nextEpoch_structure_competition

Installation

!! First fork this repo so you can push your changes !!

pip install -r requirements.txt

Usage

This repo is a template for the competition. You need to write your pytorch model in src/model.py and the training loop in train.py.

Provided code

  1. Dataloaders: The function get_dataloaders() will construct a training and validation dataloader with the specified batch size, maximum sequence length and dataset size. Each batch is a dictionary with reference, sequence, and structure elements
  2. Template model: The class RNA_net embeds the integer encoded sequence into the sequence and matrix representations, and should be completed to crete a secondary structure model You can use some utils in util.py for general logging and metrics
  3. Metrics: we provide functions to compute the Precision, Recall and F1 score metrics in util.py
  4. Plotting: we provide a function to save a plot of a structure prediction compared to its ground truth in util.py