mindspore 2.6.0rc1__cp39-cp39-manylinux1_x86_64.whl → 2.7.0__cp39-cp39-manylinux1_x86_64.whl

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

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