{
  "id": 671195,
  "title": "If using nnunet for training, how do people usually set the number of epochs? Is the default 1000? Do you train with the full dataset for 1000 epochs?",
  "url": "/competitions/vesuvius-challenge-surface-detection/discussion/671195",
  "author_name": "",
  "post_date": "2026-01-31T14:03:00.688856500Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>If using nnunet for training, how do people usually set the number of epochs? Is the default 1000? Do you train with the full dataset for 1000 epochs?</p>",
  "messages": [
    {
      "id": "3399911",
      "postDate": "01/31/2026 14:03:00",
      "content": "<p>If using nnunet for training, how do people usually set the number of epochs? Is the default 1000? Do you train with the full dataset for 1000 epochs?</p>",
      "rawMarkdown": "If using nnunet for training, how do people usually set the number of epochs? Is the default 1000? Do you train with the full dataset for 1000 epochs?",
      "votes": null
    },
    {
      "id": "3399924",
      "postDate": "01/31/2026 14:28:18",
      "content": "<p>use default setting can give you a LB score &gt; 0.55x</p>",
      "rawMarkdown": "use default setting can give you a LB score > 0.55x",
      "votes": null
    },
    {
      "id": "3399930",
      "postDate": "01/31/2026 14:44:42",
      "content": "<p>The default settings already provide an excellent baseline. They currently rank for a bronze medal on the leaderboard. You may refer to the organizer's code for reference.</p>",
      "rawMarkdown": "The default settings already provide an excellent baseline. They currently rank for a bronze medal on the leaderboard. You may refer to the organizer's code for reference.",
      "votes": null
    },
    {
      "id": "3400020",
      "postDate": "01/31/2026 17:43:26",
      "content": "<p>Do you use <code>3d_fullres</code> and <code>fold=\"all\"</code> ?</p>",
      "rawMarkdown": "Do you use `3d_fullres` and `fold=\"all\"` ?",
      "votes": null
    },
    {
      "id": "3400256",
      "postDate": "02/01/2026 09:00:11",
      "content": "<p>I use 3d_fullres and fold 0</p>",
      "rawMarkdown": "I use 3d_fullres and fold 0",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3399924,
      "author_name": "nguyncdngs",
      "author_url": "",
      "post_date": "01/31/2026 14:28:18",
      "content": "<p>use default setting can give you a LB score &gt; 0.55x</p>",
      "votes": null,
      "replies": [
        {
          "id": 3400020,
          "author_name": "sjtuwangshuo",
          "author_url": "",
          "post_date": "01/31/2026 17:43:26",
          "content": "<p>Do you use <code>3d_fullres</code> and <code>fold=\"all\"</code> ?</p>",
          "votes": null,
          "replies": [
            {
              "id": 3400256,
              "author_name": "nguyncdngs",
              "author_url": "",
              "post_date": "02/01/2026 09:00:11",
              "content": "<p>I use 3d_fullres and fold 0</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 3399930,
      "author_name": "ggayoayogg",
      "author_url": "",
      "post_date": "01/31/2026 14:44:42",
      "content": "<p>The default settings already provide an excellent baseline. They currently rank for a bronze medal on the leaderboard. You may refer to the organizer's code for reference.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3399911": "If using nnunet for training, how do people usually set the number of epochs? Is the default 1000? Do you train with the full dataset for 1000 epochs?",
    "3399924": "use default setting can give you a LB score > 0.55x",
    "3399930": "The default settings already provide an excellent baseline. They currently rank for a bronze medal on the leaderboard. You may refer to the organizer's code for reference.",
    "3400020": "Do you use `3d_fullres` and `fold=\"all\"` ?",
    "3400256": "I use 3d_fullres and fold 0"
  },
  "source": "meta"
}