{
  "id": 598871,
  "title": "Question about run time — is a 5h submission too long?",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/598871",
  "author_name": "Zuohong Pan",
  "post_date": "2025-08-13T07:24:09.364000",
  "votes": 0,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I noticed that <em>“This leaderboard is calculated with approximately 42% of the test data. The final results will be based on the other 58%, so the final standings may be different.”</em> and also that <em>“GPU Notebook &lt;= 9 hours run-time”.</em></p>\n<p>I’m wondering: when the code is re-run on the private test set, will my current 5-hour runtime increase to around 11 hours (if it’s run on the full dataset), or only to about 6 hours (if it’s run just on the remaining 58% of the data)?</p>",
  "messages": [
    {
      "id": 3268616,
      "postDate": "2025-08-13T08:14:34.760Z",
      "content": "<blockquote>\n  <p>I’m wondering: when the code is re-run on the private test set, will my current 5-hour runtime increase to around 11 hours (if it’s run on the full dataset), or only to about 6 hours (if it’s run just on the remaining 58% of the data)?</p>\n</blockquote>\n<p>Don't worry about this issue, because your submission includes calculations based on the all test set including private dataset. You just can't see the private score yet.</p>",
      "rawMarkdown": ">I’m wondering: when the code is re-run on the private test set, will my current 5-hour runtime increase to around 11 hours (if it’s run on the full dataset), or only to about 6 hours (if it’s run just on the remaining 58% of the data)?\n\nDon't worry about this issue, because your submission includes calculations based on the all test set including private dataset. You just can't see the private score yet.\n\n\n\n",
      "votes": 3,
      "replies": [
        {
          "id": 3268756,
          "postDate": "2025-08-13T11:02:39.297Z",
          "content": "<p>Thank you for your two replies.</p>\n<p>After comparing with the open-source notebook, I think it’s because I’ve included too many models — about eight — and also have two sets of feature engineering, which makes the runtime quite costly. I’m considering removing a few of these models.</p>",
          "rawMarkdown": "Thank you for your two replies.\n\nAfter comparing with the open-source notebook, I think it’s because I’ve included too many models — about eight — and also have two sets of feature engineering, which makes the runtime quite costly. I’m considering removing a few of these models."
        }
      ]
    },
    {
      "id": 3268593,
      "postDate": "2025-08-13T07:24:09.363Z",
      "content": "<p>I noticed that <em>“This leaderboard is calculated with approximately 42% of the test data. The final results will be based on the other 58%, so the final standings may be different.”</em> and also that <em>“GPU Notebook &lt;= 9 hours run-time”.</em></p>\n<p>I’m wondering: when the code is re-run on the private test set, will my current 5-hour runtime increase to around 11 hours (if it’s run on the full dataset), or only to about 6 hours (if it’s run just on the remaining 58% of the data)?</p>",
      "rawMarkdown": "I noticed that *“This leaderboard is calculated with approximately 42% of the test data. The final results will be based on the other 58%, so the final standings may be different.”* and also that *“GPU Notebook <= 9 hours run-time”.*\n\nI’m wondering: when the code is re-run on the private test set, will my current 5-hour runtime increase to around 11 hours (if it’s run on the full dataset), or only to about 6 hours (if it’s run just on the remaining 58% of the data)?"
    },
    {
      "id": 3268605,
      "postDate": "2025-08-13T07:56:25.710Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 3268616,
      "author_name": "Timmy Juicehouse",
      "author_url": "",
      "post_date": "2025-08-13T08:14:34.760000",
      "content": "<blockquote>\n  <p>I’m wondering: when the code is re-run on the private test set, will my current 5-hour runtime increase to around 11 hours (if it’s run on the full dataset), or only to about 6 hours (if it’s run just on the remaining 58% of the data)?</p>\n</blockquote>\n<p>Don't worry about this issue, because your submission includes calculations based on the all test set including private dataset. You just can't see the private score yet.</p>",
      "votes": 3,
      "replies": [
        {
          "id": 3268756,
          "author_name": "Zuohong Pan",
          "author_url": "",
          "post_date": "2025-08-13T11:02:39.297000",
          "content": "<p>Thank you for your two replies.</p>\n<p>After comparing with the open-source notebook, I think it’s because I’ve included too many models — about eight — and also have two sets of feature engineering, which makes the runtime quite costly. I’m considering removing a few of these models.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 3268605,
      "author_name": "",
      "author_url": "",
      "post_date": "2025-08-13T07:56:25.710000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3268616": ">I’m wondering: when the code is re-run on the private test set, will my current 5-hour runtime increase to around 11 hours (if it’s run on the full dataset), or only to about 6 hours (if it’s run just on the remaining 58% of the data)?\n\nDon't worry about this issue, because your submission includes calculations based on the all test set including private dataset. You just can't see the private score yet.\n\n\n\n",
    "3268593": "I noticed that *“This leaderboard is calculated with approximately 42% of the test data. The final results will be based on the other 58%, so the final standings may be different.”* and also that *“GPU Notebook <= 9 hours run-time”.*\n\nI’m wondering: when the code is re-run on the private test set, will my current 5-hour runtime increase to around 11 hours (if it’s run on the full dataset), or only to about 6 hours (if it’s run just on the remaining 58% of the data)?",
    "3268605": ""
  }
}