fix: run test_single_chat_session_image_embeds on L40S

Signed-off-by: ixlmar <206748156+ixlmar@users.noreply.github.com>

cpp/kernels/fmha_v2/pytest.ini

@@ -6,6 +6,7 @@ markers =
     fmhca
     debug
     bench
+    needs_l40s
 # bin: unit tests
 # test: python script for invoking fmha.exe
 testpaths = bin test

tests/integration/defs/test_e2e.py

@@ -1675,9 +1675,13 @@ def test_openai_lora(llm_root, llm_venv):
 
 def test_openai_chat_multimodal_example(llm_root, llm_venv):
     test_root = unittest_path() / "llmapi" / "apps"
-    llm_venv.run_cmd(
-        ["-m", "pytest",
-         str(test_root / "_test_openai_chat_multimodal.py")])
+    llm_venv.run_cmd([
+        "-m",
+        "pytest",
+        str(test_root / "_test_openai_chat_multimodal.py"),
+        "-m",
+        "not needs_l40s",
+    ])
 
 
 def test_openai_mmencoder_example(llm_root, llm_venv):

tests/integration/test_lists/test-db/l0_l40s.yml

@@ -28,6 +28,7 @@ l0_l40s:
   - test_e2e.py::test_ptp_quickstart_multimodal_phi4mm[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-audio]
   - test_e2e.py::test_ptp_quickstart_multimodal_phi4mm[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-image]
   - test_e2e.py::test_ptp_quickstart_multimodal_phi4mm[phi4-multimodal-instruct-multimodals/Phi-4-multimodal-instruct-image_audio]
+  - unittest/llmapi/apps/_test_openai_chat_multimodal.py::test_single_chat_session_image_embeds -m needs_l40s
   # MMMU sanity check
   - accuracy/test_llm_api_pytorch_multimodal.py::TestQwen2_5_VL_7B::test_auto_dtype
   - accuracy/test_llm_api_pytorch_multimodal.py::TestVILA1_5_3B::test_auto_dtype

tests/unittest/llmapi/apps/_test_openai_chat_multimodal.py

@@ -168,6 +168,7 @@ def server_patched(model_name: str, temp_extra_llm_api_options_file: str):
         yield remote_server
 
 
+@pytest.mark.needs_l40s
 @pytest.mark.asyncio(loop_scope="module")
 def test_single_chat_session_image_embeds(
     server_patched: RemoteOpenAIServer,