{
  "id": 372389,
  "title": "Does this competition have a runtime limit? Worried about \"Notebook Timeout\" error",
  "url": "/competitions/g2net-detecting-continuous-gravitational-waves/discussion/372389",
  "author_name": "Jim Woo",
  "post_date": "2022-12-15T18:55:29.957000",
  "votes": 0,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I am working on a CPU based method that could take a while to complete on the competition's hidden data set. I am worried about \"Notebook Timeout\" errors. If needed, I can adapt my method to be faster but less accurate. But how do I know what is needed?</p>\n<p>Do you have any information on how much time a model is allowed to spend per test sample?</p>\n<p>I searched the competition rules but did not find anything on this partcular subject.</p>\n<p>Thanks</p>",
  "messages": [
    {
      "id": 2066582,
      "postDate": "2022-12-15T21:18:39.060Z",
      "content": "<p>This competition is not a \"code competition\", so you can run all of your inference locally, and just upload the final csv; so you don't need to worry about timeout errors :) </p>",
      "rawMarkdown": "This competition is not a \"code competition\", so you can run all of your inference locally, and just upload the final csv; so you don't need to worry about timeout errors :) ",
      "votes": 3,
      "replies": [
        {
          "id": 2067079,
          "postDate": "2022-12-16T11:30:54.193Z",
          "content": "<p><a href=\"https://www.kaggle.com/jimwoo\" target=\"_blank\">@jimwoo</a> In case you dont have the resources available locally you can also run it on Kaggle till the maximum runtime allowance, save your model weights and continue your training with a refreshed session (after loading your weights from the previous session). </p>",
          "rawMarkdown": "@jimwoo In case you dont have the resources available locally you can also run it on Kaggle till the maximum runtime allowance, save your model weights and continue your training with a refreshed session (after loading your weights from the previous session). ",
          "votes": 2
        },
        {
          "id": 2067463,
          "postDate": "2022-12-16T18:23:38.407Z",
          "content": "<p>Thanks. Can't believe I had not spotted that. I see now that the public leaderboard is simply based on 24% of the test data.</p>",
          "rawMarkdown": "Thanks. Can't believe I had not spotted that. I see now that the public leaderboard is simply based on 24% of the test data."
        }
      ]
    },
    {
      "id": 2066493,
      "postDate": "2022-12-15T18:55:29.957Z",
      "content": "<p>I am working on a CPU based method that could take a while to complete on the competition's hidden data set. I am worried about \"Notebook Timeout\" errors. If needed, I can adapt my method to be faster but less accurate. But how do I know what is needed?</p>\n<p>Do you have any information on how much time a model is allowed to spend per test sample?</p>\n<p>I searched the competition rules but did not find anything on this partcular subject.</p>\n<p>Thanks</p>",
      "rawMarkdown": "I am working on a CPU based method that could take a while to complete on the competition's hidden data set. I am worried about \"Notebook Timeout\" errors. If needed, I can adapt my method to be faster but less accurate. But how do I know what is needed?\n\nDo you have any information on how much time a model is allowed to spend per test sample?\n\nI searched the competition rules but did not find anything on this partcular subject.\n\nThanks"
    },
    {
      "id": 2068520,
      "postDate": "2022-12-18T04:34:52.583Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 2066582,
      "author_name": "chris",
      "author_url": "",
      "post_date": "2022-12-15T21:18:39.060000",
      "content": "<p>This competition is not a \"code competition\", so you can run all of your inference locally, and just upload the final csv; so you don't need to worry about timeout errors :) </p>",
      "votes": 3,
      "replies": [
        {
          "id": 2067079,
          "author_name": "Ali Abdin",
          "author_url": "",
          "post_date": "2022-12-16T11:30:54.193000",
          "content": "<p><a href=\"https://www.kaggle.com/jimwoo\" target=\"_blank\">@jimwoo</a> In case you dont have the resources available locally you can also run it on Kaggle till the maximum runtime allowance, save your model weights and continue your training with a refreshed session (after loading your weights from the previous session). </p>",
          "votes": 2,
          "replies": []
        },
        {
          "id": 2067463,
          "author_name": "Jim Woo",
          "author_url": "",
          "post_date": "2022-12-16T18:23:38.407000",
          "content": "<p>Thanks. Can't believe I had not spotted that. I see now that the public leaderboard is simply based on 24% of the test data.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 2068520,
      "author_name": "",
      "author_url": "",
      "post_date": "2022-12-18T04:34:52.583000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2066582": "This competition is not a \"code competition\", so you can run all of your inference locally, and just upload the final csv; so you don't need to worry about timeout errors :) ",
    "2066493": "I am working on a CPU based method that could take a while to complete on the competition's hidden data set. I am worried about \"Notebook Timeout\" errors. If needed, I can adapt my method to be faster but less accurate. But how do I know what is needed?\n\nDo you have any information on how much time a model is allowed to spend per test sample?\n\nI searched the competition rules but did not find anything on this partcular subject.\n\nThanks",
    "2068520": ""
  }
}