Method state_load_file

subprojects/llama.cpp/src/llama-context.cpp:2368–2409

}

bool llama_context::state_load_file(const char * filepath, llama_token * tokens_out, size_t n_token_capacity, size_t * n_token_count_out) {
    llama_file file(filepath, "rb");

    // sanity checks
    {
        const uint32_t magic   = file.read_u32();
        const uint32_t version = file.read_u32();

        if (magic != LLAMA_SESSION_MAGIC || version != LLAMA_SESSION_VERSION) {
            LLAMA_LOG_ERROR("%s: unknown (magic, version) for session file: %08x, %08x\n", __func__, magic, version);
            return false;
        }
    }

    // load the prompt
    {
        const uint32_t n_token_count = file.read_u32();

        if (n_token_count > n_token_capacity) {
            LLAMA_LOG_ERROR("%s: token count in session file exceeded capacity! %u > %zu\n", __func__, n_token_count, n_token_capacity);
            return false;
        }

        file.read_raw(tokens_out, sizeof(llama_token) * n_token_count);
        *n_token_count_out = n_token_count;
    }

    // restore the context state
    {
        const size_t n_state_size_cur = file.size() - file.tell();

        llama_io_read_file io( &file);
        const size_t n_read = state_read_data(io);

        if (n_read != n_state_size_cur) {
            LLAMA_LOG_ERROR("%s: did not read all of the session file data! size %zu, got %zu\n", __func__, n_state_size_cur, n_read);
            return false;
        }
    }

    return true;
}

bool llama_context::state_save_file(const char * filepath, const llama_token * tokens, size_t n_token_count) {
    llama_file file(filepath, "wb");