[Bug-Report|缺陷反馈]: unique_consecutive存在嵌套调用L0接口情况，请分析是否修改-ops-nn-AtomGit

已关闭

[Bug-Report|缺陷反馈]: unique_consecutive存在嵌套调用L0接口情况，请分析是否修改 #5545

liuhanwen8创建于 9月3日关闭于 9月28日

liuhanwen8

9月3日创建

Thanks for sending an issue! Please fill in the following template to help quickly solve your problem.

Describe the current behavior / 问题描述 (Mandatory / 必填)

https://gitcode.com/cann/ops-nn/blob/master/index/unique_consecutive/op_api/unique_consecutive.cpp

aclnnStatus UniqueConsecutive(const aclTensor* self, bool returnInverse, bool returnCounts, int64_t dim,
aclTensor* valueOut, aclTensor* inverseOut, aclTensor* countsOut, aclOpExecutor* executor)
{
L0_DFX(UniqueConsecutive, self, returnInverse, returnCounts, dim, valueOut, inverseOut, countsOut);

if (CheckSupport4Aicore(self, returnInverse, dim, valueOut, inverseOut, countsOut)) {
    op::DataType outIdxType = op::DataType::DT_INT64;
    if (returnInverse) {
        outIdxType = inverseOut->GetDataType();
    } else if (returnCounts) {
        outIdxType = countsOut->GetDataType();
    }
    return UniqueConsecutiveAiCore(self, returnInverse, returnCounts, dim, valueOut, inverseOut, countsOut,
                                   outIdxType, executor);
}

static internal::AicpuTaskSpace space("UniqueConsecutive", ge::DEPEND_SHAPE_RANGE);
aclnnStatus ret = ACLNN_SUCCESS;
if (dim == NoneN) {
    ret = ADD_TO_LAUNCHER_LIST_AICPU(UniqueConsecutive, OP_ATTR_NAMES({"return_idx", "return_counts"}),
                                     OP_INPUT(self), OP_OUTPUT(valueOut, inverseOut, countsOut),
                                     OP_ATTR(returnInverse, returnCounts));
} else {
    ret = ADD_TO_LAUNCHER_LIST_AICPU(UniqueConsecutive, OP_ATTR_NAMES({"return_idx", "return_counts", "axis"}),
                                     OP_INPUT(self), OP_OUTPUT(valueOut, inverseOut, countsOut),
                                     OP_ATTR(returnInverse, returnCounts, dim));
}
return ret;

}
} // namespace l0op

UniqueConsecutiveAiCore也是L0函数，存在嵌套调用