如何在LibTorch（C++）中对张量数值使用比较运算符？

问题代码 #

尝试遍历1D张量并判断每个元素是否大于0.5，但编译报错：

#include <torch/torch.h>
using namespace torch::indexing;

torch::Tensor frogs;   

int main() {

    frogs = torch::rand({11});

    for (int i = 0; i<10; ++i) {

        if (frogs.index({i}).item() > 0.5) {

            std::cout << frogs.index({i}).item() << " \n";
        }
    }   

    return 0;
}

编译错误信息 #

Consolidate compiler generated dependencies of target mujoco_gym
[ 50%] Building CXX object CMakeFiles/mujoco_gym.dir/tester.cpp.o
/home/iii/tor/m_gym/tester.cpp: In function ‘int main()’:
/home/iii/tor/m_gym/tester.cpp:18:37: error: no match for ‘operator>’ (operand types are ‘c10::Scalar’ and ‘double’)
   18 |         if (frogs.index({i}).item() > 0.5) {
      |             ~~~~~~~~~~~~~~~~~~~~~~~ ^ ~~~
      |                                  |    |
      |                                  |    double
      |                                  c10::Scalar
In file included from /home/iii/tor/m_gym/libtorch/include/c10/util/string_view.h:5,
                 from /home/iii/tor/m_gym/libtorch/include/c10/util/StringUtil.h:6,
                 from /home/iii/tor/m_gym/libtorch/include/c10/util/Exception.h:6,
                 from /home/iii/tor/m_gym/libtorch/include/c10/core/Device.h:5,
                 from /home/iii/tor/m_gym/libtorch/include/ATen/core/TensorBody.h:11,
                 from /home/iii/tor/m_gym/libtorch/include/ATen/core/Tensor.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/ATen/Tensor.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/autograd/function_hook.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/autograd/cpp_hook.h:2,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/autograd/variable.h:6,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/autograd/autograd.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/api/include/torch/autograd.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/api/include/torch/all.h:7,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/api/include/torch/torch.h:3,
                 from /home/iii/tor/m_gym/tester.cpp:1:
/home/iii/tor/m_gym/libtorch/include/c10/util/reverse_iterator.h:200:23: note: candidate: ‘template<class _Iterator> constexpr bool c10::operator>(const c10::reverse_iterator<_Iterator>&, const c10::reverse_iterator<_Iterator>&)’
  200 | inline constexpr bool operator>(
      |                       ^~~~~~~~
/home/iii/tor/m_gym/libtorch/include/c10/util/reverse_iterator.h:200:23: note:   template argument deduction/substitution failed:
/home/iii/tor/m_gym/tester.cpp:18:39: note:   ‘c10::Scalar’ is not derived from ‘const c10::reverse_iterator<_Iterator>’
   18 |         if (frogs.index({i}).item() > 0.5) {
      |                                       ^~~
In file included from /home/iii/tor/m_gym/libtorch/include/c10/util/string_view.h:5,
                 from /home/iii/tor/m_gym/libtorch/include/c10/util/StringUtil.h:6,
                 from /home/iii/tor/m_gym/libtorch/include/c10/util/Exception.h:6,
                 from /home/iii/tor/m_gym/libtorch/include/c10/core/Device.h:5,
                 from /home/iii/tor/m_gym/libtorch/include/ATen/core/TensorBody.h:11,
                 from /home/iii/tor/m_gym/libtorch/include/ATen/core/Tensor.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/ATen/Tensor.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/autograd/function_hook.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/autograd/cpp_hook.h:2,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/autograd/variable.h:6,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/autograd/autograd.h:3,
                 from /home/iii/tor/m_gym/libtorch/include/torch/csrc/api/include/torch/autograd.h:3,

错误原因 #

frogs.index({i}).item()返回的是c10::Scalar类型，该类型没有定义与double类型直接比较的operator>，因此编译时无法找到匹配的运算符号。

解决方案 #

方案1：显式转换Scalar为数值类型 #

将item()改为item<double>()，明确指定转换为double类型，即可与0.5进行比较。同时修正循环条件避免遗漏元素：

#include <torch/torch.h>
using namespace torch::indexing;

torch::Tensor frogs;   

int main() {

    frogs = torch::rand({11});

    for (int i = 0; i < 11; ++i) {

        if (frogs.index({i}).item<double>() > 0.5) {

            std::cout << frogs.index({i}).item<double>() << " \n";
        }
    }   

    return 0;
}

item<T>()会将c10::Scalar强制转换为指定的数值类型T，这里用double匹配0.5的类型，即可正常执行比较逻辑。

方案2：使用LibTorch向量化操作（推荐） #

LibTorch原生支持向量化运算，无需手动遍历，代码更简洁且效率更高（GPU环境下可并行加速）：

#include <torch/torch.h>
#include <iostream>

int main() {
    torch::Tensor frogs = torch::rand({11});
    // 生成布尔掩码，标记每个元素是否大于0.5
    torch::Tensor mask = frogs > 0.5;
    // 过滤出符合条件的元素
    torch::Tensor filtered = frogs.masked_select(mask);
    
    // 打印结果
    std::cout << filtered << std::endl;
    return 0;
}

通过张量直接比较生成布尔掩码，再用masked_select过滤元素，完全利用LibTorch的内置运算能力，符合框架设计理念。

内容的提问来源于stack exchange，提问作者Ant