[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
@@ -110,13 +110,22 @@ class QuantizationConfig(ABC):
@classmethod
def override_quantization_method(
cls, hf_quant_cfg, user_quant
cls,
hf_quant_cfg: dict[str, Any],
user_quant: str | None,
hf_config: Any = None,
) -> QuantizationMethods | None:
"""
Detects if this quantization method can support a given checkpoint
format by overriding the user specified quantization method --
this method should only be overwritten by subclasses in exceptional
circumstances
circumstances.
Args:
hf_quant_cfg: The checkpoint's quantization config dict.
user_quant: The user-specified quantization method string.
hf_config: The HuggingFace model config object (e.g. for
model_type checks). May be None if not available.
"""
return None