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
[Chore] dedup logs (#19955)
Aaron Pham committed
Jun 22, 2025 at 15:43 UTC
e91386cde110e690d754684a2b03707925605f90
2 files changed
+1
-5
vllm/config.py
+1
-1
@@ -1970,7 +1970,7 @@ class ParallelConfig:
1970
1971
if not current_platform.use_custom_allreduce():
1972
self.disable_custom_all_reduce = True
1973
- logger.info(
1973
+ logger.debug(
1974
"Disabled the custom all-reduce kernel because it is not "
1975
"supported on current platform.")
1976
if self.ray_workers_use_nsight and not self.use_ray:
vllm/triton_utils/importing.py
-4
@@ -74,10 +74,6 @@ class TritonPlaceholder(types.ModuleType):
74
self.heuristics = self._dummy_decorator("heuristics")
75
self.Config = self._dummy_decorator("Config")
76
self.language = TritonLanguagePlaceholder()
77
- logger.warning_once(
78
- "Triton is not installed. Using dummy decorators. "
79
- "Install it via `pip install triton` to enable kernel"
80
- " compilation.")
77
78
def _dummy_decorator(self, name):
79