@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / 0583578f42
Fork 0

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] breaks down V1 Test into 3 groups of approx 30 minutes runtime (#23757)

Signed-off-by: Jean Schmidt <contato@jschmidt.me>

Jean Schmidt committed Aug 28, 2025 at 17:59 UTC 0583578f42fb23cc8a6d612e041c6be402551282
1 file changed +20 -6
.buildkite/test-pipeline.yaml
+20 -6
@@ -234,16 +234,33 @@ steps:
234 # OOM in the CI unless we run this separately
235 - pytest -v -s tokenization
236
237 -- label: V1 Test
237 +- label: V1 Test e2e + engine
238 mirror_hardwares: [amdexperimental]
239 source_file_dependencies:
240 - vllm/
241 - tests/v1
242 commands:
243 - # split the test to avoid interference
244 - - pytest -v -s v1/core
243 + # TODO: accuracy does not match, whether setting
244 + # VLLM_USE_FLASHINFER_SAMPLER or not on H100.
245 + - pytest -v -s v1/e2e
246 - pytest -v -s v1/engine
247 +
248 +- label: V1 Test entrypoints
249 + mirror_hardwares: [amdexperimental]
250 + source_file_dependencies:
251 + - vllm/
252 + - tests/v1
253 + commands:
254 - pytest -v -s v1/entrypoints
255 +
256 +- label: V1 Test others
257 + mirror_hardwares: [amdexperimental]
258 + source_file_dependencies:
259 + - vllm/
260 + - tests/v1
261 + commands:
262 + # split the test to avoid interference
263 + - pytest -v -s v1/core
264 - pytest -v -s v1/executor
265 - pytest -v -s v1/sample
266 - pytest -v -s v1/logits_processors
@@ -256,9 +273,6 @@ steps:
273 - pytest -v -s v1/test_utils.py
274 - pytest -v -s v1/test_oracle.py
275 - pytest -v -s v1/test_metrics_reader.py
259 - # TODO: accuracy does not match, whether setting
260 - # VLLM_USE_FLASHINFER_SAMPLER or not on H100.
261 - - pytest -v -s v1/e2e
276 # Integration test for streaming correctness (requires special branch).
277 - pip install -U git+https://github.com/robertgshaw2-redhat/lm-evaluation-harness.git@streaming-api
278 - pytest -v -s entrypoints/openai/correctness/test_lmeval.py::test_lm_eval_accuracy_v1_engine