[Multi-host] Fix bugs in the deployment script

Signed-off-by: Lihao Ran <[email protected]>

Author: Lumosis

scripts/multihost/deploy_cluster.sh

@@ -75,13 +75,15 @@ if [ ! -f "${SCRIPT_PATH}" ]; then
     exit 1
 fi
 
+
 # Convert comma-separated worker string to a bash array for SSH access
 IFS=',' read -r -a WORKER_SSH_IPS <<< "$WORKER_SSH_IP_STRING"
 # Create a list of all SSH IPs for cleanup and script distribution
 ALL_SSH_IPS=("${HEAD_SSH_IP}" "${WORKER_SSH_IPS[@]}")
 
 if [ -z "${SSH_USER}" ]; then
-    read -pr "Enter the SSH username for all hosts: " SSH_USER
+    printf "Enter the SSH username for all hosts: "
+    read -r SSH_USER
 fi
 
 # --- STEP 1: Pre-Deployment Cleanup ---
@@ -96,7 +98,8 @@ fi
 
 # --- STEP 2: Distribute the script ---
 echo "📦 Distributing '${SCRIPT_PATH}' to all hosts..."
-REMOTE_SCRIPT_DIR="$HOME/tpu-inference/scripts/multihost"
+REMOTE_HOME_DIR="~"
+REMOTE_SCRIPT_DIR="${REMOTE_HOME_DIR}/tpu-inference/scripts/multihost"
 REMOTE_SCRIPT_PATH="${REMOTE_SCRIPT_DIR}/run_cluster.sh"
 SSH_OPTIONS="-o StrictHostKeyChecking=no -o BatchMode=yes"