diff --git a/bpu_model_perf_images/example_fs_bayese/input/yolov8_640x640_nv12.bin b/bpu_model_perf_images/example_fs_bayese/input/yolov8_640x640_nv12.bin new file mode 100644 index 0000000..52e43ea Binary files /dev/null and b/bpu_model_perf_images/example_fs_bayese/input/yolov8_640x640_nv12.bin differ diff --git a/bpu_model_perf_images/example_fs_bayese/output/result.json b/bpu_model_perf_images/example_fs_bayese/output/result.json new file mode 100644 index 0000000..1047cf9 --- /dev/null +++ b/bpu_model_perf_images/example_fs_bayese/output/result.json @@ -0,0 +1,48 @@ +[ + { + "model_name": "yolov8_640x640_nv12.bin", + "model_path": "/workspace/input/yolov8_640x640_nv12.bin", + "perf_results": [ + { + "thread_num": 1, + "frame_count": 20, + "run_time_ms": 115.618, + "total_latency_ms": 114.53302, + "avg_latency_ms": 5.726651, + "fps": 172.983445, + "raw_output": "hrt_model_exec perf --model_file /workspace/input/yolov8_640x640_nv12.bin --thread_num 1 --frame_count 20\n[BPU_PLAT]BPU Platform Version(1.3.6)! soc info(x5)\n[HBRT] set log level as 0. version = 3.15.55.0\n[DNN] Runtime version = 1.24.5_(3.15.55 HBRT)\n[A][DNN][packed_model.cpp:247][Model](2026-03-18,10:09:48.457.812) [HorizonRT] The model builder version = 1.23.6\n[W][DNN]bpu_model_info.cpp:491][Version](2026-03-18,10:09:48.544.186) Model: yolov8n_640x640_nv12. Inconsistency between the hbrt library version 3.15.55.0 and the model build version 3.15.49.0 detected, in order to ensure correct model results, it is recommended to use compilation tools and the BPU SDK from the same OpenExplorer package.\nLoad model to DDR cost 160.029ms.\n\n\u001b[1;33m [Warning]: These operators have range limitations on input data: \u001b[0m\n\u001b[1;33m [Div, Tan, Acos, Asin, Sqrt, Gather, GatherElements, GatherND, GridSample, Log, Onehot, PsroiPooling, Range, ReverseSequence, RoiPooling, RoiAlign, ScatterElements, ScatterND, Slice, Tile, Topk, Upsample]. \u001b[0m\n\u001b[1;33m Please make sure that these operators are not in your model, when no input data is provided to the tool. \u001b[0m\n\u001b[1;33m [Suggestion]: Using --input_file command to specify perf input data, which can appoint valid input data. \u001b[0m\n\nFrame count: 20, Thread Average: 5.726651 ms, thread max latency: 7.971000 ms, thread min latency: 5.530000 ms, FPS: 173.265182\n\nRunning condition:\n Thread number is: 1\n Frame count is: 20\n Program run time: 115.618000 ms\nPerf result:\n Frame totally latency is: 114.533020 ms\n Average latency is: 5.726651 ms\n Frame rate is: 172.983445 FPS\nI0000 00:00:00.000000 9 vlog_is_on.cc:197] RAW: Set VLOG level for \"*\" to 3\nI0318 10:09:48.550940 9 function_util.cpp:323] get model handle success\nI0318 10:09:48.551020 9 function_util.cpp:656] get model input count success\nI0318 10:09:48.551142 9 function_util.cpp:687] prepare input tensor success!\nI0318 10:09:48.551162 9 function_util.cpp:697] get model output count success\n", + "returncode": 0 + }, + { + "thread_num": 2, + "frame_count": 20, + "run_time_ms": 78.113, + "total_latency_ms": 150.414993, + "avg_latency_ms": 7.52075, + "fps": 256.039328, + "raw_output": "hrt_model_exec perf --model_file /workspace/input/yolov8_640x640_nv12.bin --thread_num 2 --frame_count 20\n[BPU_PLAT]BPU Platform Version(1.3.6)! soc info(x5)\n[HBRT] set log level as 0. version = 3.15.55.0\n[DNN] Runtime version = 1.24.5_(3.15.55 HBRT)\n[A][DNN][packed_model.cpp:247][Model](2026-03-18,10:09:49.861.354) [HorizonRT] The model builder version = 1.23.6\n[W][DNN]bpu_model_info.cpp:491][Version](2026-03-18,10:09:49.941.333) Model: yolov8n_640x640_nv12. Inconsistency between the hbrt library version 3.15.55.0 and the model build version 3.15.49.0 detected, in order to ensure correct model results, it is recommended to use compilation tools and the BPU SDK from the same OpenExplorer package.\nLoad model to DDR cost 148.921ms.\n\n\u001b[1;33m [Warning]: These operators have range limitations on input data: \u001b[0m\n\u001b[1;33m [Div, Tan, Acos, Asin, Sqrt, Gather, GatherElements, GatherND, GridSample, Log, Onehot, PsroiPooling, Range, ReverseSequence, RoiPooling, RoiAlign, ScatterElements, ScatterND, Slice, Tile, Topk, Upsample]. \u001b[0m\n\u001b[1;33m Please make sure that these operators are not in your model, when no input data is provided to the tool. \u001b[0m\n\u001b[1;33m [Suggestion]: Using --input_file command to specify perf input data, which can appoint valid input data. \u001b[0m\n\nFrame count: 20, Thread Average: 7.520750 ms, thread max latency: 10.315000 ms, thread min latency: 6.401000 ms, FPS: 256.561554\n\nRunning condition:\n Thread number is: 2\n Frame count is: 20\n Program run time: 78.113000 ms\nPerf result:\n Frame totally latency is: 150.414993 ms\n Average latency is: 7.520750 ms\n Frame rate is: 256.039328 FPS\nI0000 00:00:00.000000 28 vlog_is_on.cc:197] RAW: Set VLOG level for \"*\" to 3\nI0318 10:09:49.942950 28 function_util.cpp:323] get model handle success\nI0318 10:09:49.943027 28 function_util.cpp:656] get model input count success\nI0318 10:09:49.943101 28 function_util.cpp:687] prepare input tensor success!\nI0318 10:09:49.943121 28 function_util.cpp:697] get model output count success\n", + "returncode": 0 + }, + { + "thread_num": 3, + "frame_count": 20, + "run_time_ms": 78.147, + "total_latency_ms": 219.791016, + "avg_latency_ms": 10.989551, + "fps": 255.927931, + "raw_output": "hrt_model_exec perf --model_file /workspace/input/yolov8_640x640_nv12.bin --thread_num 3 --frame_count 20\n[BPU_PLAT]BPU Platform Version(1.3.6)! soc info(x5)\n[HBRT] set log level as 0. version = 3.15.55.0\n[DNN] Runtime version = 1.24.5_(3.15.55 HBRT)\n[A][DNN][packed_model.cpp:247][Model](2026-03-18,10:09:51.211.152) [HorizonRT] The model builder version = 1.23.6\n[W][DNN]bpu_model_info.cpp:491][Version](2026-03-18,10:09:51.301.611) Model: yolov8n_640x640_nv12. Inconsistency between the hbrt library version 3.15.55.0 and the model build version 3.15.49.0 detected, in order to ensure correct model results, it is recommended to use compilation tools and the BPU SDK from the same OpenExplorer package.\nLoad model to DDR cost 159.794ms.\n\n\u001b[1;33m [Warning]: These operators have range limitations on input data: \u001b[0m\n\u001b[1;33m [Div, Tan, Acos, Asin, Sqrt, Gather, GatherElements, GatherND, GridSample, Log, Onehot, PsroiPooling, Range, ReverseSequence, RoiPooling, RoiAlign, ScatterElements, ScatterND, Slice, Tile, Topk, Upsample]. \u001b[0m\n\u001b[1;33m Please make sure that these operators are not in your model, when no input data is provided to the tool. \u001b[0m\n\u001b[1;33m [Suggestion]: Using --input_file command to specify perf input data, which can appoint valid input data. \u001b[0m\n\nFrame count: 20, Thread Average: 10.989551 ms, thread max latency: 13.497000 ms, thread min latency: 7.165000 ms, FPS: 256.410248\n\nRunning condition:\n Thread number is: 3\n Frame count is: 20\n Program run time: 78.147000 ms\nPerf result:\n Frame totally latency is: 219.791016 ms\n Average latency is: 10.989551 ms\n Frame rate is: 255.927931 FPS\nI0000 00:00:00.000000 48 vlog_is_on.cc:197] RAW: Set VLOG level for \"*\" to 3\nI0318 10:09:51.303458 48 function_util.cpp:323] get model handle success\nI0318 10:09:51.303537 48 function_util.cpp:656] get model input count success\nI0318 10:09:51.303613 48 function_util.cpp:687] prepare input tensor success!\nI0318 10:09:51.303634 48 function_util.cpp:697] get model output count success\n", + "returncode": 0 + }, + { + "thread_num": 4, + "frame_count": 20, + "run_time_ms": 78.287, + "total_latency_ms": 285.552979, + "avg_latency_ms": 14.277649, + "fps": 255.470257, + "raw_output": "hrt_model_exec perf --model_file /workspace/input/yolov8_640x640_nv12.bin --thread_num 4 --frame_count 20\n[BPU_PLAT]BPU Platform Version(1.3.6)! soc info(x5)\n[HBRT] set log level as 0. version = 3.15.55.0\n[DNN] Runtime version = 1.24.5_(3.15.55 HBRT)\n[A][DNN][packed_model.cpp:247][Model](2026-03-18,10:09:52.572.259) [HorizonRT] The model builder version = 1.23.6\n[W][DNN]bpu_model_info.cpp:491][Version](2026-03-18,10:09:52.631.569) Model: yolov8n_640x640_nv12. Inconsistency between the hbrt library version 3.15.55.0 and the model build version 3.15.49.0 detected, in order to ensure correct model results, it is recommended to use compilation tools and the BPU SDK from the same OpenExplorer package.\nLoad model to DDR cost 128.734ms.\n\n\u001b[1;33m [Warning]: These operators have range limitations on input data: \u001b[0m\n\u001b[1;33m [Div, Tan, Acos, Asin, Sqrt, Gather, GatherElements, GatherND, GridSample, Log, Onehot, PsroiPooling, Range, ReverseSequence, RoiPooling, RoiAlign, ScatterElements, ScatterND, Slice, Tile, Topk, Upsample]. \u001b[0m\n\u001b[1;33m Please make sure that these operators are not in your model, when no input data is provided to the tool. \u001b[0m\n\u001b[1;33m [Suggestion]: Using --input_file command to specify perf input data, which can appoint valid input data. \u001b[0m\n\nFrame count: 20, Thread Average: 14.277649 ms, thread max latency: 17.122000 ms, thread min latency: 7.361000 ms, FPS: 255.973785\n\nRunning condition:\n Thread number is: 4\n Frame count is: 20\n Program run time: 78.287000 ms\nPerf result:\n Frame totally latency is: 285.552979 ms\n Average latency is: 14.277649 ms\n Frame rate is: 255.470257 FPS\nI0000 00:00:00.000000 69 vlog_is_on.cc:197] RAW: Set VLOG level for \"*\" to 3\nI0318 10:09:52.633414 69 function_util.cpp:323] get model handle success\nI0318 10:09:52.633491 69 function_util.cpp:656] get model input count success\nI0318 10:09:52.633570 69 function_util.cpp:687] prepare input tensor success!\nI0318 10:09:52.633590 69 function_util.cpp:697] get model output count success\n", + "returncode": 0 + } + ] + } +] \ No newline at end of file