a100-qwen-api-runtime / scripts /check_threads.sh
ShoaibSSM's picture
Bound llama.cpp HTTP and compute threads; enable four slots
d861a68 verified
Raw History Blame Contribute Delete
443 Bytes
#!/usr/bin/env bash
set -u
echo '=== CGROUP TASKS ==='
printf 'current: '; cat /sys/fs/cgroup/pids.current
printf 'max: '; cat /sys/fs/cgroup/pids.max
echo
echo '=== TOP THREAD USERS ==='
ps -eo pid,ppid,comm,nlwp,%cpu,%mem --sort=-nlwp | head -20
echo
echo '=== LLAMA PROCESSES ==='
ps -eo pid,ppid,comm,nlwp,%cpu,%mem,args | grep -E '[l]lama'
echo
echo '=== THREAD ENV ==='
env | grep -E 'TOKIO|OMP|MKL|OPENBLAS|NUMEXPR|VECLIB|BLIS' || true