UTR_VAE

Multi-task learning for sequence modeling.

• A: Intigrating multiple-dataset to learn common regularoty feature
• B : Embed multiple sequence characteristics into one joint space

prerequistes

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
conda install pandas matplotlib seaborn configparser scikit-learn=0.23
conda install Biopython logomaker -c bioconda 

Model Configuring

click into any ini file under the log/ directory, copy one and change the parameter in the file. Some template configs were provided for your reference.

Training

• for multi-dataset model:

  python script/iter_train.py --config_file  <CONFIG_PATH>

  • if you are running on a GPU feasible machine:

  python script/iter_train.py --config_file  <CONFIG_PATH> --CUDA <GPU DEVICE INDEX> --kfold_index <which CV fold>

• for softshared multi-task model :

  python script/main_train.py --config_file