fix: address PR comment + fix obtaining ray cluster

Author: pawelpaszki

demo-notebooks/guided-demos/6_rayjob_checkpointing_example.ipynb

@@ -211,21 +211,20 @@
       "outputs": [],
       "source": [
         "print(job.status())\n",
-        "\n",
-        "# Resolve RayCluster created by the RayJob (retry until it exists after Kueue admission).\n",
+        "# KubeRay assigns a generated name — not job.cluster_name (the template).\n",
         "cluster = None\n",
         "for _ in range(36):\n",
-        "    try:\n",
-        "        cluster = get_cluster(job.cluster_name, namespace=NAMESPACE, verify_tls=False)\n",
-        "        break\n",
-        "    except Exception:\n",
-        "        time.sleep(5)\n",
-        "\n",
+        "    status_data = job._api.get_job_status(name=job.name, k8s_namespace=NAMESPACE)\n",
+        "    ray_cluster_name = (status_data or {}).get(\"rayClusterName\")\n",
+        "    if ray_cluster_name:\n",
+        "        cluster = get_cluster(ray_cluster_name, namespace=NAMESPACE, verify_tls=False)\n",
+        "        if cluster is not None:\n",
+        "            break\n",
+        "    time.sleep(5)\n",
         "if cluster is None:\n",
         "    raise RuntimeError(\n",
         "        \"RayCluster not ready — check RayJob / Workload admission and operator logs.\"\n",
         "    )\n",
-        "\n",
         "print(f\"Ray Dashboard (open in browser): {cluster.cluster_dashboard_uri()}\")\n",
         "print(\"In the dashboard, open Jobs and stream logs for the training driver.\")\n",
         "print(\n",

demo-notebooks/guided-demos/train_with_checkpoints.py

@@ -137,7 +137,7 @@ def train_func(config):
             "lr": 0.001,
         },
         scaling_config=ScalingConfig(
-            num_workers=1,
+            num_workers=2,
             use_gpu=False,
         ),
         run_config=RunConfig(