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

[Feature] support qkv&gate linear fusion (#6455)

Browse Source 
* [Feature] support qkv&gate linear fusion

* add test
This commit is contained in:
Longzhi Wang
2026-02-24 15:20:29 +08:00
committed by GitHub
parent 38c3e02470
commit 22566168c3
3 changed files with 692 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fastdeploy/model_executor/layers/quantization/block_wise_fp8.py
+2
View File
@@ -23,6 +23,7 @@ from fastdeploy import envs
from fastdeploy.model_executor.layers.linear import (
MergedColumnParallelLinear,
MergedReplicatedLinear,
QKVGateParallelLinear,
QKVParallelLinear,
)
from fastdeploy.model_executor.layers.moe import FusedMoE
@@ -160,6 +161,7 @@ class BlockWiseFP8LinearMethod(QuantMethodBase):
isinstance(layer, MergedColumnParallelLinear)
or isinstance(layer, QKVParallelLinear)
or isinstance(layer, MergedReplicatedLinear)
or isinstance(layer, QKVGateParallelLinear)
):
tensor_output_dim = (self.model_format == "torch") ^ quant_attrs.get("output_dim", True)
quant_attrs = {