#include "op_plugin/AclOpsInterface.h"
#include "op_plugin/OpInterface.h"
#include "op_plugin/utils/op_api_common.h"
namespace op_api {
using npu_preparation = at_npu::native::OpPreparation;
#if VERSION_BETWEEN(V2R1, V2R1)
std::tuple<at::Tensor, at::Tensor> _unique(
const at::Tensor& self,
bool sorted,
bool return_inverse)
{
DO_COMPATIBILITY(aclnnUnique, acl_op::_unique(self, sorted, return_inverse));
at::Tensor y = npu_preparation::apply_tensor_without_format(self, self.numel());
at::Tensor y_inverse = return_inverse
? npu_preparation::apply_tensor_without_format(self.sizes(),
self.options().dtype(at::kLong))
: npu_preparation::apply_tensor_without_format({0}, self.options().dtype(at::kLong));
static auto opApiFuncAddr = []() {
auto ret = GetOpApiFuncAddr("aclGetViewShape");
TORCH_CHECK(ret != nullptr, OPS_ERROR(ErrCode::PTR));
return ret;
}();
using aclGetViewShapeFunc = int (*)(const aclTensor *tensor, int64_t **view_dims, uint64_t *view_dims_num);
auto aclGetViewShape = reinterpret_cast<aclGetViewShapeFunc>(opApiFuncAddr);
OP_EXEC_LOG(aclnnUnique, "EXEC_NPU_CMD_SYNC", self, sorted, return_inverse, y, y_inverse);
auto npuAclParams = EXEC_NPU_CMD_SYNC(aclnnUnique, self, sorted, return_inverse, y, y_inverse);
int64_t *view_dims = nullptr;
uint64_t view_dim_num = 0;
auto ret = aclGetViewShape(npuAclParams.Get<3>(), &view_dims, &view_dim_num);
TORCH_CHECK(ret == 0, "aclGetViewShape failed.", OPS_ERROR(ErrCode::ACL));
c10::SmallVector<int64_t, SIZE> output_size(view_dims, view_dims + view_dim_num);
y.resize_(output_size);
delete[] view_dims;
view_dims = nullptr;
return std::tuple<at::Tensor, at::Tensor>(y, y_inverse);
}
#endif
#if VERSION_BETWEEN(V2R2, VERSION_NEWEST)
std::tuple<at::Tensor, at::Tensor> _unique(
const at::Tensor& self,
bool sorted,
bool return_inverse)
{
DO_COMPATIBILITY(aclnnUnique, acl_op::_unique(self, sorted, return_inverse));
at::Tensor y = npu_preparation::apply_tensor_without_format(self, self.numel());
at::Tensor y_inverse = return_inverse
? npu_preparation::apply_tensor_without_format(self.sizes(),
self.options().dtype(at::kLong))
: npu_preparation::apply_tensor_without_format({0}, self.options().dtype(at::kLong));
static auto opApiFuncAddr = []() {
auto ret = GetOpApiFuncAddr("aclGetViewShape");
TORCH_CHECK(ret != nullptr, OPS_ERROR(ErrCode::PTR));
return ret;
}();
using aclGetViewShapeFunc = int (*)(const aclTensor *tensor, int64_t **view_dims, uint64_t *view_dims_num);
auto aclGetViewShape = reinterpret_cast<aclGetViewShapeFunc>(opApiFuncAddr);
sorted = true;
OP_EXEC_LOG(aclnnUnique, "EXEC_NPU_CMD_SYNC", self, sorted, return_inverse, y, y_inverse);
auto npuAclParams = EXEC_NPU_CMD_SYNC(aclnnUnique, self, sorted, return_inverse, y, y_inverse);
int64_t *view_dims = nullptr;
uint64_t view_dim_num = 0;
auto ret = aclGetViewShape(npuAclParams.Get<3>(), &view_dims, &view_dim_num);
TORCH_CHECK(ret == 0, "aclGetViewShape failed.", OPS_ERROR(ErrCode::ACL));
c10::SmallVector<int64_t, SIZE> output_size(view_dims, view_dims + view_dim_num);
y.resize_(output_size);
delete[] view_dims;
view_dims = nullptr;
return std::tuple<at::Tensor, at::Tensor>(y, y_inverse);
}
#endif
}