[Quantization] [Refactor] Create special "GptOssMxfp4MoeMethod" (#39604)
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
This commit is contained in:
@@ -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":
|
||||
|
||||
Reference in New Issue
Block a user