Retrain a fine tuned model

If I have huge training data or if I want to retrain the existing fine-tuned BERT/Longformer model on new data, can I use it (like we are using the bert-base-uncased model) to fine-tune it again? I tried saving the fine-tuned BERT/Longformer model and loading it again, but I was not able to do it. If not, what would be a good way to solve the memory error that we get by using huge data? clearing the GPU memory?