Addition of validation per batch

[RobvanGastel]

Is there an a possibility to add a validation set to every batch, besides the training and test set? I'm stuck on how to adjust the ClassSplitter class to experiment with adding the validation set.

dataset = omniglot("data", ways=5, shots=5, test_shots=15, meta_train=True, download=True)
dataloader = BatchMetaDataLoader(dataset, batch_size=16, num_workers=4)

for batch in dataloader:
    train_inputs, train_targets = batch["train"]
    val_inputs, val_targets = batch["val"]
    test_inputs, test_targets = batch["test"]

[RobvanGastel]

I figured that the proper approach to do is, is by taking some samples from the training set if possible and use it as validation set. As is done in this paper, ''Neural Voice Cloning with a Few Samples'', https://arxiv.org/abs/1802.06006.