作者小头像 Lv.3
更多个人资料
264 成长值
2 关注
3 粉丝
+ 关注 私信

个人介绍

一个00后机器人爱好者。

感兴趣或擅长的领域

人工智能、编程语言
个人勋章
TA还没获得勋章~
成长雷达
105
129
0
30
0

个人资料

个人介绍

一个00后机器人爱好者。

感兴趣或擅长的领域

人工智能、编程语言

达成规则

以上满足项可达成此勋章

主题 | 回复
模型推理的过程中用opencv和ascendcl的预处理方式最后模型预测的的效果差异明显
发布时间 2022/09/23 23:29:02 最后回复 zaq 2022/09/27 08:57:33 版块 CANN
126 6 0
他的回复:
不指定参数--precision_mode=allow_fp32_to_fp16，模型转换不会报错，但是指定后会报错NanoDet模型原始文件ONNX格式上面为原始模型的链接，因为上传附件有大小限制，请自行下载aipp文件aipp_op { related_input_rank : 0src_image_size_w : 416src_image_size_h : 416crop : falsepadding : falseinput_format : YUV420SP_U8aipp_mode: staticcsc_switch : truerbuv_swap_switch : falsematrix_r0c0 : 256matrix_r0c1 : 454matrix_r0c2 : 0matrix_r1c0 : 256matrix_r1c1 : -88matrix_r1c2 : -183matrix_r2c0 : 256matrix_r2c1 : 0matrix_r2c2 : 359input_bias_0 : 0input_bias_1 : 128input_bias_2 : 128mean_chn_0 : 104mean_chn_1 : 117mean_chn_2 : 123min_chn_0 : 0.0min_chn_1 : 0.0min_chn_2 : 0.0var_reci_chn_0 : 1var_reci_chn_1 : 1var_reci_chn_2 : 1}
模型推理的过程中用opencv和ascendcl的预处理方式最后模型预测的的效果差异明显
发布时间 2022/09/23 23:29:02 最后回复 zaq 2022/09/27 08:57:33 版块 CANN
126 6 0
他的回复:
HwHiAiUser@ecs-748f:~/code/nanodet_dvpp/model$ atc --model=./nanodet-plus-m-1.5x_416.onnx --framework=5 --output=nanodet-plus-m-1.5x_416 --soc_version=Ascend310 --insert_op_conf=insert_op.cfg --precision_mode=allow_fp32_to_fp16ATC start working now, please wait for a moment.ATC run failed, Please check the detail log, Try 'atc --help' for more informationE29999: Inner Error!E29999 [SubGraphOpt][ParseJson][PkgTvmJsInfo] ReadJsonFile failed.[FUNC:PackageTvmJsonInfo][FILE:tbe_json_parse.cc][LINE:670] [SubGraphOpt][Compile][Normal] Failed to parse json or compress op for graph[partition0_rank20_new_sub_graph241].[FUNC:ReCompileWithNoFusionStrategy][FILE:op_compiler_normal.cc][LINE:100] [SubGraphOpt][Compile][RunCmplProc] Failed to re-compile op with no lx fusion for graph [partition0_rank20_new_sub_graph241][FUNC:RunCompileProcess][FILE:op_compiler_normal.cc][LINE:164] [GraphOpt][FusedGraph][RunCompile] Failed to compile graph with compiler Normal mode Op Compiler[FUNC:SubGraphCompile][FILE:fe_graph_optimizer.cc][LINE:1295] Call OptimizeFusedGraph failed, ret:-1, engine_name:AIcoreEngine, graph_name:partition0_rank20_new_sub_graph241[FUNC:OptimizeSubGraph][FILE:graph_optimize.cc][LINE:127] [SubGraphOpt][Compile][Normal] Failed to parse json or compress op for graph[partition0_rank52_new_sub_graph273].[FUNC:ReCompileWithNoFusionStrategy][FILE:op_compiler_normal.cc][LINE:100] [SubGraphOpt][Compile][RunCmplProc] Failed to re-compile op with no lx fusion for graph [partition0_rank52_new_sub_graph273][FUNC:RunCompileProcess][FILE:op_compiler_normal.cc][LINE:164] Call OptimizeFusedGraph failed, ret:-1, engine_name:AIcoreEngine, graph_name:partition0_rank52_new_sub_graph273[FUNC:OptimizeSubGraph][FILE:graph_optimize.cc][LINE:127] subgraph 19 optimize failed[FUNC:OptimizeSubGraphWithMultiThreads][FILE:graph_manager.cc][LINE:720] build graph failed, graph id:0, ret:-1[FUNC:BuildModel][FILE:ge_generator.cc][LINE:1429]