aws-neuron/Qwen3-1.7B-TP4-BS4-SEQ2048
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: aws-neuron/Qwen3-1.7B-TP4-BS4-SEQ2048
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-08-30 21:56:23 +08:00
commit fd57cdfd17
114 changed files with 237628 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
token_generation_model/_tp0_bk0/command.txt Normal file
View File

@@ -0,0 +1 @@
neuronx-cc compile --framework=XLA model.MODULE_5e3d0ce8963512f9ea2d+937cd7a2.hlo_module.pb --output model.MODULE_5e3d0ce8963512f9ea2d+937cd7a2.neff --target=trn2 --auto-cast=none --model-type=transformer '--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ' --lnc=2 -O2 --internal-hlo2tensorizer-options=--verify-hlo=true --logfile=log-neuron-cc.txt --enable-internal-neff-wrapper --verbose=35

1
token_generation_model/_tp0_bk0/compile_flags.MODULE_5e3d0ce8963512f9ea2d+937cd7a2.json Normal file
View File

@@ -0,0 +1 @@
["--target=trn2", "--auto-cast=none", "--model-type=transformer", "--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ", "--lnc=2", "-O2", "--internal-hlo2tensorizer-options=--verify-hlo=true", "--logfile=/models/Qwen3-1.7B-compiled/token_generation_model/_tp0_bk0/log-neuron-cc.txt", "--enable-internal-neff-wrapper"]

590
token_generation_model/_tp0_bk0/global_metric_store.json Normal file
View File

@@ -0,0 +1,590 @@
{
"Average": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 96.97119140625,
"StaticProfiler::AveragePartitionUtilization": 88.53791809082031,
"StaticProfiler::AveragePeUtilization": 81.30671691894531,
"StaticProfiler::LocalizationEfficiency": 161.8649139404297,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 167.2097930908203,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0
}
},
"Count": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 1,
"StaticProfiler::AveragePartitionUtilization": 1,
"StaticProfiler::AveragePeUtilization": 1,
"StaticProfiler::LocalizationEfficiency": 1,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 1,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 1,
"TilingProfiler::AveragePeUtilizationAfterTiling": 1
}
},
"Sum": {
"compiletime": {
"AGOrderingAnalysisPass": 2.377948760986328,
"AffinePredicateResolution": 0.03525829315185547,
"AliasDependencyElimination": 0.0024094581604003906,
"AliasDependencyInduction": 0.2979590892791748,
"AliasDependencyReset": 0.30464816093444824,
"BFComputeCutting": 0.07316970825195313,
"BirCodeGenLoop": 2.482311487197876,
"CCOpFusion": 0.5325038433074951,
"CanonicalizeConv": 0.00023499999952036887,
"CanonicalizeDAGForPGTiling": 0.1513075828552246,
"CanonicalizeForTensorizer": 0.00026500000967644155,
"CanonicalizeIR": 0.04851794242858887,
"Canonicalizer": 0.004513999912887812,
"CoalesceCCOp": 0.16019058227539063,
"CommuteConcat": 0.024104833602905273,
"DMALocalityOpt": 0.037362098693847656,
"DMAProfiler": 0.07356810569763184,
"DMATilingProfiler": 0.09021782875061035,
"DataLocalityOpt": 3.034395694732666,
"DataStreaming": 0.1286299228668213,
"DeConcat": 0.04242873191833496,
"DeadCodeElimination": 0.02465653419494629,
"DeadStoreElimination": 0.8050529956817627,
"DelinearIndices": 0.46034955978393555,
"Delinearization": 0.11224937438964844,
"DelinearizeSPMD": 0.14354729652404785,
"DoNothing": 0.00033402442932128906,
"DramToDramTranspose": 0.2800898551940918,
"DumpGraphAndMetadata": 0.14458084106445313,
"EliminateDivs": 0.11128425598144531,
"ExpandBatchNorm": 0.04912686347961426,
"ExpandISAMacro": 0.07862186431884766,
"FactorizeBlkDims": 0.49788737297058105,
"FactorizeThreadAxesInFreeDims": 0.05206179618835449,
"FlattenMacroLoop": 0.08013129234313965,
"GenericAccessSimplifier": 0.022362232208251953,
"HoistCompute": 3.899999865097925e-05,
"IdentifyCrossPassTensors": 0.00022899999748915434,
"InferInitValue": 1.3044743537902832,
"InferIntrinsicOnCC": 0.26564502716064453,
"InferNeuronTensor": 1.4273626804351807,
"InferNonlocalTensors": 3.070617198944092,
"InferPSumTensor": 1.877610206604004,
"InferShardAxis": 4.046135902404785,
"InferSharedMemLoc": 0.10418868064880371,
"InlineNativeKernels": 0.046288251876831055,
"InsertCoreBarrier": 0.12113451957702637,
"InsertIOTransposes": 1.5809462070465088,
"InsertImplicitShardAxisBeforeISel": 0.37781310081481934,
"InsertLocalTransposes": 0.8543226718902588,
"InsertOffloadedTransposes": 0.0779869556427002,
"LICM": 0.10844612121582031,
"LateLegalizeInst": 0.1468517780303955,
"LateLegalizePostSplit": 0.08665323257446289,
"LateLowerReshapeOp": 0.02942204475402832,
"LateLowerTensorOp": 0.22415471076965332,
"LateNeuronInstComb": 0.9669840335845947,
"LayoutPreprocessing": 0.7032866477966309,
"LayoutPreprocessingAndAnalysis": 1.1142585277557373,
"LayoutRequirementAnalysis": 0.40787601470947266,
"LegalizeCCOpLayout": 0.052059173583984375,
"LegalizeOpLevelAlias": 0.022985219955444336,
"LegalizePartitionReduce": 0.036779165267944336,
"LegalizeSundaAccess": 0.9409115314483643,
"LegalizeSundaMacro": 0.7163646221160889,
"LegalizeType": 0.1376960277557373,
"LocalLayoutOpt": 0.5141463279724121,
"LoopFusion": 0.2369997501373291,
"LoopSplitting": 0.026223182678222656,
"LowerBroadcast": 0.04980111122131348,
"LowerCCOpBlockAxis": 0.17850804328918457,
"LowerComplexBroadcast": 0.06136465072631836,
"LowerIntrinsics": 0.9023723602294922,
"LowerShardAxis": 0.19546747207641602,
"LowerTensorOp": 0.38861823081970215,
"LowerToSendRecv": 0.1479358673095703,
"LowerTranspose": 0.41634607315063477,
"MacroGeneration": 1.907278060913086,
"MaskPropagation": 0.10989689826965332,
"MemcastMotion": 0.00011899999663000926,
"MemcpyElimination": 3.3127715587615967,
"MutateDataType": 0.031205177307128906,
"NeuronAliasDependencyInduction": 0.014397859573364258,
"NeuronAliasDependencyReset": 0.018595218658447266,
"NeuronInstComb": 0.29677605628967285,
"NeuronLICM": 0.2591841220855713,
"NeuronLoopFusion": 0.9968361854553223,
"NeuronLoopInterchange": 0.049851179122924805,
"NeuronSimplifier": 0.4482152462005615,
"NeuronSimplifyPredicates": 0.13344645500183105,
"NeuronValueNumbering": 0.10327720642089844,
"OptimizeAliasedCopyChain": 0.010796785354614258,
"OptimizeNKIKernels": 0.9786763191223145,
"PAGLayoutOpt": 11.095458984375,
"PComputeCutting": 0.28275370597839355,
"PGLayoutTilingPipeline": 27.67894172668457,
"PGTiling": 4.957591533660889,
"PadElimination": 0.009732246398925781,
"ParAxesAnnotation": 10.236833572387695,
"PartialLoopFusion": 1.0972201824188232,
"PartialSimdFusion": 0.529569149017334,
"PenguinizeFunctions": 0.00012799999967683107,
"PerfectLoopNest": 0.059952735900878906,
"PruneFunctions": 0.000391999987186864,
"RecognizeOpIdiom": 0.120391845703125,
"Recompute": 0.005687713623046875,
"RelaxPredicates": 0.09513998031616211,
"Rematerialization": 0.14099407196044922,
"RemoveOptimizationBarriers": 0.0004239999980200082,
"RemoveShardedPartitionAxes": 0.6145329475402832,
"ReshapeWeights": 0.0214080810546875,
"ResolveAccessConflict": 0.1656482219696045,
"ResolveComplicatePredicates": 0.03499007225036621,
"RewriteReplicationMatmul": 0.03796839714050293,
"RewriteWeights": 0.06430387496948242,
"SFKVectorizer": 5.783308029174805,
"ScatterMotion": 0.002827000105753541,
"ShardingPropagationAnalysis": 0.5885381698608398,
"SimpleAllReduceTiling": 0.05830836296081543,
"Simplifier": 0.07847452163696289,
"SimplifyMacroPredicates": 0.25547003746032715,
"SimplifyNeuronTensor": 0.3350214958190918,
"SimplifySlice": 0.02300572395324707,
"SimplifyTensor": 0.2370157241821289,
"SpillPSum": 0.4782223701477051,
"SplitAPUnionSets": 0.6796882152557373,
"SplitAccGrp": 0.03665971755981445,
"StaticProfiler": 0.13864398002624512,
"StaticTransposeLocalTensor": 0.2623753547668457,
"SundaISel": 1.3397884368896484,
"TCTransform": 0.025269269943237305,
"TensorInitialization": 0.16660237312316895,
"TensorOpSimplifier": 0.15259718894958496,
"TensorOpTransform": 0.8282039165496826,
"TensorizerLegalizationPass": 0.00011800000356743112,
"TileCCOps": 0.3066561222076416,
"TilingProfiler": 0.33243393898010254,
"TransformConvOp": 0.05726790428161621,
"TritiumFusion": 0.13151931762695313,
"ValueNumbering": 0.06717801094055176,
"VectorizeDMA": 0.5187788009643555,
"VectorizeMatMult": 0.04601097106933594,
"VerifySupportedOps": 0.000195999993593432,
"WeightCoalescing": 0.05271005630493164,
"ZeroSizeTensorElimination": 0.0003685951232910156,
"algsimp": 0.0011439999798312783,
"batchnorm_expander": 0.0005740000051446259,
"boundary-marker-removal": 0.00017699999443721026,
"call-inliner": 0.00013800000306218863,
"canonicalize-boundary-marker": 0.0003129999968223274,
"collective-stream-id-checker": 7.899999764049426e-05,
"comparison-expander": 0.000195999993593432,
"computation-deduplicator": 0.00031900001340545714,
"config-lowering": 0.00015100000018719584,
"constant_folding": 9.40000027185306e-05,
"cse": 0.00038400001358240843,
"dce": 2.4000000848900527e-05,
"dynamic-slice-transpose": 8.099999831756577e-05,
"eliminate-redundant-compare": 7.999999797903001e-05,
"emit-offloaded-dropout": 0.00013899999612476677,
"flatten-call-graph": 0.0001720000000204891,
"fuse-send-recv": 0.0008440000237897038,
"hilo-conditional-to-select": 5.2999999752501026e-05,
"hilo::LegalizeAlias": 0.0019950000569224358,
"hilo::NeuronInstCombine": 0.0008609999786131084,
"hilo::NeuronOpFusion": 0.0003239999932702631,
"hilo::ReplaceTokenTypeWithU8Pass": 0.00031400000443682075,
"hilo::ScheduleFusion": 2.9000000722589903e-05,
"hilo::SixtyFourHack": 0.00023200000578071922,
"hilo::VerifyAliasing": 5.0999999075429514e-05,
"hlo-mac-count": 0.006806999910622835,
"io-con-pipe-begin": 9.999999747378752e-06,
"io-con-pipe-end": 0.0,
"io-layout-normalization": 0.0014359999913722277,
"legalize-ccops-for-tensorizer": 1.4999999621068127e-05,
"legalize-compare": 0.000155999994603917,
"lower-argminmax-custom-call": 7.599999662488699e-05,
"map-inline": 0.0003760000108741224,
"metadata-naming": 0.0007949999999254942,
"mlir::detail::OpToOpPassAdaptor": 0.00014600000577047467,
"mlir::hlo::MhloToPyPenguin": 0.057739999145269394,
"mlir::mhlo::LowerComplexExtraPass": 0.0017610000213608146,
"mlir::mhlo::LowerComplexPass": 0.00240899994969368,
"native-to-custom-softmax": 0.00022000000171829015,
"native-to-custom-softmax-dx": 0.00023200000578071922,
"neuron-hlo-verifier": 0.01620600000023842,
"operand_upcaster": 0.0005629999795928597,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 0.0,
"post-partition-simplification": 0.03471200168132782,
"pre-hlo-begin": 3.000000106112566e-06,
"pre-hlo-end": 0.0,
"replace-minimum-constant": 0.00010599999950500205,
"reshape-mover": 4.5000000682193786e-05,
"simplify-concat": 0.0008580000139772892,
"simplify-while-loops": 3.300000025774352e-05,
"transform-variadic-reduce": 0.00025599999935366213,
"tuple-simplifier": 8.900000102585182e-05,
"unpack-nested-aws-ntwsr": 0.00014000000373926014,
"unroll-while-loop": 6.000000212225132e-06
},
"hilo": {
"HloMacCount": 1751965696.0,
"Traffic": 1252616320.0
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 41972,
"StaticProfiler::AifUb": 17.11549949645996,
"StaticProfiler::ArithmeticIntensityTensorizer": 27.70398712158203,
"StaticProfiler::AverageDmaLength": 2723.7119140625,
"StaticProfiler::DDRTransferBytes": 918042224,
"StaticProfiler::InternalTransferBytes": 174444896,
"StaticProfiler::LoadExpanded": 238040,
"StaticProfiler::StoreExpanded": 18609,
"StaticProfiler::TotalDMAExpanded": 256649,
"StaticProfiler::TotalDynamicInstancesCount": 57730,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 50535,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 179,
"TilingProfiler::MatMultInstructionsAfterTiling": 30304,
"TilingProfiler::NumPfTransposes": 348,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 198,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 7835,
"TilingProfiler::PfTransposeInstructionsForIo": 5217,
"TilingProfiler::PfTransposeInstructionsForLocal": 508,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 2110,
"TilingProfiler::ReduceInstructionsAfterTiling": 59,
"TilingProfiler::SimdInstructionsAfterTiling": 2266,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"all": {
"compiletime": {
"CanonicalizeConv": 0.00023499999952036887,
"CanonicalizeForTensorizer": 0.00026500000967644155,
"Canonicalizer": 0.004513999912887812,
"HoistCompute": 3.899999865097925e-05,
"IdentifyCrossPassTensors": 0.00022899999748915434,
"MemcastMotion": 0.00011899999663000926,
"PenguinizeFunctions": 0.00012799999967683107,
"PruneFunctions": 0.000391999987186864,
"RemoveOptimizationBarriers": 0.0004239999980200082,
"ScatterMotion": 0.002827000105753541,
"TensorizerLegalizationPass": 0.00011800000356743112,
"VerifySupportedOps": 0.000195999993593432,
"algsimp": 0.0011439999798312783,
"batchnorm_expander": 0.0005740000051446259,
"boundary-marker-removal": 0.00017699999443721026,
"call-inliner": 0.00013800000306218863,
"canonicalize-boundary-marker": 0.0003129999968223274,
"collective-stream-id-checker": 7.899999764049426e-05,
"comparison-expander": 0.000195999993593432,
"computation-deduplicator": 0.00031900001340545714,
"config-lowering": 0.00015100000018719584,
"constant_folding": 9.40000027185306e-05,
"cse": 0.00038400001358240843,
"dce": 2.4000000848900527e-05,
"dynamic-slice-transpose": 8.099999831756577e-05,
"eliminate-redundant-compare": 7.999999797903001e-05,
"emit-offloaded-dropout": 0.00013899999612476677,
"flatten-call-graph": 0.0001720000000204891,
"fuse-send-recv": 0.0008440000237897038,
"hilo-conditional-to-select": 5.2999999752501026e-05,
"hilo::LegalizeAlias": 0.0019950000569224358,
"hilo::NeuronInstCombine": 0.0008609999786131084,
"hilo::NeuronOpFusion": 0.0003239999932702631,
"hilo::ReplaceTokenTypeWithU8Pass": 0.00031400000443682075,
"hilo::ScheduleFusion": 2.9000000722589903e-05,
"hilo::SixtyFourHack": 0.00023200000578071922,
"hilo::VerifyAliasing": 5.0999999075429514e-05,
"hlo-mac-count": 0.006806999910622835,
"io-con-pipe-begin": 9.999999747378752e-06,
"io-con-pipe-end": 0.0,
"io-layout-normalization": 0.0014359999913722277,
"legalize-ccops-for-tensorizer": 1.4999999621068127e-05,
"legalize-compare": 0.000155999994603917,
"lower-argminmax-custom-call": 7.599999662488699e-05,
"map-inline": 0.0003760000108741224,
"metadata-naming": 0.0007949999999254942,
"mlir::detail::OpToOpPassAdaptor": 0.00014600000577047467,
"mlir::hlo::MhloToPyPenguin": 0.057739999145269394,
"mlir::mhlo::LowerComplexExtraPass": 0.0017610000213608146,
"mlir::mhlo::LowerComplexPass": 0.00240899994969368,
"native-to-custom-softmax": 0.00022000000171829015,
"native-to-custom-softmax-dx": 0.00023200000578071922,
"neuron-hlo-verifier": 0.01620600000023842,
"operand_upcaster": 0.0005629999795928597,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 0.0,
"post-partition-simplification": 0.03471200168132782,
"pre-hlo-begin": 3.000000106112566e-06,
"pre-hlo-end": 0.0,
"replace-minimum-constant": 0.00010599999950500205,
"reshape-mover": 4.5000000682193786e-05,
"simplify-concat": 0.0008580000139772892,
"simplify-while-loops": 3.300000025774352e-05,
"transform-variadic-reduce": 0.00025599999935366213,
"tuple-simplifier": 8.900000102585182e-05,
"unpack-nested-aws-ntwsr": 0.00014000000373926014,
"unroll-while-loop": 6.000000212225132e-06
}
},
"cumsum": {
"compiletime": {
"CoalesceCCOp": 0.0002205371856689453,
"DMALocalityOpt": 0.00016808509826660156,
"DMAProfiler": 0.0006198883056640625,
"DataStreaming": 0.0002627372741699219,
"DoNothing": 0.00013875961303710938,
"ExpandISAMacro": 0.0005130767822265625,
"FactorizeBlkDims": 0.0005578994750976563,
"InferPSumTensor": 0.0005154609680175781,
"InferSharedMemLoc": 0.00028228759765625,
"InsertCoreBarrier": 0.00024247169494628906,
"LateLegalizeInst": 0.0003752708435058594,
"LateNeuronInstComb": 0.0005624294281005859,
"LegalizeSundaAccess": 0.0013544559478759766,
"LegalizeType": 0.00024437904357910156,
"LowerBroadcast": 0.00022649765014648438,
"LowerIntrinsics": 0.00021767616271972656,
"LowerTranspose": 0.0002415180206298828,
"NeuronInstComb": 0.0007607936859130859,
"NeuronLICM": 0.00041222572326660156,
"NeuronSimplifyPredicates": 0.002053976058959961,
"NeuronValueNumbering": 0.0004031658172607422,
"SFKVectorizer": 0.002403736114501953,
"SimpleAllReduceTiling": 0.0002071857452392578,
"SimplifyNeuronTensor": 0.0004944801330566406,
"SpillPSum": 0.0004985332489013672,
"WeightCoalescing": 0.000213623046875
}
},
"sg00": {
"hilo": {
"ArithmeticIntensity": 2.797290325164795,
"HloMacCount": 1751965696.0,
"Traffic": 1252616320.0
}
},
"sg0000": {
"compiletime": {
"AGOrderingAnalysisPass": 2.377948760986328,
"AffinePredicateResolution": 0.03525829315185547,
"AliasDependencyElimination": 0.0024094581604003906,
"AliasDependencyInduction": 0.2979590892791748,
"AliasDependencyReset": 0.30464816093444824,
"BFComputeCutting": 0.07316970825195313,
"BirCodeGenLoop": 2.482311487197876,
"CCOpFusion": 0.5325038433074951,
"CanonicalizeDAGForPGTiling": 0.1513075828552246,
"CanonicalizeIR": 0.04851794242858887,
"CoalesceCCOp": 0.1563706398010254,
"CommuteConcat": 0.024104833602905273,
"DMALocalityOpt": 0.03418087959289551,
"DMAProfiler": 0.06951689720153809,
"DMATilingProfiler": 0.09021782875061035,
"DataLocalityOpt": 3.034395694732666,
"DataStreaming": 0.12200498580932617,
"DeConcat": 0.04242873191833496,
"DeadCodeElimination": 0.02465653419494629,
"DeadStoreElimination": 0.8050529956817627,
"DelinearIndices": 0.46034955978393555,
"Delinearization": 0.11224937438964844,
"DelinearizeSPMD": 0.14354729652404785,
"DoNothing": 6.198883056640625e-05,
"DramToDramTranspose": 0.2800898551940918,
"DumpGraphAndMetadata": 0.14458084106445313,
"EliminateDivs": 0.11128425598144531,
"ExpandBatchNorm": 0.04912686347961426,
"ExpandISAMacro": 0.07448863983154297,
"FactorizeBlkDims": 0.48647499084472656,
"FactorizeThreadAxesInFreeDims": 0.05206179618835449,
"FlattenMacroLoop": 0.08013129234313965,
"GenericAccessSimplifier": 0.022362232208251953,
"InferInitValue": 1.3044743537902832,
"InferIntrinsicOnCC": 0.26564502716064453,
"InferNeuronTensor": 1.4273626804351807,
"InferNonlocalTensors": 3.070617198944092,
"InferPSumTensor": 1.8667116165161133,
"InferShardAxis": 4.046135902404785,
"InferSharedMemLoc": 0.10098028182983398,
"InlineNativeKernels": 0.046288251876831055,
"InsertCoreBarrier": 0.1175682544708252,
"InsertIOTransposes": 1.5809462070465088,
"InsertImplicitShardAxisBeforeISel": 0.37781310081481934,
"InsertLocalTransposes": 0.8543226718902588,
"InsertOffloadedTransposes": 0.0779869556427002,
"LICM": 0.10844612121582031,
"LateLegalizeInst": 0.13918232917785645,
"LateLegalizePostSplit": 0.08665323257446289,
"LateLowerReshapeOp": 0.02942204475402832,
"LateLowerTensorOp": 0.22415471076965332,
"LateNeuronInstComb": 0.9580295085906982,
"LayoutPreprocessing": 0.7032866477966309,
"LayoutPreprocessingAndAnalysis": 1.1142585277557373,
"LayoutRequirementAnalysis": 0.40787601470947266,
"LegalizeCCOpLayout": 0.052059173583984375,
"LegalizeOpLevelAlias": 0.022985219955444336,
"LegalizePartitionReduce": 0.036779165267944336,
"LegalizeSundaAccess": 0.9256985187530518,
"LegalizeSundaMacro": 0.7163646221160889,
"LegalizeType": 0.12823128700256348,
"LocalLayoutOpt": 0.5141463279724121,
"LoopFusion": 0.2369997501373291,
"LoopSplitting": 0.026223182678222656,
"LowerBroadcast": 0.04629969596862793,
"LowerCCOpBlockAxis": 0.17850804328918457,
"LowerComplexBroadcast": 0.06136465072631836,
"LowerIntrinsics": 0.898535966873169,
"LowerShardAxis": 0.19546747207641602,
"LowerTensorOp": 0.38861823081970215,
"LowerToSendRecv": 0.1479358673095703,
"LowerTranspose": 0.4128274917602539,
"MacroGeneration": 1.907278060913086,
"MaskPropagation": 0.10989689826965332,
"MemcpyElimination": 3.3127715587615967,
"MutateDataType": 0.031205177307128906,
"NeuronAliasDependencyInduction": 0.014397859573364258,
"NeuronAliasDependencyReset": 0.018595218658447266,
"NeuronInstComb": 0.28766393661499023,
"NeuronLICM": 0.24966120719909668,
"NeuronLoopFusion": 0.9968361854553223,
"NeuronLoopInterchange": 0.049851179122924805,
"NeuronSimplifier": 0.4482152462005615,
"NeuronSimplifyPredicates": 0.12775635719299316,
"NeuronValueNumbering": 0.09903788566589355,
"OptimizeAliasedCopyChain": 0.010796785354614258,
"OptimizeNKIKernels": 0.9786763191223145,
"PAGLayoutOpt": 11.095458984375,
"PComputeCutting": 0.28275370597839355,
"PGLayoutTilingPipeline": 27.67894172668457,
"PGTiling": 4.957591533660889,
"PadElimination": 0.009732246398925781,
"ParAxesAnnotation": 10.236833572387695,
"PartialLoopFusion": 1.0972201824188232,
"PartialSimdFusion": 0.529569149017334,
"PerfectLoopNest": 0.059952735900878906,
"RecognizeOpIdiom": 0.120391845703125,
"Recompute": 0.005687713623046875,
"RelaxPredicates": 0.09513998031616211,
"Rematerialization": 0.14099407196044922,
"RemoveShardedPartitionAxes": 0.6145329475402832,
"ReshapeWeights": 0.0214080810546875,
"ResolveAccessConflict": 0.1656482219696045,
"ResolveComplicatePredicates": 0.03499007225036621,
"RewriteReplicationMatmul": 0.03796839714050293,
"RewriteWeights": 0.06430387496948242,
"SFKVectorizer": 5.7498369216918945,
"ShardingPropagationAnalysis": 0.5885381698608398,
"SimpleAllReduceTiling": 0.05400824546813965,
"Simplifier": 0.07847452163696289,
"SimplifyMacroPredicates": 0.25547003746032715,
"SimplifyNeuronTensor": 0.2766604423522949,
"SimplifySlice": 0.02300572395324707,
"SimplifyTensor": 0.2370157241821289,
"SpillPSum": 0.4576537609100342,
"SplitAPUnionSets": 0.6796882152557373,
"SplitAccGrp": 0.03665971755981445,
"StaticProfiler": 0.13864398002624512,
"StaticTransposeLocalTensor": 0.2623753547668457,
"SundaISel": 1.3397884368896484,
"TCTransform": 0.025269269943237305,
"TensorInitialization": 0.16660237312316895,
"TensorOpSimplifier": 0.15259718894958496,
"TensorOpTransform": 0.8282039165496826,
"TileCCOps": 0.3066561222076416,
"TilingProfiler": 0.33243393898010254,
"TransformConvOp": 0.05726790428161621,
"TritiumFusion": 0.13151931762695313,
"ValueNumbering": 0.06717801094055176,
"VectorizeDMA": 0.5187788009643555,
"VectorizeMatMult": 0.04601097106933594,
"WeightCoalescing": 0.049105167388916016,
"ZeroSizeTensorElimination": 0.0003685951232910156
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 41972,
"StaticProfiler::AifUb": 17.11549949645996,
"StaticProfiler::ArithmeticIntensityTensorizer": 27.70398712158203,
"StaticProfiler::AverageDmaLength": 2723.7119140625,
"StaticProfiler::AverageFractalPeUtilization": 96.97119140625,
"StaticProfiler::AveragePartitionUtilization": 88.53791809082031,
"StaticProfiler::AveragePeUtilization": 81.30671691894531,
"StaticProfiler::DDRTransferBytes": 918042224,
"StaticProfiler::InternalTransferBytes": 174444896,
"StaticProfiler::LoadExpanded": 238040,
"StaticProfiler::LocalizationEfficiency": 161.8649139404297,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 167.2097930908203,
"StaticProfiler::StoreExpanded": 18609,
"StaticProfiler::TotalDMAExpanded": 256649,
"StaticProfiler::TotalDynamicInstancesCount": 57730,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 50535,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 179,
"TilingProfiler::MatMultInstructionsAfterTiling": 30304,
"TilingProfiler::NumPfTransposes": 348,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 198,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 7835,
"TilingProfiler::PfTransposeInstructionsForIo": 5217,
"TilingProfiler::PfTransposeInstructionsForLocal": 508,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 2110,
"TilingProfiler::ReduceInstructionsAfterTiling": 59,
"TilingProfiler::SimdInstructionsAfterTiling": 2266,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"topk": {
"compiletime": {
"CoalesceCCOp": 0.003599405288696289,
"DMALocalityOpt": 0.003013134002685547,
"DMAProfiler": 0.0034313201904296875,
"DataStreaming": 0.006362199783325195,
"DoNothing": 0.00013327598571777344,
"ExpandISAMacro": 0.003620147705078125,
"FactorizeBlkDims": 0.010854482650756836,
"InferPSumTensor": 0.010383129119873047,
"InferSharedMemLoc": 0.0029261112213134766,
"InsertCoreBarrier": 0.003323793411254883,
"LateLegalizeInst": 0.007294178009033203,
"LateNeuronInstComb": 0.008392095565795898,
"LegalizeSundaAccess": 0.013858556747436523,
"LegalizeType": 0.009220361709594727,
"LowerBroadcast": 0.0032749176025390625,
"LowerIntrinsics": 0.0036187171936035156,
"LowerTranspose": 0.0032770633697509766,
"NeuronInstComb": 0.008351325988769531,
"NeuronLICM": 0.009110689163208008,
"NeuronSimplifyPredicates": 0.0036361217498779297,
"NeuronValueNumbering": 0.0038361549377441406,
"SFKVectorizer": 0.031067371368408203,
"SimpleAllReduceTiling": 0.0040929317474365234,
"SimplifyNeuronTensor": 0.057866573333740234,
"SpillPSum": 0.02007007598876953,
"WeightCoalescing": 0.003391265869140625
}
}
}

3
token_generation_model/_tp0_bk0/graph.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:29d9d8ef7e5a0b735090051a93d177f1b4866fe07fc6988421cb47de73a1953f
size 3185664

4225
token_generation_model/_tp0_bk0/log-neuron-cc.txt Normal file
View File

File diff suppressed because it is too large Load Diff

3
token_generation_model/_tp0_bk0/metaneff.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:2751bdd6298e760b3ed04356f31f6ffbb7d7c37e3002f4e686a4fba8c24c2240
size 2465286

3
token_generation_model/_tp0_bk0/model.MODULE_5e3d0ce8963512f9ea2d+937cd7a2.hlo_module.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:166776f75e13c12167868e06c05edd9ecd84b1f5005be8ff59ffb090e4b60e00
size 2443667

3
token_generation_model/_tp0_bk0/model.MODULE_5e3d0ce8963512f9ea2d+937cd7a2.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:29d9d8ef7e5a0b735090051a93d177f1b4866fe07fc6988421cb47de73a1953f
size 3185664

224
token_generation_model/_tp0_bk0/neuron_config.json Normal file
View File

@@ -0,0 +1,224 @@
{
"_attn_implementation_autoset": false,
"_name_or_path": "/models/Qwen3-1.7B/",
"add_cross_attention": false,
"architectures": [
"Qwen3ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"attribute_map": {},
"bad_words_ids": null,
"begin_suppress_tokens": null,
"bos_token_id": 151643,
"chunk_size_feed_forward": 0,
"cross_attention_hidden_size": null,
"decoder_start_token_id": null,
"diversity_penalty": 0.0,
"do_sample": false,
"early_stopping": false,
"encoder_no_repeat_ngram_size": 0,
"eos_token_id": 151645,
"exponential_decay_length_penalty": null,
"finetuning_task": null,
"forced_bos_token_id": null,
"forced_eos_token_id": null,
"fused_spec_config": null,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 2048,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"initializer_range": 0.02,
"intermediate_size": 6144,
"is_decoder": false,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"length_penalty": 1.0,
"max_length": 20,
"max_position_embeddings": 40960,
"max_window_layers": 28,
"metadata": null,
"min_length": 0,
"model_type": "qwen3",
"neuron_config": {
"activation_quantization_type": null,
"allow_input_truncation": false,
"apply_seq_ids_mask": false,
"async_mode": false,
"attention_dp_degree": 1,
"attention_dtype": null,
"attn_block_cte_nki_kernel_enabled": false,
"attn_block_tkg_nki_kernel_cache_update": false,
"attn_block_tkg_nki_kernel_cascaded_attention": false,
"attn_block_tkg_nki_kernel_enabled": false,
"attn_cls": {
"__module__": "neuronx_distributed_inference.models.qwen3.modeling_qwen3",
"__name__": "NeuronQwen3Attention"
},
"attn_kernel_enabled": null,
"attn_tkg_builtin_kernel_enabled": false,
"attn_tkg_nki_kernel_enabled": false,
"batch_size": 4,
"bucket_n_active_tokens": false,
"buckets": [
128
],
"cast_type": "config",
"cc_pipeline_tiling_factor": 1,
"chunked_prefill_config": null,
"context_encoding_buckets": null,
"cp_degree": 1,
"ctx_batch_size": 1,
"disable_kv_cache_tiling": false,
"draft_model_modules_to_not_convert": null,
"enable_bucketing": true,
"enable_cte_modular_flow": false,
"enable_eagle_draft_input_norm": false,
"enable_eagle_speculation": false,
"enable_fused_speculation": false,
"enable_long_context_mode": false,
"enable_output_completion_notifications": false,
"enable_spill_reload_dge": false,
"enable_token_tree": false,
"ep_degree": 1,
"expert_mlp_nki_kernel_enabled": null,
"flash_decoding_enabled": false,
"fused_qkv": false,
"fused_rmsnorm_skip_gamma": false,
"is_block_kv_layout": null,
"is_chunked_prefill": false,
"is_continuous_batching": true,
"is_eagle_draft": false,
"is_medusa": false,
"is_prefill_stage": false,
"is_prefix_caching": false,
"k_cache_transposed": false,
"kv_cache_batch_size": 4,
"kv_cache_padding_size": 0,
"kv_cache_quant": false,
"kv_cache_tiling": false,
"layer_boundary_markers": false,
"lm_head_pad": true,
"lm_head_pad_alignment_size": 1,
"local_ranks_size": 4,
"logical_nc_config": 2,
"lora_config": null,
"max_batch_size": 4,
"max_context_length": 2048,
"max_length": 2048,
"max_new_tokens": null,
"medusa_speculation_length": 0,
"medusa_tree": null,
"mlp_kernel_enabled": false,
"mlp_kernel_fuse_residual_add": false,
"modules_to_not_convert": null,
"moe_fused_nki_kernel_enabled": null,
"n_active_tokens": 1,
"n_positions": 2048,
"num_medusa_heads": 0,
"on_cpu": false,
"on_device_sampling_config": {
"deterministic": false,
"do_sample": false,
"dynamic": true,
"global_topk": 256,
"on_device_sampling_config": true,
"temperature": 1.0,
"top_k": 1,
"top_k_kernel_enabled": false,
"top_p": 1.0
},
"output_logits": false,
"overrides_torch_dtype": true,
"pa_block_size": 2048,
"pa_num_blocks": 4,
"padding_side": "right",
"pp_degree": 1,
"prefix_buckets": null,
"qk_layernorm": false,
"qkv_kernel_enabled": false,
"qkv_kernel_fuse_residual_add": false,
"qkv_kernel_nbsd_layout": false,
"quantization_dtype": "int8",
"quantization_type": "per_tensor_symmetric",
"quantize_clamp_bound": Infinity,
"quantized": false,
"quantized_checkpoints_path": null,
"quantized_mlp_kernel_enabled": false,
"rmsnorm_quantize_kernel_enabled": false,
"router_topk_nki_kernel_enabled": null,
"rpl_reduce_dtype": null,
"save_sharded_checkpoint": true,
"scratchpad_page_size": null,
"seq_len": 2048,
"seq_len_threshold_for_cc_tiling": 16384,
"sequence_parallel_enabled": false,
"shared_mlp_nki_kernel_enabled": null,
"skip_sharding": false,
"skip_warmup": false,
"spec_batch_size": 4,
"speculation_length": 0,
"start_rank_id": 0,
"strided_context_parallel_kernel_enabled": false,
"target": null,
"tensor_capture_config": null,
"tile_cc": false,
"tkg_batch_size": 4,
"token_generation_buckets": [
128
],
"token_tree_config": null,
"torch_dtype": "bfloat16",
"tp_degree": 4,
"vocab_parallel": false,
"weight_gather_seq_len_threshold": 32768,
"weights_to_skip_layout_optimization": [],
"world_size": 4
},
"no_repeat_ngram_size": 0,
"num_attention_heads": 16,
"num_beam_groups": 1,
"num_beams": 1,
"num_cores_per_group": 1,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"num_return_sequences": 1,
"output_attentions": false,
"output_hidden_states": false,
"output_scores": false,
"pad_token_id": 0,
"prefix": null,
"problem_type": null,
"pruned_heads": {},
"remove_invalid_values": false,
"repetition_penalty": 1.0,
"return_dict": true,
"return_dict_in_generate": false,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000,
"sep_token_id": null,
"sliding_window": null,
"suppress_tokens": null,
"task_specific_params": null,
"temperature": 1.0,
"tf_legacy_loss": false,
"tie_encoder_decoder": false,
"tie_word_embeddings": true,
"tokenizer_class": null,
"top_k": 50,
"top_p": 1.0,
"torchscript": false,
"transformers_version": "4.51.0",
"typical_p": 1.0,
"use_bfloat16": false,
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}

3
token_generation_model/_tp0_bk0/wrapped_neff.hlo Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:3d7d771432f2a5b315c36747141a382cd677d56bd52622fbc82564ea1e8d3890
size 3380710

1
token_generation_model/_tp0_bk1/command.txt Normal file
View File

@@ -0,0 +1 @@
neuronx-cc compile --framework=XLA model.MODULE_f53407701fa4882a24c0+55c11e15.hlo_module.pb --output model.MODULE_f53407701fa4882a24c0+55c11e15.neff --target=trn2 --auto-cast=none --model-type=transformer '--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ' --lnc=2 -O2 --internal-hlo2tensorizer-options=--verify-hlo=true --logfile=log-neuron-cc.txt --verbose=35

1
token_generation_model/_tp0_bk1/compile_flags.MODULE_f53407701fa4882a24c0+55c11e15.json Normal file
View File

@@ -0,0 +1 @@
["--target=trn2", "--auto-cast=none", "--model-type=transformer", "--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ", "--lnc=2", "-O2", "--internal-hlo2tensorizer-options=--verify-hlo=true", "--logfile=/models/Qwen3-1.7B-compiled/token_generation_model/_tp0_bk1/log-neuron-cc.txt"]

590
token_generation_model/_tp0_bk1/global_metric_store.json Normal file
View File

@@ -0,0 +1,590 @@
{
"Average": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 97.62759399414063,
"StaticProfiler::AveragePartitionUtilization": 90.06017303466797,
"StaticProfiler::AveragePeUtilization": 82.14405822753906,
"StaticProfiler::LocalizationEfficiency": 160.6605224609375,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 165.92486572265625,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0
}
},
"Count": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 1,
"StaticProfiler::AveragePartitionUtilization": 1,
"StaticProfiler::AveragePeUtilization": 1,
"StaticProfiler::LocalizationEfficiency": 1,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 1,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 1,
"TilingProfiler::AveragePeUtilizationAfterTiling": 1
}
},
"Sum": {
"compiletime": {
"AGOrderingAnalysisPass": 5.160665512084961,
"AffinePredicateResolution": 0.7100803852081299,
"AliasDependencyElimination": 0.0032160282135009766,
"AliasDependencyInduction": 17.29242706298828,
"AliasDependencyReset": 18.28659439086914,
"BFComputeCutting": 0.24476861953735352,
"BirCodeGenLoop": 6.947810173034668,
"CCOpFusion": 0.6929187774658203,
"CanonicalizeConv": 2.9000000722589903e-05,
"CanonicalizeDAGForPGTiling": 0.21501421928405762,
"CanonicalizeForTensorizer": 0.0008399999933317304,
"CanonicalizeIR": 0.8826379776000977,
"Canonicalizer": 0.014088000170886517,
"CoalesceCCOp": 0.5657715797424316,
"CommuteConcat": 0.028485536575317383,
"DMALocalityOpt": 0.042168617248535156,
"DMAProfiler": 0.1519615650177002,
"DMATilingProfiler": 0.09857678413391113,
"DataLocalityOpt": 2.827857255935669,
"DataStreaming": 0.20627403259277344,
"DeConcat": 0.08719110488891602,
"DeadCodeElimination": 0.03139615058898926,
"DeadStoreElimination": 1.1847929954528809,
"DelinearIndices": 0.4130244255065918,
"Delinearization": 0.35245656967163086,
"DelinearizeSPMD": 0.4181056022644043,
"DoNothing": 0.00046539306640625,
"DramToDramTranspose": 0.30753135681152344,
"DumpGraphAndMetadata": 0.2011098861694336,
"EliminateDivs": 1.3804805278778076,
"ExpandBatchNorm": 1.1923854351043701,
"ExpandISAMacro": 0.08926725387573242,
"FactorizeBlkDims": 0.5252444744110107,
"FactorizeThreadAxesInFreeDims": 0.1739037036895752,
"FlattenMacroLoop": 0.07839345932006836,
"GenericAccessSimplifier": 0.025495052337646484,
"HoistCompute": 6.199999916134402e-05,
"IdentifyCrossPassTensors": 0.0005760000203736126,
"InferInitValue": 1.3857665061950684,
"InferIntrinsicOnCC": 0.646845817565918,
"InferNeuronTensor": 2.0493955612182617,
"InferNonlocalTensors": 6.630102634429932,
"InferPSumTensor": 1.2878854274749756,
"InferShardAxis": 10.588101387023926,
"InferSharedMemLoc": 0.10591363906860352,
"InlineNativeKernels": 0.0488896369934082,
"InsertCoreBarrier": 0.3818776607513428,
"InsertIOTransposes": 0.8546113967895508,
"InsertImplicitShardAxisBeforeISel": 0.36914730072021484,
"InsertLocalTransposes": 1.5642907619476318,
"InsertOffloadedTransposes": 0.12299966812133789,
"LICM": 0.12427496910095215,
"LateLegalizeInst": 0.4006483554840088,
"LateLegalizePostSplit": 0.09814929962158203,
"LateLowerReshapeOp": 0.0431976318359375,
"LateLowerTensorOp": 5.883858680725098,
"LateNeuronInstComb": 1.04233980178833,
"LayoutPreprocessing": 1.2663531303405762,
"LayoutPreprocessingAndAnalysis": 1.9893858432769775,
"LayoutRequirementAnalysis": 0.6988728046417236,
"LegalizeCCOpLayout": 1.499981164932251,
"LegalizeOpLevelAlias": 0.6842360496520996,
"LegalizePartitionReduce": 0.09090352058410645,
"LegalizeSundaAccess": 3.133746862411499,
"LegalizeSundaMacro": 0.6402661800384521,
"LegalizeType": 0.1407451629638672,
"LocalLayoutOpt": 0.7939743995666504,
"LoopFusion": 0.29726386070251465,
"LoopSplitting": 0.08333015441894531,
"LowerBroadcast": 0.06013798713684082,
"LowerCCOpBlockAxis": 1.4568629264831543,
"LowerComplexBroadcast": 0.06818914413452148,
"LowerIntrinsics": 0.9435675144195557,
"LowerShardAxis": 0.22619009017944336,
"LowerTensorOp": 3.0802407264709473,
"LowerToSendRecv": 0.19478487968444824,
"LowerTranspose": 0.5133018493652344,
"MacroGeneration": 4.587047100067139,
"MaskPropagation": 0.1672959327697754,
"MemcastMotion": 0.00015999999595806003,
"MemcpyElimination": 29.92277717590332,
"MutateDataType": 0.04070854187011719,
"NeuronAliasDependencyInduction": 0.018942594528198242,
"NeuronAliasDependencyReset": 0.02425408363342285,
"NeuronInstComb": 0.4308168888092041,
"NeuronLICM": 0.29584765434265137,
"NeuronLoopFusion": 1.117248296737671,
"NeuronLoopInterchange": 0.0636141300201416,
"NeuronSimplifier": 0.704033374786377,
"NeuronSimplifyPredicates": 0.2319803237915039,
"NeuronValueNumbering": 0.11443066596984863,
"OptimizeAliasedCopyChain": 0.2889397144317627,
"OptimizeNKIKernels": 1.2374975681304932,
"PAGLayoutOpt": 18.6531982421875,
"PComputeCutting": 0.5685915946960449,
"PGLayoutTilingPipeline": 53.59619903564453,
"PGTiling": 11.116703033447266,
"PadElimination": 0.012248039245605469,
"ParAxesAnnotation": 17.06294822692871,
"PartialLoopFusion": 1.3277881145477295,
"PartialSimdFusion": 0.7600483894348145,
"PenguinizeFunctions": 0.001062000053934753,
"PerfectLoopNest": 0.061771392822265625,
"PruneFunctions": 0.0004830000107176602,
"RecognizeOpIdiom": 0.12406373023986816,
"Recompute": 0.00886678695678711,
"RelaxPredicates": 0.11265873908996582,
"Rematerialization": 0.18548583984375,
"RemoveOptimizationBarriers": 0.009134000167250633,
"RemoveShardedPartitionAxes": 1.4166390895843506,
"ReshapeWeights": 0.02198624610900879,
"ResolveAccessConflict": 0.20142865180969238,
"ResolveComplicatePredicates": 0.6187732219696045,
"RewriteReplicationMatmul": 0.04161477088928223,
"RewriteWeights": 0.0676727294921875,
"SFKVectorizer": 11.220393180847168,
"ScatterMotion": 0.003527000080794096,
"ShardingPropagationAnalysis": 0.8252537250518799,
"SimpleAllReduceTiling": 0.1859297752380371,
"Simplifier": 0.09509730339050293,
"SimplifyMacroPredicates": 0.26697540283203125,
"SimplifyNeuronTensor": 0.41252875328063965,
"SimplifySlice": 0.02629375457763672,
"SimplifyTensor": 0.42005443572998047,
"SpillPSum": 0.5884366035461426,
"SplitAPUnionSets": 0.4571385383605957,
"SplitAccGrp": 0.052629709243774414,
"StaticProfiler": 0.12954020500183105,
"StaticTransposeLocalTensor": 0.36740827560424805,
"SundaISel": 1.4737660884857178,
"TCTransform": 0.031054019927978516,
"TensorInitialization": 0.18288874626159668,
"TensorOpSimplifier": 3.2880165576934814,
"TensorOpTransform": 18.079126358032227,
"TensorizerLegalizationPass": 0.0004990000161342323,
"TileCCOps": 0.17596793174743652,
"TilingProfiler": 0.3990769386291504,
"TransformConvOp": 1.2077322006225586,
"TritiumFusion": 0.2976958751678467,
"ValueNumbering": 0.10138130187988281,
"VectorizeDMA": 0.8864037990570068,
"VectorizeMatMult": 0.05278921127319336,
"VerifySupportedOps": 0.0004729999927803874,
"WeightCoalescing": 0.08338785171508789,
"ZeroSizeTensorElimination": 0.0006525516510009766,
"algsimp": 0.0014349999837577343,
"batchnorm_expander": 0.000818000000435859,
"boundary-marker-removal": 0.0003020000003743917,
"call-inliner": 0.00018099999579135329,
"canonicalize-boundary-marker": 0.0005639999872073531,
"collective-stream-id-checker": 0.002294000005349517,
"comparison-expander": 0.0012120000319555402,
"computation-deduplicator": 0.0011109999613836408,
"config-lowering": 0.00042299999040551484,
"constant_folding": 0.0001289999927394092,
"cse": 0.0007570000016130507,
"dce": 7.300000288523734e-05,
"dynamic-slice-transpose": 0.0002640000020619482,
"eliminate-redundant-compare": 0.0001250000059371814,
"emit-offloaded-dropout": 0.0006559999892488122,
"flatten-call-graph": 0.0003330000035930425,
"fuse-send-recv": 0.013015000149607658,
"hilo-conditional-to-select": 0.00016799999866634607,
"hilo::LegalizeAlias": 0.0036430000327527523,
"hilo::NeuronInstCombine": 1.1000000085914508e-05,
"hilo::NeuronOpFusion": 0.00011600000289035961,
"hilo::ReplaceTokenTypeWithU8Pass": 0.0003260000084992498,
"hilo::ScheduleFusion": 3.300000025774352e-05,
"hilo::SixtyFourHack": 0.0013909999979659915,
"hilo::VerifyAliasing": 0.00014600000577047467,
"hlo-mac-count": 0.019317999482154846,
"io-con-pipe-begin": 0.0002589999930933118,
"io-con-pipe-end": 9.999999974752427e-07,
"io-layout-normalization": 0.07799100130796432,
"legalize-ccops-for-tensorizer": 5.2999999752501026e-05,
"legalize-compare": 0.0006939999875612557,
"lower-argminmax-custom-call": 0.0002629999944474548,
"map-inline": 0.06807900220155716,
"metadata-naming": 0.07221200317144394,
"mlir::detail::OpToOpPassAdaptor": 0.007284999825060368,
"mlir::hlo::MhloToPyPenguin": 0.6235949993133545,
"mlir::mhlo::LowerComplexExtraPass": 0.00916799996048212,
"mlir::mhlo::LowerComplexPass": 0.0006799999973736703,
"native-to-custom-softmax": 0.0017719999887049198,
"native-to-custom-softmax-dx": 0.0017500000540167093,
"neuron-hlo-verifier": 0.2390509992837906,
"operand_upcaster": 0.070872001349926,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 9.999999747378752e-06,
"post-partition-simplification": 0.21146699786186218,
"pre-hlo-begin": 8.600000001024455e-05,
"pre-hlo-end": 9.999999974752427e-07,
"replace-minimum-constant": 0.00019700000120792538,
"reshape-mover": 5.400000009103678e-05,
"simplify-concat": 0.002297000028192997,
"simplify-while-loops": 4.600000102072954e-05,
"transform-variadic-reduce": 0.00037900000461377203,
"tuple-simplifier": 0.0001340000017080456,
"unpack-nested-aws-ntwsr": 0.00025400001322850585,
"unroll-while-loop": 7.000000096013537e-06
},
"hilo": {
"HloMacCount": 1766645760.0,
"Traffic": 1252618368.0
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 42842,
"StaticProfiler::AifUb": 17.765029907226563,
"StaticProfiler::ArithmeticIntensityTensorizer": 28.541391372680664,
"StaticProfiler::AverageDmaLength": 3512.794677734375,
"StaticProfiler::DDRTransferBytes": 924925552,
"StaticProfiler::InternalTransferBytes": 182108768,
"StaticProfiler::LoadExpanded": 181148,
"StaticProfiler::StoreExpanded": 4273,
"StaticProfiler::TotalDMAExpanded": 185421,
"StaticProfiler::TotalDynamicInstancesCount": 57693,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 50217,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 123,
"TilingProfiler::MatMultInstructionsAfterTiling": 30976,
"TilingProfiler::NumPfTransposes": 348,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 198,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 7892,
"TilingProfiler::PfTransposeInstructionsForIo": 5666,
"TilingProfiler::PfTransposeInstructionsForLocal": 564,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 1662,
"TilingProfiler::ReduceInstructionsAfterTiling": 115,
"TilingProfiler::SimdInstructionsAfterTiling": 2295,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"all": {
"compiletime": {
"CanonicalizeConv": 2.9000000722589903e-05,
"CanonicalizeForTensorizer": 0.0008399999933317304,
"Canonicalizer": 0.014088000170886517,
"HoistCompute": 6.199999916134402e-05,
"IdentifyCrossPassTensors": 0.0005760000203736126,
"MemcastMotion": 0.00015999999595806003,
"PenguinizeFunctions": 0.001062000053934753,
"PruneFunctions": 0.0004830000107176602,
"RemoveOptimizationBarriers": 0.009134000167250633,
"ScatterMotion": 0.003527000080794096,
"TensorizerLegalizationPass": 0.0004990000161342323,
"VerifySupportedOps": 0.0004729999927803874,
"algsimp": 0.0014349999837577343,
"batchnorm_expander": 0.000818000000435859,
"boundary-marker-removal": 0.0003020000003743917,
"call-inliner": 0.00018099999579135329,
"canonicalize-boundary-marker": 0.0005639999872073531,
"collective-stream-id-checker": 0.002294000005349517,
"comparison-expander": 0.0012120000319555402,
"computation-deduplicator": 0.0011109999613836408,
"config-lowering": 0.00042299999040551484,
"constant_folding": 0.0001289999927394092,
"cse": 0.0007570000016130507,
"dce": 7.300000288523734e-05,
"dynamic-slice-transpose": 0.0002640000020619482,
"eliminate-redundant-compare": 0.0001250000059371814,
"emit-offloaded-dropout": 0.0006559999892488122,
"flatten-call-graph": 0.0003330000035930425,
"fuse-send-recv": 0.013015000149607658,
"hilo-conditional-to-select": 0.00016799999866634607,
"hilo::LegalizeAlias": 0.0036430000327527523,
"hilo::NeuronInstCombine": 1.1000000085914508e-05,
"hilo::NeuronOpFusion": 0.00011600000289035961,
"hilo::ReplaceTokenTypeWithU8Pass": 0.0003260000084992498,
"hilo::ScheduleFusion": 3.300000025774352e-05,
"hilo::SixtyFourHack": 0.0013909999979659915,
"hilo::VerifyAliasing": 0.00014600000577047467,
"hlo-mac-count": 0.019317999482154846,
"io-con-pipe-begin": 0.0002589999930933118,
"io-con-pipe-end": 9.999999974752427e-07,
"io-layout-normalization": 0.07799100130796432,
"legalize-ccops-for-tensorizer": 5.2999999752501026e-05,
"legalize-compare": 0.0006939999875612557,
"lower-argminmax-custom-call": 0.0002629999944474548,
"map-inline": 0.06807900220155716,
"metadata-naming": 0.07221200317144394,
"mlir::detail::OpToOpPassAdaptor": 0.007284999825060368,
"mlir::hlo::MhloToPyPenguin": 0.6235949993133545,
"mlir::mhlo::LowerComplexExtraPass": 0.00916799996048212,
"mlir::mhlo::LowerComplexPass": 0.0006799999973736703,
"native-to-custom-softmax": 0.0017719999887049198,
"native-to-custom-softmax-dx": 0.0017500000540167093,
"neuron-hlo-verifier": 0.2390509992837906,
"operand_upcaster": 0.070872001349926,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 9.999999747378752e-06,
"post-partition-simplification": 0.21146699786186218,
"pre-hlo-begin": 8.600000001024455e-05,
"pre-hlo-end": 9.999999974752427e-07,
"replace-minimum-constant": 0.00019700000120792538,
"reshape-mover": 5.400000009103678e-05,
"simplify-concat": 0.002297000028192997,
"simplify-while-loops": 4.600000102072954e-05,
"transform-variadic-reduce": 0.00037900000461377203,
"tuple-simplifier": 0.0001340000017080456,
"unpack-nested-aws-ntwsr": 0.00025400001322850585,
"unroll-while-loop": 7.000000096013537e-06
}
},
"cumsum": {
"compiletime": {
"CoalesceCCOp": 0.0002148151397705078,
"DMALocalityOpt": 0.0001659393310546875,
"DMAProfiler": 0.0008566379547119141,
"DataStreaming": 0.00026869773864746094,
"DoNothing": 0.00015687942504882813,
"ExpandISAMacro": 0.0004825592041015625,
"FactorizeBlkDims": 0.0004360675811767578,
"InferPSumTensor": 0.0005650520324707031,
"InferSharedMemLoc": 0.0002880096435546875,
"InsertCoreBarrier": 0.0002799034118652344,
"LateLegalizeInst": 0.0003802776336669922,
"LateNeuronInstComb": 0.0006039142608642578,
"LegalizeSundaAccess": 0.001421213150024414,
"LegalizeType": 0.0002644062042236328,
"LowerBroadcast": 0.0002384185791015625,
"LowerIntrinsics": 0.0002446174621582031,
"LowerTranspose": 0.00022077560424804688,
"NeuronInstComb": 0.0005753040313720703,
"NeuronLICM": 0.00037217140197753906,
"NeuronSimplifyPredicates": 0.0022978782653808594,
"NeuronValueNumbering": 0.0004074573516845703,
"SFKVectorizer": 0.002466440200805664,
"SimpleAllReduceTiling": 0.0002262592315673828,
"SimplifyNeuronTensor": 0.0005323886871337891,
"SpillPSum": 0.00045490264892578125,
"WeightCoalescing": 0.0002465248107910156
}
},
"sg00": {
"hilo": {
"ArithmeticIntensity": 2.8207247257232666,
"HloMacCount": 1766645760.0,
"Traffic": 1252618368.0
}
},
"sg0000": {
"compiletime": {
"AGOrderingAnalysisPass": 5.160665512084961,
"AffinePredicateResolution": 0.7100803852081299,
"AliasDependencyElimination": 0.0032160282135009766,
"AliasDependencyInduction": 17.29242706298828,
"AliasDependencyReset": 18.28659439086914,
"BFComputeCutting": 0.24476861953735352,
"BirCodeGenLoop": 6.947810173034668,
"CCOpFusion": 0.6929187774658203,
"CanonicalizeDAGForPGTiling": 0.21501421928405762,
"CanonicalizeIR": 0.8826379776000977,
"CoalesceCCOp": 0.5618836879730225,
"CommuteConcat": 0.028485536575317383,
"DMALocalityOpt": 0.03933858871459961,
"DMAProfiler": 0.14742827415466309,
"DMATilingProfiler": 0.09857678413391113,
"DataLocalityOpt": 2.827857255935669,
"DataStreaming": 0.19969892501831055,
"DeConcat": 0.08719110488891602,
"DeadCodeElimination": 0.03139615058898926,
"DeadStoreElimination": 1.1847929954528809,
"DelinearIndices": 0.4130244255065918,
"Delinearization": 0.35245656967163086,
"DelinearizeSPMD": 0.4181056022644043,
"DoNothing": 0.0001385211944580078,
"DramToDramTranspose": 0.30753135681152344,
"DumpGraphAndMetadata": 0.2011098861694336,
"EliminateDivs": 1.3804805278778076,
"ExpandBatchNorm": 1.1923854351043701,
"ExpandISAMacro": 0.08506417274475098,
"FactorizeBlkDims": 0.5093107223510742,
"FactorizeThreadAxesInFreeDims": 0.1739037036895752,
"FlattenMacroLoop": 0.07839345932006836,
"GenericAccessSimplifier": 0.025495052337646484,
"InferInitValue": 1.3857665061950684,
"InferIntrinsicOnCC": 0.646845817565918,
"InferNeuronTensor": 2.0493955612182617,
"InferNonlocalTensors": 6.630102634429932,
"InferPSumTensor": 1.275925636291504,
"InferShardAxis": 10.588101387023926,
"InferSharedMemLoc": 0.102783203125,
"InlineNativeKernels": 0.0488896369934082,
"InsertCoreBarrier": 0.3783996105194092,
"InsertIOTransposes": 0.8546113967895508,
"InsertImplicitShardAxisBeforeISel": 0.36914730072021484,
"InsertLocalTransposes": 1.5642907619476318,
"InsertOffloadedTransposes": 0.12299966812133789,
"LICM": 0.12427496910095215,
"LateLegalizeInst": 0.3925168514251709,
"LateLegalizePostSplit": 0.09814929962158203,
"LateLowerReshapeOp": 0.0431976318359375,
"LateLowerTensorOp": 5.883858680725098,
"LateNeuronInstComb": 1.0332238674163818,
"LayoutPreprocessing": 1.2663531303405762,
"LayoutPreprocessingAndAnalysis": 1.9893858432769775,
"LayoutRequirementAnalysis": 0.6988728046417236,
"LegalizeCCOpLayout": 1.499981164932251,
"LegalizeOpLevelAlias": 0.6842360496520996,
"LegalizePartitionReduce": 0.09090352058410645,
"LegalizeSundaAccess": 3.1170387268066406,
"LegalizeSundaMacro": 0.6402661800384521,
"LegalizeType": 0.1309196949005127,
"LocalLayoutOpt": 0.7939743995666504,
"LoopFusion": 0.29726386070251465,
"LoopSplitting": 0.08333015441894531,
"LowerBroadcast": 0.05664563179016113,
"LowerCCOpBlockAxis": 1.4568629264831543,
"LowerComplexBroadcast": 0.06818914413452148,
"LowerIntrinsics": 0.9396772384643555,
"LowerShardAxis": 0.22619009017944336,
"LowerTensorOp": 3.0802407264709473,
"LowerToSendRecv": 0.19478487968444824,
"LowerTranspose": 0.5097734928131104,
"MacroGeneration": 4.587047100067139,
"MaskPropagation": 0.1672959327697754,
"MemcpyElimination": 29.92277717590332,
"MutateDataType": 0.04070854187011719,
"NeuronAliasDependencyInduction": 0.018942594528198242,
"NeuronAliasDependencyReset": 0.02425408363342285,
"NeuronInstComb": 0.42151784896850586,
"NeuronLICM": 0.2860753536224365,
"NeuronLoopFusion": 1.117248296737671,
"NeuronLoopInterchange": 0.0636141300201416,
"NeuronSimplifier": 0.704033374786377,
"NeuronSimplifyPredicates": 0.22606325149536133,
"NeuronValueNumbering": 0.11009359359741211,
"OptimizeAliasedCopyChain": 0.2889397144317627,
"OptimizeNKIKernels": 1.2374975681304932,
"PAGLayoutOpt": 18.6531982421875,
"PComputeCutting": 0.5685915946960449,
"PGLayoutTilingPipeline": 53.59619903564453,
"PGTiling": 11.116703033447266,
"PadElimination": 0.012248039245605469,
"ParAxesAnnotation": 17.06294822692871,
"PartialLoopFusion": 1.3277881145477295,
"PartialSimdFusion": 0.7600483894348145,
"PerfectLoopNest": 0.061771392822265625,
"RecognizeOpIdiom": 0.12406373023986816,
"Recompute": 0.00886678695678711,
"RelaxPredicates": 0.11265873908996582,
"Rematerialization": 0.18548583984375,
"RemoveShardedPartitionAxes": 1.4166390895843506,
"ReshapeWeights": 0.02198624610900879,
"ResolveAccessConflict": 0.20142865180969238,
"ResolveComplicatePredicates": 0.6187732219696045,
"RewriteReplicationMatmul": 0.04161477088928223,
"RewriteWeights": 0.0676727294921875,
"SFKVectorizer": 11.18409252166748,
"ShardingPropagationAnalysis": 0.8252537250518799,
"SimpleAllReduceTiling": 0.18207907676696777,
"Simplifier": 0.09509730339050293,
"SimplifyMacroPredicates": 0.26697540283203125,
"SimplifyNeuronTensor": 0.3594348430633545,
"SimplifySlice": 0.02629375457763672,
"SimplifyTensor": 0.42005443572998047,
"SpillPSum": 0.5635168552398682,
"SplitAPUnionSets": 0.4571385383605957,
"SplitAccGrp": 0.052629709243774414,
"StaticProfiler": 0.12954020500183105,
"StaticTransposeLocalTensor": 0.36740827560424805,
"SundaISel": 1.4737660884857178,
"TCTransform": 0.031054019927978516,
"TensorInitialization": 0.18288874626159668,
"TensorOpSimplifier": 3.2880165576934814,
"TensorOpTransform": 18.079126358032227,
"TileCCOps": 0.17596793174743652,
"TilingProfiler": 0.3990769386291504,
"TransformConvOp": 1.2077322006225586,
"TritiumFusion": 0.2976958751678467,
"ValueNumbering": 0.10138130187988281,
"VectorizeDMA": 0.8864037990570068,
"VectorizeMatMult": 0.05278921127319336,
"WeightCoalescing": 0.07968640327453613,
"ZeroSizeTensorElimination": 0.0006525516510009766
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 42842,
"StaticProfiler::AifUb": 17.765029907226563,
"StaticProfiler::ArithmeticIntensityTensorizer": 28.541391372680664,
"StaticProfiler::AverageDmaLength": 3512.794677734375,
"StaticProfiler::AverageFractalPeUtilization": 97.62759399414063,
"StaticProfiler::AveragePartitionUtilization": 90.06017303466797,
"StaticProfiler::AveragePeUtilization": 82.14405822753906,
"StaticProfiler::DDRTransferBytes": 924925552,
"StaticProfiler::InternalTransferBytes": 182108768,
"StaticProfiler::LoadExpanded": 181148,
"StaticProfiler::LocalizationEfficiency": 160.6605224609375,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 165.92486572265625,
"StaticProfiler::StoreExpanded": 4273,
"StaticProfiler::TotalDMAExpanded": 185421,
"StaticProfiler::TotalDynamicInstancesCount": 57693,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 50217,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 123,
"TilingProfiler::MatMultInstructionsAfterTiling": 30976,
"TilingProfiler::NumPfTransposes": 348,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 198,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 7892,
"TilingProfiler::PfTransposeInstructionsForIo": 5666,
"TilingProfiler::PfTransposeInstructionsForLocal": 564,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 1662,
"TilingProfiler::ReduceInstructionsAfterTiling": 115,
"TilingProfiler::SimdInstructionsAfterTiling": 2295,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"topk": {
"compiletime": {
"CoalesceCCOp": 0.003673076629638672,
"DMALocalityOpt": 0.0026640892028808594,
"DMAProfiler": 0.0036766529083251953,
"DataStreaming": 0.00630640983581543,
"DoNothing": 0.00016999244689941406,
"ExpandISAMacro": 0.003720521926879883,
"FactorizeBlkDims": 0.015497684478759766,
"InferPSumTensor": 0.011394739151000977,
"InferSharedMemLoc": 0.002842426300048828,
"InsertCoreBarrier": 0.0031981468200683594,
"LateLegalizeInst": 0.0077512264251708984,
"LateNeuronInstComb": 0.008512020111083984,
"LegalizeSundaAccess": 0.015286922454833984,
"LegalizeType": 0.00956106185913086,
"LowerBroadcast": 0.003253936767578125,
"LowerIntrinsics": 0.003645658493041992,
"LowerTranspose": 0.0033075809478759766,
"NeuronInstComb": 0.008723735809326172,
"NeuronLICM": 0.009400129318237305,
"NeuronSimplifyPredicates": 0.0036191940307617188,
"NeuronValueNumbering": 0.003929615020751953,
"SFKVectorizer": 0.033834218978881836,
"SimpleAllReduceTiling": 0.003624439239501953,
"SimplifyNeuronTensor": 0.05256152153015137,
"SpillPSum": 0.024464845657348633,
"WeightCoalescing": 0.003454923629760742
}
}
}

3
token_generation_model/_tp0_bk1/graph.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:cda6f9bc52718ce93e1fb58a7aa253d606cf545bda03b396e531716e6a9ce678
size 3154944

4469
token_generation_model/_tp0_bk1/log-neuron-cc.txt Normal file
View File

File diff suppressed because it is too large Load Diff

3
token_generation_model/_tp0_bk1/metaneff.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:fd8b7f01387ce1b80997167a938d7a05e712ade8a8271b5da3b01c446894cbd8
size 2464555

3
token_generation_model/_tp0_bk1/model.MODULE_f53407701fa4882a24c0+55c11e15.hlo_module.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:d9364e7ab828d2cbc2c983e088e9c293cdaddc1137d948ee2d3c625e9eab02f7
size 2550451

3
token_generation_model/_tp0_bk1/model.MODULE_f53407701fa4882a24c0+55c11e15.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:cda6f9bc52718ce93e1fb58a7aa253d606cf545bda03b396e531716e6a9ce678
size 3154944

224
token_generation_model/_tp0_bk1/neuron_config.json Normal file
View File

@@ -0,0 +1,224 @@
{
"_attn_implementation_autoset": false,
"_name_or_path": "/models/Qwen3-1.7B/",
"add_cross_attention": false,
"architectures": [
"Qwen3ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"attribute_map": {},
"bad_words_ids": null,
"begin_suppress_tokens": null,
"bos_token_id": 151643,
"chunk_size_feed_forward": 0,
"cross_attention_hidden_size": null,
"decoder_start_token_id": null,
"diversity_penalty": 0.0,
"do_sample": false,
"early_stopping": false,
"encoder_no_repeat_ngram_size": 0,
"eos_token_id": 151645,
"exponential_decay_length_penalty": null,
"finetuning_task": null,
"forced_bos_token_id": null,
"forced_eos_token_id": null,
"fused_spec_config": null,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 2048,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"initializer_range": 0.02,
"intermediate_size": 6144,
"is_decoder": false,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"length_penalty": 1.0,
"max_length": 20,
"max_position_embeddings": 40960,
"max_window_layers": 28,
"metadata": null,
"min_length": 0,
"model_type": "qwen3",
"neuron_config": {
"activation_quantization_type": null,
"allow_input_truncation": false,
"apply_seq_ids_mask": false,
"async_mode": false,
"attention_dp_degree": 1,
"attention_dtype": null,
"attn_block_cte_nki_kernel_enabled": false,
"attn_block_tkg_nki_kernel_cache_update": false,
"attn_block_tkg_nki_kernel_cascaded_attention": false,
"attn_block_tkg_nki_kernel_enabled": false,
"attn_cls": {
"__module__": "neuronx_distributed_inference.models.qwen3.modeling_qwen3",
"__name__": "NeuronQwen3Attention"
},
"attn_kernel_enabled": null,
"attn_tkg_builtin_kernel_enabled": false,
"attn_tkg_nki_kernel_enabled": false,
"batch_size": 4,
"bucket_n_active_tokens": false,
"buckets": [
256
],
"cast_type": "config",
"cc_pipeline_tiling_factor": 1,
"chunked_prefill_config": null,
"context_encoding_buckets": null,
"cp_degree": 1,
"ctx_batch_size": 1,
"disable_kv_cache_tiling": false,
"draft_model_modules_to_not_convert": null,
"enable_bucketing": true,
"enable_cte_modular_flow": false,
"enable_eagle_draft_input_norm": false,
"enable_eagle_speculation": false,
"enable_fused_speculation": false,
"enable_long_context_mode": false,
"enable_output_completion_notifications": false,
"enable_spill_reload_dge": false,
"enable_token_tree": false,
"ep_degree": 1,
"expert_mlp_nki_kernel_enabled": null,
"flash_decoding_enabled": false,
"fused_qkv": false,
"fused_rmsnorm_skip_gamma": false,
"is_block_kv_layout": null,
"is_chunked_prefill": false,
"is_continuous_batching": true,
"is_eagle_draft": false,
"is_medusa": false,
"is_prefill_stage": false,
"is_prefix_caching": false,
"k_cache_transposed": false,
"kv_cache_batch_size": 4,
"kv_cache_padding_size": 0,
"kv_cache_quant": false,
"kv_cache_tiling": false,
"layer_boundary_markers": false,
"lm_head_pad": true,
"lm_head_pad_alignment_size": 1,
"local_ranks_size": 4,
"logical_nc_config": 2,
"lora_config": null,
"max_batch_size": 4,
"max_context_length": 2048,
"max_length": 2048,
"max_new_tokens": null,
"medusa_speculation_length": 0,
"medusa_tree": null,
"mlp_kernel_enabled": false,
"mlp_kernel_fuse_residual_add": false,
"modules_to_not_convert": null,
"moe_fused_nki_kernel_enabled": null,
"n_active_tokens": 1,
"n_positions": 2048,
"num_medusa_heads": 0,
"on_cpu": false,
"on_device_sampling_config": {
"deterministic": false,
"do_sample": false,
"dynamic": true,
"global_topk": 256,
"on_device_sampling_config": true,
"temperature": 1.0,
"top_k": 1,
"top_k_kernel_enabled": false,
"top_p": 1.0
},
"output_logits": false,
"overrides_torch_dtype": true,
"pa_block_size": 2048,
"pa_num_blocks": 4,
"padding_side": "right",
"pp_degree": 1,
"prefix_buckets": null,
"qk_layernorm": false,
"qkv_kernel_enabled": false,
"qkv_kernel_fuse_residual_add": false,
"qkv_kernel_nbsd_layout": false,
"quantization_dtype": "int8",
"quantization_type": "per_tensor_symmetric",
"quantize_clamp_bound": Infinity,
"quantized": false,
"quantized_checkpoints_path": null,
"quantized_mlp_kernel_enabled": false,
"rmsnorm_quantize_kernel_enabled": false,
"router_topk_nki_kernel_enabled": null,
"rpl_reduce_dtype": null,
"save_sharded_checkpoint": true,
"scratchpad_page_size": null,
"seq_len": 2048,
"seq_len_threshold_for_cc_tiling": 16384,
"sequence_parallel_enabled": false,
"shared_mlp_nki_kernel_enabled": null,
"skip_sharding": false,
"skip_warmup": false,
"spec_batch_size": 4,
"speculation_length": 0,
"start_rank_id": 0,
"strided_context_parallel_kernel_enabled": false,
"target": null,
"tensor_capture_config": null,
"tile_cc": false,
"tkg_batch_size": 4,
"token_generation_buckets": [
256
],
"token_tree_config": null,
"torch_dtype": "bfloat16",
"tp_degree": 4,
"vocab_parallel": false,
"weight_gather_seq_len_threshold": 32768,
"weights_to_skip_layout_optimization": [],
"world_size": 4
},
"no_repeat_ngram_size": 0,
"num_attention_heads": 16,
"num_beam_groups": 1,
"num_beams": 1,
"num_cores_per_group": 1,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"num_return_sequences": 1,
"output_attentions": false,
"output_hidden_states": false,
"output_scores": false,
"pad_token_id": 0,
"prefix": null,
"problem_type": null,
"pruned_heads": {},
"remove_invalid_values": false,
"repetition_penalty": 1.0,
"return_dict": true,
"return_dict_in_generate": false,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000,
"sep_token_id": null,
"sliding_window": null,
"suppress_tokens": null,
"task_specific_params": null,
"temperature": 1.0,
"tf_legacy_loss": false,
"tie_encoder_decoder": false,
"tie_word_embeddings": true,
"tokenizer_class": null,
"top_k": 50,
"top_p": 1.0,
"torchscript": false,
"transformers_version": "4.51.0",
"typical_p": 1.0,
"use_bfloat16": false,
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}

1
token_generation_model/_tp0_bk2/command.txt Normal file
View File

@@ -0,0 +1 @@
neuronx-cc compile --framework=XLA model.MODULE_c60e20a111715e620aa2+6bb8acc9.hlo_module.pb --output model.MODULE_c60e20a111715e620aa2+6bb8acc9.neff --target=trn2 --auto-cast=none --model-type=transformer '--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ' --lnc=2 -O2 --internal-hlo2tensorizer-options=--verify-hlo=true --logfile=log-neuron-cc.txt --verbose=35

1
token_generation_model/_tp0_bk2/compile_flags.MODULE_c60e20a111715e620aa2+6bb8acc9.json Normal file
View File

@@ -0,0 +1 @@
["--target=trn2", "--auto-cast=none", "--model-type=transformer", "--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ", "--lnc=2", "-O2", "--internal-hlo2tensorizer-options=--verify-hlo=true", "--logfile=/models/Qwen3-1.7B-compiled/token_generation_model/_tp0_bk2/log-neuron-cc.txt"]

590
token_generation_model/_tp0_bk2/global_metric_store.json Normal file
View File

@@ -0,0 +1,590 @@
{
"Average": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 97.78141784667969,
"StaticProfiler::AveragePartitionUtilization": 90.33787536621094,
"StaticProfiler::AveragePeUtilization": 81.002685546875,
"StaticProfiler::LocalizationEfficiency": 155.71730041503906,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 160.65768432617188,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0
}
},
"Count": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 1,
"StaticProfiler::AveragePartitionUtilization": 1,
"StaticProfiler::AveragePeUtilization": 1,
"StaticProfiler::LocalizationEfficiency": 1,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 1,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 1,
"TilingProfiler::AveragePeUtilizationAfterTiling": 1
}
},
"Sum": {
"compiletime": {
"AGOrderingAnalysisPass": 6.775559425354004,
"AffinePredicateResolution": 0.7124242782592773,
"AliasDependencyElimination": 0.0037508010864257813,
"AliasDependencyInduction": 17.88080596923828,
"AliasDependencyReset": 19.01345443725586,
"BFComputeCutting": 0.14321279525756836,
"BirCodeGenLoop": 4.049878120422363,
"CCOpFusion": 0.9675219058990479,
"CanonicalizeConv": 0.0,
"CanonicalizeDAGForPGTiling": 0.18202924728393555,
"CanonicalizeForTensorizer": 0.0005849999724887311,
"CanonicalizeIR": 0.8085732460021973,
"Canonicalizer": 0.015991000458598137,
"CoalesceCCOp": 0.25977158546447754,
"CommuteConcat": 0.027615070343017578,
"DMALocalityOpt": 0.048050880432128906,
"DMAProfiler": 0.09339690208435059,
"DMATilingProfiler": 0.09855914115905762,
"DataLocalityOpt": 2.8309335708618164,
"DataStreaming": 0.2996983528137207,
"DeConcat": 0.08213448524475098,
"DeadCodeElimination": 0.030428647994995117,
"DeadStoreElimination": 1.051145315170288,
"DelinearIndices": 0.5529322624206543,
"Delinearization": 0.14493751525878906,
"DelinearizeSPMD": 0.181105375289917,
"DoNothing": 0.0003535747528076172,
"DramToDramTranspose": 0.5176758766174316,
"DumpGraphAndMetadata": 0.1749565601348877,
"EliminateDivs": 1.6087908744812012,
"ExpandBatchNorm": 1.096595048904419,
"ExpandISAMacro": 0.09282922744750977,
"FactorizeBlkDims": 0.5121583938598633,
"FactorizeThreadAxesInFreeDims": 0.08890652656555176,
"FlattenMacroLoop": 0.08152222633361816,
"GenericAccessSimplifier": 0.02545022964477539,
"HoistCompute": 9.899999713525176e-05,
"IdentifyCrossPassTensors": 0.00011500000255182385,
"InferInitValue": 1.2978432178497314,
"InferIntrinsicOnCC": 0.7616860866546631,
"InferNeuronTensor": 2.6053829193115234,
"InferNonlocalTensors": 7.113053321838379,
"InferPSumTensor": 4.450908660888672,
"InferShardAxis": 12.849608421325684,
"InferSharedMemLoc": 0.19225406646728516,
"InlineNativeKernels": 0.08442234992980957,
"InsertCoreBarrier": 0.15249347686767578,
"InsertIOTransposes": 0.8695223331451416,
"InsertImplicitShardAxisBeforeISel": 0.49620485305786133,
"InsertLocalTransposes": 1.934779167175293,
"InsertOffloadedTransposes": 0.23423099517822266,
"LICM": 0.1213834285736084,
"LateLegalizeInst": 0.28924560546875,
"LateLegalizePostSplit": 0.22411513328552246,
"LateLowerReshapeOp": 0.033789873123168945,
"LateLowerTensorOp": 4.297260761260986,
"LateNeuronInstComb": 1.0230679512023926,
"LayoutPreprocessing": 1.8572075366973877,
"LayoutPreprocessingAndAnalysis": 2.2904410362243652,
"LayoutRequirementAnalysis": 0.4237644672393799,
"LegalizeCCOpLayout": 1.1935856342315674,
"LegalizeOpLevelAlias": 0.7793729305267334,
"LegalizePartitionReduce": 0.09432244300842285,
"LegalizeSundaAccess": 1.0004379749298096,
"LegalizeSundaMacro": 0.7198967933654785,
"LegalizeType": 0.1766369342803955,
"LocalLayoutOpt": 0.6666848659515381,
"LoopFusion": 0.29988932609558105,
"LoopSplitting": 0.05208778381347656,
"LowerBroadcast": 0.08215570449829102,
"LowerCCOpBlockAxis": 0.2862672805786133,
"LowerComplexBroadcast": 0.09356045722961426,
"LowerIntrinsics": 1.3606979846954346,
"LowerShardAxis": 0.39397311210632324,
"LowerTensorOp": 4.018380165100098,
"LowerToSendRecv": 0.27350616455078125,
"LowerTranspose": 0.5171041488647461,
"MacroGeneration": 2.0922586917877197,
"MaskPropagation": 0.12857651710510254,
"MemcastMotion": 0.0002789999998640269,
"MemcpyElimination": 30.280338287353516,
"MutateDataType": 0.03657245635986328,
"NeuronAliasDependencyInduction": 0.026292085647583008,
"NeuronAliasDependencyReset": 0.03253936767578125,
"NeuronInstComb": 0.36768054962158203,
"NeuronLICM": 0.3019218444824219,
"NeuronLoopFusion": 1.711080551147461,
"NeuronLoopInterchange": 0.08134770393371582,
"NeuronSimplifier": 0.5104148387908936,
"NeuronSimplifyPredicates": 0.24121522903442383,
"NeuronValueNumbering": 0.11496901512145996,
"OptimizeAliasedCopyChain": 0.40732264518737793,
"OptimizeNKIKernels": 1.0493371486663818,
"PAGLayoutOpt": 18.23680305480957,
"PComputeCutting": 0.5901892185211182,
"PGLayoutTilingPipeline": 54.30891036987305,
"PGTiling": 10.472368240356445,
"PadElimination": 0.0187380313873291,
"ParAxesAnnotation": 16.293506622314453,
"PartialLoopFusion": 1.5408625602722168,
"PartialSimdFusion": 0.8349573612213135,
"PenguinizeFunctions": 0.0005789999850094318,
"PerfectLoopNest": 0.07371997833251953,
"PruneFunctions": 0.00018899999849963933,
"RecognizeOpIdiom": 0.12681221961975098,
"Recompute": 0.007548809051513672,
"RelaxPredicates": 0.11768078804016113,
"Rematerialization": 0.15080475807189941,
"RemoveOptimizationBarriers": 0.0004290000069886446,
"RemoveShardedPartitionAxes": 1.1734652519226074,
"ReshapeWeights": 0.022508859634399414,
"ResolveAccessConflict": 0.20145153999328613,
"ResolveComplicatePredicates": 0.3191838264465332,
"RewriteReplicationMatmul": 0.039293527603149414,
"RewriteWeights": 0.0773766040802002,
"SFKVectorizer": 11.08578872680664,
"ScatterMotion": 0.005127000156790018,
"ShardingPropagationAnalysis": 0.665086030960083,
"SimpleAllReduceTiling": 0.07452917098999023,
"Simplifier": 0.09160900115966797,
"SimplifyMacroPredicates": 0.28479433059692383,
"SimplifyNeuronTensor": 0.5389096736907959,
"SimplifySlice": 0.02684783935546875,
"SimplifyTensor": 0.2644212245941162,
"SpillPSum": 1.0592920780181885,
"SplitAPUnionSets": 1.0491511821746826,
"SplitAccGrp": 0.055588722229003906,
"StaticProfiler": 0.39172840118408203,
"StaticTransposeLocalTensor": 0.7211995124816895,
"SundaISel": 1.5357580184936523,
"TCTransform": 0.029030561447143555,
"TensorInitialization": 0.20336008071899414,
"TensorOpSimplifier": 3.100177049636841,
"TensorOpTransform": 20.175758361816406,
"TensorizerLegalizationPass": 0.000455000001238659,
"TileCCOps": 0.17463970184326172,
"TilingProfiler": 0.7014575004577637,
"TransformConvOp": 0.9923229217529297,
"TritiumFusion": 0.33277034759521484,
"ValueNumbering": 0.08796954154968262,
"VectorizeDMA": 0.520815372467041,
"VectorizeMatMult": 0.08555078506469727,
"VerifySupportedOps": 0.0002469999890308827,
"WeightCoalescing": 0.09020423889160156,
"ZeroSizeTensorElimination": 0.0006709098815917969,
"algsimp": 0.0018210000125691295,
"batchnorm_expander": 0.0022700000554323196,
"boundary-marker-removal": 0.0009800000116229057,
"call-inliner": 0.0002410000015515834,
"canonicalize-boundary-marker": 0.0009430000209249556,
"collective-stream-id-checker": 0.0008960000122897327,
"comparison-expander": 0.000623999978415668,
"computation-deduplicator": 0.003625999903306365,
"config-lowering": 0.00017899999511428177,
"constant_folding": 0.0668639987707138,
"cse": 0.001184999942779541,
"dce": 6.399999983841553e-05,
"dynamic-slice-transpose": 0.0003530000103637576,
"eliminate-redundant-compare": 0.00014400000509340316,
"emit-offloaded-dropout": 0.0007820000173524022,
"flatten-call-graph": 0.0009519999730400741,
"fuse-send-recv": 0.07950299978256226,
"hilo-conditional-to-select": 0.00015700000221841037,
"hilo::LegalizeAlias": 0.003444999922066927,
"hilo::NeuronInstCombine": 0.0008500000112690032,
"hilo::NeuronOpFusion": 0.00020799999765586108,
"hilo::ReplaceTokenTypeWithU8Pass": 0.0007050000131130219,
"hilo::ScheduleFusion": 4.3000000005122274e-05,
"hilo::SixtyFourHack": 0.0005150000215508044,
"hilo::VerifyAliasing": 0.00035099999513477087,
"hlo-mac-count": 0.08947300165891647,
"io-con-pipe-begin": 6.199999916134402e-05,
"io-con-pipe-end": 9.999999974752427e-07,
"io-layout-normalization": 0.0706389993429184,
"legalize-ccops-for-tensorizer": 0.00013800000306218863,
"legalize-compare": 0.00027200000477023423,
"lower-argminmax-custom-call": 0.0004130000015720725,
"map-inline": 0.0010939999483525753,
"metadata-naming": 0.009581999853253365,
"mlir::detail::OpToOpPassAdaptor": 0.0004569999873638153,
"mlir::hlo::MhloToPyPenguin": 0.5751969814300537,
"mlir::mhlo::LowerComplexExtraPass": 0.00508299982175231,
"mlir::mhlo::LowerComplexPass": 0.0012469999492168427,
"native-to-custom-softmax": 0.0004349999944679439,
"native-to-custom-softmax-dx": 0.00044800000614486635,
"neuron-hlo-verifier": 0.3175100088119507,
"operand_upcaster": 0.0018560000462457538,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 0.0,
"post-partition-simplification": 0.48272499442100525,
"pre-hlo-begin": 1.4000000192027073e-05,
"pre-hlo-end": 1.9999999949504854e-06,
"replace-minimum-constant": 0.00022400000307243317,
"reshape-mover": 5.700000110664405e-05,
"simplify-concat": 0.002764000091701746,
"simplify-while-loops": 9.40000027185306e-05,
"transform-variadic-reduce": 0.001218999968841672,
"tuple-simplifier": 0.00048099999548867345,
"unpack-nested-aws-ntwsr": 0.0007040000054985285,
"unroll-while-loop": 1.2000000424450263e-05
},
"hilo": {
"HloMacCount": 1796005888.0,
"Traffic": 1252622464.0
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 45730,
"StaticProfiler::AifUb": 19.087068557739258,
"StaticProfiler::ArithmeticIntensityTensorizer": 29.72186851501465,
"StaticProfiler::AverageDmaLength": 2694.496826171875,
"StaticProfiler::DDRTransferBytes": 954289776,
"StaticProfiler::InternalTransferBytes": 196796000,
"StaticProfiler::LoadExpanded": 281500,
"StaticProfiler::StoreExpanded": 4273,
"StaticProfiler::TotalDMAExpanded": 285773,
"StaticProfiler::TotalDynamicInstancesCount": 62161,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 53675,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 123,
"TilingProfiler::MatMultInstructionsAfterTiling": 32320,
"TilingProfiler::NumPfTransposes": 292,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 142,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 8762,
"TilingProfiler::PfTransposeInstructionsForIo": 6564,
"TilingProfiler::PfTransposeInstructionsForLocal": 536,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 1662,
"TilingProfiler::ReduceInstructionsAfterTiling": 171,
"TilingProfiler::SimdInstructionsAfterTiling": 2465,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"all": {
"compiletime": {
"CanonicalizeConv": 0.0,
"CanonicalizeForTensorizer": 0.0005849999724887311,
"Canonicalizer": 0.015991000458598137,
"HoistCompute": 9.899999713525176e-05,
"IdentifyCrossPassTensors": 0.00011500000255182385,
"MemcastMotion": 0.0002789999998640269,
"PenguinizeFunctions": 0.0005789999850094318,
"PruneFunctions": 0.00018899999849963933,
"RemoveOptimizationBarriers": 0.0004290000069886446,
"ScatterMotion": 0.005127000156790018,
"TensorizerLegalizationPass": 0.000455000001238659,
"VerifySupportedOps": 0.0002469999890308827,
"algsimp": 0.0018210000125691295,
"batchnorm_expander": 0.0022700000554323196,
"boundary-marker-removal": 0.0009800000116229057,
"call-inliner": 0.0002410000015515834,
"canonicalize-boundary-marker": 0.0009430000209249556,
"collective-stream-id-checker": 0.0008960000122897327,
"comparison-expander": 0.000623999978415668,
"computation-deduplicator": 0.003625999903306365,
"config-lowering": 0.00017899999511428177,
"constant_folding": 0.0668639987707138,
"cse": 0.001184999942779541,
"dce": 6.399999983841553e-05,
"dynamic-slice-transpose": 0.0003530000103637576,
"eliminate-redundant-compare": 0.00014400000509340316,
"emit-offloaded-dropout": 0.0007820000173524022,
"flatten-call-graph": 0.0009519999730400741,
"fuse-send-recv": 0.07950299978256226,
"hilo-conditional-to-select": 0.00015700000221841037,
"hilo::LegalizeAlias": 0.003444999922066927,
"hilo::NeuronInstCombine": 0.0008500000112690032,
"hilo::NeuronOpFusion": 0.00020799999765586108,
"hilo::ReplaceTokenTypeWithU8Pass": 0.0007050000131130219,
"hilo::ScheduleFusion": 4.3000000005122274e-05,
"hilo::SixtyFourHack": 0.0005150000215508044,
"hilo::VerifyAliasing": 0.00035099999513477087,
"hlo-mac-count": 0.08947300165891647,
"io-con-pipe-begin": 6.199999916134402e-05,
"io-con-pipe-end": 9.999999974752427e-07,
"io-layout-normalization": 0.0706389993429184,
"legalize-ccops-for-tensorizer": 0.00013800000306218863,
"legalize-compare": 0.00027200000477023423,
"lower-argminmax-custom-call": 0.0004130000015720725,
"map-inline": 0.0010939999483525753,
"metadata-naming": 0.009581999853253365,
"mlir::detail::OpToOpPassAdaptor": 0.0004569999873638153,
"mlir::hlo::MhloToPyPenguin": 0.5751969814300537,
"mlir::mhlo::LowerComplexExtraPass": 0.00508299982175231,
"mlir::mhlo::LowerComplexPass": 0.0012469999492168427,
"native-to-custom-softmax": 0.0004349999944679439,
"native-to-custom-softmax-dx": 0.00044800000614486635,
"neuron-hlo-verifier": 0.3175100088119507,
"operand_upcaster": 0.0018560000462457538,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 0.0,
"post-partition-simplification": 0.48272499442100525,
"pre-hlo-begin": 1.4000000192027073e-05,
"pre-hlo-end": 1.9999999949504854e-06,
"replace-minimum-constant": 0.00022400000307243317,
"reshape-mover": 5.700000110664405e-05,
"simplify-concat": 0.002764000091701746,
"simplify-while-loops": 9.40000027185306e-05,
"transform-variadic-reduce": 0.001218999968841672,
"tuple-simplifier": 0.00048099999548867345,
"unpack-nested-aws-ntwsr": 0.0007040000054985285,
"unroll-while-loop": 1.2000000424450263e-05
}
},
"cumsum": {
"compiletime": {
"CoalesceCCOp": 0.0002155303955078125,
"DMALocalityOpt": 0.00016641616821289063,
"DMAProfiler": 0.0007414817810058594,
"DataStreaming": 0.0002970695495605469,
"DoNothing": 0.0001289844512939453,
"ExpandISAMacro": 0.0005061626434326172,
"FactorizeBlkDims": 0.0004284381866455078,
"InferPSumTensor": 0.0005972385406494141,
"InferSharedMemLoc": 0.0002605915069580078,
"InsertCoreBarrier": 0.000263214111328125,
"LateLegalizeInst": 0.00037741661071777344,
"LateNeuronInstComb": 0.0005505084991455078,
"LegalizeSundaAccess": 0.0013835430145263672,
"LegalizeType": 0.00022840499877929688,
"LowerBroadcast": 0.0002124309539794922,
"LowerIntrinsics": 0.0002105236053466797,
"LowerTranspose": 0.00021958351135253906,
"NeuronInstComb": 0.000583648681640625,
"NeuronLICM": 0.00035643577575683594,
"NeuronSimplifyPredicates": 0.002285003662109375,
"NeuronValueNumbering": 0.0004379749298095703,
"SFKVectorizer": 0.0024003982543945313,
"SimpleAllReduceTiling": 0.00020265579223632813,
"SimplifyNeuronTensor": 0.0004928112030029297,
"SpillPSum": 0.00045371055603027344,
"WeightCoalescing": 0.0002079010009765625
}
},
"sg00": {
"hilo": {
"ArithmeticIntensity": 2.867593288421631,
"HloMacCount": 1796005888.0,
"Traffic": 1252622464.0
}
},
"sg0000": {
"compiletime": {
"AGOrderingAnalysisPass": 6.775559425354004,
"AffinePredicateResolution": 0.7124242782592773,
"AliasDependencyElimination": 0.0037508010864257813,
"AliasDependencyInduction": 17.88080596923828,
"AliasDependencyReset": 19.01345443725586,
"BFComputeCutting": 0.14321279525756836,
"BirCodeGenLoop": 4.049878120422363,
"CCOpFusion": 0.9675219058990479,
"CanonicalizeDAGForPGTiling": 0.18202924728393555,
"CanonicalizeIR": 0.8085732460021973,
"CoalesceCCOp": 0.2558913230895996,
"CommuteConcat": 0.027615070343017578,
"DMALocalityOpt": 0.04517340660095215,
"DMAProfiler": 0.08888816833496094,
"DMATilingProfiler": 0.09855914115905762,
"DataLocalityOpt": 2.8309335708618164,
"DataStreaming": 0.29266810417175293,
"DeConcat": 0.08213448524475098,
"DeadCodeElimination": 0.030428647994995117,
"DeadStoreElimination": 1.051145315170288,
"DelinearIndices": 0.5529322624206543,
"Delinearization": 0.14493751525878906,
"DelinearizeSPMD": 0.181105375289917,
"DoNothing": 7.724761962890625e-05,
"DramToDramTranspose": 0.5176758766174316,
"DumpGraphAndMetadata": 0.1749565601348877,
"EliminateDivs": 1.6087908744812012,
"ExpandBatchNorm": 1.096595048904419,
"ExpandISAMacro": 0.08580923080444336,
"FactorizeBlkDims": 0.49906015396118164,
"FactorizeThreadAxesInFreeDims": 0.08890652656555176,
"FlattenMacroLoop": 0.08152222633361816,
"GenericAccessSimplifier": 0.02545022964477539,
"InferInitValue": 1.2978432178497314,
"InferIntrinsicOnCC": 0.7616860866546631,
"InferNeuronTensor": 2.6053829193115234,
"InferNonlocalTensors": 7.113053321838379,
"InferPSumTensor": 4.437821388244629,
"InferShardAxis": 12.849608421325684,
"InferSharedMemLoc": 0.18901801109313965,
"InlineNativeKernels": 0.08442234992980957,
"InsertCoreBarrier": 0.1489267349243164,
"InsertIOTransposes": 0.8695223331451416,
"InsertImplicitShardAxisBeforeISel": 0.49620485305786133,
"InsertLocalTransposes": 1.934779167175293,
"InsertOffloadedTransposes": 0.23423099517822266,
"LICM": 0.1213834285736084,
"LateLegalizeInst": 0.2814202308654785,
"LateLegalizePostSplit": 0.22411513328552246,
"LateLowerReshapeOp": 0.033789873123168945,
"LateLowerTensorOp": 4.297260761260986,
"LateNeuronInstComb": 1.014098882675171,
"LayoutPreprocessing": 1.8572075366973877,
"LayoutPreprocessingAndAnalysis": 2.2904410362243652,
"LayoutRequirementAnalysis": 0.4237644672393799,
"LegalizeCCOpLayout": 1.1935856342315674,
"LegalizeOpLevelAlias": 0.7793729305267334,
"LegalizePartitionReduce": 0.09432244300842285,
"LegalizeSundaAccess": 0.972252607345581,
"LegalizeSundaMacro": 0.7198967933654785,
"LegalizeType": 0.16599535942077637,
"LocalLayoutOpt": 0.6666848659515381,
"LoopFusion": 0.29988932609558105,
"LoopSplitting": 0.05208778381347656,
"LowerBroadcast": 0.07862257957458496,
"LowerCCOpBlockAxis": 0.2862672805786133,
"LowerComplexBroadcast": 0.09356045722961426,
"LowerIntrinsics": 1.3567829132080078,
"LowerShardAxis": 0.39397311210632324,
"LowerTensorOp": 4.018380165100098,
"LowerToSendRecv": 0.27350616455078125,
"LowerTranspose": 0.513293981552124,
"MacroGeneration": 2.0922586917877197,
"MaskPropagation": 0.12857651710510254,
"MemcpyElimination": 30.280338287353516,
"MutateDataType": 0.03657245635986328,
"NeuronAliasDependencyInduction": 0.026292085647583008,
"NeuronAliasDependencyReset": 0.03253936767578125,
"NeuronInstComb": 0.35875773429870605,
"NeuronLICM": 0.29233455657958984,
"NeuronLoopFusion": 1.711080551147461,
"NeuronLoopInterchange": 0.08134770393371582,
"NeuronSimplifier": 0.5104148387908936,
"NeuronSimplifyPredicates": 0.2326216697692871,
"NeuronValueNumbering": 0.11057472229003906,
"OptimizeAliasedCopyChain": 0.40732264518737793,
"OptimizeNKIKernels": 1.0493371486663818,
"PAGLayoutOpt": 18.23680305480957,
"PComputeCutting": 0.5901892185211182,
"PGLayoutTilingPipeline": 54.30891036987305,
"PGTiling": 10.472368240356445,
"PadElimination": 0.0187380313873291,
"ParAxesAnnotation": 16.293506622314453,
"PartialLoopFusion": 1.5408625602722168,
"PartialSimdFusion": 0.8349573612213135,
"PerfectLoopNest": 0.07371997833251953,
"RecognizeOpIdiom": 0.12681221961975098,
"Recompute": 0.007548809051513672,
"RelaxPredicates": 0.11768078804016113,
"Rematerialization": 0.15080475807189941,
"RemoveShardedPartitionAxes": 1.1734652519226074,
"ReshapeWeights": 0.022508859634399414,
"ResolveAccessConflict": 0.20145153999328613,
"ResolveComplicatePredicates": 0.3191838264465332,
"RewriteReplicationMatmul": 0.039293527603149414,
"RewriteWeights": 0.0773766040802002,
"SFKVectorizer": 11.047651290893555,
"ShardingPropagationAnalysis": 0.665086030960083,
"SimpleAllReduceTiling": 0.07064080238342285,
"Simplifier": 0.09160900115966797,
"SimplifyMacroPredicates": 0.28479433059692383,
"SimplifyNeuronTensor": 0.4839820861816406,
"SimplifySlice": 0.02684783935546875,
"SimplifyTensor": 0.2644212245941162,
"SpillPSum": 1.035245418548584,
"SplitAPUnionSets": 1.0491511821746826,
"SplitAccGrp": 0.055588722229003906,
"StaticProfiler": 0.39172840118408203,
"StaticTransposeLocalTensor": 0.7211995124816895,
"SundaISel": 1.5357580184936523,
"TCTransform": 0.029030561447143555,
"TensorInitialization": 0.20336008071899414,
"TensorOpSimplifier": 3.100177049636841,
"TensorOpTransform": 20.175758361816406,
"TileCCOps": 0.17463970184326172,
"TilingProfiler": 0.7014575004577637,
"TransformConvOp": 0.9923229217529297,
"TritiumFusion": 0.33277034759521484,
"ValueNumbering": 0.08796954154968262,
"VectorizeDMA": 0.520815372467041,
"VectorizeMatMult": 0.08555078506469727,
"WeightCoalescing": 0.08340644836425781,
"ZeroSizeTensorElimination": 0.0006709098815917969
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 45730,
"StaticProfiler::AifUb": 19.087068557739258,
"StaticProfiler::ArithmeticIntensityTensorizer": 29.72186851501465,
"StaticProfiler::AverageDmaLength": 2694.496826171875,
"StaticProfiler::AverageFractalPeUtilization": 97.78141784667969,
"StaticProfiler::AveragePartitionUtilization": 90.33787536621094,
"StaticProfiler::AveragePeUtilization": 81.002685546875,
"StaticProfiler::DDRTransferBytes": 954289776,
"StaticProfiler::InternalTransferBytes": 196796000,
"StaticProfiler::LoadExpanded": 281500,
"StaticProfiler::LocalizationEfficiency": 155.71730041503906,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 160.65768432617188,
"StaticProfiler::StoreExpanded": 4273,
"StaticProfiler::TotalDMAExpanded": 285773,
"StaticProfiler::TotalDynamicInstancesCount": 62161,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 53675,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 123,
"TilingProfiler::MatMultInstructionsAfterTiling": 32320,
"TilingProfiler::NumPfTransposes": 292,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 142,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 8762,
"TilingProfiler::PfTransposeInstructionsForIo": 6564,
"TilingProfiler::PfTransposeInstructionsForLocal": 536,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 1662,
"TilingProfiler::ReduceInstructionsAfterTiling": 171,
"TilingProfiler::SimdInstructionsAfterTiling": 2465,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"topk": {
"compiletime": {
"CoalesceCCOp": 0.003664731979370117,
"DMALocalityOpt": 0.002711057662963867,
"DMAProfiler": 0.003767251968383789,
"DataStreaming": 0.0067331790924072266,
"DoNothing": 0.00014734268188476563,
"ExpandISAMacro": 0.006513833999633789,
"FactorizeBlkDims": 0.012669801712036133,
"InferPSumTensor": 0.012490510940551758,
"InferSharedMemLoc": 0.0029754638671875,
"InsertCoreBarrier": 0.00330352783203125,
"LateLegalizeInst": 0.007447957992553711,
"LateNeuronInstComb": 0.008418560028076172,
"LegalizeSundaAccess": 0.02680182456970215,
"LegalizeType": 0.010413169860839844,
"LowerBroadcast": 0.0033206939697265625,
"LowerIntrinsics": 0.003704547882080078,
"LowerTranspose": 0.0035905838012695313,
"NeuronInstComb": 0.008339166641235352,
"NeuronLICM": 0.009230852127075195,
"NeuronSimplifyPredicates": 0.006308555603027344,
"NeuronValueNumbering": 0.003956317901611328,
"SFKVectorizer": 0.03573727607727051,
"SimpleAllReduceTiling": 0.0036857128143310547,
"SimplifyNeuronTensor": 0.054434776306152344,
"SpillPSum": 0.02359294891357422,
"WeightCoalescing": 0.0065898895263671875
}
}
}

3
token_generation_model/_tp0_bk2/graph.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:2e9104a1923d9a888389573352849a655d6f7005c5782380d0fa3306dab803f3
size 3390464

4468
token_generation_model/_tp0_bk2/log-neuron-cc.txt Normal file
View File

File diff suppressed because it is too large Load Diff

3
token_generation_model/_tp0_bk2/metaneff.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:a4a34ac27ace8c75df52c6f7fb5e0c1736f327638e3ac42ff906c6b52e125a5e
size 2464555

3
token_generation_model/_tp0_bk2/model.MODULE_c60e20a111715e620aa2+6bb8acc9.hlo_module.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:0baed4147c67e64b22e5d725b3e55d70e3e16ed2fe838758927f1cb592ebdc7a
size 2550451

3
token_generation_model/_tp0_bk2/model.MODULE_c60e20a111715e620aa2+6bb8acc9.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:2e9104a1923d9a888389573352849a655d6f7005c5782380d0fa3306dab803f3
size 3390464

224
token_generation_model/_tp0_bk2/neuron_config.json Normal file
View File

@@ -0,0 +1,224 @@
{
"_attn_implementation_autoset": false,
"_name_or_path": "/models/Qwen3-1.7B/",
"add_cross_attention": false,
"architectures": [
"Qwen3ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"attribute_map": {},
"bad_words_ids": null,
"begin_suppress_tokens": null,
"bos_token_id": 151643,
"chunk_size_feed_forward": 0,
"cross_attention_hidden_size": null,
"decoder_start_token_id": null,
"diversity_penalty": 0.0,
"do_sample": false,
"early_stopping": false,
"encoder_no_repeat_ngram_size": 0,
"eos_token_id": 151645,
"exponential_decay_length_penalty": null,
"finetuning_task": null,
"forced_bos_token_id": null,
"forced_eos_token_id": null,
"fused_spec_config": null,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 2048,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"initializer_range": 0.02,
"intermediate_size": 6144,
"is_decoder": false,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"length_penalty": 1.0,
"max_length": 20,
"max_position_embeddings": 40960,
"max_window_layers": 28,
"metadata": null,
"min_length": 0,
"model_type": "qwen3",
"neuron_config": {
"activation_quantization_type": null,
"allow_input_truncation": false,
"apply_seq_ids_mask": false,
"async_mode": false,
"attention_dp_degree": 1,
"attention_dtype": null,
"attn_block_cte_nki_kernel_enabled": false,
"attn_block_tkg_nki_kernel_cache_update": false,
"attn_block_tkg_nki_kernel_cascaded_attention": false,
"attn_block_tkg_nki_kernel_enabled": false,
"attn_cls": {
"__module__": "neuronx_distributed_inference.models.qwen3.modeling_qwen3",
"__name__": "NeuronQwen3Attention"
},
"attn_kernel_enabled": null,
"attn_tkg_builtin_kernel_enabled": false,
"attn_tkg_nki_kernel_enabled": false,
"batch_size": 4,
"bucket_n_active_tokens": false,
"buckets": [
512
],
"cast_type": "config",
"cc_pipeline_tiling_factor": 1,
"chunked_prefill_config": null,
"context_encoding_buckets": null,
"cp_degree": 1,
"ctx_batch_size": 1,
"disable_kv_cache_tiling": false,
"draft_model_modules_to_not_convert": null,
"enable_bucketing": true,
"enable_cte_modular_flow": false,
"enable_eagle_draft_input_norm": false,
"enable_eagle_speculation": false,
"enable_fused_speculation": false,
"enable_long_context_mode": false,
"enable_output_completion_notifications": false,
"enable_spill_reload_dge": false,
"enable_token_tree": false,
"ep_degree": 1,
"expert_mlp_nki_kernel_enabled": null,
"flash_decoding_enabled": false,
"fused_qkv": false,
"fused_rmsnorm_skip_gamma": false,
"is_block_kv_layout": null,
"is_chunked_prefill": false,
"is_continuous_batching": true,
"is_eagle_draft": false,
"is_medusa": false,
"is_prefill_stage": false,
"is_prefix_caching": false,
"k_cache_transposed": false,
"kv_cache_batch_size": 4,
"kv_cache_padding_size": 0,
"kv_cache_quant": false,
"kv_cache_tiling": false,
"layer_boundary_markers": false,
"lm_head_pad": true,
"lm_head_pad_alignment_size": 1,
"local_ranks_size": 4,
"logical_nc_config": 2,
"lora_config": null,
"max_batch_size": 4,
"max_context_length": 2048,
"max_length": 2048,
"max_new_tokens": null,
"medusa_speculation_length": 0,
"medusa_tree": null,
"mlp_kernel_enabled": false,
"mlp_kernel_fuse_residual_add": false,
"modules_to_not_convert": null,
"moe_fused_nki_kernel_enabled": null,
"n_active_tokens": 1,
"n_positions": 2048,
"num_medusa_heads": 0,
"on_cpu": false,
"on_device_sampling_config": {
"deterministic": false,
"do_sample": false,
"dynamic": true,
"global_topk": 256,
"on_device_sampling_config": true,
"temperature": 1.0,
"top_k": 1,
"top_k_kernel_enabled": false,
"top_p": 1.0
},
"output_logits": false,
"overrides_torch_dtype": true,
"pa_block_size": 2048,
"pa_num_blocks": 4,
"padding_side": "right",
"pp_degree": 1,
"prefix_buckets": null,
"qk_layernorm": false,
"qkv_kernel_enabled": false,
"qkv_kernel_fuse_residual_add": false,
"qkv_kernel_nbsd_layout": false,
"quantization_dtype": "int8",
"quantization_type": "per_tensor_symmetric",
"quantize_clamp_bound": Infinity,
"quantized": false,
"quantized_checkpoints_path": null,
"quantized_mlp_kernel_enabled": false,
"rmsnorm_quantize_kernel_enabled": false,
"router_topk_nki_kernel_enabled": null,
"rpl_reduce_dtype": null,
"save_sharded_checkpoint": true,
"scratchpad_page_size": null,
"seq_len": 2048,
"seq_len_threshold_for_cc_tiling": 16384,
"sequence_parallel_enabled": false,
"shared_mlp_nki_kernel_enabled": null,
"skip_sharding": false,
"skip_warmup": false,
"spec_batch_size": 4,
"speculation_length": 0,
"start_rank_id": 0,
"strided_context_parallel_kernel_enabled": false,
"target": null,
"tensor_capture_config": null,
"tile_cc": false,
"tkg_batch_size": 4,
"token_generation_buckets": [
512
],
"token_tree_config": null,
"torch_dtype": "bfloat16",
"tp_degree": 4,
"vocab_parallel": false,
"weight_gather_seq_len_threshold": 32768,
"weights_to_skip_layout_optimization": [],
"world_size": 4
},
"no_repeat_ngram_size": 0,
"num_attention_heads": 16,
"num_beam_groups": 1,
"num_beams": 1,
"num_cores_per_group": 1,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"num_return_sequences": 1,
"output_attentions": false,
"output_hidden_states": false,
"output_scores": false,
"pad_token_id": 0,
"prefix": null,
"problem_type": null,
"pruned_heads": {},
"remove_invalid_values": false,
"repetition_penalty": 1.0,
"return_dict": true,
"return_dict_in_generate": false,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000,
"sep_token_id": null,
"sliding_window": null,
"suppress_tokens": null,
"task_specific_params": null,
"temperature": 1.0,
"tf_legacy_loss": false,
"tie_encoder_decoder": false,
"tie_word_embeddings": true,
"tokenizer_class": null,
"top_k": 50,
"top_p": 1.0,
"torchscript": false,
"transformers_version": "4.51.0",
"typical_p": 1.0,
"use_bfloat16": false,
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}

1
token_generation_model/_tp0_bk3/command.txt Normal file
View File

@@ -0,0 +1 @@
neuronx-cc compile --framework=XLA model.MODULE_02e08d5fdadfbcb8a569+a6f96e4f.hlo_module.pb --output model.MODULE_02e08d5fdadfbcb8a569+a6f96e4f.neff --target=trn2 --auto-cast=none --model-type=transformer '--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ' --lnc=2 -O2 --internal-hlo2tensorizer-options=--verify-hlo=true --logfile=log-neuron-cc.txt --verbose=35

1
token_generation_model/_tp0_bk3/compile_flags.MODULE_02e08d5fdadfbcb8a569+a6f96e4f.json Normal file
View File

@@ -0,0 +1 @@
["--target=trn2", "--auto-cast=none", "--model-type=transformer", "--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ", "--lnc=2", "-O2", "--internal-hlo2tensorizer-options=--verify-hlo=true", "--logfile=/models/Qwen3-1.7B-compiled/token_generation_model/_tp0_bk3/log-neuron-cc.txt"]

590
token_generation_model/_tp0_bk3/global_metric_store.json Normal file
View File

@@ -0,0 +1,590 @@
{
"Average": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 97.88961791992188,
"StaticProfiler::AveragePartitionUtilization": 89.79463958740234,
"StaticProfiler::AveragePeUtilization": 78.904052734375,
"StaticProfiler::LocalizationEfficiency": 146.6905975341797,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 151.06674194335938,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0
}
},
"Count": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 1,
"StaticProfiler::AveragePartitionUtilization": 1,
"StaticProfiler::AveragePeUtilization": 1,
"StaticProfiler::LocalizationEfficiency": 1,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 1,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 1,
"TilingProfiler::AveragePeUtilizationAfterTiling": 1
}
},
"Sum": {
"compiletime": {
"AGOrderingAnalysisPass": 6.388453960418701,
"AffinePredicateResolution": 0.40378904342651367,
"AliasDependencyElimination": 0.016310930252075195,
"AliasDependencyInduction": 15.282819747924805,
"AliasDependencyReset": 16.618453979492188,
"BFComputeCutting": 0.1391315460205078,
"BirCodeGenLoop": 4.402980804443359,
"CCOpFusion": 0.8881416320800781,
"CanonicalizeConv": 4.999999873689376e-06,
"CanonicalizeDAGForPGTiling": 0.20721173286437988,
"CanonicalizeForTensorizer": 0.00076299998909235,
"CanonicalizeIR": 1.028043270111084,
"Canonicalizer": 0.08714199811220169,
"CoalesceCCOp": 0.2357621192932129,
"CommuteConcat": 0.0287783145904541,
"DMALocalityOpt": 0.07430553436279297,
"DMAProfiler": 0.08057951927185059,
"DMATilingProfiler": 0.09187006950378418,
"DataLocalityOpt": 2.820284605026245,
"DataStreaming": 0.4295690059661865,
"DeConcat": 0.06640625,
"DeadCodeElimination": 0.053946495056152344,
"DeadStoreElimination": 1.0540196895599365,
"DelinearIndices": 0.3954615592956543,
"Delinearization": 0.13945698738098145,
"DelinearizeSPMD": 0.17487859725952148,
"DoNothing": 0.0004494190216064453,
"DramToDramTranspose": 0.34574389457702637,
"DumpGraphAndMetadata": 0.23639750480651855,
"EliminateDivs": 1.4790117740631104,
"ExpandBatchNorm": 1.123610019683838,
"ExpandISAMacro": 0.12250304222106934,
"FactorizeBlkDims": 0.49267148971557617,
"FactorizeThreadAxesInFreeDims": 0.08273863792419434,
"FlattenMacroLoop": 0.08472108840942383,
"GenericAccessSimplifier": 0.025115966796875,
"HoistCompute": 0.0,
"IdentifyCrossPassTensors": 0.0003549999964889139,
"InferInitValue": 1.3054986000061035,
"InferIntrinsicOnCC": 0.7679235935211182,
"InferNeuronTensor": 1.671952724456787,
"InferNonlocalTensors": 7.110470771789551,
"InferPSumTensor": 4.2829790115356445,
"InferShardAxis": 12.786405563354492,
"InferSharedMemLoc": 0.16948676109313965,
"InlineNativeKernels": 0.05566692352294922,
"InsertCoreBarrier": 0.15158963203430176,
"InsertIOTransposes": 2.3931386470794678,
"InsertImplicitShardAxisBeforeISel": 0.5525217056274414,
"InsertLocalTransposes": 0.7429313659667969,
"InsertOffloadedTransposes": 0.19022369384765625,
"LICM": 0.14770793914794922,
"LateLegalizeInst": 0.31015753746032715,
"LateLegalizePostSplit": 0.22165155410766602,
"LateLowerReshapeOp": 0.03522944450378418,
"LateLowerTensorOp": 6.813569068908691,
"LateNeuronInstComb": 1.2720398902893066,
"LayoutPreprocessing": 1.8736376762390137,
"LayoutPreprocessingAndAnalysis": 2.3256783485412598,
"LayoutRequirementAnalysis": 0.4448714256286621,
"LegalizeCCOpLayout": 1.317772388458252,
"LegalizeOpLevelAlias": 0.5106728076934814,
"LegalizePartitionReduce": 0.08073091506958008,
"LegalizeSundaAccess": 0.9463906288146973,
"LegalizeSundaMacro": 0.7451527118682861,
"LegalizeType": 0.13950800895690918,
"LocalLayoutOpt": 0.6608095169067383,
"LoopFusion": 0.3101928234100342,
"LoopSplitting": 0.036502838134765625,
"LowerBroadcast": 0.06515121459960938,
"LowerCCOpBlockAxis": 0.32977890968322754,
"LowerComplexBroadcast": 0.15041875839233398,
"LowerIntrinsics": 1.0637662410736084,
"LowerShardAxis": 0.4116206169128418,
"LowerTensorOp": 3.285423755645752,
"LowerToSendRecv": 0.25904369354248047,
"LowerTranspose": 0.5238301753997803,
"MacroGeneration": 2.1192569732666016,
"MaskPropagation": 0.14647436141967773,
"MemcastMotion": 0.0,
"MemcpyElimination": 29.3009090423584,
"MutateDataType": 0.037621498107910156,
"NeuronAliasDependencyInduction": 0.06313967704772949,
"NeuronAliasDependencyReset": 0.08403873443603516,
"NeuronInstComb": 0.3617970943450928,
"NeuronLICM": 0.2975320816040039,
"NeuronLoopFusion": 1.491703987121582,
"NeuronLoopInterchange": 0.06900668144226074,
"NeuronSimplifier": 0.5315425395965576,
"NeuronSimplifyPredicates": 0.2570762634277344,
"NeuronValueNumbering": 0.11507940292358398,
"OptimizeAliasedCopyChain": 0.6118557453155518,
"OptimizeNKIKernels": 1.0786433219909668,
"PAGLayoutOpt": 18.570398330688477,
"PComputeCutting": 0.6222004890441895,
"PGLayoutTilingPipeline": 55.62285614013672,
"PGTiling": 10.0595703125,
"PadElimination": 0.021188735961914063,
"ParAxesAnnotation": 17.818796157836914,
"PartialLoopFusion": 1.5088038444519043,
"PartialSimdFusion": 0.8573262691497803,
"PenguinizeFunctions": 0.000754999986384064,
"PerfectLoopNest": 0.07143402099609375,
"PruneFunctions": 0.0012410000199452043,
"RecognizeOpIdiom": 0.1291654109954834,
"Recompute": 0.007833480834960938,
"RelaxPredicates": 0.14508056640625,
"Rematerialization": 0.17157793045043945,
"RemoveOptimizationBarriers": 0.0005370000144466758,
"RemoveShardedPartitionAxes": 1.3566563129425049,
"ReshapeWeights": 0.022662878036499023,
"ResolveAccessConflict": 0.2033238410949707,
"ResolveComplicatePredicates": 0.49439358711242676,
"RewriteReplicationMatmul": 0.03938794136047363,
"RewriteWeights": 0.07809877395629883,
"SFKVectorizer": 10.350098609924316,
"ScatterMotion": 0.0,
"ShardingPropagationAnalysis": 0.7199399471282959,
"SimpleAllReduceTiling": 0.07370853424072266,
"Simplifier": 0.09367704391479492,
"SimplifyMacroPredicates": 0.28262877464294434,
"SimplifyNeuronTensor": 0.9568257331848145,
"SimplifySlice": 0.027447223663330078,
"SimplifyTensor": 0.2734675407409668,
"SpillPSum": 1.2486350536346436,
"SplitAPUnionSets": 0.8627588748931885,
"SplitAccGrp": 0.1022341251373291,
"StaticProfiler": 0.4753732681274414,
"StaticTransposeLocalTensor": 0.6612956523895264,
"SundaISel": 1.6735970973968506,
"TCTransform": 0.029105186462402344,
"TensorInitialization": 0.1944262981414795,
"TensorOpSimplifier": 3.3376035690307617,
"TensorOpTransform": 19.926769256591797,
"TensorizerLegalizationPass": 0.0006489999941550195,
"TileCCOps": 0.1856687068939209,
"TilingProfiler": 0.5146362781524658,
"TransformConvOp": 1.4806337356567383,
"TritiumFusion": 0.2828052043914795,
"ValueNumbering": 0.08587479591369629,
"VectorizeDMA": 0.5646088123321533,
"VectorizeMatMult": 0.08723974227905273,
"VerifySupportedOps": 0.0004870000120718032,
"WeightCoalescing": 0.06492829322814941,
"ZeroSizeTensorElimination": 0.0008580684661865234,
"algsimp": 0.0014750000555068254,
"batchnorm_expander": 0.0015930000226944685,
"boundary-marker-removal": 0.0007229999755509198,
"call-inliner": 0.00018899999849963933,
"canonicalize-boundary-marker": 0.0011119999689981341,
"collective-stream-id-checker": 8.900000102585182e-05,
"comparison-expander": 0.0007440000190399587,
"computation-deduplicator": 0.0037249999586492777,
"config-lowering": 0.06129100173711777,
"constant_folding": 0.00012700000661425292,
"cse": 0.0007800000021234155,
"dce": 5.199999941396527e-05,
"dynamic-slice-transpose": 0.0002530000056140125,
"eliminate-redundant-compare": 0.0001140000022132881,
"emit-offloaded-dropout": 0.0006549999816343188,
"flatten-call-graph": 0.0013830000534653664,
"fuse-send-recv": 0.07738199830055237,
"hilo-conditional-to-select": 0.00014400000509340316,
"hilo::LegalizeAlias": 0.004616000223904848,
"hilo::NeuronInstCombine": 0.0015979999443516135,
"hilo::NeuronOpFusion": 0.0001250000059371814,
"hilo::ReplaceTokenTypeWithU8Pass": 0.0016670000040903687,
"hilo::ScheduleFusion": 9.999999747378752e-06,
"hilo::SixtyFourHack": 0.0008529999759048223,
"hilo::VerifyAliasing": 0.000291000003926456,
"hlo-mac-count": 0.02539600059390068,
"io-con-pipe-begin": 7.999999979801942e-06,
"io-con-pipe-end": 0.0,
"io-layout-normalization": 0.0035389999393373728,
"legalize-ccops-for-tensorizer": 0.00014200000441633165,
"legalize-compare": 0.0007239999831654131,
"lower-argminmax-custom-call": 0.0003229999856557697,
"map-inline": 0.0010969999711960554,
"metadata-naming": 0.009553000330924988,
"mlir::detail::OpToOpPassAdaptor": 9.999999747378752e-05,
"mlir::hlo::MhloToPyPenguin": 0.6092489957809448,
"mlir::mhlo::LowerComplexExtraPass": 0.0049149999395012856,
"mlir::mhlo::LowerComplexPass": 0.008914999663829803,
"native-to-custom-softmax": 0.0010550000006332994,
"native-to-custom-softmax-dx": 0.001585999969393015,
"neuron-hlo-verifier": 0.22506099939346313,
"operand_upcaster": 0.003352999920025468,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 0.0,
"post-partition-simplification": 0.37791401147842407,
"pre-hlo-begin": 3.999999989900971e-06,
"pre-hlo-end": 9.999999974752427e-07,
"replace-minimum-constant": 0.00019299999985378236,
"reshape-mover": 5.2999999752501026e-05,
"simplify-concat": 0.06822799891233444,
"simplify-while-loops": 5.0999999075429514e-05,
"transform-variadic-reduce": 0.0008849999867379665,
"tuple-simplifier": 0.00012099999730708078,
"unpack-nested-aws-ntwsr": 0.000615999975707382,
"unroll-while-loop": 7.000000096013537e-06
},
"hilo": {
"HloMacCount": 1854726144.0,
"Traffic": 1252630656.0
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 51674,
"StaticProfiler::AifUb": 21.735599517822266,
"StaticProfiler::ArithmeticIntensityTensorizer": 31.884082794189453,
"StaticProfiler::AverageDmaLength": 2843.9130859375,
"StaticProfiler::DDRTransferBytes": 1013018224,
"StaticProfiler::InternalTransferBytes": 226623072,
"StaticProfiler::LoadExpanded": 281500,
"StaticProfiler::StoreExpanded": 4273,
"StaticProfiler::TotalDMAExpanded": 285773,
"StaticProfiler::TotalDynamicInstancesCount": 67989,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 57540,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 123,
"TilingProfiler::MatMultInstructionsAfterTiling": 35008,
"TilingProfiler::NumPfTransposes": 292,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 142,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 10670,
"TilingProfiler::PfTransposeInstructionsForIo": 8360,
"TilingProfiler::PfTransposeInstructionsForLocal": 648,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 1662,
"TilingProfiler::ReduceInstructionsAfterTiling": 283,
"TilingProfiler::SimdInstructionsAfterTiling": 2805,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"all": {
"compiletime": {
"CanonicalizeConv": 4.999999873689376e-06,
"CanonicalizeForTensorizer": 0.00076299998909235,
"Canonicalizer": 0.08714199811220169,
"HoistCompute": 0.0,
"IdentifyCrossPassTensors": 0.0003549999964889139,
"MemcastMotion": 0.0,
"PenguinizeFunctions": 0.000754999986384064,
"PruneFunctions": 0.0012410000199452043,
"RemoveOptimizationBarriers": 0.0005370000144466758,
"ScatterMotion": 0.0,
"TensorizerLegalizationPass": 0.0006489999941550195,
"VerifySupportedOps": 0.0004870000120718032,
"algsimp": 0.0014750000555068254,
"batchnorm_expander": 0.0015930000226944685,
"boundary-marker-removal": 0.0007229999755509198,
"call-inliner": 0.00018899999849963933,
"canonicalize-boundary-marker": 0.0011119999689981341,
"collective-stream-id-checker": 8.900000102585182e-05,
"comparison-expander": 0.0007440000190399587,
"computation-deduplicator": 0.0037249999586492777,
"config-lowering": 0.06129100173711777,
"constant_folding": 0.00012700000661425292,
"cse": 0.0007800000021234155,
"dce": 5.199999941396527e-05,
"dynamic-slice-transpose": 0.0002530000056140125,
"eliminate-redundant-compare": 0.0001140000022132881,
"emit-offloaded-dropout": 0.0006549999816343188,
"flatten-call-graph": 0.0013830000534653664,
"fuse-send-recv": 0.07738199830055237,
"hilo-conditional-to-select": 0.00014400000509340316,
"hilo::LegalizeAlias": 0.004616000223904848,
"hilo::NeuronInstCombine": 0.0015979999443516135,
"hilo::NeuronOpFusion": 0.0001250000059371814,
"hilo::ReplaceTokenTypeWithU8Pass": 0.0016670000040903687,
"hilo::ScheduleFusion": 9.999999747378752e-06,
"hilo::SixtyFourHack": 0.0008529999759048223,
"hilo::VerifyAliasing": 0.000291000003926456,
"hlo-mac-count": 0.02539600059390068,
"io-con-pipe-begin": 7.999999979801942e-06,
"io-con-pipe-end": 0.0,
"io-layout-normalization": 0.0035389999393373728,
"legalize-ccops-for-tensorizer": 0.00014200000441633165,
"legalize-compare": 0.0007239999831654131,
"lower-argminmax-custom-call": 0.0003229999856557697,
"map-inline": 0.0010969999711960554,
"metadata-naming": 0.009553000330924988,
"mlir::detail::OpToOpPassAdaptor": 9.999999747378752e-05,
"mlir::hlo::MhloToPyPenguin": 0.6092489957809448,
"mlir::mhlo::LowerComplexExtraPass": 0.0049149999395012856,
"mlir::mhlo::LowerComplexPass": 0.008914999663829803,
"native-to-custom-softmax": 0.0010550000006332994,
"native-to-custom-softmax-dx": 0.001585999969393015,
"neuron-hlo-verifier": 0.22506099939346313,
"operand_upcaster": 0.003352999920025468,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 0.0,
"post-partition-simplification": 0.37791401147842407,
"pre-hlo-begin": 3.999999989900971e-06,
"pre-hlo-end": 9.999999974752427e-07,
"replace-minimum-constant": 0.00019299999985378236,
"reshape-mover": 5.2999999752501026e-05,
"simplify-concat": 0.06822799891233444,
"simplify-while-loops": 5.0999999075429514e-05,
"transform-variadic-reduce": 0.0008849999867379665,
"tuple-simplifier": 0.00012099999730708078,
"unpack-nested-aws-ntwsr": 0.000615999975707382,
"unroll-while-loop": 7.000000096013537e-06
}
},
"cumsum": {
"compiletime": {
"CoalesceCCOp": 0.00023365020751953125,
"DMALocalityOpt": 0.00019097328186035156,
"DMAProfiler": 0.0009503364562988281,
"DataStreaming": 0.00032329559326171875,
"DoNothing": 0.00015044212341308594,
"ExpandISAMacro": 0.0005333423614501953,
"FactorizeBlkDims": 0.0005185604095458984,
"InferPSumTensor": 0.0006201267242431641,
"InferSharedMemLoc": 0.0002682209014892578,
"InsertCoreBarrier": 0.0003275871276855469,
"LateLegalizeInst": 0.0004162788391113281,
"LateNeuronInstComb": 0.0005996227264404297,
"LegalizeSundaAccess": 0.0016045570373535156,
"LegalizeType": 0.0002541542053222656,
"LowerBroadcast": 0.00023365020751953125,
"LowerIntrinsics": 0.00023651123046875,
"LowerTranspose": 0.0002307891845703125,
"NeuronInstComb": 0.0006008148193359375,
"NeuronLICM": 0.00036597251892089844,
"NeuronSimplifyPredicates": 0.002484560012817383,
"NeuronValueNumbering": 0.0004477500915527344,
"SFKVectorizer": 0.002768278121948242,
"SimpleAllReduceTiling": 0.00021409988403320313,
"SimplifyNeuronTensor": 0.0005936622619628906,
"SpillPSum": 0.0005390644073486328,
"WeightCoalescing": 0.00023293495178222656
}
},
"sg00": {
"hilo": {
"ArithmeticIntensity": 2.961329698562622,
"HloMacCount": 1854726144.0,
"Traffic": 1252630656.0
}
},
"sg0000": {
"compiletime": {
"AGOrderingAnalysisPass": 6.388453960418701,
"AffinePredicateResolution": 0.40378904342651367,
"AliasDependencyElimination": 0.016310930252075195,
"AliasDependencyInduction": 15.282819747924805,
"AliasDependencyReset": 16.618453979492188,
"BFComputeCutting": 0.1391315460205078,
"BirCodeGenLoop": 4.402980804443359,
"CCOpFusion": 0.8881416320800781,
"CanonicalizeDAGForPGTiling": 0.20721173286437988,
"CanonicalizeIR": 1.028043270111084,
"CoalesceCCOp": 0.23186254501342773,
"CommuteConcat": 0.0287783145904541,
"DMALocalityOpt": 0.07128620147705078,
"DMAProfiler": 0.07605504989624023,
"DMATilingProfiler": 0.09187006950378418,
"DataLocalityOpt": 2.820284605026245,
"DataStreaming": 0.42258405685424805,
"DeConcat": 0.06640625,
"DeadCodeElimination": 0.053946495056152344,
"DeadStoreElimination": 1.0540196895599365,
"DelinearIndices": 0.3954615592956543,
"Delinearization": 0.13945698738098145,
"DelinearizeSPMD": 0.17487859725952148,
"DoNothing": 0.00014019012451171875,
"DramToDramTranspose": 0.34574389457702637,
"DumpGraphAndMetadata": 0.23639750480651855,
"EliminateDivs": 1.4790117740631104,
"ExpandBatchNorm": 1.123610019683838,
"ExpandISAMacro": 0.11814379692077637,
"FactorizeBlkDims": 0.4797940254211426,
"FactorizeThreadAxesInFreeDims": 0.08273863792419434,
"FlattenMacroLoop": 0.08472108840942383,
"GenericAccessSimplifier": 0.025115966796875,
"InferInitValue": 1.3054986000061035,
"InferIntrinsicOnCC": 0.7679235935211182,
"InferNeuronTensor": 1.671952724456787,
"InferNonlocalTensors": 7.110470771789551,
"InferPSumTensor": 4.2699432373046875,
"InferShardAxis": 12.786405563354492,
"InferSharedMemLoc": 0.16634631156921387,
"InlineNativeKernels": 0.05566692352294922,
"InsertCoreBarrier": 0.1479203701019287,
"InsertIOTransposes": 2.3931386470794678,
"InsertImplicitShardAxisBeforeISel": 0.5525217056274414,
"InsertLocalTransposes": 0.7429313659667969,
"InsertOffloadedTransposes": 0.19022369384765625,
"LICM": 0.14770793914794922,
"LateLegalizeInst": 0.30222225189208984,
"LateLegalizePostSplit": 0.22165155410766602,
"LateLowerReshapeOp": 0.03522944450378418,
"LateLowerTensorOp": 6.813569068908691,
"LateNeuronInstComb": 1.2631995677947998,
"LayoutPreprocessing": 1.8736376762390137,
"LayoutPreprocessingAndAnalysis": 2.3256783485412598,
"LayoutRequirementAnalysis": 0.4448714256286621,
"LegalizeCCOpLayout": 1.317772388458252,
"LegalizeOpLevelAlias": 0.5106728076934814,
"LegalizePartitionReduce": 0.08073091506958008,
"LegalizeSundaAccess": 0.9303104877471924,
"LegalizeSundaMacro": 0.7451527118682861,
"LegalizeType": 0.12957215309143066,
"LocalLayoutOpt": 0.6608095169067383,
"LoopFusion": 0.3101928234100342,
"LoopSplitting": 0.036502838134765625,
"LowerBroadcast": 0.06164741516113281,
"LowerCCOpBlockAxis": 0.32977890968322754,
"LowerComplexBroadcast": 0.15041875839233398,
"LowerIntrinsics": 1.0597736835479736,
"LowerShardAxis": 0.4116206169128418,
"LowerTensorOp": 3.285423755645752,
"LowerToSendRecv": 0.25904369354248047,
"LowerTranspose": 0.5202181339263916,
"MacroGeneration": 2.1192569732666016,
"MaskPropagation": 0.14647436141967773,
"MemcpyElimination": 29.3009090423584,
"MutateDataType": 0.037621498107910156,
"NeuronAliasDependencyInduction": 0.06313967704772949,
"NeuronAliasDependencyReset": 0.08403873443603516,
"NeuronInstComb": 0.3526570796966553,
"NeuronLICM": 0.28780031204223633,
"NeuronLoopFusion": 1.491703987121582,
"NeuronLoopInterchange": 0.06900668144226074,
"NeuronSimplifier": 0.5315425395965576,
"NeuronSimplifyPredicates": 0.25080394744873047,
"NeuronValueNumbering": 0.11057853698730469,
"OptimizeAliasedCopyChain": 0.6118557453155518,
"OptimizeNKIKernels": 1.0786433219909668,
"PAGLayoutOpt": 18.570398330688477,
"PComputeCutting": 0.6222004890441895,
"PGLayoutTilingPipeline": 55.62285614013672,
"PGTiling": 10.0595703125,
"PadElimination": 0.021188735961914063,
"ParAxesAnnotation": 17.818796157836914,
"PartialLoopFusion": 1.5088038444519043,
"PartialSimdFusion": 0.8573262691497803,
"PerfectLoopNest": 0.07143402099609375,
"RecognizeOpIdiom": 0.1291654109954834,
"Recompute": 0.007833480834960938,
"RelaxPredicates": 0.14508056640625,
"Rematerialization": 0.17157793045043945,
"RemoveShardedPartitionAxes": 1.3566563129425049,
"ReshapeWeights": 0.022662878036499023,
"ResolveAccessConflict": 0.2033238410949707,
"ResolveComplicatePredicates": 0.49439358711242676,
"RewriteReplicationMatmul": 0.03938794136047363,
"RewriteWeights": 0.07809877395629883,
"SFKVectorizer": 10.313403129577637,
"ShardingPropagationAnalysis": 0.7199399471282959,
"SimpleAllReduceTiling": 0.06984925270080566,
"Simplifier": 0.09367704391479492,
"SimplifyMacroPredicates": 0.28262877464294434,
"SimplifyNeuronTensor": 0.9048190116882324,
"SimplifySlice": 0.027447223663330078,
"SimplifyTensor": 0.2734675407409668,
"SpillPSum": 1.2256271839141846,
"SplitAPUnionSets": 0.8627588748931885,
"SplitAccGrp": 0.1022341251373291,
"StaticProfiler": 0.4753732681274414,
"StaticTransposeLocalTensor": 0.6612956523895264,
"SundaISel": 1.6735970973968506,
"TCTransform": 0.029105186462402344,
"TensorInitialization": 0.1944262981414795,
"TensorOpSimplifier": 3.3376035690307617,
"TensorOpTransform": 19.926769256591797,
"TileCCOps": 0.1856687068939209,
"TilingProfiler": 0.5146362781524658,
"TransformConvOp": 1.4806337356567383,
"TritiumFusion": 0.2828052043914795,
"ValueNumbering": 0.08587479591369629,
"VectorizeDMA": 0.5646088123321533,
"VectorizeMatMult": 0.08723974227905273,
"WeightCoalescing": 0.06122708320617676,
"ZeroSizeTensorElimination": 0.0008580684661865234
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 51674,
"StaticProfiler::AifUb": 21.735599517822266,
"StaticProfiler::ArithmeticIntensityTensorizer": 31.884082794189453,
"StaticProfiler::AverageDmaLength": 2843.9130859375,
"StaticProfiler::AverageFractalPeUtilization": 97.88961791992188,
"StaticProfiler::AveragePartitionUtilization": 89.79463958740234,
"StaticProfiler::AveragePeUtilization": 78.904052734375,
"StaticProfiler::DDRTransferBytes": 1013018224,
"StaticProfiler::InternalTransferBytes": 226623072,
"StaticProfiler::LoadExpanded": 281500,
"StaticProfiler::LocalizationEfficiency": 146.6905975341797,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 151.06674194335938,
"StaticProfiler::StoreExpanded": 4273,
"StaticProfiler::TotalDMAExpanded": 285773,
"StaticProfiler::TotalDynamicInstancesCount": 67989,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 57540,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 123,
"TilingProfiler::MatMultInstructionsAfterTiling": 35008,
"TilingProfiler::NumPfTransposes": 292,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 142,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 10670,
"TilingProfiler::PfTransposeInstructionsForIo": 8360,
"TilingProfiler::PfTransposeInstructionsForLocal": 648,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 1662,
"TilingProfiler::ReduceInstructionsAfterTiling": 283,
"TilingProfiler::SimdInstructionsAfterTiling": 2805,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"topk": {
"compiletime": {
"CoalesceCCOp": 0.003665924072265625,
"DMALocalityOpt": 0.002828359603881836,
"DMAProfiler": 0.0035741329193115234,
"DataStreaming": 0.006661653518676758,
"DoNothing": 0.00015878677368164063,
"ExpandISAMacro": 0.0038259029388427734,
"FactorizeBlkDims": 0.012358903884887695,
"InferPSumTensor": 0.01241612434387207,
"InferSharedMemLoc": 0.0028722286224365234,
"InsertCoreBarrier": 0.0033416748046875,
"LateLegalizeInst": 0.0075190067291259766,
"LateNeuronInstComb": 0.008240699768066406,
"LegalizeSundaAccess": 0.014475584030151367,
"LegalizeType": 0.00968170166015625,
"LowerBroadcast": 0.0032701492309570313,
"LowerIntrinsics": 0.0037560462951660156,
"LowerTranspose": 0.0033812522888183594,
"NeuronInstComb": 0.008539199829101563,
"NeuronLICM": 0.00936579704284668,
"NeuronSimplifyPredicates": 0.0037877559661865234,
"NeuronValueNumbering": 0.0040531158447265625,
"SFKVectorizer": 0.03392672538757324,
"SimpleAllReduceTiling": 0.003645181655883789,
"SimplifyNeuronTensor": 0.05141305923461914,
"SpillPSum": 0.02246880531311035,
"WeightCoalescing": 0.0034682750701904297
}
}
}

3
token_generation_model/_tp0_bk3/graph.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:0c81d96024b571f29492feb3e8f54a86bb85699475d770c8027b26b3b25d6d30
size 3605504

4471
token_generation_model/_tp0_bk3/log-neuron-cc.txt Normal file
View File

File diff suppressed because it is too large Load Diff

3
token_generation_model/_tp0_bk3/metaneff.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:6f9340d294060545b7fd3affe72e9160857b035623ad5e43d5186704d48fc4cd
size 2464555

3
token_generation_model/_tp0_bk3/model.MODULE_02e08d5fdadfbcb8a569+a6f96e4f.hlo_module.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:25287d0dc180171fa1e47e9880d832a32168762120f79d5891ed820ee9b4fcda
size 2550451

3
token_generation_model/_tp0_bk3/model.MODULE_02e08d5fdadfbcb8a569+a6f96e4f.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:0c81d96024b571f29492feb3e8f54a86bb85699475d770c8027b26b3b25d6d30
size 3605504

224
token_generation_model/_tp0_bk3/neuron_config.json Normal file
View File

@@ -0,0 +1,224 @@
{
"_attn_implementation_autoset": false,
"_name_or_path": "/models/Qwen3-1.7B/",
"add_cross_attention": false,
"architectures": [
"Qwen3ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"attribute_map": {},
"bad_words_ids": null,
"begin_suppress_tokens": null,
"bos_token_id": 151643,
"chunk_size_feed_forward": 0,
"cross_attention_hidden_size": null,
"decoder_start_token_id": null,
"diversity_penalty": 0.0,
"do_sample": false,
"early_stopping": false,
"encoder_no_repeat_ngram_size": 0,
"eos_token_id": 151645,
"exponential_decay_length_penalty": null,
"finetuning_task": null,
"forced_bos_token_id": null,
"forced_eos_token_id": null,
"fused_spec_config": null,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 2048,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"initializer_range": 0.02,
"intermediate_size": 6144,
"is_decoder": false,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"length_penalty": 1.0,
"max_length": 20,
"max_position_embeddings": 40960,
"max_window_layers": 28,
"metadata": null,
"min_length": 0,
"model_type": "qwen3",
"neuron_config": {
"activation_quantization_type": null,
"allow_input_truncation": false,
"apply_seq_ids_mask": false,
"async_mode": false,
"attention_dp_degree": 1,
"attention_dtype": null,
"attn_block_cte_nki_kernel_enabled": false,
"attn_block_tkg_nki_kernel_cache_update": false,
"attn_block_tkg_nki_kernel_cascaded_attention": false,
"attn_block_tkg_nki_kernel_enabled": false,
"attn_cls": {
"__module__": "neuronx_distributed_inference.models.qwen3.modeling_qwen3",
"__name__": "NeuronQwen3Attention"
},
"attn_kernel_enabled": null,
"attn_tkg_builtin_kernel_enabled": false,
"attn_tkg_nki_kernel_enabled": false,
"batch_size": 4,
"bucket_n_active_tokens": false,
"buckets": [
1024
],
"cast_type": "config",
"cc_pipeline_tiling_factor": 1,
"chunked_prefill_config": null,
"context_encoding_buckets": null,
"cp_degree": 1,
"ctx_batch_size": 1,
"disable_kv_cache_tiling": false,
"draft_model_modules_to_not_convert": null,
"enable_bucketing": true,
"enable_cte_modular_flow": false,
"enable_eagle_draft_input_norm": false,
"enable_eagle_speculation": false,
"enable_fused_speculation": false,
"enable_long_context_mode": false,
"enable_output_completion_notifications": false,
"enable_spill_reload_dge": false,
"enable_token_tree": false,
"ep_degree": 1,
"expert_mlp_nki_kernel_enabled": null,
"flash_decoding_enabled": false,
"fused_qkv": false,
"fused_rmsnorm_skip_gamma": false,
"is_block_kv_layout": null,
"is_chunked_prefill": false,
"is_continuous_batching": true,
"is_eagle_draft": false,
"is_medusa": false,
"is_prefill_stage": false,
"is_prefix_caching": false,
"k_cache_transposed": false,
"kv_cache_batch_size": 4,
"kv_cache_padding_size": 0,
"kv_cache_quant": false,
"kv_cache_tiling": false,
"layer_boundary_markers": false,
"lm_head_pad": true,
"lm_head_pad_alignment_size": 1,
"local_ranks_size": 4,
"logical_nc_config": 2,
"lora_config": null,
"max_batch_size": 4,
"max_context_length": 2048,
"max_length": 2048,
"max_new_tokens": null,
"medusa_speculation_length": 0,
"medusa_tree": null,
"mlp_kernel_enabled": false,
"mlp_kernel_fuse_residual_add": false,
"modules_to_not_convert": null,
"moe_fused_nki_kernel_enabled": null,
"n_active_tokens": 1,
"n_positions": 2048,
"num_medusa_heads": 0,
"on_cpu": false,
"on_device_sampling_config": {
"deterministic": false,
"do_sample": false,
"dynamic": true,
"global_topk": 256,
"on_device_sampling_config": true,
"temperature": 1.0,
"top_k": 1,
"top_k_kernel_enabled": false,
"top_p": 1.0
},
"output_logits": false,
"overrides_torch_dtype": true,
"pa_block_size": 2048,
"pa_num_blocks": 4,
"padding_side": "right",
"pp_degree": 1,
"prefix_buckets": null,
"qk_layernorm": false,
"qkv_kernel_enabled": false,
"qkv_kernel_fuse_residual_add": false,
"qkv_kernel_nbsd_layout": false,
"quantization_dtype": "int8",
"quantization_type": "per_tensor_symmetric",
"quantize_clamp_bound": Infinity,
"quantized": false,
"quantized_checkpoints_path": null,
"quantized_mlp_kernel_enabled": false,
"rmsnorm_quantize_kernel_enabled": false,
"router_topk_nki_kernel_enabled": null,
"rpl_reduce_dtype": null,
"save_sharded_checkpoint": true,
"scratchpad_page_size": null,
"seq_len": 2048,
"seq_len_threshold_for_cc_tiling": 16384,
"sequence_parallel_enabled": false,
"shared_mlp_nki_kernel_enabled": null,
"skip_sharding": false,
"skip_warmup": false,
"spec_batch_size": 4,
"speculation_length": 0,
"start_rank_id": 0,
"strided_context_parallel_kernel_enabled": false,
"target": null,
"tensor_capture_config": null,
"tile_cc": false,
"tkg_batch_size": 4,
"token_generation_buckets": [
1024
],
"token_tree_config": null,
"torch_dtype": "bfloat16",
"tp_degree": 4,
"vocab_parallel": false,
"weight_gather_seq_len_threshold": 32768,
"weights_to_skip_layout_optimization": [],
"world_size": 4
},
"no_repeat_ngram_size": 0,
"num_attention_heads": 16,
"num_beam_groups": 1,
"num_beams": 1,
"num_cores_per_group": 1,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"num_return_sequences": 1,
"output_attentions": false,
"output_hidden_states": false,
"output_scores": false,
"pad_token_id": 0,
"prefix": null,
"problem_type": null,
"pruned_heads": {},
"remove_invalid_values": false,
"repetition_penalty": 1.0,
"return_dict": true,
"return_dict_in_generate": false,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000,
"sep_token_id": null,
"sliding_window": null,
"suppress_tokens": null,
"task_specific_params": null,
"temperature": 1.0,
"tf_legacy_loss": false,
"tie_encoder_decoder": false,
"tie_word_embeddings": true,
"tokenizer_class": null,
"top_k": 50,
"top_p": 1.0,
"torchscript": false,
"transformers_version": "4.51.0",
"typical_p": 1.0,
"use_bfloat16": false,
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}

1
token_generation_model/_tp0_bk4/command.txt Normal file
View File

@@ -0,0 +1 @@
neuronx-cc compile --framework=XLA model.MODULE_b4b44a47076a167bd9a5+795ee7cc.hlo_module.pb --output model.MODULE_b4b44a47076a167bd9a5+795ee7cc.neff --target=trn2 --auto-cast=none --model-type=transformer '--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ' --lnc=2 -O2 --internal-hlo2tensorizer-options=--verify-hlo=true --logfile=log-neuron-cc.txt --verbose=35

1
token_generation_model/_tp0_bk4/compile_flags.MODULE_b4b44a47076a167bd9a5+795ee7cc.json Normal file
View File

@@ -0,0 +1 @@
["--target=trn2", "--auto-cast=none", "--model-type=transformer", "--tensorizer-options=--enable-ccop-compute-overlap --cc-pipeline-tiling-factor=1 --vectorize-strided-dma ", "--lnc=2", "-O2", "--internal-hlo2tensorizer-options=--verify-hlo=true", "--logfile=/models/Qwen3-1.7B-compiled/token_generation_model/_tp0_bk4/log-neuron-cc.txt"]

590
token_generation_model/_tp0_bk4/global_metric_store.json Normal file
View File

@@ -0,0 +1,590 @@
{
"Average": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 98.06681823730469,
"StaticProfiler::AveragePartitionUtilization": 88.63555145263672,
"StaticProfiler::AveragePeUtilization": 75.70149230957031,
"StaticProfiler::LocalizationEfficiency": 131.45083618164063,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 134.9540252685547,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0
}
},
"Count": {
"tensorizer": {
"StaticProfiler::AverageFractalPeUtilization": 1,
"StaticProfiler::AveragePartitionUtilization": 1,
"StaticProfiler::AveragePeUtilization": 1,
"StaticProfiler::LocalizationEfficiency": 1,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 1,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 1,
"TilingProfiler::AveragePeUtilizationAfterTiling": 1
}
},
"Sum": {
"compiletime": {
"AGOrderingAnalysisPass": 5.95287561416626,
"AffinePredicateResolution": 0.49277400970458984,
"AliasDependencyElimination": 0.01732039451599121,
"AliasDependencyInduction": 18.183685302734375,
"AliasDependencyReset": 18.999177932739258,
"BFComputeCutting": 0.3126344680786133,
"BirCodeGenLoop": 1.8802330493927002,
"CCOpFusion": 0.8809163570404053,
"CanonicalizeConv": 0.0,
"CanonicalizeDAGForPGTiling": 0.17032384872436523,
"CanonicalizeForTensorizer": 0.0005619999719783664,
"CanonicalizeIR": 0.7969982624053955,
"Canonicalizer": 0.011158999986946583,
"CoalesceCCOp": 0.6349403858184814,
"CommuteConcat": 0.02859354019165039,
"DMALocalityOpt": 0.03904151916503906,
"DMAProfiler": 0.14334964752197266,
"DMATilingProfiler": 0.08906126022338867,
"DataLocalityOpt": 4.2657470703125,
"DataStreaming": 0.1790633201599121,
"DeConcat": 0.06563258171081543,
"DeadCodeElimination": 0.0297544002532959,
"DeadStoreElimination": 1.2178454399108887,
"DelinearIndices": 0.35585713386535645,
"Delinearization": 0.6069140434265137,
"DelinearizeSPMD": 0.6448159217834473,
"DoNothing": 0.0005474090576171875,
"DramToDramTranspose": 0.3358180522918701,
"DumpGraphAndMetadata": 0.17146539688110352,
"EliminateDivs": 1.3107364177703857,
"ExpandBatchNorm": 0.48049426078796387,
"ExpandISAMacro": 0.08499646186828613,
"FactorizeBlkDims": 0.6006636619567871,
"FactorizeThreadAxesInFreeDims": 0.09288716316223145,
"FlattenMacroLoop": 0.0829305648803711,
"GenericAccessSimplifier": 0.026407241821289063,
"HoistCompute": 9.699999645818025e-05,
"IdentifyCrossPassTensors": 0.00019700000120792538,
"InferInitValue": 1.2918815612792969,
"InferIntrinsicOnCC": 0.31558942794799805,
"InferNeuronTensor": 1.773860216140747,
"InferNonlocalTensors": 6.451204299926758,
"InferPSumTensor": 1.442307472229004,
"InferShardAxis": 10.398513793945313,
"InferSharedMemLoc": 0.12981677055358887,
"InlineNativeKernels": 0.051079511642456055,
"InsertCoreBarrier": 0.3666720390319824,
"InsertIOTransposes": 1.088529348373413,
"InsertImplicitShardAxisBeforeISel": 0.4208717346191406,
"InsertLocalTransposes": 0.6978082656860352,
"InsertOffloadedTransposes": 0.13763642311096191,
"LICM": 0.11658525466918945,
"LateLegalizeInst": 0.5649778842926025,
"LateLegalizePostSplit": 0.10899925231933594,
"LateLowerReshapeOp": 0.03313422203063965,
"LateLowerTensorOp": 6.405031204223633,
"LateNeuronInstComb": 1.0954866409301758,
"LayoutPreprocessing": 1.1262352466583252,
"LayoutPreprocessingAndAnalysis": 1.4945013523101807,
"LayoutRequirementAnalysis": 0.36240172386169434,
"LegalizeCCOpLayout": 1.2990221977233887,
"LegalizeOpLevelAlias": 1.2160041332244873,
"LegalizePartitionReduce": 0.0940859317779541,
"LegalizeSundaAccess": 3.4260854721069336,
"LegalizeSundaMacro": 0.6391239166259766,
"LegalizeType": 0.1378321647644043,
"LocalLayoutOpt": 0.6706066131591797,
"LoopFusion": 0.28072071075439453,
"LoopSplitting": 0.07517576217651367,
"LowerBroadcast": 0.06826472282409668,
"LowerCCOpBlockAxis": 0.20669817924499512,
"LowerComplexBroadcast": 0.1663343906402588,
"LowerIntrinsics": 0.9762301445007324,
"LowerShardAxis": 0.24805045127868652,
"LowerTensorOp": 3.8965094089508057,
"LowerToSendRecv": 0.18735170364379883,
"LowerTranspose": 0.4830780029296875,
"MacroGeneration": 2.835282802581787,
"MaskPropagation": 0.1340806484222412,
"MemcastMotion": 0.00024199999461416155,
"MemcpyElimination": 28.533254623413086,
"MutateDataType": 0.05120110511779785,
"NeuronAliasDependencyInduction": 0.05350494384765625,
"NeuronAliasDependencyReset": 0.07738471031188965,
"NeuronInstComb": 0.4167485237121582,
"NeuronLICM": 0.2737100124359131,
"NeuronLoopFusion": 1.1017746925354004,
"NeuronLoopInterchange": 0.05889534950256348,
"NeuronSimplifier": 0.5097112655639648,
"NeuronSimplifyPredicates": 0.2256777286529541,
"NeuronValueNumbering": 0.1252288818359375,
"OptimizeAliasedCopyChain": 0.31018686294555664,
"OptimizeNKIKernels": 1.1799430847167969,
"PAGLayoutOpt": 17.110271453857422,
"PComputeCutting": 1.0609712600708008,
"PGLayoutTilingPipeline": 50.19084930419922,
"PGTiling": 11.0619478225708,
"PadElimination": 0.012235879898071289,
"ParAxesAnnotation": 16.403249740600586,
"PartialLoopFusion": 1.436572551727295,
"PartialSimdFusion": 0.8934595584869385,
"PenguinizeFunctions": 0.0003769999893847853,
"PerfectLoopNest": 0.07143855094909668,
"PruneFunctions": 0.00010900000052060932,
"RecognizeOpIdiom": 0.13362383842468262,
"Recompute": 0.0345001220703125,
"RelaxPredicates": 0.11409568786621094,
"Rematerialization": 0.22487568855285645,
"RemoveOptimizationBarriers": 0.00021300000662449747,
"RemoveShardedPartitionAxes": 0.8885042667388916,
"ReshapeWeights": 0.02379584312438965,
"ResolveAccessConflict": 0.46589231491088867,
"ResolveComplicatePredicates": 0.5794906616210938,
"RewriteReplicationMatmul": 0.040407657623291016,
"RewriteWeights": 0.08104610443115234,
"SFKVectorizer": 11.748018264770508,
"ScatterMotion": 0.004219000227749348,
"ShardingPropagationAnalysis": 0.7053203582763672,
"SimpleAllReduceTiling": 0.22031855583190918,
"Simplifier": 0.12051248550415039,
"SimplifyMacroPredicates": 0.27616286277770996,
"SimplifyNeuronTensor": 0.460674524307251,
"SimplifySlice": 0.026907682418823242,
"SimplifyTensor": 0.25261998176574707,
"SpillPSum": 0.7110695838928223,
"SplitAPUnionSets": 0.49262213706970215,
"SplitAccGrp": 0.04794502258300781,
"StaticProfiler": 0.13262128829956055,
"StaticTransposeLocalTensor": 0.6367781162261963,
"SundaISel": 1.6440446376800537,
"TCTransform": 0.030546188354492188,
"TensorInitialization": 0.24484515190124512,
"TensorOpSimplifier": 3.413067102432251,
"TensorOpTransform": 18.72884178161621,
"TensorizerLegalizationPass": 0.0004299999854993075,
"TileCCOps": 0.18365025520324707,
"TilingProfiler": 0.8507044315338135,
"TransformConvOp": 1.1984179019927979,
"TritiumFusion": 1.2713313102722168,
"ValueNumbering": 0.08294677734375,
"VectorizeDMA": 2.601423501968384,
"VectorizeMatMult": 0.10021328926086426,
"VerifySupportedOps": 0.0005520000122487545,
"WeightCoalescing": 0.06738448143005371,
"ZeroSizeTensorElimination": 0.00041866302490234375,
"algsimp": 0.0031190000008791685,
"batchnorm_expander": 0.0024679999332875013,
"boundary-marker-removal": 0.0005530000198632479,
"call-inliner": 0.00031400000443682075,
"canonicalize-boundary-marker": 0.000981000019237399,
"collective-stream-id-checker": 0.0006099999882280827,
"comparison-expander": 0.00041000000783242285,
"computation-deduplicator": 0.0009740000241436064,
"config-lowering": 0.00040600000647827983,
"constant_folding": 0.00021100000594742596,
"cse": 0.0007169999880716205,
"dce": 6.500000017695129e-05,
"dynamic-slice-transpose": 0.0003060000017285347,
"eliminate-redundant-compare": 0.00017600000137463212,
"emit-offloaded-dropout": 0.00033400001120753586,
"flatten-call-graph": 0.0003760000108741224,
"fuse-send-recv": 0.011651000007987022,
"hilo-conditional-to-select": 0.00012700000661425292,
"hilo::LegalizeAlias": 0.0029070000164210796,
"hilo::NeuronInstCombine": 0.0015899999998509884,
"hilo::NeuronOpFusion": 0.000295000005280599,
"hilo::ReplaceTokenTypeWithU8Pass": 0.00023499999952036887,
"hilo::ScheduleFusion": 4.600000102072954e-05,
"hilo::SixtyFourHack": 0.0004560000088531524,
"hilo::VerifyAliasing": 0.00018600000475998968,
"hlo-mac-count": 0.02524399943649769,
"io-con-pipe-begin": 0.00023999999393709004,
"io-con-pipe-end": 9.999999974752427e-07,
"io-layout-normalization": 0.08203800022602081,
"legalize-ccops-for-tensorizer": 0.00014600000577047467,
"legalize-compare": 0.0009549999958835542,
"lower-argminmax-custom-call": 0.0003000000142492354,
"map-inline": 0.0008740000193938613,
"metadata-naming": 0.06593800336122513,
"mlir::detail::OpToOpPassAdaptor": 0.00038400001358240843,
"mlir::hlo::MhloToPyPenguin": 0.5922709703445435,
"mlir::mhlo::LowerComplexExtraPass": 0.005900000222027302,
"mlir::mhlo::LowerComplexPass": 0.004220999777317047,
"native-to-custom-softmax": 0.0007440000190399587,
"native-to-custom-softmax-dx": 0.0006929999799467623,
"neuron-hlo-verifier": 0.3141849935054779,
"operand_upcaster": 0.0021929999347776175,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 0.0,
"post-partition-simplification": 0.3108839988708496,
"pre-hlo-begin": 1.9999999494757503e-05,
"pre-hlo-end": 9.999999974752427e-07,
"replace-minimum-constant": 0.0002690000110305846,
"reshape-mover": 0.00011000000085914508,
"simplify-concat": 0.06884700059890747,
"simplify-while-loops": 8.499999967170879e-05,
"transform-variadic-reduce": 0.0008679999737069011,
"tuple-simplifier": 0.00019299999985378236,
"unpack-nested-aws-ntwsr": 0.00036700000055134296,
"unroll-while-loop": 1.1000000085914508e-05
},
"hilo": {
"HloMacCount": 1972166656.0,
"Traffic": 1252647040.0
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 63618,
"StaticProfiler::AifUb": 27.032575607299805,
"StaticProfiler::ArithmeticIntensityTensorizer": 35.5345458984375,
"StaticProfiler::AverageDmaLength": 2744.118896484375,
"StaticProfiler::DDRTransferBytes": 1130475120,
"StaticProfiler::InternalTransferBytes": 288112224,
"StaticProfiler::LoadExpanded": 338844,
"StaticProfiler::StoreExpanded": 4273,
"StaticProfiler::TotalDMAExpanded": 343117,
"StaticProfiler::TotalDynamicInstancesCount": 81003,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 66628,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 123,
"TilingProfiler::MatMultInstructionsAfterTiling": 40384,
"TilingProfiler::NumPfTransposes": 292,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 142,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 14486,
"TilingProfiler::PfTransposeInstructionsForIo": 11952,
"TilingProfiler::PfTransposeInstructionsForLocal": 872,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 1662,
"TilingProfiler::ReduceInstructionsAfterTiling": 563,
"TilingProfiler::SimdInstructionsAfterTiling": 3485,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"all": {
"compiletime": {
"CanonicalizeConv": 0.0,
"CanonicalizeForTensorizer": 0.0005619999719783664,
"Canonicalizer": 0.011158999986946583,
"HoistCompute": 9.699999645818025e-05,
"IdentifyCrossPassTensors": 0.00019700000120792538,
"MemcastMotion": 0.00024199999461416155,
"PenguinizeFunctions": 0.0003769999893847853,
"PruneFunctions": 0.00010900000052060932,
"RemoveOptimizationBarriers": 0.00021300000662449747,
"ScatterMotion": 0.004219000227749348,
"TensorizerLegalizationPass": 0.0004299999854993075,
"VerifySupportedOps": 0.0005520000122487545,
"algsimp": 0.0031190000008791685,
"batchnorm_expander": 0.0024679999332875013,
"boundary-marker-removal": 0.0005530000198632479,
"call-inliner": 0.00031400000443682075,
"canonicalize-boundary-marker": 0.000981000019237399,
"collective-stream-id-checker": 0.0006099999882280827,
"comparison-expander": 0.00041000000783242285,
"computation-deduplicator": 0.0009740000241436064,
"config-lowering": 0.00040600000647827983,
"constant_folding": 0.00021100000594742596,
"cse": 0.0007169999880716205,
"dce": 6.500000017695129e-05,
"dynamic-slice-transpose": 0.0003060000017285347,
"eliminate-redundant-compare": 0.00017600000137463212,
"emit-offloaded-dropout": 0.00033400001120753586,
"flatten-call-graph": 0.0003760000108741224,
"fuse-send-recv": 0.011651000007987022,
"hilo-conditional-to-select": 0.00012700000661425292,
"hilo::LegalizeAlias": 0.0029070000164210796,
"hilo::NeuronInstCombine": 0.0015899999998509884,
"hilo::NeuronOpFusion": 0.000295000005280599,
"hilo::ReplaceTokenTypeWithU8Pass": 0.00023499999952036887,
"hilo::ScheduleFusion": 4.600000102072954e-05,
"hilo::SixtyFourHack": 0.0004560000088531524,
"hilo::VerifyAliasing": 0.00018600000475998968,
"hlo-mac-count": 0.02524399943649769,
"io-con-pipe-begin": 0.00023999999393709004,
"io-con-pipe-end": 9.999999974752427e-07,
"io-layout-normalization": 0.08203800022602081,
"legalize-ccops-for-tensorizer": 0.00014600000577047467,
"legalize-compare": 0.0009549999958835542,
"lower-argminmax-custom-call": 0.0003000000142492354,
"map-inline": 0.0008740000193938613,
"metadata-naming": 0.06593800336122513,
"mlir::detail::OpToOpPassAdaptor": 0.00038400001358240843,
"mlir::hlo::MhloToPyPenguin": 0.5922709703445435,
"mlir::mhlo::LowerComplexExtraPass": 0.005900000222027302,
"mlir::mhlo::LowerComplexPass": 0.004220999777317047,
"native-to-custom-softmax": 0.0007440000190399587,
"native-to-custom-softmax-dx": 0.0006929999799467623,
"neuron-hlo-verifier": 0.3141849935054779,
"operand_upcaster": 0.0021929999347776175,
"post-par-pipe-begin": 9.999999974752427e-07,
"post-par-pipe-end": 0.0,
"post-partition-simplification": 0.3108839988708496,
"pre-hlo-begin": 1.9999999494757503e-05,
"pre-hlo-end": 9.999999974752427e-07,
"replace-minimum-constant": 0.0002690000110305846,
"reshape-mover": 0.00011000000085914508,
"simplify-concat": 0.06884700059890747,
"simplify-while-loops": 8.499999967170879e-05,
"transform-variadic-reduce": 0.0008679999737069011,
"tuple-simplifier": 0.00019299999985378236,
"unpack-nested-aws-ntwsr": 0.00036700000055134296,
"unroll-while-loop": 1.1000000085914508e-05
}
},
"cumsum": {
"compiletime": {
"CoalesceCCOp": 0.0003490447998046875,
"DMALocalityOpt": 0.00033211708068847656,
"DMAProfiler": 0.0012180805206298828,
"DataStreaming": 0.0005357265472412109,
"DoNothing": 0.00019812583923339844,
"ExpandISAMacro": 0.0012791156768798828,
"FactorizeBlkDims": 0.000789642333984375,
"InferPSumTensor": 0.001064300537109375,
"InferSharedMemLoc": 0.00042891502380371094,
"InsertCoreBarrier": 0.00044727325439453125,
"LateLegalizeInst": 0.0007555484771728516,
"LateNeuronInstComb": 0.0008594989776611328,
"LegalizeSundaAccess": 0.0021588802337646484,
"LegalizeType": 0.0003390312194824219,
"LowerBroadcast": 0.0003108978271484375,
"LowerIntrinsics": 0.0003314018249511719,
"LowerTranspose": 0.0003635883331298828,
"NeuronInstComb": 0.0009188652038574219,
"NeuronLICM": 0.0006546974182128906,
"NeuronSimplifyPredicates": 0.004039287567138672,
"NeuronValueNumbering": 0.0005695819854736328,
"SFKVectorizer": 0.006131172180175781,
"SimpleAllReduceTiling": 0.0004286766052246094,
"SimplifyNeuronTensor": 0.0011782646179199219,
"SpillPSum": 0.0007271766662597656,
"WeightCoalescing": 0.00041985511779785156
}
},
"sg00": {
"hilo": {
"ArithmeticIntensity": 3.148798704147339,
"HloMacCount": 1972166656.0,
"Traffic": 1252647040.0
}
},
"sg0000": {
"compiletime": {
"AGOrderingAnalysisPass": 5.95287561416626,
"AffinePredicateResolution": 0.49277400970458984,
"AliasDependencyElimination": 0.01732039451599121,
"AliasDependencyInduction": 18.183685302734375,
"AliasDependencyReset": 18.999177932739258,
"BFComputeCutting": 0.3126344680786133,
"BirCodeGenLoop": 1.8802330493927002,
"CCOpFusion": 0.8809163570404053,
"CanonicalizeDAGForPGTiling": 0.17032384872436523,
"CanonicalizeIR": 0.7969982624053955,
"CoalesceCCOp": 0.6308276653289795,
"CommuteConcat": 0.02859354019165039,
"DMALocalityOpt": 0.036058664321899414,
"DMAProfiler": 0.1385490894317627,
"DMATilingProfiler": 0.08906126022338867,
"DataLocalityOpt": 4.2657470703125,
"DataStreaming": 0.1718294620513916,
"DeConcat": 0.06563258171081543,
"DeadCodeElimination": 0.0297544002532959,
"DeadStoreElimination": 1.2178454399108887,
"DelinearIndices": 0.35585713386535645,
"Delinearization": 0.6069140434265137,
"DelinearizeSPMD": 0.6448159217834473,
"DoNothing": 0.00018167495727539063,
"DramToDramTranspose": 0.3358180522918701,
"DumpGraphAndMetadata": 0.17146539688110352,
"EliminateDivs": 1.3107364177703857,
"ExpandBatchNorm": 0.48049426078796387,
"ExpandISAMacro": 0.07998967170715332,
"FactorizeBlkDims": 0.5875980854034424,
"FactorizeThreadAxesInFreeDims": 0.09288716316223145,
"FlattenMacroLoop": 0.0829305648803711,
"GenericAccessSimplifier": 0.026407241821289063,
"InferInitValue": 1.2918815612792969,
"InferIntrinsicOnCC": 0.31558942794799805,
"InferNeuronTensor": 1.773860216140747,
"InferNonlocalTensors": 6.451204299926758,
"InferPSumTensor": 1.4296114444732666,
"InferShardAxis": 10.398513793945313,
"InferSharedMemLoc": 0.12653231620788574,
"InlineNativeKernels": 0.051079511642456055,
"InsertCoreBarrier": 0.36295533180236816,
"InsertIOTransposes": 1.088529348373413,
"InsertImplicitShardAxisBeforeISel": 0.4208717346191406,
"InsertLocalTransposes": 0.6978082656860352,
"InsertOffloadedTransposes": 0.13763642311096191,
"LICM": 0.11658525466918945,
"LateLegalizeInst": 0.5564777851104736,
"LateLegalizePostSplit": 0.10899925231933594,
"LateLowerReshapeOp": 0.03313422203063965,
"LateLowerTensorOp": 6.405031204223633,
"LateNeuronInstComb": 1.0862395763397217,
"LayoutPreprocessing": 1.1262352466583252,
"LayoutPreprocessingAndAnalysis": 1.4945013523101807,
"LayoutRequirementAnalysis": 0.36240172386169434,
"LegalizeCCOpLayout": 1.2990221977233887,
"LegalizeOpLevelAlias": 1.2160041332244873,
"LegalizePartitionReduce": 0.0940859317779541,
"LegalizeSundaAccess": 3.4090025424957275,
"LegalizeSundaMacro": 0.6391239166259766,
"LegalizeType": 0.12801194190979004,
"LocalLayoutOpt": 0.6706066131591797,
"LoopFusion": 0.28072071075439453,
"LoopSplitting": 0.07517576217651367,
"LowerBroadcast": 0.06457281112670898,
"LowerCCOpBlockAxis": 0.20669817924499512,
"LowerComplexBroadcast": 0.1663343906402588,
"LowerIntrinsics": 0.9722909927368164,
"LowerShardAxis": 0.24805045127868652,
"LowerTensorOp": 3.8965094089508057,
"LowerToSendRecv": 0.18735170364379883,
"LowerTranspose": 0.47939133644104004,
"MacroGeneration": 2.835282802581787,
"MaskPropagation": 0.1340806484222412,
"MemcpyElimination": 28.533254623413086,
"MutateDataType": 0.05120110511779785,
"NeuronAliasDependencyInduction": 0.05350494384765625,
"NeuronAliasDependencyReset": 0.07738471031188965,
"NeuronInstComb": 0.4073524475097656,
"NeuronLICM": 0.2638108730316162,
"NeuronLoopFusion": 1.1017746925354004,
"NeuronLoopInterchange": 0.05889534950256348,
"NeuronSimplifier": 0.5097112655639648,
"NeuronSimplifyPredicates": 0.21783137321472168,
"NeuronValueNumbering": 0.12063908576965332,
"OptimizeAliasedCopyChain": 0.31018686294555664,
"OptimizeNKIKernels": 1.1799430847167969,
"PAGLayoutOpt": 17.110271453857422,
"PComputeCutting": 1.0609712600708008,
"PGLayoutTilingPipeline": 50.19084930419922,
"PGTiling": 11.0619478225708,
"PadElimination": 0.012235879898071289,
"ParAxesAnnotation": 16.403249740600586,
"PartialLoopFusion": 1.436572551727295,
"PartialSimdFusion": 0.8934595584869385,
"PerfectLoopNest": 0.07143855094909668,
"RecognizeOpIdiom": 0.13362383842468262,
"Recompute": 0.0345001220703125,
"RelaxPredicates": 0.11409568786621094,
"Rematerialization": 0.22487568855285645,
"RemoveShardedPartitionAxes": 0.8885042667388916,
"ReshapeWeights": 0.02379584312438965,
"ResolveAccessConflict": 0.46589231491088867,
"ResolveComplicatePredicates": 0.5794906616210938,
"RewriteReplicationMatmul": 0.040407657623291016,
"RewriteWeights": 0.08104610443115234,
"SFKVectorizer": 11.707208633422852,
"ShardingPropagationAnalysis": 0.7053203582763672,
"SimpleAllReduceTiling": 0.2163074016571045,
"Simplifier": 0.12051248550415039,
"SimplifyMacroPredicates": 0.27616286277770996,
"SimplifyNeuronTensor": 0.4094421863555908,
"SimplifySlice": 0.026907682418823242,
"SimplifyTensor": 0.25261998176574707,
"SpillPSum": 0.6899855136871338,
"SplitAPUnionSets": 0.49262213706970215,
"SplitAccGrp": 0.04794502258300781,
"StaticProfiler": 0.13262128829956055,
"StaticTransposeLocalTensor": 0.6367781162261963,
"SundaISel": 1.6440446376800537,
"TCTransform": 0.030546188354492188,
"TensorInitialization": 0.24484515190124512,
"TensorOpSimplifier": 3.413067102432251,
"TensorOpTransform": 18.72884178161621,
"TileCCOps": 0.18365025520324707,
"TilingProfiler": 0.8507044315338135,
"TransformConvOp": 1.1984179019927979,
"TritiumFusion": 1.2713313102722168,
"ValueNumbering": 0.08294677734375,
"VectorizeDMA": 2.601423501968384,
"VectorizeMatMult": 0.10021328926086426,
"WeightCoalescing": 0.06349992752075195,
"ZeroSizeTensorElimination": 0.00041866302490234375
},
"tensorizer": {
"DMATilingProfiler::TotalInstructionsAfterTiling": 63618,
"StaticProfiler::AifUb": 27.032575607299805,
"StaticProfiler::ArithmeticIntensityTensorizer": 35.5345458984375,
"StaticProfiler::AverageDmaLength": 2744.118896484375,
"StaticProfiler::AverageFractalPeUtilization": 98.06681823730469,
"StaticProfiler::AveragePartitionUtilization": 88.63555145263672,
"StaticProfiler::AveragePeUtilization": 75.70149230957031,
"StaticProfiler::DDRTransferBytes": 1130475120,
"StaticProfiler::InternalTransferBytes": 288112224,
"StaticProfiler::LoadExpanded": 338844,
"StaticProfiler::LocalizationEfficiency": 131.45083618164063,
"StaticProfiler::LocalizationEfficiencyIgnoreNonlocal": 134.9540252685547,
"StaticProfiler::StoreExpanded": 4273,
"StaticProfiler::TotalDMAExpanded": 343117,
"StaticProfiler::TotalDynamicInstancesCount": 81003,
"StaticProfiler::TotalDynamicInstancesWithMmPackedCount": 66628,
"StaticProfiler::TotalLNCComm": 0,
"StaticProfiler::TotalLNCCommTransfer": 0,
"TilingProfiler::AveragePartitionUtilizationAfterTiling": 0,
"TilingProfiler::AveragePeUtilizationAfterTiling": 0,
"TilingProfiler::BatchnormInstructionsAfterTiling": 0,
"TilingProfiler::DmaInstructionsAfterTiling": 0,
"TilingProfiler::GenericInstructionsAfterTiling": 123,
"TilingProfiler::MatMultInstructionsAfterTiling": 40384,
"TilingProfiler::NumPfTransposes": 292,
"TilingProfiler::NumPfTransposesForIo": 30,
"TilingProfiler::NumPfTransposesForLocal": 142,
"TilingProfiler::NumPfTransposesForNonlocal": 120,
"TilingProfiler::PfTransposeInstructions": 14486,
"TilingProfiler::PfTransposeInstructionsForIo": 11952,
"TilingProfiler::PfTransposeInstructionsForLocal": 872,
"TilingProfiler::PfTransposeInstructionsForNonlocal": 1662,
"TilingProfiler::ReduceInstructionsAfterTiling": 563,
"TilingProfiler::SimdInstructionsAfterTiling": 3485,
"TilingProfiler::TotalInstructionsAfterTiling": 0,
"TransformConvOp::Conv1d_depthwise_bf01_oi01_bf01": 0,
"TransformConvOp::Conv2d_dw_fb01_io01_01bf_rep_nhwc_Pcinh": 0,
"TransformConvOp::Conv2d_pbp_0f1b_0i1o_01fb_experimental_1": 0,
"TransformConvOp::Conv2d_pbp_fb01_io01_01bf_experimental_1": 0,
"TransformConvOp::conv2d_column_packing": 0,
"TransformConvOp::conv2d_column_packing_1": 0,
"TransformConvOp::conv2d_column_packing_io10": 0,
"TransformConvOp::conv2d_depthwise_f01b_o01i_bf01": 0
}
},
"topk": {
"compiletime": {
"CoalesceCCOp": 0.0037636756896972656,
"DMALocalityOpt": 0.002650737762451172,
"DMAProfiler": 0.003582477569580078,
"DataStreaming": 0.006698131561279297,
"DoNothing": 0.00016760826110839844,
"ExpandISAMacro": 0.0037276744842529297,
"FactorizeBlkDims": 0.012275934219360352,
"InferPSumTensor": 0.01163172721862793,
"InferSharedMemLoc": 0.002855539321899414,
"InsertCoreBarrier": 0.0032694339752197266,
"LateLegalizeInst": 0.007744550704956055,
"LateNeuronInstComb": 0.008387565612792969,
"LegalizeSundaAccess": 0.014924049377441406,
"LegalizeType": 0.009481191635131836,
"LowerBroadcast": 0.003381013870239258,
"LowerIntrinsics": 0.0036077499389648438,
"LowerTranspose": 0.003323078155517578,
"NeuronInstComb": 0.008477210998535156,
"NeuronLICM": 0.009244441986083984,
"NeuronSimplifyPredicates": 0.00380706787109375,
"NeuronValueNumbering": 0.004020214080810547,
"SFKVectorizer": 0.03467822074890137,
"SimpleAllReduceTiling": 0.003582477569580078,
"SimplifyNeuronTensor": 0.050054073333740234,
"SpillPSum": 0.02035689353942871,
"WeightCoalescing": 0.0034646987915039063
}
}
}

3
token_generation_model/_tp0_bk4/graph.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:2840765467a4771e8bfbd46ac14d381ab5b0ca2a5728c2c4e45bcaecd45c75d2
size 4097024

4484
token_generation_model/_tp0_bk4/log-neuron-cc.txt Normal file
View File

File diff suppressed because it is too large Load Diff

3
token_generation_model/_tp0_bk4/metaneff.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:d29e9e51be8d730fedcaafdffddee32ee55e95335b0070a546ddb2f8bc39ed32
size 2464555

3
token_generation_model/_tp0_bk4/model.MODULE_b4b44a47076a167bd9a5+795ee7cc.hlo_module.pb Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:0e834b15d13ccd9cce8994a607f90a7511bc74cf2044170c619756122300d806
size 2550451

3
token_generation_model/_tp0_bk4/model.MODULE_b4b44a47076a167bd9a5+795ee7cc.neff Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:2840765467a4771e8bfbd46ac14d381ab5b0ca2a5728c2c4e45bcaecd45c75d2
size 4097024

224
token_generation_model/_tp0_bk4/neuron_config.json Normal file
View File

@@ -0,0 +1,224 @@
{
"_attn_implementation_autoset": false,
"_name_or_path": "/models/Qwen3-1.7B/",
"add_cross_attention": false,
"architectures": [
"Qwen3ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"attribute_map": {},
"bad_words_ids": null,
"begin_suppress_tokens": null,
"bos_token_id": 151643,
"chunk_size_feed_forward": 0,
"cross_attention_hidden_size": null,
"decoder_start_token_id": null,
"diversity_penalty": 0.0,
"do_sample": false,
"early_stopping": false,
"encoder_no_repeat_ngram_size": 0,
"eos_token_id": 151645,
"exponential_decay_length_penalty": null,
"finetuning_task": null,
"forced_bos_token_id": null,
"forced_eos_token_id": null,
"fused_spec_config": null,
"head_dim": 128,
"hidden_act": "silu",
"hidden_size": 2048,
"id2label": {
"0": "LABEL_0",
"1": "LABEL_1"
},
"initializer_range": 0.02,
"intermediate_size": 6144,
"is_decoder": false,
"is_encoder_decoder": false,
"label2id": {
"LABEL_0": 0,
"LABEL_1": 1
},
"length_penalty": 1.0,
"max_length": 20,
"max_position_embeddings": 40960,
"max_window_layers": 28,
"metadata": null,
"min_length": 0,
"model_type": "qwen3",
"neuron_config": {
"activation_quantization_type": null,
"allow_input_truncation": false,
"apply_seq_ids_mask": false,
"async_mode": false,
"attention_dp_degree": 1,
"attention_dtype": null,
"attn_block_cte_nki_kernel_enabled": false,
"attn_block_tkg_nki_kernel_cache_update": false,
"attn_block_tkg_nki_kernel_cascaded_attention": false,
"attn_block_tkg_nki_kernel_enabled": false,
"attn_cls": {
"__module__": "neuronx_distributed_inference.models.qwen3.modeling_qwen3",
"__name__": "NeuronQwen3Attention"
},
"attn_kernel_enabled": null,
"attn_tkg_builtin_kernel_enabled": false,
"attn_tkg_nki_kernel_enabled": false,
"batch_size": 4,
"bucket_n_active_tokens": false,
"buckets": [
2048
],
"cast_type": "config",
"cc_pipeline_tiling_factor": 1,
"chunked_prefill_config": null,
"context_encoding_buckets": null,
"cp_degree": 1,
"ctx_batch_size": 1,
"disable_kv_cache_tiling": false,
"draft_model_modules_to_not_convert": null,
"enable_bucketing": true,
"enable_cte_modular_flow": false,
"enable_eagle_draft_input_norm": false,
"enable_eagle_speculation": false,
"enable_fused_speculation": false,
"enable_long_context_mode": false,
"enable_output_completion_notifications": false,
"enable_spill_reload_dge": false,
"enable_token_tree": false,
"ep_degree": 1,
"expert_mlp_nki_kernel_enabled": null,
"flash_decoding_enabled": false,
"fused_qkv": false,
"fused_rmsnorm_skip_gamma": false,
"is_block_kv_layout": null,
"is_chunked_prefill": false,
"is_continuous_batching": true,
"is_eagle_draft": false,
"is_medusa": false,
"is_prefill_stage": false,
"is_prefix_caching": false,
"k_cache_transposed": false,
"kv_cache_batch_size": 4,
"kv_cache_padding_size": 0,
"kv_cache_quant": false,
"kv_cache_tiling": false,
"layer_boundary_markers": false,
"lm_head_pad": true,
"lm_head_pad_alignment_size": 1,
"local_ranks_size": 4,
"logical_nc_config": 2,
"lora_config": null,
"max_batch_size": 4,
"max_context_length": 2048,
"max_length": 2048,
"max_new_tokens": null,
"medusa_speculation_length": 0,
"medusa_tree": null,
"mlp_kernel_enabled": false,
"mlp_kernel_fuse_residual_add": false,
"modules_to_not_convert": null,
"moe_fused_nki_kernel_enabled": null,
"n_active_tokens": 1,
"n_positions": 2048,
"num_medusa_heads": 0,
"on_cpu": false,
"on_device_sampling_config": {
"deterministic": false,
"do_sample": false,
"dynamic": true,
"global_topk": 256,
"on_device_sampling_config": true,
"temperature": 1.0,
"top_k": 1,
"top_k_kernel_enabled": false,
"top_p": 1.0
},
"output_logits": false,
"overrides_torch_dtype": true,
"pa_block_size": 2048,
"pa_num_blocks": 4,
"padding_side": "right",
"pp_degree": 1,
"prefix_buckets": null,
"qk_layernorm": false,
"qkv_kernel_enabled": false,
"qkv_kernel_fuse_residual_add": false,
"qkv_kernel_nbsd_layout": false,
"quantization_dtype": "int8",
"quantization_type": "per_tensor_symmetric",
"quantize_clamp_bound": Infinity,
"quantized": false,
"quantized_checkpoints_path": null,
"quantized_mlp_kernel_enabled": false,
"rmsnorm_quantize_kernel_enabled": false,
"router_topk_nki_kernel_enabled": null,
"rpl_reduce_dtype": null,
"save_sharded_checkpoint": true,
"scratchpad_page_size": null,
"seq_len": 2048,
"seq_len_threshold_for_cc_tiling": 16384,
"sequence_parallel_enabled": false,
"shared_mlp_nki_kernel_enabled": null,
"skip_sharding": false,
"skip_warmup": false,
"spec_batch_size": 4,
"speculation_length": 0,
"start_rank_id": 0,
"strided_context_parallel_kernel_enabled": false,
"target": null,
"tensor_capture_config": null,
"tile_cc": false,
"tkg_batch_size": 4,
"token_generation_buckets": [
2048
],
"token_tree_config": null,
"torch_dtype": "bfloat16",
"tp_degree": 4,
"vocab_parallel": false,
"weight_gather_seq_len_threshold": 32768,
"weights_to_skip_layout_optimization": [],
"world_size": 4
},
"no_repeat_ngram_size": 0,
"num_attention_heads": 16,
"num_beam_groups": 1,
"num_beams": 1,
"num_cores_per_group": 1,
"num_hidden_layers": 28,
"num_key_value_heads": 8,
"num_return_sequences": 1,
"output_attentions": false,
"output_hidden_states": false,
"output_scores": false,
"pad_token_id": 0,
"prefix": null,
"problem_type": null,
"pruned_heads": {},
"remove_invalid_values": false,
"repetition_penalty": 1.0,
"return_dict": true,
"return_dict_in_generate": false,
"rms_norm_eps": 1e-06,
"rope_scaling": null,
"rope_theta": 1000000,
"sep_token_id": null,
"sliding_window": null,
"suppress_tokens": null,
"task_specific_params": null,
"temperature": 1.0,
"tf_legacy_loss": false,
"tie_encoder_decoder": false,
"tie_word_embeddings": true,
"tokenizer_class": null,
"top_k": 50,
"top_p": 1.0,
"torchscript": false,
"transformers_version": "4.51.0",
"typical_p": 1.0,
"use_bfloat16": false,
"use_cache": true,
"use_sliding_window": false,
"vocab_size": 151936
}