How to force finishing the training loop and moving to the validation loop from within the trainer?

[motykatomasz]

Is it possible to finish the training loop from within one of the callbacks in the trainer, even though the data might still be in the data loader? I know you can return -1 from 'on_train_batch_end' but that moves the training the next epoch and skips the validation in the current one.