{
  "id": 303380,
  "title": "About the time limit for the evaluation period",
  "url": "/competitions/g-research-crypto-forecasting/discussion/303380",
  "author_name": "",
  "post_date": "2022-01-27T09:30:52.887655400Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Is this assumption correct: </p>\n<p>Since both the public and the evaluation period handles 3 months of data, a kernel which successfully manage the time limit for the public test set will also manage the total/final evaluation period without any problem. </p>\n<p>Or is more to it, to take account for?</p>",
  "messages": [
    {
      "id": "1665978",
      "postDate": "01/27/2022 09:30:52",
      "content": "<p>Is this assumption correct: </p>\n<p>Since both the public and the evaluation period handles 3 months of data, a kernel which successfully manage the time limit for the public test set will also manage the total/final evaluation period without any problem. </p>\n<p>Or is more to it, to take account for?</p>",
      "rawMarkdown": "Is this assumption correct: \n\nSince both the public and the evaluation period handles 3 months of data, a kernel which successfully manage the time limit for the public test set will also manage the total/final evaluation period without any problem. \n\nOr is more to it, to take account for?",
      "votes": null
    },
    {
      "id": "1666611",
      "postDate": "01/27/2022 23:40:26",
      "content": "<p>In principle, yes. One thing to be aware of is that there will be an extra week of data in the supplemental train (possible marginal effect on compute time). I have also noticed that different compute nodes can have slightly different iteration speeds (even on the same code and container image) so I would leave a bit of a buffer.</p>",
      "rawMarkdown": "In principle, yes. One thing to be aware of is that there will be an extra week of data in the supplemental train (possible marginal effect on compute time). I have also noticed that different compute nodes can have slightly different iteration speeds (even on the same code and container image) so I would leave a bit of a buffer.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1666611,
      "author_name": "jagofc",
      "author_url": "",
      "post_date": "01/27/2022 23:40:26",
      "content": "<p>In principle, yes. One thing to be aware of is that there will be an extra week of data in the supplemental train (possible marginal effect on compute time). I have also noticed that different compute nodes can have slightly different iteration speeds (even on the same code and container image) so I would leave a bit of a buffer.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1665978": "Is this assumption correct: \n\nSince both the public and the evaluation period handles 3 months of data, a kernel which successfully manage the time limit for the public test set will also manage the total/final evaluation period without any problem. \n\nOr is more to it, to take account for?",
    "1666611": "In principle, yes. One thing to be aware of is that there will be an extra week of data in the supplemental train (possible marginal effect on compute time). I have also noticed that different compute nodes can have slightly different iteration speeds (even on the same code and container image) so I would leave a bit of a buffer."
  },
  "source": "meta"
}