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, 01:00 UTC
[Misc] Do not print async output warning for v1 (#21151)
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Woosuk Kwon committed
Jul 17, 2025 at 21:57 UTC
54cf1cae6222ece444673fd083ee7b930d4e1d52
2 files changed
+2
-2
vllm/platforms/cuda.py
+1
-1
@@ -99,7 +99,7 @@ class CudaPlatformBase(Platform):
99
100
@classmethod
101
def is_async_output_supported(cls, enforce_eager: Optional[bool]) -> bool:
102
- if enforce_eager:
102
+ if enforce_eager and not envs.VLLM_USE_V1:
103
logger.warning(
104
"To see benefits of async output processing, enable CUDA "
105
"graph. Since, enforce-eager is enabled, async output "
vllm/platforms/rocm.py
+1
-1
@@ -299,7 +299,7 @@ class RocmPlatform(Platform):
299
300
@classmethod
301
def is_async_output_supported(cls, enforce_eager: Optional[bool]) -> bool:
302
- if enforce_eager:
302
+ if enforce_eager and not envs.VLLM_USE_V1:
303
logger.warning(
304
"To see benefits of async output processing, enable CUDA "
305
"graph. Since, enforce-eager is enabled, async output "