A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 21:00 UTC
[CI/Build] Use spawn multiprocessing mode for V1 test pipeline (#14243)
Signed-off-by: Russell Bryant <rbryant@redhat.com>
Russell Bryant committed
Mar 5, 2025 at 20:08 UTC
ffad94397dd3fe5478919d2e4c785273378b4aab
1 file changed
+3
.buildkite/test-pipeline.yaml
+3
@@ -78,6 +78,7 @@ steps:
78
- tests/basic_correctness/test_preemption
79
- tests/basic_correctness/test_cumem.py
80
commands:
81
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
82
- pytest -v -s basic_correctness/test_cumem.py
83
- pytest -v -s basic_correctness/test_basic_correctness.py
84
- pytest -v -s basic_correctness/test_cpu_offload.py
@@ -112,6 +113,7 @@ steps:
113
- tests/entrypoints/test_chat_utils
114
- tests/entrypoints/offline_mode
115
commands:
116
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
117
- pytest -v -s entrypoints/llm --ignore=entrypoints/llm/test_lazy_outlines.py --ignore=entrypoints/llm/test_generate.py --ignore=entrypoints/llm/test_generate_multiple_loras.py --ignore=entrypoints/llm/test_guided_generate.py --ignore=entrypoints/llm/test_collective_rpc.py
118
- pytest -v -s entrypoints/llm/test_lazy_outlines.py # it needs a clean process
119
- pytest -v -s entrypoints/llm/test_generate.py # it needs a clean process
@@ -196,6 +198,7 @@ steps:
198
- tests/v1
199
commands:
200
# split the test to avoid interference
201
+ - export VLLM_WORKER_MULTIPROC_METHOD=spawn
202
- VLLM_USE_V1=1 pytest -v -s v1/core
203
- VLLM_USE_V1=1 pytest -v -s v1/engine
204
- VLLM_USE_V1=1 pytest -v -s v1/sample