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 9, 02:30 UTC
[CI][Bugfix] Fix flaky `tests/v1/kv_connector/unit/test_multi_connector.py::test_multi_example_connector_consistency` (#33555)
Signed-off-by: NickLucche <nlucches@redhat.com>
Nicolò Lucchesi committed
Feb 2, 2026 at 12:01 UTC
528b3076afaac5ed47da85ee28966f9fd845dd78
1 file changed
+4
-3
tests/v1/kv_connector/unit/test_multi_connector.py
+4
-3
@@ -151,7 +151,8 @@ def test_multi_example_connector_consistency():
151
kv_transfer_config=kv_transfer_config,
152
)
153
# Run generation - this should trigger saving KV cache
154
- _ = llm.generate(PROMPTS, SAMPLING_PARAMS)
154
+ # Use a single prompt to avoid race conditions depending on the order of scheduling
155
+ _ = llm.generate(PROMPTS[0], SAMPLING_PARAMS)
156
157
# --- Verification ---
158
@@ -221,7 +222,7 @@ def test_multi_example_connector_consistency():
222
223
# Run generation again - this should trigger loading from the first
224
# connector.
224
- _ = llm.generate(PROMPTS, SAMPLING_PARAMS)
225
+ _ = llm.generate(PROMPTS[1], SAMPLING_PARAMS)
226
227
events = get_connector_events()
228
# get_num_new_matched_tokens will return new tokens from the first
@@ -247,7 +248,7 @@ def test_multi_example_connector_consistency():
248
249
# Run generation again - this should trigger loading from the first
250
# connector.
250
- _ = llm.generate(PROMPTS, SAMPLING_PARAMS)
251
+ _ = llm.generate(PROMPTS[0], SAMPLING_PARAMS)
252
253
events = get_connector_events()
254
# get_num_new_matched_tokens will be called for both connectors but will