Predicted embds

Hi,

in predict_embs.py file, there are several unnecessary empty lines causing the indent error. They can be deleted as follows::

if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('-pos_low_file', type=str, default=None, help='csv file with positive mode [M+H]+ low energy spectrum')
parser.add_argument('-pos_high_file', type=str, default=None, help='csv file with positive mode [M+H]+ high energy spectrum')
parser.add_argument('-neg_low_file', type=str, default=None, help='csv file with positive mode [M-H]+ low energy spectrum')
parser.add_argument('-neg_high_file', type=str, default=None, help='csv file with positive mode [M-H]- low energy spectrum')
args = parser.parse_args()
main(args)

In addition, a command for saving pre_emb is missing in predict_embs.py (e.g. torch.save(pred_emb, 'sample_my.pt')).

Finally, it would be very helpful to write a wrapper to couple the embs predictor with the decoder and to add an automatic setting of the unknown keys when processing the example input data.