chore(rl-train): 🔧 Update RL training script with new environment variables and execution flow for dynamic environment selection and training modes

Co-Authored-By: Lilith Autocommit <noreply@atlilith.com>

scripts/rl-train.sh

@@ -86,28 +86,56 @@ case "$cmd" in
     ;;
 
   launch)
+    # Launch as a transient systemd --user .service under heavy-tests.slice.
+    # The slice (CPUWeight=20, MemoryMax=32G, TasksMax=4096) prevents the godot
+    # workers spawned by the python parent from starving sshd/interactive work.
+    # Every child process (flatpak, bwrap, godot-bin) inherits the cgroup, so a
+    # 3000-proc explosion stays contained — exactly the wedge mode seen on
+    # 2026-05-18 and 2026-05-19.
+    #
+    # Unit name includes epoch so re-launches with the same RL_RUN_NAME don't
+    # collide with a stopped-but-not-yet-collected unit.
+    RL_UNIT="rl-train-${RL_RUN_NAME}-$(date +%s)"
     remote "
-      cd ${RL_WORKTREE} || exit 1
-      existing=\$(ps -eo pid,comm,args | awk '\$2 ~ /^python/ && /rl_self_play.train/ {print \$1}')
+      set -e
+      cd ${RL_WORKTREE}
+      existing=\$(systemctl --user list-units --type=service --no-legend --state=running 'rl-train-*' 2>/dev/null | awk '{print \$1}' | head -1)
       if [ -n \"\$existing\" ]; then
-        echo 'training already running; run kill first'
-        echo \"\$existing\"
+        echo \"training already running: \$existing — run 'kill' first\"
         exit 1
       fi
-      nohup python3 -m tooling.rl_self_play.train \
-        --device ${RL_DEVICE} --num-envs ${RL_ENVS} \
-        --total-steps ${RL_TOTAL_STEPS} --eval-freq ${RL_EVAL_FREQ} \
-        --eval-episodes ${RL_EVAL_EPS} --max-turns ${RL_MAX_TURNS} \
-        --run-name ${RL_RUN_NAME} > ${LOG_REMOTE} 2>&1 &
-      echo \$! > ${RL_PIDFILE}
+      systemd-run --user \\
+        --slice=heavy-tests.slice \\
+        --unit=${RL_UNIT} \\
+        --collect --quiet \\
+        --working-directory=${RL_WORKTREE} \\
+        --setenv=PYTHONUNBUFFERED=1 \\
+        --property=StandardOutput=append:${LOG_REMOTE} \\
+        --property=StandardError=append:${LOG_REMOTE} \\
+        -- python3 -m tooling.rl_self_play.train \\
+            --device ${RL_DEVICE} --num-envs ${RL_ENVS} \\
+            --total-steps ${RL_TOTAL_STEPS} --eval-freq ${RL_EVAL_FREQ} \\
+            --eval-episodes ${RL_EVAL_EPS} --max-turns ${RL_MAX_TURNS} \\
+            --run-name ${RL_RUN_NAME}
+      echo ${RL_UNIT} > ${RL_PIDFILE}
       sleep 3
-      ps -eo pid,comm,args | awk '\$2 ~ /^python/ && /rl_self_play.train/' \\
-        || (echo 'launch failed; check log'; tail -20 ${LOG_REMOTE})
+      systemctl --user status ${RL_UNIT} --no-pager --lines=0
+      systemctl --user show ${RL_UNIT} --property=MainPID
     "
     ;;
 
   kill)
+    # Stop all rl-train-* transient services. systemd cascades SIGTERM through
+    # the cgroup, then SIGKILL after TimeoutStopSec, reaping all godot children.
+    # Falls back to pkill for any procs not in a unit (legacy runs / orphans).
     remote "
+      units=\$(systemctl --user list-units --type=service --no-legend 'rl-train-*' 2>/dev/null | awk '{print \$1}')
+      if [ -n \"\$units\" ]; then
+        echo \"stopping units:\"
+        echo \"\$units\"
+        echo \"\$units\" | xargs -r systemctl --user stop
+      fi
+      # Legacy / out-of-unit sweep
       pkill -f 'rl_self_play.train' 2>/dev/null || true
       sleep 5
       pkill -9 -f 'rl_self_play.train' 2>/dev/null || true