未验证 提交 4629401e 编写于 作者: C crystal 提交者: GitHub

[cherry-pick] FixEighOP; Unified MatrixEighFunctor function (#35812) (#35919)

cherry-pick #35812,修复Eigh OP
上级 91f25ee3
......@@ -47,13 +47,10 @@ class EighOp : public framework::OperatorWithKernel {
input_dim[rank - 2], input_dim[rank - 1]));
std::vector<int64_t> values_dim;
if (rank > 2) {
for (auto i = 0; i < rank - 1; i++) {
values_dim.emplace_back(input_dim[i]);
}
} else {
values_dim = {input_dim[1]};
}
ctx->SetOutputDim("Eigenvalues", framework::make_ddim(values_dim));
ctx->SetOutputDim("Eigenvectors", input_dim);
......@@ -99,9 +96,9 @@ class EighGradOp : public framework::OperatorWithKernel {
"EighGrad");
OP_INOUT_CHECK(ctx->HasInput("Eigenvectors"), "Input", "Eigenvectors",
"EighGrad");
OP_INOUT_CHECK(ctx->HasInputs(framework::GradVarName("Eigenvalues")),
OP_INOUT_CHECK(ctx->HasInput(framework::GradVarName("Eigenvalues")),
"Input", "Eigenvalues@GRAD", "EighGrad");
OP_INOUT_CHECK(ctx->HasInputs(framework::GradVarName("Eigenvectors")),
OP_INOUT_CHECK(ctx->HasInput(framework::GradVarName("Eigenvectors")),
"Input", "Eigenvectors@GRAD", "EighGrad");
auto dims = ctx->GetInputDim("Eigenvectors");
auto x_grad_name = framework::GradVarName("X");
......
......@@ -14,34 +14,14 @@ limitations under the License. */
#include "paddle/fluid/operators/eigh_op.h"
namespace paddle {
namespace operators {
using Tensor = framework::Tensor;
template <typename ValueType, typename T>
class EighGPUKernel : public framework::OpKernel<T> {
public:
void Compute(const framework::ExecutionContext &ctx) const override {
auto input_var = ctx.Input<Tensor>("X");
auto output_w_var = ctx.Output<Tensor>("Eigenvalues");
auto output_v_var = ctx.Output<Tensor>("Eigenvectors");
std::string lower = ctx.Attr<std::string>("UPLO");
bool is_lower = (lower == "L");
math::MatrixEighFunctor<ValueType, T> functor;
functor(ctx, *input_var, output_w_var, output_v_var, is_lower, true);
}
};
} // namespace operators
} // namespace paddle
namespace ops = paddle::operators;
REGISTER_OP_CUDA_KERNEL(
eigh, ops::EighGPUKernel<float, float>, ops::EighGPUKernel<double, double>,
ops::EighGPUKernel<float, paddle::platform::complex<float>>,
ops::EighGPUKernel<double, paddle::platform::complex<double>>);
eigh, ops::EighKernel<paddle::platform::CUDADeviceContext, float, float>,
ops::EighKernel<paddle::platform::CUDADeviceContext, double, double>,
ops::EighKernel<paddle::platform::CUDADeviceContext, float,
paddle::platform::complex<float>>,
ops::EighKernel<paddle::platform::CUDADeviceContext, double,
paddle::platform::complex<double>>);
REGISTER_OP_CUDA_KERNEL(
eigh_grad,
......
......@@ -22,24 +22,17 @@ namespace operators {
using Tensor = framework::Tensor;
template <typename T, size_t D, int MajorType = Eigen::RowMajor,
typename IndexType = Eigen::DenseIndex>
using EigenTensor = framework::EigenTensor<T, D, MajorType, IndexType>;
template <typename T, int MajorType = Eigen::RowMajor,
typename IndexType = Eigen::DenseIndex>
using EigenVector = framework::EigenVector<T, MajorType, IndexType>;
template <typename DeviceContext, typename ValueType, typename T>
class EighKernel : public framework::OpKernel<T> {
public:
void Compute(const framework::ExecutionContext& ctx) const override {
auto input_var = ctx.Input<Tensor>("X");
auto output_w_var = ctx.Output<Tensor>("Eigenvalues");
auto output_v_var = ctx.Output<Tensor>("Eigenvectors");
auto input = ctx.Input<Tensor>("X");
auto output_w = ctx.Output<Tensor>("Eigenvalues");
auto output_v = ctx.Output<Tensor>("Eigenvectors");
std::string lower = ctx.Attr<std::string>("UPLO");
bool is_lower = (lower == "L");
math::MatrixEighFunctorCPU<DeviceContext, ValueType, T> functor;
functor(ctx, *input_var, output_w_var, output_v_var, is_lower, true);
math::MatrixEighFunctor<DeviceContext, ValueType, T> functor;
functor(ctx, *input, output_w, output_v, is_lower, true);
}
};
......@@ -49,30 +42,30 @@ class EighGradKernel : public framework::OpKernel<T> {
void Compute(const framework::ExecutionContext& ctx) const override {
auto& x_grad = *ctx.Output<framework::Tensor>(framework::GradVarName("X"));
x_grad.mutable_data<T>(ctx.GetPlace());
auto& output_w_var = *ctx.Input<Tensor>("Eigenvalues");
auto& output_v_var = *ctx.Input<Tensor>("Eigenvectors");
auto& output_w = *ctx.Input<Tensor>("Eigenvalues");
auto& output_v = *ctx.Input<Tensor>("Eigenvectors");
auto& output_w_grad =
*ctx.Input<Tensor>(framework::GradVarName("Eigenvalues"));
auto& output_v_grad =
*ctx.Input<Tensor>(framework::GradVarName("Eigenvectors"));
auto& dims = output_v_var.dims();
auto& dims = output_v.dims();
const int m = dims[dims.size() - 1];
auto dito =
math::DeviceIndependenceTensorOperations<DeviceContext, T, ValueType>(
ctx);
auto tV = dito.Transpose(dito.Conj(output_v_var));
auto W = dito.Sub_(dito.Unsqueeze(output_w_var, -2),
dito.Unsqueeze(output_w_var, -1));
auto tV = dito.Transpose(dito.Conj(output_v));
auto W = dito.template Sub<ValueType>(dito.Unsqueeze(output_w, -2),
dito.Unsqueeze(output_w, -1));
Tensor result = dito.Matmul(tV, output_v_grad);
result.mutable_data<T>(dims, ctx.GetPlace());
std::vector<int> out_shape = framework::vectorize<int>(dims);
auto constant = dito.Fill(out_shape, 0.5);
result = dito.Sub(result, dito.Conj(dito.Transpose(result)));
result = dito.Mul(result, constant);
result = dito.Div_(result, W);
result = dito.Div(result, W);
result = dito.DiagFill(m, m, m, 0, output_w_grad, result);
x_grad = dito.Matmul(output_v_var, dito.Matmul(result, tV));
x_grad = dito.Matmul(output_v, dito.Matmul(result, tV));
}
};
......
......@@ -16,7 +16,6 @@
#include "Eigen/Core"
#include "paddle/fluid/memory/memory.h"
#include "paddle/fluid/operators/math/complex_functors.h"
#include "paddle/fluid/operators/svd_helper.h"
#ifdef PADDLE_WITH_CUDA
#include "paddle/fluid/platform/dynload/cusolver.h"
......@@ -26,10 +25,6 @@ namespace paddle {
namespace operators {
namespace math {
template <typename T, size_t D, int MajorType = Eigen::RowMajor,
typename IndexType = Eigen::DenseIndex>
using EigenTensor = framework::EigenTensor<T, D, MajorType, IndexType>;
template <typename T, int MajorType = Eigen::RowMajor,
typename IndexType = Eigen::DenseIndex>
using InputMatrixMap = Eigen::Map<
......@@ -67,7 +62,7 @@ inline void ComputeFloatEigenvaluesAndVectors(ValueType *x_data,
eigenvalues = eigen_solver.eigenvalues().transpose();
if (has_vectors) {
eigenvectors = eigen_solver.eigenvectors().transpose();
eigenvectors = eigen_solver.eigenvectors();
}
}
}
......@@ -103,7 +98,7 @@ inline void ComputeComplexEigenvaluesAndVectors(T *x_data,
eigenvalues = eigen_solver.eigenvalues().transpose();
if (has_vectors) {
eigenvectors = eigen_solver.eigenvectors().transpose();
eigenvectors = eigen_solver.eigenvectors();
}
}
}
......@@ -117,11 +112,18 @@ inline int64_t GetBatchSize(framework::DDim dims) {
return batch_size;
}
template <typename DeviceContext, typename ValueType, typename T>
struct MatrixEighFunctor {
void operator()(const framework::ExecutionContext &ctx, const Tensor &input,
Tensor *eigen_values, Tensor *eigen_vectors, bool is_lower,
bool has_vectors);
};
// Calculates the eigenvalues ​​and eigenvectors of Hermitian or real
// symmetric matrices, and uses the variable has_vectors to
// control whether to return the eigenvectors.
template <typename DeviceContext, typename ValueType, typename T>
struct MatrixEighFunctorCPU {
template <typename ValueType, typename T>
struct MatrixEighFunctor<platform::CPUDeviceContext, ValueType, T> {
public:
void operator()(const framework::ExecutionContext &ctx, const Tensor &input,
Tensor *eigen_values, Tensor *eigen_vectors, bool is_lower,
......@@ -134,7 +136,8 @@ struct MatrixEighFunctorCPU {
for (int64_t i = 0; i < dim_size - 2; i++) {
batch_size *= dims[i];
}
auto dito = DeviceIndependenceTensorOperations<DeviceContext, T>(ctx);
auto dito =
DeviceIndependenceTensorOperations<platform::CPUDeviceContext, T>(ctx);
Tensor input_tensor;
TensorCopy(input, ctx.GetPlace(), &input_tensor);
if (!is_lower) {
......@@ -157,9 +160,6 @@ struct MatrixEighFunctorCPU {
ComputeFloatEigenvaluesAndVectors<ValueType>(
x_data, value_data, vector_data, batch_size, rows, rows, has_vectors);
}
if (has_vectors) {
*eigen_vectors = dito.Transpose(*eigen_vectors);
}
}
};
......@@ -169,7 +169,7 @@ struct MatrixEighFunctorCPU {
// symmetric matrices on GPU, and uses the variable has_vectors
// to control whether to return the eigenvectors.
template <typename ValueType, typename T>
struct MatrixEighFunctor {
struct MatrixEighFunctor<platform::CUDADeviceContext, ValueType, T> {
public:
void operator()(const framework::ExecutionContext &ctx, const Tensor &input,
Tensor *eigen_values, Tensor *eigen_vectors, bool is_lower,
......@@ -278,7 +278,8 @@ struct MatrixEighFunctor {
#define EVDBUFFER_INSTANCE(ValueType, T, C, CastType) \
template <> \
inline void MatrixEighFunctor<ValueType, T>::EvdBuffer( \
inline void \
MatrixEighFunctor<platform::CUDADeviceContext, ValueType, T>::EvdBuffer( \
cusolverDnHandle_t handle, cusolverEigMode_t jobz, \
cublasFillMode_t uplo, int n, const T *A, int lda, const ValueType *W, \
int *lwork) const { \
......@@ -292,7 +293,8 @@ FUNC_WITH_TYPES(EVDBUFFER_INSTANCE);
#define EVD_INSTANCE(ValueType, T, C, CastType) \
template <> \
inline void MatrixEighFunctor<ValueType, T>::Evd( \
inline void \
MatrixEighFunctor<platform::CUDADeviceContext, ValueType, T>::Evd( \
cusolverDnHandle_t handle, cusolverEigMode_t jobz, \
cublasFillMode_t uplo, int n, T *A, int lda, ValueType *W, T *work, \
int lwork, int *devInfo) const { \
......
......@@ -289,10 +289,20 @@ struct DeviceIndependenceTensorOperations {
framework::Tensor Div(const framework::Tensor& x,
const framework::Tensor& y) {
framework::Tensor ret;
if (x.type() != y.type()) {
ret.mutable_data<T>(x.dims(), context.GetPlace());
auto x_vector = EigenVector<T>::Flatten(x);
auto y_vector = EigenVector<ValueType>::Flatten(y);
auto out_vector = EigenVector<T>::Flatten(ret);
auto& place =
*context.template device_context<DeviceContext>().eigen_device();
out_vector.device(place) = x_vector / y_vector;
} else {
std::vector<int> out_shape = GetBroadcastShape({&x, &y});
ret.Resize(framework::make_ddim(out_shape));
ElementwiseComputeEx<DivFunctor<T>, DeviceContext, T>(
context, &x, &y, -1, DivFunctor<T>(), &ret);
}
return ret;
}
framework::Tensor Add(const framework::Tensor& x,
......@@ -330,7 +340,8 @@ struct DeviceIndependenceTensorOperations {
NameInTensorMap inputs({{"X", {&x}}});
return CreateOpRunAndReturnTensor("reduce_max", inputs, attrs, out_dim);
}
// Support float and complex type subtraction,the default is T type
template <typename InT = T>
framework::Tensor Sub(const framework::Tensor& x,
const framework::Tensor& y) {
framework::Tensor ret;
......@@ -340,18 +351,18 @@ struct DeviceIndependenceTensorOperations {
#if defined(__NVCC__) || defined(__HIPCC__)
// For GPU, there is no need to define XxxInverseFunctor and call
// ElementwiseComputeEx in two branches.
ElementwiseComputeEx<SubFunctor<T>, DeviceContext, T>(
context, &x, &y, -1, SubFunctor<T>(), &ret);
ElementwiseComputeEx<SubFunctor<InT>, DeviceContext, InT>(
context, &x, &y, -1, SubFunctor<InT>(), &ret);
#endif
} else {
if (x.dims().size() >= y.dims().size()) {
ElementwiseComputeEx<SubFunctor<T>, DeviceContext, T>(
context, &x, &y, -1, SubFunctor<T>(), &ret);
ElementwiseComputeEx<SubFunctor<InT>, DeviceContext, InT>(
context, &x, &y, -1, SubFunctor<InT>(), &ret);
} else {
ElementwiseComputeEx<InverseSubFunctor<T>, DeviceContext, T>(
// This is copyed from elementwise_sub, which means we
// need reverse will xrank < yrank
context, &x, &y, -1, InverseSubFunctor<T>(), &ret);
ElementwiseComputeEx<InverseSubFunctor<InT>, DeviceContext, InT>(
context, &x, &y, -1, InverseSubFunctor<InT>(), &ret);
}
}
return ret;
......@@ -461,37 +472,6 @@ struct DeviceIndependenceTensorOperations {
return out;
}
// Support x and y are different data types
Tensor Div_(const Tensor& x, const Tensor& y) {
Tensor out;
out.mutable_data<T>(x.dims(), context.GetPlace());
auto x_vector = EigenVector<T>::Flatten(x);
auto y_vector = EigenVector<ValueType>::Flatten(y);
auto out_vector = EigenVector<T>::Flatten(out);
auto& place =
*context.template device_context<DeviceContext>().eigen_device();
out_vector.device(place) = x_vector / y_vector;
return out;
}
framework::Tensor Sub_(const framework::Tensor& x,
const framework::Tensor& y) {
framework::Tensor ret;
std::vector<int> out_shape = GetBroadcastShape({&x, &y});
ret.Resize(framework::make_ddim(out_shape));
if (x.dims().size() >= y.dims().size()) {
ElementwiseComputeEx<SubFunctor<ValueType>, DeviceContext, ValueType>(
context, &x, &y, -1, SubFunctor<ValueType>(), &ret);
} else {
ElementwiseComputeEx<InverseSubFunctor<ValueType>, DeviceContext,
ValueType>(
// This is copyed from elementwise_sub, which means we
// need reverse will xrank < yrank
context, &x, &y, -1, InverseSubFunctor<ValueType>(), &ret);
}
return ret;
}
private:
const framework::ExecutionContext& context;
BlasT<DeviceContext, T> GetBlas() {
......
......@@ -140,7 +140,7 @@ class TestEighAPI(unittest.TestCase):
self.check_static_complex_result()
def test_in_dynamic_mode(self):
paddle.disable_static(self.place)
paddle.disable_static()
input_real_data = paddle.to_tensor(self.real_data)
expected_w, expected_v = np.linalg.eigh(self.real_data)
actual_w, actual_v = paddle.linalg.eigh(input_real_data)
......@@ -152,7 +152,7 @@ class TestEighAPI(unittest.TestCase):
self.compare_result(actual_w, actual_v.numpy(), expected_w, expected_v)
def test_eigh_grad(self):
paddle.disable_static(self.place)
paddle.disable_static()
x = paddle.to_tensor(self.complex_data, stop_gradient=False)
w, v = paddle.linalg.eigh(x)
(w.sum() + paddle.abs(v).sum()).backward()
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册