{
  "id": 426251,
  "title": "Is it best to use external datasets of predictions for generalized pretrained models such as DeepGOZero?",
  "url": "/competitions/cafa-5-protein-function-prediction/discussion/426251",
  "author_name": "SMC",
  "post_date": "2023-07-22T10:19:31.386000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I'm fairly new to ML and was wondering how people run the code for DeepGOZero (and other generalised pre-trained models) in their Kaggle notebooks. I was thinking it'd be as simple as importing the model and fitting it onto the training set, but I've noticed that people simply take external datasets of predictions made from these models. </p>\n<p>So, with that in consideration:</p>\n<p><strong>Are the generalised pre-trained models impossible to run in the Kaggle environment? Is it computationally too expensive to run these models in the Kaggle environment? If either is true, is it best practice to use these external datasets of predictions?</strong></p>\n<p>Feel free to correct me if I made any mistakes in my interpretation of the competition so far. </p>",
  "messages": [
    {
      "id": 2354285,
      "postDate": "2023-07-22T10:19:31.387Z",
      "content": "<p>I'm fairly new to ML and was wondering how people run the code for DeepGOZero (and other generalised pre-trained models) in their Kaggle notebooks. I was thinking it'd be as simple as importing the model and fitting it onto the training set, but I've noticed that people simply take external datasets of predictions made from these models. </p>\n<p>So, with that in consideration:</p>\n<p><strong>Are the generalised pre-trained models impossible to run in the Kaggle environment? Is it computationally too expensive to run these models in the Kaggle environment? If either is true, is it best practice to use these external datasets of predictions?</strong></p>\n<p>Feel free to correct me if I made any mistakes in my interpretation of the competition so far. </p>",
      "rawMarkdown": "I'm fairly new to ML and was wondering how people run the code for DeepGOZero (and other generalised pre-trained models) in their Kaggle notebooks. I was thinking it'd be as simple as importing the model and fitting it onto the training set, but I've noticed that people simply take external datasets of predictions made from these models. \n\nSo, with that in consideration:\n\n**Are the generalised pre-trained models impossible to run in the Kaggle environment? Is it computationally too expensive to run these models in the Kaggle environment? If either is true, is it best practice to use these external datasets of predictions?**\n\nFeel free to correct me if I made any mistakes in my interpretation of the competition so far. \n\n",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2354285": "I'm fairly new to ML and was wondering how people run the code for DeepGOZero (and other generalised pre-trained models) in their Kaggle notebooks. I was thinking it'd be as simple as importing the model and fitting it onto the training set, but I've noticed that people simply take external datasets of predictions made from these models. \n\nSo, with that in consideration:\n\n**Are the generalised pre-trained models impossible to run in the Kaggle environment? Is it computationally too expensive to run these models in the Kaggle environment? If either is true, is it best practice to use these external datasets of predictions?**\n\nFeel free to correct me if I made any mistakes in my interpretation of the competition so far. \n\n"
  }
}