Hello,
I’m trying to use the TFLITE demo in the torizon-samples repository. I built the image and added to Portainer, but when I launch the container, i have the error
Traceback (most recent call last):
File "/home/object-detection/object-detection.py", line 13, in <module>
import tflite_runtime.interpreter as tf #Tensorflow_Lite
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ModuleNotFoundError: No module named 'tflite_runtime'
Can somebody check if the docker image is building tensorflow lite properly ? There is a bunch of sh scripts and it’s a bit complicated to go through every step.
Thanks
Hi @luciolis, thanks for reaching out.
Which branch are you using to build the image from? Bookworm or Bullseye?
Thanks
Bookworm, the current main branch. I think the tutorial does not specify a branch.
Hi @luciolis ,
I am currently trying to reproduce the issue.
I will come back to you as soon as possible.
Best Regards
Kevin
Hi @luciolis, @kevin.tx,
I was able to reproduce the issue yesterday. The issue regarding the reported bug is that due to PEP-668 (PEP 668 – Marking Python base environments as “externally managed” | peps.python.org), the python script couldn’t detect the tflite package installed with pip.
I also found out ./tensorflow-lite-vx-delegate_2.8.0.sh
is not properly building the necessary libs after fixing the bug described above, so I’ve raised the issue internally with the team for us to properly work on it.
I’ll post an updated on this thread when this is done.
Regards,
1 Like
Hi @leon.tx ,
can confirm, I saw the same behaviour yesterday as well.
BR,
Kevin
Hi again @luciolis,
Just wanted to let you know that we pushed the fixes for the tflite-rtsp
sample.
Feel free to just git pull
or go directly to GitHub - toradex/torizon-samples: All sample code related to TorizonCore project., copy the code locally and re-build the container.
Cheers,