Surprising with result of yolov5s model

YOLOv5 🚀
1

I trying to validate yolov5s and my pruned model with coco128 dataset. I got strange result of mAP@.5.

  1. yolov5s.onnx - 0.711
  2. yolov5s_pruned - 0.92

I am using the above query.
python val.py --weights yolo_models/yolov5s.onnx --data coco128.yaml --img 640 --batch-size 1

Basically the values of the mAP@.5 of the pruned one is going up. But it should go down logically.
Could you please tell why this is happening? Did I do something wrong?