Hi,
I started training a fpn-ssd-mobilenetv2 model(which is one of the base models provided by the eIQ software) using custom dataset on the software recently. However, when I converted it into a tflite model and implemented it on a board with NNAPI delegate, it tells me that I'm "attempting to use a delegate that only supports static-sized tensors with a graph that has dynamic-sized tensors" and it failed to apply NNAPI delegate.
I was just wondering what is the cause of this problem and what should I do to avoid it?
Thanks very much!
Hello jasonliu1,
One possible reason here is that when you use eIQ software to convert the custom model, you choose MLIR as tflite converter. I will recommend to use TOCO instead and see whether this problem is avoided. Thank you.