{
  "id": 201397,
  "title": "Out of Time Submission when Doing Ensemble...",
  "url": "/competitions/hubmap-kidney-segmentation/discussion/201397",
  "author_name": "",
  "post_date": "2020-12-04T18:27:16.293441500Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>What thecniques are you using in order to reduce execution time?</p>",
  "messages": [
    {
      "id": "1102264",
      "postDate": "12/04/2020 18:27:16",
      "content": "<p>What thecniques are you using in order to reduce execution time?</p>",
      "rawMarkdown": "What thecniques are you using in order to reduce execution time?",
      "votes": null
    },
    {
      "id": "1102303",
      "postDate": "12/04/2020 19:14:02",
      "content": "<p><a href=\"https://www.kaggle.com/marcelosanchezortega\" target=\"_blank\">@marcelosanchezortega</a>,</p>\n<p>You can see some ideas in my notebook <a href=\"https://www.kaggle.com/marcosnovaes/hubmap-memory-efficient-submission-using-disk\" target=\"_blank\">https://www.kaggle.com/marcosnovaes/hubmap-memory-efficient-submission-using-disk</a></p>\n<p>Notice that I am using the disk space to avoid running out of memory. To speed up compute time, I think the trick is to make prediction in batches and not one tile at a time. In my submission I use a tf.Data.Dataset and I ask it to process 64 tiles at a time. Hope that helps! Cheers!</p>",
      "rawMarkdown": "marcelosanchezortega,\n\nYou can see some ideas in my notebook [https://www.kaggle.com/marcosnovaes/hubmap-memory-efficient-submission-using-disk](https://www.kaggle.com/marcosnovaes/hubmap-memory-efficient-submission-using-disk)\n\nNotice that I am using the disk space to avoid running out of memory. To speed up compute time, I think the trick is to make prediction in batches and not one tile at a time. In my submission I use a tf.Data.Dataset and I ask it to process 64 tiles at a time. Hope that helps! Cheers!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1102303,
      "author_name": "marcosnovaes",
      "author_url": "",
      "post_date": "12/04/2020 19:14:02",
      "content": "<p><a href=\"https://www.kaggle.com/marcelosanchezortega\" target=\"_blank\">@marcelosanchezortega</a>,</p>\n<p>You can see some ideas in my notebook <a href=\"https://www.kaggle.com/marcosnovaes/hubmap-memory-efficient-submission-using-disk\" target=\"_blank\">https://www.kaggle.com/marcosnovaes/hubmap-memory-efficient-submission-using-disk</a></p>\n<p>Notice that I am using the disk space to avoid running out of memory. To speed up compute time, I think the trick is to make prediction in batches and not one tile at a time. In my submission I use a tf.Data.Dataset and I ask it to process 64 tiles at a time. Hope that helps! Cheers!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1102264": "What thecniques are you using in order to reduce execution time?",
    "1102303": "marcelosanchezortega,\n\nYou can see some ideas in my notebook [https://www.kaggle.com/marcosnovaes/hubmap-memory-efficient-submission-using-disk](https://www.kaggle.com/marcosnovaes/hubmap-memory-efficient-submission-using-disk)\n\nNotice that I am using the disk space to avoid running out of memory. To speed up compute time, I think the trick is to make prediction in batches and not one tile at a time. In my submission I use a tf.Data.Dataset and I ask it to process 64 tiles at a time. Hope that helps! Cheers!"
  },
  "source": "meta"
}