[CI][Bugfix] Reduce max_model_len in OOT embedding test to fix KV-cache OOM on small GPUs (#49351)
Signed-off-by: sfeng33 <4florafeng@gmail.com>
Flora Feng committed
Jul 21, 2026 at 15:00 UTC
fca252d59efdb627549e5677263c69709590c982
1 file changed
+1
-1
tests/plugins_tests/test_oot_registration_offline.py
+1
-1
@@ -50,7 +50,7 @@ def test_oot_registration_embedding(
50
m.setenv("VLLM_PLUGINS", "register_dummy_model")
51
prompts = ["Hello, my name is", "The text does not matter"]
52
llm = LLM(
53
- model=dummy_gemma2_embedding_path, load_format="dummy", max_model_len=2048
53
+ model=dummy_gemma2_embedding_path, load_format="dummy", max_model_len=512
54
)
55
outputs = llm.embed(prompts)
56