Implementation of mixup: Beyond Empirical Minimization by Zhang et al. using PyTorch and tested on CIFAR-10 Dataset.
No saving checkpoints saving/loading implemented. Please adjust to your needs accordingly.
The current runnable code is in train_local.py. The graviti_API.py is the ongoing SDK for Graviti. The SDK for faster retrieval of datasets from their platform is currently being implemented. You may update the dataloaders accordingly once the SDK implementation is refined.