{
  "id": 301694,
  "title": "FFCV - a library to train fast! ⚡",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/301694",
  "author_name": "",
  "post_date": "2022-01-19T02:49:32.246315Z",
  "votes": 14,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I haven't seen this around, but I thought people in this competition might be interested in checking out Fast Forward Computer Vision (FFCV).  They are posting some ridiculously fast times! ⚡ </p>\n<p>I don't know if it works in Kaggle notebooks, but it could be useful to quickly train outside of Kaggle! Training goes so fast on A100s that it is relatively cheap.</p>\n<p>Website: <a href=\"https://ffcv.io/\" target=\"_blank\">https://ffcv.io/</a><br>\nGitHub: <a href=\"https://github.com/libffcv/ffcv\" target=\"_blank\">https://github.com/libffcv/ffcv</a></p>\n<p><img src=\"https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv1.png\" alt=\"\"><br>\n<img src=\"https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv2.png\" alt=\"\"><br>\n<img src=\"https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv3.png\" alt=\"\"><br>\n<img src=\"https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv4.png\" alt=\"\"></p>",
  "messages": [
    {
      "id": "1655986",
      "postDate": "01/19/2022 02:49:32",
      "content": "<p>I haven't seen this around, but I thought people in this competition might be interested in checking out Fast Forward Computer Vision (FFCV).  They are posting some ridiculously fast times! ⚡ </p>\n<p>I don't know if it works in Kaggle notebooks, but it could be useful to quickly train outside of Kaggle! Training goes so fast on A100s that it is relatively cheap.</p>\n<p>Website: <a href=\"https://ffcv.io/\" target=\"_blank\">https://ffcv.io/</a><br>\nGitHub: <a href=\"https://github.com/libffcv/ffcv\" target=\"_blank\">https://github.com/libffcv/ffcv</a></p>\n<p><img src=\"https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv1.png\" alt=\"\"><br>\n<img src=\"https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv2.png\" alt=\"\"><br>\n<img src=\"https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv3.png\" alt=\"\"><br>\n<img src=\"https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv4.png\" alt=\"\"></p>",
      "rawMarkdown": "I haven't seen this around, but I thought people in this competition might be interested in checking out Fast Forward Computer Vision (FFCV).  They are posting some ridiculously fast times! ⚡ \n\nI don't know if it works in Kaggle notebooks, but it could be useful to quickly train outside of Kaggle! Training goes so fast on A100s that it is relatively cheap.\n\nWebsite: https://ffcv.io/\nGitHub: https://github.com/libffcv/ffcv\n\n![](https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv1.png)\n![](https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv2.png)\n![](https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv3.png)\n![](https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv4.png)",
      "votes": null
    },
    {
      "id": "1660026",
      "postDate": "01/22/2022 10:24:53",
      "content": "<p>How to use that? Could you make a notebook for it~</p>",
      "rawMarkdown": "How to use that? Could you make a notebook for it~",
      "votes": null
    },
    {
      "id": "1661360",
      "postDate": "01/23/2022 12:18:22",
      "content": "<p>tbh I don’t know if I have time. I’ll see what I can do!</p>",
      "rawMarkdown": "tbh I don’t know if I have time. I’ll see what I can do!",
      "votes": null
    },
    {
      "id": "1665259",
      "postDate": "01/26/2022 17:32:45",
      "content": "<p>i will try it <a href=\"https://www.kaggle.com/calvchen\" target=\"_blank\">@calvchen</a> </p>",
      "rawMarkdown": "i will try it @calvchen",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1660026,
      "author_name": "calvchen",
      "author_url": "",
      "post_date": "01/22/2022 10:24:53",
      "content": "<p>How to use that? Could you make a notebook for it~</p>",
      "votes": null,
      "replies": [
        {
          "id": 1661360,
          "author_name": "nbroad",
          "author_url": "",
          "post_date": "01/23/2022 12:18:22",
          "content": "<p>tbh I don’t know if I have time. I’ll see what I can do!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1665259,
          "author_name": "bibhabasumohapatra",
          "author_url": "",
          "post_date": "01/26/2022 17:32:45",
          "content": "<p>i will try it <a href=\"https://www.kaggle.com/calvchen\" target=\"_blank\">@calvchen</a> </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1655986": "I haven't seen this around, but I thought people in this competition might be interested in checking out Fast Forward Computer Vision (FFCV).  They are posting some ridiculously fast times! ⚡ \n\nI don't know if it works in Kaggle notebooks, but it could be useful to quickly train outside of Kaggle! Training goes so fast on A100s that it is relatively cheap.\n\nWebsite: https://ffcv.io/\nGitHub: https://github.com/libffcv/ffcv\n\n![](https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv1.png)\n![](https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv2.png)\n![](https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv3.png)\n![](https://raw.githubusercontent.com/nbroad1881/kaggle-images/main/tensorflow_great_barrier_reef/ffcv4.png)",
    "1660026": "How to use that? Could you make a notebook for it~",
    "1661360": "tbh I don’t know if I have time. I’ll see what I can do!",
    "1665259": "i will try it @calvchen"
  },
  "source": "meta"
}