[ Misc ] fbgemm checkpoints (#6559)
This commit is contained in:
@@ -52,8 +52,8 @@ class SqueezeLLMConfig(QuantizationConfig):
|
||||
weight_bits = cls.get_from_keys(config, ["wbits"])
|
||||
return cls(weight_bits)
|
||||
|
||||
def get_quant_method(
|
||||
self, layer: torch.nn.Module) -> Optional[QuantizeMethodBase]:
|
||||
def get_quant_method(self, layer: torch.nn.Module,
|
||||
prefix: str) -> Optional[QuantizeMethodBase]:
|
||||
if isinstance(layer, LinearBase):
|
||||
return SqueezeLLMLinearMethod(self)
|
||||
return None
|
||||
|
||||
Reference in New Issue
Block a user