{
  "id": 534410,
  "title": "Is it possible to train a model offline and submit it alongside the notebook?",
  "url": "/competitions/rsna-2024-lumbar-spine-degenerative-classification/discussion/534410",
  "author_name": "",
  "post_date": "2024-09-16T14:20:28.561289700Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>The nine hours running time makes it difficult to train such complex models. Is it possible to train models offline (on our computer resources), and then submit it alongside the submitted Jupyter notebook?</p>\n<p>Thank you<br>\nSimon</p>",
  "messages": [
    {
      "id": "2990555",
      "postDate": "09/16/2024 14:20:28",
      "content": "<p>Hi,</p>\n<p>The nine hours running time makes it difficult to train such complex models. Is it possible to train models offline (on our computer resources), and then submit it alongside the submitted Jupyter notebook?</p>\n<p>Thank you<br>\nSimon</p>",
      "rawMarkdown": "Hi,\n\nThe nine hours running time makes it difficult to train such complex models. Is it possible to train models offline (on our computer resources), and then submit it alongside the submitted Jupyter notebook?\n\nThank you\nSimon",
      "votes": null
    },
    {
      "id": "2990556",
      "postDate": "09/16/2024 14:26:17",
      "content": "<p>You can train your model on your local computer then use the trained model for inference. It’s allowed in the competition </p>",
      "rawMarkdown": "You can train your model on your local computer then use the trained model for inference. It’s allowed in the competition",
      "votes": null
    },
    {
      "id": "2991024",
      "postDate": "09/17/2024 00:03:32",
      "content": "<p>Definitely. You just need to upload your model as a dataset and link it as an input to your inference notebook. Many people do this, since, as you say, the training time (and hardware, to some extent) is limited on Kaggle. Would 100% recommend if you possess the resources to do so.</p>",
      "rawMarkdown": "Definitely. You just need to upload your model as a dataset and link it as an input to your inference notebook. Many people do this, since, as you say, the training time (and hardware, to some extent) is limited on Kaggle. Would 100% recommend if you possess the resources to do so.",
      "votes": null
    },
    {
      "id": "2992900",
      "postDate": "09/19/2024 08:22:52",
      "content": "<p>short answer, yes </p>",
      "rawMarkdown": "short answer, yes",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2990556,
      "author_name": "tabassumnova",
      "author_url": "",
      "post_date": "09/16/2024 14:26:17",
      "content": "<p>You can train your model on your local computer then use the trained model for inference. It’s allowed in the competition </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2991024,
      "author_name": "bharatraghavan03",
      "author_url": "",
      "post_date": "09/17/2024 00:03:32",
      "content": "<p>Definitely. You just need to upload your model as a dataset and link it as an input to your inference notebook. Many people do this, since, as you say, the training time (and hardware, to some extent) is limited on Kaggle. Would 100% recommend if you possess the resources to do so.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2992900,
      "author_name": "rohitchaudhari25",
      "author_url": "",
      "post_date": "09/19/2024 08:22:52",
      "content": "<p>short answer, yes </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2990555": "Hi,\n\nThe nine hours running time makes it difficult to train such complex models. Is it possible to train models offline (on our computer resources), and then submit it alongside the submitted Jupyter notebook?\n\nThank you\nSimon",
    "2990556": "You can train your model on your local computer then use the trained model for inference. It’s allowed in the competition",
    "2991024": "Definitely. You just need to upload your model as a dataset and link it as an input to your inference notebook. Many people do this, since, as you say, the training time (and hardware, to some extent) is limited on Kaggle. Would 100% recommend if you possess the resources to do so.",
    "2992900": "short answer, yes"
  },
  "source": "meta"
}