#include "op_plugin/AclOpsInterface.h"
#include "op_plugin/utils/OpAdapter.h"
namespace acl_op {
using npu_preparation = at_npu::native::OpPreparation;
using npu_utils = at_npu::native::NpuUtils;
namespace {
c10::SmallVector<int64_t, SIZE> renorm_npu_output_size(
const at::Tensor& self,
int64_t dim) {
c10::SmallVector<int64_t, SIZE> out_size;
for (int64_t i = 0; i < self.dim(); i++) {
if (i != dim) {
out_size.emplace_back(1);
} else {
out_size.emplace_back(self.sizes()[i]);
}
}
return out_size;
}
at::Tensor& renorm_compute(
at::Tensor& result,
const at::Tensor& self,
at::Scalar p,
int64_t dim,
at::Scalar maxnorm) {
float p_value = op_plugin::utils::get_scalar_float_value(p);
float maxnorm_value = op_plugin::utils::get_scalar_float_value(maxnorm);
at_npu::native::OpCommand cmd;
cmd.Name("Renorm")
.Input(self)
.Output(result)
.Attr("p", p_value)
.Attr("maxnorm", maxnorm_value)
.Attr("dim", dim)
.Run();
return result;
}
at::Tensor& renorm_out_nocheck(
at::Tensor& result,
const at::Tensor& self,
at::Scalar p,
int64_t dim,
at::Scalar maxnorm) {
auto ori_type = self.scalar_type();
if (ori_type != c10::ScalarType::Half && ori_type != c10::ScalarType::Float) {
TORCH_CHECK(false, "Renorm only support float16 or float32 type." + OPS_ERROR(ErrCode::TYPE));
}
TORCH_CHECK(result.scalar_type() == ori_type, "result's type must be equal to input's."
+ OPS_ERROR(ErrCode::TYPE));
dim = op_plugin::utils::make_warp_dim(dim, self.dim());
auto output_size = renorm_npu_output_size(self, dim);
at::Tensor result_bak = npu_preparation::apply_tensor_with_format(
output_size,
self.options().dtype(at::kFloat),
npu_preparation::get_tensor_npu_format(self));
if (ori_type == c10::ScalarType::Half) {
at::Tensor self_no_name = self.rename(c10::nullopt);
at::Tensor result_no_name = result.rename(c10::nullopt);
self_no_name = at_npu::native::custom_ops::_npu_dtype_cast(self_no_name, c10::ScalarType::Float);
result_no_name = at_npu::native::custom_ops::_npu_dtype_cast(result_no_name, c10::ScalarType::Float);
renorm_compute(
result_bak,
self_no_name,
p,
dim,
maxnorm);
at::Tensor result_broadcast = acl_op::npu_broadcast(result_bak, self.sizes());
at::mul_out(result_no_name, result_broadcast, self_no_name);
acl_op::npu_dtype_cast_(result, result_no_name);
} else {
renorm_compute(
result_bak,
self,
p,
dim,
maxnorm);
at::Tensor result_broadcast = acl_op::npu_broadcast(result_bak, self.sizes());
at::mul_out(result, result_broadcast, self);
}
return result;
}
}
at::Tensor& renorm_out(
const at::Tensor& self,
const at::Scalar& p,
int64_t dim,
const at::Scalar& maxnorm,
at::Tensor& result) {
npu_preparation::CheckOut(
{self},
result,
self);
if (!npu_utils::check_match(&self)) {
at::Tensor contiguous_self = npu_utils::format_contiguous(self);
renorm_out_nocheck(contiguous_self, contiguous_self, p, dim, maxnorm);
npu_utils::format_fresh_view(result, contiguous_self);
} else {
renorm_out_nocheck(result, self, p, dim, maxnorm);
}
return result;
}
at::Tensor renorm(const at::Tensor& self, const at::Scalar& p, int64_t dim, const at::Scalar& maxnorm) {
at::Tensor result = npu_preparation::apply_tensor(self);
renorm_out_nocheck(result, self, p, dim, maxnorm);
return result;
}
at::Tensor& renorm_(at::Tensor& self, const at::Scalar& p, int64_t dim, const at::Scalar& maxnorm) {
return acl_op::renorm_out(self, p, dim, maxnorm, self);
}
}