{
  "id": 115082,
  "title": "What's the difference between code-only and kernel-only?",
  "url": "/competitions/tensorflow2-question-answering/discussion/115082",
  "author_name": "",
  "post_date": "2019-10-31T06:00:29.867115700Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I have some questions about this QA competition, I hope to get Kagglers' help.\nWhich format of result should I upload?\nAs the competition said ,external data is allowed, so how large is allowed?\nUsually competitions are in kernel-only or csv-only mode, so what's the difference between code-only and them?\nHope to get some help!</p>",
  "messages": [
    {
      "id": "662136",
      "postDate": "10/31/2019 06:00:29",
      "content": "<p>I have some questions about this QA competition, I hope to get Kagglers' help.\nWhich format of result should I upload?\nAs the competition said ,external data is allowed, so how large is allowed?\nUsually competitions are in kernel-only or csv-only mode, so what's the difference between code-only and them?\nHope to get some help!</p>",
      "rawMarkdown": "I have some questions about this QA competition, I hope to get Kagglers' help.\nWhich format of result should I upload?\nAs the competition said ,external data is allowed, so how large is allowed?\nUsually competitions are in kernel-only or csv-only mode, so what's the difference between code-only and them?\nHope to get some help!",
      "votes": null
    },
    {
      "id": "662154",
      "postDate": "10/31/2019 06:29:21",
      "content": "<p><a href=\"/qhd0081\">@qhd0081</a> The code only and kernel only is the same i think. Cause in Kaggle <a href=\"https://www.kaggle.com/docs/competitions\">documentation </a> they have replaced kernel only with code only and the only example for code only competition attached there is <a href=\"https://www.kaggle.com/c/quora-insincere-questions-classification\">Quora Insincere Questions Classification</a> which is a kernel only competition.</p>",
      "rawMarkdown": "qhd0081 The code only and kernel only is the same i think. Cause in Kaggle [documentation ](https://www.kaggle.com/docs/competitions) they have replaced kernel only with code only and the only example for code only competition attached there is [Quora Insincere Questions Classification](https://www.kaggle.com/c/quora-insincere-questions-classification) which is a kernel only competition.",
      "votes": null
    },
    {
      "id": "662158",
      "postDate": "10/31/2019 06:38:57",
      "content": "<p>If you check the starter kernel output <a href=\"https://www.kaggle.com/philculliton/using-tensorflow-2-0-w-bert-on-nq/output\">here</a>, for each example_id there is a long ans and short ans. You are going to add a Prediction string column which will have start_token_no: end_token_no for that long or short ans. For example, check this\n<code>\nanswer.append(str(short_answer[\"start_token\"]) + \":\" + str(short_answer[\"end_token\"]))\n</code></p>",
      "rawMarkdown": "If you check the starter kernel output [here](https://www.kaggle.com/philculliton/using-tensorflow-2-0-w-bert-on-nq/output), for each example_id there is a long ans and short ans. You are going to add a Prediction string column which will have start_token_no: end_token_no for that long or short ans. For example, check this\n```\nanswer.append(str(short_answer[\"start_token\"]) + \":\" + str(short_answer[\"end_token\"]))\n```",
      "votes": null
    },
    {
      "id": "662161",
      "postDate": "10/31/2019 06:44:24",
      "content": "<p><a href=\"/khairulislam\">@khairulislam</a> Got it, thank you very much!</p>",
      "rawMarkdown": "khairulislam Got it, thank you very much!",
      "votes": null
    },
    {
      "id": "662163",
      "postDate": "10/31/2019 06:47:10",
      "content": "<p>Thank you too</p>",
      "rawMarkdown": "Thank you too",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 662154,
      "author_name": "khairulislam",
      "author_url": "",
      "post_date": "10/31/2019 06:29:21",
      "content": "<p><a href=\"/qhd0081\">@qhd0081</a> The code only and kernel only is the same i think. Cause in Kaggle <a href=\"https://www.kaggle.com/docs/competitions\">documentation </a> they have replaced kernel only with code only and the only example for code only competition attached there is <a href=\"https://www.kaggle.com/c/quora-insincere-questions-classification\">Quora Insincere Questions Classification</a> which is a kernel only competition.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 662158,
      "author_name": "khairulislam",
      "author_url": "",
      "post_date": "10/31/2019 06:38:57",
      "content": "<p>If you check the starter kernel output <a href=\"https://www.kaggle.com/philculliton/using-tensorflow-2-0-w-bert-on-nq/output\">here</a>, for each example_id there is a long ans and short ans. You are going to add a Prediction string column which will have start_token_no: end_token_no for that long or short ans. For example, check this\n<code>\nanswer.append(str(short_answer[\"start_token\"]) + \":\" + str(short_answer[\"end_token\"]))\n</code></p>",
      "votes": null,
      "replies": [
        {
          "id": 662161,
          "author_name": "qhd0081",
          "author_url": "",
          "post_date": "10/31/2019 06:44:24",
          "content": "<p><a href=\"/khairulislam\">@khairulislam</a> Got it, thank you very much!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 662163,
          "author_name": "khairulislam",
          "author_url": "",
          "post_date": "10/31/2019 06:47:10",
          "content": "<p>Thank you too</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "662136": "I have some questions about this QA competition, I hope to get Kagglers' help.\nWhich format of result should I upload?\nAs the competition said ,external data is allowed, so how large is allowed?\nUsually competitions are in kernel-only or csv-only mode, so what's the difference between code-only and them?\nHope to get some help!",
    "662154": "qhd0081 The code only and kernel only is the same i think. Cause in Kaggle [documentation ](https://www.kaggle.com/docs/competitions) they have replaced kernel only with code only and the only example for code only competition attached there is [Quora Insincere Questions Classification](https://www.kaggle.com/c/quora-insincere-questions-classification) which is a kernel only competition.",
    "662158": "If you check the starter kernel output [here](https://www.kaggle.com/philculliton/using-tensorflow-2-0-w-bert-on-nq/output), for each example_id there is a long ans and short ans. You are going to add a Prediction string column which will have start_token_no: end_token_no for that long or short ans. For example, check this\n```\nanswer.append(str(short_answer[\"start_token\"]) + \":\" + str(short_answer[\"end_token\"]))\n```",
    "662161": "khairulislam Got it, thank you very much!",
    "662163": "Thank you too"
  },
  "source": "meta"
}