{
  "id": 448162,
  "title": "run time limit, preprocessing, training and inference? ",
  "url": "/competitions/child-mind-institute-detect-sleep-states/discussion/448162",
  "author_name": "",
  "post_date": "2023-10-18T16:02:11.613353800Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello, I have a question about the time limit. Can we train the model beforehand and, during submission, have the notebook use the pre-trained model's weights? This way, we can utilize the full 9 hours for test feature calculation and inferences. Is this approach allowed, or are we required to perform data preprocessing(277 series in training, 200 in test), training, and inference within the 9-hour time frame?</p>",
  "messages": [
    {
      "id": "2487491",
      "postDate": "10/18/2023 16:02:11",
      "content": "<p>Hello, I have a question about the time limit. Can we train the model beforehand and, during submission, have the notebook use the pre-trained model's weights? This way, we can utilize the full 9 hours for test feature calculation and inferences. Is this approach allowed, or are we required to perform data preprocessing(277 series in training, 200 in test), training, and inference within the 9-hour time frame?</p>",
      "rawMarkdown": "Hello, I have a question about the time limit. Can we train the model beforehand and, during submission, have the notebook use the pre-trained model's weights? This way, we can utilize the full 9 hours for test feature calculation and inferences. Is this approach allowed, or are we required to perform data preprocessing(277 series in training, 200 in test), training, and inference within the 9-hour time frame?",
      "votes": null
    },
    {
      "id": "2487582",
      "postDate": "10/18/2023 16:56:14",
      "content": "<p>Yes, you can train offline, then upload your model as a private dataset, and use that dataset during inference.</p>",
      "rawMarkdown": "Yes, you can train offline, then upload your model as a private dataset, and use that dataset during inference.",
      "votes": null
    },
    {
      "id": "2487623",
      "postDate": "10/18/2023 17:19:10",
      "content": "<p>Thank you 😎</p>",
      "rawMarkdown": "Thank you 😎",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2487582,
      "author_name": "chris62",
      "author_url": "",
      "post_date": "10/18/2023 16:56:14",
      "content": "<p>Yes, you can train offline, then upload your model as a private dataset, and use that dataset during inference.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2487623,
          "author_name": "fredyrojas",
          "author_url": "",
          "post_date": "10/18/2023 17:19:10",
          "content": "<p>Thank you 😎</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2487491": "Hello, I have a question about the time limit. Can we train the model beforehand and, during submission, have the notebook use the pre-trained model's weights? This way, we can utilize the full 9 hours for test feature calculation and inferences. Is this approach allowed, or are we required to perform data preprocessing(277 series in training, 200 in test), training, and inference within the 9-hour time frame?",
    "2487582": "Yes, you can train offline, then upload your model as a private dataset, and use that dataset during inference.",
    "2487623": "Thank you 😎"
  },
  "source": "meta"
}