scDGD pretrained models

Each file presents a compressed collection of files for each of the pre-trained models presented in the `Deep Generative Decoder` paper (https://doi.org/10.1093/bioinformatics/btad497). Random seeds for all models was 0 and these instances were trained for 1000 epochs. For the two main models (dgd_Zheng2017 and dgd_mousebrain_1m), we also provide metadata information with cell types for the representations. In the case of the 1M mousebrain model, cell type labels were predicted using CellTypist.

Training data sources and processing are described in the corresponding paper: https://doi.org/10.1093/bioinformatics/btad49.