PyTorch batching at char-level and token-level

Question

I\'m trying to train an NLP model that uses a character level and token level LSTM to create and concatenate embeddings. The problem I\'m having is batching the data so that