Simple transformer encoder based classifier for amino acid sequences. Assumes data in similar format to dataset.csv.
Training can be done with the train_model.py script:
python train_model.py dataset.csv -m model_parameters.yaml -s trainer_parameters.yaml -o ./training_output -a gpu
The script automatically splits the data into training, validation and testing, but a separate test set can be given via the -t argument.
Example model and trainer parameters can be seen in e.g. data/model1.
If training on cpu, use -a cpu.
5-fold cross-validation can be done with the cross_validate.py script, which has the same input as above (except the -t argument).
The script trains 5 models and gathers the predictions and mean accuracy.