Class ov::pass::low_precision::ResolvePrecisionAttribute

class ResolvePrecisionAttribute : public ov::pass::MatcherPass

ResolvePrecisionAttribute resolves multi-precision PrecisionsAttribute on FakeQuantize outputs to a single precision based on the FQ’s output ranges.

After MarkupOptimizations propagates a set of possible PrecisionsAttribute values (e.g. {u8, i8}), this pass visits each FakeQuantize and selects the most suitable precision from the existing set, based on the FQ’s output_low / output_high ranges (signed vs. unsigned). The selection prefers representations that avoid unnecessary zero-points, resulting in a single, resolved precision for downstream consumers.