I would like to train one of the efficientnetv2-bn networks on a custom dataset, and I noticed the following in effnetv2_configs.py line 202:

# For fair comparison to EfficientNetV1, using the same scaling and autoaug. 'efficientnetv2-b0': # 78.7% @ 7M params (v2_base_block, 1.0, 1.0, 192, 224, 0.2, 0, 0, 'effnetv1_autoaug'), 'efficientnetv2-b1': # 79.8% @ 8M params (v2_base_block, 1.0, 1.1, 192, 240, 0.2, 0, 0, 'effnetv1_autoaug'), 'efficientnetv2-b2': # 80.5% @ 10M params (v2_base_block, 1.1, 1.2, 208, 260, 0.3, 0, 0, 'effnetv1_autoaug'), 'efficientnetv2-b3': # 82.1% @ 14M params (v2_base_block, 1.2, 1.4, 240, 300, 0.3, 0, 0, 'effnetv1_autoaug'), 

I would really like to use the progressive learning mentioned in the paper instead of the augmentation used in V1. Would I just change the aug field to 'randaug' or are additional changes needed?