{
  "id": 254251,
  "title": "Uploading previously trained models",
  "url": "/competitions/rsna-miccai-brain-tumor-radiogenomic-classification/discussion/254251",
  "author_name": "",
  "post_date": "2021-07-20T21:15:52.705415800Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I have a quick question - I am looking at the submission portal right now. Looks like we have to make a jupyter notebook and a fresh run is performed. There doesn't seem to be a way to upload trained models? Does that mean we have to train our models on the Kaggle server or is there a way to upload previously trained models?</p>\n<p>Thanks!</p>",
  "messages": [
    {
      "id": "1395062",
      "postDate": "07/20/2021 21:15:52",
      "content": "<p>I have a quick question - I am looking at the submission portal right now. Looks like we have to make a jupyter notebook and a fresh run is performed. There doesn't seem to be a way to upload trained models? Does that mean we have to train our models on the Kaggle server or is there a way to upload previously trained models?</p>\n<p>Thanks!</p>",
      "rawMarkdown": "I have a quick question - I am looking at the submission portal right now. Looks like we have to make a jupyter notebook and a fresh run is performed. There doesn't seem to be a way to upload trained models? Does that mean we have to train our models on the Kaggle server or is there a way to upload previously trained models?\n\nThanks!",
      "votes": null
    },
    {
      "id": "1395071",
      "postDate": "07/20/2021 21:46:10",
      "content": "<p>Make a dataset with your model weights and reference that dataset in your notebook. </p>",
      "rawMarkdown": "Make a dataset with your model weights and reference that dataset in your notebook.",
      "votes": null
    },
    {
      "id": "1396628",
      "postDate": "07/22/2021 09:54:01",
      "content": "<p>No, you can also train your model at your local machine and upload the weights. Then you use the weights/model to predict the label</p>",
      "rawMarkdown": "No, you can also train your model at your local machine and upload the weights. Then you use the weights/model to predict the label",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1395071,
      "author_name": "richardepstein",
      "author_url": "",
      "post_date": "07/20/2021 21:46:10",
      "content": "<p>Make a dataset with your model weights and reference that dataset in your notebook. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1396628,
      "author_name": "lucamtb",
      "author_url": "",
      "post_date": "07/22/2021 09:54:01",
      "content": "<p>No, you can also train your model at your local machine and upload the weights. Then you use the weights/model to predict the label</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1395062": "I have a quick question - I am looking at the submission portal right now. Looks like we have to make a jupyter notebook and a fresh run is performed. There doesn't seem to be a way to upload trained models? Does that mean we have to train our models on the Kaggle server or is there a way to upload previously trained models?\n\nThanks!",
    "1395071": "Make a dataset with your model weights and reference that dataset in your notebook.",
    "1396628": "No, you can also train your model at your local machine and upload the weights. Then you use the weights/model to predict the label"
  },
  "source": "meta"
}