{
  "id": 235597,
  "title": "About inference time consuming",
  "url": "/competitions/hubmap-kidney-segmentation/discussion/235597",
  "author_name": "",
  "post_date": "2021-04-30T10:13:09.823869900Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Can somebody share how much time your notebook took for inference after submitting ?<br>\nIn my case, the heavy one(3 TTAs, 5 Fold model) took about 7 hours to get the final score. <br>\nI'm worried about if this heavy one would over the time limitation when running on the private dataset?<br>\nAnd could someone tell me the recommended inference time? Thanks a lot!!</p>",
  "messages": [
    {
      "id": "1288767",
      "postDate": "04/30/2021 10:13:09",
      "content": "<p>Can somebody share how much time your notebook took for inference after submitting ?<br>\nIn my case, the heavy one(3 TTAs, 5 Fold model) took about 7 hours to get the final score. <br>\nI'm worried about if this heavy one would over the time limitation when running on the private dataset?<br>\nAnd could someone tell me the recommended inference time? Thanks a lot!!</p>",
      "rawMarkdown": "Can somebody share how much time your notebook took for inference after submitting ?\nIn my case, the heavy one(3 TTAs, 5 Fold model) took about 7 hours to get the final score. \nI'm worried about if this heavy one would over the time limitation when running on the private dataset?\nAnd could someone tell me the recommended inference time? Thanks a lot!!",
      "votes": null
    },
    {
      "id": "1288771",
      "postDate": "04/30/2021 10:17:38",
      "content": "<p>My understanding is those 7 hours already include inference on the private dataset. That is, the private leaderboard results are currently \"known\" to Kaggle, but not yet released. I believe as long as your notebook completes and successfully posts a public leaderboard score, you are safe.</p>\n<p>The only competition I know in recent memory that hasn't run this way is the Deepfake Detection Challenge, where private test set inference was done offline after competition closing.</p>",
      "rawMarkdown": "My understanding is those 7 hours already include inference on the private dataset. That is, the private leaderboard results are currently \"known\" to Kaggle, but not yet released. I believe as long as your notebook completes and successfully posts a public leaderboard score, you are safe.\n\nThe only competition I know in recent memory that hasn't run this way is the Deepfake Detection Challenge, where private test set inference was done offline after competition closing.",
      "votes": null
    },
    {
      "id": "1288780",
      "postDate": "04/30/2021 10:26:08",
      "content": "<p>OK! I understand!! Thanks a lot!! Now I can choose the heavy one to submit for this endless competition. :)</p>",
      "rawMarkdown": "OK! I understand!! Thanks a lot!! Now I can choose the heavy one to submit for this endless competition. :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1288771,
      "author_name": "jamesphoward",
      "author_url": "",
      "post_date": "04/30/2021 10:17:38",
      "content": "<p>My understanding is those 7 hours already include inference on the private dataset. That is, the private leaderboard results are currently \"known\" to Kaggle, but not yet released. I believe as long as your notebook completes and successfully posts a public leaderboard score, you are safe.</p>\n<p>The only competition I know in recent memory that hasn't run this way is the Deepfake Detection Challenge, where private test set inference was done offline after competition closing.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1288780,
          "author_name": "shinewine",
          "author_url": "",
          "post_date": "04/30/2021 10:26:08",
          "content": "<p>OK! I understand!! Thanks a lot!! Now I can choose the heavy one to submit for this endless competition. :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1288767": "Can somebody share how much time your notebook took for inference after submitting ?\nIn my case, the heavy one(3 TTAs, 5 Fold model) took about 7 hours to get the final score. \nI'm worried about if this heavy one would over the time limitation when running on the private dataset?\nAnd could someone tell me the recommended inference time? Thanks a lot!!",
    "1288771": "My understanding is those 7 hours already include inference on the private dataset. That is, the private leaderboard results are currently \"known\" to Kaggle, but not yet released. I believe as long as your notebook completes and successfully posts a public leaderboard score, you are safe.\n\nThe only competition I know in recent memory that hasn't run this way is the Deepfake Detection Challenge, where private test set inference was done offline after competition closing.",
    "1288780": "OK! I understand!! Thanks a lot!! Now I can choose the heavy one to submit for this endless competition. :)"
  },
  "source": "meta"
}