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][Core] Remove unused speculative config from scheduler (#15818)
Signed-off-by: Mark McLoughlin <markmc@redhat.com>
Mark McLoughlin committed
Mar 31, 2025 at 20:15 UTC
f98a4920f968b1514c185bcdb881125903059c70
3 files changed
+1
-6
tests/v1/core/test_scheduler.py
-1
@@ -71,7 +71,6 @@ def create_scheduler(
71
scheduler_config,
72
model_config,
73
cache_config,
74
- speculative_config=None,
74
lora_config=None,
75
log_stats=True,
76
structured_output_manager=StructuredOutputManager(vllm_config),
vllm/v1/core/sched/scheduler.py
+1
-4
@@ -7,8 +7,7 @@ from collections import deque
7
from collections.abc import Iterable
8
from typing import Optional, Union
9
10
-from vllm.config import (CacheConfig, LoRAConfig, ModelConfig, SchedulerConfig,
11
- SpeculativeConfig)
10
+from vllm.config import CacheConfig, LoRAConfig, ModelConfig, SchedulerConfig
11
from vllm.logger import init_logger
12
from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalRegistry
13
from vllm.v1.core.encoder_cache_manager import (EncoderCacheManager,
@@ -36,7 +35,6 @@ class Scheduler(SchedulerInterface):
35
model_config: ModelConfig,
36
cache_config: CacheConfig,
37
lora_config: Optional[LoRAConfig],
39
- speculative_config: Optional[SpeculativeConfig],
38
structured_output_manager: StructuredOutputManager,
39
mm_registry: MultiModalRegistry = MULTIMODAL_REGISTRY,
40
include_finished_set: bool = False,
@@ -45,7 +43,6 @@ class Scheduler(SchedulerInterface):
43
self.scheduler_config = scheduler_config
44
self.cache_config = cache_config
45
self.lora_config = lora_config
48
- self.speculative_config = speculative_config
46
self.log_stats = log_stats
47
self.structured_output_manager = structured_output_manager
48
vllm/v1/engine/core.py
-1
@@ -95,7 +95,6 @@ class EngineCore:
95
model_config=vllm_config.model_config,
96
cache_config=vllm_config.cache_config,
97
lora_config=vllm_config.lora_config,
98
- speculative_config=vllm_config.speculative_config,
98
include_finished_set=vllm_config.parallel_config.data_parallel_size
99
> 1,
100
log_stats=self.log_stats,