{
  "id": 414865,
  "title": "Win without using external GPUs?",
  "url": "/competitions/birdclef-2023/discussion/414865",
  "author_name": "",
  "post_date": "2023-06-03T17:43:06.553705100Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Has anyone among the Top 20 winners not used any external GPUs, besides Kaggle GPUs?.</p>\n<p>Thank you in advance!.</p>",
  "messages": [
    {
      "id": "2286729",
      "postDate": "06/03/2023 17:43:06",
      "content": "<p>Has anyone among the Top 20 winners not used any external GPUs, besides Kaggle GPUs?.</p>\n<p>Thank you in advance!.</p>",
      "rawMarkdown": "Has anyone among the Top 20 winners not used any external GPUs, besides Kaggle GPUs?.\n\nThank you in advance!.",
      "votes": null
    },
    {
      "id": "2286785",
      "postDate": "06/03/2023 19:02:23",
      "content": "<p>I think most top Kaggle users have baroque PCs and good hardware/ are sponsored by leading companies. I am not sure if Kaggle resources are sufficient to provide this level of success <a href=\"https://www.kaggle.com/pablolarrosa\" target=\"_blank\">@pablolarrosa</a> </p>",
      "rawMarkdown": "I think most top Kaggle users have baroque PCs and good hardware/ are sponsored by leading companies. I am not sure if Kaggle resources are sufficient to provide this level of success @pablolarrosa",
      "votes": null
    },
    {
      "id": "2286800",
      "postDate": "06/03/2023 19:21:38",
      "content": "<p>I think like you, kfold and augmentation require a lot of processing.</p>\n<p>Thank for your answer.</p>",
      "rawMarkdown": "I think like you, kfold and augmentation require a lot of processing.\n\nThank for your answer.",
      "votes": null
    },
    {
      "id": "2286806",
      "postDate": "06/03/2023 19:37:33",
      "content": "<p><a href=\"https://www.kaggle.com/pablolarrosa\" target=\"_blank\">@pablolarrosa</a>, a local deep learning PC is perhaps a good option to do well at most competitions on Kaggle. Deep learning based challenges need lots of relevant hardware. Participating in teams is a good way to circumnavigate this to an extent. Colab pro+/ paper-space is a sub-optimal but a decent solution </p>",
      "rawMarkdown": "pablolarrosa, a local deep learning PC is perhaps a good option to do well at most competitions on Kaggle. Deep learning based challenges need lots of relevant hardware. Participating in teams is a good way to circumnavigate this to an extent. Colab pro+/ paper-space is a sub-optimal but a decent solution",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2286785,
      "author_name": "ravi20076",
      "author_url": "",
      "post_date": "06/03/2023 19:02:23",
      "content": "<p>I think most top Kaggle users have baroque PCs and good hardware/ are sponsored by leading companies. I am not sure if Kaggle resources are sufficient to provide this level of success <a href=\"https://www.kaggle.com/pablolarrosa\" target=\"_blank\">@pablolarrosa</a> </p>",
      "votes": null,
      "replies": [
        {
          "id": 2286800,
          "author_name": "pablolarrosa",
          "author_url": "",
          "post_date": "06/03/2023 19:21:38",
          "content": "<p>I think like you, kfold and augmentation require a lot of processing.</p>\n<p>Thank for your answer.</p>",
          "votes": null,
          "replies": [
            {
              "id": 2286806,
              "author_name": "ravi20076",
              "author_url": "",
              "post_date": "06/03/2023 19:37:33",
              "content": "<p><a href=\"https://www.kaggle.com/pablolarrosa\" target=\"_blank\">@pablolarrosa</a>, a local deep learning PC is perhaps a good option to do well at most competitions on Kaggle. Deep learning based challenges need lots of relevant hardware. Participating in teams is a good way to circumnavigate this to an extent. Colab pro+/ paper-space is a sub-optimal but a decent solution </p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2286729": "Has anyone among the Top 20 winners not used any external GPUs, besides Kaggle GPUs?.\n\nThank you in advance!.",
    "2286785": "I think most top Kaggle users have baroque PCs and good hardware/ are sponsored by leading companies. I am not sure if Kaggle resources are sufficient to provide this level of success @pablolarrosa",
    "2286800": "I think like you, kfold and augmentation require a lot of processing.\n\nThank for your answer.",
    "2286806": "pablolarrosa, a local deep learning PC is perhaps a good option to do well at most competitions on Kaggle. Deep learning based challenges need lots of relevant hardware. Participating in teams is a good way to circumnavigate this to an extent. Colab pro+/ paper-space is a sub-optimal but a decent solution"
  },
  "source": "meta"
}