mlxops-aug 0.1.5

Data augmentation methods for deep learning (CutMix, MixUp, PuzzleMix)

mlxops-aug

Data augmentation methods for deep learning, including CutMix, MixUp, and PuzzleMix.

Installation

pip install mlxops-aug

For PuzzleMix support (requires pygco):

pip install "mlxops-aug[puzzlemix]"

Note: pygco requires a C compiler. See pyGCO for installation instructions.

Usage

CutMix / MixUp

from mlxops_aug import CutMixUp

aug = CutMixUp(
    num_classes=10,
    config={
        "use_cutmix": True,
        "use_mixup": True,
        "alpha": 1.0,
        "prob": 0.5,
    },
)

x_aug, y_aug = aug(x, y)

PuzzleMix

from mlxops_aug import PuzzleMix

aug = PuzzleMix(
    num_classes=10,
    config={
        "block_num": 2,
        "transport": True,
        "prob": 1.0,
    },
)

Requirements

• Python >= 3.10
• PyTorch >= 2.0
• torchvision >= 0.15

License

MIT