Unable to profile or deploy a simple TensorFlow SVC

Report bugs
#1

Mamdouh / myanni-project-1

Running environment with following details:
Python: 3.11.4 | packaged by Anaconda, Inc. | (main, Jul 5 2023, 13:38:37) [MSC v.1916 64 bit (AMD64)]
scipy: 1.10.1
numpy: 1.24.3
matplotlib: 3.7.1
pandas: 1.5.3
sklearn: 1.3.1

I am trying to profile and deploy a simple TensorFlow SVC (Support Vector Classifier) using the following command:
profile = ei.model.profile(model = model, device = ‘cortex-m4f-80mhz’)

And the resulting debug output from Edge Impulse using Spyder 5.4.3 is:
edgeimpulse.exceptions.InvalidModelException: Was unable to load_model of type <class ‘sklearn.svm._classes.SVC’> with exception Unexpected model type

I am not sure where my mistake is. Please help!

Warm regards
Mamdouh

#2

Hi @myanni,

Edge Impulse does not support support vector machines (SVM) at this time. Only TFLite operations for neural networks are supported. If you are looking to deploy to a microcontroller, only a subset of those operations are supported (for a reference, please see this documentation).

#3

Thank you Shawn for the quick response. It is most appreciated.

Any idea if there any converters out there that can take an TensorFlow SVC and provide C/C++ code (deploy) as well as show RAM/FLASH/Inference Execution time (profile)? Thank you again.

Warm regards
Mamdouh

#4

Hi @myanni,

I do not off the top of my head, but Google turned up some results that might be useful:

#5

Thank you Shawn. I have looked through those provided links before, but thank you nonetheless.

Warm regards
Mamdouh

1 Like