what does the pretrain=True flag do in the mini-imagenet data loader?

[brando90]

@WangYueFt @HobbitLong what does the pretrain=True flag do?

e.g.

        if self.pretrain:
            self.file_pattern = 'miniImageNet_category_split_train_phase_%s.pickle'
        else:
            self.file_pattern = 'miniImageNet_category_split_%s.pickle'

rfs/dataset/mini_imagenet.py

Line 42 in f8c837b

if self.pretrain:

[txw1997]

Hello! Have you trained on the tieredImageNet yet? Can you share some information about the validation accuracy when training on the trainset?
My validation accuracy just 0.5%, but training accuracy is gradually ascending, do you have the same issue?