xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-14 12:57:03 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Model] Use the same fused_moe configs for all H200 devices (#23642)

Browse Source 
Signed-off-by: Roger Meier <r.meier@siemens.com>
This commit is contained in:
Roger Meier 2025-10-31 01:36:56 +08:00 committed by GitHub
parent 1004205795
commit 2918c1b49c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 3 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
vllm/model_executor/layers/fused_moe/fused_moe.py
View File

@ -818,6 +818,9 @@ def get_config_file_name(
E: int, N: int, dtype: str | None, block_shape: list[int] | None = None
) -> str:
device_name = current_platform.get_device_name().replace(" ", "_")
# Set device_name to H200 if a device from the H200 family is detected
if "H200" in device_name:
device_name = "H200"
dtype_selector = "" if not dtype else f",dtype={dtype}"
block_shape_selector = (
"" if not block_shape or not all(block_shape) else f",block_shape={block_shape}"