apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity

[Iluvatar] refactor attn and moe code (#6887)

Browse Source
This commit is contained in:
yzwu
2026-03-18 10:31:00 +08:00
committed by GitHub
parent 0359794e08
commit 8b890c0d72
16 changed files with 877 additions and 140 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/layers/quantization/weight_only.py
+16
View File
@@ -149,6 +149,22 @@ class WeightOnlyConfig(QuantConfigBase):
else:
return GPUWeightOnlyLinearMethod(self)
elif current_platform.is_iluvatar():
if isinstance(layer, FusedMoE):
if layer.use_method == "cutlass":
from fastdeploy.model_executor.layers.backends import (
IluvatarCutlassWeightOnlyMoEMethod,
)
return IluvatarCutlassWeightOnlyMoEMethod(self)
else:
raise ValueError(f"Unsupported MOE backend {layer.use_method}")
else:
from fastdeploy.model_executor.layers.backends import (
IluvatarWeightOnlyLinearMethod,
)
return IluvatarWeightOnlyLinearMethod(self)
else:
if isinstance(layer, FusedMoE):
if layer.use_method == "cutlass":