[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
@@ -84,7 +84,7 @@ class GGUFConfig(QuantizationConfig):
@classmethod
def override_quantization_method(
cls, hf_quant_cfg: dict[str, Any], user_quant: str | None
cls, hf_quant_cfg: dict[str, Any], user_quant: str | None, hf_config=None
) -> "QuantizationMethods | None":
# When user explicitly specifies --quantization gguf, override
# whatever quantization method is in the HF model config (e.g. fp8).