{
  "id": 107820,
  "title": "Does own dataset occupy disk space during private reruning stage?",
  "url": "/competitions/aptos2019-blindness-detection/discussion/107820",
  "author_name": "",
  "post_date": "2019-09-07T05:10:11.293577500Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hi all, When my kernel ran  with editable mode, I found that my own dataset didn't occupy any disk space in monitor. </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1031751%2Fd004e1617a5226c9781816b2869f8737%2FWeChat%20Image_20190907130728.png?generation=1567832873835424&amp;alt=media\" alt=\"\"></p>\n\n<p>I wish to know if we will get same situation during private reruning stage, coz it looks like we can hv 4.9GB in one session, our kernel could be broken if we don't control our disk space.</p>",
  "messages": [
    {
      "id": "620137",
      "postDate": "09/07/2019 05:10:11",
      "content": "<p>Hi all, When my kernel ran  with editable mode, I found that my own dataset didn't occupy any disk space in monitor. </p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1031751%2Fd004e1617a5226c9781816b2869f8737%2FWeChat%20Image_20190907130728.png?generation=1567832873835424&amp;alt=media\" alt=\"\"></p>\n\n<p>I wish to know if we will get same situation during private reruning stage, coz it looks like we can hv 4.9GB in one session, our kernel could be broken if we don't control our disk space.</p>",
      "rawMarkdown": "Hi all, When my kernel ran  with editable mode, I found that my own dataset didn't occupy any disk space in monitor. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1031751%2Fd004e1617a5226c9781816b2869f8737%2FWeChat%20Image_20190907130728.png?generation=1567832873835424&amp;alt=media)\n\n\nI wish to know if we will get same situation during private reruning stage, coz it looks like we can hv 4.9GB in one session, our kernel could be broken if we don't control our disk space.",
      "votes": null
    },
    {
      "id": "620169",
      "postDate": "09/07/2019 05:38:47",
      "content": "<p>When you add data, the data stores as 'input only' by default, it will only occupy your disk space if you copy it to other directory(../working).\nAlso, in this kind of competition your kernel won't be rerun. All your submissions has already been fed with private test data(if you do inferece with test.csv) when submitting.</p>",
      "rawMarkdown": "When you add data, the data stores as 'input only' by default, it will only occupy your disk space if you copy it to other directory(../working).\nAlso, in this kind of competition your kernel won't be rerun. All your submissions has already been fed with private test data(if you do inferece with test.csv) when submitting.",
      "votes": null
    },
    {
      "id": "620226",
      "postDate": "09/07/2019 07:09:48",
      "content": "<p>wait, do u mean our kernel are fed with whole private test data when submitting even we are still in public stage?</p>",
      "rawMarkdown": "wait, do u mean our kernel are fed with whole private test data when submitting even we are still in public stage?",
      "votes": null
    },
    {
      "id": "620234",
      "postDate": "09/07/2019 07:24:04",
      "content": "<p>YES.</p>",
      "rawMarkdown": "YES.",
      "votes": null
    },
    {
      "id": "620237",
      "postDate": "09/07/2019 07:29:44",
      "content": "<p>well, that's y each submission spend weird long time even with fast inference</p>",
      "rawMarkdown": "well, that's y each submission spend weird long time even with fast inference",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 620169,
      "author_name": "homoalways",
      "author_url": "",
      "post_date": "09/07/2019 05:38:47",
      "content": "<p>When you add data, the data stores as 'input only' by default, it will only occupy your disk space if you copy it to other directory(../working).\nAlso, in this kind of competition your kernel won't be rerun. All your submissions has already been fed with private test data(if you do inferece with test.csv) when submitting.</p>",
      "votes": null,
      "replies": [
        {
          "id": 620226,
          "author_name": "mikelkl",
          "author_url": "",
          "post_date": "09/07/2019 07:09:48",
          "content": "<p>wait, do u mean our kernel are fed with whole private test data when submitting even we are still in public stage?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 620234,
          "author_name": "homoalways",
          "author_url": "",
          "post_date": "09/07/2019 07:24:04",
          "content": "<p>YES.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 620237,
          "author_name": "mikelkl",
          "author_url": "",
          "post_date": "09/07/2019 07:29:44",
          "content": "<p>well, that's y each submission spend weird long time even with fast inference</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "620137": "Hi all, When my kernel ran  with editable mode, I found that my own dataset didn't occupy any disk space in monitor. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1031751%2Fd004e1617a5226c9781816b2869f8737%2FWeChat%20Image_20190907130728.png?generation=1567832873835424&amp;alt=media)\n\n\nI wish to know if we will get same situation during private reruning stage, coz it looks like we can hv 4.9GB in one session, our kernel could be broken if we don't control our disk space.",
    "620169": "When you add data, the data stores as 'input only' by default, it will only occupy your disk space if you copy it to other directory(../working).\nAlso, in this kind of competition your kernel won't be rerun. All your submissions has already been fed with private test data(if you do inferece with test.csv) when submitting.",
    "620226": "wait, do u mean our kernel are fed with whole private test data when submitting even we are still in public stage?",
    "620234": "YES.",
    "620237": "well, that's y each submission spend weird long time even with fast inference"
  },
  "source": "meta"
}