From fca252d59efdb627549e5677263c69709590c982 Mon Sep 17 00:00:00 2001 From: Flora Feng <4florafeng@gmail.com> Date: Tue, 21 Jul 2026 15:00:06 -0400 Subject: [PATCH] [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> --- tests/plugins_tests/test_oot_registration_offline.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/plugins_tests/test_oot_registration_offline.py b/tests/plugins_tests/test_oot_registration_offline.py index 3f483a5706..ac5361c66f 100644 --- a/tests/plugins_tests/test_oot_registration_offline.py +++ b/tests/plugins_tests/test_oot_registration_offline.py @@ -50,7 +50,7 @@ def test_oot_registration_embedding( m.setenv("VLLM_PLUGINS", "register_dummy_model") prompts = ["Hello, my name is", "The text does not matter"] llm = LLM( - model=dummy_gemma2_embedding_path, load_format="dummy", max_model_len=2048 + model=dummy_gemma2_embedding_path, load_format="dummy", max_model_len=512 ) outputs = llm.embed(prompts)