diff --git a/object_detection/detectron2_training-kfold.ipynb b/object_detection/detectron2_training-kfold.ipynb
index 66c1567..5e4e8aa 100644
--- a/object_detection/detectron2_training-kfold.ipynb
+++ b/object_detection/detectron2_training-kfold.ipynb
@@ -215,6 +215,8 @@
     "  \n",
     "    def data_loader_mapper(self, batch):\n",
     "        return batch\n",
+    "    \n",
+    "\n",
     "\n",
     "    def run_hooks(self):\n",
     "        val_loss = self.validation()\n",
@@ -229,8 +231,9 @@
     "                self._trainer.has_finished = True\n",
     "\n",
     "    def validation(self):\n",
-    "        val_loader = build_detection_test_loader(self.cfg, self.cfg.DATASETS.TEST[0], evaluators=[evaluator])\n",
+    "        # Define evaluator here\n",
     "        evaluator = COCOEvaluator(self.cfg.DATASETS.TEST[0], self.cfg, True, output_dir=\"./output/\")\n",
+    "        val_loader = build_detection_test_loader(self.cfg, self.cfg.DATASETS.TEST[0], evaluators=[evaluator])\n",
     "        val_results = self._trainer.test(self.cfg, self.model, evaluators=[evaluator])[0]\n",
     "        val_loss = val_results[\"total_loss\"]\n",
     "        return val_loss\n",
@@ -256,9 +259,13 @@
     "    cfg.SOLVER.STEPS = []        # milestones where LR is reduced, in this case there's no decay\n",
     "    cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 128   # The \"RoIHead batch size\". \n",
     "    cfg.MODEL.ROI_HEADS.NUM_CLASSES = 80  \n",
-    "    cfg.TEST.EVAL_PERIOD = 500\n",
+    "    cfg.TEST.EVAL_PERIOD = 15000\n",
     "    os.makedirs(cfg.OUTPUT_DIR, exist_ok=True)\n",
     "    trainer = Early_stopping(cfg, early_stop_patience=5, model_checkpoint_path=\"model_checkpoint.pth\")\n",
+    "    # Specify evaluators during testing\n",
+    "    evaluator = COCOEvaluator(cfg.DATASETS.TEST[0], cfg, True, output_dir=\"./output/\")\n",
+    "    trainer.resume_or_load(resume=False)\n",
+    "    trainer.test(cfg, trainer.model, evaluators=[evaluator])\n",
     "    trainer.resume_or_load(resume=False)\n",
     "    trainer.train();\n",
     "    return cfg\n"