{
  "id": 662500,
  "title": "How to carry out the two-stage process of training and prediction？",
  "url": "/competitions/recodai-luc-scientific-image-forgery-detection/discussion/662500",
  "author_name": "",
  "post_date": "2025-12-13T09:04:21.651748600Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hey, everyone! I believe everyone has noticed that the running time of the notebook for this competition is not allowed to exceed 4 hours. This naturally leads one to think of first training a model locally or on the Kaggle platform, and then directly writing a prediction script and submitting it to the competition. \nAs a beginner, I attempted to train a model using the Kaggle platform, but during the training process, the browser often crashed. I would like to know if anyone has any good solutions?\nOr, does the organizing committee allow the participants to train a model locally on their own machines, and then use it to submit predictions?</p>",
  "messages": [
    {
      "id": "3375865",
      "postDate": "12/13/2025 09:04:21",
      "content": "<p>Hey, everyone! I believe everyone has noticed that the running time of the notebook for this competition is not allowed to exceed 4 hours. This naturally leads one to think of first training a model locally or on the Kaggle platform, and then directly writing a prediction script and submitting it to the competition. \nAs a beginner, I attempted to train a model using the Kaggle platform, but during the training process, the browser often crashed. I would like to know if anyone has any good solutions?\nOr, does the organizing committee allow the participants to train a model locally on their own machines, and then use it to submit predictions?</p>",
      "rawMarkdown": "Hey, everyone! I believe everyone has noticed that the running time of the notebook for this competition is not allowed to exceed 4 hours. This naturally leads one to think of first training a model locally or on the Kaggle platform, and then directly writing a prediction script and submitting it to the competition. \nAs a beginner, I attempted to train a model using the Kaggle platform, but during the training process, the browser often crashed. I would like to know if anyone has any good solutions?\nOr, does the organizing committee allow the participants to train a model locally on their own machines, and then use it to submit predictions?",
      "votes": null
    },
    {
      "id": "3376007",
      "postDate": "12/13/2025 13:40:45",
      "content": "<p>when you write your script to train, in the top right you can click \"Save version\" to Save and Run All (Commit) and then it will start running in the background for you. You can close your browser and it will keep running, just log in a few hours later to see the results</p>\n<p>alternatively u can train a model locally on your own machine too and upload it as a private model</p>",
      "rawMarkdown": "when you write your script to train, in the top right you can click \"Save version\" to Save and Run All (Commit) and then it will start running in the background for you. You can close your browser and it will keep running, just log in a few hours later to see the results\n\nalternatively u can train a model locally on your own machine too and upload it as a private model",
      "votes": null
    },
    {
      "id": "3376031",
      "postDate": "12/13/2025 14:24:31",
      "content": "<p>Thank you！I will have a try. wish you good results in the competition！</p>",
      "rawMarkdown": "Thank you！I will have a try. wish you good results in the competition！",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3376007,
      "author_name": "returnofsputnik",
      "author_url": "",
      "post_date": "12/13/2025 13:40:45",
      "content": "<p>when you write your script to train, in the top right you can click \"Save version\" to Save and Run All (Commit) and then it will start running in the background for you. You can close your browser and it will keep running, just log in a few hours later to see the results</p>\n<p>alternatively u can train a model locally on your own machine too and upload it as a private model</p>",
      "votes": null,
      "replies": [
        {
          "id": 3376031,
          "author_name": "yichiisallyouneed",
          "author_url": "",
          "post_date": "12/13/2025 14:24:31",
          "content": "<p>Thank you！I will have a try. wish you good results in the competition！</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3375865": "Hey, everyone! I believe everyone has noticed that the running time of the notebook for this competition is not allowed to exceed 4 hours. This naturally leads one to think of first training a model locally or on the Kaggle platform, and then directly writing a prediction script and submitting it to the competition. \nAs a beginner, I attempted to train a model using the Kaggle platform, but during the training process, the browser often crashed. I would like to know if anyone has any good solutions?\nOr, does the organizing committee allow the participants to train a model locally on their own machines, and then use it to submit predictions?",
    "3376007": "when you write your script to train, in the top right you can click \"Save version\" to Save and Run All (Commit) and then it will start running in the background for you. You can close your browser and it will keep running, just log in a few hours later to see the results\n\nalternatively u can train a model locally on your own machine too and upload it as a private model",
    "3376031": "Thank you！I will have a try. wish you good results in the competition！"
  },
  "source": "meta"
}