[V0 Deprecation] Deprecate --disable-frontend-multiprocessing (#37612)

Signed-off-by: sfeng33 <4florafeng@gmail.com>

tests/entrypoints/instrumentator/test_basic.py

@@ -28,7 +28,7 @@ def server_args(request: pytest.FixtureRequest) -> list[str]:
     >>> @pytest.mark.parametrize(
     >>>     "server_args",
     >>>     [
-    >>>         ["--disable-frontend-multiprocessing"],
+    >>>         ["--max-model-len", "10100"],
     >>>         [
     >>>             "--model=NousResearch/Hermes-3-Llama-3.1-70B",
     >>>             "--enable-auto-tool-choice",
@@ -40,7 +40,7 @@ def server_args(request: pytest.FixtureRequest) -> list[str]:
     >>>     ...
 
     This will run `test_foo` twice with servers with:
-    - `--disable-frontend-multiprocessing`
+    - `--max-model-len 10100`
     - `--model=NousResearch/Hermes-3-Llama-3.1-70B --enable-auto-tool-choice`.
 
     """
@@ -79,17 +79,6 @@ async def client(server):
         yield async_client
 
 
-@pytest.mark.parametrize(
-    "server_args",
-    [
-        pytest.param([], id="default-frontend-multiprocessing"),
-        pytest.param(
-            ["--disable-frontend-multiprocessing"],
-            id="disable-frontend-multiprocessing",
-        ),
-    ],
-    indirect=True,
-)
 @pytest.mark.asyncio
 async def test_show_version(server: RemoteOpenAIServer):
     response = requests.get(server.url_for("version"))
@@ -98,17 +87,6 @@ async def test_show_version(server: RemoteOpenAIServer):
     assert response.json() == {"version": VLLM_VERSION}
 
 
-@pytest.mark.parametrize(
-    "server_args",
-    [
-        pytest.param([], id="default-frontend-multiprocessing"),
-        pytest.param(
-            ["--disable-frontend-multiprocessing"],
-            id="disable-frontend-multiprocessing",
-        ),
-    ],
-    indirect=True,
-)
 @pytest.mark.asyncio
 async def test_check_health(server: RemoteOpenAIServer):
     response = requests.get(server.url_for("health"))
@@ -119,13 +97,7 @@ async def test_check_health(server: RemoteOpenAIServer):
 @pytest.mark.parametrize(
     "server_args",
     [
-        pytest.param(
-            ["--max-model-len", "10100"], id="default-frontend-multiprocessing"
-        ),
-        pytest.param(
-            ["--disable-frontend-multiprocessing", "--max-model-len", "10100"],
-            id="disable-frontend-multiprocessing",
-        ),
+        pytest.param(["--max-model-len", "10100"]),
     ],
     indirect=True,
 )

tests/entrypoints/instrumentator/test_metrics.py

@@ -50,7 +50,6 @@ def default_server_args():
     params=[
         "",
         "--enable-chunked-prefill",
-        "--disable-frontend-multiprocessing",
         f"--show-hidden-metrics-for-version={PREV_MINOR_VERSION}",
     ],
 )

tests/entrypoints/openai/completion/test_completion_with_prompt_embeds.py

@@ -83,11 +83,8 @@ def example_prompt_embeds(hf_runner):
     return [_encode_embeds(item) for item in example_embeddings]
 
 
-@pytest.fixture(scope="module", params=["", "--disable-frontend-multiprocessing"])
-def server_with_prompt_embeds(default_server_args, request):
-    if request.param:
-        default_server_args.append(request.param)
-
+@pytest.fixture(scope="module")
+def server_with_prompt_embeds(default_server_args):
     with RemoteOpenAIServer(MODEL_NAME, default_server_args) as remote_server:
         yield remote_server
 

tests/entrypoints/openai/completion/test_shutdown.py

@@ -150,7 +150,6 @@ async def test_shutdown_on_engine_failure():
             "0.05",
             "--max-num-seqs",
             "2",
-            "--disable-frontend-multiprocessing",
         ],
         # ROCm: Disable stdout/stderr pipe capture. Subprocess hangs when
         # stdout/stderr pipes are enabled during ROCm GPU initialization.