how to train with 100% label?

[gehaoran-dev]

Hi @Na-Z，thanks for sharing this great work.
how to train with 100% label?
When I set labeled_ratio_list to [1. 0] for training，valueerror: num samples should be a positive integer value, but got num sample = 0.Then I set “shuffle” in dateloader to false, an error is still reported. I hope to get your suggestions

[Na-Z]

Hi @gehaoran-dev,

If you would like to train with 100% labels, you just need to remove those parts related to UNLABELED_DATASET in train_sess.py .

For example:

1. Do not create UNLABELED_DATASET in train_sess.py .
2. Modify

   sess/train_sess.py

   Lines 241 to 251 in f1bbb44

   	unlabeled_dataloader_iterator = iter(UNLABELED_DATALOADER)
   	for batch_idx, batch_data_label in enumerate(LABELED_DATALOADER):
   	try:
   	batch_data_unlabeled = next(unlabeled_dataloader_iterator)
   	except StopIteration:
   	unlabeled_dataloader_iterator = iter(UNLABELED_DATALOADER)
   	batch_data_unlabeled = next(unlabeled_dataloader_iterator)
   	for key in batch_data_unlabeled:
   	batch_data_label[key] = torch.cat((batch_data_label[key], batch_data_unlabeled[key]), dim=0)

   as for batch_idx, batch_data_label in enumerate(LABELED_DATALOADER):, since we only iterate over labeled dataset.

Hope this solves your problem.

[gehaoran-dev]

Thank you for the quick reply~