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, 23:45 UTC
Correctly kill vLLM processes after finishing serving benchmarks (#21641)
Signed-off-by: Huy Do <huydhn@gmail.com>
Huy Do committed
Jul 25, 2025 at 19:06 UTC
a55c95096b3537edfbbb7a5eafae0b0475c5ef07
1 file changed
+8
-6
.buildkite/nightly-benchmarks/scripts/run-nightly-benchmarks.sh
+8
-6
@@ -95,12 +95,14 @@ json2args() {
95
}
96
97
kill_gpu_processes() {
98
- pkill -f python
99
- pkill -f python3
100
- pkill -f tritonserver
101
- pkill -f pt_main_thread
102
- pkill -f text-generation
103
- pkill -f lmdeploy
98
+ pkill -f '[p]ython'
99
+ pkill -f '[p]ython3'
100
+ pkill -f '[t]ritonserver'
101
+ pkill -f '[p]t_main_thread'
102
+ pkill -f '[t]ext-generation'
103
+ pkill -f '[l]mdeploy'
104
+ # vLLM now names the process with VLLM prefix after https://github.com/vllm-project/vllm/pull/21445
105
+ pkill -f '[V]LLM'
106
107
while [ "$(nvidia-smi --query-gpu=memory.used --format=csv,noheader,nounits | head -n 1)" -ge 1000 ]; do
108
sleep 1