In this work, we present a deep learning based full-spectrum prediction method and demonstrate the merits of using full-spectrum predicted approaches for library searching. Our proposed model provides flexibility to accommodate post-translational modifications and fills the current gap for long peptide predictions.