xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-16 16:46:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

[CPU][CI] Fallback sliding window to v0 and fix CPU pooling model tests (#19901)

Browse Source 
Signed-off-by: jiang1.li <jiang1.li@intel.com>
This commit is contained in:
Li, Jiang 2025-06-20 23:30:36 +08:00 committed by GitHub
parent 2e3e3c86dc
commit 79f2f1c2a1
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 13 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
tests/models/language/pooling/test_embedding.py
View File

@ -1,5 +1,7 @@
# SPDX-License-Identifier: Apache-2.0 # SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
import os
import pytest import pytest
from vllm.config import PoolerConfig from vllm.config import PoolerConfig
@ -33,7 +35,7 @@ def v1(run_with_both_engines):
# To avoid this problem, for now we skip v0 since it will be # To avoid this problem, for now we skip v0 since it will be
# deprecated anyway. # deprecated anyway.
pytest.param("ssmits/Qwen2-7B-Instruct-embed-base", pytest.param("ssmits/Qwen2-7B-Instruct-embed-base",
marks=[pytest.mark.skip_v0]), marks=[pytest.mark.skip_v0, pytest.mark.cpu_model]),
# [Encoder-only] # [Encoder-only]
pytest.param("BAAI/bge-base-en-v1.5", pytest.param("BAAI/bge-base-en-v1.5",
marks=[ marks=[
@ -58,6 +60,9 @@ def test_models(
model, model,
monkeypatch, monkeypatch,
) -> None: ) -> None:
if model == "intfloat/e5-mistral-7b-instruct" and current_platform.is_cpu(
) and os.environ.get("VLLM_USE_V1", "0") == "1":
pytest.skip("CPU V1 doesn't support sliding window")
if model == "BAAI/bge-multilingual-gemma2" and current_platform.is_rocm(): if model == "BAAI/bge-multilingual-gemma2" and current_platform.is_rocm():
# ROCm Triton FA does not currently support sliding window attention # ROCm Triton FA does not currently support sliding window attention

7
vllm/engine/arg_utils.py
View File

@ -1449,6 +1449,13 @@ class EngineArgs:
model_config=model_config) and _warn_or_fallback( model_config=model_config) and _warn_or_fallback(
current_platform.device_name): current_platform.device_name):
return False return False
if (current_platform.is_cpu()
and model_config.get_sliding_window() is not None):
_raise_or_fallback(feature_name="sliding window (CPU backend)",
recommend_to_remove=False)
return False
############################################################# #############################################################
return True return True