$NINJA_PATH/ninja step error issue

[kkhsuk]

when I install the darknet-nnpack

I have an issue on step of "$NINJA_PATH"

[62/147] LINK bin/transform-benchmark FAILED: /home/pi/ninja/NNPACK-darknet/bin/transform-benchmark g++ -pthread -o /home/pi/ninja/NNPACK-darknet/bin/transform-benchmark /home/pi/ninja/NNPACK-darknet/build/bench/transform.c.o /home/pi/ninja/NNPACK-darknet/build/bench/median.c.o /home/pi/ninja/NNPACK-darknet/build/bench/memread.c.o -lrt /home/pi/ninja/NNPACK-darknet/lib/libnnpack.a /home/pi/ninja/NNPACK-darknet/lib/libpthreadpool.a /home/pi/ninja/NNPACK-darknet/lib/libpthreadpool.a(pthreads.c.o): In function pthreadpool_create': /home/pi/ninja/NNPACK-darknet/deps/pthreadpool/src/pthreads.c:258: undefined reference to pthreadpool_allocate' /home/pi/ninja/NNPACK-darknet/lib/libpthreadpool.a(pthreads.c.o): In function pthreadpool_destroy': /home/pi/ninja/NNPACK-darknet/deps/pthreadpool/src/pthreads.c:459: undefined reference to pthreadpool_deallocate' /usr/bin/ld: /home/pi/ninja/NNPACK-darknet/bin/transform-benchmark: internal symbol pthreadpool_deallocate' isn't defined /usr/bin/ld: final link failed: Bad value collect2: error: ld returned 1 exit status [63/147] LINK bin/convolution-benchmark FAILED: /home/pi/ninja/NNPACK-darknet/bin/convolution-benchmark g++ -pthread -o /home/pi/ninja/NNPACK-darknet/bin/convolution-benchmark /home/pi/ninja/NNPACK-darknet/build/bench/convolution.c.o /home/pi/ninja/NNPACK-darknet/build/bench/median.c.o /home/pi/ninja/NNPACK-darknet/build/bench/memread.c.o -lrt /home/pi/ninja/NNPACK-darknet/lib/libnnpack.a /home/pi/ninja/NNPACK-darknet/lib/libpthreadpool.a /home/pi/ninja/NNPACK-darknet/lib/libpthreadpool.a(pthreads.c.o): In function pthreadpool_create': /home/pi/ninja/NNPACK-darknet/deps/pthreadpool/src/pthreads.c:258: undefined reference to pthreadpool_allocate' /home/pi/ninja/NNPACK-darknet/lib/libpthreadpool.a(pthreads.c.o): In function pthreadpool_destroy': /home/pi/ninja/NNPACK-darknet/deps/pthreadpool/src/pthreads.c:459: undefined reference to pthreadpool_deallocate' /usr/bin/ld: /home/pi/ninja/NNPACK-darknet/bin/convolution-benchmark: internal symbol pthreadpool_deallocate' isn't defined /usr/bin/ld: final link failed: Bad value collect2: error: ld returned 1 exit status [64/147] CXX test/convolution-output/alexnet+relu.cc ninja: build stopped: subcommand failed.

somebody help me

[Shivankit99]

Just avoid using Ninja and build NNPACK using cmake and make install.

[GiorgioSgl]

Same here, as you solved in some way?

[kkhsuk]

someone send to me a this solution

Just avoid using Ninja and build NNPACK using cmake and make install.

this solution is working but we had some problem of the darknet-nnpack so we didn't use darknet-nnpack

-----Original Message----- From: "Giorgio Segalla"[email protected] To: "digitalbrain79/darknet-nnpack"[email protected]; Cc: "kkhsuk"[email protected]; "Author"[email protected]; Sent: 2021-03-10 (수) 00:21:46 (GMT+09:00) Subject: Re: [digitalbrain79/darknet-nnpack] $NINJA_PATH/ninja step error issue (#60)

Same here, as you solved in some way? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[GiorgioSgl]

someone send to me a this solution Just avoid using Ninja and build NNPACK using cmake and make install. this solution is working but we had some problem of the darknet-nnpack so we didn't use darknet-nnpack … -----Original Message----- From: "Giorgio Segalla"[email protected] To: "digitalbrain79/darknet-nnpack"[email protected]; Cc: "kkhsuk"[email protected]; "Author"[email protected]; Sent: 2021-03-10 (수) 00:21:46 (GMT+09:00) Subject: Re: [digitalbrain79/darknet-nnpack] $NINJA_PATH/ninja step error issue (#60) Same here, as you solved in some way? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

Okay seem I solved it follow your suggestion, thank you very much.

These are the command for who doesn't know hot to run cmake with make, like me ahhahah

All these in NNPACK directory mkdir build cd build cmake -G "Unix Makefiles" .. make

[GiorgioSgl]

After I get an error when try to build the darknet-nnpack I create a new issue try to check it got solved before test the comment above

[kkhsuk]

I didn't see this issue by my rasp pi 3B

after the solved cmake issue and we gave up use darknet-nnpack when I send the email to digitalbrain79 he said try to use the rasp pi os image in the Readme file link and he said darknet-nnpack is old system so he suggest the another framework like tensorflow

when i gave up the darknet-nnpack, I use the tf lite withe the google coral usb accelerator

this is the only advice I can give you

-----Original Message----- From: "Giorgio Segalla"[email protected] To: "digitalbrain79/darknet-nnpack"[email protected]; Cc: "kkhsuk"[email protected]; "Author"[email protected]; Sent: 2021-03-10 (수) 20:29:38 (GMT+09:00) Subject: Re: [digitalbrain79/darknet-nnpack] $NINJA_PATH/ninja step error issue (#60)

After I get an error when try to build the darknet-nnpack I create a new issue try to check it got solved before test the comment above — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[GiorgioSgl]

I didn't see this issue by my rasp pi 3B after the solved cmake issue and we gave up use darknet-nnpack when I send the email to digitalbrain79 he said try to use the rasp pi os image in the Readme file link and he said darknet-nnpack is old system so he suggest the another framework like tensorflow when i gave up the darknet-nnpack, I use the tf lite withe the google coral usb accelerator this is the only advice I can give you … -----Original Message----- From: "Giorgio Segalla"[email protected] To: "digitalbrain79/darknet-nnpack"[email protected]; Cc: "kkhsuk"[email protected]; "Author"[email protected]; Sent: 2021-03-10 (수) 20:29:38 (GMT+09:00) Subject: Re: [digitalbrain79/darknet-nnpack] $NINJA_PATH/ninja step error issue (#60) After I get an error when try to build the darknet-nnpack I create a new issue try to check it got solved before test the comment above — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

Yeah I agree, but I have already implemented yolov4-tiny in the tflite and I got very bad result in the mAP50 (29%) while by state of the art it would get 40%. That's the main reason I was trying to use the darknetNNPACK, because with the normal version of darknet I got 0.1 FPS very bad result. I will try somenew things and hopefully get some better result.