apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-05 08:37:06 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

load hadamard_block_size from config (#3797)

Browse Source
This commit is contained in:
Yuan Xiaolan
2025-09-05 17:07:58 +08:00
committed by GitHub
parent 41aee08982
commit 2cf55168ca
10 changed files with 60 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
fastdeploy/model_executor/layers/quantization/w4a8.py
View File

@@ -25,9 +25,10 @@ class W4A8Config(QuantConfigBase):
quantization config for weight 4bits and activation 8bits
"""
def __init__(self, is_permuted) -> None:
def __init__(self, is_permuted, hadamard_block_size) -> None:
super().__init__()
self.is_permuted = is_permuted
self.hadamard_block_size = hadamard_block_size
def name(self) -> str:
return "w4a8"
@@ -35,7 +36,8 @@ class W4A8Config(QuantConfigBase):
@classmethod
def from_config(cls, config: dict) -> "W4A8Config":
is_permuted = config.get("is_permuted", True)
return cls(is_permuted)
hadamard_block_size = config.get("hadamard_block_size", 128)
return cls(is_permuted, hadamard_block_size)
def get_quant_method(self, layer) -> Optional[QuantMethodBase]:
if isinstance(layer, FusedMoE):