{
  "id": 670306,
  "title": "Runtime safety margin",
  "url": "/competitions/csiro-biomass/discussion/670306",
  "author_name": "",
  "post_date": "2026-01-27T12:11:44.194668900Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Since this is a code competition and the notebook must finish in &lt; 9 hours for the Submit button / valid submission, can we treat the public submission runtime as a reliable proxy for the final submission rerun runtime?</p>\n<p>Concretely: if my public submission completes in ≤ 9h, is it generally safe to assume the final rerun will also stay under 9h? If not, what “headspace” margin do people recommend (e.g., target ≤ 8h, ≤ 8.5h, etc.) to be safe? </p>\n<p>Tnx</p>",
  "messages": [
    {
      "id": "3397502",
      "postDate": "01/27/2026 12:11:44",
      "content": "<p>Since this is a code competition and the notebook must finish in &lt; 9 hours for the Submit button / valid submission, can we treat the public submission runtime as a reliable proxy for the final submission rerun runtime?</p>\n<p>Concretely: if my public submission completes in ≤ 9h, is it generally safe to assume the final rerun will also stay under 9h? If not, what “headspace” margin do people recommend (e.g., target ≤ 8h, ≤ 8.5h, etc.) to be safe? </p>\n<p>Tnx</p>",
      "rawMarkdown": "Since this is a code competition and the notebook must finish in < 9 hours for the Submit button / valid submission, can we treat the public submission runtime as a reliable proxy for the final submission rerun runtime?\n\nConcretely: if my public submission completes in ≤ 9h, is it generally safe to assume the final rerun will also stay under 9h? If not, what “headspace” margin do people recommend (e.g., target ≤ 8h, ≤ 8.5h, etc.) to be safe? \n\nTnx",
      "votes": null
    },
    {
      "id": "3397503",
      "postDate": "01/27/2026 12:13:41",
      "content": "<p>There is no rerun, when you submit the inference is done on all the test data, but the score on the private set isn't shown yet.</p>",
      "rawMarkdown": "There is no rerun, when you submit the inference is done on all the test data, but the score on the private set isn't shown yet.",
      "votes": null
    },
    {
      "id": "3397508",
      "postDate": "01/27/2026 12:21:03",
      "content": "<p>Got it, thanks for clarifying. So when I submit a notebook, Kaggle runs inference on the full test set (including the private portion); the private score is just withheld until the end.\nAnd if I “Submit” one of my already-committed notebooks and it completes successfully, then I know it produces a valid submission within the runtime limits.</p>",
      "rawMarkdown": "Got it, thanks for clarifying. So when I submit a notebook, Kaggle runs inference on the full test set (including the private portion); the private score is just withheld until the end.\nAnd if I “Submit” one of my already-committed notebooks and it completes successfully, then I know it produces a valid submission within the runtime limits.",
      "votes": null
    },
    {
      "id": "3397511",
      "postDate": "01/27/2026 12:29:41",
      "content": "<p>Correct, that's how it works for most competitions, except when the test data is generated after the deadline (e.g. time series).</p>",
      "rawMarkdown": "Correct, that's how it works for most competitions, except when the test data is generated after the deadline (e.g. time series).",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3397503,
      "author_name": "tennogh",
      "author_url": "",
      "post_date": "01/27/2026 12:13:41",
      "content": "<p>There is no rerun, when you submit the inference is done on all the test data, but the score on the private set isn't shown yet.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3397508,
          "author_name": "davidvikstrand",
          "author_url": "",
          "post_date": "01/27/2026 12:21:03",
          "content": "<p>Got it, thanks for clarifying. So when I submit a notebook, Kaggle runs inference on the full test set (including the private portion); the private score is just withheld until the end.\nAnd if I “Submit” one of my already-committed notebooks and it completes successfully, then I know it produces a valid submission within the runtime limits.</p>",
          "votes": null,
          "replies": [
            {
              "id": 3397511,
              "author_name": "tennogh",
              "author_url": "",
              "post_date": "01/27/2026 12:29:41",
              "content": "<p>Correct, that's how it works for most competitions, except when the test data is generated after the deadline (e.g. time series).</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3397502": "Since this is a code competition and the notebook must finish in < 9 hours for the Submit button / valid submission, can we treat the public submission runtime as a reliable proxy for the final submission rerun runtime?\n\nConcretely: if my public submission completes in ≤ 9h, is it generally safe to assume the final rerun will also stay under 9h? If not, what “headspace” margin do people recommend (e.g., target ≤ 8h, ≤ 8.5h, etc.) to be safe? \n\nTnx",
    "3397503": "There is no rerun, when you submit the inference is done on all the test data, but the score on the private set isn't shown yet.",
    "3397508": "Got it, thanks for clarifying. So when I submit a notebook, Kaggle runs inference on the full test set (including the private portion); the private score is just withheld until the end.\nAnd if I “Submit” one of my already-committed notebooks and it completes successfully, then I know it produces a valid submission within the runtime limits.",
    "3397511": "Correct, that's how it works for most competitions, except when the test data is generated after the deadline (e.g. time series)."
  },
  "source": "meta"
}