Wrong implementation of shake-drop

[ArlindKadra]

Auto-PyTorch/autoPyTorch/pipeline/components/setup/network_backbone/utils.py

Lines 121 to 125 in 9002937

	if not is_training:
	# Move to torch.randn(1) for reproducibility
	bl = torch.tensor(1.0) if torch.randn(1) <= pl else torch.tensor(0.0)
	if is_training:
	bl = torch.tensor(pl)

We noticed with @ravinkohli that the implementation is not right for shake-drop. The if conditions should be reversed, only during test it should be fixed.