I am converting a keras model to uff to run in int8 precision.
When I look at the keras model’s sumary and the list of layers, there are no layers called “(Unnamed Layer* 3) [Shuffle]” or “input reformatter 0”, which I find in the layerwise profiling.
My question is what are these new layers and what are their purpose?
- input reformatter 0
- (Unnamed Layer* 3) [Shuffle]
What do they do and why are they there in the uff model?