{
  "id": 679641,
  "title": "GPU for Local Training",
  "url": "/competitions/deep-past-initiative-machine-translation/discussion/679641",
  "author_name": "",
  "post_date": "2026-03-02T19:21:16.996117300Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello, I have a question I'm curious about. Since I usually participate in playgrounds or competitions held in my country, I haven't had much opportunity to use GPUs other than the one Kaggle provides us. I also want to participate in global competitions, but while the Kaggle GPU is sufficient for inference, I'm not sure if Kaggle GPUs are enough for training. Or generally, I'm not sure if Kaggle is used for training in global competitions. If the Kaggle GPU is not enough, could you suggest a good GPU that I would need to participate in competitions?</p>",
  "messages": [
    {
      "id": "3416413",
      "postDate": "03/02/2026 19:21:16",
      "content": "<p>Hello, I have a question I'm curious about. Since I usually participate in playgrounds or competitions held in my country, I haven't had much opportunity to use GPUs other than the one Kaggle provides us. I also want to participate in global competitions, but while the Kaggle GPU is sufficient for inference, I'm not sure if Kaggle GPUs are enough for training. Or generally, I'm not sure if Kaggle is used for training in global competitions. If the Kaggle GPU is not enough, could you suggest a good GPU that I would need to participate in competitions?</p>",
      "rawMarkdown": "Hello, I have a question I'm curious about. Since I usually participate in playgrounds or competitions held in my country, I haven't had much opportunity to use GPUs other than the one Kaggle provides us. I also want to participate in global competitions, but while the Kaggle GPU is sufficient for inference, I'm not sure if Kaggle GPUs are enough for training. Or generally, I'm not sure if Kaggle is used for training in global competitions. If the Kaggle GPU is not enough, could you suggest a good GPU that I would need to participate in competitions?",
      "votes": null
    },
    {
      "id": "3416524",
      "postDate": "03/03/2026 05:32:50",
      "content": "<p>For this competition, Kaggle is sufficient for both training and inference—you can refer to the public notebooks for reference. If you plan to use a rental GPU, you can choose one based on your requirements. This mainly depends on the competition dataset and the batch size you intend to use. In most cases, an RTX 3090 or RTX 4090 with 24 GB of GPU memory is more than adequate.</p>",
      "rawMarkdown": "For this competition, Kaggle is sufficient for both training and inference—you can refer to the public notebooks for reference. If you plan to use a rental GPU, you can choose one based on your requirements. This mainly depends on the competition dataset and the batch size you intend to use. In most cases, an RTX 3090 or RTX 4090 with 24 GB of GPU memory is more than adequate.",
      "votes": null
    },
    {
      "id": "3416540",
      "postDate": "03/03/2026 06:34:25",
      "content": "<p>Thank you for your answer. I didn't know which GPU to use for this competition or for general use. Your answer is quite sufficient.</p>",
      "rawMarkdown": "Thank you for your answer. I didn't know which GPU to use for this competition or for general use. Your answer is quite sufficient.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3416524,
      "author_name": "xaviertce",
      "author_url": "",
      "post_date": "03/03/2026 05:32:50",
      "content": "<p>For this competition, Kaggle is sufficient for both training and inference—you can refer to the public notebooks for reference. If you plan to use a rental GPU, you can choose one based on your requirements. This mainly depends on the competition dataset and the batch size you intend to use. In most cases, an RTX 3090 or RTX 4090 with 24 GB of GPU memory is more than adequate.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3416540,
          "author_name": "deryaumutkulal",
          "author_url": "",
          "post_date": "03/03/2026 06:34:25",
          "content": "<p>Thank you for your answer. I didn't know which GPU to use for this competition or for general use. Your answer is quite sufficient.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3416413": "Hello, I have a question I'm curious about. Since I usually participate in playgrounds or competitions held in my country, I haven't had much opportunity to use GPUs other than the one Kaggle provides us. I also want to participate in global competitions, but while the Kaggle GPU is sufficient for inference, I'm not sure if Kaggle GPUs are enough for training. Or generally, I'm not sure if Kaggle is used for training in global competitions. If the Kaggle GPU is not enough, could you suggest a good GPU that I would need to participate in competitions?",
    "3416524": "For this competition, Kaggle is sufficient for both training and inference—you can refer to the public notebooks for reference. If you plan to use a rental GPU, you can choose one based on your requirements. This mainly depends on the competition dataset and the batch size you intend to use. In most cases, an RTX 3090 or RTX 4090 with 24 GB of GPU memory is more than adequate.",
    "3416540": "Thank you for your answer. I didn't know which GPU to use for this competition or for general use. Your answer is quite sufficient."
  },
  "source": "meta"
}