{
  "id": 77466,
  "title": "How much runtime buffer is enough for final submission?",
  "url": "/competitions/quora-insincere-questions-classification/discussion/77466",
  "author_name": "",
  "post_date": "2019-01-13T04:25:54.614681900Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Runtime will be affected by\n1. runtime variability, as we can see, runtime of same kernel may differ\n2. size of private test set, this will affect inference time</p>\n\n<p>So, how much runtime buffer will you leave for final submission?</p>",
  "messages": [
    {
      "id": "455149",
      "postDate": "01/13/2019 04:25:54",
      "content": "<p>Runtime will be affected by\n1. runtime variability, as we can see, runtime of same kernel may differ\n2. size of private test set, this will affect inference time</p>\n\n<p>So, how much runtime buffer will you leave for final submission?</p>",
      "rawMarkdown": "Runtime will be affected by\n1. runtime variability, as we can see, runtime of same kernel may differ\n2. size of private test set, this will affect inference time\n\nSo, how much runtime buffer will you leave for final submission?",
      "votes": null
    },
    {
      "id": "455164",
      "postDate": "01/13/2019 05:32:50",
      "content": "<p>For ours, it will probably increase by 100s.</p>",
      "rawMarkdown": "For ours, it will probably increase by 100s.",
      "votes": null
    },
    {
      "id": "455165",
      "postDate": "01/13/2019 05:34:40",
      "content": "<p>What about 30  (for runtime variability) + #models* 2(estimated seconds to make predictions per model on current test set)* (376/56-1)  seconds?  </p>",
      "rawMarkdown": "What about 30  (for runtime variability) + #models* 2(estimated seconds to make predictions per model on current test set)* (376/56-1)  seconds?",
      "votes": null
    },
    {
      "id": "455203",
      "postDate": "01/13/2019 08:29:48",
      "content": "<p>Depends on your preprocessing routines.</p>",
      "rawMarkdown": "Depends on your preprocessing routines.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 455164,
      "author_name": "xiaobai1123q",
      "author_url": "",
      "post_date": "01/13/2019 05:32:50",
      "content": "<p>For ours, it will probably increase by 100s.</p>",
      "votes": null,
      "replies": [
        {
          "id": 455203,
          "author_name": "philippsinger",
          "author_url": "",
          "post_date": "01/13/2019 08:29:48",
          "content": "<p>Depends on your preprocessing routines.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 455165,
      "author_name": "jiangm",
      "author_url": "",
      "post_date": "01/13/2019 05:34:40",
      "content": "<p>What about 30  (for runtime variability) + #models* 2(estimated seconds to make predictions per model on current test set)* (376/56-1)  seconds?  </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "455149": "Runtime will be affected by\n1. runtime variability, as we can see, runtime of same kernel may differ\n2. size of private test set, this will affect inference time\n\nSo, how much runtime buffer will you leave for final submission?",
    "455164": "For ours, it will probably increase by 100s.",
    "455165": "What about 30  (for runtime variability) + #models* 2(estimated seconds to make predictions per model on current test set)* (376/56-1)  seconds?",
    "455203": "Depends on your preprocessing routines."
  },
  "source": "meta"
}