My question is same as the title.
As my understanding, eIQ is software where we can train model, convert model (to tflite fro example) and visualize it. Before I traing Tensorflow model and convert to tflite model (not using eIQ) and deploy on MIMXRT board. Can I use this model to deploy on MCXN947 with NPU mode? Is there any special things to develop tensorflow lite model for MCXN947 with NPU mode?
Solved! Go to Solution.
Hi @nnxxpp
When you need to deploy it on MCXN947, you need to convert TensorFlow lite for Neutron using convert tool.
You can refer to it.
eIQ Neutron NPU Lab Guides - NXP Community
BR
Harry
Thank you. I could convert to NPU Tflite. I see that th output model is also .tflite model with custom operators NeutronGraph. How can I evaluate this NPU tflite model on PC with Python?
Do you have any guide? I known that NeutronGraph is the custom operator and we need have its implementation to use in Python.
Thank you. I understood.
Hi @nnxxpp
When you need to deploy it on MCXN947, you need to convert TensorFlow lite for Neutron using convert tool.
You can refer to it.
eIQ Neutron NPU Lab Guides - NXP Community
BR
Harry