diff --git a/mods/drop-caches/run.sh b/mods/drop-caches/run.sh new file mode 100644 index 0000000..7defc59 --- /dev/null +++ b/mods/drop-caches/run.sh @@ -0,0 +1,17 @@ +#!/bin/bash + +# This mod will drop the FS caches every minute - useful to unstuck Qwen3.5-397B or other similar models during loading + +CMD='sync; echo 3 > /proc/sys/vm/drop_caches' +LOG="/tmp/drop_caches.log" +PIDFILE="/tmp/drop_caches.pid" + +nohup bash -c ' + while true; do + '"$CMD"' >> "'"$LOG"'" 2>&1 + sleep 60 + done +' >/dev/null 2>&1 & + +echo $! > "$PIDFILE" +echo "Started drop_caches loop with PID $(cat "$PIDFILE"); log is available in $LOG" \ No newline at end of file diff --git a/recipes/qwen3.5-397b-int4-autoround.yaml b/recipes/qwen3.5-397b-int4-autoround.yaml index b6c77f3..750accd 100644 --- a/recipes/qwen3.5-397b-int4-autoround.yaml +++ b/recipes/qwen3.5-397b-int4-autoround.yaml @@ -23,6 +23,7 @@ mods: - mods/fix-qwen3.5-autoround - mods/fix-qwen3.5-chat-template - mods/gpu-mem-util-gb + - mods/drop-caches # Default settings (can be overridden via CLI) defaults: