Our paper presents FPLearner, a deep learning-based tool designed for predicting the performance and accuracy of mixed-precision programs. FPLearner can be integrated into dynamic precision tuners to optimize time efficiency. The accompanying executable artifact aims to provide users with details of the raw data used/generated in our study, along with instructions demonstrating how to replicate/reproduce our evaluation results from scratch. The comprehensive documentation of the data and instructions can be found in our publicly accessible repository: https://github.com/ucd-plse/FPLearner, whose release has been archived in Zenodo: https://zenodo.org/doi/10.5281/zenodo.10426441.