Hi Intel team,
Could you please describe here (point, copy-paste if this is a duplicate)
exact steps to reproduce INT8 quantization benefit which is described here
under the section Performance Counters ?
Since there are a lot of links about how-to (FP32->INT8) but there is no any successful story so far as I see.
Either config file is missing and should be created self or the same should be done for annotation sets.
Shouldn't it all be placed into some one place for this feature verification/estimation by end user before any development?
For more complete information about compiler optimizations, see our Optimization Notice.