Wasn't able to get it to work: Pi Zero image available?

[Tylersuard]

Hello. I wasn't able to get this to work. Do you have an image of a Pi Zero I can download and use instead? Thank you!

Ps, here is my error message:

• cd /home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../..
• CC_PREFIX=arm-linux-gnueabihf-
• make -j 3 -f tensorflow/lite/tools/make/Makefile TARGET=rpi TARGET_ARCH=armv6
  arm-linux-gnueabihf-g++ -O3 -DNDEBUG -fPIC --std=c++11 -march=armv6 -mfpu=vfp -funsafe-math-optimizations -ftree-vectorize -fPIC -I. -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../../../ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../../../../ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/eigen -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/absl -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/gemmlowp -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/neon_2_sse -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/farmhash/src -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/flatbuffers/include -I -I/usr/local/include -c tensorflow/lite/arena_planner.cc -o /home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/arena_planner.o
  arm-linux-gnueabihf-gcc -O3 -DNDEBUG -fPIC -I. -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../../../ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../../../../ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/eigen -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/absl -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/gemmlowp -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/neon_2_sse -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/farmhash/src -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/flatbuffers/include -I -I/usr/local/include -c tensorflow/lite/c/c_api_internal.c -o /home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/c/c_api_internal.o
  arm-linux-gnueabihf-g++ -O3 -DNDEBUG -fPIC --std=c++11 -march=armv6 -mfpu=vfp -funsafe-math-optimizations -ftree-vectorize -fPIC -I. -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../../../ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../../../../ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/eigen -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/absl -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/gemmlowp -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/neon_2_sse -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/farmhash/src -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/flatbuffers/include -I -I/usr/local/include -c tensorflow/lite/allocation.cc -o /home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/allocation.o
  arm-linux-gnueabihf-g++ -O3 -DNDEBUG -fPIC --std=c++11 -march=armv6 -mfpu=vfp -funsafe-math-optimizations -ftree-vectorize -fPIC -I. -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../../../ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/../../../../../../ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/ -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/eigen -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/absl -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/gemmlowp -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/neon_2_sse -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/farmhash/src -I/home/tdoggo/tensorflow/tensorflow/lite/tools/make/downloads/flatbuffers/include -I -I/usr/local/include -c tensorflow/lite/core/api/error_reporter.cc -o /home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/core/api/error_reporter.o
  In file included from tensorflow/lite/core/api/error_reporter.cc:15:0:
  ./tensorflow/lite/core/api/error_reporter.h: In destructor ‘virtual tflite::ErrorReporter::~ErrorReporter()’:
  ./tensorflow/lite/core/api/error_reporter.h:37:28: sorry, unimplemented: Thumb-1 hard-float VFP ABI
  virtual ~ErrorReporter() {}
  ^
  In file included from /usr/arm-linux-gnueabihf/include/c++/5/bits/stl_algobase.h:62:0,
  from /usr/arm-linux-gnueabihf/include/c++/5/memory:62,
  from ./tensorflow/lite/arena_planner.h:18,
  from tensorflow/lite/arena_planner.cc:15:
  /usr/arm-linux-gnueabihf/include/c++/5/ext/type_traits.h: In function ‘bool __gnu_cxx::__is_null_pointer(std::nullptr_t)’:
  /usr/arm-linux-gnueabihf/include/c++/5/ext/type_traits.h:160:35: sorry, unimplemented: Thumb-1 hard-float VFP ABI
  __is_null_pointer(std::nullptr_t)
  ^
  tensorflow/lite/tools/make/Makefile:216: recipe for target '/home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/core/api/error_reporter.o' failed
  make: *** [/home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/core/api/error_reporter.o] Error 1
  make: *** Waiting for unfinished jobs....
  In file included from /usr/arm-linux-gnueabihf/include/stdio.h:933:0,
  from /usr/arm-linux-gnueabihf/include/c++/5/cstdio:42,
  from ./tensorflow/lite/allocation.h:20,
  from tensorflow/lite/allocation.cc:16:
  /usr/arm-linux-gnueabihf/include/bits/stdio.h: In function ‘int getchar()’:
  /usr/arm-linux-gnueabihf/include/bits/stdio.h:44:14: sorry, unimplemented: Thumb-1 hard-float VFP ABI
  getchar (void)
  ^
  tensorflow/lite/tools/make/Makefile:216: recipe for target '/home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/allocation.o' failed
  make: *** [/home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/allocation.o] Error 1
  tensorflow/lite/tools/make/Makefile:216: recipe for target '/home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/arena_planner.o' failed
  make: *** [/home/tdoggo/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/arena_planner.o] Error 1

[rushil791]

Hi Tyler,

I had a similar issue. I found that reverting back to a specific version of Tensorflow Lite (1.13.1 worked for me) fixed all issues and the Makefile worked as expected.

Let me know if you still have issues, I've misplaced my SD card but I have documented all the steps I took so I'm happy to send them over.

[Tylersuard]

I would appreciate knowing the steps! I'm having a hard time finding Tflow 1.13.1 Thank you!

[cloudwiser]

Apologies for the slow response folks - not getting GH notifications for some reason.
Let me know if you are still blocked and I'll investigate - I may have missed some obvious steps and/or the dependencies have changed...

[ONLYstcm]

I seem to have a similar problem. I get the following error when I try to complete the compilation:

collect2: error: ld returned 1 exit status
make: *** [tensorflow/lite/tools/make/Makefile:316: /home/pi/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/bin/label_image] Error 1

'benchmark' executable is in the bin folder but 'minimal' and 'label_image' are missing. 'libtensorflow-lite.a' and 'benchmark-lib.a' are in the lib folder.

[ONLYstcm]

Just downloaded Tensorflow v1.13.1. When I try to compile I get the following error:

make: *** No rule to make target '/home/pi/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/experimental/ruy/allocator.o', needed by '/home/pi/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/lib/libtensorflow-lite.a'. Stop.

[cloudwiser]

Just to confirm - these errors were the output of ./tensorflow/lite/tools/make/build_rpi_armv6_lib.sh running against TF 1.13.1 on Pi Zero W?

[ONLYstcm]

Yes. I used the script: ./tensorflow/lite/tools/make/build_rpi_armv6_lib.sh, I actually tried compiling again with TF 1.13.1 and an older Makefile model (the one from 13 March) and it compiled for a longer time but eventually got the follwing error

collect2: error: ld returned 1 exit status
make: *** [tensorflow/lite/tools/make/Makefile:252: /home/pi/tensorflow-1.13.1/tensorflow/lite/tools/make/gen/rpi_armv6/bin/benchmark_model] Error 1

'libtensorflow-lite.a' and 'benchmark-lib.a' are present in the lib folder.

[cloudwiser]

collect2: error: ld returned 1 exit status

This is usually the consequence of previous errors so you'd need to look earlier in the build process to see if other dependencies are missing/failing.
Unfortunately I'm running TF 1.14.1 but is this on a clean install of TF 1.13.1 on the Pi Zero?

[ONLYstcm]

Yes, this was a clean install of TF 1.13.1. There were alot of undefined reference messages to files starting with `_atomic'. However, I can try compiling with TF 1.14.1 and see if there's any success. Is this the recommended Makefile for TF 1.14.1?

[cloudwiser]

Yes - this was the original makefile in TF 1.14.1 but only builds the lib, minimal and benchmark.
You would still need build_rpi_armv6_lib.sh or its command line flags to set the target and target arch for rpi and armv6 if you are building on a RPi Zero W
I'll try a clean build on 1.14.1 tomorrow and double check

[cloudwiser]

OK - after a make clean, I have re-run .../build_rpi_armv6_lib.sh with the repo Makefile overnight and I have benchmark.lib and libtensorflow-lite.a in /home/pi/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/lib and label_image, benchmark-model and minimal binaries in /home/pi/tensorflow/tensorflow/lite/tools/make/gen/rpi_armv6/bin. Note you need to stop, change swap space size and start the swapfile service as per the README.mdfor this to complete without errors.
Having checked the RELEASE.mdin tensorflow root directory, I'm running 1.14.0...so I'll check to see if the incremental changes to 1.14.1 may have impacted the build process

[ONLYstcm]

Swap space is at 2GB, double checked with htop. I've just downloaded 1.14.0 source file from the tensorflow library and very early during compilation (with Makefile and Makefile 1.14.0) I get the following error:

In file included from ./tensorflow/lite/core/api/op_resolver.h:20,
from ./tensorflow/lite/core/api/flatbuffer_conversions.h:24,
from tensorflow/lite/core/api/flatbuffer_conversions.cc:16:
./tensorflow/lite/schema/schema_generated.h:21:10: fatal error: flatbuffers/flatbuffers.h: No such file or directory
#include "flatbuffers/flatbuffers.h"
^~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
make: *** [tensorflow/lite/tools/make/Makefile:266: /home/pi/tensorflow-1.14.0/tensorflow/lite/tools/make/gen/rpi_armv6/obj/tensorflow/lite/core/api/flatbuffer_conversions.o] Error 1

If it means anything I'm running 4.19.79+ kernel

[cloudwiser]

That sounds like a fundamental TF Lite build issue - it might be worth a search here or on StackOverflow in case others have reported the same.
Anyway, I'll zip my TF 1.14.0 root folder from my RPi Zero W and host either here or on a G-Drive to hopefully unblock anyone with this or other TF-Lite-on-armv6 build issues

[cloudwiser]

I have shared a zip of my Tensorflow 1.14.0 folders installed at /home/pi/<tensorflow_unzip_goes_here>/... on my RPi Zero W

This also includes the binaries, model files and sample input images as described in the ObjectDetectionRPiZero repo on GH.

Hope this helps.