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 10, 01:30 UTC
[CI] Move Distributed Tests from H200 -> H100 (#32555)
Robert Shaw committed
Jan 18, 2026 at 13:25 UTC
afc362260257ab16c2ffc6105678e1403c0475cb
1 file changed
+5
-4
.buildkite/test-pipeline.yaml
+5
-4
@@ -1400,18 +1400,19 @@ steps:
1400
- pytest -v -s tests/distributed/test_sequence_parallel.py
1401
- pytest -v -s tests/compile/distributed/test_sequence_parallelism.py
1402
1403
-##### H200 test #####
1404
-- label: Distributed Tests (H200) # optional
1405
- gpu: h200
1403
+- label: Distributed Tests (H100) # optional
1404
+ gpu: h100
1405
optional: true
1406
working_dir: "/vllm-workspace/"
1407
num_gpus: 2
1408
commands:
1409
- VLLM_TEST_CLEAN_GPU_MEMORY=1 pytest -v -s tests/compile/distributed/test_async_tp.py
1410
- pytest -v -s tests/distributed/test_context_parallel.py
1412
- - CUDA_VISIBLE_DEVICES=1,2 VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model=Qwen/Qwen1.5-MoE-A2.7B -tp=1 -dp=2 --max-model-len=2048 --all2all-backend=deepep_high_throughput
1411
+ - VLLM_USE_DEEP_GEMM=1 VLLM_LOGGING_LEVEL=DEBUG python3 examples/offline_inference/data_parallel.py --model=Qwen/Qwen1.5-MoE-A2.7B -tp=1 -dp=2 --max-model-len=2048 --all2all-backend=deepep_high_throughput
1412
- pytest -v -s tests/v1/distributed/test_dbo.py
1413
1414
+##### H200 test #####
1415
+
1416
- label: LM Eval Large Models (H200) # optional
1417
timeout_in_minutes: 60
1418
gpu: h200