{
  "id": 103095,
  "title": "Pre trained data",
  "url": "/competitions/aptos2019-blindness-detection/discussion/103095",
  "author_name": "",
  "post_date": "2019-08-07T04:50:27.346966800Z",
  "votes": -3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I have seen some kernel. Do we need pre trained data</p>",
  "messages": [
    {
      "id": "593763",
      "postDate": "08/07/2019 04:50:27",
      "content": "<p>I have seen some kernel. Do we need pre trained data</p>",
      "rawMarkdown": "I have seen some kernel. Do we need pre trained data",
      "votes": null
    },
    {
      "id": "594450",
      "postDate": "08/08/2019 03:37:13",
      "content": "<p>If you have high computational resource then you can train it from scratch but it will be efficient  if you could freeze few weights and train the remaining layers</p>",
      "rawMarkdown": "If you have high computational resource then you can train it from scratch but it will be efficient  if you could freeze few weights and train the remaining layers",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 594450,
      "author_name": "nandhuelan",
      "author_url": "",
      "post_date": "08/08/2019 03:37:13",
      "content": "<p>If you have high computational resource then you can train it from scratch but it will be efficient  if you could freeze few weights and train the remaining layers</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "593763": "I have seen some kernel. Do we need pre trained data",
    "594450": "If you have high computational resource then you can train it from scratch but it will be efficient  if you could freeze few weights and train the remaining layers"
  },
  "source": "meta"
}