{
  "id": 568438,
  "title": "Torch Template Iteration for Stanford RNA 3D Folding",
  "url": "/competitions/stanford-rna-3d-folding/discussion/568438",
  "author_name": "Kostas Anthoulis",
  "post_date": "2025-03-15T22:37:39.659000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hello all! This being the first Kaggle competition I'm participating in, and wanting to help out as much as I can, I created a template notebook for anyone to train their own iteration using Pytorch! Over the following days I will also tackle loading the synthetic (~400k sequences) dataset, as well as some fixes on the custom collate function for batching sequences together when not one-hot encoded. Hopefully this is is helping out anyone stuck trying to make their own Torch iteration. I've also included a basic LSTM as a baseline, though you're encouraged to change the model to anything you fancy, as long as input and output channels remain the same.</p>\n<p>Notebook link: <a href=\"https://www.kaggle.com/code/kostasanthoulis/stanford-rna-3d-folding-torch-template/notebook\" target=\"_blank\">https://www.kaggle.com/code/kostasanthoulis/stanford-rna-3d-folding-torch-template/notebook</a></p>",
  "messages": [
    {
      "id": 3150782,
      "postDate": "2025-03-15T22:37:39.660Z",
      "content": "<p>Hello all! This being the first Kaggle competition I'm participating in, and wanting to help out as much as I can, I created a template notebook for anyone to train their own iteration using Pytorch! Over the following days I will also tackle loading the synthetic (~400k sequences) dataset, as well as some fixes on the custom collate function for batching sequences together when not one-hot encoded. Hopefully this is is helping out anyone stuck trying to make their own Torch iteration. I've also included a basic LSTM as a baseline, though you're encouraged to change the model to anything you fancy, as long as input and output channels remain the same.</p>\n<p>Notebook link: <a href=\"https://www.kaggle.com/code/kostasanthoulis/stanford-rna-3d-folding-torch-template/notebook\" target=\"_blank\">https://www.kaggle.com/code/kostasanthoulis/stanford-rna-3d-folding-torch-template/notebook</a></p>",
      "rawMarkdown": "Hello all! This being the first Kaggle competition I'm participating in, and wanting to help out as much as I can, I created a template notebook for anyone to train their own iteration using Pytorch! Over the following days I will also tackle loading the synthetic (~400k sequences) dataset, as well as some fixes on the custom collate function for batching sequences together when not one-hot encoded. Hopefully this is is helping out anyone stuck trying to make their own Torch iteration. I've also included a basic LSTM as a baseline, though you're encouraged to change the model to anything you fancy, as long as input and output channels remain the same.\n\nNotebook link: https://www.kaggle.com/code/kostasanthoulis/stanford-rna-3d-folding-torch-template/notebook"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3150782": "Hello all! This being the first Kaggle competition I'm participating in, and wanting to help out as much as I can, I created a template notebook for anyone to train their own iteration using Pytorch! Over the following days I will also tackle loading the synthetic (~400k sequences) dataset, as well as some fixes on the custom collate function for batching sequences together when not one-hot encoded. Hopefully this is is helping out anyone stuck trying to make their own Torch iteration. I've also included a basic LSTM as a baseline, though you're encouraged to change the model to anything you fancy, as long as input and output channels remain the same.\n\nNotebook link: https://www.kaggle.com/code/kostasanthoulis/stanford-rna-3d-folding-torch-template/notebook"
  }
}