This release accompanies the pub Deep learning scaling behaviour in a quantitative genetics framework This dataset contains the underlying simulated data used for all three in-silico 'experiments' reported in the pub. Files 'alphasimr_dilution_input.tar.xz', and 'alphasimr_pleio_input.tar.xz' correspoding to data generated for the dilution and pleiotropy/genetic correlation experiments in the pub respecitvely. The two scaling files split the data generated for the first scaling experiment. The file named 'alphasimr_scaling_1e06_input.tar.xz' contains the largest simulation replicates with 1,000,000 sampled genomes. The file 'alphasimr_scaling_input.tar.xz' contains all other smaller simulation replicates. Each compressed directory includes simulated genotypes, phenotypes, and QTL effect sizes saved in .txt format for each simulation replicate.