Code for "Onboard deep lossless and near-lossless predictive coding of hyperspectral images with line-based attention" paper. The method implements a low-complexity neural network for compression of hyperspectral images. The architecture works on a line-by-line basis using a hybrid recurrent-attentive scheme to limit memory requirements, and a predictive coding principle. The code allows to train the model on the HyspecNet-11k dataset, or test pretrained models on new hyperspectral images.