Closed dlmacedo closed 3 years ago
I guess that, even using seed, during the program using the following line generates new OOD from the 80 million available:
train_loader_out.dataset.offset = np.random.randint(len(train_loader_out.dataset))
I guess this may be overestimenting results as we may be using much more OOD data than originally thought.
I guess that, even using seed, during the program using the following line generates new OOD from the 80 million available:
train_loader_out.dataset.offset = np.random.randint(len(train_loader_out.dataset))
I guess this may be overestimenting results as we may be using much more OOD data than originally thought.