[Quantization] Support online quantization with partially pre-quantized checkpoints (#51392)
Signed-off-by: Felix Marty <Felix.Marty@amd.com> Signed-off-by: mgoin <mgoin64@gmail.com> Co-authored-by: Codex <codex@openai.com> Co-authored-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
co-authored by
Codex
mgoin
parent
a11dfcff89
commit
5690b02c03
@@ -56,5 +56,5 @@ class ExpertsInt8Config(QuantizationConfig):
|
||||
if isinstance(layer, LinearBase):
|
||||
return UnquantizedLinearMethod()
|
||||
elif isinstance(layer, RoutedExperts):
|
||||
return Int8OnlineMoEMethod(layer=layer)
|
||||
return Int8OnlineMoEMethod(moe=layer.moe_config)
|
||||
return None
|
||||
|
||||
Reference in New Issue
Block a user