在C中使用张量浮点值++

我可以这样做:

std::vector<Tensor> outputs;
    Status run_status = session->Run(feed_dict, {"fc2/Softmax"}, {}, &outputs);
    if (!run_status.ok()) {
        cout << "Running model failed: " << run_status;
        return -1;
    }
    for (auto &t : outputs) // access by reference to avoid copying
    {  
        cout << t.DebugString() + "\n";
        tensorflow::TTypes<float, 2>::Tensor scores = t.flat_inner_dims<float>();
        auto dims = scores.dimensions();
        int imgCount = dims[0];
        int classesCount = dims[1];
        for(int i = 0; i<imgCount; i++) {
            float maxVal = scores(i,0);
            int maxIndex = 0;
            for(int j = 1; j<classesCount; j++) {
                float val = scores(i,j);
                if(val > maxVal) {
                    maxVal = val;
                    maxIndex = j;
                }
            }
            cout << "Img" + to_string(i) + " prediction: " + to_string(maxIndex) + ", score: " + to_string(maxVal) + "\n";
        }
    }

功能 flat_inner_dims() 允许检索给定形状的特征::张量 <float, 2> 表示一个二维浮点数数组。然后,我运行了这个数组。

您可以找到有关如何使用特征::张量的更多信息 here 。

希望这能有所帮助。