PyPI - mindspore - Versions diffs - 2.6.0rc1__cp311-none-any.whl → 2.7.0__cp311-none-any.whl - Mend - Supply Chain Defender

mindspore 2.6.0rc1cp311-none-any.whl → 2.7.0cp311-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.

This version of mindspore might be problematic. Click here for more details.

Files changed (4997) hide show

mindspore/include/mindspore/ccsrc/pyboost/functions/auto_generate/auto_grad_op_reg.h CHANGED Viewed

@@ -26,1447 +26,1651 @@ namespace mindspore {
 namespace kernel {
 namespace pyboost {
 enum class OpType {
-  kFmodTensor = 0,
-  kIndex = 1,
-  kScatterValue = 2,
-  kEmbeddingDenseBackward = 3,
-  kSoftMarginLoss = 4,
-  kLogicalXor = 5,
-  kInplaceSubExt = 6,
-  kAllFinite = 7,
-  kSinc = 8,
-  kInplaceDivMods = 9,
-  kCummax = 10,
-  kAtanExt = 11,
-  kExpandAs = 12,
-  kMaxUnpool2DExt = 13,
-  kGreaterEqual = 14,
-  kMax = 15,
-  kNotEqual = 16,
-  kErf = 17,
-  kBatchNormElemtGrad = 18,
-  kInplaceNormal = 19,
-  kNLLLoss2d = 20,
-  kUpsampleNearest3D = 21,
-  kLogSoftmax = 22,
-  kOuter = 23,
-  kInplaceExp = 24,
-  kSiLUGrad = 25,
-  kUpsampleNearest3DGrad = 26,
+  kUpsampleNearest1D = 0,
+  kSquare = 1,
+  kNormalTensorTensor = 2,
+  kSilentCheckV3 = 3,
+  kConvolutionStrGrad = 4,
+  kDistCommAllGather = 5,
+  kMaximum = 6,
+  kFloorDivScalar = 7,
+  kLerpScalar = 8,
+  kDistCommAllReduce = 9,
+  kFmodTensor = 10,
+  kInplaceElu = 11,
+  kInplaceScatterAdd = 12,
+  kErfc = 13,
+  kBitwiseAndScalar = 14,
+  kBatchNormGatherStatsWithCounts = 15,
+  kNewFull = 16,
+  kBitwiseOrScalar = 17,
+  kNarrow = 18,
+  kInplaceRemainderTensorTensor = 19,
+  kDistCommAllGatherIntoTensorUneven = 20,
+  kEmpty = 21,
+  kReplicationPad1DGrad = 22,
+  kNewZeros = 23,
+  kSigmoid = 24,
+  kRepeat = 25,
+  kCustomExt = 26,
   kUpsampleBicubic2DGrad = 27,
-  kConcat = 28,
-  kAdaptiveAvgPool1D = 29,
-  kUpsampleBilinear2D = 30,
-  kLog2 = 31,
-  kGmmV2Backward = 32,
-  kMSELossExt = 33,
-  kLerp = 34,
-  kDiagExt = 35,
-  kFrac = 36,
-  kGluGrad = 37,
-  kGenerator = 38,
-  kFlattenExt = 39,
-  kSoftmax = 40,
-  kDivMod = 41,
-  kSoftMarginLossGrad = 42,
-  kMaskedFill = 43,
-  kGridSampler3DGrad = 44,
-  kArgMaxExt = 45,
-  kInplacePut = 46,
-  kTopkExt = 47,
-  kConv1DExt = 48,
-  kReplicationPad3D = 49,
-  kReplicationPad3DGrad = 50,
-  kTExt = 51,
-  kCountNonZero = 52,
-  kInplaceMaskedFillTensor = 53,
-  kXlogy = 54,
-  kSoftShrinkGrad = 55,
-  kReduceMin = 56,
-  kMatrixInverseExt = 57,
-  kInplaceElu = 58,
-  kUpsampleTrilinear3D = 59,
-  kInplaceHardtanh = 60,
-  kAddLayerNormV2 = 61,
-  kBitwiseAndTensor = 62,
-  kInplaceClampScalar = 63,
-  kLayerNormGradExt = 64,
-  kInplaceScatterAdd = 65,
-  kSmoothL1Loss = 66,
-  kReflectionPad3DGrad = 67,
-  kClampTensor = 68,
-  kMoeTokenUnpermuteGrad = 69,
-  kLogSoftmaxExt = 70,
-  kCeil = 71,
-  kInplaceMul = 72,
-  kGreater = 73,
-  kScatter = 74,
-  kReciprocal = 75,
-  kMuls = 76,
-  kInplaceUniform = 77,
-  kLayerNormExt = 78,
-  kBaddbmm = 79,
-  kConv2DPadding = 80,
-  kMultinomialExt = 81,
-  kConv2DExt = 82,
-  kDense = 83,
-  kConvolutionStrGrad = 84,
-  kOnes = 85,
-  kRepeatInterleaveGrad = 86,
-  kPow = 87,
-  kMin = 88,
-  kSmoothL1LossGrad = 89,
-  kFlashAttentionScore = 90,
-  kInplaceCopy = 91,
-  kInplaceScatterValue = 92,
-  kUnstackExt = 93,
-  kAvgPool3DGradExt = 94,
-  kMaxPoolWithMask = 95,
-  kHSigmoidGrad = 96,
-  kSilentCheckV2 = 97,
-  kErfc = 98,
-  kBinaryCrossEntropy = 99,
-  kReduceAll = 100,
-  kLogSumExp = 101,
-  kInplaceDivMod = 102,
-  kGeluGradExt = 103,
-  kIndexFillScalar = 104,
-  kSub = 105,
-  kMaximum = 106,
-  kNLLLoss = 107,
-  kXLogYScalarOther = 108,
-  kIdentity = 109,
-  kBitwiseXorScalar = 110,
-  kLogSigmoidGrad = 111,
-  kScatterAddExt = 112,
-  kAdaptiveAvgPool2DGradExt = 113,
-  kNormalFloatTensor = 114,
-  kBitwiseXorTensor = 115,
-  kGLU = 116,
-  kUpsampleNearest1D = 117,
-  kDivMods = 118,
-  kFFNExt = 119,
-  kAdd = 120,
-  kCopy = 121,
-  kAsinExt = 122,
-  kReplicationPad2DGrad = 123,
-  kNeScalar = 124,
-  kGmmBackward = 125,
-  kGroupNorm = 126,
-  kVarMean = 127,
-  kConvolutionGrad = 128,
-  kMaxPoolGradWithIndices = 129,
-  kInplaceReLU = 130,
-  kLinSpaceExt = 131,
-  kNewOnes = 132,
-  kArgMinWithValue = 133,
-  kStd = 134,
-  kSin = 135,
-  kSortExt = 136,
-  kDropoutDoMaskExt = 137,
-  kSoftmaxBackward = 138,
-  kTypeAs = 139,
-  kGcd = 140,
-  kAvgPool1D = 141,
-  kSlice = 142,
-  kFloorDivScalar = 143,
-  kMoeTokenPermute = 144,
-  kMatMulExt = 145,
-  kInnerNonZero = 146,
-  kReflectionPad3D = 147,
-  kRepeatInterleaveTensor = 148,
-  kMedianDim = 149,
-  kUpsampleNearest2DGrad = 150,
-  kAdaptiveMaxPool2D = 151,
-  kInplaceZero = 152,
-  kErfinv = 153,
-  kInplaceLog = 154,
-  kHShrinkGrad = 155,
-  kKthvalue = 156,
-  kNormalTensorTensor = 157,
-  kReplicationPad1D = 158,
-  kBatchMatMul = 159,
-  kAddmm = 160,
-  kSelectV2 = 161,
-  kSoftplusGradExt = 162,
-  kBatchNormExt = 163,
-  kGroupNormGrad = 164,
-  kRandIntLike = 165,
-  kMinDim = 166,
-  kContiguous = 167,
-  kArgMaxWithValue = 168,
-  kMaskedSelectGrad = 169,
-  kSpeedFusionAttentionGrad = 170,
-  kLogicalOr = 171,
-  kRemainderTensorTensor = 172,
-  kAdaptiveAvgPool3DExt = 173,
-  kRandpermExt = 174,
-  kGeLUGrad = 175,
-  kLogicalAnd = 176,
-  kCast = 177,
-  kMishExt = 178,
-  kBCEWithLogitsLoss = 179,
-  kPolar = 180,
-  kReluGrad = 181,
-  kNorm = 182,
-  kRemainderScalarTensor = 183,
-  kAvgPool2D = 184,
-  kXLogYScalarSelf = 185,
-  kDropoutGradExt = 186,
-  kConvTranspose2D = 187,
-  kHShrink = 188,
-  kL1LossBackwardExt = 189,
-  kBitwiseOrScalar = 190,
-  kExp = 191,
-  kBatchNormGatherStatsWithCounts = 192,
-  kZeros = 193,
-  kEqualExt = 194,
-  kClone = 195,
-  kMaxPoolWithIndices = 196,
-  kAdaptiveAvgPool2DExt = 197,
-  kConv3DPadding = 198,
-  kChunk = 199,
-  kAbs = 200,
-  kCumsumExt = 201,
-  kTraceExt = 202,
-  kSign = 203,
-  kInplaceClampTensor = 204,
-  kAtan2Ext = 205,
-  kInplaceFillTensor = 206,
-  kArgMinExt = 207,
-  kOnesLikeExt = 208,
-  kAcoshExt = 209,
-  kMedianExt = 210,
-  kInplaceFloorDivide = 211,
-  kReplicationPad2D = 212,
-  kReduceAny = 213,
-  kBincountExt = 214,
-  kRepeat = 215,
-  kRemainderTensorScalar = 216,
-  kTanh = 217,
-  kTrilExt = 218,
-  kKLDiv = 219,
-  kMeshgrid = 220,
-  kAddmv = 221,
-  kNansum = 222,
-  kHSigmoid = 223,
-  kThreshold = 224,
-  kSqrt = 225,
-  kRotaryPositionEmbedding = 226,
-  kUpsampleTrilinear3DGrad = 227,
-  kSliceExt = 228,
-  kSplitWithSize = 229,
-  kInplaceAddsExt = 230,
-  kExpm1 = 231,
-  kInplaceDivs = 232,
-  kNonZero = 233,
-  kInplaceAddExt = 234,
-  kRound = 235,
-  kAdaptiveAvgPool3DGradExt = 236,
-  kInplaceMuls = 237,
-  kReLU = 238,
-  kLerpScalar = 239,
-  kCosh = 240,
-  kIsInf = 241,
-  kMv = 242,
-  kInnerIndex = 243,
-  kLog10 = 244,
-  kDivs = 245,
-  kIndexFillTensor = 246,
-  kGatherD = 247,
-  kAvgPool2DGrad = 248,
-  kIm2ColExt = 249,
-  kNanToNum = 250,
-  kLinalgVectorNorm = 251,
-  kSigmoidGrad = 252,
-  kProdExt = 253,
-  kAddScalar = 254,
-  kNLLLoss2dGrad = 255,
-  kLog1p = 256,
-  kConv3DExt = 257,
-  kSilentCheckV3 = 258,
-  kMultiScaleDeformableAttnGrad = 259,
-  kAddExt = 260,
-  kInplaceFloor = 261,
-  kInplaceTanh = 262,
-  kAsinhExt = 263,
-  kConvolution = 264,
-  kUpsampleBicubic2D = 265,
-  kEmbedding = 266,
-  kBinaryCrossEntropyGrad = 267,
-  kAddbmm = 268,
-  kMoeTokenUnpermute = 269,
-  kHardtanhGrad = 270,
-  kAcosExt = 271,
-  kSwigluGrad = 272,
-  kMultiScaleDeformableAttn = 273,
-  kMinimum = 274,
-  kHistcExt = 275,
-  kMm = 276,
-  kReflectionPad1DGrad = 277,
-  kGridSampler2D = 278,
-  kBitwiseNot = 279,
-  kUpsampleLinear1D = 280,
-  kRotaryPositionEmbeddingGrad = 281,
-  kUnique2 = 282,
-  kSpeedFusionAttention = 283,
-  kClampScalar = 284,
-  kCol2ImExt = 285,
-  kGridSampler3D = 286,
-  kLogAddExp = 287,
-  kInplaceFloorDivides = 288,
-  kBatchNormReduceGrad = 289,
-  kView = 290,
-  kBatchMatMulExt = 291,
-  kInplaceErfinv = 292,
-  kMaxPoolGradWithMask = 293,
-  kTile = 294,
-  kSoftShrink = 295,
-  kReshape = 296,
-  kBatchNormElemt = 297,
-  kLinalgQr = 298,
-  kSigmoid = 299,
-  kL1LossExt = 300,
-  kViewAs = 301,
-  kTranspose = 302,
-  kUpsampleNearest2D = 303,
-  kCustomExt = 304,
-  kDropoutGenMaskExt = 305,
-  kInplaceIndexAddExt = 306,
-  kElu = 307,
-  kSinh = 308,
-  kSelect = 309,
-  kInplaceMaskedFillScalar = 310,
-  kGeluExt = 311,
-  kBatchNormGradExt = 312,
-  kNormalTensorFloat = 313,
-  kRmsNormGrad = 314,
-  kSeluGrad = 315,
-  kNonZeroExt = 316,
-  kSquare = 317,
-  kReduceMax = 318,
-  kBitwiseAndScalar = 319,
-  kInplaceScatterValueReduce = 320,
-  kThresholdGrad = 321,
-  kUniqueDim = 322,
-  kHSwishGrad = 323,
-  kLeakyReLUGradExt = 324,
-  kCumminExt = 325,
-  kIsClose = 326,
-  kTake = 327,
-  kReflectionPad1D = 328,
-  kAddcdivExt = 329,
-  kVar = 330,
-  kMul = 331,
-  kInplaceThreshold = 332,
-  kRmsNorm = 333,
-  kBinaryCrossEntropyWithLogitsBackward = 334,
-  kStdMean = 335,
-  kMoeTokenPermuteGrad = 336,
-  kRandInt = 337,
-  kSoftplusExt = 338,
-  kReflectionPad2D = 339,
-  kPowScalarTensor = 340,
-  kInplaceScatterSrcReduce = 341,
-  kEqual = 342,
-  kRandn = 343,
-  kRoll = 344,
-  kGeLU = 345,
-  kAsStrided = 346,
-  kTensorScatterElements = 347,
-  kFlashAttentionScoreGrad = 348,
-  kAtanh = 349,
-  kInplaceStopGradient = 350,
-  kLogicalNot = 351,
-  kDropoutExt = 352,
-  kNormalFloatFloat = 353,
-  kConv1DPadding = 354,
-  kRsqrt = 355,
-  kConstantPadND = 356,
-  kBitwiseOrTensor = 357,
-  kMSELossGradExt = 358,
-  kMeanExt = 359,
-  kLessEqual = 360,
-  kPromptFlashAttention = 361,
-  kSelectExt = 362,
-  kEluExt = 363,
-  kCol2ImGrad = 364,
-  kTriangularSolve = 365,
-  kFillTensor = 366,
-  kSubScalar = 367,
-  kIndexAddExt = 368,
-  kFillScalar = 369,
-  kLeakyReLUExt = 370,
-  kRandExt = 371,
-  kDiv = 372,
-  kTransposeExt = 373,
-  kInplaceFillScalar = 374,
-  kGatherDGradV2 = 375,
-  kPReLU = 376,
-  kInplaceDiv = 377,
-  kInplaceIndexPut = 378,
-  kKLDivGrad = 379,
-  kUpsampleNearest1DGrad = 380,
-  kUniformExt = 381,
-  kIsFinite = 382,
-  kLog = 383,
-  kBroadcastTo = 384,
-  kSplitTensor = 385,
-  kExpandDims = 386,
-  kFloor = 387,
-  kLogSoftmaxGrad = 388,
-  kLess = 389,
-  kReplicationPad1DGrad = 390,
-  kAdamW = 391,
-  kInplaceRandom = 392,
-  kSplit = 393,
-  kAvgPool3DExt = 394,
-  kNarrow = 395,
-  kSqueeze = 396,
-  kLogAddExp2 = 397,
-  kNeg = 398,
-  kInplaceFillDiagonal = 399,
-  kInplaceGroupedMatmulAdd = 400,
-  kBernoulliExt = 401,
-  kReflectionPad2DGrad = 402,
-  kFullLike = 403,
-  kEye = 404,
-  kConvolutionStr = 405,
-  kGridSampler2DGrad = 406,
-  kHSwish = 407,
-  kUpsampleLinear1DGrad = 408,
-  kUpsampleBilinear2DGrad = 409,
-  kAllGatherMatmul = 410,
-  kMaxDim = 411,
-  kIncreFlashAttention = 412,
-  kTriu = 413,
-  kFloorDiv = 414,
-  kAddRmsNorm = 415,
-  kTanhGrad = 416,
-  kExp2 = 417,
-  kHardtanh = 418,
-  kIndexSelect = 419,
-  kLogSigmoid = 420,
-  kStackExt = 421,
-  kMaskedSelect = 422,
-  kMishGradExt = 423,
-  kReverseV2 = 424,
-  kOneHotExt = 425,
-  kPReLUGrad = 426,
-  kDot = 427,
-  kIsNegInf = 428,
-  kTrunc = 429,
-  kMatMul = 430,
-  kNLLLossGrad = 431,
-  kSwiglu = 432,
-  kRepeatInterleaveInt = 433,
-  kGreaterEqualScalar = 434,
-  kRandnLike = 435,
-  kZerosLikeExt = 436,
-  kNewZeros = 437,
-  kArgSort = 438,
-  kSearchSorted = 439,
-  kSiLU = 440,
-  kInplaceSubScalar = 441,
-  kInnerInplaceIndexPut = 442,
-  kSumExt = 443,
-  kCos = 444,
-  kSubExt = 445,
-  kArange = 446,
-  kInplaceScatterSrc = 447,
-  kAddcmulExt = 448,
-  kCross = 449,
-  kAdaptiveMaxPool1D = 450,
-  kMatmulReduceScatter = 451,
-  kTan = 452,
-  kRandLikeExt = 453,
-  kBatchNormStats = 454,
-  kPowTensorScalar = 455,
-  kInplaceAddmm = 456,
-  kSeLUExt = 457,
-  kUniqueConsecutive = 458,
-  kFmodScalar = 459,
-  kAddLayerNormGrad = 460,
-  kEluGradExt = 461,
-  kGroupedMatmulV2 = 462,
-  kKVCacheScatterUpdate = 463,
-  kFusedInferAttentionScore = 464,
-  kDynamicQuantExt = 465,
-  kQuantBatchMatmul = 466,
-  kMoeFinalizeRouting = 467,
-  kGroupedMatmulV4 = 468,
-  kMatmulAllReduceAddRmsNorm = 469,
-  kQuantV2 = 470,
-  kAddRmsNormQuantV2 = 471,
-  kMoeComputeExpertTokens = 472,
-  kGroupedMatmul = 473,
-  kWeightQuantBatchMatmul = 474,
-  kMoeGatingTopKSoftmax = 475,
-  kMoeInitRoutingV2 = 476,
-  kMoeInitRouting = 477,
-  kPixelShuffle = 478,
+  kTake = 28,
+  kAddcmulExt = 29,
+  kFlattenExt = 30,
+  kInplacePut = 31,
+  kCumsumExt = 32,
+  kCountNonZero = 33,
+  kAdaptiveAvgPool1D = 34,
+  kLogSoftmaxGrad = 35,
+  kDense = 36,
+  kMaxDim = 37,
+  kSoftmaxBackward = 38,
+  kInplaceGroupedMatmulAdd = 39,
+  kReciprocal = 40,
+  kInplaceFloorDivide = 41,
+  kConvTranspose2D = 42,
+  kMoeTokenPermuteGrad = 43,
+  kSliceExtView = 44,
+  kEye = 45,
+  kSoftMarginLoss = 46,
+  kRmsNorm = 47,
+  kTypeAs = 48,
+  kAllGatherMatmul = 49,
+  kInnerCommAllGather = 50,
+  kInplaceRemainderTensorScalar = 51,
+  kSlice = 52,
+  kConvolutionGrad = 53,
+  kInplaceHardtanh = 54,
+  kDistCommAllGatherIntoTensor = 55,
+  kSoftplusExt = 56,
+  kStdMean = 57,
+  kInplaceScatterValueReduce = 58,
+  kOuter = 59,
+  kMaskedFill = 60,
+  kSoftShrink = 61,
+  kInnerCommAllReduce = 62,
+  kMoeDistributeDispatch = 63,
+  kLogAddExp = 64,
+  kMoeTokenUnpermuteGrad = 65,
+  kInplaceMatmulAdd = 66,
+  kIndexAddExt = 67,
+  kSin = 68,
+  kBatchNormReduceGrad = 69,
+  kAcoshExt = 70,
+  kArgMinExt = 71,
+  kMoeDistributeCombine = 72,
+  kProdExt = 73,
+  kCross = 74,
+  kBitwiseNot = 75,
+  kBinaryCrossEntropyGrad = 76,
+  kEmbedding = 77,
+  kMishExt = 78,
+  kMuls = 79,
+  kIsClose = 80,
+  kPagedAttention = 81,
+  kReduceMin = 82,
+  kSoftShrinkGrad = 83,
+  kUpsampleTrilinear3D = 84,
+  kErf = 85,
+  kMaskedSelect = 86,
+  kReluGrad = 87,
+  kMv = 88,
+  kLogSigmoidGrad = 89,
+  kLeakyReLUExt = 90,
+  kSmoothL1Loss = 91,
+  kBatchNormExt = 92,
+  kNewOnes = 93,
+  kMedianDim = 94,
+  kNLLLoss = 95,
+  kNotEqual = 96,
+  kReduceMax = 97,
+  kStackExt = 98,
+  kConv1DExt = 99,
+  kSumExt = 100,
+  kMatMul = 101,
+  kReLU = 102,
+  kBitwiseOrTensor = 103,
+  kMedianExt = 104,
+  kReplicationPad2D = 105,
+  kInnerCommIsend = 106,
+  kReshape = 107,
+  kGluGrad = 108,
+  kThresholdGrad = 109,
+  kScatterAddExt = 110,
+  kIndexFillTensor = 111,
+  kInplaceDivMod = 112,
+  kAtanExt = 113,
+  kLayerNormExt = 114,
+  kAdamW = 115,
+  kSubScalar = 116,
+  kLogSumExp = 117,
+  kPReLU = 118,
+  kHShrink = 119,
+  kFrac = 120,
+  kSub = 121,
+  kInplaceDivs = 122,
+  kRandn = 123,
+  kInplaceZero = 124,
+  kEmptyLike = 125,
+  kLogSoftmaxExt = 126,
+  kFillTensor = 127,
+  kLogSoftmax = 128,
+  kGatherDGradV2 = 129,
+  kAtan2Ext = 130,
+  kReflectionPad1DGrad = 131,
+  kSelectExtView = 132,
+  kSoftMarginLossGrad = 133,
+  kKLDivGrad = 134,
+  kNansum = 135,
+  kBitwiseAndTensor = 136,
+  kGreaterEqualScalar = 137,
+  kNormalFloatFloat = 138,
+  kFloorDiv = 139,
+  kTrunc = 140,
+  kLayerNormGradExt = 141,
+  kHardtanhGrad = 142,
+  kGridSampler2D = 143,
+  kAddbmm = 144,
+  kChunkView = 145,
+  kUpsampleNearest1DGrad = 146,
+  kUniqueConsecutive = 147,
+  kAdaptiveAvgPool3DGradExt = 148,
+  kMm = 149,
+  kAddmm = 150,
+  kRandnLike = 151,
+  kBernoulliExt = 152,
+  kReflectionPad2D = 153,
+  kSpeedFusionAttentionGrad = 154,
+  kInnerInplaceIndexPut = 155,
+  kInplaceLog = 156,
+  kAvgPool2D = 157,
+  kFmodScalar = 158,
+  kDistCommScatterTensor = 159,
+  kAvgPool1D = 160,
+  kMoeTokenPermute = 161,
+  kReplicationPad1D = 162,
+  kCol2ImGrad = 163,
+  kCast = 164,
+  kRotaryPositionEmbeddingGrad = 165,
+  kReplicationPad3DGrad = 166,
+  kGcd = 167,
+  kBatchMatMul = 168,
+  kReplicationPad2DGrad = 169,
+  kIm2ColExt = 170,
+  kRotaryPositionEmbedding = 171,
+  kRoll = 172,
+  kThreshold = 173,
+  kMeanExt = 174,
+  kInplaceThreshold = 175,
+  kConv1DPadding = 176,
+  kLogicalXor = 177,
+  kMSELossGradExt = 178,
+  kLogSigmoid = 179,
+  kNormalFloatTensor = 180,
+  kCumminExt = 181,
+  kFlashAttentionScoreGrad = 182,
+  kInplaceFloor = 183,
+  kHShrinkGrad = 184,
+  kAsinhExt = 185,
+  kRepeatInterleaveInt = 186,
+  kMul = 187,
+  kArange = 188,
+  kFullLike = 189,
+  kRandExt = 190,
+  kXlogy = 191,
+  kIsNegInf = 192,
+  kBroadcastToView = 193,
+  kDistCommGather = 194,
+  kDistCommIsend = 195,
+  kBaddbmm = 196,
+  kSelectV2 = 197,
+  kRandInt = 198,
+  kExp2 = 199,
+  kArgSort = 200,
+  kInplaceFloorDivides = 201,
+  kNLLLossGrad = 202,
+  kAtanh = 203,
+  kInplaceSubScalar = 204,
+  kAvgPool2DGrad = 205,
+  kSinc = 206,
+  kMaxPoolGradWithMask = 207,
+  kInplaceDivMods = 208,
+  kMaxUnpool2DExt = 209,
+  kDivMods = 210,
+  kInplaceScatterSrc = 211,
+  kInnerCommAllToAllV = 212,
+  kTanhGrad = 213,
+  kMaxPoolWithIndices = 214,
+  kDistCommIrecv = 215,
+  kKLDiv = 216,
+  kGeluExt = 217,
+  kPReLUGrad = 218,
+  kGLU = 219,
+  kViewAs = 220,
+  kCellBackwardHook = 221,
+  kAsinExt = 222,
+  kAddLayerNormGrad = 223,
+  kExpandAs = 224,
+  kRandLikeExt = 225,
+  kNLLLoss2dGrad = 226,
+  kSign = 227,
+  kVar = 228,
+  kConvolutionStr = 229,
+  kIndexSelect = 230,
+  kInnerNonZero = 231,
+  kInplaceFillScalar = 232,
+  kCosh = 233,
+  kL1LossExt = 234,
+  kPromptFlashAttention = 235,
+  kInplaceMaskedFillScalar = 236,
+  kMaxPoolGradWithIndices = 237,
+  kConv3DExt = 238,
+  kRemainderScalarTensor = 239,
+  kTransposeExtView = 240,
+  kMinimum = 241,
+  kDropoutDoMaskExt = 242,
+  kPow = 243,
+  kRandIntLike = 244,
+  kUpsampleBicubic2D = 245,
+  kSwiglu = 246,
+  kDivMod = 247,
+  kLinalgQr = 248,
+  kCummax = 249,
+  kIsFinite = 250,
+  kLog = 251,
+  kSpeedFusionAttention = 252,
+  kAdaptiveAvgPool3DExt = 253,
+  kTriangularSolve = 254,
+  kRsqrt = 255,
+  kIndexFillScalar = 256,
+  kMultiScaleDeformableAttnGrad = 257,
+  kAddLayerNormV2 = 258,
+  kMaskedSelectGrad = 259,
+  kHistcExt = 260,
+  kSqueeze = 261,
+  kDistCommAllToAllVSingle = 262,
+  kSilentCheckV2 = 263,
+  kInplaceScatterSrcReduce = 264,
+  kRmsNormGrad = 265,
+  kXLogYScalarSelf = 266,
+  kLeakyReLUGradExt = 267,
+  kLogAddExp2 = 268,
+  kClampTensor = 269,
+  kSoftplusGradExt = 270,
+  kUpsampleNearest2DGrad = 271,
+  kAddScalar = 272,
+  kNanToNum = 273,
+  kReflectionPad3DGrad = 274,
+  kInplaceAddsExt = 275,
+  kReflectionPad3D = 276,
+  kDiagonalView = 277,
+  kSplit = 278,
+  kAsStrided = 279,
+  kTExt = 280,
+  kInplaceSubExt = 281,
+  kBinaryCrossEntropy = 282,
+  kDropoutGradExt = 283,
+  kBinaryCrossEntropyWithLogitsBackward = 284,
+  kInplaceBernoulliTensor = 285,
+  kOnes = 286,
+  kAvgPool3DGradExt = 287,
+  kNLLLoss2d = 288,
+  kFFNExt = 289,
+  kSoftmax = 290,
+  kInplaceRandom = 291,
+  kCol2ImExt = 292,
+  kDot = 293,
+  kExpm1 = 294,
+  kLerp = 295,
+  kUnstackExtView = 296,
+  kInplaceMuls = 297,
+  kHSwishGrad = 298,
+  kLess = 299,
+  kSliceExt = 300,
+  kInplaceExp = 301,
+  kAcosExt = 302,
+  kRandpermExt = 303,
+  kCrossEntropyLoss = 304,
+  kTile = 305,
+  kTanh = 306,
+  kCeil = 307,
+  kMinDim = 308,
+  kGreaterEqual = 309,
+  kMla = 310,
+  kAbs = 311,
+  kCrossEntropyLossGrad = 312,
+  kEluGradExt = 313,
+  kSwigluGrad = 314,
+  kOnesLikeExt = 315,
+  kBroadcastTo = 316,
+  kConcat = 317,
+  kInplaceIndexPut = 318,
+  kSearchSorted = 319,
+  kGroupNorm = 320,
+  kUniformExt = 321,
+  kRepeatInterleaveGrad = 322,
+  kGridSampler3DGrad = 323,
+  kOneHotExt = 324,
+  kSinh = 325,
+  kKthvalue = 326,
+  kUpsampleLinear1DGrad = 327,
+  kBatchNormElemtGrad = 328,
+  kSeLUExt = 329,
+  kInplaceNormal = 330,
+  kLogicalAnd = 331,
+  kInplaceUniform = 332,
+  kDiagExt = 333,
+  kRemainderTensorScalar = 334,
+  kConv3DPadding = 335,
+  kGridSampler2DGrad = 336,
+  kReplicationPad3D = 337,
+  kAddcdivExt = 338,
+  kInplaceCopy = 339,
+  kReshapeAndCache = 340,
+  kAdaptiveMaxPool1D = 341,
+  kPolar = 342,
+  kNorm = 343,
+  kSeluGrad = 344,
+  kSubExt = 345,
+  kCos = 346,
+  kIdentity = 347,
+  kArgMaxExt = 348,
+  kMultinomialExt = 349,
+  kRemainderTensorTensor = 350,
+  kReduceAny = 351,
+  kFlashAttentionScore = 352,
+  kChunk = 353,
+  kMaxPoolWithMask = 354,
+  kConv2DExt = 355,
+  kSplitWithSize = 356,
+  kSelect = 357,
+  kNormalTensorFloat = 358,
+  kUpsampleTrilinear3DGrad = 359,
+  kMatrixInverseExt = 360,
+  kBatchNormStats = 361,
+  kMishGradExt = 362,
+  kInplaceSiLU = 363,
+  kInplaceStopGradient = 364,
+  kArgMinWithValue = 365,
+  kInplaceScatterValue = 366,
+  kExpandDims = 367,
+  kAvgPool3DExt = 368,
+  kAddmv = 369,
+  kSplitTensorView = 370,
+  kRepeatInterleaveTensor = 371,
+  kGeluGradExt = 372,
+  kReduceAll = 373,
+  kDropoutExt = 374,
+  kUpsampleNearest2D = 375,
+  kUpsampleBilinear2DGrad = 376,
+  kTriu = 377,
+  kMatMulExt = 378,
+  kXLogYScalarOther = 379,
+  kUpsampleNearest3D = 380,
+  kInnerCommReduceScatter = 381,
+  kNonZero = 382,
+  kIsInf = 383,
+  kMSELossExt = 384,
+  kNeg = 385,
+  kUnique2 = 386,
+  kAdd = 387,
+  kL1LossBackwardExt = 388,
+  kNonZeroExt = 389,
+  kEqual = 390,
+  kGenerator = 391,
+  kGeLUGrad = 392,
+  kAdaptiveAvgPool2DExt = 393,
+  kSplitWithSizeView = 394,
+  kInplaceClampScalar = 395,
+  kSortExt = 396,
+  kZeros = 397,
+  kEmbeddingDenseBackward = 398,
+  kRingAttentionUpdate = 399,
+  kGeLU = 400,
+  kLinSpaceExt = 401,
+  kInplaceFillTensor = 402,
+  kInplaceAddmm = 403,
+  kGridSampler3D = 404,
+  kDistCommReduceScatter = 405,
+  kGroupNormGrad = 406,
+  kSiLUGrad = 407,
+  kDistCommGatherIntoTensor = 408,
+  kLog10 = 409,
+  kSplitTensor = 410,
+  kInplaceMaskedFillTensor = 411,
+  kExp = 412,
+  kNarrowView = 413,
+  kHSwish = 414,
+  kInplaceErfinv = 415,
+  kDistCommAllToAllV = 416,
+  kVarMean = 417,
+  kErfinv = 418,
+  kTranspose = 419,
+  kBatchNormGradExt = 420,
+  kMax = 421,
+  kPowTensorScalar = 422,
+  kLessEqual = 423,
+  kDivs = 424,
+  kHSigmoidGrad = 425,
+  kExpandDimsView = 426,
+  kBitwiseXorTensor = 427,
+  kTransposeView = 428,
+  kTrilExt = 429,
+  kConvolution = 430,
+  kFillScalar = 431,
+  kView = 432,
+  kTopkExt = 433,
+  kBCEWithLogitsLoss = 434,
+  kReflectionPad1D = 435,
+  kDistCommBarrier = 436,
+  kSmoothL1LossGrad = 437,
+  kLogicalOr = 438,
+  kReverseV2 = 439,
+  kInnerMoeTokenUnpermute = 440,
+  kInplaceClampTensor = 441,
+  kIndex = 442,
+  kAddExt = 443,
+  kNeScalar = 444,
+  kCopy = 445,
+  kMeshgrid = 446,
+  kUpsampleBilinear2D = 447,
+  kClampScalar = 448,
+  kElu = 449,
+  kDistCommReduce = 450,
+  kBatchMatMulExt = 451,
+  kFloor = 452,
+  kGatherD = 453,
+  kBatchNormElemt = 454,
+  kPowScalarTensor = 455,
+  kLog2 = 456,
+  kInplaceAddExt = 457,
+  kEqualExt = 458,
+  kConv2DPadding = 459,
+  kEluExt = 460,
+  kScatterValue = 461,
+  kDiv = 462,
+  kInplaceMul = 463,
+  kGreater = 464,
+  kAdaptiveMaxPool2D = 465,
+  kAllFinite = 466,
+  kInplaceReLU = 467,
+  kReflectionPad2DGrad = 468,
+  kDistCommReduceScatterTensor = 469,
+  kRound = 470,
+  kApplyRotaryPosEmb = 471,
+  kInplaceFillDiagonal = 472,
+  kAdaptiveAvgPool2DGradExt = 473,
+  kDistCommBatchIsendIrecv = 474,
+  kInplaceDiv = 475,
+  kDistCommScatter = 476,
+  kZerosLikeExt = 477,
+  kMatmulReduceScatter = 478,
+  kMin = 479,
+  kMultiScaleDeformableAttn = 480,
+  kInplaceBernoulliScalar = 481,
+  kLinalgVectorNorm = 482,
+  kBitwiseXorScalar = 483,
+  kHSigmoid = 484,
+  kBincountExt = 485,
+  kInnerIndex = 486,
+  kTensorScatterElements = 487,
+  kSiLU = 488,
+  kStd = 489,
+  kDistCommBroadcast = 490,
+  kHardtanh = 491,
+  kMaskedScatter = 492,
+  kConstantPadND = 493,
+  kUniqueDim = 494,
+  kUpsampleNearest3DGrad = 495,
+  kTraceExt = 496,
+  kLogicalNot = 497,
+  kLog1p = 498,
+  kNewEmpty = 499,
+  kSigmoidGrad = 500,
+  kInplaceTanh = 501,
+  kSqrt = 502,
+  kAddRmsNorm = 503,
+  kScatter = 504,
+  kUpsampleLinear1D = 505,
+  kInplaceIndexAddExt = 506,
+  kIncreFlashAttention = 507,
+  kClone = 508,
+  kArgMaxWithValue = 509,
+  kDropoutGenMaskExt = 510,
+  kTan = 511,
+  kDistCommReduceScatterTensorUneven = 512,
+  kInnerCommIrecv = 513,
+  kContiguous = 514,
+  kQuantV2 = 515,
+  kAddRmsNormQuantV2 = 516,
+  kKVCacheScatterUpdate = 517,
+  kQuantBatchMatmul = 518,
+  kFusedInferAttentionScore = 519,
+  kMoeInitRouting = 520,
+  kGroupedMatmul = 521,
+  kMoeGatingTopKSoftmax = 522,
+  kMatmulAllReduceAddRmsNorm = 523,
+  kWeightQuantBatchMatmul = 524,
+  kMoeComputeExpertTokens = 525,
+  kMoeInitRoutingQuantV2 = 526,
+  kDynamicQuantExt = 527,
+  kGroupedMatmulV4 = 528,
+  kMoeInitRoutingV2 = 529,
+  kGroupedMatmulV2 = 530,
+  kQuantMatmul = 531,
+  kMoeFinalizeRouting = 532,
+  kFuncDropoutExt = 533,
+  kAny = 534,
+  kFuncMaxPool2D = 535,
+  kGmmV2 = 536,
+  kGmm = 537,
+  kGmmBackward = 538,
+  kGmmV2Backward = 539,
+  kAnyExt = 540,
+  kGmmV2BackwardFusion = 541,
+  kInplaceExponential = 542,
+  kPixelShuffle = 543,
+  kEinsumExt = 544,
+  kGmmBackwardFusion = 545,
+  kMoeTokenUnpermute = 546,
 };
-using FmodTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using IndexGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using ScatterValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::Int64ImmPtr &)>;
-using EmbeddingDenseBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
-using SoftMarginLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using LogicalXorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceSubExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using AllFiniteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &)>;
-using SincGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceDivModsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using CummaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using AtanExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ExpandAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaxUnpool2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using GreaterEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NotEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ErfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BatchNormElemtGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceNormalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NLLLoss2dGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using UpsampleNearest3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using LogSoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using OuterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SiLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using UpsampleNearest3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using UpsampleBicubic2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using ConcatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using AdaptiveAvgPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using UpsampleBilinear2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using Log2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using GmmV2BackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-using MSELossExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using LerpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DiagExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using FracGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using GluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using GeneratorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
-using FlattenExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using SoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using DivModGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using SoftMarginLossGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using MaskedFillGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using GridSampler3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &)>;
-using ArgMaxExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
-using InplacePutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &)>;
-using TopkExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using Conv1DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using ReplicationPad3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using ReplicationPad3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using TExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using CountNonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using InplaceMaskedFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using XlogyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SoftShrinkGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using ReduceMinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
-using MatrixInverseExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceEluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using UpsampleTrilinear3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using InplaceHardtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using AddLayerNormV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
-using BitwiseAndTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceClampScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ScalarPtr> &, const std::optional<mindspore::ScalarPtr> &)>;
-using LayerNormGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using InplaceScatterAddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SmoothL1LossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using ReflectionPad3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using ClampTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-using MoeTokenUnpermuteGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using LogSoftmaxExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using CeilGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using GreaterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using ReciprocalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MulsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using InplaceUniformGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LayerNormExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::FP32ImmPtr &)>;
-using BaddbmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using Conv2DPaddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using MultinomialExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using Conv2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using DenseGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-using ConvolutionStrGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
-using OnesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using RepeatInterleaveGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using PowGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SmoothL1LossGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using FlashAttentionScoreGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceCopyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceScatterValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using UnstackExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using AvgPool3DGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using MaxPoolWithMaskGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using HSigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SilentCheckV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using ErfcGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BinaryCrossEntropyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-using ReduceAllGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using LogSumExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
-using InplaceDivModGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using GeluGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using IndexFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using SubGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaximumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NLLLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using XLogYScalarOtherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using IdentityGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BitwiseXorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using LogSigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ScatterAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AdaptiveAvgPool2DGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NormalFloatTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::FP32ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BitwiseXorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using GLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using UpsampleNearest1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using DivModsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using FFNExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using AddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using CopyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AsinExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReplicationPad2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using NeScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using GmmBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using GroupNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::FP32ImmPtr &)>;
-using VarMeanGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using ConvolutionGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
-using MaxPoolGradWithIndicesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LinSpaceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using NewOnesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using ArgMinWithValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using StdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using SinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SortExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using DropoutDoMaskExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using SoftmaxBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using TypeAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using GcdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AvgPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using SliceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
-using FloorDivScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using MoeTokenPermuteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
-using MatMulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InnerNonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReflectionPad3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using RepeatInterleaveTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using MedianDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using UpsampleNearest2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using AdaptiveMaxPool2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using InplaceZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ErfinvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceLogGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using HShrinkGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using KthvalueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using NormalTensorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReplicationPad1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using BatchMatMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using AddmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using SelectV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SoftplusGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using BatchNormExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &)>;
-using GroupNormGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using RandIntLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using MinDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using ContiguousGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ArgMaxWithValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using MaskedSelectGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SpeedFusionAttentionGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using LogicalOrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using RemainderTensorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AdaptiveAvgPool3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using RandpermExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using GeLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogicalAndGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using CastGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using MishExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BCEWithLogitsLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-using PolarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using RemainderScalarTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AvgPool2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using XLogYScalarSelfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DropoutGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using ConvTranspose2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
-using HShrinkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using L1LossBackwardExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using BitwiseOrScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using ExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BatchNormGatherStatsWithCountsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-using ZerosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using EqualExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using CloneGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaxPoolWithIndicesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using AdaptiveAvgPool2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using Conv3DPaddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using ChunkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using AbsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using CumsumExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using TraceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SignGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceClampTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-using Atan2ExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ArgMinExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
-using OnesLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using AcoshExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MedianExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceFloorDivideGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReplicationPad2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using ReduceAnyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
-using BincountExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-using RepeatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using RemainderTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using TanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using TrilExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using KLDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using MeshgridGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using AddmvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using NansumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using HSigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ThresholdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &)>;
-using SqrtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using RotaryPositionEmbeddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using UpsampleTrilinear3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using SliceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using SplitWithSizeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceAddsExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using Expm1GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceDivsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using NonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using RoundGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using AdaptiveAvgPool3DGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceMulsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using ReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LerpScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using CoshGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using IsInfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InnerIndexGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using Log10GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DivsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using IndexFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using GatherDGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AvgPool2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using Im2ColExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
-using NanToNumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::FP32ImmPtr> &, const std::optional<mindspore::FP32ImmPtr> &, const std::optional<mindspore::FP32ImmPtr> &)>;
-using LinalgVectorNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using SigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ProdExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using AddScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using NLLLoss2dGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using Log1pGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using Conv3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using SilentCheckV3GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using MultiScaleDeformableAttnGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using InplaceFloorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceTanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AsinhExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ConvolutionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using UpsampleBicubic2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using EmbeddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::FP32ImmPtr> &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
-using BinaryCrossEntropyGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-using AddbmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using MoeTokenUnpermuteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using HardtanhGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using AcosExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SwigluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using MultiScaleDeformableAttnGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MinimumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using HistcExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using MmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReflectionPad1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using GridSampler2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using BitwiseNotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using UpsampleLinear1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using RotaryPositionEmbeddingGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-using Unique2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using SpeedFusionAttentionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using ClampScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ScalarPtr> &, const std::optional<mindspore::ScalarPtr> &)>;
-using Col2ImExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
-using GridSampler3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using LogAddExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceFloorDividesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using BatchNormReduceGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using ViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using BatchMatMulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceErfinvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaxPoolGradWithMaskGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using TileGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using SoftShrinkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using ReshapeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using BatchNormElemtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::FP32ImmPtr &)>;
-using LinalgQrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using SigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using L1LossExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using ViewAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using TransposeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using UpsampleNearest2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using UpsampleNearest1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using SquareGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using NormalTensorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using SilentCheckV3GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using ConvolutionStrGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
+using DistCommAllGatherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using MaximumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using FloorDivScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using LerpScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using DistCommAllReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using FmodTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceEluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using InplaceScatterAddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ErfcGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BitwiseAndScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using BatchNormGatherStatsWithCountsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using NewFullGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using BitwiseOrScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using NarrowGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &, const int64_t &)>;
+using InplaceRemainderTensorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DistCommAllGatherIntoTensorUnevenGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using EmptyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using ReplicationPad1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using NewZerosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using SigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using RepeatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
 using CustomExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &)>;
-using DropoutGenMaskExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::FP32ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceIndexAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using EluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using SinhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceMaskedFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using GeluExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using BatchNormGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::ValueTuplePtr &)>;
-using NormalTensorFloatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using RmsNormGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SeluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NonZeroExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SquareGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReduceMaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
-using BitwiseAndScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using InplaceScatterValueReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::Int64ImmPtr &)>;
-using ThresholdGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using UniqueDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using HSwishGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LeakyReLUGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::BoolImmPtr &)>;
-using CumminExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using IsCloseGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
-using TakeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReflectionPad1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using AddcdivExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using VarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using MulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceThresholdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &)>;
-using RmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using BinaryCrossEntropyWithLogitsBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-using StdMeanGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using MoeTokenPermuteGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using RandIntGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using SoftplusExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using ReflectionPad2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using PowScalarTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceScatterSrcReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using EqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using RandnGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using RollGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using GeLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AsStridedGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using TensorScatterElementsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using FlashAttentionScoreGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using AtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceStopGradientGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogicalNotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DropoutExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NormalFloatFloatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using Conv1DPaddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using RsqrtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ConstantPadNDGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ScalarPtr &)>;
-using BitwiseOrTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MSELossGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using MeanExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using LessEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using PromptFlashAttentionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using SelectExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using EluExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using Col2ImGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
-using TriangularSolveGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using FillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using SubScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using IndexAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using FillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using LeakyReLUExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using RandExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using DivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using TransposeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using GatherDGradV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using PReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceIndexPutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &)>;
-using KLDivGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using UpsampleNearest1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using UniformExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using IsFiniteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BroadcastToGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using SplitTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using ExpandDimsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using FloorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogSoftmaxGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using LessGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReplicationPad1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using AdamWGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using InplaceRandomGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SplitGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using AvgPool3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using NarrowGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using SqueezeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using LogAddExp2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NegGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceFillDiagonalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::BoolImmPtr &)>;
-using InplaceGroupedMatmulAddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BernoulliExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReflectionPad2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using FullLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using UpsampleBicubic2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using TakeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AddcmulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using FlattenExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using InplacePutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &)>;
+using CumsumExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using CountNonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using AdaptiveAvgPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using LogSoftmaxGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using DenseGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using MaxDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using SoftmaxBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceGroupedMatmulAddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReciprocalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceFloorDivideGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ConvTranspose2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
+using MoeTokenPermuteGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using SliceExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &, const int64_t &, const int64_t &)>;
 using EyeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using ConvolutionStrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using GridSampler2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &)>;
-using HSwishGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using UpsampleLinear1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using UpsampleBilinear2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using AllGatherMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using MaxDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using IncreFlashAttentionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using TriuGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using FloorDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AddRmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using TanhGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using Exp2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using HardtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using IndexSelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogSigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
+using SoftMarginLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using RmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using TypeAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AllGatherMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using InnerCommAllGatherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using InplaceRemainderTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using SliceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &, const std::vector<int64_t> &)>;
+using ConvolutionGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceHardtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using DistCommAllGatherIntoTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using SoftplusExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using StdMeanGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using InplaceScatterValueReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::Int64ImmPtr &)>;
+using OuterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MaskedFillGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using SoftShrinkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using InnerCommAllReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using MoeDistributeDispatchGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::StringImmPtr> &, const std::optional<mindspore::StringImmPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using LogAddExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MoeTokenUnpermuteGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using InplaceMatmulAddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using IndexAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using SinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BatchNormReduceGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using AcoshExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ArgMinExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
+using MoeDistributeCombineGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::StringImmPtr> &, const std::optional<mindspore::StringImmPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using ProdExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using CrossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using BitwiseNotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BinaryCrossEntropyGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using EmbeddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::FP32ImmPtr> &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
+using MishExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MulsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using IsCloseGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
+using PagedAttentionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using ReduceMinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
+using SoftShrinkGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using UpsampleTrilinear3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using ErfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MaskedSelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using LogSigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using LeakyReLUExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using SmoothL1LossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using BatchNormExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &)>;
+using NewOnesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using MedianDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using NLLLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using NotEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReduceMaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
 using StackExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using MaskedSelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MishGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReverseV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using OneHotExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using PReLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using IsNegInfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using TruncGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MatMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using NLLLossGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using SwigluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using RepeatInterleaveIntGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using GreaterEqualScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using RandnLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using ZerosLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using NewZerosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using ArgSortGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using SearchSortedGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using SiLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceSubScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using InnerInplaceIndexPutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &)>;
-using SumExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using CosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SubExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
+using Conv1DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using SumExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using MatMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using ReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BitwiseOrTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MedianExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReplicationPad2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using InnerCommIsendGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &)>;
+using ReshapeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using GluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using ThresholdGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using ScatterAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using IndexFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceDivModGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using AtanExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using LayerNormExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::FP32ImmPtr &)>;
+using AdamWGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using SubScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using LogSumExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
+using PReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using HShrinkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using FracGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SubGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceDivsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using RandnGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using InplaceZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using EmptyLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using LogSoftmaxExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using FillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using LogSoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using GatherDGradV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using Atan2ExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReflectionPad1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using SelectExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using SoftMarginLossGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using KLDivGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using NansumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using BitwiseAndTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using GreaterEqualScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using NormalFloatFloatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using FloorDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using TruncGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using LayerNormGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using HardtanhGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using GridSampler2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using AddbmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using ChunkViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using UpsampleNearest1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using UniqueConsecutiveGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using AdaptiveAvgPool3DGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AddmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using RandnLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using BernoulliExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReflectionPad2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using SpeedFusionAttentionGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using InnerInplaceIndexPutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &)>;
+using InplaceLogGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using AvgPool2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using FmodScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using DistCommScatterTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using AvgPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using MoeTokenPermuteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
+using ReplicationPad1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using Col2ImGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
+using CastGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using RotaryPositionEmbeddingGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using ReplicationPad3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using GcdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using BatchMatMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using ReplicationPad2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using Im2ColExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
+using RotaryPositionEmbeddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using RollGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using ThresholdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using MeanExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using InplaceThresholdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using Conv1DPaddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using LogicalXorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MSELossGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using LogSigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using NormalFloatTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using CumminExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using FlashAttentionScoreGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceFloorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using HShrinkGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using AsinhExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using RepeatInterleaveIntGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using MulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
 using ArangeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using InplaceScatterSrcGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AddcmulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using CrossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using AdaptiveMaxPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using MatmulReduceScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using TanGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using RandLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using BatchNormStatsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using PowTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using InplaceAddmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using SeLUExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using UniqueConsecutiveGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using FmodScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using AddLayerNormGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using EluGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
+using FullLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using RandExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using XlogyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using IsNegInfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BroadcastToViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using DistCommGatherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using DistCommIsendGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &)>;
+using BaddbmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using SelectV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using RandIntGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using Exp2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ArgSortGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using InplaceFloorDividesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using NLLLossGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using AtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceSubScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using AvgPool2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using SincGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MaxPoolGradWithMaskGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceDivModsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using MaxUnpool2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using DivModsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using InplaceScatterSrcGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InnerCommAllToAllVGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using TanhGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MaxPoolWithIndicesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using DistCommIrecvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using KLDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using GeluExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using PReLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using GLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using ViewAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using CellBackwardHookGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &)>;
+using AsinExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using AddLayerNormGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ExpandAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using RandLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using NLLLoss2dGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &)>;
+using SignGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using VarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using ConvolutionStrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using IndexSelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &)>;
+using InnerNonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using CoshGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using L1LossExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using PromptFlashAttentionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceMaskedFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using MaxPoolGradWithIndicesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using Conv3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using RemainderScalarTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &)>;
+using TransposeExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using MinimumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DropoutDoMaskExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using PowGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using RandIntLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using UpsampleBicubic2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using SwigluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using DivModGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using LinalgQrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using CummaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using IsFiniteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using LogGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SpeedFusionAttentionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using AdaptiveAvgPool3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using TriangularSolveGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using RsqrtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using IndexFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using MultiScaleDeformableAttnGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AddLayerNormV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
+using MaskedSelectGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using HistcExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using SqueezeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using DistCommAllToAllVSingleGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using SilentCheckV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceScatterSrcReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using RmsNormGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using XLogYScalarSelfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &)>;
+using LeakyReLUGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::BoolImmPtr &)>;
+using LogAddExp2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ClampTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using SoftplusGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using UpsampleNearest2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using AddScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using NanToNumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::FP32ImmPtr> &, const std::optional<mindspore::FP32ImmPtr> &, const std::optional<mindspore::FP32ImmPtr> &)>;
+using ReflectionPad3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceAddsExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using ReflectionPad3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using DiagonalViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &, const int64_t &)>;
+using SplitGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using AsStridedGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &, const std::vector<int64_t> &, const int64_t &)>;
+using TExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceSubExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using BinaryCrossEntropyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using DropoutGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using BinaryCrossEntropyWithLogitsBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using InplaceBernoulliTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using OnesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using AvgPool3DGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using NLLLoss2dGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using FFNExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using SoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceRandomGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using Col2ImExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
+using DotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using Expm1GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using LerpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using UnstackExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &)>;
+using InplaceMulsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using HSwishGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using LessGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using SliceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &, const int64_t &, const int64_t &)>;
+using InplaceExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using AcosExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using RandpermExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using CrossEntropyLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
+using TileGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using TanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using CeilGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MinDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using GreaterEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MlaGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using AbsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using CrossEntropyLossGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &)>;
+using EluGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
+using SwigluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using OnesLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using BroadcastToGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using ConcatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceIndexPutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &)>;
+using SearchSortedGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using GroupNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::FP32ImmPtr &)>;
+using UniformExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using RepeatInterleaveGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using GridSampler3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &)>;
+using OneHotExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using SinhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using KthvalueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using UpsampleLinear1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using BatchNormElemtGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using SeLUExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceNormalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using LogicalAndGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceUniformGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DiagExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using RemainderTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using Conv3DPaddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using GridSampler2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &)>;
+using ReplicationPad3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using AddcdivExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using InplaceCopyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &)>;
+using ReshapeAndCacheGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using AdaptiveMaxPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using PolarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using NormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using SeluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using SubExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using CosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using IdentityGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ArgMaxExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
+using MultinomialExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using RemainderTensorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReduceAnyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
+using FlashAttentionScoreGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using ChunkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using MaxPoolWithMaskGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using Conv2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using SplitWithSizeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &, const int64_t &)>;
+using SelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using NormalTensorFloatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using UpsampleTrilinear3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using MatrixInverseExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BatchNormStatsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using MishGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceSiLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceStopGradientGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ArgMinWithValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using InplaceScatterValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using ExpandDimsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &)>;
+using AvgPool3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using AddmvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using SplitTensorViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using RepeatInterleaveTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using GeluGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using ReduceAllGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using DropoutExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using UpsampleNearest2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using UpsampleBilinear2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using TriuGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using MatMulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using XLogYScalarOtherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using UpsampleNearest3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using InnerCommReduceScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using NonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using IsInfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MSELossExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using NegGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using Unique2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using AddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using L1LossBackwardExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using NonZeroExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using EqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using GeneratorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
+using GeLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AdaptiveAvgPool2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using SplitWithSizeViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &, const int64_t &)>;
+using InplaceClampScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ScalarPtr> &, const std::optional<mindspore::ScalarPtr> &)>;
+using SortExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using ZerosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using EmbeddingDenseBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
+using RingAttentionUpdateGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using GeLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using LinSpaceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using InplaceFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceAddmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using GridSampler3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using DistCommReduceScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using GroupNormGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using SiLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DistCommGatherIntoTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using Log10GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SplitTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using InplaceMaskedFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using NarrowViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &, const int64_t &)>;
+using HSwishGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceErfinvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using DistCommAllToAllVGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using VarMeanGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using ErfinvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using TransposeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using BatchNormGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::ValueTuplePtr &)>;
+using MaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using PowTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using LessEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DivsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using HSigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ExpandDimsViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &)>;
+using BitwiseXorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using TransposeViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using TrilExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using ConvolutionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using FillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using ViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using TopkExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using BCEWithLogitsLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using ReflectionPad1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using DistCommBarrierGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::StringImmPtr &)>;
+using SmoothL1LossGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using LogicalOrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReverseV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using InnerMoeTokenUnpermuteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using InplaceClampTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using IndexGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using AddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using NeScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using CopyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MeshgridGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const int64_t &)>;
+using UpsampleBilinear2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using ClampScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ScalarPtr> &, const std::optional<mindspore::ScalarPtr> &)>;
+using EluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using DistCommReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using BatchMatMulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using FloorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using GatherDGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &)>;
+using BatchNormElemtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::FP32ImmPtr &)>;
+using PowScalarTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &)>;
+using Log2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using EqualExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using Conv2DPaddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using EluExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using ScatterValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::Int64ImmPtr &)>;
+using DivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using GreaterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AdaptiveMaxPool2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using AllFiniteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReflectionPad2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using DistCommReduceScatterTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using RoundGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using ApplyRotaryPosEmbGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceFillDiagonalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::BoolImmPtr &)>;
+using AdaptiveAvgPool2DGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DistCommBatchIsendIrecvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::StringImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DistCommScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using ZerosLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using MatmulReduceScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using MinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MultiScaleDeformableAttnGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceBernoulliScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using LinalgVectorNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using BitwiseXorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using HSigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BincountExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using InnerIndexGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using TensorScatterElementsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using SiLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using StdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using DistCommBroadcastGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using HardtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using MaskedScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ConstantPadNDGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ScalarPtr &)>;
+using UniqueDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using UpsampleNearest3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using TraceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using LogicalNotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using Log1pGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using NewEmptyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using SigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceTanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SqrtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using AddRmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using ScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using UpsampleLinear1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
+using InplaceIndexAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using IncreFlashAttentionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using CloneGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ArgMaxWithValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using DropoutGenMaskExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::FP32ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using TanGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using DistCommReduceScatterTensorUnevenGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using InnerCommIrecvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &)>;
+using ContiguousGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using QuantV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using AddRmsNormQuantV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using KVCacheScatterUpdateGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using QuantBatchMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using FusedInferAttentionScoreGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using MoeInitRoutingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using GroupedMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using MoeGatingTopKSoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using MatmulAllReduceAddRmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using WeightQuantBatchMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using MoeComputeExpertTokensGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using MoeInitRoutingQuantV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using DynamicQuantExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using GroupedMatmulV4GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using MoeInitRoutingV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
 using GroupedMatmulV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using KVCacheScatterUpdateGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using FusedInferAttentionScoreGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using DynamicQuantExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-using QuantBatchMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using MoeFinalizeRoutingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-using GroupedMatmulV4GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using MatmulAllReduceAddRmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using QuantV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using AddRmsNormQuantV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using MoeComputeExpertTokensGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using GroupedMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using WeightQuantBatchMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using MoeGatingTopKSoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-using MoeInitRoutingV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using MoeInitRoutingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using PixelShuffleGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
+using QuantMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using MoeFinalizeRoutingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using FuncDropoutExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AnyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using FuncMaxPool2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using GmmV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using GmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using GmmBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &)>;
+using GmmV2BackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using AnyExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using GmmV2BackwardFusionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using InplaceExponentialGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using PixelShuffleGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using EinsumExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::StringImmPtr &, const mindspore::ValueTuplePtr &)>;
+using GmmBackwardFusionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &)>;
+using MoeTokenUnpermuteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
 struct OpsAutoGradRegisters {
+  UpsampleNearest1DGradFunc UpsampleNearest1DGradFuncObj;
+  SquareGradFunc SquareGradFuncObj;
+  NormalTensorTensorGradFunc NormalTensorTensorGradFuncObj;
+  SilentCheckV3GradFunc SilentCheckV3GradFuncObj;
+  ConvolutionStrGradGradFunc ConvolutionStrGradGradFuncObj;
+  DistCommAllGatherGradFunc DistCommAllGatherGradFuncObj;
+  MaximumGradFunc MaximumGradFuncObj;
+  FloorDivScalarGradFunc FloorDivScalarGradFuncObj;
+  LerpScalarGradFunc LerpScalarGradFuncObj;
+  DistCommAllReduceGradFunc DistCommAllReduceGradFuncObj;
   FmodTensorGradFunc FmodTensorGradFuncObj;
-  IndexGradFunc IndexGradFuncObj;
-  ScatterValueGradFunc ScatterValueGradFuncObj;
-  EmbeddingDenseBackwardGradFunc EmbeddingDenseBackwardGradFuncObj;
-  SoftMarginLossGradFunc SoftMarginLossGradFuncObj;
-  LogicalXorGradFunc LogicalXorGradFuncObj;
-  InplaceSubExtGradFunc InplaceSubExtGradFuncObj;
-  AllFiniteGradFunc AllFiniteGradFuncObj;
-  SincGradFunc SincGradFuncObj;
-  InplaceDivModsGradFunc InplaceDivModsGradFuncObj;
-  CummaxGradFunc CummaxGradFuncObj;
-  AtanExtGradFunc AtanExtGradFuncObj;
-  ExpandAsGradFunc ExpandAsGradFuncObj;
-  MaxUnpool2DExtGradFunc MaxUnpool2DExtGradFuncObj;
-  GreaterEqualGradFunc GreaterEqualGradFuncObj;
-  MaxGradFunc MaxGradFuncObj;
-  NotEqualGradFunc NotEqualGradFuncObj;
-  ErfGradFunc ErfGradFuncObj;
-  BatchNormElemtGradGradFunc BatchNormElemtGradGradFuncObj;
-  InplaceNormalGradFunc InplaceNormalGradFuncObj;
-  NLLLoss2dGradFunc NLLLoss2dGradFuncObj;
-  UpsampleNearest3DGradFunc UpsampleNearest3DGradFuncObj;
-  LogSoftmaxGradFunc LogSoftmaxGradFuncObj;
-  OuterGradFunc OuterGradFuncObj;
-  InplaceExpGradFunc InplaceExpGradFuncObj;
-  SiLUGradGradFunc SiLUGradGradFuncObj;
-  UpsampleNearest3DGradGradFunc UpsampleNearest3DGradGradFuncObj;
+  InplaceEluGradFunc InplaceEluGradFuncObj;
+  InplaceScatterAddGradFunc InplaceScatterAddGradFuncObj;
+  ErfcGradFunc ErfcGradFuncObj;
+  BitwiseAndScalarGradFunc BitwiseAndScalarGradFuncObj;
+  BatchNormGatherStatsWithCountsGradFunc BatchNormGatherStatsWithCountsGradFuncObj;
+  NewFullGradFunc NewFullGradFuncObj;
+  BitwiseOrScalarGradFunc BitwiseOrScalarGradFuncObj;
+  NarrowGradFunc NarrowGradFuncObj;
+  InplaceRemainderTensorTensorGradFunc InplaceRemainderTensorTensorGradFuncObj;
+  DistCommAllGatherIntoTensorUnevenGradFunc DistCommAllGatherIntoTensorUnevenGradFuncObj;
+  EmptyGradFunc EmptyGradFuncObj;
+  ReplicationPad1DGradGradFunc ReplicationPad1DGradGradFuncObj;
+  NewZerosGradFunc NewZerosGradFuncObj;
+  SigmoidGradFunc SigmoidGradFuncObj;
+  RepeatGradFunc RepeatGradFuncObj;
+  CustomExtGradFunc CustomExtGradFuncObj;
   UpsampleBicubic2DGradGradFunc UpsampleBicubic2DGradGradFuncObj;
-  ConcatGradFunc ConcatGradFuncObj;
-  AdaptiveAvgPool1DGradFunc AdaptiveAvgPool1DGradFuncObj;
-  UpsampleBilinear2DGradFunc UpsampleBilinear2DGradFuncObj;
-  Log2GradFunc Log2GradFuncObj;
-  GmmV2BackwardGradFunc GmmV2BackwardGradFuncObj;
-  MSELossExtGradFunc MSELossExtGradFuncObj;
-  LerpGradFunc LerpGradFuncObj;
-  DiagExtGradFunc DiagExtGradFuncObj;
-  FracGradFunc FracGradFuncObj;
-  GluGradGradFunc GluGradGradFuncObj;
-  GeneratorGradFunc GeneratorGradFuncObj;
+  TakeGradFunc TakeGradFuncObj;
+  AddcmulExtGradFunc AddcmulExtGradFuncObj;
   FlattenExtGradFunc FlattenExtGradFuncObj;
-  SoftmaxGradFunc SoftmaxGradFuncObj;
-  DivModGradFunc DivModGradFuncObj;
-  SoftMarginLossGradGradFunc SoftMarginLossGradGradFuncObj;
-  MaskedFillGradFunc MaskedFillGradFuncObj;
-  GridSampler3DGradGradFunc GridSampler3DGradGradFuncObj;
-  ArgMaxExtGradFunc ArgMaxExtGradFuncObj;
   InplacePutGradFunc InplacePutGradFuncObj;
-  TopkExtGradFunc TopkExtGradFuncObj;
-  Conv1DExtGradFunc Conv1DExtGradFuncObj;
-  ReplicationPad3DGradFunc ReplicationPad3DGradFuncObj;
-  ReplicationPad3DGradGradFunc ReplicationPad3DGradGradFuncObj;
-  TExtGradFunc TExtGradFuncObj;
+  CumsumExtGradFunc CumsumExtGradFuncObj;
   CountNonZeroGradFunc CountNonZeroGradFuncObj;
-  InplaceMaskedFillTensorGradFunc InplaceMaskedFillTensorGradFuncObj;
-  XlogyGradFunc XlogyGradFuncObj;
-  SoftShrinkGradGradFunc SoftShrinkGradGradFuncObj;
-  ReduceMinGradFunc ReduceMinGradFuncObj;
-  MatrixInverseExtGradFunc MatrixInverseExtGradFuncObj;
-  InplaceEluGradFunc InplaceEluGradFuncObj;
-  UpsampleTrilinear3DGradFunc UpsampleTrilinear3DGradFuncObj;
+  AdaptiveAvgPool1DGradFunc AdaptiveAvgPool1DGradFuncObj;
+  LogSoftmaxGradGradFunc LogSoftmaxGradGradFuncObj;
+  DenseGradFunc DenseGradFuncObj;
+  MaxDimGradFunc MaxDimGradFuncObj;
+  SoftmaxBackwardGradFunc SoftmaxBackwardGradFuncObj;
+  InplaceGroupedMatmulAddGradFunc InplaceGroupedMatmulAddGradFuncObj;
+  ReciprocalGradFunc ReciprocalGradFuncObj;
+  InplaceFloorDivideGradFunc InplaceFloorDivideGradFuncObj;
+  ConvTranspose2DGradFunc ConvTranspose2DGradFuncObj;
+  MoeTokenPermuteGradGradFunc MoeTokenPermuteGradGradFuncObj;
+  SliceExtViewGradFunc SliceExtViewGradFuncObj;
+  EyeGradFunc EyeGradFuncObj;
+  SoftMarginLossGradFunc SoftMarginLossGradFuncObj;
+  RmsNormGradFunc RmsNormGradFuncObj;
+  TypeAsGradFunc TypeAsGradFuncObj;
+  AllGatherMatmulGradFunc AllGatherMatmulGradFuncObj;
+  InnerCommAllGatherGradFunc InnerCommAllGatherGradFuncObj;
+  InplaceRemainderTensorScalarGradFunc InplaceRemainderTensorScalarGradFuncObj;
+  SliceGradFunc SliceGradFuncObj;
+  ConvolutionGradGradFunc ConvolutionGradGradFuncObj;
   InplaceHardtanhGradFunc InplaceHardtanhGradFuncObj;
-  AddLayerNormV2GradFunc AddLayerNormV2GradFuncObj;
-  BitwiseAndTensorGradFunc BitwiseAndTensorGradFuncObj;
-  InplaceClampScalarGradFunc InplaceClampScalarGradFuncObj;
-  LayerNormGradExtGradFunc LayerNormGradExtGradFuncObj;
-  InplaceScatterAddGradFunc InplaceScatterAddGradFuncObj;
-  SmoothL1LossGradFunc SmoothL1LossGradFuncObj;
-  ReflectionPad3DGradGradFunc ReflectionPad3DGradGradFuncObj;
-  ClampTensorGradFunc ClampTensorGradFuncObj;
+  DistCommAllGatherIntoTensorGradFunc DistCommAllGatherIntoTensorGradFuncObj;
+  SoftplusExtGradFunc SoftplusExtGradFuncObj;
+  StdMeanGradFunc StdMeanGradFuncObj;
+  InplaceScatterValueReduceGradFunc InplaceScatterValueReduceGradFuncObj;
+  OuterGradFunc OuterGradFuncObj;
+  MaskedFillGradFunc MaskedFillGradFuncObj;
+  SoftShrinkGradFunc SoftShrinkGradFuncObj;
+  InnerCommAllReduceGradFunc InnerCommAllReduceGradFuncObj;
+  MoeDistributeDispatchGradFunc MoeDistributeDispatchGradFuncObj;
+  LogAddExpGradFunc LogAddExpGradFuncObj;
   MoeTokenUnpermuteGradGradFunc MoeTokenUnpermuteGradGradFuncObj;
-  LogSoftmaxExtGradFunc LogSoftmaxExtGradFuncObj;
-  CeilGradFunc CeilGradFuncObj;
-  InplaceMulGradFunc InplaceMulGradFuncObj;
-  GreaterGradFunc GreaterGradFuncObj;
-  ScatterGradFunc ScatterGradFuncObj;
-  ReciprocalGradFunc ReciprocalGradFuncObj;
+  InplaceMatmulAddGradFunc InplaceMatmulAddGradFuncObj;
+  IndexAddExtGradFunc IndexAddExtGradFuncObj;
+  SinGradFunc SinGradFuncObj;
+  BatchNormReduceGradGradFunc BatchNormReduceGradGradFuncObj;
+  AcoshExtGradFunc AcoshExtGradFuncObj;
+  ArgMinExtGradFunc ArgMinExtGradFuncObj;
+  MoeDistributeCombineGradFunc MoeDistributeCombineGradFuncObj;
+  ProdExtGradFunc ProdExtGradFuncObj;
+  CrossGradFunc CrossGradFuncObj;
+  BitwiseNotGradFunc BitwiseNotGradFuncObj;
+  BinaryCrossEntropyGradGradFunc BinaryCrossEntropyGradGradFuncObj;
+  EmbeddingGradFunc EmbeddingGradFuncObj;
+  MishExtGradFunc MishExtGradFuncObj;
   MulsGradFunc MulsGradFuncObj;
-  InplaceUniformGradFunc InplaceUniformGradFuncObj;
-  LayerNormExtGradFunc LayerNormExtGradFuncObj;
-  BaddbmmGradFunc BaddbmmGradFuncObj;
-  Conv2DPaddingGradFunc Conv2DPaddingGradFuncObj;
-  MultinomialExtGradFunc MultinomialExtGradFuncObj;
-  Conv2DExtGradFunc Conv2DExtGradFuncObj;
-  DenseGradFunc DenseGradFuncObj;
-  ConvolutionStrGradGradFunc ConvolutionStrGradGradFuncObj;
-  OnesGradFunc OnesGradFuncObj;
-  RepeatInterleaveGradGradFunc RepeatInterleaveGradGradFuncObj;
-  PowGradFunc PowGradFuncObj;
-  MinGradFunc MinGradFuncObj;
-  SmoothL1LossGradGradFunc SmoothL1LossGradGradFuncObj;
-  FlashAttentionScoreGradFunc FlashAttentionScoreGradFuncObj;
-  InplaceCopyGradFunc InplaceCopyGradFuncObj;
-  InplaceScatterValueGradFunc InplaceScatterValueGradFuncObj;
-  UnstackExtGradFunc UnstackExtGradFuncObj;
-  AvgPool3DGradExtGradFunc AvgPool3DGradExtGradFuncObj;
-  MaxPoolWithMaskGradFunc MaxPoolWithMaskGradFuncObj;
-  HSigmoidGradGradFunc HSigmoidGradGradFuncObj;
-  SilentCheckV2GradFunc SilentCheckV2GradFuncObj;
-  ErfcGradFunc ErfcGradFuncObj;
-  BinaryCrossEntropyGradFunc BinaryCrossEntropyGradFuncObj;
-  ReduceAllGradFunc ReduceAllGradFuncObj;
-  LogSumExpGradFunc LogSumExpGradFuncObj;
-  InplaceDivModGradFunc InplaceDivModGradFuncObj;
-  GeluGradExtGradFunc GeluGradExtGradFuncObj;
-  IndexFillScalarGradFunc IndexFillScalarGradFuncObj;
-  SubGradFunc SubGradFuncObj;
-  MaximumGradFunc MaximumGradFuncObj;
-  NLLLossGradFunc NLLLossGradFuncObj;
-  XLogYScalarOtherGradFunc XLogYScalarOtherGradFuncObj;
-  IdentityGradFunc IdentityGradFuncObj;
-  BitwiseXorScalarGradFunc BitwiseXorScalarGradFuncObj;
+  IsCloseGradFunc IsCloseGradFuncObj;
+  PagedAttentionGradFunc PagedAttentionGradFuncObj;
+  ReduceMinGradFunc ReduceMinGradFuncObj;
+  SoftShrinkGradGradFunc SoftShrinkGradGradFuncObj;
+  UpsampleTrilinear3DGradFunc UpsampleTrilinear3DGradFuncObj;
+  ErfGradFunc ErfGradFuncObj;
+  MaskedSelectGradFunc MaskedSelectGradFuncObj;
+  ReluGradGradFunc ReluGradGradFuncObj;
+  MvGradFunc MvGradFuncObj;
   LogSigmoidGradGradFunc LogSigmoidGradGradFuncObj;
-  ScatterAddExtGradFunc ScatterAddExtGradFuncObj;
-  AdaptiveAvgPool2DGradExtGradFunc AdaptiveAvgPool2DGradExtGradFuncObj;
-  NormalFloatTensorGradFunc NormalFloatTensorGradFuncObj;
-  BitwiseXorTensorGradFunc BitwiseXorTensorGradFuncObj;
-  GLUGradFunc GLUGradFuncObj;
-  UpsampleNearest1DGradFunc UpsampleNearest1DGradFuncObj;
-  DivModsGradFunc DivModsGradFuncObj;
-  FFNExtGradFunc FFNExtGradFuncObj;
-  AddGradFunc AddGradFuncObj;
-  CopyGradFunc CopyGradFuncObj;
-  AsinExtGradFunc AsinExtGradFuncObj;
-  ReplicationPad2DGradGradFunc ReplicationPad2DGradGradFuncObj;
-  NeScalarGradFunc NeScalarGradFuncObj;
-  GmmBackwardGradFunc GmmBackwardGradFuncObj;
-  GroupNormGradFunc GroupNormGradFuncObj;
-  VarMeanGradFunc VarMeanGradFuncObj;
-  ConvolutionGradGradFunc ConvolutionGradGradFuncObj;
-  MaxPoolGradWithIndicesGradFunc MaxPoolGradWithIndicesGradFuncObj;
-  InplaceReLUGradFunc InplaceReLUGradFuncObj;
-  LinSpaceExtGradFunc LinSpaceExtGradFuncObj;
+  LeakyReLUExtGradFunc LeakyReLUExtGradFuncObj;
+  SmoothL1LossGradFunc SmoothL1LossGradFuncObj;
+  BatchNormExtGradFunc BatchNormExtGradFuncObj;
   NewOnesGradFunc NewOnesGradFuncObj;
-  ArgMinWithValueGradFunc ArgMinWithValueGradFuncObj;
-  StdGradFunc StdGradFuncObj;
-  SinGradFunc SinGradFuncObj;
-  SortExtGradFunc SortExtGradFuncObj;
-  DropoutDoMaskExtGradFunc DropoutDoMaskExtGradFuncObj;
-  SoftmaxBackwardGradFunc SoftmaxBackwardGradFuncObj;
-  TypeAsGradFunc TypeAsGradFuncObj;
-  GcdGradFunc GcdGradFuncObj;
-  AvgPool1DGradFunc AvgPool1DGradFuncObj;
-  SliceGradFunc SliceGradFuncObj;
-  FloorDivScalarGradFunc FloorDivScalarGradFuncObj;
-  MoeTokenPermuteGradFunc MoeTokenPermuteGradFuncObj;
-  MatMulExtGradFunc MatMulExtGradFuncObj;
-  InnerNonZeroGradFunc InnerNonZeroGradFuncObj;
-  ReflectionPad3DGradFunc ReflectionPad3DGradFuncObj;
-  RepeatInterleaveTensorGradFunc RepeatInterleaveTensorGradFuncObj;
   MedianDimGradFunc MedianDimGradFuncObj;
-  UpsampleNearest2DGradGradFunc UpsampleNearest2DGradGradFuncObj;
-  AdaptiveMaxPool2DGradFunc AdaptiveMaxPool2DGradFuncObj;
+  NLLLossGradFunc NLLLossGradFuncObj;
+  NotEqualGradFunc NotEqualGradFuncObj;
+  ReduceMaxGradFunc ReduceMaxGradFuncObj;
+  StackExtGradFunc StackExtGradFuncObj;
+  Conv1DExtGradFunc Conv1DExtGradFuncObj;
+  SumExtGradFunc SumExtGradFuncObj;
+  MatMulGradFunc MatMulGradFuncObj;
+  ReLUGradFunc ReLUGradFuncObj;
+  BitwiseOrTensorGradFunc BitwiseOrTensorGradFuncObj;
+  MedianExtGradFunc MedianExtGradFuncObj;
+  ReplicationPad2DGradFunc ReplicationPad2DGradFuncObj;
+  InnerCommIsendGradFunc InnerCommIsendGradFuncObj;
+  ReshapeGradFunc ReshapeGradFuncObj;
+  GluGradGradFunc GluGradGradFuncObj;
+  ThresholdGradGradFunc ThresholdGradGradFuncObj;
+  ScatterAddExtGradFunc ScatterAddExtGradFuncObj;
+  IndexFillTensorGradFunc IndexFillTensorGradFuncObj;
+  InplaceDivModGradFunc InplaceDivModGradFuncObj;
+  AtanExtGradFunc AtanExtGradFuncObj;
+  LayerNormExtGradFunc LayerNormExtGradFuncObj;
+  AdamWGradFunc AdamWGradFuncObj;
+  SubScalarGradFunc SubScalarGradFuncObj;
+  LogSumExpGradFunc LogSumExpGradFuncObj;
+  PReLUGradFunc PReLUGradFuncObj;
+  HShrinkGradFunc HShrinkGradFuncObj;
+  FracGradFunc FracGradFuncObj;
+  SubGradFunc SubGradFuncObj;
+  InplaceDivsGradFunc InplaceDivsGradFuncObj;
+  RandnGradFunc RandnGradFuncObj;
   InplaceZeroGradFunc InplaceZeroGradFuncObj;
-  ErfinvGradFunc ErfinvGradFuncObj;
-  InplaceLogGradFunc InplaceLogGradFuncObj;
-  HShrinkGradGradFunc HShrinkGradGradFuncObj;
-  KthvalueGradFunc KthvalueGradFuncObj;
-  NormalTensorTensorGradFunc NormalTensorTensorGradFuncObj;
+  EmptyLikeGradFunc EmptyLikeGradFuncObj;
+  LogSoftmaxExtGradFunc LogSoftmaxExtGradFuncObj;
+  FillTensorGradFunc FillTensorGradFuncObj;
+  LogSoftmaxGradFunc LogSoftmaxGradFuncObj;
+  GatherDGradV2GradFunc GatherDGradV2GradFuncObj;
+  Atan2ExtGradFunc Atan2ExtGradFuncObj;
+  ReflectionPad1DGradGradFunc ReflectionPad1DGradGradFuncObj;
+  SelectExtViewGradFunc SelectExtViewGradFuncObj;
+  SoftMarginLossGradGradFunc SoftMarginLossGradGradFuncObj;
+  KLDivGradGradFunc KLDivGradGradFuncObj;
+  NansumGradFunc NansumGradFuncObj;
+  BitwiseAndTensorGradFunc BitwiseAndTensorGradFuncObj;
+  GreaterEqualScalarGradFunc GreaterEqualScalarGradFuncObj;
+  NormalFloatFloatGradFunc NormalFloatFloatGradFuncObj;
+  FloorDivGradFunc FloorDivGradFuncObj;
+  TruncGradFunc TruncGradFuncObj;
+  LayerNormGradExtGradFunc LayerNormGradExtGradFuncObj;
+  HardtanhGradGradFunc HardtanhGradGradFuncObj;
+  GridSampler2DGradFunc GridSampler2DGradFuncObj;
+  AddbmmGradFunc AddbmmGradFuncObj;
+  ChunkViewGradFunc ChunkViewGradFuncObj;
+  UpsampleNearest1DGradGradFunc UpsampleNearest1DGradGradFuncObj;
+  UniqueConsecutiveGradFunc UniqueConsecutiveGradFuncObj;
+  AdaptiveAvgPool3DGradExtGradFunc AdaptiveAvgPool3DGradExtGradFuncObj;
+  MmGradFunc MmGradFuncObj;
+  AddmmGradFunc AddmmGradFuncObj;
+  RandnLikeGradFunc RandnLikeGradFuncObj;
+  BernoulliExtGradFunc BernoulliExtGradFuncObj;
+  ReflectionPad2DGradFunc ReflectionPad2DGradFuncObj;
+  SpeedFusionAttentionGradGradFunc SpeedFusionAttentionGradGradFuncObj;
+  InnerInplaceIndexPutGradFunc InnerInplaceIndexPutGradFuncObj;
+  InplaceLogGradFunc InplaceLogGradFuncObj;
+  AvgPool2DGradFunc AvgPool2DGradFuncObj;
+  FmodScalarGradFunc FmodScalarGradFuncObj;
+  DistCommScatterTensorGradFunc DistCommScatterTensorGradFuncObj;
+  AvgPool1DGradFunc AvgPool1DGradFuncObj;
+  MoeTokenPermuteGradFunc MoeTokenPermuteGradFuncObj;
   ReplicationPad1DGradFunc ReplicationPad1DGradFuncObj;
+  Col2ImGradGradFunc Col2ImGradGradFuncObj;
+  CastGradFunc CastGradFuncObj;
+  RotaryPositionEmbeddingGradGradFunc RotaryPositionEmbeddingGradGradFuncObj;
+  ReplicationPad3DGradGradFunc ReplicationPad3DGradGradFuncObj;
+  GcdGradFunc GcdGradFuncObj;
   BatchMatMulGradFunc BatchMatMulGradFuncObj;
-  AddmmGradFunc AddmmGradFuncObj;
+  ReplicationPad2DGradGradFunc ReplicationPad2DGradGradFuncObj;
+  Im2ColExtGradFunc Im2ColExtGradFuncObj;
+  RotaryPositionEmbeddingGradFunc RotaryPositionEmbeddingGradFuncObj;
+  RollGradFunc RollGradFuncObj;
+  ThresholdGradFunc ThresholdGradFuncObj;
+  MeanExtGradFunc MeanExtGradFuncObj;
+  InplaceThresholdGradFunc InplaceThresholdGradFuncObj;
+  Conv1DPaddingGradFunc Conv1DPaddingGradFuncObj;
+  LogicalXorGradFunc LogicalXorGradFuncObj;
+  MSELossGradExtGradFunc MSELossGradExtGradFuncObj;
+  LogSigmoidGradFunc LogSigmoidGradFuncObj;
+  NormalFloatTensorGradFunc NormalFloatTensorGradFuncObj;
+  CumminExtGradFunc CumminExtGradFuncObj;
+  FlashAttentionScoreGradGradFunc FlashAttentionScoreGradGradFuncObj;
+  InplaceFloorGradFunc InplaceFloorGradFuncObj;
+  HShrinkGradGradFunc HShrinkGradGradFuncObj;
+  AsinhExtGradFunc AsinhExtGradFuncObj;
+  RepeatInterleaveIntGradFunc RepeatInterleaveIntGradFuncObj;
+  MulGradFunc MulGradFuncObj;
+  ArangeGradFunc ArangeGradFuncObj;
+  FullLikeGradFunc FullLikeGradFuncObj;
+  RandExtGradFunc RandExtGradFuncObj;
+  XlogyGradFunc XlogyGradFuncObj;
+  IsNegInfGradFunc IsNegInfGradFuncObj;
+  BroadcastToViewGradFunc BroadcastToViewGradFuncObj;
+  DistCommGatherGradFunc DistCommGatherGradFuncObj;
+  DistCommIsendGradFunc DistCommIsendGradFuncObj;
+  BaddbmmGradFunc BaddbmmGradFuncObj;
   SelectV2GradFunc SelectV2GradFuncObj;
-  SoftplusGradExtGradFunc SoftplusGradExtGradFuncObj;
-  BatchNormExtGradFunc BatchNormExtGradFuncObj;
-  GroupNormGradGradFunc GroupNormGradGradFuncObj;
+  RandIntGradFunc RandIntGradFuncObj;
+  Exp2GradFunc Exp2GradFuncObj;
+  ArgSortGradFunc ArgSortGradFuncObj;
+  InplaceFloorDividesGradFunc InplaceFloorDividesGradFuncObj;
+  NLLLossGradGradFunc NLLLossGradGradFuncObj;
+  AtanhGradFunc AtanhGradFuncObj;
+  InplaceSubScalarGradFunc InplaceSubScalarGradFuncObj;
+  AvgPool2DGradGradFunc AvgPool2DGradGradFuncObj;
+  SincGradFunc SincGradFuncObj;
+  MaxPoolGradWithMaskGradFunc MaxPoolGradWithMaskGradFuncObj;
+  InplaceDivModsGradFunc InplaceDivModsGradFuncObj;
+  MaxUnpool2DExtGradFunc MaxUnpool2DExtGradFuncObj;
+  DivModsGradFunc DivModsGradFuncObj;
+  InplaceScatterSrcGradFunc InplaceScatterSrcGradFuncObj;
+  InnerCommAllToAllVGradFunc InnerCommAllToAllVGradFuncObj;
+  TanhGradGradFunc TanhGradGradFuncObj;
+  MaxPoolWithIndicesGradFunc MaxPoolWithIndicesGradFuncObj;
+  DistCommIrecvGradFunc DistCommIrecvGradFuncObj;
+  KLDivGradFunc KLDivGradFuncObj;
+  GeluExtGradFunc GeluExtGradFuncObj;
+  PReLUGradGradFunc PReLUGradGradFuncObj;
+  GLUGradFunc GLUGradFuncObj;
+  ViewAsGradFunc ViewAsGradFuncObj;
+  CellBackwardHookGradFunc CellBackwardHookGradFuncObj;
+  AsinExtGradFunc AsinExtGradFuncObj;
+  AddLayerNormGradGradFunc AddLayerNormGradGradFuncObj;
+  ExpandAsGradFunc ExpandAsGradFuncObj;
+  RandLikeExtGradFunc RandLikeExtGradFuncObj;
+  NLLLoss2dGradGradFunc NLLLoss2dGradGradFuncObj;
+  SignGradFunc SignGradFuncObj;
+  VarGradFunc VarGradFuncObj;
+  ConvolutionStrGradFunc ConvolutionStrGradFuncObj;
+  IndexSelectGradFunc IndexSelectGradFuncObj;
+  InnerNonZeroGradFunc InnerNonZeroGradFuncObj;
+  InplaceFillScalarGradFunc InplaceFillScalarGradFuncObj;
+  CoshGradFunc CoshGradFuncObj;
+  L1LossExtGradFunc L1LossExtGradFuncObj;
+  PromptFlashAttentionGradFunc PromptFlashAttentionGradFuncObj;
+  InplaceMaskedFillScalarGradFunc InplaceMaskedFillScalarGradFuncObj;
+  MaxPoolGradWithIndicesGradFunc MaxPoolGradWithIndicesGradFuncObj;
+  Conv3DExtGradFunc Conv3DExtGradFuncObj;
+  RemainderScalarTensorGradFunc RemainderScalarTensorGradFuncObj;
+  TransposeExtViewGradFunc TransposeExtViewGradFuncObj;
+  MinimumGradFunc MinimumGradFuncObj;
+  DropoutDoMaskExtGradFunc DropoutDoMaskExtGradFuncObj;
+  PowGradFunc PowGradFuncObj;
   RandIntLikeGradFunc RandIntLikeGradFuncObj;
-  MinDimGradFunc MinDimGradFuncObj;
-  ContiguousGradFunc ContiguousGradFuncObj;
-  ArgMaxWithValueGradFunc ArgMaxWithValueGradFuncObj;
-  MaskedSelectGradGradFunc MaskedSelectGradGradFuncObj;
-  SpeedFusionAttentionGradGradFunc SpeedFusionAttentionGradGradFuncObj;
-  LogicalOrGradFunc LogicalOrGradFuncObj;
-  RemainderTensorTensorGradFunc RemainderTensorTensorGradFuncObj;
+  UpsampleBicubic2DGradFunc UpsampleBicubic2DGradFuncObj;
+  SwigluGradFunc SwigluGradFuncObj;
+  DivModGradFunc DivModGradFuncObj;
+  LinalgQrGradFunc LinalgQrGradFuncObj;
+  CummaxGradFunc CummaxGradFuncObj;
+  IsFiniteGradFunc IsFiniteGradFuncObj;
+  LogGradFunc LogGradFuncObj;
+  SpeedFusionAttentionGradFunc SpeedFusionAttentionGradFuncObj;
   AdaptiveAvgPool3DExtGradFunc AdaptiveAvgPool3DExtGradFuncObj;
-  RandpermExtGradFunc RandpermExtGradFuncObj;
-  GeLUGradGradFunc GeLUGradGradFuncObj;
-  LogicalAndGradFunc LogicalAndGradFuncObj;
-  CastGradFunc CastGradFuncObj;
-  MishExtGradFunc MishExtGradFuncObj;
-  BCEWithLogitsLossGradFunc BCEWithLogitsLossGradFuncObj;
-  PolarGradFunc PolarGradFuncObj;
-  ReluGradGradFunc ReluGradGradFuncObj;
-  NormGradFunc NormGradFuncObj;
-  RemainderScalarTensorGradFunc RemainderScalarTensorGradFuncObj;
-  AvgPool2DGradFunc AvgPool2DGradFuncObj;
+  TriangularSolveGradFunc TriangularSolveGradFuncObj;
+  RsqrtGradFunc RsqrtGradFuncObj;
+  IndexFillScalarGradFunc IndexFillScalarGradFuncObj;
+  MultiScaleDeformableAttnGradGradFunc MultiScaleDeformableAttnGradGradFuncObj;
+  AddLayerNormV2GradFunc AddLayerNormV2GradFuncObj;
+  MaskedSelectGradGradFunc MaskedSelectGradGradFuncObj;
+  HistcExtGradFunc HistcExtGradFuncObj;
+  SqueezeGradFunc SqueezeGradFuncObj;
+  DistCommAllToAllVSingleGradFunc DistCommAllToAllVSingleGradFuncObj;
+  SilentCheckV2GradFunc SilentCheckV2GradFuncObj;
+  InplaceScatterSrcReduceGradFunc InplaceScatterSrcReduceGradFuncObj;
+  RmsNormGradGradFunc RmsNormGradGradFuncObj;
   XLogYScalarSelfGradFunc XLogYScalarSelfGradFuncObj;
+  LeakyReLUGradExtGradFunc LeakyReLUGradExtGradFuncObj;
+  LogAddExp2GradFunc LogAddExp2GradFuncObj;
+  ClampTensorGradFunc ClampTensorGradFuncObj;
+  SoftplusGradExtGradFunc SoftplusGradExtGradFuncObj;
+  UpsampleNearest2DGradGradFunc UpsampleNearest2DGradGradFuncObj;
+  AddScalarGradFunc AddScalarGradFuncObj;
+  NanToNumGradFunc NanToNumGradFuncObj;
+  ReflectionPad3DGradGradFunc ReflectionPad3DGradGradFuncObj;
+  InplaceAddsExtGradFunc InplaceAddsExtGradFuncObj;
+  ReflectionPad3DGradFunc ReflectionPad3DGradFuncObj;
+  DiagonalViewGradFunc DiagonalViewGradFuncObj;
+  SplitGradFunc SplitGradFuncObj;
+  AsStridedGradFunc AsStridedGradFuncObj;
+  TExtGradFunc TExtGradFuncObj;
+  InplaceSubExtGradFunc InplaceSubExtGradFuncObj;
+  BinaryCrossEntropyGradFunc BinaryCrossEntropyGradFuncObj;
   DropoutGradExtGradFunc DropoutGradExtGradFuncObj;
-  ConvTranspose2DGradFunc ConvTranspose2DGradFuncObj;
-  HShrinkGradFunc HShrinkGradFuncObj;
-  L1LossBackwardExtGradFunc L1LossBackwardExtGradFuncObj;
-  BitwiseOrScalarGradFunc BitwiseOrScalarGradFuncObj;
-  ExpGradFunc ExpGradFuncObj;
-  BatchNormGatherStatsWithCountsGradFunc BatchNormGatherStatsWithCountsGradFuncObj;
-  ZerosGradFunc ZerosGradFuncObj;
-  EqualExtGradFunc EqualExtGradFuncObj;
-  CloneGradFunc CloneGradFuncObj;
-  MaxPoolWithIndicesGradFunc MaxPoolWithIndicesGradFuncObj;
-  AdaptiveAvgPool2DExtGradFunc AdaptiveAvgPool2DExtGradFuncObj;
-  Conv3DPaddingGradFunc Conv3DPaddingGradFuncObj;
-  ChunkGradFunc ChunkGradFuncObj;
+  BinaryCrossEntropyWithLogitsBackwardGradFunc BinaryCrossEntropyWithLogitsBackwardGradFuncObj;
+  InplaceBernoulliTensorGradFunc InplaceBernoulliTensorGradFuncObj;
+  OnesGradFunc OnesGradFuncObj;
+  AvgPool3DGradExtGradFunc AvgPool3DGradExtGradFuncObj;
+  NLLLoss2dGradFunc NLLLoss2dGradFuncObj;
+  FFNExtGradFunc FFNExtGradFuncObj;
+  SoftmaxGradFunc SoftmaxGradFuncObj;
+  InplaceRandomGradFunc InplaceRandomGradFuncObj;
+  Col2ImExtGradFunc Col2ImExtGradFuncObj;
+  DotGradFunc DotGradFuncObj;
+  Expm1GradFunc Expm1GradFuncObj;
+  LerpGradFunc LerpGradFuncObj;
+  UnstackExtViewGradFunc UnstackExtViewGradFuncObj;
+  InplaceMulsGradFunc InplaceMulsGradFuncObj;
+  HSwishGradGradFunc HSwishGradGradFuncObj;
+  LessGradFunc LessGradFuncObj;
+  SliceExtGradFunc SliceExtGradFuncObj;
+  InplaceExpGradFunc InplaceExpGradFuncObj;
+  AcosExtGradFunc AcosExtGradFuncObj;
+  RandpermExtGradFunc RandpermExtGradFuncObj;
+  CrossEntropyLossGradFunc CrossEntropyLossGradFuncObj;
+  TileGradFunc TileGradFuncObj;
+  TanhGradFunc TanhGradFuncObj;
+  CeilGradFunc CeilGradFuncObj;
+  MinDimGradFunc MinDimGradFuncObj;
+  GreaterEqualGradFunc GreaterEqualGradFuncObj;
+  MlaGradFunc MlaGradFuncObj;
   AbsGradFunc AbsGradFuncObj;
-  CumsumExtGradFunc CumsumExtGradFuncObj;
-  TraceExtGradFunc TraceExtGradFuncObj;
-  SignGradFunc SignGradFuncObj;
-  InplaceClampTensorGradFunc InplaceClampTensorGradFuncObj;
-  Atan2ExtGradFunc Atan2ExtGradFuncObj;
-  InplaceFillTensorGradFunc InplaceFillTensorGradFuncObj;
-  ArgMinExtGradFunc ArgMinExtGradFuncObj;
+  CrossEntropyLossGradGradFunc CrossEntropyLossGradGradFuncObj;
+  EluGradExtGradFunc EluGradExtGradFuncObj;
+  SwigluGradGradFunc SwigluGradGradFuncObj;
   OnesLikeExtGradFunc OnesLikeExtGradFuncObj;
-  AcoshExtGradFunc AcoshExtGradFuncObj;
-  MedianExtGradFunc MedianExtGradFuncObj;
-  InplaceFloorDivideGradFunc InplaceFloorDivideGradFuncObj;
-  ReplicationPad2DGradFunc ReplicationPad2DGradFuncObj;
-  ReduceAnyGradFunc ReduceAnyGradFuncObj;
-  BincountExtGradFunc BincountExtGradFuncObj;
-  RepeatGradFunc RepeatGradFuncObj;
+  BroadcastToGradFunc BroadcastToGradFuncObj;
+  ConcatGradFunc ConcatGradFuncObj;
+  InplaceIndexPutGradFunc InplaceIndexPutGradFuncObj;
+  SearchSortedGradFunc SearchSortedGradFuncObj;
+  GroupNormGradFunc GroupNormGradFuncObj;
+  UniformExtGradFunc UniformExtGradFuncObj;
+  RepeatInterleaveGradGradFunc RepeatInterleaveGradGradFuncObj;
+  GridSampler3DGradGradFunc GridSampler3DGradGradFuncObj;
+  OneHotExtGradFunc OneHotExtGradFuncObj;
+  SinhGradFunc SinhGradFuncObj;
+  KthvalueGradFunc KthvalueGradFuncObj;
+  UpsampleLinear1DGradGradFunc UpsampleLinear1DGradGradFuncObj;
+  BatchNormElemtGradGradFunc BatchNormElemtGradGradFuncObj;
+  SeLUExtGradFunc SeLUExtGradFuncObj;
+  InplaceNormalGradFunc InplaceNormalGradFuncObj;
+  LogicalAndGradFunc LogicalAndGradFuncObj;
+  InplaceUniformGradFunc InplaceUniformGradFuncObj;
+  DiagExtGradFunc DiagExtGradFuncObj;
   RemainderTensorScalarGradFunc RemainderTensorScalarGradFuncObj;
-  TanhGradFunc TanhGradFuncObj;
-  TrilExtGradFunc TrilExtGradFuncObj;
-  KLDivGradFunc KLDivGradFuncObj;
-  MeshgridGradFunc MeshgridGradFuncObj;
-  AddmvGradFunc AddmvGradFuncObj;
-  NansumGradFunc NansumGradFuncObj;
-  HSigmoidGradFunc HSigmoidGradFuncObj;
-  ThresholdGradFunc ThresholdGradFuncObj;
-  SqrtGradFunc SqrtGradFuncObj;
-  RotaryPositionEmbeddingGradFunc RotaryPositionEmbeddingGradFuncObj;
-  UpsampleTrilinear3DGradGradFunc UpsampleTrilinear3DGradGradFuncObj;
-  SliceExtGradFunc SliceExtGradFuncObj;
+  Conv3DPaddingGradFunc Conv3DPaddingGradFuncObj;
+  GridSampler2DGradGradFunc GridSampler2DGradGradFuncObj;
+  ReplicationPad3DGradFunc ReplicationPad3DGradFuncObj;
+  AddcdivExtGradFunc AddcdivExtGradFuncObj;
+  InplaceCopyGradFunc InplaceCopyGradFuncObj;
+  ReshapeAndCacheGradFunc ReshapeAndCacheGradFuncObj;
+  AdaptiveMaxPool1DGradFunc AdaptiveMaxPool1DGradFuncObj;
+  PolarGradFunc PolarGradFuncObj;
+  NormGradFunc NormGradFuncObj;
+  SeluGradGradFunc SeluGradGradFuncObj;
+  SubExtGradFunc SubExtGradFuncObj;
+  CosGradFunc CosGradFuncObj;
+  IdentityGradFunc IdentityGradFuncObj;
+  ArgMaxExtGradFunc ArgMaxExtGradFuncObj;
+  MultinomialExtGradFunc MultinomialExtGradFuncObj;
+  RemainderTensorTensorGradFunc RemainderTensorTensorGradFuncObj;
+  ReduceAnyGradFunc ReduceAnyGradFuncObj;
+  FlashAttentionScoreGradFunc FlashAttentionScoreGradFuncObj;
+  ChunkGradFunc ChunkGradFuncObj;
+  MaxPoolWithMaskGradFunc MaxPoolWithMaskGradFuncObj;
+  Conv2DExtGradFunc Conv2DExtGradFuncObj;
   SplitWithSizeGradFunc SplitWithSizeGradFuncObj;
-  InplaceAddsExtGradFunc InplaceAddsExtGradFuncObj;
-  Expm1GradFunc Expm1GradFuncObj;
-  InplaceDivsGradFunc InplaceDivsGradFuncObj;
+  SelectGradFunc SelectGradFuncObj;
+  NormalTensorFloatGradFunc NormalTensorFloatGradFuncObj;
+  UpsampleTrilinear3DGradGradFunc UpsampleTrilinear3DGradGradFuncObj;
+  MatrixInverseExtGradFunc MatrixInverseExtGradFuncObj;
+  BatchNormStatsGradFunc BatchNormStatsGradFuncObj;
+  MishGradExtGradFunc MishGradExtGradFuncObj;
+  InplaceSiLUGradFunc InplaceSiLUGradFuncObj;
+  InplaceStopGradientGradFunc InplaceStopGradientGradFuncObj;
+  ArgMinWithValueGradFunc ArgMinWithValueGradFuncObj;
+  InplaceScatterValueGradFunc InplaceScatterValueGradFuncObj;
+  ExpandDimsGradFunc ExpandDimsGradFuncObj;
+  AvgPool3DExtGradFunc AvgPool3DExtGradFuncObj;
+  AddmvGradFunc AddmvGradFuncObj;
+  SplitTensorViewGradFunc SplitTensorViewGradFuncObj;
+  RepeatInterleaveTensorGradFunc RepeatInterleaveTensorGradFuncObj;
+  GeluGradExtGradFunc GeluGradExtGradFuncObj;
+  ReduceAllGradFunc ReduceAllGradFuncObj;
+  DropoutExtGradFunc DropoutExtGradFuncObj;
+  UpsampleNearest2DGradFunc UpsampleNearest2DGradFuncObj;
+  UpsampleBilinear2DGradGradFunc UpsampleBilinear2DGradGradFuncObj;
+  TriuGradFunc TriuGradFuncObj;
+  MatMulExtGradFunc MatMulExtGradFuncObj;
+  XLogYScalarOtherGradFunc XLogYScalarOtherGradFuncObj;
+  UpsampleNearest3DGradFunc UpsampleNearest3DGradFuncObj;
+  InnerCommReduceScatterGradFunc InnerCommReduceScatterGradFuncObj;
   NonZeroGradFunc NonZeroGradFuncObj;
-  InplaceAddExtGradFunc InplaceAddExtGradFuncObj;
-  RoundGradFunc RoundGradFuncObj;
-  AdaptiveAvgPool3DGradExtGradFunc AdaptiveAvgPool3DGradExtGradFuncObj;
-  InplaceMulsGradFunc InplaceMulsGradFuncObj;
-  ReLUGradFunc ReLUGradFuncObj;
-  LerpScalarGradFunc LerpScalarGradFuncObj;
-  CoshGradFunc CoshGradFuncObj;
   IsInfGradFunc IsInfGradFuncObj;
-  MvGradFunc MvGradFuncObj;
-  InnerIndexGradFunc InnerIndexGradFuncObj;
-  Log10GradFunc Log10GradFuncObj;
-  DivsGradFunc DivsGradFuncObj;
-  IndexFillTensorGradFunc IndexFillTensorGradFuncObj;
-  GatherDGradFunc GatherDGradFuncObj;
-  AvgPool2DGradGradFunc AvgPool2DGradGradFuncObj;
-  Im2ColExtGradFunc Im2ColExtGradFuncObj;
-  NanToNumGradFunc NanToNumGradFuncObj;
-  LinalgVectorNormGradFunc LinalgVectorNormGradFuncObj;
-  SigmoidGradGradFunc SigmoidGradGradFuncObj;
-  ProdExtGradFunc ProdExtGradFuncObj;
-  AddScalarGradFunc AddScalarGradFuncObj;
-  NLLLoss2dGradGradFunc NLLLoss2dGradGradFuncObj;
-  Log1pGradFunc Log1pGradFuncObj;
-  Conv3DExtGradFunc Conv3DExtGradFuncObj;
-  SilentCheckV3GradFunc SilentCheckV3GradFuncObj;
-  MultiScaleDeformableAttnGradGradFunc MultiScaleDeformableAttnGradGradFuncObj;
-  AddExtGradFunc AddExtGradFuncObj;
-  InplaceFloorGradFunc InplaceFloorGradFuncObj;
-  InplaceTanhGradFunc InplaceTanhGradFuncObj;
-  AsinhExtGradFunc AsinhExtGradFuncObj;
-  ConvolutionGradFunc ConvolutionGradFuncObj;
-  UpsampleBicubic2DGradFunc UpsampleBicubic2DGradFuncObj;
-  EmbeddingGradFunc EmbeddingGradFuncObj;
-  BinaryCrossEntropyGradGradFunc BinaryCrossEntropyGradGradFuncObj;
-  AddbmmGradFunc AddbmmGradFuncObj;
-  MoeTokenUnpermuteGradFunc MoeTokenUnpermuteGradFuncObj;
-  HardtanhGradGradFunc HardtanhGradGradFuncObj;
-  AcosExtGradFunc AcosExtGradFuncObj;
-  SwigluGradGradFunc SwigluGradGradFuncObj;
-  MultiScaleDeformableAttnGradFunc MultiScaleDeformableAttnGradFuncObj;
-  MinimumGradFunc MinimumGradFuncObj;
-  HistcExtGradFunc HistcExtGradFuncObj;
-  MmGradFunc MmGradFuncObj;
-  ReflectionPad1DGradGradFunc ReflectionPad1DGradGradFuncObj;
-  GridSampler2DGradFunc GridSampler2DGradFuncObj;
-  BitwiseNotGradFunc BitwiseNotGradFuncObj;
-  UpsampleLinear1DGradFunc UpsampleLinear1DGradFuncObj;
-  RotaryPositionEmbeddingGradGradFunc RotaryPositionEmbeddingGradGradFuncObj;
+  MSELossExtGradFunc MSELossExtGradFuncObj;
+  NegGradFunc NegGradFuncObj;
   Unique2GradFunc Unique2GradFuncObj;
-  SpeedFusionAttentionGradFunc SpeedFusionAttentionGradFuncObj;
-  ClampScalarGradFunc ClampScalarGradFuncObj;
-  Col2ImExtGradFunc Col2ImExtGradFuncObj;
+  AddGradFunc AddGradFuncObj;
+  L1LossBackwardExtGradFunc L1LossBackwardExtGradFuncObj;
+  NonZeroExtGradFunc NonZeroExtGradFuncObj;
+  EqualGradFunc EqualGradFuncObj;
+  GeneratorGradFunc GeneratorGradFuncObj;
+  GeLUGradGradFunc GeLUGradGradFuncObj;
+  AdaptiveAvgPool2DExtGradFunc AdaptiveAvgPool2DExtGradFuncObj;
+  SplitWithSizeViewGradFunc SplitWithSizeViewGradFuncObj;
+  InplaceClampScalarGradFunc InplaceClampScalarGradFuncObj;
+  SortExtGradFunc SortExtGradFuncObj;
+  ZerosGradFunc ZerosGradFuncObj;
+  EmbeddingDenseBackwardGradFunc EmbeddingDenseBackwardGradFuncObj;
+  RingAttentionUpdateGradFunc RingAttentionUpdateGradFuncObj;
+  GeLUGradFunc GeLUGradFuncObj;
+  LinSpaceExtGradFunc LinSpaceExtGradFuncObj;
+  InplaceFillTensorGradFunc InplaceFillTensorGradFuncObj;
+  InplaceAddmmGradFunc InplaceAddmmGradFuncObj;
   GridSampler3DGradFunc GridSampler3DGradFuncObj;
-  LogAddExpGradFunc LogAddExpGradFuncObj;
-  InplaceFloorDividesGradFunc InplaceFloorDividesGradFuncObj;
-  BatchNormReduceGradGradFunc BatchNormReduceGradGradFuncObj;
-  ViewGradFunc ViewGradFuncObj;
-  BatchMatMulExtGradFunc BatchMatMulExtGradFuncObj;
+  DistCommReduceScatterGradFunc DistCommReduceScatterGradFuncObj;
+  GroupNormGradGradFunc GroupNormGradGradFuncObj;
+  SiLUGradGradFunc SiLUGradGradFuncObj;
+  DistCommGatherIntoTensorGradFunc DistCommGatherIntoTensorGradFuncObj;
+  Log10GradFunc Log10GradFuncObj;
+  SplitTensorGradFunc SplitTensorGradFuncObj;
+  InplaceMaskedFillTensorGradFunc InplaceMaskedFillTensorGradFuncObj;
+  ExpGradFunc ExpGradFuncObj;
+  NarrowViewGradFunc NarrowViewGradFuncObj;
+  HSwishGradFunc HSwishGradFuncObj;
   InplaceErfinvGradFunc InplaceErfinvGradFuncObj;
-  MaxPoolGradWithMaskGradFunc MaxPoolGradWithMaskGradFuncObj;
-  TileGradFunc TileGradFuncObj;
-  SoftShrinkGradFunc SoftShrinkGradFuncObj;
-  ReshapeGradFunc ReshapeGradFuncObj;
-  BatchNormElemtGradFunc BatchNormElemtGradFuncObj;
-  LinalgQrGradFunc LinalgQrGradFuncObj;
-  SigmoidGradFunc SigmoidGradFuncObj;
-  L1LossExtGradFunc L1LossExtGradFuncObj;
-  ViewAsGradFunc ViewAsGradFuncObj;
+  DistCommAllToAllVGradFunc DistCommAllToAllVGradFuncObj;
+  VarMeanGradFunc VarMeanGradFuncObj;
+  ErfinvGradFunc ErfinvGradFuncObj;
   TransposeGradFunc TransposeGradFuncObj;
-  UpsampleNearest2DGradFunc UpsampleNearest2DGradFuncObj;
-  CustomExtGradFunc CustomExtGradFuncObj;
-  DropoutGenMaskExtGradFunc DropoutGenMaskExtGradFuncObj;
-  InplaceIndexAddExtGradFunc InplaceIndexAddExtGradFuncObj;
-  EluGradFunc EluGradFuncObj;
-  SinhGradFunc SinhGradFuncObj;
-  SelectGradFunc SelectGradFuncObj;
-  InplaceMaskedFillScalarGradFunc InplaceMaskedFillScalarGradFuncObj;
-  GeluExtGradFunc GeluExtGradFuncObj;
   BatchNormGradExtGradFunc BatchNormGradExtGradFuncObj;
-  NormalTensorFloatGradFunc NormalTensorFloatGradFuncObj;
-  RmsNormGradGradFunc RmsNormGradGradFuncObj;
-  SeluGradGradFunc SeluGradGradFuncObj;
-  NonZeroExtGradFunc NonZeroExtGradFuncObj;
-  SquareGradFunc SquareGradFuncObj;
-  ReduceMaxGradFunc ReduceMaxGradFuncObj;
-  BitwiseAndScalarGradFunc BitwiseAndScalarGradFuncObj;
-  InplaceScatterValueReduceGradFunc InplaceScatterValueReduceGradFuncObj;
-  ThresholdGradGradFunc ThresholdGradGradFuncObj;
-  UniqueDimGradFunc UniqueDimGradFuncObj;
-  HSwishGradGradFunc HSwishGradGradFuncObj;
-  LeakyReLUGradExtGradFunc LeakyReLUGradExtGradFuncObj;
-  CumminExtGradFunc CumminExtGradFuncObj;
-  IsCloseGradFunc IsCloseGradFuncObj;
-  TakeGradFunc TakeGradFuncObj;
+  MaxGradFunc MaxGradFuncObj;
+  PowTensorScalarGradFunc PowTensorScalarGradFuncObj;
+  LessEqualGradFunc LessEqualGradFuncObj;
+  DivsGradFunc DivsGradFuncObj;
+  HSigmoidGradGradFunc HSigmoidGradGradFuncObj;
+  ExpandDimsViewGradFunc ExpandDimsViewGradFuncObj;
+  BitwiseXorTensorGradFunc BitwiseXorTensorGradFuncObj;
+  TransposeViewGradFunc TransposeViewGradFuncObj;
+  TrilExtGradFunc TrilExtGradFuncObj;
+  ConvolutionGradFunc ConvolutionGradFuncObj;
+  FillScalarGradFunc FillScalarGradFuncObj;
+  ViewGradFunc ViewGradFuncObj;
+  TopkExtGradFunc TopkExtGradFuncObj;
+  BCEWithLogitsLossGradFunc BCEWithLogitsLossGradFuncObj;
   ReflectionPad1DGradFunc ReflectionPad1DGradFuncObj;
-  AddcdivExtGradFunc AddcdivExtGradFuncObj;
-  VarGradFunc VarGradFuncObj;
-  MulGradFunc MulGradFuncObj;
-  InplaceThresholdGradFunc InplaceThresholdGradFuncObj;
-  RmsNormGradFunc RmsNormGradFuncObj;
-  BinaryCrossEntropyWithLogitsBackwardGradFunc BinaryCrossEntropyWithLogitsBackwardGradFuncObj;
-  StdMeanGradFunc StdMeanGradFuncObj;
-  MoeTokenPermuteGradGradFunc MoeTokenPermuteGradGradFuncObj;
-  RandIntGradFunc RandIntGradFuncObj;
-  SoftplusExtGradFunc SoftplusExtGradFuncObj;
-  ReflectionPad2DGradFunc ReflectionPad2DGradFuncObj;
+  DistCommBarrierGradFunc DistCommBarrierGradFuncObj;
+  SmoothL1LossGradGradFunc SmoothL1LossGradGradFuncObj;
+  LogicalOrGradFunc LogicalOrGradFuncObj;
+  ReverseV2GradFunc ReverseV2GradFuncObj;
+  InnerMoeTokenUnpermuteGradFunc InnerMoeTokenUnpermuteGradFuncObj;
+  InplaceClampTensorGradFunc InplaceClampTensorGradFuncObj;
+  IndexGradFunc IndexGradFuncObj;
+  AddExtGradFunc AddExtGradFuncObj;
+  NeScalarGradFunc NeScalarGradFuncObj;
+  CopyGradFunc CopyGradFuncObj;
+  MeshgridGradFunc MeshgridGradFuncObj;
+  UpsampleBilinear2DGradFunc UpsampleBilinear2DGradFuncObj;
+  ClampScalarGradFunc ClampScalarGradFuncObj;
+  EluGradFunc EluGradFuncObj;
+  DistCommReduceGradFunc DistCommReduceGradFuncObj;
+  BatchMatMulExtGradFunc BatchMatMulExtGradFuncObj;
+  FloorGradFunc FloorGradFuncObj;
+  GatherDGradFunc GatherDGradFuncObj;
+  BatchNormElemtGradFunc BatchNormElemtGradFuncObj;
   PowScalarTensorGradFunc PowScalarTensorGradFuncObj;
-  InplaceScatterSrcReduceGradFunc InplaceScatterSrcReduceGradFuncObj;
-  EqualGradFunc EqualGradFuncObj;
-  RandnGradFunc RandnGradFuncObj;
-  RollGradFunc RollGradFuncObj;
-  GeLUGradFunc GeLUGradFuncObj;
-  AsStridedGradFunc AsStridedGradFuncObj;
-  TensorScatterElementsGradFunc TensorScatterElementsGradFuncObj;
-  FlashAttentionScoreGradGradFunc FlashAttentionScoreGradGradFuncObj;
-  AtanhGradFunc AtanhGradFuncObj;
-  InplaceStopGradientGradFunc InplaceStopGradientGradFuncObj;
-  LogicalNotGradFunc LogicalNotGradFuncObj;
-  DropoutExtGradFunc DropoutExtGradFuncObj;
-  NormalFloatFloatGradFunc NormalFloatFloatGradFuncObj;
-  Conv1DPaddingGradFunc Conv1DPaddingGradFuncObj;
-  RsqrtGradFunc RsqrtGradFuncObj;
-  ConstantPadNDGradFunc ConstantPadNDGradFuncObj;
-  BitwiseOrTensorGradFunc BitwiseOrTensorGradFuncObj;
-  MSELossGradExtGradFunc MSELossGradExtGradFuncObj;
-  MeanExtGradFunc MeanExtGradFuncObj;
-  LessEqualGradFunc LessEqualGradFuncObj;
-  PromptFlashAttentionGradFunc PromptFlashAttentionGradFuncObj;
-  SelectExtGradFunc SelectExtGradFuncObj;
+  Log2GradFunc Log2GradFuncObj;
+  InplaceAddExtGradFunc InplaceAddExtGradFuncObj;
+  EqualExtGradFunc EqualExtGradFuncObj;
+  Conv2DPaddingGradFunc Conv2DPaddingGradFuncObj;
   EluExtGradFunc EluExtGradFuncObj;
-  Col2ImGradGradFunc Col2ImGradGradFuncObj;
-  TriangularSolveGradFunc TriangularSolveGradFuncObj;
-  FillTensorGradFunc FillTensorGradFuncObj;
-  SubScalarGradFunc SubScalarGradFuncObj;
-  IndexAddExtGradFunc IndexAddExtGradFuncObj;
-  FillScalarGradFunc FillScalarGradFuncObj;
-  LeakyReLUExtGradFunc LeakyReLUExtGradFuncObj;
-  RandExtGradFunc RandExtGradFuncObj;
+  ScatterValueGradFunc ScatterValueGradFuncObj;
   DivGradFunc DivGradFuncObj;
-  TransposeExtGradFunc TransposeExtGradFuncObj;
-  InplaceFillScalarGradFunc InplaceFillScalarGradFuncObj;
-  GatherDGradV2GradFunc GatherDGradV2GradFuncObj;
-  PReLUGradFunc PReLUGradFuncObj;
-  InplaceDivGradFunc InplaceDivGradFuncObj;
-  InplaceIndexPutGradFunc InplaceIndexPutGradFuncObj;
-  KLDivGradGradFunc KLDivGradGradFuncObj;
-  UpsampleNearest1DGradGradFunc UpsampleNearest1DGradGradFuncObj;
-  UniformExtGradFunc UniformExtGradFuncObj;
-  IsFiniteGradFunc IsFiniteGradFuncObj;
-  LogGradFunc LogGradFuncObj;
-  BroadcastToGradFunc BroadcastToGradFuncObj;
-  SplitTensorGradFunc SplitTensorGradFuncObj;
-  ExpandDimsGradFunc ExpandDimsGradFuncObj;
-  FloorGradFunc FloorGradFuncObj;
-  LogSoftmaxGradGradFunc LogSoftmaxGradGradFuncObj;
-  LessGradFunc LessGradFuncObj;
-  ReplicationPad1DGradGradFunc ReplicationPad1DGradGradFuncObj;
-  AdamWGradFunc AdamWGradFuncObj;
-  InplaceRandomGradFunc InplaceRandomGradFuncObj;
-  SplitGradFunc SplitGradFuncObj;
-  AvgPool3DExtGradFunc AvgPool3DExtGradFuncObj;
-  NarrowGradFunc NarrowGradFuncObj;
-  SqueezeGradFunc SqueezeGradFuncObj;
-  LogAddExp2GradFunc LogAddExp2GradFuncObj;
-  NegGradFunc NegGradFuncObj;
-  InplaceFillDiagonalGradFunc InplaceFillDiagonalGradFuncObj;
-  InplaceGroupedMatmulAddGradFunc InplaceGroupedMatmulAddGradFuncObj;
-  BernoulliExtGradFunc BernoulliExtGradFuncObj;
+  InplaceMulGradFunc InplaceMulGradFuncObj;
+  GreaterGradFunc GreaterGradFuncObj;
+  AdaptiveMaxPool2DGradFunc AdaptiveMaxPool2DGradFuncObj;
+  AllFiniteGradFunc AllFiniteGradFuncObj;
+  InplaceReLUGradFunc InplaceReLUGradFuncObj;
   ReflectionPad2DGradGradFunc ReflectionPad2DGradGradFuncObj;
-  FullLikeGradFunc FullLikeGradFuncObj;
-  EyeGradFunc EyeGradFuncObj;
-  ConvolutionStrGradFunc ConvolutionStrGradFuncObj;
-  GridSampler2DGradGradFunc GridSampler2DGradGradFuncObj;
-  HSwishGradFunc HSwishGradFuncObj;
-  UpsampleLinear1DGradGradFunc UpsampleLinear1DGradGradFuncObj;
-  UpsampleBilinear2DGradGradFunc UpsampleBilinear2DGradGradFuncObj;
-  AllGatherMatmulGradFunc AllGatherMatmulGradFuncObj;
-  MaxDimGradFunc MaxDimGradFuncObj;
-  IncreFlashAttentionGradFunc IncreFlashAttentionGradFuncObj;
-  TriuGradFunc TriuGradFuncObj;
-  FloorDivGradFunc FloorDivGradFuncObj;
-  AddRmsNormGradFunc AddRmsNormGradFuncObj;
-  TanhGradGradFunc TanhGradGradFuncObj;
-  Exp2GradFunc Exp2GradFuncObj;
-  HardtanhGradFunc HardtanhGradFuncObj;
-  IndexSelectGradFunc IndexSelectGradFuncObj;
-  LogSigmoidGradFunc LogSigmoidGradFuncObj;
-  StackExtGradFunc StackExtGradFuncObj;
-  MaskedSelectGradFunc MaskedSelectGradFuncObj;
-  MishGradExtGradFunc MishGradExtGradFuncObj;
-  ReverseV2GradFunc ReverseV2GradFuncObj;
-  OneHotExtGradFunc OneHotExtGradFuncObj;
-  PReLUGradGradFunc PReLUGradGradFuncObj;
-  DotGradFunc DotGradFuncObj;
-  IsNegInfGradFunc IsNegInfGradFuncObj;
-  TruncGradFunc TruncGradFuncObj;
-  MatMulGradFunc MatMulGradFuncObj;
-  NLLLossGradGradFunc NLLLossGradGradFuncObj;
-  SwigluGradFunc SwigluGradFuncObj;
-  RepeatInterleaveIntGradFunc RepeatInterleaveIntGradFuncObj;
-  GreaterEqualScalarGradFunc GreaterEqualScalarGradFuncObj;
-  RandnLikeGradFunc RandnLikeGradFuncObj;
+  DistCommReduceScatterTensorGradFunc DistCommReduceScatterTensorGradFuncObj;
+  RoundGradFunc RoundGradFuncObj;
+  ApplyRotaryPosEmbGradFunc ApplyRotaryPosEmbGradFuncObj;
+  InplaceFillDiagonalGradFunc InplaceFillDiagonalGradFuncObj;
+  AdaptiveAvgPool2DGradExtGradFunc AdaptiveAvgPool2DGradExtGradFuncObj;
+  DistCommBatchIsendIrecvGradFunc DistCommBatchIsendIrecvGradFuncObj;
+  InplaceDivGradFunc InplaceDivGradFuncObj;
+  DistCommScatterGradFunc DistCommScatterGradFuncObj;
   ZerosLikeExtGradFunc ZerosLikeExtGradFuncObj;
-  NewZerosGradFunc NewZerosGradFuncObj;
-  ArgSortGradFunc ArgSortGradFuncObj;
-  SearchSortedGradFunc SearchSortedGradFuncObj;
-  SiLUGradFunc SiLUGradFuncObj;
-  InplaceSubScalarGradFunc InplaceSubScalarGradFuncObj;
-  InnerInplaceIndexPutGradFunc InnerInplaceIndexPutGradFuncObj;
-  SumExtGradFunc SumExtGradFuncObj;
-  CosGradFunc CosGradFuncObj;
-  SubExtGradFunc SubExtGradFuncObj;
-  ArangeGradFunc ArangeGradFuncObj;
-  InplaceScatterSrcGradFunc InplaceScatterSrcGradFuncObj;
-  AddcmulExtGradFunc AddcmulExtGradFuncObj;
-  CrossGradFunc CrossGradFuncObj;
-  AdaptiveMaxPool1DGradFunc AdaptiveMaxPool1DGradFuncObj;
   MatmulReduceScatterGradFunc MatmulReduceScatterGradFuncObj;
+  MinGradFunc MinGradFuncObj;
+  MultiScaleDeformableAttnGradFunc MultiScaleDeformableAttnGradFuncObj;
+  InplaceBernoulliScalarGradFunc InplaceBernoulliScalarGradFuncObj;
+  LinalgVectorNormGradFunc LinalgVectorNormGradFuncObj;
+  BitwiseXorScalarGradFunc BitwiseXorScalarGradFuncObj;
+  HSigmoidGradFunc HSigmoidGradFuncObj;
+  BincountExtGradFunc BincountExtGradFuncObj;
+  InnerIndexGradFunc InnerIndexGradFuncObj;
+  TensorScatterElementsGradFunc TensorScatterElementsGradFuncObj;
+  SiLUGradFunc SiLUGradFuncObj;
+  StdGradFunc StdGradFuncObj;
+  DistCommBroadcastGradFunc DistCommBroadcastGradFuncObj;
+  HardtanhGradFunc HardtanhGradFuncObj;
+  MaskedScatterGradFunc MaskedScatterGradFuncObj;
+  ConstantPadNDGradFunc ConstantPadNDGradFuncObj;
+  UniqueDimGradFunc UniqueDimGradFuncObj;
+  UpsampleNearest3DGradGradFunc UpsampleNearest3DGradGradFuncObj;
+  TraceExtGradFunc TraceExtGradFuncObj;
+  LogicalNotGradFunc LogicalNotGradFuncObj;
+  Log1pGradFunc Log1pGradFuncObj;
+  NewEmptyGradFunc NewEmptyGradFuncObj;
+  SigmoidGradGradFunc SigmoidGradGradFuncObj;
+  InplaceTanhGradFunc InplaceTanhGradFuncObj;
+  SqrtGradFunc SqrtGradFuncObj;
+  AddRmsNormGradFunc AddRmsNormGradFuncObj;
+  ScatterGradFunc ScatterGradFuncObj;
+  UpsampleLinear1DGradFunc UpsampleLinear1DGradFuncObj;
+  InplaceIndexAddExtGradFunc InplaceIndexAddExtGradFuncObj;
+  IncreFlashAttentionGradFunc IncreFlashAttentionGradFuncObj;
+  CloneGradFunc CloneGradFuncObj;
+  ArgMaxWithValueGradFunc ArgMaxWithValueGradFuncObj;
+  DropoutGenMaskExtGradFunc DropoutGenMaskExtGradFuncObj;
   TanGradFunc TanGradFuncObj;
-  RandLikeExtGradFunc RandLikeExtGradFuncObj;
-  BatchNormStatsGradFunc BatchNormStatsGradFuncObj;
-  PowTensorScalarGradFunc PowTensorScalarGradFuncObj;
-  InplaceAddmmGradFunc InplaceAddmmGradFuncObj;
-  SeLUExtGradFunc SeLUExtGradFuncObj;
-  UniqueConsecutiveGradFunc UniqueConsecutiveGradFuncObj;
-  FmodScalarGradFunc FmodScalarGradFuncObj;
-  AddLayerNormGradGradFunc AddLayerNormGradGradFuncObj;
-  EluGradExtGradFunc EluGradExtGradFuncObj;
-  GroupedMatmulV2GradFunc GroupedMatmulV2GradFuncObj;
-  KVCacheScatterUpdateGradFunc KVCacheScatterUpdateGradFuncObj;
-  FusedInferAttentionScoreGradFunc FusedInferAttentionScoreGradFuncObj;
-  DynamicQuantExtGradFunc DynamicQuantExtGradFuncObj;
-  QuantBatchMatmulGradFunc QuantBatchMatmulGradFuncObj;
-  MoeFinalizeRoutingGradFunc MoeFinalizeRoutingGradFuncObj;
-  GroupedMatmulV4GradFunc GroupedMatmulV4GradFuncObj;
-  MatmulAllReduceAddRmsNormGradFunc MatmulAllReduceAddRmsNormGradFuncObj;
+  DistCommReduceScatterTensorUnevenGradFunc DistCommReduceScatterTensorUnevenGradFuncObj;
+  InnerCommIrecvGradFunc InnerCommIrecvGradFuncObj;
+  ContiguousGradFunc ContiguousGradFuncObj;
   QuantV2GradFunc QuantV2GradFuncObj;
   AddRmsNormQuantV2GradFunc AddRmsNormQuantV2GradFuncObj;
-  MoeComputeExpertTokensGradFunc MoeComputeExpertTokensGradFuncObj;
+  KVCacheScatterUpdateGradFunc KVCacheScatterUpdateGradFuncObj;
+  QuantBatchMatmulGradFunc QuantBatchMatmulGradFuncObj;
+  FusedInferAttentionScoreGradFunc FusedInferAttentionScoreGradFuncObj;
+  MoeInitRoutingGradFunc MoeInitRoutingGradFuncObj;
   GroupedMatmulGradFunc GroupedMatmulGradFuncObj;
-  WeightQuantBatchMatmulGradFunc WeightQuantBatchMatmulGradFuncObj;
   MoeGatingTopKSoftmaxGradFunc MoeGatingTopKSoftmaxGradFuncObj;
+  MatmulAllReduceAddRmsNormGradFunc MatmulAllReduceAddRmsNormGradFuncObj;
+  WeightQuantBatchMatmulGradFunc WeightQuantBatchMatmulGradFuncObj;
+  MoeComputeExpertTokensGradFunc MoeComputeExpertTokensGradFuncObj;
+  MoeInitRoutingQuantV2GradFunc MoeInitRoutingQuantV2GradFuncObj;
+  DynamicQuantExtGradFunc DynamicQuantExtGradFuncObj;
+  GroupedMatmulV4GradFunc GroupedMatmulV4GradFuncObj;
   MoeInitRoutingV2GradFunc MoeInitRoutingV2GradFuncObj;
-  MoeInitRoutingGradFunc MoeInitRoutingGradFuncObj;
+  GroupedMatmulV2GradFunc GroupedMatmulV2GradFuncObj;
+  QuantMatmulGradFunc QuantMatmulGradFuncObj;
+  MoeFinalizeRoutingGradFunc MoeFinalizeRoutingGradFuncObj;
+  FuncDropoutExtGradFunc FuncDropoutExtGradFuncObj;
+  AnyGradFunc AnyGradFuncObj;
+  FuncMaxPool2DGradFunc FuncMaxPool2DGradFuncObj;
+  GmmV2GradFunc GmmV2GradFuncObj;
+  GmmGradFunc GmmGradFuncObj;
+  GmmBackwardGradFunc GmmBackwardGradFuncObj;
+  GmmV2BackwardGradFunc GmmV2BackwardGradFuncObj;
+  AnyExtGradFunc AnyExtGradFuncObj;
+  GmmV2BackwardFusionGradFunc GmmV2BackwardFusionGradFuncObj;
+  InplaceExponentialGradFunc InplaceExponentialGradFuncObj;
   PixelShuffleGradFunc PixelShuffleGradFuncObj;
+  EinsumExtGradFunc EinsumExtGradFuncObj;
+  GmmBackwardFusionGradFunc GmmBackwardFusionGradFuncObj;
+  MoeTokenUnpermuteGradFunc MoeTokenUnpermuteGradFuncObj;
 };
 }  // namespace pyboost
 }  // namespace kernel