Hi, I'm trying to reproduce your great work.
And I noticed that the args.zca should be set as True when distill the cifar10/100 dataset.
But I have a question that, there seems no zca.inverse_transform when save the distilled dataset, which may cause the final accuracy is low (only about 0.25).
I'd appreciate it if you can provide some information about it.
Thank you.
Hi, I'm trying to reproduce your great work.
And I noticed that the args.zca should be set as True when distill the cifar10/100 dataset.
But I have a question that, there seems no zca.inverse_transform when save the distilled dataset, which may cause the final accuracy is low (only about 0.25).
I'd appreciate it if you can provide some information about it.
Thank you.