(hr_api.cpp:460): HorizonRT version = 1.1.30(hr_api.cpp:465): hbrt version = 3.9.6[HBRT] set log level as 0. version = 3.9.6[BPU_PLAT]Makesure BPU Core Opened!!!)[BPU_PLAT]BPU Platform Version(1.1.1)![BPU_PLAT]Makesure BPU Core(1) Opened!!!(model.cpp:165): The model builder version = 1.1.24(bpu_model_info.cpp:37): Hybrid model hbm_hbrt_version = 3.9.6HR:ERROR: can not find according hbm node input feature for graph input with same nameHR:ERROR: [/home/users/daofu.zhang/bpu-predict/3rdparty/horizonrt/src/engine/executor.cpp:31](Init) Error_code is 10047. Error_msg is hrErrorHbmInputFeatureNameNotSameWithGraphInputNameHR:ERROR: [/home/users/daofu.zhang/bpu-predict/3rdparty/horizonrt/src/hr_api.cpp:51](hrLoadHBMFromAddr) Error_code is 10047. Error_msg is hrErrorHbmInputFeatureNameNotSameWithGraphInputName[ERROR:548538851344: 02:01:47]/home/users/daofu.zhang/bpu-predict/src/bpu_instance.cc:209: load hbm file failed : hrErrorHbmInputFeatureNameNotSameWithGraphInputNameE0109 02:01:47.155158 2736 utils.cc:134] Load model failed, bpu load hbm file failedF0109 02:01:47.159913 2736 multi_input_example.cc:71] Load model failed*** Check failure stack trace: ***
您好，经过我们内部测试确认，该问题是我们 runtime 代码上的一个 bug。
原因是之前没考虑到 BPU 多输入的情况，所以发生了内存踩踏，我们会在下个版本修复该问题。-
谢谢您的模型为我们提供了一种新的测试场景。