mmdeploy/csrc/codebase/mmcls/linear_cls.cpp

// Copyright (c) OpenMMLab. All rights reserved.

#include <numeric>

#include "codebase/mmcls/mmcls.h"
#include "core/tensor.h"
#include "core/utils/formatter.h"
#include "experimental/module_adapter.h"

using std::vector;

namespace mmdeploy::mmcls {

class LinearClsHead : public MMClassification {
 public:
  explicit LinearClsHead(const Value& cfg) : MMClassification(cfg) {
    if (cfg.contains("params")) {
      topk_ = cfg["params"].value("topk", 1);
      if (topk_ <= 0) {
        ERROR("'topk' should be greater than 0, but got '{}'", topk_);
        throw_exception(eInvalidArgument);
      }
    }
  }

  Result<Value> operator()(const Value& infer_res) {
    DEBUG("infer_res: {}", infer_res);
    auto output_tensor = infer_res["output"].get<Tensor>();
    assert(output_tensor.shape().size() >= 2);
    auto class_num = (int)output_tensor.shape()[1];

    if (output_tensor.device().is_host()) {
      vector<float> scores(output_tensor.data<float>(),
                           output_tensor.data<float>() + output_tensor.size());
      OUTCOME_TRY(stream().Wait());
      return GetLabels(scores, class_num);
    } else {
      vector<float> scores(output_tensor.size());
      OUTCOME_TRY(output_tensor.CopyTo(scores.data(), stream()));
      OUTCOME_TRY(stream().Wait());
      return GetLabels(scores, class_num);
    }
  }

 private:
  Value GetLabels(const vector<float>& scores, int class_num) const {
    ClassifyOutput output;
    output.labels.reserve(topk_);
    std::vector<int> idx(class_num);
    iota(begin(idx), end(idx), 0);
    partial_sort(begin(idx), begin(idx) + topk_, end(idx),
                 [&](int i, int j) { return scores[i] > scores[j]; });
    for (int i = 0; i < topk_; ++i) {
      auto label = ClassifyOutput::Label{idx[i], scores[idx[i]]};
      DEBUG("label_id: {}, score: {}", label.label_id, label.score);
      output.labels.push_back(label);
    }
    return to_value(std::move(output));
  }

 private:
  int topk_{1};
};

REGISTER_CODEBASE_COMPONENT(MMClassification, LinearClsHead);

}  // namespace mmdeploy::mmcls