{
  "id": 670979,
  "title": "Feel Free to Use This To Write Your Suggestions",
  "url": "/competitions/vesuvius-challenge-surface-detection/discussion/670979",
  "author_name": "",
  "post_date": "2026-01-30T16:08:44.015354100Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi, wanted to create a topic for everyone to share how you approach competitions in general.\nI saw some really nice posts where people even linked VSCode with Kaggle, it's nice.</p>\n<p>Most people say that both the dataset and processing it takes a lot of time, especially in big data like this one. So I wanna know would buying Colab pro is enough? Do you download datasets to your local and train locally? Do people buy better hardware just to compete in these competitions?</p>\n<p>Would love to learn from you! Thanks in advance and all the best ⭐️</p>",
  "messages": [
    {
      "id": "3399316",
      "postDate": "01/30/2026 16:08:44",
      "content": "<p>Hi, wanted to create a topic for everyone to share how you approach competitions in general.\nI saw some really nice posts where people even linked VSCode with Kaggle, it's nice.</p>\n<p>Most people say that both the dataset and processing it takes a lot of time, especially in big data like this one. So I wanna know would buying Colab pro is enough? Do you download datasets to your local and train locally? Do people buy better hardware just to compete in these competitions?</p>\n<p>Would love to learn from you! Thanks in advance and all the best ⭐️</p>",
      "rawMarkdown": "Hi, wanted to create a topic for everyone to share how you approach competitions in general.\nI saw some really nice posts where people even linked VSCode with Kaggle, it's nice.\n\nMost people say that both the dataset and processing it takes a lot of time, especially in big data like this one. So I wanna know would buying Colab pro is enough? Do you download datasets to your local and train locally? Do people buy better hardware just to compete in these competitions?\n\nWould love to learn from you! Thanks in advance and all the best ⭐️",
      "votes": null
    },
    {
      "id": "3399384",
      "postDate": "01/30/2026 19:14:19",
      "content": "<p>Kaggle hardware is enough tbh, you can train models effectively, and then post and preprocessaing strategies are to be tested on cpu anyways, I do all training on notebooks on kaggle only, and it's sufficient. Obviously you can't compare the number of iterations to someone with a good pc, but you certainly can have fun and explore ideas just with kaggle.</p>",
      "rawMarkdown": "Kaggle hardware is enough tbh, you can train models effectively, and then post and preprocessaing strategies are to be tested on cpu anyways, I do all training on notebooks on kaggle only, and it's sufficient. Obviously you can't compare the number of iterations to someone with a good pc, but you certainly can have fun and explore ideas just with kaggle.",
      "votes": null
    },
    {
      "id": "3399419",
      "postDate": "01/30/2026 20:28:43",
      "content": "<p>That's incredible! Can I know the training time for the model you're currently using, if it's not a secret?</p>",
      "rawMarkdown": "That's incredible! Can I know the training time for the model you're currently using, if it's not a secret?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3399384,
      "author_name": "choudharymanas",
      "author_url": "",
      "post_date": "01/30/2026 19:14:19",
      "content": "<p>Kaggle hardware is enough tbh, you can train models effectively, and then post and preprocessaing strategies are to be tested on cpu anyways, I do all training on notebooks on kaggle only, and it's sufficient. Obviously you can't compare the number of iterations to someone with a good pc, but you certainly can have fun and explore ideas just with kaggle.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3399419,
          "author_name": "lllleeeo",
          "author_url": "",
          "post_date": "01/30/2026 20:28:43",
          "content": "<p>That's incredible! Can I know the training time for the model you're currently using, if it's not a secret?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3399316": "Hi, wanted to create a topic for everyone to share how you approach competitions in general.\nI saw some really nice posts where people even linked VSCode with Kaggle, it's nice.\n\nMost people say that both the dataset and processing it takes a lot of time, especially in big data like this one. So I wanna know would buying Colab pro is enough? Do you download datasets to your local and train locally? Do people buy better hardware just to compete in these competitions?\n\nWould love to learn from you! Thanks in advance and all the best ⭐️",
    "3399384": "Kaggle hardware is enough tbh, you can train models effectively, and then post and preprocessaing strategies are to be tested on cpu anyways, I do all training on notebooks on kaggle only, and it's sufficient. Obviously you can't compare the number of iterations to someone with a good pc, but you certainly can have fun and explore ideas just with kaggle.",
    "3399419": "That's incredible! Can I know the training time for the model you're currently using, if it's not a secret?"
  },
  "source": "meta"
}