Datasets

PiNN uses TensorFlow’s dataset class to handle data. Each data point is a dictionary of tensors. The data point should at least include the 'coord': coordinates and 'elems': atomic numbers as its features. For training potentials, 'e_data': energy and optionally 'f_data': forces are used as labels.

• Loading data
• Manipulating datasets
• Implemented loaders