导出yolact模型时报错

[NULL037]

现象：
按照官方提供的yolact模型转换的文档操作，在导出模型时报如下错误：

I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/yolact.2012.vcxproj
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/makefile.linux
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/.cproject
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/.project
D Generate fake input /home/ubuntu/yolact/ovxlib/yolact/input.1_219_0.tensor
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -c vnn_pre_process.c
cc1: warning: command line option ‘-std=c++11’ is valid for C++/ObjC++ but not for C
vnn_pre_process.c: In function ‘_handle_multiple_inputs’:
vnn_pre_process.c:749:11: warning: variable ‘p1’ set but not used [-Wunused-but-set-variable]
  749 |     char *p1 = NULL;
      |           ^~
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -c vnn_yolact.c
cc1: warning: command line option ‘-std=c++11’ is valid for C++/ObjC++ but not for C
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -c main.c
cc1: warning: command line option ‘-std=c++11’ is valid for C++/ObjC++ but not for C
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -c vnn_post_process.c
cc1: warning: command line option ‘-std=c++11’ is valid for C++/ObjC++ but not for C
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -O3  vnn_pre_process.o vnn_yolact.o main.o vnn_post_process.o -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib/vsim -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib/x64_linux -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib/x64_linux/vsim -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib/x64_linux/vsim -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/../common/lib/ -lvdtproxy /home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/../common/lib/libjpeg.a -o gen_nbg
Create Neural Network: 213ms or 213529us
Verify...
E [kernel/evis/resize_bilinear_evis.c:_resize_bilinear_initializer_impl:1051]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E [main.c:vnn_VerifyGraph:91]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E [main.c:main:243]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E Fatal model generation error: 65280
W ----------------Error(1),Warning(0)----------------
Traceback (most recent call last):
  File "pegasus.py", line 131, in <module>
  File "pegasus.py", line 116, in main
  File "acuitylib/app/exporter/commands.py", line 178, in execute
  File "acuitylib/vsi_nn.py", line 650, in export_ovxlib
  File "acuitylib/app/exporter/ovxlib_case/export_ovxlib.py", line 73, in run
  File "acuitylib/app/exporter/ovxlib_case/casegenerator.py", line 701, in generate
  File "acuitylib/app/exporter/ovxlib_case/casegenerator.py", line 660, in _gen_special_case
  File "acuitylib/app/exporter/ovxlib_case/casegenerator.py", line 637, in _gen_nb_file
  File "acuitylib/acuitylog.py", line 263, in e
acuitylib.acuityerror.AcuityError: ('Fatal model generation error: 65280', 'nbg_generate')
[5240] Failed to execute script 'pegasus' due to unhandled exception!

目前可知的提示信息：

E [kernel/evis/resize_bilinear_evis.c:_resize_bilinear_initializer_impl:1051]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E [main.c:vnn_VerifyGraph:91]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E [main.c:main:243]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E Fatal model generation error: 65280
W ----------------Error(1),Warning(0)----------------

前面的操作都是按照文档的操作，请问各位大佬这啥问题导致的？

[NULL037]

@null037 反复对比检查了一下不知道是否是下面的警告导致的报错

[NULL037]

@null037 找了一段时间没什么头绪，这放一下根据文档转换后的ONNX格式的yolact模型文件，可以尝试按照文档进行模型转换，看一下各位大佬在转换时是否也会遇到此问题：

yolact模型文件：yolact.onnx

yolact模型转换工程文件：yolact.tar.gz

我是在导出模型时遇到的问题，前面没有出现报错信息。导出模型命令为：

pegasus export ovxlib --model yolact-sim.json --model-data yolact-sim.data --dtype
quantized --model-quantize yolact-sim.quantilize --batch-size 1 --save-fused-graph --
target-ide-project 'linux64' --with-input-meta yolact-sim-inputemeta.yml --postprocess-file yolact-sim-postprocess-file.yml --output-path ovxlib/yolact/yolact --pack-nbg-unify -
-optimize "VIP9000PICO_PID0XEE" --vivsdk${VIV_SDK}

报错信息为：

D Quantize @Conv_Conv_97_183:bias to asymmetric_affine.
D Quantize @Conv_Conv_97_183:weight to asymmetric_affine.
D Packing Conv_Conv_99_162 ...
D Quantize @Conv_Conv_99_162:bias to asymmetric_affine.
D Quantize @Conv_Conv_99_162:weight to asymmetric_affine.
D Packing Initializer_769_6 ...
D Packing Initializer_onnx//Add_528_104 ...
I Saving data to ovxlib/yolact/yolact.export.data

D Get binary package acuity_path [/home/ubuntu/VeriSilicon/acuity-toolkit-binary-6.6.1/bin]
D Get binary package acuity_path [/home/ubuntu/VeriSilicon/acuity-toolkit-binary-6.6.1/bin]
D Get binary package acuity_path [/home/ubuntu/VeriSilicon/acuity-toolkit-binary-6.6.1/bin]
I Export ovxlib case with version 1.1.36
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/vnn_yolact.c
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/vnn_yolact.h
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/vnn_post_process.c
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/vnn_post_process.h
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/vnn_pre_process.c
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/vnn_pre_process.h
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/vnn_global.h
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/main.c
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/BUILD
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/yolact.vcxproj
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/yolact.2012.vcxproj
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/makefile.linux
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/.cproject
I Save vx network source file to /home/ubuntu/yolact/ovxlib/yolact/.project
D Generate fake input /home/ubuntu/yolact/ovxlib/yolact/input.1_219_0.tensor
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -c vnn_pre_process.c
cc1: warning: command line option ‘-std=c++11’ is valid for C++/ObjC++ but not for C
vnn_pre_process.c: In function ‘_handle_multiple_inputs’:
vnn_pre_process.c:749:11: warning: variable ‘p1’ set but not used [-Wunused-but-set-variable]
  749 |     char *p1 = NULL;
      |           ^~
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -c vnn_yolact.c
cc1: warning: command line option ‘-std=c++11’ is valid for C++/ObjC++ but not for C
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -c main.c
cc1: warning: command line option ‘-std=c++11’ is valid for C++/ObjC++ but not for C
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -c vnn_post_process.c
cc1: warning: command line option ‘-std=c++11’ is valid for C++/ObjC++ but not for C
gcc -Wall -std=c++0x -I. -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/CL -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/VX -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/ovxlib -I/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/include/jpeg -D__linux__ -DLINUX -O3 -O3  vnn_pre_process.o vnn_yolact.o main.o vnn_post_process.o -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib/vsim -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib/x64_linux -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib/x64_linux/vsim -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/lib/x64_linux/vsim -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -L/home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/../common/lib/ -lvdtproxy /home/ubuntu/VeriSilicon/VivanteIDE5.7.0/cmdtools/vsimulator/../common/lib/libjpeg.a -o gen_nbg
Create Neural Network: 1494ms or 1494420us
Verify...
E [kernel/evis/resize_bilinear_evis.c:_resize_bilinear_initializer_impl:1051]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E [main.c:vnn_VerifyGraph:91]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E [main.c:main:243]CHECK STATUS(-1:A generic error code, used when no other describes the error.)
E Fatal model generation error: 65280
W ----------------Error(1),Warning(0)----------------
Traceback (most recent call last):
  File "pegasus.py", line 131, in <module>
  File "pegasus.py", line 116, in main
  File "acuitylib/app/exporter/commands.py", line 178, in execute
  File "acuitylib/vsi_nn.py", line 650, in export_ovxlib
  File "acuitylib/app/exporter/ovxlib_case/export_ovxlib.py", line 73, in run
  File "acuitylib/app/exporter/ovxlib_case/casegenerator.py", line 701, in generate
  File "acuitylib/app/exporter/ovxlib_case/casegenerator.py", line 660, in _gen_special_case
  File "acuitylib/app/exporter/ovxlib_case/casegenerator.py", line 637, in _gen_nb_file
  File "acuitylib/acuitylog.py", line 263, in e
acuitylib.acuityerror.AcuityError: ('Fatal model generation error: 65280', 'nbg_generate')
[3550] Failed to execute script 'pegasus' due to unhandled exception!

[NULL037]

@null037 通过测试转换yolov5的模型发现

问题并不是出现在GCC和G++的问题，而是在校验是没有通过，不太清楚校验主要验证模型的什么内容

[DabuGege]

@null037 hello，看了里头的yolact-sim.onnx模型，是模型的输出shape还没有固定导致的错误；请重新获取onnx模型并将模型设置为静态shape再尝试~

[NULL037]

@dabugege 您好，很感谢您的回复让我找到了思路。我现在根据您的建议又进行尝试了，将输出和输入都固定，使用onnxsim，如下所示：

模型简化前后结构对比如下图所示：

我发现简化后固定输入输出只有四个输出节点，提供的文档中有5个输出节点。但我觉得这个不会影响。

继续进行模型转换发现还是会出现同样的问题

源模型文件：yolact.onnx

简化后模型文件：yolact-sim.onnx

这是我固定模型输出有问题吗？期待您的回复

[DabuGege]

@null037 hi~，应是这个版本的IDE某个op支持有bug，试下这个IDE版本；或是联系全志获取最新的NPU配套软件。
IDE v5.7.1: https://netstorage.allwinnertech.com:5001/sharing/9gbytImwj
pwd：allwinnertech

[yzr1992]

@dabugege 怎么联系全志获取最新的NPU配套软件？

[WhycanService]

@yzr1992 联系对接的FAE

[yzr1992]

@whycanservice 直接从百问网的淘宝店购买的，没有FAE，全志的官方QQ群也只是个吹水群，都让我到论坛发帖子。。

[WhycanService]

@yzr1992 那就用这个链接
IDE v5.7.1: https://netstorage.allwinnertech.com:5001/sharing/9gbytImwj
pwd：allwinnertech