From 8996733307c5c84e297da10440c68fd32af32e7c Mon Sep 17 00:00:00 2001 From: Mengqing Cao Date: Mon, 24 Mar 2025 16:09:06 +0800 Subject: [PATCH] [CI] fix vllm test (#365) fix vllm test Signed-off-by: MengqingCao --- .github/workflows/vllm_ascend_test.yaml | 11 ++++++----- pytest.ini | 5 +++++ 2 files changed, 11 insertions(+), 5 deletions(-) diff --git a/.github/workflows/vllm_ascend_test.yaml b/.github/workflows/vllm_ascend_test.yaml index a5a9a4d..aa51f20 100644 --- a/.github/workflows/vllm_ascend_test.yaml +++ b/.github/workflows/vllm_ascend_test.yaml @@ -130,11 +130,12 @@ jobs: run: | VLLM_USE_V1=0 pytest -sv tests - # FIXME: make vllm test pass - #- name: Checkout vllm-project/vllm repo - #- name: Run vllm-project/vllm test - # run: | - # VLLM_USE_V1=0 pytest -sv + - name: Run vllm-project/vllm test + env: + VLLM_USE_V1: 0 + PYTORCH_NPU_ALLOC_CONF: max_split_size_mb:256 + run: | + pytest -sv post_cleanup: name: vLLM Ascend test (post-cleanup) diff --git a/pytest.ini b/pytest.ini index a174294..23b1279 100644 --- a/pytest.ini +++ b/pytest.ini @@ -53,6 +53,11 @@ addopts = --ignore=vllm-empty/tests/test_utils.py --ignore=vllm-empty/tests/multimodal/test_processing.py --ignore=vllm-empty/tests/multimodal/test_processor_kwargs.py --ignore=vllm-empty/tests/multimodal/test_utils.py + ; Both V1 and V0 engine will be run in detokenizer/test_stop_reason + ; VLLM_USE_V1=1 is not supported with device type=npu. + --ignore=vllm-empty/tests/detokenizer/test_stop_reason.py + ; oom on llama-2-7b-hf + --ignore=vllm-empty/tests/detokenizer/test_stop_strings.py testpaths = vllm-empty/tests