CCj2SrSSKJr SSKrSSKrSSKrSSKrSSKrSSKrSSK r SSK r SSK r SSK r SSK r SSKrSSKrSSKrSSKJr SSKJr SSKJrJrJrJr SSKrSSKJr SSKrSSKJr SSK J!r! SS K"J#r# SS KJ$r$ SS K%J&r& SS K'J(r( SS K)J*r*J+r+ SSK,J-r- SSKJ.r.J/r/J0r0 \(aSSK1J2r2J3r3 SSK4J5r5 SSK6J7r7 \ Rp"\95r:\"S5r;\"S5r<\\>(aSSK?r@/rASrBSrC\>(a`/SQrA\@RRRR5RSS5rCSRS\A55rJS\JS3rB/SQrK"SS 5rLSDS!jrMS"rN"S#S$\O5rP"S%S&5rQ\RSDS'j5rSS(S).SES*jjrTS(S).SES+jjrUSDS,jrVSFS-jrW"S.S/\X5rYSGS0jrZSHSIS1jjr[SJS(S(S2.SKS3jjjr\SLS4jr]SMS5jr^SNS6jr_SJS(S(S2.SOS7jjjr`SPS8jraSQS9jrb\b"\R5rd\b"\R"S:55rf\b"S5rg\b"S(5rh\b"S(5ri"S;S<5rj"S=S>5rkSRS?jrl"S@SA5rmSSSTSBjjrnSUSCjrog)Va Debug utilities for TorchDynamo compilation and execution. This module provides various debugging tools and utilities for TorchDynamo, including: - Minification support for reducing test cases while preserving bugs - Input/output handling via InputReader and InputWriter for reproducible testing - Accuracy checking between original and compiled models - Neural network module string conversion via NNModuleToString - Profiling tools and system information collection - Buck build system integration for Meta-internal testing Key classes: - InputReader/InputWriter: Handle serialization of model inputs/outputs - NNModuleToString: Converts nn.Modules to string representations - BuckTargetWriter: Manages Buck build system integration ) annotationsN)Counter) import_module)Anycast TYPE_CHECKINGTypeVar)Tensor) rand_strided)normalize_path_separator)is_float_dtype)StorageWeakRef) ConfigModule)ContentStoreReaderContentStoreWriter)config) clone_inputs get_debug_dir warn_once)CallableSequence)tqdm)UntypedStorageTztorch._inductor.config)z1//caffe2/torch/fb/sparsenn:sparsenn_operators_gpuz-//caffe2/torch/fb/sparsenn:sparsenn_operatorsz///deeplearning/fbgemm/fbgemm_gpu:sparse_ops_cpuz+//deeplearning/fbgemm/fbgemm_gpu:sparse_opszfbcode:// c#.# UH nSUS3v M g7f)z "",N).0xs ڧC:\Users\johnn\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.13_qbz5n2kfra8p0\LocalCache\local-packages\Python313\site-packages\torch/_dynamo/debug_utils.py r%Ys C 152 zfor _extra_dep in [ zY ]: try: torch.ops.load_library(_extra_dep) except OSError: pass )buck2runz@mode/dev-nosanc6\rSrSrSSjrSSjrS S SjjrSrg) BuckTargetWriterfc,[RR[RRU55uUlUlUR R SS5UlURR SS5SUR3UlURURRS5SUlURSSUlURnX"RS5SSSnSUS UR3Ul g) Nz.pyr/.zfbcode.zfbcode/r:) ospathsplitabspathsubdirpy_filereplacetargetfind cmd_line_path)selffilenametmps r$__init__BuckTargetWriter.__init__gs$&GGMM"''//(2K$L! T\ll**5"5 {{**345Qt{{mD IIdiinnY79: IIabM kk((9%'(,!#a }5c SR[Vs/sH nSUS3PM sn5n[R"SURSUR S[ SUSURS 3 5$s snf) Nrz "r za load("@fbcode_macros//build_defs:python_binary.bzl", "python_binary") python_binary( name="z", srcs = ["z"], compile = False, deps = [ "//caffe2:torch", "//caffe2:libtorch", "//caffe2/functorch:functorch", "//triton:triton", "z", ], cpp_deps = [ z ], main_module = "z", par_style = "xar", ) )join extra_depstextwrapdedentr8r6 cur_targetr2)r;r#extra_cpp_depss r$buildBuckTargetWriter.buildusz#Jz!is"$5z#JK  ;;-ll^  II;#   $KsA3cd[RRURS5n[ US5nUR UR 55 SSS5 [UR/-nU(a%[RSSRU55 U$!,(df  NP=f)NTARGETSwzFFound an example that reproduces the error. Run this cmd to repro - %s ) r1r2rBr5openwriterHBUCK_CMD_PREFIXr:logwarning)r; print_msg target_filefd cmd_splits r$rOBuckTargetWriter.writesggll4;; : +s #r HHTZZ\ "$$t'9'9&::  KKX# $ #s B!! B/)r:r2r6r5r8N)r<strreturnNonerYrX)T)rSboolrY list[str])__name__ __module__ __qualname____firstlineno__r>rHrO__static_attributes__r!r@r$r*r*fs 6 4  r@r*c([RR[5S5nUc-[R "5S[ R"53n[RRU5(d[R"USS9 U$)Nminifierz /minifier_T)exist_ok) r1r2rBrtempfile gettempdirgetpassgetuserexistsmakedirs)r2s r$ minifier_dirrlse 77<<  4D |%%'( 7??3D2EF 77>>$   D4( Kr@c\rSrSrSrg)UnsupportedNNModuleErrorr!Nr^r_r`rarbr!r@r$roror@roc\rSrSrSS1r1SkrSS1rSr\RR\RR\RR\RR\RR\RR\RR \RR"\RR$\RR&\RR(\RR*\RR,\RR.\RR0\RR2\RR4\RR6\RR8\RR:\RR</r\ SSj5r!\ SS j5r"\ SS j5r#\ SS j5r$\ SS j5r%\ SS j5r&\ SSj5r'\ SSj5r(\ SS.SSjj5r)\ SS.SSjj5r*Sr+g)NNModuleToStringz0torch.ao.quantization.fake_quantize.FakeQuantizezAtorch.ao.quantization.fake_quantize.FusedMovingAvgObsFakeQuantize> #torch.ao.nn.qat.modules.conv.Conv1d#torch.ao.nn.qat.modules.conv.Conv2d#torch.ao.nn.qat.modules.conv.Conv3d5torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvBn1d5torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvBn2d5torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvBn3d7torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvReLU1d7torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvReLU2d7torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvReLU3d9torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvBnReLU1d9torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvBnReLU2d9torch.ao.nn.intrinsic.qat.modules.conv_fused.ConvBnReLU3dz%torch.ao.nn.qat.modules.linear.Linearz8torch.ao.nn.intrinsic.qat.modules.linear_relu.LinearReLU)fbgemmx86qnnpackonednnc>[5nUR5HYup#[U[RR 5(aM0[ RU5bMHURU5 M[ [U5S:a[RSU5 gg)Nrz2Cannot safely convert some modules to strings - %sFT) setnamed_children isinstancetorchfx GraphModulert_module_constructor_stringaddlenrQrR)gm cant_convert_modules r$can_convert_to_string&NNModuleToString.can_convert_to_stringsu **,IA&%(("6"677::6BJ  ( - | q KKDl r@c\[U5RS[U5R3$Nr.)typer_r^)rs r$_module_type_name"NNModuleToString._module_type_names)v,))*!DL,A,A+BCCr@c8URSUR3$r)r_r^)type_s r$ _type_nameNNModuleToString._type_names""#1U^^$455r@c[RU5nU[R;agURn[ [ UR 5n[ [ UR5n[USS5(a US-nUS-S-nS[R[U553SU<3SU<3SUR<3S UR<3S UR<3/n[US5(aURS UR <35 [US 5(aURS UR"<35 USSR%U5S3$)N reduce_rangeFrz observer=z quant_min=z quant_max=dtype=zqscheme=z is_dynamic=z reduce_range=ch_axiszch_axis=(, ))rtrfake_quant_modulesactivation_post_processrint quant_min quant_maxgetattrrrdtypeqscheme is_dynamichasattrappendrrrB)r module_typeobserverrrargss r$_fake_quant_constructor(NNModuleToString._fake_quant_constructorsJ&88@ .AA A11f../ f../ 8^U 3 3 NI!A )I(33DNCD E & &V\\$ %v~~( )&++. /   8^ , , KK-(=(='@A B 8Y ' ' KK(8#3#3"67 8a $022r@c[USS5n[RHNn[RR R U5n[U5[U5:XdMGSU<S3s $ g![a M`f=f)Nqconfigz6qconfig=torch.ao.quantization.get_default_qat_qconfig(r) rrtdefault_qat_qconfig_backendsrao quantizationget_default_qat_qconfigAssertionErrorrepr)rrbackenddefault_qconfigs r$_qat_qconfig_constructor)NNModuleToString._qat_qconfig_constructors&)T2'DDG "'(("7"7"O"OPW"XG}_ 55{!% E"  s)A11 A?>A?c [RU5nU[R;ag[RU5nUcg[ UR 5[ UR 5SUR<3SUR<3SUR<3SUR<3SUR<3SURSL3SUR<3/ n[USS5nUb>URS UR <3S UR"<3S UR$<3/5 UR'U5 US S R)U5S3$)Nz kernel_size=zstride=zpadding=z dilation=zgroups=bias=z padding_mode=bnzeps=z momentum=z freeze_bn=rrr)rtrqat_conv_modulesrr in_channels out_channels kernel_sizestridepaddingdilationgroupsbias padding_moderextendepsmomentum freeze_bnrrB)rrrrrs r$_qat_conv_constructor&NNModuleToString._qat_conv_constructor&sP&88@ .?? ?";;FC ? ## $ $$ %6--0 1fmm& 'v~~( )* +fmm& 'FKKt+, -F//2 3  VT4 ( > KK266*% / !1!1 45  Ga $022r@c,[RU5nU[R;ag[RU5nUcg[ UR 5[ UR 5SURSL3U/nUSSRU5S3$)Nrrrr) rtrqat_linear_modulesrr in_features out_featuresrrB)rrrrs r$_qat_linear_constructor(NNModuleToString._qat_linear_constructorFs&88@ .AA A";;FC ? ## $ $$ %FKKt+, -    a $022r@c[U5[R;ag[SU3SS5 g![a gf=f)NFzmod = z execT)rrt safe_reprscompile SyntaxError)r module_strs r$_can_emit_module_constructor-NNModuleToString._can_emit_module_constructorWsI </:: :  fZL)?F C  s 1 >>Fallow_unsafe_reprc[U5n[RX5(aU$[RU5=(d1 [R U5=(d [R U5nUbU$U(aU$gN)rrtrrrr)rrrconstructor_strs r$r+NNModuleToString._module_constructor_stringbsy&\  8 8 L L   4 4V < @55f= @77?   &" "  r@c^^^^^^^SSKJm Sm0m/mS UUU4SjjmS UUUU4SjjmST"U53nSSRTU/-5-$) Nr) _addindent c>[U5T;aT[U5$US[T53nUT[U5'T"U5nTRSUSU35 U$)Nrzclass z(torch.nn.Module): )idrr) name_hintsub class_namebodyemit_module_bodysubgraph_classes subgraph_defss r$register_subgraph3NNModuleToString.convert..register_subgraphss#w**'300%;a,<(='>?J(2 RW %#C(D  6*5I$!P Q r@c>TSTS-S3nUR5Hup#[U[RR5(aT"X#5nUTS-SUSUS3- nMI[ R UTS9nUc[SU<35e[UR5S5nUbUO[UR5S5nUb,URRS :waUS URS 3nUTS-SUSUS 3- nM URR5GHupU cM U R5[ ::aFS SKJn U R&[ :a [)SU R&S[ S35e[+U 5n Oh[R,"U 5(a'S[/U R05SU R2S3n O&S[/U R05SU R2S3n U RRS :waU S U RS 3n UTS-SUSU S3- nGM UR4R5HoupU cM SnU RRS :waSU RS3nS[/U R05SU R2US3n UTS-SU SU S 3- nMq UT"UR6[9T55S 3- nU$)Nzdef __init__(self) -> None: rzsuper().__init__() zself.z = z() rz!Cannot convert module to string: cpuz.to("z")rr) PRINT_OPTSzPRINT_OPTS.threshold (z() must be >= MAX_CONSTANT_NUMEL_INLINE (rz torch.randn(z, dtype=ztorch.randint(1, size=zself.register_buffer('z', z) rz , device=""ztorch.nn.Parameter(torch.randn(z)))rrrrrrtrronext parametersbuffersdevicer_buffersitemsnumelMAX_CONSTANT_NUMEL_INLINEtorch._tensor_strr thresholdrris_floating_pointlistshaper _parameterscoder)r model_str module_namechildrr example_paramexample_tensor buffer_namebufferr tensor_str param_nameparam maybe_devicerrrtabs r$r2NNModuleToString.convert..emit_module_bodys;%4S1WI=QR '-&;&;&=" eUXX%9%9::!2;!FJC!G9E+c*T!RRI-HH->I %2;E9E !%U%5%5%7 > %0"emmot4 "-.2G2G2L2LPU2U$.0(.'<'<'># ><<>%>><!++.GG,4Z5I5I4JK::S9TTUW"&fJ,,V44&tFLL'9&:(6<<.PQR$:$v||:L9MXV\VbVbUccd!eJ==%%.$.tEKK?P>QQYZ_ZeZeYfgsfttvw ayj\ZLKK &@ Jv{{CH=>bA AI r@zclass Repro(torch.nn.Module): z from torch.nn import * r)rrXrtorch.fx.GraphModulerYrXrtorch.nn.ModulerYrX)torch.nn.modules.modulerrB) rr repro_classrrrrrrs ` @@@@@@r$convertNNModuleToString.convertts_6,.#%   @ @ D88H8L7MN +dii  8U.VVVr@r!N)rrrYr\r)rz type[object]rYrX)rrrY str | None)rrrrXrYr\)rrrr\rYr)rrrr\rYrX),r^r_r`rarrrrrnnLinearConv1dConv2dConv3d BatchNorm1d BatchNorm2d BatchNorm3d LayerNormDropoutSoftmaxReLUGELUIdentity MaxPool2d EmbeddingTanhConvTranspose1dGLULSTMFlattenAdaptiveAvgPool2dr staticmethodrrrrrrrrrrrbr!r@r$rtrts%:K  0B$J                             ""+J0"DD66332  33>33 >C7; "GL]W]Wr@rtc[RR5(dgSn[RRcl[ R "SS/5nUR5RS5nSRUVs/sHo3S:wdM SUS3PM sn5nXS3- nOUS - n[S [R5555nUS - nUR5HupgUSUS US3- nM US- nU$s snf![[ R4a US - nNf=f)Nz:# torch.cuda.is_available()==False, no GPU info collected z# CUDA Info: nvccz --versionrrz# z z'# Not searching for nvcc on ROCM setup z# nvcc not found c3`# UH$n[RRU5v M& g7fr)rcudaget_device_name)r"is r$r%,_cuda_system_info_comment..s&/O! ""1%%/Os,.z# GPU Hardware Info: z : )rr: is_availableversionhip subprocess check_outputdecoder3rBOSErrorCalledProcessErrorrrange device_countr)r cuda_version_outcuda_version_linesscomment gpu_namesnamecounts r$_cuda_system_info_commentrOsJ :: " " $ $L!I * ==   $)66 7LM !1!8!8!:!@!@!F gg4FR4Fqr'{A3c{4FRSG 9B 'I C CI/4UZZ5L5L5N/OI**I ( r$s5'-- ) I S Z22 3*)) *s0A!D" D D D"/D"D""EEF) stable_outputc p^^^U(ag/SQm/SQmSS/mSUUU4Sjjn[RR5VVs/sH-up#U"U5(dMSUSURS S 5S 3PM/ nnnUR S T55 S R U5n[ S US35$s snnf)zd Generate a string configuration for environment variables related to Dynamo, Inductor, and Triton. z+# env var omitted due to stable_output=True)TORCHDYNAMOINDUCTORTRITON)TRITON_LIBDEVICE_PATHTRITON_PTXAS_PATHTRITON_LIBCUDA_PATHTORCHDYNAMO_REPRO_AFTERTORCHDYNAMO_REPRO_LEVELc`>^[U4SjT55=(a TT;=(a TT;$)Nc3,># UH oT;v M g7frr!)r"stringkeys r$r%;generate_env_vars_string..filter..s7J&# Js)any)r^ allow_listrepro_env_vars skip_lists`r$filter(generate_env_vars_string..filters0 7J7 7 *9$ *>) r@z os.environ['z'] = ''rc3.# UH nSUS3v M g7f)zos.environ.pop('z', None)Nr!)r"r^s r$r%+generate_env_vars_string.. sSNS*3%x8Nr&rz import os z )r^rXrYr\)r1environrr7rrBr ) rPrdr^value config_lines config_stringrarbrcs @@@r$generate_env_vars_stringrms<:JUI/1JKN  ****,,JC #; ALVEMM#s$;# ??r@ch[5n[RSU5 [(a[ U5R 5 [ US5nUR U5 SSS5 g!,(df  g=f![a)n[RS5 [SU35UeSnAff=f)NzWriting minified repro to: %srLrzCould not write to ) rrQrRuse_buckr*rOrNrD exceptionNotImplementedError)contentsminified_repro_pathrUes r$helper_for_dump_minifyr7s13KK02EFx,-335V %s +r HHX , + + V b!$78K7L"MNTUUVs6 A>A-$A>- A;7A>;A>> B1$B,,B1c\rSrSrSrg) AccuracyErroriFr!Nrqr!r@r$rrFrrr@rc[U5n[[U55HEn[X[R 5(dM&XR XR5 MG U$)z This clone inputs is different from utils clone_input. In case of minifier, all the tensors are leaf tensors while creating a new graph. So, we set the requires_grad field w/o checking the leafness of the tensor. )rrFrrrr requires_grad_ requires_grad)example_inputs cloned_inputsidxs r$clone_inputs_retaining_gradnessrJsX !0MS() m(%,, 7 7   - -n.A.O.O P* r@cXSSKJnJnJn [R "U5nU(d [ U5n[US5(aURS5 [USS5(aU"U5OU"U6nU(aU$U"U5(aU"U5nUR5 U"XSU5$)z Runs a forward and possibly backward iteration for a given mod and args. When disable_clone is True, we will use args as-is without cloning. This is higher fidelity but we may destroy the args in the process. r)collect_resultsreduce_to_scalar_lossrequires_bwd_pass zero_gradT _boxed_callFN) testingrrrcopydeepcopyrrrrbackward) rronly_fwd disable_clonerrroutlosss r$run_fwd_maybe_bwdrWsSR r B .t4r; Tb-77"T(RYC $S)  2D$ //r@ require_fp64 ignore_non_fpc SSKJn [XU5nSn[R(a8[ [ R"U5[U55up[XU5n[XU5n U"UU U[RSUS9n U $![a* U(a [S5e[RS5 N[f=f![a [RS5 gf=f) aI Check two models have same accuracy. require_fp64: if True, raise an error if we unable to calculate the fp64 reference ignore_non_fp: if True, do not compare outputs which are not floating point. This is mostly useful for the minifier (which wants to avoid quantizing floating point error into integer/boolean error) r)sameNzfCould not generate fp64 outputs, workaround with torch._dynamo.config.same_two_models_use_fp64 = FalsezCould not generate fp64 outputszWhile minifying the program in accuracy minification mode, ran into a runtime exception which is likely an unrelated issue. Skipping this graph.T)tol equal_nanr)utilsrrrsame_two_models_use_fp64 cast_to_fp64rrr Exception RuntimeErrorrQrRrrepro_tolerance) ropt_gmrrrrrreffp64_ref fp64_model fp64_examplesrespassings r$same_two_modelsrws" B 9CH && ;(4 b!#B>#R) %J)HMH A    " "# G N7 ;"| KK9 :  ;   $ s#7B # C 1B>=B>C#"C#cNURRGH_nURS:XaUR[R R RRLa[UR5S:wa![S[UR535e[URS5(aEURS[R:wa$URS[R4Ul URS:XdMURRS5nUcGM[U5(dGM1[!UR5n[RUS'X1lGMb URR#5 UR%5 U$)N call_functionrz"Expected node to have 2 args, got rrr)graphnodesopr8ropsprimsconvert_element_typedefaultrrrr float64kwargsgetdictlint recompile)modelnoder new_kwargss r$cast_dtype_args_to_fp64rs" !! GG & uyyCCKKK499~"$8TYY8HIdiil++ !  0M!YYq\5==9 77o %KKOOG,E ^E%:%:!$++. &+mm 7#( !"$ KK OO Lr@c^SSKJn URT5nT[R:Xa [ U5nU"U4SjU5nX4$)Nr)tree_mapc>[U[R5(a&UR5(aUR T5$U$r)rrr rto)r#rs r$cast_to..s< a & &1+>+>+@+@$$u+ r@)torch.utils._pytreerrrrr)rrinputsrs` r$cast_torsK- HHUOE  (.    F =r@c6[[RX5$r)rrr)rrs r$rrs 5==% 00r@c U"[R"U5[U55n[UUUUUUS9(+$![a [ R S5 gf=f)NrzWhile minifying the program in accuracy minification mode, ran into a runtime exception which is likely an unrelated issue. Skipping this graphF)rrrrrrQr)rr compiler_fnrrr compiled_gms r$backend_accuracy_failsrst! MM" >~N #    %'       # s8;AAc8UbU$[R"U5$r)rmake_contiguous_strides_for)rrs r$_stride_or_defaultr s '6UU-N-Nu-UUr@c^U4Sj$)Nc>UbU$T$rr!)r#ds r$r_mk_defaulter..s!-Q.Q.r@r!)rs`r$ _mk_defaulterrs ..r@rc\rSrSrS SjrSSS.SSjjrSSjrSSjrSSjrSS jr SS jr SS jr S r g)NopInputReadericSUlg)Nrtotal)r;s r$r>NopInputReader.__init__ s  r@Nr dtype_hintc.U=RS- slgNrr)r; storage_hashnbytesrrs r$storageNopInputReader.storage#s a r@cgrr!r;rrs r$tensorNopInputReader.tensor- r@cgrr!rs r$symintNopInputReader.symint0rr@cgrr!r;rMs r$constNopInputReader.const3rr@cgrr!rs r$ unsupportedNopInputReader.unsupported6rr@cgrr!)r; device_type device_indexs r$ generatorNopInputReader.generator9rr@c.U=RS- slgrrr;script_class_names r$opaqueNopInputReader.opaque<s a r@rrYrZ) rrrrr)torch._prims_common.DeviceLikeType | Nonertorch.dtype | NonerYrZ)rrrrrYztorch.Tensor | None)rrrrrY int | NonerMrXrYrZ)rrXrrrYrZrrXrYrZ r^r_r`rar>rrrrrrrrbr!r@r$rrsc=A)-   :  '       r@rc\rSrSrSSS.SSjjjrSSS.SSjjrSSSSSS.SSjjjrSS .SS jjrSS jrSS jr SS jr SSjr Sr g) InputReaderiBN)pbarcxUc[RS5 Ub [U5OSUl/UlX lg)Nz0no save_dir specified, will generate random data)rQrRrstorerr)r;save_dirrs r$r>InputReader.__init__Cs7   KKJ K5=5I'1t !  r@rcURbURRS5 [U5n[U5nURbPUbMURR U5nX5R :wa [RSX5R 5 U$[SUS35 X$R-4n[SUS9n[XgXC5R5$![a NNf=f)Nrzdevice mismatch: %s != %szcould not load z , generating random data insteadr)rupdate_device_or_default_dtype_or_defaultr read_storagerrQrRFileNotFoundErrorritemsizerr untyped_storage)r;rrrrrrrs r$rInputReader.storagePs 99 II  Q #F+&z2 :: !l&> **11,?^^+KK ;V^^TOL>1QRS...0#D6E:>NNPP%  sC C+*C+)storage_offsetrris_leafc [X2S9n[U5n[U5n[U5n[ U5n[ R "/XQRUS9n [ R"5 U RXX#5 SSS5 U(dk[ R"5 U R[ RS9n SSS5 [ R"5 U RXX#5 SSS5 [ RRRU 5U:wa8[!S[ RRRU 5SU35e[ R"R%X5 UR&R)U 5 U $!,(df  GN"=f!,(df  N=f!,(df  N=f)Nr)rrr) memory_formatz/Tensor leaf status mismatch: safe_is_leaf(t) = z , expected )r_storage_offset_or_defaultr_is_leaf_or_default_requires_grad_or_defaultrrrno_gradset_ enable_gradclonepreserve_format _subclasses meta_utils safe_is_leafr_utilsset_tensor_metadatarr) r;rrrrrrr metadatats r$rInputReader.tensormsM$F83NC!%(%g.1-@ LL eNN- ]]_ FF7E :""$GG%*?*?G@%w>!    ' ' 4 4Q 77 B A$$//<&%,,''H'' (8(899 <r@c\rSrSrSS.SSjjrSSjrSSS.SSjjrSS jrSS jrSS jr SS jr SS jr SSjr Sr g) InputWriteriF stable_hashc/Ul[R"5UlXlUb [ XS9OSUl0Ulg)NrP)_lines itertoolsrNstorage_counterrrr seen_storages)r;rrQs r$r>InputWriter.__init__sC!# (0 # x A 9;r@crS/nURSUR55 URS5 U$)Nzdef load_args(reader):c3,# UH nSU3v M g7f)rNr!)r"ls r$r%$InputWriter.lines..s1[4s[szload_args._version = 0)rrSr)r;rs r$linesInputWriter.liness8 $  1T[[11 )*r@N) device_hintrc ~[U5nURRU5nUbU$S[UR53nSn[ S5[ U5:waSU<3nSnUR nURS:XaUc [S5eUn[S5U:waSU<3nUR5n [U 5n Sn URb5UR RS:waURRU5n URRUSU <SU UUS 35 XPRU'U$) Nbufrz , dtype_hint=metaz:device_hint must be provided when storage device is 'meta'z , device=z = reader.storage(rr)rrVrrrUrrrrrrrMr write_storagerSr) r;rr_rwsvmaybe_dtype_hintrrr nbytes_sourcers r$rInputWriter.storagesXO ,    " "2 & =H$t++,- . T "&7 &C C!.zn=   '' ;;& "$P!F d #v -&vj1L '')1&9  :: !o&<&<&A&AV&K::33ODL c#L#32m_\NScRdde f "#2r@c SSKJnJn URUR 5UR UR S9n/nU"U"[SURS9UR555(d1UR[[UR5555 [S5UR :waURSUR <35 U"[S5UR5:H5(d#URSUR5<35 [ R"R%U5nU(a&UR'SUR)555 [+S5UR,:waURSUR,<35 [ R.R0R3U5n[5S5U:waURS U<35 UR6RS S R9U[[UR55/UQ5-S U3-5 g) Nr)statically_known_truesym_eq)rr_rrzstorage_offset=c36# UHupUSU<3v M g7f)=Nr!)r"kres r$r%%InputWriter.tensor..sI1H1#Qqe 1Hszrequires_grad=zis_leaf=zreader.tensor(r) # )rKrjrkrrrrrrrrrXtuplerr#rrr.get_tensor_metadatarrr%rr+r,r-r$rSrB) r;rMr1rjrkrrtensor_metadatar s r$rInputWriter.tensorsW,,   AGG $ %d!'':AHHJ G   KKE!((*-. / T "agg - KK& , -$ &t ,0@0@0B B   KK/!*:*:*<)?@ A,,::1=  KKI1F1F1HI I $T *aoo = KK.(;< =##..;;A> t $ / KK(7+. /  ii#eAGGn"5==> ?dVn  r@c,URRSU<S[U535 [U[[ 45(aURRS5 [ U5Hup4USUS3n[U[R5(aURXT5 M?[U[[R45(aURXT5 MwURXT5 M URRS5 gg)Nzreader.unsupported(z#) # unsupported type for dumping: z"""[])rSrrrrrq enumeraterr rrrJrr)r;rMargr<aname_is r$rInputWriter.unsupported s !$)LTRUYK X cD%= ) ) KK  u %!# 61#Qa..KK*C#677KK*$$V/' KK  u % *r@cJURRSU<SUS35 g)Nz reader.const(rpz!, filtered out during compilationrSrrs r$rInputWriter.const4s' D85.O P r@c>[U[R5(a\[URR 5nURR nURRSU<SU<SU35 gURRSU<SU35 g)Nzreader.symint(z, expr=rp) rrrJrXrr3hintrSr)r;rMr6expr_strrs r$rInputWriter.symint:sw c5<< ( (388==)H88==D KK  xwxl%PTvV W KK  weD6B Cr@cURnURRSUR<SUR<SU35 g)Nzreader.generator(rrp)rrSrrindex)r;rMryrs r$rInputWriter.generatorBs<  b0@dV L r@cHURRSU<SU35 g)Nzreader.opaque(rpr~)r;rMrs r$rInputWriter.opaqueHs# ^,=+@dVLMr@)rSrrVrUr)rrrQr\rYrZ)rYr])rrr_rrr rYrX)rMrXr1rGrYrZ)rMrXryrrYrZr )rMrXr6rrYrZ)rMrXryztorch._C.GeneratorrYrZ)rMrXrrXrYrZ)r^r_r`rar>r]rrrrrrrrbr!r@r$rOrOsfDI ;&BF)- $'$? $ ' $ $L @&( D Nr@rOc ^^^^SSKJn UR5VVs0sHupVXe_M nnnSRUR 55n[ R "U5n SUS3n SUS3n Sn "S S 5n 0nU=(d 0mSUU4S jjmSUU4S jjnURnUR"5HunnUS :XaM[R"U U5nU(a>UR5unn[URS55nUUnU"UU5UU'[R"U U5nU(dMT"URS55UU'M S[ R"U5R;aqU "5nUUS'[R "X5HKnUR5unnnn[URS55nUUn[#UUU"UU55 MM U$s snnf)a Takes in a function which has been printed with print_readable() and constructs kwargs to run it. Handles Tensor inputs, Symints, and a graph module which might have tensor constants. Consider a function `forward` defined as follows: def forward(self, primals_1: "f32[1001, 6]", primals_2: "f32[s0]", primals_3: "Sym(s0)",): _tensor_constant0: "i64[4190]" = self._tensor_constant0 # Further implementation kwargs = aot_graph_input_parser(forward) forward(**kwargs) r) dtype_abbrs|z(_tensor_constant\d+): \"(z0)\[\s*(.*?)\s*\]\" = self\.(_tensor_constant\d+)rz)\[\s*(.*?)\s*\]z Sym\((s\d+)\)c\rSrSrSrSrg)/aot_graph_input_parser..TensorContaineripz#Container for tensors as attributesr!N)r^r_r`ra__doc__rbr!r@r$TensorContainerrps-r@rrYc|>^[R"TT;=(d TSLU4Sj5 TRTT5$)Nc>TS3$)Nz; not in symbolic_shapes and default sym shape not passed inr!)rsr$r=aot_graph_input_parser..get_sym_int..{s vhYZr@)r_checkr)rdefault_sym_shapesym_shapes_dicts`r$ get_sym_int+aot_graph_input_parser..get_sym_intxs; o % F):$)F Z ""6+<==r@c>/n/n[U5HjupEUR5nSU;a,T "U5nURU5 URU5 MGU(dMPUR[U55 Ml UR(a[ R O[ RnU"X!T S9nUH"n [ RRX5 M$ U$)NrJ)rr) rxstriprrrrrandnzerosrx mark_dynamic) rrresolved_shape dynamic_dimsr<dimrJ constructorrrrrs r$ gen_tensor*aot_graph_input_parser..gen_tensors &FA))+Ccz$%%a(##A&3"))#c(3'&+%<%>*&&K(..0z H   , 3 #(<<> Iy)//#./Ei(E&ue4F5M /:6 5' A7F5M1"""4(333#% "v[[!8AE16 .Iy)Q)//#./Ei(E Iy*UE*B C B M[)sG7c^^[R"5m[RR [RR T55mSU4SjjnSUU4Sjjn[ R"U5 U$)z Decorator to cProfile a given function and save the result to disk on process exit. Args: filename: filename to save profile to cN>^[R"T5SUU4Sjj5nU$)Nc>TR5 T"U0UD6TR5 $!TR5 f=fr)enabledisable)rrfnprofs r$wrapper3profile_to_file..decorator..wrappers1 KKM 4*6*  s+=)rrrrrYr) functoolswraps)rrrs` r$ decorator"profile_to_file..decorators%     r@c >TRT5 [RR[R "STSTS355 g)Nz! Wrote profile to z+, view with: snakeviz z ) dump_statssysstderrrOrDrE)r<rsr$save_it profile_to_file..save_itsK !  OO""*,&Z(  r@)rrrYrr)cProfileProfiler1r2r4 expanduseratexitregister)r<rrrs` @r$profile_to_filersW    Dwwrww11(;rs$#   #44 ##.@.;3M992,! CL 89  $ $ &     J $$..==?GG SWXJ yy C CC  012 644n ~ eWeWP  67< @5:D@ V I    00 00 0  0H 7 77 7"7 7  77 7t0 3=F+(1 1)21+1  !S    RV 1V )V$ V/"%--0"5<<#67*1-)%0#E*FeehKNKN`(,$( d 0d d&d" d  dN#r@