{
  "id": 294540,
  "title": "Train then Infer Notebook",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/294540",
  "author_name": "",
  "post_date": "2021-12-11T07:09:34.372282100Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I notice notebook public submissions that split the problem in two. A train notebook that accesses the Internet and an infer notebook that doesn't. Models like Yolov5 seem to need the Internet - so this appears to be one reason. Does the allowance for pre-trained models allow us to split the problem in half? Run and save output from a a train notebook with Internet - then using the saved files - run an infer notebook on the saved now pre-trained model in the 9 hour time limit w/o Internet?</p>",
  "messages": [
    {
      "id": "1614526",
      "postDate": "12/11/2021 07:09:34",
      "content": "<p>I notice notebook public submissions that split the problem in two. A train notebook that accesses the Internet and an infer notebook that doesn't. Models like Yolov5 seem to need the Internet - so this appears to be one reason. Does the allowance for pre-trained models allow us to split the problem in half? Run and save output from a a train notebook with Internet - then using the saved files - run an infer notebook on the saved now pre-trained model in the 9 hour time limit w/o Internet?</p>",
      "rawMarkdown": "I notice notebook public submissions that split the problem in two. A train notebook that accesses the Internet and an infer notebook that doesn't. Models like Yolov5 seem to need the Internet - so this appears to be one reason. Does the allowance for pre-trained models allow us to split the problem in half? Run and save output from a a train notebook with Internet - then using the saved files - run an infer notebook on the saved now pre-trained model in the 9 hour time limit w/o Internet?",
      "votes": null
    },
    {
      "id": "1614527",
      "postDate": "12/11/2021 07:12:20",
      "content": "<p>“Run and save output from a a train notebook with Internet - then using the saved files - run an infer notebook on the saved now pre-trained model in the 9 hour time limit w/o Internet?“<br>\nyes, you are right</p>",
      "rawMarkdown": "“Run and save output from a a train notebook with Internet - then using the saved files - run an infer notebook on the saved now pre-trained model in the 9 hour time limit w/o Internet?“\nyes, you are right",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1614527,
      "author_name": "businiao",
      "author_url": "",
      "post_date": "12/11/2021 07:12:20",
      "content": "<p>“Run and save output from a a train notebook with Internet - then using the saved files - run an infer notebook on the saved now pre-trained model in the 9 hour time limit w/o Internet?“<br>\nyes, you are right</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1614526": "I notice notebook public submissions that split the problem in two. A train notebook that accesses the Internet and an infer notebook that doesn't. Models like Yolov5 seem to need the Internet - so this appears to be one reason. Does the allowance for pre-trained models allow us to split the problem in half? Run and save output from a a train notebook with Internet - then using the saved files - run an infer notebook on the saved now pre-trained model in the 9 hour time limit w/o Internet?",
    "1614527": "“Run and save output from a a train notebook with Internet - then using the saved files - run an infer notebook on the saved now pre-trained model in the 9 hour time limit w/o Internet?“\nyes, you are right"
  },
  "source": "meta"
}