biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Kernel] Delegate construction of FusedMoEQuantConfig to FusedMoEMethodBase subclasses (#22537)

Browse Source 
Signed-off-by: Bill Nell <bnell@redhat.com>
This commit is contained in:
bnellnm
2025-09-17 19:43:31 -04:00
committed by GitHub
parent e6585ddb45
commit 5963b98b46
68 changed files with 2698 additions and 2526 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/kernels/moe/modular_kernel_tools/make_feature_matrix.py
View File

@@ -10,7 +10,8 @@ import torch
from tqdm import tqdm
from vllm.config import VllmConfig, set_current_vllm_config
from vllm.model_executor.layers.fused_moe.config import FusedMoEQuantConfig
from vllm.model_executor.layers.fused_moe.config import (
FUSED_MOE_UNQUANTIZED_CONFIG)
from vllm.platforms import current_platform
from .common import (Config, RankTensors, WeightTensors, reference_moe_impl,
@@ -86,7 +87,7 @@ def make_feature_matrix(csv_file_path: str):
quant_config_dict = config_dict['quant_config']
del config_dict['quant_config']
if quant_config_dict is None:
quant_config = FusedMoEQuantConfig(None)
quant_config = FUSED_MOE_UNQUANTIZED_CONFIG
quant_config_dict = asdict(quant_config)
config_dict |= quant_config_dict