mindspore 2.6.0__cp310-cp310-manylinux1_x86_64.whl → 2.7.0rc1__cp310-cp310-manylinux1_x86_64.whl

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

@@ -55,963 +55,1047 @@ enum class OpType {
   kSplitTensorView = 26,
   kUpsampleNearest3DGrad = 27,
   kAddRmsNorm = 28,
-  kBatchNormReduceGrad = 29,
-  kInplaceScatterSrc = 30,
-  kUpsampleNearest2DGrad = 31,
-  kSplitWithSizeView = 32,
-  kKLDiv = 33,
-  kEluExt = 34,
-  kBaddbmm = 35,
-  kReduceAny = 36,
-  kLogSigmoid = 37,
-  kLeakyReLUGradExt = 38,
-  kKLDivGrad = 39,
-  kTriu = 40,
-  kMultiScaleDeformableAttnGrad = 41,
-  kLogSoftmax = 42,
-  kInplaceFloor = 43,
-  kInplaceStopGradient = 44,
-  kBroadcastToView = 45,
-  kIndexFillTensor = 46,
-  kSilentCheckV3 = 47,
-  kAddExt = 48,
-  kRsqrt = 49,
-  kViewAs = 50,
-  kErfinv = 51,
-  kTrilExt = 52,
-  kCumsumExt = 53,
-  kLogSigmoidGrad = 54,
-  kAvgPool3DGradExt = 55,
-  kBCEWithLogitsLoss = 56,
-  kCeil = 57,
-  kHardtanhGrad = 58,
-  kUnique2 = 59,
-  kHSwishGrad = 60,
-  kMedianExt = 61,
-  kReplicationPad1D = 62,
-  kCol2ImExt = 63,
-  kReLU = 64,
-  kAllGatherMatmul = 65,
-  kLerp = 66,
-  kAdaptiveAvgPool2DGradExt = 67,
-  kMatrixInverseExt = 68,
-  kRotaryPositionEmbeddingGrad = 69,
-  kEye = 70,
-  kMaxPoolGradWithMask = 71,
-  kBatchMatMulExt = 72,
-  kGatherDGradV2 = 73,
-  kNorm = 74,
-  kInplaceIndexAddExt = 75,
-  kInnerIndex = 76,
-  kGLU = 77,
-  kInplaceCopy = 78,
-  kReplicationPad3D = 79,
-  kLog10 = 80,
-  kSeLUExt = 81,
-  kInplaceClampScalar = 82,
-  kLog1p = 83,
-  kGridSampler3D = 84,
-  kFmodTensor = 85,
-  kArgMaxWithValue = 86,
-  kInplaceGroupedMatmulAdd = 87,
-  kLogAddExp = 88,
-  kMaskedSelectGrad = 89,
-  kBroadcastTo = 90,
-  kCopy = 91,
-  kSoftmax = 92,
-  kStdMean = 93,
-  kSoftShrink = 94,
-  kAddcmulExt = 95,
-  kSiLUGrad = 96,
-  kDiv = 97,
-  kMatmulReduceScatter = 98,
-  kSliceExt = 99,
-  kSelectV2 = 100,
-  kThresholdGrad = 101,
-  kDivs = 102,
-  kBinaryCrossEntropyWithLogitsBackward = 103,
-  kCos = 104,
-  kSliceExtView = 105,
-  kCross = 106,
-  kMSELossExt = 107,
-  kInplaceDivMods = 108,
-  kGridSampler2DGrad = 109,
-  kBatchNormElemtGrad = 110,
-  kBatchNormGradExt = 111,
-  kAbs = 112,
-  kConv3DExt = 113,
-  kInplaceThreshold = 114,
-  kBatchNormElemt = 115,
-  kLogSoftmaxGrad = 116,
-  kRemainderTensorScalar = 117,
-  kInplaceScatterSrcReduce = 118,
-  kRandn = 119,
-  kHShrink = 120,
-  kSortExt = 121,
-  kPReLU = 122,
-  kInplaceExp = 123,
-  kGeluExt = 124,
-  kAcoshExt = 125,
-  kReduceAll = 126,
-  kRemainderScalarTensor = 127,
-  kXlogy = 128,
-  kTraceExt = 129,
-  kInplaceAddmm = 130,
-  kBatchNormGatherStatsWithCounts = 131,
-  kInplaceLog = 132,
-  kOnesLikeExt = 133,
-  kDropoutGenMaskExt = 134,
-  kAddcdivExt = 135,
-  kDropoutDoMaskExt = 136,
-  kInplaceMul = 137,
-  kBitwiseNot = 138,
-  kSquare = 139,
-  kZeros = 140,
-  kRound = 141,
-  kLogSumExp = 142,
-  kInnerInplaceIndexPut = 143,
-  kConvolutionStr = 144,
-  kDot = 145,
-  kTrunc = 146,
-  kAddLayerNormGrad = 147,
-  kConv2DPadding = 148,
-  kGatherD = 149,
-  kMatMulExt = 150,
-  kUpsampleBilinear2DGrad = 151,
-  kMaskedFill = 152,
-  kExp = 153,
-  kTensorScatterElements = 154,
-  kTanhGrad = 155,
-  kBitwiseXorScalar = 156,
-  kSoftMarginLossGrad = 157,
-  kAvgPool2D = 158,
-  kRepeat = 159,
-  kAdaptiveAvgPool2DExt = 160,
-  kFillTensor = 161,
-  kSinc = 162,
-  kExpandDimsView = 163,
-  kAvgPool1D = 164,
-  kReshape = 165,
-  kIndexSelect = 166,
-  kReluGrad = 167,
-  kBernoulliExt = 168,
-  kConv1DExt = 169,
-  kLogSoftmaxExt = 170,
-  kSqrt = 171,
-  kSigmoidGrad = 172,
-  kReflectionPad3DGrad = 173,
-  kPowScalarTensor = 174,
-  kInplaceReLU = 175,
-  kMaxPoolWithIndices = 176,
-  kHSigmoid = 177,
-  kNLLLossGrad = 178,
-  kUniqueConsecutive = 179,
-  kUniformExt = 180,
-  kGroupNormGrad = 181,
-  kBitwiseOrScalar = 182,
-  kConv3DPadding = 183,
-  kAddLayerNormV2 = 184,
-  kGridSampler2D = 185,
-  kMaxPoolGradWithIndices = 186,
-  kConv1DPadding = 187,
-  kSoftplusGradExt = 188,
-  kGeLUGrad = 189,
-  kNormalTensorTensor = 190,
-  kMaskedSelect = 191,
-  kMSELossGradExt = 192,
-  kSwigluGrad = 193,
-  kUpsampleNearest1D = 194,
-  kBatchNormExt = 195,
-  kNeScalar = 196,
-  kNansum = 197,
-  kReflectionPad1D = 198,
-  kAvgPool3DExt = 199,
-  kBitwiseOrTensor = 200,
-  kHistcExt = 201,
-  kSmoothL1LossGrad = 202,
-  kLess = 203,
-  kMoeTokenUnpermute = 204,
-  kSign = 205,
-  kTriangularSolve = 206,
-  kPReLUGrad = 207,
-  kInplaceNormal = 208,
-  kSelectExtView = 209,
-  kInplaceFillTensor = 210,
-  kEluGradExt = 211,
-  kAdaptiveMaxPool2D = 212,
-  kAtanExt = 213,
-  kSubExt = 214,
-  kLogicalNot = 215,
-  kNLLLoss = 216,
-  kSin = 217,
-  kUpsampleTrilinear3D = 218,
-  kView = 219,
-  kChunkView = 220,
-  kFloorDivScalar = 221,
-  kSinh = 222,
-  kSwiglu = 223,
-  kClampScalar = 224,
-  kRandInt = 225,
-  kNormalFloatFloat = 226,
-  kAddScalar = 227,
-  kReplicationPad2DGrad = 228,
-  kDense = 229,
-  kSiLU = 230,
-  kLog = 231,
-  kUniqueDim = 232,
-  kSlice = 233,
-  kSmoothL1Loss = 234,
-  kCustomExt = 235,
-  kAsinhExt = 236,
-  kMaxDim = 237,
-  kInplaceFloorDivides = 238,
-  kNewZeros = 239,
-  kRandpermExt = 240,
-  kTExt = 241,
-  kPromptFlashAttention = 242,
-  kIndexAddExt = 243,
-  kUpsampleNearest3D = 244,
-  kUpsampleLinear1D = 245,
-  kSoftMarginLoss = 246,
-  kGeluGradExt = 247,
-  kConcat = 248,
-  kAcosExt = 249,
-  kIndex = 250,
-  kAddmm = 251,
-  kOnes = 252,
-  kScatter = 253,
-  kChunk = 254,
-  kReflectionPad2D = 255,
-  kInplaceIndexPut = 256,
-  kUpsampleLinear1DGrad = 257,
-  kGenerator = 258,
-  kTransposeExtView = 259,
-  kInplaceScatterValue = 260,
-  kLerpScalar = 261,
-  kLayerNormGradExt = 262,
-  kArgSort = 263,
-  kReplicationPad2D = 264,
-  kAdaptiveMaxPool1D = 265,
-  kNLLLoss2dGrad = 266,
-  kConvolution = 267,
-  kClampTensor = 268,
-  kSubScalar = 269,
-  kReflectionPad2DGrad = 270,
-  kTransposeView = 271,
-  kEmbeddingDenseBackward = 272,
-  kArgMaxExt = 273,
-  kNormalTensorFloat = 274,
-  kExp2 = 275,
-  kLogicalOr = 276,
-  kAdamW = 277,
-  kEmbedding = 278,
-  kLinSpaceExt = 279,
-  kCummax = 280,
-  kAtan2Ext = 281,
-  kInplaceRandom = 282,
-  kRandLikeExt = 283,
-  kXLogYScalarOther = 284,
-  kIsClose = 285,
-  kMeshgrid = 286,
-  kNonZeroExt = 287,
-  kMaximum = 288,
-  kAvgPool2DGrad = 289,
-  kAdaptiveAvgPool1D = 290,
-  kReplicationPad3DGrad = 291,
-  kInplaceFillScalar = 292,
-  kSoftplusExt = 293,
-  kReflectionPad3D = 294,
-  kBinaryCrossEntropy = 295,
-  kUpsampleTrilinear3DGrad = 296,
-  kRandnLike = 297,
-  kUpsampleBicubic2DGrad = 298,
-  kMultiScaleDeformableAttn = 299,
-  kInplaceAddsExt = 300,
-  kPolar = 301,
-  kInplacePut = 302,
-  kLogicalAnd = 303,
-  kFillScalar = 304,
-  kScatterAddExt = 305,
-  kExpandDims = 306,
-  kFrac = 307,
-  kInplaceScatterAdd = 308,
-  kEqualExt = 309,
-  kAtanh = 310,
-  kMishGradExt = 311,
-  kErfc = 312,
-  kInplaceDivs = 313,
-  kClone = 314,
-  kArgMinExt = 315,
-  kMul = 316,
-  kAdaptiveAvgPool3DExt = 317,
-  kInplaceZero = 318,
-  kRepeatInterleaveTensor = 319,
-  kRotaryPositionEmbedding = 320,
-  kGcd = 321,
-  kOuter = 322,
-  kSplitTensor = 323,
-  kAsStrided = 324,
-  kSilentCheckV2 = 325,
-  kAsinExt = 326,
-  kRepeatInterleaveInt = 327,
-  kNarrowView = 328,
-  kSplit = 329,
-  kFullLike = 330,
-  kInplaceMaskedFillTensor = 331,
-  kFlattenExt = 332,
-  kMoeTokenPermute = 333,
-  kKthvalue = 334,
-  kRmsNorm = 335,
-  kConvTranspose2D = 336,
-  kRoll = 337,
-  kStackExt = 338,
-  kMeanExt = 339,
-  kExpm1 = 340,
-  kInplaceFloorDivide = 341,
-  kMaxPoolWithMask = 342,
-  kZerosLikeExt = 343,
-  kBatchNormStats = 344,
-  kConvolutionStrGrad = 345,
-  kIsNegInf = 346,
-  kMaxUnpool2DExt = 347,
-  kTranspose = 348,
-  kTake = 349,
-  kNLLLoss2d = 350,
-  kMinimum = 351,
-  kReplicationPad1DGrad = 352,
-  kMoeTokenPermuteGrad = 353,
-  kL1LossExt = 354,
-  kInplaceHardtanh = 355,
-  kRandIntLike = 356,
-  kIndexFillScalar = 357,
-  kNormalFloatTensor = 358,
-  kLinalgQr = 359,
-  kCosh = 360,
-  kInplaceAddExt = 361,
-  kUpsampleBilinear2D = 362,
-  kIsInf = 363,
-  kTan = 364,
-  kEqual = 365,
-  kRmsNormGrad = 366,
-  kContiguous = 367,
-  kGreaterEqual = 368,
-  kMinDim = 369,
-  kInplaceFillDiagonal = 370,
-  kNanToNum = 371,
-  kSpeedFusionAttentionGrad = 372,
-  kNotEqual = 373,
-  kFloorDiv = 374,
-  kDiagExt = 375,
-  kSoftmaxBackward = 376,
-  kThreshold = 377,
-  kNarrow = 378,
-  kInplaceSubExt = 379,
-  kSumExt = 380,
-  kConvolutionGrad = 381,
-  kMin = 382,
-  kRepeatInterleaveGrad = 383,
-  kBitwiseAndTensor = 384,
-  kHSwish = 385,
-  kMatMul = 386,
-  kStd = 387,
-  kElu = 388,
-  kIm2ColExt = 389,
-  kLessEqual = 390,
-  kLog2 = 391,
-  kProdExt = 392,
-  kFmodScalar = 393,
-  kTanh = 394,
-  kHShrinkGrad = 395,
-  kArgMinWithValue = 396,
-  kInplaceSubScalar = 397,
-  kArange = 398,
-  kGreaterEqualScalar = 399,
-  kBitwiseAndScalar = 400,
-  kInnerNonZero = 401,
-  kNewOnes = 402,
-  kBitwiseXorTensor = 403,
-  kAdd = 404,
-  kDropoutGradExt = 405,
-  kInplaceScatterValueReduce = 406,
-  kGroupNorm = 407,
-  kIsFinite = 408,
-  kFFNExt = 409,
-  kMoeTokenUnpermuteGrad = 410,
-  kVar = 411,
-  kAddmv = 412,
-  kSpeedFusionAttention = 413,
-  kTopkExt = 414,
-  kInplaceElu = 415,
-  kLinalgVectorNorm = 416,
-  kXLogYScalarSelf = 417,
-  kDivMods = 418,
-  kInplaceUniform = 419,
-  kLayerNormExt = 420,
-  kSigmoid = 421,
-  kMultinomialExt = 422,
-  kInplaceDivMod = 423,
-  kPowTensorScalar = 424,
-  kFloor = 425,
-  kSplitWithSize = 426,
-  kInplaceTanh = 427,
-  kReflectionPad1DGrad = 428,
-  kErf = 429,
-  kBincountExt = 430,
-  kUpsampleBicubic2D = 431,
-  kMedianDim = 432,
-  kInplaceDiv = 433,
-  kInplaceClampTensor = 434,
-  kInplaceMuls = 435,
-  kReduceMin = 436,
-  kReciprocal = 437,
-  kSeluGrad = 438,
-  kVarMean = 439,
-  kLogicalXor = 440,
-  kMm = 441,
-  kGluGrad = 442,
-  kGeLU = 443,
-  kAddbmm = 444,
-  kSelect = 445,
-  kInplaceErfinv = 446,
-  kNeg = 447,
-  kAdaptiveAvgPool3DGradExt = 448,
-  kConstantPadND = 449,
-  kCol2ImGrad = 450,
-  kReduceMax = 451,
-  kUpsampleNearest1DGrad = 452,
-  kL1LossBackwardExt = 453,
-  kBinaryCrossEntropyGrad = 454,
-  kIncreFlashAttention = 455,
-  kFlashAttentionScoreGrad = 456,
-  kIdentity = 457,
-  kPow = 458,
-  kExpandAs = 459,
-  kUpsampleNearest2D = 460,
-  kCast = 461,
-  kDropoutExt = 462,
-  kGreater = 463,
-  kMax = 464,
-  kFlashAttentionScore = 465,
-  kUnstackExtView = 466,
-  kHSigmoidGrad = 467,
-  kDynamicQuantExt = 468,
-  kKVCacheScatterUpdate = 469,
-  kWeightQuantBatchMatmul = 470,
-  kMoeGatingTopKSoftmax = 471,
-  kMatmulAllReduceAddRmsNorm = 472,
-  kAddRmsNormQuantV2 = 473,
-  kMoeFinalizeRouting = 474,
-  kMoeComputeExpertTokens = 475,
-  kGroupedMatmulV4 = 476,
-  kGroupedMatmul = 477,
-  kQuantBatchMatmul = 478,
-  kFusedInferAttentionScore = 479,
-  kMoeInitRouting = 480,
-  kMoeInitRoutingV2 = 481,
-  kGroupedMatmulV2 = 482,
-  kQuantV2 = 483,
-  kGmmBackwardFusion = 484,
-  kGmmV2BackwardFusion = 485,
-  kGmmV2 = 486,
-  kGmmBackward = 487,
-  kGmm = 488,
-  kGmmV2Backward = 489,
-  kPixelShuffle = 490,
-  kInplaceExponential = 491,
+  kNewFull = 29,
+  kBatchNormReduceGrad = 30,
+  kInplaceScatterSrc = 31,
+  kUpsampleNearest2DGrad = 32,
+  kSplitWithSizeView = 33,
+  kKLDiv = 34,
+  kEluExt = 35,
+  kBaddbmm = 36,
+  kReduceAny = 37,
+  kLogSigmoid = 38,
+  kLeakyReLUGradExt = 39,
+  kKLDivGrad = 40,
+  kTriu = 41,
+  kMultiScaleDeformableAttnGrad = 42,
+  kLogSoftmax = 43,
+  kInplaceFloor = 44,
+  kInplaceStopGradient = 45,
+  kBroadcastToView = 46,
+  kDiagonalView = 47,
+  kIndexFillTensor = 48,
+  kSilentCheckV3 = 49,
+  kAddExt = 50,
+  kRsqrt = 51,
+  kViewAs = 52,
+  kErfinv = 53,
+  kTrilExt = 54,
+  kCumsumExt = 55,
+  kLogSigmoidGrad = 56,
+  kAvgPool3DGradExt = 57,
+  kEmptyLike = 58,
+  kBCEWithLogitsLoss = 59,
+  kCeil = 60,
+  kHardtanhGrad = 61,
+  kUnique2 = 62,
+  kReshapeAndCache = 63,
+  kHSwishGrad = 64,
+  kMedianExt = 65,
+  kDistCommAllGather = 66,
+  kReplicationPad1D = 67,
+  kCol2ImExt = 68,
+  kReLU = 69,
+  kAllGatherMatmul = 70,
+  kLerp = 71,
+  kAdaptiveAvgPool2DGradExt = 72,
+  kDistCommGatherIntoTensor = 73,
+  kMatrixInverseExt = 74,
+  kRotaryPositionEmbeddingGrad = 75,
+  kEye = 76,
+  kMaxPoolGradWithMask = 77,
+  kBatchMatMulExt = 78,
+  kGatherDGradV2 = 79,
+  kNorm = 80,
+  kInplaceIndexAddExt = 81,
+  kInnerIndex = 82,
+  kGLU = 83,
+  kInplaceCopy = 84,
+  kDistCommAllReduce = 85,
+  kReplicationPad3D = 86,
+  kLog10 = 87,
+  kSeLUExt = 88,
+  kInplaceClampScalar = 89,
+  kLog1p = 90,
+  kGridSampler3D = 91,
+  kFmodTensor = 92,
+  kArgMaxWithValue = 93,
+  kInplaceGroupedMatmulAdd = 94,
+  kLogAddExp = 95,
+  kMaskedSelectGrad = 96,
+  kBroadcastTo = 97,
+  kCopy = 98,
+  kSoftmax = 99,
+  kStdMean = 100,
+  kDistCommBarrier = 101,
+  kSoftShrink = 102,
+  kAddcmulExt = 103,
+  kSiLUGrad = 104,
+  kDiv = 105,
+  kMatmulReduceScatter = 106,
+  kSliceExt = 107,
+  kSelectV2 = 108,
+  kDistCommScatterTensor = 109,
+  kThresholdGrad = 110,
+  kDivs = 111,
+  kBinaryCrossEntropyWithLogitsBackward = 112,
+  kCos = 113,
+  kSliceExtView = 114,
+  kCross = 115,
+  kMSELossExt = 116,
+  kInplaceDivMods = 117,
+  kGridSampler2DGrad = 118,
+  kBatchNormElemtGrad = 119,
+  kBatchNormGradExt = 120,
+  kAbs = 121,
+  kConv3DExt = 122,
+  kInplaceThreshold = 123,
+  kBatchNormElemt = 124,
+  kLogSoftmaxGrad = 125,
+  kRemainderTensorScalar = 126,
+  kInplaceScatterSrcReduce = 127,
+  kRandn = 128,
+  kHShrink = 129,
+  kSortExt = 130,
+  kPReLU = 131,
+  kInplaceExp = 132,
+  kGeluExt = 133,
+  kAcoshExt = 134,
+  kReduceAll = 135,
+  kRemainderScalarTensor = 136,
+  kXlogy = 137,
+  kTraceExt = 138,
+  kInplaceAddmm = 139,
+  kBatchNormGatherStatsWithCounts = 140,
+  kInplaceLog = 141,
+  kOnesLikeExt = 142,
+  kDropoutGenMaskExt = 143,
+  kAddcdivExt = 144,
+  kMoeDistributeCombine = 145,
+  kDropoutDoMaskExt = 146,
+  kInplaceMul = 147,
+  kBitwiseNot = 148,
+  kSquare = 149,
+  kZeros = 150,
+  kRound = 151,
+  kLogSumExp = 152,
+  kInnerInplaceIndexPut = 153,
+  kConvolutionStr = 154,
+  kDot = 155,
+  kDistCommAllToAllV = 156,
+  kTrunc = 157,
+  kAddLayerNormGrad = 158,
+  kConv2DPadding = 159,
+  kGatherD = 160,
+  kMatMulExt = 161,
+  kUpsampleBilinear2DGrad = 162,
+  kDistCommBroadcast = 163,
+  kMaskedFill = 164,
+  kExp = 165,
+  kTensorScatterElements = 166,
+  kTanhGrad = 167,
+  kBitwiseXorScalar = 168,
+  kSoftMarginLossGrad = 169,
+  kAvgPool2D = 170,
+  kRepeat = 171,
+  kAdaptiveAvgPool2DExt = 172,
+  kFillTensor = 173,
+  kSinc = 174,
+  kExpandDimsView = 175,
+  kAvgPool1D = 176,
+  kReshape = 177,
+  kIndexSelect = 178,
+  kReluGrad = 179,
+  kBernoulliExt = 180,
+  kConv1DExt = 181,
+  kLogSoftmaxExt = 182,
+  kSqrt = 183,
+  kSigmoidGrad = 184,
+  kDistCommReduce = 185,
+  kReflectionPad3DGrad = 186,
+  kPowScalarTensor = 187,
+  kEmpty = 188,
+  kInplaceReLU = 189,
+  kMaxPoolWithIndices = 190,
+  kHSigmoid = 191,
+  kNLLLossGrad = 192,
+  kUniqueConsecutive = 193,
+  kUniformExt = 194,
+  kGroupNormGrad = 195,
+  kInnerCommAllReduce = 196,
+  kBitwiseOrScalar = 197,
+  kConv3DPadding = 198,
+  kAddLayerNormV2 = 199,
+  kGridSampler2D = 200,
+  kMaxPoolGradWithIndices = 201,
+  kConv1DPadding = 202,
+  kDistCommIsend = 203,
+  kSoftplusGradExt = 204,
+  kGeLUGrad = 205,
+  kNormalTensorTensor = 206,
+  kDistCommIrecv = 207,
+  kMaskedSelect = 208,
+  kMSELossGradExt = 209,
+  kSwigluGrad = 210,
+  kUpsampleNearest1D = 211,
+  kBatchNormExt = 212,
+  kNeScalar = 213,
+  kNansum = 214,
+  kReflectionPad1D = 215,
+  kAvgPool3DExt = 216,
+  kBitwiseOrTensor = 217,
+  kHistcExt = 218,
+  kSmoothL1LossGrad = 219,
+  kLess = 220,
+  kSign = 221,
+  kTriangularSolve = 222,
+  kPReLUGrad = 223,
+  kInplaceNormal = 224,
+  kSelectExtView = 225,
+  kInplaceFillTensor = 226,
+  kEluGradExt = 227,
+  kAdaptiveMaxPool2D = 228,
+  kAtanExt = 229,
+  kSubExt = 230,
+  kLogicalNot = 231,
+  kNLLLoss = 232,
+  kSin = 233,
+  kUpsampleTrilinear3D = 234,
+  kView = 235,
+  kChunkView = 236,
+  kFloorDivScalar = 237,
+  kSinh = 238,
+  kSwiglu = 239,
+  kClampScalar = 240,
+  kRandInt = 241,
+  kNormalFloatFloat = 242,
+  kAddScalar = 243,
+  kReplicationPad2DGrad = 244,
+  kDense = 245,
+  kSiLU = 246,
+  kLog = 247,
+  kUniqueDim = 248,
+  kSlice = 249,
+  kSmoothL1Loss = 250,
+  kCustomExt = 251,
+  kAsinhExt = 252,
+  kMaxDim = 253,
+  kInplaceFloorDivides = 254,
+  kNewZeros = 255,
+  kRandpermExt = 256,
+  kTExt = 257,
+  kPromptFlashAttention = 258,
+  kIndexAddExt = 259,
+  kUpsampleNearest3D = 260,
+  kUpsampleLinear1D = 261,
+  kApplyRotaryPosEmb = 262,
+  kSoftMarginLoss = 263,
+  kGeluGradExt = 264,
+  kConcat = 265,
+  kDistCommReduceScatter = 266,
+  kAcosExt = 267,
+  kIndex = 268,
+  kAddmm = 269,
+  kOnes = 270,
+  kScatter = 271,
+  kChunk = 272,
+  kDistCommAllGatherIntoTensor = 273,
+  kReflectionPad2D = 274,
+  kInplaceIndexPut = 275,
+  kUpsampleLinear1DGrad = 276,
+  kGenerator = 277,
+  kTransposeExtView = 278,
+  kInplaceScatterValue = 279,
+  kLerpScalar = 280,
+  kLayerNormGradExt = 281,
+  kArgSort = 282,
+  kReplicationPad2D = 283,
+  kAdaptiveMaxPool1D = 284,
+  kNLLLoss2dGrad = 285,
+  kConvolution = 286,
+  kClampTensor = 287,
+  kSubScalar = 288,
+  kReflectionPad2DGrad = 289,
+  kTransposeView = 290,
+  kEmbeddingDenseBackward = 291,
+  kArgMaxExt = 292,
+  kNormalTensorFloat = 293,
+  kExp2 = 294,
+  kLogicalOr = 295,
+  kAdamW = 296,
+  kEmbedding = 297,
+  kLinSpaceExt = 298,
+  kCummax = 299,
+  kAtan2Ext = 300,
+  kInplaceRandom = 301,
+  kRandLikeExt = 302,
+  kXLogYScalarOther = 303,
+  kIsClose = 304,
+  kInnerCommAllGather = 305,
+  kMeshgrid = 306,
+  kNonZeroExt = 307,
+  kMaximum = 308,
+  kAvgPool2DGrad = 309,
+  kAdaptiveAvgPool1D = 310,
+  kReplicationPad3DGrad = 311,
+  kInplaceFillScalar = 312,
+  kSoftplusExt = 313,
+  kReflectionPad3D = 314,
+  kBinaryCrossEntropy = 315,
+  kUpsampleTrilinear3DGrad = 316,
+  kRandnLike = 317,
+  kUpsampleBicubic2DGrad = 318,
+  kPagedAttention = 319,
+  kMultiScaleDeformableAttn = 320,
+  kInplaceAddsExt = 321,
+  kPolar = 322,
+  kInplacePut = 323,
+  kLogicalAnd = 324,
+  kFillScalar = 325,
+  kScatterAddExt = 326,
+  kExpandDims = 327,
+  kFrac = 328,
+  kInplaceScatterAdd = 329,
+  kEqualExt = 330,
+  kAtanh = 331,
+  kMishGradExt = 332,
+  kErfc = 333,
+  kInplaceDivs = 334,
+  kClone = 335,
+  kDistCommScatter = 336,
+  kArgMinExt = 337,
+  kMul = 338,
+  kAdaptiveAvgPool3DExt = 339,
+  kInplaceZero = 340,
+  kInplaceSiLU = 341,
+  kRepeatInterleaveTensor = 342,
+  kRotaryPositionEmbedding = 343,
+  kGcd = 344,
+  kOuter = 345,
+  kDistCommReduceScatterTensor = 346,
+  kSplitTensor = 347,
+  kAsStrided = 348,
+  kSilentCheckV2 = 349,
+  kDistCommAllToAllVSingle = 350,
+  kAsinExt = 351,
+  kRepeatInterleaveInt = 352,
+  kNarrowView = 353,
+  kSplit = 354,
+  kFullLike = 355,
+  kInplaceMaskedFillTensor = 356,
+  kFlattenExt = 357,
+  kMoeTokenPermute = 358,
+  kKthvalue = 359,
+  kRmsNorm = 360,
+  kConvTranspose2D = 361,
+  kRoll = 362,
+  kInnerCommIsend = 363,
+  kStackExt = 364,
+  kMeanExt = 365,
+  kExpm1 = 366,
+  kInplaceFloorDivide = 367,
+  kNewEmpty = 368,
+  kMaxPoolWithMask = 369,
+  kZerosLikeExt = 370,
+  kBatchNormStats = 371,
+  kConvolutionStrGrad = 372,
+  kIsNegInf = 373,
+  kMaxUnpool2DExt = 374,
+  kTranspose = 375,
+  kTake = 376,
+  kNLLLoss2d = 377,
+  kMinimum = 378,
+  kReplicationPad1DGrad = 379,
+  kMoeTokenPermuteGrad = 380,
+  kL1LossExt = 381,
+  kInplaceHardtanh = 382,
+  kRandIntLike = 383,
+  kIndexFillScalar = 384,
+  kNormalFloatTensor = 385,
+  kLinalgQr = 386,
+  kCosh = 387,
+  kInplaceAddExt = 388,
+  kUpsampleBilinear2D = 389,
+  kIsInf = 390,
+  kTan = 391,
+  kInnerCommReduceScatter = 392,
+  kEqual = 393,
+  kRmsNormGrad = 394,
+  kContiguous = 395,
+  kGreaterEqual = 396,
+  kMinDim = 397,
+  kInnerMoeTokenUnpermute = 398,
+  kInplaceFillDiagonal = 399,
+  kNanToNum = 400,
+  kSpeedFusionAttentionGrad = 401,
+  kNotEqual = 402,
+  kFloorDiv = 403,
+  kDiagExt = 404,
+  kSoftmaxBackward = 405,
+  kThreshold = 406,
+  kNarrow = 407,
+  kInplaceSubExt = 408,
+  kSumExt = 409,
+  kConvolutionGrad = 410,
+  kMin = 411,
+  kRepeatInterleaveGrad = 412,
+  kBitwiseAndTensor = 413,
+  kHSwish = 414,
+  kMatMul = 415,
+  kStd = 416,
+  kElu = 417,
+  kIm2ColExt = 418,
+  kLessEqual = 419,
+  kLog2 = 420,
+  kProdExt = 421,
+  kFmodScalar = 422,
+  kTanh = 423,
+  kHShrinkGrad = 424,
+  kArgMinWithValue = 425,
+  kInplaceSubScalar = 426,
+  kArange = 427,
+  kGreaterEqualScalar = 428,
+  kBitwiseAndScalar = 429,
+  kInnerNonZero = 430,
+  kNewOnes = 431,
+  kBitwiseXorTensor = 432,
+  kAdd = 433,
+  kDropoutGradExt = 434,
+  kInplaceScatterValueReduce = 435,
+  kGroupNorm = 436,
+  kIsFinite = 437,
+  kFFNExt = 438,
+  kMoeTokenUnpermuteGrad = 439,
+  kVar = 440,
+  kAddmv = 441,
+  kSpeedFusionAttention = 442,
+  kTopkExt = 443,
+  kInplaceElu = 444,
+  kLinalgVectorNorm = 445,
+  kInnerCommAllToAllV = 446,
+  kXLogYScalarSelf = 447,
+  kDivMods = 448,
+  kInplaceUniform = 449,
+  kLayerNormExt = 450,
+  kSigmoid = 451,
+  kMultinomialExt = 452,
+  kInplaceDivMod = 453,
+  kInnerCommIrecv = 454,
+  kPowTensorScalar = 455,
+  kFloor = 456,
+  kSplitWithSize = 457,
+  kInplaceTanh = 458,
+  kDistCommGather = 459,
+  kReflectionPad1DGrad = 460,
+  kErf = 461,
+  kBincountExt = 462,
+  kUpsampleBicubic2D = 463,
+  kMedianDim = 464,
+  kInplaceDiv = 465,
+  kInplaceClampTensor = 466,
+  kInplaceMuls = 467,
+  kReduceMin = 468,
+  kReciprocal = 469,
+  kSeluGrad = 470,
+  kVarMean = 471,
+  kLogicalXor = 472,
+  kMm = 473,
+  kGluGrad = 474,
+  kGeLU = 475,
+  kAddbmm = 476,
+  kDistCommBatchIsendIrecv = 477,
+  kSelect = 478,
+  kInplaceErfinv = 479,
+  kNeg = 480,
+  kAdaptiveAvgPool3DGradExt = 481,
+  kConstantPadND = 482,
+  kCol2ImGrad = 483,
+  kReduceMax = 484,
+  kUpsampleNearest1DGrad = 485,
+  kInplaceMatmulAdd = 486,
+  kL1LossBackwardExt = 487,
+  kBinaryCrossEntropyGrad = 488,
+  kIncreFlashAttention = 489,
+  kFlashAttentionScoreGrad = 490,
+  kIdentity = 491,
+  kMoeDistributeDispatch = 492,
+  kPow = 493,
+  kExpandAs = 494,
+  kUpsampleNearest2D = 495,
+  kCast = 496,
+  kDropoutExt = 497,
+  kGreater = 498,
+  kMax = 499,
+  kFlashAttentionScore = 500,
+  kUnstackExtView = 501,
+  kHSigmoidGrad = 502,
+  kQuantMatmul = 503,
+  kDynamicQuantExt = 504,
+  kKVCacheScatterUpdate = 505,
+  kWeightQuantBatchMatmul = 506,
+  kMoeGatingTopKSoftmax = 507,
+  kMatmulAllReduceAddRmsNorm = 508,
+  kAddRmsNormQuantV2 = 509,
+  kMoeFinalizeRouting = 510,
+  kMoeComputeExpertTokens = 511,
+  kGroupedMatmulV4 = 512,
+  kGroupedMatmul = 513,
+  kQuantBatchMatmul = 514,
+  kFusedInferAttentionScore = 515,
+  kMoeInitRoutingQuantV2 = 516,
+  kMoeInitRouting = 517,
+  kMoeInitRoutingV2 = 518,
+  kGroupedMatmulV2 = 519,
+  kQuantV2 = 520,
+  kGmmBackwardFusion = 521,
+  kAnyExt = 522,
+  kGmmV2BackwardFusion = 523,
+  kGmmV2 = 524,
+  kGmmBackward = 525,
+  kGmm = 526,
+  kMoeTokenUnpermute = 527,
+  kGmmV2Backward = 528,
+  kPixelShuffle = 529,
+  kAny = 530,
+  kFuncMaxPool2D = 531,
+  kEinsumExt = 532,
+  kInplaceExponential = 533,
 };
 
-using LeakyReLUExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using BatchMatMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using CumminExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-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 LogAddExp2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MulsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using RemainderTensorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
+using LeakyReLUExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using BatchMatMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using CumminExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using RandExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using LogAddExp2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MulsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using RemainderTensorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
 using AllFiniteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &)>;
-using SoftShrinkGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using SubGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using Conv2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using 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 ReverseV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using TileGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using CountNonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using ScatterValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::Int64ImmPtr &)>;
-using MishExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using TypeAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceMaskedFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using 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 SqueezeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using MvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 DivModGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using HardtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using SplitTensorViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-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 AddRmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-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 InplaceScatterSrcGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 SplitWithSizeViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using KLDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using EluExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using BaddbmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using ReduceAnyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
-using LogSigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LeakyReLUGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::BoolImmPtr &)>;
-using 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 TriuGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using MultiScaleDeformableAttnGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogSoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceFloorGradFunc = 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 BroadcastToViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-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 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 AddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using RsqrtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ViewAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ErfinvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using TrilExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using CumsumExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using LogSigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 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 CeilGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using HardtanhGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using Unique2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using HSwishGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MedianExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReplicationPad1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-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 ReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 LerpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, 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 MatrixInverseExtGradFunc = 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 SoftShrinkGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using SubGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using NonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 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 ReverseV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using TileGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using CountNonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
+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 MishExtGradFunc = 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 InplaceMaskedFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const 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 SqueezeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using MvGradFunc = 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 DivModGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using HardtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using SplitTensorViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using 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 AddRmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+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 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 InplaceScatterSrcGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 SplitWithSizeViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &, const int64_t &)>;
+using KLDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using EluExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+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 ReduceAnyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
+using LogSigmoidGradFunc = 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 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 TriuGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+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 LogSoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceFloorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceStopGradientGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BroadcastToViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using DiagonalViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &, const int64_t &)>;
+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 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 AddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using RsqrtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ViewAsGradFunc = 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 TrilExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using CumsumExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using LogSigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 EmptyLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, 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 CeilGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using HardtanhGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using Unique2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using ReshapeAndCacheGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+using HSwishGradGradFunc = 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 DistCommAllGatherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using ReplicationPad1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+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 ReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using AllGatherMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using LerpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AdaptiveAvgPool2DGradExtGradFunc = 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 MatrixInverseExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 EyeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-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 BatchMatMulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, 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 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 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 InnerIndexGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using GLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceCopyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReplicationPad3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using Log10GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SeLUExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceClampScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ScalarPtr> &, const std::optional<mindspore::ScalarPtr> &)>;
-using Log1pGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 FmodTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ArgMaxWithValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using InplaceGroupedMatmulAddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogAddExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaskedSelectGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BroadcastToGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using CopyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-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 SoftShrinkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-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 SiLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 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 SelectV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, 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 DivsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-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 CosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SliceExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using CrossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using MSELossExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceDivModsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using GridSampler2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &)>;
-using 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 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 AbsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using Conv3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceThresholdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &)>;
-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 LogSoftmaxGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using RemainderTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-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 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 HShrinkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using SortExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using PReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using GeluExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using AcoshExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReduceAllGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
-using RemainderScalarTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using XlogyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using TraceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 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 InplaceLogGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using OnesLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<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 AddcdivExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using DropoutDoMaskExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using InplaceMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BitwiseNotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SquareGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
+using 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 BatchMatMulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 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 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 InnerIndexGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using GLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceCopyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DistCommAllReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using ReplicationPad3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using Log10GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SeLUExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceClampScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ScalarPtr> &, const std::optional<mindspore::ScalarPtr> &)>;
+using Log1pGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 FmodTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ArgMaxWithValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+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 LogAddExpGradFunc = 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 BroadcastToGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using CopyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+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 DistCommBarrierGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::StringImmPtr &)>;
+using SoftShrinkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+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 SiLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, 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 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 SelectV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, 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 ThresholdGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using DivsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+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 CosGradFunc = 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 CrossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using MSELossExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, 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 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 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 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 AbsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 InplaceThresholdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+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 LogSoftmaxGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using RemainderTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+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 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 HShrinkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using SortExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using PReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using GeluExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using AcoshExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 RemainderScalarTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &)>;
+using XlogyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using TraceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceAddmmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using 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 InplaceLogGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using OnesLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+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 AddcdivExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+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 DropoutDoMaskExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using InplaceMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using BitwiseNotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SquareGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
 using ZerosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using RoundGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using LogSumExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
-using InnerInplaceIndexPutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &)>;
-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 DotGradFunc = 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 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 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 GatherDGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MatMulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 MaskedFillGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 TanhGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 SoftMarginLossGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, 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 RepeatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using AdaptiveAvgPool2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using FillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using SincGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ExpandDimsViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-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 ReshapeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using IndexSelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using BernoulliExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using 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 LogSoftmaxExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using SqrtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, 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 PowScalarTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceReLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaxPoolWithIndicesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using HSigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 UniqueConsecutiveGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-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 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 BitwiseOrScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-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 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 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 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 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 SoftplusGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using GeLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, 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 MaskedSelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MSELossGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using SwigluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using UpsampleNearest1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using 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 NeScalarGradFunc = 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 ReflectionPad1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-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 BitwiseOrTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using HistcExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using 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 LessGradFunc = 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 SignGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 PReLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceNormalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SelectExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using EluGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
-using AdaptiveMaxPool2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using AtanExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SubExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using LogicalNotGradFunc = 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 SinGradFunc = 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 ViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using ChunkViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using FloorDivScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using SinhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SwigluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using ClampScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ScalarPtr> &, const std::optional<mindspore::ScalarPtr> &)>;
-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 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 AddScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using ReplicationPad2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using DenseGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-using SiLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using UniqueDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
-using SliceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
-using SmoothL1LossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using RoundGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using LogSumExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
+using InnerInplaceIndexPutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &)>;
+using ConvolutionStrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using DotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DistCommAllToAllVGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using TruncGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using AddLayerNormGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using 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 GatherDGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &)>;
+using MatMulExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 DistCommBroadcastGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using MaskedFillGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ExpGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using 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 TanhGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using BitwiseXorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 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 RepeatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using AdaptiveAvgPool2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using FillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using SincGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using ExpandDimsViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &)>;
+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 ReshapeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using IndexSelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using BernoulliExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, 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 LogSoftmaxExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using SqrtGradFunc = 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 DistCommReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using ReflectionPad3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using PowScalarTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, 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 InplaceReLUGradFunc = 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 HSigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 UniqueConsecutiveGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+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 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 InnerCommAllReduceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using BitwiseOrScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using Conv3DPaddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
+using 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 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 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 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 DistCommIsendGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &)>;
+using SoftplusGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using GeLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, 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 DistCommIrecvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using MaskedSelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MSELossGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using SwigluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using UpsampleNearest1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+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 NeScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+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 ReflectionPad1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+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 BitwiseOrTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using HistcExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using 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 LessGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using SignGradFunc = 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 PReLUGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceNormalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using SelectExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using InplaceFillTensorGradFunc = 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 AdaptiveMaxPool2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using AtanExtGradFunc = 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 LogicalNotGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 SinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 ViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using ChunkViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using FloorDivScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using SinhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SwigluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, 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 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 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 AddScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using ReplicationPad2DGradGradFunc = 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 SiLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using LogGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using UniqueDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using SliceGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &, const std::vector<int64_t> &)>;
+using SmoothL1LossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &)>;
 using CustomExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &)>;
-using AsinhExtGradFunc = 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 InplaceFloorDividesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using NewZerosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<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 TExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using PromptFlashAttentionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using 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 UpsampleNearest3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-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 SoftMarginLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using GeluGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
+using AsinhExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MaxDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using InplaceFloorDividesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using NewZerosGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using RandpermExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using TExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 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 UpsampleNearest3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+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 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 SoftMarginLossGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using GeluGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
 using ConcatGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using AcosExtGradFunc = 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 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 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 AcosExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using IndexGradFunc = 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 OnesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-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 ChunkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using ReflectionPad2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using InplaceIndexPutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &)>;
-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 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 ChunkGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const 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 ReflectionPad2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceIndexPutGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &)>;
+using UpsampleLinear1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &)>;
 using GeneratorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
-using TransposeExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceScatterValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using LerpScalarGradFunc = 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 ArgSortGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using ReplicationPad2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using AdaptiveMaxPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-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 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 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 SubScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using ReflectionPad2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using TransposeViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-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 ArgMaxExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
-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 Exp2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using LogicalOrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 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 TransposeExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using InplaceScatterValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using LerpScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using 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 ArgSortGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using ReplicationPad2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using AdaptiveMaxPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+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 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 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 SubScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using ReflectionPad2DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using TransposeViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+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 ArgMaxExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
+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 Exp2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using LogicalOrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 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 LinSpaceExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &, const mindspore::Int64ImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using CummaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using Atan2ExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 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 XLogYScalarOtherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-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 MeshgridGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using NonZeroExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaximumGradFunc = 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 AdaptiveAvgPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using ReplicationPad3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using InplaceFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using SoftplusExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using ReflectionPad3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 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 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 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 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 InplaceAddsExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-using PolarGradFunc = 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 LogicalAndGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
+using CummaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using Atan2ExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 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 XLogYScalarOtherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using IsCloseGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::BoolImmPtr &)>;
+using InnerCommAllGatherGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &)>;
+using MeshgridGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const int64_t &)>;
+using NonZeroExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MaximumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 AdaptiveAvgPool1DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using ReplicationPad3DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using SoftplusExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using ReflectionPad3DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using 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 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 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 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 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 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 InplaceAddsExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using PolarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, 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 LogicalAndGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, 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 ScatterAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ExpandDimsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using FracGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceScatterAddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using EqualExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MishGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ErfcGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceDivsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using CloneGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ArgMinExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
-using MulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AdaptiveAvgPool3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using InplaceZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 RotaryPositionEmbeddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using GcdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 SplitTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using AsStridedGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-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 AsinExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 NarrowViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using SplitGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using FullLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using InplaceMaskedFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using FlattenExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-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 KthvalueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using RmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using ConvTranspose2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &, const mindspore::ValueTuplePtr &)>;
-using RollGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
+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 ExpandDimsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &)>;
+using FracGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 EqualExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MishGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ErfcGradFunc = 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 CloneGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 ArgMinExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &)>;
+using MulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AdaptiveAvgPool3DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceSiLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 RotaryPositionEmbeddingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using GcdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using OuterGradFunc = 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 SplitTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using AsStridedGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &, const std::vector<int64_t> &, const int64_t &)>;
+using 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 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 AsinExtGradFunc = 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 NarrowViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &, const int64_t &)>;
+using SplitGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &)>;
+using FullLikeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using InplaceMaskedFillTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using FlattenExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+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 KthvalueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using RmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+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 RollGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using InnerCommIsendGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &)>;
 using StackExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, 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 Expm1GradFunc = 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 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 ZerosLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using BatchNormStatsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using 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 IsNegInfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaxUnpool2DExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using TransposeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using TakeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NLLLoss2dGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using MinimumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReplicationPad1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using MoeTokenPermuteGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using L1LossExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceHardtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
-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 IndexFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-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 LinalgQrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using CoshGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using 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 IsInfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using TanGradFunc = 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 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 ContiguousGradFunc = 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 MinDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using InplaceFillDiagonalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::BoolImmPtr &)>;
-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 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 NotEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using FloorDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 SoftmaxBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using ThresholdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::FP32ImmPtr &)>;
-using NarrowGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceSubExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, 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 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 MinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using RepeatInterleaveGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using BitwiseAndTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using HSwishGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MatMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
-using StdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using EluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-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 LessEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using Log2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ProdExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using FmodScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using TanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using HShrinkGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using ArgMinWithValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using InplaceSubScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using MeanExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using Expm1GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceFloorDivideGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using 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 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 ZerosLikeExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using BatchNormStatsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+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 IsNegInfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 TransposeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &)>;
+using TakeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 MinimumGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ReplicationPad1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using MoeTokenPermuteGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using L1LossExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using InplaceHardtanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+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 IndexFillScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using NormalFloatTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::FP32ImmPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using LinalgQrGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using CoshGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceAddExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using 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 IsInfGradFunc = 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 InnerCommReduceScatterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::StringImmPtr &, const mindspore::StringImmPtr &)>;
+using EqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 ContiguousGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using GreaterEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MinDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+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 InplaceFillDiagonalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::BoolImmPtr &)>;
+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 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 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 DiagExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using SoftmaxBackwardGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using ThresholdGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::ScalarPtr &)>;
+using NarrowGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &, const int64_t &, const int64_t &)>;
+using InplaceSubExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+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 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 MinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 BitwiseAndTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using HSwishGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using MatMulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+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 EluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+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 LessEqualGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using Log2GradFunc = 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 FmodScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using TanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using HShrinkGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+using ArgMinWithValueGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using InplaceSubScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, 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 GreaterEqualScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using BitwiseAndScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using InnerNonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using NewOnesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using BitwiseXorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using AddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DropoutGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using 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 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 IsFiniteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 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 VarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-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 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 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 InplaceEluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &)>;
-using LinalgVectorNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::FP32ImmPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::BoolImmPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using XLogYScalarSelfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DivModsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const std::optional<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 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 SigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 InplaceDivModGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
-using PowTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using FloorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SplitWithSizeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceTanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ReflectionPad1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &)>;
-using ErfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 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 MedianDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
-using InplaceDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceClampTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-using InplaceMulsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &)>;
-using ReduceMinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
-using ReciprocalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using SeluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 LogicalXorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 GeLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-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 SelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using InplaceErfinvGradFunc = 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 AdaptiveAvgPool3DGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ConstantPadNDGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, 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 ReduceMaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
-using UpsampleNearest1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
-using L1LossBackwardExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-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 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 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 IdentityGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using PowGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using ExpandAsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, 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 CastGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-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 GreaterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using MaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using 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 UnstackExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using HSigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
-using DynamicQuantExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &)>;
-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 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 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 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 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 MoeComputeExpertTokensGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, 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::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 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 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 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 MoeInitRoutingGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using MoeInitRoutingV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using GreaterEqualScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using BitwiseAndScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using InnerNonZeroGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using NewOnesGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+using BitwiseXorTensorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using AddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using DropoutGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+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 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 IsFiniteGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 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 VarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+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 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 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 InplaceEluGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &)>;
+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 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 XLogYScalarSelfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &)>;
+using DivModsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+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 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 SigmoidGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 InplaceDivModGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::Int64ImmPtr> &)>;
+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 PowTensorScalarGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using FloorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SplitWithSizeGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::vector<int64_t> &, const int64_t &)>;
+using InplaceTanhGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 ReflectionPad1DGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &)>;
+using ErfGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using BincountExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using 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 MedianDimGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using InplaceDivGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 InplaceMulsGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &)>;
+using ReduceMinGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
+using ReciprocalGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using SeluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 LogicalXorGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MmGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using GluGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using GeLUGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 DistCommBatchIsendIrecvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::StringImmPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &)>;
+using SelectGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using InplaceErfinvGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using NegGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using AdaptiveAvgPool3DGradExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using ConstantPadNDGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ScalarPtr &)>;
+using 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 ReduceMaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &)>;
+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 InplaceMatmulAddGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using L1LossBackwardExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using 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 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 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 IdentityGradFunc = std::function<void(const kernel::pyboost::OpPtr &, 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 PowGradFunc = 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 UpsampleNearest2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &)>;
+using CastGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+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 GreaterGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+using MaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+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 UnstackExtViewGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const int64_t &)>;
+using HSigmoidGradGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
+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 DynamicQuantExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &)>;
+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 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 MoeGatingTopKSoftmaxGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using MatmulAllReduceAddRmsNormGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::FP32ImmPtr &, const mindspore::StringImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using 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 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 MoeComputeExpertTokensGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, 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 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 QuantBatchMatmulGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &)>;
+using FusedInferAttentionScoreGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::FP32ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using 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 MoeInitRoutingV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
 using GroupedMatmulV2GradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
-using 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 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 GmmBackwardFusionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::Int64ImmPtr &)>;
-using GmmV2BackwardFusionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::BaseTensorPtr> &, const mindspore::Int64ImmPtr &)>;
-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::BaseTensorPtr> &, const mindspore::Int64ImmPtr &, const mindspore::Int64ImmPtr &)>;
+using AnyExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &, const mindspore::BoolImmPtr &)>;
+using GmmV2BackwardFusionGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::tensor::TensorPtr> &, const mindspore::Int64ImmPtr &)>;
+using 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 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 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 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 PixelShuffleGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::Int64ImmPtr &)>;
-using InplaceExponentialGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::BaseTensorPtr &, const mindspore::tensor::BaseTensorPtr &)>;
+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 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 PixelShuffleGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::Int64ImmPtr &)>;
+using AnyGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &)>;
+using FuncMaxPool2DGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ValueTuplePtr &, const std::optional<mindspore::ValueTuplePtr> &, const mindspore::ValueTuplePtr &, const mindspore::ValueTuplePtr &, const mindspore::BoolImmPtr &, const mindspore::BoolImmPtr &)>;
+using EinsumExtGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::StringImmPtr &, const mindspore::ValueTuplePtr &)>;
+using InplaceExponentialGradFunc = std::function<void(const kernel::pyboost::OpPtr &, const mindspore::tensor::TensorPtr &, const mindspore::ScalarPtr &, const mindspore::tensor::TensorPtr &, const mindspore::tensor::TensorPtr &)>;
 
 struct OpsAutoGradRegisters {
   LeakyReLUExtGradFunc LeakyReLUExtGradFuncObj;
@@ -1043,6 +1127,7 @@ struct OpsAutoGradRegisters {
   SplitTensorViewGradFunc SplitTensorViewGradFuncObj;
   UpsampleNearest3DGradGradFunc UpsampleNearest3DGradGradFuncObj;
   AddRmsNormGradFunc AddRmsNormGradFuncObj;
+  NewFullGradFunc NewFullGradFuncObj;
   BatchNormReduceGradGradFunc BatchNormReduceGradGradFuncObj;
   InplaceScatterSrcGradFunc InplaceScatterSrcGradFuncObj;
   UpsampleNearest2DGradGradFunc UpsampleNearest2DGradGradFuncObj;
@@ -1060,6 +1145,7 @@ struct OpsAutoGradRegisters {
   InplaceFloorGradFunc InplaceFloorGradFuncObj;
   InplaceStopGradientGradFunc InplaceStopGradientGradFuncObj;
   BroadcastToViewGradFunc BroadcastToViewGradFuncObj;
+  DiagonalViewGradFunc DiagonalViewGradFuncObj;
   IndexFillTensorGradFunc IndexFillTensorGradFuncObj;
   SilentCheckV3GradFunc SilentCheckV3GradFuncObj;
   AddExtGradFunc AddExtGradFuncObj;
@@ -1070,18 +1156,22 @@ struct OpsAutoGradRegisters {
   CumsumExtGradFunc CumsumExtGradFuncObj;
   LogSigmoidGradGradFunc LogSigmoidGradGradFuncObj;
   AvgPool3DGradExtGradFunc AvgPool3DGradExtGradFuncObj;
+  EmptyLikeGradFunc EmptyLikeGradFuncObj;
   BCEWithLogitsLossGradFunc BCEWithLogitsLossGradFuncObj;
   CeilGradFunc CeilGradFuncObj;
   HardtanhGradGradFunc HardtanhGradGradFuncObj;
   Unique2GradFunc Unique2GradFuncObj;
+  ReshapeAndCacheGradFunc ReshapeAndCacheGradFuncObj;
   HSwishGradGradFunc HSwishGradGradFuncObj;
   MedianExtGradFunc MedianExtGradFuncObj;
+  DistCommAllGatherGradFunc DistCommAllGatherGradFuncObj;
   ReplicationPad1DGradFunc ReplicationPad1DGradFuncObj;
   Col2ImExtGradFunc Col2ImExtGradFuncObj;
   ReLUGradFunc ReLUGradFuncObj;
   AllGatherMatmulGradFunc AllGatherMatmulGradFuncObj;
   LerpGradFunc LerpGradFuncObj;
   AdaptiveAvgPool2DGradExtGradFunc AdaptiveAvgPool2DGradExtGradFuncObj;
+  DistCommGatherIntoTensorGradFunc DistCommGatherIntoTensorGradFuncObj;
   MatrixInverseExtGradFunc MatrixInverseExtGradFuncObj;
   RotaryPositionEmbeddingGradGradFunc RotaryPositionEmbeddingGradGradFuncObj;
   EyeGradFunc EyeGradFuncObj;
@@ -1093,6 +1183,7 @@ struct OpsAutoGradRegisters {
   InnerIndexGradFunc InnerIndexGradFuncObj;
   GLUGradFunc GLUGradFuncObj;
   InplaceCopyGradFunc InplaceCopyGradFuncObj;
+  DistCommAllReduceGradFunc DistCommAllReduceGradFuncObj;
   ReplicationPad3DGradFunc ReplicationPad3DGradFuncObj;
   Log10GradFunc Log10GradFuncObj;
   SeLUExtGradFunc SeLUExtGradFuncObj;
@@ -1108,6 +1199,7 @@ struct OpsAutoGradRegisters {
   CopyGradFunc CopyGradFuncObj;
   SoftmaxGradFunc SoftmaxGradFuncObj;
   StdMeanGradFunc StdMeanGradFuncObj;
+  DistCommBarrierGradFunc DistCommBarrierGradFuncObj;
   SoftShrinkGradFunc SoftShrinkGradFuncObj;
   AddcmulExtGradFunc AddcmulExtGradFuncObj;
   SiLUGradGradFunc SiLUGradGradFuncObj;
@@ -1115,6 +1207,7 @@ struct OpsAutoGradRegisters {
   MatmulReduceScatterGradFunc MatmulReduceScatterGradFuncObj;
   SliceExtGradFunc SliceExtGradFuncObj;
   SelectV2GradFunc SelectV2GradFuncObj;
+  DistCommScatterTensorGradFunc DistCommScatterTensorGradFuncObj;
   ThresholdGradGradFunc ThresholdGradGradFuncObj;
   DivsGradFunc DivsGradFuncObj;
   BinaryCrossEntropyWithLogitsBackwardGradFunc BinaryCrossEntropyWithLogitsBackwardGradFuncObj;
@@ -1150,6 +1243,7 @@ struct OpsAutoGradRegisters {
   OnesLikeExtGradFunc OnesLikeExtGradFuncObj;
   DropoutGenMaskExtGradFunc DropoutGenMaskExtGradFuncObj;
   AddcdivExtGradFunc AddcdivExtGradFuncObj;
+  MoeDistributeCombineGradFunc MoeDistributeCombineGradFuncObj;
   DropoutDoMaskExtGradFunc DropoutDoMaskExtGradFuncObj;
   InplaceMulGradFunc InplaceMulGradFuncObj;
   BitwiseNotGradFunc BitwiseNotGradFuncObj;
@@ -1160,12 +1254,14 @@ struct OpsAutoGradRegisters {
   InnerInplaceIndexPutGradFunc InnerInplaceIndexPutGradFuncObj;
   ConvolutionStrGradFunc ConvolutionStrGradFuncObj;
   DotGradFunc DotGradFuncObj;
+  DistCommAllToAllVGradFunc DistCommAllToAllVGradFuncObj;
   TruncGradFunc TruncGradFuncObj;
   AddLayerNormGradGradFunc AddLayerNormGradGradFuncObj;
   Conv2DPaddingGradFunc Conv2DPaddingGradFuncObj;
   GatherDGradFunc GatherDGradFuncObj;
   MatMulExtGradFunc MatMulExtGradFuncObj;
   UpsampleBilinear2DGradGradFunc UpsampleBilinear2DGradGradFuncObj;
+  DistCommBroadcastGradFunc DistCommBroadcastGradFuncObj;
   MaskedFillGradFunc MaskedFillGradFuncObj;
   ExpGradFunc ExpGradFuncObj;
   TensorScatterElementsGradFunc TensorScatterElementsGradFuncObj;
@@ -1187,8 +1283,10 @@ struct OpsAutoGradRegisters {
   LogSoftmaxExtGradFunc LogSoftmaxExtGradFuncObj;
   SqrtGradFunc SqrtGradFuncObj;
   SigmoidGradGradFunc SigmoidGradGradFuncObj;
+  DistCommReduceGradFunc DistCommReduceGradFuncObj;
   ReflectionPad3DGradGradFunc ReflectionPad3DGradGradFuncObj;
   PowScalarTensorGradFunc PowScalarTensorGradFuncObj;
+  EmptyGradFunc EmptyGradFuncObj;
   InplaceReLUGradFunc InplaceReLUGradFuncObj;
   MaxPoolWithIndicesGradFunc MaxPoolWithIndicesGradFuncObj;
   HSigmoidGradFunc HSigmoidGradFuncObj;
@@ -1196,15 +1294,18 @@ struct OpsAutoGradRegisters {
   UniqueConsecutiveGradFunc UniqueConsecutiveGradFuncObj;
   UniformExtGradFunc UniformExtGradFuncObj;
   GroupNormGradGradFunc GroupNormGradGradFuncObj;
+  InnerCommAllReduceGradFunc InnerCommAllReduceGradFuncObj;
   BitwiseOrScalarGradFunc BitwiseOrScalarGradFuncObj;
   Conv3DPaddingGradFunc Conv3DPaddingGradFuncObj;
   AddLayerNormV2GradFunc AddLayerNormV2GradFuncObj;
   GridSampler2DGradFunc GridSampler2DGradFuncObj;
   MaxPoolGradWithIndicesGradFunc MaxPoolGradWithIndicesGradFuncObj;
   Conv1DPaddingGradFunc Conv1DPaddingGradFuncObj;
+  DistCommIsendGradFunc DistCommIsendGradFuncObj;
   SoftplusGradExtGradFunc SoftplusGradExtGradFuncObj;
   GeLUGradGradFunc GeLUGradGradFuncObj;
   NormalTensorTensorGradFunc NormalTensorTensorGradFuncObj;
+  DistCommIrecvGradFunc DistCommIrecvGradFuncObj;
   MaskedSelectGradFunc MaskedSelectGradFuncObj;
   MSELossGradExtGradFunc MSELossGradExtGradFuncObj;
   SwigluGradGradFunc SwigluGradGradFuncObj;
@@ -1218,7 +1319,6 @@ struct OpsAutoGradRegisters {
   HistcExtGradFunc HistcExtGradFuncObj;
   SmoothL1LossGradGradFunc SmoothL1LossGradGradFuncObj;
   LessGradFunc LessGradFuncObj;
-  MoeTokenUnpermuteGradFunc MoeTokenUnpermuteGradFuncObj;
   SignGradFunc SignGradFuncObj;
   TriangularSolveGradFunc TriangularSolveGradFuncObj;
   PReLUGradGradFunc PReLUGradGradFuncObj;
@@ -1260,15 +1360,18 @@ struct OpsAutoGradRegisters {
   IndexAddExtGradFunc IndexAddExtGradFuncObj;
   UpsampleNearest3DGradFunc UpsampleNearest3DGradFuncObj;
   UpsampleLinear1DGradFunc UpsampleLinear1DGradFuncObj;
+  ApplyRotaryPosEmbGradFunc ApplyRotaryPosEmbGradFuncObj;
   SoftMarginLossGradFunc SoftMarginLossGradFuncObj;
   GeluGradExtGradFunc GeluGradExtGradFuncObj;
   ConcatGradFunc ConcatGradFuncObj;
+  DistCommReduceScatterGradFunc DistCommReduceScatterGradFuncObj;
   AcosExtGradFunc AcosExtGradFuncObj;
   IndexGradFunc IndexGradFuncObj;
   AddmmGradFunc AddmmGradFuncObj;
   OnesGradFunc OnesGradFuncObj;
   ScatterGradFunc ScatterGradFuncObj;
   ChunkGradFunc ChunkGradFuncObj;
+  DistCommAllGatherIntoTensorGradFunc DistCommAllGatherIntoTensorGradFuncObj;
   ReflectionPad2DGradFunc ReflectionPad2DGradFuncObj;
   InplaceIndexPutGradFunc InplaceIndexPutGradFuncObj;
   UpsampleLinear1DGradGradFunc UpsampleLinear1DGradGradFuncObj;
@@ -1300,6 +1403,7 @@ struct OpsAutoGradRegisters {
   RandLikeExtGradFunc RandLikeExtGradFuncObj;
   XLogYScalarOtherGradFunc XLogYScalarOtherGradFuncObj;
   IsCloseGradFunc IsCloseGradFuncObj;
+  InnerCommAllGatherGradFunc InnerCommAllGatherGradFuncObj;
   MeshgridGradFunc MeshgridGradFuncObj;
   NonZeroExtGradFunc NonZeroExtGradFuncObj;
   MaximumGradFunc MaximumGradFuncObj;
@@ -1313,6 +1417,7 @@ struct OpsAutoGradRegisters {
   UpsampleTrilinear3DGradGradFunc UpsampleTrilinear3DGradGradFuncObj;
   RandnLikeGradFunc RandnLikeGradFuncObj;
   UpsampleBicubic2DGradGradFunc UpsampleBicubic2DGradGradFuncObj;
+  PagedAttentionGradFunc PagedAttentionGradFuncObj;
   MultiScaleDeformableAttnGradFunc MultiScaleDeformableAttnGradFuncObj;
   InplaceAddsExtGradFunc InplaceAddsExtGradFuncObj;
   PolarGradFunc PolarGradFuncObj;
@@ -1329,17 +1434,21 @@ struct OpsAutoGradRegisters {
   ErfcGradFunc ErfcGradFuncObj;
   InplaceDivsGradFunc InplaceDivsGradFuncObj;
   CloneGradFunc CloneGradFuncObj;
+  DistCommScatterGradFunc DistCommScatterGradFuncObj;
   ArgMinExtGradFunc ArgMinExtGradFuncObj;
   MulGradFunc MulGradFuncObj;
   AdaptiveAvgPool3DExtGradFunc AdaptiveAvgPool3DExtGradFuncObj;
   InplaceZeroGradFunc InplaceZeroGradFuncObj;
+  InplaceSiLUGradFunc InplaceSiLUGradFuncObj;
   RepeatInterleaveTensorGradFunc RepeatInterleaveTensorGradFuncObj;
   RotaryPositionEmbeddingGradFunc RotaryPositionEmbeddingGradFuncObj;
   GcdGradFunc GcdGradFuncObj;
   OuterGradFunc OuterGradFuncObj;
+  DistCommReduceScatterTensorGradFunc DistCommReduceScatterTensorGradFuncObj;
   SplitTensorGradFunc SplitTensorGradFuncObj;
   AsStridedGradFunc AsStridedGradFuncObj;
   SilentCheckV2GradFunc SilentCheckV2GradFuncObj;
+  DistCommAllToAllVSingleGradFunc DistCommAllToAllVSingleGradFuncObj;
   AsinExtGradFunc AsinExtGradFuncObj;
   RepeatInterleaveIntGradFunc RepeatInterleaveIntGradFuncObj;
   NarrowViewGradFunc NarrowViewGradFuncObj;
@@ -1352,10 +1461,12 @@ struct OpsAutoGradRegisters {
   RmsNormGradFunc RmsNormGradFuncObj;
   ConvTranspose2DGradFunc ConvTranspose2DGradFuncObj;
   RollGradFunc RollGradFuncObj;
+  InnerCommIsendGradFunc InnerCommIsendGradFuncObj;
   StackExtGradFunc StackExtGradFuncObj;
   MeanExtGradFunc MeanExtGradFuncObj;
   Expm1GradFunc Expm1GradFuncObj;
   InplaceFloorDivideGradFunc InplaceFloorDivideGradFuncObj;
+  NewEmptyGradFunc NewEmptyGradFuncObj;
   MaxPoolWithMaskGradFunc MaxPoolWithMaskGradFuncObj;
   ZerosLikeExtGradFunc ZerosLikeExtGradFuncObj;
   BatchNormStatsGradFunc BatchNormStatsGradFuncObj;
@@ -1379,11 +1490,13 @@ struct OpsAutoGradRegisters {
   UpsampleBilinear2DGradFunc UpsampleBilinear2DGradFuncObj;
   IsInfGradFunc IsInfGradFuncObj;
   TanGradFunc TanGradFuncObj;
+  InnerCommReduceScatterGradFunc InnerCommReduceScatterGradFuncObj;
   EqualGradFunc EqualGradFuncObj;
   RmsNormGradGradFunc RmsNormGradGradFuncObj;
   ContiguousGradFunc ContiguousGradFuncObj;
   GreaterEqualGradFunc GreaterEqualGradFuncObj;
   MinDimGradFunc MinDimGradFuncObj;
+  InnerMoeTokenUnpermuteGradFunc InnerMoeTokenUnpermuteGradFuncObj;
   InplaceFillDiagonalGradFunc InplaceFillDiagonalGradFuncObj;
   NanToNumGradFunc NanToNumGradFuncObj;
   SpeedFusionAttentionGradGradFunc SpeedFusionAttentionGradGradFuncObj;
@@ -1431,6 +1544,7 @@ struct OpsAutoGradRegisters {
   TopkExtGradFunc TopkExtGradFuncObj;
   InplaceEluGradFunc InplaceEluGradFuncObj;
   LinalgVectorNormGradFunc LinalgVectorNormGradFuncObj;
+  InnerCommAllToAllVGradFunc InnerCommAllToAllVGradFuncObj;
   XLogYScalarSelfGradFunc XLogYScalarSelfGradFuncObj;
   DivModsGradFunc DivModsGradFuncObj;
   InplaceUniformGradFunc InplaceUniformGradFuncObj;
@@ -1438,10 +1552,12 @@ struct OpsAutoGradRegisters {
   SigmoidGradFunc SigmoidGradFuncObj;
   MultinomialExtGradFunc MultinomialExtGradFuncObj;
   InplaceDivModGradFunc InplaceDivModGradFuncObj;
+  InnerCommIrecvGradFunc InnerCommIrecvGradFuncObj;
   PowTensorScalarGradFunc PowTensorScalarGradFuncObj;
   FloorGradFunc FloorGradFuncObj;
   SplitWithSizeGradFunc SplitWithSizeGradFuncObj;
   InplaceTanhGradFunc InplaceTanhGradFuncObj;
+  DistCommGatherGradFunc DistCommGatherGradFuncObj;
   ReflectionPad1DGradGradFunc ReflectionPad1DGradGradFuncObj;
   ErfGradFunc ErfGradFuncObj;
   BincountExtGradFunc BincountExtGradFuncObj;
@@ -1459,6 +1575,7 @@ struct OpsAutoGradRegisters {
   GluGradGradFunc GluGradGradFuncObj;
   GeLUGradFunc GeLUGradFuncObj;
   AddbmmGradFunc AddbmmGradFuncObj;
+  DistCommBatchIsendIrecvGradFunc DistCommBatchIsendIrecvGradFuncObj;
   SelectGradFunc SelectGradFuncObj;
   InplaceErfinvGradFunc InplaceErfinvGradFuncObj;
   NegGradFunc NegGradFuncObj;
@@ -1467,11 +1584,13 @@ struct OpsAutoGradRegisters {
   Col2ImGradGradFunc Col2ImGradGradFuncObj;
   ReduceMaxGradFunc ReduceMaxGradFuncObj;
   UpsampleNearest1DGradGradFunc UpsampleNearest1DGradGradFuncObj;
+  InplaceMatmulAddGradFunc InplaceMatmulAddGradFuncObj;
   L1LossBackwardExtGradFunc L1LossBackwardExtGradFuncObj;
   BinaryCrossEntropyGradGradFunc BinaryCrossEntropyGradGradFuncObj;
   IncreFlashAttentionGradFunc IncreFlashAttentionGradFuncObj;
   FlashAttentionScoreGradGradFunc FlashAttentionScoreGradGradFuncObj;
   IdentityGradFunc IdentityGradFuncObj;
+  MoeDistributeDispatchGradFunc MoeDistributeDispatchGradFuncObj;
   PowGradFunc PowGradFuncObj;
   ExpandAsGradFunc ExpandAsGradFuncObj;
   UpsampleNearest2DGradFunc UpsampleNearest2DGradFuncObj;
@@ -1482,6 +1601,7 @@ struct OpsAutoGradRegisters {
   FlashAttentionScoreGradFunc FlashAttentionScoreGradFuncObj;
   UnstackExtViewGradFunc UnstackExtViewGradFuncObj;
   HSigmoidGradGradFunc HSigmoidGradGradFuncObj;
+  QuantMatmulGradFunc QuantMatmulGradFuncObj;
   DynamicQuantExtGradFunc DynamicQuantExtGradFuncObj;
   KVCacheScatterUpdateGradFunc KVCacheScatterUpdateGradFuncObj;
   WeightQuantBatchMatmulGradFunc WeightQuantBatchMatmulGradFuncObj;
@@ -1494,17 +1614,23 @@ struct OpsAutoGradRegisters {
   GroupedMatmulGradFunc GroupedMatmulGradFuncObj;
   QuantBatchMatmulGradFunc QuantBatchMatmulGradFuncObj;
   FusedInferAttentionScoreGradFunc FusedInferAttentionScoreGradFuncObj;
+  MoeInitRoutingQuantV2GradFunc MoeInitRoutingQuantV2GradFuncObj;
   MoeInitRoutingGradFunc MoeInitRoutingGradFuncObj;
   MoeInitRoutingV2GradFunc MoeInitRoutingV2GradFuncObj;
   GroupedMatmulV2GradFunc GroupedMatmulV2GradFuncObj;
   QuantV2GradFunc QuantV2GradFuncObj;
   GmmBackwardFusionGradFunc GmmBackwardFusionGradFuncObj;
+  AnyExtGradFunc AnyExtGradFuncObj;
   GmmV2BackwardFusionGradFunc GmmV2BackwardFusionGradFuncObj;
   GmmV2GradFunc GmmV2GradFuncObj;
   GmmBackwardGradFunc GmmBackwardGradFuncObj;
   GmmGradFunc GmmGradFuncObj;
+  MoeTokenUnpermuteGradFunc MoeTokenUnpermuteGradFuncObj;
   GmmV2BackwardGradFunc GmmV2BackwardGradFuncObj;
   PixelShuffleGradFunc PixelShuffleGradFuncObj;
+  AnyGradFunc AnyGradFuncObj;
+  FuncMaxPool2DGradFunc FuncMaxPool2DGradFuncObj;
+  EinsumExtGradFunc EinsumExtGradFuncObj;
   InplaceExponentialGradFunc InplaceExponentialGradFuncObj;
 };
 }  // namespace pyboost