[RPI Zero] ModuleNotFoundError: No module named 'tflite_runtime'

[rajurimu]

Please make sure that this is a build/installation issue. As per our GitHub Policy, we only address code/doc bugs, performance issues, feature requests and build/installation issues on GitHub. tag:build_template

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Raspbian GNU/Linux 10 (buster)
• Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device: Raspberry Pi Zero
• TensorFlow installed from (source or binary): Compile natively on Raspberry Pi using https://www.tensorflow.org/lite/guide/build_rpi
• TensorFlow version: Trying to install and Tensorflow lite as a standalone
• Python version: 3.7.3
• Installed using virtualenv? pip? conda?:
• Bazel version (if compiling from source): None
• GCC/Compiler version (if compiling from source):
• CUDA/cuDNN version:
• GPU model and memory:

Describe the problem
Trying to install Tensorflow Lite on RPI Zero as a standalone. However once installation is complete as per the following steps,
sudo apt-get install build-essential
git clone https://github.com/tensorflow/tensorflow.git tensorflow_src
cd tensorflow_src && ./tensorflow/lite/tools/make/download_dependencies.sh
./tensorflow/lite/tools/make/build_rpi_lib.sh

I can not import,
import tflite_runtime.interpreter as tflite

Error I am getting is
ModuleNotFoundError: No module named 'tflite_runtime'

In addition as per the instructions, I should see a static library. I do not see this either.
tensorflow/lite/tools/make/gen/lib/rpi_armv6/libtensorflow-lite.a

Any other info / logs
Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached.

[rajurimu]

I am still facing this issue on RPI Zero.

[terryheo]

This are prebuilt Python wheels of tflite_runtime package. https://www.tensorflow.org/lite/guide/python
But I'm not sure it's compatible with RPI Zero.

If it doesn't work, you might need to build it by yourself.
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/tools/pip_package will help.

[rajurimu]

t

This are prebuilt Python wheels of tflite_runtime package. https://www.tensorflow.org/lite/guide/python
But I'm not sure it's compatible with RPI Zero.

If it doesn't work, you might need to build it by yourself.
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/tools/pip_package will help.

I am using https://www.tensorflow.org/lite/guide/build_rpi. For Compile natively on Raspberry Pi.

The site states that "The following instructions have been tested on Raspberry Pi Zero, Raspbian GNU/Linux 10 (buster), gcc version 8.3.0 (Raspbian 8.3.0-6+rpi1)".

So I am puzzled at this thing not working on rpi zero and giving the error. I have done it a few times with the same error. The site which you are referring to does not have arm6l whl.

[terryheo]

libtensorflow-lite.a will work with your RPI zero. But it's C library not Python.

You might be able to build it on RPI zero with the following command.

sudo apt install swig libjpeg-dev zlib1g-dev python3-dev python3-numpy
pip install numpy pybind11
sh tensorflow/lite/tools/make/download_dependencies.sh
sh tensorflow/lite/tools/pip_package/build_pip_package.sh

[rajurimu]

libtensorflow-lite.a will work with your RPI zero. But it's C library not Python.

You might be able to build it on RPI zero with the following command.

sudo apt install swig libjpeg-dev zlib1g-dev python3-dev python3-numpy
pip install numpy pybind11
sh tensorflow/lite/tools/make/download_dependencies.sh
sh tensorflow/lite/tools/pip_package/build_pip_package.sh

terryheo...it ran something as it says on webpage but it is giving error. I am trying on RP4 now as RP0 is slow and if I can make whl on RP4, I can try on RP0 later.

I had to modify commands like this 'cd tensorflow_src && ./tensorflow/lite/tools/pip_package/build_pip_package.sh'

collect2: error: ld returned 1 exit status
make: *** [tensorflow/lite/tools/make/Makefile:341: /home/pi/tensorflow_src/tensorflow/lite/tools/make/gen/linux_armv7l/bin/minimal] Error 1
make: Leaving directory '/home/pi/tensorflow_src'
Traceback (most recent call last):
File "setup.py", line 227, in
'build_py': CustomBuildPy,
File "/usr/local/lib/python3.7/dist-packages/setuptools/init.py", line 165, in setup
return distutils.core.setup(**attrs)
File "/usr/lib/python3.7/distutils/core.py", line 148, in setup
dist.run_commands()
File "/usr/lib/python3.7/distutils/dist.py", line 966, in run_commands
self.run_command(cmd)
File "/usr/lib/python3.7/distutils/dist.py", line 985, in run_command
cmd_obj.run()
File "/usr/lib/python3.7/distutils/command/bdist.py", line 143, in run
self.run_command(cmd_name)
File "/usr/lib/python3.7/distutils/cmd.py", line 313, in run_command
self.distribution.run_command(command)
File "/usr/lib/python3.7/distutils/dist.py", line 985, in run_command
cmd_obj.run()
File "/usr/lib/python3.7/distutils/command/bdist_dumb.py", line 81, in run
self.run_command('build')
File "/usr/lib/python3.7/distutils/cmd.py", line 313, in run_command
self.distribution.run_command(command)
File "/usr/lib/python3.7/distutils/dist.py", line 985, in run_command
cmd_obj.run()
File "/usr/lib/python3.7/distutils/command/build.py", line 135, in run
self.run_command(cmd_name)
File "/usr/lib/python3.7/distutils/cmd.py", line 313, in run_command
self.distribution.run_command(command)
File "/usr/lib/python3.7/distutils/dist.py", line 985, in run_command
cmd_obj.run()
File "setup.py", line 133, in run
self.run_command('build_ext')
File "/usr/lib/python3.7/distutils/cmd.py", line 313, in run_command
self.distribution.run_command(command)
File "/usr/lib/python3.7/distutils/dist.py", line 985, in run_command
cmd_obj.run()
File "setup.py", line 125, in run
make()
File "setup.py", line 105, in make
subprocess.check_call(make_args(quiet=False))
File "/usr/lib/python3.7/subprocess.py", line 347, in check_call
raise CalledProcessError(retcode, cmd)
subprocess.CalledProcessError: Command '['make', 'SHELL=/bin/bash', 'BUILD_WITH_NNAPI=false', '-C', '/home/pi/tensorflow_src/tensorflow/lite/tools/pip_package/../../../..', '-f', 'tensorflow/lite/tools/make/Makefile', '-j', '1']' returned non-zero exit status 2.

[rajurimu]

and where does it create whl file, when it creates.

[terryheo]

If you're using RPI4, just use prebuilt from https://www.tensorflow.org/lite/guide/python

[rajurimu]

If you're using RPI4, just use prebuilt from https://www.tensorflow.org/lite/guide/python

I did that but my end goal is to work on RPi Zero and I am struggling. Not an expert so learning as I go. The instructions are failing from the link. Not sure if I am missing something. The sh commands are not working as it is so I am modifying as mentioned above.

[freeedom67]

FYI

I was able to find a working solution for running tensorflow(lite as well) on the ARM6 based raspberry pi (like the zero or the Pi1):

• get a pre compilied tf python wheel from https://github.com/lhelontra/tensorflow-on-arm/releases or compile it on your own.
• Install it on the device using pip3.

in python:

import tensorflow as  tf

print(str(tf.__version__)) # should print the version
interpreter = tf.lite.Interpreter(model_path) # for example if you just need the python tf lite runtime 

After a long time looking for a good solution and struggling with static libs and cross compiling, I found this working pretty well and easy!

[mohantym]

Hi @rajurimu ! Inline with above comment.
You can also import like below.
import tflite-runtime.interpreter as tflite instead of
import tflite_runtime.interpreter as tflite

Attached relevant threads for reference. link1,link2,link3,link4 . Thanks!

[google-ml-butler]

This issue has been automatically marked as stale because it has no recent activity. It will be closed if no further activity occurs. Thank you.

[google-ml-butler]

Closing as stale. Please reopen if you'd like to work on this further.

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No