我使用的昇腾架构部署大模型,cpu使用鲲鹏920*2,显卡使用的是Altas 300I DUO*4,使用容器部署mindie,在我使用单张卡推理时可以正常响应数据,但是当我配置多卡进行推理时报如下错误,我应该怎么解决问题,使得多张卡也可以进行推理。 [2025-03-28 09:32:28.026+08:00] [177082] [183396] [server] [ERROR] [master_IPC_communicator.cpp:182] : the required reponses count 1 is not same as connector return count 0 [2025-03-28 09:32:28.026+08:00] [177082] [183396] [server] [ERROR] [master_IPC_communicator.cpp:467] : DeSerialize Execute Response failed [2025-03-28 09:32:28.026+08:00] [177082] [183396] [server] [ERROR] [model_backend.cpp:478] : [ModelBackend::InstanceExecute] Failed to send execute message to connector. [2025-03-28 09:32:28.026439+08:00] [177082] [183396] [server] [ERROR] [EndPoint base_wrapper.cpp:41] Failed to get eos info by output 'IBIS_EOS_ATTR' not found in response [2025-03-28 09:32:27.493+08:00] [177082] [183399] [server] [INFO] [ibis_request.cpp:62] : Create IBIS request: 0 successfully with batchSize: 1, isSync: 0, maxOutputLen: 20, priority: 5 [2025-03-28 09:32:27.493+08:00] [177082] [183399] [server] [INFO] [ibis_batch_scheduler.cpp:176] : IBIS Pipeline [0] of WorkFlow Standard_LLM: enqueue sequence: 0 [2025-03-28 09:32:28.026+08:00] [177082] [183396] [server] [ERROR] [ibis_model_instance.cpp:58] : GMIS executes modelInstance failed [2025-03-28 09:32:28.026484+08:00] [177082] [183448] [server] [ERROR] [EndPoint base_wrapper.cpp:41] Failed to get eos info by output 'IBIS_EOS_ATTR' not found in response [2025-03-28 09:32:28.026+08:00] [177082] [183396] [server] [ERROR] [gmis_agent_base_plugin.cpp:37] : Failed to execute modelInstance [2025-03-28 09:32:28.026510+08:00] [177082] [183448] [server] [ERROR] [EndPoint common_wrapper.cpp:251] Failed to get engine response [2025-03-28 09:32:28.026+08:00] [177082] [183396] [server] [ERROR] [gmis_agent_base_plugin.cpp:51] : Failed to save output ids, reqId: 0 [2025-03-28 09:32:28.026+08:00] [177082] [183396] [server] [ERROR] [ibis_request_manager.cpp:229] : GMIS request executed failed, id: 0 [2025-03-28 09:32:28.026+08:00] [177082] [183397] [server] [INFO] [ibis_request_manager.cpp:181] : IBIS requestManager release request: 0 [2025-03-28 09:32:28.026624+08:00] [177082] [183448] [server] [INFO] [EndPoint http_server.cpp:56] Http receive response from 172.17.0.1:1025 ,status code 500, body len 79
Atlas 300I DUO多卡问题
已解决



