{
  "id": 613366,
  "title": "My Kaggle Submission Times Out, Even Though “Save Version” Only Takes 3 Minutes",
  "url": "/competitions/physionet-ecg-image-digitization/discussion/613366",
  "author_name": "",
  "post_date": "2025-10-26T09:56:04.584724900Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hello everyone.</p>\n<p>I’m facing a strange issue, when I Save Version of my notebook, the whole process finishes in around 3 minutes.<br>\nHowever, when I Submit to Competition, the same notebook runs out of time or gets killed due to time limit exceeded.</p>\n<p>I don’t retrain the model at all, it only does inference using a pretrained model from /kaggle/input.</p>\n<p>Has anyone experienced this before?<br>\nWhy would the same notebook run fine in “Save Version” mode but fail in “Submit” mode?</p>",
  "messages": [
    {
      "id": "3307147",
      "postDate": "10/26/2025 09:56:04",
      "content": "<p>Hello everyone.</p>\n<p>I’m facing a strange issue, when I Save Version of my notebook, the whole process finishes in around 3 minutes.<br>\nHowever, when I Submit to Competition, the same notebook runs out of time or gets killed due to time limit exceeded.</p>\n<p>I don’t retrain the model at all, it only does inference using a pretrained model from /kaggle/input.</p>\n<p>Has anyone experienced this before?<br>\nWhy would the same notebook run fine in “Save Version” mode but fail in “Submit” mode?</p>",
      "rawMarkdown": "Hello everyone.\n\nI’m facing a strange issue, when I Save Version of my notebook, the whole process finishes in around 3 minutes.\nHowever, when I Submit to Competition, the same notebook runs out of time or gets killed due to time limit exceeded.\n\nI don’t retrain the model at all, it only does inference using a pretrained model from /kaggle/input.\n\nHas anyone experienced this before?\nWhy would the same notebook run fine in “Save Version” mode but fail in “Submit” mode?",
      "votes": null
    },
    {
      "id": "3307171",
      "postDate": "10/26/2025 11:16:12",
      "content": "<p>When submitted, the test set will be replaced by hidden test set that has 1000 images,</p>",
      "rawMarkdown": "When submitted, the test set will be replaced by hidden test set that has 1000 images,",
      "votes": null
    },
    {
      "id": "3307710",
      "postDate": "10/27/2025 15:48:27",
      "content": "<p>I'm facing the same issue as well, and as <a href=\"https://www.kaggle.com/evanarlenhandy\" target=\"_blank\">@evanarlenhandy</a> said and as i have implemented, why would a notebook with 2-3 minute of just inference give TLE</p>\n<p>Did you submit the whole model training notebook as well for scoring?</p>",
      "rawMarkdown": "I'm facing the same issue as well, and as @evanarlenhandy said and as i have implemented, why would a notebook with 2-3 minute of just inference give TLE\n\nDid you submit the whole model training notebook as well for scoring?",
      "votes": null
    },
    {
      "id": "3307782",
      "postDate": "10/27/2025 19:12:25",
      "content": "<p>I think the TLE happens because the preprocessing and computations get much heavier when the full test data is processed during submission, like when using deep learning models, but that’s just my assumption for now.</p>",
      "rawMarkdown": "I think the TLE happens because the preprocessing and computations get much heavier when the full test data is processed during submission, like when using deep learning models, but that’s just my assumption for now.",
      "votes": null
    },
    {
      "id": "3325228",
      "postDate": "11/15/2025 05:01:52",
      "content": "<p>1000 images in under 9 hours mean each image needs to be processed in ~30 seconds or less.</p>",
      "rawMarkdown": "1000 images in under 9 hours mean each image needs to be processed in ~30 seconds or less.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3307171,
      "author_name": "llkh0a",
      "author_url": "",
      "post_date": "10/26/2025 11:16:12",
      "content": "<p>When submitted, the test set will be replaced by hidden test set that has 1000 images,</p>",
      "votes": null,
      "replies": [
        {
          "id": 3307710,
          "author_name": "avish006",
          "author_url": "",
          "post_date": "10/27/2025 15:48:27",
          "content": "<p>I'm facing the same issue as well, and as <a href=\"https://www.kaggle.com/evanarlenhandy\" target=\"_blank\">@evanarlenhandy</a> said and as i have implemented, why would a notebook with 2-3 minute of just inference give TLE</p>\n<p>Did you submit the whole model training notebook as well for scoring?</p>",
          "votes": null,
          "replies": [
            {
              "id": 3307782,
              "author_name": "evanarlenhandy",
              "author_url": "",
              "post_date": "10/27/2025 19:12:25",
              "content": "<p>I think the TLE happens because the preprocessing and computations get much heavier when the full test data is processed during submission, like when using deep learning models, but that’s just my assumption for now.</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 3325228,
      "author_name": "davidlist",
      "author_url": "",
      "post_date": "11/15/2025 05:01:52",
      "content": "<p>1000 images in under 9 hours mean each image needs to be processed in ~30 seconds or less.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3307147": "Hello everyone.\n\nI’m facing a strange issue, when I Save Version of my notebook, the whole process finishes in around 3 minutes.\nHowever, when I Submit to Competition, the same notebook runs out of time or gets killed due to time limit exceeded.\n\nI don’t retrain the model at all, it only does inference using a pretrained model from /kaggle/input.\n\nHas anyone experienced this before?\nWhy would the same notebook run fine in “Save Version” mode but fail in “Submit” mode?",
    "3307171": "When submitted, the test set will be replaced by hidden test set that has 1000 images,",
    "3307710": "I'm facing the same issue as well, and as @evanarlenhandy said and as i have implemented, why would a notebook with 2-3 minute of just inference give TLE\n\nDid you submit the whole model training notebook as well for scoring?",
    "3307782": "I think the TLE happens because the preprocessing and computations get much heavier when the full test data is processed during submission, like when using deep learning models, but that’s just my assumption for now.",
    "3325228": "1000 images in under 9 hours mean each image needs to be processed in ~30 seconds or less."
  },
  "source": "meta"
}