[ci] Move 3 entrypoints tests to h200_35gb queue (#43164)
Signed-off-by: Simon Mo <simon@inferact.ai> Signed-off-by: Simon Mo <simon@simon-mac-mini-9.local> Co-authored-by: Simon Mo <simon@inferact.ai> Co-authored-by: OpenAI Codex <noreply@openai.com>
Kevin H. Luu committed
Jul 18, 2026 at 08:43 UTC
29c0ec4d63d1869f54a9fbcdf082f77534d9211a
2 files changed
+4
-1
.buildkite/test_areas/entrypoints.yaml
+3
@@ -3,6 +3,7 @@ depends_on:
3
- image-build
4
steps:
5
- label: Entrypoints Unit Tests
6
+ device: h200_35gb
7
key: entrypoints-unit-tests
8
timeout_in_minutes: 25
9
working_dir: "/vllm-workspace/tests"
@@ -15,6 +16,7 @@ steps:
16
- pytest -v -s entrypoints/weight_transfer
17
18
- label: Entrypoints Integration (LLM)
19
+ device: h200_35gb
20
key: entrypoints-integration-llm
21
timeout_in_minutes: 60
22
working_dir: "/vllm-workspace/tests"
@@ -114,6 +116,7 @@ steps:
116
- image-build-amd
117
118
- label: Entrypoints Integration (Responses API)
119
+ device: h200_35gb
120
key: entrypoints-integration-responses-api
121
timeout_in_minutes: 50
122
working_dir: "/vllm-workspace/tests"
tests/entrypoints/openai/responses/test_parsable_context.py
+1
-1
@@ -68,7 +68,7 @@ async def client(server):
68
async def test_basic(client: OpenAI, model_name: str):
69
response = await client.responses.create(
70
model=model_name,
71
- input="What is 123 * 456?",
71
+ input="What is 123 * 456? Answer with only the number.",
72
temperature=0.0,
73
)
74
assert response is not None