reference/cpp/onnx_wrapper.h

#include <vector>
#include <string>

#if defined(__APPLE__)
#include <onnxruntime/onnxruntime_cxx_api.h>
#else
#include "onnxruntime_run_options_config_keys.h"
#include "onnxruntime_cxx_api.h"
#endif

#ifdef _WIN32

#define ORTSTRING(str) StrToWstr(str)
#define ORTCHAR(str) StrToWstr(str).c_str()

inline std::wstring String2wstring(const std::string& str, const std::string& locale)
{
    typedef std::codecvt_byname<wchar_t, char, std::mbstate_t> F;
    std::wstring_convert<F> strCnv(new F(locale));
    return strCnv.from_bytes(str);
}

inline std::wstring  StrToWstr(std::string str) {
    if (str.length() == 0)
        return L"";
    return  String2wstring(str, "zh-CN");

}

#else

#define ORTSTRING(str) str
#define ORTCHAR(str) str

#endif

class OnnxVadWrapper {
public:
    explicit OnnxVadWrapper(const std::string& model_path, bool force_cpu = false, int thread_num = 1);
    ~OnnxVadWrapper();

    // 重载 operator()，使得对象可以像函数一样调用
    std::pair<std::vector<float>, std::vector<float>> operator()(const std::vector<float>& x, int sr);

    // 批量处理整个音频
    std::vector<float> audio_forward(const std::vector<float>& audio, int sr);

    // 重置 RNN 状态
    void reset_states(int batch_size = 1);

private:
    Ort::Env env_;

    std::unique_ptr<Ort::Session> session_;
    std::vector<std::string> input_names_, output_names_;
    std::vector<const char *> vad_in_names_;
    std::vector<const char *> vad_out_names_;

    std::vector<int> sample_rates_;
    std::string model_path_;

    std::vector<float> state_;     // RNN State
    std::vector<float> context_;   // Context buffer
    int last_sr_ = 0;
    int last_batch_size_ = 0;

    void read_model();
    bool supports_cpu();
    void validate_input(const std::vector<float>& x, int sr);
};