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 6, 09:30 UTC
[CPU][UX] Do not crash when tcmalloc/libiomp are not ldpreloaded (#37561)
Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>
Fadi Arafeh committed
Mar 19, 2026 at 16:35 UTC
2890aecce5d1fe1dcdb61be4bedbe2d46700e51c
1 file changed
+6
-4
vllm/v1/worker/cpu_worker.py
+6
-4
@@ -57,11 +57,13 @@ class CPUWorker(Worker):
57
def check_preloaded_libs(name: str):
58
ld_preload_list = os.environ.get("LD_PRELOAD", "")
59
if name not in ld_preload_list:
60
- raise RuntimeError(
61
- f"{name} is not found in LD_PRELOAD. "
62
- "Please follow the section `set LD_PRELOAD` in "
60
+ logger.warning(
61
+ "%s is not found in LD_PRELOAD. "
62
+ "For best performance, please follow the section "
63
+ "`set LD_PRELOAD` in "
64
"https://docs.vllm.ai/en/latest/getting_started/installation/cpu/ "
64
- "to setup required pre-loaded libraries."
65
+ "to setup required pre-loaded libraries.",
66
+ name,
67
)
68
69
if sys.platform.startswith("linux"):