{
  "id": 381372,
  "title": "DL projects: cloud or physical GPUs?",
  "url": "/competitions/rsna-breast-cancer-detection/discussion/381372",
  "author_name": "rbrgAlou",
  "post_date": "2023-01-26T12:42:54.623000",
  "votes": 0,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I would like to know how you run your Deep Learning projects: cloud GPUs or physical GPUs? Which one? </p>",
  "messages": [
    {
      "id": 2116484,
      "postDate": "2023-01-26T14:30:30.177Z",
      "content": "<p>For me both.</p>\n<ul>\n<li>local for fast iteration - I can create many datasets (no upload time), flexibility, always ready enviroment, run 24h/day (no limits)</li>\n<li>cloud with more resources - for final model training (milestones model training) or occasionally when problem needs such setup (e.g. large dataset, large model, large bs etc.)</li>\n</ul>",
      "rawMarkdown": "For me both.\n- local for fast iteration - I can create many datasets (no upload time), flexibility, always ready enviroment, run 24h/day (no limits)\n- cloud with more resources - for final model training (milestones model training) or occasionally when problem needs such setup (e.g. large dataset, large model, large bs etc.)",
      "votes": 3,
      "replies": [
        {
          "id": 2116967,
          "postDate": "2023-01-26T21:00:43.457Z",
          "content": "<p>Ok thx, which physical GPU do you have and which cloud service do you use ? How much does it costs to you to train you r model for this competition for instance ? </p>",
          "rawMarkdown": "Ok thx, which physical GPU do you have and which cloud service do you use ? How much does it costs to you to train you r model for this competition for instance ? ",
          "replies": [
            {
              "id": 2116974,
              "postDate": "2023-01-26T21:04:14.533Z",
              "content": "<p>Cloud:</p>\n<ul>\n<li>Paperspace Gradio - mostly free but during this competition I had to pay for hours (in total 2 months - ~200 USD)</li>\n<li>sometimes Colab Pro+ (not for this competition)</li>\n</ul>\n<p>Laptop:</p>\n<ul>\n<li>Dell Allienware R17 - RTX3080 / 64GB RAM </li>\n</ul>\n<p>PC:</p>\n<ul>\n<li>Z by HP 1xA6000</li>\n</ul>",
              "rawMarkdown": "Cloud:\n- Paperspace Gradio - mostly free but during this competition I had to pay for hours (in total 2 months - ~200 USD)\n- sometimes Colab Pro+ (not for this competition)\n\nLaptop:\n- Dell Allienware R17 - RTX3080 / 64GB RAM \n\nPC:\n- Z by HP 1xA6000",
              "votes": 1
            }
          ]
        }
      ]
    },
    {
      "id": 2116353,
      "postDate": "2023-01-26T12:42:54.623Z",
      "content": "<p>I would like to know how you run your Deep Learning projects: cloud GPUs or physical GPUs? Which one? </p>",
      "rawMarkdown": "I would like to know how you run your Deep Learning projects: cloud GPUs or physical GPUs? Which one? "
    }
  ],
  "comments": [
    {
      "id": 2116484,
      "author_name": "Remek Kinas",
      "author_url": "",
      "post_date": "2023-01-26T14:30:30.177000",
      "content": "<p>For me both.</p>\n<ul>\n<li>local for fast iteration - I can create many datasets (no upload time), flexibility, always ready enviroment, run 24h/day (no limits)</li>\n<li>cloud with more resources - for final model training (milestones model training) or occasionally when problem needs such setup (e.g. large dataset, large model, large bs etc.)</li>\n</ul>",
      "votes": 3,
      "replies": [
        {
          "id": 2116967,
          "author_name": "rbrgAlou",
          "author_url": "",
          "post_date": "2023-01-26T21:00:43.457000",
          "content": "<p>Ok thx, which physical GPU do you have and which cloud service do you use ? How much does it costs to you to train you r model for this competition for instance ? </p>",
          "votes": 0,
          "replies": [
            {
              "id": 2116974,
              "author_name": "Remek Kinas",
              "author_url": "",
              "post_date": "2023-01-26T21:04:14.533000",
              "content": "<p>Cloud:</p>\n<ul>\n<li>Paperspace Gradio - mostly free but during this competition I had to pay for hours (in total 2 months - ~200 USD)</li>\n<li>sometimes Colab Pro+ (not for this competition)</li>\n</ul>\n<p>Laptop:</p>\n<ul>\n<li>Dell Allienware R17 - RTX3080 / 64GB RAM </li>\n</ul>\n<p>PC:</p>\n<ul>\n<li>Z by HP 1xA6000</li>\n</ul>",
              "votes": 1,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2116484": "For me both.\n- local for fast iteration - I can create many datasets (no upload time), flexibility, always ready enviroment, run 24h/day (no limits)\n- cloud with more resources - for final model training (milestones model training) or occasionally when problem needs such setup (e.g. large dataset, large model, large bs etc.)",
    "2116353": "I would like to know how you run your Deep Learning projects: cloud GPUs or physical GPUs? Which one? "
  }
}