{
  "id": 381505,
  "title": "Submission is taking more than 9h",
  "url": "/competitions/rsna-breast-cancer-detection/discussion/381505",
  "author_name": "Alfredo Maussa",
  "post_date": "2023-01-27T02:25:45.058000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>It takes around 1 second by image to read, I expect the 28% of the test set which presumably is around 8000 patients with 4 images each to process in total around 8960 images, which means 2.5 hours (2h30min), but my code is running more than 9h.</p>\n<p>I am not sure if is the 28% of 8000 patients for the public scoreboard or if the 28% is already the 8000. In this case, it would take 32000 images, 8.8 hours.</p>\n<p>However, I have tried to replicate the same amount of data before submission and it takes 21min for 10%x28%x8000x4 images (expecting 3.5h for 100%).</p>\n<hr>\n<p>Just scored my 15th submission which uses 10% from predicted labels and 90% random, it took around 2 hours. I assume the 100% would take 20 hours.</p>\n<hr>\n<p>Let me know too if the &lt;9h run time requirement is only for the \"submit\" code for public scoring, not for training or the private score (as it has more data it would take more time).</p>",
  "messages": [
    {
      "id": 2117107,
      "postDate": "2023-01-27T02:25:45.057Z",
      "content": "<p>It takes around 1 second by image to read, I expect the 28% of the test set which presumably is around 8000 patients with 4 images each to process in total around 8960 images, which means 2.5 hours (2h30min), but my code is running more than 9h.</p>\n<p>I am not sure if is the 28% of 8000 patients for the public scoreboard or if the 28% is already the 8000. In this case, it would take 32000 images, 8.8 hours.</p>\n<p>However, I have tried to replicate the same amount of data before submission and it takes 21min for 10%x28%x8000x4 images (expecting 3.5h for 100%).</p>\n<hr>\n<p>Just scored my 15th submission which uses 10% from predicted labels and 90% random, it took around 2 hours. I assume the 100% would take 20 hours.</p>\n<hr>\n<p>Let me know too if the &lt;9h run time requirement is only for the \"submit\" code for public scoring, not for training or the private score (as it has more data it would take more time).</p>",
      "rawMarkdown": "It takes around 1 second by image to read, I expect the 28% of the test set which presumably is around 8000 patients with 4 images each to process in total around 8960 images, which means 2.5 hours (2h30min), but my code is running more than 9h.\n\nI am not sure if is the 28% of 8000 patients for the public scoreboard or if the 28% is already the 8000. In this case, it would take 32000 images, 8.8 hours.\n\nHowever, I have tried to replicate the same amount of data before submission and it takes 21min for 10%x28%x8000x4 images (expecting 3.5h for 100%).\n\n---\n\nJust scored my 15th submission which uses 10% from predicted labels and 90% random, it took around 2 hours. I assume the 100% would take 20 hours.\n\n---\n\nLet me know too if the <9h run time requirement is only for the \"submit\" code for public scoring, not for training or the private score (as it has more data it would take more time).",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2117107": "It takes around 1 second by image to read, I expect the 28% of the test set which presumably is around 8000 patients with 4 images each to process in total around 8960 images, which means 2.5 hours (2h30min), but my code is running more than 9h.\n\nI am not sure if is the 28% of 8000 patients for the public scoreboard or if the 28% is already the 8000. In this case, it would take 32000 images, 8.8 hours.\n\nHowever, I have tried to replicate the same amount of data before submission and it takes 21min for 10%x28%x8000x4 images (expecting 3.5h for 100%).\n\n---\n\nJust scored my 15th submission which uses 10% from predicted labels and 90% random, it took around 2 hours. I assume the 100% would take 20 hours.\n\n---\n\nLet me know too if the <9h run time requirement is only for the \"submit\" code for public scoring, not for training or the private score (as it has more data it would take more time)."
  }
}