[Bugifx][INC] Fix INC quantization method selection for non-quantized layers (#47237)

Signed-off-by: lvliang-intel <liang1.lv@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Liang Lv
2026-09-01 10:40:47 +00:00
committed by GitHub
co-authored by Kunshang Ji mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 754d5e1f65
commit 92ccd2c306
2 changed files with 91 additions and 2 deletions
@@ -328,9 +328,11 @@ class INCConfig(QuantizationConfig):
if (
layer_name == prefix or layer_name == f"model.{prefix}"
) and self.extra_config[layer_name].get("bits", 16) >= 16:
if isinstance(layer, (LinearBase, ParallelLMHead)):
return UnquantizedLinearMethod()
if isinstance(layer, RoutedExperts):
return UnquantizedFusedMoEMethod(layer.moe_config)
return UnquantizedLinearMethod()
return None
layer_config = self.config_parser.resolve(layer, prefix)
if not layer_config.quantized: