Right now, the trained emulators are saved in npz files. However, this is done using a single array that chains together the weights and biases of all layers. A more clever approach would be to save them using Dicts and arrays. In this way, it should be easier to read the files.
Right now, the trained emulators are saved in
npzfiles. However, this is done using a single array that chains together the weights and biases of all layers. A more clever approach would be to save them usingDictsand arrays. In this way, it should be easier to read the files.