniform_atenr[+ s+ EMAJJsIJ. HrzN Constructs a tensor filled with values drawn uniformly from low to high. zuniform(SymInt[] shape, *, Scalar low, Scalar high, ScalarType dtype, Device device, SymInt[] stride, Generator? generator=None) -> TensorcHV^8dQhR\R\R\R\/#rrsronesidedr5rrrr)rs"rrrO s8VV V V V  Vrc\P!VPV4p\P!V4\ VP 4pV'd"VR,pW4,^,^,W4&\P !VP4p\P!V4p\W6WPPR7#rrr) r8rrvalidate_no_repeating_dimsr:rcorresponding_complex_dtyperrrr)rsrr^rlast_dimrrs&$$ r _fft_r2c_metareO s  ! !%**c 2C $$S)  Er7/Q.2  - -ekk :E//6G E% UUrcHV^8dQhR\R\R\R\/#r]r_)rs"rrrb s2?? ? ? ?  ?rc4^p\P!WW24#r&)r_fft_r2c)rsrr^ normalizations&$$ r _fft_r2c_atenrjb s M >>%m >>rz7 Performs a real to complex Fast Fourier Transform z;fft_r2c(Tensor self, *, int[] dim, bool onesided) -> TensorcHV^8dQhR\R\R\R\/#rrsrforwardr5r_)rs"rrrz s2         rc\P!VPV4p\P!V4VPp\P !V4p\ W4VPVPR7#r) r8rrrbrrrrr)rsrrmrrs&$$ r _fft_c2c_metaroz s]  ! !%**c 2C $$S) KKE//6G EKK  rcHV^8dQhR\R\R\R\/#rlr_)rs"rrr s2>> > > >  >rc4^p\P!WW24#r&)r_fft_c2c)rsrrmris&$$ r _fft_c2c_atenrs s M >>%m ==rz> Performs either a Fast Fourier Transform, or its inverse z:fft_c2c(Tensor self, *, int[] dim, bool forward) -> TensorcHV^8dQhR\R\R\R\/#rrsr last_dim_sizer5rrrGr)rs"rrr s8 V V  V  V V  VrcJ\P!VPV4p\P!V4\ VP 4pW#VR,&\P !VP4p\P!V4p\W5W@PR7#ra) r8rrrbr:rrDrrrr)rsrrvrrrs&$$ r _fft_c2r_metary sv  ! !%**c 2C $$S)  E"#b'N  * *5;; 7E//6G E% UUrcHV^8dQhR\R\R\R\/#rurw)rs"rrr s8DD D D D  Drc4^p\P!WW24#r&)r_fft_c2r)rsrrvris&$$ r _fft_c2r_atenr} s M >>%m CCrz? Performs a complex to real Inverse Fast Fourier Transform zBfft_c2r(Tensor self, *, int[] dim, SymInt last_dim_size) -> TensorcRV^8dQhR\R\\\3,/#)rselfr5)rr)rs"rrr s(MMnM~~/M)NMrc\P!VPPR4\P!V4\P!V\P R73#)cR#)z1torch.frexp() only supports floating-point dtypesrrrrrR_frexp_meta.. sCrr)rrTrrriint32)rs&r _frexp_metar sG LL $$C   D !5#3#3D #L LLrz8frexp(Tensor self) -> (Tensor mantissa, Tensor exponent)c$V^8dQhR\/#)rr5rt)rs"rrr s.rc\4#rr rrr_make_token_atenr s  rz_make_token() -> Tensorz7Creates a token used for keeping track of side effects.cV^8dQhRR/#)rr5Nr)rs"rrr s   rcR#rr)tokenss&r_sink_tokens_atenr srz#_sink_tokens(Tensor[] tokens) -> ()zTSink all of the tokens which were previously used for keeping track of side effects.rr)Nr(]rcollections.abcrrenumr functoolsrrtypingrr rtorch._prims_commonrqr8 torch.libraryr r torch._Cr torch._higher_order_ops.effectsrtorch._library.utilsrtorch._prims.debug_primsrtorch._prims.rng_primsrrrrrrrrrrrrrtorch._prims_common.wrappersrrr r!torch.overridesr"r#torch.utils._pytreer$r%r&rLibraryrrrrr__all__rr#r%rOrVr`rcrgr.r0r/r-r0r1r2r3r4r5r6specialr;r<i0r7i0er8i1r9i1er:r=rqr>r?rx_conj_physicalr^r@r}rrrArBerfinvrCrDrErFrGrHrrIrJrrKrLr/rMrNrOrPrQrTrUrRrSrVrzrWrXrYrZr[r\r]r^r_r`rarbrcrdrrerfrgrhrirkrlrmrngammaincro gammainccrprqrrrrsrrtrurvrwrxrybitwise_left_shiftr{bitwise_right_shiftr|r}r~rrr_as_strided_docrrr_broadcast_in_dim_docrrrrrr_collapse_view_docrr _conj_docrrr r_split_dim_docrr _squeeze_docrrr_transpose_docrr r$ _view_of_docrr+r0_view_element_type_docrr<_as_strided_scatter_docrr@rF _collapse_docrrTrW_cat_docrr]re _reshape_docrrj_rev_docfliprrn _where_docrrsrz_convert_element_type_docr pointwiserrr_device_put_docrr _item_docrrrr_maximum_value_docrrr_minimum_value_docrrr _copy_to_docINPLACErrr_copy_strided_docrrr _resize_docrrr_sum_doc _xor_sum_doc _prod_doc _amax_doc _amin_doc_var_docrrrrrrrrrrr _iota_docrrrrr _empty_docr~r_empty_strided_docrr_empty_permuted_docrr r _full_docr rr_full_like_docrrr _scalar_tensor_docrr<r@_svd_docrrLrQ _normal_docrrWr[ _uniform_docrrerj _fft_r2c_docrrors _fft_c2c_docrryr} _fft_c2r_docrrrjrrrNONErfxnodehas_side_effectrrrrs.%" ##(<595    ADEFF }}We, MM ! !'63N O  ==00&/R]]**7FJG&&w?p  p  p  p  p  p  p  p  p  p   !p "#p $%p &'p ()p *+p ,-p ./p 0 1p 2 3p 45p 67p 8 9p :;p < =p > ?p @ Ap B Cp D Ep F Gp H Ip J Kp LMp N Op P Qp R Sp T Up V Wp X Yp Z [p \ ]p ^_p ` ap b cp dep f gp h ip jkp l mp n op p qp r sp z {p | }p ~p @Ap BCp F Gp H Ip J Kp L Mp N Op P Qp R Sp T Up V Wp X Yp Z [p \]p ^ _p ` ap bcp dep f gp h ip jkp l mp nop p qp rsp tup vwp x yp z {p BCp DEp FGp H Ip JKp N Op PQp RSp TUp VWp XYp `ap hip j kp lmp n op v wp ~ p @Ap BCp D Ep FGp HIp JKp RSp T Up ^ _p ` ap b cp d ep fgp h ip pqp rsp tup v wp ~ p F Gp HIp PQp RSp TUp \]p ^_p f5K#5K"& 5K !% 5K )- 5K5Kpn(,n$)n+0nb4_AE_D3  # ii 7HH $ jj 7??  % kk 7??  $ jj 7??  % kk 7??  $ jj 7??  % kk 7??  # ii 7?? $ jj 7??  )mm%% 7??    )mm%% 7??    )hh 7??   *mm 7??   )mm 7??   *mm 7??  + 7??   8$  7?? $ jj 7?? . 1 "">   CHCXCX 0  N kk&"&   ' mm 7??  # ii 7??  ' mm"" 7??  $ mm   7??  % mm!! 7??  # ii 7??  % mm!! 7??  $ mm   7??  6 jj  % kk 7??   . &;LL   jj   (nn 7CC   & ll 7??  # ii 7??  % kk 7??  $ jj 7??  % kk 7??   . &;LL   jj  * 7??   % mm!! 7??  # ii 7??  % kk 7??   % kk 7??  $ jj 7??  ' mm 7??  # ii 7?? $ jj 7?? 3mm// 7?? $ jj 7?? # ii 7?? $ jj 7??  % kk 7??  $ ii 7??  & kk 7??  , 7??  + 7??  , 7??  '$  7?? #hh 7CC % jj 7?? % jj 7?? % jj 7?? $ ii 7?? #hh 7CC #hh 7CC  & kk 7??   ' mm$$ 7??   ( mm%% 7??  #hh 7CC #hh 7CC  (  7??   (  7??  $ ii 7?? #hh 7CC  *oo 7??   $ ii 7??  *oo 7??   +&& 7??  7'' 7??  # ii 7?? % mm   7?? =,=   j     T@n   c $     =&P"f D " H !       + jj   @@4D:  N      D.  @ mm    R$)  C      A   .      F %   /8  y !&& "  =   @#   5   XI   8   D   0 jj    = kk   7 "$" G #(! ))     C C33  S   '   2 % - ' & 6 !  ' & 6 !    *  : ##"&        >       ? ##  t$A           iiH!% HH   ,!% ,,$ F   jj  jj  ,   }  PPWW    d   PP  ~ !!  %%P  E ""  P P$   v   F F$  u    XX C C  ^ !  ""J<<    T +//;??C P)-PP0 )-  "   W   O)- O O  )-     O   VV&??  H      >>  G    V V DD  O   MM  E +//2kk   $ A     0   ^   l+r