{
  "id": 109765,
  "title": "What's your source of compute?",
  "url": "/competitions/recursion-cellular-image-classification/discussion/109765",
  "author_name": "",
  "post_date": "2019-09-22T04:16:30.058308600Z",
  "votes": 3,
  "comment_count": 1,
  "views": 0,
  "content": "<p>This is inspired by the other best Kaggle kernel score thread. What do you use for this competition? Local GPUs? Cloud GPUs? TPUs? </p>\n\n<p>We have been running 2x 2080 Tis since we entered the competition, and in the last couple of weeks added two 1080 Tis. My heart says I need moar GPUs, but my wallet says no...</p>",
  "messages": [
    {
      "id": "631460",
      "postDate": "09/22/2019 04:16:30",
      "content": "<p>This is inspired by the other best Kaggle kernel score thread. What do you use for this competition? Local GPUs? Cloud GPUs? TPUs? </p>\n\n<p>We have been running 2x 2080 Tis since we entered the competition, and in the last couple of weeks added two 1080 Tis. My heart says I need moar GPUs, but my wallet says no...</p>",
      "rawMarkdown": "This is inspired by the other best Kaggle kernel score thread. What do you use for this competition? Local GPUs? Cloud GPUs? TPUs? \n\nWe have been running 2x 2080 Tis since we entered the competition, and in the last couple of weeks added two 1080 Tis. My heart says I need moar GPUs, but my wallet says no...",
      "votes": null
    },
    {
      "id": "633351",
      "postDate": "09/24/2019 19:43:57",
      "content": "<p>Cloud GPU with 1 1080Ti. And I don't really feel like I need to train more fat models, so 1 Ti is kind of enough for experimentation.</p>",
      "rawMarkdown": "Cloud GPU with 1 1080Ti. And I don't really feel like I need to train more fat models, so 1 Ti is kind of enough for experimentation.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 633351,
      "author_name": "devvindan",
      "author_url": "",
      "post_date": "09/24/2019 19:43:57",
      "content": "<p>Cloud GPU with 1 1080Ti. And I don't really feel like I need to train more fat models, so 1 Ti is kind of enough for experimentation.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "631460": "This is inspired by the other best Kaggle kernel score thread. What do you use for this competition? Local GPUs? Cloud GPUs? TPUs? \n\nWe have been running 2x 2080 Tis since we entered the competition, and in the last couple of weeks added two 1080 Tis. My heart says I need moar GPUs, but my wallet says no...",
    "633351": "Cloud GPU with 1 1080Ti. And I don't really feel like I need to train more fat models, so 1 Ti is kind of enough for experimentation."
  },
  "source": "meta"
}