Skip to content

MLBEDSW-10626: Fix quantization when decomposing transpose

Johan Gunnarsson requested to merge MLBEDSW-10626 into main

When decomposing a transpose op into a sequence of new ops, make all new ops use unit quantization, except ops that write to the original output tensor. Those ops should use the original IFM/OFM quantization.

Signed-off-by: Johan Gunnarsson johan.gunnarsson@arm.com Change-Id: I1cfb52280dff9365548e15c68f7975927e44a104

Merge request reports