{
  "id": 315414,
  "title": "Anyone monitor Top1/5 Acc when training?",
  "url": "/competitions/happy-whale-and-dolphin/discussion/315414",
  "author_name": "",
  "post_date": "2022-03-28T06:44:15.309153500Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I monitor Top 1 and Top 5 accuracy in training. If anyone also doing this, would you mind sharing your Top 1/5 ACC and LB score? My Top 5 cv is 0.64 and my LB is 0.6. I wonder if this is normal or if I made some mistake when inference. Thanks.</p>",
  "messages": [
    {
      "id": "1737157",
      "postDate": "03/28/2022 06:44:15",
      "content": "<p>I monitor Top 1 and Top 5 accuracy in training. If anyone also doing this, would you mind sharing your Top 1/5 ACC and LB score? My Top 5 cv is 0.64 and my LB is 0.6. I wonder if this is normal or if I made some mistake when inference. Thanks.</p>",
      "rawMarkdown": "I monitor Top 1 and Top 5 accuracy in training. If anyone also doing this, would you mind sharing your Top 1/5 ACC and LB score? My Top 5 cv is 0.64 and my LB is 0.6. I wonder if this is normal or if I made some mistake when inference. Thanks.",
      "votes": null
    },
    {
      "id": "1737532",
      "postDate": "03/28/2022 14:04:12",
      "content": "<p>My recent model achieved LB 0.783, top 1 validation acc. 79.321, top 5 validation acc. 89.290. Used single model, holdout for validation. In my experiments, top 1 validation accuracies are similar with LB scores, +- 0.01~0.02.</p>",
      "rawMarkdown": "My recent model achieved LB 0.783, top 1 validation acc. 79.321, top 5 validation acc. 89.290. Used single model, holdout for validation. In my experiments, top 1 validation accuracies are similar with LB scores, +- 0.01~0.02.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1737532,
      "author_name": "hwigeon",
      "author_url": "",
      "post_date": "03/28/2022 14:04:12",
      "content": "<p>My recent model achieved LB 0.783, top 1 validation acc. 79.321, top 5 validation acc. 89.290. Used single model, holdout for validation. In my experiments, top 1 validation accuracies are similar with LB scores, +- 0.01~0.02.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1737157": "I monitor Top 1 and Top 5 accuracy in training. If anyone also doing this, would you mind sharing your Top 1/5 ACC and LB score? My Top 5 cv is 0.64 and my LB is 0.6. I wonder if this is normal or if I made some mistake when inference. Thanks.",
    "1737532": "My recent model achieved LB 0.783, top 1 validation acc. 79.321, top 5 validation acc. 89.290. Used single model, holdout for validation. In my experiments, top 1 validation accuracies are similar with LB scores, +- 0.01~0.02."
  },
  "source": "meta"
}