[Kernels] Clean up FusedMoeMethodBase and modular kernel setup. Remove extra arguments from modular kernel methods. (#22035)
Signed-off-by: Bill Nell <bnell@redhat.com> Co-authored-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
@@ -241,7 +241,7 @@ class AutoRoundConfig(QuantizationConfig):
|
||||
|
||||
if isinstance(layer, FusedMoE):
|
||||
if use_marlin:
|
||||
return AWQMoEMethod(quant_args_marlin)
|
||||
return AWQMoEMethod(quant_args_marlin, layer.moe)
|
||||
from vllm.model_executor.layers.quantization.moe_wna16 import (
|
||||
MoeWNA16Config)
|
||||
|
||||
@@ -339,7 +339,7 @@ class AutoRoundConfig(QuantizationConfig):
|
||||
}
|
||||
return MoeWNA16Config.from_config(config).get_quant_method(
|
||||
layer, prefix)
|
||||
return GPTQMarlinMoEMethod(quant_args_marlin)
|
||||
return GPTQMarlinMoEMethod(quant_args_marlin, layer.moe)
|
||||
|
||||
if isinstance(layer, (LinearBase, ParallelLMHead)):
|
||||
if use_marlin:
|
||||
|
||||
Reference in New Issue
Block a user