{
  "id": 305355,
  "title": "Why is my code running multiple times locally with very different results？",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/305355",
  "author_name": "",
  "post_date": "2022-02-05T00:21:01.229937500Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>The object detection model I used is YOLOv5. I ran the same model twice. After submitting, the LB was 0.46 and 0.51 respectively. Why are the results so different? All hyperparameters of my two models are the same</p>",
  "messages": [
    {
      "id": "1676398",
      "postDate": "02/05/2022 00:21:01",
      "content": "<p>The object detection model I used is YOLOv5. I ran the same model twice. After submitting, the LB was 0.46 and 0.51 respectively. Why are the results so different? All hyperparameters of my two models are the same</p>",
      "rawMarkdown": "The object detection model I used is YOLOv5. I ran the same model twice. After submitting, the LB was 0.46 and 0.51 respectively. Why are the results so different? All hyperparameters of my two models are the same",
      "votes": null
    },
    {
      "id": "1676721",
      "postDate": "02/05/2022 08:56:41",
      "content": "<p>I believe YOLOV5 uses mosaic and mixup data augmentation which will cause differences in the training data. This is probably the cause of the f2 discrepancy</p>",
      "rawMarkdown": "I believe YOLOV5 uses mosaic and mixup data augmentation which will cause differences in the training data. This is probably the cause of the f2 discrepancy",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1676721,
      "author_name": "maxvandijck",
      "author_url": "",
      "post_date": "02/05/2022 08:56:41",
      "content": "<p>I believe YOLOV5 uses mosaic and mixup data augmentation which will cause differences in the training data. This is probably the cause of the f2 discrepancy</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1676398": "The object detection model I used is YOLOv5. I ran the same model twice. After submitting, the LB was 0.46 and 0.51 respectively. Why are the results so different? All hyperparameters of my two models are the same",
    "1676721": "I believe YOLOV5 uses mosaic and mixup data augmentation which will cause differences in the training data. This is probably the cause of the f2 discrepancy"
  },
  "source": "meta"
}