p85971406
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Bug fix

This commit is contained in:
panzhihui 2023-02-06 14:37:47 +08:00
parent e0915062cf
commit 2f8db25bb7
4 changed files with 13 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
mindspore/ccsrc/plugin/device/ascend/kernel/aicpu/aicpu_ops/cpu_kernel/ms_kernel/matrix_exp.cc
View File

@ -124,8 +124,8 @@ uint32_t MatrixExpCpuKernel::MatrixExpCheck(CpuKernelContext &ctx) {
}
template <typename Derived1, typename Derived2, typename Derived3>
void MatrixExpCpuKernel::MTaylorApproximant(const Eigen::MatrixBase<Derived1> &A, const Eigen::MatrixBase<Derived2> &I,
int order, Eigen::MatrixBase<Derived3> &E) {
void MatrixExpCpuKernel::MTaylorApproximant(Eigen::MatrixBase<Derived1> &A, Eigen::MatrixBase<Derived2> &I, int order,
Eigen::MatrixBase<Derived3> &E) {
constexpr int expension_order_1 = 1;
constexpr int expension_order_2 = 2;
constexpr int expension_order_4 = 4;
@ -167,7 +167,7 @@ void MatrixExpCpuKernel::MTaylorApproximant(const Eigen::MatrixBase<Derived1> &A
}
template <typename Derived1, typename Derived2>
void MatrixExpCpuKernel::MexpImpl(const Eigen::MatrixBase<Derived1> &A, const Eigen::MatrixBase<Derived2> &I,
void MatrixExpCpuKernel::MexpImpl(Eigen::MatrixBase<Derived1> &A, Eigen::MatrixBase<Derived2> &I,
Eigen::MatrixBase<Derived1> &mexp, CpuKernelContext &ctx) {
const auto norm = A.cwiseAbs().colwise().sum().maxCoeff();
constexpr std::array<int, total_n_degs> m_vals = {1, 2, 4, 8, 12, 18};
@ -203,8 +203,8 @@ void MatrixExpCpuKernel::MexpImpl(const Eigen::MatrixBase<Derived1> &A, const Ei
}
if (s >= 0) {
const auto pow2s = pow(2, s);
const auto A_scaled = A / pow2s;
MTaylorApproximant(A_scaled, I, m_vals[total_n_degs - 1], mexp);
A /= pow2s;
MTaylorApproximant(A, I, m_vals[total_n_degs - 1], mexp);
for (int k = 0; k < s; k++) {
mexp = mexp * mexp;
}

6
mindspore/ccsrc/plugin/device/ascend/kernel/aicpu/aicpu_ops/cpu_kernel/ms_kernel/matrix_exp.h
View File

@ -31,12 +31,12 @@ class MatrixExpCpuKernel : public CpuKernel {
uint32_t MatrixExpCheck(CpuKernelContext &ctx);
template <typename Derived1, typename Derived2, typename Derived3>
void MTaylorApproximant(const Eigen::MatrixBase<Derived1> &A, const Eigen::MatrixBase<Derived2> &I, int order,
void MTaylorApproximant(Eigen::MatrixBase<Derived1> &A, Eigen::MatrixBase<Derived2> &I, int order,
Eigen::MatrixBase<Derived3> &E);
template <typename Derived1, typename Derived2>
void MexpImpl(const Eigen::MatrixBase<Derived1> &A, const Eigen::MatrixBase<Derived2> &I,
Eigen::MatrixBase<Derived1> &mexp, CpuKernelContext &ctx);
void MexpImpl(Eigen::MatrixBase<Derived1> &A, Eigen::MatrixBase<Derived2> &I, Eigen::MatrixBase<Derived1> &mexp,
CpuKernelContext &ctx);
template <typename T>
uint32_t MatrixExpCompute(CpuKernelContext &ctx);

7
mindspore/core/ops/unravel_index.cc
View File

@ -79,8 +79,11 @@ TypePtr UravelIndexInferType(const PrimitivePtr &prim, const std::vector<Abstrac
MIND_API_OPERATOR_IMPL(UnravelIndex, BaseOperator);
AbstractBasePtr UnravelIndexInfer(const abstract::AnalysisEnginePtr &, const PrimitivePtr &primitive,
const std::vector<AbstractBasePtr> &input_args) {
return std::make_shared<abstract::AbstractTensor>(UravelIndexInferType(primitive, input_args),
UravelIndexInferShape(primitive, input_args));
const int64_t input_num = 2;
CheckAndConvertUtils::CheckInputArgs(input_args, kEqual, input_num, primitive->name());
auto infer_type = UravelIndexInferType(primitive, input_args);
auto infer_shape = UravelIndexInferShape(primitive, input_args);
return std::make_shared<abstract::AbstractTensor>(infer_type, infer_shape);
}
// AG means auto generated

8
mindspore/python/mindspore/ops/operations/nn_ops.py
View File

@ -8667,10 +8667,6 @@ class SparseApplyCenteredRMSProp(Primitive):
``Ascend`` ``GPU`` ``CPU``
Examples:
>>> import numpy as np
>>> from mindspore import Tensor
>>> import mindspore.common.dtype as mstype
>>> import mindspore.ops.operations.nn_ops as nn_ops
>>> var = Tensor(np.array([[0.6, 0.4], [0.1, 0.5]]).astype(np.float32))
>>> mg = Tensor(np.array([[0.1, 0.3], [0.1, 0.5]]).astype(np.float32))
>>> ms = Tensor(np.array([[0.2, 0.1], [0.1, 0.2]]).astype(np.float32))
@ -9761,10 +9757,6 @@ class SparseApplyAdagradDA(Primitive):
``Ascend`` ``GPU`` ``CPU``
Examples:
>>> import numpy as np
>>> from mindspore import Tensor
>>> import mindspore.common.dtype as mstype
>>> import mindspore.ops.operations.nn_ops as nn_ops
>>> var = Parameter(Tensor(np.array([[1,2], [1,2]]).astype(np.float32)))
>>> grad_accum = Parameter(Tensor(np.array([[2,1], [3,1]]).astype(np.float32)))
>>> grad_square_accum = Parameter(Tensor(np.array([[4,1], [5,1]]).astype(np.float32)))