A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 04:45 UTC
[Fix][V1] Remove --scheduling-policy oracle (#20010)
Signed-off-by: amit <amit.man@gmail.com>
amit committed
Jun 24, 2025 at 19:52 UTC
981eeca41aed8202074f36baa1e46eb9c2c82b16
2 files changed
-11
tests/v1/test_oracle.py
-6
@@ -74,12 +74,6 @@ def test_unsupported_configs(monkeypatch):
74
disable_async_output_proc=True,
75
).create_engine_config()
76
77
- with pytest.raises(NotImplementedError):
78
- AsyncEngineArgs(
79
- model=MODEL,
80
- scheduling_policy="priority",
81
- ).create_engine_config()
82
-
77
with pytest.raises(NotImplementedError):
78
AsyncEngineArgs(
79
model=MODEL,
vllm/engine/arg_utils.py
-5
@@ -1289,11 +1289,6 @@ class EngineArgs:
1289
recommend_to_remove=True)
1290
return False
1291
1292
- if self.scheduling_policy != SchedulerConfig.policy:
1293
- _raise_or_fallback(feature_name="--scheduling-policy",
1294
- recommend_to_remove=False)
1295
- return False
1296
-
1292
if self.num_scheduler_steps != SchedulerConfig.num_scheduler_steps:
1293
_raise_or_fallback(feature_name="--num-scheduler-steps",
1294
recommend_to_remove=True)