[Quantization] [Refactor] Create special "GptOssMxfp4MoeMethod" (#39604)

Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
This commit is contained in:
Yongye Zhu
2026-04-13 12:53:58 -04:00
committed by GitHub
parent 4d042ed85f
commit 739e5945dc
18 changed files with 129 additions and 34 deletions
@@ -130,7 +130,7 @@ class MoeWNA16Config(QuantizationConfig):
@classmethod
def override_quantization_method(
cls, hf_quant_cfg, user_quant
cls, hf_quant_cfg, user_quant, hf_config=None
) -> QuantizationMethods | None:
can_convert = cls.is_moe_wna16_compatible(hf_quant_cfg)
if can_convert and user_quant == "moe_wna16":