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 14, 05:30 UTC
Code quality improvements: version update, type annotation enhancement, and enum usage simplification (#27581)
Signed-off-by: Bradley <bradley.b.pitt@gmail.com>
usberkeley committed
Oct 28, 2025 at 01:50 UTC
69f064062ba78a0ac44962f55a46a9d79cfb9ce0
3 files changed
+7
-9
docs/deployment/docker.md
+2
-2
@@ -41,11 +41,11 @@ You can add any other [engine-args](../configuration/engine_args.md) you need af
41
create a custom Dockerfile on top of the base image with an extra layer that installs them:
42
43
```Dockerfile
44
- FROM vllm/vllm-openai:v0.9.0
44
+ FROM vllm/vllm-openai:v0.11.0
45
46
# e.g. install the `audio` optional dependencies
47
# NOTE: Make sure the version of vLLM matches the base image!
48
- RUN uv pip install --system vllm[audio]==0.9.0
48
+ RUN uv pip install --system vllm[audio]==0.11.0
49
```
50
51
!!! tip
vllm/multimodal/profiling.py
+1
-1
@@ -368,7 +368,7 @@ class MultiModalProfiler(Generic[_I]):
368
self,
369
seq_len: int,
370
mm_counts: Mapping[str, int] | None = None,
371
- ):
371
+ ) -> Mapping[str, int]:
372
"""
373
Returns the maximum length of the multimodal (image placeholders+text)
374
tokens, including any break/text tokens in-between image embeddings.
vllm/v1/core/sched/scheduler.py
+4
-6
@@ -113,14 +113,12 @@ class Scheduler(SchedulerInterface):
113
# req_id -> Request
114
self.requests: dict[str, Request] = {}
115
# Scheduling policy
116
- if self.scheduler_config.policy == "priority":
117
- self.policy = SchedulingPolicy.PRIORITY
118
- elif self.scheduler_config.policy == "fcfs":
119
- self.policy = SchedulingPolicy.FCFS
120
- else:
116
+ try:
117
+ self.policy = SchedulingPolicy(self.scheduler_config.policy)
118
+ except ValueError as e:
119
raise ValueError(
120
f"Unknown scheduling policy: {self.scheduler_config.policy}"
123
- )
121
+ ) from e
122
# Priority queues for requests.
123
self.waiting = create_request_queue(self.policy)
124
self.running: list[Request] = []