{
  "id": 588336,
  "title": " notebook inference server never started error ",
  "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/588336",
  "author_name": "",
  "post_date": "2025-07-05T17:14:59.399734Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>i constantly got notebook inference server never started error in the scoring phase despite that my model is passing the submission phase </p>",
  "messages": [
    {
      "id": "3242200",
      "postDate": "07/05/2025 17:14:59",
      "content": "<p>i constantly got notebook inference server never started error in the scoring phase despite that my model is passing the submission phase </p>",
      "rawMarkdown": "i constantly got notebook inference server never started error in the scoring phase despite that my model is passing the submission phase",
      "votes": null
    },
    {
      "id": "3242424",
      "postDate": "07/05/2025 21:09:54",
      "content": "<p>the predict function should start in 15 minutes, if your notebook run for 15 minutes before it reach the predict function you will get this error, so train in another notebook and load everything in another notebook for inference</p>",
      "rawMarkdown": "the predict function should start in 15 minutes, if your notebook run for 15 minutes before it reach the predict function you will get this error, so train in another notebook and load everything in another notebook for inference",
      "votes": null
    },
    {
      "id": "3243155",
      "postDate": "07/06/2025 19:52:38",
      "content": "<p>Indeed. You can <code>pickle.dump</code> the predict function (if it is properly self contained) that is generated in your training notebook and then <code>pickle.load</code> it from the training notebook attached as an input to a separate (inference) notebook.</p>",
      "rawMarkdown": "Indeed. You can `pickle.dump` the predict function (if it is properly self contained) that is generated in your training notebook and then `pickle.load` it from the training notebook attached as an input to a separate (inference) notebook.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3242424,
      "author_name": "tahaalshatiri",
      "author_url": "",
      "post_date": "07/05/2025 21:09:54",
      "content": "<p>the predict function should start in 15 minutes, if your notebook run for 15 minutes before it reach the predict function you will get this error, so train in another notebook and load everything in another notebook for inference</p>",
      "votes": null,
      "replies": [
        {
          "id": 3243155,
          "author_name": "nnjjpp",
          "author_url": "",
          "post_date": "07/06/2025 19:52:38",
          "content": "<p>Indeed. You can <code>pickle.dump</code> the predict function (if it is properly self contained) that is generated in your training notebook and then <code>pickle.load</code> it from the training notebook attached as an input to a separate (inference) notebook.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3242200": "i constantly got notebook inference server never started error in the scoring phase despite that my model is passing the submission phase",
    "3242424": "the predict function should start in 15 minutes, if your notebook run for 15 minutes before it reach the predict function you will get this error, so train in another notebook and load everything in another notebook for inference",
    "3243155": "Indeed. You can `pickle.dump` the predict function (if it is properly self contained) that is generated in your training notebook and then `pickle.load` it from the training notebook attached as an input to a separate (inference) notebook."
  },
  "source": "meta"
}