New augmentation

In DC-Bench, the following commonly used augmentations are provided

• Autoaug
• Randaug
• Imagenet_aug
• DSA

All augmentations are applied on the fly, in order to add a new augmentation method

Step 1

In evaluator_utils.py inside evaluator folder, find method: custom_aug

Step 2

Add new data augmentation methods in the form of Pytorch Transform

One example is

data_transforms = transforms.Compose([transforms.RandAugment(num_ops=1)])

Step 3

Run the eval commands with new aug

python evaluator/evaluator.py --method DC  --dataset tinyimagenet --aug NEW_AUG