htjb / margarine

Code to replicate posterior probability distributions with bijectors/KDEs and perform marginal KL/bayesian dimensionality calculations.
MIT License
13 stars 8 forks source link

Unable to save flow.loss_history and flow.test_loss_history in MAF objects #56

Open DilyOng opened 5 months ago

DilyOng commented 5 months ago

After MAF objects (e.g. flow = MAF(data, weights=weights)) are trained (e.g. flow.train(epoch, early_stop=True)) and saved (e.g. using pickle), the information on flow.loss_history and flow.test_loss_history are lost.

htjb commented 4 months ago

Thanks @DilyOng this is a good spot! Sorry I haven't got around to fixing this. I will try and have a look asap.