{
  "id": 181417,
  "title": "Question regarding running time",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/181417",
  "author_name": "",
  "post_date": "2020-09-08T18:06:24.432155300Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>The requirement of the submission is described as followed.</p>\n<p>Submissions to this competition must be made through Notebooks. In order for the \"Submit to Competition\" button to be active after a commit, the following conditions must be met:</p>\n<pre><code>CPU Notebook &lt;= 9 hours run-time\nGPU Notebook &lt;= 4 hours run-time\n</code></pre>\n<p>So I am wondering if it will be OK if my notebook requires more than 9 hours running on the private test set. Thanks a lot for helping.</p>",
  "messages": [
    {
      "id": "1003186",
      "postDate": "09/08/2020 18:06:24",
      "content": "<p>The requirement of the submission is described as followed.</p>\n<p>Submissions to this competition must be made through Notebooks. In order for the \"Submit to Competition\" button to be active after a commit, the following conditions must be met:</p>\n<pre><code>CPU Notebook &lt;= 9 hours run-time\nGPU Notebook &lt;= 4 hours run-time\n</code></pre>\n<p>So I am wondering if it will be OK if my notebook requires more than 9 hours running on the private test set. Thanks a lot for helping.</p>",
      "rawMarkdown": "The requirement of the submission is described as followed.\n\nSubmissions to this competition must be made through Notebooks. In order for the \"Submit to Competition\" button to be active after a commit, the following conditions must be met:\n\n    CPU Notebook <= 9 hours run-time\n    GPU Notebook <= 4 hours run-time\n\nSo I am wondering if it will be OK if my notebook requires more than 9 hours running on the private test set. Thanks a lot for helping.",
      "votes": null
    },
    {
      "id": "1003214",
      "postDate": "09/08/2020 18:41:26",
      "content": "<p>You can train your model without limit and any ways. To reduce your prediction time, you can do the training in another notebook for example.  But for the prediction the notebook must be as required, i.e: if you didn't choose any accelerator (GPU or TPU), your notebook should run under 9h. If you selected the GPU accelerator, then it should run under 4h.</p>",
      "rawMarkdown": "You can train your model without limit and any ways. To reduce your prediction time, you can do the training in another notebook for example.  But for the prediction the notebook must be as required, i.e: if you didn't choose any accelerator (GPU or TPU), your notebook should run under 9h. If you selected the GPU accelerator, then it should run under 4h.",
      "votes": null
    },
    {
      "id": "1003229",
      "postDate": "09/08/2020 18:52:39",
      "content": "<p><a href=\"https://www.kaggle.com/yohannwattiez\" target=\"_blank\">@yohannwattiez</a> I understand, thanks a lot.</p>",
      "rawMarkdown": "yohannwattiez I understand, thanks a lot.",
      "votes": null
    },
    {
      "id": "1004982",
      "postDate": "09/10/2020 07:19:35",
      "content": "<p>Hi Yohann, </p>\n<p>if I do that (train my model in a different notebook) do I then just have to make my model/weights a dataset that I set as as input to my submission notebook?  Or is there another way please?</p>",
      "rawMarkdown": "Hi Yohann, \n\nif I do that (train my model in a different notebook) do I then just have to make my model/weights a dataset that I set as as input to my submission notebook?  Or is there another way please?",
      "votes": null
    },
    {
      "id": "1005059",
      "postDate": "09/10/2020 08:31:56",
      "content": "<p>You could indeed create a dataset with just your trained model and load it. It's also possible to add the output of a notebook as an input to another notebook. You will just need to save the trained model in your first notebook.</p>",
      "rawMarkdown": "You could indeed create a dataset with just your trained model and load it. It's also possible to add the output of a notebook as an input to another notebook. You will just need to save the trained model in your first notebook.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1003214,
      "author_name": "yohannwattiez",
      "author_url": "",
      "post_date": "09/08/2020 18:41:26",
      "content": "<p>You can train your model without limit and any ways. To reduce your prediction time, you can do the training in another notebook for example.  But for the prediction the notebook must be as required, i.e: if you didn't choose any accelerator (GPU or TPU), your notebook should run under 9h. If you selected the GPU accelerator, then it should run under 4h.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1004982,
          "author_name": "cascadenite",
          "author_url": "",
          "post_date": "09/10/2020 07:19:35",
          "content": "<p>Hi Yohann, </p>\n<p>if I do that (train my model in a different notebook) do I then just have to make my model/weights a dataset that I set as as input to my submission notebook?  Or is there another way please?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1005059,
          "author_name": "yohannwattiez",
          "author_url": "",
          "post_date": "09/10/2020 08:31:56",
          "content": "<p>You could indeed create a dataset with just your trained model and load it. It's also possible to add the output of a notebook as an input to another notebook. You will just need to save the trained model in your first notebook.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1003229,
      "author_name": "lanmaomao",
      "author_url": "",
      "post_date": "09/08/2020 18:52:39",
      "content": "<p><a href=\"https://www.kaggle.com/yohannwattiez\" target=\"_blank\">@yohannwattiez</a> I understand, thanks a lot.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1003186": "The requirement of the submission is described as followed.\n\nSubmissions to this competition must be made through Notebooks. In order for the \"Submit to Competition\" button to be active after a commit, the following conditions must be met:\n\n    CPU Notebook <= 9 hours run-time\n    GPU Notebook <= 4 hours run-time\n\nSo I am wondering if it will be OK if my notebook requires more than 9 hours running on the private test set. Thanks a lot for helping.",
    "1003214": "You can train your model without limit and any ways. To reduce your prediction time, you can do the training in another notebook for example.  But for the prediction the notebook must be as required, i.e: if you didn't choose any accelerator (GPU or TPU), your notebook should run under 9h. If you selected the GPU accelerator, then it should run under 4h.",
    "1003229": "yohannwattiez I understand, thanks a lot.",
    "1004982": "Hi Yohann, \n\nif I do that (train my model in a different notebook) do I then just have to make my model/weights a dataset that I set as as input to my submission notebook?  Or is there another way please?",
    "1005059": "You could indeed create a dataset with just your trained model and load it. It's also possible to add the output of a notebook as an input to another notebook. You will just need to save the trained model in your first notebook."
  },
  "source": "meta"
}