{
  "id": 487249,
  "title": "Pre-training a model and loading it for inference?",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/487249",
  "author_name": "",
  "post_date": "2024-03-28T10:03:47.739250200Z",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Since there is a time constraint and low memory seems to be an issue, are we allowed to train a model, say, logistic regression on our machine, upload the saved model on Kaggle as private and then load it on inference notebook? This way, we will not need to load the training data, significantly reducing the notebook's run time and memory requirements. Does it violate any rule?</p>",
  "messages": [
    {
      "id": "2720389",
      "postDate": "03/28/2024 10:03:47",
      "content": "<p>Since there is a time constraint and low memory seems to be an issue, are we allowed to train a model, say, logistic regression on our machine, upload the saved model on Kaggle as private and then load it on inference notebook? This way, we will not need to load the training data, significantly reducing the notebook's run time and memory requirements. Does it violate any rule?</p>",
      "rawMarkdown": "Since there is a time constraint and low memory seems to be an issue, are we allowed to train a model, say, logistic regression on our machine, upload the saved model on Kaggle as private and then load it on inference notebook? This way, we will not need to load the training data, significantly reducing the notebook's run time and memory requirements. Does it violate any rule?",
      "votes": null
    },
    {
      "id": "2720434",
      "postDate": "03/28/2024 10:43:36",
      "content": "<p>I have already started doing it.</p>",
      "rawMarkdown": "I have already started doing it.",
      "votes": null
    },
    {
      "id": "2720440",
      "postDate": "03/28/2024 10:51:27",
      "content": "<p>Not at all, it is the best way to handle the competition <a href=\"https://www.kaggle.com/basu1999\" target=\"_blank\">@basu1999</a> </p>",
      "rawMarkdown": "Not at all, it is the best way to handle the competition @basu1999",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2720434,
      "author_name": "yunsuxiaozi",
      "author_url": "",
      "post_date": "03/28/2024 10:43:36",
      "content": "<p>I have already started doing it.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2720440,
      "author_name": "ravi20076",
      "author_url": "",
      "post_date": "03/28/2024 10:51:27",
      "content": "<p>Not at all, it is the best way to handle the competition <a href=\"https://www.kaggle.com/basu1999\" target=\"_blank\">@basu1999</a> </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2720389": "Since there is a time constraint and low memory seems to be an issue, are we allowed to train a model, say, logistic regression on our machine, upload the saved model on Kaggle as private and then load it on inference notebook? This way, we will not need to load the training data, significantly reducing the notebook's run time and memory requirements. Does it violate any rule?",
    "2720434": "I have already started doing it.",
    "2720440": "Not at all, it is the best way to handle the competition @basu1999"
  },
  "source": "meta"
}