首先,atlas200加速模块RC模式, 物理机安装了python3.7.5+mindspore2.0+mindspore_serving2.0和cann-toolkits,能够实现正常运行mindspore_serving推理服务。
随后,构造docker推理容器,挂载物理机python环境目录,挂载物理机cann-toolkits工作目录,npm-smi info、atc --help、pyacl加载om模型等功能都能够正常运行,但是mindspore_serving推理服务报错,关键信息如下:
日志内容
首先,atlas200加速模块RC模式, 物理机安装了python3.7.5+mindspore2.0+mindspore_serving2.0和cann-toolkits,能够实现正常运行mindspore_serving推理服务。
随后,构造docker推理容器,挂载物理机python环境目录,挂载物理机cann-toolkits工作目录,npm-smi info、atc --help、pyacl加载om模型等功能都能够正常运行,但是mindspore_serving推理服务报错,关键信息如下:
root@a1ef4bd5548d: python serving_server.py --- Serving: Master server start success, listening on unix:unix_socket_files/serving_master_284 --- Traceback (most recent call last): File "mytest/serving_server.py", line 16, in <module> start() File "mytest/serving_server.py", line 10, in start server.start_servables(servable_configs=servable_config) File "/root/miniconda3/envs/py375/lib/python3.7/site-packages/mindspore_serving/server/master/_master.py", line 63, in handle_except func(*args, **kwargs) File "/root/miniconda3/envs/py375/lib/python3.7/site-packages/mindspore_serving/server/_server.py", line 119, in start_servables _listening_workers_when_startup(worker_list) File "/root/miniconda3/envs/py375/lib/python3.7/site-packages/mindspore_serving/server/_server.py", line 231, in _listening_workers_when_startup raise RuntimeError(f"Fail to start workers: {worker.get_notified_error()}") --- RuntimeError: Fail to start workers: {servable name:mytest, device id:0, <Raise failed: Load model failed, servable directory: '/home/HwHiAiUser/xc200/mytest', servable name: 'mytest', model file: '[mytest-018.mindir]', version number 1,model context: {}, load error details: Common error code.>} bash 输出日志内容