Function _get_load_info

lightllm/server/httpserver/pd_loop.py:190–208

()

# 获取节点负载信息
def _get_load_info() -> dict:

    from lightllm.server.api_http import g_objs

    assert g_objs.shared_token_load is not None, "shared_token_load is not initialized"
    args = g_objs.args
    dp_size_in_node = max(1, args.dp // args.nnodes)

    # 获取当前每个 dp 的负载，数值含义为当前的 token 总容量使用率， 上报给 PD_Master 用于做
    # 调度决策。
    current_load = [
        float(g_objs.shared_token_load.get_dynamic_max_load(dp_index)) for dp_index in range(dp_size_in_node)
    ]
    mean_node_load = sum(current_load) / len(current_load)
    load_info = {
        "total_token_usage_rate": mean_node_load,
        "client_ip_port": f"{g_objs.httpserver_manager.host_ip}:{g_objs.args.port}",
    }
    return load_info