[V0 Deprecation] Remove async_output_proc, preemption mode, delay factor (#25334)

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>

tests/detokenizer/test_stop_strings.py

@@ -32,10 +32,6 @@ def _test_stopping(llm: LLM,
     assert output.stop_reason == expected_reason
 
 
-def _set_async_mode(llm, is_async):
-    llm.llm_engine.scheduler[0].use_async_output_proc = is_async
-
-
 def _stop_basic(llm):
     _test_stopping(llm,
                    stop=["."],
@@ -103,40 +99,8 @@ def test_stop_strings():
     # async output processing below.
     llm = LLM(MODEL, enforce_eager=envs.VLLM_USE_V1)
 
-    if envs.VLLM_USE_V1:
-        _stop_basic(llm)
-    else:
-        _set_async_mode(llm, True)
-        _stop_basic(llm)
-
-        _set_async_mode(llm, False)
-        _stop_basic(llm)
-
-    if envs.VLLM_USE_V1:
-        _stop_multi_tokens(llm)
-    else:
-        _set_async_mode(llm, True)
-        _stop_multi_tokens(llm)
-
-        _set_async_mode(llm, False)
-        _stop_multi_tokens(llm)
-
-    if envs.VLLM_USE_V1:
-        _stop_partial_token(llm)
-    else:
-        _set_async_mode(llm, True)
-        _stop_partial_token(llm)
-
-        _set_async_mode(llm, False)
-        _stop_partial_token(llm)
-
-    if envs.VLLM_USE_V1:
-        # FIXME: this does not respect include_in_output=False
-        # _stop_token_id(llm)
-        pass
-    else:
-        _set_async_mode(llm, True)
-        _stop_token_id(llm)
-
-        _set_async_mode(llm, False)
-        _stop_token_id(llm)
+    _stop_basic(llm)
+    _stop_multi_tokens(llm)
+    _stop_partial_token(llm)
+    # FIXME: this does not respect include_in_output=False
+    # _stop_token_id(llm)

tests/v1/engine/test_processor_multi_modal_uuids.py

@@ -6,7 +6,6 @@ import pytest
 from vllm.assets.image import ImageAsset
 from vllm.assets.video import VideoAsset
 from vllm.config import CacheConfig, DeviceConfig, ModelConfig, VllmConfig
-from vllm.platforms.interface import UnspecifiedPlatform
 from vllm.sampling_params import SamplingParams
 from vllm.v1.engine import processor as processor_mod
 from vllm.v1.engine.processor import Processor
@@ -33,15 +32,6 @@ def _mk_processor(monkeypatch,
                         "__post_init__",
                         lambda self, *args: None,
                         raising=True)
-    monkeypatch.setattr(UnspecifiedPlatform,
-                        "is_async_output_supported",
-                        classmethod(lambda cls, enforce_eager: True),
-                        raising=True)
-    monkeypatch.setattr(
-        ModelConfig,
-        "verify_async_output_proc",
-        lambda self, parallel_config, speculative_config, device_config: None,
-        raising=True)
     monkeypatch.setattr(ModelConfig,
                         "verify_with_parallel_config",
                         lambda self, parallel_config: None,

tests/v1/test_oracle.py

@@ -29,24 +29,6 @@ def test_unsupported_configs(monkeypatch):
                 },
             ).create_engine_config()
 
-        with pytest.raises(NotImplementedError):
-            AsyncEngineArgs(
-                model=MODEL,
-                preemption_mode="swap",
-            ).create_engine_config()
-
-        with pytest.raises(NotImplementedError):
-            AsyncEngineArgs(
-                model=MODEL,
-                disable_async_output_proc=True,
-            ).create_engine_config()
-
-        with pytest.raises(NotImplementedError):
-            AsyncEngineArgs(
-                model=MODEL,
-                scheduler_delay_factor=1.2,
-            ).create_engine_config()
-
 
 def test_enable_by_default_fallback(monkeypatch):
     with monkeypatch.context() as m: