#include "op_plugin/AclOpsInterface.h"
#include "op_plugin/utils/OpAdapter.h"
#include "op_plugin/utils/custom_functions/aclops/inner_compute.h"
namespace acl_op {
using npu_preparation = at_npu::native::OpPreparation;
std::tuple<at::Tensor, at::Tensor, at::Tensor> npu_deep_norm(const at::Tensor& x,
const at::Tensor& gx,
const at::Tensor& beta,
const at::Tensor& gamma,
double alpha,
double epsilon)
{
at::SmallVector<int64_t, SIZE> shape;
auto param_dim = x.dim() - gamma.dim();
for (int64_t index = 0; index < x.dim(); index++) {
if (index < param_dim) {
shape.emplace_back(x.size(index));
} else {
shape.emplace_back(1);
}
}
at::Tensor y = npu_preparation::apply_tensor(x);
at::Tensor mean = npu_preparation::apply_tensor(shape, x.options().dtype(at::kFloat), x);
at::Tensor rstd = npu_preparation::apply_tensor(shape, x.options().dtype(at::kFloat), x);
at_npu::native::OpCommand cmd;
cmd.Name("DeepNorm")
.Input(x, "x")
.Input(gx, "gx")
.Input(beta, "beta")
.Input(gamma, "gamma")
.Output(mean, "mean")
.Output(rstd, "rstd")
.Output(y, "y")
.Attr("alpha", static_cast<float>(alpha))
.Attr("epsilon", static_cast<float>(epsilon))
.Run();
return std::make_tuple(mean, rstd, y);
}
}