char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Zijun Yuopenvino: OV 2026.2, context-shift, Q5_1 support, gemma4 dense/embedding, and -fa off (#24503)890f1a27e2026-06-17

#pragma once

#include "input_model.h"
#include "node_context.h"

namespace ov {
namespace frontend {
namespace ggml {

class TranslateSession {
public:
    TranslateSession(const frontend::InputModel::Ptr & input_model,
                     const std::unordered_map<std::string, CreatorFunction> & translator_map,
                     bool naive = false);

    std::shared_ptr<Model> get_converted_model();
    std::shared_ptr<Model> translate_graph(const frontend::InputModel::Ptr & input_model);

private:
    std::shared_ptr<Model> apply_transformations(std::shared_ptr<Model> model);
    const frontend::InputModel::Ptr m_input_model;
    const std::unordered_map<std::string, CreatorFunction> & m_translator_map;
    std::shared_ptr<Model> m_ov_model;
    bool m_naive;
};

}  // namespace ggml
}  // namespace frontend
}  // namespace ov