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, 17:15 UTC
[V1][Minor] Use `SchedulerInterface` type for engine scheduler field (#15499)
Signed-off-by: Nick Hill <nhill@redhat.com>
Nick Hill committed
Mar 25, 2025 at 14:21 UTC
6aa196c8dc977abebbb1a721279944d21f9ca174
1 file changed
+2
-1
vllm/v1/engine/core.py
+2
-1
@@ -23,6 +23,7 @@ from vllm.utils import (get_exception_traceback, resolve_obj_by_qualname,
23
zmq_socket_ctx)
24
from vllm.v1.core.kv_cache_utils import (get_kv_cache_config,
25
unify_kv_cache_configs)
26
+from vllm.v1.core.sched.interface import SchedulerInterface
27
from vllm.v1.core.sched.output import SchedulerOutput
28
from vllm.v1.core.sched.scheduler import Scheduler as V1Scheduler
29
from vllm.v1.engine import (EngineCoreOutputs, EngineCoreRequest,
@@ -84,7 +85,7 @@ class EngineCore:
85
"compatibility may not be maintained.",
86
vllm_config.scheduler_config.scheduler_cls)
87
87
- self.scheduler = Scheduler(
88
+ self.scheduler: SchedulerInterface = Scheduler(
89
scheduler_config=vllm_config.scheduler_config,
90
model_config=vllm_config.model_config,
91
cache_config=vllm_config.cache_config,