search

nelson1425 / EfficientAD

Unofficial implementation of EfficientAD https://arxiv.org/abs/2303.14535
https://arxiv.org/abs/2303.14535
Apache License 2.0
273 stars 72 forks source link

Asking why use batch size =1 ......? Please...help #20

Open yujins02 opened 1 year ago

yujins02 commented 1 year ago

Why use batch size =1 ? What is for it? train_loader = DataLoader(train_set, batch_size=1, shuffle=True, num_workers=4, pin_memory=True) train_loader_infinite = InfiniteDataloader(train_loader) validation_loader = DataLoader(validation_set, batch_size=1)

abc-125 commented 1 year ago

This batch size is specifically mentioned in the paper, and it is necessary to use size 1 for training EfficientAD. For validation, it's also 1 because of how this implementation works, but it can be a different number if you change the code.

PHaiJun commented 9 months ago

what if batch size > 1 in the train process?

Kaze816 commented 7 months ago

because it need 10Gb+ memory when training by one

Kaze816 commented 7 months ago

what if batch size > 1 in the train process?

maybe out of memory size.