The datasets are associated with the non-orthogonal pilot assignment case in the article. The input and labelled output for training the models therein, and the computed SE performance for the conventional optimization approaches are provided. A brief description of the datasets: 'APpositions.npy': Represents the access point (AP) 2D locations utilized to generate the datasets. 'dataset_betas.npy': Large-scale fading coefficients in linear scale [W]. Represents the input to the DNN models. 'dataset_mu_XX_WMMSE_ADMM.npy': Locally optimal square roots of power coefficients [sqrt(W)] for the sum-SE maximization objective with; (1) XX = MR, and (2) XX = RZF precoding schemes. Represents the labelled output of the DNN models. 'dataset_mu_XX_WMMSE_PF_ADMM.npy': Locally optimal square roots of power coefficients [sqrt(W)] for the proportional fairness (PF) maximization objective with; (1) XX = MR, and (2) XX = RZF precoding schemes. Represents the labelled output of the DNN models. 'dataset_SE_XX_WMMSE_ADMM.npy': Per user spectral efficiency (SE) in [bits/s/Hz] for the sum-SE maximization objective with ; (1) XX = MR, and (2) XX = RZF precoding schemes. 'dataset_SE_XX_WMMSE_PF_ADMM.npy': Per user spectral efficiency (SE) in [bits/s/Hz] for the proportional fairness (PF) maximization objective with ; (1) XX = MR, and (2) XX = RZF precoding schemes. If you in anyway use this code for research that results in publications, please cite our original article listed below. The article can be found at: 10.1109/TWC.2022.3192203. Also on arXiv at: https://arxiv.org/pdf/2109.03128.pdf. The simulation code is available here on GitHub for training and testing the DNN models.