{
  "id": 126729,
  "title": "Submission Takes Longer Than Commit?",
  "url": "/competitions/deepfake-detection-challenge/discussion/126729",
  "author_name": "",
  "post_date": "2020-01-19T18:43:20.075362300Z",
  "votes": null,
  "comment_count": 5,
  "views": 0,
  "content": "<p>My commit took 3.5 hours to process, but my submission is taking &gt; 7 hours. Is the submission dataset bigger than the sample test dataset?</p>",
  "messages": [
    {
      "id": "723258",
      "postDate": "01/19/2020 18:43:20",
      "content": "<p>My commit took 3.5 hours to process, but my submission is taking &gt; 7 hours. Is the submission dataset bigger than the sample test dataset?</p>",
      "rawMarkdown": "My commit took 3.5 hours to process, but my submission is taking &gt; 7 hours. Is the submission dataset bigger than the sample test dataset?",
      "votes": null
    },
    {
      "id": "723262",
      "postDate": "01/19/2020 18:55:31",
      "content": "<p>I just read: <a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/122280\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/122280</a> which states the test set is 10x the size? Do we have 10x the time for the difference or still limited to 9 hours?</p>",
      "rawMarkdown": "I just read: https://www.kaggle.com/c/deepfake-detection-challenge/discussion/122280 which states the test set is 10x the size? Do we have 10x the time for the difference or still limited to 9 hours?",
      "votes": null
    },
    {
      "id": "723265",
      "postDate": "01/19/2020 19:02:26",
      "content": "<p>Validation size is 400, public test is 4000. The limit is still 9h. </p>",
      "rawMarkdown": "Validation size is 400, public test is 4000. The limit is still 9h.",
      "votes": null
    },
    {
      "id": "723334",
      "postDate": "01/19/2020 21:59:05",
      "content": "<p>You have 8 seconds on average per video (9 hours divided by 4000 videos). I added a little timing loop to my kernel that runs on 10 videos and then computes the average time per video. If it's more than 8 seconds, there is no point in submitting this kernel because it will be too slow.</p>",
      "rawMarkdown": "You have 8 seconds on average per video (9 hours divided by 4000 videos). I added a little timing loop to my kernel that runs on 10 videos and then computes the average time per video. If it's more than 8 seconds, there is no point in submitting this kernel because it will be too slow.",
      "votes": null
    },
    {
      "id": "724113",
      "postDate": "01/20/2020 20:18:05",
      "content": "<p>I enabled GPU on the notebook and OpenCV works so much faster :) From 3.5 hours to 25 minutes for the 400 which should translate to 250 minutes = 4.2 hours :)</p>",
      "rawMarkdown": "I enabled GPU on the notebook and OpenCV works so much faster :) From 3.5 hours to 25 minutes for the 400 which should translate to 250 minutes = 4.2 hours :)",
      "votes": null
    },
    {
      "id": "743047",
      "postDate": "02/11/2020 18:37:41",
      "content": "<p>sorry, never mind.</p>",
      "rawMarkdown": "sorry, never mind.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 723262,
      "author_name": "sethkitchen",
      "author_url": "",
      "post_date": "01/19/2020 18:55:31",
      "content": "<p>I just read: <a href=\"https://www.kaggle.com/c/deepfake-detection-challenge/discussion/122280\">https://www.kaggle.com/c/deepfake-detection-challenge/discussion/122280</a> which states the test set is 10x the size? Do we have 10x the time for the difference or still limited to 9 hours?</p>",
      "votes": null,
      "replies": [
        {
          "id": 723265,
          "author_name": "zaharch",
          "author_url": "",
          "post_date": "01/19/2020 19:02:26",
          "content": "<p>Validation size is 400, public test is 4000. The limit is still 9h. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 723334,
      "author_name": "humananalog",
      "author_url": "",
      "post_date": "01/19/2020 21:59:05",
      "content": "<p>You have 8 seconds on average per video (9 hours divided by 4000 videos). I added a little timing loop to my kernel that runs on 10 videos and then computes the average time per video. If it's more than 8 seconds, there is no point in submitting this kernel because it will be too slow.</p>",
      "votes": null,
      "replies": [
        {
          "id": 743047,
          "author_name": "leonidboytsov",
          "author_url": "",
          "post_date": "02/11/2020 18:37:41",
          "content": "<p>sorry, never mind.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 724113,
      "author_name": "sethkitchen",
      "author_url": "",
      "post_date": "01/20/2020 20:18:05",
      "content": "<p>I enabled GPU on the notebook and OpenCV works so much faster :) From 3.5 hours to 25 minutes for the 400 which should translate to 250 minutes = 4.2 hours :)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "723258": "My commit took 3.5 hours to process, but my submission is taking &gt; 7 hours. Is the submission dataset bigger than the sample test dataset?",
    "723262": "I just read: https://www.kaggle.com/c/deepfake-detection-challenge/discussion/122280 which states the test set is 10x the size? Do we have 10x the time for the difference or still limited to 9 hours?",
    "723265": "Validation size is 400, public test is 4000. The limit is still 9h.",
    "723334": "You have 8 seconds on average per video (9 hours divided by 4000 videos). I added a little timing loop to my kernel that runs on 10 videos and then computes the average time per video. If it's more than 8 seconds, there is no point in submitting this kernel because it will be too slow.",
    "724113": "I enabled GPU on the notebook and OpenCV works so much faster :) From 3.5 hours to 25 minutes for the 400 which should translate to 250 minutes = 4.2 hours :)",
    "743047": "sorry, never mind."
  },
  "source": "meta"
}