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 6, 03:15 UTC
[CI] Add batch invariant test: Block FP8 + small MOE (#37895)
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Wentao Ye committed
Mar 23, 2026 at 21:16 UTC
83c9d525b623566f45d161593e7610cb65aaad60
1 file changed
+3
-1
.buildkite/test_areas/misc.yaml
+3
-1
@@ -192,7 +192,7 @@ steps:
192
- pytest -v -s config
193
194
- label: Batch Invariance (H100)
195
- timeout_in_minutes: 25
195
+ timeout_in_minutes: 30
196
device: h100
197
source_file_dependencies:
198
- vllm/v1/attention
@@ -203,6 +203,8 @@ steps:
203
- pip install pytest-timeout pytest-forked
204
- pytest -v -s v1/determinism/test_batch_invariance.py
205
- pytest -v -s v1/determinism/test_rms_norm_batch_invariant.py
206
+ - VLLM_TEST_MODEL=deepseek-ai/DeepSeek-V2-Lite-Chat pytest -v -s v1/determinism/test_batch_invariance.py::test_v1_generation_is_deterministic_across_batch_sizes_with_needle[TRITON_MLA]
207
+ - VLLM_TEST_MODEL=Qwen/Qwen3-30B-A3B-Thinking-2507-FP8 pytest -v -s v1/determinism/test_batch_invariance.py::test_v1_generation_is_deterministic_across_batch_sizes_with_needle[FLASH_ATTN]
208
209
- label: Acceptance Length Test (Large Models) # optional
210
timeout_in_minutes: 25