{
  "id": 75308,
  "title": "Private test set",
  "url": "/competitions/quora-insincere-questions-classification/discussion/75308",
  "author_name": "",
  "post_date": "2018-12-20T12:32:04.831280200Z",
  "votes": 2,
  "comment_count": 6,
  "views": 0,
  "content": "<p>in the data section, it is mentioned like this</p>\n\n<p>In the second stage of the competition, we will re-run your selected Kernels. […] This file will have ~56k rows in stage 1 and ~376k rows in stage 2. The public leaderboard data remains the same for both versions. The file name will be the same (both test.csv) to ensure that your code will run.</p>\n\n<p>what is the meaning of this public leaderboard data remains the same?</p>\n\n<p>Does this mean that out of 376 k rows 56 k rows will be the same (which are given to us) and remaining 320k rows will be new \nor total 376k rows will be different than this public test set.</p>\n\n<p>Can anyone clarify more about this ?</p>",
  "messages": [
    {
      "id": "442744",
      "postDate": "12/20/2018 12:32:04",
      "content": "<p>in the data section, it is mentioned like this</p>\n\n<p>In the second stage of the competition, we will re-run your selected Kernels. […] This file will have ~56k rows in stage 1 and ~376k rows in stage 2. The public leaderboard data remains the same for both versions. The file name will be the same (both test.csv) to ensure that your code will run.</p>\n\n<p>what is the meaning of this public leaderboard data remains the same?</p>\n\n<p>Does this mean that out of 376 k rows 56 k rows will be the same (which are given to us) and remaining 320k rows will be new \nor total 376k rows will be different than this public test set.</p>\n\n<p>Can anyone clarify more about this ?</p>",
      "rawMarkdown": "in the data section, it is mentioned like this\n\nIn the second stage of the competition, we will re-run your selected Kernels. […] This file will have ~56k rows in stage 1 and ~376k rows in stage 2. The public leaderboard data remains the same for both versions. The file name will be the same (both test.csv) to ensure that your code will run.\n\nwhat is the meaning of this public leaderboard data remains the same?\n\nDoes this mean that out of 376 k rows 56 k rows will be the same (which are given to us) and remaining 320k rows will be new \nor total 376k rows will be different than this public test set.\n\nCan anyone clarify more about this ?",
      "votes": null
    },
    {
      "id": "442758",
      "postDate": "12/20/2018 13:03:07",
      "content": "<p>I would like to know if there is a time limit in the second phase?</p>",
      "rawMarkdown": "I would like to know if there is a time limit in the second phase?",
      "votes": null
    },
    {
      "id": "442877",
      "postDate": "12/20/2018 16:21:22",
      "content": "<p>same time limits. there is no change. 2hrs for gpu kernel and 6hrs for cpu kernel</p>",
      "rawMarkdown": "same time limits. there is no change. 2hrs for gpu kernel and 6hrs for cpu kernel",
      "votes": null
    },
    {
      "id": "443198",
      "postDate": "12/21/2018 07:44:41",
      "content": "<p>Yes, this was confirmed in the welcome thread, so if your current kernel runs tight on 2hrs, you probably would have a problem.</p>",
      "rawMarkdown": "Yes, this was confirmed in the welcome thread, so if your current kernel runs tight on 2hrs, you probably would have a problem.",
      "votes": null
    },
    {
      "id": "443297",
      "postDate": "12/21/2018 11:35:33",
      "content": "<p>Philip \nWhat do you think about the question I asked ?</p>",
      "rawMarkdown": "Philip \nWhat do you think about the question I asked ?",
      "votes": null
    },
    {
      "id": "443308",
      "postDate": "12/21/2018 11:55:29",
      "content": "<p>I can't say. Usually it is different, and I really hope it will be.</p>",
      "rawMarkdown": "I can't say. Usually it is different, and I really hope it will be.",
      "votes": null
    },
    {
      "id": "443529",
      "postDate": "12/21/2018 20:08:42",
      "content": "<p>My understanding is that for the leaderboard nothing will change. Our kernels will take slightly more time, but the leaderboard will still be calculated on the same subset. So we will not get any additional information regarding the distribution... Which sucks honestly. The current test set is so fickle, having some more data to verify hypotheses would be very helpful.</p>",
      "rawMarkdown": "My understanding is that for the leaderboard nothing will change. Our kernels will take slightly more time, but the leaderboard will still be calculated on the same subset. So we will not get any additional information regarding the distribution... Which sucks honestly. The current test set is so fickle, having some more data to verify hypotheses would be very helpful.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 442758,
      "author_name": "xiaobai1123q",
      "author_url": "",
      "post_date": "12/20/2018 13:03:07",
      "content": "<p>I would like to know if there is a time limit in the second phase?</p>",
      "votes": null,
      "replies": [
        {
          "id": 442877,
          "author_name": "suchith0312",
          "author_url": "",
          "post_date": "12/20/2018 16:21:22",
          "content": "<p>same time limits. there is no change. 2hrs for gpu kernel and 6hrs for cpu kernel</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 443198,
          "author_name": "philippsinger",
          "author_url": "",
          "post_date": "12/21/2018 07:44:41",
          "content": "<p>Yes, this was confirmed in the welcome thread, so if your current kernel runs tight on 2hrs, you probably would have a problem.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 443297,
          "author_name": "suchith0312",
          "author_url": "",
          "post_date": "12/21/2018 11:35:33",
          "content": "<p>Philip \nWhat do you think about the question I asked ?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 443308,
          "author_name": "philippsinger",
          "author_url": "",
          "post_date": "12/21/2018 11:55:29",
          "content": "<p>I can't say. Usually it is different, and I really hope it will be.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 443529,
      "author_name": "mschumacher",
      "author_url": "",
      "post_date": "12/21/2018 20:08:42",
      "content": "<p>My understanding is that for the leaderboard nothing will change. Our kernels will take slightly more time, but the leaderboard will still be calculated on the same subset. So we will not get any additional information regarding the distribution... Which sucks honestly. The current test set is so fickle, having some more data to verify hypotheses would be very helpful.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "442744": "in the data section, it is mentioned like this\n\nIn the second stage of the competition, we will re-run your selected Kernels. […] This file will have ~56k rows in stage 1 and ~376k rows in stage 2. The public leaderboard data remains the same for both versions. The file name will be the same (both test.csv) to ensure that your code will run.\n\nwhat is the meaning of this public leaderboard data remains the same?\n\nDoes this mean that out of 376 k rows 56 k rows will be the same (which are given to us) and remaining 320k rows will be new \nor total 376k rows will be different than this public test set.\n\nCan anyone clarify more about this ?",
    "442758": "I would like to know if there is a time limit in the second phase?",
    "442877": "same time limits. there is no change. 2hrs for gpu kernel and 6hrs for cpu kernel",
    "443198": "Yes, this was confirmed in the welcome thread, so if your current kernel runs tight on 2hrs, you probably would have a problem.",
    "443297": "Philip \nWhat do you think about the question I asked ?",
    "443308": "I can't say. Usually it is different, and I really hope it will be.",
    "443529": "My understanding is that for the leaderboard nothing will change. Our kernels will take slightly more time, but the leaderboard will still be calculated on the same subset. So we will not get any additional information regarding the distribution... Which sucks honestly. The current test set is so fickle, having some more data to verify hypotheses would be very helpful."
  },
  "source": "meta"
}