youngkingdom/vllm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[Misc] auto_tune: kill specific vllm process (#26304)

Browse Source 
Signed-off-by: Karan Goel <karangoel@google.com>
This commit is contained in:
Karan Goel
2025-10-06 11:02:51 -07:00
committed by GitHub
parent 05f6846ede
commit 824a3f403f
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
benchmarks/auto_tune/auto_tune.sh
View File

@ -74,7 +74,7 @@ start_server() {
local vllm_log=$4
local profile_dir=$5
pkill -if vllm
pkill -if "vllm serve" || true
# Define the common arguments as a bash array.
# Each argument and its value are separate elements.
@ -139,7 +139,7 @@ run_benchmark() {
echo "vllm_log: $vllm_log"
echo
rm -f $vllm_log
pkill -if vllm
pkill -if "vllm serve" || true
echo "starting server..."
# Call start_server without a profile_dir to avoid profiling overhead
@ -232,7 +232,7 @@ run_benchmark() {
echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput"
pkill -if vllm
pkill -if "vllm serve" || true
sleep 10
echo "===================="
return 0
@ -308,6 +308,6 @@ if (( $(echo "$best_throughput > 0" | bc -l) )); then
else
echo "No configuration met the latency requirements. Skipping final profiling run."
fi
pkill -if vllm
pkill -if "vllm serve" || true
echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH"
echo "best_max_num_seqs: $best_max_num_seqs, best_num_batched_tokens: $best_num_batched_tokens, best_throughput: $best_throughput, profile saved in: $PROFILE_PATH" >> "$RESULT"