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 14, 05:30 UTC
[CI]Add transformers_utils to Async Engine, Inputs, Utils, Worker Test (#24615)
Signed-off-by: Xingyu Liu <charlotteliu12x@gmail.com>
Xingyu Liu committed
Sep 11, 2025 at 01:52 UTC
0fc36463e0faed724b20cb45561382cbf2a44972
1 file changed
+4
-2
.buildkite/test-pipeline.yaml
+4
-2
@@ -54,6 +54,7 @@ steps:
54
- tests/utils_
55
- tests/worker
56
- tests/standalone_tests/lazy_imports.py
57
+ - tests/transformers_utils
58
commands:
59
- python3 standalone_tests/lazy_imports.py
60
- pytest -v -s mq_llm_engine # MQLLMEngine
@@ -63,6 +64,7 @@ steps:
64
- pytest -v -s multimodal
65
- pytest -v -s utils_ # Utils
66
- pytest -v -s worker # Worker
67
+ - pytest -v -s transformers_utils # transformers_utils
68
69
- label: Python-only Installation Test # 10min
70
timeout_in_minutes: 20
@@ -822,8 +824,8 @@ steps:
824
# Avoid importing model tests that cause CUDA reinitialization error
825
- pytest models/test_transformers.py -v -s -m 'distributed(num_gpus=2)'
826
- pytest models/language -v -s -m 'distributed(num_gpus=2)'
825
- - pytest models/multimodal -v -s -m 'distributed(num_gpus=2)' --ignore models/multimodal/generation/test_whisper.py
826
- - VLLM_WORKER_MULTIPROC_METHOD=spawn pytest models/multimodal/generation/test_whisper.py -v -s -m 'distributed(num_gpus=2)'
827
+ - pytest models/multimodal -v -s -m 'distributed(num_gpus=2)' --ignore models/multimodal/generation/test_whisper.py
828
+ - VLLM_WORKER_MULTIPROC_METHOD=spawn pytest models/multimodal/generation/test_whisper.py -v -s -m 'distributed(num_gpus=2)'
829
# test sequence parallel
830
- pytest -v -s distributed/test_sequence_parallel.py
831
# this test fails consistently.