Can I use Flex Delegate with tflite_runtime on iMX93?

[dpcivl]

Hi,

I'm currently working on my iMX93 EVK board to test some deep learning models. However, I am encountering difficulties using TensorFlow Select Ops for my FastSpeech2 model. To work with FastSpeech2, I need to use the Flex Delegate.

I've noticed that the iMX93 EVK board does not come with any delegate files (.so). I referred to the IMX Machine Learning User Guide, but I couldn't find any guidelines on this issue.

Could you please let me know if there is a way to build the delegate for the iMX93 EVK board? If not, does this mean that the board does not support Flex Delegate at all?

Additionally, if I compile my application with the Flex Delegate (using the TensorFlow Lite C++ API), will I be able to run my model on the iMX93 board?

I appreciate your prompt response. Thank you for your assistance.

Best regards,

[Zhiming_Liu]

Hi @dpcivl 

The steps about building Flex Delegate applies to i.MX8 and i.MX9 platforms. Of course, you can run your model on i.MX93.

Best Regards
Zhiming

[dpcivl]

I found that I could apply Flex Delegate on the i.MX93 EVK board. However, I have another question: can I use the delegate with Python as well? It seems that the documentation for building Flex Delegate is primarily for C/C++. I would like to know if it's possible to use the delegate for a Python application. Please let me know. Thank you.