Problem of saving trained network parameters after network training
hanghoo opened this issue · 2 comments
hanghoo commented
Hello, if I want to save the network parameters, henceforth make test convenience, how can I do.
IshitaTakeshi commented
I think the easiest way is just saving the entire model :pcanet
by pickle.
If you want to save only network parameters, trained parameters appear in pcanet.pca_l1.components_
and pcanet.pca_l2.components_
. So just save them with hyperparameters listed from L141.
I have no idea about saving network parameters with HDF5 etc. since I'm not familiar with other deep learning frameworks, sorry.
hanghoo commented
@IshitaTakeshi Thank you very much.