Pytorch-lightning: Best way to use mixup in lightning?

[s-rog]

Just wondering what the best way to implement mixup in lightning is, possibly in the dataset?

[jeremyjordan]

Let's use the facebookresearch/mixup-cifar10 as an example. They implement it in the training loop.

for batch_idx, (inputs, targets) in enumerate(trainloader):
    if use_cuda:
        inputs, targets = inputs.cuda(), targets.cuda()

    inputs, targets_a, targets_b, lam = mixup_data(inputs, targets,
                                                    args.alpha, use_cuda)
    # rest of code

In PyTorch lightning, you would do this in your training step.

def training_step(self, batch, batch_idx):
    # we don't need to worry about CUDA, PyTorch Lightning does this for us
    inputs, targets = batch
    inputs, targets_a, targets_b, lam = mixup_data(inputs, targets,
                                                    args.alpha, use_cuda)
    # rest of code

[s-rog]

I'll give it a shot, thanks.