running swin-transformer tflite on npu

[noway]

hi,  community

# iMX 8MP

@ Harvey021

I missed an issue during model deployment.

When running the TFLite model with label_image.py, it indicates that the flex op is not supported.

By the way, I have already copied "libtensorflow_lite.so" to "/usr/lib". The dynamic lib is complied according to the official documentation.

target image version: linux6.6.36 full

target machine: i.MX 8mp

model: swin-transformer, tflite, quant

[noway]

Sorry, I need modify a clerical error: ”libtensorflow_lite.so“ -> "libtensorflow_flex.so".

[danielchen]

Hi noway:

This error means your model contains Flex ops -TensorFlow operations that are not natively supported by TFLite and require the Flex delegate to run.

Can you try the latest version  BSP 6.12.34 and try it again?

[noway]

Thank you for your reply. Are there any other solutions?

Can I running the model that have the "flex " by python in the BSP 6.12.34? If not, is there a demo with c++？